Nuclear Reactor Physics

Weston M. Stacey Nuclear Reactor Physics Second Edition, Completely Revised and Enlarged Weston M. Stacey Nuclear Reactor Physics 1807–2007 Knowledge for Generations Each generation has its unique needs and aspirations. When Charles Wiley first opened his small printing shop in lower Manhattan in 1807, it was a generation of boundless potential searching for an identity. And we were there, helping to define a new American literary tradition. Over half a century later, in the midst of the Second Industrial Revolution, it was a generation focused on building the future. Once again, we were there, supplying the critical scientific, technical, and engineering knowledge that helped frame the world. Throughout the 20th Century, and into the new millennium, nations began to reach out beyond their own borders and a new international community was born. Wiley was there, expanding its operations around the world to enable a global exchange of ideas, opinions, and know-how. For 200 years, Wiley has been an integral part of each generation’s journey, enabling the flow of information and understanding necessary to meet their needs and fulfill their aspirations. Today, bold new technologies are changing the way we live and learn. Wiley will be there, providing you the must-have knowledge you need to imagine new worlds, new possibilities, and new opportunities. Generations come and go, but you can always count on Wiley to provide you the knowledge you need, when and where you need it! William J. Pesce President and Chief Executive Officer Peter Booth Wiley Chairman of the Board Weston M. Stacey Nuclear Reactor Physics Second Edition, Completely Revised and Enlarged The Author Prof. Weston M. Stacey Georgia Institute of Technology Nuclear & Radiological Engineering 900 Atlantic Drive, NW Atlanta, GA 30332-0425 USA Cover Four-assembly fuel module for a boiling water reactor (Courtesy of General Electric Company). All books published by Wiley-VCH are carefully produced. Nevertheless, authors, editors, and publisher do not warrant the information contained in these books, including this book, to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate. Library of Congress Card No.: applied for British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at <http://dnb.ddb.de>. © 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Typesetting VTEX, Vilnius, Lithuania Printing betz-druck GmbH, Darmstadt Binding Litges & Dopf Buchbinderei GmbH, Heppenheim Printed in the Federal Republic of Germany Printed on acid-free paper ISBN 978-3-527-40679-1 To Penny, Helen, Billy, and Lucia vii Contents Preface xxiii Preface to 2nd Edition PART 1 1 1.1 1.2 1.3 1.4 1.5 1.6 2 2.1 xxvii BASIC REACTOR PHYSICS Neutron Nuclear Reactions 3 Neutron-Induced Nuclear Fission 3 Stable Nuclides 3 Binding Energy 3 Threshold External Energy for Fission 4 Neutron-Induced Fission 5 Neutron Fission Cross Sections 5 Products of the Fission Reaction 8 Energy Release 10 Neutron Capture 13 Radiative Capture 13 Neutron Emission 19 Neutron Elastic Scattering 20 Summary of Cross-Section Data 24 Low-Energy Cross Sections 24 Spectrum-Averaged Cross Sections 24 Evaluated Nuclear Data Files 24 Elastic Scattering Kinematics 27 Correlation of Scattering Angle and Energy Loss Average Energy Loss 29 28 Neutron Chain Fission Reactors 33 Neutron Chain Fission Reactions 33 Capture-to-Fission Ratio 33 Number of Fission Neutrons per Neutron Absorbed in Fuel 33 Contents viii 2.2 2.3 2.4 3 3.1 3.2 3.3 3.4 3.5 3.6 Neutron Utilization 34 Fast Fission 34 Resonance Escape 36 Criticality 37 Effective Multiplication Constant 37 Effect of Fuel Lumping 37 Leakage Reduction 38 Time Dependence of a Neutron Fission Chain Assembly Prompt Fission Neutron Time Dependence 38 Source Multiplication 39 Effect of Delayed Neutrons 39 Classification of Nuclear Reactors 40 Physics Classification by Neutron Spectrum 40 Engineering Classification by Coolant 41 38 Neutron Diffusion Theory 43 Derivation of One-Speed Diffusion Theory 43 Partial and Net Currents 43 Diffusion Theory 45 Interface Conditions 46 Boundary Conditions 46 Applicability of Diffusion Theory 47 Solutions of the Neutron Diffusion Equation in Nonmultiplying Media 48 Plane Isotropic Source in an Infinite Homogeneous Medium 48 Plane Isotropic Source in a Finite Homogeneous Medium 48 Line Source in an Infinite Homogeneous Medium 49 Homogeneous Cylinder of Infinite Axial Extent with Axial Line Source 49 Point Source in an Infinite Homogeneous Medium 49 Point Source at the Center of a Finite Homogeneous Sphere 50 Diffusion Kernels and Distributed Sources in a Homogeneous Medium 50 Infinite-Medium Diffusion Kernels 50 Finite-Slab Diffusion Kernel 51 Finite Slab with Incident Neutron Beam 52 Albedo Boundary Condition 52 Neutron Diffusion and Migration Lengths 53 Thermal Diffusion-Length Experiment 53 Migration Length 55 Bare Homogeneous Reactor 57 Slab Reactor 57 Right Circular Cylinder Reactor 59 Contents 3.7 3.8 3.9 3.10 3.11 3.12 4 4.1 Interpretation of Criticality Condition 60 Optimum Geometries 61 Reflected Reactor 62 Reflected Slab Reactor 62 Reflector Savings 64 Reflected Spherical, Cylindrical, and Rectangular Parallelepiped Cores 65 Homogenization of a Heterogeneous Fuel–Moderator Assembly Spatial Self-Shielding and Thermal Disadvantage Factor 65 Effective Homogeneous Cross Sections 69 Thermal Utilization 71 Measurement of Thermal Utilization 72 Local Power Peaking Factor 73 Control Rods 73 Effective Diffusion Theory Cross Sections for Control Rods 73 Windowshade Treatment of Control Rods 76 Numerical Solution of Diffusion Equation 77 Finite Difference Equations in One Dimension 78 Forward Elimination/Backward Substitution Spatial Solution Procedure 79 Power Iteration on Fission Source 79 Finite-Difference Equations in Two Dimensions 80 Successive Relaxation Solution of Two-Dimensional Finite-Difference Equations 82 Power Outer Iteration on Fission Source 82 Limitations on Mesh Spacing 83 Nodal Approximation 83 Transport Methods 85 Transmission and Absorption in a Purely Absorbing Slab Control Plate 87 Escape Probability in a Slab 87 Integral Transport Formulation 87 Collision Probability Method 88 Differential Transport Formulation 89 Spherical Harmonics Methods 90 Discrete Ordinates Method 94 Neutron Energy Distribution 101 Analytical Solutions in an Infinite Medium 101 Fission Source Energy Range 102 Slowing-Down Energy Range 102 Moderation by Hydrogen Only 103 Energy Self-Shielding 103 Slowing Down by Nonhydrogenic Moderators with No Absorption 65 104 ix Contents x 4.2 4.3 4.4 5 5.1 5.2 5.3 5.4 Slowing-Down Density 105 Slowing Down with Weak Absorption 106 Fermi Age Neutron Slowing Down 107 Neutron Energy Distribution in the Thermal Range 108 Summary 111 Multigroup Calculation of Neutron Energy Distribution in an Infinite Medium 111 Derivation of Multigroup Equations 111 Mathematical Properties of the Multigroup Equations 113 Solution of Multigroup Equations 114 Preparation of Multigroup Cross-Section Sets 115 Resonance Absorption 117 Resonance Cross Sections 117 Doppler Broadening 119 Resonance Integral 122 Resonance Escape Probability 122 Multigroup Resonance Cross Section 122 Practical Width 122 Neutron Flux in Resonance 123 Narrow Resonance Approximation 123 Wide Resonance Approximation 124 Resonance Absorption Calculations 124 Temperature Dependence of Resonance Absorption 127 Multigroup Diffusion Theory 127 Multigroup Diffusion Equations 127 Two-Group Theory 128 Two-Group Bare Reactor 129 One-and-One-Half-Group Theory 129 Two-Group Theory of Two-Region Reactors 130 Two-Group Theory of Reflected Reactors 133 Numerical Solutions for Multigroup Diffusion Theory 137 Nuclear Reactor Dynamics 143 Delayed Fission Neutrons 143 Neutrons Emitted in Fission Product Decay 143 Effective Delayed Neutron Parameters for Composite Mixtures Photoneutrons 146 Point Kinetics Equations 147 Period–Reactivity Relations 148 Approximate Solutions of the Point Neutron Kinetics Equations One Delayed Neutron Group Approximation 150 Prompt-Jump Approximation 153 Reactor Shutdown 154 145 150 Contents 5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 Delayed Neutron Kernel and Zero-Power Transfer Function 155 Delayed Neutron Kernel 155 Zero-Power Transfer Function 155 Experimental Determination of Neutron Kinetics Parameters 156 Asymptotic Period Measurement 156 Rod Drop Method 157 Source Jerk Method 157 Pulsed Neutron Methods 157 Rod Oscillator Measurements 158 Zero-Power Transfer Function Measurements 159 Rossi-α Measurement 159 Reactivity Feedback 161 Temperature Coefficients of Reactivity 162 Doppler Effect 162 Fuel and Moderator Expansion Effect on Resonance Escape Probability 164 Thermal Utilization 165 Nonleakage Probability 166 Representative Thermal Reactor Reactivity Coefficients 166 Startup Temperature Defect 167 Perturbation Theory Evaluation of Reactivity Temperature Coefficients 168 Perturbation Theory 168 Sodium Void Effect in Fast Reactors 169 Doppler Effect in Fast Reactors 169 Fuel and Structure Motion in Fast Reactors 170 Fuel Bowing 171 Representative Fast Reactor Reactivity Coefficients 171 Reactor Stability 171 Reactor Transfer Function with Reactivity Feedback 171 Stability Analysis for a Simple Feedback Model 172 Threshold Power Level for Reactor Stability 174 More General Stability Conditions 175 Power Coefficients and Feedback Delay Time Constants 178 Measurement of Reactor Transfer Functions 179 Rod Oscillator Method 179 Correlation Methods 179 Reactor Noise Method 181 Reactor Transients with Feedback 183 Step Reactivity Insertion (ρex < β): Prompt Jump 184 Step Reactivity Insertion (ρex < β): Post-Prompt-Jump Transient 185 Reactor Fast Excursions 186 Step Reactivity Input: Feedback Proportional to Fission Energy 186 Ramp Reactivity Input: Feedback Proportional to Fission Energy 187 xi Contents xii 5.13 6 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 6.10 Step Reactivity Input: Nonlinear Feedback Proportional to Cumulative Energy Release 187 Bethe–Tait Model 188 Numerical Methods 190 Fuel Burnup 197 Changes in Fuel Composition 197 Fuel Transmutation–Decay Chains 198 Fuel Depletion–Transmutation–Decay Equations 199 Fission Products 203 Solution of the Depletion Equations 204 Measure of Fuel Burnup 205 Fuel Composition Changes with Burnup 205 Reactivity Effects of Fuel Composition Changes 206 Compensating for Fuel-Depletion Reactivity Effects 208 Reactivity Penalty 208 Effects of Fuel Depletion on the Power Distribution 209 In-Core Fuel Management 210 Samarium and Xenon 211 Samarium Poisoning 211 Xenon Poisoning 213 Peak Xenon 215 Effect of Power-Level Changes 216 Fertile-to-Fissile Conversion and Breeding 217 Availability of Neutrons 217 Conversion and Breeding Ratios 219 Simple Model of Fuel Depletion 219 Fuel Reprocessing and Recycling 221 Composition of Recycled LWR Fuel 221 Physics Differences of MOX Cores 222 Physics Considerations with Uranium Recycle 224 Physics Considerations with Plutonium Recycle 225 Reactor Fueling Characteristics 225 Radioactive Waste 226 Radioactivity 226 Hazard Potential 226 Risk Factor 226 Burning Surplus Weapons-Grade Uranium and Plutonium Composition of Weapons-Grade Uranium and Plutonium Physics Differences Between Weapons- and Reactor-Grade Plutonium-Fueled Reactors 234 Utilization of Uranium Energy Content 235 Transmutation of Spent Nuclear Fuel 237 Closing the Nuclear Fuel Cycle 244 233 233 Contents 7 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 7.10 7.11 7.12 7.13 8 8.1 8.2 Nuclear Power Reactors 249 Pressurized Water Reactors 249 Boiling Water Reactors 250 Pressure Tube Heavy Water–Moderated Reactors 255 Pressure Tube Graphite-Moderated Reactors 258 Graphite-Moderated Gas-Cooled Reactors 260 Liquid-Metal Fast Breeder Reactors 261 Other Power Reactors 265 Characteristics of Power Reactors 265 Advanced Generation-III Reactors 265 Advanced Boiling Water Reactors (ABWR) 266 Advanced Pressurized Water Reactors (APWR) 267 Advanced Pressure Tube Reactor 268 Modular High-Temperature Gas-Cooled Reactors (GT-MHR) 268 Advanced Generation-IV Reactors 269 Gas-Cooled Fast Reactors (GFR) 270 Lead-Cooled Fast Reactors (LFR) 271 Molten Salt Reactors (MSR) 271 Super-Critical Water Reactors (SCWR) 272 Sodium-Cooled Fast Reactors (SFR) 272 Very High Temperature Reactors (VHTR) 272 Advanced Sub-critical Reactors 273 Nuclear Reactor Analysis 275 Construction of Homogenized Multigroup Cross Sections 275 Criticality and Flux Distribution Calculations 276 Fuel Cycle Analyses 277 Transient Analyses 278 Core Operating Data 279 Criticality Safety Analysis 279 Interaction of Reactor Physics and Reactor Thermal Hydraulics 280 Power Distribution 280 Temperature Reactivity Effects 281 Coupled Reactor Physics and Thermal-Hydraulics Calculations 281 Reactor Safety 283 Elements of Reactor Safety 283 Radionuclides of Greatest Concern 283 Multiple Barriers to Radionuclide Release Defense in Depth 285 Energy Sources 285 Reactor Safety Analysis 285 Loss of Flow or Loss of Coolant 287 Loss of Heat Sink 287 283 xiii Contents xiv 8.3 8.4 8.5 Reactivity Insertion 287 Anticipated Transients without Scram Quantitative Risk Assessment 288 Probabilistic Risk Assessment 288 Radiological Assessment 291 Reactor Risks 291 Reactor Accidents 293 Three Mile Island 294 Chernobyl 297 Passive Safety 299 Pressurized Water Reactors 299 Boiling Water Reactors 299 Integral Fast Reactors 300 Passive Safety Demonstration 300 PART 2 9 9.1 9.2 9.3 9.4 288 ADVANCED REACTOR PHYSICS Neutron Transport Theory 305 Neutron Transport Equation 305 Boundary Conditions 310 Scalar Flux and Current 310 Partial Currents 310 Integral Transport Theory 310 Isotropic Point Source 311 Isotropic Plane Source 311 Anisotropic Plane Source 312 Transmission and Absorption Probabilities 314 Escape Probability 314 First-Collision Source for Diffusion Theory 315 Inclusion of Isotropic Scattering and Fission 315 Distributed Volumetric Sources in Arbitrary Geometry 316 Flux from a Line Isotropic Source of Neutrons 317 Bickley Functions 318 Probability of Reaching a Distance t from a Line Isotropic Source without a Collision 318 Collision Probability Methods 319 Reciprocity Among Transmission and Collision Probabilities 320 Collision Probabilities for Slab Geometry 320 Collision Probabilities in Two-Dimensional Geometry 321 Collision Probabilities for Annular Geometry 322 Interface Current Methods in Slab Geometry 323 Emergent Currents and Reaction Rates Due to Incident Currents 323 Emergent Currents and Reaction Rates Due to Internal Sources 326 Contents 9.5 9.6 9.7 9.8 9.9 9.10 9.11 9.12 Total Reaction Rates and Emergent Currents 327 Boundary Conditions 329 Response Matrix 329 Multidimensional Interface Current Methods 330 Extension to Multidimension 330 Evaluation of Transmission and Escape Probabilities 332 Transmission Probabilities in Two-Dimensional Geometries 333 Escape Probabilities in Two-Dimensional Geometries 335 Simple Approximations for the Escape Probability 337 Spherical Harmonics (PL ) Methods in One-Dimensional Geometries 338 Legendre Polynomials 338 Neutron Transport Equation in Slab Geometry 339 339 PL Equations Boundary and Interface Conditions 340 342 P1 Equations and Diffusion Theory 343 Simplified PL or Extended Diffusion Theory 344 PL Equations in Spherical and Cylindrical Geometries Diffusion Equations in One-Dimensional Geometry 347 Half-Angle Legendre Polynomials 347 348 Double-PL Theory 349 D-P0 Equations 350 Multidimensional Spherical Harmonics (PL ) Transport Theory Spherical Harmonics 350 Spherical Harmonics Transport Equations in Cartesian Coordinates 351 352 P1 Equations in Cartesian Geometry Diffusion Theory 353 Discrete Ordinates Methods in One-Dimensional Slab Geometry 354 355 PL and D-PL Ordinates Spatial Differencing and Iterative Solution 357 Limitations on Spatial Mesh Size 358 Discrete Ordinates Methods in One-Dimensional Spherical Geometry 359 Representation of Angular Derivative 360 Iterative Solution Procedure 360 Acceleration of Convergence 362 Calculation of Criticality 362 Multidimensional Discrete Ordinates Methods 363 Ordinates and Quadrature Sets 363 366 SN Method in Two-Dimensional x–y Geometry Further Discussion 369 Even-Parity Transport Formulation 369 Monte Carlo Methods 371 Probability Distribution Functions 371 xv Contents xvi Analog Simulation of Neutron Transport Statistical Estimation 373 Variance Reduction 375 Tallying 377 Criticality Problems 378 Source Problems 379 Random Numbers 380 372 10.5 Neutron Slowing Down 385 Elastic Scattering Transfer Function 385 Lethargy 385 Elastic Scattering Kinematics 385 Elastic Scattering Kernel 386 Isotropic Scattering in Center-of-Mass System 388 Linearly Anisotropic Scattering in Center-of-Mass System 389 390 P1 and B1 Slowing-Down Equations Derivation 390 Solution in Finite Uniform Medium 393 394 B1 Equations Few-Group Constants 395 Diffusion Theory 396 Lethargy-Dependent Diffusion Theory 396 Directional Diffusion Theory 397 Multigroup Diffusion Theory 398 Boundary and Interface Conditions 399 Continuous Slowing-Down Theory 400 400 P1 Equations in Slowing-Down Density Formulation Slowing-Down Density in Hydrogen 403 Heavy Mass Scatterers 404 Age Approximation 404 Selengut–Goertzel Approximation 405 405 Consistent P1 Approximation Extended Age Approximation 405 Grueling–Goertzel Approximation 406 407 Summary of Pl Continuous Slowing-Down Theory Inclusion of Anisotropic Scattering 407 Inclusion of Scattering Resonances 409 410 Pl Continuous Slowing-Down Equations Multigroup Discrete Ordinates Transport Theory 411 11 11.1 Resonance Absorption 415 Resonance Cross Sections 415 10 10.1 10.2 10.3 10.4 Contents 11.2 11.3 11.4 11.5 11.6 12 12.1 12.2 12.3 Widely Spaced Single-Level Resonances in a Heterogeneous Fuel–Moderator Lattice 415 Neutron Balance in Heterogeneous Fuel–Moderator Cell 415 Reciprocity Relation 418 Narrow Resonance Approximation 419 Wide Resonance Approximation 420 Evaluation of Resonance Integrals 420 Infinite Dilution Resonance Integral 422 Equivalence Relations 422 Heterogeneous Resonance Escape Probability 423 Homogenized Multigroup Resonance Cross Section 423 Improved and Intermediate Resonance Approximations 424 Calculation of First-Flight Escape Probabilities 424 Escape Probability for an Isolated Fuel Rod 425 Closely Packed Lattices 427 Unresolved Resonances 428 Multigroup Cross Sections for Isolated Resonances 430 Self-Overlap Effects 431 Overlap Effects for Different Sequences 432 Multiband Treatment of Spatially Dependent Self-Shielding 433 Spatially Dependent Self-Shielding 433 Multiband Theory 434 Evaluation of Multiband Parameters 436 Calculation of Multiband Parameters 437 Interface Conditions 439 Resonance Cross-Section Representations 439 R-Matrix Representation 439 Practical Formulations 441 Generalization of the Pole Representation 445 Doppler Broadening of the Generalized Pole Representation 448 Neutron Thermalization 453 Double Differential Scattering Cross Section for Thermal Neutrons 453 Neutron Scattering from a Monatomic Maxwellian Gas 454 Differential Scattering Cross Section 454 Cold Target Limit 455 Free-Hydrogen (Proton) Gas Model 455 455 Radkowsky Model for Scattering from H2 O Heavy Gas Model 456 Thermal Neutron Scattering from Bound Nuclei 457 Pair Distribution Functions and Scattering Functions 457 Intermediate Scattering Functions 458 Incoherent Approximation 459 xvii Contents xviii 12.4 12.5 12.6 13 13.1 13.2 13.3 13.4 13.5 13.6 13.7 13.8 13.9 Gaussian Representation of Scattering 459 Measurement of the Scattering Function 460 Applications to Neutron Moderating Media 460 Calculation of the Thermal Neutron Spectra in Homogeneous Media 462 Wigner–Wilkins Proton Gas Model 463 Heavy Gas Model 466 Numerical Solution 468 Moments Expansion Solution 470 Multigroup Calculation 473 Applications to Moderators 474 Calculation of Thermal Neutron Energy Spectra in Heterogeneous Lattices 474 Pulsed Neutron Thermalization 477 Spatial Eigenfunction Expansion 477 Energy Eigenfunctions of the Scattering Operator 477 Expansion in Energy Eigenfunctions of the Scattering Operator 479 Perturbation and Variational Methods 483 Perturbation Theory Reactivity Estimate 483 Multigroup Diffusion Perturbation Theory 483 Adjoint Operators and Importance Function 486 Adjoint Operators 486 Importance Interpretation of the Adjoint Function 487 Eigenvalues of the Adjoint Equation 489 Variational/Generalized Perturbation Reactivity Estimate 489 One-Speed Diffusion Theory 490 Other Transport Models 493 Reactivity Worth of Localized Perturbations in a Large PWR Core Model 494 Higher-Order Variational Estimates 495 Variational/Generalized Perturbation Theory Estimates of Reaction Rate Ratios in Critical Reactors 495 Variational/Generalized Perturbation Theory Estimates of Reaction Rates 497 Variational Theory 498 Stationarity 498 Roussopolos Variational Functional 498 Schwinger Variational Functional 499 Rayleigh Quotient 499 Construction of Variational Functionals 500 Variational Estimate of Intermediate Resonance Integral 500 Heterogeneity Reactivity Effects 502 Variational Derivation of Approximate Equations 503 Contents 13.10 Variational Even-Parity Transport Approximations 505 Variational Principle for the Even-Parity Transport Equation Ritz Procedure 506 Diffusion Approximation 507 One-Dimensional Slab Transport Equation 508 13.11 Boundary Perturbation Theory 508 14 14.1 14.2 14.3 14.4 14.5 14.6 14.7 14.8 15 15.1 15.2 15.3 15.4 15.5 15.6 505 Homogenization 515 Equivalent Homogenized Cross Sections 516 ABH Collision Probability Method 517 Blackness Theory 520 Fuel Assembly Transport Calculations 522 Pin Cells 522 Wigner–Seitz Approximation 523 Collision Probability Pin-Cell Model 524 Interface Current Formulation 527 Multigroup Pin-Cell Collision Probabilities Model 528 Resonance Cross Sections 529 Full Assembly Transport Calculation 529 Homogenization Theory 529 Homogenization Considerations 530 Conventional Homogenization Theory 531 Equivalence Homogenization Theory 531 Multiscale Expansion Homogenization Theory 535 Flux Detail Reconstruction 538 Nodal and Synthesis Methods 541 General Nodal Formalism 542 Conventional Nodal Methods 544 Transverse Integrated Nodal Diffusion Theory Methods 547 Transverse Integrated Equations 547 Polynomial Expansion Methods 549 Analytical Methods 553 Heterogeneous Flux Reconstruction 554 Transverse Integrated Nodal Integral Transport Theory Models Transverse Integrated Integral Transport Equations 554 Polynomial Expansion of Scalar Flux 557 Isotropic Component of Transverse Leakage 558 558 Double-Pn Expansion of Surface Fluxes Angular Moments of Outgoing Surface Fluxes 559 Nodal Transport Equations 561 Transverse Integrated Nodal Discrete Ordinates Method 561 Finite-Element Coarse Mesh Methods 563 554 xix Contents xx Variational Functional for the P1 Equations 563 One-Dimensional Finite-Difference Approximation 564 Diffusion Theory Variational Functional 566 Linear Finite-Element Diffusion Approximation in One Dimension 567 Higher-Order Cubic Hermite Coarse-Mesh Diffusion Approximation 569 Multidimensional Finite-Element Coarse-Mesh Methods 570 15.7 Variational Discrete Ordinates Nodal Method 571 Variational Principle 571 Application of the Method 579 15.8 Variational Principle for Multigroup Diffusion Theory 580 15.9 Single-Channel Spatial Synthesis 583 15.10 Multichannel Spatial Synthesis 589 15.11 Spectral Synthesis 591 16 16.1 16.2 16.3 16.4 Space–Time Neutron Kinetics 599 Flux Tilts and Delayed Neutron Holdback 599 Modal Eigenfunction Expansion 600 Flux Tilts 601 Delayed Neutron Holdback 602 Spatially Dependent Point Kinetics 602 Derivation of Point Kinetics Equations 604 Adiabatic and Quasistatic Methods 605 Variational Principle for Static Reactivity 606 Variational Principle for Dynamic Reactivity 607 Time Integration of the Spatial Neutron Flux Distribution 609 Explicit Integration: Forward-Difference Method 610 Implicit Integration: Backward-Difference Method 611 Implicit Integration: θ Method 612 Implicit Integration: Time-Integrated Method 615 Implicit Integration: GAKIN Method 616 Alternating Direction Implicit Method 619 Stiffness Confinement Method 622 Symmetric Successive Overrelaxation Method 623 Generalized Runge–Kutta Methods 624 Stability 625 Classical Linear Stability Analysis 625 Lyapunov’s Method 627 Lyapunov’s Method for Distributed Parameter Systems 629 Control 631 Variational Methods of Control Theory 631 Dynamic Programming 633 Pontryagin’s Maximum Principle 634 Contents 16.5 16.6 Variational Methods for Spatially Dependent Control Problems 636 Dynamic Programming for Spatially Continuous Systems 638 Pontryagin’s Maximum Principle for a Spatially Continuous System 639 Xenon Spatial Oscillations 641 Linear Stability Analysis 642 μ-Mode Approximation 644 λ-Mode Approximation 645 Nonlinear Stability Criterion 649 Control of Xenon Spatial Power Oscillations 650 Variational Control Theory of Xenon Spatial Oscillations 650 Stochastic Kinetics 652 Forward Stochastic Model 653 Means, Variances, and Covariances 656 Correlation Functions 658 Physical Interpretation, Applications, and Initial and Boundary Conditions 659 Numerical Studies 660 Startup Analysis 663 APPENDICES A Physical Constants and Nuclear Data 669 B Some Useful Mathematical Formulas 675 C C.1 C.2 Step Functions, Delta Functions, and Other Functions Introduction 677 Properties of the Dirac δ-Function 678 A. Alternative Representations 678 B. Properties 678 C. Derivatives 679 D Some Properties of Special Functions E E.1 E.2 Introduction to Matrices and Matrix Algebra Some Definitions 687 Matrix Algebra 689 681 687 677 xxi Contents xxii F F.1 F.2 Introduction to Laplace Transforms 691 Motivation 691 “Cookbook” Laplace transforms 694 Index 697 xxiii Preface Nuclear reactor physics is the physics of neutron fission chain reacting systems. It encompasses those applications of nuclear physics and radiation transport and interaction with matter that determine the behavior of nuclear reactors. As such, it is both an applied physics discipline and the core discipline of the field of nuclear engineering. As a distinct applied physics discipline, nuclear reactor physics originated in the middle of the twentieth century in the wartime convergence of international physics efforts in the Manhattan Project. It developed vigorously for roughly the next third of the century in various government, industrial, and university R&D and design efforts worldwide. Nuclear reactor physics is now a relatively mature discipline, in that the basic physical principles governing the behavior of nuclear reactors are well understood, most of the basic nuclear data needed for nuclear reactor analysis have been measured and evaluated, and the computational methodology is highly developed and validated. It is now possible to accurately predict the physics behavior of existing nuclear reactor types under normal operating conditions. Moreover, the basic physical concepts, nuclear data, and computational methodology needed to develop an understanding of new variants of existing reactor types or of new reactor types exist for the most part. As the core discipline of nuclear engineering, nuclear reactor physics is fundamental to the major international nuclear power undertaking. As of 2000, there are 434 central station nuclear power reactors operating worldwide to produce 350,442 MWe of electrical power. This is a substantial fraction of the world’s electrical power (e.g., more than 80% of the electricity produced in France and more than 20% of the electricity produced in the United States). The world’s electrical power requirements will continue to increase, particularly as the less developed countries strive to modernize, and nuclear power is the only proven technology for meeting these growing electricity requirements without dramatically increasing the already unacceptable levels of greenhouse gas emission into the atmosphere. Nuclear reactors have additional uses other than central station electricity production. There are more than 100 naval propulsion reactors in the U.S. fleet (plus others in foreign fleets). Nuclear reactors are also employed for basic neutron physics research, for materials testing, for radiation therapy, for the production of radio-isotopes for medical, industrial, and national security applications, and xxiv Preface as mobile power sources for remote stations. In the future, nuclear reactors may power deep space missions. Thus nuclear reactor physics is a discipline important to the present and future well-being of the world. This book is intended as both a textbook and a comprehensive reference on nuclear reactor physics. The basic physical principles, nuclear data, and computational methodology needed to understand the physics of nuclear reactors are developed and applied to explain the static and dynamic behavior of nuclear reactors in Part 1. This development is at a level that should be accessible to seniors in physics or engineering (i.e., requiring a mathematical knowledge only through ordinary and partial differential equations and Laplace transforms and an undergraduatelevel knowledge of atomic and nuclear physics). Mastery of the material presented in Part 1 provides an understanding of the physics of nuclear reactors sufficient for nuclear engineering graduates at the B.S. and M.S. levels, for most practicing nuclear engineers and for others interested in acquiring a broad working knowledge of nuclear reactor physics. The material in Part 1 was developed in the process of teaching undergraduate and first-year graduate courses in nuclear reactor physics at Georgia Tech for a number of years. The emphasis in the presentation is on conveying the basic physical concepts and their application to explain nuclear reactor behavior, using the simplest mathematical description that will suffice to illustrate the physics. Numerous examples are included to illustrate the step-by-step procedures for carrying out the calculations discussed in the text. Problems at the end of each chapter have been chosen to provide physical insight and to extend the material discussed in the text, while providing practice in making calculations; they are intended as an integral part of the textbook. Part 1 is suitable for an undergraduate semesterlength course in nuclear reactor physics; the material in Part 1 is also suitable for a semester-length first-year graduate course, perhaps with selective augmentation from Part 2. The purpose of Part 2 is to augment Part 1 to provide a comprehensive, detailed, and advanced development of the principal topics of nuclear reactor physics. There is an emphasis in Part 2 on the theoretical bases for the advanced computational methods of reactor physics. This material provides a comprehensive, though necessarily abridged, reference work on advanced nuclear reactor physics and the theoretical bases for its computational methods. Although the material stops short of descriptions of specific reactor physics codes, it provides the basis for understanding the code manuals. There is more than enough material in Part 2 for a semester-length advanced graduate course in nuclear reactor physics. The treatment is necessarily somewhat more mathematically intense than in Part 1. Part 2 is intended primarily for those who are or would become specialists in nuclear reactor physics and reactor physics computations. Mastery of this material provides the background for creating the new physics concepts necessary for developing new reactor types and for understanding and extending the computational methods in existing reactor physics codes (i.e., the stock-in-trade for the professional reactor physicist). Moreover, the extensive treatment of neutron transport computational methods also provides an important component of the background Preface necessary for specialists in radiation shielding, for specialists in the applications of neutrons and photons in medicine and industry, and for specialists in neutron, photon, and neutral atom transport in industrial, astrophysical, and thermonuclear plasmas. Any book of this scope owes much to many people besides the author, and this one is no exception. The elements of the subject of reactor physics were developed by many talented people over the past half-century, and the references can only begin to recognize their contributions. In this regard, I note the special contribution of R.N. Hwang, who helped prepare certain sections on resonance theory. The selection and organization of material has benefited from the example of previous authors of textbooks on reactor physics. The feedback from a generation of students has assisted in shaping the organization and presentation. Several people (C. Nickens, B. Crumbly, S. Bennett-Boyd) supported the evolution of the manuscript through at least three drafts, and several other people at Wiley transformed the manuscript into a book. I am grateful to all of these people, for without them there would be no book. Atlanta, Georgia October 2000 Weston M. Stacey xxv xxvii Preface to 2nd Edition This second edition differs from the original in two important ways. First, a section on neutron transport methods has been added in Chapter 3 to provide an introduction to that subject in the first section of the book on basic reactor physics that is intended as the text for an advanced undergraduate course. My original intention was to use diffusion theory to introduce reactor physics, without getting into the mathematical complexities of transport theory. I think this works reasonably well from a pedagogical point of view, but it has the disadvantage of sending BS graduates into the workplace without an exposure to transport theory. So, a short section on transport methods in slab geometry was added at the end of the diffusion theory chapter to provide an introduction. Second, there has been a resurgence in interest and activity in the improvement of reactor designs and in the development of new reactor concepts that are more inherently safe, better utilize the uranium resources, discharge less long-lived waste and are more resistant to the diversion of fuels to other uses. A section has been added in Chapter 7 on the improved Generation-III designs that will be coming online over the next decade or so, and a few sections have been added in Chapters 6 and 7 on the new reactor concepts being developed under the Generation-IV and Advanced Fuel Cycle Initiatives with the objective of closing the nuclear fuel cycle. The text was amplified for the sake of explication in a few places, some additional homework problems were included, and numerous typos, omissions and other errors that slipped through the final proof-reading of the first edition were corrected. I am grateful to colleagues, students and particularly the translators preparing a Russian edition of the book for calling several such mistakes to my attention. Otherwise, the structure and context of the book remains unchanged. The first eight chapters on basic reactor physics provide the text for a first course in reactor physics at the advanced undergraduate or graduate level. The second eight chapters on advanced reactor physics provide a text suitable for graduate courses on neutron transport theory and reactor physics. I hope that this second edition will serve to introduce to the field the new generation of scientists and engineers who will carry forward the emerging resurgence of xxviii Preface to 2nd Edition nuclear power to meet the growing energy needs of mankind in a safe, economical, environmentally sustainable and proliferation-resistant way. Weston M. Stacey Atlanta, Georgia May 2006 Part 1: Basic Reactor Physics 3 1 Neutron Nuclear Reactions The physics of nuclear reactors is determined by the transport of neutrons and their interaction with matter within a reactor. The basic neutron nucleus reactions of importance in nuclear reactors and the nuclear data used in reactor physics calculations are described in this chapter. 1.1 Neutron-Induced Nuclear Fission Stable Nuclides Short-range attractive nuclear forces acting among nucleons (neutrons and protons) are stronger than the Coulomb repulsive forces acting among protons at distances on the order of the nuclear radius (R ≈ 1.25 × 10−13 A1/3 cm) in a stable nucleus. These forces are such that the ratio of the atomic mass A (the number of neutrons plus protons) to the atomic number Z (the number of protons) increases with Z; in other words, the stable nuclides become increasingly neutron-rich with increasing Z, as illustrated in Fig. 1.1. The various nuclear species are referred to as nuclides, and nuclides with the same atomic number are referred to as isotopes of the element corresponding to Z. We use the notation A XZ (e.g., 235 U92 ) to identify nuclides. Binding Energy The actual mass of an atomic nucleus is not the sum of the masses (mp ) of the Z protons and the masses (mn ) of A − Z neutrons of which it is composed. The stable nuclides have a mass defect  = [Zmp + (A − Z)mn ] − A mz (1.1) This mass defect is conceptually thought of as having been converted to energy (E = c2 ) at the time that the nucleus was formed, putting the nucleus into a negative energy state. The amount of externally supplied energy that would have 4 1 Neutron Nuclear Reactions Fig. 1.1 Nuclear stability curve. (From Ref. 1; used with permission of McGraw-Hill.) to be converted to mass in disassembling a nucleus into its separate nucleons is known as the binding energy of the nucleus, BE = c2 . The binding energy per nucleon (BE/A) is shown in Fig. 1.2. Any process that results in nuclides being converted to other nuclides with more binding energy per nucleon will result in the conversion of mass into energy. The combination of low A nuclides to form higher A nuclides with a higher BE/A value is the basis for the fusion process for the release of nuclear energy. The splitting of very high A nuclides to form intermediate-A nuclides with a higher BE/A value is the basis of the fission process for the release of nuclear energy. Threshold External Energy for Fission The probability of any nuclide undergoing fission (reconfiguring its A nucleons into two nuclides of lower A) can become quite large if a sufficient amount of external energy is supplied to excite the nucleus. The minimum, or threshold, amount of such excitation energy required to cause fission with high probability depends on the nuclear structure and is quite large for nuclides with Z < 90. For nuclides with Z > 90, the threshold energy is about 4 to 6 MeV for even-A nuclides, and generally is much lower for odd-A nuclides. Certain of the heavier nuclides (e.g., 240 Pu94 and 252 Cf98 ) exhibit significant spontaneous fission even in the absence of any externally supplied excitation energy. 1.1 Neutron-Induced Nuclear Fission Fig. 1.2 Binding energy per nucleon. (From Ref. 1; used with permission of McGraw-Hill.) Neutron-Induced Fission When a neutron is absorbed into a heavy nucleus (A, Z) to form a compound nucleus (A + 1, Z), the BE/A value is lower for the compound nucleus than for the original nucleus. For some nuclides (e.g., 233 U92 , 235 U92 , 239 Pu94 , 241 Pu94 ), this reduction in BE/A value is sufficient that the compound nucleus will undergo fission, with high probability, even if the neutron has very low energy. Such nuclides are referred to as fissile; that is, they can be caused to undergo fission by the absorption of a low-energy neutron. If the neutron had kinetic energy prior to being absorbed into a nucleus, this energy is transformed into additional excitation energy of the compound nucleus. All nuclides with Z > 90 will undergo fission with high probability when a neutron with kinetic energy in excess of about 1 MeV is absorbed. Nuclides such as 232 Th90 , 238 U92 , and 240 Pu94 will undergo fission with neutrons with energy of about 1 MeV or higher, with high probability. Neutron Fission Cross Sections The probability of a nuclear reaction, in this case fission, taking place can be expressed in terms of a quantity σ which expresses the probable reaction rate for n neutrons traveling with speed v a distance dx in a material with N nuclides per unit volume: σ≡ reaction rate nvN dx (1.2) The units of σ are area, which gives rise to the concept of σ as a cross-sectional area presented to the neutron by the nucleus, for a particular reaction process, and 5 6 1 Neutron Nuclear Reactions Fig. 1.3 Fission cross sections for 233 U92 . (From http://www.nndc.bnl.gov/.) to the designation of σ as a cross section. Cross sections are usually on the order of 10−24 cm2 , and this unit is referred to as a barn, for historical reasons. The fission cross section, σf , is a measure of the probability that a neutron and a nucleus interact to form a compound nucleus which then undergoes fission. The probability that a compound nucleus will be formed is greatly enhanced if the relative energy of the neutron and the original nucleus, plus the reduction in the nuclear binding energy, corresponds to the difference in energy of the ground state and an excited state of the compound nucleus, so that the energetics are just right for formation of a compound nucleus in an excited state. The first excited states of the compound nuclei resulting from neutron absorption by odd-A fissile nuclides are generally lower lying (nearer to the ground state) than are the first excited states of the compound nuclei resulting from neutron absorption by the heavy even-A nuclides, which accounts for the odd-A nuclides having much larger absorption and fission cross sections for low-energy neutrons than do the even-A nuclides. Fission cross sections for some of the principal fissile nuclides of interest for nuclear reactors are shown in Figs. 1.3 to 1.5. The resonance structure corresponds to the formation of excited states of the compound nuclei, the lowest lying of which are at less than 1 eV. The nature of the resonance cross section can be shown to give rise to a 1/E 1/2 or 1/υ dependence of the cross section at off-resonance neutron energies below and above the resonance range, as is evident in these figures. The fission cross sections are largest in the thermal energy region E < ∼1 eV. The thermal fission cross section for 239 Pu94 is larger than that of 235 U92 or 233 U92 . 1.1 Neutron-Induced Nuclear Fission Fig. 1.4 Fission cross sections for 235 U92 . (From http://www.nndc.bnl.gov/.) Fig. 1.5 Fission cross sections for 239 Pu94 . (From http://www.nndc.bnl.gov/.) 7 8 1 Neutron Nuclear Reactions Fig. 1.6 Fission cross sections for 238 U92 . (From http://www.nndc.bnl.gov/.) Fission cross sections for 238 U92 and 240 Pu94 are shown in Figs. 1.6 and 1.7. Except for resonances, the fission cross section is insignificant below about 1 MeV, above which it is about 1 barn. The fission cross sections for these and other even-A heavy mass nuclides are compared in Fig. 1.8, without the resonance structure. Products of the Fission Reaction A wide range of nuclides are formed by the fission of heavy mass nuclides, but the distribution of these fission fragments is sharply peaked in the mass ranges 90 < A < 100 and 135 < A < 145, as shown in Fig. 1.9. With reference to the curvature of the trajectory of the stable isotopes on the n versus p plot of Fig. 1.1, most of these fission fragments are above the stable isotopes (i.e., are neutron rich) and will decay, usually by β-decay (electron emission), which transmutes the fission fragment nuclide (A, Z) to (A, Z + 1), or sometimes by neutron emission, which transmutes the fission fragment nuclide (A, Z) to (A − 1, Z), in both instances toward the range of stable isotopes. Sometimes several decay steps are necessary to reach a stable isotope. Usually either two or three neutrons will be emitted promptly in the fission event, and there is a probability of one or more neutrons being emitted subsequently upon the decay of neutron-rich fission fragments over the next second or so. The number of neutrons, on average, which are emitted in the fission process, ν, depends on the fissioning nuclide and on the energy of the neutron inducing fission, as shown in Fig. 1.10. 1.1 Neutron-Induced Nuclear Fission Fig. 1.7 Fission cross sections for 240 Pu94 . (From http://www.nndc.bnl.gov/.) Fig. 1.8 Fission cross sections for principal nonfissile heavy mass nuclides. (From Ref. 15; used with permission of Argonne National Laboratory.) 9 10 1 Neutron Nuclear Reactions Fig. 1.9 Yield versus mass number for 235 U92 fission. (From Ref. 15.) Energy Release The majority of the nuclear energy created by the conversion of mass to energy in the fission event (207 MeV for 233 U92 ) is in the form of the kinetic energy (168 MeV) of the recoiling fission fragments. The range of these massive, highly charged particles in the fuel element is a fraction of a millimeter, so that the recoil energy is effectively deposited as heat at the point of fission. Another 5 MeV is in the form of kinetic energy of prompt neutrons released in the fission event, distributed in energy as shown in Fig. 1.11, with a most likely energy of 0.7 MeV (for 235 U ). This energy is deposited in the surrounding material within 10 to 100 cm 92 as the neutron diffuses, slows down by scattering collisions with nuclei, and is finally absorbed. A fraction of these neutron absorption events result in neutron capture followed by gamma emission, producing on average about 7 MeV in the form of energetic capture gammas per fission. This secondary capture gamma en- 1.1 Neutron-Induced Nuclear Fission Fig. 1.10 Average number of neutrons emitted per fission. (From Ref. 12; used with permission of Wiley.) Fig. 1.11 Fission spectrum for thermal neutron-induced fission in 235 U92 . (From Ref. 12; used with permission of Wiley.) 11 12 1 Neutron Nuclear Reactions Table 1.1 235 U92 Fission Energy Release Form Energy (MeV) Range Kinetic energy fission products Kinetic energy prompt gammas Kinetic energy prompt neutrons Kinetic energy capture gammas Decay of fission products Kinetic energy electrons Kinetic energy neutrinos 168 7 5 7 < mm 10–100 cm 10–100 cm 10–100 cm ∼mm ∞ 8 12 ergy is transferred as heat to the surrounding material over a range of 10 to 100 cm by gamma interactions. There is also on average about 7 MeV of fission energy directly released as gamma rays in the fission event, which is deposited as heat within the surrounding 10 to 100 cm. The remaining 20 MeV of fission energy is in the form of kinetic energy of electrons (8 MeV) and neutrinos (12 MeV) from the decay of the fission fragments. The electron energy is deposited, essentially in the fuel element, within about 1 mm of the fission fragment, but since neutrinos rarely interact with matter, the neutrino energy is lost. Although the kinetic energy of the neutrons emitted by the decay of fission products is almost as great as that of the prompt fission neutrons, there are so few delayed neutrons from fission product decay that their contribution to the fission energy distribution is negligible. This fission energy distribution for 235 U92 is summarized in Table 1.1. The recoverable energy released from fission by thermal and fission spectrum neutrons is given in Table 1.2. Table 1.2 Recoverable Energy from Fission Isotope Thermal Neutron Fission Neutron 233 U 190.0 192.9 198.5 200.3 – – – – – – – – – – – – 184.2 188.9 191.4 193.9 193.6 196.9 196.9 200.0 235 U 239 Pu 241 Pu 232 Th 234 U 236 U 238 U 237 Np 238 Pu 240 Pu 242 Pu Source: Data from Ref. 12; used with permission of Wiley. 1.2 Neutron Capture Thus, in total, about 200 MeV per fission of heat energy is produced. One Watt of heat energy then corresponds to the fission of 3.1 × 1010 nuclei per second. Since 1 g of any fissile nuclide contains about 2.5 × 1021 nuclei, the fissioning of 1 g of fissile material produces about 1 megawatt-day (MWd) of heat energy. Because some fissile nuclei will also be transmuted by neutron capture, the amount of fissile material destroyed is greater than the amount fissioned. 1.2 Neutron Capture Radiative Capture When a neutron is absorbed by a nucleus to form a compound nucleus, a number of reactions are possible, in addition to fission, in the heavy nuclides. We have already mentioned radiative capture, in which the compound nucleus decays by the emission of a gamma ray, and we now consider this process in more detail. An energy-level diagram for the compound nucleus formation and decay associated with the prominent 238 U92 resonance for incident neutron energies of about 6.67 eV is shown in Fig. 1.12. The energy in the center-of-mass (CM) system of an incident neutron with energy EL in the lab system is Ec = [A/(1 + A)]EL . The reduction in binding energy due to the absorbed neutron is EB . If Ec + EB is close to an excited energy level of the compound nucleus, the probability for com- Fig. 1.12 Energy-level diagram for compound nucleus formation. (From Ref. 12; used with permission of Wiley.) 13 14 1 Neutron Nuclear Reactions Fig. 1.13 Radiative capture cross section for 232 Th90 . (From http://www.nndc.bnl.gov/.) pound nucleus formation is greatly enhanced. The excited compound nucleus will generally decay by emission of one or more gamma rays, the combined energy of which is equal to the difference in the excited- and ground-state energy levels of the compound nucleus. Radiative capture cross sections, denoted σγ , for some nuclei of interest for nuclear reactors are shown in Figs. 1.13 to 1.21. The resonance nature of the cross sections over certain ranges correspond to the discrete excited states of the compound nucleus that is formed upon neutron capture. These excited states correspond to neutron energies in the range of a fraction of an eV to 103 eV for the fissile nuclides, generally correspond to neutron energies of 10 to 104 eV for even-A heavy mass nuclides (with the notable exception of thermal 240 Pu94 resonance), and correspond to much higher neutron energies for the lower mass nuclides. The 1/ν “off-resonance” cross-section dependence is apparent. The Breit–Wigner single-level resonance formula for the neutron capture cross section is  1/2 1 2 γ E0 , y = (Ec − E0 ) (1.3) σγ (Ec ) = σ0 Ec 1 + y2 where E0 is the energy (in the CM) system at which the resonance peak occurs (i.e., Ec + EB matches the energy of an excited state of the compound nucleus), the full width at half-maximum of the resonance, σ0 the maximum value of the total cross section (at E0 ), and γ the radiative capture width ( γ / is the probability that the compound nucleus, once formed, will decay by gamma emission). The 1.2 Neutron Capture Fig. 1.14 Radiative capture cross section for 233 U92 . (From http://www.nndc.bnl.gov/.) Fig. 1.15 Radiative capture cross section for 235 U92 . (From http://www.nndc.bnl.gov/.) 15 16 1 Neutron Nuclear Reactions Fig. 1.16 Radiative capture cross section for 239 Pu94 . (From http://www.nndc.bnl.gov/.) Fig. 1.17 Radiative capture cross section for 238 U92 . (From http://www.nndc.bnl.gov/.) 1.2 Neutron Capture Fig. 1.18 Radiative capture cross section for 240 Pu94 . (From http://www.nndc.bnl.gov/.) Fig. 1.19 Radiative capture cross section for 56 Fe26 . (From http://www.nndc.bnl.gov/.) 17 18 1 Neutron Nuclear Reactions Fig. 1.20 Radiative capture cross section for 23 Na11 . (From http://www.nndc.bnl.gov/.) Fig. 1.21 Radiative capture cross section for 1 H1 . (From http://www.nndc.bnl.gov/.) 1.2 Neutron Capture fission resonance cross section can be represented by a similar expression with the fission width f , defined such that f / is the probability that the compound nucleus, once formed, will decay by fission. Equation (1.3) represents the cross section describing the interaction of a neutron and nucleus with relative (CM) energy Ec . However, the nuclei in a material are distributed in energy (approximately a Maxwellian distribution characterized by the temperature of the material). What is needed is a cross section averaged over the motion of the nuclei:  1 dE  |v(E) − v(E  )| σ (Ec ) fmax (E  , T ) (1.4) σ̄ (E, T ) = υ(E) where E and E  are the neutron and nuclei energies, respectively, in the lab system, and fmax (E  ) is the Maxwellian energy distribution: fmax (E  ) = √ 2π  E  e−E /kT 3/2 (πkT ) (1.5) Using Eqs. (1.3) and (1.5), Eq. (1.4) becomes σ̄γ (E, T ) = σ0  γ E0 E 1/2 (1.6) (ξ, x) where x= 2 ξ= (E − E0 ), (4E0 kT /A)1/2 (1.7) A is the atomic mass (amu) of the nuclei, and ξ (ξ, x) = √ 2 π  ∞ −∞ e−(1/4)(x−y) 2ξ 2 dy 1 + y2 (1.8) Neutron Emission When the compound nucleus formed by neutron capture decays by the emission of one neutron, leaving the nucleus in an excited state which subsequently undergoes further decays, the event is referred to as inelastic scattering and the cross section is denoted σin . Since the nucleus is left in an excited state, the energy of the emitted neutron can be considerably less than the energy of the incident neutron. If the compound nucleus decays by the emission of two or more neutrons, the events are referred to as n − 2n, n − 3n, and so on, events, and the cross sections are denoted σn,2n , σn,3n , on so on. Increasingly higher incident neutron energies are required to provide enough excitation energy for single, double, triple, and so on, neutron emission. Inelastic scattering is the most important of these events in nuclear reactors, but it is most important for neutrons 1 MeV and higher in energy. 19 20 1 Neutron Nuclear Reactions 1.3 Neutron Elastic Scattering Elastic scattering may take place via compound nucleus formation followed by the emission of a neutron that returns the compound nucleus to the ground state of the original nucleus. In such a resonance elastic scattering event the kinetic energy of the original neutron–nuclear system is conserved. The neutron and the nucleus may also interact without neutron absorption and the formation of a compound nucleus, which is referred to as potential scattering. Although quantum mechanical (s-wave) in nature, the latter event may be visualized and treated as a classical hard-sphere scattering event, away from resonance energies. Near resonance energies, there is quantum mechanical interference between the potential and resonance scattering, which is constructive just above and destructive just below the resonance energy. The single-level Breit–Wigner form of the scattering cross section, modified to include potential and interference scattering, is σs (Ec ) = σ0 n  E0 Ec 1/2 1 σ0 2R y + + 4πR 2 2 λ0 1 + y 2 1+y (1.9) where ( n / ) is the probability that, once formed, the compound nucleus decays to the ground state of the original nucleus by neutron emission, R  1.25 × 10−13 A1/3 centimeters is the nuclear radius, and λ0 is the reduced neutron wavelength. Averaging over a Maxwellian distribution of nuclear motion yields the scattering cross section for neutron lab energy E and material temperature T : σ̄s (E, T ) = σ0 n ψ(ξ, x) + σ0 R χ(ξ, x) + 4πR 2 λ0 (1.10) where ξ χ(ξ, x) = √ π  ∞ −∞ ye−(1/4)(x−y) 1 + y2 2ξ 2 dy (1.11) The elastic scattering cross sections for a number of nuclides of interest in nuclear reactors are shown in Figs. 1.22 to 1.26. In general, the elastic scattering cross section is almost constant in energy below the neutron energies corresponding to the excited states of the compound nucleus. The destructive interference effects just below the resonance energy are very evident in Fig. 1.26. The energy dependence of the carbon scattering cross section is extended to very low neutron energies in Fig. 1.27 to illustrate another phenomenon. At sufficiently small neutron energy, the neutron wavelength λ0 = 2.86 × 10−9 h h cm = √ =√ p E(eV) 2mE (1.12) 1.3 Neutron Elastic Scattering Fig. 1.22 Elastic scattering cross section for 1 H1 . (From http://www.nndc.bnl.gov/.) Fig. 1.23 Elastic scattering cross section for 16 O8 . (From http://www.nndc.bnl.gov/.) 21 22 1 Neutron Nuclear Reactions Fig. 1.24 Elastic scattering cross section for 23 Na11 . (From http://www.nndc.bnl.gov/.) Fig. 1.25 Elastic scattering cross section for 56 Fe26 . (From http://www.nndc.bnl.gov/.) 1.3 Neutron Elastic Scattering Fig. 1.26 Elastic scattering cross section for 238 U92 . (From http://www.nndc.bnl.gov/.) Fig. 1.27 Total scattering cross section of 12 C6 . (From Ref. 12; used with permission of Wiley.) 23 24 1 Neutron Nuclear Reactions becomes comparable to the interatomic spacing, and the neutron interacts not with a single nucleus but with an aggregate of bound nuclei. If the material has a regular structure, as graphite does, the neutron will be diffracted and the energy dependence of the cross section will reflect the neutron energies corresponding to multiples of interatomic spacing. For sufficiently small energies, diffraction becomes impossible and the cross section is once again insensitive to neutron energy. 1.4 Summary of Cross-Section Data Low-Energy Cross Sections The low-energy total cross sections for several nuclides of interest in nuclear reactors are plotted in Fig. 1.28. Gadolinium is sometimes used as a “burnable poison,” and xenon and samarium are fission products with large thermal cross sections. Spectrum-Averaged Cross Sections Table 1.3 summarizes the cross-section data for a number of important nuclides in nuclear reactors. The first three columns give fission, radiative capture, and elastic scattering cross sections averaged over a Maxwellian distribution with T = 0.0253 eV, corresponding to a representative thermal energy spectrum. The next two columns give the infinite dilution fission and radiative capture resonance integrals, which are averages of the respective resonance cross sections over a 1/E spectrum typical of the resonance energy region in the limit of an infinitely dilute concentration of the resonance absorber. The final five columns give cross sections averaged over the fission spectrum. Example 1.1: Calculation of Macroscopic Cross Section. The macroscopic cross section  = Nσ , where N is the number density. The number density is related to the density ρ and atomic number A by N = (ρ/A)N0 , where N0 = 6.022 × 1023 is Avogadro’s number, the number of atoms in a mole. For a mixture of isotopes with volume fractions υi , the macroscopic cross section is  = i υi (ρ/A)i N0 σi ; for example, for a 50:50 vol % mixture of carbon and 238 U, the macroscopic thermal absorption cross section is a = 0.5(ρC /AC )N0 σaC + 0.5(ρU /AU )N0 σaU = 0.5(1.60 g/cm3 per 12 g/mol)(6.022 × 1023 atom/mol)(0.003 × 10−24 cm2 ) + 0.5(18.9 g/cm3 per 238 g/mol)(6.022 × 1023 atom/mol)(2.4 × 10−24 cm2 ) = 0.0575 cm−1 . 1.5 Evaluated Nuclear Data Files Published experimental and theoretical results on neutron–nuclear reactions are collected by several collaborating nuclear data agencies worldwide. Perhaps the 469 507 698 938 – – 0.05 – – – – – – – – – – – – 233 U 92 235 U 92 239 Pu 94 241 Pu 94 232 Th 90 238 U 92 240 Pu 94 242 Pu 94 1H 1 2H 1 10 B 5 12 C 6 16 O 8 23 Na 11 56 Fe 26 91 Zr 40 135 Xe 54 149 Sm 62 157 Gd 64 41 87 274 326 6.5 2.4 264 16.8 0.29 5 × 10−4 443 0.003 2 × 10−4 0.47 2.5 1.1 2.7 × 106 6.0 × 104 1.9 × 103 11.9 15.0 7.8 11.1 13.7 9.4 1.5 8.3 20.5 3.4 2.1 4.7 3.8 3.0 12.5 10.6 3.8 × 105 373 819 Thermal Cross Section σγ σ el Source: Data from http://www.nndc.bnl.gov/. σf Nuclide 774 278 303 573 – 2 8.9 5.6 – – – – – – – – – – – 138 133 182 180 84 278 8103 1130 0.15 3 × 10−4 0.22 0.002 6 × 10−4 0.31 1.4 6.9 7.6 × 103 3.5 × 103 761 Resonance Cross Section σf σγ Table 1.3 Spectrum-Averaged Thermal, Resonance, and Fast Neutron Cross Sections (barns) 1.9 1.2 1.8 1.6 0.08 0.31 1.4 1.1 – – – – – – – – – – – σf 0.07 0.09 0.05 0.12 0.09 0.07 0.09 0.09 4 × 10−5 7 × 10−6 8 × 10−5 2 × 10−5 9 × 10−5 2 × 10−4 3 × 10−3 0.01 0.01 0.22 0.11 4.4 4.6 4.4 5.2 4.6 4.8 4.3 4.8 3.9 2.5 2.1 2.3 2.7 2.7 3.0 5.0 4.9 4.6 4.7 1.2 1.8 1.5 0.9 2.9 2.6 2.0 1.9 – – 0.07 0.01 – 0.5 0.7 0.7 1.0 2.2 2.2 Fission Spectrum Cross Section σγ σ el σ in 4 × 10−3 12 × 10−3 4 × 10−3 21 × 10−3 14 × 10−3 12 × 10−3 4 × 10−3 7 × 10−3 – – – – – – – – – – 11 × 10−3 σ n,2n 1.5 Evaluated Nuclear Data Files 25 26 1 Neutron Nuclear Reactions Fig. 1.28 Low-energy absorption (fission + capture) cross sections for several important nuclides. (From Ref. 12; used with permission of Wiley.) most comprehensive computerized compilation of experimental data is the EXFOR computer library (Ref. 11). The computerized card index file CINDA (Ref. 8), which contains comprehensive information on measurements, calculations, and evaluations of neutron–nuclear data, is updated annually. The plethora of sometimes contradictory nuclear data must be evaluated before it can be used confidently in reactor physics calculations. Such evaluation consists of intercomparison of data, use of data to calculate benchmark experiments, critical assessment of statistical and systematic errors, checks for internal consistency and consistency with standard neutron cross sections, and the derivation of consistent preferred values by appropriate averaging procedures. Several large evaluated nuclear data files are maintained: 1.6 Elastic Scattering Kinematics Fig. 1.29 Scattering event in lab and CM systems. (From Ref. 12; used with permission of Wiley.) (1) United States Evaluated Nuclear Data File (ENDF/B), (2) Evaluated Nuclear Data Library of the Lawrence Livermore National Laboratory (ENDL), (3) United Kingdom Nuclear Data Library (UKNDL), (4) Japanese Evaluated Nuclear Data Library (JENDL), (5) Karlsruhe Nuclear Data File (KEDAK), (6) Russian (formerly Soviet) Evaluated Nuclear Data File (BROND), and (7) Joint Evaluated File of NEA Countries (JEF). Processing codes are used to convert these data to a form that can be used in reactor physic calculations, as discussed in subsequent chapters. 1.6 Elastic Scattering Kinematics Consider a neutron with energy EL = 12 mv2L in the laboratory (L) system incident upon a stationary nucleus of mass M. Since only the relative masses are important in the kinematics, we set m = 1 and M = A. It is convenient to convert to the center-of-mass (CM) system, as indicated in Fig. 1.29, because the elastic scattering event is usually isotropic in the CM system. 27 28 1 Neutron Nuclear Reactions The velocity of the CM system in the L system is vcm = 1 vL (vL + AVL ) = 1+A 1+A (1.13) and the velocities of the neutron and the nucleus in the CM system are A vL A+1 −1 vL Vc = −vcm = A+1 vc = vL − vcm = (1.14) The energy of the neutron in the CM system, Ec , is related to the energy of the neutron in the lab, EL , by 1 1 A 1 2 A v = EL Ec = v2c + AVc2 = 2 2 A+1 2 L A+1 (1.15) Correlation of Scattering Angle and Energy Loss From consideration of conservation of momentum and kinetic energy, it can be shown that the speeds of the neutron and the nucleus in the center-of-mass system do not change during the scattering event: vc = vc = A vL A+1 Vc = Vc = (1.16) −1 vL A+1 With reference to Fig. 1.30, the scattering angles in the lab and CM systems are related by tan θL = vc sin θc sin θc = vcm + vc cos θc (1/A) + cos θc (1.17) The law of cosines yields cos(π − θc ) = (vc )2 + (vcm )2 − (vL )2 2vcm vc Fig. 1.30 Relation between lab and CM scattering angles. (From Ref. 12; used with permission of Wiley.) (1.18) 1.6 Elastic Scattering Kinematics which may be combined with Eqs. (1.13) and (1.16) to obtain a relationship between the incident and final energies of the neutron in the lab system and the scattering angle in the CM system: 1  2 2 m(vL ) 1 2 2 m(vL ) ≡ EL A2 + 1 + 2A cos θc (1 + α) + (1 − α) cos θc = = EL 2 (A + 1)2 (1.19) where α ≡ (A − 1)2 /(A + 1)2 . Average Energy Loss Equation (1.19) states that the ratio of final to incident energies in an elastic scattering event is correlated to the scattering angle in the CM system, which in turn is correlated via Eq. (1.17) to the scattering angle in the lab system. The maximum energy loss (minimum value of EL /EL ) occurs for θc = π (i.e., backward scattering in the CM system), in which case EL = αEL . For hydrogen (A = 1), α = 0 and all of the neutron energy can be lost in a single collision. For other nuclides, only a fraction (1 − α) of the neutron energy can be lost in a single collision, and for heavy nuclides (α → 1) this fraction becomes very small. The probability that a neutron scatters from energy EL to within a differential band of energies dEL about energy EL is equivalent to the probability that a neutron scatters into a cone 2π sin θc dθc about θc : σs (EL )P (EL → EL )dEL = −σcm (EL , θc )2π sin θc dθc (1.20) where the negative sign takes into account that an increase in angle corresponds to a decrease in energy, σs is the elastic scattering cross section, and σcm (θc ) is the cross section for scattering through angle θc . Using Eq. (1.19) to evaluate dEL /dθc , this becomes  4πσcm (EL , θc )   (1.21) P (EL → EL ) = (1 − α)EL σs (EL ) , αEL ≤ EL ≤ EL 0, otherwise Except for very high energy neutrons scattering from heavy mass nuclides, elastic scattering in the CM is isotropic, σcm (θc ) = σs /4π . In this case, Eq. (1.21) may be written σs (EL → EL ) ≡ σs (EL )P (EL → EL ) = = 0, σs (EL ) , (1 − α)EL αEL ≤ EL ≤ EL (1.22) otherwise The average energy loss in an elastic scattering event may be calculated from  EL ≡ EL − EL 1 dEL EL P (EL → EL ) = (1 − α)EL 2 αEL (1.23) 29 30 1 Neutron Nuclear Reactions Table 1.4 Number of Collisions, on Average, to Moderate a Neutron from 2 MeV to 1 eV Moderator ξ H D H2 O D2 O He Be C Na Fe 238 U 1.0 0.725 0.920 0.509 0.425 0.209 0.158 0.084 0.035 0.008 Number of Collisions 14 20 16 29 43 69 91 171 411 1730 ξ  s / a – – 71 5670 83 143 192 1134 35 0.0092 and the average logarithmic energy loss may be calculated from  EL P (EL → EL ) EL αEL   (A − 1)2 A+1 α ln α = 1 − ln = 1+ 1−α 2A A−1 ξ≡  EL dEL ln  (1.24) The number of collisions, on average, required for a neutron of energy E0 to be moderated to thermal energies, say 1 eV, can be estimated from no. collisions  ln[E0 (eV )/1.0] ξ (1.25) The results are shown in Table 1.4 for E0 = 2 MeV. The parameter ξ , which is a measure of the moderating ability, decreases with nuclide mass, with the result that the number of collisions that are needed to moderate a fast neutron increases with nuclide mass. However, the effectiveness of a nuclide (or molecule) in moderating a neutron also depends on the relative probability that a collision will result in a scattering reaction, not a capture reaction, which would remove the neutron. Thus the parameter ξ s /a , referred to as the moderating ratio, is a measure of the effectiveness of a moderating material. Even though H2 O is the better moderator in terms of the number of collisions required to thermalize a fast neutron, D2 O is the more effective moderator because the absorption cross section for D is much less than that for H. Example 1.2: Moderation by a Mixture. The moderating parameters for a mixture of isotopes is constructed by weighting the moderating parameters of the individual isotopes by their concentrations in the mixture. For example, in a mixture of 12 C and 238 U the average value of ξ s = NC ξC σsC + NU ξU σsU = NC (0.158) (2.3 × 10−24 cm2 ) + NU (0.008)(4.8 × 10−24 cm2 ), where the fission spectrum range elastic scattering cross sections of Table 1.3 have been assumed to hold also in the References slowing-down range. The total absorption cross section is a = NC σaC + NU σaU = NC (0.002 × 10−24 cm2 ) + NU (280 × 10−24 cm2 ) in the slowing-down range, where the resonance range cross sections from Table 1.3 have been used. References 1 H. Cember, Introduction to Health Physics, 3rd ed., McGraw-Hill, New York (1996). 2 C. Nordborg and M. Salvatores, “Status of the JEF Evaluated Nuclear Data Library,” Proc. Int. Conf. Nuclear Data for Science and Technology, Gatlinburg, TN, Vol. 2 (1994), p. 680. 3 R. W. Roussin, P. G. Young, and R. McKnight, “Current Status of ENDF/B-VI, Proc. Int. Conf. Nuclear Data for Science and Technology, Gatlinburg, TN, Vol. 2 (1994), p. 692. 4 Y. Kikuchi, “JENDL-3 Revision 2: JENDL 3-2,” Proc. Int. Conf. Nuclear Data for Science and Technology, Gatlinburg, TN, Vol. 2 (1994), p. 685. 5 R. A. Knief, Nuclear Engineering, Taylor & Francis, Washington, DC (1992). 6 J. J. Schmidt, “Nuclear Data: Their Importance and Application in Fission Reactor Physics Calculations,” in D. E. Cullen, R. Muranaka, and J. Schmidt, eds., Reactor Physics Calculations for Applications in Nuclear Technology, World Scientific, Singapore (1990). 7 A. Trkov, “Evaluated Nuclear Data Processing and Nuclear Reactor Calculations,” in D. E. Cullen, R. Muranaka, and J. Schmidt, eds., Reactor Physics Calculations for Applications in Nuclear Technology, World Scientific, Singapore (1990). 8 CINDA: An Index to the Literature on Microscopic Neutron Data, International Atomic Energy Agency, Vienna; CINDA-A, 1935–1976 (1979); CINDA-B, 1977–1981 (1984); CINDA-89 (1989). 9 D. E. Cullen, “Nuclear Cross Section Preparation,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations I, CRC Press, Boca Raton, FL (1986). 10 J. L. Rowlands and N. Tubbs, “The Joint Evaluated File: A New Nuclear Data Library for Reactor Calculations,” Proc. Int. Conf. Nuclear Data for Basic and Applied Science, Santa Fe, NM, Vol. 2 (1985), p. 1493. 11 A. Calamand and H. D. Lemmel, Short Guide to EXFOR, IAEA-NDS-1, Rev. 3, International Atomic Energy Agency, Vienna (1981). 12 J. J. Duderstadt and L. G. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976), Chap. 2. 13 H. C. Honeck, ENDF/B: Specifications for an Evaluated Data File for Reactor Applications, USAEC report BNL-50066, Brookhaven National Laboratory, Upton, NY (1966). 14 I. Kaplan, Nuclear Physics, 2nd ed., Addison-Wesley, Reading, MA (1963). 15 L. J. Templin, ed., Reactor Physics Constants, 2nd ed., ANL-5800, Argonne National Laboratory, Argonne, IL (1963). Problems 1.1. Demonstrate that the speeds of the neutron and nucleus in the CM system do not change in an elastic scattering event by using conservation of momentum and kinetic energy. 31 32 1 Neutron Nuclear Reactions 1.2. Estimate the probability that a 1-MeV neutron will be moderated to thermal without being captured in a mixture of uranium and water with NH /NU = 1:1. Repeat for a 1:1 mixture of uranium and carbon. 1.3. Neutrons are slowed down to thermal energies in a 1:1 mixture of H2 O and 4% enriched uranium (4%235 U, 96%238 U). Estimate the thermal value of η = νσf /(σc + σf ). Repeat the calculation for a mixture of (2%235 U, 2%239 Pu, 96%238 U). 1.4. Estimate the probability that a fission neutron will have a scattering collision with H2 O in the mixtures of Problem 1.3. 1.5. Calculate the average energy loss for neutrons at 1-MeV, 100-keV, 10-keV, and 1-keV scattering from carbon. Repeat the calculation for scattering from iron and from uranium. 1.6. Repeat Problem 1.5 for scattering from hydrogen and sodium. 1.7. Calculate the moderating ratio and the average number of collisions required to moderate a fission neutron to thermal for a 1:1 mixture of 12 C : 238 U. Repeat for a 10:1 mixture. 1.8. Calculate the thermal absorption cross section for a 1:1 wt% mixture of carbon and 4% enriched uranium (e.g., 4%235 U, 96%238 U). 1.9. Derive Eq. (1.21) from Eqs. (1.20) and (1.19). 1.10. Calculate the average number of scattering events required to moderate a neutron’s energy from above the resonance range to below the resonance range of 238 U for carbon, H2 O and D2 O moderators. 33 2 Neutron Chain Fission Reactors 2.1 Neutron Chain Fission Reactions Since two or three neutrons are released in every neutron-induced fission reaction, the possibility of a sustained neutron chain reaction is obvious, as illustrated in Fig. 2.1. To sustain a fission chain reaction, one or more of the neutrons produced in the fission event must, on average, survive to produce another fission event. There is competition for the fission neutrons in any assembly—some will be absorbed in fuel nuclides as radiative capture events rather than fission events, some will be absorbed by nonfuel nuclides, and some will leak out of the assembly. A scattering event does not compete for a neutron because the scattered neutron remains in the assembly and available for causing a fission event, but a scattering event does change a neutron’s energy and thus, because the various cross sections are energy dependent, does change the relative likelihood of the next collision being a fission event. Capture-to-Fission Ratio The fission cross sections for the fissile nuclides increase approximately as 1/υ with decreasing neutron energy, but then so do the capture cross sections of the fissile nuclides. The probability that a neutron that is captured in a fissile nuclide causes a fission is just σf /(σf + σγ ) = 1/(1 + σγ /σf ) = 1/(1 + α), where α ≡ σγ /σf is referred to as the capture-to-fission ratio. The capture-to-fission ratio for the principal fissile nuclides decreases as the neutron energy increases. For high neutron energies, the fission probability, which varies as (1 + α)−1 , is larger for 239 Pu than for 235 U or 233 U, but the situation is reversed for low-energy thermal neutrons. Number of Fission Neutrons per Neutron Absorbed in Fuel The product of the fission probability for a neutron absorbed in the fuel and the average number of neutrons released per fission, η ≡ νσf /(σf + σγ ) = ν/(1 + α), provides a somewhat better characterization of the relative capabilities of the var- 34 2 Neutron Chain Fission Reactors Fig. 2.1 Schematic of a fission chain reaction. (From Ref. 3; used with permission of Wiley.) ious fissile nuclides to sustain a fission chain reaction. This quantity is plotted in Fig. 2.2 for the principal fissile nuclides. For high neutron energies, η is larger for 239 Pu than for 235 U or 233 U, but the situation is reversed for low-energy thermal neutrons. Neutron Utilization The probability that a neutron is absorbed in a fissile nuclide instead of being absorbed in another nuclide or leaking from the assembly is absorb fissile absorb fissile + absorb nonfissile + leak = absorb fissile 1 ≡ f PNL absorb total (1 + leak/absorb total) (2.1) where f is the fraction of the absorbed neutrons which are absorbed in the fissile nuclides, or the utilization: f= Nfis σafis Nfis σafis + Nother σaother (2.2) and PNL refers to the nonleakage probability. Since the absorption cross section, σa = σf + σγ , is much greater for thermal neutrons than for fast neutrons for the fissile nuclides, but comparable for fast and thermal neutrons for the nonfissile fuel nuclides and for structural nuclides, the utilization for a given composition is much greater for thermal neutrons than for fast neutrons (and, in fact, is usually referred to as the thermal utilization). Fast Fission The product ηf is the number of neutrons produced, on average, from the fission of fissile nuclides for each neutron absorbed in the assembly. There will also 2.1 Neutron Chain Fission Reactions Fig. 2.2 η for the principal fissile nuclides. (From Ref. 9; used with permission of Electric Power Research Institute.) be neutrons produced by the fission of the nonfissile fuel nuclides, mostly by fast neutrons. Defining the fast fission factor ε ≡ total fission neutron production rate/fission neutron production rate in fissile nuclides, ηf ε is the total number of fission neutrons produced for each neutron absorbed in the assembly, and ηf εPNL is the total number of fission neutrons produced, on average, for each neutron introduced into the assembly by a previous fission event. 35 36 2 Neutron Chain Fission Reactors Fig. 2.3 Neutron balance in a thermal neutron fission assembly. (From Ref. 1; used with permission of Taylor & Francis.) Resonance Escape The parameters ηf ε must be evaluated by averaging over the energy of the neutrons in the assembly, of course. When the neutron population consists predominantly of thermal neutrons, the thermal spectrum-averaged cross sections given in Table 1.3 may be used to estimate η and f , and the cross sections averaged over the fission spectrum may be used in estimating ε, which should now also include fast fission in the fissile nuclides. In this case, it is necessary to take into account separately the capture of fission neutrons while they are slowing down to the thermal energy range, predominantly by the capture resonances of the fuel nuclides. The probability that a neutron is not captured during the slowingdown process is referred to as the resonance escape probability and denoted p. This competition for neutrons is illustrated schematically in Fig. 2.3 (leakage is neglected). 2.2 Criticality 2.2 Criticality Effective Multiplication Constant The product ηf εpPNL is the total number of fission neutrons produced, on average, by one fast neutron from a previous fission event. This quantity is referred to as the effective multiplication constant of the assembly: k = ηf εpPNL ≡ k∞ PNL (2.3) where k∞ refers to the multiplication constant of an infinite assembly with no leakage. If exactly one neutron, on average, survives to cause another fission, a condition referred to as criticality (k = 1), the neutron population in the assembly will remain constant. If less than one neutron, on average, survives to produce another fission event, a condition referred to as subcriticality (k < 1), the neutron population in the assembly will decrease. If more than one fission neutron, on average, survives to cause another fission, a condition referred to as supercriticality (k > 1), the neutron population in the assembly will increase. The effective multiplication constant depends on the composition (k∞ ) and size (PNL ) of an assembly and on the arrangement of the materials within the assembly (f and p). The composition affects k both by the relative number of nuclides of different species that are present and by the determination of the neutron energy distribution, which determines the average cross sections for each nuclide. The arrangement of materials determines the spatial neutron distribution and hence the relative number of neutrons at the locations of the various nuclides. The fissile nuclide 235 U is only 0.72% of natural uranium. Fuel enrichment to achieve a higher fissile content, hence larger value of f , is a major means of increasing the multiplication constant. The number of fission neutrons produced for each neutron absorbed in fissile material, η, is significantly larger for fast neutrons than for thermal neutrons, because the capture-to-fission ratio is smaller and the number of neutrons per fission is larger. On the other hand, for a given fuel enrichment, the utilization, f , is greater for thermal neutrons than for fast neutrons because the absorption cross section is much greater for thermal neutrons than for fast neutrons for the fissile nuclides, but comparable for fast and thermal neutrons for the nonfissile fuel nuclides and for structural nuclides. On the whole, the amount of fissile material necessary to achieve a given value of the multiplication constant is substantially less in a fast neutron spectrum than in a thermal neutron spectrum. Effect of Fuel Lumping Lumping the fuel rather than distributing it uniformly can have a significant effect on the multiplication constant. For example, if natural uranium is distributed 37 38 2 Neutron Chain Fission Reactors uniformly in a graphite lattice, the values of the various parameters are η ≈ 1.33, f ≈ 0.9, ε ≈ 1.05, and p ≈ 0.7, yielding k∞ ≈ 0.88 (i.e., the assembly is subcritical). If the fuel is lumped, the strong resonance absorption at the exterior of the fuel elements reduces the number of neutrons that reach the interior of the fuel elements, increasing the resonance escape probability to p ≈ 0.9. Lumping the fuel also reduces the thermal utilization f , for the same reason, but the effect is not so significant. Lumping the fuel was the key to achieving criticality (k = 1) in the first graphite-moderated natural uranium reactors and is crucial in achieving criticality in present-day D2 O-moderated natural uranium reactors. Leakage Reduction The multiplication constant can be increased by reducing the leakage, most of which is due to fast neutrons. This can be done simply by increasing the size. The leakage can also be reduced by choosing a composition that moderates the neutrons quickly before they can travel far or by surrounding the assembly with a material with a large scattering cross section (e.g., graphite), which will reflect leaking neutrons back into the assembly. Example 2.1: Effective Multiplication Factor for a PWR. For a typical pressurized water reactor (PWR), the various parameters are η ≈ 1.65, f ≈ 0.71, ε ≈ 1.02, and p ≈ 0.87, yielding k∞ ≈ 1.04. The nonleakage factors for fast and thermal neutrons are typically 0.97 and 0.99, yielding k ≈ 1.00. 2.3 Time Dependence of a Neutron Fission Chain Assembly Prompt Fission Neutron Time Dependence If there are N0 fission neutrons introduced into an assembly at t = 0, and if l is the average time required for a fission neutron to slow down and be absorbed or leak out, l = ν1f υ , the number of neutrons, on average, in the assembly at time t = l is (k)N0 . Continuing in this fashion, the number of neutrons in the assembly at time t = ml (m integer) is (k)m N0 . The quantity l is typically ≈ 10−4 s for assemblies in which the neutrons slow down to thermal before causing another fission, and is typically ≈ 10−6 s for assemblies in which the fission is produced by fast neutrons. For example, a 12 % change in absorption cross section, which could be produced by control rod motion, causes an approximately 0.005 change in k. The neutron population after 0.1 s in a thermal assembly (0.1 s = 103 l) in which k = 1.005, would be N(0.1) = (1.005)1000 N0 ≈ 150N0 . In a thermal assembly with k = 0.995, the neutron population after 0.1 s would be N(0.1) = (0.995)1000 N0 ≈ 0.0066N0 . An equation governing the neutron kinetics described above is dN(t) k − 1 = N(t) + S(t) dt l (2.4) 2.3 Time Dependence of a Neutron Fission Chain Assembly which simply states that the time rate of change of the neutron population is equal to the excess of neutron production (by fission) minus neutron loss by absorption or leakage in a neutron lifetime plus any external source that is present. For a constant source, Eq. (2.4) has the solution N(t) = N(0)e(k−1)t/ l + Sl [e(k−1)t/ l − 1] k−1 (2.5) which displays an exponential time behavior. Using the same example as above, with the source set to zero, leads to N(0.1) = N(0) exp(5.0) = 148N(0) for k = 1.005 and N(0.1) = N(0) exp(−5.0) = 0.00677N(0) for k = 0.995. Source Multiplication Equation (2.4) does not have a steady-state solution for k > 1 and does not have a unique steady-state solution for k = 1. However, for k < 1, the asymptotic solution is Nasymptotic = lS0 1−k (2.6) This equation provides a method to measure the effective multiplication factor k when k < 1 by measuring the asymptotic neutron population which results from placing a source S0 in a multiplying medium. Effect of Delayed Neutrons It would be very difficult, if not impossible, to control a neutron fission chain assembly which responded so dramatically to a 12 % change in absorption cross section. Fortunately, a small fraction (β ≈ 0.0075 for 235 U fueled reactors) of the fission neutrons are delayed until the decay (λ ≈ 0.08 s−1 ) of the fission fragments. For an assembly that was critical prior to t = 0, the equilibrium concentration of such delayed neutron precursor fission fragments is found from the balance equation: dC0 β = 0 = βνNf σf υN0 − λC0 = N0 − λC0 dt l (2.7) where Nf is the density of fuel nuclei, N0 the neutron population, and C0 the population of delayed neutron precursor fission fragments. When the 12 % change in cross section occurs at t = 0, the multiplication of the prompt neutrons after 0.1 s (1000l) is [(1 − β)k]1000 . During each multiplication interval l there is a source λlC of delayed neutrons from the decay of fission fragments. This source results in (1 − β)kλlC neutrons in the following multiplication interval, [(1 − β)k]2 λlC neutrons in the second following multiplication interval, and so on. There is such a delayed neutron source in each of the 1000 multiplication intervals in our example. To simplify the problem, we assume that the fission 39 40 2 Neutron Chain Fission Reactors fragment concentration does not change (i.e., C = C0 ). Thus the number of neutrons after 0.1 s (1000l) is N(1000l) = [(1 − β)k]m N0 + λlC0 [(1 − β)k]m−1 + λlC0 [(1 − β)k]m−2 + · · · + λlC0 (1 − β)k + λlC0 = [(1 − β)k]m N0 + λlC0   = [(1 − β)k]m 1 − (1 − k[1 − β])m−1 1 − k(1 − β)  β β + N0 k(1 − β)[1 − k(1 − β)] 1 − k(1 − β) (2.8) where Eq. (2.7) has been used in the last step. Evaluating this expression for k = 1.005 yields N(t = 0.1 s) = 3.03N0 , instead of the 150N0 found without taking the delayed neutrons into account. If we had taken into account the changing fission fragment population, we would have found a slightly larger number. Nevertheless, the fact that some of the neutrons emitted in fission are delayed results in a rather slow and hence controllable response of a neutron chain fission reacting assembly, provided that (1 − β)k < 1. 2.4 Classification of Nuclear Reactors Physics Classification by Neutron Spectrum From the physics viewpoint, the main differences among reactor types arise from differences in the neutron energy distribution, or spectrum, which causes differences in the neutron–nuclear reaction rates and the competition for neutrons. The first level of physics classification categories are then thermal reactors and fast reactors, corresponding to the majority of the neutron–nuclear reactions involving neutrons in the thermal energy range (E < 1 eV) and to the majority of the neutron–nuclear reactions involving neutrons in the fast energy range (E > 1 keV), respectively. Representative neutron spectra for thermal (LWR) and fast (LMFBR) reactor cores are shown in Fig. 2.4. There are important physics differences among the different thermal reactors and among the different fast reactors, but these differences are not so great as the physics differences between a thermal reactor and a fast reactor. The capture-tofission ratio, α, is lower and the number of neutrons produced per fission, ν, is larger in fast reactors than in thermal reactors. This generally results in a larger value of k for a given amount of fuel in a fast reactor than in a thermal reactor, or, more to the point, a smaller critical mass of fuel in a fast reactor than in a thermal reactor. Because of the larger neutron–nuclear reaction rates for thermal neutrons than for fast neutrons, the mean distance that a neutron travels before absorption is greater in a fast reactor than in a thermal reactor. This implies that the detailed 2.4 Classification of Nuclear Reactors Fig. 2.4 Representative fast (LMFBR) and thermal (LWR) reactor neutron energy distributions. Flux ≡ nv (From Ref. 1; used with permission of Taylor & Francis.) distribution of fuel, coolant, and control elements has a much greater effect on the local competition for neutrons in a thermal reactor than in a fast reactor and that the neutron populations in the different regions of the core are more tightly coupled in a fast reactor than in a thermal reactor. Engineering Classification by Coolant The neutron spectrum is determined primarily by the principal neutron moderating material present, and in many cases this material is the coolant. Because the heat transport system is such a major aspect of a nuclear reactor, it is also common to classify reactors according to coolant. Water-cooled reactors, such as the pressurized water (PWR) and boiling water (BWR) reactors, which use H2 O coolant, and the pressurized heavy water reactor (PHWR), which uses D2 O coolant, have thermal neutron spectra because of the excellent moderating properties of hydrogen. Since gas is too diffuse to serve as an effective moderator, gas-cooled reactors can be either thermal or fast, depending on whether or not a moderator, commonly graphite, is included. The early Magnox and subsequent advanced gas reactors (AGR) are cooled with CO2 , and the advanced high-temperature gas-cooled reactor (HTGR) is cooled with helium; all are moderated with graphite to achieve a thermal spectrum. Designs have been developed for a helium-cooled reactor without graphite, which is known as the gas-cooled fast reactor (GCFR). The pressure tube graphite-moderated reactor (PTGR) is cooled with pressurized or boiling water in pressure tubes, but it is necessary to include graphite to achieve a thermal spectrum. The molten salt breeder reactor (MSBR) employs a molten salt fluid which acts as both the fuel and the primary coolant loop, and is moderated by graphite to achieve a thermal spectrum. The advanced liquid-metal reactor (ALMR) and the liquid-metal fast breeder reactor (LMFBR) are cooled with sodium, which is not a particularly effective moderator, and the neutron spectrum is fast. 41 42 2 Neutron Chain Fission Reactors References 1 R. A. Knief, Nuclear Engineering, 2nd ed., Taylor & Francis, Washington, DC (1992). 2 J. R. Lamarsh, Introduction to Nuclear Reactor Theory, 2nd ed., AddisonWesley, Reading, MA (1983). 3 J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976). 4 A. F. Henry, Nuclear-Reactor Analysis, MIT Press, Cambridge, MA (1975). 5 G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand Reinhold, New York (1970). 6 R. V. Meghreblian and D. K. Holmes, Reactor Analysis, McGrawHill, New York (1960), pp. 160–267 and 626–747. 7 A. M. Weinberg and E. P. Wigner, The Physical Theory of Neutron Chain Reactors, University of Chicago Press, Chicago (1958). 8 S. Glasstone and M. C. Edlund, Nuclear Reactor Theory, D. Van Nostrand, Princeton, NJ (1952). 9 N. L. Shapiro et al., Electric Power Research Institute Report, EPRI-NP-359, Electric Power Research Institute, Palo Alto, CA (1977). Problems 2.1. Calculate and plot the thermal value of η for a uranium-fueled reactor as a function of enrichment (e.g., percentage 235 U in uranium) over the range 0.07 to 5.0%. 2.2. Calculate the thermal utilization in a homogeneous 50:50 volume % mixture of carbon and natural uranium. Repeat the calculation for 4% enriched uranium. 43 3 Neutron Diffusion Theory In this chapter we develop a one-speed diffusion theory mathematical description of nuclear reactors. Such a relatively simple description has the great advantage of illustrating many of the important features of nuclear reactors without the complexity that is introduced by the treatment of important effects associated with the neutron energy spectrum and with highly directional neutron transport, which are the subjects of subsequent chapters. Moreover, diffusion theory is sufficiently accurate to provide a quantitative understanding of many physics features of nuclear reactors and is, in fact, the workhorse computational method of nuclear reactor physics. 3.1 Derivation of One-Speed Diffusion Theory Calculation of the rates of the different reactions of neutrons with the materials in the various parts of a nuclear reactor is the fundamental task of nuclear reactor physics. This calculation requires a knowledge of nuclear cross sections and their energy dependence (Chapter 1) and of the distribution of neutrons in space and energy throughout the reactor. The neutron distribution depends on the neutron source distribution, which in the case of the fission source depends on the neutron distribution itself, and on the interactions with atomic nuclei experienced by the neutrons as they move away from the source. The simplest and most widely used mathematical description of the neutron distribution in nuclear reactors is provided by neutron diffusion theory. For simplicity of explication, the neutrons are treated as if they are all of one effective speed, and effects associated with changes in neutron energy are suppressed. Such a simplification would be justified in practice if the cross sections were averaged over the appropriate neutron energy distribution. As a further simplification, the medium is initially assumed to be uniform. Partial and Net Currents With respect to Fig. 3.1, the rate at which neutrons are scattering in the differential volume element dr = r 2 dr dμ dψ is s φ dr, where μ ≡ cos θ , the macroscopic 44 3 Neutron Diffusion Theory Fig. 3.1 Definition of coordinate system. (From Ref. 10; used with permission of McGraw-Hill.) cross section  ≡ Nσ is the product of the number density of atomic nuclei and the microscopic cross section discussed previously, and the neutron scalar flux φ ≡ nυ is the product of the neutron density and the neutron speed. The fraction of the isotropically scattered neutrons leaving dr headed toward the differential area dA at the origin is −(r/r) · dA/4πr 2 = μdA/4πr 2 . Not all of these neutrons reach dA, however; some are absorbed and others are scattered again so that they do not cross dA. The probability that a neutron leaving dr in the direction of dA actually reaches dA is e−r . The differential current j− (0 : r, μ, ψ)dr dA of neutrons passing downward through dA which had their last scattering collision in dr is thus j− (0 : r, μ, ψ)dr dA = μe−r s φ(r, μ, ψ)dr dA 4πr 2 (3.1) The total current passing downward through dA is found by integrating this expression over the entire upper half-plane (x > 0):  j− (0)dA = s  ∞ dr 0  2π dψ 0 1 dμ μe−s r φ(r, μ, ψ) 0 dA 4π (3.2) Now, the first major approximation leading to diffusion theory is made—for the purpose of evaluating the integral in Eq. (3.2), the flux is assumed to be sufficiently slowly varying in space that it can be approximated by expansion in a Taylor series about the origin: φ(r) = φ(0) + r · ∇φ(0) + 1 2 2 r ∇ φ(0) + · · · 2 (3.3) 3.1 Derivation of One-Speed Diffusion Theory in which only the first two terms are retained. Using this approximation and the trigonometric identity cos β = cos θx cos θ +sin θx sin θ cos(ψy −ψ), and making the second major approximation—that absorption is small relative to scattering (e.g.,  ∼ s )—Eq. (3.2) can be integrated to obtain the diffusion theory expression for the partial downward current density: j− (0) = ≡ 1 1 dφ(0) 1 1 φ(0) + |∇φ(0)| cos θx = φ(0) + 4 6s 4 6s dx 1 dφ(0) 1 φ(0) + D 4 2 dx (3.4) A similar derivation leads to an expression for the partial upward current density, 1 dφ(0) 1 j+ (0) = φ(0) − D 4 2 dx (3.5) where D is known as the diffusion coefficient. The diffusion theory expression for the net current at the origin (positive sign up) is Jx (0) = j+ (0) − j− (0) = dφ(0) 1 dφ(0) = −D 3s dx dx (3.6) Carrying out a similar derivation for dA in the x–y and x–z planes leads immediately to the three-dimensional generalization J(0) = − 1 ∇φ(0) ≡ −D∇φ(0) 3s (3.7) A third assumption—that the neutrons are scattered isotropically—was used in the derivation above. The last form of Eq. (3.7) is known as Fick’s law, which governs the diffusion of many other quantities as well as neutrons. A more accurate derivation of diffusion theory from transport theory (Section 3.12) reveals that a better approximation for the diffusion coefficient which takes into account anisotropy in scattering is given by D= 1 1 ≡ 3(t − μ̄0 s ) 3tr (3.8) 2 where t and s are the total and scattering cross sections and μ̄0 ≈ 3A is the average cosine of the scattering angle (A is the atomic mass number of the scattering nuclei). Diffusion Theory The mathematical formulation of neutron diffusion theory is then obtained by using the diffusion theory expression for the neutron current in the neutron balance 45 46 3 Neutron Diffusion Theory equation on a differential volume element: ∂n = S + νf φ − a φ − ∇ · J ∂t = S + νf φ − a φ + ∇ · D∇φ (3.9) which states that the time rate of change of the neutron density within a differential volume is equal to the external rate at which neutrons are produced in the volume by an external source (S) and by fission (νf φ) minus the rate at which neutrons are lost by absorption (a φ) and minus the net leakage of neutrons out of the volume (∇ · J). Proof that the net leakage out of a differential volume element is ∇ · J follows from considering the difference of outward and inward currents in a cube of dimensions x y z . The net transport of particles out of the cube is [Jx (x + x) − Jx (x)]y z + [Jy (y + y ) − Jy (y)]x z + [Jz (z + z )       ∂Jy ∂Jx ∂Jz  x y  z + y x  z + y  x  y − Jz (z)]x y  ∂x ∂y ∂z ≡ ∇ · Jx y z where a Taylor’s series expansion of the current has been made. Interface Conditions At an interface between regions 1 and 2 at which there is an isotropic source S0 , the partial currents on both side of the interface must be related by (2) (1) j+ (0) = 12 S0 + j+ (0) j−(1) (0) = 12 S0 + j−(2) (0) (3.10) Subtracting these two equations and using Eqs. (3.4) and (3.5) yields an interface condition of continuity of neutron flux: φ2 (0) = φ1 (0) (3.11) Adding the two equations yields J2 (0) = J1 (0) + S0 (3.12) which, in the absence of an interface source, is a continuity of neutron net current condition. Boundary Conditions At an external boundary, the appropriate boundary condition is found by equating the expression for the inward partial current to the known incident current, j in , for example, from the right at xb , 1 1 dφ(xb ) j−in = φ(xb ) + D 4 2 dx (3.13) 3.1 Derivation of One-Speed Diffusion Theory When the diffusing medium is surrounded by a vacuum or nonreflecting region, j in = 0 and Eq. (3.13) may be written 1 dφ φ dx =− xb 3tr 1 =− 2D 2 (3.14) A widely used but more approximate vacuum boundary condition is obtained by noting that this expression relates the flux and the flux slope at the boundary. If the slope of the flux versus x at the boundary (xb ) is used to extrapolate the flux outside the boundary, the extrapolated flux will vanish at a distance λextrap = 2 2 −1 3 λtr = 3 tr outside the external boundary. A more accurate result from neutron transport theory is λextrap = 0.7104λtr . This result gives rise to the approximate vacuum boundary condition of zero neutron flux at a distance λextrap outside the physical boundary at x = a, or φ(a + λextrap ) ≡ φ(aex ) = 0, where we have defined the extrapolated boundary aex ≡ a + λextrap (3.15) Since λextrap is usually very small compared to the typical dimensions of a diffusing medium encountered in reactor physics, it is common to use the even more approximate vacuum boundary condition of zero flux at the physical external boundary. Example 3.1: Typical Values of Thermal Extrapolation Distance. The thermal neutron extrapolation distance λextrap = 0.7104/tr = 0.7104/[a + (1 − μ0 )s ] for some typical diffusing media are 0.30 cm for H2 O, 1.79 cm for D2 O, 1.95 cm for C, and 6.34 cm for Na. The approximation that the neutron flux vanishes at the boundary of the diffusing medium is valid when the dimension L of the diffusing medium is much larger than the extrapolation distance, L  λextrap . Applicability of Diffusion Theory Diffusion theory provides a strictly valid mathematical description of the neutron flux when the assumptions made in its derivation—absorption much less likely than scattering, linear spatial variation of the neutron distribution, isotropic scattering—are satisfied. The first condition is satisfied for most of the moderating (e.g., water, graphite) and structural materials found in a nuclear reactor, but not for the fuel and control elements. The second condition is satisfied a few mean free paths away from the boundary of large (relative to the mean free path) homogeneous media with relatively uniform source distributions. The third condition is satisfied for scattering from heavy atomic mass nuclei. One might well ask at this point how diffusion theory can be used in reactor physics when a modern nuclear reactor consists of thousands of small elements, many of them highly absorbing, with dimensions on the order of a few mean free paths or less. Yet diffusion theory is widely used in nuclear reactor analysis and makes accurate predictions. The secret is that a more accurate transport theory is used to “make diffusion theory 47 48 3 Neutron Diffusion Theory work” where it would be expected to fail. The many small elements in a large region are replaced by a homogenized mixture with effective averaged cross sections and diffusion coefficients, thus creating a computational model for which diffusion theory is valid. Highly absorbing control elements are represented by effective diffusion theory cross sections which reproduce transport theory absorption rates. 3.2 Solutions of the Neutron Diffusion Equation in Nonmultiplying Media Plane Isotropic Source in an Infinite Homogeneous Medium Consider an infinite homogeneous nonmultiplying (f = 0) medium in which a plane isotropic source (infinite in the y–z plane) with strength S0 is located at x = 0. Everywhere except at x = 0 the time-independent diffusion equation can be written d 2 φ(x) 1 − 2 φ(x) = 0 dx 2 L (3.16) where L2 ≡ D/a is the neutron diffusion length squared. This equation has a general solution φ = A exp(x/L) + B exp(−x/L). For x > 0, the physical requirement for a finite solution at large x requires that A = 0, and the physical requirement that the net current must approach 12 S0 as x approaches 0 requires that B = LS0 /2D. Following a similar procedure for x < 0 leads to similar results, so that the solution may be written φ(x) = S0 Le−|x|/L 2D (3.17) Plane Isotropic Source in a Finite Homogeneous Medium Consider next a finite slab medium extending from x = 0 to x = +a with an isotropic plane source at x = 0. In this case, the general solution of Eq. (3.16) is more conveniently written as φ = A sinh(x/L) + B cosh(x/L). The appropriate boundary conditions are that the inward partial current vanishes at x = a [i.e., j − (a) = 0] and that the outward partial current equals 12 the isotropic source strength as x → 0 [i.e., j + (0) = 12 S0 ]. The resulting solution is φ(x) = 4S0 sinh[(a − x)/L] + (2D/L) cosh[(a − x)/L] [2(D/L) + 1]2 ea/L − [2(D/L) − 1]2 e−a/L (3.18) If instead of j − (a) = 0, the extrapolated boundary condition φ(aex ) = 0 is used, the resulting solution is φ(x) = 4S0 sinh[(aex − x)/L] sinh(aex /L) + (2D/L) cosh(aex /L) (3.19) 3.2 Solutions of the Neutron Diffusion Equation in Nonmultiplying Media When 0.71λtr /a  1 and 2(a /3tr )1/2  1 (i.e., when the transport mean free path is small compared to the dimension of the medium and the absorption cross section is small relative to the scattering cross section), these two solutions agree. These conditions must also be satisfied in order for diffusion theory to be valid, so we conclude that use of the extrapolated zero flux boundary condition instead of the zero inward current boundary condition is acceptable. Line Source in an Infinite Homogeneous Medium Consider an isotropic line source (e.g., infinite along the z-axis) of strength S0 (per centimeter per second) located at r = 0. The general solution of   1 d dφ(r) 1 (3.20) r − 2 φ(r) = 0 r dr dr L is φ = AI0 (r/L) + BK0 (r/L), where I0 and K0 are the modified Bessel functions of order zero of the first and second kind, respectively. The physical requirement for a finite solution at large r requires that A = 0. The source condition is limr→0 2πrJ = S0 . The resulting solution for an isotropic line source in an infinite homogeneous nonmultiplying medium is φ(r) = S0 K0 (r/L) 2πD (3.21) Homogeneous Cylinder of Infinite Axial Extent with Axial Line Source Consider an infinitely long cylinder of radius a with an isotropic source on axis. The source condition limr→0 2πrJ = S0 still obtains, but now A = 0 no longer holds and the other boundary condition is a zero incident current condition at r = a or a zero flux condition at r = a + λextrap . The latter vacuum boundary condition leads to the solution for the neutron flux distribution in an infinite homogeneous nonmultiplying cylinder with an isotropic axial line source: φ(r) = S0 [I0 (aex /L)K0 (r/L) − K0 (aex /L)I0 (r/L)] 2πDI0 (aex /L) (3.22) Point Source in an Infinite Homogeneous Medium The neutron diffusion equation in spherical coordinates is   1 d 2 dφ(r) 1 r − 2 φ(r) = 0 dr r 2 dr L (3.23) This equation has the general solution φ = (Aer/L + Be−r/L )/r. The source condition is limr→0 4πr 2 J = S0 , and the physical requirement for a finite solution at large r requires that A = 0, yielding φ(r) = S0 e−r/L 4πrD (3.24) 49 50 3 Neutron Diffusion Theory Point Source at the Center of a Finite Homogeneous Sphere Consider a finite sphere of radius a with a point source at the center. The same general solution φ = (Aer/L + Be−r/L )/r of Eq. (3.23) is applicable, but the A = 0 condition must be replaced by a vacuum boundary condition at r = a. Using an extrapolated zero flux condition yields φ(r) = S0 sinh[(aex − r)/L] 4πrD sinh(aex /L) (3.25) for the neutron distribution in a finite sphere of homogeneous nonmultiplying material with a point source at the center. 3.3 Diffusion Kernels and Distributed Sources in a Homogeneous Medium Infinite-Medium Diffusion Kernels The previous solutions for plane, line, and point sources at the origin of slab, cylindrical, and spherical coordinate systems in an infinite medium can be generalized immediately to slab, line, and point sources located away from the origin (i.e., the location of the coordinate axis in an infinite medium can be offset without changing the functional form of the result). The resulting solutions for the neutron flux at location x or r due to a unit isotropic source at x  and r  may be thought of as kernels. The infinite-medium kernels for a plane isotropic source of one neutron per unit area per second, a point isotropic source of one neutron per second, a line isotropic source of one neutron per unit length, a cylindrical shell source of one neutron per shell per unit length per second, and a spherical shell source of one neutron per shell per second are Plane: φpl (x : x  ) = Line: φl (r : r  ) = L −|x−x  |L e 2D K0 (|r − r  |/L) 2πD Point: φpt (r : r ) =  e−|r−r |/L 4π|r − r |D Cylindrical shell: φcyl (r : r  ) = Spherical shell: φsph (r : r  ) =  1 K0 (r/L)I0 (r  /L), × K0 (r  /L)I0 (r/L), 2πD (3.26) r > r , r < r L   (e−|r−r |/L − e−|r+r |/L ) 8πrr  D These kernels may be used to construct the neutron flux in an infinite homogeneous nonmultiplying medium due to an arbitrary source distribution S0 :  (3.27) φ(r) = φ(r : r )S0 (r )dr 3.3 Diffusion Kernels and Distributed Sources in a Homogeneous Medium For a planar source distribution this takes the form  φ(x) = ∞ −∞ S0 (x  )L −|x−x  |/L  dx e 2D (3.28) and for the more general point source,  ∞ φ(r) = S0 (r ) 0  e−|r−r |/L dr 4π|r − r |D (3.29) Finite-Slab Diffusion Kernel Consider a slab infinite in the y- and z-directions extending from x = −a to x = +a with a unit isotropic source at x  . The neutron diffusion equation d 2 φ(x) 1 − 2 φ(x) = 0 2 dx L (3.30) holds everywhere in −a < x < +a except at x = x  , the source plane. The interface conditions at the source plane, x = x  , are, from Eqs. (3.11) and (3.12), φ(x  + ε) = φ(x  − ε) (3.31) J (x  + ε) = J (x  − ε) + 1 where x  + ε indicates an infinitesimal distance to the right of x  , and so on. For the vacuum boundary conditions at x = −a and x = a we use the approximate zero flux conditions φ(−a) = φ(a) = 0 (3.32) Solving Eq. (3.30) as before and using these source and boundary conditions yields the following expressions for the flux at x due to a unit isotropic source at x  , or the finite-slab diffusion kernel: sinh[(a + x  )/L] sinh[(a − x)/L] , (D/L) sinh(2a/L) x > x sinh[(a − x  )/L] sinh[(a + x)/L] φ− (x : x ) = , (D/L) sinh(2a/L)  φ+ (x : x  ) =  (3.33) x<x These kernels may be used to calculate the neutron flux distribution in the slab due to a distributed source, S0 (x  ):  φ(x) = x −a S0 (x  )φ+ (x : x  )dx  +  x a S0 (x  )φ− (x : x  )dx  (3.34) 51 52 3 Neutron Diffusion Theory Finite Slab with Incident Neutron Beam As a further relevant example, consider the first-collision source distribution in a slab due to a beam incident from the left at x = −a: S0 (x) = q0 s e−t (x+a) (3.35) Using this source in Eq. (3.34) yields the neutron flux distribution within the slab: φ(x) = q0 s e−t a − (1/L2 )) sinh(2a/L)        a−x a+x 2a t a −t a −t x × e sinh sinh sinh +e −e L L L D(t2 (3.36) By using a first-collision source, the highly anisotropic incident beam neutrons are treated by first-flight transport theory until they have had a scattering collision which (at least partially) converts the beam to a nearly isotropic neutron distribution which is amenable to treatment by diffusion theory. The solution for the nearly isotropic neutron distribution given by Eq. (3.36) has a maximum some distance into the slab at 0 > x > −a. 3.4 Albedo Boundary Condition Consider a slab that is infinite in the y- and z-directions located between x = 0 + . Upon solving for the and x = a with a known inward partial current j + (0) = jin neutron flux distribution for an extrapolated zero flux vacuum boundary condition φ(a + λextrap ) ≡ φ(aex ) = 0, it is possible to evaluate the reflection coefficient, or albedo, for neutrons entering the slab from the left at x = 0. α≡ j− (0) 1 − (2D/L) coth(aex /L) = j+ (0) 1 + (2D/L) coth(aex /L) (3.37) As a/L becomes large, coth[(a +λextrap )/L] → 1, and α → (1−2D/L)/(1+2D/L), the infinite-medium value. Now consider two adjacent slabs, one denoted B and located in the range −b ≤ x ≤ 0 and the other denoted A and located in the range 0 ≤ x ≤ a. If we are not interested in the neutron flux distribution in slab A but only in the effect of slab A on the neutron flux distribution in slab B, the albedo of slab A can be used as an albedo boundary condition for the neutron flux solution in slab B. From Eqs. (3.4) and (3.5),   dφB j+ (0) − j− (0) 1 1 − αA 1 (3.38) DB =− =− φB dx x=0 2[j+ (0) + j− (0)] 2 1 + αA 3.5 Neutron Diffusion and Migration Lengths This albedo boundary condition can also be simplified by a geometric interpretation. If the flux in slab B at the interface between slabs A and B (x = 0) is extrapolated (into slab A) to zero using the slope at the interface given by Eq. (3.38), an approximate albedo boundary condition for the flux solution in slab B (−b < x < 0) becomes φB (λalbedo ) = 0, where   1 + αA (3.39) λalbedo ≡ 0.71λB tr 1 − αA 3.5 Neutron Diffusion and Migration Lengths The distribution of neutrons within a finite or infinite medium is determined by the source distribution, the geometry (in a finite medium), and the neutron diffusion length, L = (D/a )1/2 . The (thermal) diffusion length is related to the meansquared distance that a thermal neutron travels from the source point to the point at which it is absorbed, as may be seen by computing the mean-squared distance to capture for (thermal) neutrons emitted by a point source in an infinite medium: r̄ 2 ≡ ∞ 2 2 0 r (4πr a φ)dr ∞ 2 0 (4πr a φ)dr = ∞ 3 −r/L dr 0 r e ∞ −r/L dr 0 re = 6L2 (3.40) where Eq. (3.24) has been used for the neutron flux due to a point source at r = 0. It is also apparent from the exp (±x/L) nature of many of the solutions above that L is the physical distance over which the neutron flux can change by a significant amount (i.e., e−1 ). Thermal Diffusion-Length Experiment The thermal neutron diffusion length can be determined experimentally by measuring the axial neutron flux distribution in a long (with respect to mean free path) block of material with an isotropic thermal neutron flux incident on one end (e.g., from the thermal column of a reactor). With reference to Fig. 3.2, consider a rectangular parallelepiped of length c and cross section 2a × 2b with an incident isotropic thermal neutron source S0 (x, y) at z = 0 which is symmetric in x and y about x = 0 and y = 0. The neutron flux in the block satisfies ∂ 2φ ∂ 2φ ∂ 2φ 1 + 2 + 2 − 2 =0 ∂x 2 ∂y ∂z L (3.41) and the boundary conditions 1 j+ (x, y, 0) = S0 (x, y) 2 φ(±aex , y, z) = 0 (3.42a) (3.42b) 53 54 3 Neutron Diffusion Theory Fig. 3.2 Geometry for diffusion-length experiment. (From Ref. 10.) φ(x, ±bex , z) = 0 (3.42c) φ(x, y, cex ) = 0 (3.42d) We seek a separable solution to Eq. (3.41) of the form φ(x, y, z) = X(x)Y (y)Z(z). Substitution of this form into Eq. (3.41) and division by XY Z yields X (x) Y  (y) Z  (z) 1 + + = 2 X(x) Y (y) Z(z) L (3.43) where the double prime indicates a second derivative with respect to the respective spatial variables. In general, this equation can only be satisfied if each of the terms on the left is separately equal to a constant: X (x) = −k12 , X(x) Y  (y) = −k22 , Y (y) Z  (z) = k32 Z(z) (3.44) in which case Eq. (3.43) becomes k32 = 1 + k12 + k22 L2 (3.45) The general solutions to Eqs. (3.44) are X(x) = A1 sin k1 x + C1 cos k1 x Y (y) = A2 sin k2 y + C2 cos k2 y Z(z) = A3 e −k3 z + C3 e (3.46) k3 z The x–y symmetry requirement determines that A1 = A2 = 0. The end condition of Eq. (3.42d) may be used to eliminate C3 to obtain Z(z) = A3 e−k3 z [1 − e−2k3 (cex −z) ] (3.47) 3.5 Neutron Diffusion and Migration Lengths The extrapolated boundary conditions of Eqs. (3.42b) and (3.42c) require that cos k1 aex = cos k2 bex = 0, which can only be satisfied if k1 and k2 have the discrete values π (2n + 1), 2aex π k2m = (2m + 1), 2bex k1n = n = 0, 1, . . . (3.48) m = 0, 1, . . . This result, together with Eq. (3.45), requires that k3 can only take on discrete values 2 2 2 = k1n + k2m + k32 → k3nm 1 L2 (3.49) Thus the most general solution of the neutron diffusion equation that satisfies the extrapolated boundary conditions of Eqs. (3.42b) to (3.42d) is φ(x, y, z) = ∞  Amn cos k1n x cos k2m ye−k3nm z [1 − e−2k3nm (cex −z) ] (3.50) n,m=0 where Amn is a constant that can be determined from Eq. (3.42a), but that is not necessary for our purposes. Noting that k3nm increases with m and n, the asymptotic form of the neutron flux distribution along the z-axis that persists at large distances from z = 0 is φ(0, 0, z)  A00 e−k300 z [1 − e−2k300 (cex −z) ] (3.51) For very long blocks (large cex ), the term in brackets is unimportant except near the end, and the flux decreases exponentially, so that a measurement of the axial flux distribution far away from both the source at z = 0 and the end at z = cex should provide for experimental determination of k300 . The diffusion length then is determined from     l π 2 π 2 2 = k − − (3.52) 300 2aex 2bex L2 The measured diffusion lengths L for thermal neutrons in H2 O, D2 O, and graphite are about 2.9, 170, and 60 cm, respectively. The implication of these measurements is that thermal neutrons would diffuse a root-mean-square distance from the point at which they appear (are thermalized) to the point at which they are absorbed of 7.1, 416, and 147 cm, respectively, in these three moderators. Migration Length In a water- or graphite-moderated reactor, the fission neutrons are born fast (average energy about 1.0 MeV) and diffuse as fast neutrons while they are in the process of slowing down to become thermal neutrons. In fast reactors, the neutrons are absorbed before thermalizing. In a later chapter we return to calculation 55 56 3 Neutron Diffusion Theory Table 3.1 Diffusion Parameters for Common Moderators Moderator Density (g/cm3 ) D (cm)  a (cm−1 ) L (cm) τ th 1/2 (cm) M (cm) 1.00 1.10 1.60 0.16 0.87 0.84 2.0 × 10−2 2.9 × 10−5 2.4 × 10−4 2.9 170 59 5.1 11.4 19 5.8 170 62 H2 O D2 O Graphite Source: Data from Ref. 4; used with permission of Wiley. of the diffusion of these fast neutrons, but for now we simply indicate that there is an equivalent for fast neutrons of the thermal diffusion length, which for historical reasons is identified as the square root of the “age to thermal,” τth . For intermediate to heavy mass moderators, this quantity can be shown to be equal to one-sixth the mean-squared distance a fast neutron diffuses before it thermalizes (for hydrogenous moderators, this is the definition of the quantity). The mean-squared distance that a neutron travels from birth as a fast fission neutron until capture as a thermal neutron is given by r̄ 2 = 6(τth + L2 ) ≡ 6M 2 (3.53) where M = (τth + L2 )1/2 is known as the migration length. Example 3.2: Characteristic Diffusion Parameters. Diffusion characteristics for some common moderators are given in Table 3.1. The values of D, a , and L are for thermal neutrons. Diffusion characteristics for compositions representative of pressurized water (H2 O) reactors (PWRs), boiling water (H2 O) reactors (BWRs), high-temperature graphite thermal reactors (HTGRs), sodium-cooled fast reactors (LMFRs), and gas-cooled fast reactors (GCFRs) are given in Table 3.2. Typical core diameters, measured in thermal diffusion lengths and in migration lengths for the thermal reactors and measured in fast diffusion lengths for the fast reactors, are also given. It is clear from these numbers that most of the diffusion displacement undergone by a fission neutron occurs during the slowing-down process. Table 3.2 Diffusion Parameters for Representative Reactor Core Types Reactor L (cm) PWR BWR HTGR LMFR GCFR 1.8 2.2 12 5.0* 6.6* 1/2 τ th (cm) M (cm) Diameter (L) Diameter (M) 6.3 7.1 17 6.6 7.3 21 5.0 6.6 190 180 63 35 35 56 50 40 35 35 Source: Data from Ref. 4; used with permission of Wiley. ∗ Fast neutron diffusion length. 3.6 Bare Homogeneous Reactor 3.6 Bare Homogeneous Reactor In a fission chain reacting medium (i.e., a medium in which neutron absorption can lead to fission and the production of more neutrons), the diffusion equation may or may not have an equilibrium steady-state solution, depending on the precise amount of multiplication. Thus we must consider the time-dependent diffusion equation 1 ∂φ(r, t) − D∇ 2 φ(r, t) + a φ(r, t) = νf φ(r, t) v ∂t (3.54) In a finite homogeneous medium (i.e., a bare reactor) the appropriate boundary condition is the extrapolated zero flux condition φ(aex , t) = 0 (3.55) where aex denotes the external boundaries. We further specify an initial condition φ(r, 0) = φ0 (r) (3.56) where φ0 denotes the initial spatial flux distribution at t = 0. We use the separation-of-variables technique and look for a solution of the form φ(r, t) = ψ(r)T (t) (3.57) Substituting Eq. (3.57) into Eq. (3.54) and dividing by φ = ψT yields 1 ∂T v D∇ 2 ψ + (νf − a )ψ = = −λ ψ T ∂t (3.58) where we have indicated that an expression which depends only on the spatial variable and an expression which depends only on the time variable can be equal at all spatial locations and times only if both expressions are equal to the same constant, −λ. The second form of Eq. (3.58) has the solution T (t) = T (0)e−λt (3.59) We look for spatial solutions ψ that satisfy ∇ 2 ψ(r) = −Bg2 ψ(r) (3.60) and the extrapolated spatial boundary conditions of Eq. (3.55). The constant Bg , known as geometric buckling, depends only on the geometry. Slab Reactor For example, in a slab reactor extending from x = −a/2 to x = +a/2 and infinite in the y- and z-directions, Eqs. (3.60) and (3.55) become     d 2ψ −aex aex 2 = ψ =0 (3.61) (x) = −B ψ(x), ψ g 2 2 dx 2 57 58 3 Neutron Diffusion Theory which have solutions ψ = ψn only for the (infinite) set of discrete spatial eigenvalues of Bg = Bn :  ψn (x) = cos Bn x, Bn2 = nπ aex 2 , n = 1, 3, 5, . . . (3.62) Using this result in Eq. (3.58) implies that solutions of that equation exist only for discrete-time eigenvalues λn given by   λn = v a + DBn2 − νf (3.63) Thus the solution of Eq. (3.54) for a slab reactor is φ(x, t) =  n=odd An Tn (t) cos nπx aex (3.64) where Tn is given by Eq. (3.59) with λ = λn and An is a constant which may be determined from the initial condition of Eq. (3.56) and orthogonality:  aex/2 2 nπx dx φ0 (x) cos (3.65) An (x) = aex −aex/2 aex Since B12 < B32 < · · · < Bn2 = (nπ/aex )2 , the time eigenvalues are ordered λ1 < λ3 < · · · < λn = v(a + DBn2 − νf ). Thus, after a sufficiently long time (t  1/λ3 ), the solution becomes φ(x, t) → A1 e−λ1 t cos B1 x = A1 e−λ1 t cos πx aex (3.66) This result implies that, independent of the initial distribution (as long as A1 = 0), the asymptotic shape will be the fundamental mode solution corresponding to the smallest spatial and time eigenvalues. The asymptotic solution is steady-state only if λ1 = 0. If λ1 > 0, the asymptotic solution is decaying in time, and if λ1 < 0, it is increasing in time. When the neutron population is sustained precisely in steadystate by the fission chain reaction, the reactor is said to be critical; when the neutron population is increasing in time, the reactor is said to be supercritical; and when the neutron population is dying away in time, the reactor is said to be subcritical. Defining the material buckling, Bm , 2 Bm ≡ νf − a νf /a − 1 = D L2 (3.67) the criticality condition for a bare homogeneous reactor may be written: Supercritical: 2 > B2 λ1 < 0, Bm 1 Critical: 2 = B2 λ1 = 0, Bm 1 Subcritical: 2 < B2 λ1 > 0, Bm 1 (3.68) 3.6 Bare Homogeneous Reactor Right Circular Cylinder Reactor The slab reactor results can be extended immediately to more general geometries by replacing Eqs. (3.61) and (3.62) with the corresponding equations for the other geometries. For example, for the more realistic core geometry of a right circular cylinder of radius a and height H , the equation corresponding to Eq. (3.60) is   ∂ψ(r, z) ∂ 2 ψ(r, z) 1 ∂ r + = −Bg2 ψ(r, z) (3.69) r ∂r ∂r ∂z2 and the extrapolated boundary conditions are   Hex =0 ψ(aex , z) = ψ r, ± 2 (3.70) We make further use of the separation-of-variables technique to write ψ(r, z) = R(r)Z(z) (3.71) Substituting Eq. (3.71) into Eq. (3.69) and dividing by RZ yields   1 1 ∂ ∂R(r) 1 ∂ 2 Z(z) = −ν 2 − κ 2 = −Bg2 r + R(r) r ∂r ∂r Z(z) ∂z2 (3.72) where the second form of the equation indicates that the only way in which the sum of an expression which depends only on the r-variable plus an expression which depends only on the z-variable can everywhere equal a constant is if the two expressions separately are equal to constants. Solutions of these two equations— the first expression equal to the first constant and the second expression equal to the second constant—which satisfy the corresponding boundary condition of Eqs. (3.70), exist only for discrete values of the constants νm (the roots of J0 (νm ) = 0, m = 1, 2, . . .) and κn (κn = nπ/Hex , n = 1, 3, . . .). Since the roots of J0 are ordered, ν1 < ν2 < · · · < νn , the smallest of the corresponding discrete eigen2 = ( νm )2 + (nπ/H )2 is B 2 = ( ν1 )2 + (π/H )2 , and the smallest time values Bmn ex ex 11 Rex Rex eigenvalue is       ν1 2 π 2 − νf + λ1 = v  a + D Rex Hex   2 = v Bg2 − Bm The corresponding asymptotic solution is     ν1 r πz −λ1 t cos e φ(r, z, t) → A11 J0 aex Hex  2  2 = v B11 − Bm (3.73) (3.74) 2 = B 2 = B 2 . The first zeroThe criticality condition, λ1 = 0, corresponds to Bm g 11 crossing for the Bessel function J0 (ν) occurs at ν = ν1 = 2.405. The geometric bucklings and asymptotic flux solutions are given for the common geometries in Table 3.3. 59 60 3 Neutron Diffusion Theory Table 3.3 Geometric Bucklings and Critical Flux Profiles Characterizing Some Common Core Geometries Geometric Buckling B 2g Geometry   π 2 aex   cos aπexx   ν1 2 Rex   ν r J0 R1 ex   π 2 Rex   r −1 sin Rπ r      π 2+ π 2+ π 2 aex cex bex       cos aπexx cos bπy cos cπexz    ν1 2 + Hπ Rex ex     ν r J0 R1 cos Hπ z ex ex Slab Infinite cylinder Sphere  Rectangular Flux Profile ex ex parallelepiped Finite cylinder  Source: Adapted from Ref. 4; used with permission of Wiley. Interpretation of Criticality Condition 2 = B 2 , can be rearranged to yield The criticality condition λ1 = 0, or Bm g 1= νf /a k∞ ≡ = k∞ PNL 2 2 1 + L Bg 1 + L2 Bg2 (3.75) where k∞ is the infinite-medium multiplication constant and PNL = (1 + L2 Bg2 )−1 is interpreted as the nonleakage probability. If λ1 = 0, the reactor is not critical and the asymptotic solution will either grow indefinitely or decay away in time, because the multiplication of neutrons (the ratio of the neutron population in successive generations) is greater or less than, respectively, unity. Since Eq. (3.75) applies only when k = 1, we can more generally write k= νf /a ≡ k∞ PNL 1 + L2 Bg2 (3.76) 3.6 Bare Homogeneous Reactor The situation λ1 < 0, in which the asymptotic solution increases in time, corresponds to k > 1, and the situation λ1 > 0, in which the asymptotic solution decays in time, corresponds to k < 1. From Eqs. (3.63) and (3.76),       νf /a 2 (3.77) D = va 1 + L2 B12 1 − λ1 = v B12 − Bm 1 + L2 B12 Since the mean free path to absorption is 1/a , the lifetime of a neutron that remains in the reactor until absorption is 1/va . Defining an effective lifetime of a neutron in the reactor which takes into account the possibility of leakage before absorption, l= PNL 1 = va (1 + L2 B12 ) va (3.78) enables Eq. (3.77) to be written λ1 = 1−k l (3.79) Thus the asymptotic solution of Eq. (3.54) that satisfies the extrapolated boundary conditions of Eq. (3.55) can be written φasy (r, t) → A1 ψ1 (r)e[(k−1)/ l]t (3.80) where ψ is the fundamental mode spatial distribution for the specific geometry given in Table 3.3. Optimum Geometries The minimum size for a bare reactor of a given composition that will be critical depends on the leakage, hence on the surface-to-volume ratio. The minimum critical volume for a rectangular parallelepiped bare reactor occurs for a cube and is 3 . For a right circular cylinder, the minimum critical volume bare V ≈ 161.11/Bm 3. reactor occurs for a radius a = 21/2 × 2.405H /π ≈ 1.08H and is V ≈ 148.31/Bm 3 The minimum critical volume for a spherical bare reactor is 129.88/Bm . It is generally desirable for the neutron flux to be distributed as uniformly as possible over the reactor core. A measure of non-uniformity is the peak-to-volume average value. For a homogeneous bare core, the peak value occurs at the center, and the peak-to-volume average is (π/2)2 = 3.88 for a rectangular parallelepiped, −2.405πν1 /4J1 (ν1 ) = 3.65 for a right circular cylinder, and π 2 /3 = 3.29 for a sphere. Example 3.3: Critical Size of a Bare Cylindrical Reactor. Although the above formalism has been developed for a one-speed description of neutron diffusion, it can be generalized to energy-dependent diffusion by using cross sections that are averaged over the neutron energy distribution. A typical composition and set of 61 62 3 Neutron Diffusion Theory Fig. 3.3 Thermal neutron flux in a spherical 235 U water-moderated reactor with and without a beryllium oxide reflector. (From Ref. 11; used with permission of University of Chicago Press.) spectrum-averaged cross sections for a PWR are given in Table 3.4. From the table a number of important materials parameters can be determined: D = 31 tr = 2 = (ν −  )/D = 4.13 × 10−4 cm−2 , 9.21 cm, L2 = D/a = 60.1 cm2 , Bm j a 2 = k∞ = νf /a = 1.025, and λextrap = 19.6 cm. The criticality condition is Bm 2 2 2 Bg = (π/Hex ) + (2.405/Rex ) . Fixing the height at 370 cm, the criticality condition requires that Rex = 127.6 cm or R = 108 cm. 3.7 Reflected Reactor Since the dimensions of a critical core of a given composition depend on the fraction of the neutrons that leak out, these dimensions can be reduced if some of the leaking neutrons are reflected back into the core. A reflector has the added benefit of making the neutron flux distribution in the core more uniform by increasing the neutron population in the outer region due to reflected neutrons which otherwise would have escaped. Figure 3.3 illustrates the neutron flux distributions in bare and reflected cores of the same composition and dimension. Reflected Slab Reactor The mathematical treatment of a reflected reactor can be illustrated most simply by considering a slab core of thickness a extending from x = −a/2 to x = +a/2 reflected on both sides by a nonmultiplying slab of thickness b. If we were to solve the time-dependent equations in both the core and reflector as we did for the bare core, but now also requiring that the solutions satisfied continuity of flux and current conditions at x = +a/2, we would find a similar but more complicated result as before—that the solution consists of a sum of spatial eigenfunctions corresponding to discrete geometrical eigenvalues, and at long times the dominant component is the fundamental mode. Rather than carry through the entire calculation, we examine the fundamental mode that obtains at long times. σ tr (10−24 cm2 ) 0.650 0.260 0.787 0.554 1.62 1.06 0.877 n (1024 cm−3 ) 2.748 × 10−2 2.757 × 10−2 3.694 × 10−3 1.710 × 10−3 1.909 × 10−4 6.592 × 10−3 1.001 × 10−5 0.294 1.78 × 10−4 0.190 2.33 484.0 2.11 3.41 × 103 σa (10−24 cm2 ) Source: Data from Ref. 4; used with permission of Wiley. Sum 10 B 238 U 235 U H O Zr Fe Isotope 0 0 0 0 312.0 0.638 0 σf (10−24 cm2 ) Table 3.4 Typical PWR Core Composition and Spectrum-Averaged Cross Sections 0 0 0 0 2.43 2.84 0 ν 1.79 × 10−2 7.16 × 10−3 2.91 × 10−3 9.46 × 10−4 3.08 × 10−4 6.93 × 10−3 8.77 × 10−6 3.62 × 10−2  tr (cm−1 ) 8.08 × 10−3 4.90 × 10−6 7.01 × 10−4 3.99 × 10−3 9.24 × 10−2 1.39 × 10−2 3.41 × 10−2 0.1532  a (cm−1 ) 0 0 0 0 0.145 1.20 × 10−2 0 0.1570 νf (cm−1 ) 3.7 Reflected Reactor 63 64 3 Neutron Diffusion Theory The neutron diffusion equations in the core and reflector are Core: −DC d 2 φC + (aC − νf C )φC = 0 dx 2 d 2 φR Reflector: −DR + aR φR = 0 dx 2 (3.81) The appropriate interface and boundary conditions are symmetry at x = 0, continuity of flux and current at x = a/2, and zero flux at the extrapolated boundary a/2 + bex : dφC =0 dx x=0     a a φC = φR 2 2     a a JC = JR 2 2   a φR + bex = 0 2 (3.82a) (3.82b) (3.82c) (3.82d) The solution in the core satisfying the symmetry boundary condition Eq. (3.82a) is φC (x) = AC cos BmC x (3.83) and the solution in the reflector satisfying the extrapolated boundary condition Eq. (3.82d) is φR = AR sinh (a/2) + bex − x LR (3.84) 2 = (ν 2 where BmC f C − aC )/DC and LR = DR /aC . Using these general solutions in the interface conditions of Eqs. (3.82b) and (3.82c), dividing the two equations, and rearranging leads to the criticality condition which must be satisfied in order for a steady-state solution to exist: BmC a BmC a DR a bex tan = coth 2 2 2DC LR LR (3.85) The smallest value of a for which a solution of this equation exists is less than π/BmC , as can be seen by plotting both sides of Eq. (3.85), in Fig. 3.4. Since the criticality condition for the bare slab was BmC = π/aex , this result confirms that the addition of a reflector reduces the dimension necessary for criticality. Reflector Savings The difference in the reflected and unreflected critical dimensions is known as the reflector savings, δ:   DC BmC 1 bex (3.86) δ ≡ a(bare) − a(reflected) = tan−1 LR tanh BmC DR LR 3.8 Homogenization of a Heterogeneous Fuel–Moderator Assembly Fig. 3.4 Plot of criticality equation for reflected reactor. (From Ref. 10; used with permission of McGraw-Hill.) In the limit of a reflector that is thick in comparison to the neutron diffusion length (b  LR ), this reduces to δ ≈ DC LR /DR . Reflected Spherical, Cylindrical, and Rectangular Parallelepiped Cores A similar calculation can be performed for other core geometries, but with reflection in only one direction. The resulting criticality conditions are given in Table 3.5. 3.8 Homogenization of a Heterogeneous Fuel–Moderator Assembly In our previous treatment of a homogeneous core, we have implicitly assumed that the actual core—consisting of thousands of fuel and control elements, coolant, and structure (Fig. 3.5)—can be represented by some effective homogeneous mixture. Spatial Self-Shielding and Thermal Disadvantage Factor We might be tempted to construct this homogeneous mixture by simply volumeweighting the number densities of the various fuel, control, moderator, coolant, and structural materials, but this procedure would fail to take into account the reduction of the neutron population in the region of strong absorbers, a phenomenon known as spatial self-shielding. We illustrate this phenomenon by considering the thermal neutron flux distribution in a large fuel–moderator assembly consisting of a repeating array of slab fuel elements of width 2a interspersed with moderating regions of thickness 2(b − a). Since the moderator is much more effective than the fuel at slowing down neutrons, we specify a uniform source SM of thermal neutrons in the moderator and no thermal neutron source in the fuel. We take as a calculational model one-half of the slab fuel element, extending from x = 0 to 65 Cylinder: side-reflected Sphere Table 3.5 Criticality Condition for Reflected Reactors Geometry (νf −a ) D R  (Continued) 2 2  ν −a π 2 = 1 + κR , κc2 = fD − 2hπ 2 2h ex ex L ex L0 (ρ) = I0 (κR ρ1ex )K0 (κR ρ) − I0 (κR ρ)K0 (κR ρ1ex ) 2  π 2 = κ2 + BmC c 2h Dc J0 (κc ρ0 ) DR L0 (ρ0 ) J0 (κc ρ0 ) = L0 (ρ0 ) B2 =   R −R0 R Dc (BR0 cot BR0 − 1) = −DR L 0 coth R1ex 1ex +1 LR R 66 3 Neutron Diffusion Theory Block: end-reflected Cylinder: end-reflected Table 3.5 (Continued) Geometry LR 1ex ex 1ex ex DC κ1 tan κ1 a = DR μ1 coth μ1 (dex − a) 2  2  + 2cπ μ21 = 12 + 2bπ ex ex LR 2  2  π 2 = κ2 + BmC + 2cπ 1 2bex ex 2  2  ν −a κ12 = fD − 2bπ − 2cπ 1ex ν1 = 2.405 2  ν −   n ν f − ρ1 μ2c = D Dc μc tan μc h = DR coth μR (aex − h) 2 2   ν 2 = μ2 + ν 1 μ2R = 12 + ρ 1 , BmC ρ C 3.8 Homogenization of a Heterogeneous Fuel–Moderator Assembly 67 68 3 Neutron Diffusion Theory Fig. 3.5 Heterogeneous nuclear reactor fuel assemblies. (From Ref. 4; used with permission of Wiley.) x = a, and one-half of the moderating region, extending from x = a to x = b. The neutron diffusion equations in the fuel and moderator are Fuel: −DF Moderator: −DM d 2 φF (x) + aF φF (x) = 0 dx 2 d 2 φM (x) + aM φM (x) = SM dx 2 (3.87) The appropriate boundary conditions are symmetry at the fuel and moderator midplanes at x = 0 and x = b, respectively. The other two conditions that must be 3.8 Homogenization of a Heterogeneous Fuel–Moderator Assembly satisfied are continuity of flux and current at the fuel–moderator interface at x = a dφF (0) =0 dx dφM (b) =0 dx φF (a) = φM (a) DF (3.88a) (3.88b) (3.88c) dφF (a) dφM (a) = DM dx dx (3.88d) The solutions to Eqs. (3.87) that satisfy the conditions of Eqs. (3.88) are φF (x) = SM cosh(x/LF ) {(LF /DF ) coth(a/LF ) + (LM /DM ) coth[(b − a)/LM ]}(DF /LF )aM sinh(a/LF ) φM (x) =   cosh[(b − x)/LM ] SM 1− aM {(LF /DF ) coth(a/LF ) + (LM /DM ) coth[(b − a)/LM ]}(DM /LM ) sinh[(b − a)/LM ] (3.89) The thermal flux disadvantage factor is defined as the ratio of the average flux in the moderator to the average flux in the fuel: b ξ≡ a a φM (x)dx VF aF φ̄M = = a VM aM (b − a) 0 φF (x)dx φ̄F   VM aM F +E−1 VF aF (3.90) where VF = a, VM = b − a, and F= a a coth , LF LF E=   b−a b−a coth LM LM (3.91) for slab geometry. Thermal flux disadvantage factors for repeating arrays formed by other simple geometries can be calculated in the same manner and represented by the second form of Eq. (3.90). The results for the lattice functions E and F in other geometries 2 − πρ 2 are given in Table 3.6. The volumes are VF = πρF2 and 43 πrf3 and VM = πρM F 3 − r 3 ), for the cylinder and sphere, respectively. and 43 π(rM F Effective Homogeneous Cross Sections An effective homogeneous fuel cross section averaged over the fuel–moderator lattice can be constructed by using the thermal disadvantage factor of Eq. (3.90) in the definition aF φ̄F VF aF VF VF + VM = V F + VM V F + ξ VM φ̄F VF + φ̄M VM   hom 1 + VM /VF = aF 1 + ξ VM /VF eff aF ≡ (3.92) 69 Source: Adapted from Ref. 10: used with permission of McGraw-Hill. Spherical Cylindrical Slab Geometry Table 3.6 Functions E and F for Various Cell Geometries F F E= 3 rM −rF 1−(rM /LM ) coth[(rM −rF )/LM ] 3rF L2M 1−rM rF /L2M −[(rM −rF )/LM ] coth[(rM −rF )/LM ] 3 F = 3[(r F/L F)−tanh(r F /L )] F F F F (r /L )2 tanh(r /L) 2 −ρ 2 ) (1/LM )(ρM I0 (ρF /LM )K1 (ρM /LM )+K0 (ρF /LM )I1 (ρM /LM ) F 2ρF I1 (ρM /LM )K1 (ρF /LM )−K1 (ρM /LM )I1 (ρF /LM ) E=  (ρF /LF )I0 (ρF /LF ) 2I1 (ρF /LF ) F= E = bL− a coth bL− a M M F = La coth La E and F Functions  70 3 Neutron Diffusion Theory 3.8 Homogenization of a Heterogeneous Fuel–Moderator Assembly An effective homogeneous absorption cross section for the moderator can obviously be constructed by exchanging the F and M subscripts and replacing ξ by ξ −1 . These fuel and moderator effective cross sections can then be combined eff +  eff ) to obtain an effective homogeneous cross section for the (aeff = aF aM fuel–moderator assembly to be used in one of the previous homogeneous core calculations. Effective homogeneous scattering and transport cross sections can be constructed in a similar manner. Example 3.4: Flux Disadvantage Factor and Effective Homogenized Cross Section in a Slab Lattice. Consider a lattice consisting of a large number of 1-cm-thick slab fuel plates separated by 1 cm of water at room temperature. The fuel is 10% enriched uranium. The fuel and water number densities are n235 = 0.00478 × 1024 cm−3 , n238 = 0.0430 × 1024 cm−3 , and nH2 O = 0.0334 × 1024 cm−3 . Using the spectrumaveraged cross sections of Table 3.4 (and constructing effective H2 O σ ’s as two times the H σ ’s plus the O σ ) yields the following material properties for the uranium fuel: tr = 0.0534 cm−1 , a = 3.220 cm−1 , D = 6.17 cm, and L = 1.38 cm, and for water: tr = 0.0521 cm−1 , a = 0.0196 cm−1 , D = 6.40 cm, and L = 18.06 cm. The geometric parameters in Eqs. (3.90) and (3.91) are VF = VM = a = b − a = 0.5 cm. Evaluating Eq. (3.90) yields ξ = 1.04 for the thermal disadvantage factor. The effective homogenized fuel absorption and transport cross sections calculated eff = 1.575 cm−1 and  eff = 0.0264 cm−1 . A simple hofrom Eq. (3.92) are aF trF hom = 1.610 cm−1 and mogenization (implicitly assuming that ξ = 1) yields aF hom −1 trF = 0.0267 cm , so the effect of the spatial self-shielding (ξ ) is significant. The effective homogenized cross section for the water (moderator) is derived by a procedure similar to that in Eq. (3.92) and results in an expression similar to Eq. (3.92) but with the M and F subscripts interchanged and ξ replaced by ξ −1 . The effective homogenized water absorption and transport cross sections are eff = 0.010 cm−1 and  eff = 0.0266 cm−1 , so that the total effective absorption aM trM eff +  eff = 1.575 + 0.010 = and transport cross sections for the lattice are aeff = aF aM eff eff eff −1 1.585 cm and tr = trF + trM = 0.0264 + 0.0266 = 0.053 cm−1 . Note that diffusion theory is not really suitable for calculating the diffusion of neutrons in such a lattice because λtr = 1/tr  0.5 cm, the dimension of the diffusing medium, in both the fuel and the water; and that this example serves more to illustrate the application of the methodology than to provide accurate quantitative results. The transport methods introduced in Section 3.12 and described more fully in Chapter 9 must be used to calculate flux disadvantage factors, in general. Thermal Utilization Another use of the thermal disadvantage factor is to calculate the thermal utilization for the fuel–moderator lattice: fhet ≡ aF VF + aM VM aF φ̄F VF aF VF = aF VF + aM VM aF VF + ξ aM VM aF φ̄F VF + aM φ̄M VM 71 72 3 Neutron Diffusion Theory = fhom aF VF + aM VM aF VF + ξ aM VM (3.93) In both Eqs. (3.92) and (3.93), the first term is the result that would be obtained with simple volume-weighted homogenization of the fuel and moderator number densities, and the second term is a correction that accounts for the flux selfshielding in the fuel. Measurement of Thermal Utilization In a finite fuel–moderator assembly with geometry characterized by the geometric buckling Bg and neutrons becoming thermal at a rate qM (per second per cubic centimeter) in the moderator, the thermal neutron balance is   (3.94) qM VM = (aM φ̄M VM + aF φ̄F VF ) 1 + L2 Bg2 and the thermal utilization is just the fraction of those thermal neutrons which are absorbed that are absorbed by the fuel: f= aF φ̄F VF aF φ̄F VF + aM φ̄M VM (3.95) The ratio of the slowing-down source to the thermal flux at some point in the moderator, qM /φM (x) can be determined by irradiating an indium foil (indium has an absorption resonance just above thermal) at that point and then measuring the total foil activation Atot . Then another indium foil clad in a cadmium jacket, which will absorb all the thermal neutrons before they can reach the foil but will pass the epithermal neutrons, is irradiated at the same location to determine the epithermal activation Aepi . The thermal component of the total activation, Ath = Atot − Aepi , is proportional to the thermal flux at the location of the foil, Ath = cth φM (x). The epithermal activation is proportional to the slowing-down source, Aepi = cepi qM . Thus qM /φM (x) = (cepi /cth )(Aepi /Ath ). The quantity CR = Aepi /Ath is determined by the foil measurements and is known as the cadmium ratio. The ratio of constants (cepi /cth ) can be determined by irradiating many clad and unclad indium foils in a large block of pure moderator that has a source emitting Q neutrons per second. The neutron balance is   aM φ(x)dx  q(x)dx = Q (3.96) and the ratio of integrated thermal and epithermal activities is ρ≡ cepi φ(x)dx cepi 1 Ath (x)dx = = cth aM cth q(x)dx Aepi (x)dx (3.97) These results can be combined to write an expression for the thermal utilization, f =1− aM φ̄M (1 + L2 B 2 ) 1 + L2 B 2 φ̄M =1− qM ρCR φM (x) (3.98) 3.9 Control Rods in terms of the experimentally determined quantities CR and ρ and the local-toaverage moderator flux ratio, which can be calculated using the foregoing formalism. Local Power Peaking Factor Once effective homogenized cross sections are constructed, the fuel–moderator assembly may be treated as a homogeneous region, and the average flux distribution in the assembly may be calculated using one of the other techniques discussed in this chapter. The average power density in the fuel–moderator assembly is then feffF φav , where feffF is given by an expression such as Eq. (3.92) and φav is the average flux in the fuel–moderator assembly: φav = 1 + ξ(VM /VF ) φ̄F VF + φ̄M VM = φ̄F V F + VM 1 + (VM /VF ) (3.99) The peak power density will occur at the location of the maximum neutron flux in the fuel element, which is at x = a, as may be seen from Eq. (3.89). The power peaking factor—the ratio of the peak to average power densities in the assembly—is given by     f F φF (a) VM φF (a) VM a a Fpp ≡ = 1+ = 1+ coth eff V V L L φ̄ f F φav F F F F F        1 a 2 VM 1 a 4 1+ = 1+ − + ··· , VF 3 LF 45 LF a <π LF (3.100) where the form of φF (a)/φF for a slab fuel–moderator lattice has been used to arrive at the second form of the equation. The power peaking is minimized by minimizing a/LF and VM /VF . 3.9 Control Rods Effective Diffusion Theory Cross Sections for Control Rods Localized highly absorbing control elements such as control rods cannot be calculated directly using diffusion theory. However, transport theory can be used to determine effective diffusion theory cross sections for use with diffusion theory. We illustrate this by considering the BWR example shown in Fig. 3.5 of a core consisting of a repeating array of four fuel–moderator assemblies surrounding a cruciform control rod. First, the fuel–moderator assemblies must be homogenized, using the procedure of Section 3.8 or some more sophisticated procedure based on transport theory, yielding a model of a cruciform control rod embedded in a square cell of homogeneous fuel–moderator, as shown in Fig. 3.6. If the span, l, of the control blade is large compared to the neutron diffusion length in the fuel–moderator 73 74 3 Neutron Diffusion Theory Fig. 3.6 One-dimensional model of a cruciform control blade cell. (From Ref. 4; used with permission of Wiley.) region, the diffusion of neutrons into the rod is essentially one-dimensional. We take advantage of this fact to replace the two-dimensional problem by an equivalent one-dimensional problem that preserves both the ratio of the control rod surface to the fuel–moderator volume and the thickness of the control blade. We construct an equivalent model consisting of a repeating array of fuel–moderator slabs of thickness 2a and control rod slabs of thickness 2t , where a = (m2 − 2tl + t 2 )/2l, as shown in Fig. 3.6. Our calculational model then is a fuel–moderator (half) slab from x = 0 to x = a and a control (half) slab from x = a to x = a + t , with symmetry boundary conditions at x = 0 and x = a + t . The neutron diffusion equation −D d 2 φ(x) + a φ(x) = S0 dx 2 (3.101) is valid in the fuel–moderator slab, where S0 is a uniform source of neutrons slowing down in the fuel–moderator region. The symmetry boundary condition for the diffusion theory calculation is dφ(0) =0 dx (3.102) and a transport boundary condition J (a) =α φ(a) (3.103) is used at the fuel–moderator interface with the control rod. The parameter α must be determined from a transport theory calculation of the control rod region (Section 3.12). For a slab of width 2t , such a calculation yields α= 1 − 2E3 (2ac t) 2[1 + 3E4 (2ac t)] (3.104) 3.9 Control Rods where ac is the control rod absorption cross section and En is the exponential integral function:  ∞ e−gu u−n du (3.105) En (g) = 1 The solution to Eq. (3.101) which satisfies Eqs. (3.102) and (3.103) is   α cosh(x/L) S0 1− φ(x) = a α cosh(a/L) + (D/L) sinh(a/L) (3.106) We now define an effective diffusion theory cross section for the control rod by requiring that the diffusion theory and transport theory calculations of the neutron absorption rate in the control rod agree: ceff φav Acell = Pc Jc (3.107) where φav is the average diffusion theory flux in the fuel–moderator region, Acell = (a + t)b is the area of the fuel–moderator plus control rod cell of arbitrary transverse direction b, Pc = b is the perimeter of the control rod interface with the fuel– moderator region, and Jc is the neutron current from the fuel–moderator region at the surface of the control rod. It is assumed that all neutrons which enter the control rod are absorbed. Combining the results above yields ceff = Pc Jc 1 φ(a) a = α = Acell φ̄ a a[a /α + (1/L) coth(a/L)] − 1 φ̄ (3.108) for the effective homogeneous control rod cross section to be used in a diffusion theory calculation. Note that the a in this equation is the effective fuel–moderator homogenized cross section, and the control rod cross section is hidden in the parameter α. Example 3.5: Slab Control Plate Effective Cross Section. Consider again the lattice of alternating 10% enriched uranium fuel and water slabs, each 1 cm thick, discussed in Section 3.8. The effective homogenized lattice cross sections are aeff = 0.4144 cm−1 and treff = 0.0525 cm−1 , leading to D eff = 6.35 cm and Leff = 3.91 cm in the fuel–water lattice. Now consider the placement of 1-cmthick slab natural boron plates (19.9% 10 B) every 10.5 cm in the lattice. With respect to Fig. 3.6, t = 0.5 cm and a = 5 cm. The 10 B density in the control slab is nB10 = 0.199(2.45/10.8)(0.6022 × 1024 ) = 0.0271 × 1024 cm−3 , the absorption cross section from Table 3.4 is σB10 = 3.41 × 10−21 cm2 , and the macroscopic control slab absorption cross section is ac = 92.411 cm−1 . For such large values of 2tac , the exponential integrals approach zero and the transport boundary condition parameter α → 0.5. Evaluation of Eq. (3.108) with these parameters yields for the effective homogenized control cross section ceff = 0.0894 cm−1 . Thus, in the homogenized representation of the lattice, the effective macroscopic absorption cross section is 0.414 cm−1 with the control plates removed and 0.493 cm−1 with the control plates inserted. The effective transport cross section is 0.0525 cm−1 and is assumed to be the same with or without the control plates. 75 76 3 Neutron Diffusion Theory Fig. 3.7 Insertion of a control rod bank into a bare cylindrical core. (From Ref. 4; used with permission of Wiley.) Windowshade Treatment of Control Rods Now that we know how to obtain effective homogenized cross sections for the fuel– moderator assemblies and for control rods, we can represent the partial insertion of a bank of control rods (from the top) into a bare cylindrical core as a two-region core diffusion problem, as indicated in Fig. 3.7. The lower, unrodded region is represented by the homogenized fuel–moderator cross sections, and the upper rodded region is represented by the homogenized fuel–moderator cross sections plus the effective control rod cross section. The neutron diffusion equation in both the rodded and unrodded regions is of the form of Eq. (3.69), and we can anticipate from the development of Section 3.6 that a separation of variables solution that satisfies a zero flux boundary condition at r = R (we assume that the reactor is sufficiently large that the zero flux condition at the external boundary is equivalent to the zero flux condition at the extrapolated boundary) will be of the form   ν1 r (3.109) ψ(r, z) = Z(z)J0 R and the function Z(z) will satisfy d 2Z + Bz2 (Z)(z) = 0 dz2 where Bz2 ≡   νf /a − 1 ν1 2 − Rex L2 (3.110) (3.111) 3.10 Numerical Solution of Diffusion Equation and ν1 = 2.405 is the smallest root of J0 (ν) = 0. Solving the diffusion equation separately in the rodded and unrodded regions and requiring that the solutions vanish at z = 0 and z = H yields   Zun (z) = Aun sin Bzun z , 0≤z≤h (3.112) rod Zrod (z) = Arod sinh Bz (H − z) , h ≤ z ≤ H We require continuity of flux and current at z = h, the interface between the rodded and unrodded regions, Zun (h) = Zrod (h) (3.113) dZun (h) dZrod (h) Dun = Drod dz dz The first condition leads to the relationship sin(Bzun h) Arod = Aun sinh[Bzrod (H − h)] (3.114) and dividing the two conditions leads to the criticality condition, 1 1 tan Bzun h = − tanh Bzrod (H − h) Dun Bzun Drod Bzrod (3.115) which may be solved for the rod insertion distance (H − h), for which the reactor is just critical. The axial neutron flux solution is sketched in Fig. 3.7 for several rod insertions. As might be expected, the axial flux distribution is symmetric when the rod bank is fully withdrawn and becomes progressively more peaked toward the bottom of the core as the rod bank is inserted farther downward. Note that in case of rod insertion from the bottom, the situation is just reversed. 3.10 Numerical Solution of Diffusion Equation Although the semianalytical techniques for solving the neutron diffusion equation that we have developed can be extended to treat reactor models consisting of a larger number of different homogeneous regions than we have considered, realistic reactor models may consist of hundreds or thousands of different homogenized regions, even after the local fuel–moderator homogenization has taken place. The fuel concentration may vary from assembly to assembly and within an assembly in order to make the power distribution more uniform, and even within initially uniform assemblies the composition will change differently from location to location with fuel burnup. The standard practice today is to use numerical techniques to solve the neutron diffusion equation. 77 78 3 Neutron Diffusion Theory Finite Difference Equations in One Dimension The neutron diffusion equation in a one-dimensional slab reactor model is − 1 dφ(x) d D(x) + a (x)φ(x) = νf (x)φ(x) dx dx λ (3.116) The first step in developing a numerical solution procedure is to replace the continuous spatial dependence of the flux, φ(x), with the values of the flux at a number of discrete spatial locations, φi ≡ φ(xi ), the solution for which will be the objective of the numerical technique. There are many ways to do this, and we will use a simple finite-difference approximation. We subdivide the interval 0 ≤ x ≤ a of interest into I subintervals of length  = a/I . (A more general development would use nonuniform subintervals.) A general rule of thumb is that  < L (the neutron diffusion length) sets an upper limit on the subinterval length (or mesh spacing). ←→ x0 x1 x2 . . . xi−1 xi xi+1 ... xI −1 xI xi−1/2 xi+1/2 Next, the terms in Eq. (3.116) are each integrated from xi−1/2 to xi+1/2 , using the following approximations:  xi +(1/2) dxa (x)φ(x)  ai φi   xi −(1/2) xi +(1/2) xi −(1/2)  dφ d D(x) dx dx dx  D  dφ dx xi +1/2 −D dφ dx (3.117) xi +1/2 1 φi+1 − φi (Di + Di+1 ) 2  φi − φi−1 1 − (Di−1 + Di ) 2  where we have associated ai , Di , and so on, with the subinterval xi−1/2 ≤ x ≤ xi+1/2 . The discrete equation associated with xi may be written ai,i−1 φi−1 + ai,i φi + ai,i+1 φi+1 = where 1 fi φi ≡ Si , λ    c 1 Di + Di−1 1 − 2 2i − 1 2   1 Di−1 + 2Di + Di+1 ai,i = ai + 2 2    c 1 Di+1 + Di 1 + ai,i+1 = − 2 2i − 1 2 i = 1, . . . , I − 1 (3.118) ai,i−1 = − fi = νf i (3.119) 3.10 Numerical Solution of Diffusion Equation We have generalized to other one-dimensional geometries, where c = 0, 1, and 2 for slab, cylindrical, and spherical, respectively. The significant feature of the set of Eqs. (3.118) is nearest-neighbor coupling—the flux at any xi is only directly coupled to the flux at the adjacent points xi−1 and xi+1 , which greatly facilitates their solution. Note that the difference equations are formulated only for the I − 1 interior mesh points at x1 , x2 , . . . , xI −1 . The boundary conditions determine the exterior mesh points. A zero flux boundary condition at the left boundary corresponds to φ0 = 0, for example. A zero current or symmetry boundary condition at the left boundary corresponds to φ0 = φ1 and would be implemented by setting a1,0 = 0 and a1,1 = a1 + (D1 + D2 )/2 . Forward Elimination/Backward Substitution Spatial Solution Procedure The set of I −1 Eqs. (3.118) can readily be solved by Gaussian elimination, or forward elimination backward substitution, for a known fission source Si . The Gaussian elimination solution is implemented by subtracting ai,i−1 /ai−1,i−1 times the (i − 1)th equation from the ith equation to eliminate the ai,i−1 element in the ith equation. The modified ith equation is then divided by ai,i . This process is repeated successively for i = 1 through i = I − 1. Then the manipulated equations can be solved successively from i = I − 1 to i = 1 using the algorithms φI −1 = αI −1 φI −2 = −AI −2 φI −1 + αI −2 .. . (3.120) φi = −Ai φi+1 + αi for the backward substitution, where A1 = a1,2 , a1,1 S1 α1 = , a1,1 Ai = ai,i+1 ai,i − ai,i−1 Ai−1 Si − ai,i−1 αi−1 αi = ai,i − ai,i−1 Ai−1 (3.121) had previously been constructed on the forward elimination. Power Iteration on Fission Source The fission source is not known a priori, of course, so the Gaussian elimination must be embedded in an iteration on the fission source term, as follows. An initial (0) guess of the flux φi at each point and of the eigenvalue λ(0) is made and an ini(0) (0) tial fission source at each point is constructed Si = νf i φi /λ(0) . The Gaussian (1) elimination is performed to determine φi . A new estimate of the eigenvalue is 79 80 3 Neutron Diffusion Theory made from I −1 (1) i=1 νf i φi  (0) (0) (0) i=1 [ai,i−1 φi−1 + ai,i φi + ai,i+1 φi+1 ] λ(1) = I −1 I −1 (1) νf i φi   I −1i=1 (0) (0) i=1 νf i φi /λ (3.122) and a new fission source is constructed from (1) Si 1 (1) νf i φi λ(1) = (3.123) This iteration process is continued [using Eqs. (3.122) and (3.123) with 0 → n − 1 and 1 → n] until the eigenvalues obtained on two successive iterates differ by less than some convergence criterion, say ε = 10−5 : λ(n) − λ(n−1) <ε λ(n−1) (3.124) Finite-Difference Equations in Two Dimensions In rectangular geometry, the neutron diffusion equation is − ∂φ(x, y) ∂ ∂φ(x, y) ∂ D(x, y) − D(x, y) + a (x, y)φ(x, y) ∂x ∂x ∂y ∂y = 1 νf (x, y)φ(x, y) λ (3.125) To extend the procedure for developing finite-difference equations which was discussed for the one-dimensional case, we consider a rectangle with x-dimension a and y-dimension b. We subdivide a into I intervals of length x = a/I and subdivide b into J intervals of length y = b/J . • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • (i, j ) • • • • • • • • • x0 • • • • x1 • • • • x2 • • • • ... • • • • • • • • • • • • ... • • • • • • • • xI xi−1 • • • • xi xi+1 xI −1 yJ yJ −1 .. . yj +1 yj yj −1 .. . y2 y1 y0 3.10 Numerical Solution of Diffusion Equation The diffusion equation is integrated over the mesh box (xi−1/2 ≤ x ≤ xi+1/2 , yj −1/2 / ≤ y ≤ yj +1/2 ) and the approximations of Eqs. (3.117) are extended to two dimensions to obtain the finite-difference equations:     1 Di,j + Di+1,j 1 Di−1,j + Di,j φ φi+1,j − − i−1,j 2 2 2x 2x     1 Di,j −1 + Di,j 1 Di,j + Di,j +1 φi,j −1 − φi,j +1 − 2 2 2y 2y   1 1 1 Di−1,j + Di,j + 12 Di+1,j 2 Di,j −1 + Di,j + 2 Di,j +1 φi,j + aij + 2 + 2x 2y = 1 νf i,j φi,j , λ i = 1, . . . , I − 1; j = 1, . . . , J − 1 (3.126) The significant feature of these equations is, once again, nearest-neighbor coupling —the flux at (i, j ) is only directly coupled to the fluxes at (i, j + 1), (i, j − 1), (i + 1, j ), and (i − 1, j ). The boundary conditions are used to specify φ0,j , φI,j , φi,0 , and φi,J , as discussed for the one-dimensional case. In order to simplify the notation somewhat, we replace the (i, j ) identification of a spatial location with a (p) identification. The total number of spatial locations is P = (I − 1) × (J − 1). We will choose p = 1 for (i = 1, j = 1), p = 2 for (i = 2, j = 1), . . . , p = I − 1 for (i = I − 1, j = 1), p = I for (i = 1, j = 2), . . . , p = 2(I − 1) for (i = I − 1, j = 2), and so on. Then the set of finite difference equations may be written a1,1 φ1 + a1,2 φ2 + a1,3 φ3 + · · · + a1,p φp + · · · + a1,P φP = Sf 1 a2,1 φ1 + a2,2 φ2 + a2,3 φ3 + · · · + a2,p φp + · · · + a2,P φP = Sf 2 a3,1 φ1 + a3,2 φ2 + a3,3 φ3 + · · · + a3,p φp + · · · + a3,P φP = Sf 3 (3.127) .. . aP ,1 φ1 + aP ,2 φ2 + aP ,3 φ3 + · · · + aP ,p φp + · · · + aP ,P φP = Sf P where 1 1 + Dp + 12 Dp+1 2 Dp−I + Dp + 2 Dp+I + 2x 2y     1 Dp−1 + Dp 1 Dp + Dp+1 , a ap,p−1 = − = − p,p+1 2 2 2x 2x     1 Dp + Dp−I 1 Dp+I + Dp , a = − ap,p−I = − p,p+I 2 2 2y 2y ap,p = ap + ap,q = 0, Sfp = 1 2 Dp−1 q = p − 1, p + 1, p − I, p + I 1 νfp φp λ (3.128) 81 82 3 Neutron Diffusion Theory Successive Relaxation Solution of Two-Dimensional Finite-Difference Equations There are a number of possible ways to solve the set of Eqs. (3.127). We describe here the widely used Gauss–Seidel or successive relaxation method. This is an iterative method that proceeds by solving the first equation for φ1 , assuming S1 is known and guessing a value for φ2 , . . . , φP ; then solving the second equation for φ2 , assuming that S2 is known, using the value just calculated for φ1 , and using the same guessed values for φ3 , . . . , φP ; then solving the third equation for φ3 , assuming that S3 is known, using the just calculated values for φ1 and φ2 , and using the same guessed values for φ4 , . . . , φP ; and continuing thusly until the last equation is solved for φP , assuming that SP is known, and using the just calculated values for φ1 , . . . , φP −1 . The set of new values of φ1 , . . . , φP thus calculated provides a new guess to be used in a repeated iteration. The general algorithm for the solution at each step is φp(m+1) = 1  ap,p Sfp − p−1  ap,q φq(m+1) q=1 P  −  ap,q φq(m) (3.129) q=p+1 where m is the iteration index. This inner iteration is continued until the flux solution at each location has converged to within a specified tolerance, ε ≈ 10−2 , which may be chosen smaller in regions where exact knowledge of the neutron flux is important than in, for example, reflector regions: (m+1) φp (m) − φp (m) φp (3.130) < εp It is possible to accelerate the convergence of the relaxation iteration by using as a new flux guess a mixture of the previous flux and the relaxation result of Eq. (3.129): φp(m+1) = (1 − ω)φp(m) + ω ap,p  Sfp − p−1  q=1 ap,q φq(m+1) − P   ap,q φq(m) q=p+1 (3.131) The acceleration parameter ω may be chosen in a number of ways (see Ref. 8), but generally varies between 1 and 2. The algorithm of Eq. (3.131) is known as successive overrelaxation (SOR). Another widely used method for solving the two-dimensional diffusion equations is the alternating direction implicit iteration scheme described in Section 16.3. Power Outer Iteration on Fission Source The power iteration on the fission source proceeds as described above [i.e., in Eqs. (3.122) to (3.124)] but with i replaced by p in the notation of this section, and with  replaced by x , y . 3.11 Nodal Approximation Thus the solution of the finite-difference equations has a two-level iteration hierarchy. There is an outer power iteration on the fission source and the eigenvalue, described by Eqs. (3.122) to (3.124). Then for each of the outer iterations, there is a series of inner relaxation iterations—described by Eq. (3.129) or (3.131) and (3.130)—to converge the flux solution for that outer iterate of the fission source. Limitations on Mesh Spacing We can obtain some insight as to limitations on mesh spacing by considering the source-free diffusion equation in one dimension: d 2φ φ − 2 =0 dx 2 L (3.132) which can be solved exactly over the mesh interval  = xi+1/2 − xi−1/2 centered on xi : φ(xi+1/2 ) = e−/2L φ(xi ) = e−/L φ(xi−1/2 ) (3.133) The central difference finite-difference approximation (which we have been using) of Eq. (3.132) on this interval can be written   2   φ(xi ) φ(xi+1/2 ) + φ(xi−1/2 ) = 2 + L (3.134) Comparing the right side of Eq. (3.134) to the exact expression for the left side constructed from Eq. (3.133) allows us to define the difference as a measure of the error in the finite-difference approximation: error =   3  2 + ··· 4 L (3.135) Clearly, the mesh spacing should be less than the diffusion length. 3.11 Nodal Approximation In principle, once the local fuel cell heterogeneity in each fuel assembly is replaced by effective homogenized cross sections and effective cross sections are constructed for the control rods, the three-dimensional finite-difference diffusion equations can be solved for the effective multiplication constant and the neutron flux distribution everywhere in a reactor. In practice, it is seldom practical to do so because of the large number of simultaneous equations that must be solved. As we have seen, accuracy in the finite-difference solution requires that the mesh spacing be smaller than the diffusion length, and a typical LWR core is about 200 thermal diffusion lengths in each of the three dimensions, which results in several million mesh points, hence several million simultaneous equations. 83 84 3 Neutron Diffusion Theory Fig. 3.8 Division of a reactor into nodes. (From Ref. 4; used with permission of Wiley.) One means to deal with this situation is to divide the flux solution into two parts. The reactor core (and reflector, etc.) is divided into a relatively small number (on the order of 100 or less) large regions, or nodes, as depicted in Fig. 3.8. The detailed flux distribution within each node is determined from a finite-difference calculation just within the node (or set of contiguous nodes); such calculations need be performed only for every different type of node, since the solution for different nodes that have the same internal material distribution and the same boundary conditions will be identical. The global flux distribution (i.e., the average value of the flux in the different nodes) and the effective multiplication factor are then determined from a nodal calculation. The general derivation of nodal diffusion theory methods may be illustrated by integrating the diffusion equation ν −∇ · D∇φ + a φ = f φ k over the spatial domain of each node n to obtain    1 − ds · D∇φ + a φ dr = νf φ dr, k Vn Sn Vn (3.136) n = 1, . . . , N (3.137) where Gauss’s law has been used to replace the volume integral over node n, Vn , of the divergence with the surface integral over the surface Sn bounding node n of the normal component of the current. In general, the surface Sn bounding node n consists of the several interfaces Snn between node n and the contiguous nodes n . Defining the average nodal flux as  1 φ(r) dr (3.138) φn ≡ Vn V n the definition of average nodal cross section follows immediately:   1 1 a (r)φ(r) dr, νfn ≡ νf (r)φ(r) dr (3.139) a n ≡ φn Vn Vn φn Vn Vn 3.12 Transport Methods The treatment of the surface integral term, which represents node-to-node leakage, is not so obvious. However, it is plausible that the gradient of the flux across the surface between two adjacent nodes is proportional to the difference in the two average nodal fluxes:  ds · D∇φ (3.140) −αnn (φn − φn ) ≡ Snn The accuracy of the nodal methods depends to a large extent on the actual evaluation of the nodal coupling coefficients αnn , which is discussed in some detail in Chapter 15. A simple approximation results from using an average value 1  2 (Dn + Dn ) for the diffusion coefficient on the interface between nodes n and n , and assuming the average diffusion coefficient and the flux gradient are both constant over the interface, which yields αnn  Snn · 12 (Dn + Dn ) lnn (3.141) where lnn is the distance between the centers of contiguous nodes n and n . Collecting these results leads to the set of N nodal equations for the nodal average fluxes and the effective multiplication constant: −  αnn φn n ∈n  n ∈n  1 αnn + an Vn φn = νf n Vn φn , k n = 1, . . . , N (3.142) where n ∈ n indicates that the sum is over nodes n which are contiguous to node n. For those nodes n located adjacent to the exterior boundary of the reactor, the nodal equations contain the flux φn for a nonexistent node on the other side of the boundary. For vacuum boundary conditions, this flux φn , would be set to zero in the equation for node n. For symmetry boundary conditions, φn = φn would be used in the equation for node n. 3.12 Transport Methods For many situations of interest (e.g. a strongly absorbing control rod), the conditions for the validity of neutron diffusion theory are not satisfied, and a more rigorous approximation for the transport of neutrons than the diffusion approximation developed in Section 3.1 is required. Limiting consideration to a “one-dimensional” medium, i.e. one which is uniform in two dimensions (y and z in Fig. 3.9), the number of uncollided neutrons arising from an isotropic source, S0 , in the differential area dA = ρdθdρ in the y–z plane at x = 0 that passes through unit area at a point on the x-axis per unit time is (S0 e−t R dA)/4πR 2 . Defining the cosine of the angle between R and x, μ ≡ nx · nR , we can write R = x/μ and make use of ρ 2 + x 2 = R 2 to write the total flux of uncollided neutrons arising from a ring of a 85 86 3 Neutron Diffusion Theory Fig. 3.9 Coordinate system for plane isotropic source calculation. (From Ref. 10; used with permission of McGraw-Hill.) uniform plane isotropic source that passes through a point on the x-axis as d( μx ) dR 1 dμ 1 1 = S0 e−t (x/μ) x = − S0 e−t (x/μ) ψ(x, μ : 0) = S0 e−t R 2 R 2 (μ) 2 μ (3.143) This expression can be used to construct the total flux of uncollided neutrons passing through unit area at a point on the x-axis by integrating over all such rings to obtain 1 φ(x : 0) = S0 2  ∞ e−t R x dR 1 = S0 R 2  1 e−t x/μ 0 dμ 1 ≡ S0 E1 (t x) (3.144) μ 2 and a similar expression for the total current of uncollided neutrons 1 J (x : 0) = S0 2  ∞ μe−t R x dR 1 = S0 R 2  1 0 1 e−t x/μ dμ = S0 E2 (t x) 2 (3.145) where  En (y) ≡ 1 μn−2 e−y/μ dμ (3.146) 0 is the exponential integral function, which has the useful differentiation property dEn (y) = −En−1 (y), dy n = 1, 2, 3, . . . (3.147) 3.12 Transport Methods Transmission and Absorption in a Purely Absorbing Slab Control Plate Consider a purely absorbing slab of thickness a inserted in a diffusing medium; i.e. a medium in which diffusion theory is appropriate, such as a homogenized fuel–moderator lattice. Assuming that the scattering source external to the slab is isotropic, the probability that a neutron incident on the slab is transmitted across the absorbing slab is T ≡ J (a : 0)/J (0 : 0) = E2 (t a), and the probability that an incident neutron is absorbed in the slab is A ≡ 1 − T = 1 − E2 (t a). Such an approximation was used in deriving the effective control cross section in Section 3.9. Escape Probability in a Slab The probability that a neutron that enters a slab region of thickness a from a diffusing medium and suffers an isotropic scattering event at a location 0 ≤ x  ≤ a and then escapes from the slab across the surface at x = a without another collision can be calculated by treating the scattering event as an isotropic source; i.e. the escaping current at x = a of ‘uncollided’ neutrons arising from the isotropic scattering at x = x  is J (a : x  ) = (1/2)S0 E2 (t (a − x  )). If the scattering source of incident neutrons is uniform over the slab, then the total current of scattered neutrons which escapes across the surface at x = a without another collision is  a  a   1 Jout (a) = J (a : x  )dx  = S0 E2 t (a − x  ) dx  2 0 0 = 1 S0 [E3 (0) − E3 (t a)] 2 t (3.148) Noting that E3 (0) = 1/2 and that the current of neutrons which escapes across the surface at x = 0 without another collision must be the same as escape across the surface at x = a, the ‘first-flight’ escape probability is the sum of these two escaping currents divided by the total scattering source, aS0   Jout (a) + Jout (0) 1 1 − E3 (t a) P0 ≡ = (3.149) aS0 at 2 Integral Transport Formulation Consider again the slab with a distributed isotropic source of neutrons, but now with isotropic elastic scattering and fission, as well as absorption, represented explicitly. The flux of uncollided source neutrons is  a   S0 (x  )E1 t (|x − x  |) dx  (3.150) φ0 (x) = 0 If the first-collision rate at x = x  is considered as a plane isotropic source of oncecollided neutrons at x  , then the flux of once-collided neutrons at x due to the once-collided source at x  is φ1 (x : x  ) =    1 s (x  ) + νf (x  ) φ0 (x  )E1 t (|x − x  |) 2 (3.151) 87 88 3 Neutron Diffusion Theory and the total flux of once-collided neutrons at x is found by integrating over the distribution of first-collision sources  a φ1 (x) = φ1 (x : x  )dx  0 =  a 0     1 s (x  ) + νf (x  ) φ0 (x  ) E1 t (|x − x  |) dx  2 (3.152) Continuing in this vein, the flux of n-collided neutrons is given by   a    1    s (x ) + νf (x ) φn−1 (x ) E1 t (|x − x  |) dx  , φn (x) = 2 0 n = 1, 2, 3, . . . (3.153) The total neutron flux is the sum of the uncollided, once-collided, twice-collided, etc. fluxes φ(x) ≡ φ0 (x) + ∞  φn (x) n=1  ∞    1 s (x  ) + νf (x  ) φn−1 (x  )E1 t (|x − x  |) dx  2 a = 0 n=0  a +   S0 (x  )E1 t (|x − x  |) dx  0  a = 0     1 s (x  ) + νf (x  ) φ(x  )E1 t (|x − x  |) dx  2 + a   S0 (x  )E1 t (|x − x  |) dx  (3.154) 0 Thus, we have found an integral equation for the neutron flux in a slab with isotropic scattering and fission, with a kernel (1/2)(s (x  ) + νf (x  ))E1 (t (|x − x  |)) and a first collision or external source S0 E1 (t x  ). Collision Probability Method If the volume of the slab is partitioned into discrete slab regions centered at xi with widths i ≡ xi+1/2 − xi−1/2 within each of which uniform average cross sections and a flat flux is assumed, Eq. (3.154) can be integrated over the volume of each region, Vi = const. × i and the resulting equation can be divided by Vi to obtain  φi = T j →i (sj + νfj )φj + S0j (3.155) j which relates the fluxes in the various volumes by the ‘first-flight transmission probabilities’ T j →i    1 1  T j →i ≡ dx  dx E1 α(x  , x) (3.156)  j i 2 j 3.12 Transport Methods where the “optical thickness” is defined α(x  , x) ≡  x x t (x  )dx  (3.157) Since α(xj , xi ) = α(xi , xj )—i.e. the optical path is the same no matter which way the neutron traverses the straight-line distance between xi and xj —there is a reciprocity relation between the transmission probabilities i T j →i = j T i→j (3.158) Upon multiplication by ti i , Eq. (3.155) can be written ti i φi =  j P ji [(sj + νfj )φj + S0j ] tj (3.159) where the collision rate in cell i is related to the neutrons introduced by scattering, fission and an external source in all cells j by the ‘collision probabilities’    1  ji  P = ti tj dx dx E1 α(x  , x) (3.160) 2 i j Because α(xi, xj ) = α(xj , xi ), there is also reciprocity between the collision probabilities; i.e. P ij = P j i (3.161) For j = i, the probability that a neutron introduced in cell j has its next collision in cell i is P ji = 1 E3 (αi+1/2,j +1/2 ) − E3 (αi−1/2,j +1/2 ) − E3 (αi+1/2,j −1/2 ) 2 + E3 (αi−1/2,j −1/2 ) (3.162) where αi,j ≡ α(xi , xj ). For j = i, a similar development leads to an expression for the probability that a neutron introduced in cell i has its next collision in cell i    1  (3.163) 1 − 2E3 (ti i ) P ii = ti i 1 − 2ti i The set of Eqs. (3.159) can be solved for the neutron flux in all the cells. Differential Transport Formulation Another formulation of neutron transport theory follows an approach similar to that used in Section 3.1 to derive diffusion theory, but without some of the limiting assumptions. Referring to Fig. 3.9, the change in the flux of particles moving along 89 90 3 Neutron Diffusion Theory the cone of radius vectors R making the same angle (μ = cos θ) with respect to a point on the x-axis within the differential distance R can be written  ψ(R + R, μ) = ψ(R, μ) + R 1 + R 2  1 −1 1 −1 s (R, μ → μ)ψ(R, μ ) dμ νf (R)ψ(R, μ ) dμ − R s (R) + a (R) ψ(R, μ) (3.164) The second term on the right is the source of neutrons within R with direction μ due to scattering within R by neutrons with other directions μ (including μ). The third term on the right is the number of source neutrons within R with direction μ due to isotropic fission produced by neutrons with other directions μ (including μ). The last term on the right represents the rate at which neutrons within R with direction μ are being lost by scattering to some other direction μ (including μ) and by absorption. The first assumption made is that the directional flux can be represented by the first two terms of a Taylor’s series, ψ(R + R, μ)  ψ(R, μ) + R[dψ(R, μ)/dR]. Noting that the spatial non-uniformity depends on the variable x and that μ = x/R, Eq. (3.164) becomes in the differential limit R → dR the one-dimensional slab version of the Boltzmann transport equation μ dψ(R, μ) + [s (R) + a (R)]ψ(R, μ) dx   1 1 1 s (R, μ → μ)ψ(R, μ ) dμ + νf (R)ψ(R, μ ) dμ (3.165) = 2 −1 −1 Spherical Harmonics Methods The spherical harmonics, or PL , approximation is developed by expansion of the angular dependence of the angular flux and of the differential scattering cross section in Legendre polynomials. The first few Legendre polynomials are  1 2 3μ − 1 2 P0 (μ) = 1, P2 (μ) = P1 (μ) = μ,  1 P3 (μ) = 5μ3 − 3μ 2 (3.166) and higher order polynomials can be generated from the recursion relation (2n + 1)μPn (μ) = (n + 1)Pn+1 (μ) + nPn−1 (μ) (3.167) The Legendre polynomials satisfy the orthogonality relation  1 −1 dμPm (μ)Pn (μ) = 2δmn 2n + 1 (3.168) 3.12 Transport Methods The cosine of the scattering angle between μ and μ, can be expressed in terms of the Legendre polynomials of μ and μ by the addition theorem in one-dimension Pn (μ0 ) = Pn (μ)Pn (μ ) (3.169) The PL equations are based on the approximation that the angular dependence of the neutron flux can be expanded in the first L + 1 Legendre polynomials ψ(x, μ) = L  2l + 1 2 l=0 φl (x)Pl (μ) (3.170) The angular dependence of the differential scattering cross section is also expanded in Legendre polynomials M  2m + 1 sm (x)Pm (μ0 ) 2 s (x, μ0 ) = (3.171) m=0 When these expansions are used in Eq. (3.165), the addition theorem of Eq. (3.169) is used to replace Pm (μ0 ) with Pm (μ)Pm (μ ), the recursion relation of Eq. (3.167) is used to replace μPn (μ) terms with Pn±1 (μ) terms, the resulting equation is multiplied in turn by Pk (μ) (k = 0 to L) and integrated over −1 ≤ μ ≤ 1, and the orthogonality relation of Eq. (3.168) is used, the L + 1 PL equations dφ1 (x) + (t − so )φ0 (x) = S0 (x), n = 0 dx n dφn−1 (n + 1) dφn+1 (x) + + (t − sn )φn (x) = Sn (x), (2n + 1) dx (2n + 1) dx n = 1, . . . , L (3.172) are obtained. The n subscript indicates the nth Legendre moment of the direction flux, source and scattering cross section  1 φn (x) ≡ dμPn (μ)ψ(x, μ)  Sn (x) ≡ −1 1 −1  sn (x) ≡ dμPn (μ)S(x, μ) (3.173) 1 −1 dμ0 Pn (μ0 )s (x, μ0 ) This set of L + 1 equations has a closure problem—they involve L + 2 unknowns. This problem is usually resolved by ignoring the term dφL+1 /dx which appears in the n = L equation. Boundary and interface conditions The true boundary condition at the left boundary xL ψ(xL , μ) = ψin (xL , μ), μ>0 (3.174) 91 92 3 Neutron Diffusion Theory where ψin (xL , μ > 0) is a known incident flux (ψin (xL , μ > 0) = 0 is the vacuum boundary condition), cannot be satisfied exactly by the angular flux approximation of Eq. (3.170), for finite L. The most obvious way to develop approximate boundary conditions which are consistent with the flux approximation is to substitute Eq. (3.170) into the exact boundary condition of (3.174), multiply by Pm (μ), and integrate over 0 ≤ μ ≤ 1. Since it is the odd Legendre polynomials which represent directionality (i.e. are different for μ and −μ), this procedure is repeated for all the odd Legendre polynomials m = 1, 3, . . . , L (or L − 1) as weighting functions to obtain, with the use of the orthogonality relation of Eq. (3.168), the Marshak boundary conditions  1 dμPm (μ) 0  N   2n + 1 φn (xL )Pn (μ) ≡ φm (xL ) 2 n=0  1 = m = 1, 3, . . . , L (or L − 1) dμPm (μ)ψin (xL , μ), (3.175) 0 Equations (3.175) constitute a set of (L + 1)/2 boundary conditions. An additional (L + 1)/2 boundary conditions are similarly obtain for the right boundary. The Marshak boundary conditions insure that the exact inward partial current at the boundary is incorporated into the solution; i.e. J + (xL ) ≡  1 dμP1 (μ) 0  n=0 1 = 0  N   2n + 1 2 φn (xL )Pn (μ) + dμP1 (μ)ψin (xL , μ) ≡ Jin (xL ) (3.176) A symmetry, or reflective, boundary condition ψ(xL , μ) = ψ(xL , −μ) obviously requires that all odd moments of the flux vanish; i.e. φn (xL ) = 0 for n = 1, 3, . . . odd. The exact interface condition of continuity of angular flux ψ(xs − ε, μ) = ψ(xs + ε, μ) (3.177) where ε is a vanishingly small distance, cannot, of course, be satisfied exactly by the flux approximation of Eq. (3.170), for finite L. Following the same procedure as for Marshak boundary conditions, we replace the exact flux with the expansion of Eq. (3.170) and require that the first L + 1 Legendre moments of this relation be satisfied (i.e. multiply by Pm and integrate over −1 ≤ μ ≤ 1, for m = 0, . . . , L). Using the orthogonality relation of Eq. (3.168) then leads to the interface conditions of continuity of the moments φn (x − ε) = φn (xs + ε), n = 0, 1, 2, . . . , L (3.178) 3.12 Transport Methods P1 equations and diffusion theory Neglecting the dφ2 /dx term, the first two of Eqs. (3.172) constitute the P1 equations dφ1 + (t − so )φ0 = S0 dx (3.179) 1 dφ0 + (t − s1 )φ1 = S1 3 dx Noting that s0 = s , the total scattering cross section, and that s1 = μ̄0 s , where μ̄0 is the average cosine of the scattering angle, and assuming that the source is isotropic (i.e. S1 = 0), the second of these P1 equations yields a Fick’s law for neutron diffusion  1 1 dφ0 μψ(x, μ) dμ ≡ J (x) = − (3.180) φ1 (x) = 3(t − μ̄0 s ) dx −1 which, when used in the first of the P1 equations yields the neutron diffusion equation   d dφ0 (x) D0 (x) + (t − s )φ0 (x) = S0 (x) − (3.181) dx dx where the diffusion coefficient and the transport cross section are defined by D0 ≡ 1 1 ≡ 3(t − μ̄0 s ) 3tr (3.182) The basic assumptions made in this derivation of diffusion theory are that the angular dependence of the neutron flux is linearly anisotropic 3 1 ψ(x, μ)  φ0 (x) + μφ1 (x) 2 2 (3.183) and that the neutron source is isotropic, or at least has no linearly anisotropic component (S1 = 0). Diffusion theory should be a good approximation when these assumptions are valid; i.e. in media for which the distribution is almost isotropic because of the preponderance of randomizing scattering collisions, away from interfaces with dissimilar media, and in the absence of anisotropic sources. The boundary conditions for diffusion theory follow directly from the Marshak condition (3.175) + (xL ) = Jin =   1 dμP1 (μ) 0  1 3 φ0 (xL ) + μφ1 (xL ) 2 2 1 dφ0 (xL ) 1 φ0 (xL ) − D 4 2 dx (3.184) + = 0, the vacuum boundary condition for When the prescribed incident current, Jin diffusion theory can be constructed from a geometrical interpretation of the ratio 93 94 3 Neutron Diffusion Theory of the flux gradient to the flux in this equation to obtain the condition that the extrapolated flux vanishes a distance λtr = 1/tr outside the boundary φ(xL − λex ) = 0, λex = 2 3tr (3.185) The interface conditions of Eq. (3.178) become in the diffusion approximation φ0 (xs + ε) = φ(xs + ε) dφ0 (xs + ε) dφ0 (xs − ε) −D(xs + ε) = −D(xs − ε) dx dx (3.186) Discrete Ordinates Method The discrete ordinate method is based on a conceptually straightforward evaluation of the transport equation at a few discrete angular directions, or ordinates, and the use of quadrature relationships to replace scattering and fission neutron source integrals over angle with summations over ordinates. The essence of the method is the choice of ordinates, quadrature weights, differencing schemes and iterative solution procedures. In one dimension, the ordinates can be chosen such that the discrete ordinates methods are completely equivalent to the PL method discussed above, and in fact the use of discrete ordinates is probably the most effective way to solve the PL and D − PL equations in one dimension. Making use of the spherical harmonics expansion of the differential scattering cross section of Eq. (3.171) and the addition theorem for Legendre polynomials of Eq. (3.169), the one dimensional neutron transport equation (3.165) in slab geometry becomes μ dψ (x, μ) + t (x)ψ(x, μ) dx  1  2l  + 1  sl  (x)Pl  (μ) dμ Pl  (μ )ψ(x, μ ) + S(x, μ) (3.187) = 2 −1  l =0 where the source term includes an external source and, in the case of a multiplying medium such as a reactor core, a fission source. We will first discuss the solution of the fixed external source problem (which implicitly assumes a subcritical reactor) and then return to the solution of the critical reactor problem, in which the solution of the fixed source problem constitutes part of the iteration strategy. Defining N ordinate directions, μn , and corresponding quadrature weights, wn , the integral over angle in Eq. (3.187) can be replaced by  1  dμPl (μ)ψ(x, μ) wn Pl (μn )ψn (x) (3.188) φl (x) ≡ −1 n where ψn ≡ ψ(μn ). The quadrature weights are normalized by N  n=1 wn = 2 (3.189) 3.12 Transport Methods It is convenient to choose ordinates and quadrature weights that are symmetric about μ = 0, hence providing equal detail in the description of forward and backward neutron fluxes. This can be accomplished by choosing μN+1−n = −μn , wN+1−n = wn , μn > 0, n = 1, 2, . . . , N/2 wn > 0, n = 1, 2, . . . , N/2 (3.190) With such even ordinates, reflective boundary conditions are simply prescribed ψn = ψN+1−n , n = 1, 2, . . . , N/2 (3.191) Known incident flux, ψin (μ), boundary conditions, including vacuum conditions when ψin (μ) = 0, are ψn = ψin (μn ), n = 1, 2, . . . , N/2 (3.192) Normally, an even number of ordinates is used (N = even), because this results in the correct number of boundary conditions and avoids certain other problems encountered with N = odd. Even with these restrictions, there remains considerable freedom in the choice of ordinates and weights. If the ordinates are chosen to be the L roots of the Legendre polynomial of order N PN (μi ) = 0 (3.193) and the weights are chosen to correctly integrate all Legendre polynomials up to PN −1  1 Pl (μ)dμ = −1 N  wn Pl (μn ) = 2δl0 , l = 0, 1, . . . , N − 1 (3.194) n=1 then the discrete ordinates equations with N ordinates are entirely equivalent to the PN −1 equations. To establish this, we multiply Eq. (3.187) by wn Pl (μn ) for 0 ≤ l ≤ N − 1, in turn, and use the recursion relation of Eq. (3.167) to obtain      l+1 l dψn Pl+1 (μn ) + Pl−1 (μn ) + wn t ψn 2l + 1 2l + 1 dx N−1   2l  + 1  sl  wn Pl  (μn )Pl (μn )φl  + wn Pl (μn )S(μn ), = 2  wn l =0 l = 0, . . . , N − 1, n = 1, . . . , N Summing these equations over 1 ≤ n ≤ N yields     l + 1 dφl+1 l dφl−1 + + t φl 2l + 1 dx 2l + 1 dx (3.195) 95 96 3 Neutron Diffusion Theory = N−1  l  =0 N   N   2l  + 1 wn Pl  (μn )Pl (μn ) + wn Pl (μn )S(μn ), sl  φl  2 n=0 n=1 l = 0, . . . , N − 1 (3.196) Weights chosen to satisfy Eqs. (3.194) obviously correctly integrate all polynomials through order N (any polynomial of order n can be written as a sum of Legendre polynomials through order n), but fortuitously they also integrate correctly all polynomials through order less than 2N . Thus, the term in the scattering integral becomes N   wn Pl  (μn )Pl (μn ) = 1 −1 n=1 Pl  (μ)Pl (μ)dμ = 2δll  2l + 1 (3.197) and assuming that the angular dependence of the source term can be represented by a polynomial of order < 2N N   wn Pl (μn )S(μn ) = n=1 1 −1 Pl (μ)S(μ)dμ = 2Sl 2l + 1 (3.198) where Sl is the Legendre moment of the source given by Eq. (3.173). Using Eqs. (3.197) and (3.198), Eqs. (3.196) become     l + 1 dφl+1 l dφl−1 + + (t − sl )φl = Sl 2l + 1 dx 2l + 1 dx l = 0, . . . , N − 2 dφ(N−1)−1 N −1 + (t − s,N−1 )φN−1 = SN−1 2(N − 1) + 1 dx (3.199) l =N −1 which, when φ−1 is set to zero, are identically the PL equations (3.172) for L = N − 1. References 1 D. R. Vondy, “Diffusion Theory,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations I, CRC Press, Boca Raton, FL (1986). 2 R. J. J. Stamm’ler and M. J. Abbate, Methods of Steady-State Reactor Physics in Nuclear Design, Academic Press, London (1983), Chap. 5. 3 J. R. Lamarsh, Introduction to Nuclear Reactor Theory, 2nd ed., Addison- Wesley, Reading, MA (1983), Chaps. 5, 6, 8, 9, and 10. 4 J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976), pp. 149–232 and 537–556. 5 A. F. Henry, Nuclear-Reactor Analysis, MIT Press, Cambridge, MA (1975), pp. 115–199. 6 G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand Rein- Problems hold, New York (1970), pp. 89–91, 104–105, and 151–157. 7 M. K. Butler and J. M. Cook, “One Dimensional Diffusion Theory,” and A. Hassitt, “Diffusion Theory in Two and Three Dimensions,” in H. Greenspan, C. N. Kelber, and D. Okrent, eds., Computing Methods in Reactor Physics, Gordon and Breach, New York (1968). 8 E. L. Wachspress, Iterative Solution of Elliptic Systems, Prentice Hall, Englewood Cliffs, NJ (1966). Another widely used method for solving the two-dimensional diffusion equations is the alternating direction implicit iteration scheme described in Section 14.3. 9 M. Clarke and K. F. Hansen, Numerical Methods of Reactor Analysis, Academic Press, New York (1964). 10 R. V. Meghreblian and D. K. Holmes, Reactor Analysis, McGrawHill, New York (1960), pp. 160–267 and 626–747. 11 A. M. Weinberg and E. P. Wigner, The Physical Theory of Neutron Chain Reactors, University of Chicago Press, Chicago (1958), pp. 181–218, 495–500, and 615–655. 12 S. Glasstone and M. C. Edlund, Nuclear Reactor Theory, D. Van Nostrand, Princeton, NJ (1952), pp. 90–136, 236– 272, and 279–289. Problems 3.1. Plot the neutron flux distribution given by Eq. (3.24) from r = 0 to r = 25 cm away from a point thermal neutron source in an infinite medium of (a) H2 O (L = 2.9 cm, D = 0.16 cm); (b) D2 O (L = 170 cm, D = 0.9 cm); and (c) graphite (L = 60 cm, D = 0.8 cm). 3.2. Plot the neutron flux distribution in a finite slab of width 2a = 10 cm with an incident thermal neutron beam from the left, as given by Eq. (3.36), for an iron slab (t = 1.15 cm−1 , D = 0.36 cm, L = 1.3 cm). 3.3. Derive the albedo boundary condition of Eq. (3.38) from the definition of the albedo, α ≡ j− /j+ , and the diffusion theory expressions for partial currents, Eqs. (3.4) and (3.5). 3.4. A thermal diffusion-length experiment is performed by placing a block of diffusing medium with aex = bex = 175.7 cm adjacent to a reactor thermal column port and irradiating a series of indium foils placed along the z-axis of the block. The saturation activity (disintegrations/min) of foils at various locations is (40,000 at z = 28 cm), (29,000 at z = 40 cm), (20,000 at z = 45 cm), (17,000 at z = 56 cm), (10,000 at z = 70 cm), (8500 at z = 76 cm), (5800 at 90 cm), and (3500 at 100 cm). The experimental error is ±10%. Determine the thermal neutron diffusion length. 97 98 3 Neutron Diffusion Theory 3.5. Derive the criticality condition for a bare rectangular parallelepiped core of x-dimension a, y-dimension b, and z-dimension c. 3.6. A typical composition for a PWR core is: H, 2.75 × 1022 cm−3 ; O, 2.76 × 1022 cm−3 ; Zr, 3.69 × 1021 cm−3 ; Fe, 1.71 × 1021 cm−3 ; 235 U 1.91 × 1020 cm−3 ; 238 U, 6.59 × 1021 cm−3 ; and 10 B, 1 × 1019 cm−3 . Appropriate spectrum-averaged microscopic cross sections (barns) for these isotopes are σtr /σa /νσf = 0.65/0.29/0.0 for H, 0.26/0.0002/0.0 for O, 0.79/0.19/0.0 for Zr, 0.55/2.33/0.0 for Fe, 1.62/484.0/758.0 for 235 U, 1.06/2.11/1.82 for 238 U, and 0.89/3410.0/0.0 for 10 B. Calculate the critical radius for a right circular cylindrical bare core of fixed height H = 375 cm. 3.7. Calculate the critical radius for the right circular cylindrical core of Problem 3.6 with a 20-cm-thick side reflector with DR = 1 cm and aR = 0.01 cm−1 . 3.8. Calculate the thermal flux disadvantage factor for UO2 rods varying from 0.5 to 2.0 cm in diameter in an H2 O moderator for VM /VF varying from 1.0 to 4.0. Calculate the corresponding effective homogeneous absorption cross sections and thermal utilization. Plot the results. 3.9. Derive an expression analogous to Eq. (3.100) for the power peaking factor in a fuel–moderator assembly with cylindrical fuel elements. 3.10. Derive an expression for the effective diffusion theory absorption cross section for a cylindrical control rod of radius a surrounded by an annular region of fuel–moderator extending from r = a to r = R. The transport parameter for this geometry is given by (1/3α) = 0.7104 + 0.2524/aaC + 0.0949/(aaC )2 + · · · . * 3.11. Jezebel is a bare, critical, spherical fast reactor assembly with radius 6.3 cm constructed of 239 Pu metal (density 15.4 g/cm3 ). Using the one-group constants ν = 2.98, σf = 1.85 barns (1 barn = 10−24 cm2 ) σa = 2.11 barns, and σtr = 6.8 barns and the finite-difference numerical method, calculate the effective multiplication constant, λ = keff , predicted by diffusion theory. λ − 1 is a measure of the accuracy of diffusion theory for this assembly. Should diffusion theory be valid for this assembly? * Problems 11 to 13 are longer problems suitable for take-home projects. Problems 3.12.* Solve numerically for the eigenvalue and neutron flux distribution in a slab reactor consisting of two adjacent core regions each of thickness 50 cm, with a 25-cm-thick reflector on each side. The nuclear parameters of the two core regions are (D = 0.65 cm, a = 0.12 cm−1 , and νf = 0.125 cm− 1) and (D = 0.75 cm, a = 0.10 cm−1 , and νf = 0.12 cm−1 ), and the parameters of the reflector are (D = 1.15 cm, a = 0.01 cm−1 , and νf = 0.0 cm−1 ). Solve this problem analytically and compare the answers. 3.13.* Calculate numerically the effective multiplication constant and the flux distribution in a reactor with rectangular (x, y) cross section which is sufficiently tall that axial (z) leakage can be neglected. The core cross section in the x–y plane consists of four symmetric quadrants. The upper right quadrant consists of core region 1, rectangular (0 < x < 50 cm, 60 < y < 100 cm); core region 2, rectangular (0 < x < 50 cm, 0 < y < 60 cm); and reflector region 3, also rectangular (50 < x < 100 cm, 0 < y < 100 cm). The nuclear parameters are: core region 1 (D = 0.65 cm, a = 0.12 cm−1 , νf = 0.185); core region 2 (D = 0.75 cm, a = 0.10 cm−1 , νf = 0.15); and reflector region 3 (D = 1.15 cm, a = 0.01 cm−1 , νf = 0.0 cm−1 ). Use vacuum boundary conditions except on the boundaries (x = 0, 0 < y < 100 cm) and (0 < x < 100 cm, y = 0), where a symmetry condition should be used. (This is a model for one-half of the symmetric reactor cross section.) Plot the x-direction flux distribution at y = 30 and 80 cm. 3.14. Calculate the thermal extrapolation distance λextrap for H2 O and for a 1:1 wt% homogeneous mixture of H2 O and 4% enriched uranium. 3.15. Estimate the maximum size of the mesh spacing that can be used in a finite-difference solution for the thermal neutron flux distribution in an H2 O medium and in a 1:1 wt% homogeneous mixture of H2 O and 4% enriched uranium. 3.16. Calculate and plot the thermal neutron flux distribution arising from a plane neutron source in an H2 O medium and in a 1:1 wt% homogeneous mixture of H2 O and 4% enriched uranium. 3.17. Repeat the calculation of Problem 3.16 in a carbon medium and in a 1:1 wt% homogeneous mixture of carbon and 4% enriched uranium. 3.18. Calculate the albedo boundary condition for the thermal neutron flux in a 1-m-thick slab medium with a 1:1 wt% 99 100 3 Neutron Diffusion Theory 3.19. 3.20. 3.21. 3.22. 3.23. 3.24. 3.25. 3.26. homogeneous mixture of H2 O and 4% enriched uranium which is bounded on both sides by very thick graphite slabs. Using the microscopic cross sections and number densities (except for 235 U) of Table 3.4, determine the critical 235 U enrichment for a bare cylindrical core of height H = 350 cm and radius R = 110 cm. Repeat the calculation for R = 100 and 120 cm. Repeat the calculation in Section 3.8 (Example 3.4) of flux disadvantage factor and effective homogenized fuel absorption cross section for a water thickness of 2 and 5 cm between fuel plates. Calculate the power peaking factor in the slab lattices of Problem 3.20. Repeat the calculation of the effective control slab cross section given in Section 3.9 (Example 3.5) for a control blade that contains only 2% natural boron. Solve Problem 3.12 using a four-node model, one node for each reflector and core region. Compare the result with the results of Problem 3.12. Discuss the approximations made in the derivation of neutron diffusion theory and how these approximations limit the validity of diffusion theory in reactor calculations. Explain why diffusion theory is not directly valid for the calculation of the flux distribution within the highly heterogeneous assembly of fuel, moderator, structure and control elements that make up a typical LWR fuel region. Describe how an effective homogeneous model of this heterogeneous assembly can be constructed for which diffusion theory is valid. Calculate the critical (k = 1) dimension of a bare cube of multiplying medium with average macroscopic cross sections νf = 0.115 cm−1 , a = 0.113 cm−1 , tr = 0.3 cm−1 . A single rod containing a spontaneous fission source is placed in a large volume of moderating material and surrounded at different radial distances with foils containing a material with a large 1/υ capture cross section. The foil activation is measured. Explain how the radial distribution of the foil activation rate could be used to determine the migration length of neutrons in the moderating material. 101 4 Neutron Energy Distribution Because the cross sections for neutron–nucleus reactions depend on energy, it is necessary to determine the energy distribution of neutrons in order to determine the rate of interactions of neutrons with matter, which in turn determines the transport of neutrons. We first address this problem by considering the neutron energy distribution in an infinite homogeneous medium, for which some analytical results can be obtained to provide physical insight. Then the important multigroup method for calculating an approximate neutron energy distribution is described. Methods for dealing with the rapidly varying neutron energy distribution in the energy range of cross-section resonances are described. Then the multigroup calculation of the neutron energy distribution is combined with the diffusion theory calculation of the spatial neutron distribution to obtain a powerful method for calculating the space- and energy-dependent neutron flux distribution in a nuclear reactor. 4.1 Analytical Solutions in an Infinite Medium We start our investigation of the neutron energy distribution in a nuclear reactor by considering an infinite homogeneous medium in which spatial effects may be ignored. The neutron flux within a differential energy interval dE is determined by a balance between the source of fission neutrons being created within dE plus neutrons being scattered into dE from some other energy interval dE  and the loss of neutrons from within dE due to absorption and to scattering from dE into some other energy interval dE  : a (E) + s (E) φ(E)dE   ∞     χ(E) ∞      dE  s E  → E φ E  + dE  νk E  φ E  dE (4.1) = k∞ 0 0 where we have included the infinite medium multiplication constant which may be adjusted to ensure that a steady-state solution exists. 102 4 Neutron Energy Distribution Fission Source Energy Range At very high energies, the direct source of fission neutrons into dE is much larger than the source of fission neutrons which have been created at higher energies and are slowing down into dE, in which case the first term on the right in Eq. (4.1) can be neglected in comparison to the second term, leading to φ (0) (E)  χ(E) K∞ t (E)  0 ∞     χ(E) dE  νf E  φ E  = × const. t (E) (4.2) where t = a + s . Thus the neutron flux distribution at the higher energies, where direct fission neutrons are the principal source, is proportional to the fission spectrum divided by the total cross section. This solution can be improved by using Eq. (4.2) as a first iterate on the right side of Eq. (4.1) to evaluate  ∞      1 dE  s E  → E φ (0) E  + χ(E) × const. φ (1) (E) = t (E) E  ∞   χ(E  )  1 dE  s E  → E + χ(E) × const. = t (E) E t (E  )    ∞  χ(E  )   1 χ(E)  × const. 1 + (4.3) dE s E → E = t (E) χ(E) E t (E  ) where we have taken advantage of the fact that scattering of very energetic neutrons with much less energetic nuclei will result in an energy loss for the neutron to place a lower limit of E on the energies from which a neutron can scatter into dE. At the higher energies, where the fission source is important, inelastic scattering is also important and must be included in calculation of the correction factor. The improved energy distribution is also of the form of the fission spectrum divided by the total cross section times 1 plus a correction factor that obviously becomes large at lower energies where χ(E) becomes small. Numerical evaluation of the correction factor for typical compositions indicates that φ(E) = χ(E)/t (E) represents the energy distribution rather well for energies E > 0.5 MeV. Slowing-Down Energy Range Very few fission neutrons are produced with energy less than about 50 keV. There is very little inelastic scattering in this energy range, so the elastic scattering transfer function  s (E  ) E     , E ≤ E ≤ (4.4) s E → E = E  (1 − α) α  0, otherwise can be used, where α ≡ [(A − 1)/(A + 1)]2 and A is the mass of the scattering nucleus in amu. If we limit our attention further to neutron energies greater than 4.1 Analytical Solutions in an Infinite Medium only about 1 eV, the neutrons will lose energy in a scattering collision, and we can write the slowing-down equation for the neutron energy distribution t (E)φ(E) =  E/αj dE  E j s (E  )φ(E  ) E  (1 − αj ) j (4.5) where the sum is over the various nuclear species present. Moderation by Hydrogen Only Consider a mixture of hydrogen αH ≡ [(AH − 1)/(AH + 1)]2 = 0 and very heavy mass nuclei α ≡ [(A − 1)/(A + 1)]2 ≈ 1, for which Eq. (4.5) becomes  t (E)φ(E) = E   ∞ ∞ E  j   dE   E/αj s (E  )φ(E  ) sH φ E  + dE    E E (1 − αj ) E j =H  sH φ E   dE  E +  1 j s (E)φ(E) αj (4.6) j =H where the range of integration E < E  < E/αj is so small for the heavy mass nuj j clei that the approximation s (E  )φ(E  )/E  ≈ s (E)φ(E)/E can be made. This equation can be rearranged:  a (E) + sH φ(E) = ∞ E sH φ(E  )  dE E (4.7) Differentiating Eq. (4.7), dividing both sides by (a +sH )φ and integrating from E to some arbitrary upper energy E1 leads to φ(E) =    E1 a (E  )dE  [a (E1 ) + sH ]E1 φ(E1 ) exp −  H  [a (E) + sH ]E E [a (E ) + s ]E (4.8) The neutron energy distribution varies with energy as φ(E) ∼ 1/(a (E)+sH )E and is exponentially attenuated in magnitude relative to the value at E1 by any absorption that occurs over the interval E1 > E  > E. The overall 1/E energy dependence of the flux is modified by the energy dependence of a (E). Energy Self-Shielding In particular, if a large narrow resonance is present, a (E) will increase sharply over the range of the resonance, causing φ(E) ∼ 1/(a (E) + sH ) to dip sharply over this range where the resonance cross section is large, as indicated in Fig. 4.1. At energies just below the resonance, where a (E) becomes small again, the flux recovers almost to the value just above the resonance, the difference being due to the absorption in the resonance. Physically, only those neutrons that are scattered into the energy range of the resonance will be absorbed, but those neutrons that 103 104 4 Neutron Energy Distribution Fig. 4.1 Energy self-shielding of the neutron flux in a large absorption resonance. (From Ref. 6; used with permission of MIT Press.) are scattered from energies above the resonance to energies below the resonance will not be affected by the presence of the resonance. This reduction in the neutron flux in the energy range of the resonance reduces the resonance absorption relative to what it would be if the effect of the resonance on the flux was not present, a phenomenon known as energy self-shielding. We can obtain a rough estimate of the importance of energy self-shielding by calculating the exponential attenuation due to the resonance under the simplifying assumption that the resonance is very large over an energy width E. Then the attenuation factor can be approximated:   exp − E1 E    E+E  E a (E  )dE  dE   exp − =  [a (E  ) + sH ]E  E E + E E (4.9) For the first large absorption resonance in 238 U at E = 6.67 eV, the width of the resonance is about E = 0.027 eV, which would absorb only about 4% of the neutrons slowing down past the resonance energy according to Eq. (4.9). Slowing Down by Nonhydrogenic Moderators with No Absorption The case of slowing down by only hydrogen provides valuable physical insight into features of the neutron energy distribution in the slowing-down energy range, most notably φ ∼ 1/E and energy self-shielding of resonances, which remains valid under other moderating conditions. To gain some insight into the effect of various moderators on the neutron energy distribution, we now consider the case of mod- 4.1 Analytical Solutions in an Infinite Medium eration by nonhydrogen isotopes, first in the absence of absorption. The slowingdown balance equation is j  j   E/αj s (E  )φ(E  ) s (E)φ(E) = dE  (4.10) s (E)φ(E) ≡ (1 − αj )E  E j j Guided by the result for slowing down from hydrogen, we look for a solution of the form φ(E) = C C =  j Es (E) E j s (E) (4.11) Substituting this into Eq. (4.10) leads to the identity  E/αj j   sj 1 C  s (E ) dE  C = = C  C  ) (1 − α )(E  )2 E  (E  E E s s j E j (4.12) j when it is assumed that the energy dependence of the scattering cross section is the same for all isotopes present, establishing that a solution of the form of Eq. (4.11) satisfies Eq. (4.10) under this assumption. Slowing-Down Density The slowing-down density at energy E, q(E), is defined as the rate at which neutrons are scattered from energies E  above E to energies E  below E. With reference to Fig. 4.2, this may be written  E j   E/αj s (E  )φ(E  )  dE dE   (4.13) q(E) ≡ E (1 − αj ) αj E  E j Fig. 4.2 Energy intervals for neutron slowing-down density. 105 106 4 Neutron Energy Distribution The maximum energy E  from which a neutron may scatter elastically below E is E/α, and the minimum energy E  of a neutron that scatters from an energy E  > E to an energy E  < E is E  = E  α. Without absorption, the slowing-down density is obviously constant in energy. Substituting Eq. (4.11) into Eq. (4.13) leads to q(E) =  dE   E j =C E/αj  j 1+ E αj E  αj ln αj 1 − αj dE   Cs (E  )  s (E )(1 − αj )(E  )2 j  s (E) s ≡C ξj ≡ C ξ̄ (E) s s (E) j j (4.14) j where the assumption of a common energy dependence of all scattering cross sections has been used again. The quantity ξj is the average logarithmic energy loss in a collision with a nucleus of species j that was discussed in Chapter 2, and ξ̄ is the effective logarithmic energy loss for the mixture of moderators. Using this result with Eq. (4.11) leads to the very important relationship between neutral slowingdown density and neutron flux: q(E) = ξ̄ s (E)Eφ(E) (4.15) Slowing Down with Weak Absorption Absorption removes neutrons from the slowing-down process and thereby reduces the slowing-down density as the energy decreases. Noting that decreasing energy corresponds to −dE, the change in slowing-down density due to absorption is described by dq(E) = a (E)φ(E) dE (4.16) Assuming weak absorption or localized (resonance) absorption near energy E, so that the flux given by Eq. (4.15) can be used to evaluate the scattering-in integral, the neutron balance equation yields a generalization of Eq. (4.15) for the case of weak or resonance absorption a (E) + s (E) φ(E) =  dE  E j q E/αj  j E/αj E s (E  )φ(E  ) E  (1 − αj ) j s (E  ) dE  1 q = (4.17)   2 ξ̄ (1 − αj ) s (E ) (E ) ξ̄ E j where again the assumption of similar energy dependence of the scattering cross sections for all species present has been used. Combining Eqs. (4.16) and (4.17) yields dq(E) a (E)q(E) = dE E ξ̄ [a (E) + s (E)] (4.18) 4.1 Analytical Solutions in an Infinite Medium which may be integrated from energy E up to some energy E1 to obtain   E1 a (E  )dE  q(E) = q(E1 ) exp −    E ξ̄ [a (E ) + s (E )]E (4.19) which describes the attenuation of the neutron slowing-down density due to absorption. Making use of Eq. (4.17) yields an expression for the energy dependence of the neutron flux φ(E) = [a (E1 ) + s (E1 )]ξ̄ (E1 )E1 φ(E1 ) [a (E) + s (E)]ξ̄ (E)E   E1 a (E  )dE  × exp −     E ξ̄ (E )[a (E ) + s (E )]E (4.20) The neutron flux with nonhydrogenic moderators and weak or resonant absorption has an energy dependence φ ∼ 1/ξ̄ t (E)E and is exponentially attenuated, a result very similar to that obtained for moderation by hydrogen only [Eq. (4.8)— note that ξ = 1 for hydrogen]. In particular, the energy self-shielding of resonances discussed previously is contained in the 1/t (E) dependence of the neutron flux. Fermi Age Neutron Slowing Down The assumption that the scattering cross sections of all moderating isotopes had the same energy dependence, which was made to obtain a relatively simple solution for slowing down by nonhydrogenic moderators, can be avoided in the case of heavy moderators. The neutron balance equation for slowing down by a mixture of moderators is j   E/αj s (E  )φ(E  ) (4.21) dE   t (E)φ(E) = E (1 − αj ) E j Based on the previous results, we expect that ξ̄ t (E)Eφ is a slowly varying j function of E. Thus we make a Taylor’s series expansion of s (E  )E  φ(E  ) about j s (E)Eφ(E): j    j E  s E  φ E  = Es (E)φ(E) + d E j Es (E)φ(E) ln + ··· d ln E E (4.22) in the scattering-in integrals on the right of Eq. (4.21). If the scattering-in interval E/αj to E is small (i.e., if αj ≡ [(Aj − 1)/(Aj + 1)]2 ≈ 1), it should be sufficient to retain only the first two terms in the Taylor’s series expansion, leading to   E/αj dE  t (E)φ(E) = (E  )2 (1 − αj ) E j  E d j j Es (E)φ(E) + · · · × Es (E)φ(E) + ln E d ln E 107 108 4 Neutron Energy Distribution =  j s (E)φ(E) j +    αj ln αj d 1 j 1+ Es (E)φ(E) + · · · E 1 − αj d ln E d E ξ̄ (E)s (E)φ(E) + · · · dE (4.23)    E1 a (E  )dE  E1 ξ̄ (E1 )s (E1 )φ(E1 ) exp −    E ξ̄ (E)s (E) E E ξ̄ (E )s (E ) (4.24) = s (E)φ(E) + which can be integrated to obtain φ(E) = This result for the energy distribution of the neutron flux is identical to the result obtained in Eq. (4.20) when a  s , but obtained under quite different assumptions. The assumptions in deriving Eq. (4.20) were that the absorption was weak, so that the no-absorption relationship between the slowing-down density and the flux could be used and that the energy dependence of the scattering cross sections was the same for all moderators in the mixture, in order to evaluate the scatteringj in integrals. The only assumption in deriving Eq. (4.24) was that s (E  )E  φ(E  ) varied slowly over the scattering-in interval E to E/αj . The important results we have obtained about the neutron energy distribution in the slowing-down region are φ(E) ∼ 1/ξ̄ (E)t (E)E, q ≈ ξ̄ (E)t (E)Eφ(E) and that both the neutron slowing-down density, q, and the neutron flux, φ, are attenuated exponentially by absorption during the slowing-down process. The expressions that we have developed for this exponential attenuation are qualitatively correct, but need to be refined to explicitly treat the resonance absorption which dominates in the slowing-down region. We return to this in Section 4.3. Neutron Energy Distribution in the Thermal Range Determination of the neutron energy distribution in the “thermal” range (E less than 1 eV or so) is complicated by a number of factors. The thermal motion of the nuclei is comparable to the neutron motion, with the consequences that the cross sections must be averaged over the nuclear motion and that a scattering event can increase, as well as decrease, the energy of the neutron. Since the thermal neutron energy is comparable to the binding energy of nuclei in material lattices, the recoil of the nucleus will be affected by the binding of the nucleus in the lattice, and the neutron scattering kinematics is more complex. Inelastic scattering in which the molecular rotational or vibrational states or the crystal lattice vibration state is changed also affects the scattering kinematics. At very low energies the neutron wavelength is comparable to the interatomic spacing of the scattering nuclei, and diffraction effects become important. Accurate calculation of thermal reaction rates requires both the calculation of appropriate cross sections characterizing thermal neutron scattering and calculation of the energy distribution of neutrons in the thermal range. Fortunately, most of the complex details of thermal neutron cross 4.1 Analytical Solutions in an Infinite Medium sections are not of great importance in nuclear reactor calculations, and reasonable accuracy can be obtained with relatively simple models. In this section we characterize the thermal neutron distribution and reaction rates from relatively simple physical considerations. We return to a more detailed discussion of thermal neutron cross sections and distributions in Chapter 12. The neutron balance equation in the thermal energy range is  a (E) + s (E) φ(E) = Eth     dE  s E  → E φ E  + S(E) (4.25) 0 where the scattering-in integral is from all energies in the thermal range E < Eth , and S(E) is the source of neutrons scattered into the thermal energy range from E > Eth . An equilibrium solution requires that the total number of neutrons downscattered into the thermal energy range be absorbed, assuming for the moment no leakage and no upscatter above Eth :  0 Eth dE a (E)φ(E) = q(Eth ) (4.26) where q(Eth ) is the neutron slowing-down density past Eth . Consider the situation that would obtain if there were no absorption and slowingdown source; that is, the neutron flux balance is  ∞     s (E)φ(E) = dE s E  → E φ E  (4.27) 0 where we have extended the upper limit on the integral to infinity under the assumption that the scattering to energies greater than Eth is zero. The principle of detailed balance places the following constraint on the scattering transfer cross section for a neutron distribution that is in equilibrium, regardless of the physical details of the scattering event:       v s E  → E M E  , T = vs E → E  M(E, T ) where M(E, T ) is the Maxwellian neutron distribution   √ E 2π E exp − M(E, T ) = kT (πkT )3/2 (4.28) (4.29) It can be shown that the Maxwellian neutron flux distribution,  1/2   2 2πn E E exp − kT (πkT )3/2 m   E E exp − ≡ φT kT (kT )2 φM (E, T ) = nv(E)M(E) = (4.30) satisfies Eq. (4.27). Thus the principal of detailed balance is sufficient to ensure that the equilibrium neutron distribution, in the absence of absorption, leakage, 109 110 4 Neutron Energy Distribution or sources, is a Maxwellian distribution characterized by the temperature T of the medium (i.e., the neutrons will come into thermal equilibrium with the moderator nuclei). The most probable energy of neutrons in a Maxwellian distribution is kT , and the corresponding neutron speed is vT = (2kT /m)1/2 . However, absorption, leakage, and a slowing-down source will distort the actual neutron distribution from a Maxwellian. Since most absorption cross sections vary as 1/v = 1/(E)1/2 , absorption preferentially removes lower-energy neutrons, effectively shifting the neutron distribution to higher energies than a Maxwellian at the moderator temperature T . A shift to higher energies can be represented approximately by a Maxwellian distribution with an effective “neutron temperature”   Ca Tn = T 1 + ξ s (4.31) where C must be determined experimentally. Leakage can be represented by modifying the absorption cross section to a → a + DB 2 . Since D = 1/3tr , leakage will preferentially remove higher-energy neutrons, offsetting the effect of absorption to some extent. In the slowing-down region E > Eth , the neutron flux distribution is 1/E, and the slowing-down source into the upper part of the thermal energy range will tend to make the flux 1/E. Thus the hardened Maxwellian must be corrected by the addition of a joining term  which is about unity for values of E/kTn > 10 and vanishes for values of E/kTn < 5: φ(E) = φM (E, Tn ) + λ (E/kTn ) E where λ is a normalization factor √ ( π/2) λ = φT 1 − ξ s /a (4.32) (4.33) The Maxwellian distribution has some useful properties insofar as calculation of the neutron absorption rate in the thermal energy range is concerned. Most absorption cross sections are 1/v; that is, a (E, T ) = a0 a (E0 )v0 = v v (4.34) where E0 = kT = 0.025 eV and v0 = (2kT /mn )1/2 = 2200 m/s. The total absorption rate integrated over the thermal energy range is  Ra = Eth dEa (E, T )vnM(E, Tn ) = a (E0 )v0 n0 ≡ a (E0 )φ0 (4.35) 0 The quantity φ0 = nv0 is the 2200 m/s flux, which when multiplied by the cross section evaluated at E = 0.025 eV yields the total absorption rate integrated over the thermal energy range. Most thermal data compilations include the 2200 m/s value 4.2 Multigroup Calculation of Neutron Energy Distribution in an Infinite Medium of the cross section (see Appendix A). From the definitions of φT = (2/π 1/2 )nv [Eq. (4.30)] and of φ0 = nv0 , the appropriate thermal group absorption cross section (the quantity that is multiplied by the integral of the neutron flux over the thermal energy range to recover Ra ) for a 1/v absorber in a Maxwellian neutron distribution at neutron temperature Tn is ath √  1/2 π T0 = a (E0 ) 2 Tn (4.36) Non-1/v correction factors have been developed to correct this expression for absorbers that are not 1/v. Summary The fission spectrum divided by the total cross section, φ(E) = χ(E)/t (E), represents the energy distribution rather well for energies E > 0.5 MeV. In the slowingdown range below the fission spectrum, E < 50 keV, and above the thermal range, E > 1 eV, φ(E) ∼ 1/ξ̄ (E)t (E)E represents the neutron energy distribution. In the thermal range, E < 1 eV, a hardened Maxwellian plus a 1/E correction at higher energies, φ(E) = φM (E, Tn ) + λ(E/kTn )/E, represents the neutron energy distribution. 4.2 Multigroup Calculation of Neutron Energy Distribution in an Infinite Medium Derivation of Multigroup Equations While the neutron energy dependences derived in Section 4.1 provide a qualitative, even semiquantitative description of the neutron energy distribution in nuclear reactors, the multigroup method is widely used for the quantitative calculation of the neutron energy distribution. As we will see, the qualitative results of Section 4.1 will provide valuable insight as to the choice of weighting functions to be used in the preparation of multigroup constants. To develop a multigroup calculational method for the energy distribution, we divide the energy interval of interest, say 10 MeV down to zero, into G intervals, or groups, as indicated in Fig. 4.3. The equation describing the neutron energy distribution in a very large homogeneous region of a nuclear reactor (where spatial and leakage effects may be neglected) is a (E) + s (E) φ(E)   ∞    χ(E) ∞      dE  s E  → E φ E  + dE  νf E  φ E  = k∞ 0 0 (4.37) This equation can be integrated over the energy interval Eg < E < Eg−1 of group g to obtain 111 112 4 Neutron Energy Distribution Fig. 4.3 Multigroup energy structure.  Eg−1 Eg  = dE a (E) + s (E) φ(E) G   Eg−1 dE Eg + Eg  −1 g  =1 Eg  1 k∞      dE  s E  → E φ E  Eg−1 dEχ(E) Eg G   Eg  −1 g  =1 Eg      dE  νf E  φ E  (4.38) where we have made use of the fact that the sum of integrals over the groups is equal to the integral over 0 < E < ∞. Defining the integral terms in Eq. (4.38) in a natural way,  φg ≡ g a ≡ g  →g s  Eg−1 χg ≡ dE φ(E), Eg Eg−1 Eg dE a (E)φ(E) φg ≡ Eg−1 Eg dE Eg  −1 Eg  Eg−1 dE χ(E) Eg g νf ≡ , dE  s (E  → E)φ(E  ) φg  , g s ≡ Eg−1 Eg G  dE νf (E) φg g→g  s g  =1 (4.39) Eq. (4.38) can be written as  G G  χg  g g g  →g g a + s φg = s φg  + νf φg  , k ∞   g =1 g =1 g = 1, . . . , G (4.40) 4.2 Multigroup Calculation of Neutron Energy Distribution in an Infinite Medium Equations (4.40) are the multigroup neutron spectrum equations for an infinite medium, one in which spatial and leakage effects are unimportant. There are G equations and G unknowns, the group fluxes φg , so the problem is well posed. This overlooks the fact that the group constants  g depend on the neutron flux and hence are also unknown. Actually, the group constants depend only on the energy dependence of the neutron flux within the group, not on the magnitude of the neutron flux, which appears in both the numerator and denominator of the definition of the group constants. In practice, some assumption is made about this energy dependence, so that the group constants are known. From the results of the preceding section, we have a pretty good idea about the energy dependence of the neutron flux in the fission, slowing-down, and thermal energy ranges, which can be used to evaluate group constants. Summing Eqs. (4.40) over groups yields G g g=1 νf φg g g=1 a φg k∞ = G (4.41) which identifies k∞ as the ratio of the total neutron production rate by fission to the total neutron absorption rate, in accord with our previous discussion of the multiplication constant. Mathematical Properties of the Multigroup Equations The set of equations (4.40) may be written in matrix notation as Aφ −   1 1 Fφ = A − F φ=0 k∞ k∞ (4.42) where A and F are G × G matrices and φ is a G-element column vector: ⎡ ⎤ −s2→1 −s3→1 · · · −sG→1 a1 + s1 − s1→1 ⎢ ⎥ −s1→2 a2 + s2 − s2→2 −s3→2 · · · −sG→2 ⎢ ⎥ A=⎢ ⎥ .. .. .. .. .. ⎣ ⎦ . . . . . −s2→G −s3→G · · · aG + sG − sG→G −s1→G (4.43) ⎤ ⎡ ⎤ ⎡ χ1 νf1 χ1 νf2 χ1 νf3 · · · χ1 νfG φ1 ⎢ χ ν 1 χ ν 2 χ ν 3 · · · χ ν G ⎥ ⎢ φ2 ⎥ 2 2 2 ⎢ 2 f f f f ⎥ ⎥ ⎢ ⎥, φ=⎢ . ⎥ F =⎢ .. .. .. .. .. ⎥ ⎢ . ⎦ ⎣ ⎦ ⎣ . . . . . . G 3 1 2 φ χG νf χG νf χG νf · · · χg νf G g g→g Note that the scattering terms on the diagonal are of the form s − s , leadg g g g→g ing to the concept of a removal cross section r ≡ a + s − ss to represent the net loss of neutrons from group g by absorption plus scattering. 113 114 4 Neutron Energy Distribution Equations (4.40) or (4.42) are homogeneous equations and thus, by Cramer’s rule, have nontrivial solutions only if the determinant of the coefficient matrix vanishes:   1 det A − F =0 (4.44) k∞ This condition defines an eigenvalue problem for the determination of k∞ —there are only a certain set of G discrete values of k∞ for which a nontrivial solution exists. [Note that we have included k∞ in the formulation for just this reason. If we had not included k∞ , Eq. (4.44) would be a requirement on the composition of the reactor for criticality, and we would be faced with the cumbersome requirement to adjust the composition by trial and error until Eq. (4.44) was satisfied.] It is possible to prove that the inverse of the matrix A exists for any physically real set of cross sections and number densities. Multiplying Eq. (4.42) by k∞ A−1 yields k∞ φ = A−1 F φ (4.45) which is the standard form for a matrix eigenvalue problem. It is possible to prove (e.g., Refs. 8, 11, and 12) for this equation that (1) there is a unique real, positive eigenvalue greater in magnitude than any other eigenvalue; (2) all of the elements—the group fluxes—of the eigenvector corresponding to this largest eigenvalue are real and positive; and (3) the eigenvectors corresponding to all other eigenvalues have zero or negative elements. Thus the largest value of k∞ for which Eq. (4.44) is satisfied is real and positive and the associated group fluxes given by Eq. (4.45) are real and positive (i.e., physical). Solution of Multigroup Equations The multigroup equations have been written in their full generality, allowing upscatter (the terms above the diagonal in A) as well as downscatter (the terms below the diagonal in A) and a fission spectrum contribution in every group. In fact, upscatter takes place only for those groups that are in the thermal energy range E 1 eV, and the fission spectrum contributes only to the higher-energy groups E 50 keV. Taking these physical considerations into account greatly simplifies solution of the multigroup equations. Consider, as the simplest example of a multigroup description, the representation of the neutrons in a nuclear reactor as being either in a thermal group (E 1 eV) or in a fast group (E 1 eV). All of the fission neutrons are deposited in the fast group, and there is no upscatter from the thermal to the fast group. The two-group equations are    1  1 νf φ1 + νf2 φ2 a1 + s1→2 φ1 = k∞ a2 φ2 = s1→2 φ1 (4.46) 4.2 Multigroup Calculation of Neutron Energy Distribution in an Infinite Medium which may readily be solved for φ1 = a2 φ2 , s1→2 k∞ = νf1 + (s1→2 /a2 )νf2 a1 + s1→2 (4.47) Note that a critical reactor may operate at many power levels, so the absolute magnitude of the group fluxes quite properly cannot be determined by the set of homogeneous multigroup equations, but the relative magnitudes of the different group fluxes can be determined. A somewhat better multigroup description results from representing the fission interval (E > 50 keV) as a fast group into which all fission neutrons are introduced, the slowing-down interval (50 keV > E > 1 eV) as an intermediate group, and the thermal region (E < 1 eV) as a thermal group. There would be no upscattering in such a group structure, allowing the three-group equations to be written    1 1  1 νf φ1 + νf2 φ2 + νf3 φ3 a + s1→2 + s1→3 φ1 = k∞  2  2→3 1→2 a + s φ2 = s φ1 a3 φ3 = s1→3 φ1 (4.48) + s2→3 φ2 with solutions   φ2 = s1→2 / a2 + s2→3 φ1 , !   s2→3 1→3 1→2 + 2  φ3 = s a2 φ1 a + s2→3 s  k∞ = νf1 + νf2 s1→2 2 a + s2→3  + νf3 s1→3 + s2→3  1→2 a2 + s2→3 s ! (4.49)   1 a + s1→2 + s1→3 Example 4.1: Two-Group Fluxes and k∞ . A representative set of two-group cross sections for a PWR fuel assembly are (s1→2 = 0.0241 cm−1 , a1 = 0.0121 cm−1 , νf1 = 0.0085) and (a2 = 0.121 cm−1 , νf2 = 0.185). From Eq. (4.47) the fast/thermal flux ratio is φ1 /φ2 = 0.121/0.241 = 5.02, and k∞ = (0.0085 + 0.185/5.02)/(0.0121 + 0.0241) = 1.253. The spectrum-averaged one-group absorption cross section is a = (a1 φ1 + a2 φ2 )/(φ1 + φ2 ) = 0.0302 cm−1 . Preparation of Multigroup Cross-Section Sets There exist in the world several sets of evaluated nuclear data (e.g., Refs. 7 and 9), which have been both checked for consistency and benchmarked extensively in the calculation of experiments designed for data testing. Representation of the crosssection data in such data files is generally as follows: 115 116 4 Neutron Energy Distribution 1. σ (Ei ) are tabulated pointwise in energy at low energies below the resonance region. 2. Resolved resonance parameters and background cross sections in the resolved resonance region. 3. Unresolved resonance statistical parameters and background cross section in the unresolved resonance region. 4. σ (Ei ) are tabulated pointwise in energy at energies above the resonance region. 5. Scattering transfer functions p(Ei , μs ) are tabulated pointwise in energy and either pointwise in angle (μsj ) or as Legendre coefficients. The resonance parameters and the construction of multigroup cross sections from them are discussed in Section 4.3 and in Chapter 11. The scattering transfer function—the probability that a neutron will undergo a scattering event which changes its direction from direction  to direction  (μs =  ·  ) and its energy from E to E  —is represented as     (4.50) σs μs , E → E  = m(E)σs (E)p(E, μs )g μs , E → E  where m(E) = 1 for elastic and inelastic scattering, 2 for (n, 2n), ν for fission; p(E, μs ) is the angular distribution for scattering of a neutron of energy E; and g(μs , E → E  ) is the final energy distribution of a neutron at energy E which has scattered through μs . When the scattering angle and energy loss are correlated, as they are for elastic scattering, E  /E = [(1 + α) + (1 − α) cos θ]/2 and g(μs , E → E  ) = δ(μs − μ(E, E  )). Otherwise, g(μsi , Ej → Ek ) is tabulated. The angular distribution may be tabulated as p(Ei , μsj ), or the Legendre components may be tabulated pointwise in energy  1 dμs Pn (μs )p(Ei , μs ) (4.51) pn (Ei ) = −1 where Pn is the Legendre polynomial. There are a number of codes (e.g., Refs. 2, 4, and 5) which directly process the evaluated nuclear data files to prepare multigroup cross sections. These codes numerically calculate integrals of the type g σ = g→g  σn Eg−1 dEσ (E)W (E) Eg Eg−1 dEW (E) Eg  = Eg−1 Eg dEσs (E)W (E)  Eg  −1 Eg  "   dE  pn E  (4.52) Eg−1 dEW (E) Eg for a specified weighting function, W (E), which may be a constant, 1/E, χ(E), and so on. These codes are used to calculate fine-group cross sections in a few hundred groups for thermal reactors or ultrafine-group cross sections in a few thousand groups for fast reactors. These fine- or ultrafine-group structures are chosen such 4.3 Resonance Absorption that the results of calculations using the fine- or ultrafine-group cross sections are essentially independent of the choice of weighting function, W (E), used in the cross-section preparation. Once the fine- or ultrafine-group cross sections are prepared, a fine- or ultrafinegroup spectrum (φg ) is calculated for a representative homogenized medium. The unit cell heterogeneous structure of the region must be taken into account in homogenizing the medium. Resonances must be treated specially, as discussed in Chapter 11. This fine- or ultrafine-group spectrum can then be used to weight the fine- or ultrafine-group cross sections to obtain few-group (2 to 10) cross sections for thermal reactors or many-group (20 to 30) cross sections for fast reactors:  g g∈k σ φg σk =  g∈k φg (4.53)   g→g  φ  ∈k  σn g  g∈k g  σnk→k = g∈k φg The notation g ∈ k indicates that the sum is over all fine or ultrafine groups g within few or many group k. The few- or many-group cross sections may be calculated for several different large regions in a reactor. They are then used in a few- or many-group diffusion or transport theory calculation of the entire reactor to determine the effective multiplication constant, power distribution, and so on. Because many such calculations must be made, a number of parameterizations of few- and many-group cross sections have been developed (e.g., Ref. 10) to avoid the necessity of making the fineor ultrafine-group spectrum calculation numerous times. 4.3 Resonance Absorption Resonance Cross Sections When the relative (center-of-mass) energy of an incident neutron and a nucleus plus the neutron binding energy match an energy level of the compound nucleus that would be formed upon neutron capture, the probability of capture is quite large. The lowest-energy excited states are only a fraction of 1 eV above the ground state and extend up to about 100 keV for heavy mass fuel nuclei (fissile and fertile), but start at about 10 eV for intermediate mass nuclei and at about 10 keV for lighter mass nuclei. The heavier mass isotopes have many relatively low energy excited states, which give rise to resonances in the neutron absorption and scattering cross sections (Fig. 4.4). The neutron resonance absorption phenomena constitute one of the most fundamental subjects in nuclear reactor physics. One of the most effective means of treating these phenomena is in terms of the resonance integral concept, which has a fundamental premise that the resonance cross sections are representable 117 118 4 Neutron Energy Distribution Fig. 4.4 238 U capture cross section. (From http://www.nndc.bnl.gov/.) by superposition of many Breit–Wigner resonances with known parameters. This premise allows the complex resonance structure to be characterized in a reasonably simple way by calculating the contributions of each individual resonance. The discussion in this section concentrates on s-wave neutron cross sections in the low-energy range. As shown in Chapter 1, the (n, γ ) capture cross section averaged over the motion of the nucleus is given by σγ (E, T ) = σ0 γ  E0 E 1/2 ψ(ξ, x) (4.54) and the total scattering cross section, including resonance and potential scattering and interference between the two, can be written σs (E, T ) = σ0 n ψ(ξ, x) + σ0 R χ(ξ, x) + 4πR 2 λ0 (4.55) where R is the nuclear radius, λ0 the neutron DeBroglie wavelength, the functions  ∞ dy ξ 2 2 e−1/4(x−y) ξ (4.56) ψ(ξ, x) = √ 1 + y2 2 π −∞  ∞ ξ 2 2 y dy e−1/4(x−y) ξ (4.57) χ(ξ, x) = √ 1 + y2 π −∞ are integrals over the relative motion of the neutron and nucleus, x = 2(E −E0 )/ , it has been assumed that the nuclear motion can be characterized by a Maxwellian 4.3 Resonance Absorption Fig. 4.5 Temperature broadening of the ψ -function. (From Ref. 3; used with permission of Wiley.) distribution with temperature T , and E is the energy of the neutron in the lab system. The parameters characterizing the resonance are σ0 , the peak value of the cross section; E0 , the neutron energy in the center-of-mass system at which it occurs; , the resonance width; γ , the partial width for neutron capture; f , the partial width for fission; and n , the partial width for scattering. The resonance occurs when the center of mass energy Ecm = (A/(A + 1))E plus the change in binding energy upon neutron capture equals the energy above ground state of an excited level of the compound nucleus. Doppler Broadening The temperature characterizing the nuclear motion is contained in the parameter ξ= (4E0 kT /A)1/2 (4.58) where A is the atomic mass (amu) and k is the Boltzmann constant. The general dependence of the ψ -function on temperature is indicated in Fig. 4.5. As the temperature increases, the peak magnitude of ψ at E0 decreases and the magnitude away from peak increases. This broadening of the cross section is known as Doppler broadening. It can be shown that the area under the curve of the ψ -function remains constant as the temperature changes. Similar behavior results for the χ -function. The ψ - and χ -functions are tabulated in Tables 4.1 and 4.2. The assumption that the nuclear motion can be characterized by a Maxwellian is only approximately correct for atoms bound in a crystalline state. Investigation of this point indicates that a Maxwellian is a good approximation, but with a slightly higher temperature which corresponds to the average energy per vibrational degree 119 0.04309 0.08384 0.12239 0.15889 0.19347 0.22624 0.25731 0.28679 0.31477 0.34135 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.04308 0.08379 0.12223 0.15854 0.19281 0.22516 0.25569 0.28450 0.31168 0.33733 0.5 0.04306 0.08364 0.12176 0.15748 0.19086 0.22197 0.25091 0.27776 0.30261 0.32557 1 Source: Data from Ref. 3; used with permission of Wiley. 0 ξ Table 4.1 ψ -Function 0.04298 0.08305 0.11989 0.15331 0.18324 0.20968 0.23271 0.25245 0.26909 0.28286 2 0.04267 0.08073 0.11268 0.13777 0.15584 0.16729 0.17288 0.17359 0.17052 0.16469 4 x 0.04216 0.07700 0.10165 0.11540 0.11934 0.11571 0.10713 0.09604 0.08439 0.07346 6 0.04145 0.07208 0.08805 0.09027 0.08277 0.07042 0.05724 0.04566 0.03670 0.03025 8 0.04055 0.06623 0.07328 0.06614 0.05253 0.03880 0.02815 0.02109 0.01687 0.01446 10 0.03380 0.03291 0.01695 0.00713 0.00394 0.00314 0.00289 0.00277 0.00270 0.00266 20 0.01639 0.00262 0.00080 0.00070 0.00067 0.00065 0.00064 0.00064 0.00064 0.00063 40 120 4 Neutron Energy Distribution 0 0 0 0 0 0 0 0 0 0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.00120 0.00458 0.00986 0.01680 0.02515 0.03470 0.04529 0.05674 0.06890 0.08165 0.5 0.00239 0.00915 0.01968 0.03344 0.04994 0.06873 0.08940 0.11160 0.13498 0.15927 1 2 0.00478 0.01821 0.03894 0.06567 0.09714 0.13219 0.16976 0.20890 0.24880 0.28875 Source: Data from Ref. 3; used with permission of Wiley. 0 ξ Table 4.2 χ -Function 0.00951 0.03573 0.07470 0.12219 0.17413 0.22694 0.27773 0.32442 0.36563 0.40075 4 x 0.01415 0.05192 0.10460 0.16295 0.21909 0.26757 0.30564 0.33286 0.35033 0.35998 6 0.01865 0.06626 0.12690 0.18538 0.23168 0.26227 0.27850 0.28419 0.28351 0.27979 8 0.02297 0.07833 0.14096 0.19091 0.22043 0.23199 0.23236 0.22782 0.22223 0.21729 10 0.04076 0.10132 0.12219 0.11754 0.11052 0.10650 0.10437 0.10316 0.10238 0.10185 20 0.05221 0.05957 0.05341 0.05170 0.05103 0.05069 0.05049 0.05037 0.05028 0.05022 40 4.3 Resonance Absorption 121 122 4 Neutron Energy Distribution of freedom of the lattice, including the zero-point energy. In practice, the actual material temperature is widely used. Resonance Integral The total absorption rate per nuclei by a resonance absorber is known as the resonance integral,  Iγ = σγ (E)φ(E) dE (4.59) Resonance Escape Probability The absorption probability for a single resonance depends on the balance between absorption and moderation and is given by Rabs = Nres I /q0 , where q0 = ξ s Eφasy is the asymptotic slowing-down density above the resonance and Nres is the number density of the resonance absorber. If we use φasy = 1/E to evaluate the resonance integral, then Rabs = I /ξ σs , where the denominator is the moderating power per absorber nucleus. The resonance escape probability is p = 1 − Rabs = 1 − I /ξ σs ≈ exp(−I /ξ σs ), where Rabs is assumed small for any one resonance. The quantity σs = (Nres σsres + Nnon σsnon )/Nres is the effective scattering cross section per resonance absorbing atom, including both the resonance and non-resonance species present. The total resonance integral for all resonances is a sum over the individual resonance integrals, and the total resonance escape probability is   # 1  (4.60) p= pi = exp − Ii ξ σs i i Multigroup Resonance Cross Section The resonances within a given energy group in a multigroup treatment can be treated as a group capture cross section given by σγg = Eg−1 dEσγ (E)φ(E) Eg Eg−1 dEφ(E) Eg  = i∈g Ii ln(Eg−1 /Eg ) (4.61) where φ(E) ∼ 1/E has been used. Practical Width The practical width of a resonance is defined as the energy range over which the resonance cross section is larger than the nonresonance part of the cross section of the resonance nuclide, which from the Breit–Wigner formula is $ $ σ0 σ0 = (4.62) p 2 σp 4πR 4.3 Resonance Absorption Typically, for low-energy resonances σ0 /4πR 2 ≡ σ0 /σp ∼ 103 , so the practical width is much larger than the total width. The practical width provides a measure of the range of influence of the resonance, which we will see is important in evaluating the neutron flux in the resonance. Neutron Flux in Resonance The resonance region is well below the fission spectrum, so the neutron balance in the vicinity of the resonance can be written  tres (E) + sM φ(E) = E/αM E  + dE sM φ(E  ) E  1 − αm E/αres E dE  sres (E  )φ(E  ) E 1 − αres (4.63) where the moderator scattering cross section is assumed to be much larger than its absorption cross section and to be effectively constant. The practical width of the resonance will generally be much less than the scattering-in interval of the moderator, p  E0 (1 − αM ). For widely spaced resonances, this allows the approximate evaluation of the moderator scattering source term with the asymptotic form of the neutron flux in the absence of resonances, φasy ∼ 1/ξ sM E. We choose the normalization φasy = 1/E above the resonance energy to obtain tres (E) + sM φ(E) = sM + E  E/αres E dE  sres (E  )φ(E  ) E 1 − αres (4.64) Narrow Resonance Approximation If the practical width of the resonance is also small compared to the scattering-in interval of the resonance absorber, p  E0 (1 − αres ), then the second scattering source term can be approximated in the same fashion to obtain φNR (E) = sM + pres res [t (E) + sM ]E (4.65) which can be used in Eq. (4.59) to evaluate the resonance integral:  sM + pres dE σγ (E) res E t (E) + sM   ∞ ψ(ξ, x) dx γ  res = σp + σsM 2E0 ψ(ξ, x) + θχ(ξ, x) + β −∞ γ INR = (4.66) where β= σsM + σpres σ0  , θ= n σpres σ0 1/2 (4.67) 123 124 4 Neutron Energy Distribution σsM is the moderator scattering cross section per absorber nucleus and σpres = 4πR 2 is the potential scattering cross section of the resonance absorber. If interference between resonance and potential scattering is neglected, the resonance integral can be written γ γ INR = E0  res  σp + σsM J (ξ, β) where the function  ∞ ψ(ξ, x) dx J (ξ, β) ≡ ψ(ξ, x) + β 0 (4.68) (4.69) is tabulated in Table 4.3. A generalization of the J -function which includes the interference scattering term has been devised, but the form given above is more commonly used. Wide Resonance Approximation If the practical width of the resonance is large compared to the scattering-in interval of the resonance absorber, p  E0 (1 − αres ), the second scattering source term in Eq. (4.64) can be approximated by assuming that sres (E  )φ(E  )/E  ≈ sres (E)φ(E)/E, which leads to φWR (E) = sM res [t (E) − sres (E) + sM ]E (4.70) Using this result to evaluate the resonance integral defined by Eq. (4.59) yields  IWR =   dE sM σγ (E) res = σ M J ξ, β  E t (E) − sres (E) + sM E0 s (4.71) where β = σsM σ0 (4.72) γ Resonance Absorption Calculations Data for several of the low-energy resonances in 238 U are given in Table 4.4. Also shown is a comparison of the absorption probabilities calculated with the narrow and wide resonance approximations with an “exact” solution obtained numerically, for a representative fuel-to-moderator ratio for a thermal reactor. The WR approximation is more suitable for the lowest-energy resonances, but the narrow resonance approximation generally is preferable for all but the lowest-energy resonances. ξ = 0.1 4.979(2) 3.532 2.514 1.801 1.307 9.667(1) 7.355 5.773 4.647 3.781 3.045 2.367 1.730 1.164 7.172(0) 4.088 2.204 1.148 5.862(−1) 2.963 1.490 j 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 4.970(2) 3.517 2.491 1.767 1.257 8.993(1) 6.501 4.777 3.589 2.759 2.153 1.676 1.268 9.081(0) 6.014 3.658 2.067 1.109 5.757(−1) 2.936 1.483 ξ = 0.2 Table 4.3 J -Function (β = 2j × 10−5 )∗ 4.969(2) 3.514 2.487 1.761 1.248 8.872(1) 6.335 4.562 3.328 2.471 1.867 1.423 1.074 7.815(0) 5.342 3.371 1.966 1.078 5.671(−1) 2.913 1.477 ξ = 0.3 4.968(2) 3.513 2.485 1.759 1.245 8.831(1) 6.278 4.485 3.230 2.354 1.741 1.301 9.718(0) 7.087 4.914 3.169 1.889 1.053 5.599(−1) 2.894 1.472 ξ = 0.4 ξ = 0.6 4.968(2) 3.513 2.484 1.757 1.243 8.802(1) 6.238 4.430 3.158 2.265 1.638 1.194 8.739(0) 6.322 4.419 2.911 1.781 1.016 5.488(−1) 2.863 1.464 J (ξ , β) 4.968(2) 3.513 2.485 1.758 1.244 8.812(1) 6.252 4.450 3.183 2.297 1.675 1.235 9.119(0) 6.629 4.624 3.022 1.829 1.033 5.539(−1) 2.877 1.468 ξ = 0.5 4.967(2) 3.513 2.484 1.757 1.243 8.796(1) 6.230 4.419 3.143 2.245 1.614 1.168 8.484(0) 6.107 4.268 2.826 1.743 1.002 5.445(−1) 2.851 1.461 ξ = 0.7 4.967(2) 3.513 2.484 1.757 1.243 8.792(1) 6.225 4.412 3.133 2.232 1.598 1.151 8.304(0) 5.950 4.154 2.759 1.712 9.904(−1) 5.408 2.840 1.458 ξ = 0.8 4.967(2) 3.513 2.484 1.757 1.242 8.790(1) 6.221 4.407 3.126 2.223 1.587 1.138 8.174(0) 5.833 4.066 2.706 1.687 9.805(−1) 5.376 2.831 1.455 ξ = 0.9 4.967(2) 3.513 2.484 1.757 1.242 8.788(1) 6.218 4.403 3.121 2.217 1.579 1.129 8.077(0) 5.744 3.997 2.663 1.666 9.722(−1) 5.348 2.823 1.453 ξ = 1.0 4.3 Resonance Absorption 125 7.468(−2) 3.739 1.871 9.358(−3) 4.680 2.340 1.170 5.851(−4) 2.925 1.463 7.314(−5) 21 22 23 24 25 26 27 28 29 30 31 7.452(−2) 3.735 1.870 9.356(−3) 4.680 2.340 1.170 5.851(−4) 2.926 1.463 7.314(−5) ξ = 0.2 7.437(−2) 3.732 1.869 9.355(−3) 4.679 2.340 1.170 5.851(−4) 2.926 1.463 7.315(−5) ξ = 0.3 7.424(−2) 3.728 1.868 9.352(−3) 4.679 2.340 1.170 5.851(−4) 2.926 1.463 7.315(−5) ξ = 0.4 Source: Data from Ref. 3; used with permission of Wiley. ∗ Numbers in parentheses are powers of 10, which multiply the entry next to which they stand and all unmarked entries below it. ξ = 0.1 j Table 4.3 (Continued) ξ = 0.6 7.403(−2) 3.723 1.867 9.349(−3) 4.678 2.340 1.170 5.851(−4) 2.926 1.463 7.315(−5) J (ξ, β) 7.413(−2) 3.726 1.868 9.350(−3) 4.678 2.340 1.170 5.851(−4) 2.926 1.463 7.315(−5) ξ = 0.5 7.395(−2) 3.721 1.867 9.348(−3) 4.678 2.340 1.170 5.851(−4) 2.926 1.463 7.314(−5) ξ = 0.7 7.388(−2) 3.719 1.866 9.346(−3) 4.677 2.340 1.170 5.851(−4) 2.926 1.463 7.314(−5) ξ = 0.8 7.381(−2) 3.718 1.866 9.345(−3) 4.677 2.340 1.170 5.851(−4) 2.926 1.463 7.314(−5) ξ = 0.9 7.375(−2) 3.716 1.865 9.344(−3) 4.677 2.340 1.170 5.851(−4) 2.926 1.463 7.314(−5) ξ = 1.0 126 4 Neutron Energy Distribution 4.4 Multigroup Diffusion Theory Table 4.4 Low-Energy 238 U Resonances (1 − αres E 0 ) E 0 (eV) 6.67 20.90 36.80 116.85 208.46 n (eV) 0.00152 0.0087 0.032 0.030 0.053 γ (eV) σ0 (barns) p 0.026 0.025 0.025 0.022 0.022 2.15 × 105 3.19 × 104 3.98 × 104 1.30 × 104 8.86 × 103 1.26 1.95 3.65 1.32 2.63 (eV) 1−p (eV) NR WR Exact 0.110 0.348 0.612 0.966 1.73 0.2376 0.07455 0.04739 0.00904 0.00444 0.1998 0.07059 0.06110 0.00950 0.00769 0.1963 0.06755 0.05820 0.00917 0.00502 Source: Data from Ref. 3; used with permission of Wiley. Example 4.2: Group Capture Cross Section for 6.67-eV 238 U Resonance. The contribution of the 6.67-eV 238 U resonance to the capture cross section of an energy group extending from 1 to 10 eV is calculated in the narrow resonance approxγ γ g imation from σγ = INR / ln(10/1), where INR = ( γ /E0 )(σpres + σsM )J (ξ, β). For uranium, σpres = 8.3 barns. For a moderator cross section per fuel atom of σsM = sM /NF = 60 barns and a temperature T = 330◦ C, ξ = ( /2)/(E0 kT /A)1/2 = (0.0275 eV/2)/[(6.67 eV × 603 K × 0.86 × 10−4 eV/K)/238]1/2 = 0.361 and β = (σpres + σsM )/σ0 = (60 + 8.3)/2.15 × 105 = 31.8 × 10−5 = 2j × 105 , or j = 4.98. Inγ terpolating on j and ξ in Table 4.3 yields J ≈ 88. With these values, INR ≈ 23 and g as σγ ≈ 10 barns. Temperature Dependence of Resonance Absorption Examination of the function J (ξ, β) of Eq. (4.69) reveals that for any value of β, the value of J increases or remains constant as ξ decreases. Since ξ ∼ 1/T 1/2 , the resonance absorption must increase or remain unchanged when the temperature increases. The physical reason for this is that as the temperature increases, the cross section (averaged over nuclear motion) decreases in peak value and broadens in energy in such a manner as to preserve the area under the cross-section curve, as indicated in Fig. 4.5, but the decreasing value of the cross section results in a decreasing depression in the neutron flux in the resonance region. This increase in absorption cross section with increasing fuel temperature introduces an important negative-feedback Doppler temperature coefficient of reactivity, which is important for reactor safety, as discussed in Chapter 5. 4.4 Multigroup Diffusion Theory Multigroup Diffusion Equations We consider cohorts of neutrons of different energies diffusing within a nuclear reactor. The basic diffusion equation for each cohort, or group, of neutrons is the 127 128 4 Neutron Energy Distribution same as derived in Chapter 3, but with absorption generalized to all processes that remove the neutron from the cohort or group (i.e., absorption plus scattering to another group) and with the source of neutrons for each group specialized to include the in-scatter of neutrons from other groups, which are also diffusing within the reactor: g −∇ · D g (r)∇φg (r) + r (r)φg (r) = G  g  →g s g  =g G  1 g (r)φ(r)g  + χ g νf (r)φg  (r), k  g = 1, . . . , G (4.73) g =1 The definition of group constants given by Eqs. (4.39) is applicable. For the group diffusion coefficient there are two plausible definitions:  Eg−1 g dE D(r, E)φ(r, E)/φg (r) D (r) = Eg 1 = 3  Eg−1 dE Eg φ(r, E) ! φg (r) tr (r, E) (4.74) or D g (r) = 1 = g 3tr (r) 3 1 Eg−1 g dEtr (r, E)φ(r, E)/φg (r) (4.75) We return to this issue in Chapter 10, where the multigroup diffusion equations are formally derived from energy-dependent transport theory. Equations (4.73) constitute a set of homogeneous equations, the solutions of which are nontrivial only for certain discrete values of the effective multiplication constant, k. It has been shown (Refs. 8 and 12) that the mathematical properties of the multigroup diffusion equations are such that the largest such discrete eigenvalue is real and positive. The corresponding eigenfunction is unique and nonnegative everywhere within the reactor. In other words, mathematically, these equations have a physically correct solution corresponding to the largest value of the eigenvalue. Two-Group Theory The simplest example of multigroup diffusion theory is two-group theory in which the fast group contains all neutrons with E 1 eV and the thermal group contains the neutrons that have slowed down into the thermal interval E 1 eV. This model is described by    1 −∇ · D 1 ∇φ1 + a1 + s1→2 φ1 = νf1 φ1 + νf2 φ2 k (4.76) −∇ · D 2 ∇φ2 + a2 φ2 = s1→2 φ1 and the boundary conditions of the neutron fluxes in both groups vanishing on the boundary of the reactor. 4.4 Multigroup Diffusion Theory Two-Group Bare Reactor For a uniform reactor, the vanishing of the neutron flux on the boundary requires that the neutron flux in both groups satisfies ∇ 2 ψ(r) + Bg2 ψ(r) = 0 (4.77) where Bg is the geometric buckling of Table 3.3. Using this form for the group fluxes in Eqs. (4.76) leads to a pair of homogeneous algebraic equations that can be solved for the effective multiplication constant k= νf1 a1 + s1→2 + D 1 Bg2 + νf2 s1→2 · a1 + s1→2 + D 1 Bg2 a2 + D 2 Bg2 (4.78) and the flux ratio φ1 = a2 + D 2 Bg2 s1→2 φ2 (4.79) Extending the definition of the diffusion length for the fast group to include removal by scattering to the thermal group L21 = D1 D1 = a1 + s1→2 r1 (4.80) Eq. (4.78) for the effective multiplication constant can be rearranged into a form from which the definition of terms in the six-factor formula are apparent:  2      νf1 φ1 νf s1→2 1 1 k= 1 + a2 a1 + s1→2 νf2 φ2 1 + L21 Bg2 1 + L22 Bg2  1  2  PNL (4.81) = (ηf )(ε)(p) PNL 1 ) and thermal (P 2 ) non-leakage probabilities are identified where the fast (PNL NL separately. One-and-One-Half-Group Theory Because the thermal group absorption cross section is generally much larger than the fast-group cross section, D 2  D 1 . This suggests approximating the two-group equations by neglecting D 2 and using the resulting solution of the thermal group equation φ 2 = (s1→2 /a2 )φ 1 in the fast-group equation to obtain    1→2 1 −∇ · D 1 ∇φ1 + r1 φ1 = νf1 + νf2 s 2 φ1 (4.82) k a which has the form of a one-group diffusion equation for the fast neutrons. This method may be extended to account for the diffusion of thermal neutrons by using an effective value of the fast diffusion coefficient, 1 = D1 + Deff a1 + s1→2 2 D a2 (4.83) 129 130 4 Neutron Energy Distribution which has the effect of replacing the fast diffusion length L1 by the migration length; that is, L21 → M 2 = a1 D1 D2 + 2 1→2 + s a (4.84) The solutions discussed in Chapter 3 for the one-speed neutron diffusion equation can be applied immediately to 1 12 -group theory merely by replacing νf → νf1 + 1 . νf2 (s1→2 /a2 ) and D → Deff Two-Group Theory of Two-Region Reactors Consider a rectangular parallelepiped core consisting of a uniform central region (material 1) bounded on both ends by regions of the same composition (material 2), as depicted in Fig. 4.6. The two-group equations in each material (subscript k) are 1 −Dk1 ∇ 2 φ1k (x, y, z) + rk φ1k (x, y, z) = 1 νf1 k φ1k (x, y, z) + νf2 φ2k (x, y, z) k (4.85) 2 1→2 −Dk2 ∇ 2 φ2k (x, y, z) + ak φ2k (x, y, z) = sk φ1 (x, y, z) where group 2 is assumed to be below the fission spectrum. We seek a solution by separation of variables, and recalling the results of Chapter 3 look for a solution of the form φgk (x, y, z) = Xgk (x) cos πy πy cos 2Y1 2Z1 (4.86) The y- and z-components of the gradient operators acting on the trial solutions of Eq. (4.86) give rise to a transverse buckling term,  2 = Byz π 2Y1 2  + π 2Z1 2 Fig. 4.6 Three-region reactor model. (From Ref. 6; used with permission of MIT Press.) (4.87) 4.4 Multigroup Diffusion Theory These trial solutions are substituted into Eqs. (4.85) to obtain equations for the Xgk :  1  ∂2 1 2 X1k (x) + rk + Dk1 Byz X1k (x) = νf1 k X1k (x) + νf2 k X2k (x) k ∂x 2 (4.88) 2 X (x)   ∂ 2k 2 2 1→2 −Dk2 + ak + Dk2 Byz X1k (x) X2k (x) = sk ∂x 2 −Dk1 These equations must satisfy symmetry boundary conditions at x = 0, continuity of flux and current interface conditions at x = x1 , and zero flux at x = x1 + x2 : dXg1 (0) = 0, dx Xg2 (x1 + x2 ) = 0 Xg1 (x1 ) = Xg2 (x1 ), (4.89) g dXg1 (x1 ) −D1 dx g dXg2 (x1 ) = −D2 dx The procedure for solving Eqs. (4.88) is to look for solutions of a particular form with arbitrary constants and then to establish conditions on the arbitrary constants by requiring the form to satisfy Eqs. (4.88). In particular, we look for solutions that satisfy d 2 Xgk (x) + Bk2 Xgk (x) = 0 dx 2 (4.90) in each region k. Note that we require that Eq. (4.90) be satisfied with the same value of Bk2 by both the fast (X1k ) and thermal (X2k ) fluxes in each region k. Substituting the solution of the form that satisfies Eqs. (4.90) into Eqs. (4.88) leads to a set of equations for each region k:    1 1 2 1 + rk − νf1 k X1k (x) − νf2 k X2k (x) = 0 Dk1 Bk2 + Dk1 Byz k k (4.91)  2 2    1→2 2 2 2 −sk X1k (x) + Dk Bk + Dk Byz + ak X2k (x) = 0 which must be satisfied if the solution of Eqs. (4.88) within each material is to have the form that satisfies Eqs. (4.90). These are homogeneous equations, which have a nontrivial solution only if the determinant of the coefficient matrix vanishes, which defines two values Bk2 = μ2k and Bk2 = −νk2 for which Eqs. (4.88) have solutions of the form that satisfies Eqs. (4.90):  2 1 − k −1 ν 1  rk 1 ak fk + 2 Dk2 Dk1  2 1 + k −1 ν 1 2 1→2 1/2 rk k −1 νf2 k sk ak fk − + + 2Dk2 2Dk1 Dk1 Dk2 2 μ2k = −Byz −  2 1 − k −1 ν 1  rk 1 ak fk 2 νk2 = Byz + + 2 Dk2 Dk1  2 1 − k −1 ν 1 2 1→2 1/2 rk k −1 νf2 k sk ak fk − + + 2Dk2 Dk1 Dk1 Dk2 (4.92) 131 132 4 Neutron Energy Distribution The quantity −νk2 is always negative, but μ2k can be positive or negative, depending on the value of the two-group constants. Thus there are solutions of Eqs. (4.88) that satisfy Eqs. (4.90), which are of the form X2k (x) = A12k sin μk x + A22k cos μk x + A32k sinh νk x + A42k cosh νk x (4.93) X1k (x) = sk A12k sin μk x + sk A22k cos μk x + tk A32k sinh νk x + tk A42k cosh νk x where the second of Eqs. (4.91) has been used to determine the ratio of fast-tothermal group components: sk = tk = 2 ) + 2 Dk2 (μ2k + Byz ak 1→2 sk 2 ) + 2 Dk2 (−νk2 + Byz ak (4.94) 1→2 sk The symmetry conditions x = 0 require that A121 = A321 = 0, and the zero flux conditions at x = x1 + x2 require that the solution in region 2 be of the form 1 2 X22 (x) = C22 sin μ2 (x1 + x2 − x) + C22 sinh ν2 (x1 + x2 − x) (4.95) 1 2 sin μ2 (x1 + x2 − x) + t2 C22 sinh ν2 (x1 + x2 − x) X12 (x) = s2 C22 Requiring the solution in region 1 given by Eqs. (4.93) and the solution in region 2 given by Eqs. (4.95) to satisfy the continuity of flux and current interface conditions 1 , and results in a set of four homogeneous equations for the constants A221 , A421 , C22 2 C22 . The requirement for a nontrivial solution, the vanishing of the determinant of the coefficients, then, is the criticality condition   s1 cos μ1 x1 t1 cosh ν1 x1 −s2 sin μ2 x2 −t2 sinh ν2 x2 1 1 1 1 s1 D μ1 sin μ1 x1 −t1 D ν1 sinh ν1 x1 −s2 D μ2 cos μ2 x2 −t2 D cosh ν2 x2  1 1 2 2  det   cos μ1 x1 cosh ν1 x1 − sin μ2 x2 − sinh ν2 x2  D12 μ1 sin μ1 x1 −D12 ν1 sinh ν1 x1 −D22 μ2 cos μ2 x2 −D22 ν2 cosh ν2 x2 =0 (4.96) which may be solved for the effective multiplication constant, k. The four equations can be solved for three of the constants in terms of the one remaining constant, which must be determined from the total reactor power level. This procedure could be extended to multiregion reactors, but it becomes extremely cumbersome, and direct numerical solution of Eqs. (4.85) becomes preferable. 4.4 Multigroup Diffusion Theory Two-Group Theory of Reflected Reactors The results above can be specialized to the situation of a reflected reactor by setting f = 0 in region 2, in which case Eq. (4.92) reduces to 2 − μ22 = −Byz 1 r2 D21 , 2 −ν22 = −Byz − 1 a2 D22 (4.97) A solution of the type just described can be carried out in spherical and cylindrical geometry (reflected axially or radially, but not both), as well as in the block geometry. The results are summarized in Table 4.5, where Z(R = R, {R, z} or [x, y, z}) and W are spatial flux shapes in the core and U and V are spatial flux shapes in the reflector. The thermal flux in the core of a spherical reflected reactor is given by φc2 (r) = φ0c (sin μc r + a sinh νc r) r (4.98) and the thermal flux in the spherical shell reflector is given by φR2 (r) =     φ0R sinh μR R  − r + b sinh νR R  − r r (4.99) The corresponding fast fluxes are related to the thermal fluxes by the factors sk and tk given by Eqs. (4.94). These fluxes are plotted for a representative set of two-group constants in Fig. 4.7. The much larger ratio s1→2 /a2 in the reflector than in the core causes a peaking of the thermal flux in the reflector at the core–reflector interface. Physically, fast neutrons are diffusing out of the core and being slowed down into the thermal group in the reflector, where the thermal absorption is greatly reduced relative to the core. This same type of peaking of the thermal flux would occur in a water gap next to a fuel assembly within the core. Fig. 4.7 Fast and thermal fluxes in a reflected spherical reactor with properties (core: D1 = D2 = 1 cm, s1→2 = 0.009 cm−1 , a1 = 0.001, a2 = 0.05 cm−1 , νf2 = 0.057; reflector: D1 = D2 = 1 cm, s1→2 = 0.009 cm−1 , a1 = 0.001, a2 = 0.0049 cm−1 , νf2 = 0.0. (From Ref. 13; used with permission of McGraw-Hill.) 133 Spherical Geometry Table 4.5 Flux Shapes in Reflected Reactors in Two-Group Theory sin μR R μ cos μR sin μR  Z (R) = − R R2 sinh λR W (R) = R λ cosh λR sinh λR  W (R) = − R R2 sinh κ3 (R̃  − R) U (R) = R κ3 cosh κ3 (R̃  − R) sinh κ3 (R̃  − R)  − U (R) = − R R2  sinh κ4 (R̃ − R) V (R) = R2 κ cosh κ4 (R̃  − R) sinh κ4 (R̃  − R) V  (R) = − 4 − R R2 Z(R) = 134 4 Neutron Energy Distribution Cylinder: End Reflectors Cylinder: Side Reflectors Table 4.5 (Continued) Geometry Z(R) = J0 (l1 R) cos l2 z l2 ≡ V  (R) = −m5 J0 (m1 ρ) cosh m5 (d̃ − h) V (R) = J0 (m1 ρ) sinh m5 (d̃ − h) U  (R) = −m4 J0 (m1 ρ) cosh m4 (d̃ − h) W (R) = J0 (m1 ρ) cosh m3 h W  (R) = m3 J0 (m1 ρ) sinh m3 h 2.405 m22 ≡ μ2 − m21 m1 ≡ R̃  U (R) = J0 (m1 ρ) sinh m4 (d̃ − h) Z  (R) = −m2 J0 (m1 ρ) sin m2 h l42 ≡ κ32 + l22 l52 ≡ κ42 + l22 Z(R) = J0 (m1 ρ) cos m2 h m25 ≡ κ42 + m21 m24 ≡ κ32 + m21 m23 ≡ λ2 + m21 V  (R) = l5 [I1 (l5 R)K0 (l5 R̃  ) + I0 (l5 R̃  )K1 (l5 R)] cos l2 z V (R) = [I0 (l5 R)K0 (l5 R̃  ) − I0 (l5 R̃  )K0 (l5 R)] cos l2 z U  (R) = l4 [I1 (l4 R)K0 (l4 R̃  ) + I0 (l4 R̃  )K1 (l4 R)] cos l2 z l12 ≡ μ2 − l22 π l32 ≡ λ2 + l22 2h̃ U (R) = [I0 (l4 R)K0 (l4 R̃  ) − I0 (l4 R̃  )K0 (l4 R)] cos l2 z W  (R) = l3 I1 (l3 R) cos l2 z W (R) = I0 (l3 R) cos l2 z Z  (R) = −l1 J1 (l1 R) cos l2 z 4.4 Multigroup Diffusion Theory 135 Geometry Source: Adapted from Ref. 13; used with permission of McGraw-Hill. Prime on Z  , W  , U  , V  indicates spatial derivative. Tilde on symbol indicates extrapolated boundary. Block Table 4.5 (Continued) n25 ≡ κ32 + n22 + n23 n26 ≡ κ42 + n22 + n23 V  (R) = −n6 cosh n6 (d̃ − a) cos n2 y cos n3 z V (R) = sinh n6 (d̃ − a) cos n2 y cos n3 z U  (R) = −n5 cosh n5 (d̃ − a) cos n2 y cos n3 z W  (R) = n4 sinh n4 a cos n2 y cos n3 z π π n3 ≡ n2 ≡ n24 ≡ λ2 + n22 + n23 2c̃ 2b̃ U (R) = sinh n5 (d̃ − a) cos n2 y cos n3 z n21 ≡ μ2 − n22 − n23 W (R) = cosh n4 a cos n2 y cos n3 z Z(R) = cos n1 a cos n2 y cos n3 z Z  (R) = −n1 sin n1 a cos n2 y cos n3 z 136 4 Neutron Energy Distribution 4.4 Multigroup Diffusion Theory Numerical Solutions for Multigroup Diffusion Theory The numerical solution procedures discussed for the one-speed diffusion equation in Section 3.10 are readily extended to the solution of the multigroup diffusion equations. The G multigroup equations for the case of G − 1 fast groups and a thermal group G are 1 −∇ · D 1 ∇φ1 + r1 φ1 = χ1 Sf k 1 −∇ · D 2 ∇φ2 + r2 φ2 = χ2 Sf + s1→2 φ1 k −∇ · D 3 ∇φ3 + r3 φ3 (4.100) 1 = χ3 Sf + s1→3 φ1 + s2→3 φ2 k −∇ · D G ∇φG + aG φG = s1→G φ1 + s2→G φ2 + · · · + sG−1→G φG−1 where the fission source is Sf (r) = G  g (4.101) νf (r)φg (r) g=1 (0) The solution procedure is initiated by guessing a fission source distribution, Sf , and an effective multiplication constant, k (0) , and solving the group 1 equation for (1) the first iterate flux, φ1 : (1) (1) −∇ · D 1 ∇φ1 + r1 φ1 = 1 (0) χ1 Sf k (0) (4.102) Equation (4.102) is solved iteratively (e.g., by the successive relaxation method described in Section 3.10). Next, the group 2 equation is solved for the first iterate flux, φ2(1) : (1) (1) −∇ · D 2 ∇φ2 + r2 φ2 = 1 (0) (1) χ2 Sf + s1→2 φ1 k (0) (4.103) using the just calculated φ1(1) and an iteration procedure of the type described in Section 3.10. This procedure is continued successively to all the lower groups, using the just calculated values of the fluxes for higher-energy groups to calculate the scattering-in source, to determine the first iterate of all G group fluxes (1) (1) (1) [φ1 , φ2 , . . . , φG ], which are then used to compute a first iterate fission source: Sf(1) (r) = G  g νf (e)φg(1) (r) (4.104) g=1 and a first iterate effective multiplication constant: k (1) = k (0) drSf(1) (r) (1) drSf (r) (4.105) 137 138 4 Neutron Energy Distribution The iterations are continued until the effective multiplication constant converges, as described in Section 3.10. If a multigroup structure is chosen in which there is more than one group in the thermal energy interval E < 1 eV, there is upscattering among the thermal groups and the successive-group solution procedure above must be modified by solving simultaneously for the fluxes in the thermal groups or by an iterative solution for the thermal group fluxes. References 1 D. E. Cullen, “Nuclear Cross Section Preparation,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations I, CRC Press, Boca Raton, FL (1986). 2 R. E. MacFarlane, D. W. Muir, and R. M. Boicourt, The NJOY Nuclear Data Processing System, Vols. I and II, LA-9303-M, Los Alamos National Laboratory, Los Alamos, NM (1982). 3 J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976). 4 B. J. Toppel, “The New Multigroup Cross Section Code, MC2 -II,” in Proc. Conf. New Developments of Reactor Mathematics and Applications, CONF710302, Idaho Falls, ID (1971); H. Henryson et al., MC 2 -II: A Code to Calculate Fast Neutron Spectra and Multigroup Cross Sections, ANL-8144, Argonne National Laboratory, Argonne, IL (1976). 5 C. R. Weisbin et al., MINX: A Multigroup Interpretation of Nuclear Cross Sections from ENDF/B, LA-6486-MS(ENDF-237), Los Alamos National Laboratory, Los Alamos, NM (1976). 6 A. F. Henry, Nuclear-Reactor Analysis, MIT Press, Cambridge, MA (1975). 7 R. Kinsey, Data Formats and Procedures for the Evaluated Nuclear Data File, ENDF, BNL-NCS-50496, ENDG1021, 2nd ed., ENDF/B-V, Brookhaven National Laboratory, Upton, NY (1970); C. Brewster, ENDF/B Cross Sections, BNL-17100 (ENDF-200), 2nd ed., Brookhaven National Laboratory, Upton, NY (1975). 8 E. L. Wachspress, Iterative Solutions of Elliptic Systems and Applications to the Neutron Diffusion Equations of Reactor Physics, Prentice Hall, Englewood Cliffs, NJ (1973). 9 R. H. Howerton et al., Evaluation Techniques and Documentation of Specific Evaluations of the LLL Evaluated Nuclear Data Library (ENDL), UCRL50400, Vol. 15, Lawrence Livermore Laboratory, Livermore, CA (1970). 10 I. I. Bondarenko et al., Group Constants for Nuclear Reactor Calculations, Consultants Bureau, New York (1964). 11 R. S. Varga, Matrix Iterative Analysis, Prentice Hall, Englewood Cliffs, NJ (1962). 12 G. J. Habetler and M. A. Martino, Proc. Symp. Appl. Math. IX, 127 (1961). 13 R. V. Meghreblian and D. K. Holmes, Reactor Analysis, McGrawHill, New York (1960). Problems 4.1. Solve the neutron balance equation in the slowing-down range for the neutron flux, and determine the neutron Problems 4.2. 4.3. 4.4. 4.5. 4.6. 4.7. 4.8. slowing-down density, for a mixture of nonhydrogenic moderators and no absorption. Compare the result with Eq. (4.20) in the no-absorption limit. Consider a very large block of material with composition (235 U = 0.002 × 1024 at/cm3 , 238 U = 0.040 × 1024 at/cm3 , H2 O = 0.022 × 1024 at/cm3 , Fe = 0.009 × 1024 at/cm3 ) and temperature T = 400◦ C. Calculate and plot the neutron flux energy distribution in the fission, slowing-down, and thermal regions. Carry out the steps to demonstrate that the Maxwellian distribution of Eq. (4.29) satisfies the equilibrium neutron balance equation of Eq. (4.27). Calculate the thermal group absorption cross section for 235 U at T = 300, 400, and 500◦ C. n Calculate the infinite multiplication constant and the relative group fluxes in a very large fuel assembly with the four-group constants given in Table P4.5. The radiative capture cross section for a certain isotope is measured at the following energies: 50 eV, 200 barns; 100 eV, 245 barns; 150 eV, 275 barns; 300 eV, 200 barns; 350 eV, 180 barns; 400 eV, 210 barns. Calculate a multigroup capture cross section for the group Eg = 75 eV, Eg−1 = 425 eV. Calculate the resonance escape probability for the 6.67 eV 238 U resonance at T = 300◦ C when the moderator scattering cross section per uranium nucleus is sM /Nres = 50 barns. Calculate the resonance integral using either the narrow or wide resonance approximation; explain your choice. Calculate the contribution of each of the resonances in Table 4.4 to the multigroup capture cross section for a group extending from 1 to 300 eV when the moderator scattering cross section per uranium nucleus is sM /Nres = 75 barns and the temperature is 300◦ C. Table P4.5 Group Constant Group 1: 1.35–10 MeV Group 2: 9.1 keV–1.35 MeV Group 3: 0.4 eV–9.1 keV Group 4: 0.0–0.4 eV χ νf (cm−1 ) a (cm−1 ) 0.575 0.0096 0.0049 0.425 0.0012 0.0028 0 0.0177 0.0305 0 0.1851 0.1210 s D (cm) 0.0831 2.162 0.0585 1.087 0.0651 0.632 – 0.354 g→g+1 (cm−1 ) 139 140 4 Neutron Energy Distribution Table P4.11 Core Group Constant Water/Structure Group 1 Group 2 Group 1 Group 2 1.0 0.0085 0.0121 0.0241 1.267 0.0 0.1851 0.121 – 0.354 0.0 0.0 0.0004 0.0493 1.130 0.0 0.0 0.020 – 0.166 χ νf (cm−1 ) a (cm−1 ) s1→2 (cm−1 ) D (cm) 4.9. Repeat the calculation of Problem 4.8 for sM /Nres = 25 barns. Repeat the calculation for 500◦ C. 4.10. Calculate the total resonance escape probability for the resonances in Table 4.4 when the moderator scattering cross section per uranium nucleus is sM /Nres = 75 barns and the temperature is 300◦ C. 4.11. Consider a large repeating array of slab fuel assemblies of width 50 cm separated by 10 cm water–structure slabs. Calculate the thermal and fast flux distributions and the infinite multiplication factor for the fuel–water–structure array using the two-group cross sections given in Table P4.11. 4.12.∗ Write a computer code to solve numerically for the fast and thermal flux distributions and the effective multiplication constant in a two-dimensional cut through a very tall reactor core. The reactor core extends from −50 cm < x < +50 cm. Region 1 of the core extends from 15 cm < y < 55 cm, and region 2 of the core extends from 55 cm < y < 105 cm. The core is entirely surrounded by a 15-cm-thick reflector. The two-group constants for the core and reflector are given in Table P4.12. Table P4.12 Core 1 Group Constant χ νf (cm−1 ) a (cm−1 ) s1→2 (cm−1 ) D (cm) * Core 2 Reflector Group 1 Group 2 Group 1 Group 2 Group 1 1.0 0.0085 0.0121 0.0241 1.267 0.0 0.1851 0.121 – 0.354 1.0 0.006 0.010 0.016 1.280 0.0 0.150 0.100 0.000 0.400 0.0 0.0 0.0004 0.0493 1.130 Problem 4.12 is a longer problem suitable for a take-home project. Group 2 0.0 0.0 0.020 – 0.166 Problems 4.13. Calculate the reduction of the slowing-down density as a function of energy below 50 keV in a 1:1 homogeneous mixture of H2 O and 3% enriched uranium. (Use the resonance range cross sections of Table 1.3, assuming them to be constant in energy.) 4.14. Calculate and plot the hardened Maxwellian component of the thermal spectrum for in a 1:1 homogeneous mixture of H2 O and uranium for natural uranium and 4% enriched uranium. (Use the thermal range cross sections of Table 1.3, assuming them to be constant in energy, and use C = 1.5.) 4.15. Calculate the spectrum averaged one group cross sections for Problem 4.5. 4.16. Extend the development of Section 3.11 to derive the equations for a multigroup nodal model. 4.17. Calculate the node average fluxes and the effective multiplication constant of Problem 4.12 using a two-group nodal model. Compare with the results of Problem 4.12. 4.18. Calculate in two-group theory the critical radius of a 3.5-m-high bare cylindrical core with the cross sections given for core 1 in Problem 4.12. 4.19. Repeat the calculation of Problem 4.18 for the situation in which the core is surrounded by a 15-cm-thick annular reflector with the properties given in Problem 4.12. Compare the result with the result that would be obtained by subtracting the reflector savings from the critical radius for the bare core calculated in Problem 4.18. 4.20. Solve Problem 4.18 in 1 12 group theory. 4.21. Calculate the multiplication constant for a bare, cylindrical reactor of height H = 3.5 m and radius R = 1.1 m using 2-group theory and the group constants given for the core in Table P4.11 on p. 140 of the text. 4.22. You have access to an evaluated nuclear data library in which the cross-sections are tabulated pointwise in energy, except in the resonance region where single-level resonance parameters are given. You need to construct a 3-group cross section set for analyzing a PWR, but you do not have access to a many-group cross section processing code, nor do you have time to write one. Describe how you might construct a 3-group cross section set from the tabulated cross-section data and the resonance parameters. In particular, describe how you would weight the pointwise data. 4.23. Describe how you would construct a set of 2-group cross-sections for a fast reactor in which there were essentially no neutrons moderated below 1 keV. What group structure would you use? What weighting functions? Define 141 142 4 Neutron Energy Distribution the group cross sections that could be calculated from a library of cross sections given at discrete energies. 4.24. Calculate the critical radius of a bare, homogeneous cylindrical reactor of height H = 3 m. Use 2-group theory and the core cross sections (νf = 0.0085, s1→2 = 0.0241, a = 0.0121 cm−1 , D = 1.267 cm, χ = 1 for fast group 1) and (νf = 0.1851, a = 0.121 cm−1 , D = 0.354 cm, χ = 0 for thermal group 2). 143 5 Nuclear Reactor Dynamics An understanding of the time-dependent behavior of the neutron population in a nuclear reactor in response to either a planned change in the reactor conditions or to unplanned and abnormal conditions is of the utmost importance to the safe and reliable operation of nuclear reactors. We saw in Chapter 2 that the response of the prompt neutrons is very rapid indeed. However, unless the reactor is supercritical on prompt neutrons alone, the delayed emission of a small fraction of the fission neutrons can slow the increase in neutron population to the delayed neutron precursor decay time scale of seconds, providing time for corrective control measures to be taken. If a change in conditions makes a reactor supercritical on prompt neutrons alone, only intrinsic negative feedback mechanisms that automatically provide a compensating change in reactor conditions in response to an increase in the neutron population can prevent a runaway increase in neutron population (and fission power level). However, some of the intrinsic changes in reactor conditions in response to a change in power level may enhance the power excursion (positive feedback), and others may be negative but delayed sufficiently long that the compensatory reactivity feedback is out of phase with the actual condition of the neutron population in the reactor, leading to power-level instabilities. These reactor dynamics phenomena, the methods used for their analysis, and the experimental techniques for determining the basic kinetics parameters that govern them are discussed in this chapter. 5.1 Delayed Fission Neutrons Neutrons Emitted in Fission Product Decay The dynamics of a nuclear reactor or any other fission chain-reacting system under normal operation is determined primarily by the characteristics of the delayed emission of neutrons from the decay of fission products. The total yield of delayed neutrons per fission, νd , depends on the fissioning isotope and generally increases with the energy of the neutron causing fission. Although there are a relatively large number of fission products which subsequently decay via neutron emission, the 144 5 Nuclear Reactor Dynamics Table 5.1 Delayed Neutron Parameters Group Fast Neutrons Decay Constant Relative Yield λi (s−1 ) βi /β Thermal Neutrons Decay Constant Relative Yield λi (s−1 ) βi /β 0.0125 0.0360 0.138 0.318 1.22 3.15 νd = 0.00731 β = 0.0026 0.096 0.208 0.242 0.327 0.087 0.041 0.0126 0.0337 0.139 0.325 1.13 2.50 νd = 0.00667 β = 0.0026 0.086 0.299 0.252 0.278 0.051 0.034 0.0127 0.0317 0.115 0.311 1.40 3.87 νd = 0.01673 β = 0.0064 0.038 0.213 0.188 0.407 0.128 0.026 0.0124 0.0305 0.111 0.301 1.14 3.01 νd = 0.01668 β = 0.0067 0.033 0.219 0.196 0.395 0.115 0.042 0.0129 0.0311 0.134 0.331 1.26 3.21 νd = 0.0063 β = 0.0020 0.038 0.280 0.216 0.328 0.103 0.035 0.0128 0.0301 0.124 0.325 1.12 2.69 νd = 0.00645 β = 0.0022 0.035 0.298 0.211 0.326 0.086 0.044 0.0128 0.0297 0.124 0.352 1.61 3.47 νd = 0.0157 β = 0.0054 0.010 0.229 0.173 0.390 0.182 0.016 233 U 1 2 3 4 5 6 235 U 1 2 3 4 5 6 239 Pu 1 2 3 4 5 6 241 Pu 1 2 3 4 5 6 νd = 0.0152 — — — — — — — — — — — — 0.0124 0.0334 0.121 0.321 1.21 3.29 νd = 0.0531 β = 0.0203 0.034 0.150 0.155 0.446 0.172 0.043 232 Th 1 2 3 4 5 6 5.1 Delayed Fission Neutrons Table 5.1 (Continued) Group Fast Neutrons Decay Constant Relative Yield βi /β λi (s−1 ) 238 U 1 2 3 4 5 6 0.0132 0.0321 0.139 0.358 1.41 4.02 νd = 0.0460 β = 0.0164 0.013 0.137 0.162 0.388 0.225 0.075 0.0129 0.0313 0.135 0.333 1.36 4.04 νd = 0.0090 β = 0.0029 0.028 0.273 0.192 0.350 0.128 0.029 240 Pu 1 2 3 4 5 6 Thermal Neutrons Decay Constant Relative Yield λi (s−1 ) βi /β observed composite emission characteristics can be well represented by defining six effective groups of delayed neutron precursor fission products. Each group can be characterized by a decay constant, λi , and a relative yield fraction, βi /β. The fraction of the total fission neutrons that are delayed is β = νd /ν. The parameters of delayed neutrons emitted by fission product decay of several relevant isotopes are given in Table 5.1. Effective Delayed Neutron Parameters for Composite Mixtures The delayed neutrons emitted by the decay of fission products are generally less energetic (average energy about 0.5 MeV) than the prompt neutrons (average energy about 1 MeV) released directly in the fission event. Thus these delayed neutrons will slow down quicker than the prompt neutrons and experience less probability for absorption and leakage in the process (i.e., the delayed and prompt neutrons have a difference in their effectiveness in producing a subsequent fission event). Since the energy distribution of the delayed neutrons differs from group to group, the different groups of delayed neutrons will also have a different effectiveness. Furthermore, a nuclear reactor will, of course, contain a mixture of fissionable isotopes (e.g., a uranium-fueled reactor will initially contain 235 U and 238 U, and after operation for some time will also contain some admixture of 239 Pu, 240 Pu, and so on; see Chapter 6). To deal with this situation, it is necessary to define an importance function, φ + (r, E), which is the probability that a neutron introduced at position r and en- 145 146 5 Nuclear Reactor Dynamics ergy E will ultimately result in a fission (Chapter 13). Then the relative importance (to the production of a subsequent fission) of delayed neutrons in group i emitted q with energy distribution χdi (E) and prompt neutrons from the fission of isotope q q emitted with energy distribution χp (E) are  q  I di = q 0  ∞  Ip= ∞ dV dV dE χdi (E)φ + (r, E) q dE χp (E)φ + (r, E) q 0   0 ∞ 0 ∞ dE  νσf (E  )Nq (r)φ(r, E  ) q dE  νσf (E  )Nq (r)φ(r, E  ) q (5.1) (5.2) The relative effective delayed neutron yield of group i delayed neutrons for fisq q q sionable isotope q is I di βi , where βi is the group i delayed neutron yield of fissionable isotope q given in Table 5.1. The effective group i delayed neutron fraction for isotope q in a mixture of fissionable isotopes is then q γi βi =I q q di βi  % &  6 6 !   q q q q I p 1− βi + I di βi q i=1 (5.3) i=1 The effectiveness of delayed neutron group i of fissionable isotope q in a specific q q q q admixture of fissionable isotopes and reactor geometry is then γi = γi βi /βi . In the remainder of the book, except when specifically stated otherwise, it is assumed that the delayed neutron effectiveness is included in the evaluation of βi and β, and the effectiveness parameter will be suppressed. Photoneutrons Fission products also emit gamma rays when they undergo β-decay. A photon can eject a neutron from a nucleus when its energy exceeds the neutron binding energy. Although most nuclei have neutron binding energies in excess of 6 MeV, which is above the energy of most gamma rays from fission, there are four nuclei that have sufficiently low neutron binding energy, En , to be of practical interest: 2 D (En = 2.2 MeV), 9 Be (En = 1.7 MeV), 6 Li (En = 5.4 MeV), and 13 C (En = 4.9 MeV). The photoneutrons can be considered as additional groups of delayed neutrons. Since the β-decay of fission products is generally much slower than the direct neutron decay, the photoneutron precursor decay constants are much smaller than the delayed neutron precursor decay constants shown in Table 5.1. The only reactors in which photoneutrons are of practical importance are D2 O-moderated reactors. As we shall see, the dynamic response time of a reactor under normal operation is largely determined by the inverse decay constants, and consequently, D2 O reactors are quite sluggish compared to other reactor types. 5.2 Point Kinetics Equations 5.2 Point Kinetics Equations The delayed neutron precursors satisfy an equation of the form ∂ Ĉi (r, t) = βi νf (r, t)φ(r, t) − λi Ĉi (r, t), ∂t i = 1, . . . , 6 (5.4) The one-speed neutron diffusion equation is now written 1 ∂φ(r, t) − D(r, t)∇ 2 φ(r, t) + a (r, t)φ(r, t) v ∂t = (1 − β)νF (r, t)φ(r, t) + 6  λi Ĉi (r, t) (5.5) i=1 where we have taken into account that a fraction β of the fission neutrons is delayed and that there is a source of neutrons due to the decay of the delayed neutron precursors. Based on the results of Chapter 3, we assume a separation-of-variables solution φ(r, t) = vn(t)ψ1 (r), Ĉi (r, t) = Ci (t)ψ1 (r) (5.6) where ψ1 is the fundamental mode solution of ∇ 2 ψn + Bg2 ψn = 0 (5.7) and Bg is the geometric buckling appropriate for the reactor geometry, as discussed in Chapter 3. Using this in Eqs. (5.4) and (5.5) leads to the point kinetics equations  dn(t) ρ(t) − β = n(t) + λi Ci (t) dt  6 i=1 dCi (t) βi = n(t) − λi Ci (t), dt  (5.8) i = 1, . . . , 6 where  ≡ (vνF )−1 (5.9) is the mean generation time between the birth of a fission neutron and the subsequent absorption leading to another fission, and ρ(t) ≡ νF − a (1 + L2 Bg2 ) νF ≡ k(t) − 1 k(t) (5.10) is the reactivity. The quantity k is the effective multiplication constant, given by k≡ νF /a 1 + L2 Bg2 (5.11) 147 148 5 Nuclear Reactor Dynamics Fig. 5.1 Plot of the function R(ω) = ω[ + βi /(ω + λi )], which appears in the in-hour equation. (From Ref. 4; used with permission of MIT Press.) For predominantly thermal reactors, νf and a are thermal cross sections, and L2 should be replaced by M 2 = L2 + τth to include the fast diffusion while the neutron is slowing down, τth , as well as the thermal diffusion length, L2 . For fast reactors, all cross sections are averaged over the appropriate fast spectrum. The limiting assumption for the validity of the point kinetics equations is the assumption of a constant spatial shape. As we will see, this assumption is reasonable for transients caused by uniform changes in reactor properties or for reactors with dimensions that are only a few migration lengths, M (or diffusion lengths, L), but is poor for reactors with dimensions that are very large compared to M in which the transient is caused by localized changes in reactor properties (e.g., a nonsymmetric control rod withdrawal). However, as we will see in Chapter 16, such spatial shape changes can be taken into account in computation of the reactivity and the mean generation time, and the point kinetics equations can be extended to have a much wider range of validity. 5.3 Period–Reactivity Relations Equations (5.8) may be solved for the case of an initially critical reactor in which the properties are changed at t = 0 in such a way as to introduce a reactivity ρ0 which is then constant over time, by Laplace transforming, or equivalently assuming an exponential time dependence e−st . The equations for the time-dependent parts of n and Ci are  ρ0 − β n(s) + λi Ci (s) + n0  6 sn(s) = i=1 βi sCi (s) = n(s) − λi Ci (s) + Ci0 ,  (5.12) i = 1, . . . , 6 5.3 Period–Reactivity Relations which can be reduced to n(s) = f (s, n0 , Ci0 ) Y (s) (5.13) where % Y (s) ≡ ρ0 − s  + 6  i=1 & βi s + λi (5.14) The poles of the right side—the roots of Y (s) = 0—determine the time dependence of the neutron and precursor populations. Y (s) = 0 is a seventh-order equation, known as the inverse hour, or more succinctly, the inhour, equation, the solutions of which are best visualized graphically, as indicated in Fig. 5.1, where the right-hand side of % ρ0 = s  + 6  βi s + λi i=1 & (5.15) is plotted. The left-hand side, ρ0 , would plot as a straight horizontal line, of course, and the points at which it crosses the right-hand side are the solutions (roots of the equation). For ρ0 < 0, indicated by the circles in Fig. 5.1, all the solutions sj < 0. For ρ0 > 0, indicated by the crosses, there are one positive and six negative solutions. The solution for the time-dependent neutron flux is of the form n(t) = 6  Aj esj t (5.16) j =0 where the sj are the roots of Y (s) = 0 and the Aj are given by % Aj =  + 6  i=1 βi sj + λi &' 1+k 6  i=1 βi λi (sj + λi )2  (5.17) After a sufficient time, the solution will be dominated by the largest root s0 (s0 > 0 when ρ0 > 0, s0 is the least negative root when ρ0 < 0): n(t)  A0 es0 t ≡ A0 et/T (5.18) where T ≡ s0−1 is referred to as the asymptotic period. Measurement of the asymptotic period then provides a means for the experimental determination of the reactivity   6  βi 1 + ρ0 = T (1/T ) + λi i=1 (5.19) 149 150 5 Nuclear Reactor Dynamics 5.4 Approximate Solutions of the Point Neutron Kinetics Equations One Delayed Neutron Group Approximation To simplify the problem so that we can gain insight into the nature of the solution of the point kinetics equations, we assume that the six groups of delayed neutrons can be replaced by one delayed neutron group with an effective yield fraction β = i γi βi and an effective decay constant λ = i βi λi /β, so that the point kinetics equations become dn ρ − β = n + λC, dt  dC β = n − λC dt  (5.20) Proceeding as in Section 5.3 by Laplace transforming or assuming an est form of the solution, the equivalent of Eq. (5.13) for the determination of the roots of the reduced in-hour equation is   ρ −β λρ s2 − −λ s − =0 (5.21)   which has the solution s1,2   1 ρ −β = −λ ± 2    1 ρ −β −λ ± = 2  ( (  2 1 ρ −β βλ +λ + 4    2 1 ρ −β λρ −λ + 4   (5.22) For ρ > 0, one root is positive and the other negative; for ρ = 0, one root is zero and the other is negative; and for ρ < 0, both roots are negative. The assumed est time dependence, when used in Eqs. (5.20), requires that for each of the two roots, s1 and s2 , there is a fixed relation between the precursor and the neutron populations: !  C(t) β ρ −β = =− − s1,2 λ (5.23) n(t) (s1,2 + λ)  which means that the solution of Eqs. (5.20) is of the form n(t) = A1 es1 t + A2 es2 t , C(t) = A1 β β es1 t + A2 es2 t (s1 + λ) (s2 + λ) (5.24) Now, let us take some parameters typical of a light water reactor: β = 0.0075, λ = 0.08 s−1 ,  = 6 × 10−5 s. Except for |ρ − β| ≈ 0, one root of Eq. (5.21) will be of very large magnitude, and the other will be of very small magnitude. For the larger root, s12  λρ/ and λρ/ can be neglected in Eq. (5.21); and for the 5.4 Approximate Solutions of the Point Neutron Kinetics Equations smaller root, s22  λρ/ and s22 can be neglected. Assuming that |ρ − β|/  λ, the solutions of Eq. (5.21) are s1 = ρ −β ,  s2 = − λρ ρ −β (5.25) The constants A1 and A2 can be evaluated by requiring that the solution satisfy the initial condition at t = 0 that is determined by setting ρ = 0 in Eqs. (5.24), which identifies A1 ≈ n0 ρ/(ρ − β) and A2 ≈ −n0 β/(ρ − β), where n0 is the initial neutron population before the reactivity insertion, so that the solutions of Eqs. (5.24) become      ρ ρ −β β λρ exp t − exp − t n(t) = n0 ρ −β  ρ −β ρ −β (5.26)      ρβ ρ −β β λρ exp t + exp − t c(t) = n0  λ ρ −β (ρ − β)2 At t = 0, before the reactivity insertion, C0 = βn0 /λ ≈ 1600n0 . Thus the population of delayed neutron precursors, hence the latent source of neutrons, is about 1600 times greater than the neutron population in a critical reactor. It is not surprising that this large latent neutron source controls the dynamics of the neutron population under normal conditions, as we shall now see. Example 5.1: Step Negative Reactivity Insertion, ρ < 0. Equations (5.26) enable us to investigate the neutron kinetics of a nuclear reactor. We first consider the case of a large negative reactivity insertion ρ = −0.05 into a critical reactor at t = 0, such as might be produced by scramming (rapid insertion) of a control rod bank. With the representative light water reactor parameters (β = 0.0075, λ = 0.08 s−1 ,  = 6 × 10−5 s), Eqs. (5.26) become n(t) = n0 (0.87e−958t + 0.13e−0.068t ) C(t) = n0 (0.0113e−958t + 1563e−0.068t ) (5.27) which is plotted in Fig. 5.2, with T ≡ n. The first term goes promptly to zero on a time scale t ≈ , corresponding physically to readjustment of the prompt neutron population to the subcritical condition of the reactor on the neutron generation time scale. The second term decays slowly, corresponding to the slow decay of the delayed neutron precursor source of neutrons. The neutron population drops promptly from n0 to n0 /(1 − ρ/β)—the prompt jump—then slowly decays as e−[λ/(1−β/ρ)]t . Thus, scramming a control rod bank cannot immediately shut down (reduce the neutron population or the fission rate to near zero) a nuclear reactor or other fission chain reacting medium. The delayed neutron precursors decay as e−[λ/(1−β/ρ)]t . Example 5.2: Subprompt-Critical (Delayed Critical) Step Positive Reactivity Insertion, 0 < ρ < β. Next, consider a positive reactivity insertion ρ = 0.0015 < β, such as 151 152 5 Nuclear Reactor Dynamics Fig. 5.2 Neutron and delayed neutron precursor decay following negative reactivity insertion ρ = −0.05 into a critical nuclear reactor. (From Ref. 4; used with permission of MIT Press.) might occur as a result of control rod withdrawal. Equations (5.26) now become n(t) = n0 (−0.25e−100t + 1.25e0.02t ) C(t) = n0 (0.3125e−100t + 1562.5e0.02t ) (5.28) which is plotted in Fig. 5.3. The neutron population increases promptly, on the neutron generation time scale—the prompt jump—from n0 to n0 /(1 − ρ/β), as the prompt neutron population adjusts to the supercritical condition of the reactor, then increases as e−[λ/(1−β/ρ)]t , governed by the rate of increase in the delayed neutron source. The relatively slow rate of increase of the neutron population, following the prompt jump, allows time for corrective control action to be taken before the fission rate becomes excessive. Example 5.3: Superprompt-Critical Step Positive Reactivity Insertion, ρ > β. Now consider a step increase of reactivity ρ = 0.0115 > β, such as might occur as the result Fig. 5.3 Neutron and delayed neutron precursor increase following subprompt-critical positive reactivity insertion ρ = 0.0015 < β into a critical nuclear reactor. (From Ref. 4; used with permission of MIT Press.) 5.4 Approximate Solutions of the Point Neutron Kinetics Equations of the ejection of a bank of control rods from a reactor. Equations (5.26) now become n(t) = n0 (2.9e66.7t − 1.9e−0.23t ) C(t) = n0 (5.4e66.7t + 1563e−0.23t ) (5.29) The neutron population in the reactor grows exponentially on the neutron generation time scale, n ∼ e[(ρ−β)/]t , because the reactor is supercritical on prompt neutrons alone [i.e., k(1 − β) > 1]. In this example, the neutron population would increase by almost a factor of 800 in a tenth of a second, and it would be impossible to take corrective action quickly enough to prevent excessive fission heating and destruction of the reactor. Fortunately, there are inherent feedback mechanisms that introduce negative reactivity instantaneously in response to an increase in the fission heating (e.g., the Doppler effect discussed in Sections 5.7 and 5.8), and the neutron population will first increase rapidly, then decrease. However, conditions that would lead to superprompt-critical reactivity insertion are to be avoided for reasons of safety. Since β = 0.0026 for 233 U, 0.0067 for 235 U, and 0.0022 239 Pu, the safe operating range for positive reactivity insertions, 0 < ρ < β, is much larger for reactors fueled with 235 U than for reactors fueled with 233 U or 239 Pu. Prompt-Jump Approximation We found that with a reactivity insertion for which the reactor condition is less than prompt critical (ρ < β) the neutron population changed sharply on the neutron generation time scale, then changed slowly on the delayed neutron inverse decay constant time scale. If we are not interested in the details of the prompt neutron kinetics during the prompt jump, we can simplify the equations by assuming that the prompt jump takes place instantaneously in response to any reactivity change, and afterward, the neutron population changes instantaneously in response to changes in the delayed neutron source (i.e., we set the time derivative to zero in the neutron equation). 0 = [ρ(t) − β]n(t) +  6  λi Ci (t) (5.30) i=1 Since the delayed neutron precursor population does not respond instantaneously to a change in reactivity, Eq. (5.30) is valid with the same delayed precursor population both before and just after a change in reactivity from ρ0 to ρ1 < β, from which we conclude that the ratio of the neutron populations just after and before the reactivity change is n1 β − ρ0 = n0 β − ρ1 (5.31) 153 154 5 Nuclear Reactor Dynamics Use of Eq. (5.30) to eliminate n(t) in the second of Eqs. (5.8) yields a coupled set of equations for the time dependence of the precursor density: 6  βi dCi λj Cj (t) − λi Ci (t) = dt ρ(t) − βi (5.32) j =1 which in the one delayed precursor group approximation takes on the simple form −λC(t) dC(t) = dt 1 − β/ρ(t) (5.33) The prompt-jump approximation is convenient for numerical solutions because it eliminates the fast time scale due to , which introduces difficulties in time differencing methods. Numerical solutions of the point kinetics equations with and without the prompt-jump approximation for a variety of reactivity insertions indicate that the prompt-jump approximation is accurate to within about 1% for reactivities ρ < 0.5β. Using the one delayed precursor group approximation, the equivalent of Eq. (5.30) can be solved for C(t) and used in the second of Eqs. (5.20) to obtain   dρ(t) dn(t) + + λρ(t) n(t) = 0 (5.34) [ρ(t) − β] dt dt which for a given reactivity variation ρ(t) can be solved for the neutron population   t ρ̇(t  ) + λρ(t  )  n(t) = n0 exp (5.35) dt β − ρ(t  ) 0 Example 5.4: Reactivity Worth of Rod Insertion. The neutron flux measured by a detector is observed to drop instantaneously from n0 to 0.5 n0 when a control rod is dropped into a cold highly enriched critical nuclear reactor, in which ρ0 = 0. Using the one-delayed group model with β = 0.0065, Eq. (5.31) yields ρ1 = β(1 − n0 /n1 ) = 0.0065(1 − 2) = −0.0065k/k. Reactor Shutdown We mentioned that the large step negative reactivity insertion considered previously might be representative of the situation encountered in a reactor shutdown, or scram. However, the time required to fully insert control rods is very long compared to the prompt neutron generation time that governs the time scale of the prompt jump. We can improve on the representation of the control rod insertion by considering a ramp reactivity insertion ρ(t) = −εt . If we are only interested in calculating the initial rapid decrease in the neutron population, we can make the assumption that the initial precursor concentration remains constant; hence the precursor source of delayed neutrons remains constant at its pre-insertion value 6  i=1 λi Ci (0) = β n0  (5.36) 5.5 Delayed Neutron Kernel and Zero-Power Transfer Function Using this approximation, the equation governing the prompt neutron response to the reactivity insertion—the first of Eqs. (5.8)—can be integrated to obtain     1 1 2 εt + βt n(t) = n0 exp −  2      t 1 ε 2 β (5.37) (t − (t  )2 ) + β(t − t  ) dt  exp − +  0  2 This provides a somewhat better description of the initial reduction in the neutron population than do Eqs. (5.27), which, however, would still govern the long-time decay after completion of the rod insertion. 5.5 Delayed Neutron Kernel and Zero-Power Transfer Function Delayed Neutron Kernel The delayed neutron precursor equations, the second of Eqs. (5.8), can be formally integrated to obtain (assuming that Ci = 0 at −∞)  t  ∞ βi βi −λi τ  −λi (t−t  )  n(t )e e dt = n(t − τ ) dτ. (5.38) Ci (t) =   −∞ 0 Using this result in the neutron kinetics equation, the first of Eqs. (5.8), yields    ∞ dn(t) β ρ(t) − β = n(t) + D(τ )n(t − τ ) dτ (5.39) dt   0 where we have defined the delayed neutron kernel D(τ ) ≡ 6  λi βi i=1 β e−λi τ (5.40) Zero-Power Transfer Function If the neutron population is expanded about the initial neutron population in the critical reactor at t = 0, n(t) = n0 + n1 (t) (5.41) Eq. (5.39) may be rewritten dn1 (t) ρ(t)n0 ρ(t)n1 (t) = + + dt    0 ∞ β D(τ )[n1 (t − τ ) − n1 (t)] dτ  The Laplace transform of a function of time A(t) is defined as  ∞ A(t)e−st dt A(s) = 0 (5.42) (5.43) 155 156 5 Nuclear Reactor Dynamics Laplace transforming Eq. (5.42) and using the convolution theorem  ∞  L A(t)B(τ − t) dt = A(s)B(s) (5.44) 0 yields, upon assuming that the term ρ(t)n1 (t) is a product of two small terms and can be neglected relative to ρ(t)n0 , n1 (s) = n0 Z(s)ρ(s) where (5.45) &−1 % 6  1 βi Z(s) ≡ + s s + λi (5.46) i=1 is the zero-power transfer function, which defines the response of the density n1 to the reactivity. The inverse Laplace transformation of Eq. (5.45) and the convolution theorem yield the solution for the time dependence of the neutron population as a function of the time dependence of the reactivity:  t dτ Z(t − τ )ρ(τ ) (5.47) n1 (t) = n0 0 where the inverse Laplace transform of the zero-power transfer function is 1  esj (t−τ ) + 6 2  i=1 [βi λi /(sj + λi ) ]} j =2 sj { + 7 Z(t − τ ) = (5.48) and the sj are the roots of the inhour equation, Y (s) = 0, with Y (s) given by Eq. (5.14). 5.6 Experimental Determination of Neutron Kinetics Parameters Asymptotic Period Measurement When a critical reactor is perturbed by a step change in properties, the asymptotic period may be determined from the response R(t) of neutron detectors by T −1 = d(ln R)/dt ; then the period–reactivity relation of Eq. (5.19) can be used to infer the reactivity. For negative reactivities, the asymptotic period, the largest root of the inhour equation, is dominated by the largest delayed neutron period and is relatively insensitive to the value of the reactivity, so this method is limited practically to supercritical reactivity (0 < ρ) measurements, for which Eq. (5.19) may be written  βi /β  βi /β  ρ = +  β βT 1 + λi T 1 + λi T 6 6 i=1 i=1 (5.49) 5.6 Experimental Determination of Neutron Kinetics Parameters where the fact that safety considerations further limit the practical applicability of this method to the delayed critical regime (0 < ρ < β) has been taken into account in writing the second form of the equation. Rod Drop Method The responses of a neutron detector immediately before (R0 ∼ n0 ) and after (R1 ∼ n1 ) a control rod is dropped into a critical reactor (ρ0 = 0) are related by Eq. (5.31), which allows determination of the reactivity worth of the rod ρ1 R0 =1− β R1 (5.50) Source Jerk Method Consider a subcritical system that is maintained at equilibrium neutron, n0 , and precursor, Ci0 , populations by an extraneous neutron source rate, S. The neutron balance equation is   6  ρ −β λi Ci0 + S = 0 n0 +  (5.51) i=1 If the source is jerked, the prompt-jump approximation for the neutron density immediately after the source jerk is   6  ρ −β n1 + λi Ci0 = 0  (5.52) i=1 because the delayed neutron precursor population will not change immediately. These equations and the equilibrium precursor concentrations Ci0 = βi n0 /λi  may be used to relate the responses of a neutron detector immediately before (R0 ∼ n0 ) and after (R1 ∼ n1 ) the source jerk to the reactivity of the system: ρ R0 =1− β R1 (5.53) Pulsed Neutron Methods The time dependence of the prompt neutron population in a subcritical fission chain reacting medium following the introduction of a burst of neutrons is described by 1 ∂φ(r, t) = D∇ 2 φ(r, t) − [a − (1 − β)νf ]φ(r, t) v ∂t (5.54) since the delayed neutrons will not contribute until later. As discussed in Section 3.6, the asymptotic solution that remains after higher-order spatial transients decay is the fundamental mode, which decays exponentially: n(r, t)  A1 ψ1 (r)e−v[a −(1−β)νf +DBg ]t 2 (5.55) 157 158 5 Nuclear Reactor Dynamics where Bg is the fundamental mode geometric buckling for the geometry of the system. If the neutron detector response, R(r, t) ∼ n(r, t), is measured as a function of time, then α0 = 1 dR ρ −β = v νf (1 − β) − a − DBg2 = R dt  (5.56) Thus the pulsed neutron method can be used to determine ∼ ρ/, assuming that β/ is known. If the experiment is performed in a critical system (ρ = 0), the measurement yields a value for β/. In practice, a correction must be made to account for transport- and energy-dependent effects which have been neglected in this analysis, so that α0 = v νf (1 − β) − a − DBg2 − CBg4 + · · · (5.57) Rod Oscillator Measurements The response of the neutron population, as measured by a neutron detector R(t) ∼ n(t), to a sinusoidal oscillation of a control rod that produces a sinusoidal reactivity perturbation ρ(t) = ρ0 sin ωt (5.58) can be used to determine a number of neutron kinetics parameters. The response of the neutron population to a sinusoidal reactivity perturbation can be calculated from Eq. (5.45) by first computing the Laplace transform of Eq. (5.58): ρ(s) = ρ0 ω ρ0 ω = 2 (s + iω)(s − iω) +ω (5.59) s2 and then Laplace inverting Eq. (5.45), or equivalently, by using Eq. (5.58) in Eq. (5.47), to obtain n1 (t) = n0 ρ0 [|Z(iω)| sin(ωt + φ)] + ω 6  j =0 esj t (ω2 + sj2 )(dY/ds)sj (5.60) where φ is the phase angle, defined by tan φ ≡ Im{Z(iω)} Re{Z(iω)} (5.61) The first term in Eq. (5.60) arises from the poles of the reactivity [Eq. (5. 59)] at s = ±iω, and the remaining terms arise from the poles of the zero-power transfer function Z(s) [i.e., the roots of the inhour equation Y (s) = 0 given by Eq. (5.13)]. For a critical system, the largest root of the inhour equation is s = 0, so that after sufficient time the solution given by Eq. (5.60) approaches   1 (5.62) n1 (t)  n0 ρ0 |Z(iω)| sin(ωt + φ) + ω 5.6 Experimental Determination of Neutron Kinetics Parameters The average neutron detector response will be (ρ0 /ω)R0 , where R0 is the average detector response before the oscillation began. At high oscillation frequency, the contribution of the first term in Eq. (5.62) to the detector response will average to zero and the detector response will reflect the second term. In both cases, this provides a means for the experimental determination of ρ0 / in terms of the average detector response R : ρ0 R − R0 =ω  R0 (5.63) Zero-Power Transfer Function Measurements By varying the frequency of rod oscillation, ω, the zero-power transfer function, Z(iω), can be measured for a reactor or other critical fission chain reacting system by interpreting the detector reading R(t) as   1 (5.64) R(t) − R0 = R0 ρ0 |Z(iω)| sin(ωt + φ) + ω Such measurements, when compared with calculation of the transfer function, provide an indirect means of determining or confirming the parameters , βi , and λi . At low frequencies the amplitude of the transfer function approaches |Z(iω)| → ω 6  βi /β i=1 (5.65) λ2i for ω  λi , and the phase angle φ approaches tan φ → − 6  βi /β i=1 λi ' ω 6  βi /β i=1 λ2i (5.66) Rossi-α Measurement The prompt neutron decay constant α≡ 1 dn k(1 − β) − 1 = n dt l (5.67) can be measured by observing the decay of individual fission reaction chains in succession if the process is continued long enough to observe a statistically significant number of decay chains. Assume that a neutron count from a decay chain is observed at t = 0. The probability of another neutron count being observed at a later time t is the sum of the probability of a count from a chain-related neutron, Q exp(αt)t, plus the probability of a neutron from another chain, Ct , where C is the average counting rate: P (t) dt = C dt + Qeαt dt (5.68) 159 160 5 Nuclear Reactor Dynamics We use a statistical argument to determine Q. The probability of a count occurring at t0 is F dt0 , where F is just the average fission rate in the system. The probability of another detector count at t1 > t0 that is chain related to the count at t0 is P (t1 )dt1 = ενp vf eα(t1 −t0 ) dt1 (5.69) where νp is the number of prompt neutrons per fission and ε is the detector efficiency. The probability of a second chain-related count at t2 > t1 is P (t2 )dt2 = ε(νp − 1)vf eα(t2 −t0 ) dt2 (5.70) where (νp − 1) takes account of the chain-related fission required to produce the count at t1 . The three probabilities F dt0 , P (t1 )dt1 and P (t2 )dt2 are treated as independent probabilities. Hence the probability for a count in dt1 followed by a count in dt2 , both in the chain that produced the count in dt0 , is obtained by multiplying the three probabilities and integrating over −∞ < t < t1 :  P (t1 , t2 )dt1 dt2 = t1 −∞   F ε2 νp2 − ν̄p (vf )2 eα(t1 +t2 −2t0 ) dt0 dt1 dt2   (vf )2 α(t −t ) = F ε2 νp2 − ν̄p e 2 1 dt1 dt2 −2α (5.71) where an overbar indicates an average over the prompt neutron emission distribution function. Noting that νp = kp a /f = kp /(vlf ) and including the probability F 2 ε2 dt1 dt2 of a random pair of counts, this becomes P (t1 , t2 )dt1 dt2 = F 2 ε2 dt1 dt2 + F ε2 (νp2 − ν̄p )kp2 eα(t2 −t1 ) dt1 dt2 2νp2 (1 − kp )l (5.72) Since the overall probability of a count in dt1 is F εdt1 , we need to normalize this conditional probability by division by F εdt1 , which yields, upon rescaling time from t1 = 0, P (t1 , t2 )dt1 dt2 = ε(νp2 − ν̄p ) kp2 νp2 2(1 − kp )l eαt dt (5.73) This is the Q exp(αt)dt term in Eq. (5.68), so Q= ε(νp2 − ν̄p ) kp2 νp2 2(1 − kp )l (5.74) In a Rossi-α experiment, the function P (t) of Eq. (5.68) is measured by a time analyzer and the random count rate C dt is subtracted. The parameter α is then determined from the remaining Q exp(αt)dt term. 5.7 Reactivity Feedback 5.7 Reactivity Feedback Up to this point, we have discussed neutron kinetics—the response of the neutron population in a nuclear reactor or other fission chain reacting system to an external reactivity input—under the implicit assumption that the level of the neutron population does not affect the properties of the system that determine the neutron kinetics, most notably the reactivity. This is the situation when the neutron population is sufficiently small that the fission heat does not affect the temperature of the system (i.e., at zero power). However, in an operating nuclear reactor the neutron population is large enough that any change in fission heating resulting from a change in neutron population will produce changes in temperature, which in turn will produce changes in reactivity, or reactivity feedback. The combined and coupled response of the neutron population and of the temperatures, densities, and displacements of the various materials in a nuclear reactor is properly the subject of reactor dynamics, but the term is commonly used to also include neutron kinetics. When the neutron population increases, the fission heating increases. Since this heating is deposited in the fuel element, the fuel temperature will increase immediately. An increase in fuel temperature will broaden the effective resonance absorption (and fission) cross section, generally resulting in an increase in neutron absorption and a corresponding reduction in reactivity—the Doppler effect. The fuel element will also expand and, depending on the constraints, bend or bow slightly, thus changing the local fuel–moderator geometry and flux disadvantage factor (the ratio of the flux in the fuel to the flux in the moderator), thereby producing a change in reactivity. If the increase in fission heating is large enough to raise the fuel temperature above the melting point, fuel slumping will occur, resulting in a large change in the local fuel–moderator geometry and a corresponding change in flux disadvantage factor and fuel absorption, producing a further change in reactivity. Some of the increased fission heat will be transported out of the fuel element (time constant of tenths of seconds to seconds) into the surrounding moderator/coolant and structure, causing a delayed increase in moderator/coolant and structure temperature. An increase in moderator/coolant temperature will produce a decrease in moderator/coolant density, which causes a change in the local fuel–moderator properties and a corresponding change in both the moderator absorption and the flux disadvantage factor. In addition, a decrease in moderator density will reduce the moderating effectiveness and produce a hardening (shift to higher energies) in the neutron energy distribution, which will change the effective energy-averaged absorption cross sections for the fuel, control elements, and so on. An increase in structure temperature will cause expansion and deformation, producing a change in the local geometry that will further affect the flux disadvantage factor. These various moderator/coolant changes all produce changes in reactivity. The reduction in moderator/coolant density increases the diffusion of neutrons, and the increase in temperature causes an expansion of the reactor. The effect of 161 162 5 Nuclear Reactor Dynamics increased diffusion is to increase the leakage, and the effect of increased size is to reduce the leakage, producing offsetting negative and positive reactivity effects. In addition to these internal (to the core) reactivity feedback effects, there are external feedback effects caused by changes in the coolant outlet temperature that will produce changes in the coolant inlet temperature. Temperature Coefficients of Reactivity The temperature coefficient of reactivity is defined as   ∂ k−1 1 ∂k 1 ∂k ∂ρ = = 2  αT ≡ ∂T ∂T k k ∂T k ∂T (5.75) To gain physical insight into the various physical phenomena that contribute to the reactivity feedback, we first use the one-speed diffusion theory expression for the effective multiplication constant for a bare reactor, but extend it to account for fast fission by including the ratio ε = total fission/thermal fission, to account for the resonance absorption of neutrons during the slowing down to thermal energies by including the resonance escape probability p, and to account for the leakage of fast as well as thermal neutrons by replacing the diffusion length with the migration length M: k = k∞ PNL = νf νf  F 1 1 = F a 2 2 a 1 + L B a a (1 + L2 B 2 ) = ηf PNL → ηf εpPNL (5.76) which allows us to write αT = 1 ∂η 1 ∂ε 1 ∂f 1 ∂p 1 ∂PNL + + + + η ∂T ε ∂T f ∂T p ∂T PNL ∂T (5.77) This formalism lends itself to physical interpretation and can provide quantitative estimates of reactivity coefficients for thermal reactors, but it is not directly applicable to fast reactors. We discuss fast reactor reactivity coefficients in the next section, where a perturbation theory formalism that is more appropriate for the quantitative evaluation of reactivity coefficients in both fast and thermal reactors is introduced. We now discuss reactivity feedback effects on p, f , and PNL ; there are also smaller reactivity effects associated with η due to shifts in the thermal neutron energy distribution and associated with ε, which latter are similar to the effects associated with the thermal utilization factor. Doppler Effect The resonance capture cross section (one-level Breit–Wigner) is $ σγ = σ0 E0 E γ ψ(x, ξ ) (5.78) 5.7 Reactivity Feedback where ψ is the Doppler broadening shape function, which takes into account the averaging of the neutron–nucleus interaction cross section over the thermal motion of the nucleus,  ∞ dy ξ 2 2 e−[(x−y) ξ /4] (5.79) ψ(x, ξ ) = √ 1 + y2 4π −∞ σ0 is the peak resonance cross section, γ and are the capture and total widths of the resonance, x = (E − E0 )/ , ξ = /(4E0 kT /A)1/2 , E and E0 are the energies of the neutron and of the resonance peak, and A is the mass of the nucleus in amu. The total capture in the resonance is given by the resonance integral  Iγ ≡ σγ (E)φ(E)dE (5.80) The function ψ broadens with increasing temperature, T , characterizing the motion of the nucleus. A broadening of the ψ function reduces the energy selfshielding in the resonance and increases the resonance integral. Thus an increase in fuel temperature due to an increase in fission heating will cause an increase in the effective capture cross section σγ ∼ Iγ . A similar result is found for the fission resonances. In thermal reactors, the Doppler effect is due primarily to epithermal capture resonances in the nonfissionable fuel isotopes (232 Th, 238 U, 240 Pu) and can be estimated by considering the change in resonance escape probability p = e−(NF Iγ /ξ p ) (5.81) where ξ p /NF is the average moderating power per fuel atom, with a sum over resonance integrals for all fuel resonances implied, the function  ∞ ψ(x, ξ ) dx (5.82) J (ξ, β  ) ≡ ψ(x, ξ ) + β 0 is tabulated in Table 4.3, and β  = (p /NF )( /σ0 γ ). The Doppler temperature coefficient of reactivity for a thermal reactor can then be calculated as   ∂ρ 1 ∂k 1 ∂p 1 ∂I D (5.83) αTF =  = = ln p ∂TF k ∂TF p ∂TF I ∂TF Since the additional fission heating is deposited in the fuel, the fuel temperature, TF , increases immediately, and the Doppler effect immediately reduces the reactivity. The Doppler effect is a very strong contributor to the safety and operational stability of thermal reactors. There are useful fits to the total resonance integrals for 238 UO2 and 232 ThO2 :   SF I (300 K) = 11.6 + 22.8 MF √  ) I (TF ) = I (300 K) 1 + β  T (K) − 300 163 164 5 Nuclear Reactor Dynamics   238  −4 −4 SF UO2 : β = 61 × 10 + 47 × 10 MF   SF 232 ThO2 : β  = 97 × 10−4 + 120 × 10−4 MF (5.84) where SF and MF are surface area and mass of the fuel element. Using this fit, Eq. (5.83) becomes   1 β  D (5.85) αTF = − ln √ p(300 K) 2 TF (K) Fuel and Moderator Expansion Effect on Resonance Escape Probability When the fuel temperature increases, the fuel will expand, causing among other things a decrease in the fuel density, which affects the resonance escape probability and contributes an immediate temperature coefficient of reactivity:   1 ∂p ∂Nf 1 ∂NF p = −3θF ln p αTF = = ln p (5.86) p ∂NF ∂Tf NF ∂TF where (dN/dT )/T = −3(dl/dT )/ l = −3θ , with θ being the linear coefficient of expansion of the material. Since the fuel density decreases upon expansion, the resonance absorption decreases, and this reactivity coefficient contribution is positive (note that since p < 1, ln p < 0). After the increase in fission heating has been transported out of the fuel element into the coolant/moderator, the moderator temperature, TM , will increase, which causes the moderator to expand and contributes a delayed temperature coefficient of reactivity:   1 ∂p ∂NM 1 ∂NM p = 3θM ln p αTM = = − ln p (5.87) p ∂NM ∂TM NM ∂TM The decreased moderator density reduces the moderating power, reducing the probability that the neutrons will be scattered to energies beneath the resonance, hence increasing the resonance absorption and contributing a negative reactivity coefficient. Example 5.5: Resonance Escape Probability Fuel Temperature Coefficient for UO2 . The prompt feedback resulting immediately from an increase in power is associated with the increase in fuel temperature, the most significant part of which is due to the change in the resonance escape probability due to the Doppler broadening of resonances, as given by Eq. (5.85), and due to the fuel expansion, as given by Eq. (5.86). For a UO2 reactor consisting of assemblies of 1-cm-diameter fuel pins of height H in a water lattice with p /NF = 100 and fuel density ρ = 10 g/cm3 , SF /MF = π dH /π(d/2)2 Hρ = 0.4, I (300 K) = 11.6 + 22.8 × 0.4 = 20.72, and β  = 61 + 47(SF /MF ) × 10−4 = 79.8 × 10−4 . The resonance escape probability at 5.7 Reactivity Feedback 300 K is p = exp(−NF I /ξ p ) = exp[−20.72/(100 × 0.948)] = 0.8036, and ln(p) = −0.2186. The Doppler temperature coefficient of reactivity at 300 K is αTDF = ln(p)β  /2T 1/2 = (−0.2186)(79.8 × 10−4 )/(2)(17.32) = −5.036 × 10−5 k/k. The linear thermal expansion coefficient for UO2 is θF = 1.75 × 10−5 K−1 , and the fuel expansion contribution to the resonance escape probability temperature coefp ficient of reactivity is αT F = −3θF ln(p) = −3(1.75 × 10−5 ) · (−0.2186) = 1.148 × −5 10 k/k. Thus the total prompt fuel temperature coefficient of reactivity due to p the resonance escape probability is αTDF + αT F = −3.888 × 10−5 k/k. Thermal Utilization The thermal utilization can be written simply in terms of the effective cell-averaged fuel and moderator absorption cross sections discussed in Section 3.8: f= eff aF eff aF eff + aM → aF aF + aM (5.88) Recalling that  ≡ Nσ , the reactivity coefficient associated with the thermal utilization has an immediate negative component associated with the fuel temperature increase and a delayed positive contribution associated with the moderator density decrease:   1 ∂σaF 1 ∂aF ∂ξ 1 ∂NF 1 ∂f = (1 − f ) + + f ∂T σaF ∂TF aF ∂ξ ∂TF NF ∂TF  1 ∂σaM 1 ∂aM ∂ξ 1 ∂NM − + + σaM ∂TM aM ∂ξ ∂TM NM ∂TM    1 1 ∂aF ∂ξ  (1 − f ) − F + 3θF 2TF a ∂ξ ∂TF   1 ∂aM ∂ξ + 3θM − − M a ∂ξ ∂TM f f ≡ αTF + αTM (5.89) Account has been taken in writing Eq. (5.89) of the fact that the thermal disadvantage factor, ξ , which is used in the definition of effective homogenized fuel and moderator cross sections, will also be affected by a change in temperature. An increase in fuel temperature hardens (makes more energetic) the thermal neutron energy distribution, which reduces the spectrum average of the 1/v thermal fuel cross section and thus reduces the thermal utilization. An increase in the fuel temperature also reduces the fuel density, further reducing the thermal utilization. An increase in moderator temperature has little effect on the moderator cross section but reduces the moderator density, which increases the thermal utilization. 165 166 5 Nuclear Reactor Dynamics Nonleakage Probability The nonleakage probability can be represented by PNL  1 1 + M 2 Bg2 (5.90) Temperature increases can affect the nonleakage probability by changing the characteristic neutron migration length, or the mean distance that a neutron is displaced before absorption, and by changing the size of the reactor. Assuming that both of these effects are associated primarily with changes in the moderator temperature, we write   M 2 Bg2 1 ∂M 2 1 ∂Bg2 1 ∂PNL =− + PNL ∂TM 1 + M 2 Bg2 M 2 ∂TM Bg2 ∂TM (5.91) An increase in moderator temperature causes a decrease in moderator density, which affects the migration area as 1 ∂M 2 1 ∂DM 1 ∂aM (1 − f ) = − M 2 DM ∂TM a ∂TM M ∂TM = 6θM + 1 1 ∂f − 2TF 1 − f ∂T (5.92) where we have used a = aM + aF = aM (1 − f ). The geometric buckling Bg = G/ lR , where G is a constant depending on geometry (Table 3.3) and lR is a characteristic physical dimension of the reactor. Thus 1 ∂Bg2 = Bg2 ∂TM  lR G 2   ∂(G/ lR )2 1 ∂lR = −2 ∂TM lR ∂TM (5.93) and Eq. (5.91) becomes αTPMNL = = 1 ∂PNL PNL ∂TM M 2 Bg2 1 + M 2 Bg2  2 ∂lR 1 1 ∂f − 6θM − + lR ∂TM 2TF 1 − f ∂T  (5.94) A decrease in moderator density allows neutrons to travel farther before absorption, which increases the leakage and contributes a negative reactivity coefficient component. Expansion of the reactor means that a neutron must travel farther to escape, which contributes a positive reactivity coefficient component. Representative Thermal Reactor Reactivity Coefficients Reactivity coefficients calculated for representative thermal reactors are given in Table 5.2. 5.7 Reactivity Feedback Table 5.2 Representative Reactivity Temperature Coefficients in Thermal Reactors Doppler (k/k × 10−6 K−1 ) Coolant void (k/k × 10−6 /% void) Moderator (k/k × 10−6 K−1 ) Expansion (k/k × 10−6 K−1 ) BWR PWR HTGR −4 to −1 −200 to −100 −50 to −8 ≈0 −4 to −1 — −50 to −8 ≈0 −7 — +1 ≈0 Source: Data from Ref. 3; used with permission of Wiley. Example 5.6: UO2 Fuel Heat Removal Time Constant. It is important to emphasize that the temperature reactivity feedback associated with the various mechanisms that have been discussed take place at different times. The feedback associated with changes in the fuel temperature take place essentially instantaneously, since an increase in fission rate produces an immediate increase in fuel temperature. However, the increase in moderator/coolant temperature occurs later, after some of the additional heat is conducted out of the fuel element. The heat balance equation in the fuel element,     ∂ rκ∂T ∂T = r −1 + q  (5.95) ρC ∂t ∂r ∂r where ρ is the fuel density, κ the heat conductivity, C the heat capacity, and q  the volumetric fission heat source, can be used to estimate a time constant characterizing the conduction of heat out of the fuel element to the interface with the coolant/moderator for a fuel pin of radius a, τ ≈ ρCa 2 /κ. Typical parameters for a UO2 fuel element in a thermal reactor are a = 0.5 cm, κ = 0.024 W/cm · K, ρ = 10.0 g/cm3 , and C = 220 J/kg · K. The heat conduction time constant for heat removal from the fuel into the coolant is τ = ρCa 2 /κ = (10 g/cm3 )(220 J/kg · K)/(0.024 J/s · cm · K)(103 g/kg) = 22.9 s. For a smaller fuel pin characteristic of a fast reactor with a = 0.25 cm, the UO2 fuel time constant would be about 6 s. With a metal fuel instead of UO2 , the heat conductivity is much larger, and the heat removal time constants are on the order of 0.1 to 1.0 s. Startup Temperature Defect A reactor is initially started up from a cold condition by withdrawing control rods until the reactor is slightly supercritical, thus producing an exponentially increasing neutron population on a very long period. As the neutron population increases, the fission heating and thus the reactor temperature increase. This increase in temperature produces a decrease in reactivity (almost all reactors are designed to have a negative temperature coefficient) that would cause the neutron population to decrease and the reactor to shut down if the control rods were not withdrawn further to maintain an increasing neutron population. The total amount of feedback reactivity that must be offset by control rod withdrawal during the course of the startup to operating power level is known as the temperature defect. The temperature defects 167 168 5 Nuclear Reactor Dynamics for water-moderated reactors, graphite-moderated reactors, and sodium-cooled fast reactors are about k/k = 2–3 × 10−2 , 0.7 × 10−2 , and 0.5 × 10−2 , respectively. 5.8 Perturbation Theory Evaluation of Reactivity Temperature Coefficients Perturbation Theory The multigroup diffusion equations (Chapter 4) are −∇ · Dg ∇φg + rg φg = G  g  =g G 1  g  →g φg  + χg νf g  φg  , k  g =1 g = 1, . . . , G (5.96) where g  →g is the cross section for scattering a neutron from group g  to group g, rg is the removal cross section for group g, which is equal to the absorption cross section plus the cross section for scattering to all other groups, χg is the fraction of the fission neutrons in group g, Dg and νf g are the diffusion coefficient and the nu-fission cross section in group g, and φg is the neutron flux in group g. We now consider a perturbation in materials properties (e.g., as would be caused by a change in local temperature) such that the reactor is described by an equation like Eq. (5.96), but with Dg → Dg + Dg , g → g + g , where the  terms include changes in densities, changes in the energy averaging of the cross-section data and energy self-shielding, changes in spatial self-shielding, and changes in geometry. If we assume that the perturbation in materials properties is sufficiently small that it does not significantly alter the group fluxes, we can multiply the unperturbed and perturbed equations by φg+ , subtract the two, integrate over the volume of the reactor, and sum the resulting equations for all groups to obtain the perturbation theory estimate for the change in reactivity associated with the perturbation in material properties: k   k G   dr φg+ ∇ · (Dg ∇φg ) − φg+ rg φg + φg+ g=1 + φg+ χg ÷ G   g=1 % dr φg+ χg G   g  →g φg  g  =g (νf g  )φg  g  =1 G  G  νf g  φg  & (5.97) g  =1 The quantity φg+ , the importance of neutrons in group g in producing a subsequent fission event, is discussed in Chapter 13. This expression, together with the 5.8 Perturbation Theory Evaluation of Reactivity Temperature Coefficients subsidiary calculation of the g and Dg terms, including all the effects mentioned above, provides a practical means for the quantitative evaluation of reactivity coefficients in nuclear reactors. Example 5.7: Reactivity Worth of Uniform Change in Thermal Absorption Cross Section. With the assumption that all of the fission occurs in the thermal group, the reactivity worth of a uniform change in thermal absorption cross section in a uniform thermal reactor is k/k = ath Ith /νjth Ith ≈ ath /ath , because Ith , the integral over the reactor of the product of the thermal group importance function and flux, appears in both the numerator and denominator, and because in a critical reactor ath ≈ νfth . We now discuss some fast reactor reactivity coefficients that could not be treated by the more approximate method of the preceding section, although we emphasize that this perturbation theory calculation is also used for thermal reactor reactivity coefficient evaluation. Sodium Void Effect in Fast Reactors The reactivity change that occurs when sodium is voided from a fast reactor can be separated into leakage, absorption, and spectral components. The leakage and spectral components correspond to the first (Dg ) and third (g  →g ) terms, respectively, in Eq. (5.97). The absorption component corresponds to the second (rg ) and fourth (νf ) terms in Eq. (5.97), although the change in fission cross section is usually small and therefore neglected, and this component is usually referred to as the capture component. The spectral and capture components are normally largest in the center of the core, where the neutron flux and importance function are largest, and the leakage component is normally largest in the outer part of the core, where the flux gradient is largest. The magnitude of the sodium void coefficient varies directly with the ratio of the number of sodium atoms removed to the number of fuel atoms present. The spectral component of the sodium void coefficient is generally positive, is more positive for 239 Pu than for 235 U, and becomes increasingly positive as fissile material concentration decreases relative to sodium content. The capture component tends to become more positive with softer neutron spectra because of the 2.85-keV resonance in 23 Na, hence to become more positive with increasing sodium concentration relative to fuel concentration. The negative leakage component is generally smaller than the other two components, although the leakage component can be enhanced by the choice of geometrical configuration. As a result, the overall reactivity effect of voiding the central part of the core is positive, and may be positive for voiding of the entire core. This poses a serious safety concern that must be offset by proper design to ensure that other negative reactivity coefficients are dominant. Doppler Effect in Fast Reactors In fast reactors, the neutron energy spectrum includes the resonance regions of both the fissionable (235 U, 233 U, 239 Pu, 241 Pu) and nonfissionable (232 Th, 238 U, 169 170 5 Nuclear Reactor Dynamics 240 Pu) fuel isotopes. The Doppler effect in fast reactors is due almost entirely to resonances below about 25 keV. An increase in fuel temperature will produce an increase in both the fission and absorption cross sections, and the resulting change in reactivity can be positive or negative, depending on the exact composition. The temperature coefficient of reactivity can be estimated from     ∂σγ ∂σf ∂σf ∂k φ(E)dE = NF φf+ − φ + (E) + ∂TF ∂TF ∂TF ∂TF  1 ∂σf  NF (ν − 1 − α)φ(E)dE (5.98) ν ∂TF where NF is the density of fuel nuclei (sum over species implied), φ + (E) and φf+ are the importance of a neutron at energy E and of a fission neutron (i.e., the number of fissions the neutron subsequently produces). Since in a critical system each neutron will on average produce 1/ν fissions, φ + ≈ φf+ ≈ 1/ν is used in the second form of the estimate, and α ≡ σγ /σf has also been used. Since α generally decreases with increasing neutron energy (Chapter 2), the reactivity change will tend to be more positive/less negative for metal-fueled cores with a relatively hard spectrum. The oxygen in UO2 fuel softens (makes less energetic) the energy spectrum and thereby makes the reactivity change more negative/less positive. Detailed design calculations, using methods benchmarked against critical experiments, indicate that in larger reactors with a high fertile-to-fissile ratio the Doppler coefficient is sufficiently negative to provide a prompt shutdown mechanism in the event of excess fission heating of the fuel. Fuel and Structure Motion in Fast Reactors The increased fission heating coincident with an increase in the neutron population causes the fuel to expand radially and axially and to distort (e.g., bow) due to constraints. The expanding fuel first compresses, then ejects, sodium. The additional fission heat is transferred to the structure, producing a delayed expansion and distortion of the structure. The radial expansion, which is cumulative from the core center outward, results in a general outward radial movement of the fuel and in an expansion of the size of the reactor. The reactivity effect of this fuel and structure motion is highly dependent on the details of the design. However, a few simple estimates provide a sense of the magnitude of the effects. Example 5.8: Reactivity Effects of Fuel and Structure Expansion. Radial motion of the fuel by an amount r from an initial radial location r causes a reduction in local fuel density which varies as r 2 , leading to a local density change NF /NF ≈ (r 2 − (r + r)2 /r 2 ≈ −2r/r. Axial fuel expansion leads to linear fuel density decreases. The overall expansion reactivity coefficient is a combination of the negative effect of reduced fuel density and the positive effect of increased core size, hence reduced leakage. An overall expansion reactivity coefficient is of the form     R H NF NF exp + c (5.99) +b +d αTM = a R NF radial H NF axial 5.9 Reactor Stability Table 5.3 Reactivity Coefficients in a 1000-MWe Oxide-Fueled Fast Reactor Sodium expansion core Sodium expansion reflector Doppler Radial fuel pin expansion Axial core expansion Radial core expansion Temperature k/k×10−6 ◦ C−1 Power: k/k×10−6 MW−1 +3.0 −1.6 −3.2 +0.4 −4.1 −6.8 +0.085 −0.081 −0.628 +0.117 −0.181 −0.182 Source: Data from Ref. 9; used with permission of American Nuclear Society. where, for the example of a 1000-MWe UO2 reactor with H/D = 0.6, the constants are (a = 0.143, b = 0.282, c = 0.131, d = 0.281). Fuel Bowing Fuel distortion (e.g., bowing) is very much a function of how the fuel is constrained. The calculated reactivity effect of inward bowing in the metal fueled EBR-II was k/k ≈ −0.35V /V ≈ −0.7R/R ≈ 0.0013. This predicted positive reactivity due to bowing exceeded the combined negative reactivity from all other effects at full flow and intermediate power, suggesting the possibility of a positive reactivity coefficient over the intermediate power range, consistent with experimental observation. Representative Fast Reactor Reactivity Coefficients Reactivity coefficients calculated for a representative fast reactor design are given in Table 5.3. 5.9 Reactor Stability Reactor Transfer Function with Reactivity Feedback Since the reactor power is related directly to the neutron population, we can rewrite the neutron kinetics equations, in particular Eq. (5.39), in terms of the power, P = Ef nvνf · Vol, where Ef is the energy release per fission. If we expand the power about the equilibrium power P0 as P (t) = P0 + P1 (t) and limit consideration to the situation |P1 /P0 |  1, we find that   ∞ dP1 (t) 1 = ρ(t)P0 + dτβD(τ )[P1 (t − τ ) − P1 (t)] dt  0 (5.100) 171 172 5 Nuclear Reactor Dynamics Representing the reactivity as the sum of an external reactivity, ρex , such as may be caused by control rod motion, and a feedback reactivity, ρf , caused by the inherent reactivity feedback mechanisms discussed in the preceding two sections, the total reactivity may be written ρ(t) = ρex (t) + ρf (t)  t f (t − τ )P1 (τ )dτ = ρex (t) + −∞  = ρex (t) + ∞ f (τ )P1 (t − τ )dτ (5.101) 0 where f (t − τ ) is the feedback kernel that relates the power deviation P1 = P − P0 at time t − τ to the resulting reactivity at time t . Using the last form of Eq. (5.101) in Eq. (5.100), Laplace transforming (equivalently, assuming an est time dependence), and rearranging yields a transfer function, H (s), relating the external reactivity input to the power deviation from equilibrium: P1 (s) = Z(s) P0 ρex (s) ≡ H (s)P0 ρex (s) 1 − P0 F (s)Z(s) (5.102) This new transfer function contains the zero-power transfer function, Z(s), which relates the prompt and delayed neutron response to the external reactivity, and the feedback transfer function, F (s), which relates the feedback reactivity to the power deviation P1 = P − P0 : ρf (s) = F (s)P1 (s) (5.103) Note that when P0 → 0, H (s) → Z(s). The linear stability of a nuclear reactor can be determined by locating the poles of H (s) in the complex s-plane. This follows from noting that when Eq. (5.102) is Laplace inverted, the solutions for P1 (t) ∼ exp(sj t), where the sj are the poles of H (s). Any poles located in the right half of the complex s-plane (i.e., with a positive real part) indicate a growing value of P1 (t)—an instability. Since Z(s) appears in the numerator and denominator of H (s), its poles (the roots of the inhour equation) cancel in H (s), and the poles of H (s) are the roots of 1 − P0 F (s)Z(s) = 0 (5.104) We can anticipate from Eq. (5.104) that the poles of H (s), hence the linear stability of the reactor, will depend on the equilibrium power level, P0 . Stability Analysis for a Simple Feedback Model To determine the roots of Eq. (5.104), we must first specify a feedback model in order to determine the feedback transfer function, F (s). We consider a two- 5.9 Reactor Stability temperature model in which the deviation in the fuel temperature from the equilibrium value satisfies dTF (t) = aP1 (t) − ωF TF (t) dt (5.105) where a involves the heat capacity and density of the fuel and ωF is the inverse of the heat transfer time constant of the fuel element (i.e., the time constant for removal of heat from the fuel element into the coolant/moderator). The temperature deviation about the equilibrium value in the coolant/moderator satisfies dTM (t) = bTF (t − t) − ωM TM (t) dt (5.106) where b involves the mechanism governing the response of the coolant/moderator temperature to a change in the fuel temperature, ωM is the inverse of the heat removal time constant for the moderator, and for the sake of generality we assume that the coolant mass flow rate is varied in response to the fuel temperature at an earlier time (t − t ). The same model could be applied to any two-temperature representation of a reactor core. For example, we could consider TF to be the temperature of a simultaneously heated fuel–coolant region and TM to represent the temperature of the structure in a fast reactor model. Writing  ρ(t) = ρex (t) + αF TF (t) + αM TM (t) ≡ ρex (t) + t f (t − τ )P1 (τ )dτ (5.107) 0 defines the feedback kernel, f (t − τ ), where TF (t) and TM (t) are deviations from the equilibrium temperatures. Laplace transforming these three equations, using the convolution theorem, and combining leads to identification of the feedback transfer function: F (s) = XM e−st XF + 1 + s/ωF (1 + s/ωF )(1 + s/ωM ) (5.108) where XF = aαF /ωF and XM = (abαM /ωF ωM ) are the steady-state reactivity power coefficient for the fuel and coolant/moderator, respectively. Using the zeropower transfer function, Z(s), of Eq. (5.46), but in the one-delayed neutron group approximation, and the feedback transfer function, F (s), of Eq. (5.108), Eq. (5.104) for the poles of the reactor transfer function with feedback, H (s), becomes 1−   XF XM e−st P0 =0 + s[ + β/(s + λ)] 1 + s/ωF (1 + s/ωF )(1 + s/ωM ) (5.109) There are a number of powerful mathematical techniques from the field of linear control theory (Nyquist diagrams, root-locus plots, Routh–Hurwitz criterion, iterative root finding methods, etc.) for finding the roots of Eq. (5.109), or of the more complex equations that would result from more detailed reactivity feedback models. Some simplification results from limiting attention to growth rates that are 173 174 5 Nuclear Reactor Dynamics small compared to the inverse neutron generation time (s  −1 ), allowing neglect of the  term. We now consider two additional approximations which allow us to obtain valuable physical insights. If we set XM ∼ αM = 0 (i.e., neglect the coolant/moderator feedback), Eq. (5.109) can be solved analytically to obtain (    P0 XF 1 4(P0 XF /β)(λ/ωF ) s± = ωF (5.110) −1 1± 1+ 2 β (P0 XF /β − 1)2 If the fuel power coefficient is positive (XF ∼ αF > 0), the term under the radical is positive and greater than unity, both roots are real, and one root is positive, indicating an instability. If the fuel power coefficient is negative (XF ∼ αF < 0), the real parts of both roots are negative, indicating stability. Threshold Power Level for Reactor Stability If we retain XM finite but restrict our consideration to instabilities with growth rates much less than the inverse fuel heat removal time constant, s  ωF , and set the time delay to zero, t = 0, we can again solve Eq. (5.109) analytically for the poles of the reactor transfer function, H (s): 1− 1 s± = − ωM 2 P 0 XF β  XM λ X F + 1 + ωM 1 − P0βXF      *1/2  4 ωλM P0βXF 1 − P0βXF × 1± 1+  M 2 λ 1 − P0βXF X X F + 1 + ωM   (5.111) This expression reveals the existence of a threshold equilibrium power level, P0 , above which a reactor becomes unstable. As P0 → 0, the two roots approach 0 and −ωM , a marginally stable condition, and do not depend on the reactivity power coefficients XM and XF . As P0 increases, the nature of the solution depends on XM and XF . Suppose that the fuel power coefficient is positive, XF > 0, and the moderator power coefficient is negative, XM < 0; this situation might arise, for example, in a fast reactor when XF represents the combined Doppler, fuel expansion, and sodium void coefficients of the fuel–coolant mixture and XM represents the structure expansion coefficient. Taking XF /XM = − 12 and ωM = 14 , the roots of Eq. (5.111) are plotted as a function of |XM |P0 /β (denoted at P0 ) in Fig. 5.4. As P0 increases from zero, the marginally stable (s = 0) root moves into the left-half complex s-plane and the (s = ωM ) root becomes less negative, indicating that the reactor would be stable. At |XM |P0 /β = 0.0962, the roots become complex conjugates with a real part that increases with P0 . At |XM |P0 /β > 23 , the real part of the two roots becomes positive, indicating that the reactor would become unstable above a certain threshold operating power level. At |XM |P0 /β > 1.664, the roots become real and positive, with one increasing and the other decreasing with increasing P0 , continuing to indicate instability. 5.9 Reactor Stability Fig. 5.4 Characteristic roots s+ and s− of Eq. (5.111) as a function of critical power level P0 (|XM |P0 /β) (XF > 0, XF /XM = − 12 , WM = 14 S). (From Ref. 8; used with permission of Van Nostrand.) The total power coefficient at steady state is negative (F (0) = XF + XM < 0), but the reactor in this example was unstable above a certain threshold power level. The positive fuel power feedback was instantaneous because the fuel temperature increases instantaneously in response to an increase in fission heating. However, the coolant/moderator temperature does not increase instantaneously because of moderator heat removal, but increases on a time scale governed by the moderator −1 following a change in fuel temperature, as may be heat removal time constant ωM seen by solving Eq. (5.106) for a step increase TF at t = 0:  t < t  0, (5.112) TM = bTF  (1 − e−ωM (t+t) ), t ≥ t ωM The delay of the moderator temperature response to an increase in the temperature of the fuel was neglected; its inclusion would contribute further to the possibility of instabilities. It is clear that heat removal time constants play an important role in the stability of a reactor. The two-temperature feedback model can be generalized to investigate the stability of a variety of different feedback models that can be characterized by a fast (f )- and a slow (s)- responding temperature. For a fast temperature response that was either prompt (ωf = 0) or zero (Xf = 0) plus a slow temperature response with a finite time constant (ωs = 0) determined either by heat conduction or heat convection, the results are given in Table 5.4. More General Stability Conditions A necessary condition for stability is  ∞ F (0) = f (t)dt < 0 0 (5.113) 175 176 5 Nuclear Reactor Dynamics Table 5.4 Instability Conditions for Some Simple Two-Temperature Feedback Models Reactivity Coefficients Fast (ωf = 0) Slow (ωs = 0) Heat Removal F (s) Instability Xs 1+s/ωs Xs e−s/ωs None Xf = 0 Xs < 0 Conduction Xf = 0 Xs < 0 Convection P0 > Pthresh Xf > 0 Xs < 0 Convection Xs Xf + 1+s/ω s Xs Xf + 1+s/ω s Xf + Xs e−s/ωs Xf < 0 Xs < 0 Convection Xf + Xs e−s/ωs P0 > Pthresh Xf = 0 Xs1 < 0 Convection s2 Xs1 e−s/ωs1 + 1+s/ω s2 P0 > Pthresh Xs2 < or > 0 Conduction Xf > 0 Xs < 0 Conduction Xf < 0 Xs < 0 Conduction X P0 > Pthresh None P0 > Pthresh Source: Data from Ref. 9; used with permission of American Nuclear Society. However, this is not a sufficient condition, as the analysis above, in which F (0) = XF + XM < 0, demonstrates. The result discussed in the preceding example suggests a useful generalization—a reactor is on the verge of becoming unstable when the transfer function, H (s), has a pole with purely imaginary s [i.e., when Eq. (5.104) has a purely imaginary root s = iω]. Except for values of ω for which Z(iω) = 0, Eq. (5.104), which determines the poles of the transfer function, can be rewritten in the case s = iω:  βj iω 1 − P0 F (iω) = iω + − P0 F (iω) = 0 Z(iω) iω + λj 6 G(iω) = (5.114) j =1 If this equation has a solution, it corresponds to a condition for which the reactor is on the verge of instability. A necessary condition for a solution is that Z −1 (iω) and F (iω) have the same ratio of real to imaginary parts (i.e., the same phase). If Z −1 (iω) and F (iω) do have the same phase at some ω = ωres , there will be some value of P0 for which Eq. (5.114) has a solution. If this value of P0 is physically reasonable (P0 ≥ 0), there is instability onset at this (P0 , ωres ) condition. The real and imaginary parts of 1/Z(iω) are  Re  Im 1 Z(iω) = 6  ω 2 βj j =1 ω2 + λ2j (5.115) 1 Z(iω) = ω + 6  ωβj λj j =1 ω2 + λ2j which are both real and positive, thus are in the upper right quadrant of the complex plane. Therefore, a necessary condition for G(iω) = 0 to have a solution is that 5.9 Reactor Stability Fig. 5.5 Plot of R = Re{F (iω)} + iI {F (iω)} of Eq. (5.108) with t = 0: case (a) XF = 0, XM < 0; case (b), XF < 0, XM < 0; case (c), |XM | > XF > 0, XM < 0. (From Ref. 8; used with permission of Van Nostrand.) the real and imaginary parts of the feedback transfer function, F (iω), also lie in the same quadrant (i.e., both be real and positive). Hence a necessary condition for an instability is Re{F (iω)} > 0 and Im{F (iω)} > 0 (5.116) We now consider the example above with the simple feedback model of Eqs. (5.105) to (5.108), but with the delay term t = 0. The qualitative behavior of the real and imaginary parts of F (iω) of Eq. (5.108) are plotted in Fig. 5.5 for three different cases, all of which have a negative moderator power coefficient, XM < 0. Case (a) corresponds to no reactivity feedback from the fuel (XF = 0); the instability criterion of Eq. (5.116) is satisfied for ω > (ωM ωf )1/2 , even though the steady-state power coefficient X(0) = XM < 0. For case (b), with a sufficiently large negative value of the fuel power coefficient, XF < 0, the criterion of Eq. (5.116) is never satisfied and the reactor is stable. In case (c), the fuel reactivity power coefficient is positive but smaller in magnitude than the negative moderator reactivity power coefficient, |XM | > |XF | > 0, which is the situation leading to the solution of Eq. (5.111); the reactor can become unstable, as found above from examination of the roots given by Eq. (5.111). A sufficient condition for unconditional stability (i.e., no power threshold) has been shown to be  ∞ f (t) cos(ωt)dt ≤ 0 (5.117) Re{F (iω)} = 0 which is a requirement that the phase angle of the feedback transfer function, −F (s), along the iω-axis is between −90◦ < φ < +90◦ ; thus the feedback response is negative and less than 90◦ out of phase with the power change that produced it. This phase constraint places constraints on the time delays. This sufficient criterion for stability has been found to be over restrictive, however. 177 178 5 Nuclear Reactor Dynamics Table 5.5 Sufficient Conditions for Unconditional Stability of Two-Temperature Feedback Models Reactivity Coefficients Coupled prompt Xf , conduction Xs Uncoupled conduction Xf and Xs Coupled conduction Xf and Xs Coupled prompt Xf , convection Xs Coupled conduction Xf , convection Xs F (iω) Stability Criterion Xs Xf + 1+iω/ω s Xf ≤ 0 and Xf + Xs < 0 Xf Xs 1+iω/ωf + 1+iω/ωs Xf Xs 1+iω/ωf + (1+iω/ωf )(1+iω/ωs ) Xf + Xs e−iω/ωs Xf Xs e−iω/ωs 1+iω/ωf + 1+iω/ωs Xf + Xs ≤ 0, Xf ωf + Xs ωs < 0, and Xf ωf2 + Xs ωs2 ≤ 0 Xf < 0, Xf + Xs ≤ 0, and Xf ωf − Xs ωs ≤ 0 Xf < 0 and −Xf ≥ |Xs | Never unconditionally stable The unconditional stability sufficient condition of Eq. (5.117) has been used to determine unconditional stability criteria for a variety of feedback models that can be characterized by a fast (f ) and a slow (s) responding temperature. The fast temperature response was either prompt (ωf = 0) or determined by heat conduction, and the slow temperature response was with a finite time constant (ωs = 0) determined by either heat conduction or heat convection. The results are given in Table 5.5. Power Coefficients and Feedback Delay Time Constants It is clear from the previous discussion that the reactivity temperature coefficients actually enter the analysis as reactor power coefficients, associated with which there are time delays related to heat transfer and removal time constants, and that the results of the analysis are dependent on the delay times as well as on the temperature coefficients. We can generalize the two-temperature model to define a general reactor power coefficient: X(t) =    ∂ρ ∂Tj (t) ∂ρ ∂Tj (t) + ∂Tj ∂P ∂Tj ∂P (5.118) j where ∂ρ/∂Tj are the reactivity temperature coefficients corresponding to a change in local temperature Tj . The quantities ∂ρ/∂Tj are reactivity temperature gradient coefficients denoting the change in reactivity due to a change in temperature gradient (e.g., as would produce bowing of a fuel element). These reactivity coefficients can be calculated as discussed in the two preceding sections. The quantities ∂Tj /∂P and ∂Tj /∂P are the time-dependent changes in local temperature and temperature gradients resulting from a change in reactor power and must be cal- 5.10 Measurement of Reactor Transfer Functions culated from models of the distributed temperature response to a change in reactor power. The time constants that determine the time delays in the various local temperature responses to a power increase depend on the specific reactor design. Some simple estimates suffice to establish orders of magnitude. The time constant for heat transfer out of a fuel pin of radius r or plate of thickness r, density ρ, heat capacity C, and thermal conductivity κ is τf = ρCr 2 /κ, which generally varies from a few tenths to a few tens of seconds. The effect of cladding and the surface film drop is to increase the time constant for the fuel element. The lumped time constant for the coolant temperature is τc = Cc / h + (Z/2v)(1 + Cf /Cc ), where Cc and Cf are the heat capacities per unit length of the coolant and fuel, respectively, h is the heat transfer coefficient between fuel and coolant, Z is the core height, and v is the coolant flow speed. Typical values of τc vary from a few tenths to a few seconds. 5.10 Measurement of Reactor Transfer Functions Measurement of the reactor transfer function provides useful information about a reactor. A measurement at low power can identify incipient instabilities which produce peaks in the transfer function. Provided that the feedback mechanisms do not change abruptly with power, the low-power transfer function measurements can identify conditions that would be hazardous at high power, thus allowing for their correction. Information about the feedback mechanisms can be extracted from measurement of the amplitude and phase of the transfer function. Any component malfunction that altered the heat removal characteristics of the reactor would affect the transfer function, so periodic transfer function measurements provide a means to monitor for component malfunction. Rod Oscillator Method The sinusoidal oscillation of a control rod over a range of frequencies can be used to measure the transfer function, as described in Section 5.6. The results of Eqs. (5.60) to (5.64) apply to a reactor with feedback when n0 Z(iω) is replaced by P0 H (iω). There are some practical problems in measuring the transfer function with rod oscillation. There will be noise in the detector response, which will require a sufficiently large reactivity oscillation for the detector response to be separable from the noise, and nonlinear effects [i.e., the term ρn1 which was neglected in Eq. (5.42)] may invalidate the interpretation. Furthermore, the oscillation will not be perfectly sinusoidal, and it will be necessary to Fourier analyze the detector response to isolate the fundamental sinusoidal component. Correlation Methods It is possible to measure the reactor transfer function with a nonperiodic rod oscillation. Consider the inverse Laplace transform of Eq. (5.102): 179 180 5 Nuclear Reactor Dynamics  P1 (t) = t −∞  ρex (τ )h(t − τ )dτ = ∞ ρex (t − τ )h(τ )dτ (5.119) 0 which relates the relative power variation from equilibrium [P1 /P0 = (P − P0 )/P0 ] to the time history of the external reactivity—the rod oscillation in this case— including the effect of feedback. The kernel h(t) is the inverse Laplace transform of the transfer function, H (s). The cross correlation between the external reactivity and the power variation is defined as  T  T 1 1 ρex (t)P1 (t + τ )dt = ρex (t − τ )P1 (t)dt (5.120) φρP ≡ 2T −T 2T −T where T is the period if ρex and P1 are periodic and T goes to infinity if not. Using Eq. (5.119) in Eq. (5.120) yields   ∞  T      1  φρP = ρex (t − τ ) ρex t − t h t dt dt 2T −T 0    ∞  T   1 = h(t  ) ρex (t − τ )ρex t − t  dt dt  2T −T 0  ∞ ≡ h(t  )φρρ (τ − t  )dt  (5.121) 0 where φρρ is the reactivity autocorrelation function. Taking the Fourier transform of Eq. (5.121) yields an expression for the transfer function H (−iω) = ∓ F {φρP (τ )} F {φρρ (τ )} where the transforms  F {φρP (τ )} ≡ ∞  ∞ −∞ F {φρρ (τ )} ≡ −∞ (5.122) eiωτ φρP (τ )dτ e (5.123) iωτ φρρ (τ )dτ are known as the cross spectral density and the input or reactivity spectral density, respectively. If the control rod (or other neutron absorber) position is varied randomly over a narrow range and a neutron detector response is recorded, the reactivity autocorrelation function, φρρ , and the reactivity-power cross-correlation function, φρP , can be constructed by numerically evaluating the defining integrals over a period of about 5 min using a series of delay intervals, τ , increasing in discrete steps of about τ = 0.01 s. The cross spectral density and reactivity spectral density can then be calculated by numerically evaluating the defining Fourier transform; for example,  F {φρP (τ )}  φρP (nτ )(cos nωτ + i sin nωτ )τ (5.124) n 5.10 Measurement of Reactor Transfer Functions where n varies from a large negative integer to a large positive integer. There are sophisticated fast Fourier transform methods which are used in practice for evaluation of the cross and reactivity spectral densities. Experimentally, it is convenient to use a reactivity variation that changes from positive to negative at definite times, so that the reactivity autocorrelation function is nearly a delta function. For such a pseudorandom binary reactivity variation,     φρρ τ − t   const δ t − t  (5.125) In this case, it follows from Eq. (5.121) that φρP (τ )  const h(τ ) (5.126) F {φρP (τ )}  const H (−iω) and that the amplitude and phase of the transfer function can be extracted from the computation of only the cross correlation function. By repeating the Fourier transforms of Eq. (5.123) for different values of ω, the frequency dependence of H (−iω) can be determined. Reactor Noise Method Minor and essentially random variations in temperature and density within a nuclear reactor, such as bubble formation in boiling water reactors, produce small and essentially random reactivity variations. Autocorrelation of the response of a neutron detector, which is proportional to the reactor neutron population or power, provides a means of determining the amplitude of the reactor transfer function from this noise. Writing the power autocorrelation function φP P (τ ) ≡ 1 2T  T −T P1 (t)P1 (t + τ )dt (5.127) and using Eq. (5.119) yields   ∞  T      ∞      1  dt h t ρex t − t dt h t ρex t + τ − t  dt  φP P (τ ) = 2T −T 0 0     ∞  T    ∞    1     = h t h t ρex t − t  ρex t + τ − t  dt dt  dt  2T −T 0 0  ∞  ∞       = h t  dt  h t  φρρ τ + t  − t  dt  (5.128) 0 0 Fourier transformation then gives H (−iω)H (iω) = |H (iω)|2 = F {φP P (τ )} F {φP P (τ )}  F {φρρ (τ )} const (5.129) 181 182 5 Nuclear Reactor Dynamics where the fact that the autocorrelation function of a random reactivity input is a delta function, the Fourier transform of which is a constant, has been used in writing the final form. Thus the amplitude, but not the phase, of the reactor transfer function can be determined from autocorrelation of the reactor noise. Again, the frequency dependence is determined by taking the Fourier transform with respect to various frequencies, ω. This provides a powerful technique for online, nonintrusive monitoring of an operating reactor for component malfunction and incipient problems. Example 5.9: Reactor Transfer Function Measurement in EBR-I. The reactor transfer function measurement on the early EBR-I sodium-cooled, metal fuel fast reactor provides a good example of the physical insight provided by transfer function measurements. The Mark II core was stable at lower power levels, but at moderate power levels an oscillatory power was observed. The measured transfer function is shown in Fig. 5.6: in part (a) for several values of the coolant flow rate (gallons Fig. 5.6 Reactor transfer function EBR-1: (a) as a function of coolant flow rate; (b, c) as a function of reactor power. (From Ref. 9; used with permission of American Nuclear Society.) 5.11 Reactor Transients with Feedback per minute), and in parts (b) and (c) for several values of the reactor power level. At the lower coolant flow rates and the higher power levels there is a pronounced resonance in the transfer function, suggesting an incipient instability, which is not present at the higher flow rates and lower power levels. The Mark II core was known to have a prompt reactivity feedback which added reactivity with an increase in power or a decrease in coolant flow. However, when steady state was achieved following an increase in power at constant flow, the net change in reactivity was negative, indicating an overall asymptotic power coefficient that was negative. Calculations indicated that the Doppler effect was negligible, that bowing of the fuel rods toward the center of the core contributed significant positive reactivity, and that the outward expansion of the structural plates supporting the fuel rods led to a delayed outward movement of the fuel rods that contributed negative reactivity. A three-temperature model was used to explain the phenomena observed. The fast positive reactivity was modeled as due to the fuel bowing, and the delayed negative reactivity was modeled as the fuel motion due to the delayed outward motion of the fuel rods upon expansion of the structural plates. Heat conduction plus convection for the two separate structural effects led to a three-term representation of the power feedback. After correcting for the frequency dependence of the oscillatory heat flow, the model achieved very good agreement with the transfer function measurements. 5.11 Reactor Transients with Feedback The dynamics equations are intrinsically nonlinear when feedback effects are included. The calculation of reactor transients is carried out with very sophisticated computer codes which model in detail the coupled dynamics of the neutrons, temperature, flow, structural motion, change of state, and so on. However, some physical insight as to the effects of feedback can be obtained by considering the simple model of Section 5.4 in the presence of feedback. The point kinetics equations with feedback may be written in the one delayed neutron group approximation as   ρex + αF T (t) − β dn(t) = n(t) + λC(t) dt  (5.130) dC(t) β = n(t) − λC(t) dt  where a feedback reactivity ρf (t) = αf T (t) has been added to the step reactivity insertion ρex . We will treat the temperature, T , as either a fuel temperature or a lumped fuel–moderator temperature which satisfies ρCP dT (t) = Ef f vn(t) − θT (t) dt (5.131) 183 184 5 Nuclear Reactor Dynamics where ρ is the density, Ef the deposited energy per fission, and θ ≈ κ/(heat transfer distance) account for conductive heat removal. In Section 5.4 we found that the response to a step subprompt-critical (ρex < β) reactivity insertion into a critical reactor was a prompt jump that changed the neutron density from n0 to n0 /(1 − ρex /β) in a time on the order of the neutron generation time, , followed by a slow rise (ρex > 0) or decay (ρex < 0) of the neutron density on the delayed neutron decay constant time scale. We examine these two phases of the transient separately in the presence of feedback. Step Reactivity Insertion (ρex < β): Prompt Jump During the initial phase of the transient for a few  following the reactivity insertion, the delayed neutron precursor decay source is constant at the critical equilibrium value λC0 = (β/)n0 . In the absence of feedback, the solution of Eq. (5.130) in this case is  n(t) = n0 exp ρex − β t   1+ β   0 t    ρex − β  t dt  exp −  n0  1 − ρex /β (5.132) Assuming that the feedback is on the fuel temperature, which responds instantaneously to an increase in the fission rate, the corresponding solution with feedback reactivity is   ρex + αf T (t) − β t       t ρex + αf T (t  ) − β  β t dt  exp − × 1+  0  n(t) = n0 exp (5.133) On this short time scale t ∼   ρCp /θ , the solution of Eq. (5.131) is Ef vf T (t)  ρCp  t   n t  dt  (5.134) 0 If the feedback is negative (αf < 0), the effect of the feedback is to reduce the magnitude of the input reactivity step. If ρex > 0, n and T increase in time and ρf = αf T < 0; if ρex < 0, n and T decrease in time and ρf = αf T > 0; (T0 = 0). If the feedback is positive (αf > 0), the effect of the feedback is to enhance the magnitude of the input reactivity step. If ρex > 0, n and T increase in time and ρf = αf T > 0; if ρex < 0, n and T decrease in time and ρf = αf T < 0. Thus negative feedback reactivity would reduce the magnitude of the prompt jump and perhaps reverse the sign if the feedback reactivity exceeds the input reactivity; positive feedback reactivity would enhance the magnitude of the prompt jump. 5.11 Reactor Transients with Feedback Step Reactivity Insertion (ρex < β): Post-Prompt-Jump Transient We saw in Section 5.4 that in the absence of feedback, after the initial prompt jump in the neutron density on the prompt neutron time scale, the subsequent transient evolves on the slower time scale of the delayed neutron precursor decay: n(t) = n0 exp{(λρex /β)t/(1 − ρex /β)} 1 − ρex /β (5.135) For the problem with feedback, we make use of the prompt-jump approximation (set dn/dt = 0) and solve Eqs. (5.130) to obtain n(t)  +  n0 exp −λ t − , t dt  0 1−[ρex +αf T (t  )]/β 1 − [ρex + αf T (t)]/β (5.136) which reduces to Eq. (5.135) when αf = 0. Note that Eq. (5.136) is valid only for the time after the prompt jump in neutron density between t = 0 and t = tpj ≈ . This equation evaluated at tpj implies an effective prompt jump from n0 → n0 /[1 − (ρex + αf T (tpj ))/β], to be compared with the effective prompt jump from n0 → n0 /(1−ρex /β) in the case without feedback implied by Eq. (5.135). Equation (5.131) can be solved formally for the temperature T (t) = Ef vf ρCp  t     n t  exp −(θ/ρCp ) t − t  dt  (5.137) 0 The presence of feedback can have a dramatic effect on the course of the transient. Consider a positive step reactivity insertion, 0 < ρex < β, which without feedback would result in an exponentially increasing neutron density with period (β/ρex − 1)/λ. With negative reactivity feedback (αf < 0), the period becomes longer (the rate of increase is slower), or even becomes negative (the neutron density decreases in time) if |αf |T (t) becomes greater than ρex . For a negative step reactivity insertion, ρex < 0, and negative reactivity feedback, the presence of feedback with the decreasing temperature causes the decay in the neutron density to become slower and even reverse and start increasing if |αf T (t)| becomes greater than |ρex |. Thus a reactor with a negative temperature coefficient of reactivity will adjust automatically to a step reactivity insertion by seeking a new critical condition. For example, when a cold reactor is started up by withdrawing the control rods to produce an increasing neutron population and increasing fission heating, the negative reactivity will increase also, until the reactor reaches a new temperature and neutron population at which it is just critical. A negative temperature coefficient of reactivity also allows a reactor to automatically load follow (an increase in power output demand will result in a decrease in coolant inlet temperature, which produces a positive reactivity that causes the neutron population and the fission rate to increase until a new critical condition is reached at higher power). 185 186 5 Nuclear Reactor Dynamics 5.12 Reactor Fast Excursions The examination of hypothetical accidents requires the analysis of fast, supercritical excursions in the neutron population in a reactor. Although this analysis is done with sophisticated computer codes, which solve the coupled neutron–thermodynamics–hydrodynamics equation of state equations, there are several analytical models which provide physical insight into the phenomena of fast supercritical reactor excursions. Delayed neutron precursors respond too slowly to be important in such transients and may be neglected. Step Reactivity Input: Feedback Proportional to Fission Energy The prompt neutron kinetics equation for a step reactivity input k0 > kβ and a feedback negative reactivity proportional to the cumulative fission energy release is described by  1 dP (t) k − 1 k0 − αE E(t) k0 αE t     P t dt (5.138) = = = − P dt     0 where k0 is measured relative to prompt critical and  t   ∂k E(t) ≡ P t  dt  , αE ≡ ∂E 0 (5.139) The solution of Eq. (5.139) is E(t) = where (k0 /) + R αE/ (  R≡ k0  2  +2 1 − e−Rt [(R+k0 /)] −Rt [(R−k0 /)] e +1  αE P0  (5.140) (5.141) For transients initiated from low initial power level, P0 , R ≈ k0 / and    ! 2(k0 /)2 −(k /)t −(k0 /)t 0 E(t)  2(k0 /αE )(1 − e ) e + 1 (5.142) (αE /)P0 The instantaneous power is " 2    R + (k0 /) −Rt 2R 2 R + (k0 /) −Rt +1 e e αE / R − (k0 /) R − (k0 /) " 2 (k0 /)2 −(k0 /)t 4(k0 /)4 −(k0 /)t 2  e e +1 (5.143) (αE /)P0 (αE /)2 P0 P (t) = Ė(t) = where the second form is valid only for low initial power. 5.12 Reactor Fast Excursions Equation (5.143) describes a symmetrical power excursion that increases to a maximum power Pmax = (k0 /)2 /2(αE /) at t ≈ 1.3/(k0 /) and then decreases to zero. The width of the power burst at half maximum is ≈3.52/(k0 /), and the total fission energy produced in the burst is 2k0 /αE . Ramp Reactivity Input: Feedback Proportional to Fission Energy If, instead of a step reactivity input, the external reactivity input is a ramp (e.g., as might occur in rod withdrawal), Eq. (5.138) becomes 1 dP (t) at − αE E(t) at αE = = − P dt     t   P t  dt  (5.144) 0 which has a solution of the form E(t) = a t + periodic function αE (5.145) The power level has a background (a/αE ) upon which is superimposed a series of oscillations as the net external plus feedback reactivity oscillates about prompt critical (ρ = β). We now examine one of the power oscillations. Differentiating Eq. (5.144) yields an equation for the instantaneous period θ ≡ (dP /dt)/P : a αE dθ(t) = − P (t) dt   (5.146) which may be combined with Eq. (5.144) to obtain dP θP = dθ a/ − (αE /)P (5.147) This equation has the solution 1 2 a P (t) αE − θ (t) = ln [P (t) − P0 ] 2  P0  (5.148) The maximum power at the peak of the oscillation occurs when θ = 0 and thus satisfies Pmax = P0 + a Pmax a Pmax ln  ln αE P0 αE P0 (5.149) where the second form is only valid for P0  Pmax , where P0 = a/αE now refers to the background power at the beginning of the oscillation. Step Reactivity Input: Nonlinear Feedback Proportional to Cumulative Energy Release The Doppler feedback coefficient in large fast power reactors is not constant but is calculated to vary approximately inversely with fuel temperature, and theoretical considerations suggest that it varies inversely with fuel temperature to the 32 power. 187 188 5 Nuclear Reactor Dynamics If we assume no heat loss from the fuel and constant specific heat to relate the fuel temperature increase during a transient to the cumulative fission energy release, we can represent a broad class of temperature-dependent feedback reactivities as αE E n , where αE now refers to the value of the feedback coefficient at the temperature at which the transient is initiated. In this case, the prompt neutron dynamics equation for a step external reactivity input k0 is  1 dP (t) k0 αE = − [E(t)]n P dt   (5.150) This equation has the solution for the cumulative fission energy release   k0 1/n ! E(t) = (n + 1)  1 + ne−(nk0 /)t αE 1/n (5.151) which can be differentiated to obtain the instantaneous power P (t) = Ė(t)     k0 1/n 2 k0 −(nk0 /)t ! n e = (n + 1)  1 + ne−(nk0 /)t αE  1/(n+1) (5.152) Once again, the power increases to a maximum value, in this case Pmax =   n (k0 /)n+1 1/n  / 1+n αE and then decreases to zero. The total energy release in the burst is Etot =  ]1/n . [(1 + n)k0 /αE Bethe–Tait Model It is clear that the course of a reactor excursion produced by a given external reactivity insertion is very sensitive to the feedback reactivity, hence to the evolution of the thermodynamic, hydrodynamic, and geometric condition of the reactor. The coupled evolution of these variables is calculated numerically in modern analyses. However, we can gain valuable physical insight by considering an early semianalytical model developed for fast metal fuel reactors. The prompt neutron dynamics are determined by k−1−β k 1 dP (t) = = P dt   = k0 + kinput (t) + kdispl (t) + kother (t) (5.153) where k0 is the initiating step reactivity (relative to prompt critical), kinput is any control rod input, kdispl is the reactivity associated with a displacement of 5.12 Reactor Fast Excursions core material due to pressure buildup, and kother includes the Doppler effect and other nonhydrodynamic reactivity changes. The displacement reactivity is given by  kdispl (t) = ρ(r, t)u(r, t) · ∇w+ (r)dr (5.154) Here ρ is the material density, u(r, t) represents a material displacement from r to r + r, and w + (r, t) is the importance of a unit mass of material at location r to producing subsequent fission events. (The importance function is discussed in Chapter 13.) The displacement is related to the pressure by the hydrodynamic equations ρ ∂ 2 u(r, t) = −∇p(r, t) ∂t 2 (5.155) and   ∂ρ(r, t) ∂u(r, t) + ∇ · ρ(r, t) =0 ∂t ∂t (5.156) An equation of state, represented symbolically as p(r, t) = p(e(r, t), ρ(r, t)) (5.157) relates the pressure to the energy density, e(r, t), and to the density. We neglect changes in density and work done in expansion or compression. Differentiating Eq. (5.154) twice and using Eq. (5.155) yields ∂ 2 kdispl =− ∂t 2  ∇p(r, t) · ∇w+ (r)dr (5.158) The analysis proceeds by postulating that there is no feedback, except the Doppler effect, until the total energy generated in the core reaches a threshold value, E ∗ , at which point the core material begins to vaporize, thereby building up pressure, which causes the core to expand until the negative reactivity associated with expansion eventually terminates the excursion. Rather than carry through the rather involved derivation (see Ref. 9), we summarize the main results for a spherical core. When the energy, E, exceeds the threshold value, it subsequently increases as   ∗ E − E ∗ = E ∗ e(k/)(t−t ) − 1 (5.159) The pressure near the center of the core is proportional to E − E ∗ ≈ E, so that once it becomes large the pressure varies as p ∼ ENe(k/)t (5.160) 189 190 5 Nuclear Reactor Dynamics The pressure gradient that tends to blow the core apart is proportional to p/R. Thus the radial acceleration produced by the pressure gradient goes as R̈ ∼ |∇p| = C1 (k/)t e R (5.161) Integrating this expression twice yields an expression for the instantaneous core radius   C1 2 (k/)t (5.162) e R  (t)  R 1 + (k)2 R 2 The excursion terminates when the expansion increases the negative reactivity sufficiently to offset the initiating reactivity less any negative Doppler or rod input reactivity:   kdispl R  − R = k0 − kother − kinput = k  (5.163) which occurs at time t given by  e(k /)t = (k  )3 R 2 C1 kdispl 2 (5.164) The energy generated up to the time of termination is E∼ (k  )3 R 2 2 (5.165) Numerical calculations indicate that the approximate relationships above represent quite well excursions resulting from large initial reactivity insertions. For modest initiating reactivities, the expression     (k  )3 R 2 2/9 E − 1 ∼ E∗ 2 (5.166) is in better qualitative agreement with numerical results. 5.13 Numerical Methods In practice, numerical methods are used to solve the neutron dynamics equations. The solution is made difficult by the difference in time scales involved. The prompt neutron time scale is on the order of  = 10−4 to 10−5 s for thermal reactors or 10−6 to 10−7 s for fast reactors, while the delayed neutron time scales vary from tenths of seconds to tens of seconds. When ρ is significantly less than β, making the prompt jump approximation removes the prompt neutron time scale from the problem, and straightforward time-differencing schemes are satisfactory. When it References is necessary to retain the prompt neutron dynamics (i.e., for transients near or above prompt critical), the usual numerical methods for solving ordinary differential equations (e.g., Runge–Kutta) are limited by solution stability to extremely small time steps over which there is little change in the neutron population. However, a class of methods for solving stiff sets of ordinary differential equations (sets with very different time constants) have been developed (Refs. 2 and 7) and are now widely used for solution of the neutron dynamics equations. References 1 D. Saphier, “Reactor Dynamics,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations II, CRC Press, Boca Raton, FL (1986). 2 G. Hall and J. M. Watts, Modern Numerical Methods for Ordinary Differential Equations, Clarendon Press, Oxford (1976). 3 J. L. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976), Chap. 6 and pp. 556–565. 4 A. F. Henry, Nuclear-Reactor Analysis, MIT Press, Cambridge, MA (1975), Chap. 7. 5 D. L. Hetrick, ed., Dynamics of Nuclear Systems, University of Arizona Press, Tucson, AZ (1972). 6 A. Z. Akcasu, G. S. Lellouche, and M. L. Shotkin, Mathematical Methods in Nuclear Reactor Dynamics, Academic Press, New York (1971). 7 C. W. Gear, Numerical Initial Value Problems in Ordinary Differential Equations, Prentice Hall, Englewood Cliffs, NJ (1971). 8 G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Wiley (Van Nostrand Reinhold), New York (1970), Chap. 9. 9 H. H. Hummel and D. Okrent, Reactivity Coefficients in Large Fast Power Reactors, American Nuclear Society, La Grange Park, IL (1970). 10 L. E. Weaver, Reactor Dynamics and Control, Elsevier, New York (1968). 11 H. P. Flatt, “Reactor Kinetics Calculations,” in H. Greenspan, C. N. Kelber, and D. Okrent, eds., Computational Methods in Reactor Physics, Gordon and Breach, New York (1968). 12 D. L. Hetrick and L. E. Weaver, eds., Neutron Dynamics and Control, USAEC-CONF-650413, U.S. Atomic Energy Commission, Washington, DC (1966). 13 M. Ash, Nuclear Reactor Kinetics, McGraw-Hill, New York (1965). 14 G. R. Keepin, Physics of Nuclear Kinetics, Addison-Wesley, Reading, MA (1965). 15 A. Radkowsky, ed., Naval Reactors Physics Handbook, U.S. Atomic Energy Commission, Washington, DC (1964), Chap. 5. 16 T. J. Thompson and J. G. Beckerly, eds., The Technology of Nuclear Reactor Safety, MIT Press, Cambridge, MA (1964). 17 L. E. Weaver, ed., Reactor Kinetics and Control, USAEC-TID-7662, U.S. Atomic Energy Commission, Washington, DC (1964). 18 J. A. Thie, Reactor Noise, Rowman & Littlefield, Totowa, NJ (1963). 19 J. Lasalle and S. Lefschetz, Stability by Liapunov’s Direct Methods and Applications, Academic Press, New York (1961). 20 R. V. Meghreblian and D. K. Holmes, Reactor Analysis, McGrawHill, New York (1960), Chap. 9. 191 192 5 Nuclear Reactor Dynamics Problems 5.1. The absorption cross section in a bare, critical thermal reactor is decreased by 0.5% by removing a purely absorbing material. Calculate the associated reactivity. 5.2. A bare metal sphere of essentially pure 235 U is assembled, and the output of a neutron detector is observed, after an initial transient, to be increasing exponentially with a period T = 1 s. The neutron effectiveness values for the six delayed neutron groups are calculated to be γi = 1.10, 1.03, 1.05, 1.03, 1.01, and 1.01. What is the effective multiplication constant, k, for the assembly? 5.3 Using the one-delayed precursor group approximation, prompt-jump approximation, and the reactor parameters β = 0.0075, λ = 0.08 s−1 ,  = 6 × 10−5 s, solve for the time dependence of the neutron population over the interval 0 < t < 10 s following the introduction of a ramp reactivity ρ(t) = 0.1βt into a critical reactor for 0 < t < 5 s. Such a reactivity insertion might result from partial withdrawal of a control rod bank. 5.4. A pulsed neutron measurement was performed in an assembly with β = 0.0075 and  = 6 × 10−5 . An exponential prompt neutron decay constant α0 = −100 s−1 was measured. What are the reactivity and effective multiplication constant of the assembly? 5.5. A control rod was partially withdrawn from a critical nuclear reactor for 5 s, then reinserted to bring the reactor back to critical. The reactivity worth of the partial rod withdrawal was ρ = 0.0025. Use the prompt-jump approximation and a one delayed neutron group approximation to calculate the neutron and precursor populations, relative to the initial critical populations, for times 0 < t < 10 s. Use the neutron kinetic parameters β = 0.0075, λ = 0.08 s−1 , and  = 6 × 10−5 s. 5.6. A control rod bank is scrammed in an initially critical reactor. The signal of a neutron detector drops instantaneously to one-third of its prescram level, then decays exponentially. Assume one group of delayed neutrons with β = 0.0075 and λ = 0.08 s, and use  = 10−4 s for the reactor lifetime. What is the reactivity worth of the control rod bank? How long is needed for the power level to reach 1% of the initial prescram level? 5.7. Plot the real and imaginary parts of the zero-power transfer function versus ω (s = iω) for a 235 U reactor using a one Problems 5.8 5.9. 5.10. 5.11. 5.12. 5.13. 5.14. 5.15. 5.16. delayed neutron group model with β = 0.0075, λ = 0.08 s−1 , and  = 6 × 10−5 s. Calculate the Doppler reactivity temperature coefficient for a UO2 -fueled, H2 O-cooled thermal reactor with long fuel rods 1 cm in diameter operating with a fuel temperature of 450 K. The moderator macroscopic scattering cross section per atom of 238 U is 100. Take the resonance integral at 300 K as I = 10 barns. Derive an expression for the calculation of a void temperature coefficient of reactivity for a pressurized water reactor (i.e., the temperature coefficient associated with a small fraction of the moderator being replaced with void). Repeat the calculation for when the water contains 1000 ppm 10 B as a “chemical shim.” Calculate the nonleakage reactivity temperature coefficient for a bare cylindrical graphite reactor with height-to-diameter ratio H /D = 1.0, k∞ = 1.10, migration area M 2 = 400 cm2 , and moderator linear expansion coefficient θM = 1 × 10−5 ◦ C−1 . Calculate the reactivity defect in a PWR with fuel and moderator temperature coefficients of αF = −1.0 × 10−5 k/k/◦ F and αM = −2.0 × 10−4 k/k/◦ F when the reactor goes from hot zero power (TF = TM = 530◦ F) to hot full power (TF = 1200◦ F and TM = 572◦ F). A critical reactor is operating at steady state when there is a step reactivity insertion ρ = k/k = 0.0025. Use one group of delayed neutrons, the parameters β = 0.0075, λ = 0.08 s−1 , and  = 6 × 10−5 s, and a temperature coefficient of reactivity αT = −2.5 × 10−4 ◦ C−1 . Assume that the heat removal is proportional to the temperature. Write the coupled set of equations that describe the dynamics of the prompt and delayed neutrons and the temperature. Linearize and solve these equations (e.g., by Laplace transform). Calculate the power threshold for linear stability (in units of P0 XF /β) from Eq. (5.111) for XF /XM = −0.25 and −0.50 and for ωM = 0.1, 0.25, and 0.5. Analyze the linear stability of a one-temperature model for a nuclear reactor in which the heat is removed by conduction −1 and in which there is an overall with time constant ωR negative steady-state power coefficient, XR < 0. Is the reactor stable at all power levels? Repeat problem 5.14 for convective heat removal. Calculate and plot the power burst described by Eq. (5.143) for a fast reactor with generation time  = 1 × 10−6 s and negative energy feedback coefficient αE = −0.5 × 10−6 193 194 5 Nuclear Reactor Dynamics 5.17. 5.18. 5.19. 5.20.∗ 5.21. 5.22. * k/k/MJ into which a step reactivity insertion of k0 = +0.02 takes place at t = 0. Use P0 = 100 MW. A control rod is partially withdrawn (assume instantaneously) from a 235 U-fueled nuclear reactor that is critical and at low power at room temperature. The signal measured by a neutron detector is observed to increase immediately to 125% of its value prior to rod withdrawal, and then to increase approximately exponentially. What is the reactivity worth of the control rod? What is the value of the exponent that governs the long-time exponential increase of the signal measured by the neutron detector? In a cold critical PWR fueled with 4% enriched UO2 , the control rod bank is withdrawn a fraction of a centimeter, introducing a positive reactivity of ρ = 0.0005. The neutron flux begins to increase, increasing the fission rate. Discuss the feedback reactivity effects that occur as a result of the increasing fission heating. Use the temperature coefficients of reactivity given in Table 5.3 to calculate the change in reactivity when the core temperature in an oxide-fueled fast reactor increases from 300◦ C to 500◦ C. Assume uniform temperatures in fuel, coolant, and structure. Repeat the calculation for a fuel temperature increase to 800◦ C and a coolant and structure temperature increase to 350◦ C. Solve Eqs. (5.133) and (5.134) to calculate the response of the neutron population in a UO2 -fueled PWR to step rod withdrawal with reactivity worth ρ = 0.002, taking into account a negative fuel Doppler feedback coefficient of −2 × 10−6 k/k/K. The reactor has neutronics properties (β = 0.0065, λ = 0.08 s−1 ,  = 1.0 × 10−4 ), fission heat deposition in the fuel vnf Ef = 250 W/cm3 , and fuel properties ρ = 10.0 g/cm3 and Cp = 220 J/kg. (Hint: It is probably easiest to do this numerically.) Evaluate the resonance escape probability moderator temperature coefficient of reactivity of Eq. (5.87) for a UO2 reactor consisting of assemblies of 1-cm-diameter fuel pins of height H in a water lattice with p /NM = 100 and fuel density ρ = 10 g/cm3 . Use θM = 1 × 10−4 /K for the linear coefficient of expansion for water. Derive an explicit expression for the thermal utilization temperature coefficient of reactivity of Eq. (5.89) by using Eqs. (3.90) and (3.92) to evaluate the ∂aF /∂ξ and ∂ξ/∂TF Problem 5.20 is a longer problem suitable for a take-home project. Problems terms and equivalent relations to evaluate the daM /dξ and dξ/dTM terms. 5.23. In a “rod-drop” experiment, a control rod is dropped into a cold, critical reactor. The neutron flux is observed to immediately drop to one-half of its value prior to the rod drop and then decay slowly. Using a one-delayed-group model with delayed neutron fraction β = 0.0065 and decay constant λ = 0.08/s, determine the reactivity worth of the control rod. 195 197 6 Fuel Burnup The long-term changes in the properties of a nuclear reactor over its lifetime are determined by the changes in composition due to fuel burnup and the manner in which these are compensated. The economics of nuclear power is strongly affected by the efficiency of fuel utilization to produce power, which in turn is affected by these long-term changes associated with fuel burnup. In this chapter we describe the changes in fuel composition that take place in an operating reactor and their effects on the reactor, the effects of the samarium and xenon fission products with large thermal neutron cross sections, the conversion of fertile material to fissionable material by neutron transmutation, the effects of using plutonium from spent fuel and from weapons surplus as fuel, the production of radioactive waste, the extraction of the residual energy from spent fuel, and the destruction of long-lived actinides. 6.1 Changes in Fuel Composition The initial composition of a fuel element will depend on the source of the fuel. For reactors operating on the uranium cycle, fuel developed directly from natural uranium will contain a mixture of 234 U, 235 U, and 238 U, with the fissile 235 U content varying from 0.72% (for natural uranium) to more than 90%, depending on the enrichment. Recycled fuel from reprocessing plants will also contain the various isotopes produced in the transmutation–decay process of uranium. Reactors operating on the thorium cycle will contain 232 Th and 233 U or 235 U, and if the fuel is from a reprocessing plant, isotopes produced in the transmutation–decay process of thorium. During the operation of a nuclear reactor a number of changes occur in the composition of the fuel. The various fuel nuclei are transmuted by neutron capture and subsequent decay. For a uranium-fueled reactor, this process produces a variety of transuranic elements in the actinide series of the periodic table. For a thorium-fueled reactor, a number of uranium isotopes are produced. The fission event destroys a fissile nucleus, of course, and in the process produces two intermediate mass fission products. The fission products tend to be neutron-rich and 198 6 Fuel Burnup Fig. 6.1 Transmutation–decay chains for 238 U and 232 Th. (From Ref. 3; used with permission of Taylor & Francis.) subsequently decay by beta or neutron emission (usually accompanied by gamma emission) and undergo neutron capture to be transmuted into a heavier isotope, which itself undergoes radioactive decay and neutron transmutation, and so on. The fissile nuclei also undergo neutron transmutation via radiative capture followed by decay or further transmutation. Fuel Transmutation–Decay Chains Uranium-235, present 0.72% in natural uranium, is the only naturally occurring isotope that is fissionable by thermal neutrons. However, three other fissile (fissionable by thermal neutrons) isotopes of major interest as nuclear reactor fuel are produced as the result of transmutation–decay chains. Isotopes that can be converted to fissile isotopes by neutron transmutation and decay are known as fertile isotopes. 239 Pu and 241 Pu are products of the transmutation–decay chain beginning with the fertile isotope 238 U, and 233 U is a product of the transmutation–decay 6.1 Changes in Fuel Composition chain beginning with the fertile isotope 232 Th. These two transmutation–decay chains are shown in Fig. 6.1. Isotopes are in rows with horizontal arrows representing (n, γ ) transmutation reactions, with the value of the cross section (in barns) shown. Downward arrows indicate β-decay, with the half-lives shown. Thermal neutron fission is represented by a dashed diagonal arrow, and the thermal cross section is shown. (Fast fission also occurs but is relatively less important in thermal reactors.) Natural abundances, decay half-lifes, modes of decay, decay energies, spontaneous fission yields, thermal capture and fission cross sections averaged over a Maxwellian distribution with kT = 0.0253 eV (σ th ), infinite-dilution capture and fission resonance integrals (RIs), and capture and fission cross sections averaged over the fission spectrum (σ χ ) are given in Table 6.1. Fuel Depletion–Transmutation–Decay Equations Concentrations of the various fuel isotopes in a reactor are described by a coupled set of production–destruction equations. We adopt the two-digit superscript convention for identifying isotopes in which the first digit is the last digit in the atomic number and the second digit is the last digit in the atomic mass. We represent the neutron reaction rate by σxnm ϕnnm , although the actual calculation may involve a sum over energy groups of such terms. For reactors operating on the uranium cycle, the isotopic concentrations are described by ∂n24 = −σa24 φn24 ∂t ∂n25 = σγ24 φn24 − σa25 φn25 ∂t ∂n26 36 = σγ25 φn25 − σa26 φn26 + λ36 ec n ∂t ∂n27 28 = σγ26 φn26 + σn,2n φn28 − λ27 n27 ∂t ∂n28 = −σa28 φn28 ∂t   ∂n29 = σγ28 φn28 − λ29 + σa29 φ n29 ∂t   ∂n36 37 φn37 − λ36 + σa36 φ n36 = σn,2n ∂t ∂n37 = λ27 n27 − σa27 φn37 ∂t   ∂n38 = σγ37 φn37 − λ38 + σa38 φ n38 ∂t   ∂n39 = λ29 n29 − λ39 + σa39 φ n39 ∂t (6.1) 199 200 6 Fuel Burnup ∂n48 = λ38 n38 − σa48 φn48 ∂t ∂n49 = λ39 n39 − σa49 φn49 + σγ48 φn48 ∂t ∂n40 = σγ49 φn49 − σa40 φn40 + σγ29 φn29 + σγ39 φn39 ∂t   ∂n41 = σγ40 φn40 − λ41 + σa41 φ n41 ∂t ∂n42 = σγ41 φn41 − σa42 φn42 ∂t   ∂n43 = σγ42 φn42 − λ43 + σa43 φ n43 ∂t   ∂n51 = λ41 n41 − λ51 + σa51 φ n51 ∂t ∂n52 = σγ51 φn51 − σa52 φn52 ∂t ∂n53 = λ43 n43 − σa53 φn53 + σγ52 φn52 ∂t With respect to Fig. 6.1, a few approximations have been made in writing Eqs. (6.1). The neutron capture in 239 U to produce 240 U followed by the decay (t1/2 = 14 h) into 240 Np and the subsequent decay (t1/2 = 7 min) into 240 Pu is treated as the direct production of 240 Pu by neutron capture in 239 U, and the production of 240 Np by neutron capture in 239 Np followed by the subsequent decay (t1/2 = 7 min) of 240 Np into 240 Pu is treated as the direct production of 240 Pu by neutron capture in 239 Np. These approximations have the beneficial effect for numerical solution techniques of removing short time scales from the set of equations, without sacrificing information of interest on the longer time scale of fuel burnup. For reactors operating on the thorium cycle, the isotopic concentrations are described by ∂n02 = −σa02 φn02 ∂t   ∂n03 = σγ02 φn02 − λ03 + σa03 φ n03 ∂t   ∂n13 = λ03 n03 − λ13 + σa13 φ n13 ∂t   ∂n22 = − λ22 + σa22 φ n22 ∂t ∂n23 = σγ22 φn22 + λ13 n13 − σa25 φn23 ∂t ∂n24 = σγ23 φn23 + σγ13 φn13 − σa24 φn24 ∂t (6.2) 238 Np 237 Np 236 Np 240 U 239 U 238 U 237 U 236 U 235 U 234 U 233 U 232 U 234 Pa 233 Pa 234 Th 233 Th 232 Th Isotope – – 100 – – – – – – 0.0057 0.719 – – 99.27 – – – Abundance (%) 2.14 × 106 y 2.12 d 1.41 × 1010 y 22.3 m 24.1 d 27.0 d 6.7 h 68.9 y 1.59 × 105 y 2.46 × 105 y 7.04 × 108 y 2.34 × 106 y 6.75 d 4.47 × 109 y 23.5 m 14.1 h 1.54 × 105 y t1/2 α β β β β α α α α α β α β β ec∗ β∗ α β Decay Mode Table 6.1 Cross Section and Decay Data for Fuel Isotopes 4.1 1.2 0.27 0.57 2.2 5.4 4.9 4.9 4.7 4.6 0.52 4.3 1.3 0.39 0.94 0.49 5.0 1.3 Energy (MeV) σ th γ (barns) 7 1285 2 35 – 64 41 88 87 5 392 2 – – 621 – 144 399 Spontaneous Fission Yield (%) <1 × 10−9 – – – – – <6 × 10−9 1.7 × 10−9 7.0 × 10−9 9.6 × 10−8 – 5 × 10−5 – – – – <2 × 10−10 – – 13 – – – 66 469 6 507 54 1 10 − − 2453 − 20 1835 σ fth (barns) 84 643 94 864 – 173 138 631 133 346 1084 278 – – 259 – 661 201 RIγ (barns) – 11 – – – 364 774 7 278 8 49 2 − − 1032 – 7 940 RIf (barns) 0.09 0.09 0.11 0.28 – 0.03 0.07 0.22 0.09 0.11 0.93 0.07 – – 0.19 – 0.17 0.11 χ σγ (barns) 0.08 0.11 0.04 0.33 – 2.01 1.95 1.22 1.24 0.59 0.74 0.31 – – 1.92 – 1.33 1.42 (Continued) χ σf (barns) 6.1 Changes in Fuel Composition 201 – – – – – – – – – – – 2.86 y 45 87.7 y 2.41 × 104 y 6.56 × 103 y 14.4 y 3.73 × 105 y 432 y 236 d t 1/2 β – α ec∗ α α α β α α Decay Mode 0.72 – 5.9 0.22 5.6 5.2 5.3 0.02 5.0 5.6 Energy (MeV) Source: Brookhaven National Laboratory Nuclear Data Center, http:/ /www.dne.bnl.gov/CoN/index.html. * 87.3% electron capture, 12.5% β. 241 Am 242 Pu 241 Pu 240 Pu 239 Pu 238 Pu 237 Pu 236 Pu 240 Np 239 Np Isotope Abundance (%) Table 6.1 (Continued) – – 1.4 × 10−7 – 1.9 × 10−7 3 × 10−10 5.7 × 10−6 <2 × 10−14 >5.5 × 10−4 4 × 10−10 Spontaneous Fission Yield (%) 33 – 126 – 458 274 264 326 17 532 σ th γ (barns) – – 146 − 15 698 53 938 – 3 σ fth (barns) 445 – 401 – 154 182 8103 180 1130 1305 RIγ (barns) – – 59 − 33 303 9 576 – 14 RIf (barns) 0.19 – 0.15 – 0.10 0.05 0.10 0.12 0.09 0.23 χ σγ (barns) 1.46 – 2.08 – 1.99 1.80 1.36 1.65 1.13 1.38 χ σf (barns) 202 6 Fuel Burnup 6.1 Changes in Fuel Composition ∂n25 = σγ24 φn24 − σa25 φn25 ∂t ∂n26 = σγ25 φn25 − σa26 φn26 ∂t   ∂n27 = σγ26 φn26 − λ27 + σa27 φ n27 ∂t ∂n37 = λ27 n27 − σa37 φn37 ∂t Another short-time-scale elimination approximation that neutron capture in 233 Pa leads directly to 234 U has been made. Example 6.1: Depletion of a Pure 235 U-Fueled Reactor. As an example of the nature of the solution of the equations above, consider the hypothetical case of a reactor initially fueled with pure 235 U which operates for 1 year with a constant neutron flux of 1014 n/cm2 ·s. The solution of the second of Eqs. (6.1) is n25 (t) = n25 (0) exp(−σa25 φt ), where at the end of 1 year, σa25 φt = (594 × 10−24 cm2 ) × (1 × 1014 /cm2 ·s) × (3.15 × 107 s) = 1.87 and n25 (t) = 0.154n25 (0). The number of atoms that have fissioned in this 1 year is (n(1) − n(0)) × [σf /(σf + σγ )] = [0.846n25 (0)](507/594) = 0.722n25 (0). Each fission event releases 192.9 MeV of recoverable energy, so the total recoverable fission energy release is [0.722n25 (0) fissions] × (192.9 MeV/fission) × (1.6 × 10−19 MJ/MeV) = 2.23 × 10−17 × n25 (0) MJ. If the initial core loading is 100 kg of 235 U, this corresponds to (2.23 × 10−17 ) × (105 /235) × (6.02 × 1023 ) = 0.95 × 109 MJ = 1.1 × 104 MWd of recoverable fission energy. Neglecting the production of 236 U by electron capture decay of 236 Np, the solution for n25 (t) can be used to solve the third of Eqs. (6.1) to obtain n26 (t) = [n25 (0)σγ25 /(σa25 − σa26 )][exp(−σa26 φt) − exp(−σa25 φt)]. This expression for n26 (t) can be used in the fourth of Eqs. (6.1) to obtain a similar, but more complicated solution for n27 (t), since we have assumed that n28 = 0; and so on. Fission Products The fission event usually produces two intermediate mass nuclei, in addition to releasing two or three neutrons. Interestingly, the fission product masses are not usually equal to about half the mass of the fissioning species, but are distributed in mass with peaks at about 100 and 140 amu, as shown in Fig. 6.2. The isotopes produced by fission tend to be neutron-rich and undergo radioactive decay. They also undergo neutron capture, with cross sections ranging from a few tenths of a barn to millions of barns. The general production–destruction equation satisfied by a fission product species j is    dnj j  = γj f φ + (6.3) λi→j + σ i→j φ ni − λj + σa φ nj dt i where γj is the fraction of fission events that produces a fission product species j , λi→j is the decay rate of isotope i to produce isotope j (beta, alpha, neutron, etc., 203 204 6 Fuel Burnup Fig. 6.2 Fission yields for 235 U and 239 Pu. (From Ref. 15.) decay) and σ i→j is the transmutation cross section for the production of isotope j by neutron capture in isotope j . Even though the fission products undergo transmutation and decay, the total inventory of direct fission products plus their progeny increases in time as dnfp  dnj  = = γj f φ dt dt j (6.4) j Solution of the Depletion Equations The equations above can be integrated to determine composition changes over the lifetime of the reactor core loading if the time dependence of the flux is known. However, the flux distribution depends on the composition. In practice, a neutron flux distribution is calculated for the beginning-of-cycle composition and critical control rod position or soluble boron concentration (PWR), and this flux distribution is used to integrate the composition equations above over a depletion-time step, tburn . Then the new critical control rod position or soluble boron concentration is determined (by trial and error) and the flux distribution is recalculated for use in integrating the production–destruction equations over the next depletion time step, and so on, until the end of cycle is reached. The maximum value of tburn depends on how fast the composition is changing and the effect of that composition change on the neutron flux distribution and on the accuracy of the numerical integration scheme. Excluding, for the moment, the relatively short time scale phenomena associated with the xenon and samarium fission products, the time scale of significant composition and flux changes is typically several hundred hours or more. 6.1 Changes in Fuel Composition The typical process of advancing the depletion solution from time ti , at which the composition is known, to time ti+1 is: (1) determine the multigroup constants appropriate for the composition at ti , (2) determine the critical control rod positions or soluble poison concentration by solving the multigroup diffusion equations for the flux at ti (adjusting the control rod positions or boron concentration until the reactor is critical), and (3) integrate the various fuel and fission product production–destruction equations from ti to ti+1 . (The neutron flux solution could be made with a multigroup transport calculation or with multigroup or continuousenergy Monte Carlo calculation, and the preparation of cross sections could involve infinite media spectra and unit cell homogenization calculations or could be based on fitted, precomputed constants.) The integration of the production–destruction equations can be for a large number of points, using the neutron flux at each point; for each fuel pin, using the average flux in the fuel pin; for each fuel assembly, using the average flux over the fuel assembly; and so on. Assuming that the flux is constant in the interval ti < t < ti+1 , the production– destruction equations can be written in matrix notation as N (t) dN N (t) + F (φ(ti )), = A (φ(ti ))N dt ti ≤ t ≤ ti+1 (6.5) The general solution to these equations is of the form A(ti )t]N N (ti ) + A −1 (ti ){exp[A A(ti )t] − 1}F F (ti ) N (ti+1 ) = exp[A (6.6) In general, the accuracy of the solution depends on tburn being chosen so that (λi + σai ϕ)tburn  1 for all of the isotopes involved. For this reason, it is economical to reformulate the physical production–destruction equations to eliminate short-time-scale phenomena that do not act the overall result, as discussed previously. There exist a number of computer codes that solve the production– destruction equations for input neutron fluxes (e.g., Ref. 7). Measure of Fuel Burnup The most commonly used measure of fuel burnup is the fission energy release per unit mass of fuel. The fission energy release in megawatt-days divided by the total mass (in units of 1000 kg or 1 tonne) of fuel nuclei (fissile plus fertile) in the initial loading is referred to as megawatt-days per tonne (MWd/T). An equivalent unit is MWd/kg—10−3 MWd/T. For example, a reactor with 100,000 kg of fuel operating at 3000 MW power level for 1000 days would have a burnup of 30,000 MWd/T. For LWRs the typical fuel burnup is 30,000 to 50,000 MWd/T. Fuel burnup in fast reactors is projected up to be about 100,000 to 150,000 MWd/T. Fuel Composition Changes with Burnup The original fissionable isotope (e.g., 235 U) naturally decreases as the reactor operates. However, the neutron transmutation of the fertile isotope (e.g., 238 U) produces 205 206 6 Fuel Burnup Fig. 6.3 Buildup of Pu isotopes in 4 wt% enriched UO2 in an LWR. (From Ref. 1; used with permission of Nuclear Energy Agency, Paris.) the fissionable isotope 239 Pu, which in turn is transmuted by neutron capture into 240 Pu and higher actinide isotopes. The buildup of the various Pu isotopes as a function of fuel burnup for a typical LWR is shown in Fig. 6.3. Compositions of spent fuel discharged from representative LWR and LMFBR designs are given in Table 6.2. The units are densities (cgs units) times 10−24 , which allows construction of macroscopic cross section upon multiplication by the microscopic cross section in barns. The composition for the average enrichment and burnup of PWR spent fuel is shown in the first column for fuel discharged before 1995 and in the second column for fuel discharged after 1995. Reactivity Effects of Fuel Composition Changes There are a variety of reactivity effects associated with the change in fuel composition. The fission of fuel nuclei produces two negative reactivity effects; the number 6.1 Changes in Fuel Composition Table 6.2 Heavy Metal Composition of Spent UO2 Fuel at Discharge∗ Reactor Type LWR LWR LMFBR LMFBR Initial enrichment (wt %) Power (MW/MTU) Burnup (GWd/T) Actinides (1 × 1024 cm−3 ) 234 U 235 U 236 U 237 U 238 U 237 Np 239 Np 238 Pu 239 Pu 240 Pu 241 Pu 242 Pu 241 Am 243 Am 242 Cm 244 Cm 3.13 21.90 32 4.11 27.99 46 20 54.76 100 20 54.76 150 3.92 × 10−6 1.92 × 10−4 8.73 × 10−5 † 4.51 × 10−6 1.72 × 10−4 1.23 × 10−4 2.48 × 10−7 2.08 × 10−2 1.64 × 10−5 1.55 × 10−6 6.56 × 10−6 1.23 × 10−4 4.28 × 10−5 4.07 × 10−5 1.69 × 10−5 1.62 × 10−6 4.46 × 10−6 5.66 × 10−7 1.39 × 10−6 3.37 × 10−5 2.17 × 10−3 4.58 × 10−4 5.71 × 10−7 1.63 × 10−2 5.11 × 10−5 2.93 × 10−6 3.84 × 10−6 1.04 × 10−3 7.83 × 10−5 2.60 × 10−6 † 2.88 × 10−5 1.37 × 10−3 5.62 × 10−4 7.89 × 10−7 1.53 × 10−2 1.01 × 10−4 3.16 × 10−6 1.20 × 10−5 1.36 × 10−3 1.71 × 10−4 8.37 × 10−6 4.70 × 10−7 6.87 × 10−7 † † † 2.12 × 10−2 1.01 × 10−5 1.25 × 10−6 3.36 × 10−6 1.23 × 10−4 4.05 × 10−5 3.44 × 10−5 1.05 × 10−5 1.45 × 10−6 2.12 × 10−6 3.71 × 10−7 4.81 × 10−7 1.50 × 10−7 † † † * Calculated with ORIGEN (Ref. 7). † < 0.001%. of fuel nuclei is reduced and fission products are created, many of which have large neutron capture cross sections. The transmutation–decay chain of fertile fuel nuclei of a given species produces a sequence of actinides (uranium-fueled reactor) or uranium isotopes (thorium-fueled reactor), some of which are fissile. The transmutation of one fertile isotope into another nonfissile isotope can have a positive or negative reactivity effect, depending on the cross sections for the isotopes involved, but the transmutation of a fertile isotope into a fissile isotope has a positive reactivity effect. Depending on the initial enrichment, the transmutation–decay process generally produces more fissile nuclei than are destroyed early in the cycle, causing a positive reactivity effect, until the concentration of transmuted fissile nuclei comes into equilibrium. The buildup of 239 Pu early in life of a uranium-fueled reactor produces a large positive reactivity effect which may be greater than the negative reactivity effect of 235 U depletion and fission product buildup. For thermal reactors, η49 < η25 , so the buildup of 239 Pu must exceed the burnup of 235 U in order for a positive reactivity effect. For fast reactors, η49 > η25 for neutron energies in excess of about 10 keV, and there can be an initial positive reactivity effect even if the decrease in 235 U is greater than the buildup of 239 Pu. However, the 239 Pu concentration will saturate 207 208 6 Fuel Burnup at a value determined by the balance between the 238 U transmutation rate and the 239 Pu depletion rate, at which point the continued depletion of 235 U and buildup of fission products produce a negative reactivity effect that accrues over the lifetime of the fuel in the reactor. Compensating for Fuel-Depletion Reactivity Effects The reactivity effects of fuel depletion must be compensated to maintain criticality over the fuel burnup cycle. The major compensating elements are the control rods, which can be inserted to compensate positive depletion reactivity effects and withdrawn to compensate negative depletion reactivity effects. Adjustment of the concentration of a neutron absorber (e.g., boron in the form of boric acid) in the water coolant is another means used to compensate for fuel-depletion reactivity effects. Soluble poisons are used to compensate fuel-depletion reactivity in PWRs but not in BWRs, because of the possibility that they will plate out on boiling surfaces. Since a soluble poison introduces a positive coolant temperature reactivity coefficient because an increase in temperature decreases the density of the soluble neutron absorber, the maximum concentration (hence the amount of fuel depletion reactivity that can be compensated) is limited. Burnable poisons (e.g., boron, erbium, or gadolinium elements located in the fuel lattice), which themselves deplete over time, can be used to compensate the negative reactivity effects of fuel depletion. The concentration of burnable poison can be described by dnbp = −fbp nbp σbp φ dt (6.7) where fbp is the self-shielding of the poison element (i.e., the ratio of the neutron flux in the poison element to the neutron flux in the adjacent fuel assembly). The poison concentration is chosen so that the spatial self-shielding of the poison elements is large enough (fbp  1) early in the burnup cycle to shield the poison from neutron capture, and the neutron capture rate remains constant in time. After a certain time the concentration of the poison nuclei is sufficiently reduced that fbp increases and the poison burns out, resulting in an increasing reactivity. If the poison starts to burn out at about the same time that the overall fuel depletion reactivity effect starts to become progressively more negative (i.e., when the 239 Pu concentration saturates), the burnout of the poison will at least partially compensate the fuel-depletion reactivity decrease. Reactivity Penalty The buildup of actinides in the 238 U transmutation–decay process introduces a fuel reactivity penalty because some of actinides act primarily as parasitic absorbers. While 239 Pu and 241 Pu are fissionable in a thermal reactor, and 240 Pu transmutes into 241 Pu, 242 Pu transmutes into 243 Pu with a rather small cross section, and 243 Pu 6.1 Changes in Fuel Composition Fig. 6.4 235 U neutron transmutation–decay chain. (From Ref. 4; used with permission of American Nuclear Society.) has a rather small fission cross section, so that 242 Pu is effectively a parasitic absorber that builds up in time. The 243 Am also accumulates and acts primarily as a parasitic absorber. Whereas the 243 Am, which is produced by the decay of 243 Pu, can be separated readily, it is difficult to separate the different plutonium isotopes from each other, so the negative 242 Pu reactivity effect is exacerbated if the plutonium is recycled with uranium. A similar problem arises with the 236 U produced by radiative capture in 235 U, as shown in Fig. 6.4, which is difficult to separate from 235 U, and with 237 Np, which is produced by transmutation of 236 U into 237 U followed by beta decay. The 237 Np can be separated readily, however, and does not need to accumulate in recycled fuel. End-of-cycle reactivity penalties calculated for the recycle of BWR fuel are shown in Table 6.3 after one, two, and three cycles. It was assumed that the 237 Np and 243 Am were removed between cycles, but there was a cycle-to-cycle increase in the 237 Np and 243 Am reactivity penalties due to the accumulation of 236 U and 242 Pu, respectively. Effects of Fuel Depletion on the Power Distribution Fuel depletion and the compensating control actions affect the reactor power distribution over the lifetime of the fuel in the core. Depletion of fuel will be greatest Table 6.3 Reactivity Penalties with Recycled BWR Fuel (% k/k) End of Cycle: 236 U 237 Np 242 Pu 243 Am 1 2 3 0.62 0.90 1.12 0.13 0.59 0.73 0.65 1.53 2.04 0.36 0.57 0.89 Source: Data from Ref. 16. 209 210 6 Fuel Burnup where the power is greatest. The initial positive reactivity effect of depletion will then enhance the power peaking. At later times, the negative reactivity effects will cause the power to shift away to regions with higher kinfinity . Any strong tendency of the power distribution to peak as a result of fuel depletion must be compensated by control rod movement. However, the control rod movement to offset fuel depletion reactivity effects itself produces power peaking; the presence of the rods shields the nearby fuel from depletion and when the rods are withdrawn, the higher local k∞ causes power peaking. Similarly, burnable poisons shield the nearby fuel, producing local regions of higher k∞ and power peaking when they burn out. Determination of the proper fuel concentration zoning and distribution of burnable poisons and of the proper control rod motion to compensate fuel depletion reactivity effects without unduly large power peaking is a major nuclear analysis task. In-Core Fuel Management At any given time, the fuel in a reactor core will consist of several batches that have been in the core for different lengths of time. The choice of the number of batches is made on the basis of a trade-off between maximizing fuel burnup and minimizing the number of shutdowns for refueling, which reduces the plant capacity factor. At each refueling, the batch of fuel with the highest burnup is discharged, the batches with lower burnup may be moved to different locations, and a fresh or partially depleted batch is added to replace the discharged batch. The analysis leading to determination of the distribution of the fuel batches within the core to meet the safety, power distribution and burnup, or cycle length constraints for fuel burn cycle is known as fuel management analysis. Although fuel management may be planned in advance, it must be updated online to adjust to higher or lower capacity factors than planned (which result in lower or higher reactivity than planned at the planned refueling time) and unforeseen outages (which result in higher reactivity than planned at the planned refueling time). Typically, a PWR will have three fuel batches, and a BWR will have four fuel batches in the core at any given time and will refuel every 12 to 18 months. A number of different loading patterns have been considered, with the general conclusion that more energy is extracted from the fuel when the power distribution in the core is as flat as possible. In the in–out loading pattern, the reactor is divided into concentric annular regions loaded with different fuel batches. The fresh fuel batch is placed at the periphery, the highest burnup batch is placed at the center, and intermediate burnup batches are placed in between to counter the natural tendency of power to peak in the center of the core. At refueling, the central batch is discharged, the other batches are shifted inward, and a fresh batch is loaded on the periphery. The in–out loading pattern has been found to go too far in the sense that the power distribution is depressed in the center and peaked at the periphery. An additional difficulty is the production of a large number of fast neutrons at the periphery that leak from the core and damage the pressure vessel. In the scatter loading pattern the reactor core is divided into many small regions of four to six assemblies from different batches. At refueling, the assemblies within 6.2 Samarium and Xenon each region with the highest burnup are discharged and replaced by fresh fuel assemblies. This loading pattern has been found to produce a more uniform power distribution and to result in less fast neutron leakage than the in–out pattern. Since the pressure vessel damage by fast neutrons became recognized as a significant problem, a number of different loading patterns have been developed with the specific objective of minimizing neutron damage to the pressure vessel. These include placement of only partially depleted assemblies at the core periphery, placement of highly depleted assemblies near welds and other critical locations, using burnable poisons in peripheral assemblies, replacing peripheral fuel assemblies with dummy assemblies, and others. Better utilization of resources argues for the highest possible fuel burnup consistent with materials damage limitations, and a new higher enrichment fuel has been developed that can achieve burnups of up to 50,000 MWd/T in LWRs. The higher fuel burnup produces more actinides and fission products with large thermal neutron cross sections, which compete more effectively with control rods for thermal neutrons and reduces control rod worth, and which produces larger coolant temperature reactivity coefficients. The higher-enrichment higher-burnup fuel also provides the possibility of longer refueling cycles, which improves plant capacity factor and reduces power costs. 6.2 Samarium and Xenon The short-term time dependence of two fission product progeny, 149 Sm and 135 Xe, which have very large absorption cross sections, introduces some interesting reactivity transients when the reactor power level is changed. Samarium Poisoning Samarium-149 is produced by the beta decay of the fission product 149 Nd, as described in Fig. 6.5. It has a thermal neutron absorption cross section of 4 × 104 barns and a large epithermal absorption resonance. The 1.7-h half-life of 149 Nd is sufficiently short that 149 Pm can be assumed to be formed directly from fission in writing the production–destruction equations for 149 Sm: dP = γ Nd f φ − λP P dt dS = λP P − σaS φS dt where P and S refer to 149 Pm and solution, for constant φ, (6.8) 149 Sm, respectively. These equations have the 211 212 6 Fuel Burnup Fig. 6.5 Characteristics of 149 Sm under representative LWR conditions: (a) transmutation–decay chain; (b) fission yields; (c) time dependence. (From Ref. 3; used with permission of Taylor & Francis/Hemisphere Publishing.) P (t) = γ Nd f φ  P  P 1 − e−λ t + P (0)e−λ t P λ S(t) = S(0)e−σa φt + S −  γ Nd f  S 1 − e−σa φt S σa (6.9) γ Nd f φ − λP P (0)  −σ S φt P  e a − e−λ t P S λ − σa φ At the beginning of life in a fresh core, P (0) = S(0) = 0, and the promethium and samarium concentrations build up to equilibrium values: 6.2 Samarium and Xenon Peq = γ Nd f φ , λP Seq = γ Nd f σaS (6.10) The equilibrium value of 149 Pm depends on the neutron flux level. However, the equilibrium value of 149 Sm is determined by a balance between the fission production rate of 149 Pm and the neutron transmutation rate of 149 Sm, both of which are proportional to the neutron flux, and consequently, does not depend on the neutron flux level. The time required for the achievement of equilibrium concentrations depends on φ, σaS and λP . For typical thermal reactor flux levels (e.g., 5 × 1013 n/cm2 · s), equilibrium levels are achieved in a few hundred hours. When a reactor is shut down after running sufficiently long to build up equilibrium concentrations, the solutions of Eqs. (6.9) with P (0) = Peq , S(0) = Seq , and φ = 0 are P (t) = Peq e−λ Pt  P  S(t) = Seq + Peq 1 − e−λ t → Seq + Peq (6.11) indicating that the l49 Sm concentration will increase to Seq + Peq as the 149 Pm decays into 149 Sm with time constant 1/λP = 78 h. If the reactor is restarted, the 149 Sm burns out until the 149 Pm builds up; then the l49 Sm returns to its equilibrium value. This time dependence of the samarium concentration is illustrated in Fig. 6.5. The perturbation theory estimate for the reactivity worth of 149 Sm is (t) ρSm = − S(t)σaS a (6.12) which for the equilibrium concentration becomes eq ρSm = − γ Nd f σaS γ Nd Nd f = −γ = − σaS a a ν where we have used the approximation that k ≈ νf /a = 1. For a eq reactor, ρSm ≈ 0.0045. (6.13) 235 U-fueled Xenon Poisoning Xenon-135 has a thermal absorption cross section of 2.6 × 106 barns. It is produced directly from fission, with yield γ Xe , and from the decay of 135 I, which in turn is produced by the decay of the direct fission product 135 Te, with yield γ Te , as indicated in Fig. 6.6. The production–destruction equations may be written, with the assumption that 135 I is produced directly from fission with yield γ Te , dI (t) = γ Te f φ − λI I dt   dX(t) = γ Xe f φ + λI I − λX + σaX φ X dt (6.14) 213 214 6 Fuel Burnup Fig. 6.6 Characteristics of 135 Xe under representative LWR conditions: (a) transmutation–decay chain; (b) fission yields; (c) time dependence. (From Ref. 3; used with permission of Taylor & Francis/Hemisphere Publishing.) 6.2 Samarium and Xenon These equations have the solutions I (t) = γ Te f φ  I  I 1 − e−λ t + I (0)e−λ t I λ X(t) = (γ Te + γ Xe )f φ X X 1 − e−(λ +σa φ)t λX + σaX φ + (6.15) γ Te f φ − λI I (0) −(λX +σ X φ)t I X X a e − e−λ t + X(0)e−(λ +σa φ)t λX − λI + σaX φ When the reactor is started up from a clean condition in which X(0) = I (0) = 0, or the reactor power level is changed, the 135 I and 135 Xe concentrations approach equilibrium values: Ieq = γ Te f φ , λI Xeq = (γ Te + γ Xe )f φ λX + σaX φ (6.16) with time constants 1/λI = 0.1 h and 1/(λX + σaX ϕ) ≈ 30 h, respectively. The perturbation theory estimate of the reactivity worth of equilibrium xenon is eq ρXe = − =− σaX (γ Te + γ Xe )f φ a (λX + σaX φ) − γ Te + γ Xe ν(1 + λX /σaX φ) 0.026 1 + (0.756 × 1013 )/φ (6.17) Peak Xenon When a reactor is shut down from an equilibrium xenon condition, the iodine and xenon populations satisfy Eqs. (6.15) with I (0) = Ieq , X(0) = Xeq , and φ = 0: I (t) = Ieq e−λ It X(t) = Xeq e−λ Xt + Ieq λI  −λX t I  − e−λ t e λI − λX (6.18) If φ > (γ X /γ Te )(λX /σaX ), the xenon will build up after shutdown to a peak value at time tPk = λI 1 λI /λX ln X X I −λ 1 + (λ /λ )(λI /λX − 1)(Xeq /Ieq ) (6.19) and then decay to zero unless the reactor is restarted. For 235 U- and 233 U-fueled reactors φ > 4 × 1011 and 3 × 1012 n/cm2 · s, respectively, is sufficient for an increase in the xenon concentration following shutdown. Typical flux values (e.g., 5 × 1013 n/cm2 · s) in thermal reactors are well above these threshold levels, and for typical flux values, Eq. (6.19) yields a peak xenon time of ≈11.6 h. If the reactor is restarted before the xenon has entirely decayed, the xenon concentration will initially decrease because of the burnout of xenon and then gradually build up again 215 216 6 Fuel Burnup because of the decay of a growing iodine concentration, returning to values of Ieq and Xeq for the new power level. This time-dependence of the xenon concentration is illustrated in Fig. 6.6. Effect of Power-Level Changes When the power level changes in a reactor (e.g., in load following) the xenon concentration will change. Consider a reactor operating at equilibrium iodine Ieq (φ0 ) and xenon Xeq (φ0 ) at flux level φ0 . At t = t0 the flux changes from φ0 to φ1 . Equations (6.16) can be written  φ1 − φ0 −λI t e I (t) = Ieq (φ1 ) 1 − φ1   λX φ1 − φ0 X X e−(λ +σa φ1 )t X(t) = Xeq (φ1 ) 1 − X φ1 λ + σaX φ0  + λX + σaX φ1 γ Te I X X e−λ t − e−(λ +σa φ1 ) X I X Te Xe γ + γ λ − λ + σa φ1 (6.20)  The xenon concentration during a transient of this type is shown in Fig. 6.7. The perturbation theory estimate for the reactivity worth of xenon at any time during the transient discussed above is ρXe (t) = − σaX X(t) a − σaX X(t) νf (6.21) Example 6.2: Xenon Reactivity Worth. As an example of xenon buildup, consider a 235 U-fueled reactor that has operated at a thermal flux level of 5 × 1013 cm−2 s−1 for two months such that equilibrium xenon and iodine have built in to the levels given I = 6.6 h, t X = 9.1 h, λ = ln 2/t by Eqs. (6.16). Using σaX = 2.6 × 10−18 cm2 , t1/2 1/2 , 1/2 γTe = 0.061, and γXe = 0.003, the equilibrium values of xenon and iodine are X eq = 0.0203 × 1018 f cm−3 and I eq = 0.1051 × 1018 f cm−3 . The reactivity worth eq of equilibrium xenon is ρXe ≈ σaX Xeq /a ≈ 0.022k/k, where the approximate criticality condition νf ≈ a has been used. If the reactor is shut down for 6 h and then restarted, the xenon reactivity worth that must be compensated is, from Eqs. (6.16) and (6.21), ρXe (t = 6 h) ≈ σaX X(t = 6 h)/νf = (0.634X eq + 0.367I eq ) × σaX /νf = 0.0171 + 0.04 = 0.0571k/k. The largest contribution to the xenon worth at 6 h after shutdown clearly comes from buildup of xenon from the decay of the iodine concentration at shutdown at a faster rate than the resulting xenon decays. 6.3 Fertile-to-Fissile Conversion and Breeding Fig. 6.7 Xenon concentration following power-level changes. (From Ref. 9; used with permission of Wiley.) 6.3 Fertile-to-Fissile Conversion and Breeding Availability of Neutrons The transmutation–decay processes depicted in Fig. 6.1 hold out the potential for increasing the recoverable energy content from the world’s uranium and thorium resources by almost two orders of magnitude by converting the fertile isotopes 238 U and 232 Th, which only fission at very high neutron energies, into fissile isotopes, 239 Pu and 241 Pu in the case of 238 U, and 233 U in the case of 232 Th, which have large fission cross sections for thermal neutrons and substantial fission cross sections for fast neutrons. The rate of transmutation of fertile-to-fissile isotopes depends on the number of neutrons in excess of those needed to maintain the chain fission reaction that are available. In the absence of neutron absorption by anything other 217 218 6 Fuel Burnup Fig. 6.8 Parameter η for the principal fissile nuclei. (From Ref. 17; used with permission of Electric Power Research Institute.) than fuel and in the absence of leakage, the number of excess neutrons is η − 1. The quantity η is plotted in Fig. 6.8 for the principal fissile isotopes. The fertile-to-fissile conversion characteristics depend on the fuel cycle and the neutron energy spectrum. For a thermal neutron spectrum (E < 1 eV), 233 U has the largest value of η of the fissile nuclei. Thus the best possibility for fertile-tofissile conversion in a thermal spectrum is with the 232 Th–233 U fuel cycle. For a 6.4 Simple Model of Fuel Depletion Table 6.4 Conversion/Breeding Ratios in Different Reactor Systems Reactor System BWR PWR PHWR HTGR LMFBR Initial Fuel 2–4 wt% 235 U 2–4 wt% 235 U Natural U ≈5 wt% 235 U 10–20 wt% Pu Conversion Cycle 238 U–239 Pu 238 U–239 Pu 238 U–239 Pu 232 Th–233 U 238 U–239 Pu Conversion Ratio 0.6 0.6 0.8 0.8 1.0–1.6 Source: Data from Ref. 3; used with permission of Taylor & Francis/Hemisphere Publishing. fast neutron spectrum (E > 5 × 104 eV), 239 Pu and 241 Pu have the largest values of η of the fissile nuclei. The LMFBR, based on the 238 U–239 Pu fuel cycle, is intended to take advantage of the increase of η49 at high neutron energy. Conversion and Breeding Ratios The instantaneous conversion ratio is defined as the ratio of the rate of creation of new fissile isotopes to the rate of destruction of fissile isotopes. When this ratio is greater than unity, it is conventional to speak of a breeding ratio, because the reactor would then be producing more fissile material than it was consuming. Average conversion or breeding ratios calculated for reference reactor designs of various types are shown in Table 6.4. The values of the conversion ratios for the PWR and BWR are the same because of design similarities. The HTGR conversion ratio is somewhat higher because of the higher value of η for 233 U than for 235 U. The improved conversion ratio for the CANDU–PHWR is due to the better neutron economy provided by online refueling and consequent reduced requirements for control poisons to compensate excess reactivity. The breeding ratio in an LMFBR can vary over a rather wide range, depending on the neutron energy spectrum. Achieving a large value of η and hence a large breeding ratio favors a hard neutron spectrum. However, a softer spectrum is favored for safety reasons—the lower-energy neutrons which are subject to resonance absorption become more likely to be radiatively captured than to cause fission as the neutron energy is reduced, as discussed in Chapter 5. 6.4 Simple Model of Fuel Depletion The concepts involved in fuel depletion and the compensating control adjustment can be illustrated by a simple model in which the criticality requirement is written as 219 220 6 Fuel Burnup k = ηf = ηaF (t) fp aF (t) + aM + a (t) + c (t) =1 (6.22) where aF is the fuel macroscopic absorption cross section, aM the moderator macroscopic absorption cross section, and aC the combined (soluble and burnable poisons plus control rod) control absorption cross section. Assuming that the reactor operates at constant power νfF (t)φ(t) = νfF (0)φ(0) and that η = νfF /aF is constant in time, the fuel macroscopic absorption cross section at any time is    t NF (t  )φ(t  )dt  aF (t) = NF (t)σaF = σaF NF (0) − εσaF 0 = NF (0)σaF 1 − εφ(0)σaF t (6.23) The neutron flux is related to the beginning-of-cycle neutron flux by φ(t) = φ(0) 1 − εσaF φ(0)t (6.24) where ε < 1 is a factor that accounts for the production of new fissionable nuclei via transmutation–decay. The fission product cross section is the sum of the equilibrium xenon and samarium cross sections constructed using Eqs. (6.16) and (6.10), respectively, and an effective cross section for the other fission products, fp = σfp γfp f (t)φ(t)t = σfp γfp f (0)φ(0)t (6.25) which accumulate in time from fission with yield γfp . The quantity γfp σfp is about 40 to 50 barns per fission. Using these results, Eq. (6.22) can be solved for the value of the control cross section that is necessary to maintain criticality: c (t) = (η − 1)aF (0) 1 − σaF εφ(0)t − aM − (γ Te + γ Xe )f (0)φ(0) λX /σaX + φ(t) − γ Nd f (0) 1 − εσaF φ(0)t − σfp γfp f (0)φ(0)t (6.26) The soluble poison will be removed by the end of cycle, and the burnable poisons should be fully depleted by that time. Thus the lifetime, or cycle time, is the time at which the reactor can no longer be maintained critical with the control rods withdrawn as fully as allowed by safety considerations. This minimum control cross section is small, and we set it to zero. The end-of-cycle time can be determined from Eq. (6.26) by setting aC = 0 and solving for tEOC :  λX ηρex (1 + α) − (γ Te + γ Xe )φ(0)σaX /λX − γ Nd    , φ(t)    σaX [(η − 1)(1 + α)σaF − γ Nd σaF + γfp ]φ(0)     ηρex (1 + α) − (γ Te + γ Xe + γ Nd ) tEOC = ,  [(η − 1)(1 + α)σaF − (γ Te + γ Xe + γ Nd )σaF + γfp σfp ]φ(0)      λX    φ(t)  X σa (6.27) 6.5 Fuel Reprocessing and Recycling where α is the capture-to-fission ratio for the fuel, and ρex ≡ k∞ (0) − 1 k∞ (0) (6.28) is the excess reactivity at beginning-of-cycle without xenon, samarium, fission products, or control cross section. The initial control cross section (including soluble and burnable poisons) must be able to produce a negative reactivity greater than ρex . It is clear from Eq. (6.27) that the cycle lifetime is inversely proportional to the power, or flux, level. 6.5 Fuel Reprocessing and Recycling A substantial amount of plutonium is produced by neutron transmutation of 238 U in LWRs. About 220 kg of fissionable plutonium (mainly 239 Pu and 241 Pu) is present in the spent fuel discharged from an LWR at a burnup of 45 MWd/T. The spent fuel can be reprocessed to recover the plutonium (and remaining enriched uranium) for recycling as new fuel. Composition of Recycled LWR Fuel The potential energy content of the fissile and fertile isotopes remaining in spent reactor fuel (Table 6.2) constitutes a substantial fraction of the potential energy content of the initial fuel loading, providing an incentive to recover the uranium and plutonium isotopes for reuse as reactor fuel. The recycled plutonium concentrations calculated for successive core reloads of a PWR are shown in Table 6.5. The initial core loading and the first reload were slightly enriched UO2 . The plutonium Table 6.5 Plutonium Concentrations in a PWR Recycling Only Self-Generated Plutonium (wt%) Loading: Recycle: 235 U in UO 2 Pu in MOX MOX of fuel 235 U discharged Discharged Pu 239 Pu 240 Pu 241 Pu 242 Pu 1 2 3 1 4 2 5 3 6 4 7 5 2.14 − − 0.83 3.0 − − − 3.0 4.72 18.4 − 3.0 5.83 23.4 − 3.0 6.89 26.5 − 3.0 7.51 27.8 − 3.0 8.05 28.8 − 56.8 23.8 14.3 5.1 56.8 23.8 14.3 5.1 49.7 27.0 16.2 7.1 44.6 38.7 17.2 9.5 42.1 29.4 17.4 11.1 40.9 29.6 17.4 12.1 40.0 29.8 17.3 12.9 Source: Data from Ref. 3; used with permission of Taylor & Francis/Hemisphere Publishing. 221 222 6 Fuel Burnup Fig. 6.9 Thermal absorption cross section for 239 Pu. (From Ref. 4; used with permission of American Nuclear Society.) discharged from the first cycle was recycled in the third cycle, that in the second cycle in the fourth cycle, and so on, in separate mixed oxide (MOX) UPuO2 pins. The proportion of MOX increases from about 18% in the second reload to just under 30% in the sixth and subsequent reloads, for which reloads the plutonium recovered from spent MOX and UO2 fuel is about the same as was loaded into this fuel at beginning-of-cycle (i.e., the plutonium concentration reaches equilibrium). The percentage of plutonium in MOX increases from less than 5% on the initial recycle load to about 8% in equilibrium, in order to offset the reactivity penalty. Physics Differences of MOX Cores The use of MOX fuels in PWRs changes the physics characteristics in several ways. The variation with energy of the cross sections for the plutonium isotopes is more complex than for the uranium isotopes, as shown in Fig. 6.9. The absorption cross sections for the plutonium isotopes are about twice those of the uranium isotopes in a thermal spectrum and are characterized by large absorption resonances in the epithermal (0.3 to 1.5 eV) range and by overlapping resonances. Representative thermal neutron spectra in UO2 and MOX fuel cells are compared in Fig. 6.10. Thermal parameters for 235 U and 239 Pu, averaged over a representative LWR thermal neutron energy distribution, are given in Table 6.6. Because of the larger thermal absorption cross section for 239 Pu, the reactivity worth of control rods, burnable poisons, and soluble poisons (PWRs) will be less with MOX fuel than with UO2 , unless the MOX rods can be placed well away from control rods and 6.5 Fuel Reprocessing and Recycling Table 6.6 Thermal Parameters for 235 U and 239 Pu in a LWR Parameter Fission cross section σf (barns) Absorption cross section σa (barns) Nu-fission to absorption η Delayed neutron fraction β Generation time  (s) 235 U 365 430 2.07 0.0065 4.7×10−5 239 Pu 610 915 1.90 0.0021 2.7×10−5 Source: Data from Ref. 4; used with permission of American Nuclear Society. burnable poisons. The higher 239 Pu fission cross section will lead to greater power peaking with MOX than with UO2 , unless the MOX rods are placed well away from water gaps. There are reactivity differences between MOX and UO2 . The buildup of 240 Pu and 242 Pu with the recycling MOX fuel accumulates parasitic absorbers that results in a reactivity penalty, as discussed in Section 6.1. The average thermal value of η Fig. 6.10 Thermal neutron spectra in UO2 and MOX PWR fuel cells. (From Ref. 1; used with permission of Nuclear Energy Agency, Paris.) 223 224 6 Fuel Burnup is less for 239 Pu than for 235 U, which requires a larger fissile loading to achieve the same initial excess reactivity with MOX as with UO2 . Furthermore, the temperature defect is greater for MOX because of the large low-energy resonances in 239 Pu and 240 Pu shown in Fig. 6.9. However, the reactivity decrease with burnup is less for MOX than for UO2 , because of the lower η for 239 Pu than for 235 U, and because of the transmutation of 240 Pu into fissionable 241 Pu, so that less excess reactivity is needed. The delayed neutron fractions for 239 Pu, 241 Pu, and 235 U are in the ratio 0.0020/0.0054/0.0064, which means that the reactivity insertion required to reach prompt critical runaway conditions is less for MOX than for UO2 by a factor that depends on the 239 Pu/241 Pu/235 U ratio. As the 241 Pu builds up with repeated recycle, the difference between MOX and UO2 decreases. The neutron generation time is also shorter for MOX than for UO2 , so that any prompt supercritical excursion would have a shorter period. The fission spectrum neutrons are more energetic for 239 Pu than for 235 U. On the other hand, because of the large epithermal absorption resonances in the plutonium isotopes, the moderator and fuel Doppler temperature coefficients of reactivity tend to be more negative for MOX cores than for UO2 cores. Accumulation of actinides, which are strong emitters of energetic alpha particles, leads to higher radioactive decay heat removal requirements with MOX. These considerations would tend to limit the MOX fraction in a reload core. The yield of 135 Xe is about the same for the fission of plutonium as for the fission of uranium. Due to the higher thermal absorption cross section of the plutonium isotopes, the excess reactivity needed to start up at peak xenon conditions and the propensity for spatial flux oscillations driven by xenon oscillations (Chapter 16) are less in a MOX than a UO2 core. For plutonium recycle in other reactor types, similar types of physics considerations would enter. However, the different relative values of η for 235 U and 239 Pu in different spectra (e.g., the epithermal spectrum of a HTGR and the fast spectrum of a LMFBR) would lead to different conclusions about reactivity penalties. In fact, LMFBRs have been designed from the outset with the concept of switching from 235 U to 239 Pu as the latter was bred. Physics Considerations with Uranium Recycle Although it is relatively straightforward to separate uranium from other chemically distinct isotopes, it is impractical to separate the various uranium isotopes from each other in the reprocessing step. So recycling uranium means recycling all of the uranium isotopes, some of which are just parasitic absorbers and another of which leads through subsequent decay to the emission of an energetic gamma. Two isotopes present in relatively small concentrations in fresh fuel (234 U and 236 U) necessitate adding 235 U to enrich reprocessed uranium to a higher enrichment than is required with fresh uranium fuel. Uranium-234 has a large absorption resonance integral and, while only a tiny fraction in natural uranium, will tend to be enriched along with 235 U. Uranium-236 is produced by neutron capture in 235 U and by electron capture in 236 Np, as shown in Fig. 6.4, and is a parasitic 6.5 Fuel Reprocessing and Recycling Table 6.7 Representative Fueling Characteristics of 1000-MWt Reactors LWR Reactor Type LWR LMFBR 232 Th–233 U 238 U–239 Pu 238 U–239 Pu 0.78 1,580 35,000 720 435 285 0.71 2,150 33,000 1,000 650 350 1.32 3,160 100,000 1,480 1,690 (−210) Characteristic Fuel cycle Conversion ratio Initial core load (kg) Burnup (MWd/T) Annual reload (kg) Annual discharge (kg) Annual makeup (kg) Source: Data from Ref. 8; used with permission of International Atomic Energy Agency. neutron absorber with a significant capture resonance integral. Reprocessed uranium is made difficult to handle by the decay product 208 Tl, which emits a 2.6-MeV gamma with t1/2 = 3.1 min. This radioisotope is produced by a series of alpha decays of 232 U, which is produced by the chain shown in Fig. 6.4. Physics Considerations with Plutonium Recycle The same type of difficulties exists for plutonium reprocessing as discussed for uranium—all of the plutonium isotopes must be recycled. Plutonium-236 decays into 232 U, which leads to the emission of a 2.6-MeV gamma, as described above. Plutonium-238 is produced through neutron transmutation of 237 Np; it alphadecays with t1/2 = 88 years and constitutes a large shutdown heat source if present in sufficient quantity. Plutonium-240 has an enormous capture resonance integral. Both 238 Pu and 240 Pu contribute a large spontaneous fission neutron source. Plutonium-241, while having a large fission cross section, also decays into 241 Am, which has a large thermal capture cross section and a large capture resonance integral. Americium-241 also decays into daughter products which are energetic gamma emitters. Stored plutonium loses its potency as a fuel over time because of the decay of 241 Pu into 241 Am. Plutonium from spent LWR fuel at a typical burnup of about 35,000 MWd/T must be utilized within 3 years after discharge or it will be necessary to reprocess it again to remove the 241 Am and daughter products. Reactor Fueling Characteristics Nuclear fuel cycles with plutonium recycle have been studied extensively (e.g., Ref. 1). Representative equilibrium fueling characteristics for LWRs operating on the 238 U–239 Pu and 232 Th–233 U fuel cycles and for a LMFBR operating on the 238 U–239 Pu fuel cycle are shown in Table 6.7. Fuel is partially discharged and replenished each year (annual discharge and annual reload), requiring a net amount of new fuel (annual makeup) from outside sources. In the absence of reprocessing and 225 226 6 Fuel Burnup recycling, the annual reload would have to be supplied from outside sources. The LMFBR produces more fuel than it uses and could provide the extra fuel needed by the LWRs from the transmutation of 238 U if LMFBRs and LWRs were deployed in the ratio of about 7:5. 6.6 Radioactive Waste Radioactivity The actinides produced in the transmutation–decay of the fuel isotopes and the fission products are the major contributors to the radioactive waste produced in nuclear reactors, although activated structure and other materials are also present. The activity per ton of fuel for representative LWR and LMFBR discharges are given in Table 6.8. The fission products account for almost the entire radioactivity of spent fuel at reactor shutdown, but because of their short half-lives, this radioactivity level decays relatively quickly. In fact, the radioactivity of the spent fuel decreases substantially within the first 6 months after removal from the reactor, as shown in Table 6.8. The more troublesome fission products from the waste management point of view are those with long half-lives like 99 Tc (t1/2 = 2.1 × 105 years) and 129 I (t 7 90 1/2 = 1.59 × 10 years) and those that are gamma emitters, such as Sr and 137 Cs, which produce substantial decay heating. The actinides constitute a relatively small part of the total radioactivity at reactor shutdown but become relatively more important with time because of the longer half-lives of 239 Pu and 240 Pu and dominate the radioactivity of spent fuel after about 1000 years. Hazard Potential A simple, but useful, measure of the hazard potential of radioactive material is the hazard index, defined as the quantity of water required to dilute the material to the maximum permissible concentration for human consumption. The hazard index for spent LWR fuel is plotted against time after shutdown in Fig. 6.11. Fission products dominate the hazard index up to about 1000 years after shutdown, beyond which time the transuranics (actinides) become dominant. Including the plutonium in the recycled uranium fuel increases the hazard potential because of the continued buildup of 239 Pu and 240 Pu. Beyond 1000 to 10,000 years after shutdown, depending on burnup, the hazard potential of spent reactor fuel is less than the hazard potential of uranium ore as it is mined from the earth. Risk Factor In an effort to relate the radioactivity of a given radioisotope to a health hazard, the number of curies of radiation from a given radioisotope that would cause cancer (on the average) if swallowed by a person has been estimated and is shown as the cancer dose per curie (CD/Ci) in Table 6.9. The CD/Ci is not an absolute measure of 115 Cd 111 Ag 103m Rh 106 Ru 103 Ru 99 Tc 99m Tc 99 Mo 95 Nb 95 Zr 91 Y 90 Y 90 Sr 89 Sr 85 Kr 3H Nuclide 12.3 y 10.73 y 50.5 d 29.0 y 64.0 h 59.0 d 64.0 d 3.50 d 66.0 h 6.0 h 2.1×105 y 40.0 d 369.0 d 56.0 min 7.47 d 44.6 d Half-Life t 1/2 β β, γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ γ β, γ β, γ β, γ γ β, γ β, γ Radiations† 5.744 × 102 1.108 × 104 1.058 × 106 8.425 × 104 8.850 × 104 1.263 × 106 1.637 × 106 1.557 × 106 1.875 × 106 1.618 × 106 1.435 × 101 1.560 × 106 4.935 × 105 1.561 × 106 5.375 × 104 1.483 × 103 Discharge LWR Fuel 5.587 × 102 1.074 × 104 9.603 × 104 8.323 × 104 8.325 × 104 1.525 × 105 2.437 × 105 4.689 × 105 3.780 × 10−14 3.589 × 10−14 1.442 × 101 6.680 × 104 3.519 × 105 6.686 × 104 3.005 × 10−3 9.042 × 101 180 d 1.648 × 103 1.473 × 104 1.333 × 106 9.591 × 104 1.214 × 105 1.794 × 106 3.215 × 106 3.149 × 106 4.040 × 106 3.487 × 106 3.278 × 101 4.617 × 106 2.248 × 106 4.619 × 106 2.294 × 105 7.041 × 103 Discharge Activity (Ci/tonne Heavy Metal) LMFBR Fuel Table 6.8 Radioactivity of Representative LWR and LMFBR Spent Fuel at Discharge and at 180 Days (LWR) or 30 Days (LMFBR) After Discharge* 1.640 × 103 1.466 × 104 8.939 × 105 9.572 × 104 9.572 × 104 1.269 × 106 2.340 × 106 2.954 × 106 2.108 × 103 2.002 × 103 3.293 × 101 2.730 × 106 2.125 × 106 2.733 × 106 1.422 × 104 4.418 × 103 30 d 6.6 Radioactive Waste 227 140 La 140 Ba 137 Cs 136 Cs 134 Cs 133 Xe 132 I 131 I 129 I 132 Te 129 Te 129m Te 127 Te 127m Te 125m Te 125 Sb 124 Sb 125 Sn Nuclide 9.65 d 60.2 d 2.73 y 58.0 d 109.0 d 9.4 h 33.4 d 70.0 min 78.0 h 1.59×107 y 8.04 d 2.285 h 5.29 d 2.06 y 13.0 d 30.1 y 12.79 d 40.23 h Half-Life t 1/2 Table 6.8 (Continued) β, γ β, γ β, γ γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ β, γ Radiations† 1.081 × 104 4.147 × 102 9.525 × 103 1.976 × 103 1.384 × 104 9.920 × 104 8.508 × 104 3.211 × 105 1.486 × 106 3.219 × 10−2 1.028 × 106 1.511 × 106 2.098 × 106 2.718 × 105 6.962 × 104 1.115 × 105 1.953 × 106 2.019 × 106 Discharge LWR Fuel Discharge 3.404 × 104 2.329 × 103 5.251 × 104 1.121 × 104 4.969 × 104 3.247 × 105 2.316 × 105 8.454 × 105 3.473 × 106 1.033 × 10−1 2.602 × 106 3.546 × 106 4.414 × 106 8.283 × 104 2.577 × 105 2.522 × 105 3.636 × 106 3.698 × 104 180 d 2.624 × 10−2 5.219 × 101 8.498 × 103 2.031 × 103 4.595 × 103 4.500 × 103 2.041 × 103 1.296 × 103 3.159 × 10−11 3.268 × 10−2 1.933 × 10−1 3.254 × 10−11 1.612 × 10−4 2.303 × 105 4.719 × 100 1.102 × 105 1.133 × 102 1.303 × 102 Activity (Ci/tonne Heavy Metal) LMFBR Fuel 3.946 × 103 1.649 × 103 5.171 × 104 1.144 × 104 4.265 × 104 4.308 × 104 1.249 × 105 7.932 × 104 5.783 × 103 1.040 × 10−1 2.020 × 105 5.956 × 103 1.076 × 105 8.058 × 104 5.204 × 104 2.518 × 105 7.153 × 105 8.238 × 105 30 d 228 6 Fuel Burnup 32.53 d 284.0 d 13.58 d 10.99 d 2.62 y 53.1 h 93.0 y 13.4 y 4.8 y 72.3 d 2.35 d 87.8 y 2.44×104 y 6.54×103 y 15.0 y 433.0 y 163.0 d 17.9 d Half-Life t 1/2 β, γ β, γ β β, γ β, γ β, γ β+, β−, γ β+, β−, γ β, γ β, γ β, γ α, γ α, γ , SF α, γ , SF α, β, γ α, γ , SF α, γ , SF α, γ Radiations† ∗ Calculated with ORIGEN (Ref. 7). † α, alpha particle; β, electron; γ , gamma ray; SF, spontaneous fission. 244 Cm 242 Cm 241 Am 241 Pu 240 Pu 239 Pu 238 Pu 239 Np 160 Tb 155 Eu 152 Eu 151 Sm 149 Pm 147 Pm 147 Nd 143 Pr 144 Ce 141 Ce Nuclide Table 6.8 (Continued) 1.784 × 106 1.229 × 106 1.657 × 106 7.902 × 105 1.031 × 105 3.919 × 105 8.658 × 102 7.838 × 100 2.540 × 103 1.418 × 103 2.435 × 107 2.899 × 103 3.250 × 102 4.842 × 102 1.098 × 105 8.023 × 101 3.666 × 104 2.772 × 103 Discharge LWR Fuel 3.876 × 104 7.925 × 105 1.887 × 102 9.278 × 100 9.859 × 104 1.326 × 10−19 8.696 × 102 7.635 × 100 2.365 × 103 2.525 × 102 2.050 × 101 3.021 × 103 3.314 × 102 4.843 × 102 1.072 × 105 1.657 × 102 1.717 × 104 2.720 × 103 180 d 3.730 × 106 2.148 × 106 3.044 × 106 1.513 × 106 6.344 × 105 9.842 × 105 9.693 × 103 4.759 × 101 4.305 × 104 4.880 × 103 5.990 × 107 2.770 × 104 6.247 × 103 8.323 × 103 7.280 × 105 9.091 × 103 8.467 × 105 8.032 × 103 Discharge Activity (Ci/tonne Heavy Metal) LMFBR Fuel 1.979 × 106 1.996 × 106 7.349 × 105 2.283 × 105 6.353 × 105 8.451 × 101 9.703 × 103 4.738 × 101 4.255 × 104 3.661 × 103 8.727 × 103 2.820 × 104 6.263 × 103 8.323 × 103 7.252 × 105 9.186 × 103 7.489 × 105 8.007 × 103 30 d 6.6 Radioactive Waste 229 230 6 Fuel Burnup Table 6.9 Cancer Dose per Curie for Radioisotopes Present in Spent Fuel* Isotope 210 Pb 223 Ra 226 Ra 227 Ac 229 Th 230 Th 231 Pa 234 U 235 U 236 U 238 U 237 Np 238 Pu 239 Pu 240 Pu 242 Pu 241 Am 242m Am 243 Am 242 Cm 243 Cm 244 Cm 245 Cm 246 Cm 90 Sr 90 Y 137 Cs 99 Tc 129 I 93 Zr 135 Cs 14 C 59 Ni 63 Ni 126 Sn Toxicity Factor (CD/Ci) Half-Life (years) Actinides and Their Daughters 455.0 22.3 15.6 0.03 36.3 1.60 × 103 1185.0 21.8 127.3 7.3 × 103 19.1 7.54 × 104 372.0 3.28 × 104 7.59 2.46 × 105 7.23 7.04 × 108 7.50 2.34 × 107 6.97 4.47 × 109 197.2 2.14 × 106 246.1 87.7 267.5 2.41 × 104 267.5 6.56 × 103 267.5 3.75 × 105 272.9 433 267.5 141 272.9 7.37 × 103 6.90 0.45 196.9 29.1 163.0 18.1 284.0 8.5 × 103 284.0 4.8 × 103 Short-Lived Fission Products 16.7 29.1 0.60 7.3 × 10−3 5.77 30.2 Long-Lived Fission Products 0.17 2.13 × 105 64.8 1.57 × 107 0.095 1.5 × 106 0.84 2.3 × 106 0.20 5.73 × 103 0.08 7.6 × 104 0.03 100 1.70 1.0 × 105 Source: Data from Ref. 1: used with permission of Nuclear Energy Agency, Paris. * The toxicity factors are constructed using the methodology described by Bernard L. Cohn, “Effects of the ICRP Publication 30 and the 1980 BEIR Report of Hazard Toxicity Factor (CD/g) 3.48 × 104 7.99 × 105 3.59 × 101 8.58 × 104 2.72 × 101 3.94×10−1 1.76×10−1 4.71×10−2 1.56×10−5 4.85×10−4 2.34×10−6 1.39×10−1 4.22 × 103 1.66 × 101 6.08 × 101 1.65 × 100 9.36 × 102 2.80 × 104 5.45 × 101 2.29 × 104 9.96 × 103 1.32 × 104 4.88 × 101 8.67 × 101 2.28 × 103 3.26 × 105 4.99 × 102 2.28×10−3 1.15×10−2 2.44×10−4 9.68×10−4 8.92×10−1 6.38×10−3 1.70 × 100 4.83×10−2 6.6 Radioactive Waste Table 6.9 (Continued) Assessments of High Level Waste,” Health Phys. 42 (2) 133–143 (1982) with the following data: ICRP Publication 30, Part 4, 88, 19, and BEIR III, 80, 19. The factors stand for the fatal cancer doses per gram of isotope injected orally. They denote the hazard of the material rather than the risk because they do not include any account of pathway attenuation processes, but simply assume oral ingestion. the biological hazard of a given amount of radioactive material, because it does not contain any measure of the probability for a sequence of events that would result in individual members of a population actually swallowing exactly the amount of radioisotopes that would produce a cancer, and no more. However, the CD/Ci can be used to construct a relative measure of the biological hazard potential. The CD/ton (of heavy metal in the fuel) due to the radioisotopes in discharged fuel given in Table 6.8 can be multiplied by the number of tons of heavy metal spent fuel discharged from a reactor to construct a total cancer dose (TCD). A similar total cancer dose of natural uranium (TCDNU) as it is mined from the earth can be constructed by multiplying the Ci/ton for the radioisotopes in natural uranium by the mass of natural uranium that was required to produce the discharged fuel for which the Fig. 6.11 Hazard index for spent LWR fuel as a function of time since reactor shutdown. (From Ref. 18; used with permission of Woods Hole Oceanographic Institute.) 231 232 6 Fuel Burnup Fig. 6.12 Risk factor for LWR spent fuel without recycle. (From Ref. 5; used with permission of Elsevier Science Publishers.) TCD is calculated. (Typically, about 5 tons of natural uranium is needed to produce the fuel for a PWR.) The risk factor is then defined as RF ≡ TCD/TCDNU, which may be interpreted as the ratio of the number of cancers that would be caused by individual members of a population swallowing all of the discharged fuel in portions just sufficient to produce a cancer (on average) to the number of cancers that would be caused by individual members of a population swallowing 5 tons of natural uranium in portions just sufficient to produce a cancer (on average). The advantage of the risk factor is that the highly uncertain probability of ingestion of radioisotopes is normalized out by being treated in the same (highly questionable) way in the numerator (TCD) and denominator (TCDNU), so that RF is a measure of the relative cancer potential of spent fuel and of the natural uranium from which it was produced. The risk factor is plotted for a typical spent fuel loading from a LWR in Fig. 6.12. The short-lived fission products are dominant in the decades following discharge, but the fission product activity becomes negligible relative to the actinide activity after about 200 to 300 years. The potential α-toxicity of the actinide concentration is dominated by 241 Am over the first 5000 years, then by 240 Pu up to about 100,000 6.7 Burning Surplus Weapons-Grade Uranium and Plutonium Fig. 6.13 Risk factor for LWR spent fuel with 99.5% recycle of Pu, Am, and Np. (From Ref. 5; used with permission of Elsevier Science Publishers.) years, and thereafter by 237 Np. Note that when the risk factor becomes less than unity, the cancer potential of the spent fuel is less than the cancer potential of the natural uranium ore from which it was originally made. The long-term potential α-toxicity of spent fuel can be reduced dramatically by recycling the fuel. Figure 6.13 illustrates risk factor for the same LWR fuel as in Fig. 6.12, but now with the Pu, Am, and Np recycled to 99.5% annihilation. After about 200–300 years the potential α-toxicity of the spent fuel is less than that of the natural uranium from which it was originally produced. As discussed in Section 6.8, repeatedly recycling the spent fuel to 99.5% annihilation may be feasible, from neutron balance considerations, in a fast spectrum reactor, but does not appear to be feasible in a thermal reactor. 6.7 Burning Surplus Weapons-Grade Uranium and Plutonium Composition of Weapons-Grade Uranium and Plutonium With the reduction in nuclear weapons worldwide, surplus highly enriched, weapons-grade uranium and plutonium become available for use as fuel in nuclear 233 234 6 Fuel Burnup Table 6.10 Composition of Weapons- and Reactor-Grade Uranium and Plutonium (wt%) 238 Pu 239 Pu 240 Pu 241 Pu 242 Pu 241 Am WeaponsGrade Pu ReactorGrade Pu 0.01 93.80 5.80 0.13 0.02 0.22 1.30 60.30 24.30 5.60 5.00 3.50 WeaponsGrade U (HEU) ReactorGrade U (LEU) Natural U 234 U 0.12 94.00 5.88 0.025 3.500 96.475 235 U 238 U 0.0057 0.7193 99.2750 Source: Data from Ref. 2; used with permission of National Academy Press. reactors. The composition of typical weapons-grade uranium and plutonium is compared with the composition of reactor-grade uranium and plutonium in Table 6.10. Reactor-grade here refers to the typical enriched uranium used in LWRs and the plutonium composition created by transmutation in LWR fuel. Although it is feasible to de-enrich the weapons-grade uranium, the weapons-grade plutonium would be used as is. Physics Differences Between Weapons- and Reactor-Grade Plutonium-Fueled Reactors There are some similarities and some important differences between using weapons- and reactor-grade plutonium in an LWR designed for low-enrichment uranium fuel. The delayed neutron fractions for thermal fission of 239 Pu, 241 Pu, and 235 U are in the ratio 0.0021:0.0049:0.0065. Because the delayed neutron fraction is smaller in 239 Pu than in 235 U, the subprompt-critical reactivity range is much less for plutonium-fueled reactors than for uranium-fueled reactors, as discussed in Section 6.5; and because the delayed neutron fraction is much smaller in 239 Pu than in 241 Pu, reactors fueled with weapons-grade plutonium will have an even smaller subprompt-critical reactivity range than reactors fueled with reactorgrade plutonium. The large resonance integral of 240 Pu contributes a significant negative Doppler coefficient when reactor-grade plutonium is used, but which is absent when weapons-grade plutonium is used. Similarly, the use of weapons-grade uranium with the low 238 U content would substantially reduced the negative 238 U Doppler coefficient relative to the use of reactor-grade uranium. A resonance absorber such as tungsten can be added to weapons-grade fuel in order to recover part of the negative Doppler coefficient. Calculated Doppler coefficients for a standard LWR UO2 lattice with reactor-grade uranium and for various combinations of UO2 ZrO2 and W with weapons-grade uranium and plutonium are given in Table 6.11. Because of the higher fission cross section and higher value of η for 239 Pu than for 6.8 Utilization of Uranium Energy Content Table 6.11 Fuel Doppler Temperature Coefficients of Reactivity with Weapons-Grade Plutonium Composition R-G UO2 (3% 235 U) W-G UO2 -ZrO2 (0.6% UO2 ) W-G UO2 -ZrO2 + W (3% UO) W-G MOX-ZrO2 (2.7% UO2 , 0.3% PuO2 ) W-G PuO2 -ZrO2 (0.34% PuO2 ) W-G PuO2 -ZrO2 + W (3% PuO2 ) kk /kk (×10−5 ) −2.4720 −0.0017 −1.0357 −0.9588 −0.0009 −1.2003 235 U in a fast neutron spectrum, weapons-grade plutonium fuel projects superior performance to uranium fuel in fast breeder reactors. 6.8 Utilization of Uranium Energy Content Only about 1% of the energy content of the uranium used to produce the fuel is extracted (via fission) in a typical LWR fuel cycle. About 3% of the energy content of the mined uranium is stored as tails from the original uranium fuel production process, and about 96% remains in the depleted uranium from the enrichment process and in the discharged spent fuel in the form of uranium, plutonium, and higher-actinide isotopes. With continued reprocessing and recycling of spent fuel and depleted uranium, there is the possibility of recovering much of this remaining energy. The projected worldwide consumption of uranium based on the LWR oncethrough fuel cycle is shown in Fig. 6.14. Fig. 6.14 Worldwide Uranium Resource Utilization. 235 236 6 Fuel Burnup Table 6.12 Equilibrium Distribution of Transuranic Isotopic Masses (%) for Continuously Recycled Fuel in Thermal and Fast Reactor Neutron Spectra Isotope 237 Np 238 Pu 239 Pu 240 Pu 241 Pu 242 Pu 241 Am 242 Am 243 Am 242 Cm 243 Cm 244 Cm 245 Cm 246 Cm 247 Cm 248 Cm 249 Bk 249 Cf 250 Cf 251 Cf 252 Cf Thermal Reactor Spectrum Fast Reactor Spectrum 5.51 4.17 23.03 10.49 9.48 3.89 0.54 0.02 8.11 0.18 0.02 17.85 1.27 11.71 0.75 2.77 0.05 0.03 0.03 0.02 0.08 0.75 0.89 66.75 24.48 2.98 1.86 0.97 0.07 0.44 0.40 0.03 0.28 0.07 0.03 2. × 10−3 6. × 10−4 1. × 10−5 4. × 10−5 7. × 10−6 9. × 10−7 4. × 10−8 Source: Data from Ref. 14. To fully consume the initial uranium feedstream, for each transuranic atom fissioned there must be one neutron released to sustain the chain reaction and one neutron available for capture in 238 U (or 240 Pu) to produce 239 Pu (or 241 Pu) to replace the fissioned atom. There is, of course unavoidable parasitic capture in fission products, structure, and the transuranic elements. The continued recycling of spent fuel would lead, after long exposure, to equilibrium distributions of the transuranic isotopes in the recycled fuel, as shown for thermal and fast neutron spectra in Table 6.12. (Note that these concentrations could be altered by blending spent fuels from different numbers of recycles.) The number of neutrons per fission lost to parasitic capture in the transuranics can be estimated from their capture and fission probabilities, which are shown in Fig. 6.15 for typical LWR and LMR spectra. For the equilibrium distribution of Table 6.12, the number of neutrons per fission lost to parasitic capture is typically about 0.25 in a fast neutron spectrum and 1.25 in a thermal neutron spectrum. This means that a minimum (not accounting for parasitic capture in fission products, control elements, and structure or leakage) number of neutrons released per fission to maintain the chain reaction and transmute a fertile isotope into a fis- 6.9 Transmutation of Spent Nuclear Fuel Fig. 6.15 Probability of fission per neutron absorbed for actinide isotopes in thermal and fast neutron spectra. (From Ref. 1; used with permission of Nuclear Energy Agency, Paris.) sionable isotope to replace each fissioned isotope is 2.25 for the LMR spectrum and 3.25 for the LWR spectrum. Physically, more neutrons are wasted transmuting a transuranic nuclide into another transuranic nuclide in a thermal spectrum than in a fast spectrum. Since 2.5 < η  3.25, total energy extraction by repeated recycling in a thermal reactor is not possible, but it may be in a fast reactor. The projected uranium utilization of LWRs together with fast breeder reactors that would transmute the non-fissile 238 U into fissile 239 Pu for burning in the LWRs is compared with the utilization of LWRs operating on the once-through cycle in Fig. 6.14. Clearly the early deployment of fast breeder reactors is essential if nuclear power is to achieve its potential in meeting the world’s growing energy demands. 6.9 Transmutation of Spent Nuclear Fuel The once-through cycle (OTC), in which slightly enriched UO2 fuel (235 U increased from 0.72% in natural U to 3 to 5%) is irradiated to 30 to 50 GWd/T in a commercial reactor and then disposed of in toto as high-level waste (HLW), is the reference nuclear fuel cycle in the United States and a few other countries. With the present low uranium prices, this is the cheapest fuel cycle in the short term. Moreover, until recently (2006) U.S. government policy against reprocessing, motivated by proliferation concerns, was consistent only with the OTC. However, the long-term implications of the OTC are rather unfavorable. The potential energy content of the 237 238 6 Fuel Burnup Fig. 6.16 Worldwide Spent Fuel. residual fissile material (about 1% each Pu and 235 U) and of the 238 U (>90%) in the spent fuel and in the depleted uranium, which constitutes >90% of the potential energy content of the mined uranium, is lost in the OTC. Moreover, all the nuclides that can contribute to the potential radiotoxicity of the spent fuel are retained, together with the much greater volume of depleted U (mostly 238 U), which makes a relatively small contribution to the potential radiotoxicity, resulting in the largest possible volume of HLW, which must be stored in geological repositories for hundreds of thousands to millions of years. The projected worldwide accumulation of spent nuclear fuel in the once-through cycle is shown in Fig. 6.16. Today, there are large inventories of plutonium and other minor actinides that have accumulated in discharged spent nuclear fuel. Presently, 40,000 tonnes initial uranium of spent nuclear fuel has accumulated in the Unites States. This inventory continues to grow at a rate of 2000 tonnes/yr. At the current level of nuclear energy production in the United States using the OTC, a new repository on the scale of the presently proposed Yucca Mountain site would have to be installed about every 30 years. The objective of transmutation of spent fuel is to reduce both the mass of HLW that must be stored in geological repositories and the time of high radiotoxicity of that HLW, thus reducing the requirements for both the number of repositories and the duration of secured storage. A National Research Council (NRC) study recently concluded that the need for a geological repository could be reduced, but would not be eliminated, by transmutation. 6.9 Transmutation of Spent Nuclear Fuel The short-term radiotoxicity of the spent fuel is dominated by fission products, but after 300 to 500 years only the long-lived radionuclides (particularly 99 Tc and 129 I, but also 135 Cs, 93 Zr, and others) remain—unfortunately, some of these are relatively mobile and contribute disproportionately to the potential radiological hazard from spent fuel. However, the long-term potential radiotoxicity of spent fuel arises principally from the presence of transuranic actinides (Pu and the so-called minor actinides Np, Am, Cm, etc.) produced by transmutation–decay chains originating with neutron capture in 238 U, which constitute a significant radiation source for hundreds of thousands of years. The contributions to the radiotoxicity of typical spent nuclear fuel from actinides, fission products, and activated structure are shown in Fig. 6.17. Processing of spent UO2 fuel to recover the residual U and Pu reduces the potential long-term radiotoxicity of the remaining HLW (minor actinides, fission products, activated structure, etc.) by a factor of 10 and reduces the volume by a much larger factor, and processing technology (PUREX) capable of 99.9% efficient recovery of U and Pu is commercially available in a number of countries (United Kingdom, France, Japan, India, Russia, and China). A fuel cycle in which the recovered Pu and U was recycled as a mixed oxide (MOX) UO2 –PuO2 commercial reactor fuel has been envisioned since the beginning of the nuclear energy era, and at present a number of commercial reactors are operating with recycled Pu in western Europe. (Reprocessed uranium is not being recycled significantly because of the low cost of fresh uranium, which does not contain the neutron-absorbing 236 U that decreases the reactivity of recycled U.) Taking into account further production of minor actinides and fission products in the recycled Pu, a single recycle of the Pu in spent fuel reduces the potential radiotoxicity of the HLW associated with the original spent fuel only by a factor of 3 (rather than 10). Repeated recycling of the MOX fuel is technically feasible and would result in better fuel utilization, but the potential radiotoxicity of the HLW associated with the original spent fuel would actually increase relative to the OTC because of the further production of minor actinides and fission products. It is clear from the discussion above that to reduce the potential radiological hazard associated with spent fuel or the length of time that hazard exists, it is necessary (1) to destroy the actinides (Pu and the minor actinides) and (2) to destroy the potentially hazardous long-lived fission products. The destruction of the minor actinides and long-lived fission products, as well as the Pu, by neutron transmutation implies the requirement for separation of these nuclides from the waste stream of processed spent fuel for recycling with subsequent fuel loadings. Effective separation of Pu with 99.9% efficiency is achieved commercially with the PUREX process. The effective separation of Np is technically feasible with a modified PUREX process, but practical separation methods for Am, Cm, and the longlived fission products are still in the research stage. The pyrometallurgical (PYRO) separation technology presently under development would, unlike the PUREX process, allow separation of Np, Am, and Cm along with Pu into a code-posited metallic product that could be recycled in a metal-fuel fast reactor, resulting in a waste stream essentially free of actinides. 239 240 6 Fuel Burnup Fig. 6.17 Radiotoxic inventory of UO2 fuel as a function of time [3.7% 235 U, 45 GWD/tonne heavy metal; Becquerel (Bq) = 1 disintegration per second = 2.7 × 10−11 Ci; Sievert (Sv) = 100 rad equivalent]. (From Ref. 12; used with permission of Nuclear Energy Agency, Paris.) 6.9 Transmutation of Spent Nuclear Fuel Since all of the actinides are potentially radiotoxic and since neutron capture (n, γ ) reactions in the actinides just produce other actinides, the only effective way to destroy actinides is by neutron fission (n, f ) reactions. Some of the actinides are effectively not fissionable in a thermal neutron spectrum, such as exists in almost all commercial nuclear reactors, and the probability of fission per neutron absorbed is greater for all the actinides in a fast neutron spectrum (see Fig. 6.15). The neutron absorption cross sections for the troublesome long-lived fission products are small in a thermal neutron spectrum and even smaller in a fast neutron spectrum, implying the advantage of a very high flux of thermal neutrons for their effective destruction (effective destruction of 135 Cs may prove impractical because of the presence of other neutron-absorbing Cs isotope fission products). Several studies of minor actinide transmutation in nuclear reactors have been performed. They indicate that recycling of industrial levels of minor actinides as well as Pu in thermal neutron spectrum commercial reactors does not significantly reduce the overall radiotoxicity and requires an increase in fuel enrichment, with a corresponding increase in the cost of energy. On the other hand, recycling minor actinides as well as Pu in fast reactors is predicted to reduce the overall radiotoxicity of the HLW, but the maximum loading of minor actinides is limited by reactor safety considerations. The possibility of recycling Pu and the minor actinides first in thermal neutron spectrum commercial light water reactors (LWRs) and then in dedicated fast reactors has been calculated to be able to reduce the radio-toxic inventory in the HLW by a factor of about 100 relative to the OTC. Such studies generally indicate that the transmutation of Pu, minor actinides and fission products in critical nuclear reactors would ultimately be limited by criticality or safety constraints. While fast reactors could, in principle, burn the mix of Pu plus minor actinides and some of the fission products, the available PUREX process does not separate the minor actinides with the plutonium from the waste stream for recycling. Moreover, it would be difficult to fabricate MOX fuel containing the highly radioactive minor actinides in existing facilities. This has led in Europe and Japan to consideration of remote fuel fabrication facilities to supply fuel containing minor actinides for destruction in dedicated subcritical transmuter reactors driven by accelerator spallation neutron sources, while the Pu would be consumed in dedicated fast reactors. The U.S. ATW concept is to use remote fabrication of fuel containing separated Pu plus minor actinides (but no 238 U) for destruction in a subcritical transmuter reactor driven by an external neutron source. A variant of this concept would involve first irradiating this Pu plus minor actinide fuel by repeated recycling in a critical reactor before the final irradiation in a subcritical transmuter reactor. The small delayed neutron fraction of the minor actinides and the generally positive reactivity coefficient of fast reactors without 238 U dictates that these actinide destruction, or transmuter, reactors must remain well subcritical. The reactivity coefficient could be made negative by the addition of 238 U, which would allow the possibility of actinide destruction in critical fast reactors, but that would lead to the production of additional Pu and minor actinides by transmutation of 238 U, hence to a decreased net actinide destruction rate. 241 242 6 Fuel Burnup Fig. 6.18 Concentration of repository isotopes in HLW. Development of the PYRO separation technology would allow separation of Np, Am, and Cm along with Pu, all of which could be recycled in a metal-fuel fast reactor, resulting in a waste stream essentially free of actinides. However, it would be necessary to include 238 U in the fuel to avoid the safety problems mentioned in the preceding paragraph, which would reduce the net destruction rate of the actinides. Thus safety or net destruction rate constraints on transmutation of actinides in critical reactors could be relaxed by operating the reactors subcritical with a neutron source. Several studies of subcritical reactors driven by accelerator spallation neutron sources and a few studies of subcritical reactors driven by fusion neutron sources have predicted significantly higher levels of Pu, minor actinide, and/or long-lived fission product destruction than are predicted to be achievable in critical nuclear reactors. The optimum scenario for recycling Pu, minor actinides, and long-lived fission products in commercial thermal neutron spectrum reactors, in dedicated fast neutron spectrum reactors, and in subcritical transmuter reactors driven by neutron sources remains the subject of active investigation. The neutron spectrum in a subcritical reactor driven by a neutron source will depend more on the moderating and absorption properties, hence the material composition, of the subcritical reactor than on the energy spectrum of the source neutrons. Thus the material composition in the subcritical reactor can be optimized for the transmutation task at hand, without the criticality and safety constraints that would be present in a critical reactor. Figure 6.18 shows the mass, per metric tonne of uranium used in fabricating the original fuel, of two long-lived fission products and four actinides that would contribute significantly to the dose rates of spent fuel in a HLW repository at long times after discharge. The “Once Through Cycle” indicates LWR fuel sent directly 6.9 Transmutation of Spent Nuclear Fuel Fig. 6.19 Rate of production of decay heat. to a HLW repository after discharge from a once-through cycle. The “Single MOX Recycle” indicates the same LWR spent fuel, but with the plutonium separated out and recycled once in a LWR before being sent to the HLW repository, and with the uranium separated out and sent to a low-level-waste (LLW) repository. In the “IFR Fuel Cycle,” the original LWR spent fuel is separated into fission products (sent to HLW repository), uranium (sent to LLW repository) and transuranics (TRU) which are repeatedly recycled in a TRU-U metal fuel Integral Fast Reactor (with fission products removed and sent to a HLW repository). The presence of U in the IFR fuel is necessary to provide a prompt negative Doppler coefficient of reactivity, but also reduces the net TRU burnup because of the production of TRU by the transmutation of 238 U. If the IFR is operated subcritical to provide a large reactivity margin to prompt critical, it could be fueled with pure TRU metal fuel, in which case the net transmutation rate would be greater. The “FTWR Fuel Cycle” and “ATWR Fuel Cycle” indicate variants of the IFR Fuel Cycle in which the IFR is fueled with pure TRU and operated sub-critical with a D–T fusion and accelerator-spallation, respectively, neutron source. The small differences between the FTWR and ATWR are due to the better neutron utilization for actinide fission with Na coolant (IFR and ATWR) than with Li coolant (FTWR). Clearly, a single MOX recycle does not significantly reduce the mass of longlived radioactive material sent to the HLW repository. On the other hand, repeatedly recycling the actinides from the LWR spent fuel in an IFR (critical or sub-critical) can significantly reduce the actinide content in the material ultimately sent to the HLW repository. The HLW repository capacity will be limited by the amount of decay heat that can be removed passively. Thus, a more meaningful comparison among the above fuel 243 244 6 Fuel Burnup cycles is in terms of the decay heat at a given time after discharge of the material placed in the HLW repository, normalized by the total nuclear energy produced from the original LWR fuel in the respective fuel cycle. As shown in Fig. 6.19, the repository capacity required to achieve a given level of nuclear energy production can be reduced by a couple of orders of magnitude by repeatedly recycling the actinides from LWR spent fuel in a fast reactor, relative to depositing the LWR spent fuel directly in the repository. 6.10 Closing the Nuclear Fuel Cycle The discussions of the previous two sections on uranium utilization and transmutation of long-lived actinides in spent nuclear fuel are obviously related. The common thread is fissioning the transuranics to simultaneously recover their energy content and eliminate them as long-lived “waste.” The full utilization of the energy content of uranium can only be recovered by transmutation of 238 U into transuranics that can be fissioned. This “closing of the nuclear fuel cycle” has long been the ultimate goal of nuclear power development, but short-term economics, misguided reactions to proliferation and environmental concerns, and an apparently temporary stagnation in the development of nuclear power in the U.S. and some parts of Europe have until recently conspired to make realization of this goal but a dimly perceived possibility for the distant future. However, with the recent (2006) initiation of the Global Nuclear Energy Partnership to develop an international closed nuclear fuel cycle by development and deployment of technologies that enable recycling and consumption of spent nuclear fuel in a manner that promotes non-proliferation, enhances energy security and is environmentally responsible, this goal once again appears achievable. The Generation-IV reactor design studies to identify reactors and separations processes that could achieve this goal are discussed in Section 7.10. References 1 Physics of Plutonium Recycling, Vols. I–V, Nuclear Energy Agency, Paris (1995). 2 Management and Disposition of Excess Weapons Plutonium, National Academy Press, Washington, DC (1995). 3 R. A. Knief, Nuclear Engineering, Taylor & Francis, Washington, DC (1992), Chaps. 2 and 6. 4 R. G. Cochran and N. Tsoulfanidis, The Nuclear–Fuel Cycle: Analysis and Management, American Nuclear Society, LaGrange Park, IL (1990). 5 L. Koch, “Formation and Recycling of Minor Actinides in Nuclear Power Stations,” in A. J. Freeman and C. Keller, eds., Handbook of the Physics and Chemistry of Actinides, Vol. 4, Elsevier Science Publishers, Amsterdam (1986), Chap. 9. 6 S. H. Levine, “In-Core Fuel Management of Four Reactor Types,” in Y. Ronen, ed., CRC Handbook of Nuclear Problems Reactor Calculations II, CRC Press, Boca Raton, FL (1986). 7 A. G. Croff, ORIGEN2: A Revised and Updated Version of the Oak Ridge Isotope Generation and Depletion Code, ORNL-5621, Oak Ridge National Laboratory, Oak Ridge, TN (1980). 8 International Nuclear Fuel Cycle Evaluation, STI/PUB/534, International Atomic Energy Agency, Vienna (1980). 9 J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976), Chap. 15. Fission Product Partitioning and Transmutation, OECD/NEA, Paris (1999). 13 Proc. 1st–5th NEA International Exchange Meetings, OECD/NEA, Paris (1990, 1992, 1994, 1996, 1998). 14 D. C. Wade and R. N. Hill, “The Design Rationale of the IFR,” Prog. Nucl. Energy 31, 13 (1997). 15 L. J. Templin, ed., Reactor Physics Constants, 2nd ed., ANL-5800, Argonne National Laboratory, Argonne, IL (1963). 10 A. F. Henry, Nuclear-Reactor Analysis, MIT Press, Cambridge, MA (1975), Chap. 6. 16 A. Sesonske, Nuclear Power Plant Design Analysis, USAEC-TID-26241, U.S. Atomic Energy Commission, Washington, DC (1973). 11 National Research Council, Nuclear Wastes Technologies for Separations and Transmutation, National Academy Press, Washington, DC (1996). 17 N. L. Shapiro et al., Electric Power Research Institute Report, EPRI-NP-359, Electric Power Research Institute, Palo Alts, CA (1977). 12 First Phase P&T Systems Study: Status and Assessment Report on Actinide and 18 Oceanus, 20, Woods Hole Oceanographic Institute, Wood Hole, MA (1977). Problems 6.1. A reactor loaded initially with 125 kg of 93% enriched 235 U in the form of UO2 depletes in a constant neutron flux of φ = 5 × 1013 n/cm2 · s for one effective full power year. Assuming a thermal absorption cross section of 450 barns for 235 U, calculate the average fuel burnup in MWD/T. 6.2. Calculate the maximum enrichment at which a mixture of 235 U and 238 U will initially breed (i.e., the fissile concentration n25 + n49 will increase in time). Use σa25 = 700 barns, σa49 = 1050 barns, σγ28 = 8 barns, η5 = 2.08 and η49 = 2.12 and assume that 239 Pu is produced instantaneously by neutron capture in 238 U. * 6.3. Consider a thermal reactor with initial fuel composition 93% 235 U and 7% 238 U and fuel density 18.9 g/cm3 with an initial thermal neutron flux of 3 × 1014 n/cm2 · s. Assume a flux disadvantage factor of 2. Write a computer code to calculate the depletion of 235 U and the buildup of 239 Pu over the first 2000 h of operation, assuming operation at constant power. Estimate from perturbation theory the reactivity decrease * Problems 3 and 4 are longer problems suitable for take-home projects. 245 246 6 Fuel Burnup 6.4.* 6.5. 6.6. 6.7. 6.8. 6.9. 6.10. 6.11. associated with the 235 U depletion and 239 Pu buildup over 2000 h of operation. Plot your results as a function of time. Estimate from perturbation theory the equilibrium xenon and samarium reactivity worth and the reactivity worth of the other fission products as a function of time in the reactor of Problem 6.1. Assume that the fuel occupies 80% of the core fp and that γ fp σa = 50 barns per fission. Plot your results as a function of time. A 235 U-fueled nuclear reactor operates with a thermal flux level of 1 × 1014 n/cm2 · s. The reactor has been operating at constant power level for 2 weeks when it becomes necessary to scram the control rods to shut down the core. After detailed investigation it is determined that the scram signal was erroneous and it is now necessary to return the reactor to full-power operation: 12 h has passed since shutdown. The control rods are withdrawn to the critical prescram position and the reactor is brought to temperature, but the reactor is not critical. How much further must the control rods be withdrawn to achieve criticality if the control rod bank worth is ρ = 0.001 cm−1 . Derive the equations that determine the time dependence of the samarium concentration in a reactor that has achieved equilibrium samarium conditions at a flux φ0 when the flux is changed to φ1 . Calculate the initial excess reactivity needed for the reactor of Problem 6.2 to have a cycle lifetime of 1.0 years operating at a flux level of 5 × 10l3 n/cm2 · s. Calculate the nuclear (fission) heating density in a PWR UO2 fuel element (density 10 g/cm3 , 4% enriched) operating in a thermal neutron flux of 5 × 1013 n/cm2 · s. Calculate the α-decay heating due to plutonium at the end-of-cycle for each recycle core loading indicated in Table 6.5. A thermal reactor loaded with 100,000 kg of 3% enriched UO2 depletes in a constant thermal neutron flux of 5 × 1013 n/cm2 · s for 1 year. Using a thermal absorption cross section of 500 barns and a capture-to-fission ratio of α = 0.2 for 235 U and a density of 10 g/cm3 for UO2 , calculate the average fuel burnup in MWd/T. A 235 U-fueled reactor has been operated at a thermal flux level of 5 × 1012 n/cm2 · s for 2 months, when the power level is reduced by one-half for 10 h, then returned to full power. Calculate the reactivity worth of xenon just before the reactor is shut down; 10 h later, before it is returned to full power; Problems 6.12. 6.13. 6.14. 6.15. 6.16. 6.17. 6.18. 6.19. 6.20. and then again after it has been operating at full power for 10 h. Repeat the calculation of Problem 6.11 for full power thermal neutron flux levels of 1 × 1013 , 5 × 1013 , and 1 × 1014 n/cm2 · s. The equilibrium concentration of 149 Sm is independent of power level, and when the reactor is shut down, the 149 Sm concentration increases. Can the 149 Sm concentration ever be lower than the equilibrium concentration, once that concentration has been attained? Calculate the equilibrium and peak xenon concentrations in cores fueled with 233 U, 235 U, and 239 Pu, all operating at a thermal flux level of 1 × 1014 . A uniform bare cylindrical reactor, containing an initial loading of 125 kg of 235 U, operates until the maximum local 235 U depletion reaches 50%. Estimate the total fission energy release from the core. Calculate and plot the activity (Ci/tonne fuel) and the toxicity (cancer dose/tonne fuel) of 99 Tc, 129 I, 90 Sr, and 137 Cs in spent fuel from a LWR from the time of discharge to 104 years later. Calculate the toxicity (cancer dose/tonne fuel) of the equilibrium concentrations of the transuranic isotopes given in Table 6.12 for continuously recycled spent fuel in fast and thermal reactor spectra. Calculate the change in isotopic composition of weapons-grade plutonium that is irradiated in a thermal neutron flux of 1014 n/cm2 · s for 1 year. A thermal reactor fueled with 235 U and 232 Th in the ratio 1:20 is operated for 1 year with a neutron flux of 8 × 1013 n/cm2 · s. Calculate the concentrations of 233 U and 235 U, in terms of the initial 235 U concentration, at the end of the year. What is the annual conversion ratio? Calculate the energy content per unit mass of the original fuel loadings for the reactors in Table 6.7. Calculate the fraction of this energy content that is released by fission in a single cycle. 247 249 7 Nuclear Power Reactors As of 2000, there are 434 central station nuclear power reactors operating worldwide to produce 350,442 MWe of electrical power. Of this number, 252 are pressurized water reactors (PWRs), 92 are boiling water reactors (BWRs), 34 are gas-cooled reactors (GCRs) of all types, 39 are heavy water-cooled reactors of all types (mostly CANDUs), 15 are graphite-moderated light-water pressure tube reactors (RBMKs), and 2 are liquid-metal fast breeder reactors (LMFBRs). The general physics-related characteristics of such reactors are described in the following sections. To be quantitative, specific reactors that produce 900 to 1300 MWe (650 MWe in the case of CANDUs) were chosen, but it should be noted that reactors of each type can vary greatly in size and power output, so the numbers should be understood to be only representative. In addition to the central station power reactors mentioned above, there are more than 100 pressurized water naval propulsion reactors in the U.S. fleet (plus others in foreign fleets) and numerous research and special purpose reactors of various types worldwide. 7.1 Pressurized Water Reactors Pressurized water reactors (PWRs) were first developed in the United States based on experience from the naval reactor program. The first commercial electric power– producing unit started operation at Shippingport, Pennsylvania in 1957. The PWR is now widely distributed worldwide. The basic structure of the PWR core is the approximately 20 cm × 20 cm × 4 m high fuel assembly shown in Fig. 7.1, consisting of an array of zircaloy-clad UO2 fuel pins, or rods, of about 1 cm diameter. The enrichment varies from about 2 to 4% or more, depending on the burnup objective. A typical fuel assembly may consist of a 17 × 17 array of fuel pins of about 1 cm diameter. The coolant flows in an open lattice structure which permits some flow mixing and is under sufficient pressure that no boiling occurs under normal operation. Long-term reactivity control is provided by adjustment of the boric acid content in the coolant. The soluble poison concentration decreases with fuel burnup to compensate fuel reactivity loss and must be reduced to compensate 135 Xe and 250 7 Nuclear Power Reactors 149 Sm buildup following reactor startup. Boron addition and dilution may be used to minimize control rod motion for startup and shutdown. Soluble poisons make a positive contribution to the moderator temperature coefficient of reactivity (an increase in temperature reduces the absorption cross section), so their maximum concentration is limited, and fixed burnable poisons are used to reduce the control requirements that must be met by adjustment of the boric acid concentration. Burnable poisons consists of separate shim rods substituted for a fuel rod in the fuel assembly. These rods may consist of borosilicate glass rods with stainless steel cladding or B4 C pellets in an Al2 O3 matrix with zircaloy cladding. The shim rods burn out as the fuel depletes, which constitutes a positive reactivity contribution to compensate the negative reactivity contribution of fuel depletion, thus reducing the requirement for adjustment of the boric acid concentration. Because of the relatively short migration length (about 6 cm) of thermal neutrons in a PWR, the active control rods must be distributed. Short-term and rapid insertion (scram) reactivity control is provided by an assembly of full-length control rods driven down into the fuel assembly. For example, the control rod assembly for a 17 × 17 pin lattice consists of 24 control fingers connected by a spider, as shown in Fig. 7.1. The control rod material is either B4 C or a Ag–In–Cd mixture of somewhat weaker absorbers that produces less flux peaking upon rod withdrawal. “Part-length” rods in which only the lower 25% or so contains poison are used for controlling the axial flux distribution, which is necessary to control axial xenon oscillations (Chapter 16) as well as to minimize axial power peaking. Full-length control rods are normally designated as regulating rods, used for the normal short-term reactivity adjustments that cannot be handled by adjustment of the boric acid concentration, and shutdown or scram rods, which are held out of the core to be available for a rapid negative reactivity insertion if required for safety or a more gradual negative reactivity insertion required for normal shutdown. A typical distribution of control rods among the assemblies in a PWR core is shown in Fig. 7.2. About 190 to 240 fuel assemblies containing 90,000 to 125,000 kg of UO2 constitute a typical PWR core, which is about 3.5 m in diameter and 3.5 to 4.0 m high and is located inside a pressure vessel, as shown in Fig. 7.3. Coolant typically enters the pressure vessel near the top, flows downward between the vessel and the core, is distributed at the lower core plate, flows upward through the core, and exits the vessel at the top. The coolant, which is pressurized to about 15.5 MPa (2250 psi), typically enters the vessel with a temperature of about 290◦ C and exits at about 325◦ C. 7.2 Boiling Water Reactors Boiling water reactors (BWRs) were first developed in the United States and are now found worldwide. The physics of BWRs is similar in many respects to that of PWRs. The basic structure of the BWR core is an approximately 7.2 Boiling Water Reactors Fig. 7.1 Fuel assembly for a pressurized water reactor. (Courtesy of Westinghouse Electric Corporation.) 251 252 7 Nuclear Power Reactors Fig. 7.2 Representative control element pattern in a pressurized water reactor. (Courtesy of ABB Combustion Engineering, Inc.) 14 cm × 14 cm × 4 m high fuel assembly (Fig. 7.4) consisting of an 8 × 8 array of zircaloy-clad UO2 fuel pins, or rods, of about 1.3 cm diameter. The enrichment varies from 2 to 4% 235 U. The 8 × 8 fuel pin array is surrounded by a zircaloy fuel channel to prevent cross-flow between assemblies. A group of four fuel assemblies plus an included cruciform control rod constitutes a fuel module, out of which a typical BWR core is built up, as indicated in Fig. 7.5. Fuel pins of different enrichment are loaded into each assembly. Fuel pins of lower enrichment are located next to the control rod to suppress the flux peaking that would otherwise occur when the control rod was withdrawn, leaving a substantial water gap. The other pins are arranged to flatten the power distribution within the assembly. Long-term reactivity changes to compensate fuel depletion 7.2 Boiling Water Reactors Fig. 7.3 Pressurized water reactor. (Courtesy of Westinghouse Electric Corporation.) and reactivity changes needed for large power level changes are provided by the B4 C cruciform control rods, which are driven up from the bottom of the core because the reactivity worth is greater with the single-phase coolant in the lower part of the core than with the two-phase coolant in the upper part. Long-term compensation of the negative reactivity associated with fuel depletion is provided by mixing Gd2 O3 uniformly with the UO2 in several fuel pins in each assembly to provide a positive reactivity contribution as it burns out. 253 254 7 Nuclear Power Reactors Fig. 7.4 Fuel assembly for a boiling water reactor. 1, top fuel guide; 2, channel fastener; 3, upper tie plate; 4, expansion spring; 5, locking tab; 6, channel; 7, control rod; 8, fuel rod; 9, spacer; 10, core plate assembly; 11, lower tie plate; 12, fuel support piece; 13, fuel pellets; 14, end plug; 15, channel spacer; and 16, plenum spring. (Courtesy of General Electric Company.) 7.3 Pressure Tube Heavy Water–Moderated Reactors Fig. 7.5 Four-assembly fuel module for a boiling water reactor. (Courtesy of General Electric Company.) Short-term reactivity control is provided by recirculation flow and by control rods. Because of the negative coolant/moderator temperature coefficient of reactivity, coolant flow rate can be increased to decrease coolant temperature and the amount of boiling, making neutron moderation more effective and thus increasing reactivity. This causes the power level and the coolant temperature to increase, which in turn decreases the reactivity, until the reactor is again critical at a higher power level. Decreasing the coolant flow rate reduces the power level by a similar mechanism. Typically, about 750 fuel assemblies containing about 140,000 to 160,000 kg of UO2 constitute a BWR core, which is similar in size to a PWR core and is located inside a pressure vessel, as shown in Fig. 7.6. Coolant enters the vessel at about 7.2 MPa (1000 psi), flows downward between the vessel wall and the shroud, is distributed by the core plate, flows upward through the core and upper structure, and exits the core as steam at about 290◦ C. About 30% of the coolant flow is recirculated, which has the net effect of increasing the total coolant flow rate in the core. 7.3 Pressure Tube Heavy Water–Moderated Reactors The use of heavy water and online refueling to maintain criticality with natural uranium fuel is fundamental to CANDU reactors, which are pressure tube heavy water–moderated reactors developed in Canada but now are located in several other countries. The basic structure of the CANDU core is the fuel bundle shown in Fig. 7.7, which contains natural UO2 in 37 zircaloy-clad fuel pins about 1.3 cm in diameter and 49 cm long which are separated with spacers. Twelve fuel bundles are placed end to end in a pressure tube through which flows pressurized (10 MPa, 1450 psi) D2 O. The reactor core consists of 380 fixed calandria tubes in a vessel 255 256 7 Nuclear Power Reactors Fig. 7.6 Boiling water reactor. 1, vent and head spray; 2, steam dryer lifting lug; 3, steam dryer assembly; 4, steam outlet; 5, core spray inlet; 6, steam separator assembly; 7, feedwater inlet; 8, feedwater sparger; 9, low-pressure coolant injection inlet; 10, core spray line; 11, core spray sparger; 12, top guide; 13, jet pump assembly; 14, core shroud; 15, fuel assemblies; 16, core blade; 17, core plate; 18, jet pump/recirculation water inlet; 19, recirculation water outlet; 20, vessel support skirt; 21, shield wall; 22, control rod drives; 23, control rod drive hydraulic lines; and 24, in-core flux monitor. (Courtesy of General Electric Company.) 7.3 Pressure Tube Heavy Water–Moderated Reactors Fig. 7.7 Fuel assembly for a CANDU pressure tube heavy water reactor. (Courtesy of Atomic Energy of Canada, Ltd.) filled with D2 O moderator, as shown in Fig. 7.8. A pressure tube containing the 12 fuel bundles is loaded into each calandria tube, resulting in a core loading of about 100,000 kg of natural UO2 . The coolant enters each pressure tube at about 265◦ C and exits at about 310◦ C. A typical CANDU core is about 7 m in diameter and about 4 m high. On-line refueling is the primary means of long-term reactivity control in CANDU reactors. This is augmented by addition of soluble poison to the moderator D2 O and by the use of boron and gadolinium as burnable poisons admixed with the fuel. Because the D2 O in the reactor vessel, not the D2 O coolant in the pressure tubes, is the primary moderator, the usual negative coolant temperature coefficient of reactivity present in PWRs and BWRs is not present in the CANDU, and in fact the temperature coefficient of reactivity tends to be positive. This requires a much more precise active reactivity control system than for PWRs and BWRs. Reactivity control in each of 14 chambers is achieved by controlling the amount of H2 O (which is a poison in a D2 O system) in response to local neutron flux detector measurements. Control rods are also employed. Adjuster rods are used for flattening the power distribution and for short-term reactivity adjustments. Four cadmium rods clad in stainless steel are located above the core, which may be used to supplement the adjuster rods in achieving reactivity control or dropped to effect a rapid shutdown, or scram. A backup shutdown system consists of injection of a gadolinium nitrate solution into the moderator. 257 258 7 Nuclear Power Reactors Fig. 7.8 CANDU pressure tube heavy water reactor. (Courtesy of Atomic Energy of Canada, Ltd.) 7.4 Pressure Tube Graphite-Moderated Reactors The world’s first commercial nuclear electricity was generated near Moscow in 1954 by a graphite-moderated pressure tube light water reactor which evolved to 7.4 Pressure Tube Graphite-Moderated Reactors Fig. 7.9 Fuel assembly for a RMBK pressure tube graphite reactor. (From Ref. 8.) the reactor generally known by the acronym RBMK from the Russian for “highpower pressure tube reactor.” Reactors of this type are located in the countries of the former Soviet Union. The basic structure of the RBMK core is the fuel channel tube, made of zirconium alloyed with 2.5% niobium, shown in Fig. 7.9. Each channel tube consists of two fuel strings, which are separately cooled with H2 O at 7.2 MPa, which enters at 270◦ C and exits at 284◦ C, placed end to end. Each fuel string contains 1.8 to 2.0% enriched UO2 in 18 fuel pins about 1.3 cm in diameter and 3.6 m long, separated with spacers. Each channel tube is placed vertically into a square graphite block 0.25 m on a side and 7 m high. The graphite blocks, 1661 containing a fuel channel tube and 222 containing control rod channels, are set side by side to form an upright cylinder 12.2 m in diameter containing about 200,000 kg of UO2 . Since the migration length in graphite is about 60 cm, the core is very loosely coupled and subject to flux tilting. Furthermore, since the neutron moderation is provided by the graphite, the coolant temperature coefficient of reactivity is positive because the effect of increased coolant temperature and reduced coolant density is to reduce the coolant absorption cross section. As a consequence, the RMBK reactor is inherently unstable to power oscillations and it is necessary to control the power distribution region by region. Two hundred and eleven cylindrical B4 C control rods, with graphite extenders to enhance rod effectiveness by displacing 259 260 7 Nuclear Power Reactors H2 O that would otherwise fill the rod channel when the rod was withdrawn, are dispersed in the core. Of these, 24 are normally withdrawn from the core to be available to produce rapid shutdown. An additional 24 short absorbing rods that enter from below are used to control axial xenon oscillations (Chapter 16) and to reduce axial power peaking. With a fresh fuel loading, up to 240 additional control rods must displace fuel in the tubes in order to hold down reactivity. These control channel tubes are replaced with fuel channel tubes as fuel burnup decreases reactivity. 7.5 Graphite-Moderated Gas-Cooled Reactors The first man-made sustained fission chain reaction took place in a pile of graphite in Chicago—with air cooling—which was the prototype for the first experimental and production reactors. The original gas-cooled power reactors developed in France and England used CO2 as a coolant and graphite moderator. The original MAGNOX reactors consisted of natural uranium bars clad in a low- neutronabsorbing magnesium alloy known as magnox which were placed in holes in graphite blocks through which the CO2 coolant flows at 300 psi, leaving the core at about 400◦ C. A typical MAGNOX core is about 14 m in diameter and 8 m high. To achieve higher coolant outlet temperatures (650◦ C), the subsequent advanced gascooled reactors (AGRs) operate at 600 psi, which requires that the cladding consist of a material that can operate at higher temperature, which in turn requires that the fuel be enriched. The AGR fuel element consists of 36 tubes made up of pellets of 2.3% enriched UO2 in stainless steel cladding, which are ribbed to improve heat transfer, as shown in Fig. 7.10, encased in a graphite sleeve, and inserted in holes in graphite blocks. Excessive corrosion in piping and steam generators have led to the abandonment of CO2 as a coolant; most advanced gas-cooled reactors use helium as a coolant. As an example of a modern gas-cooled reactor, we consider the high-temperature gas-cooled reactor (HTGR). The basic structure of the HTGR core is a hexagonal graphite block containing small channels for stacks of fuel pins and for coolant flow, as shown in Fig. 7.11. The fuel consists of coated microspheres of 93% enriched UC/ThO2 contained in fuel pins about 1.6 cm in diameter and 6 cm long. About 490 fuel assemblies, each with 6.3 m active fuel height, are placed upright side by side to form a core that has a diameter of 8.4 m and contains 1,720 kg of U and 37,500 kg of Th, as shown in Fig. 7.12. The fuel assemblies are arranged in rings of six about a control assembly. Long-term reactivity control is provided by B4 C loaded into carbon rods which may be loaded into the corner locations of each fuel assembly to serve as burnable poison. Short-term reactivity control is provided by pairs of control rods that can be inserted into the two larger channels in special control assemblies. HTGRs have been deployed only on a limited scale. 7.6 Liquid-Metal Fast Breeder Reactors Fig. 7.10 Fuel assembly for advanced gas reactor. (From Ref. 4; used with permission of Taylor & Francis/Hemisphere Publishing.) 7.6 Liquid-Metal Fast Breeder Reactors The first generation of electricity from nuclear fission took place at the light bulb level in 1952 in a liquid-metal fast breeder reactor an (LMFBR), the EBR-1 in Idaho. Several LMFBRs have been operated since then, but this reactor type has not yet been deployed on a substantial scale. The physics of the LMFBR, which has a fast neutron spectrum, differs significantly from the physics of the previously discussed reactors, all of which have a thermal neutron spectrum. The basic structure of a modern LMFBR core is the fuel assembly, as indicated in Fig. 7.13. The primary fissile nuclide for fast breeders is 239 Pu, and the primary fertile nuclide is 238 U. The fuel assembly consists of about 270 fuel pins containing 10 to 30% Pu in PuO2 –UO2 in small pellet form encased in stainless steel cladding. The pins, which are about 0.9 cm in diameter and 2.7 m long, are wrapped in wire to maintain interpin spacing and placed within a stainless steel tube. The flow of liquid sodium is directed by the channel around the array. About 350 such 261 262 7 Nuclear Power Reactors Fig. 7.11 Fuel assembly for a high-temperature gas-cooled reactor. (Courtesy of General Atomics Company.) Fig. 7.12 High-temperature gas-cooled reactor. (Courtesy of General Atomics Company.) 7.6 Liquid-Metal Fast Breeder Reactors Fig. 7.13 Fuel assemblies for a liquid-metal fast breeder reactor. (Courtesy of Nuclear Engineering International.) assemblies makes up the core of an LMFBR. Another 230 similar assemblies, but with only UO2 or with a lower Pu content, are placed in a blanket around the core, as shown in Fig. 7.14. The total mass of PuO2 /UO2 is about 32,000 kg. A typical LMFBR core is about 1 m high and 2 m in diameter. Reactivity control is achieved by control bundles of B4 C rods which replace fuel assemblies, located in roughly inner and outer (radially) concentric circles. Typically, the bundles are separated into two groups, each of which is capable of shutting down the core. The fuel depletion reactivity effect of thermal reactors is reversed in LMFBRs, which produce more fissile nuclei than they consume. In addition, the negative reactivity effects of fission products, which are primarily thermal neutron absorbers such as samarium and xenon, are much less in an LMFBR than in a thermal reactor. Like the RBMK and CANDU pressure tube reactors, in which the moderator is separate from the coolant, the LMFBR tends to have a positive coolant temperature coefficient of reactivity, but for a different reason. Reduction of sodium density hardens the neutron spectrum, which results in a lower capture-to-fission ratio in the fuel and reduces the number of neutrons absorbed in the large 23 Na resonance in the keV energy range. The fast neutron spectrum means a shorter neutron lifetime (the mean time from fission to absorption or leakage of the neutron) than in a thermal reactor because the neutron is absorbed or leaks before it slows down in an LMFBR ( ≈ 10−6 s for LMFBRs as contrasted to 10−4 to 10−5 s for thermal reactors). This implies a more rapid response to superprompt-critical (ρ > β) reactivity insertions. Furthermore, the prompt-critical reactivity level (ρ = β) with plutonium in a fast spectrum (β = 0.0020 for 239 Pu and β = 0.0054 for 241 Pu) is 263 264 7 Nuclear Power Reactors Fig. 7.14 Super Phenix liquid-metal fast breeder reactor. (From Ref. 6; used with permission of CRC Press.) smaller than with 235 U in a thermal spectrum (β = 0.0067). On the other hand, the reactivity worth of perturbations such as inadvertent control rod withdrawal is generally smaller in a fast spectrum because of the smaller value of the absorption cross section for fast than for thermal neutrons. A variant of the LMFBR is the advanced liquid metal reactor (ALMR), which employs a Pu/U metal alloy fuel. The configuration consists of pairs of core modules constituting a 606 MWe power block; a core can be built up of 1, 2 or 3 power blocks. The ALMR design is based on the integral fast reactor (IFR) actinide recycle concept (see Section 7.11). An IFR would generate less actinide “waste” than do light water reactors and can recycle its own actinide waste and the actinide waste of light water reactors to recover energy which would otherwise be lost, at the same time reducing the waste disposal burden. The passive safety features of the ALMR allow extreme off-normal transients—loss of primary coolant flow without scram and loss of heat removal by the intermediate system without scram—with benign consequences to the reactor core. As will be discussed in Section 8.5, tests have shown that the ALMR can undergo these extreme events without damage. 7.7 Other Power Reactors 7.7 Other Power Reactors There are also a number of other reactors, most of which have been designed to achieve enhanced production of fissile nuclei by neutron transmutation, which have been developed through the demonstration stage but not yet implemented on a significant scale as power reactors. Two of these are basically modifications of conventional thermal light water reactors. The light water breeder reactor (LWBR) operates on a 232 Th–233 U cycle, which is more favorable than the 238 U–239 Pu cycle for the production of fissile nuclei by thermal neutron transmutation (Chapter 6). The spectral shift light water reactor operates with a mixed D2 O–H2 O coolant to achieve a slightly harder neutron spectrum to enhance the transmutation of 238 U into fissile plutonium early in the cycle and reduces the D2 O/H2 O ratio with burnup to soften the spectrum and increase the reactivity to offset reactivity loss due to fuel depletion. There are also two graphite-moderated thermal reactors designed to achieve enhanced production of fissile nuclei. The thermal molten salt breeder reactor (MSBR), which operates on the 232 Th–233 U cycle with the fuel contained in a circulating molten salt (typically, LiF–BeF2 –ThF4 –UF4 ), which also serves as the heat removal system, achieves additional enhancement of neutron utilization for fissile production by continuous removal of fission products from the recirculating fuel. The pebble bed reactor, a variant of the helium-cooled HTGR, contains the 232 Th– 233 U fuel in 6-cm-diameter graphite spheres that can be poured into and drained from a core hopper. Designs have been developed for gas-cooled fast reactors (GCFRs) which are similar to LMFBR designs, with PuO2 /UO2 fuel pins clad with stainless steel. The fuel pins are ribbed to enhance heat transfer and their spacing is about twice that of an LMFBR assembly. 7.8 Characteristics of Power Reactors Typical parameters relevant to power production are summarized for a number of reactor types in Table 7.1. 7.9 Advanced Generation-III Reactors The reactors discussed in the previous sections of this chapter may be considered to be of the first and second generations of nuclear reactors, the designs of which matured in the 1960–1990s period. Designs of a third generation of power reactors have evolved starting in the 1990s, and the following Generation III reactors 265 266 7 Nuclear Power Reactors Table 7.1 Representative Parameters Relevant to Power Production for the Major Reactor Types Type MAGNOX AGR CANDU PWR BWR RBMK LMFBR Thermal Power (MWt) Core Diameter (m) Core Height (m) Average Power Density (MW/m3 ) Linear Fuel Rating (kW/m) Average Fuel Burnup (MWd/T) 1875 1500 3425 3800 3800 3140 612 17.37 9.1 7.74 3.6 5.0 11.8 1.47 9.14 8.3 5.94 3.81 3.81 7.0 0.91 0.87 2.78 12.2 95.0 51.0 4.1 380 33.0 16.9 27.9 17.5 19.0 14.3 27.0 3,150 11,000 26,400 38,800 24,600 15,400 153,000 Source: Date from Ref. 4; used with permission of Taylor & Francis/Hemisphere Publishing. are now being deployed over roughly the 1995–2015 period. These designs have, of course, benefited from the extensive operating experience with the previous generation of power reactors. In the U.S., Europe and Japan the designs also have been driven by the major objective to incorporate passive safety features to insure safety without reliance on active control actions. In Europe, there has also been a separate emphasis on accommodating mixed oxide (MOX) fuels to take a first step towards closing the nuclear fuel cycle by recycling plutonium from spent nuclear fuel. Advanced Boiling Water Reactors (ABWR) The advanced BWR designs (and the advanced PWR designs discussed in the next section) are based on conventional UO2 fuel assemblies with negative temperature coefficients. Passive safety is enhanced by designing so that, in the event of a loss of coolant accident, the core would be flooded with enough water to provide cooling for three days, without operator action (present designs require operator response in about 20 minutes). The first two ABWRs were built in Japan and began operation in 1996 and 1997. These reactors had active core heights of 3.71 m and diameters of 5.16 m and were fueled with 3.2% enriched UO2 designed to achieve 32,000 MWd/t discharge burnup. Operating at a power density of 50.6 MW/m3 these reactors produce about 1350 MWe. The ABWRs have three independent and redundant safety systems which are physically and electronically isolated. These systems have the capability to keep the core covered at all times. This capability plus the substantial thermal margin in the fuel design is predicted to significantly reduce the frequency of transients which will lead to a scram (to less than one per year). Plant response to a LOCA has been completely automated, and operator action is not required for 72 hours (the same capability specified for passively safe plants). A US evolution of the ABWR concept, the Economically Simplified BWR (ESBWR) relies on natural circulation and passive safety features to simplify the design and enhance 7.9 Advanced Generation-III Reactors Fig. 7.15 Typical initial core loading pattern for Advanced PWR. performance in a 1560 MWe plant with 4.2% enriched UO2 fuel that is designed to achieve 50,000 MWd/t burnup. Advanced Pressurized Water Reactors (APWR) Advanced PWRs are being developed in Europe (EPR [1600 MWe class]), the US (AP-600 [600 MWe class] and AP-1000 [1000 MWe class]), Japan (APWR [1500 MWe class]) and Korea (APR-1400 [1400 MWe class]). An EPR is presently (2006) under construction in Europe, and there are plans to construct other APWRs worldwide during the coming decade. The EPR will produce 4500 MWt in 241 standard 17 × 17 array PWR fuel assemblies each containing 265 fuel rods. The fuel rods are composed of a stack of sintered pellets of UO2 or mixed U–Pu-oxide (MOX) enriched up to 5%, with or without Gd2 O3 burnable poison, within a hermetically sealed zirconium alloy cladding. The fuel rods are 0.95 cm outside diameter and 4.2 m long, and the cladding thickness is 0.57 mm. The reactor core has an active height of 4.2 m and an equivalent diameter of 3.77 m. Fuel cycle lengths up to 24 months, with in-out or out-in fuel management, are possible. A typical initial core loading pattern is indicated in Fig. 7.15. 267 268 7 Nuclear Power Reactors Soluble boron concentration in the coolant is varied to control relatively slow reactivity changes, including the effects of fuel burnup. The fast shutdown control system comprises 89 control assemblies, each located within a fuel assembly and consisting of 24 absorber rods fastened to a common driver assembly. The absorber rods consist of Ag, In, Cd-alloy and sintered B4 C pellets. Advanced Pressure Tube Reactor The Canadian CANDU line of reactors described in Section 7.3 has been improved in several respects, most notably from the reactor physics point of view by replacement of the D2 O coolant with H2 O to achieve a negative void reactivity coefficient. The new ACR-700 reactor will produce 700 MWe. Modular High-Temperature Gas-Cooled Reactors (GT-MHR) Two helium-cooled, modular, thermal reactor concepts are under development. These two concepts, while quite different in configuration, both take advantage of the heat capacity of a large amount of graphite to achieve passive safety, take advantage of triple coated TRISO fuel particles to confine fission products, and would build up full scale power plants from modular units. The Pebble Bed Modular Reactor (PBMR) illustrated in Fig. 7.16 consists of a vertical steel pressure vessel lined with a layer of graphite bricks which reflect neutrons and provides a large heat capacity. Control rods are inserted downward into vertical holes in the graphite reflector. The reactor core is 3.7 m in diameter and 9.0 m high. The core consists of an inner zone of graphite spheres, which serve as a neutron moderator and to provide heat capacity, and an outer annular zone that contains about 370,000 tennis ball size fuel pebbles. Helium flows downward around the fuel spheres. This reactor module would produce 110 MWe. Each 60 mm fuel sphere, or pebble, is coated with a 5 mm thick graphite layer and contains about 15,000 coated TRISO fuel particles each 0.92 mm in diameter, as illustrated in Fig. 7.17. The central kernel of the TRISO particle contains UO2 enriched up to 8%, surrounded by a porous carbon buffer layer to contain the gaseous fission products and then by two pyrolytic graphite layers and a SiC structural layer which prevent release of fission products. Criticality is maintained during PBMR operation by removing depleted fuel pebbles from the bottom of the reactor and replenishing with fresh or slightly burned fuel pebbles at the top of the core. The burnup of fuel spheres leaving the core would be measured, those exceeding the reference burnup would be removed to storage and the others would be recycled (about 10 times). The Gas Turbine-Modular Helium Reactor (GT-MHR) is based on an extension of the HTGR prismatic designs discussed in Section 7.5 to use coated particle TRISO fuel, which is projected to lead to fuel burnup > 100,000 MWd/t, coupled to a high efficiency direct Brayton cycle gas turbine. Each modular reactor unit would produce 600 MWth which would be converted at 48% efficiency to 286 MWe. 7.10 Advanced Generation-IV Reactors Fig. 7.16 Pebble Bed Modular Reactor. In the GT-MHR, somewhat smaller TRISO fuel particles are mixed with a carbonaceous matrix and formed into cylindrical fuel compacts 13 mm in diameter and 51 mm long which are loaded into fuel channels in hexagonal graphite fuel elements 793 mm long by 360 mm across flats. The hexagonal fuel elements are stacked 10 elements high into 102 columns to form an annular core. Graphite reflector blocks are place inside and outside of the annular core to reflect and moderate neutrons and to provide adequate heat capacity to maintain fuel temperatures below damage limits during a loss-of-coolant-accident without the need for corrective operator actions. 7.10 Advanced Generation-IV Reactors The ongoing (2006) international Generation-IV studies are intended to identify reactors and associated separations technologies that will make a major step to- 269 270 7 Nuclear Power Reactors Fig. 7.17 Pebble Bed TRISO Fuel Sphere Cross Section. wards closing the nuclear fuel cycle. The specific goals are: (1) long term availability of systems that realize effective uranium utilization; (2) minimization of the discharged nuclear waste; (3) safety, reliability, a very low likelihood and degree of core damage, and no need for off-site emergency response; (4) unattractive targets for the diversion of weapons-usable materials and acts of terrorism; and (5) a lifecycle cost advantage over and a level of financial risk comparable to other energy systems. Six reactor systems were identified for the studies in order to provide some redundancy and complementarity in the capability to satisfy the three missions (electricity production, hydrogen and process heat production, and actinide management), to attain the performance objectives and to attract commercial deployment consistent with the national priorities of the countries involved. The six reactor systems chosen were (1) the Gas-cooled Fast Reactor (GFR), (2) the Lead-cooled Fast Reactor (LFR), (3) the Sodium-cooled Fast Reactor (SFR), (4) the Very High Temperature gas-cooled Reactor (VHTR), (5) the Super-Critical Water Reactor (SCWR), and (6) the Molten Salt Reactor (MSR). The salient features and objectives of these reactors are given in Table 7.2. Gas-Cooled Fast Reactors (GFR) The GFR system features a fast neutron spectrum, helium coolant and a closed fuel cycle. The potential for high outlet helium temperature would enable the GFR to meet the hydrogen and electricity production missions, and the fast neutron spectrum would facilitate achieving the actinide fissioning mission. Core config- 7.10 Advanced Generation-IV Reactors Table 7.2 Features and Objectives of GEN-IV Reactor Designs Reactor Coolant Neutron Spectrum Electric Mission H-Prod Mission Actinide Mission Earliest Deploy GFR LFR MSR SCWR SFR VHTR Gas Lead Molten salt Water Sodium Gas Fast Fast Epi-thermal Thermal Fast Thermal X X X X X X X X X X X (fast option) X 2025 2025 2025 2025 2020 2020 X urations may be based on assemblies of prismatic blocks (as for the HTGR and GT-MHR) or pin or plate fuel elements. Composite ceramic fuel, coated TRISO fuel particles and ceramic-clad dispersion fuel are being considered. The GFR can use a direct Brayton cycle helium turbine for electricity production or can use its process heat for the thermochemical production of hydrogen. By full recycling of actinides in a fast spectrum, the GFR would minimize the discharge of long-lived waste to HLW repositories and could maximize the utilization of the energy content of uranium (including depleted U) by transmutation of 238 U into fissionable transuranics. The current (2006) reference design produces 288 MWe. Lead-Cooled Fast Reactors (LFR) The LFR system features a fast neutron spectrum, liquid lead or lead-bismuth eutectic coolant and a closed fuel cycle. The potential for high outlet liquid metal temperature (up to 850◦ C) would enable the LFR to meet the hydrogen and electricity production missions, and the fast neutron spectrum would facilitate achieving the actinide fissioning mission. Metal and nitride-based dispersion fuels, containing uranium and transuranics, would be cooled by natural convection of the liquid metal coolant. By full recycling of actinides in a fast spectrum, the LFR would minimize the discharge of long-lived waste to HLW repositories and could maximize the utilization of the energy content of uranium (including depleted U) by transmutation of 238 U into fissionable transuranics. At present (2006), a range of plant sizes are being considered, including a 1200 MWe plant, a modular system rated at 300–400 MWe, and a “battery” rated at 50–150 MWe that features a 15–20 year refueling interval and a replaceable core cassette. By full recycling of actinides in a fast spectrum, the LFR would minimize the discharge of long-lived waste to HLW repositories and could maximize the utilization of the energy content of uranium (including depleted U) by transmutation of 238 U into fissionable transuranics. Molten Salt Reactors (MSR) The MSR system features an epi-thermal neutron spectrum, a circulating molten salt fuel-coolant mixture and a full actinide recycle fuel cycle. The potential for 271 272 7 Nuclear Power Reactors coolant outlet temperatures up to 800◦ C would enable the MSR to meet the electricity and hydrogen production missions, but the epi-thermal neutron spectrum would require multiple transmutation of some actinides to arrive at species with a significant fission cross-section in meeting the actinide fissioning mission. The fuel would be a circulating liquid mixture of sodium, zirconium, uranium and actinide fluorides flowing through channels in a graphite core structure. Super-Critical Water Reactors (SCWR) The SCWR system features a high-temperature, high-pressure water-cooled reactor that operates above the critical point of water to obtain a higher thermal efficiency than current LWRs. The balance of plant is simplified by the use of a direct cycle energy conversion system. The present (2006) reference is a 1700 MWe plant operating at 25 MPa and outlet temperature of 510◦ C. A second option with a fast neutron spectrum would have a closed fuel cycle with full actinide recycle and an aqueous processing facility. Sodium-Cooled Fast Reactors (SFR) The SFR system features a fast neutron spectrum, sodium coolant and a closed fuel cycle. With an outlet coolant temperature of 550◦ C, the present (2006) SFR designs can meet the electricity mission but not the hydrogen production mission, and the fast neutron spectrum would facilitate achieving the actinide fissioning mission. The fuel cycle includes full actinide recycle, with two major options: (1) a 150–500 MWe reactor with a uranium–plutonium–actinide–zirconium metal alloy fuel and a pyrometallurgical processing facility, based on IFR technology; and (2) a 500–1500 MWe reactor with uranium–plutonium–oxide dispersion fuel and an advanced aqueous processing facility, based on LMFBR technology. By full recycling of actinides in a fast spectrum, the SFR would minimize the discharge of long-lived waste to HLW repositories and could maximize the utilization of the energy content of uranium (including depleted U) by transmutation of 238 U into fissionable transuranics. Very High Temperature Reactors (VHTR) The VHTR system features a thermal neutron spectrum, helium coolant and a once-through uranium fuel cycle. With an outlet helium temperature of 1000◦ C, the VHTR is intended to be a high-efficiency system that provides process heat for non-electrical applications such as thermochemical hydrogen production, but it could include cogeneration of electricity. The reactor core can be prismatic (like the HTGR and GT-MHR) or pebble-bed (like the PBMR). The present (2006) reference design is a 600 MWth core connected to an intermediate heat exchanger to deliver process heat. 7.11 Advanced Sub-critical Reactors 7.11 Advanced Sub-critical Reactors Sub-critical operation would allow a much wider range of fuel cycle options for any of the above reactors because remaining critical would not constrain the fuel burnup. Also, because the delayed neutron fraction for the actinides (e.g., β = 0.0020 for 239 Pu in a fast spectrum) is considerably less than for 235 U (β = 0.0064), operating a reactor with a substantial transuranic loading would significantly reduce the reactivity margin to prompt critical. Furthermore, the requirements to fission a substantial fraction of the fissile isotopes in the fuel before removing it from the reactor or to achieve a significant 238 U transmutation rate translate into requirements to compensate a substantial negative reactivity decrement. Since sub-critical operation provides an additional reactivity margin ≈ 1 − ksub and an external neutron source S can be increased to offset any decrease in reactivity and hence to maintain constant the number of neutrons in the reactor (N = (S/(1 − ksub )), where  is the neutron lifetime), achievement of both the uranium utilization and the actinide fissioning missions would be facilitated by sub-critical operation. Based on the studies to date, it is the preponderance of informed opinion that sub-critical operation will be required in at least some of the reactors in the international “fleet” in order to fully achieve the closed fuel cycle. Two types of neutron sources of the size required are being developed on a time scale that should make them available shortly after the first deployment of advanced GEN-IV reactors, (1) the proton accelerator-spallation target neutron source which is the basis for the Spallation Neutron Source which recently (2006) became operational at Oak Ridge National Laboratory, and (2) the tokamak D–T fusion neutron source which is the basis of the international ITER fusion project that is beginning the construction phase this year (2006) in France. Conceptual designs for an Accelerator Transmutation of Waste (ATW) reactor are based on a proton Linac (about 0.9 kilometer in length) accelerating protons to about 1000 MeV, at which point they would be directed downward through a vertical column in the center of the reactor into a Pb–Li target to produce via the spallation reaction copious neutrons ranging in energy up to 20 MeV that are incident onto the surrounding sub-critical reactor, as illustrated in Fig. 7.18. Projected neutron source rates are 0.1–1 × 1019 /s, and the source distribution would be highly localized about the target. The technology needed for the ATW will be tested on the Spallation Neutron Source. Conceptual designs for a Fusion Transmutation of Waste (FTW) reactor are based on a D–T plasma confined in a toroidal confinement volume of major radius R = 3–4 m producing 14 MeV neutrons. The sub-critical reactor forms an annular ring (inner radius 4–5 m, height ≈3 m, width ≈1 m) located on the outboard side of the plasma chamber, as shown in Fig. 7.19. Projected neutron source rates are 1–10 × 1019 /s, and the source would be broadly distributed over the surface of the toroidal plasma chamber. The physics and technology needed for a fusion neutron 273 274 7 Nuclear Power Reactors Fig. 7.18 Gas Cooled ADS Demo Spallation Target. Fig. 7.19 Tokamak D–T Fusion Neutron Source-Driven Sub-Critical Reactor. 7.12 Nuclear Reactor Analysis source will be tested in the ITER experiment which will begin operation in 2016– 2019. 7.12 Nuclear Reactor Analysis We now turn to a brief discussion of the application of the computational methods of reactor physics to analysis of the nuclear, or neutronics, performance of nuclear power reactors. More detailed discussions of reactor analysis procedures and a description of the various codes employed are given in Refs. 5 and 6. The advanced reactor physics calculational methods used in nuclear reactor analysis, in addition to those described in previous chapters, are described in Chapters 9 to 16. Construction of Homogenized Multigroup Cross Sections As the discussion of nuclear power reactors above illustrates, nuclear reactor cores are composed of tens of thousands of components of very different material properties, some of them highly absorbing fuel and control elements, with dimensions that are comparable to or smaller than the neutron diffusion length. Yet the major computational tool of nuclear reactor analysis is multigroup diffusion theory, which is rigorously valid only in weakly absorbing media at distances of a few diffusion lengths away from interfaces with strongly dissimilar media. Furthermore, many of the nuclear cross sections depend strongly on the details of the neutron energy distribution (e.g., resonances), which in turn are spatially dependent through the spatial distribution of materials. Thus the first major step of nuclear reactor analysis is to develop equivalent homogenized cross sections for the different fuel assemblies or fuel modules, which incorporate the effects of the detailed neutron distribution in space and energy, and to develop an equivalent representation of highly absorbing control elements. The word equivalent implies that these approximate representations would yield the same prediction of reaction rates as a detailed heterogeneous fine-energy calculation would, were it practical to perform the latter. Construction of such equivalent representations is a major and ongoing reactor physics activity. The relative importance of the treatment of spatial and energy details differs among reactor types. For thermal reactors, in which most of the neutrons are absorbed in the thermal energy range where the neutron mean free path is small, treatment of the detailed spatial heterogeneity is paramount, and treatment of the details of the energy distribution is secondary but still important. On the other hand, for fast reactors, in which most of the neutron absorption takes place with fast neutrons with long mean free paths, treatment of the details of the energy distribution is paramount, and the spatial heterogeneity is secondary. In a thermal reactor, the homogenization procedure starts at the pin-cell level of a fuel pin and the surrounding coolant, moderator, and structure. In a typical analysis, a volume-weighted homogenized fine group (30 to 60 fast groups, 15 to 275 276 7 Nuclear Power Reactors 172 thermal energy points or groups) pin-cell model is constructed, using integral transport theory to calculate heterogeneous resonance cross sections for the fuel nuclei. This model is used to calculate intermediate group cross sections to be used in a transport calculation of the heterogeneous pin cell. The spatially dependent intermediate group fluxes are used to construct volume-flux-weighted homogenized cross sections, usually with a smaller number of groups, for the pin cell. This is repeated for the various types of pin cells in a fuel assembly to obtain an intermediate group (5 to 15 groups) model for the fuel assembly which represents each fuel pin cell as an equivalent homogenized region. Then an intermediate group diffusion or transport calculation is performed for the fuel assembly or module, taking into account any water gaps, nearby structure or control elements, and so on. The intermediate group fluxes from the assembly transport calculation are then used to construct volume-flux-weighted few-group diffusion theory cross sections for the assembly. Usually, separate calculations are performed for the fast and thermal (E < 1 eV) energy regions. This process is repeated for the various fuel assemblies or modules that compose the core, resulting in equivalent few-group diffusion or transport theory cross sections which represent each homogenized fuel assembly or module. Supplemental transport calculations are used to construct effective fewgroup diffusion theory cross sections which represent the control elements in a diffusion theory model. In a fast reactor the procedure is similar, but with more emphasis on treatment of the energy structure and of overlapping resonances and less on the treatment of the spatial structure (except as it affects the resonance treatment). In a typical analysis, an entire fuel assembly or group of similar fuel assemblies is homogenized on a volume-weighted basis to obtain an ultrafine-group (≈2000) model, with integral transport calculations being used to construct heterogeneous resonance cross sections for the fuel nuclei. Ultrafine-group spectra are then calculated and used to construct fine-group cross sections for use in a multigroup (20 to 40 groups) diffusion or transport theory core calculation of the entire core. Homogenized multigroup cross sections must be constructed for the variety of conditions encountered in subsequent applications because they depend on the details of the spatial and spectral flux distributions used in their construction. The presence or absence of a control element or a strong absorber such as xenon, the change in fuel composition with burnup, the buildup of plutonium and fission products, the different temperature and coolant densities encountered in a transient, and other factors, all affect the details of the spatial and spectral distributions and must be taken into account in the preparation of equivalent homogenized multi-group cross sections. Criticality and Flux Distribution Calculations The equivalent homogenized multigroup cross sections can be used to perform global diffusion or transport theory calculations of the reactor core, with the control rod positions adjusted to achieve criticality (k = 1) or with an eigenvalue k cal- 7.12 Nuclear Reactor Analysis culated. If three-dimensional finite-difference representations of the core are used for these calculations, the calculated fluxes are averaged global flux distributions. However, detailed pin-by-pin flux distributions are needed for the calculation of pin power limits and pin fuel depletion. The detailed local flux at the fuel pin-byfuel pin level must be reconstructed by superimposing on this global average flux the detailed assembly and pin-cell transport flux distributions that were used in preparation of the homogenized multigroup cross sections, with the appropriate normalization. Frequently, further approximations are made in the calculation of global flux distributions, in the interest of computational economy (e.g., nodal models that represent the global flux distribution within a fuel assembly or module with a few parameter polynomial). In such cases, the detailed local flux on a fuel pin-by-fuel pin level again must be reconstructed by superimposing on this representation of the global average flux the detailed assembly and pin-cell transport flux distributions that were used in the preparation of the homogenized multigroup cross sections. Care must be taken that the flux reconstruction procedure is consistent with the homogenization procedures and with the procedures used in the development of the approximate global flux calculation model. Fuel Cycle Analyses Calculation of the multigroup global flux distribution and critical control rod position and reconstruction of the flux distribution on a pin-by-pin basis is coupled with the calculation of fuel composition change and fission product buildup on a pin-by-pin basis in the multistep fuel cycle analysis calculation. The sequence of calculations is first to perform a number of flux calculations to establish the critical control rod position and corresponding flux distribution for the fresh fuel loading with and without equilibrium xenon and samarium, then solution of the fuel depletion and actinide/fission product buildup equations over a depletion time step using the initial equilibrium xenon and samarium flux distribution, then solution of neutron flux equations several times to establish the critical rod position and flux distribution corresponding to the new fuel composition and fission products, then the solution of the fuel depletion and actinide/fission product buildup equations over the next depletion time step using the newly calculated flux distribution with equilibrium xenon and samarium, and so on. Typical time steps might be 150, 350, 500, several 1000, and then 2000 MWd/T, the initial small time steps taken to build up equilibrium xenon and samarium and 239 Pu. Homogenized multigroup cross sections must be redetermined at each time step, either from a recomputation as described above or from interpolation in a table of fitted cross sections. Efficient fuel management requires that the fuel cycle analysis described in the preceding paragraph be repeated many times. A series of such calculations will be made prior to fuel loading to determine the proper mixture and location of fresh and recycled fuel to achieve optimal fuel performance subject to a given set of assumptions about plant availability, power demand schedule, and refueling period. Then as the reactor operates and the assumptions are replaced by operating history, 277 278 7 Nuclear Power Reactors additional series of calculations are made to adjust the remaining operating plan and/or refueling date to achieve optimal fuel performance. The large number of criticality and flux distribution calculations needed for fuel cycle analyses places a computational efficiency requirement on the neutron flux solution method. Approximate flux solution methods, such as the nodal model, are widely used. However, it should be noted that Monte Carlo codes capable of calculating fuel depletion on a point-by-point basis are available. In fast breeder reactor calculations, there is a greater emphasis on determination of the initial plutonium concentration in the fuel and on the production and destruction of actinides with fuel depletion. Transient Analyses It is necessary to analyze a large number of planned operational transients (e.g., startup, power-level change) and potential transients that could result from offnormal or accident conditions (e.g., control rod ejection, loss of coolant flow), each subject to a variety of assumptions regarding the performance of control and reactor systems. Such calculations require solution of the time-dependent equations describing the neutron flux distribution and the reactor power level and distribution, heat conduction and the temperature distribution, the hydrodynamics and thermodynamics of the heat transport system, material expansion and movement, and in the case of extreme accident scenarios, the equations of state and the equations governing the hydrodynamics of melting and vaporizing fuel mixtures. Calculation of the neutron flux spatial distribution and level determines the reactivity, which is the driving function for any reactor transient, and the heating source level and distribution, which is the primary input to the other calculations. In the simplest point kinetics model for neutron dynamics, the neutron flux distribution is assumed to be fixed and only the amplitude, or power-level, changes. The reactivity coefficients associated with fuel and moderator temperature and density changes, fuel and structure motion, and so on, are precomputed from a series of static neutron flux and criticality calculations (or from a few such calculations supplemented by perturbation theory estimates). Then, as changing temperatures and densities, fuel and structure motion, and so on, are calculated, the reactivity worth of these changes is incorporated into the power-level calculation using the reactivity coefficients. Reference, or design basis, power distributions are often used in conjunction with point kinetics calculations to assess fuel integrity. Separate calculations are then performed to assure that the actual power distribution is not more limiting than the design basis power distribution. For certain accident simulations (e.g., those in which control elements are out of position), the design basis power distributions are inappropriate and new power distributions must be calculated based on the temperature, density, flow, and other information from the transient analysis calculation. The reactivity feedback coefficients are determined for reference control rod positions and other core conditions. If the control rod positions or the core conditions 7.12 Nuclear Reactor Analysis are altered significantly, the reactivity coefficients, which depend on a flux-adjoint volume weighting of the perturbation, will be different because the flux and adjoint distributions will be different. The most important reactivity coefficients must be computed for conditions present during the most critical stages of the transient analysis. The point-kinetics calculation cannot account for effects associated with changes in the spatial flux distribution, which may occur, for example, if there is a reduction of coolant flow only in one part of the reactor. Such changes in spatial flux distribution not only affect the local power distribution and heat source distribution but also produce changes in reactivity and in the reactivity coefficients. Thus there are situations in which calculation of the space- and time-dependent flux distribution is required. Such calculations require, in essence, a series of solutions for the spatial flux distribution, using at each step the most recent calculations of the temperature, density, and position of the materials in the reactor. Approximate flux solution methods, such as the nodal model, are normally used in such cases to make the computational requirements tractable. Core Operating Data Precalculated or on-line calculated values of various core physics parameters and responses must be available to the reactor operators to enable them to make core operational decisions, such as the control element insertion pattern, and to interpret instrument readings. Much of this information is developed in the course of fuel management and transient safety analyses, since the safety analysis considers a wide range of abnormal and normal conditions. Other information is provided by core operating data, although these are usually only for normal operating conditions. Additional power distribution and criticality calculations are necessary to fill in the database. Criticality Safety Analysis At various stages of the enrichment, fabrication, and transportation procedures prior to loading the fuel into the reactor, and at various stages of the temporary storage, processing, transportation, and permanent storage procedures for spent nuclear fuel, the nuclear fuel is distributed within a variety of configurations. Examples of such configurations are spent fuel assemblies stored in a swimming pool (to provide for decay heat removal) at the reactor site and barrels of processed fuel in liquid form arrayed on storage racks. Criticality safety requires a rigorous fuel management system to insure that the fuel inventories of each storage element is known and that the various configurations are well subcritical under all normal and conceivable off-normal conditions. Criticality calculations of the type discussed for the case when the fuel is loaded into the reactor must also be performed for these various ex-reactor configurations. While diffusion theory and the methodology discussed in previous chapters may suffice for certain of these configurations, the 279 280 7 Nuclear Power Reactors more rigorous transport methods of Chapter 9 are generally required for criticality safety analyses. 7.13 Interaction of Reactor Physics and Reactor Thermal Hydraulics Power Distribution More than 90% of the recoverable energy released in fission is in the form of kinetic energy of fission products and electrons, which is deposited in the fuel within millimeters of the site of the fission event, and somewhat less than 10% of the energy is in the form of energetic neutrons and gamma rays, which are deposited within about 10 cm around the fission site. Thus the heat deposition distribution is approximately the same as the fission rate distribution:  (r)φ(r) (7.1) q  (r) ≈ const. × f The requirement to remove this heat without violating constraints on maximum allowable values of materials temperature, heat flux from the fuel into the coolant, and so on, places limits on allowable neutron flux peaking factors, fuel element dimensions, coolant distribution, and so on. The neutron flux distribution affects the temperature in the fuel and coolant/moderator, the temperature of the fuel affects the fuel resonance cross section, and the temperature of the coolant/moderator affects the moderating power, both of which in turn affect the neutron flux distribution. An increase in the local resonance absorption in the fuel when the local fuel temperature increases results because of the Doppler broadening of the resonances. This increase in local fuel absorption cross section will generally reduce the number of neutrons that reach thermal locally in LWRs, which will tend to reduce the local fission rate and compensate the original increase in fuel temperature. The increase in local fuel resonance absorption makes the fuel compete more effectively for local neutrons, which lends to make other nearby absorbers somewhat less effective (e.g., reduces the worth of nearby control rods). The effect of coolant temperature on neutron moderation is also important. In most LWR cores, a local decrease in water density resulting from an increase in water temperature will cause a decrease in neutron moderation, which in turn causes a decrease in local power deposition. As the coolant passes up through the core, the cumulative heat input from the fuel elements causes the axial temperature distribution to increase with height: conversely, the axial density distribution decreases with height. This produces a power distribution peaked toward the bottom of the core, which is pronounced in BWRs, for which progressive coolant voiding occurs in the upper part of the core. Control rods are inserted from the bottom in BWRs to maximize rod worth and to avoid exacerbating this peaking in the axial neutron flux at the bottom of the core. The shift toward a harder spectrum associated with 7.13 Interaction of Reactor Physics and Reactor Thermal Hydraulics a local Na density decrease in a fast reactor results in an increase in local η, which increases the local heating. The coupling between reactor physics and thermal hydraulics is much weaker in gas-cooled reactors, in which the moderator is separate from the coolant. Temperature Reactivity Effects The general reactivity effects associated with changes in fuel, coolant/moderator, and structural temperatures and their effect on the reactor dynamics were discussed in Sections 5.7 to 5.12. The interaction of thermal-hydraulics and reactor physics phenomena to produce positive reactivity in the Three Mile Island and Chernobyl accidents is discussed in Section 8.4. The overall reactivity effect depends on the local changes in temperature and density in each zone of the reactor and the local neutron flux, weighted by the relative importance of these local reactivity contributions and summed over the reactor. The thermal-hydraulics characteristics of a reactor affect not only the local temperature and density changes in response to a change in the neutron flux distribution and magnitude, but also affect changes in the neutron flux distribution and magnitude in response to changes in local temperature and density. Coupled Reactor Physics and Thermal-Hydraulics Calculations It is clear from the discussion above that the power distribution and effective multiplication constant in a nuclear reactor depends not only on the distribution of material (fuel, coolant, structure, control) within a reactor core, but also on the temperature and density distribution within a reactor core. In the design process, it is necessary to determine a self-consistent material and temperature–density distribution that makes the reactor critical at operating conditions without violating thermal-hydraulics limits. The problem is further complicated by fuel depletion, which changes the materials in the fuel during the course of time; the distributions of materials and temperature–density must make the reactor critical over its entire lifetime without violating thermal-hydraulics limits. This is normally accomplished by trial and error, iterating between static neutron flux and thermal-hydraulics calculations until a self-consistent solution is found which can be made critical by adjusting control poison levels and which satisfies thermal-hydraulics and safety limits over the projected core lifetime. Once the design is fixed, it is necessary to analyze a number of operational and off-normal transients to ensure that the reactor will operate without violation of thermal-hydraulics limits under normal conditions and that it will operate safely under off-normal conditions. The transient analyses codes usually solve for the neutron power amplitude and distribution and the corresponding temperature and density distributions, in some approximation. 281 282 7 Nuclear Power Reactors References 1 “European Pressurized Water Reactor,” Nucl. Eng. Des. 187, 1–142 (1999). 2 D. C. Wade and R. N. Hill, “The Design Rationale of the IFR,” Prog. Nucl. Energy 31, 13 (1997); E. L. Gluekler, “U.S. Advanced Liquid Metal Reactor (ALMR),” Prog. Nucl. Energy 31, 43 (1997). 3 R. A. Knief, Nuclear Engineering, Taylor & Francis, Washington, DC (1992), Chaps. 8–12. 4 J. G. Collier and G. F. Hewitt, Introduction to Nuclear Power, Hemisphere Publishing, Washington, DC (1987), Chaps. 2 and 3. 5 P. J. Turinsky, “Thermal Reactor Calculations,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations, CRC Press, Boca Raton, FL (1986). 6 M. Salvatores, “Fast Reactor Calculations,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations, CRC Press, Boca Raton, FL (1986). 7 R. H. Simon and G. J. Schlueter, “From High-Temperature Gas-Cooled Reactors to Gas-Cooled Fast Breeder Reactors,” Nucl. Eng. Des. 4, 1195 (1974). 8 Report on the Accident at the Chernobyl Nuclear Power Station, NUREG-1250, U.S. Nuclear Regulatory Commission, Washington, DC (1987). Problems 7.1. Discuss the differences between ‘thermal’ and ‘fast’ reactors in terms of (a) neutron energy distribution, (b) capability to fission uranium and higher transuranic isotopes, (c) Doppler temperature coefficient of reactivity, (d) time constant for dynamic response to prompt supercritical reactivity insertions, (e) coolants that could be used, and (f) power density. 7.2. Discuss why fast reactors are necessary to close the nuclear fuel cycle. 283 8 Reactor Safety A great deal of effort is devoted to ensuring that nuclear reactors operate safely. The fundamental objective of this effort is to ensure that radionuclides are not released to create a health hazard to the general public or operating personnel. Fundamental considerations of reactor safety, the methodology of safety analyses, reactor accidents, and the design approach to reactor safety are described in this chapter, with an emphasis on the role played by reactor physics. 8.1 Elements of Reactor Safety Radionuclides of Greatest Concern The radionuclides in a nuclear reactor that could most affect public health if released are the fission products and the actinides produced by neutron transmutation. For the most part, these radionuclides are harmful only if they are inhaled or ingested and concentrated chemically in a susceptible organ. As discussed in Chapter 6, the short-lived fission products constitute the major source of such radionuclides in an operating reactor. The most significant fission products and the organs they affect are identified in Table 8.1. 90 Sr and 137 Cs and the isotopes of iodine are of particular concern. Strontium has a high fission yield and behaves chemically like calcium and is deposited in bone tissue. Both 90 Sr and its daughter 90 Y produce a very high dose per unit activity, which is quite damaging to the blood cells produced in bone marrow. The iodine radioisotopes are concentrated in the thyroid gland, where they would produce tumors. Multiple Barriers to Radionuclide Release Multiple barriers against fission product (and actinide) release are a key safety feature of nuclear reactor design. The fission products in an operating reactor are contained within UO2 pellets that are packed into clad fuel elements which are assembled within the reactor core. The reactor core is located within a pressure 4.8 5.9 5.9 6.1 2.9 4.4 6.5 7.6 5.9 2.9 0.38 1.0 5.9 8.1 d 2.4 h 20 h 52 m 6.7 h 40 d 1.0 y 34 d 33 y Fission Yield (%) 50 d 28 y 58 d 280 d Radioactive Half-Life, t1/2 Source: Data from Ref. 14. ∗ Fraction of inhaled material that deposits in the indicated tissue. † A somewhat typical average residence time for fuel in an LWR is 400 full-power days; equilibrium inventories are achieved at times that are long compared to the radionuclide half-life. Bone 89 Sr 90 Sr–90 Y 91 Y 144 Ce–144 Pr Thyroid 131 I 132 I 133 I 134 I 135 I Kidney 103 Ru–103m Rh 106 Ru–106 Rh 129m Te–129 Te Muscle 137 Cs–137m Ba Isotope 0.36 0.01 0.01 0.02 0.23 0.23 0.23 0.23 0.23 0.28 0.12 0.19 0.075 Deposition Fraction∗ Table 8.1 Significant Fission Products of Concern for Internal Doses in Reactor Accidents 17 d 13 d 19 d 10 d 7.6 d 2.4 h 20 h 52 m 6.7 h 50 d 18 y 58 d 240 d Effective Half-Life 8.6 6.9 65 46 1, 484 54 399 25 124 413 44, 200 337 1, 210 Internal Dose (mrem/μCi) 1.2 26.3 1.8 9.1 26.3 40.0 59.0 69.0 53.6 43.4 1.45 53.2 34.7 53.6 26.3 3.5 9.1 26.3 40.0 59.0 69.0 53.6 43.6 53.6 53.6 55.4 Reactor Inventory† (Ci/kWt) 400 Days Equilibrium 284 8 Reactor Safety 8.2 Reactor Safety Analysis vessel that in turn is located inside a containment building. Both the pressure vessel and the containment building are designed to withstand large overpressures. Thus the pellet, clad, pressure vessel, and containment building constitute four barriers against the release of fission products. Defense in Depth The first level of defense against fission product release is to design to prevent the occurrence of any event that could result in damage to the fuel or other reactor system. Negative reactivity coefficients that lead to inherently stable operating conditions, safety margins in design, reliable and known materials performance in structures and components, adequate instrumentation and control, and so on, are among the preventive measures employed in reactor design. The second level of defense are protective systems, which are designed to halt or bring under control any transients resulting from operator error or component failure that may lead to fuel damage and fission product release within the pressure vessel. Reactor scram systems which inject control rods into the core for rapid shutdown upon being activated by any one of several signals being outside the tolerance range, pressure relief systems, and so on, constitute the reactor protective systems. The third level of defense is provided by mitigation systems, which limit the consequences of accidents if they do occur. Emergency core cooling, emergency secondary coolant feedwater, emergency electrical power systems, systems for removing fission products that have been released into the reactor hall, and a reinforced containment building that can withstand high overpressure are elements of the mitigation system. Energy Sources The potential for the release of fission products is related directly to the amount of energy available. The primary energy source is the nuclear energy that is released in a positive reactivity insertion. However, there are other important energy sources that can play a role in an accident. The heat released in fission product decay is 7.5% of the operating power and constitutes a substantial heat source for some time after the reactor shuts down. There is thermal energy stored in the reactor materials which may become redistributed (e.g., the flashing of water to steam upon depressurization). There are several exoergic chemical reactions (Table 8.2) which may take place at elevated temperatures during the course of an accident, most of which produce hydrogen, which has an explosive potential. 8.2 Reactor Safety Analysis All reasonably conceivable failures are postulated and analyzed to design reactor protective and mitigation systems, to prevent accidents, to prevent the release of 285 ZrO2 FeO, Cr2 O3 , NiO Na2 O NaOH CO CO2 H2 O 1852† 1370† 25 25 1000 1000 1000 Zr (liquid) SS (liquid) Na (solid) Source: Data from Ref. 15; used with permission of MIT Press. ∗ Positive values indicate energy that must be added to initiate an endoergic reaction; negative values indicate energy released by exoergic reactions. † Melting point. H2 (gas) C (solid) Oxide(s) Formed Temperature (◦ C) Reactant (R) Table 8.2 Properties of Exoergic Reactions of Interest for Reactor Safety −2883 −1330 to −1430 −2162 – −2267 −7867 −29,560 −1560 −144 to −253 – −1466 +2700 +2067 – Heat of Reaction∗ with: Oxygen Water (kcal/kg◦ R) (kcal/kg◦ R) 490 440 – 490 1870 3740 – Hydrogen Produced with Water (l/kg◦ R) 286 8 Reactor Safety 8.2 Reactor Safety Analysis fission products in the event of an accident, and to investigate the consequences of various accident scenarios for the release of radionuclides. The analyses are performed with sophisticated computer code systems that model the neutron dynamics and fission power production; the temperature, density, state, and location of materials within the reactor core and the reactivity worth of changes therein: the primary and secondary heat transport system1) ; the performance of the reactor safety protective and mitigation systems; the integrity of the fuel elements, pressure vessel and containment structure intended to prevent release of radionuclides; the dispersion of any released radionuclides; and a radiological assessment of resulting health effects in the population affected. Accident scenarios are commonly classified by the initiating event, some of the major events being those discussed below. Loss of Flow or Loss of Coolant Loss-of-flow accidents (LOFAs) would be caused by failure of one or more pumps in the primary coolant system, which results in increased temperature and reduced density for the coolant. Loss-of-coolant accidents (LOCAs) can be caused by a rupture of the primary coolant line, failure of a primary coolant pump seal, inadvertent opening of a pressure relief or safety valve, and so on, and would result in increased temperature and decreased density of the coolant and possibly uncovering of the core. The negative coolant temperature reactivity coefficient of PWRs and BWRs, which would provide for an immediate power reduction, is an important feature in the early stages of such accidents. Loss of Heat Sink When steam flow in the secondary coolant system is decreased or lost due to a turbine trip (shutdown) or reduction or loss of feedwater in the secondary coolant system, an undercooling accident, or in an extreme case, a loss-of-heat sink accident (LOHA), would occur. Such an accident would result in the reduction or elimination of heat removal from the primary coolant system, causing the primary coolant temperature to increase and the density to decrease. Again, a negative coolant temperature coefficient of reactivity is an important feature in the early stages of such accidents. Reactivity Insertion Uncontrolled control rod withdrawal or ejection is the most common type of initiator for a reactivity insertion accident. However, there are other reactivity insertion 1) A PWR has a primary coolant system that removes heat from the reactor core and carries it to the steam generator, or heat exchanger, where the heat is transferred out of the primary coolant through tube walls to a cooler secondary coolant which is heated above the vaporization temperature to produce steam that is transported to turbines for the production of electricity. 287 288 8 Reactor Safety mechanisms. The startup of an inactive primary coolant pump (or recirculation loop in a BWR), which injects cold water into the primary coolant system, would cause a positive reactivity insertion in reactors with a negative coolant reactivity coefficient. A steam line break in the secondary coolant system would result in increasing coolant flow in the secondary system, hence in increasing heat removal from the primary coolant, which would also result in a positive reactivity insertion in reactors with a negative coolant reactivity coefficient. The potential for such cold-water reactivity insertions places limits on the allowed magnitude of negative coolant reactivity coefficients. Anticipated Transients without Scram Anticipated transients without scram (ATWSs) are certain transient events which may occur once or twice in a reactor lifetime, on average, which can be handled by the protective system initiating a reactor scram. When the scram system is postulated to fail, such events may initiate an accident. 8.3 Quantitative Risk Assessment Development and application of a methodology for quantification of the risk to public and worker safety associated with the occurrence of a reactor accident has provided a valuable basis for evaluating the relative safety of nuclear reactors. In broad terms, the (public safety) risk associated with a nuclear reactor may be characterized in terms of the various sequences of events, or scenarios, that could lead to the release of various quantities of radionuclides, the probabilities that each sequence of events could occur, and the public or worker health consequences of the release of various quantities of radionuclides. Probabilistic Risk Assessment Safety protective and mitigation systems are designed to minimize component damage and prevent radionuclide release for each of the potential accidentinitiating events described in the preceding section (plus others), if the system works as designed. For a given initiating event (e.g., a loss-of-coolant accident), the success or failure of the hierarchy of relevant safety systems—electric power, emergency core cooling, fission product removal from the reactor hall, containment— are considered sequentially. The frequency of occurrence of the initiating event, λ, and the failure probabilities, Pi , for each safety system are first identified. Then an event tree is constructed, as shown in the upper portion of Fig. 8.1, tracing the various pathways that the accident could follow with respect to success or failure of the various safety systems. Since the conditional failure probabilities, Pi , are small, the overall probability of any given pathway is just the initiation frequency times the product of the Pi for the different failures in the pathway, if the failure probabilities 8.3 Quantitative Risk Assessment Fig. 8.1 Event tree logic diagram for a LOCA in an LWR. (From Ref. 13.) are independent. However, the failure probabilities of the various safety systems are not independent (e.g., electrical power failure implies also failure of the emergency core cooling and fission product removal systems). Accounting for correlated failures reduces the event tree, as indicated in the lower portion of Fig. 8.1. Quantification of the initiating event frequencies and of the safety system failure probabilities is, of course, the essential part of this methodology. A deductive technique known as fault tree analysis is employed for this purpose. A given safety system failure (e.g., loss of electrical power) requires failure of both the primary (off-site power supply) and the backup (on-site diesel generator) systems. Failure of the off-site power supplies requires failure of both the power sources on the lo- 289 290 8 Reactor Safety Fig. 8.2 Estimated principal contributions to core damage frequency for Sequoyah PWR, Surrey PWR, Zion PWR, Peach Bottom BWR, and Grand Gulf BWR. (From Ref. 5.) cal grid and the tie-in with other power grids, or a failure of the local power grid. Each of these secondary failure possibilities can be related, in turn, to several possible tertiary causes, and so on. By tracing these possible failures back several levels, a fault tree can be constructed. By assigning failure probabilities at each level on the fault tree and combining these probabilities statistically, an overall failure probability for a given safety system can be constructed. Data used for fault tree analyses include component and system failure rate data, human error, maintenance and testing time when a system might not be available, and so on. Probabilistic risk assessment has proven to be a powerful tool for identifying the relative importance of various failure modes in a given plant. However, the results tend to be different for plants with different reactor types and containment systems, as shown in Fig. 8.2. 8.3 Quantitative Risk Assessment Radiological Assessment The public health consequences of the release of a given inventory of radionuclides from a containment building depends on the dose received to various body organs by the affected population and the effect of that dose on those organs. The dispersion of radionuclide fallout from the release point depends on wind and weather conditions. The population that might be affected by this fallout depends on the population pattern of the fallout zone and any evacuation measures that would be taken. Calculation of radionuclide dispersion among the affected population is relatively straightforward. Most radionuclides must be inhaled or ingested to affect public health. Immediately following their release, breathing is the most likely pathway for radionuclides to enter the body. Over the longer term, there are many possible pathways, including breathing, drinking contaminated water, and eating contaminated food at any step in the food chain, that must be considered. Calculation of radionuclide uptake by the affected population is uncertain, and worst-case assumptions must be made when information is lacking. Health effects of radiation exposure fall into three categories: early fatalities (acute), early illnesses, and latent effects. Early fatalities—defined as those that occur within a year of exposure—follow a linear dose–effect relationship varying from 0.01% fatality risk for 320 rad2) to 99.99% fatality risk for 750 rad whole-body radiation exposure has been established from radiation effects data. Early illnesses are associated primarily with the respiratory tract and lung impairment in particular. A linear dose–effect relationship varying from 5% lung impairment for 3000 rad to 100% impairment for 6000 rad internal radiation exposure to the lung has been established from radiation effects data. Latent effects of radionuclide ingestion include cancer fatalities, thyroid nodules, and genetic damage, which generally occur 10 to 40 years after the accident. Linear dose–effect relationships can be established for significant levels of radiation exposure, but there are no radiation effects data at the low levels of exposure that would be encountered in trying to determine the latent effects of radionuclide ingestion following an accident. It is common practice to extrapolate the linear dose–effect relationship to zero dose in predicting latent health effects, but this practice is controversial because theoretical studies suggest that a threshold level of radiation energy deposition is required to cause cell damage. The predicted cancer fatality rate, using the linear extrapolation to zero dose, is about 100 per 106 person-rem exposure. Reactor Risks The estimated frequencies and public health and property damage consequences of possible PWR/BWR reactor accidents are given in Table 8.3. The most likely 2) Radiation doses are measured in a variety of units. The rad corresponds to the absorption of 100 ergs/g of material, and the gray (Gy) is equal to 100 rads. The rem is equal to the rad multiplied by a quality factor (1 for x-rays, gammas, and electrons; 10 for neutrons and protons; 20 for alpha particles), and the sievert (Sv) is 100 rems. 291 < 1.0 < 1.0 < 1.0 < 1.0 < 1.0 < 0.1 < 0.3 (< 0.1) < 0.3 (< 0.1) 1 : 2 × 104∗ Source: Data from Ref. 13. ∗ This is the predicted chance of core melt per reactor year. † These rates would occur in approximately the 10 to 40-year period following a potential accident. ‡ This rate would apply to the first generation born after a potential accident. Subsequent generations would experience effects at a lower rate. Relocation area [km2 (mi2 )] Early fatalities Early illness Latent cancer fatalities (per year)† Thyroid nodules (per year)† Genetic effects (per year)‡ Total property damage ($109 ) Decontamination area [km2 (mi2 )] Consequences < 1.0 300 170 1,400 25 0.9 5,000 (2,000) 340 (130) 1 : 106 Table 8.3 Estimated Probabilities and Consequences of a Single Reactor Accident 110 3,000 460 3,500 60 3 8,000 (3,200) 650 (250) 1 : 107 900 14,000 860 6,000 110 8 8,000 (3,200) 750 (290) 1 : 108 Chance per Reactor per Year 3,300 45,000 1,500 8,000 170 14 8,000 (3,200) 750 (290) 1 : 109 – – 4.5 × 105 17,000 8,000 8,000 – – Normal incidence 292 8 Reactor Safety 8.4 Reactor Accidents Fig. 8.3 Predicted frequency of fatality due to accidents from a number of technologies. (From Ref. 13.) core meltdown accident, which has a probability of 5 × 10−5 per reactor-year of occurring, has rather modest consequences. The more serious accidents have lower probabilities of occurrence. To put the risks of reactor accidents in perspective, the same methodology was applied to estimate the public health risks of other technological and natural phenomena to which the general public are exposed. As shown in Figs. 8.3 and 8.4, the risk to public health of the approximately 100 nuclear reactors operating in the United States is miniscule by comparison. 8.4 Reactor Accidents There have been two major reactor accidents, at Three Mile Island and at Chernobyl. It is important to understand what went wrong. Examination of the causes provides a basis for the design of reactors with improved safety features and operating procedures for the future. 293 294 8 Reactor Safety Fig. 8.4 Predicted frequency of fatalities due to nuclear reactor accidents and to a number of natural events. (From Ref. 13.) Three Mile Island On March 28, 1979, a series of events took place in unit 2 of the Three Mile Island plant near Harrisburg, Pennsylvania that resulted in the only major reactor accident in the history of commercial nuclear power in the United States. The TMI-2 unit was a standard PWR. Since this accident was associated primarily with the heat removal system, a simple diagram of a PWR heat removal system is shown in Fig. 8.5 to facilitate understanding of the sequence of events. The reactor was operating at about 97% of power, but with two valves on the emergency secondary coolant feedwater lines inadvertently closed, although the records available to the operators showed them to be open. The accident was apparently initiated by unsuccessful attempts to carry out a routine procedure of clearing a demineralizer line used to maintain secondary coolant purity, which apparently caused a condensate pump trip in the secondary cooling system. This led within a second to automatic trips in the main feedwater pumps for the secondary coolant system and the turbine. The loss of secondary coolant in the steam generators reduced the rate of heat removal from the primary coolant loop and the reactor core (a loss of heat sink accident). 8.4 Reactor Accidents Fig. 8.5 Schematic diagram of a PWR heat removal system. As the primary coolant became hotter and pressure increased, the overpressure relief valve in the pressurizer in the primary coolant system opened automatically when the (15.55 MPa) set point was exceeded, and 8 s into the accident the core protective system caused the control rods to be inserted in response to high coolant pressure signals. The primary system cooled following the control rod insertion, and the pressure dropped below the 15.21 MPa set point for closure of the overpressure relief valve at about 13 s into the accident, but the valve failed to close, although the solenoid deenergized, causing the primary coolant to be lost through the open valve into the drain tank at the bottom of the containment building, which reduced the pressure in the primary coolant system as well as the coolant level. At this point there was a loss of coolant accident, unbeknown to the operators. The control panel only indicated that the solenoid had deenergized, and primary coolant continued to be lost until the operators closed the blocked valve in the pressurizer drain line 142 min into the accident. At 14 s into the accident, the emergency secondary coolant feedwater pumps reached full design pressure, but unbeknown to the operators, the two inadvertently closed valves in the emergency secondary system coolant lines prevented the emergency secondary coolant from reaching the steam generators. It was another 8 min before an operator noticed low pressure and water levels in the steam generators, discovered the closed valves, and opened them to restore secondary coolant to the steam generators. At about 2 min into the accident, the primary system pressure dropped below the 11.31-MPa set point of the high-pressure injection system, which then started pumping borated water into the core. Because of the particular design, there was no direct relationship between the coolant levels in the reactor vessel and in the pressurizer. Even with continuing loss of primary coolant, the pressurizer signal indicated a filled system, which the operators had been trained to avoid because it prevented the pressurizer from fulfilling its function. Thus the operators turned off one of the pumps and throttled back the other pump in the high-pressure injection system, resulting in emergency coolant being added at a slower rate than primary coolant was being lost through the open pressurizer valve. 295 296 8 Reactor Safety About 73 min into the accident, both primary coolant pumps in the loop to one of the two steam generators were shut down in response to indications of vibrations, low pressure, and low coolant flow. This was done to prevent destruction of seals, which the operators feared would have caused a loss of coolant accident, still being unaware that they already had one on their hands. At about 100 min into the accident, the primary coolant pumps in the other loop were shut down for similar reasons. The pump shutdown caused the steam and water in the primary coolant loop to separate and apparently prevented further coolant circulation through the steam generators. The remaining liquid did not cover the core, and decay heat caused continuing vaporization of the noncirculating coolant. At about 111 min into the accident, reactor outlet coolant temperatures rose rapidly to 325◦ C and remained there. As the core became uncovered, the clad temperatures became high enough that exoergic Zr–steam reactions occurred, adding energy to the system and producing hydrogen. The cladding, with a melting point of 2100 K, became molten and began to dissolve the UO2 fuel. The next 13 h was spent trying various means to reestablish core cooling, which was ultimately successful. The reactivation of the high-pressure injection at 200 min into the accident recovered the core and filled the reactor vessel. A major slumping of the molten core occurred at 224 min into the accident, resulting in molten debris being deposited onto the lower vessel head, where it was apparently quenched by the coolant. A sizable hydrogen bubble was created by the Zr–steam interactions involving about one-third of the zircaloy in the core, the concentration of which became large enough to support combustion, and hydrogen ignition occurred at about 9.5 h into the accident. However, the pressure was well within the design limits of the pressure vessel. The hydrogen was removed during the first week. Reactor containment was successful in limiting radionuclide releases to less than 1% of total inventory, despite extensive core damage. Radiological assessments of the radionuclide release estimated average and maximum potential off-site doses of 0.015 and 0.83 mSv. As a point of reference, a dose of 1 mSv is estimated to result in a 1 in 50,000 chance of cancer, as contrasted with the 1 in 7 normal incidence of cancer in the population. The TMI-2 accident had no significant public health impact. In hindsight, TMI-2 was a huge and costly but poorly instrumented safety experiment that provided a convincing demonstration of the safety of a properly engineered nuclear reactor. Two of the major credible accidents—loss of heat sink and loss of coolant—took place, while the operators, who were unaware of the state of the reactor, took about the worst possible actions for the actual situation in an attempt to deal with the situation they thought they had on their hands. Although the reactor was destroyed, no one got hurt. By the same token, TMI-2 exposed major deficiencies in reactor operating procedures, operator training, and exchange of safety-related operating information, which stimulated extensive subsequent improvements. 8.4 Reactor Accidents Chernobyl In the early morning hours of April 26, 1986, a test was being performed on unit 4 of the Chernobyl nuclear power station about 130 km north of Kiev. The objective was to test the use of energy in the turbine during its post-trip coastdown as a source of emergency electrical power for cooling the reactor core following a scram, ironically to enhance the safety features of the reactor system. The test plan called for the power of the RBMK reactor to be reduced from the 3200-MWt full-power level to about 1000 to 700 MWt and for bypassing some safety systems that would have prevented the test conditions from being realized. The test was initiated by inserting control rods to reduce power to about 1600 MWt, the emergency core cooling systems were shut off to prevent them from drawing power during the test, and the power reduction continued to the planned level. However, the operator failed to reprogram the computer to maintain power in the range 1000 to 700 MWt, and the power fell to 30 MWt. The majority of the control rods were withdrawn to compensate the buildup of xenon, causing the power to climb and stabilize briefly at about 200 MWt. At about 20 min into the test, all eight pumps were activated to ensure adequate post-test cooling. The normal scram trip on high flow level, which would have prevented this, was deactivated. The increase in coolant flow reduced coolant temperature and increased coolant density, which introduced negative reactivity due to increased neutron absorption in the coolant, requiring further control rod withdrawal. This increased coolant density also maximized the positive reactivity worth of coolant voiding. The combination of low power and high flow produced instability, which required numerous manual adjustments, causing the operators to deactivate other emergency shutdown signals. At about 22 min into the test, the computer indicated excess reactivity. The operators blocked the last remaining trip signal just before it would have scrammed the reactor, in order to be able to complete the test. Power started to rise and coolant voiding in the pressure tubes occurred, leading to a positive reactivity input which enhanced the power rise. The operators began control rod insertion from the fully withdrawn position. However, the fully withdrawn control rods had graphite followers below the control poison (to enhance rod worth), and these entered the active core first, displacing neutron-absorbing water with graphite and thus adding further positive reactivity, which accelerated the power increase. The power surged to 100 times design full power in the next 4 s, then decreased momentarily. There then followed repeated power pulses, one of which may have reached 500 times design full power. The fuel disintegrated, breached the cladding, and entered the water coolant, causing a steam explosion that lifted the top shield of the reactor core, shearing all the coolant pipes and removing all the control rods. The explosion was well beyond the rather modest containment design basis and penetrated the concrete walls of the reactor building, dispersing burning fuel and graphite, and releasing a plume of radioactive gases and particles. The accident resulted in 31 early fatalities. Over 1000 people received large doses of radiation. Many of the nearby population received doses greater that 0.25 Sv 297 298 8 Reactor Safety (25 rem), with the most serious in the range 0.4 to 0.5 Sv (40 to 50 rem). As a reference, recommended annual dose limits by the International Council on Radiation Protection are 50 mSv (5 rem) whole-body radiation and 500 mSv (50 rem) for any body part other than the lens of the eye. The radioactivity released into the atmosphere fell out in measurable amounts over much of the world. Estimated individual whole-body doses immediately following the accident were on the order of 100 mGy (10 rad) in the immediate vicinity of the plant, 4 mGy (400 mrad) in Poland, 1 mGy (100 mrad) in the rest of Europe, and 0.01 mGy (1 mrad) in Japan and North America. The 24,000 evacuees who received an estimated average dose of 0.43 Sv were expected to incur an additional 26 fatal leukemias over the next decade, roughly doubling the natural incidence of leukemia fatalities in that population. On a long-term basis, the predicted collective lifetime doses due to the fallout from the Chernobyl accident are 1.6 × 104 person-Gy for the evacuated population near the site, 4.7 × 105 person-Gy for the European part of the former USSR, 1.1 × 105 person-Gy for the Asian part of the former USSR, 5.8 × 105 person-Gy for Europe, 2.7 × 104 person-Gy for Asia, 1.1 × 103 person-Gy for the United States, and 1.2 × 106 person-Gy for the entire northern hemisphere. The increase in the estimated 50-year exposure doses in Europe, for example, varied from a fraction of the natural background to a few times the natural background. There is no scientific evidence on which to assess the effect, if any, of such small incremental doses. However, by extrapolating from higher dose levels, it is possible to estimate the long-term health effects of fallout from the Chernobyl accident. The estimated increase above natural incidence of fatal cancers in the respective populations due to the Chernobyl fallout is 2.4% for the evacuated population near the site, 0.12% for the European part of the former USSR, 0.01% for the Asian part of the former USSR, 0.02% for Europe, 0.00013% for Asia, and 0.00005% for the northern hemisphere. Postaccident assessments identified design-related defects as (1) positive coolant void reactivity coefficient, (2) easy-to-block safety systems, (3) slow scram (15 to 20 s for full insertion, 5 s for effective negative reactivity), and (4) absence of containment and emergency fission product control systems. These design-related defects are uniquely applicable to the RBMK reactors, which are deployed only in the former Soviet Union. Technical fixes that have been implemented subsequently on other RBMK reactors include (1) maximum allowable control rod withdrawal limitations, (2) modifications to prevent operators from manually overriding safety systems, (3) reduction of the positive coolant void reactivity coefficient, and (4) development of an alternative shutdown capability. Operator error and lax management were obviously at least partially responsible for the Chernobyl accident, and the government placed much of the blame there. Six members of plant management were subsequently tried and convicted for violation of safety rules, criminal negligence, and so on, and the station director, chief engineer, and deputy chief engineer were sentenced to 10 years in a labor camp. However, the positive coolant temperature coefficient and the absence of a 8.5 Passive Safety containment building designed to withstand overpressure events were also major contributors to the accident. 8.5 Passive Safety The experience of TMI-2 and Chernobyl has led to an emphasis on passive safety in the design of advanced reactors. Broadly speaking, the objectives of passive safety design are, to the extent possible, for the reactor to be able to maintain a balance between power production and heat removal, to shut itself down when an abnormal event occurs, and to remove decay heat, without requiring operator action or the functioning of engineered safety systems. Pressurized Water Reactors The AP-600 design features a passive emergency core cooling system consisting of water stored in large tanks above the core. During a loss-of-coolant accident, this water is injected into the core while the coolant system is still pressurized, and flows into the core under gravity when the system depressurizes, without requiring either pumps or electrical power. Decay heat, which is normally removed through the steam generators, would be removed by the natural circulation of water through the core into a large tank above the reactor vessel in the event that the steam generators were inoperable. The containment shell is cooled by gravity-driven water spray and the natural circulation of air. Because of reliance on passive safety, there are only half the number of large pumps as on a standard PWR. The PIUS reactor vessel, pressurizer, and steam generators are all immersed in borated water. If a pump fails during normal operation, the hydrostatic pressure forces the borated water into the core, where it serves both as emergency coolant and a shutdown mechanism. The natural circulation between the core and the pool of borate water would remove decay heat. Boiling Water Reactors Main coolant flow for the boiling water reactor (SWBR) design is provided by natural circulation, eliminating the need for the recirculation pumps, valves, and associated controls of a standard boiling water reactor. In the event of a loss-of-coolant accident, steam is vented into a large suppression pool located above the core to depressurize the cooling system, which allows water from the pool to gravity flow down into the core to provide emergency core cooling. Decay heat can be removed to the suppression pool by natural circulation. The entire system is enclosed in a concrete containment structure that is cooled continuously by water flow downward from the suppression pool, the evaporation of which provides passive heat removal from the core to the atmosphere. 299 300 8 Reactor Safety Integral Fast Reactors The approach to safety embodied in the integral fast reactor (IFR) includes (1) large design margins between operating conditions and safety limits, (2) reliance on passive processes to hold power production in balance with heat removal, and (3) totally passive removal of decay heat. The IFR can be designed to achieve passive power regulation, even should equipment in the control and balance-of-plant systems fail, for anticipated transient without scram scenarios. The heat transport system that removes decay heat operates at ambient pressure, has large thermal inertia, is driven by natural convection, is contained along with the core in a double-wall top-entry coolant tank, is completely independent of the balance of plant equipment, and is always in operation. The IFR system can be designed to have an inherent response that prevents release of radioactivity, even for accidents of extremely low probability far below the design basis level. Processes that are innate consequences of the materials and geometry cause dispersal of fuel early enough to avoid prompt criticality and the accompanying energy release and to ensure subcriticality and coolability inside an intact reactor vessel should significant fuel pin failures cause an accumulation of radioactive debris. Passive Safety Demonstration The passive safety features of the IFR have been demonstrated dramatically in a series of tests in the Experimental Breeder Reactor II (EBR-II), which has the same type of fuel and heat transport system as the IFR. It was demonstrated that the reactor operating at full power would be safely shut down by negative reactivity feedback, without benefit of the scram or any other safety system or of operator action, upon loss of forced coolant flow and upon loss of heat sink, two of the most demanding reactor accident scenarios. Transient temperatures during shutdown were measured to be below those of concern for fuel integrity and reactor safety. In the first test, the coolant pumps were shut off while the reactor was operating at full power with the scram system deactivated (a separate emergency scram system was operable but not used). No operator action was taken. The response of EBR-II to the loss of coolant flow is shown in Fig. 8.6. The negative reactivity feedback associated with the increase in coolant temperature following the loss of coolant flow resulted in a rapid reduction in power, which reduced the coolant temperature. Because the metal fuel has a large heat conductivity and operates at a temperature only slightly greater than that of the coolant, there is a relatively small negative Doppler reactivity coefficient and consequently, relatively little positive reactivity addition when the coolant temperature decreases later in the transient. In the second test, the ability of the system to reject heat from the primary coolant was eliminated while the reactor was at full power, with the scram system deactivated and no operator action taken. The response of EBR-II to the loss of heat sink is shown in Fig. 8.7. Again, negative reactivity feedback shut the reactor down without any danger to the plant. 8.5 Passive Safety Fig. 8.6 Response of EBR-II to loss of flow (discontinuities in reactivity are artificial). (From Ref. 6.) Fig. 8.7 Response of EBR-II to loss of heat sink (discontinuities in reactivity are artificial). (From Ref. 6.) 301 302 8 Reactor Safety References 1 D. C. Wade, R. A. Wigeland, and D. J. Hill, “The Safety of the IFR,” Prog. Nucl. Energy. 31, 63 (1997). 2 R. A. Knief, Nuclear Engineering, Taylor & Francis, Washington, DC (1992), Chaps. 13–16. 3 H. Cember, Introduction to Health Physics, 3rd ed., McGraw-Hill, New York (1996). 4 K. E. Carlson et al., RELAP5/MOD3 Code Manual, Vols. I and II, EG&G Idaho report NUREG/CR-5535, U.S. Nuclear Regulatory Commission, Washington, DC (1990). 5 Severe Accident Risks: An Assessment for Five U.S. Nuclear Plants, NUREG1150, U.S. Nuclear Regulatory Commission, Washington, DC (1989); Nucl. Eng. Des. 135, 1–135 (1992). 6 S. Fistedis, “The Experimental Breeder Reactor-II Inherent Safety Demonstration,” Nucl. Eng. Des. 101, 1 (1987); J. I. Sackett, “Operating and Test Experience with EBR-II, the IFR Prototype,” Prog. Nucl. Energy 31, 111 (1997). 7 J. G. Collier and G. F. Hewitt, Introduction to Nuclear Power, Hemisphere Publishing, Washington, DC (1987), Chaps. 4–6. 8 “Special Issue: Chernobyl,” Nucl. Safety 28 (1987). 9 “Chernobyl: A Special Report,” Nucl. News, 29, 87 (1986); “Chernobyl: The Soviet Report,” Nucl. News, Special Report (1986). 10 Report on the Accident at the Chernobyl Nuclear Power Station, NUREG-1250, U.S. Nuclear Regulatory Commission, Washington, DC (1987). 11 “The Ordeal at Three Mile Island,” Nucl. News, Special Report (1979). 12 The TMI-2 Lessons Learned Task Force Final Report, NUREG-0585, U.S. Nuclear Regulatory Commission, Washington, DC (1979). 13 WASH-1400, Reactor Safety Study: An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants, NUREG-74/014, U.S. Nuclear Regulatory Commission, Washington, DC (1975). 14 T. J. Burnett, Nucl. Sci. Eng. 2, 382 (1957). 15 T. J. Thompson and J. G. Beckerley, eds., The Technology of Nuclear Reactor Safety, MIT Press, Cambridge, MA (1964). Problems 8.1. Discuss the differences in reactivity feedback between the Three-Mile Island and Chernobyl accidents. 8.2. Discuss the important differences between a modern LWR and the Chernobyl RBMK reactors that make it extremely unlikely that a Chernobyl-type accident, with a large-scale release of radioactivity, could occur in a LWR. 8.3. What design changes would you make to the Three-Mile Island PWR reactor? 8.4. What design changes would you make to the Chernobyl RBMK reactor? Part 2: Advanced Reactor Physics 305 9 Neutron Transport Theory Calculation of the transport of neutrons and their interaction with matter are perhaps the fundamental topics of reactor physics. In this chapter, the major computational methods used for the transport of neutrons in nuclear reactors are described. 9.1 Neutron Transport Equation The distribution of neutrons in space and angle is defined by the particle distribution function N(r, , t), such that N(r, , t)dr d is the number of neutrons in volume element dr at position r moving in the cone of directions d about direction , as depicted in Fig. 9.1. An equation for N(r, , t) can be derived by considering a balance on the differential cylindrical volume element of length dl = v dt , where v is the neutron speed, and cross-section area dA surrounding the direction of neutron motion, as shown in Fig. 9.2. The rate of change of N(r, , t) within this differential volume is equal to the rate at which neutrons with direction  are flowing into the volume element (e.g., across the left face in Fig. 9.2) less the rate at which they are flowing out of the volume element (e.g., across the right face), plus the rate at which neutrons traveling in direction  are being introduced into the volume element by scattering of neutrons within the volume element from different directions  and by fission, plus the rate at which neutrons are being introduced into the volume element by an external source Sex , minus the rate at which neutrons within the volume element traveling in direction  are being absorbed or being scattered into a different direction  : ∂N (r, , t)dr d ∂t   = v N(r, , t) − N(r +  dl, , t) dA d  4π + d s (r,  → )vN(r,  , t)dr d 0  4π 1 + d νf (r)vN(r,  , t)dr d 4π 0   + Sex (r, )dr d − a (r) + s (r) vN(r, , t)dr d (9.1) 306 9 Neutron Transport Theory Fig. 9.1 Particles in dr at location r moving in the cone d about the direction . (From Ref. 2; used with permission of Wiley.) Making a Taylor’s series expansion ∂N(r, , t) dl + · · · ∂l = N(r, , t) +  · ∇N(r, , t) + · · · N(r +  dl, , t) = N(r, , t) + (9.2) to evaluate the streaming term, denning the directional flux distribution ψ(r, , t) ≡ vN(r, , t) Fig. 9.2 Incremental volume element for particles at location r moving in the direction  (From Ref. 2; used with permission of Wiley.) (9.3) 9.1 Neutron Transport Equation and taking note of the fact that the scattering from  to  depends only on  ·  ≡ μ0 , so that s (r,  → ) = 1 1 s (r,  ·  ) ≡ s (r, μ0 ) 2π 2π (9.4) and writing t = a + s , leads to the neutron transport equation 1 ∂ψ (r, , t) +  · ∇ψ(r, , t) + t (r)ψ(r, , t) v ∂t  1 dμ0 s (r, μ0 )ψ(r,  , t) = −1 + 1 4π  4π d νf (r)ψ(r, , t) + Sex (r, ) ≡ S(r, ) (9.5) 0 The representation of the neutron streaming operator,  · ∇ψ , in the common geometries is given in Table 9.1, and the respective coordinate systems are defined in Figs. 9.3 to 9.5. Fig. 9.3 Cartesian space–angle coordinate system. (From Ref. 2; used with permission of Wiley.) 307 308 9 Neutron Transport Theory Fig. 9.4 Spherical space–angle coordinate system. (From Ref. 2; used with permission of Wiley.) Fig. 9.5 Cylindrical space–angle coordinate system. (From Ref. 2; used with permission of Wiley.)  · ∇ψ μ η = (1 − μ2 )1/2 cos ω; ξ = (1 − μ2 )1/2 sin ω Streaming Operator in Rectangular Coordinates ∂ψ μ ∂x ∂ψ ∂ψ μ, η μ +η ∂x ∂y ∂ψ ∂ψ ∂ψ μ, η, ξ μ +η +ξ ∂x ∂y ∂z Streaming Operator in Cylindrical Coordinates in Conservation Form 1 ∂ μ ∂ (ρψ) − (ηψ) ω, ξ ρ ∂ρ ρ ∂ω μ ∂ η ∂ψ 1 ∂ ω, ξ (ρψ) − − (ηψ) ρ ∂ρ ρ ∂θ ρ ∂ω μ ∂ ∂ψ 1 ∂ ω, ξ (ρψ) + ξ − (ηψ) ρ ∂ρ ∂z ρ ∂ω μ ∂ η ∂ψ ∂ψ 1 ∂ ω, ξ (ρψ) − +ξ − (ηψ), μ = (1 − ξ 2 )1/2 cos ω; η = (1 − ξ 2 )1/2 sin ω ρ ∂ρ ρ ∂θ ∂z ρ ∂ω Streaming Operator in Spherical Coordinates in Conservation Form 1 ∂ μ ∂ 2 μ (ρ ψ) + [(1 − μ2 )ψ] ρ ∂μ ρ 2 ∂ρ μ ∂ 2 η cot θ ∂ ∂ 1 ∂ μ, ω (ρ ψ) + (sin θψ) + [(1 − μ2 )ψ] − (ξ ψ) ρ sin θ ∂θ ρ ∂μ ρ ∂ω ρ 2 ∂ρ ∂ η cot θ ∂ ξ ∂ψ 1 ∂ μ ∂ 2 (ρ ψ) + (sin θψ) + + [(1 − μ2 )ψ] − (ξ ψ) μ, ω ρ sin θ ∂θ ρ sin θ ∂ϕ ρ ∂μ ρ ∂ω ρ 2 ∂ρ Angular Variables Source: Data from Ref. 2; used with permission of Wiley. ρ, θ , ϕ ρ, θ ρ ρ, θ , z ρ, z (three dimensions) ρ, θ (two dimensions) ρ (one dimension) x, y, z (three dimensions) x, y (two dimensions) x (one dimension) Spatial Variables Table 9.1 Neutron Streaming Operator in Conservative Form 9.1 Neutron Transport Equation 309 310 9 Neutron Transport Theory Boundary Conditions Boundary conditions for Eq. (9.5) are generally specified by the physical situation. For a left boundary at rL with inward normal vector n, such that n ·  > 0 indicates inward, one of the following boundary conditions is usually appropriate: ·n>0 ψ(rL , ) = 0, ·n>0 ψ(rL , ) = ψin (rL , ), 4π ψ(rL , ) = 0 α( → )ψ(rL ,  )d Vacuum: Incident flux known: Reflection: (9.6) where α is a reflection or albedo function. Scalar Flux and Current The scalar flux is the product of the total number of neutrons in a differential volume, which is the integral over direction of the number of neutrons with direction within d about , times the speed:  φ(r) ≡ 4π d ψ(r, ) (9.7) 0 and the current with respect to the ξ -coordinate is the net flow of neutrons in the positive ξ -direction:  4π Jξ (r) ≡ nξ d (nξ · )ψ(r, ) (9.8) 0 Partial Currents The positive and negative partial currents, with respect to the ξ -direction, are the total neutron flows in the positive and negative ξ -directions, respectively: J+ ξ (r) ≡ nξ J− ξ (r) ≡ nξ   2π 1 dφ  0  2π dφ 0 dμ(nξ · )ψ(r, ) 0 0 −1 (9.9) dμ(nξ · )ψ(r, ) 9.2 Integral Transport Theory The steady-state version of Eq. (9.5) may be written d ψ(r, )dr d + t (r)ψ(r, )dr d = S(r, )dr d dR (9.10) 9.2 Integral Transport Theory Fig. 9.6 Incremental volume subtended by cone d at distance R = |r − r | from point r. (From Ref. 2; used with permission of Wiley.) where dR is the differential length along the direction  (i.e.,  · ∇ = d/dR). This equation may be integrated along the direction  from r0 to r, to obtain  r  ψ(r, )dr = e−α(r0 ,r) ψ(r0 , )dr0 + e−α(r ,r) S(r , )dr (9.11) r0 where α(r , r) is the optical path length along the direction  between r and r:  r t (R) dR α(r , r) ≡ (9.12) r Isotropic Point Source For an isotropic point source of strength S0 (n/s) located at r0 , the directional flux outward through the cone d about direction  is S0 (d/4π). The volume element dr subtended by this cone at distance R = |r − r | away is 4πdR 2 dR, as depicted in Fig. 9.6. From Eq. (9.11), the directional flux at r of uncollided neutrons from an isotropic point source at r (such that the direction from r to r is ) is given by    S0 e−α(r,r ) S0 e−α(R,0) = ψpt (R) = ψ |(r − r )|,  =  2 4π|r − r | 4πR 2 (9.13) Isotropic Plane Source The scalar flux of uncollided neutrons at a distance x from a uniform planar isotropic source can be constructed by treating each point in the plane as an isotropic point source and integrating over the plane, as indicated in Fig. 9.7, to obtain  ∞  ∞ 2πρψpt (R) dρ = 2πRψpt (R) dR φpl (x, 0) = 0 = S0 2  x x ∞ e−α(R,0)   dR 1 = S0 E1 α(x, 0) R 2 (9.14) 311 312 9 Neutron Transport Theory Fig. 9.7 Coordinate system for plane isotropic source calculation. (From Ref. 10; used with permission of McGraw-Hill.) where the exponential integral function is defined as  ∞ En (y) ≡ du e −yu −n u  1 = 1 dμ e−y/μ μn−2 (9.15) 0 The x-direction current of uncollided neutrons at a distance x from a uniform planar isotropic source can be constructed in a similar manner by noting that for a neutron originating on the plane with direction , the quantity μ =  · nx = x/R:  Jpl,x (x, 0) = ∞ 0 S0 = 2  x 2πρ ψpt (R) dρ = x R x ∞  ∞ x 2πRψpt (R) dR R   dR 1 e−α(R,0) 2 = S0 E2 α(x, 0) 2 R (9.16) A one-dimensional isotropic source distribution S0 (x) in a slab of thickness a can be considered as a distribution of isotropic planar sources, and the uncollided scalar flux distribution can be constructed by integrating over the contributions from each planar source:  φ(x) = 0 a 1 S0 (x )φpl (x, x )dx = 2     a   S0 (x  )E1 α(x, x  ) dx  (9.17) 0 Anisotropic Plane Source Using the relations μ = cos θ = x/R and R 2 = x 2 + ρ 2 and noting that all source neutrons in the annular region 2πρ dρ on the source plane will pass through a point at a distance x above the center of the annular region within dμ about the 9.2 Integral Transport Theory same value of μ, the directional flux of uncollided neutrons which results from an anisotropic planar source S(μ) can be constructed:   S(μ)e−α(x,0)/μ ψ(x, μ) dμ = ψpt R(ρ) 2πρ dρ = dμ μ (9.18) The scalar flux and current of uncollided neutrons at a distance x from an uniform anisotropic planar source S(μ) are  1  1 dμ φ(x) ≡ (9.19) ψ(x, μ) dμ = S(μ)e−α(x,0)/μ μ −1 0  1  1 Jx (x) ≡ μψ(x, μ) dμ = S(μ)e−α(x,0)/μ dμ (9.20) −1 0 It is convenient to expand the directional dependence of the source:  (2n + 1)pn+ (μ)Sn S(μ) = (9.21) n=0 in half-range Legendre polynomials: pn+ (μ) = Pn (2μ − 1) p0+ (μ) = 1, p1+ (μ) = 2μ − 1, p3+ (μ) = 20μ3 p2+ (μ) = 6μ2 − 6μ + 1, − 30μ2 + 12μ − 1, (9.22) etc. which have the orthogonality properties  1 δnm + pn+ (μ)pm (μ)dμ = 2n +1 0 (9.23) With these orthogonality properties, it follows immediately that Sn = 0 pn+ (μ) × S(μ)dμ. Using this expansion in Eq. (9.19), the flux of uncollided neutrons at a distance x from an uniform anisotropic planar source is  (2n + 1)Sn Bn+ (α(x, 0)) (9.24) φ(x) = 1 n=0 where  1   dμ pn+ (μ)e−α(x,0)/μ Bn+ α(x, 0) ≡ μ 0    B0+ α(x, 0) = E1 α(x, 0)       B1+ α(x, 0) = 2E2 α(x, 0) − E1 α(x, 0) , (9.25) etc. Similarly, the x-directed current of uncollided neutrons at a distance x from an uniform anisotropic planar source is  (2n + 1)Sn L+ (9.26) Jx (x) = n (α(x, 0)) n=0 313 314 9 Neutron Transport Theory where  1   L+ pn+ (μ)e−α(x,0)/μ dμ α(x, 0) = n 0     L+ 0 α(x, 0) = E2 α(x, 0)       L+ 1 α(x, 0) = 2E3 α(x, 0) − E2 α(x, 0) , (9.27) etc. Transmission and Absorption Probabilities As an example of an application of the formalism above, consider a purely absorbing slab of thickness a with an isotropic plane source of neutrons on one surface. The transmission probability for the slab is just the ratio of the exiting current on the opposite surface to the incident partial current on the other surface: T=   S0 L+ J (a) 0 (α(a, 0)) = = E2 α(a, 0) Jin (0) S0 (9.28) and the absorption probability is A = 1 − T = 1 − E2 (α(a, 0)). Escape Probability As another example, consider a uniform, purely absorbing slab of thickness a with an isotropic neutron source S0 distributed uniformly throughout. Representing the source of neutrons at x within the slab as a plane isotropic source of strength S0 /2 to the right and S0 /2 to the left, the current of neutrons produced by the source at x = x  which exit through the surface at x = a is  1      Jout (a : x  ) = 12 S0 L+ 0 α(a, x ) = 2 S0 E2 α(a, x ) (9.29) The total current of neutrons out through the surface at x = a is found by integrating this expression over the slab:  a Jout (a) = 0 1 dx  Jout (a : x  ) = S0 2  a   dx  E2 α(a, x  ) (9.30) 0 Using the differentiation property of the exponential integral function dEn = −En−1 (y), dy n = 1, 2, 3, . . . (9.31) Eq. (9.30) may be evaluated: 1 Jout (a) = − S0 2  0 a dE3 (α)  1 S0 dx = − dα 2 t  0 α(a,0) dE3 dα dα       1 S0 1 S0 1 = − E3 α(a, 0) E3 (0) − E3 α(a, 0) = 2 t 2 t 2 (9.32) 9.2 Integral Transport Theory By symmetry, the current out through the surface at x = 0 must be the same. The escape probability from the slab is the ratio of the total current out of the slab through both surfaces to the total neutron source rate aS0 in the slab: P0 =   Jout (a) + Jout (0) 1 1 − E3 (at ) = aS0 at 2 (9.33) First-Collision Source for Diffusion Theory As a further application, consider a medium with a surface source of neutrons, which is highly forward directed but almost isotropic within the forward-directional hemisphere, incident on one surface of a diffusing medium; that is, the forwarddirected neutrons incident on the medium are nearly isotropic within the forwarddirectional hemisphere, but many more neutrons are moving forward into the medium than are moving backward out of it. Diffusion theory will not be accurate for treating these source neutrons, because diffusion theory is based on an implicit assumption that the neutron flux is nearly isotropic over the full angle (this is discussed in Section 9.6), even though diffusion theory may otherwise be sufficient for the analysis of neutrons once their direction is randomized by a scattering event within the medium. The first collision of the incident source neutrons can be calculated with integral transport theory and used as a distributed first-collision source for the diffusion theory calculation:   Sfc (x) = s (x)φ(x) = S0 s (x)E1 α(x, 0) (9.34) If the distribution of incident source neutrons is more highly forward directed, so that it is anisotropic even over the forward-directional hemisphere, it may be represented by an anisotropic plane source and the first-collision source becomes Sfc (x) = s (x)    (2n + 1)Sn Bn+ α(x, 0) (9.35) n=0 Inclusion of Isotropic Scattering and Fission Consider again the slab with a distributed isotropic source of neutrons, but now with isotropic elastic scattering and fission, as well as absorption represented explicitly. The flux of uncollided source neutrons is φ0 (x) = 1 2  a   S0 (x  )E1 α(x, x  ) dx  (9.36) 0 If the first-collision rate at x = x  is considered as a plane isotropic source of oncecollided neutrons at x  , the flux of once-collided neutrons due to the once-collided source at x  is φ1 (x : x  ) = 1 2     s (x  ) + νf (x  ) φ0 (x  )E1 α(x, x  ) (9.37) 315 316 9 Neutron Transport Theory and the total flux of once-collided neutrons at x is found by integrating over the distribution of first-collision sources:  a φ1 (x : x  )dx  φ1 (x) = 0 =  a 0     1 s (x  ) + νf (x  ) φ0 (x  ) E1 α(x, x  ) dx  2 (9.38) Continuing in this vein, the flux of n-collided neutrons is given by   a    1 s (x  ) + νf (x  ) φn−1 (x  ) · E1 α(x, x  ) dx  , φn (x) = 2 0 n = 1, 2, 3, . . . , ∞ (9.39) The total neutron flux is the sum of the uncollided, once-collided, twice-collided, and so on, fluxes: φ(x) ≡ φ0 (x) + ∞  φn (x) n=1  ∞    1 φn−1 (x  )E1 α(x, x  ) dx  s (x  ) + νf (x  ) 0 2 n=1  a    1 + S0 (x  )E1 α(x, x  ) dx  2 0  a    1 = s (x  ) + νf (x  ) φ(x  )E1 α(x, x  ) dx  2 0    1 a + S0 (x  )E1 α(x, x  ) dx  2 0 = a (9.40) Thus we have found an integral equation for the neutron flux in a slab with isotropic scattering and fission, with a kernel 12 [s (x  ) + νf (x  )]E1 (α(x, x  )) and a first-collision source 12 S0 E1 (α(x, 0)). Distributed Volumetric Sources in Arbitrary Geometry The scalar flux of uncollided neutrons resulting from an arbitrary neutron source distribution can be constructed by treating each spatial location as a point source with strength given by the source distribution for that location. The uncollided directional flux at r arising from a point source at r is given by Eq. (9.13). The total uncollided directional flux at r is obtained by integrating over all source points r , and the total uncollided scalar flux is then calculated by integrating over :   S0 (r )e−α(r,r ) (9.41) φun (r) = dr 4π|r − r|2 Following the same development as that leading to Eq. (9.40), an integral equation for the total neutron flux can be developed for the case of isotropic scattering: 9.2 Integral Transport Theory  φ(r) =  = dr [s (r ) + νf (r )]φ(r )e 4π|r − r|2 −α(r,r ) + φun (r )  , e−α(r,r ) + dr s (r ) + νf (r ) φ(r ) + S0 (r ) 4π|r − r|2 (9.42) where exp[−α(r, r )]/4π|r − r |2 is the isotropic point source kernel and φun given by Eq. (9.41) is the uncollided flux contribution. The derivations leading to Eqs. (9.40) and (9.42) did not explicitly take boundary conditions into account. Since scattering source rates integrated over the volume of the reactor were used to derive successive n-collided fluxes, the implicit assumption was that neutrons which escaped from the reactor did not return. Thus these equations are valid with vacuum boundary conditions, but not with reflective boundary conditions. Flux from a Line Isotropic Source of Neutrons Consider the situation illustrated in Fig. 9.8 of a line isotropic source of neutrons of strength S0 (n/cm · s). The point source kernel can be used to construct the differential scalar flux at a point P located a distance t from the line source due to the differential element dz of the line source located at z: dφ(t) = S0 dz e−α(t,z) 4πR 2 Fig. 9.8 Geometry for calculating flux at P from a line isotropic neutron source [t = x, τ = α(x, 0)]. (From Ref. 3; used with permission of Academic Press.) (9.43) 317 318 9 Neutron Transport Theory where α(t, z) denotes the optical thickness along the path of length R from the source point at coordinate z to the point P a perpendicular distance t from the line source at z = 0. Noting that R = t/ cos θ and dz = R dθ/ cos θ = t dθ/ cos2 θ , the total flux at a point at a distance t can be found by integrating the differential flux contribution from all such differential elements dz:  ∞  ∞ S0 dz e−α(t,z) dz e−α(t,0)/ cos θ Ki1 [α(t, 0)] φ(t) = ≡ = S S0 0 2 2 2 2πt 4πR 2πt / cos θ −∞ 0 (9.44) where Ki1 (x) is the Bickley function of order one. Bickley Functions The general Bickley function is defined as  π/2  Kin (x) ≡ cosn−1 θe−x/ cos θ dθ = 0 ∞ 0 e−x cosh(u) du coshn (u) (9.45) These functions satisfy the following differential and integral laws: dKin (x) = −Kin−1 (x) dx  x  Kin−1 (x  )dx  = Kin (x) = Kin (0) − 0 x (9.46) ∞ Kin−1 (x  )dx  (9.47) and the recurrence relation nKin+1 (x) = (n − 1)Kin−1 (x) + x[Kin−2 (x) − Kin (x)] (9.48) The Bickley functions must be evaluated numerically (e.g., Ref. 3). Probability of Reaching a Distance t from a Line Isotropic Source without a Collision With reference to Fig. 9.9, the probability P that a neutron emitted isotropically from point P on the line source is able to get a perpendicular distance t away from the line source without having a collision depends on the direction in which the neutron is traveling relative to the perpendicular to the line source. The uncollided differential neutron current arising from a point on the line source and passing through a differential surface area dA = R dθt dϕ = t 2 dθ dϕ/ cos θ normal to the R-direction at a perpendicular distance t from the line source is dJ (t, θ) = e−α(t,0)/ cos θ t 2 dθ dϕ e−α(t,z) dA = 4πR 2 4π(t/ cos θ)2 cos θ (9.49) where the optical thickness α(t, z) is taken along the path length R. Integrating over all possible values of the angles, the probability of a neutron emitted isotropically from a line source crossing the cylindrical surface at a distance t from the line source is  π/2  2π   e−α(t,0)/ cos θ (t 2 / cos θ) dϕ dθ = Ki2 α(t, 0) (9.50) P (t) = 2 4π(t/ cos θ) 0 −π/2 9.3 Collision Probability Methods Fig. 9.9 Geometry for calculating probability that a neutron from an isotropic line source does not have a collision within perpendicular distance t from the line source [t = x, τ = α(x, 0)]. (From Ref. 3; used with permission of Academic Press.) where now α(t, 0) is the optical path length perpendicular to the line source out to the cylindrical surface at distance t . The Bickley and exponential integral functions arise because of the assumption of spatial symmetry. They take into account that the neutron flight path is always in three spatial dimensions, even though symmetry otherwise allows reduction in the dimensionality of the problem. 9.3 Collision Probability Methods If the volume of the problem of interest is partitioned into discrete volumes, Vi , within each of which uniform average cross sections and a flat flux are assumed, Eq. (9.42) can be integrated over Vi , and the resulting equation can be divided by Vi to obtain φi =  T j →i [(sj + νfj )φj + S0j ] (9.51) j which relates the fluxes in the various volumes by the first-flight transmission probabilities T j →i : T j →i ≡ 1 Vi   dri Vi drj Vj e−α(ri ,rj ) 4π|ri − rj |2 (9.52) 319 320 9 Neutron Transport Theory Reciprocity Among Transmission and Collision Probabilities Since α(ri , rj ) = α(rj , ri ) (i.e., the optical path is the same no matter which way the neutron traverses the straight-line distance between ri and r) there is a reciprocity relation between the transmission probabilities: Vi T j →i = Vj T i→j (9.53) Upon multiplication by ti Vi , Eq. (9.51) can be written ti Vi φi =  P ji (sj + νfj )φj + S0j (9.54) tj j where the collision rate in cell i is related to the neutrons introduced by scattering, fission, and an external source in all cells j by the collision probabilities P ji ≡ tj ti Vi T j →i  = tj ti  Vi dri Vj drj e−α(ri ,rj ) 4π|ri − rj |2 (9.55) Because α(ri , rj ) = α(rj , ri ), there is reciprocity between the collision probabilities; that is, P ij = P j i (9.56) Collision Probabilities for Slab Geometry For a slab lattice the volumes, V i , become the widths i ≡ xi+1/2 − xi−1/2 of the slab regions centered at xi , and the slab kernel E1 (α(x  , x))/2 replaces the point source kernel in Eq. (9.55), which becomes  P j i = ti tj dx  i  j  1  dx E1 α(x  , x) 2 (9.57) For j = i, the probability that a neutron introduced in cell j has its next collision in cell i is P ji = 1 2 E3 (αi+1/2,j +1/2 ) − E3 (αi−1/2,j +1/2 ) − E3 (αi+1/2,j −1/2 ) + E3 (αi−1/2,j −1/2 ) (9.58) where αi,j ≡ α(xi , xj ). For j = i, a similar development leads to an expression for the probability that a neutron introduced in cell j has its next collision in cell i is  P ii = ti i 1 −   1  1 − 2E3 (ti i ) 2ti i (9.59) 9.3 Collision Probability Methods Fig. 9.10 Geometry for calculating collision probabilities in two-dimensional geometry (τ is the optical path length α over the indicated path). (From Ref. 3; used with permission of Academic Press.) Collision Probabilities in Two-Dimensional Geometry Consider the two-dimensional cross section shown in Fig. 9.10, in which the volumes Vi and Vj extend indefinitely in the direction perpendicular to the page. With respect to Fig. 9.9, a neutron emitted at point t defined by the angle ϕ and coordinate y in volume Vi in Fig. 9.10 and traveling in the direction defined by the angle ϕ which passes through volume Vj may be traveling at any angle −π/2 ≤ θ ≤ π/2 with respect to the horizontal cross section shown in Fig. 9.10. The probability that a neutron emitted at point t will reach some point on the line perpendicular to the page which passes through the page at point t  in volume Vj is, from Eq. (9.50), given by Ki2 (α(t  , t)), where α(t  , t) is the optical path length in the horizontal plane of Fig. 9.10. With respect to Fig. 9.10, identify ti and tj as the points along the horizontal line between t and t  at which the line passes through the surfaces of volumes Vi and Vj , respectively. Thus Ki2 (ti (ti − t) + α(tj , ti )) is the probability that a neutron emitted from point t in volume Vi in direction ϕ reaches volume Vj , and Ki2 (ti (ti −t)+α(tj , ti )+α(tj +tj , tj )), with tj being the distance in the horizontal plane across volume Vj , is the probability that the neutron not only reaches volume Vj but continues through volume Vj and emerges from the opposite side without a collision, both probabilities being averaged over an isotropic distribution of neutron directions with respect to the horizontal, as measured by the angle θ . The probability that neutrons emitted from point t in volume Vi with direction ϕ have their first collision in volume Vj is then pij (t, ϕ, y) = −Ki2 (ti (ti − t) + α(tj , ti ) + α(tj + tj , tj )) + Ki2 (ti (ti − t) + α(tj , ti )). Averaging this probability over all source points along the line defined by angle ϕ within volume Vi and using the differential property of the Bickley functions given by Eq. (9.45) leads to pij (ϕ, y) = 1 ti  ti pij (t, ϕ, y)dt 0 321 322 9 Neutron Transport Theory =     1 Ki3 α(tj , ti ) − Ki3 α(tj , ti ) + α(tj + tj , tj ) ti ti    − Ki3 α(tj , ti ) + α(ti , 0) + Ki3 α(tj , ti ) + α(tj + tj , ti )  + α(ti , 0) (9.60) To obtain the average probability P ij that a neutron introduced by an isotropic source uniformly distributed over volume Vi will have its first collision in volume Vj , this expression must be multiplied by the probability that an isotropically emitted neutron source will emit a neutron in the differential direction dϕ about ϕ, which is dϕ/2π , and the probability that for a uniform source within Vi the neutron will be emitted from along the chord of length ti (y) at coordinate y, which is ti (y)dy/Vi , and integrated over all relevant values of ϕ and y. Note that the volumes Vi and Vj are actually the respective areas within the planar cross section of Fig. 9.10. The result for the collision probability is 1 2π pij =   ϕmax ymax (ϕ) dϕ   dy Ki3 α(tj , ti ) ymin (ϕ) ϕmin    − Ki3 α(tj , ti ) + α(tj + tj , tj ) − Ki3 α(tj , ti ) + α(ti , 0)   + Ki3 α(tj , ti ) + α(tj + tj , tj ) + α(ti , 0)  (9.61) A similar development leads to an expression for the probability that the next collision for a neutron introduced in volume Vi is within that same volume Vi : P ii = ti Vi − 1 2π   ϕmax ymax (ϕ) dϕ ϕmin ymin (ϕ)   dy Ki3 (0) − Ki3 α(ti , 0) (9.62) Collision Probabilities for Annular Geometry The annular geometry of a fuel pin, its clad, and the surrounding moderator is of particular interest. For the annular geometry of Fig. 9.11, Eq. (9.61) specializes to P ij = δij tj Vj + 2(Si−1,j −1 − Si−1,j − Si,j −1 + Si,j ) (9.63) where  Si,j ≡ Ri 0 [Ki3 (τij+ ) − Ki3 (τij− )]dy (9.64) with the τ being optical path lengths α over the indicated chords in the horizontal plane in Fig. 9.11: τij± ≡ . Rj2 − y 2 ∓ . Ri2 − y 2  (9.65) Methods for the numerical evaluation of these expressions are given in Ref. 3. 9.4 Interface Current Methods in Slab Geometry Fig. 9.11 Annular geometry notation for calculation of collision probabilities (τ is the optical path length α over the indicated path). (From Ref. 3; used with permission of Academic Press.) 9.4 Interface Current Methods in Slab Geometry Emergent Currents and Reaction Rates Due to Incident Currents Consider the slab geometry configuration depicted in Fig. 9.12, in which a slab − and region i is bounded by surfaces i and i + 1 with incident currents Ji+ and Ji+1 − + emergent currents Ji and Ji+1 . The angular flux of particles at x arising from an angular flux of neutrons at x  is  ψ(x, μ) = e−ti (x−x )/μ ψ(x  μ) (9.66) where it is assumed that the total cross section, t , is uniform over i , and μ is the cosine of the angle that the particle direction makes with the x-axis. Further as− , are isotropically distributed in angle suming that the incident fluxes, ψi+ and ψi+1 323 324 9 Neutron Transport Theory Fig. 9.12 Nomenclature for slab geometry interface current method. over the incident hemisphere (i.e., a double P0 approximation), the uncollided currents emergent from the opposite surface are given in terms of the incident partial − − = 12 ψi+1 ) by currents (Ji+ = 12 ψi+ , Ji+1  + Jˆun (xi+1 ) = 2Ji+ − − (xi ) = 2Ji+1 Jˆun 1 0 0  −1   μ e−ti i /μ dμ = 2E3 (i ti )Ji+   − μ e+ti i /μ dμ = 2E3 (i ti )Ji+1 (9.67) where En is the exponential integral function given by  1 En (z) ≡ μn−2 e−z/μ dμ (9.68) 0 The first collision rate for incident particles within i is given by   R̂i1 = ti 2Ji+ − + 2Ji+1   1 xi+1 dμ 0  0 −1   dx e−ti (x−xi )/μ xi xi+1 dμ   dx e−ti (x−xi+1 )/μ  xi  −  = Ji+ + Ji+1 [1 − 2E3 (i ti )] (9.69) The fraction ci of the collision rate that is due to scattering (i.e., to events that do not remove the particle) from the cohort under consideration (i.e., ci = (si + νf i )/ti ) constitutes a source of once-collided particles, which we assume to be isotropic ( 12 emerge going to the right and 12 to the left) and distributed uniformly 9.4 Interface Current Methods in Slab Geometry over i . Treating these “scattered” neutrons as a distribution of plane isotropic sources, with the source at x  producing exiting uncollided fluxes   1 R̂i1 exp(−ti (xi+1 − x  )/μ) ci 2 i μ   1 R̂i1 exp(−ti (xi − x  )/μ) ci 2 i μ and at xi+1 and xi , respectively, the emergent currents of once-collided particles are  1 R̂i1 e−ti (xi+1 −x)/μ ci dμ 2 i μ xi 0   1 ci R̂i1 1 1 = − E3 (i ti ) = ci P0i R̂i1 2 i ti 2 2  + − −  1 1 ˆ J1 (xi ) = 2 ci P0i R̂i1 = 2 ci P0i Ji + Ji+1 [1 − 2E2 (i ti )] Jˆ1+ (xi+1 ) =   xi+1  1 dx μ (9.70) where the average first-flight escape probability for source particles distributed uniformly over i has been defined as P0i ≡ 1 2  1 dx i +  xi+1 xi 1 2 i dμ μ 0   xi+1 dx 0 −1 xi e−ti (xi+1 −x)/μ μ dμ μ   1 1 − E3 (i ti ) = i ti 2 e−ti (xi −x)/μ μ (9.71) The collision rate for incident particles undergoing a second collision in i is 1 R̂i1 R̂i2 = ti ci 2 i   1 xi 0  + = xi+1 dμ 0 −1   xi  xi+1 dx xi xi+1 dμ ci R̂i1 (1 − P0i ) = ci (Ji+ dx  dx   e−ti (x−x )/μ μ x  dx xi e−ti (x−x )/μ μ  − + Ji+1 )[1 − 2E3 (i ti )](1 − P0i ) (9.72) As before, the fraction ci of this collision rate constitutes a source of twice-collided particles which are assumed to be isotropic. The emergent currents of twicecollided particles are given by Eqs. (9.70) but with R̂i1 replaced by R̂i2 : Jˆ2+ (xi+1 ) = Jˆ2− (xi ) = 12 ci R̂i2 P0i  −  = 12 ci2 Ji+ + Ji+1 [1 − 2E3 (i ti )](1 − P0i )P0i (9.73) 325 326 9 Neutron Transport Theory Continuing this line of argument, we derive general expressions for the rate at which incident particles undergo their nth collision in i :  −  (9.74) R̂in = cin−1 Ji+ + Ji+1 [1 − 2E3 (i ti )](1 − P0i )n−1 and for the emergent currents of n-collided incident particles, Jˆn+ (xi+1 ) = Jˆn− (xi )  −  = 12 cin Ji+ + Ji+1 [1 − 2E3 (i ti )](1 − P0i )n−1 P0i (9.75) The total collision rate in i due to incident currents is obtained by summing Eqs. (9.74): R̂i = ∞  ∞   −  [ci (1 − P0i )]n R̂in = Ji+ + Ji+1 [1 − 2E3 (i ti )] n=1 = (Ji+ n=0 − + Ji+1 )[1 − 2E3 (i ti )] (9.76) 1 − ci (1 − P0i ) and the total emergent currents due to incident currents are obtained by summing Eq. (9.75) and adding the uncollided contributions of Eqs. (9.67): 1  2 ci P0i [1 − 2E3 (i ti )] + ˆ + 2E3 (i ti ) Ji+ J (xi+1 ) = 1 − ci (1 − P0i ) + Jˆ− (x 1 i) = 1 2 ci P0i [1 − 2E3 (i ti )] 1 − ci (1 − P0i ) 2 ci P0i [1 − 2E3 (i ti )] 1 − ci (1 − P0i ) + (9.77)  + 2E3 (i ti ) 1 2 ci P0i [1 − 2E3 (i ti )] 1 − ci (1 − P0i ) − Ji+1 − Ji+1 Ji+ Emergent Currents and Reaction Rates Due to Internal Sources We consider a uniform distribution of particle sources within i of strength si /i per unit length. This source is allowed to be anisotropic, with a number si+ emitted to the right and si− emitted to the left. The emergent currents of uncollided source particles are   1 s + xi+1 e−ti (xi+1 −x)/μ + (xi+1 ) = i dx dμ μ = si+ P0i Jun,s i xi μ 0 (9.78)   0 s − xi+1 e−ti (xi −x)/μ − = si− P0i Jun,s (xi ) = i dx dμ μ i xi μ −1 The first collision rate of source particles within i is given by  xi+1  xi+1  1  s+ e−ti (x−x )/μ dx  dx dμ R̂i1,s = i ti  μ 0 xi x 9.4 Interface Current Methods in Slab Geometry + si− ti i  xi+1 dx  xi  + − = (si + si ) 1 −   x dx xi 1 i ti  0 −1 dμ e−ti (x−x  )/μ μ  1 − E3 (ti i ) 2 = si (1 − P0i ) (9.79) As before, treating the fraction ci of these particles that undergo scattering collisions as an isotropic plane source of once-collided particles, the emergent currents of once-collided source particles are given by  1  xi+1 1 R̂i1,s e−ti (xi+1 −x)/μ + J1s (xi+1 ) = dμ μ dx ci 2 i μ 0 xi 1 1 ci R̂i1,s P0i = ci si (1 − P0i )P0i 2 2  0  xi+1 1 R̂i1,s e−ti (xi −x)/μ − (xi ) = dμ μ dx ci J1s 2 i μ −1 xi = = (9.80) 1 1 ci Ri1,s P0i = ci si (1 − P0i )P0i 2 2 Continuing in this fashion, the general expression for the nth collision rate of source particles in i is R̂in,s = cin−1 si (1 − P0i )n (9.81) and the general expressions for the emergent currents of n-collided source particles are + − (xi+1 ) = Jns (xi ) = 12 si P0i cin (1 − P0i )n Jns (9.82) The total collision rate of source particles within i is R̂i,s = ∞  n=1 R̂in,s = si (1 − P0i ) 1 − ci (1 − P0i ) (9.83) and the total emergent currents due to an anisotropic particle source within i are obtained by summing Eqs. (9.82) and adding Eqs. (9.78):   1 1 2 si P0i Js+ (xi+1 ) = si+ − si P0i + 2 1 − ci (1 − P0i ) (9.84)   1 1 − − 2 si P0i Js (xi ) = si − si P0i + 2 1 − ci (1 − P0i ) Total Reaction Rates and Emergent Currents The total reaction rate in i due to incident currents and to internal sources is obtained by adding Eqs. (9.76) and (9.83): Ri = − (Ji+ + Ji+1 )(1 − T0i ) + si (1 − P0i ) 1 − ci (1 − P0i ) (9.85) 327 328 9 Neutron Transport Theory where the first-flight, or uncollided, transmission probability has been identified: T0i ≡ E2 (i ti ) (9.86) Further identifying the total escape probability, Pi ≡ P0i ∞  [ci (1 − P0i )]n = n=0 P0i 1 − ci (1 − P0i ) (9.87) the total reflection probability, Ri ≡ 1 2 ci P0i [1 − 2E3 (i ti )] 1 − ci (1 − P0i ) 1 = ci Pi (1 − T0i ) 2 (9.88) and the total transmission probability, Ti = T0i + Ri = T0i + 12 ci Pi (1 − T0i ) (9.89) Eqs. (9.77) and (9.84) can be summed to obtain expressions for the total emergent currents due to incident currents and internal particle sources:   + − = Ti Ji+ + Ri Ji+1 + 12 si Pi + si+ − 12 si P0i Ji+1   − Ji− = Ti Ji+1 + Ri Ji+ + 12 si Pi + si− − 12 si P0i (9.90) The inherent advantage of an interface current formulation of integral transport theory is evident from Eqs. (9.90). To solve for the currents across interface i, one needs only the currents at interface i + 1 and the source in the intervening region. This leads, in essence, to a “four-point” coupling of the unknowns, the partial currents at i and i + 1, and the evaluation of only one E3 function for each region. By contrast, in the standard collision probabilities formulation of the preceding section, the fluxes in all other regions in the problem and the transition probabilities from all of these regions to the region in question are needed in order to solve for the flux in a given region, in essence coupling all regions in the problem. In both formulations, an iterative solution is needed. As formulated above, the interface current method is based on the D-P0 assumption of an isotropic angular flux distribution within the incident hemisphere at each interface, for the purpose of calculating the uncollided transmission across the region. This assumption is physically plausible for problems with scattering (and fission) rates comparable to or larger than absorption rates, because this tends to isotropize the flux exiting from a region. However, for problems with an incident neutron source on one boundary of an almost purely absorbing medium, the flux will become increasingly forward directed with distance into the region. In the limit of a purely absorbing region with an incident isotropic neutron source at x = 0, the current attenuation at a distance x from the source plane is exactly E2 (x). If this problem is modeled in the interface current formulation and the distance x is subdivided into N intervals , the calculated current attenuation at x 9.4 Interface Current Methods in Slab Geometry / is N n=1 E2 (), which differs from the exact answer E2 (n). Thus inaccuracies might be expected in highly absorbing multiregion problems. It is informative to sum Eqs. (9.90) to obtain an intuitively obvious balance between incident and emergent currents and internal sources:  + −  + Ji− = (Ti + Ri ) Ji+ + Ji+1 + si Pi Ji+1 or (9.91) Jout = (T0i + (1 − T0i )ci Pi )Jin + si Pi Solving the first of Eqs. (9.90) for Ji+ and using the result in the second equation leads to a matrix relation among the partial currents at adjacent surfaces:   +   −1 +  Ji+1 Ti −Ti−1 Ri Ji = − Ji− Ji+1 Ri Ti−1 Ti − Ri Ti−1 Ri  % & 1 −Ti−1 + si Pi 1 − Ri Ti−1 2  *   −Ti−1 si+ − 12 si +P0i  − 1    (9.92) si − 2 si − Ri Ti−1 si+ − 12 si Equation (9.92) is well suited for numerical evaluation simply by marching from one boundary of the problem to the other. Boundary Conditions Boundary conditions take on a particularly simple form for an interface current formulation of integral transport. Let x = 0, i = 0 represent the leftmost surface of the transport medium. If a vacuum or nonscattering medium with no particle source exists for x < 0, then J0+ = 0 is the appropriate boundary condition. If, on the other hand, a source-free scattering medium exists for x < 0, an albedo or reflection condition of the form J0+ = βJ0− , where β is the reflection coefficient or albedo, is appropriate. Finally, if a known current of particles Jin is incident upon the medium from the left at x = 0, the appropriate boundary condition is J0+ = Jin . Response Matrix Suppressing internal sources, the matrix equation (9.92) may be written in more compact notation: ± J± i = R i J i+1 (9.93) (where J indicates a column vector and R indicates a matrix) and applied successively to relate the incident and exiting currents on the left boundary, J ± 0 , to the : incident and exiting currents on the right boundary, J ± I ± R 0 · R 1 · R 2 · · ·R R I −2 · R I −1 ]JJ ± J± I ≡ RJ I 0 = [R (9.94) 329 330 9 Neutron Transport Theory where the matrix R is the matrix product of the matrices R i for each slab i and has the form  11  R 12 R R= (9.95) R 21 R 22 in terms of which Eq. (9.94) may be written as the two equations J0+ = R 11 JI+ + R 12 JI− J0− = R 21 JI+ + R 22 JI− (9.96) which may be solved to obtain the response matrix relationship between the incident currents, J0+ and JI− , and the exiting currents, J0− and JI+ .  −   22  −  J0 R − R 21 (R 11 )−1 R 12 R 21 (R 11 )−1 JI = JI+ −R 21 (R 11 )−1 R 12 R 21 (R 11 )−1 J0+ (9.97) or J out = RMJ in Once the response matrix, RM RM, is evaluated, the exiting currents can be computed rapidly for a given set of incident currents. This formalism can be extended in an obvious way to treat internal sources. 9.5 Multidimensional Interface Current Methods Extension to Multidimension The interface current formulation of integral transport theory can be extended to two and, in principle, three dimensions. First, for conceptual purposes, we rewrite + out , J − = = Ji+1 Eqs. (9.90) by making the identification Ji+ = Jiin , Ji− = Jiout , Ji+1 i+1 in Ji+1 , and   si+1 si Pi ≡ 12 si Pi + si+ − 12 si P0i   si si Pi ≡ 12 si Pi + si− − 12 si P0i (9.98) where si is the fraction of escaping source neutrons that escapes to the left across surface i and si+1 is the fraction escaping to the right across surface i + 1. Then, using Eqs. (9.85) to (9.89), Eqs. (9.90) may be written   out = T J in + (1 − T ) J in + J in c P  s Ji+1 0i i 0i i i+1 i i i+1 + i+1 si Pi (9.99)   in + (1 − T ) J in + J in c P  + s s P Jiout = T0i Ji+1 0i i i i i i+1 i i i where i = i+1 = 12 is the fraction of escaping scattered incident neutrons that escape across surfaces i and i + 1, respectively. 9.5 Multidimensional Interface Current Methods Fig. 9.13 Planar projection of geometry for multidimensional interface current methods. In this form, the terms in Eqs. (9.99) for the emergent currents have a direct physical interpretation which leads immediately to a generalization to multidimensions. The outward current across surface i + 1 consists of three terms: (1) the inward current across surface i times the probability T0i that it is transmitted across region i without collision to surface i + 1; (2) the inward currents across all surfaces times the probability (1 − T0i ) that these currents are not transmitted across region i without collision, times the probability ci that the first collision is a “scattering” event, times the probability Pi that the scattered neutrons subsequently escape from region i, times the probability i+1 that escaping neutrons escape across surface i + 1; and (3) the total particle source si in region i times the probability Pi that these neutrons will escape from region i, times the probability si+1 that escaping source neutrons escape across surface i + 1. Note that i+1 and si+1 can in principle differ because an anisotropic source is allowed [i.e., i+1 = 12 and si+1 is given by Eq. (9.98) for slab geometry]. Generalization to multidimensions is straightforward, in principle. Consider the two-dimensional configuration in Fig. 9.13. The current from region k into region i is denoted Jki ( k→i in the figure), the probability that the current entering region i from region k is transmitted across region i without collision to contribute kj to the current from region i into region j is denoted T0i , and the probability that a collided or source neutron escaping from region i escapes into region j is denoted ij . The generalization of Eqs. (9.99) to two-dimensions is then Jij = i  k kj T0i Jki + i  k % 1− i  & T0ikl Jki ci Pi ij + sij si Pi (9.100) l  where the summation ik is over all regions k that are contiguous to region i. The three terms in Eq. (9.100) correspond physically to (1) the sum of the cur- 331 332 9 Neutron Transport Theory rents incident into region i from all contiguous regions times the probability that each is transmitted across region i without collision to exit into region j (note that the possibility of concave surfaces is allowed by including uncollided transmission from region j across region i back into region j ); (2) the sum of the currents incident into region i from all contiguous regions times the probability that each is not transmitted without collision across region i to any of the contiguous regions, times the probability that the first collision is a “scattering” event, times the probability that the scattered neutron eventually escapes from region i into region j ; and (3) the source of neutrons in region i times the probability that a source neutron in region i eventually escapes into region j . Evaluation of Transmission and Escape Probabilities The general form for the evaluation of transmission and escape probabilities can be developed using the point kernel discussed previously. We treat the case of incident fluxes that are distributed isotropically in the incident hemisphere of directions and volumetric neutron sources (scattering, fission, external) which are uniformly distributed over volume and emitted isotropically in direction. These results can be extended to anisotropic incident fluxes and nonuniform and anisotropic volumetric source distributions by extending the procedures indicated below. The probability that a neutron introduced isotropically at location ri within volume Vi escapes without collision across the surface Ski that defines the interface between volume Vi and contiguous volume Vk is the probability d/4π|rSki − ri |2 that the neutron is traveling within a cone of directions d which intersects that surface, times the probability exp[−α(rSki , ri )] that the neutron reaches the surface at location rSki along the direction  from ri without a collision, integrated over all  that intersect the surface Ski from point ri . This probability is then averaged over all points ri within volume Vi to obtain   1 e−α(rSki ,ri ) dri dS (9.101) P0i k = 4πVi Vi |rSki − ri |2 Ski Extension of this expression to treat anisotropic neutron emission would be accomplished by including a function f (rSki , ri ) under the integral to represent any directional dependence of neutron emission. Extension to include a spatial distribution g(ri ) of neutron sources would be accomplished by including this function in the integrand. The probability that an incident unit neutron flux which is isotropically distributed over the inward hemisphere of directions entering volume Vi from volume Vk across surface Ski is transmitted without collision across volume Vi to the surface Sj i which forms the interface with contiguous volume Vj is the product of the probability nSki · d/2π|rSki − rSj i |2 = (nSki · )d/2π|rSki − rSj i |2 that a neutron incident across Ski is traveling within a cone of directions d which intersects the surface Sj i , times the probability exp[−α(rSj i , rSki )] that the neutron reaches the surface at location rSj i along the direction  from rSki without a collision, integrated over all  that intersect the surface Sj i from point rSki . The quantity nSki 9.5 Multidimensional Interface Current Methods is the unit vector normal to the surface Ski in the direction from volume Vk into volume Vi . This probability is then averaged over all points rSki on Ski , to obtain kj T0i = Ski dS Sj i dS[e −α(rSj i ,rSki ) (nSki · )/|rSj i − rSki |2 ] nSki ·>0 d Ski dS (9.102) Extension of this expression to include an anisotropic incident neutron flux would be accomplished by including a function f (rSki , rSj i ) in the integrand. Transmission Probabilities in Two-Dimensional Geometries To develop computational algorithms, we consider geometries with symmetry in one direction, which are conventionally known as two-dimensional geometries. It is important to keep in mind, however, that neutron flight paths take place in three dimensions. Consider a volume Vi that is symmetric in the axial direction and bounded by flat vertical surfaces, so that a horizontal (x–y) planar slice is as shown in Fig. 9.14, with the vertical dimension normal to the page. We want to calculate the transmission coefficient from volume 1 through the volume i into volume 3. A three-dimensional projection and a vertical projection are shown in Fig. 9.15. The points ξ1 and ξ3 in Fig. 9.14 are the projection onto the horizontal plane of the vertical axes shown in Fig. 9.15. The differential solid angle in this coordinate system is d = 1 1 sin θ  dθ  dφ = − cos θdθdφ 4π 4π Fig. 9.14 Planar projection of geometry for transmission probability calculation in two-dimensions. (9.103) 333 334 9 Neutron Transport Theory Fig. 9.15 Three-dimensional and axial projection of geometry for transmission probability calculation in two-dimensions. The incident directional flux from volume 1 at point ξ1 , ψ(r − R, ) is attenuated when it traverses the distance R to reach the point ξ3 and enter volume 3: ψ(r, ) = ψ(r − R, )e−R (9.104) The incident partial current density (n/cm2 · s) from volume 1 at point ξ1 is  jin (ξ1 ) = d(nin · )ψ(r − R, )d nin ·>0 =− 1 4π   π dφ 0 π/2 −π/2 dθ cos2 θ sin φψ(r − R, ) (9.105) where nin ·  = cos θ sin φ has been used. When the incident flux is isotropic in the incident hemisphere (double-P0 approximation), this becomes iso (ξ1 ) = 14 ψ(r − R) jin (9.106) The incident partial current (n/s) is obtained by multiplying by the (arbitrary) axial dimension H and integrating over ξ1min ≤ ξ1 ≤ ξ1max  Jin = H ξ1max ξ1min dξ1 jin (ξ1 ) (9.107) The incident neutrons from volume 1 which enter volume Vi at ξ1 within the solid angle subtended by volume 3 and traverse volume i without collision to enter volume 3 constitute an uncollided neutron current out of volume Vi into volume 3, and hence a contribution to the incident current into volume 3 from volume i. For the moment we write this contribution to the current into volume 3 as   ξ max 1 dξ1 d( · nout )ψ(r − R, )e−R Jout = H ξ1min ·n̄out >0 φ(ξ1 )3 9.5 Multidimensional Interface Current Methods  =H ξ1max ξ1min   φmax (ξ1 ) dξ1 dφ φmin (ξ1 ) π/2 −π/2 dθ cos2 θ sin φe(−l(φ(ξ1 ))/ cos θ) × ψ(r − R, ) (9.108) where nout ·  = cos θ sin φout may differ from nin ·  = cos θ sin φ if the interfaces with volumes 1 and 3 are not parallel, and φ(ξ1 )  3 indicates angles φ from a point ξ1 which intersect the interface with region 3. When the incident flux from volume 1 is isotropic in the incident directional hemisphere, this becomes iso = Jout H 2π  ξ1max ξ1min  φmax (ξ1 ) dξ1 φmin (ξ1 )   dφ sin φout Ki3 l φ(ξ1 ) ψ(r − R) (9.109) The transmission probability for an isotropic incident flux distribution from volume 1 that is uniform over ξ1min ≤ ξ1 ≤ ξ1max can be written in a form that couples the contribution to the incident current into volume 3 with the incident current into volume i: T0i13 ≡ iso Jout iso Jin 2 = π ξ1max ξ1min dξ1 φmax (ξ1 ) φmin (ξ1 ) dφ sin φout Ki3 [l(φ(ξ1 ))] ξ1max − ξ1min (9.110) When the incident and exiting surfaces (the interfaces of volumes 1 and i and of volumes 3 and i in this example) are not parallel, there is a subtlety about the direction to take for nout in the equations above. The incident current into volume i from volume 1 was calculated on the basis of a DP-0 angular flux approximation with respect to the orientation of the incident surface. The transport of the uncollided incident DP-0 angular flux across region i is properly calculated, and by using nout = nin , the exiting uncollided partial current in the direction normal to the incident surface is properly calculated. So the neutron flow into volume 3 is properly calculated, although the direction of this current exiting volume i is not normal to the exit surface. In constructing the incident current for region 3 from region i, this uncollided contribution from region 1 is added to the uncollided contribution from regions 2 and 4 and to the collided contribution, and the combination is assumed to have a DP-0 incident angular distribution into volume 3 with respect to the orientation of this incident interface of volume 3 (the exiting interface of volume i). Thus, in the equations above, nout = nin should be used. Escape Probabilities in Two-Dimensional Geometries The neutron flux per unit surface area, dA, normal to the direction of neutron flight at a distance R away from an isotropic point source is exp(−R)/4πR 2 , and with reference to Fig. 9.15, the surface area normal to the direction  of neutron travel is dA = R dθl dφ = l 2 dθ dφ/ cos θ . Thus, with reference to Fig. 9.16, an isotropic neutron source of unit strength per axial length located at ri within volume Vi produces an outward current of uncollided neutrons over the surface labeled ξ3 335 336 9 Neutron Transport Theory Fig. 9.16 Planar projection of geometry for escape probability calculation in two dimensions. into volume 3 that is described by  3 Jout (ri ) = A⊃S3 (nout · )  = φ⊃S3  =  dφ e−R dA 4πR 2 π/2 −π/2 dθ(cos θ sin φout ) dφ sin φout φ⊃S3  = φ⊃S3 dφ sin φout 1 2π  π/2 e−l(φ)/ cos θ (l 2 / cos θ) 4π(l/ cos θ)2 dθ cos2 θe−l(φ)/ cos θ 0 Ki3 (l(φ)) 2π (9.111) where φ ⊃ S3 indicates the range of φmin < φ < φmax subtended by side S3 at location ri within volume Vi . 3 (x, y) over the planar two-dimensional area A of volThe average value of Jout i ume Vi is just the probability that an isotropic, uniform neutron source si will produce an uncollided current si si3 P0i from volume Vi into volume V3 : 1 i3 P0i = Ai = 1 Ai   Ai 3 dx dy Jout (x, y)  dx dy Ai dφ sin φout φ⊃S3 Ki3 (l(φ)) 2π (9.112) The proper value of nout is the outward normal to the surface in question, and φout is measured with respect to the orientation of that surface, whereas φ may be measured with respect to a fixed coordinate system, so that in general φout = φ, although it is convenient to orient the coordinate system so that φout = φ. 9.5 Multidimensional Interface Current Methods The total uncollided escape probability is obtained by summing Eq. (9.112) over all volumes Vk that are contiguous to volume Vi : P0i =  ik P0i (9.113) k and the directional escape fractions are calculated from ij = ij P0i ij P0i = P0i k ik P0i (9.114) Using the same arguments as were made for the one-dimensional case in the preceding section, the total escape probability, including escape after zero, one, two, . . . collisions can be calculated from Pi = P0i 1 − ci (1 − P0i ) (9.115) where ci = (si + νf i )/ti is the number of secondary neutrons produced per collision. Simple Approximations for the Escape Probability Physical considerations lead to a simple approximation for the first-flight escape probability. In the limit that the average neutron path length l in a volume V is much less than the mean free path λ for a collision, the escape probability tends to unity. In the limit when l  λ, a simple approximation for the first-flight escape probability is 1 − exp(−λ/ l ) ≈ λ/ l . If we associate the average neutron path length in the volume with the mean chord length 4V /S, where S is the surface area of the volume V , a simple rational approximation for the escape probability, first proposed by Wigner and with which his name is associated, is   1 1 1 1 = = 1− (9.116) P0 = 1 + l /λ 1 + 4V /Sλ 4V /Sλ 1 + (4V /Sλ) This Wigner rational approximation is known to underpredict the first-flight escape probability. However, extensive Monte Carlo calculations have confirmed that the first-flight escape probability depends only on the parameter 4V /Sλ, and improved rational approximations of the form  1 1 (9.117) 1− P0 = (4V /Sλ) [1 + (4V /Sλ)/c]c have been proposed. The Sauer approximation, developed from theoretical considerations for cylindrical geometry, corresponds to c = 4.58. The best fit to Monte Carlo calculations of first-flight escape probabilities for a uniform neutron source distribution in volumes with a wide range of geometries and values of the parameter 4V /Sλ was found by using c = 2.09. 337 338 9 Neutron Transport Theory 9.6 Spherical Harmonics (PL ) Methods in One-Dimensional Geometries The spherical harmonics, or PL , approximation is developed by expansion of the angular flux and the differential scattering cross section in Legendre polynomials. Legendre Polynomials The first few Legendre polynomials are P0 (μ) = 1, P2 (μ) = P1 (μ) = μ, P3 (μ) =  1 2 2 3μ  1 3 2 5μ  −1  − 3μ (9.118) and higher-order polynomials can be generated from the recursion relation (2n + 1)μPn (μ) = (n + 1)Pn+1 (μ) + nPn−1 (μ) (9.119) The Legendre polynomials satisfy the orthogonality relation  1 −1 dμPm (μ)Pn (μ) = 2δmn 2n + 1 (9.120) With reference to Fig. 9.17, the Legendre polynomials of μ0 = cos θ0 , the cosine of the angle, between μ and μ, can be expressed in terms of the Legendre polynomials of μ and μ by the addition theorem Fig. 9.17 Scattering from  to . 9.6 Spherical Harmonics (PL ) Methods in One-Dimensional Geometries Pn (μ0 ) = Pn (μ)Pn (μ ) +2 n  (n − m)! m  m P (μ )Pn (μ) cos m(φ − φ  ) (n + m)! n (9.121) m=1 where the associated Legendre functions are defined by  m/2 d m Pn (μ) Pnm (μ) ≡ 1 − μ2 dμm (9.122) Neutron Transport Equation in Slab Geometry Consider a situation in which there is symmetry in the y- and z-coordinate directions but a variation in properties in the x-coordinate direction. The steady-state neutron transport equation (9.5) in this case becomes μ ∂ψ(x, μ) + t (x, μ)ψ(x, μ) ∂x  1 s (x, μ → μ)ψ(x, μ )dμ + S(x, μ) = −1  = 1 −1 s (x, μ0 )ψ(x, μ )dμ + S(x, μ) (9.123) where, with reference to Fig. 9.17, we take advantage of the fact that the scattering from a cone of directions μ = cos θ  to a cone of directions μ = cos θ only depends on μ0 = cos θ0 , the cosine of the angle between μ and μ, and not on the incident and exiting directions for the scattering event. PL Equations The PL equations are based on the approximation that the angular dependence of the neutron flux can be expanded in the first L + 1 Legendre polynomials: ψ(x, μ) = L  2l + 1 l=0 2 φl (x)Pl (μ) (9.124) The angular dependence of the differential scattering cross section is also expanded in Legendre polynomials: s (x, μ0 ) = M  2m + 1 sm (x)Pm (μ0 ) 2 (9.125) m=0 When these expansions are used in Eq. (9.123), the addition theorem of Eq. (9.121) is used to replace Pm (μ0 ) with Pm (μ) and Pm (μ ), the recursion relation of Eq. (9.119) is used to replace μPn (μ) terms with Pn±1 (μ) terms, the resulting equation is multiplied in turn by Pk (μ) (k = 0 to L) and integrated over 339 340 9 Neutron Transport Theory −1 ≤ μ ≤ 1, and the orthogonality relation of Eq. (9.120) is used, the L + 1 PL equations dφ1 (x) + (t − so )φ0 (x) = S0 (x), n = 0 dx n + 1 dφn+1 (x) n dφn−1 + + (t − sn )φn (x) = Sn (x), 2n + 1 dx 2n + 1 dx n = 1, . . . , L (9.126) are obtained. The n subscript indicates the nth Legendre moment of the angular dependent quantity:  φn (x) ≡ Sn (x) ≡ 1 −1  1 −1  sn (x) ≡ dμPn (μ)ψ(x, μ) (9.127) dμPn (μ)S(x, μ) 1 −1 dμ0 Pn (μ0 )s (x, μ0 ) This set of L + 1 equations has a closure problem—they involve L + 2 unknowns. This problem is usually resolved by ignoring the term dφL+1 /dx which appears in the n = L equation. Boundary and Interface Conditions The true boundary condition at the left boundary xL , ψ(xL , μ) = ψin (xL , μ), (9.128) μ>0 where ψin (xL , μ > 0) is a known incident flux [ψin (xL , μ > 0) = 0 is the vacuum boundary condition], cannot be satisfied exactly by the angular flux approximation of Eq. (9.124), for finite L. The most obvious way to develop approximate boundary conditions that are consistent with the flux approximation is to substitute Eq. (9.124) into the exact boundary condition of (9.128), multiply by Pm (μ), and integrate over 0 ≤ μ ≤ 1. Since it is the odd Legendre polynomials that represent directionality (i.e., are different for μ and −μ), this procedure is repeated for all the odd Legendre polynomials m = 1, 3, . . . , L (or L − 1) as weighting functions to obtain, with the use of the orthogonality relation of Eq. (9.120), the Marshak boundary conditions  1 dμPm (μ) 0 N  2n + 1 n=0 2  φn (xL )Pn (μ) ≡ φm (xL ) = m = 1, 3, . . . , L (or L − 1) 1 dμPm (μ)ψin (xL , μ), 0 (9.129) Equations (9.129) constitute a set of (L + 1)/2 boundary conditions. An additional (L + 1)/2 boundary conditions are obtained similarly for the right boundary. The 9.6 Spherical Harmonics (PL ) Methods in One-Dimensional Geometries Marshak boundary conditions ensure that the exact inward partial current at the boundary is incorporated into the solution; that is, J + (xL ) ≡  1 dμP1 (μ) 0  ≡ 0 N  2n + 1 n=0 1 2 φn (xL )Pn (μ) + dμP1 (μ)ψin (xL , μ) ≡ Jin (xL ) (9.130) A less intuitive set of Mark boundary conditions arises from requiring that the flux expansion of Eq. (9.124) satisfy the boundary condition N  2n + 1 n=0 2 φn (xL )Pn (μi ) = ψin (xL , μi ), μi > 0 (9.131) for the (L + 1)/2 discrete values of μi in the inward direction which are the positive roots of PL+1 (μi ) = 0. Another (L + 1)/2 approximate boundary conditions are obtained at the other boundary by requiring that the flux expansion satisfy the true boundary condition for the (L + 1)/2 discrete values of μi in the inward direction which are the negative roots of PL+1 (μi ) = 0. These Mark boundary conditions are justified by the fact that analytical solution of the PL equations for a sourcefree, purely absorbing problem in a infinite half-space leads to these conditions. However, experience has shown that results obtained with the Mark boundary conditions are generally less accurate than results obtained with the Marshak boundary conditions. A symmetry, or reflective, boundary condition ψ(xL , μ) = ψ(xL , −μ) obviously requires that all odd moments of the flux vanish [i.e., φn (xL ) = 0 for n = 1, 3, . . . , odd]. The exact interface condition of continuity of angular flux ψ(xs − ε, μ) = ψ(xs + ε, μ) (9.132) where ε is a vanishingly small distance, cannot, of course, be satisfied exactly by the flux approximation of Eq. (9.124), for finite L. Following the same procedure as for Marshak boundary conditions, we replace the exact flux with the expansion of Eq. (9.124) and require that the first L + 1 Legendre moments of this relation be satisfied (i.e., multiply by Pm and integrate over −1 ≤ μ ≤ 1, for m = 0, . . . , L). Using the orthogonality relation of Eq. (9.120) then leads to the interface conditions of continuity of the moments: φn (x − ε) = φn (xs + ε), n = 0, 1, 2, . . . , L (9.133) There are some subtle reasons why this approximation is not appropriate for even-L approximations (see Ref. 6), but since odd-L approximations are almost always used, we will only raise a caution. 341 342 9 Neutron Transport Theory P1 Equations and Diffusion Theory Neglecting the dφ2 /dx term, the first two of Eqs. (9.126) constitute the P1 equations dφ1 + (t − s0 )φ0 = S0 dx 1 dφ0 + (t − s1 )φ1 = S1 3 dx (9.134) Noting that s0 = s , the total scattering cross section, and that s1 = μ̄0 s , where μ̄0 is the average cosine of the scattering angle, and assuming that the source is isotropic (i.e., S1 = 0), the second of the P1 equations yields a Fick’s law for neutron diffusion:  1 1 dφ0 μψ(x, μ) dμ ≡ J (x) = − (9.135) φ1 (x) = 3( − μ̄  ) t 0 s dx −1 which, when used in the first of the P1 equations, yields the neutron diffusion equation   dφ0 (x) d (9.136) D0 (x) + (t − s )φ0 (x) = S0 (x) − dx dx where the diffusion coefficient and the transport cross section are defined by D0 ≡ 1 1 ≡ 3(t − μ̄0 s ) 3tr (9.137) The basic assumptions made in this derivation of diffusion theory are that the angular dependence of the neutron flux is linearly anisotropic: ψ(x, μ)  12 φ0 (x) + 32 μφ1 (x) (9.138) and that the neutron source is isotropic, or at least has no linearly anisotropic component (S1 = 0). Diffusion theory should be a good approximation when these assumptions are valid (i.e., in media for which the distribution is almost isotropic because of the preponderance of randomizing scattering collisions, away from interfaces with dissimilar media, and in the absence of anisotropic sources). The boundary conditions for diffusion theory follow directly from the Marshak condition (9.130): + (xL ) = Jin =   1 dμP1 (μ) 0  1 3 φ0 (xL ) + μφ1 (xL ) 2 2 1 dφ0 (xL ) 1 φ0 (xL ) − D 4 2 dx (9.139) + = 0, the vacuum boundary condition for When the prescribed incident current, Jin diffusion theory can be constructed from a geometrical interpretation of the ratio 9.6 Spherical Harmonics (PL ) Methods in One-Dimensional Geometries of the flux gradient to the flux in this equation to obtain the condition that the extrapolated flux vanishes a distance λex outside the boundary: φ(xL − λex ) = 0, λex = 2 2 ≡ λtr 3tr 3 (9.140) The interface conditions of Eq. (9.133) become in the diffusion approximation φ0 (xs + ε) = φ0 (xs − ε) −D0 (xs + ε) dφ0 (xs + ε) dφ0 (xs − ε) = −D0 (xs − ε) dx dx (9.141) Simplified PL or Extended Diffusion Theory The same procedure used to derive diffusion theory from the P1 equations—solve the odd-order equation for the odd-order moment of the flux in terms of a gradient of the even-order flux moment and use the result to eliminate the odd-order flux—can be used to simplify odd-L PL approximations of higher order. For example, in the P3 approximation with an isotropic source and isotropic scattering, the following change of variables is made: F0 = 2φ2 + φ0 F1 = φ2 (9.142) to facilitate the reduction of the four coupled P3 equations to the two coupled diffusion equations   d dF0 − D0 + (t − s0 )F0 = S0 + 2(t − s0 )F1 dx dx     dF1 d 4 5 (9.143) − D1 + (t − s2 ) + (t − s0 ) F1 dx dx 3 3 2 2 = − S0 + (t − s0 )F0 3 3 where D1 ≡ 3 7(t − s3 ) (9.144) + = 0) boundary conditions of Eq. (9.129) become The Marshak vacuum (Jin 1 dF0 (xL ) 3 F0 (xL ) − F1 (xL ) = D0 2 8 dx dF1 (xL ) 7 1 − F0 (xL ) + F1 (xL ) = D1 8 8 dx (9.145) This formulation of the PL equations allows the powerful numerical solution techniques for diffusion theory to be used to solve a higher-order transport approximation. 343 344 9 Neutron Transport Theory PL Equations in Spherical and Cylindrical Geometries In the case of spherical symmetry, the neutron transport equation becomes μ ∂ψ(r, μ) 1 − μ2 ∂ψ(r, μ) + + t (r)ψ(r, μ) ∂r r ∂μ  = 1 −1 dμ s (r, μ → μ)ψ(r, μ ) + S(r, μ) (9.146) where r is the magnitude of the radius vector r from the center of the spherical geometry and μ =  · r. Following the same procedure as above of expanding the angular dependence of the flux and differential scattering cross section according to Eqs. (9.124) and (9.125) and making use of the addition theorem, orthogonality relations, and the recursion relation  1 − μ2  dPm dμ (μ) = (m + 1)[μPm (μ) − Pm+1 (μ)] (9.147) yields the PL equations in spherical geometry: dφ1 2 + φ1 + (t − s0 )φ0 = S0 , n = 0 dr r     n n + 1 dφn+1 n + 2 dφn−1 n − 1 + φn+1 + − φn−1 2n + 1 dr r 2n + 1 dr r + (t − sn )φn = Sn , n = 1, . . . , L (9.148) For cylindrical symmetry, the formalism becomes more complex because the angular flux depends on two components of the neutron direction vector , instead of one as in the case of slab and spherical symmetry. With reference to Fig. 9.18, μ is defined with respect to the angle θ between  and the cylindrical axis, which is taken in the z-direction, and ϕ is defined as the angle in the x–y plane between the x–y projection of  and the radius vector r, noting that p / sin θ is a unit vector: μ = cos θ =  · nz , ϕ = cos−1 r · p sin θ (9.149) The neutron transport equation in systems with cylindrical symmetry becomes   ∂ψ(r, μ, ϕ) sin θ ∂ψ(r, μ, ϕ) − + t (r)ψ(r, μ, ϕ) sin θ cos ϕ ∂r 4 ∂ϕ  4π = ds (r,  ·  )ψ(r, μ , ϕ  ) + S(r, μ, ϕ) (9.150) 0 The expansion of the angular dependence of the differential scattering cross section is written in this coordinate system as 9.6 Spherical Harmonics (PL ) Methods in One-Dimensional Geometries Fig. 9.18 Nomenclature for cylindrical geometry PL equations. s (r,  ·  ) = L 1  2l  + 1 s (r, μ0 ) = sl  (r)Pl  (μ0 ) 2π 2π  2 l =0 = 1 2π L  l  =0 2l  + 1 sl  (r) 2    l  (l  − m)! m × Pl  (μ)Pl  (μ ) + 2 P  (μ)Plm (μ ) cos m(ϕ − ϕ  ) (l  + m)! l  m=0 (9.151) where the addition theorem for Legendre polynomials has been used in the last step. Using this expansion in Eq. (9.149) and multiplying, in turn, by all functions Plm (μ) cos(mϕ) for which l ≤ L, and making use of the recursion relations m−1 m−1 (μ) − (l − m + 1)(l − m + 2)Pl+1 (μ) (l + m − 1)(l + m)Pl−1 ) = (2l + 1) 1 − μ2 Plm (μ), m = 0 ) m+1 m+1 Pl+1 (μ) − Pl−1 (μ) = (2l + 1) 1 − μ2 Plm (μ) (9.152) (where it is understood here and below that terms with negative super- or subscripts are to be omitted) and the orthogonality relations 345 346 9 Neutron Transport Theory   2π dϕ 0 1 −1  dμPlm (μ)Plm (μ) cos mϕ cos m ϕ ⎧ 2(l + m)!δll  δmm ⎪ ⎪ , m = 0 ⎨π (2l + 1)(l − m)! = 2 ⎪ ⎪ ⎩ 2π m=0 δll  δmm , 2l + 1 (9.153) leads to the PL equations with an isotropic source in systems with cylindrical symmetry: Jlm+1 + Jlm−1 + (t − sl )φlm = Sδl0 , l = 0, . . . , L − 1; m = 1, . . . , l   φ m−1 (1 − δm0 )(L + m − 1)(L + m) d m−1 φL−1 − (m − 1) L−1 2(2L + 1) dr r  m+1  φ d m+1 1 + δm0 φ + (m + 1) L−1 + (t − sL )φLm = 0, − 2(2L + 1) dr L−1 r (9.154) l = L, m = 1, . . . , L where 1 Jlm+1 ≡ (1 + δm0 ) 2  m+1 dφl+1 − + (m + 1) dr  m+1 dφl−1 dr m+1 φl+1  r + (m + 1) m+1 φl−1  ! r (2l + 1)    m−1 φ m−1 dφl−1 1 − (m − 1) l−1 Jlm−1 ≡ (1 − δm0 ) (l + m − 1)(l + m) 2 dr r (9.155) − (l − m + 1)(l − m + 2)  × m−1 dφl+1 dr − (m − 1) m−1 φl+1  ! r (2l + 1) The PL equations are equations for the L + 1 flux moments  1  2π dμPlm (μ) dϕ cos mϕψ(r, μ, ϕ) φlm (r) ≡ −1 (9.156) 0 in terms of which the angular flux distribution is given by ψ(r, μ, ϕ) = L l  2l + 1  (l − m)! m φ (r)Plm (μ) cos mϕ 4π (l + m)! l l=0 (9.157) m=−l Either the Mark or Marshak boundary conditions are applicable in spherical or cylindrical geometry, on the exterior boundary, but these provide only (L + 1)/2 conditions. The other (L + 1)/2 conditions are provided by the requirement for symmetry about the origin, which requires the odd flux moments to vanish at the origin. The Marshak form of the interface continuity conditions are also applicable in these geometries. 9.6 Spherical Harmonics (PL ) Methods in One-Dimensional Geometries Diffusion Equations in One-Dimensional Geometry The P1 equations may be reduced to a diffusion theory form for spherical and cylindrical geometries. In general, letting r be the spatial coordinate upon which the flux distribution depends, the P1 equations may be reduced to   1 d n dφ(r) − n r D0 + [t (r) − s0 ]φ(r) = S0 (r) (9.158) r dr dr where n = 0 for planar geometry, 1 for cylindrical geometry, and 2 for spherical geometry. The reduction of the PL equations to coupled diffusion equations that was discussed for slab geometry is not possible in spherical and cylindrical geometries because it is not possible to eliminate coupling terms containing spatial derivatives of the Legendre flux moments. Thus the efficient diffusion theory solution procedures cannot be employed with the spherical and cylindrical PL equations, and other, generally less efficient iterative methods must be used (e.g., Ref. 6). Half-Angle Legendre Polynomials The efficacy of the PL method depends on the validity of representing the angular dependence of the neutron flux as a low-order continuous polynomial expansion over −1 ≤ μ ≤ 1. There are situations in which the flux may be highly directional and thus not well represented by a continuous polynomial expansion over both forward and backward directions, but in which the flux may be well represented by separate low-order polynomial expansions over the forward and backward directions. The half-angle Legendre polynomials have been developed for this purpose. The forward (μ > 0) and backward (μ < 0) half-angle Legendre polynomials are defined as pl+ (μ) ≡ Pl (2μ − 1), μ>0 pl− (μ) ≡ Pl (2μ + 1), μ<0 (9.159) These polynomials clearly satisfy pl+ (0) ≡ pl− (−1) = Pl (−1) pl+ (1) = pl− (0) = Pl (+1) (9.160) and may be shown from the orthogonality and recursion relations for the full-range Legendre polynomials to satisfy the orthogonality conditions  1  0 δlm + + − (9.161) pl (μ)pm (μ)dμ = dμ pl− (μ)pm (μ) = 2l +1 0 −1 and to have the recursion relations + + (μ) + (2l + 1)pl+ (μ) + lpl−1 (μ) = 2(2l + 1)μpl+ (μ) (l + 1)pl+1 − − (μ) − (2l + 1)pl− (μ) + lpl−1 (μ) = 2(2l + 1)μpl− (μ) (l + 1)pl+1 (9.162) 347 348 9 Neutron Transport Theory Double-PL Theory Expanding the flux separately in each half-space 0 ≤ μ ≤ 1 and −1 ≤ μ ≤ 0,  L + +  l  =0 (2l + 1)φl  (x)pl  (μ), μ > 0 ψ(x, μ)   (9.163) L − −  l  =0 (2l + 1)φl  (x)pl  (μ), μ < 0 substituting into Eq. (9.123), weighting in turn by each pl+ (l ≤ L) and integrating over 0 ≤ μ ≤ 1 and by each pl− (l ≤ L) and integrating over −1 ≤ μ ≤ 0, and making use of the orthogonality and recursion relations above yields a coupled set of 2(L + 1) double-PL , or D-PL , equations: + + dφl−1 l + 1 dφl+1 2l + 1 dφl+ l + + + t φl+ 2(2l + 1) dx 2(2l + 1) dx 2(2l + 1) dx = 2L+1  2l  + 1 + Cll  sl  φl  + Sl+ 2 l  =l − − dφl−1 l + 1 dφl+1 2l + 1 dφl− l − + + t φl− 2(2l + 1) dx 2(2l + 1) dx 2(2l + 1) dx = 2L+1  2l  + 1 − Cll  sl  φl  + Sl− , 2 l  =l (9.164) l = 0, 1, . . . , L where Cll+ ≡ Sl+ ≡  1 0  1 0 pl+ (μ)Pl  (μ) dμ, Cll− ≡ pl+ (μ)S(x, μ) dμ, Sl− ≡  0 −1  0 −1 dμ pl− (μ)Pl  (μ) (9.165) dμ pl− (μ)S(x, μ) The coupling between the forward (+) and backward (−) flux moment equations comes about because of the possibility of scattering from the interval −1 ≤ μ ≤ 0 to the interval 0 ≤ μ ≤ 1, and vice versa, as indicated by the scattering sums on the right in Eqs. (9.164). The upper limits on these summations arise because the expansion of the differential scattering cross section was terminated at 2L + 1. These scattering terms contain full-range Legendre flux moments which must be represented in terms of the half-range moments by using the approximate representation of the full-range Legendre polynomials in terms of the half-range polynomials:  L,l + +  l  =0 (2l + 1)Cl  l pl  (μ), μ > 0 Pl (μ)   (9.166) L,l − −  l  =0 (2l + 1)Cl  l pl  (μ), μ < 0 where the summation extends to l or L, whichever is smaller. This representation leads to  1 dμP (μ)ψ(x, μ)dx φl (x) ≡ −1 9.6 Spherical Harmonics (PL ) Methods in One-Dimensional Geometries L,l   l  =0 (2l  + 1) Cl+ l φl+ (x) + Cl− l φl− (x) (9.167) The final form of the D-PL equations is + + dφl−1 l + 1 dφl+1 2l + 1 dφl+ l + + + t φl+ 2(2l + 1) dx 2(2l + 1) dx 2(2l + 1) dx = Sl+ + 2L+1  l  =l − dφl+1 l+1 2(2l + 1) dx = Sl− + 2L+1  l  =l  L,l  (2l  + 1) + Cll  sl  Cll+ l  φl+ + Cll− l  φl− 2  l =0 − dφl−1 2l + 1 dφl− l − + + t φl− 2(2l + 1) dx 2(2l + 1) dx (9.168)  L,l  2l  + 1 − Cll  sl  Cll+ l  φl+ + Cll− l  φl− , 2  l = 0, 1, . . . , L l =0 Interface and boundary conditions for the D-PL equations are straightforward extensions of the conditions derived for the PL equations. All of the φl+ and φl− are continuous at interfaces. A vacuum boundary condition requires that the incoming flux moments be zero at that boundary [e.g., a vacuum condition on the left boundary requires that all φl+ (xL ) = 0, and a vacuum condition on the right boundary requires that all φl− (xL ) = 0]. A symmetry, or reflective, boundary condition requires that φl+ (xL ) = φl− (xL ). Known incident flux conditions at the left boundary, ψin (xL , μ > 0), or at the right boundary, ψin (xR , μ < 0), lead to boundary conditions  1 dμ pl+ (μ)ψin (xL , μ > 0) φl+ (xL ) = 0 (9.169)  0 φl− (xR ) = dμ pl− (μ)ψin (xR , μ < 0) −1 The D-PL approximation results in 2(L+1) first-order ordinary differential equations to be solved for 2(L + 1) unknowns, the flux moments φl+ and φl− . The same number of first-order ordinary differential equations and unknown flux moments φl are obtained in the P2L approximation. In problems in which the difference in the number of neutrons moving in the forward and backward directional half-spaces is more important than the angular distribution per se, the D-PL approximation is more accurate than the P2L approximation with the same number of unknowns. Thus the D-PL approximation is to be preferred for interface and boundary problems, whereas the P2L approximation is to be preferred for deep penetration problems. D-P0 Equations This simplest and most widely used of the D-PL methods is obtained by setting ± ± = 1 and C01 = 12 : L = 0 in the equations above and noting that C00 349 350 9 Neutron Transport Theory      3 3 1 dφ0+ 1 1 + t − s0 1 + μ̄0 φ0+ = s0 1 + μ̄0 φ0− + S0+ 2 dx 2 4 2 4 (9.170)      − 3 3 1 dφ0 1 1 − + − + t − s0 1 + μ̄0 φ0 = s0 1 + μ̄0 φ0 + S0 − 2 dx 2 4 2 4 9.7 Multidimensional Spherical Harmonics (PL ) Transport Theory Spherical Harmonics The spherical harmonics are defined as (note that there are several normalizations in use) √ (l − m)! m P (μ) exp(imϕ) Ylm (μ, ϕ) = √ (9.171) (l + m)! l in terms of the previously discussed associated Legendre functions. Denoting the complex conjugate by an asterisk, it follows that ∗ Yl,−m (μ, ϕ) = (−1)m Ylm (μ, ϕ) (9.172) The first few such functions are Y00 (μ, ϕ) = P00 (μ) = P0 (μ) = 1 Y10 (μ, ϕ) = P10 (μ) = P1 (μ) = μ 1 ) Y11 (μ, ϕ) = − √ 1 − μ2 (cos ϕ + i sin ϕ) 2 1 ) Y1,−1 (μ, ϕ) = √ 1 − μ2 (cos ϕ − i sin ϕ) 2 (9.173) and the remaining spherical harmonics can be generated using the recursion relation for the associated Legendre functions defined by Eq. (9.122): m m (2l + 1)μPlm (μ) = (l − m + 1)Pl+1 (μ) + (l + m)Pl−1 (μ) (9.174) With respect to Fig. 9.19, the directional cosines along Cartesian coordinate axes are given in terms of the spherical harmonics by  z ≡  · nz = μ .  1 1  ∗ ∗ − Y1,−1 x ≡  · nx = 1 − μ2 cos ϕ = − √ (Y11 − Y1,−1 ) = √ Y11 2 2 .  ∗  i −i ∗ y ≡  · ny = 1 − μ2 sin ϕ = √ (Y11 + Y1,−1 ) = √ Y11 + Y1,−1 2 2 (9.175) 9.7 Multidimensional Spherical Harmonics (PL ) Transport Theory Fig. 9.19 Nomenclature for spherical harmonics. The spherical harmonics satisfy the orthogonality relationship  1  2π 4π δll  δmm dμ dϕYl∗ m (μ, ϕ)Ylm (μ, ϕ) = 2l +1 −1 0 (9.176) and in terms of the spherical harmonics the addition theorem for Legendre polynomials can be written l  Pl (μ0 ) = ∗ Ylm (μ, ϕ)Ylm (μ , ϕ  ) (9.177) m=−l Spherical Harmonics Transport Equations in Cartesian Coordinates Expanding the angular dependence of the neutron flux ψ(r, ) = L l  2l + 1  4π l=0 φlm (r)Ylm (μ, ϕ) (9.178) m=−l and the differential scattering cross section  s (r, μ0 ) = L  2l  + 1 l  =0 4π sl  (r)Pl  (μ0 ) (9.179) in spherical harmonics, substituting the expansions into the neutron transport equation  · ∇ψ(r, ) + t (r, )ψ(r, )   1 = d νf (r)ψ(r,  ) + d s (r)(r,  ·  )ψ(r,  ) + S(r, ) 4π 4π 4π (9.180) 351 352 9 Neutron Transport Theory ∗ in turn, integrating over d, and making use of the ormultiplying by each Ylm thogonality and recursion relations and the addition theorem yields the spherical harmonics equations for the flux moments φlm :   )  1 ∂ 1 ∂ (l + m + 2)(l + m + 1) − −i φl+1,m+1 2l + 1 2 ∂x ∂y   ∂ 1) ∂ −i φl+1,m−1 + (l − m + 2)(l − m + 1) 2 ∂x ∂y   ∂ 1) ∂ +i φl−1,m+1 + (l − m − 1)(l − m) 2 ∂x ∂y   ∂ 1) ∂ +i φl−1,m−1 + (l + m − 1)(l + m) − 2 ∂x ∂y + + ) ) (l + m + 1)(l + m − 1) (l + m)(l − m) ∂ φl+1,m ∂z ∂ φl−1,m + t φlm ∂z = sl φlm + δl0 νf φ00 + Qlm , l = 0, . . . , L, m = −l, . . . , l (9.181) ∗ moment of the external source and the other quantities have where Qlm is the Ylm been discussed previously. This formidable set of equations is rarely solved as is; however, it provides the basis for the development of a number of useful approximations. Note that the equation for each flux moment φlm contains scattering terms involving only that same flux moment, so that the coupling among equations for different flux moments is entirely through the streaming terms arising from the  · ∇ψ term. P1 Equations in Cartesian Geometry As was the case in one dimension, the spherical harmonics equations lack closure. When the spatial derivatives involving φL+1 in the l = L equation are set to zero, the three-dimensional PL approximation is obtained. We consider the lowest-order P1 approximation in more detail. Using Eqs. (9.173) and (9.175), it can be shown that the flux moments are related to the scalar flux and to the currents along the various coordinate axes:  φ(r) ≡ ψ(r, )d = φ00  1 Jx (r) ≡ (n̂x · )ψ(r, )d = √ (φ1,−1 − φ11 ) 2  (9.182) −i Jy (r) ≡ (ny · )ψ(r, )d = √ (φ1,−1 + φ11 ) 2  Jz (r) ≡ (nz · )ψ(r, )d = φ10 9.7 Multidimensional Spherical Harmonics (PL ) Transport Theory Using these relations to express the flux moments in terms of the scalar flux and the currents, Eq. (9.178) becomes (for L = 1) ψ(r, ) = 1 [φ(r) + 3 · J(r)] 4π (9.183) Using the flux moments calculated from Eqs. (9.182) in the (l = 0, m = 0) equation (9.181) yields the exact equation (i.e., it was not necessary to discard a derivative of a higher moment in this equation) ∇ · J(r) + t (r)φ(r) = s0 (r)φ(r) + νf (r)φ(r) + Q00 (r) (9.184) Adding and subtracting the (l = 1, m = 1) and (l = 1, m = −1) equations (9.181) yields the approximate (i.e., it was necessary to discard a derivative of a higher moment in these equations) equations  1 ∂φ(r) + t (r)Jx (r) − s1 (r)Jx (r) = ( · nx )Q d ≡ Q1x 3 ∂x (9.185)  1 ∂φ(r) + t (r)Jy (r) − s1 (r)Jy (r) = ( · ny )Q d ≡ Q1y 3 ∂y and the (l = 1, m = 0) equation yields the approximate equation  1 ∂φ(r) + t (r)Jz (r) − s1 (r)Jz (r) = ( · nz )Q d ≡ Q1z 3 ∂z (9.186) Equations (9.184) to (9.186) are the three-dimensional P1 equations in Cartesian geometry. Diffusion Theory The one-dimensional P1 equations led to diffusion theory, and it is of some interest to see if the same is true in three dimensions. Equations (9.185) and (9.186) can be written as a Fick’s law: J(r) = − 1 ∇φ(r) ≡ −D∇φ(r) 3[t (r) − μ̄0 s (r)] (9.187) if the anisotropic source terms Q1 vanish. Equation (9.187) can be used in Eq. (9.184) to obtain the three-dimensional diffusion equation in Cartesian coordinates −∇ · D(r)∇φ(r) + [t (r) − s0 (r)]φ(r) = νf (r)φ(r) + Q00 (r) (9.188) Equation (9.187) and hence also the diffusion equation are thus based on two major assumptions: (1) spatial derivatives of higher flux moments φ2 can be neglected; and (2) anisotropic neutron sources can be neglected. Had we carried out the development from the time-dependent transport equation, it would have also been necessary to assume that the time derivatives of the current could be neglected to obtain a Fick’s law. 353 354 9 Neutron Transport Theory 9.8 Discrete Ordinates Methods in One-Dimensional Slab Geometry The discrete ordinate methods are based on a conceptually straightforward evaluation of the transport equation at a few discrete angular directions, or ordinates, and the use of quadrature relationships to replace scattering and fission neutron source integrals over angle with summations over ordinates. The essence of the methods are the choice of ordinates, quadrature weights, differencing schemes, and iterative solution procedures. In one dimension, the ordinates can be chosen such that the discrete ordinates methods are completely equivalent to the PL and D-PL methods discussed in Section 9.6, and in fact the use of discrete ordinates is probably the most effective way to solve the PL and D-PL equations in one dimension. This equivalence does not carry over into multidimensional geometries. Making use of the spherical harmonics expansion of the differential scattering cross section of Eq. (9.125) and the addition theorem for Legendre polynomials of Eq. (9.121), the one-dimensional neutron transport equation (9.123) in slab geometry becomes μ dψ (x, μ) + t (x)ψ(x, μ) dx  1  2l  + 1 dμ Pl  (μ )ψ(x, μ ) + S(x, μ) sl  (x)Pl  (μ) = 2 −1  (9.189) l =0 where the source term includes an external source and, in the case of a multiplying medium such as a reactor core, a fission source. We first discuss the solution of the fixed external source problem (which implicitly assumes a subcritical reactor) and then return to the solution of the critical reactor problem, in which the solution of the fixed source problem constitutes part of the iteration strategy. Defining N ordinate directions, μn , and corresponding quadrature weights, wn , the integral over the angle in Eq. (9.189) can be replaced by  φl (x) ≡ 1 −1 dμPl (μ)ψ(x, μ)   wn Pl (μn )ψn (x) (9.190) n where ψn ≡ ψ(μn ). The quadrature weights are normalized by N  wn = 2 (9.191) n=1 It is convenient to choose ordinates and quadrature weights that are symmetric about μ = 0, hence providing equal detail in the description of forward and backward neutron fluxes. This can be accomplished by choosing μN+1−n = −μn , μn > 0, n = 1, 2, . . . , N/2 wN+1−n = wn , wn > 0, n = 1, 2, . . . , N/2 (9.192) 9.8 Discrete Ordinates Methods in One-Dimensional Slab Geometry With such even ordinates, reflective boundary conditions are simply prescribed: ψn = ψN+1−n , n = 1, 2, . . . , N/2 (9.193) Known incident flux, ψin (μ), boundary conditions, including vacuum conditions when ψin (μ) = 0, are ψn = ψin (μn ), n = 1, 2, . . . , N/2 (9.194) Normally, an even number of ordinates is used (N = even), because this results in the correct number of boundary conditions and avoids certain other problems encountered with N = odd. Even with these restrictions, there remains considerable freedom in the choice of ordinates and weights. PL and D-PL Ordinates If the ordinates are chosen to be the N roots of the Legendre polynomial of order N , PN (μi ) = 0 (9.195) and the weights are chosen to integrate all Legendre polynomials correctly up to PN −1  1 −1 Pl (μ)dμ = N  wn Pl (μn ) = 2δl0 , l = 0, 1, . . . , N − 1 (9.196) n=1 then the discrete ordinates equations with N ordinates are entirely equivalent to the PN −1 equations. To establish this, we multiply Eq. (9.189) by wn Pl (μn ) for 0 ≤ l ≤ N − 1, in turn, and use the recursion relation of Eq. (9.119) to obtain  wn =  l+1 l dψn Pl+1 (μn ) + Pl−1 (μn ) + wn t ψn 2l + 1 2l + 1 dx N−1  l  =0 2l  + 1 sl  wn Pl  (μn )Pl (μn )φl  + wn Pl (μn )S(μn ), 2 l = 0, . . . , N − 1, n = 1, . . . , N (9.197) Summing these equations over 1 ≤ n ≤ N yields l + 1 dφl+1 l dφl−1 + + t φl 2l + 1 dx 2l + 1 dx N  N−1 N  2l  + 1   sl  φl  = wn Pl  (μn )Pl (μn ) + wn Pl (μn )S(μn ), 2  l =0 l = 0, . . . , N − 1 n=0 n=1 (9.198) 355 356 9 Neutron Transport Theory Weights chosen to satisfy Eqs. (9.196) obviously correctly integrate all polynomials through order N (any polynomial of order n can be written as a sum of Legendre polynomials through order n), but fortuitously they also integrate correctly all polynomials through order less than 2N . Thus the term in the scattering integral becomes N   wn Pl  (μn )Pl (μn ) = 1 −1 n=1 Pl  (μ)Pl (μ)dμ = 2δll  2l + 1 (9.199) and assuming that the angular dependence of the source term can be represented by a polynomial of order < 2N : N   wn Pl (μn )S(μn ) = n=1 1 −1 Pl (μ)S(μ)dμ = 2Sl 2l + 1 (9.200) where Sl is the Legendre moment of the source given by Eq. (9.127). Using Eqs. (9.199) and (9.200), Eqs. (9.198) become l + 1 dφl+1 l dφl−1 + + (t − sl )φl = Sl , l = 0, . . . , N − 2 2l + 1 dx 2l + 1 dx (9.201) dφ(N−1)−1 N −1 + (t − s,N−1 )φN−1 = SN−1 , l = N − 1 2(N − 1) + 1 dx which, when φ−1 is set to zero, are identically the PL equations (9.126) for L = N − 1. These PL ordinates and weights are given in Table 9.2. The D-PL ordinates are the roots of the half-angle Legendre polynomials for L = N/2 − 1: PN/2 (2μn + 1) = 0, n = 1, 2, . . . , PN/2 (2μn − 1) = 0, n= N 2 (9.202) N + 1, . . . , N 2 and the corresponding weights are determined from N/2  wn Pl (2μn + 1) = δl0 , n=1 N  n=(N+2)/2 wn Pl (2μn − 1) = δl0 , l = 0, . . . , N −2 2 N −2 l = 0, . . . , 2 (9.203) These ordinates and weights may be evaluated from the data in Table 9.2. The PL ordinates and weights are preferable to the D-PL ordinates and weights for deep penetration problems in heterogeneous media and for problems in which anisotropic scattering is important, for both of which the correct calculation of a large number of Legendre moments of the flux are required. Conversely, for the 9.8 Discrete Ordinates Methods in One-Dimensional Slab Geometry Table 9.2 PN −1 Ordinates and Weights ±μn wn N =2 0.57735 1.00000 N =4 0.33998 0.86114 0.65215 0.34785 N =6 0.23862 0.66121 0.93247 0.46791 0.36076 0.17132 N =8 0.18343 0.52553 0.79667 0.96029 0.36268 0.31371 0.22238 0.10123 N = 10 0.14887 0.43340 0.67941 0.86506 0.97391 0.29552 0.26927 0.21909 0.14945 0.06667 N = 12 0.12523 0.36783 0.58732 0.76990 0.90412 0.98156 0.24915 0.23349 0.20317 0.16008 0.10694 0.04718 Source: Data from Ref. 2; used with permission of Wiley. calculation of highly anisotropic neutron fluxes near boundaries, the D-PL ordinates and weights are preferable. With either set of ordinates and weights, the discrete ordinates method in one dimension is essentially a numerical method for solving the PL or D-PL equations. Other choices of weights and ordinates can be made to specialize the discrete ordinates method to the problem to be solved (e.g., bunching ordinates to emphasize an accurate calculation of the neutron flux in a certain direction). However, care must be exercised when choosing ordinates and weights that do not correctly integrate the low-order angular polynomials, because surprising results sometimes turn up. Spatial Differencing and Iterative Solution Defining cross sections to be constant over xi−1/2 < x < xi+1/2 , Eq. (9.189), for each ordinate, can be integrated over xi−1/2 < x < xi+1/2 to obtain 357 358 9 Neutron Transport Theory  i+1/2 i−1/2  μn ψn − ψn + ti ψni i  L   2l  + 1 i i i i sl  Pl  (μn )φl  + S (μn ) = i Qn ≡ i 2  (9.204) l =0 where ψni ≡ ψ(xi , μn ), and so on, and i = xi+1/2 − xi−1/2 . Using the diamond difference relation  i+1/2 i−1/2  + ψn ψni = 12 ψn (9.205) algorithms for sweeping to the right in the direction of neutrons traveling with μn > 0,     ti i −1 i Qin i−1/2 i ψn + ψn = 1 + 2|μn | 2|μn | (9.206) i+1/2 ψn i−1/2 = 2ψni − ψn and for sweeping to the left in the direction of neutrons traveling with μn < 0,      i i −1 i Qin i+1/2 ψn ψni = 1 + t + 2|μn | 2|μn | (9.207) i−1/2 ψn i+1/2 = 2ψni − ψn are specified. The boundary conditions at the left boundary (for incident flux or vacuum conditions) are specified for the positive-direction ordinates by Eqs. (9.194) (e.g., 1/2 ψn = 0, μn > 0 for a vacuum condition). Note that the physical boundaries are located at x1/2 and xI +1/2 . Equations (9.206) are then used to sweep the solutions for ordinates μn > 0 to the right boundary, where conditions similar to Eqs. (9.194) I +1/2 = 0, μn < 0 for a vacuum condition) specify the boundary conditions (e.g., ψn for the ordinates with μn < 0, and Eqs. (9.207) are used to sweep the solutions for μn < 0 from the right to the left boundary. If there were no scattering or fission sources in Qin , the solution would be complete. However, there are, and this iterate of the fluxes must be used to update the Qin and the double-sweep repeated until convergence. If there is a reflective boundary, say on the right, the condition I +1/2 I +1/2 is used for the return sweep (the problem should be stated so ψN+1−n = ψn that the reflective boundary is on the right). If there are reflective conditions on both boundaries, the boundary conditions on the left must be initially guessed, then updated following a double-sweep, and so on, which, of course, slows convergence. Limitations on Spatial Mesh Size Truncation error determines the allowable spatial mesh size. Consider Eq. (9.189), for a given ordinate, but without the source term: μn dψn (x) + t (x)ψn (x) = 0 dx (9.208) 9.9 Discrete Ordinates Methods in One-Dimensional Spherical Geometry The exact solution for the flux at xi+1/2 in terms of the flux at xi−1/2 is i+1/2 ψn  = exp  −ti i i−1/2 ψn |μn | (9.209) The finite difference solution is found by using Eq. (9.205) to eliminate ψni in Eq. (9.204) with Qin = 0: i+1/2 ψn = 1 − ti i /2|μn | 1 + ti i /2|μn | i−1/2 ψn (9.210) The error in the approximate solution is O((ti i /2|μn |)2 ). The allowable mesh spacing is determined by the accuracy required and the smallest value of |μn |. Negative fluxes will occur if i > 2|μn |/ti . Negative flux fix-up schemes have been developed, which amount to setting negative fluxes to zero when they occur in the iteration, but this introduces difficulties. This problem is sufficiently serious to have motivated the development of a number of alternative difference schemes, but variants of the diamond differencing scheme remain the most commonly used. 9.9 Discrete Ordinates Methods in One-Dimensional Spherical Geometry The angles that specify the neutron direction in curvilinear geometry change as the neutron moves, as shown in Fig. 9.20. This leads to angular derivatives in the neutron streaming operator, making curvilinear geometries qualitatively different from slab geometry. The conservative form of the neutron transport equation in spherical geometry is  1 ∂  μ ∂ 2 ρ ψ(ρ, μ) + 1 − μ2 ψ(ρ, μ) + t (ρ)ψ(ρ, μ) = Q(ρ, μ) 2 ρ ∂μ ρ ∂ρ (9.211) Fig. 9.20 Change in angular coordinate μ = cos θ as the neutron moves. (From Ref. 2; used with permission of Wiley.) 359 360 9 Neutron Transport Theory Representation of Angular Derivative The difference scheme for the angular derivative is determined by the requirement that the sum of the angular and radial streaming terms (the first two terms in the equation above) satisfy the physical constraint of vanishing for an uniform, isotropic flux in an infinite medium. Approximating the angular derivative as  1 ∂  2 (αn+1/2 ψn+1/2 − αn−1/2 ψn−1/2 ) 1 − μ2 ψ(ρ, μ)  ρ ∂μ ρwn (9.212) and noting that for an uniform medium and an isotropic flux that ψn = ψn±1 = φn /2, the scalar flux, the requirement that the spatial plus angular derivative terms vanish is αn+1/2 = αn−1/2 − μn wn (9.213) which is an algorithm for determining the αn+1/2 once α1/2 is known. By choosing α1/2 = 0 and N even, Eq. (9.213) yields αN+1/2 = 0, which leads to closure in the angular differencing algorithm. Using this form for the angular derivative and an angular diamond difference relation ψn = 12 (ψn+1/2 + ψn−1/2 ) (9.214) in Eq. (9.211) yields μn ∂  2  ρ ψn ρ 2 ∂ρ + 2 [2αn+1/2 ψn − (αn+1/2 + αn−1/2 )ψn−1/2 ] + t ψn = Qn ρwn (9.215) The spatial differencing proceeds as for the slab case, but taking into account the variation of differential area and volume with radius. Iterative Solution Procedure The equations are solved by sweeping in the direction of neutron travel. With reference to Fig. 9.21, for an S4 (N = 4) calculation, the calculation is started on the outer surface of the sphere for the direction n = 12 . A known incident flux (including vacuum) boundary condition I +1/2 ψn = ψin (μn ), n = 1, . . . , N/2 I +1/2 provides a starting value for ψ1/2 for n = 1/2 (μ = −1) using i−1/2 ψn i+1/2 = 2ψni − ψn (9.216) . The calculation sweeps inward (decreasing i) (9.217) 9.9 Discrete Ordinates Methods in One-Dimensional Spherical Geometry Fig. 9.21 Sweep of the space–angle mesh for one-dimensional spherical geometry. (From Ref. 2; used with permission of Wiley.) i+1/2 i ψ1/2 = 2ψ1/2 + (ρi+1/2 − ρi−1/2 )Qi1/2 2 + ti (ρi+1/2 − ρi−1/2 ) , (9.218) μ<0 Next, the ψ1i row is calculated using the starting value from Eq. (9.216) and using Eq. (9.217) and −1  2 (Ai+1/2 − Ai−1/2 )αn+1/2 + Vi ti ψni = 2|μn |Ai−1/2 + wn  i+1/2 × |μn |(Ai+1/2 + Ai−1/2 )ψn + 1 i (Ai+1/2 − Ai−1/2 )(αn+1/2 + αn−1/2 )ψn−1/2 + Vi Qin wn  (9.219) i to sweep the solution inward. Then the ψ3/2 are calculated from the angular diamond difference relation i i ψn+1/2 = 2ψni − ψn−1/2 (9.220) These inward sweeps are continued, using, alternatively, Eqs. (9.217) and (9.219) i for the ψni and Eqs. (9.220) for the ψn+1/2 until all the inward (μn < 0, n ≤ N/2) fluxes are calculated. The starting fluxes at the center of the sphere (i = 1/2) for the outward (μn > 0, n > N/2) calculation are determined from the symmetry condition at the center of the sphere: 1/2 1/2 ψN +1−n = ψn , n = 1, 2, . . . , N/2 (9.221) 361 362 9 Neutron Transport Theory Then the calculation is swept outward (increasing i) using for ψni i+1/2 ψn i−1/2 = 2ψni − ψn (9.222) −1  2 (Ai+1/2 − Ai−1/2 )αn+1/2 + Vi ti ψni = 2|μn |Ai+1/2 + wn  i−1/2 × |μn |(Ai+1/2 + Ai−1/2 )ψn + 1 i (Ai+1/2 − Ai−1/2 )(αn+1/2 + αn−1/2 )ψn−1/2 + Vi Qin wn  (9.223) i : and the angular diamond difference relation for ψn+1/2 i i = 2ψni − ψn−1/2 ψn+1/2 (9.224) The A’s and V ’s in the equations above are the shell areas and differential volume elements at the radii indicated: 2 Ai+1/2 = 4πρi+1/2 , Vi =  4π  3 3 ρi+1/2 − ρi−1/2 3 (9.225) From these directional fluxes the scalar flux is calculated and the scattering and fission source terms in Q are updated for the next iteration. Acceleration of Convergence The numerical solution for the fluxes ψni on each double sweep is exact for the given scattering and fission source guess Q. The rate of convergence of the solution depends on the rate of convergence of these sources. Note from Eq. (9.204) that these sources depend only on the Legendre flux moments defined by Eq. (9.190) as a weighted sum over the ordinates of the ψni . This suggests that the iterative solution for the ψni can be accelerated by advancing the solution for the φli in a low order (e.g., diffusion theory) approximation at intermittent steps during the iteration, which is the basis of the synthetic method. Another acceleration technique—coarse mesh rebalance—makes use of the fact that the converged solution for the ψni must satisfy neutron balance. Imposing this condition on the unconverged solution over coarse mesh regions that include a number of spatial mesh points at intermittent steps in the iteration provides a means for accelerating the solution. Both acceleration methods, which are discussed in detail in Ref. 2, may become unstable if the spatial mesh spacing is not sufficiently small. The synthetic method may even become unstable with small mesh spacing. Other acceleration methods, such as Chebychev acceleration, may also be applied to accelerate the discrete ordinates solution. Calculation of Criticality Up to this point, we have discussed solving the discrete ordinates equations for a fixed external source. We now consider the critical reactor problem, in which 9.10 Multidimensional Discrete Ordinates Methods there is no external source. In this case the equations above would be modified by the inclusion of an effective multiplication constant, k −1 , as an eigenvalue in the fission term. A value k0 and an initial flux guess ψ (0) would be used to eval(0) (0) uate the fission Sf and scattering Ss sources, and the solution above would be carried out to obtain a first iterate flux solution ψ (1) . An improved fission source (1) (1) (0) Sf (ψ (1) /k1 ), an improved eigenvalue guess k1 = k0 Sf /Sf , and an improved scattering source Ss(1) (ψ (1) /k1 ) would be constructed, and the solution would be repeated to obtain ψ (2) , and so on, until the eigenvalues obtained on successive iterates converged to within a specified tolerance. There are also techniques for accelerating this power iteration procedure. 9.10 Multidimensional Discrete Ordinates Methods Ordinates and Quadrature Sets Two angular coordinates are required to specify the direction of motion in multidimensional geometries. With reference to Fig. 9.22, denote the direction cosines of the neutron direction  with respect to the x1 -, x2 -, and x3 -coordinate axes as μ, η, and ξ , respectively. Only two of these direction cosines are independent, and since  is a unit vector, μ2 + η2 + ξ 2 = 1. Fig. 9.22 Coordinate system for multidimensional discrete ordinates. (From Ref. 2; used with permission of Wiley.) 363 364 9 Neutron Transport Theory Fig. 9.23 Level symmetric S8 discrete ordinates quadrature set. (From Ref. 2; used with permission of Wiley.) In three-dimensional problems, the flux must be determined in all eight octants of the unit sphere over which  varies. In two-dimensional geometries, there is an assumption of symmetry in one of the coordinate directions, which reduces to four the number of octants over which the flux must be determined. (In onedimensional geometries, there is an assumption of symmetry in two of the coordinate directions, and the flux must be determined only within two of the octants.) It is convenient to use a set of ordinates that are symmetric in the eight octants (i.e., can satisfy reflective conditions across surfaces in the x1 –x2 plane, the x2 –x3 plane, and the x3 –x1 plane). Then, if the ordinates and weights are constructed for a set of direction cosines satisfying μ2n + ηn2 + ξn2 = 1 in one octant, the ordinates and weights for the other octants with direction cosine sets (−μn , ηn , ξn ), (μn , −ηn , ξn ), (μn , ηn , −ξn ), (−μn , −ηn , ξn ), (−μn , ηn , −ξn ), (μn , −ηn , −ξn ), and (−μn , −ηn , −ξn ) are obtained simply by changing the signs of one or more direction cosines. 9.10 Multidimensional Discrete Ordinates Methods Fig. 9.24 Equal-weighted ordinates for one octant in the SN quadrature. (From Ref. 2; used with permission of Wiley.) The level symmetric quadratures shown in Fig. 9.23 use the same set of N/2 positive values of the direction cosines with respect to each of the three axes (i.e., μn = ηn = ξn , n = 1, . . . , N/2). Use of such a quadrature set strictly defines the SN method, although the term SN is loosely used more widely as a synonym for discrete ordinates. The rotational symmetry of the level symmetric quadrature set and the requirement μ2n + ηn2 + ξn2 = 1 determines all the direction cosines except one. Once μ1 is chosen, the other μn are calculated from μ2n = μ21 + 2(n − 1) 1 − 3μ21 N −2 (9.226) and the ηn = ξn = μn . For the S2 approximation, with only one direction cosine, √ satisfaction of μ21 + η12 + ξ12 = 1 uniquely specifies η1 = ξ1 = μ1 = 1/3, and there are no degrees of freedom in the choice of ordinates. The weights in each octant are normalized by N(N+2)/8  wn = 1 (9.227) n=1 where the index n runs over all the (μi , ηj , ηk ), i, j, k = 1, . . . , N/2 ordinate combinations in the octant. For the S2 approximation, with only one ordinate per octant, w1 = 1. For other SN approximations the level symmetry condition μn = ηn = ξn requires that the weights be equal for ordinates obtained by permuting the direction cosines, as shown in Fig. 9.24, where the same value of wn is assigned to all the ordinates indicated by the same number. Note that unlike the situation in one dimension, this level symmetric quadrature set does not integrate Legendre polynomials to any given order accurately. However, even within the restrictions discussed above, there remain a few degrees of 365 366 9 Neutron Transport Theory Table 9.3 Level Symmetric SN Quadrature Set SN n μn wn S4 1 2 1 2 3 1 2 3 4 1 2 3 4 5 6 0.35002 0.86889 0.26664 0.68150 0.92618 0.21822 0.57735 0.78680 0.95119 0.16721 0.45955 0.62802 0.76002 0.87227 0.97164 0.33333 – 0.17613 0.15721 – 0.12099 0.09074 0.09259 – 0.07076 0.05588 0.03734 0.05028 0.02585 – S6 S8 S12 Source: Data from Ref. 2; used with permission of Wiley. freedom, and these may be chosen for the purpose of correctly integrating the maximum number of Legendre polynomials in each of the angular variables consistent with the number of degrees of freedom. A quadrature set so constructed is given in Table 9.3. SN Method in Two-Dimensional x–y Geometry The discrete ordinates equations in two-dimensional x–y geometry are μn ∂ψ(n ) ∂ψ(n ) + ηn + t ψ(n ) = Q(n ) ∂x ∂y (9.228) where the spatial dependence has been suppressed, n = (μn , ηn ), and the source Q includes a spherical harmonics representation of the scattering source plus a fission and external source S: Q(n ) = l L   (2 − δm0 )Ylm (n )sl φlm + S(n ) (9.229) l=0 m=0 and the discrete ordinates approximation for the flux moments are φlm = 1 4 N(N+2)/2  wn Ylm (n )ψ(n ) (9.230) n=1 Dividing the x–y domain of the problem into mesh boxes xi−1/2 < x < xi+1/2 , yj −1/2 < y < yj +1/2 centered at (xi , yj ) with constant cross sections within each 9.10 Multidimensional Discrete Ordinates Methods mesh box, integrating Eq. (9.228) over a mesh box, and defining volume-averaged quantities   1 ij ψn ≡ dx dy ψn (x, y) (9.231) xi yj i j   1 ij dx dy Qn (x, y) (9.232) Qn ≡ xi yj i j and surface-averaged fluxes  1 i+1/2,j ≡ dy ψn (xi+1/2 , y) ψn yj j  1 i,j +1/2 ≡ dx ψn (x, yj +1/2 ) ψn xi i (9.233) (9.234) yields the neutron balance equation on a mesh box: μn  i+1/2,j ηn  i,j +1/2 i−1/2,j  i,j −1/2  ij ij ij − ψn − ψn ψn + ψn + t ψn = Qn xi yj (9.235) It is necessary to relate the volume-averaged flux to the surface-averaged fluxes for each mesh box. There are several methods for doing this, the most common of which are the diamond difference method, which is used here, and the thetaweighted method. The volume- and surface-averaged fluxes are related in the diamond difference method by ij ψn = 1 2  i+1/2,j ψn i−1/2,j  + ψn = 1 2  i,j +1/2 i,j −1/2  + ψn ψn (9.236) These equations are solved by sweeping the two-dimensional mesh grid in the direction of neutron travel. With respect to Fig. 9.25, each iteration (on the scattering source) consists of four sweeps through the grid corresponding to the four octants. For the octant with (μn > 0, ηn > 0), the sweep is left to right, bottom to top; for the octant with (μn < 0, ηn > 0), the sweep is right to left, bottom to top; for the octant with (μn > 0, ηn < 0), the sweep is left to right, top to bottom; and for the octant with (μn < 0, ηn < 0), the sweep is right to left, top to bottom. For the octant with (μn > 0, ηn > 0), Eqs. (9.236) can be used to write Eqs. (9.235) as   2μn 2ηn −1 ij ij + ψn = t + xi yi   2μn i−1/2,j 2ηn i,j −1/2 ij (9.237) × ψn + ψn + Qn xi yj 1/2,j Starting with known incident flux (including vacuum) conditions ψn = ψin (xL , i,1/2 = ψin (yB , ηn > 0), i = 1, . . . , I , where xL refers μn > 0), j = 1, . . . , J and ψn 367 368 9 Neutron Transport Theory Fig. 9.25 Order of sweeping the two-dimensional (x–y) mesh grid for the octant with (μn > 0, ηn > 0). (From Ref. 2; used with permission of Wiley.) to the left boundary and yB refers to the bottom boundary, the flux ψn11 can be calculated with Eq. (9.237). The solution is then swept to the right using, al3/2,1 I +1/2,1 . Then ternatively, Eq. (9.235) and (9.236) to calculate ψn , ψn2,1 , . . . , ψn 1,3/2 2,3/2 I,3/2 Eq. (9.236) is used to calculate ψn , ψn , . . . , ψn . Using the boundary con1/2,2 = ψin (xL , μn > 0), Eqs. (9.235) and (9.236) can be used alternatively ditions ψn to sweep to the right across the j = 2 row, and then Eq. (9.236) can be used to sweep to the right across the j = 2 12 row, and so on, until all the outgoing fluxes are calculated. Sweeps through the other three octants are carried out in a similar manner but in the order indicated above and with Eqs. (9.235) and (9.236) combined in such a way as to obtain an algorithm like Eq. (9.237) appropriate for that octant. The scalar flux φ ij = 1 4 N(N+2)/2  n=1 ij wn ψn (9.238) 9.11 Even-Parity Transport Formulation and the Legendre moments ij φlm = 1 4 N(N+2)/2  ij wn Ylm (n )ψn (9.239) n=1 are then constructed and used to evaluate the scattering and fission source terms. The process is repeated until source convergence on successive iterations is within a specified tolerance. Further Discussion The discrete ordinates method in multidimensional geometries is highly geometry dependent. Because of the coupling of spatial and angular mesh intervals, the methodology was initially limited to the regular geometries: parallelepipeds, cylinders, and spheres. However, the development of triangular spatial mesh techniques enables a variety of geometries to be approximated. A number of other ordinate and weight quadrature sets have been devised for special purposes (e.g., to emphasize a given direction in a deep penetration problem). The acceleration methods discussed for the one-dimensional discrete ordinates methods are also used for multidimensional discrete ordinates solutions, but the higher dimensionality introduces complications that diminish their efficacy. In problems with optically thick regions in which the scattering cross section (within-group scattering cross section in multigroup applications) is much larger than the absorption cross section, the source convergence can become intolerably slow. In problems with very little scattering and localized neutron sources, unphysical oscillations in the angular distribution, known as ray effects, arise because of discrete directions in which the solution is calculated. There are special remedies for these ray effects, such as a semianalytical calculation of a first collision source to be used in a subsequent discrete ordinates calculation. These difficulties notwithstanding, the discrete ordinates method provides a powerful means for calculating the neutron flux distribution in a nuclear reactor core and the surrounding shield and structure, and is widely used for problems in which diffusion theory is inadequate. Detailed discussions of discrete ordinates methods can be found in Refs. 2 and 5. 9.11 Even-Parity Transport Formulation The one-group, or within-group, transport equation can be written in the case of isotropic sources and isotropic scattering:  · ∇ψ(r, ) + t (r, )ψ(r, ) = s (r)φ(r) + S(r) (9.240) Defining the (+) even- and (−−) odd-parity components of the angular flux 1 ψ ± (r, ) = [ψ(r, ) ± ψ(r, −)] 2 (9.241) 369 370 9 Neutron Transport Theory results in the following identities ψ(r, ) = [ψ + (r, ) + ψ − (r, )] ψ + (r, ) = ψ + (r, −) (9.242) ψ − (r, ) = −ψ − (r, −) which can be used to demonstrate that the scalar flux and current can be written in terms of the even and odd, respectively, components     ψ(r) ≡ d ψ + (r, ) + ψ − (r, ) = d ψ + (r, )     J(r) ≡ d  ψ + (r, ) + ψ − (r, ) = d ψ − (r, ) (9.243) Adding Eq. (9.240) written for − to the same equation written for  and using Eq. (9.241) yields  · ∇ψ − (r, ) + t (r)ψ + (r, ) = s (r)φ(r) + S(r) (9.244) and subtracting the same two equations yields  · ∇ψ + (r, ) + t (r)ψ − (r, ) = 0 (9.245) The second of these equations may be used in the first to eliminate the odd-parity flux component, resulting in an equation for the even-parity flux:  − · ∇  1  · ∇ψ + (r, ) + (r)ψ + (r, ) = s (r)φ(r) + S(r) (9.246) t (r) and Eq. (9.245) may be used to write the current in terms of the even-parity component:  J(r) = − d  1  · ∇ψ + (r, ) t (r) (9.247) The vacuum boundary condition becomes [ from Eqs. (9.242) and (9.245)] 0 = ψ(rs , ) =  · ∇ψ + (rs , ) ± t (rs )ψ + (rs , ),  · ns 0 (9.248) and the reflection boundary condition is ψ + (rs , ) = ψ + (rs ,  ) (9.249) where  is the direction of spectral reflection relative to incident direction . 9.12 Monte Carlo Methods 9.12 Monte Carlo Methods At a fundamental level, neutron transport through matter is formulated as an essentially stochastic process. The total cross section is a probability (per unit path length and unit atom density), but not a certainty, that a neutron will have a collision while traversing a certain spatial interval. If the neutron does have a collision, the cross sections for the various processes are probabilities, but not certainties, that the collision will be a scattering, radiative capture, fission, and so on, event. The neutron flux that we have discussed earlier in the chapter is actually the mean, or expectation, value of the neutron distribution function. The Monte Carlo method directly simulates neutron transport as a stochastic process. Probability Distribution Functions Let us postulate that variable x may take on various values over the interval a ≤ x ≤ b and that there exists a probability distribution function (pdf), f (x), such that f (x)dx is the probability that a variable takes on a value within dx about x. The normalization is chosen such that  a b f (x)dx = 1 (9.250) In general, f (x) ≥ 0 will not be a monotonically increasing function of x, which means that a given value for f does not correspond to a unique value of x. A more useful quantity is the cumulative probability distribution function (cdf), F (x), defined as the probability that the variable x takes on a value less than or equal to x:  F (x) = x f (x  )dx  (9.251) a which is a monotonically increasing function of x. Thus the probability of a neutron having a value of x between x and x + dx is F (x + dx) − F (x) = f (x)dx. If κ is a random number distributed between 0 and 1, the values of x determined from F (x) = κ will be distributed as f (x). In some cases, it is possible to solve directly for x = F −1 (κ). In other cases, the cumulative distribution function may be known as a large table of F (xi ) and the value of x determined by interpolation; for example, if F (xj ) < κ < F (xj −1 ) linear interpolation yields x = xj − F (xj ) − κ (xj − xj −1 ) F (xj ) − F (xj −1 ) (9.252) There are also methods of selection from the pdf, but it is generally preferable to select from the cdf. 371 372 9 Neutron Transport Theory Analog Simulation of Neutron Transport By tracing the path of an individual neutron as it traverses matter and considering the various processes that may determine its history, we can understand how a Monte Carlo calculation simulates the stochastic nature of neutron transport through matter. We begin with the source of neutrons in a nuclear reactor, which is predominantly if not entirely the fission source. The fission source has a distribution in space (we discuss calculation of the fission source distribution in Monte Carlo later), a distribution in energy given by the fission spectrum, and a distribution in direction that is isotropic. Each of these distributions may be characterized by a pdf and a cdf. Generating a random number and selecting from the cdf for the spatial fission distribution defines a location in space for the source particle. Generating another random number and selecting from the cdf for the fission spectrum determines the energy of the source particle. Generating third and fourth random numbers and selecting from the cdf’s for the two independent angular variables (say μ = cos θ and ϕ) defines the direction of the source neutron. Once launched, the source neutron will travel in a straight line until it has a collision. The probability that a neutron has a collision at a distance s along the flight path is    s (9.253) t (s  )ds  T (s) = t (s) exp − 0 which is the pdf for the collision distance s. Generating a random number λ and selecting s from the cdf  s t (s  )ds  (9.254) − ln λ = 0 locates the position of the first collision, in principle. In fact, the process is considerably complicated by the nonuniform geometry. It is necessary to know the composition at the point of the first collision. We treat the medium as piecewise homogeneous and define the lengths of each uniform segment of the straightline flight path as sj . If n−1  j =1 tj sj ≤ − ln λ < n  tj sj (9.255) j =1 the collision occurs in the nth region at a distance %  n  1 − ln λ − sn = tj sj tn (9.256) j =1 beyond the entrance of the flight path into region n. The actual procedures for treating flight paths in complex geometries are quite involved but highly developed. Modern Monte Carlo codes can essentially model any geometry exactly, which is a great strength of the method. 9.12 Monte Carlo Methods Having determined that a collision occurred at a distance sn into region n on the original flight path, it is now necessary to determine what type of nuclide and what type of reaction are involved. The probability for a reaction of type x with a nuclide of species i is Ni σix i,x Ni σix pix =  (9.257) where Ni is the number density of nuclide i in region n, σix is the microscopic cross section for reaction x for nuclide i at the energy of the neutron. Constructing a pdf and a cdf, generating a random number η, and selecting the nuclide and reaction type by equating η and the cdf [probably involving table interpolation per Eq. (9.252)], the nuclide and reaction type can be determined. If the reaction type is absorption, the neutron history is terminated, the energy and location of the absorbed neutron are recorded, and another history is started. If the reaction type is elastic scattering, another random number is generated and equated to the cdf for the cosine of the scattering angle in the center of mass (CM) to obtain μcm (it is convenient to work in the CM because the scattering is isotropic except for high-energy neutrons scattering from heavy mass nuclei, and the pdf and cdf are simple) and by transformation to obtain the scattering angle in the lab. For energies above thermal, the energy of the scattered neutron is uniquely correlated to μcm from the scattering kinematics: E = E(A2 + 2Aμcm + 1) (A + 1)2 (9.258) Knowing E  , the cosine of the scattering angle in the lab can be determined from $ $ E E 1 1 (A − 1) + μ = cos θ = (A + 1)  2 E 2 E (9.259) When inelastic scattering or elastic scattering of thermal neutrons from bound lattice atoms is involved, the cdf’s are more complicated. Generating another random number and equating it to the cdf for the azimuthal angle ϕ, the direction of the scattered neutron can be determined. The scattered neutron is treated as described above for a fission source neutron, and the calculation is repeated until the neutron either leaks from the system or is absorbed. Statistical Estimation The mean, or expectation, value of a function h(x) of x is defined in terms of the pdf for x by  h = b dxh(x)f (x) a and the standard deviation, σ , and the variance, V , are defined: (9.260) 373 374 9 Neutron Transport Theory σ (h) =  ) V (h) = b a = 1/2 dx[h(x) − h ]2 f (x) h2 − h 2 1/2 (9.261) If N random values of the variable x are chosen from the cdf, as discussed above, a statistical estimate of the mean value h is h̄ = N 1  h(xn ) N (9.262) n=1 A bound for the error in an estimate of this type is given by the central limit theorem, which states that if many estimates h̄ of h are obtained, each estimate involving N trials, the variable h̄ is normally distributed about h to terms of accuracy O(1/N 1/2 ). In the limit N → infinity, this theorem takes the form ⎧  ⎨ 0.6826, M = 1 Mσ (h) Mσ (h) Prob h − √ ≤ h̄ ≤ h + √ = 0.954, (9.263) M =2 ⎩ N N 0.997, M =3 [i.e., the probability that the statistical estimate of the mean value of Eq. (9.262) is within ±Mσ/N 1/2 of the exact value h is 68.3% for M = 1, 95.4% for M = 2, 99.7% for M = 3, etc.]. In general, the first and second moments of h(x) are unknown. The statistical data can be used to construct approximations to these moments. The expectation value of h̄ is N N  0 1 1  b 1  h(xn ) = dxn f (xn )h(xn ) h̄ = N N a n=1 n=1 = N  N 1  b 1  dx f (x)h(x) = h = h N N a n=1 (9.264) n=1 (i.e., the statistical estimate h̄ is an unbiased estimate of h since ĥ = h . The expected value of h̄2 is 2N 2N 3 3 N N N    0 21 1  2 1  h(xn ) h(xm ) = 2 h (xn ) + h(xm ) h(xm ) h̄ = 2 N N n=1 = m=1 0 1 1 N h2 + N(N − 1) h N2 n=1 2 = h2 N − n=1 N −1 h N 2 m=n (9.265) (i.e., the statistical estimate h2 is a biased estimate of h2 since h̄2 = h2 . Since h̄2 = h2 , the variance in the statistical estimate of h̄ (h-bar) can be approximated:    V (h) 1 0 2 1 V h̄ = h − h2 = N N  1  2 h − h̄2  N −1 (9.266) 9.12 Monte Carlo Methods and the mean squared fractional error associated with the statistical estimate of h̄ is   2   h 1 h2 1 ε2 = − 1  − 1 (9.267) N h2 N − 1 h̄2 Variance Reduction It is clearly important to reduce the mean-squared error in order to increase confidence in the Monte Carlo calculation of the mean value of a quantity h(x) based on a random sampling of the variable x. From Eq. (9.267), this can be accomplished by just running more histories, but that involves longer computational times. There are other methods of reducing the mean-squared error, or the related variance. We now discuss a number of such variance reduction methods. The basic idea of importance sampling is to select from a modified distribution function that yields the same mean value but a smaller variance. Suppose that instead of evaluating h and the statistical estimate from Eqs. (9.260) and (9.262), we evaluate them from  b f (x) dx h(x) ∗ f ∗ (x) (9.268) h2 = f (x) a h̄2 = N N 1  f (xn ) 1  h(xn ) ∗ h(xn )w(xn ) ≡ N f (xn ) N n=1 (9.269) n=1 where the values of xn are now selected from the distribution f ∗ (x) according to the procedures described previously. The quantity w(xn ) ≡ f (xn )/f ∗ (xn ) is known as a weight function. Obviously, the mean value h1 computed from Eq. (9.260) and the mean value h2 computed from Eq. (9.268) are the same. The statistical estimate h̄2 of Eq. (9.269) and the statistical estimate h̄1 of Eq. (9.262) both have the expectation value h . However, the variances are different, and this is the point. The variances computed by the two sampling procedures are  2 1 b  V1 (h̄1 ) = dx h(x) − h f (x) N a   b 1 (9.270) = dx h2 (x)f (x) − h 2 N a 2   1 b h(x)f (x) V2 (h̄2 ) = − h dx f ∗ (x) N a f ∗ (x)   b 1 h2 (x)f 2 (x) 2 (9.271) = dx − h N a f ∗ (x) The objective is to choose f ∗ (x) so that V2 < V1 . If the distribution h(x) and its expectation value were known, the optimum choice of f ∗ (x) would be f ∗ (x) = h(x)f (x) x (9.272) 375 376 9 Neutron Transport Theory for which V2 = 0. This suggests that a good estimate of f ∗ (x) could reduce the variance significantly. The function f ∗ (x) should be chosen to emphasize those neutrons which in some sense are the most important to the quantity that is being estimated, h , which suggests that it is an importance or adjoint function (Chapter 13). However, the variance reduction techniques which are in common use are schemes for emphasizing neutrons which are most likely to contribute to the tally for the quantity of interest, h , based on experience and intuition. Nonanalog variance reduction schemes are implemented by adjusting the neutron weight at each event in its history. An event may be a collision, crossing a boundary into a different region, and so on. An exponential transformation is useful in penetration problems to increase the number of neutron histories which penetrate deeply to contribute to the event of interest (e.g., penetration of a shield, penetration into a control rod). If the event of interest depends primarily on neutrons moving in the positive x-direction, the cross section can be artificially reduced in the x-direction to enhance penetration: tex = t (1 − p · nx ) (9.273) where 0 ≤ p ≤ 1. At a collision, the particle weight must be multiplied by a weight wex to preserve the expected weight of the collided neutron; that is, ex t e−t s ds = wex tex e−t s ds (9.274) must be satisfied, which defines the weight wex = exp[−p( · nx )s] 1 − p( · nx ) (9.275) When a reaction rate is to be calculated over a small volume in which the collision probability is small, the artifice of forced collisions is useful. A neutron entering the volume with weight w which would have to travel a distance l to cross the volume is split into two neutrons, the first of which passes through the volume without collision and the second of which is forced to collide within the volume. Since the actual probability for the particle to cross the region without collision is exp(−t l), the collided and uncollided neutrons must be given weights wc = w[1−exp(−t l)] and wun = w exp(−t l), respectively. The history of the uncollided particle with weight wun is restarted on the exiting surface of the volume. A new history is started for the collided particle. The pdf for collision of this second particle within the volume is f (s) = t e−t s 1 − e−t l (9.276) Generating a random number ξ (0 ≤ ξ ≤ 1), the distance into the volume at which the collision takes place is selected: s=−   1 ln 1 − ξ 1 − e−t l t (9.277) and the subsequent history of the collided particle with weight wc is followed. 9.12 Monte Carlo Methods In some problems, the penetration of a neutron to a particular region may be of interest, and absorption in other regions may unduly reduce the number of neutrons that survive to do so. Absorption weighting can be used as an alternative to terminating a history by an absorption event. In a collision all outcomes are treated as scattering events, but the emerging neutron is given a weight   a wa = w 1 − t (9.278) to preserve the survival probability. Since continuing the computation of histories of neutrons with small weights is inefficient, Russian roulette can be used to either increase the neutron weight or terminate the history. A random number ξ (0 ≤ ξ ≤ 1) is generated and compared with an input number v typically between 2 and 10. If ξ > 1/ν, the history is terminated; if ξ < 1/ν, the history is continued with original neutron weight w increased to wRR = wν. Splitting can be used to increase the number of histories that penetrate in deep penetration problems. When a neutron with weight w crosses a fixed surface in the direction of penetration from a region with importance Ii into a region with importance Ii+1 , the history is terminated and Ii+1 /Ii new histories are started for neutrons with the same energy and direction and weights ws = wIi /Ii+1 . Here importance refers to importance with respect to the quantity of interest h . Russian roulette can be used in conjunction with splitting to terminate histories of particles with low weights moving across the surfaces away from the direction of penetration. Tallying The calculation of reaction rates in various regions, over various energies, and by various nuclides is accomplished straightforwardly by tallying each collision event. Neutron fluxes and currents can also be constructed by tallying events and surface crossings. By definition, the collision rate in a region is equal to the product of the cross section times the flux times the volume. Thus, by tallying the collision rate (CR), the flux can be calculated from φ= CR t V (9.279) A shortcoming of this algorithm is that only particles which collide within the volume V will contribute to CR, hence to φ. Another definition of the scalar flux is the path length traversed by all particles passing through a volume per unit volume per unit time: φ̄ = N 1 1  l¯ = ln V V N n=1 (9.280) 377 378 9 Neutron Transport Theory where l¯ is the track length per unit time in the volume in question of the nth history. Taking into account the weights of neutrons at various stages of their histories, this definition of flux becomes φ̄ = 1 1  wn ln V N m (9.281) where wn is the weight the neutron on the nth history had when it traversed the volume (note that a neutron history may traverse a given volume more than once, and it should be tallied each time). The variance in the flux estimate is given by Var %N &2   n  N 1 1  2 = (wn ln ) − 2 2 wn ln N − 1 V 2N V N n=1 (9.282) n=1 Currents across surfaces are also of interest. It is straightforward to tally the rate at which particles are passing through a given surface in the positive and negative directions, pn± for history n. The total number of particles per unit time passing through the surface in the positive and negative directions can then be estimated from p± = N 1  wn pn± N (9.283) n=1 Here wn is the weight that the neutron in the nth history had when it crossed through the surface to contribute to the tally (note that a neutron in a given history may cross through a surface more than once, and it should be tallied each time). The partial currents are obtained by dividing by the surface area A, and the net current is obtained by subtracting the partial currents: J = J+ − J− = N  1   + p+ − p− = wn pn − pn− A AN (9.284) n=1 If the Monte Carlo calculation is to be used to determine small differences, such as reactivity worths of perturbations, or reactivity coefficients, special methods must be used to avoid the small difference in two calculations being masked by statistical errors. The method of correlated sampling addresses this problem by using the same sequence of random numbers to generate the sequence of events that describes the histories in the two problems. If the system is unchanged, the two calculations must yield identical results. So any difference in results is due to the perturbation. Criticality Problems Monte Carlo can be used to calculate the multiplication constant and associated eigensolution for the flux distribution. The problem is started with an arbitrary spatial distribution of neutrons distributed in the fission energy spectrum and 9.12 Monte Carlo Methods isotropically in direction. This initial spatial distribution can be uniform or a spatial distribution that is the result of a previous Monte Carlo calculation for a similar problem or of a deterministic transport (e.g., discrete ordinates) solution for the problem at hand. The history of a large number of neutrons in a given generation is followed in parallel to termination, thus obtaining a new fission distribution for the next generation of neutrons, and the process is repeated until the fission neutron spatial source distribution has settled down. The total number of neutrons in successive generations may be increased during the settling down period to obtain greater detail only after the solution has settled. Once the spatial fission neutron source distribution has settled down, the ratio of the total number of fission neutrons on successive generations is the statistical estimate of the multiplication constant. The computational effort in the period before the distribution settles down can be reduced by a number of techniques. The fission source distribution is determined from generation to generation as follows. If wn is the weight of the nth history neutron when it has an absorption event that terminates the history, then either In or In + 1 fission neutrons are produced at that location in the next generation. The selection is made by writing wn νf = In + Rn a (9.285) where In is an integer and 0 < Rn < 1. A random number ξ (0 ≤ ξ ≤ 1) is generated. If Rn > ξ , then In +1 neutrons are launched in the next generation; otherwise, In neutrons are launched. Track lengths can provide a second estimate of the total number of fission neutrons produced by history n:  wni ln i νf i = total number of secondaries produced by history n (9.286) ai where wni is the neutron weight as it crosses region i and lni is the total track length across region i. One of the problems in criticality calculations is to prevent the total neutron population from increasing or decreasing too much, which it will do if the assembly is supercritical or subcritical, respectively. One technique is to change the neutron weight at each collision by multiplying the previous weight by the expected number of secondary neutrons. A second method is simply to start off each generation with the same number of neutrons by eliminating some of the next-generation neutrons if there are more neutrons than in the previous generation or using some of the neutrons twice if there are less neutrons than in the previous generation. Source Problems A number of reactor physics problems can be formulated as source problems. The most obvious is the shielding problem, where the reactor core can be considered as a fixed neutron source. The calculation of resonance absorption of neutrons from a slowing-down source in a heterogeneous lattice, the thermalization of neutrons 379 380 9 Neutron Transport Theory from a slowing-down source into the thermal range, and the calculation of temperature coefficients of reactivity from a fixed fission source in a heterogeneous lattice are other problems which are treated as source problems. The resonance cross sections in the resolved region can be represented by values at a very large number of energy points or calculated from the Doppler-broadened Breit–Wigner formula, and the resonance cross sections in the unresolved region can be selected from a pdf based on the statistics of the nuclear level spacing and width (Chapter 11). The neutron slowing down through the resonance region is then treated by sampling the uniform distribution of neutrons scattered at energy E over the interval E to αE, sampling the path length distribution to determine the point of collision, sampling the reaction-type distribution to determine whether the collision is absorption or scattering, and so on. Effective Doppler-broadened cross sections at different temperatures can be used in conjunction with correlated sampling to compute temperature coefficients of reactivity. A source distribution in energy of neutrons slowing down into the thermal range in the moderator can be used to launch neutrons isotropically in the thermal energy region. The distribution of rotational–vibrational levels (Chapter 12) which affect inelastic scattering of neutrons from bound atoms and molecules can be used to construct pdf’s for inelastic scattering. Then the histories of thermal neutrons can be traced until termination by absorption. Path length estimators at different energies can be used to estimate the thermal flux spectrum. Random Numbers Generation of random numbers is essential to a Monte Carlo calculation. There exist a number of random number generators—algorithms for generating random numbers—and there is a great deal of controversy about just how random they are. A discussion of random number generators and several FORTRAN routines for generating random numbers are given in Ref. 1. References 1 W. H. Press et al., Numerical Recipes, Cambridge University Press, Cambridge (1989), Chap. 7. 2 E. E. Lewis and W. F. Miller, Computational Methods of Neutron Transport, Wiley-Interscience, New York (1984); reprinted by American Nuclear Society, La Grange Park, IL (1993). 3 R. J. J. Stamm’ler and M. J. Abbate, Methods of Steady-State Reactor Physics in Nuclear Design, Academic Press, London (1983), Chaps. IV and V. 4 S. O. Lindahl and Z. Weiss, “The Response Matrix Method,” in J. Lewins and M. Becker, eds., Adv. Nucl. Sci. Technol. 13 (1981). 5 B. G. Carlson and K. D. Lathrop, “Transport Theory: The Method of Discrete Ordinates,” in H. Greenspan, C. N. Kelber, and D. Okrent, eds., Computing Methods in Reactor Physics, Gordon and Breach, New York (1968). 6 E. M. Gelbard, “Spherical Harmonics Methods: PL and Double-PL Approximations,” in H. Greenspan, C. N. Kelber, and D. Okrent, eds., Computing Methods in Reactor Physics, Gordon and Breach, New York (1968). Problems 7 M. H. Kalos, F. R. Nakache, and J. Celnik, “Monte Carlo Methods in Reactor Computations,” in H. Greenspan, C. N. Kelber, and D. Okrent, eds., Computing Methods in Reactor Physics, Gordon and Breach, New York (1968). 8 J. Spanier and E. M. Gelbard, Monte Carlo Principles and Neutron Transport Problems, Addison-Wesley, Reading, MA (1964). 9 M. Clark and K. F. Hansen, Numerical Methods of Reactor Analysis, Academic Press, New York (1964). 10 R. V. Meghreblian and D. K. Holmes, Reactor Analysis, McGrawHill, New York, (1960), pp. 160–267 and 626–747. 11 B. Davison, Neutron Transport Theory, Oxford University Press, London (1957). 12 K. M. Case, F. de Hoffmann, and G. Placzek, Introduction to the Theory of Neutron Diffusion, Los Alamos National Laboratory, Los Alamos, NM (1953). 13 A. F. Henry, Nuclear Reactor Analysis, MIT Press, Cambridge, MA (1975), Chap. 6. 14 R. Sanchez, “Approximate Solution of the Two-Dimensional Integral Transport Equation by Collision Probabilities Methods,” Nucl. Sci. Engr. 64, 384 (1977); “A Transport Multicell Method for Two-Dimensional Lattices of Hexagonal Cells,” Nucl. Sci. Engr. 92, 247 (1986). Problems 9.1. Rederive the transmission and absorption probabilities for a purely absorbing slab given by Eq. (9.28) for the situation in which the incident flux is linearly anisotropic (i.e., ∼ μ). 9.2. Use the orthogonality relation for Legendre polynomials to derive the orthogonality relation for half-angle polynomials given by Eq. (9.23). 9.3. Carry through the indicated steps to derive the integral equation (9.42). 9.4. Develop analytical expressions for the two-dimensional transmission and escape probabilities of Eqs. (9.110) and (9.112) for rectangular geometry with dimension ax and ay on a side. Evaluate these transmission and first-flight escape probabilities for X = 4V /Sλ varying over the range 0.1 < X < 10.0. 9.5. Evaluate the first-flight escape probabilities given by Eq. (9.117) with c = 2.09 for X = 4V /Sλ varying over the range 0.1 < X < 10.0 and compare with the results of Problem 9.4. 9.6. Carry through the indicated steps to derive the PL equations (9.126). 9.7. Derive the simplified P3 equations (9.143) from the P3 equations and derive the boundary conditions of Eqs. (9.145). 9.8. Demonstrate that when the ordinates and weights given by Eqs. (9.202) and (9.203) are used, the discrete ordinates 381 382 9 Neutron Transport Theory 9.9. 9.10. 9.11. 9.12. 9.13. 9.14. 9.15. 9.16. 9.17. 9.18. 9.19. equations with N ordinates reduce identically to the D-PN−1 equations. Write a code to solve the one-dimensional discrete ordinates equations in slab geometry. Solve for the flux in the S2 approximation in a uniform slab 100 cm thick with vacuum boundary conditions, with t = 0.25 cm−1 , s0 = 0.15 cm−1 , and an isotropic source S0 = 1014 n/cm·s distributed over 0 < x < 25 cm. Repeat for the S4 , S8 , and S12 approximations. Repeat Problem 9.9 including anisotropic scattering s1 = 0.01 and s2 = 0.0025. Derive the spatial difference equations for the one-dimensional discrete ordinates equations in spherical geometry. Reconcile your results with the algorithms of Eqs. (9.219) and (9.223). Write a code to solve the SN equations in two-dimensional x–y geometry. Solve for the flux in the S2 approximation in a uniform square 100 cm on a side with vacuum boundary conditions, with t = 0.25 cm−1 , s0 = 0.15 cm−1 , s1 = 0.01, and s2 = 0.0025, and an isotropic source S0 = 1014 n/cm2 ·s distributed over 0 < x < 25 cm, 25 < y < 50 cm. Repeat for the S4 , S8 , and S12 approximations. The pdf for variable x is f (x) = 4/π(1 + x 2 ) with 0 ≤ x ≤ 1. Show that if a random number ξ (0 ≤ ξ ≤ 1) is generated, the corresponding value of x = tan(ξ π/4). Derive the simplified P5 diffusion equations and associated Marshak boundary conditions from the P5 equations. (Hint: Use F0 = 2φ2 + φ0 , F1 = 43 φ4 + φ2 , F2 = φ4 .) Derive the diffusion theory equation (9.158) from the one-dimensional Pl equations in cylindrical and spherical geometries. Derive the spherical harmonics approximation to the neutron transport equation in three-dimensional x–y–z geometry given by Eqs. (9.181). Plot the cumulative distribution function corresponding to the fission spectrum given approximately by √ χ(E) = 0.453 exp(−1.036E) sinh 2.29E over the energy range 104 eV ≤ E ≤ 107 eV. Calculate the maximum spatial mesh size that could be used in a one-dimensional S2 calculation for a problem with t = 0.3 cm−1 . Repeat for the S4 and S8 approximations. Plot the pdf and cdf for the cross-section distribution in a region with a = 0.15 cm−1 , and s = 0.08 cm−1 , and f = 0.08 cm−1 . Problems 9.20. Write a Monte Carlo code to calculate the multiplication constant and flux distribution for one-speed neutrons in a slab reactor of thickness a = 1.0 m with isotropic scattering for which (a = 0.12 cm−1 , s = 0.05 cm−1 , νf = 0.15 cm−1 ) over 0 < x < 50 cm and (a = 0.10 cm−1 , s = 0.05 cm−1 , νf = 0.12) over 50 < x < 100 cm. 9.21. A S(μ) ∼ μ2 neutron source is present on the left face of a slab of thickness a with absorption cross section a and isotropic scattering cross section s . Derive expressions for the uncollided and total neutron currents exiting from the right surface of the slab. 9.22. Derive the Boltzmann transport equation from particle balance considerations on a differential element of space–angle–energy phase space. Justify any assumptions. 383 385 10 Neutron Slowing Down The methods used to calculate the slowing down of fast neutrons above the thermal energy range are treated in this chapter. We also introduce the lethargy as an alternative to the energy variable and develop the formalism in terms of lethargy. 10.1 Elastic Scattering Transfer Function Lethargy It is convenient in treating neutron slowing down to replace the energy variable with the neutron lethargy u = ln E0 E (10.1) where E0 is the maximum energy that a neutron might have in a nuclear reactor, say 10 MeV. The incremental lethargy interval, du, corresponding to the incremental energy interval, dE, is du = du dE dE = − dE E (10.2) with the minus sign indicating that as the neutron energy decreases, its lethargy increases—hence the name. The fact that the total neutron flux in an incremental lethargy interval physically is the same as the neutron flux in the corresponding incremental energy interval provides a correspondence between the flux per unit energy, φ(E), and the flux per unit lethargy, φ(u): φ(u) du = −φ(E) dE ⇒ φ(u) = Eφ(E) (10.3) Elastic Scattering Kinematics The principal results obtained in Chapter 2 from the conservation of energy and momentum in an elastic scattering event were the correlation between the energy 386 10 Neutron Slowing Down Fig. 10.1 Angles involved in a scattering event. (From Ref. 2; used with permission of MIT Press.) change E  → E and the cosine of the scattering angle in the center-of-mass (CM) system μC = cos θC : A2 + 1 + 2Aμc E  = = eu −u ≡ e−U E (A + 1)2 (10.4) and the relation between the cosine of the scattering angle in the lab system, μ0 = cos θ0 , and the cosine of the scattering angle in the CM system, μ0 = 1 + Aμc (1 + 2Aμc + A2 )1/2 (10.5) which may be combined to express the correlation between the scattering angle in the lab system and the change in lethargy U = u − u : μ0 (U ) = 1 (A + 1)e−(1/2)U − (A − 1)e(1/2)U 2 (10.6) Elastic Scattering Kernel The general lethargy-angle scattering transfer function can be written σs (μ0 , u → u) = σs (u )p0 (u , μ0 )g(μ0 , u → u) (10.7) where μ0 =  ·  is the cosine of the angle in the lab system between the incident and exit directions of a neutron in a scattering collision, as shown in Fig. 10.1, p0 (u , μ0 ) is the probability that a neutron of lethargy u will scatter through an 10.1 Elastic Scattering Transfer Function angle θ0 = cos−1 μ0 and g(μ0 , u → u) is the probability that a neutron of lethargy u which scatters through an angle θ0 = cos−1 μ0 will have a final lethargy u. With the normalization  (10.8) g(μ0 , u → u)du = 1 the angular transfer function for scattering through an angle θ0 = cos−1 μ0 is σs (μ0 , u ) ≡  σs (μ0 , u → u)du = σs (u )p0 (u , μ0 ) (10.9) Writing the lethargy-angle transfer function as a function of (u , U = u − u , μ0 ) and expanding in Legendre polynomials yields σs (μ0 , U, u ) = ∞  1 l  =0 2 (2l  + 1)bl0 (u , U )Pl  (μ0 ) (10.10) where Pl (μ0 ) is the lth Legendre polynomial of the argument of the cosine of the scattering angle in the lab system, and the orthogonality properties of the Legendre polynomials can be used to identify the Legendre coefficients of the scattering transfer function: bl0 (u , U ) =  1 −1 dμ0 Pl (μ0 )σs (μ0 , U, u ) (10.11) For elastic scattering, there is a strict lethargy-angle correlation given by Eq. (10.6), which means that the probability for a scattering collision that produces a lethargy gain within dU about U is equal to the probability for scattering with a cosine of the scattering angle within dμ0 about μ0 when U and μ0 are related by Eq. (10.6) and is zero otherwise:   σs (μ0 , U, u )dU = −σs (μ0 , u )δ μ0 − μ0 (U ) dμ0 (U ) (10.12) where the minus sign reflects the fact that an increase in the cosine of the scattering angle corresponds to a decrease in the lethargy gain. Using Eq. (10.12) in Eq. (10.11) yields    dμ0 (U )  bl0 (u , U ) = σs (μ0 , u )Pl μ0 (U ) − dU (10.13) Making use of the physical fact that the probabilities for scattering through a given scattering angle in the lab system to within dμ0 about μ0 and for scattering through the corresponding [via Eq. (10.5)] scattering angle in the CM system to within dμC about μC must be equal: σs (μ0 , u )dμ0 = σsc (μc , u )dμc (10.14) 387 388 10 Neutron Slowing Down and making use of the observation that the experimental scattering data are well represented by a Legendre expansion in the cosine of the CM scattering angle θC = cos−1 μC : σ (μ0 , u ) = σsc (μc , u ) ∞ dμc  1  dμc (2l + 1)blc (u )Pl  (μc ) = dμ0 2 dμ0  (10.15) l =0 allows the Legendre moments of the lethargy gain, bl0 (u , U ), to be related to the Legendre moments of the angular scattering distribution in the CM system, blc (u ), which are tabulated in the nuclear data files:   ∞   dμ0 (U ) 1  dμc  0  c  Pl μ0 (U ) − (2l + 1)bl  (u )Pl  (μc ) bl (u , U ) = 2 dμ0 dU  l =0 = ∞  1 l  =0 ≡ ∞  l  =0 2 (2l  + 1)blc (u )Pl       dμc (U )  μc (U ) Pl μ0 (U ) − dU Tll  (U )blc (u ) (10.16) Using this result in Eq. (10.10) leads to σs (μ0 , U, u ) = 1 l,l  2 (2l + 1)Tll  (U )Pl (μ0 )blc (u ) (10.17) for the elastic scattering lethargy-angle transfer function. Integrating this result over angle yields the total probability for an elastic scattering event to cause a lethargy increase from u to u:   σs (u → u) = 1 −1 dμ0 σs (μ0 , U, u ) = ∞  l  =0 T0l  (U )blc (u ) (10.18) Isotropic Scattering in Center-of-Mass System The angular distribution of elastic scattering in the CM system may be represented by an average value of the cosine of the CM scattering angle given by μc = 0.07A2/3 E (MeV), except near scattering resonances. Hence the elastic scattering distribution is essentially isotropic in the CM system, except for high-energy neutrons scattering from heavy mass nuclei. When the scattering is taken as spherically symmetric in the CM system, the Legendre moments of the angular scattering distribution in the CM system are blc (u ) = σs (u )δl0 (10.19) In this case, Eq. (10.18) becomes σsiso (u  σs (u )eu −u → u) = σs (u )T00 (U ) = 1−α  (10.20) 10.1 Elastic Scattering Transfer Function The average lethargy increase with isotropic scattering is  ξ iso ln 1/α = 0 σs (u → u)U dU = σs (u )  ln 1/α T00 (U )U du = 1 + 0 α ln α 1−α (10.21) 2 3A (10.22) and the average cosine of the scattering angle in the lab system is  μ̄iso 0   σs (u → u)μ0 (U ) dU = σs (u ) ln 1/α = 0  ln 1/α T10 (U ) dU = 0 where A is the atomic mass in amu of the scattering nuclei and α = [(A − 1)/(A + 1)]2 . Both of these quantities are independent of lethargy for a given species of scattering nuclei. However, the composite values for a mixture, ξ =     j σsj (u)ξi / j σsj (u) and μ0 = j σsj (u)μ0j / j σsj (u), may be lethargy dependent. Linearly Anisotropic Scattering in Center-of-Mass System When only the first two Legendre components of the scattering transfer function in the center of mass system are non-zero, Eq. (10.18) becomes σsanis (u → u) = T00 b0c (u ) + T01 (U )b1c (u )      6  σs (u )eu −u  1 + μ̄c (u ) 3 − 1 − e−u −u = 1−α 1−α (U ) (10.23) In this case, the mean lethargy increase in an elastic scattering event, ξ(u ) = b0c (u ) σs (u ) =ξ iso ln 1/α dU U T00 (U ) + 0 b1c (u ) σs (u )  ln 1/α dU U T01 (U ) 0   b1c (u ) 1 A2 + 1 1 (A2 − 1)2 A − 1 + −3 c  ln b0 (u ) 4 A 8 A+1 A2 −→ ξ iso − large A  2 b1c (u ) = ξ iso 1 − μ̄c (u ) A b0c (u ) (10.24) is reduced by anisotropic scattering (i.e., the moderation in energy is reduced), and the average cosine of the scattering angle in the lab system,  b1c (u ) dU μ0 (U ) T00 (U ) + c  T01 (U ) b0 (u ) 0   3 iso + μ̄c (u ) 1 − = μ̄0 (u ) 5A2 bc (u ) μ̄0 (u ) = 0  σs (u )   ln 1/α  (10.25) is increased by anisotropic scattering (i.e., the scattering is more forward directed). Both ξ and μ0 become lethargy dependent with anisotropic scattering. 389 390 10 Neutron Slowing Down 10.2 P1 and B1 Slowing-Down Equations Derivation The transport equation of Chapter 9 can immediately be generalized to include lethargy dependence by allowing for the scattering removal of neutrons from incremental interval du and for a scattering source of neutrons into du from other incremental intervals du (in the slowing-down region above 1 eV, the in-scatter would only be from u ≤ u):  · ∇ψ(r, , u) + t (r, u)ψ(r, , u)  4π  u s (r, μ0 , U, u ) = du d ψ(r,  , u ) 2π 0 0  ∞  4π νf (r, u ) 1 + χ(u) ψ(r,  , u ) du d k 4π 0 0  4π  u sel (r, μ0 , U, u ) ≡ du d ψ(r,  , u ) + S(r, , u) (10.26) 2π u−ln 1/α 0 where μ0 =  ·  is the cosine of the angle in the lab system between the incident and exit directions of a neutron in a scattering collision. In the last step, inelastically scattered and fission neutrons are grouped into a source term and the remaining scattering term includes only elastic scattering. The macroscopic elastic scattering transfer function is a sum over nuclear species of the density times the microscopic transfer function of Eq. (10.17):  Nj (r)σsj (μ0 , U, u ) sel (r, μ0 , U, u ) = j =  j Nj (r) 1 l,l  2 (2l + 1)Tll  (U )Pl (μ0 )blc (u ) (10.27) and the lower limit of the in-scatter integral for each species is 1 − ln(1/αj ), but this is represented symbolically for notational convenience as a single 1 − ln(1/α). The Pn equations were derived in Chapter 9 for one-dimensional geometry and one-speed neutrons by expanding the directional flux in a Legendre polynomial series, and this can immediately be generalized to the lethargy-dependent neutron flux ψ(z, μ, u)  12 φ0 (z, u)P0 (μ) + 32 φ1 (z, μ)P1 (μ) = 12 φ0 (z, u) + 32 μJz (z, u) = 12 φ0 (z, u) + 32 z Jz (z, u) (10.28) where μ =  · nz = z = cos θ is the cosine of the angle made by the direction of neutron motion with the z-coordinate axis, as indicated in Fig. 10.2, and where the 10.2 P1 and B1 Slowing-Down Equations Fig. 10.2 Specification of the directional vector  in a Cartesian coordinate system. (From Ref. 2; used with permission of MIT Press.) current Jz has been associated with the n = 1 component of the flux expansion by using the orthogonality properties of the Legendre polynomials:  1 P1 (μ)ψ(z, μ, u) dμ (10.29) Jz (z, u) ≡ φ1 (z, u) ≡ −1 Further defining, x =  · nx = sin θ cos ϕ, y =  · ny = sin θ sin ϕ, and d = sin θ dθdϕ/4π , the P1 expansion of the directional neutron flux in three-dimensional geometry is ψ(r, , u)  12 φ0 (r, u) + 32  · J(r, u) (10.30) In developing the P1 equations in one dimension (Chapter 9), the expansion of Eq. (10.28) was substituted into the transport equation, and the resulting equation was weighted with P0 = 1 and integrated over μ, and then weighted with P1 (μ = z ) = μ and integrated over μ, to obtain the two P1 equations. We generalize this procedure to three dimensions by substituting Eq. (10.30) into Eq. (10.26) and weighting with 1, x = μx , y = μy , and z = μz , that is, weight with 1 and , and integrating over  to obtain the P1 equations in three-dimensional geometry:       4π 1 1 3 3 d  · ∇ φ0 +  · J + t φ0 +  · J 2 2 2 2 0    4π  u  4π sel (r, μ0 , U, u ) 1 3 d du d φ0 (u ) +  · J(u ) = 2π 2 2 0 0 0 + S0 (r, u) (10.31) 391 392 10 Neutron Slowing Down       1 1 3 3 d   · ∇ φ0 +  · J + t φ0 +  · J 2 2 2 2 0    4π  u  4π  3   sel (r, μ0 , U, u ) 1   φ0 (u ) +  · J(u ) d du d  = 2π 2 2 0 0 0 4π + S1 (r, u) (10.32) where  S0 (r, u) ≡ 4π d S(r, , u) 0  S1 (r, u) ≡ 4π (10.33) d S(r, , u) 0 To simplify these P1 equations, Eq. (10.27) is used for the scattering transfer function, and the addition theorem for Legendre polynomials, Pn (μ0 ) = Pn (μ )Pn (μ) + 2 n  (n − m)! m  m P (μ )Pn (μ) cos m(ϕ  − ϕ) (10.34) (n + m)! n m=1 is used to relate the cosine of the scattering angle μ0 = cos θ0 to the cosines of the angles that the incident and exiting neutron directions make with the z-axis, μ = cos θ  and μ = cos θ , respectively, as depicted in Fig. 10.1, where Pnm is the associated Legendre function. Using the identities   1 d = 1, ξ χ d = δξ χ 3 (10.35)   3 ξ d = ξ d = 0, ξ = x, y, z Eqs. (10.32) and (10.33) then can be reduced to  u du s0 (r, U, u )φ(r, u ) + S0 (r, u) ∇ · J(r, u) + t (r, u)φ(r, u) = u−ln 1/α 1 ∇φ(r, u) + t (r, u)J(r, u) = 3  u u−ln 1/α (10.36)    du s1 (r, U, u )J(r, u ) + S1 (r, u) (10.37) where the zero subscript on the flux has been dropped and the Legendre moments of the elastic scattering transfer functions are defined:  1 sn (r, U, u ) ≡ dμ0 sel (r, μ0 , U, u )Pn (μ0 ) −1 =  j ≡  j Nj (r)  l  =0 Tnl  (U )blc (u ) Nj (r)σsn (u → u) j (10.38) 10.2 P1 and B1 Slowing-Down Equations In particular, the isotropic and linearly anisotropic lethargy change transfer functions are   1  j  eu −u , u − ln < u < u σs (u ) j  (10.39) σs0 (u → u) = 1 − αj αj  0, otherwise σs1 (u → u)     j   σs (u )eu −u A + 1 (1/2)(u −u) A − 1 −(1/2)(u −u) , − e e = 1 − αj 2 2   0, j 1 < u < u αj otherwise (10.40) u − ln The essential approximation that has been made in deriving Eqs. (10.36) and (10.37) is that the angular dependence of the neutron flux can be represented by only a linearly anisotropic dependence on the angular variable, as given by Eq. (10.30). This approximation should be good at more than a few mean free paths away from an interface between very dissimilar media (i.e., in the interior of large homogeneous regions) and more than a few mean free paths away from an anisotropic source. Solution in Finite Uniform Medium To solve Eqs. (10.36) and (10.37), it is assumed that the medium is uniform and that the spatial dependence of the flux and the current can both be represented by a simple buckling mode [i.e., φ(z, u) = φ(u) exp(iBz), J (z, u) = J (u) exp(iBz)], so that these equations become  u iBJ (u) + t (u)φ(u) = du s0 (U, u )φ(u ) + S0 (u) (10.41) u−ln 1/α 1 iBφ(u) + t (u)J (u) = 3  u u−ln 1/α du s1 (U, u )J (u ) + S1 (u) (10.42) The parameter B may be considered to characterize the leakage from or into the medium. Note that this procedure is formally equivalent to Fourier transforming Eqs. (10.36) and (10.37). These equations may be put in multigroup form by integrating over ug = ug − ug−1 and defining  ug  ug du φ(u), Jg ≡ du J (u) φg = ug−1 g t ≡ 1 ug g  →g sn  ug−1 ug ug−1 1 ≡ ug  g Sn ≡ du t ,  ug  ug  −1   ug du ug−1  ug ug−1 (10.43) du Sn (u) du sn (u → u), n = 0, 1 393 394 10 Neutron Slowing Down Here we have used the asymptotic flux solution φ(u) ∼ 1, corresponding to φ(E) ∼ 1/E, and assumed that J (u) ∼ 1, also, in evaluating the total and scattering cross sections. The multigroup form of the P1 equations is g iBJg + t φg =  g  ≤g g  →g s0 g φg  + S0 ,  g  →g 1 g g iBφg + t Jg = s1 Jg  + S1 , 3  g = 1, . . . , G (10.44) g = 1, . . . , G (10.45) g ≤g B1 Equations The principal approximation involved in derivation of the P1 equations is the assumption of linear anisotropy in the angular dependence of the neutron flux made in Eq. (10.28) or (10.30). There is an alternative formulation that avoids this approximation but instead makes the approximation that the angular dependence of the scattering can be represented by an isotropic plus a linearly anisotropic scattering transfer function. Returning to Eq. (10.26), but simplified to one-dimensional geometry, μ ∂ψ(z, μ, u) + t (z, u)ψ(z, μ, u) ∂z  u  1  = dμ du s (z, μ0 , U, u )ψ(z, μ , u ) + S(z, μ, u) −1 (10.46) u−ln 1/α and making the same type of assumption about the spatial dependence [i.e., ψ(z, μ, u) = ψ(μ, u) exp(iBz)] in a uniform medium leads to [t (u) + iBμ]ψ(μ, u)  u  1 = du du s (μ0 , U, u )ψ(μ , u ) + S(μ, u) −1 (10.47) u−ln 1/α Dividing by (t + iBμ) and assuming linearly anisotropic scattering yields   u 1 du s0 (u → u)φ(u ) 2 u−ln 1/α  du s1 (u → u)J (u ) + S(μ, u) ψ(μ, u) = (t (u) + iBμ)−1 3 + μ 2  u u−ln 1/α (10.48) The approximation of Eq. (10.28) has not been made in deriving this result; the quantities φ and J have been identified from the definitions  φ(u) ≡  1 −1 dμ ψ(μ, u), J (u) ≡ 1 −1 dμ μψ(μ, u) (10.49) 10.2 P1 and B1 Slowing-Down Equations Now, Eq. (10.48) is multiplied by 1 and by μ and integrated over μ to obtain the two B1 equations  iBJ (u) + t (u)φ(u) = u u−ln 1/α 1 iBφ(u) + γ (u)t (u)J (u) = 3  du s0 (u → u)φ(u ) + S0 (u) u u−ln 1/α (10.50)    du s1 (u → u)J (u ) + S1 (u) where γ (u) = 4 B [B/t (u)]2 tan−1 [B/t (u)] 1+ 15 t (u) 3{B/t (u) − tan−1 [B/t (u)]} (10.51) The B1 equations differ from the P1 equations [Eqs. (10.44) and (10.45)] only by the factor γ . The essential B1 approximation is a linearly anisotropic scattering transfer function; the essential P1 approximation is linearly anisotropic neutron flux. The B1 equations have been found to be somewhat more accurate for slab geometries, but clearly the two approximations will differ only when B is significant. The multigroup P1 and B1 equations are the basis of most multigroup fast spectrum codes (e.g., Refs. 4 and 10). Typical neutron energy distributions calculated for thermal (PWR) and fast (LMFBR) reactors are shown in Fig. 10.3. Few-Group Constants The usual procedure in reactor analysis is to solve the multigroup equations (with a large number of groups varying from 50 to 100 for thermal reactors to a few 1000 for fast reactors) for one or more large homogenized regions and then to develop few-group (2 to 4 for water-moderated thermal reactors, 5 to 10 for graphitemoderated thermal reactors, 20 to 30 for fast reactors) constants which can be used in a few-group diffusion theory calculation of the neutron diffusion during the slowing-down process. The few-group constants are constructed by using the finegroup fluxes to weight the fine-group constants over the fine groups contained within a few group. Denoting the fine groups with a g and the few groups with a k, the prescriptions for the few-group capture and fission cross sections  g∈k σ σ =  k gφ g (10.52) g∈k φg and scattering transfer cross sections k  →k σsn  = g  ∈k   g  →g g∈k  σsn g  ∈k  φg   φg  (10.53) follow directly, where g ∈ k indicates a sum over fine groups g within the lethargy interval of few group k. 395 396 10 Neutron Slowing Down Fig. 10.3 Representative neutron energy distributions in a PWR and a LMFBR. (From Ref. 11; used with permission of Taylor & Francis.) There is ambiguity about the definition of the few-group diffusion coefficient, as discussed in the following section. An appropriate definition is in terms of a few-group directional transport coefficient, defined as   k σtr,ξ = g∈k g σt Jgξ −   g g  →g σs1 g∈k Jgξ Jg  ξ  , ξ = x, y, z (10.54) where Jgξ is the fine-group current in the ξ -direction. The diffusion coefficient is g related to the transport coefficient by D g = 1/3tr . Many other prescriptions for the diffusion coefficient are found in practice. 10.3 Diffusion Theory Lethargy-Dependent Diffusion Theory It was shown in Chapter 9 that the one-speed P1 equations led naturally to diffusion theory. Unfortunately, this is not the case for the lethargy-dependent P1 equations of Eqs. (10.36) and (10.37). To derive from Eq. (10.37) a relationship of the form J(r, E) = −D(r, E)∇φ(r, E), it is necessary to require further (1) that s1 (u → u) ∼ s (u )δ(u − u) or zero; and (2) that J(r, E) and ∇φ(r, E) are parallel. Neither of these relations is satisfied in general, which gives rise to a number 10.3 Diffusion Theory of ambiguities in defining the multigroup diffusion constants, in particular the diffusion coefficient. A common way to treat the anisotropic scattering difficulty is to make use of the one-speed results to approximate  u u−ln 1/α du s1 (r, U, u )J(r, u )  s (r, u)μ̄0 J(r, μ) (10.55) which is equivalent to assuming no lethargy change in anisotropic elastic scattering. If, in addition, the anisotropic source that would arise from anisotropic inelastic scattering is assumed to vanish, then J(r, u) = − =− 3(  1  ∇φ(r, u) (r, u) − t s (r, u)μ̄0 ) 1 ∇φ(r, u) = −D(r, u)∇φ(r, u) 3tr (r, u) (10.56) is obtained from Eq. (10.37). This relation, a Fick’s law, can be substituted into Eq. (10.36) to obtain lethargy-dependent diffusion theory: −∇ · D(r, u)∇φ(r, u) + t (r, u)  u du s0 (r, U, u )φ(r, u ) + S0 (r, u) = u−ln 1/α  = u u−ln 1/α du s0 (r, U, u )φ(r, u ) + 1 + χ(u) k  0  ∞ u 0 du in (r, u → u)φ(r, u ) du νf (r, u )φ(r, u ) (10.57) 0 where the inelastic and fission contributions to the isotropic source are shown explicitly in the last form. Directional Diffusion Theory In this derivation of lethargy-dependent diffusion theory from neutron transport theory, the lethargy change (energy change) in anisotropic scattering was neglected entirely. It is possible to formally include anisotropic lethargy change effects by defining tr,ξ (r, u) ≡ t (r, u) − u    u−ln 1/α du s1 (r, U, u )Jξ (r, u ) Jξ (r, u) (10.58) where Jξ is the current in the ξ -direction. Since the lethargy dependence of the current could well be different for different ξ -directions, a different tr,ξ could be defined by Eq. (10.58) for each coordinate direction ξ = x, y, and z, giving rise to directional diffusion coefficients Dξ = 1/3tr,ξ and to a directional diffusion 397 398 10 Neutron Slowing Down equation   ∂ ∂ ∂ ∂ ∂ ∂ Dx (r, u) + Dy (r, u) + Dz (r, u) − ∂x ∂x ∂y ∂y ∂z ∂z × φ(r, u) + t (r, u)φ(r, u)  u  = du s0 (r, U, u )φ(r, u ) + u−ln 1/α 1 + χ(u) k  u 0 ∞ 0 du in (r, u → u)φ(r, u ) du νf (r, u )φ(r, u ) (10.59) 0 Multigroup Diffusion Theory Multigroup diffusion theory can be formally derived from the lethargy-dependent diffusion equations [Eqs. (10.57) or (10.59)] or directly from the lethargy-dependent P1 equations [Eqs. (10.36) and (10.37)] by integrating over the lethargy interval of the group ug = ug − ug−1 . The definition of most of the group quantities is the same for all three procedures and is given by Eqs. (10.43), with fission and absorption cross sections evaluated similarly to the total cross section. However, the definition of the diffusion coefficient is different for the various derivations. In the derivation proceeding from Eq. (10.57), the multigroup diffusion term is formally defined by the replacement  g −∇ · D (r)∇φg (r) ≡ − ug ug−1 du ∇ · D(r, u)∇φ(r, u) (10.60) but this leaves open how to define Dg . Since it is unlikely that lethargy-dependent flux gradients will be available, various heuristic definitions suggest themselves; for example, D g (r) = ug ug−1 du D(r, u) (10.61) ug or D g (r) = 1 1 g = g g g 3tr 3(t − μ̄0 s ) (10.62) A similar ambiguity plagues the development of multigroup diffusion equations from Eq. (10.59). The formal definition of ξ -direction diffusion coefficient that arises from the integration of Eq. (10.37) over ug is g Dξ (r) = 3 ug ug−1 du t (r, u)Jξ (r, u) − Jξ,g (r)  ug  g  ≤g ug  −1 du s1 (r, u → u)Jξ (r, u ) (10.63) 10.3 Diffusion Theory The multigroup diffusion equations have the same form for all derivations: − ∂ g ∂φg (r) ∂ g ∂φg (r) ∂ g ∂φg (r) g Dx (r) − D (r) − Dz (r) + t (r)φg (r) ∂x ∂x ∂y ∂y ∂z ∂z =  g  →g s (r)φg  (r) + g  ≤g G χg  g νf (r)φg  (r), k  g = 1, . . . , G (10.64) g =1 where the elastic and inelastic scattering terms have been combined into a single scattering term. Boundary and Interface Conditions The appropriate transport theory boundary conditions are zero return current at external boundaries (unless there is an external beam source): φ(rb , , u) = 0, n̂b ·  < 0 (10.65) where nb is the outward unit vector to the external boundary at rb , and the appropriate interface condition is continuity of directional flux: ψ(ri − ε, , u) = ψ(ri + ε, , u) (10.66) where ε is a small quantity. These conditions obviously cannot be satisfied exactly by the diffusion theory approximation to the neutron flux. The Marshak boundary conditions discussed in Chapter 9 generalize to  (10.67) Jin = −nb · J(rb , u) = −nb · ψ(rb , , u)d = 0 Making use of the partial currents and geometric interpretation discussed in Section 3.1, this condition can be interpreted as the vanishing of the flux at an extrapolated distance 0.71/tr (u) outside the physical boundary. Given the ambiguity in defining tr (u), the computational difficulties that would ensue from an extrapolated boundary that varied with lethargy and the fact that the extrapolation distance is typically very small relative to the physical dimensions, the approximate boundary condition of vanishing flux on the physical boundary is appropriate as an approximation to Eq. (10.65): φ(rb , u) = 0, φg (rb ) = 0 (10.68) As an approximation to the interface condition of Eq. (10.66), we require that the first two Legendre moments of this equation be satisfied:   ψ(ri − ε, , u)d = ψ(ri + ε, , u)d (10.69)   ni · ψ(ri − ε, , u)d = ni · ψ(ri + ε, , u)d 399 400 10 Neutron Slowing Down Using the definitions of scalar flux and current as the first two Legendre moments of the angular flux, this may be written φ(ri − ε, u) = φ(ri + ε, u) (10.70) ni · J(ri − ε, u) = ni · J(ri + ε, u) and for multigroup diffusion theory φg (ri − ε) = φg (ri + ε) (10.71) ni · Jg (ri − ε) = ni · Jg (ri + ε) 10.4 Continuous Slowing-Down Theory Over much of the slowing-down range above (in lethargy) the fission spectrum and below the thermal range, neutron slowing down is due primarily to elastic scattering. Since there is no lethargy decrease in a scattering event below (in lethargy) the thermal range, the scatter-in integral is over lower lethargies only. It has been found convenient for computational purposes to replace the elastic scatter-in integral with a lethargy derivative of the associated elastic slowing-down density, which is computed in a coupled calculation, rather than evaluating the scatter-in integral directly. The various computational methods that have been developed for this purpose are known collectively as continuous slowing-down theory. P1 Equations in Slowing-Down Density Formulation Generalizing the definition of slowing-down density introduced in Chapter 4 to include anisotropic scattering, the isotropic slowing-down density is defined as the number of neutrons slowing down past energy E, or lethargy u, by isotropic (in the lab system) elastic scattering:  ∞  u i q0i (x, u) ≡ du du s0 (x, u → u )φ(x, u ) (10.72) u 0 and the linearly anisotropic slowing-down density is defined as the number of neutrons slowing down past lethargy u by linearly anisotropic scattering:  ∞  u i q1i (x, u) ≡ du du s1 (x, u → u )J (x, u ) (10.73) u 0 These two slowing-down densities are the zeroth and first Legendre components of the angular-dependent neutron slowing-down density. Making use of Eq. (10.20), the first of these equations can be written explicitly as  q0i (x, u) ≡ u u−ln 1/αi du  u u+ln 1/αi si (u )   eu −u φ(x, u ) 1 − αi 10.4 Continuous Slowing-Down Theory  = u u−ln 1/αi du si (u ) e u −u − αi φ(x, u ) 1 − αi (10.74) and making use of Eq. (10.23), the second of these equations can be written explicitly as  q1i (x, u) =    si (u )eu −u 1 − αi u−ln 1/αi u      2(1 − eu −u ) × 1 + 3μ̄c (u ) 1 − 1 − αi u du u+ln 1/αi du (10.75) where Ai is the atomic mass in amu of the scattering nuclei and αi = [(Ai − 1)/(Ai + 1)]2 . These slowing-down densities can be related to the scatter-in integrals in the P1 equations given by Eqs. (10.36) and (10.37): ∂q0i = ∂u  ∞ u i du s0 (x, u → u )φ(x, u) −  = si (x, u)φ(x, u) − ∂q1i = ∂u  u ∞ u 0 u 0 i du s0 (x, u → u)φ(x, u ) i du s0 (x, u → u)φ(x, u ) i du s1 (x, u → u )J (x, u) −  = μ̄i0 si (x, u)J (x, u) −  0 u  0 u (10.76) i du s1 (x, u → u)J (x, u ) i du s1 (x, u → u)J (x, u ) (10.77) Using Eqs. (10.76) and (10.77) to eliminate the scatter-in integrals in Eqs. (10.36) and (10.37) yields an equivalent form of the P1 equations (written for onedimensional slab geometry)  ∂q i ∂J (x, u) 0 + ne (x, u)φ(x, u) = − (x, u) + S0 (x, u) ∂x ∂u (10.78)  ∂q i (x, u) 1 ∂φ(x, u) 1 + tr (x, u)J (x, u) = − + S1 (x, u) 3 ∂x ∂u (10.79) I i=1 I i=1 where the nonelastic cross section ne (x, u) ≡ t (x, u) − I  si (x, u) (10.80) μ̄i0 si (x, u) (10.81) i=1 and the transport cross section tr (x, u) ≡ t (x, u) − I  i=1 401 402 10 Neutron Slowing Down have been defined in a natural way. Integrating these equations over ug = ug+1 − ug leads to the multigroup P1 equations in the slowing-down density formulation of elastic scattering:  ∂Jg (x) g g + ne (x)φg (x) = − q0i (x, ug+1 ) − q0i (x, ug ) + S0 (x) ∂x I (10.82) i=1  1 ∂φg (x) g g + tr (x)Jg (x) = − q1i (x, ug+1 ) − q1i (x, ug ) + S1 (x) 3 ∂x I i=1 g = 1, . . . , G (10.83) where the multigroup quantities are defined as  φg (x) ≡ du φ(x, u), ug  g  ug+1 Sn (x) ≡   g tr (x) ≡ ug+1 du J (x, u) ug ug+1 du Sn (x, u) ug g ne (x) ≡ Jg (x) ≡ ug+1 ug ug+1 ug (10.84) φ(x, u) du ne (x, u) φg (x) du tr (x, u) J (x, u) Jg (x) In this formulation, the natural definition of the group averaged transport equation is as a current averaged quantity. The same type of difficulty encountered previously in reducing the energydependent P1 equations to diffusion theory is present in Eq. (10.83); to obtain a Fick’s law type of relationship J = −D dφ/dx, it is necessary to require that the g anisotropic source S1 vanish and that the anisotropic slowing-down density not change over the group, which would be the case if it was assumed to be identically zero. Making these assumptions, the multigroup diffusion equation in the slowing-down density formulation is −   ∂φg (x) 1 ∂ g + ne (x)φg (x) g ∂x 3tr (x) ∂x = I  g q0i (x, ug ) − q0i (x, ug+1 ) + S0 (x) (10.85) i=1 with the diffusion coefficient being unambiguously defined in terms of a current spectrum-weighted group-averaged transport cross section, which contains some anisotropic effects—the average cosines of the scattering angle of the various nuclear species are embedded in the definition of the transport coefficient given by Eq. (10.81). 10.4 Continuous Slowing-Down Theory Making the approximation that the spatial dependence can be represented by a simple buckling [i.e., φ(x, u) = φ(u) exp(iBx), J (x, u) = J (u) exp(iBx)] in Eqs. (10.82) and (10.83) reduces these equations to the forms that are found in various multigroup spectrum codes: g iBJg + ne φg = J  g q0i (ug ) − q0i (ug+1 ) + S0 i=1  1 g g iBφg + tr φg = q1i (ug ) − q1i (ug+1 ) + S1 , 3 (10.86) J g = 1, . . . , G i=1 The asymptotic forms φasy (u) ∼ 1 and J (u)asy ∼ 1 are used in Eqs. (10.84) to define fine or ultrafine group constants. The few-group constants are then constructed from the solutions φg and Jg of the fine or ultrafine group calculation:  g g∈k ne (x)φg (x) k  ne (x) = g∈k φg (x) (10.87)  g g∈k tr (x)Jg (x) k  tr (x) = g∈k Jg (x) where g ∈ k indicates that the sum is over fine groups g within few group k. Slowing-Down Density in Hydrogen The evaluation of the slowing-down densities in hydrogen is quite straightforward because a neutron can scatter from any lethargy to any greater lethargy in a single collision, which is implicit in the fact that αH = 0. This fact allows Eqs. (10.72) and (10.73) to be written  u  q0H (x, u) = du sH (u )eu −u φ(x, u ) (10.88) 0 2 q1 (x, u) = 3  0 u  du sH (u )e3(u −u)/2 J (x, u ) (10.89) These equations may be differentiated to obtain ∂q0H + q0H (x, u) = sH (u)φ(x, u) ∂u (10.90) ∂q1H 2 3 + q1H (x, u) = sH (u)J (x, u) ∂n 2 3 (10.91) which may be put in multigroup form, to be used with Eqs. (10.86), by integration over ug : H q0H (ug+1 ) − q0H (ug ) + 1 2 q0H (ug+1 ) + q0H (ug ) ug = s g φg q1H (ug+1 ) − q1H (ug ) + 3 4 q1H (ug+1 ) + q1H (ug ) ug = 23 s g Jg H (10.92) 403 404 10 Neutron Slowing Down Heavy Mass Scatterers For moderators other than hydrogen, this procedure does not lead to such a simple differential equation for the slowing-down density, precisely because it is not possible for a neutron to lose all of its energy in a single collision, which means that the lower limits on the first integral in Eqs. (10.72) and (10.73) are u − ln(1/αi ), not zero. At the other extreme from hydrogen are heavy mass nuclei for which the maximum lethargy gain in a scattering collision is quite small and it is reasonable to expand the integrands in Eqs. (10.72) and (10.73) in Taylor’s series: ∂  i (u)φ(u) + · · · ∂u s ∂  i (u)J (u) + · · · si (u )J (u )  si (u)J (u) + (u − u) ∂u s si (u )φ(u )  si (u)φ(u) + (u − u) (10.93) (10.94) Various approximations result from keeping different terms in these expansions. Age Approximation The simplest such approximation, resulting from retaining only the first term in Eq. (10.93) and setting q1i = 0, is known as the age approximation: q0i (x, u)  ξi si (x, u)φ(x, u) (10.95) where ξi = ξiiso given by Eq. (10.21). With these approximations for q0i and q1I , Eqs. (10.78) and (10.79) become the inconsistent (because of the neglect of q1i ) P1 equations: ∂J (x, u) + ne (x, u)φ(x, u) ∂x =− I  ∂ ξi si (x, u)φ(x, u) + S0 (x, u) ∂u (10.96) i=1 1 ∂φ(x, u) + tr (x, u)J (x, u) = S1 (x, u) 3 ∂x which, with the additional assumption of zero anisotropic source (S1 = 0), can be reduced to the age-diffusion equation −   ∂ 1 ∂φ(x, u) + ne (x, u)φ(x, u) ∂x 3tr (x, u) ∂x =− I  ∂ ξi si (x, u)φ(x, u) + S0 (x, u) ∂u i=1 (10.97) 10.4 Continuous Slowing-Down Theory Selengut–Goertzel Approximation The age approximation for the slowing-down density, and hence the inconsistent P1 and age-diffusion equations, are restricted to heavy mass moderators for which the interval of the scatter-in integral, ln(1/αi ), is quite small, and certainly would not be appropriate for hydrogen. For a mixture of hydrogen and heavy mass moderators, the hydrogen can be treated exactly and the age approximation can be used for the remaining nuclei, resulting in the Selengut–Goertzel approximation I  ∂q i 0 i=1 ∂u = I   ∂q0H ∂  + ξi si φ ∂u ∂u (10.98) i=H Consistent P1 Approximation If, instead of setting q1i = 0, Eq. (10.73) is evaluated retaining the first term of the Taylor’s series expansion of Eq. (10.94), to obtain q1i (x, u)  ξi1 si (x, u)J (x, u) (10.99) where the first Legendre moment of the mean lethargy gain is defined as     1 3/2 3 1 2 1 + 1/Ai 1 − αi ln + 1 ξi = (Ai + 1) 9 2 αi    1 1/2 1 (10.100) ln + 1 − (1 − 1/Ai ) 1 − αi 2 αi the consistent P1 equations (with the Selengut–Goertzel approximation) are obtained: ∂J (x, u) + ne (x, u)φ(x, u) ∂x =− I  ∂q H ∂ ξi si (x, u)φ(x, u) − 0 (x, u) + S0 (x, u) ∂u ∂u (10.101) i=H 1 ∂φ(x, u) + tr (x, u) + ξi1 si (x, u) J (x, u) = S1 (x, u) 3 ∂x (10.102) Extended Age Approximation If the first two terms in the Taylor’s series expansion of Eq. (10.93) are retained in evaluating Eq. (10.72), the result is ai ∂q0 (x, u) ξi ∂u (10.103) eu −u − αi  αi [ln(1/αi )]2 (u − u) = − ξi 1 − αi 2(1 − αi ) (10.104) q0i (x, u)  ξi si (x, u)φ(x, u) + where  ai = u u−ln 1/αi du  405 406 10 Neutron Slowing Down Using the balance equation for the elastic slowing-down density in a very large region (neglecting leakage) ∂q0 (u) = −ne (u)φ(u) ∂u (10.105) allows Eq. (10.103) to be written   ai ξi si (x, u) − ne (x, u) φ(x, u) q0 (x, u)  ξi i  ξ t (x, u)φ(x, u) (10.106) With this extended age approximation, the summation on the right in the first of Eqs. (10.101) is replaced by −d(ξ t φ)/du in the age-diffusion equations. Grueling–Goertzel Approximation The slowing-down density for hydrogen can be calculated exactly, and the slowingdown density for heavy mass nuclei can be well approximated by one of the variants of the age approximation given above. However, light nonhydrogen moderators are not well approximated by any of the age approximations above. Greater accuracy can obviously be obtained by retaining more terms in the Taylor’s series expansions of Eqs. (10.93) and (10.94). In addition, it is possible to construct an approximate equation for the isotopic slowing-down densities which has the same form as the simple differential equation that describes the hydrogen slowing-down density and which reduces to the hydrogen equation when Ai = 1. Retaining three terms in the Taylor series expansion of Eq. (10.93) when used with Eq. (10.72) to evaluate λi0 dq0i /du + q0i yields     ∂q i ∂ ∂ 2 (si φ) ∂(si φ) i + ξ λi0 0 + q0i  λi0 ξi (si φ) + ai  φ + a i s i ∂u ∂u ∂u ∂u2 (10.107) The objective is to develop an equation for q0i which is like Eq. (10.90) for hydrogen. Neglecting ∂ 2 φ/∂u2 and choosing λi0 to make the ∂φ/∂u term vanish leads to λi0 ∂q0i (x, u) + q0i (x, u) = ξi si (x, u)φ(x, u) ∂u (10.108) which is of the same form as Eq. (10.90) for the hydrogen slowing-down density where λi0 = 1 − αi {1 + ln(1/αi ) + 12 [ln(1/αi )]2 } 1 − αi [1 + ln(1/αi )] (10.109) Retaining the first three terms in the Taylor series expansion of Eq. (10.94) when used with Eq. (10.73) in a similar calculation leads to an equation similar to the hydrogen Eq. (10.91): λi1 ∂q1i (x, u) + q1i (x, u) = ξi1 si (x, u)J (x, u) ∂u (10.110) 10.4 Continuous Slowing-Down Theory where        1 2 4 (1 + 1/Ai )2 1 + 1/Ai 8 8 1 3/2 ln − ln + − αi λi1 = − 4/Ai 3 9 2 3 αi 9   2   ! 1 1 1 1/2 8 − αi ln − 1− − 4 ln + 8 (10.111) ξi1 Ai αi αi again has been chosen to make ∂φ/∂u terms vanish. Summary of Pl Continuous Slowing-Down Theory The Pl equations ∂J (x, u) + ne (x, u)φ(x, u) ∂x ∂q H (x, u)  ∂q0i (x, u) =− 0 − + S0 (x, u) ∂u ∂u i=H 1 ∂φ(x, u) + tr (x, u)J (x, u) 3 ∂x ∂q H (x, u)  ∂q1i (x, u) − + S1 (x, u) =− 1 ∂u ∂u (10.112) i=H and the equations for the elastic slowing-down density, using the exact equations for hydrogen and the Grueling–Goertzel approximation for nonhydrogen nuclei, ∂q0H (x, u) + q0H (x, u) = sH (x, u)φ(x, u) ∂u 4 2 ∂q1H (x, u) + q1H (x, u) = sH (x, u)J (x, u) 3 ∂u 9 ∂q i λi0 0 (x, u) + q0i (x, u) ∂u λi1 = (10.113) ξi si (x, u)φ(x, u) ∂q1i (x, u) + q1i (x, u) = ξi1 si (x, u)J (x, u) ∂u represent the formulation usually referred to as Pl continuous slowing-down theory. Inclusion of Anisotropic Scattering In an ultrafine group calculation for which the group width is less than ln(1/αi ) for some of the nuclei which contribute strongly to neutron moderation, it is necessary to retain a large number of Legendre moments to accurately represent the group transfer cross sections, which actually represent the probabilities for scattering to 407 408 10 Neutron Slowing Down within relatively small angular intervals. (This situation is more likely to be found in a fast than in a thermal reactor.) The concept of slowing-down density can be extended to a higher order of anisotropy by defining the Legendre moments of the slowing-down density as the number of neutrons slowing down past lethargy u by the lth Legendre moment of the elastic scattering transfer function  ∞  u i qli (u) ≡ du du sl (u → u )φl (u ) (10.114) u 0 where, recalling Eq. (10.38), we have i sl (u → u) = Ni σsi (u )  l  =0 Tlli  (U ) blc i (u )  i i  Tll  (U )sl  (u ) c (u ) ≡ b0i  (10.115) l =0 Making a general Taylor series expansion about u in the integrand in Eq. (10.114), i   sl  (u )φl  (u ) = ∞  (u − u)n d n  i  (u)φl  (u) n! dun sl (10.116) n=0 and using Eq. (10.115) yields qli (u) = − ∞  dn (u)φl (u) Gi dun l,n+1 (10.117) n=0 where Gil,n (u) =  l  =0  +1 2n! 2l i sl  (u)  u  ∞  l  =0 u+ln 1/αi du u−ln 1/αi × Pl μc (u − u ) (u − u)n ≡  u  du Pl [μ0 (u − u )]  2eu −u 1 − αi i i sl  (u)Tll  ,n (u) (10.118) Extending the calculation that was described for the Grueling–Goertzel approximation yields an equation for each Legendre moment of the slowing-down density dql (u) + ql (u) du   I   dλl (u) i Gl,1 (u)φl (u) 1 − =− du λl (u) i + ∞  n  d [Gil,n+1 (u)φl (u)] n=2 where dun + λl (u) d n [Gil,n (u)φl (u)] dun  (10.119) 10.4 Continuous Slowing-Down Theory  ∞  λl (u)  ∞ = i i i i l  =0 Tll  ,2 (u)sl  (u) i i l  =0 Tll  ,1 (u)sl  (u) i Gil,2 (u) i Gil,1 (u) (10.120) are chosen to eliminate first derivative terms involving φl . The conventional Grueling–Goertzel theory is recovered by retaining only the l = 0, 1 slowing-down moments, neglecting terms n ≥ 2 in Eq. (10.119) and identifying ξil (u) = ali (u) = −Gil,1 (u) s (u) (10.121) −Gil,2 (u) s (u) Inclusion of Scattering Resonances An impractically large number of terms may have to be retained in the Taylor’s series expansion to obtain an accurate approximation for the slowing-down density when resonance scattering nuclei are present in the mixture, because resonance scattering in nuclei j may cause φ, hence si φ for another nuclear species i, to be a rapidly varying quantity. In this case, a better approximation may be developed by expanding the total collision density in a Taylor’s series: t (u )φl (u ) = ∞  (u − u)n d n [t (u)φl (u)] n! dun (10.122) n=0 Using this expansion to evaluate Eq. (10.114) yields ql (u) = − ∞  i i Hl,n (u) n=0 d n [t (u)φl (u)] dun (10.123) where ∞ i Hl,n (u) ≡ 1 1  (2l + 1) n!  2  l =0 × Pl  μc (u − u ) u du u−ln 1/αi   u+ln 1/αi n du Pl μ0 (u − u )  i  2eu −u sl  (u ) (u − u)n 1 − αi t (u ) (10.124) Differentiating Eq. (10.123),  ∞   d n+1 [t (u)φl (u)] dql (u) i Hl,n (u) =− du dun+1 i n=0 ∞ i (u)  dHl,n d n [t (u)φl (u)] · + du dun n=0  (10.125) 409 410 10 Neutron Slowing Down and carrying out a calculation similar to those described previously results in a hydrogen-like equation for the lth Legendre component of the slowing-down density:  dql (u) λ̂l (u) + ql (u) = − du  i Hl,0 (u) + λ̂l (u) i (u) dHl,0 i + ∞   i Hl,n (u) n=2 du  t (u)φl (u) d n [t (u)φl (u)] dun d n [t (u)φl (u)] dun * i (u) n dHl,n d [t (u)ξl (u)] (10.126) + λ̂l (u) du dun i + λ̂l (u)Hl,n−1 (u) where  λ̂l (u) = −  i i Hl,1 (u) i i i [Hl,0 (u) + dHl,1 (u)/du] (10.127) has been chosen to eliminate first derivative terms involving φl . Pl Continuous Slowing-Down Equations The lethargy-dependent Pl equations are generalized from the one-speed Pl equations of Chapter 9 by including a scattering loss term and a scatter-in source of neutrons: l + 1 ∂φl+1 (x, u) l ∂φl−1 (x, u) + + t (x, u)φl (x, u) 2l + 1 ∂x 2l + 1 ∂x  u = du s,l (x, u → u)φl (x, u ) + Sl (x, u), l = 1, . . . , L (10.128) 0 The Legendre moments of the slowing-down density are related to the Legendre moments of the scatter-in integral. Differentiating Eq. (10.114) yields ∂ql (u) = s,l (u)φl (u) − ∂u  u du s,l (u → u)φl (u ) (10.129) 0 Using this result to eliminate the elastic scatter-in integral in Eq. (10.128) leads to the Pl continuous slowing-down equations, l ∂φl−1 (x, u) l + 1 ∂φl+1 (x, u) l + + ne (x, u)φl (x, u) 2l + 1 ∂x 2l + 1 ∂x =− ∂ql (x, u) + Sl (x, u), ∂u l = 1, . . . , L (10.130) 10.5 Multigroup Discrete Ordinates Transport Theory where the nonelastic cross section is l ne (x, u) ≡ t (x, u) − s,l (x, u) (10.131) and the Legendre moments of the slowing-down density are calculated from Eqs. (10.126) for nonhydrogenic nuclei and Eqs. (10.90) and (10.91) and similarly derived higher Legendre moment equations for hydrogen. 10.5 Multigroup Discrete Ordinates Transport Theory In situations in which a high degree of angular anisotropy in the neutron flux could be expected, the low-order P1 and diffusion theory approximations might be inadequate to treat the combined slowing down and transport of neutrons. Such situations might arise in the treatment of slowing down in a highly heterogeneous lattice consisting of materials of very different properties or in the treatment of problems in which there is a highly directional flow of fast neutrons from one region to another. For such situations, the discrete ordinates methods of Chapter 9, extended to treat the neutron slowing down, are well suited. Generalizing the expansion of the differential (over scattering angle) elastic scattering cross section of Eq. (9.179) to an expansion of the double differential (over scattering angle and lethargy change) scattering cross section, and using the addition theorem for Legendre polynomials of Eq. (9.177) to relate the cosine of the scattering angle, μ0 , to the cosines of the incident, μ , and exiting, μ, directions for the scattering event, yields   s r,  · , u → u = L  2l  + 1 4π l  =0 = L  2l  + 1 4π l  =0 = L  2l  + 1 l  =0 4π sl  (r, u → u)Pl (μ0 )  l  sl  (r, u → u) m=−l  Yl  m (μ, ϕ)Yl∗ m (μ , ϕ  ) sl  (r, u → u)    l  (l  − m)! m × Pl  (μ)Pl  (μ ) + 2 P  (μ)Plm (μ ) cos m(ϕ − ϕ  ) (l  + m)! l  m=1 (10.132) Using this representation of the double differential scattering cross section in the neutron transport equation (10.26) yields 411 412 10 Neutron Slowing Down  · ∇ψ(r, , u) + t (r, u)ψ(r, , u)  ∞ L  = Sex (r, , u) + du Yl  m ()sl  (r, u → u)φl  m (r, u ) 0 + χ(u) 4π  ∞ l  =0 du νf (r, u )φ(r, u ) (10.133) 0 where the Legendre moments of the angular flux, φlm , and the scalar flux, φ, are defined as     ∗ φlm (r, u ) ≡ d Ylm ( )ψ r,  , u 4π     φ(r, u ) = φ00 (r, u ) ≡    d ψ r,  , u (10.134)  4π These equations may be reduced to a set of multigroup equations by integrating over the lethargy width ug = ug+1 − ug of group g: g  · ∇ψg (r, ) + t (r)ψg (r, ) G  L  g = Sex (r, ) + g  =1 l  =0 + g  →g Yl  m ()sl  G χg  g νf (r)φg  (r), 4π  g (r)φl  m (r) g = 1, . . . , G (10.135) g =1 where multigroup quantities have been defined  du ψ(r, , u) ψg (r, ) ≡ g φlm (r) ≡ du φlm (r, u) ug  φg (r) ≡  χg ≡ du φ(r, u) ug ug g νf (r) ≡ g  →g 4π 0 d 4π 0 ug (r) ≡  g Sex (r, ) ≡ du χ(u), g t (r) ≡ sl ug  ug d g ug du Sex (r, , u) du t (r, u)ψ(r, , u) ug du ψ(r, , u) = ug du t (r, u)φ(r, u) ug du νf (r, u)φ(r, u) ug ug du du φ(r, u) ug  du sl (r, u → u)φlm (r, u ) ug  du φlm (r, u ) (10.136) du φ(r, u) References Writing Eqs. (10.135) for each discrete ordinate, n , results in the set of multigroup discrete ordinates equations g n · ∇ψg (r, n ) + t ψg (r, n ) = Qg (r, n ), g = 1, . . . , G (10.137) where the group scattering plus fission plus external source term is Qg (r, n ) ≡ L G   g  =1 l  =0 + g  →g Yl  m (n )sl  g (r)φl  m (r) G χg  g g νf (r)φg  (r) + Sex (r, n ) 4π  (10.138) g =1 Equation (10.137), for each group, is of the same form as the discrete ordinates equation discussed in Chapter 9. Thus the methods used to solve the discrete ordinates equations in Chapter 9 can be applied to solve the multigroup discrete ordinates equations, on a group-by-group basis. For a given fission and scattering source, the multigroup discrete ordinates equations are solved group by group using the methods of Chapter 9. Then on the source iteration, the new scattering and fission source for each group are constructed by summing over contributions from all groups, and the solution for the multigroup fluxes on a group-by-group basis is repeated until convergence. The power iteration procedure for criticality eigenvalue calculations is the same as discussed in Chapter 9, but now the fission source is summed over the contributions from all groups. References 1 J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976), pp. 347–369. Multigroup Cross Sections, ANL-7318, Argonne National Laboratory, Argonne, IL (1967). 2 A. R. Henry, Nuclear Reactor Analysis, MIT Press, Cambridge, MA (1975), pp. 359–367 and 386–423. 5 J. H. Ferziger and P. F. Zweifel, The Theory of Neutron Slowing Down in Nuclear Reactors, MIT Press, Cambridge, MA (1966). 3 W. M. Stacey, “The Effect of Wide Scattering Resonances on Neutron Multigroup Cross Sections,” Nucl. Sci. Eng. 47, 29 (1972); “The Effect of Anisotropic Scattering upon the Elastic Moderation of Fast Neutrons,” Nucl. Sci. Eng. 44, 194 (1971); “Continuous Slowing Down Theory for Anisotropic Elastic Moderation in the Pn and Bn Representations,” Nucl. Sci. Eng. 41, 457 (1970). 4 B. J. Toppel, A. L. Rago, and D. M. O’Shea, MC2 : A Code to Calculate 6 M. M. R. Williams, The Slowing Down and Thermalization of Neutrons, NorthHolland, Amsterdam (1966), pp. 317– 516. 7 D. S. Selengut et al., “The Neutron Slowing Down Problem,” in A. Radkowsky, ed., Naval Reactors Physics Handbook, U.S. Atomic Energy Commission, Washington, DC (1964). 8 G. Goertzel and E. Grueling, “Approximate Method for Treating Neu- 413 414 10 Neutron Slowing Down tron Slowing Down,” Nucl. Sci. Eng. 7, 69 (1960). 9 H. J. Amster, “Heavy Moderator Approximations in Neutron Transport Theory,” J. Appl. Phys. 29, 623 (1958). 10 H. Bohl, Jr., E. M. Gelbard, and G. H. Ryan, MUFT-4: A Fast Neutron Spectrum Code, WAPD-TM-22, Bettis Atomic Power Laboratory, West Miflin, PA (1957). 11 R. A. Knief, Nuclear Engineering, 2nd ed., Taylor & Francis, Washington, DC (1992). Problems 10.1. Calculate the average cosine of the scattering angle in the CM system for neutrons at 1 MeV, 100 keV, and 1 keV colliding with uranium, iron, carbon, and hydrogen. 10.2. Calculate the values of the average lethargy increase, ξ , and the average cosine of the scattering angle in the lab system, μ0 , for the neutron energies and nuclei of Problem 10.1, for isotropic scattering and for linearly anisotropic scattering. 10.3. Carry through the steps in the derivation of the lethargy-dependent P1 equations given by Eqs. (10.36) and (10.37). 10.4. Carry through the derivation of the isotropic and linearly anisotropic lethargy transfer functions of Eqs. (10.39) and (10.40). 10.5. Divide the energy interval 10 MeV > E > 1 eV into 54 equal-lethargy intervals. Evaluate the multigroup scattering g  →g transfer functions s0 for carbon for g  = 1, 10, and 50. 10.6. Carry through the steps in the derivation of the lethargy-dependent B1 equations given by Eqs. (10.50). 10.7. Solve for the lethargy-dependent neutron flux and current in an infinite medium, using the age approximation of Eqs. (10.96). 10.8. Derive a differential equation similar to Eqs. (10.90) and (10.91) for the higher Legendre moments of the slowing-down density in hydrogen. 10.9. Derive the multigroup approximation for the P1 continuous slowing-down equations, Eqs. (10.112) and (10.113). 10.10. Write a computer code to solve the multigroup P1 continuous slowing-down equations of Problem 10.9 for an assembly consisting of 3% enriched, zircalloy-clad UO2 and water. The fuel pins are 1 cm in diameter with clad thickness of 0.05 cm in a square array with fuel pin center-to-center distance of 2.0 cm. Assume that spatial gradients can be neglected. 415 11 Resonance Absorption 11.1 Resonance Cross Sections When the relative (center-of-mass) energy of an incident neutron and a nucleus plus the neutron binding energy match an energy level of the compound nucleus that would be formed upon neutron capture, the probability of neutron absorption is quite large. For the odd-mass fissionable fuel isotopes, resonances occur from a fraction of 1 eV up to a few thousand eV, and for the even-mass fuel isotopes, resonances occur from a few eV to about 10,000 eV, as shown in Figs. 11.1 to 11.4. At the lower energies the resonances are well separated, but at the higher energies the resonances overlap and become unresolvable experimentally. We first examine the widely spaced resonances at lower energy, where spatial self-shielding, as well as energy self-shielding, is important. At higher energies, spatial self-shielding becomes less important, but resonance overlap interference effects become important. 11.2 Widely Spaced Single-Level Resonances in a Heterogeneous Fuel–Moderator Lattice Neutron Balance in Heterogeneous Fuel–Moderator Cell At lower energies in the 10-eV range, the neutron mean free path becomes comparable to the fuel and moderator dimensions, and it is important to take into account the spatial heterogeneity of the fuel–moderator cell. The fuel assembly in a nuclear reactor generally consists of a repeating array of unit cells consisting of fuel, moderator/coolant, clad, and so on. For simplicity, we consider a two-region unit cell of fuel (F ) and a separate moderator (M). We allow further for a moderator admixed with the fuel (e.g., the oxygen in UO2 fuel). We return to the problem of calculating the absorption of neutrons in widely spaced resonances which was treated for a homogeneous mixture in Section 4.3, but now take into account the important spatial self-shielding effects that are present in a heterogeneous fuel–moderator 416 11 Resonance Absorption Fig. 11.1 235 U fission cross section. (From http://www.nndc.bnl.gov/.) Fig. 11.2 235 U capture cross section. (From http://www.nndc.bnl.gov/.) 11.2 Widely Spaced Single-Level Resonances in a Heterogeneous Fuel–Moderator Lattice Fig. 11.3 238 U capture cross section. (From http://www.nndc.bnl.gov/.) Fig. 11.4 238 U elastic scattering cross section. (From http://www.nndc.bnl.gov/.) 417 418 11 Resonance Absorption lattice. Consider a repeating array of fuel–moderator cells with fuel volume VF and moderator volume VM . Define the first-flight escape probabilities PF 0 (E) = probability that a neutron that slows down to energy E in the fuel will make its next collision in the moderator PM0 (E) = probability that a neutron that slows down to energy E in the moderator will make its next collision in the fuel We assume that these probabilities are uniform over the fuel and moderator, respectively. The neutron balance equation in the fuel can be written F + tF (E) φF (E)VF m  E/αF  E/αm F    F (E  )φF (E  )  m φF (E ) = VF [1 − PF 0 (E)] dE  S + dE (1 − αF )E  (1 − αm )E  E E  E/αM  M (E  )φM (E  ) dE  S (11.1) + VM PM0 (E) (1 − αM )E  E The left side of the equation is the total reaction rate of the fuel plus admixed moderator in the fuel volume. The first term on the right side is the source of neutrons scattering to energy E in the fuel (from scattering collisions with fuel and with admixed moderator nuclei) times the probability (1 − PF 0 ) that their next collision is in the fuel, and the second term is the source of neutrons scattering into energy E in the moderator times the probability PM0 that their next collision is in the fuel. The practical width of an absorption resonance will generally be much smaller than the scattering-in interval of the moderator, p  (1 − αM )E0 , or of the admixed moderator, p  (1 − αm )E0 , which allows us to use the asymptotic form of the neutron flux above the resonance energy in the moderator and the fuel, φasy (E) ∼ 1/E, to evaluate the moderator and admixed moderator scattering integrals in Eq. (11.1), leading to F + tF (E) φF (E)VF m  E/αF F   F m  S (E )φF (E ) = VF [1 − PF 0 (E)] dE + (1 − αF )E  E E + VM PM0 (E) SM (E) E (11.2) Reciprocity Relation Define G(rF ; rM ) as the probability that a neutron isotropically scattered to energy E at location rF in the fuel travels without collision to location rM in the moderator, and G(rM ; rF ) as the probability that a neutron isotropically scattered to energy E at location rM in the moderator travels without collision to location rF in the 11.2 Widely Spaced Single-Level Resonances in a Heterogeneous Fuel–Moderator Lattice fuel. For a uniformly distributed source of neutrons scattering to energy E in each region, the following identities must obtain:   drF G(rF ; rM ) drM VF PF 0 (E) ≡ tM (E) VF  F  VM PM0 (E) ≡ m + tF (E) VM  VM  drM (11.3) VF G(rM ; rF ) drF Since G(rF ; rM ) and G(rM ; rF ) depend only on the collision probability along the path between rF and rM , and this probability is independent of the direction in which the neutron travels, G(rM ; rF ) = G(rF ; rM ), Eqs. (11.3) may be combined to obtain the reciprocity relation between the two first-flight collision probabilities: F + tF (E) VF = PM0 (E)tM (E)VM PF 0 (E) m (11.4) If we make the assumption that absorption is very small relative to scattering in the moderator, the reciprocity relation may be used to write Eq. (11.2) as F + tF (E) φF (E) m  E/αF F   F m  S (E )φF (E ) dE + = [1 − PF 0 (E)] (1 − αF )E  E E + PF 0 (E) F +  F (E) m t E (11.5) Narrow Resonance Approximation If the practical width of the resonance is much smaller than the scattering-in interval of the resonance nucleus, p  (1 − αF )E0 , the contribution of the resonance to the scattering-in integral in Eq. (11.5) can be neglected and the asymptotic flux in the fuel φ(E) ∼ 1/E can be used to evaluate the integral to obtain φFNR (E) = F [1 − PF 0 (E)]pF + PF 0 (E)tF (E) + m F +  F (E)]E [m t (11.6) Using this form for the neutron flux to evaluate the capture resonance integral  (11.7) I γ ≡ dEσγ (E)φ(E) leads to the narrow resonance approximation for the heterogeneous resonance integral:  γ INR =  = σpF + σmF + PF 0 (E)[σtF (E) − σpF ] dE σγ (E) E σmF + σtF (E) σpF + σmF + σe (E) dE σγ (E) F E σt (E) + σmF + σe (E) (11.8) 419 420 11 Resonance Absorption where we have defined an escape cross section: PF 0 (E) ≡ σe (E) σe (E) + σtF (E) + σmF ⇒ σe (E) = PF 0 (E) σtF (E) + σmF 1 − PF 0 (E) (11.9) and used the notation σtF (E) and σpF for the total and potential scattering microscopic cross sections of the resonance absorber, and σmF for the cross section of the admixed moderator per fuel nucleus. Wide Resonance Approximation If the practical width of the resonance is much larger than the scattering-in interval of the resonance nuclei, p  (1 − αF )E0 , the term sres (E  )φ(E  )/E  ≈ sres (E)φ(E)/E in the integrand of Eq. (11.5), leading to the wide resonance approximation for the flux in the fuel region, φFWR (E) = F PF 0 (E)tF (E) + m F F [t (E) + m − (1 − PF 0 (E))SF (E)]E Using this result to evaluate the resonance integral of Eq. (11.7) leads to  dE PF 0 (E)σtF (E) + σmF γ σγ (E) F IWR = E σt (E) + σmF − [1 − PF 0 (E)]σSF (E)  dE σ F + σe (E) = σγ (E) F m E σa (E) + σmF + σe (E) (11.10) (11.11) where σaF = σγF + σfF is the microscopic absorption cross section of the resonance absorber. Evaluation of Resonance Integrals Recalling from Section 4.3 the form of the single-level resonance cross section averaged over the thermal motion of the nuclei, the (n, γ ) capture cross section or fission cross section averaged over the motion of the nucleus can be written  1/2 q E0 ψ(ξ, x), q = γ , f (11.12) σq (E, T ) = σ0 E and the total scattering cross section, including resonance and potential scattering and interference between the two, can be written σs (E, T ) = σ0 n ψ(ξ, x) + σ0 R χ(ξ, x) + 4πR 2 λ0 (11.13) where R is the nuclear radius, λ0 is the neutron De Broglie wavelength, and the functions  ∞ ξ dy 2 2 ψ(ξ, x) = √ e−(1/4)(x−y) ξ (11.14) 1 + y2 2 π −∞ 11.2 Widely Spaced Single-Level Resonances in a Heterogeneous Fuel–Moderator Lattice ξ χ(ξ, x) = √ π  ∞ e−(1/4)(x−y) 2ξ 2 −∞ y dy 1 + y2 (11.15) are integrals over the relative motion of the neutron and nucleus, x = 2(Ecm − E0 )/ , assuming that the nuclear motion can be characterized by a Maxwellian distribution with temperature T , and Ecm is the energy of the neutron in the neutron–nucleus center-of-mass system. The parameters characterizing the resonance are σ0 , the peak value of the cross section; E0 , the neutron energy in the center-of-mass system at which it occurs; , the resonance width; γ , the partial width for neutron capture, f , the partial width for fission; and n , die partial width for scattering. The resonance absorption cross section is symmetric about E0 , but the scattering cross section is asymmetric because the potential and resonance scattering interfere constructively for E > E0 and destructively for E < E0 , as indicated in Fig. 11.4. The temperature characterizing the nuclear motion is contained in the parameter ξ= (11.16) (4E0 kT /A)1/2 where A is the atomic mass (amu) and k is the Boltzmann constant. Using these forms for the resonance cross sections in Eqs. (11.8) and (11.11), the resonance integrals become in the narrow resonance approximation (neglecting interference scattering) γ INR = ≡ γ 2E0 γ E0  F  σp + σmF + σe  ∞ −∞ ψ(ξ, x) dx ψ(ξ, x) + β  F  σp + σmF + σe J (ξ, β) (11.17) and in the wide resonance approximation γ IWR = γ 2E0  F  σm + σe  ∞ −∞    ψ(ξ, x) dx γ  F ≡ σm + σe J ξ, β   ψ(ξ, x) + β E0 (11.18) where β≡ σpF + σmF + σe σ0 , β ≡ σmF + σe σ0 (11.19) γ The J (ξ, β) function is tabulated in Table 4.3. The properties of the moderator region do not appear explicitly in these expressions for the resonance integral because we have assumed that a neutron which escapes the fuel will have its next collision in the moderator and because we assumed that absorption in the moderator could be ignored in using the reciprocity relation. 421 422 11 Resonance Absorption Table 11.1 Infinite Dilution Total Resonance Integrals for Some Heavy Elements∗ Isotope (n, γ ) (barns) RI(n, (n, f ) (barns) RI(n, 84 138 864 631 133 346 661 445 181 180 1305 278 8103 1130 391 – 774 – 7 278 8 7 – 302 573 14 2 9 6 1258 232 Th 233 U 233 Pa 234 U 235 U 236 U 237 Np 239 Np 239 Pu 241 Pu 241 Am 238 U 240 Pu 242 Pu 242 Am * Calculated with ORIGEN (Ref. 14). Infinite Dilution Resonance Integral In the infinite dilution limit σmF + σe  σ0 , all forms for the resonance integral approach the infinite dilution value: γ I∞ = π γ σ0 , 2 E0 f I∞ = π f σ0 2 E0 (11.20) Infinite dilution resonance integrals for a number of fuel isotopes are given in Table 11.1. Actual resonance integrals will be smaller because of self-shielding effects. Equivalence Relations For a given resonance absorbing species, assemblies with the same values of σmF + σe have the same resonance integral. Furthermore, the heterogeneous assemblies with a given value of σmF + σe have the same resonance integrals as homogeneous assemblies which have moderator scattering cross section per resonance absorber nucleus σsM = σmF + σe . Equations (11.17) and (11.18) reduce to the homogeneous resonance integrals of Eqs. (4.68) and (4.71) when σe ∼ PF 0 = 0 (i.e., in the case of a resonance absorber with a homogeneously admixed moderator). 11.2 Widely Spaced Single-Level Resonances in a Heterogeneous Fuel–Moderator Lattice Heterogeneous Resonance Escape Probability The resonance capture rate in a fuel–moderator cell with fuel volume VF and moderator volume VM is  (11.21) Rγ = VF NF σγF (E)φF (E) = VF NF I γ We have evaluated the resonance integral for an asymptotic flux above the resonance φasy = 1/E, assumed uniform over the fuel and moderator. Using the asymptotic relationship between the slowing-down density, q, and the flux q = ξ s Eφasy (11.22) where the average asymptotic moderating power of the cell is ξ s = F) VM ξM sM + VF (ξF pF + ξm m (11.23) V M + VF indicates that an asymptotic flux of 1/E implies an asymptotic slowing-down density of q = ξ s . The resonance escape probability for the cell is unity minus the resonance absorption probability, and the latter is the resonance absorption rate divided by the total number of neutrons slowing down q(VF + VM ): p=1− Rγ VF NF I γ Iγ =1− =1− q(VF + VM ) ξ s (VF + VM ) ξ σs (11.24) where ξ σs is the cell moderating power per fuel nucleus. The total resonance escape probability over an energy group g containing several resonances is P= # i∈g pi = # i∈g γ I 1− i ξ σs    1  γ  exp − Ii ξ σs (11.25) i∈g where i ∈ g indicates all of the resonances within energy group g extending from Eg to Eg−1 . By lumping the fuel, the neutron flux in the fuel is reduced relative to the flux in the separate moderator—spatial self-shielding—and it is possible to decrease the resonance integral without decreasing the slowing-down power, thus increasing the resonance escape probability relative to the value for the same fuel and moderator distributed homogeneously. In fact, lumping the natural uranium fuel in the early graphite-moderated reactors was essential to achieving criticality—the resonance escape probability increased from about 0.7 in a homogeneous graphite– natural uranium assembly to about 0.88 in a heterogeneous assembly. Homogenized Multigroup Resonance Cross Section An effective multigroup cross section for the resonance absorber can be constructed by summing the resonance absorption rates over all of the resonances 423 424 11 Resonance Absorption in the group, dividing by the fuel number density, NF , dividing by the volume of the fuel–moderator cell, and dividing by the integral of the asymptotic flux over the energy interval of the group:   γ i VF i∈g Ii i∈g Rγ /NF res (11.26) σg = = E −1 (VF + VM ) ln(Eg−1 /Eg ) (VF + VM ) Egg φasy dE Improved and Intermediate Resonance Approximations The narrow [ p  (1 − αF )E0 ] and wide [ p  (1 − αF )E0 ] resonance approximations are limiting cases. For many resonances, the actual situation is intermediate to these extremes. It is possible to improve upon the narrow resonance and wide resonance approximations using the neutron balance equation to improve the flux solution iteratively: (n) σmF + σtF (E) φF (E)   E/αF F  (n−1) (E  ) σmF  σs (E )φF = [1 − PF 0 (E)] dE + (1 − αF )E  E E + PF 0 (E) σmF + σtF (E) E (11.27) The initial flux guess can be the narrow or wide resonance approximation, or an intermediate resonance approximation which is suggested by comparison of the two: (1) φNR = σmF + σpF + σe 1 σmF + σtF + σe E (1) φWR = (1) φIR = σmF 1 σmF + σe + σaF + σe E (11.28) σmF + λσpF + σe 1 σmF + λσsF + σaF + σe E where λ, which is in the range 0 < λ < 1, is a parameter to be determined separately (Chapter 13). In practice, it is not practical to extend this procedure beyond a single iteration. 11.3 Calculation of First-Flight Escape Probabilities To evaluate the resonance integrals of Section 11.2 it is first necessary to calculate the probability PF 0 that a neutron reaching energy E in the fuel will have its next collision in the moderator. Although this can be done exactly with a Monte Carlo calculation, a large number of such calculations would be necessary, and a number of analytical approximations have been developed. 11.3 Calculation of First-Flight Escape Probabilities Fig. 11.5 Geometry notation for escape probability calculation. Escape Probability for an Isolated Fuel Rod For an isolated fuel rod surrounded by moderator, the probability PF 0 that a neutron reaching energy E in the fuel will have its next collision in the moderator is just the probability that the neutron will escape from the fuel rod without a collision, P0 . For a uniform fuel rod, the probability that a neutron created isotropically at location r0 within a fuel rod of arbitrary shape (Fig. 11.5) escapes from the fuel rod is  −l/λ e ( · ns ) ds (11.29) P0 (r0 ) = 4πl 2 where λ = t−1 is the total mean free path, l(r0 , ) the distance from r0 to the surface of the rod in the direction , ns the outward normal vector to the surface of the fuel rod, ( · ns )ds/4πl 2 (r0 , ) the solid angle that the surface element ds in the direction ω subtends, and exp(−l/λ) the probability that the neutron will reach the surface without collision. If the neutrons are created isotropically, the average escape probability is P0 = 1 V  P0 (r0 ) dr0 = 1 V   dV d e−l/λ 4π (11.30) If we represent the volume as tubular elements oriented in the  direction with cross-sectional area ds(ni · ) where ni is the inward normal unit vector on the rod surface, the volume element is dV = dl ds(ni · ), and Eq. (11.30) can be integrated over length l to obtain     λ ds d( · ni ) 1 − e−ls ()/λ (11.31) P0 = 4πV where ls () is the chord length from surface to surface of the rod in direction . For a long fuel plate of thickness a, this may be evaluated exactly: P0 =    a λ 1 − E3 a 2 λ (11.32) 425 426 11 Resonance Absorption where E3 is the exponential integral function. An approximate evaluation is possible for a sphere of radius a:     8 λ 4 2 −(3/2)(R0 /λ) 1− , + 1+ e P0 = R0 3(R0 /λ) 9(R0 /λ)2 3(R0 /λ) R0 ≡ 4a 3 (11.33) and for a long cylindrical fuel rod of radius a,     4 π/2 R0 λ 1− cos β , dβ cos βKi3 P0 = R0 π 0 λ  ∞ −X cosh u e du Kin (X) ≡ , R0 = 2a (cosh u)n 0 (11.34) A more general evaluation may be made by invoking the theory of chord distributions. The probability that the length of a chord lies between ls and ls + dls is φ(ls ) = ls =ls ( · ni ) d (·ni )>0 ds (11.35) ( · ni ) d ds where the integral over  in the numerator includes only those values of  for which ls = ls , (i.e., is a chord length of the fuel rod). The denominator is readily evaluated:   1 ( · ni ) d ds = 2πS μ dμ = πS (11.36) 0 where S is the surface area of the fuel rod. In this representation the volume of the fuel rod is  V = l( · ni ) ds, ( · ni ) > 0 and the mean chord length is      1 ( · ni ) d ds dls ls l¯s ≡ ls φ(ls ) dls = πS ls =ls (·ni )>0  1 4V = ls ( · ni ) d ds = , ( · ni ) > 0 πS S Hence 4πV φ(ls ) dls = l¯s   ls =ls  ( · ni ) d ds (11.37) (11.38) (11.39) (·ni )>0 Using these results in Eq. (11.31) yields    λ  λ P0 = 1 − e−ls /λ φ(ls ) dls = − e−ls /λ φ(ls ) dls l¯s l¯s (11.40) 11.3 Calculation of First-Flight Escape Probabilities When the dimensions of the fuel rod are small compared to the mean free path, ls  λ, this reduces to P0  1 − 1 (l¯s2 ) 1 2 l¯s λ (11.41) and when the dimensions are large compared to the mean free path, ls  λ, P0  λ l¯s (11.42) which suggests the rational approximation P0 = 1 1 = 1 + (l¯s /λ) 1 + (4V /λS) (11.43) The rational approximation is known to underestimate the escape probability. An improved approximation for a long cylindrical rod has been obtained by integrating an empirical fit for the chord length distribution function:   1 λ 1− P0 = 1 + (l¯s /4.58λ)4.58 l¯s (11.44) An improved rational approximation of the form P0 =     1 4V /λs −c 1− 1+ 4V /λs c (11.45) with c = 2.09, has been determined empirically to be more accurate than the Wigner approximation and more accurate than the Sauer approximation for all geometries other than cylindrical. Note that the approximation of Eq. (11.45) reduces to the Wigner approximation for c = 1 and to the Sauer approximation for c = 4.58. Closely Packed Lattices In a lattice of closely packed fuel elements interspersed in moderating material a neutron escaping from a fuel element without collision may traverse a distance in moderator without collision and enter another fuel element, where it may have a collision with a fuel atom or may pass through uncollided to enter moderator again, and so on. In this situation, the probability of a neutron escaping from the fuel element in which it is scattered to energy E(P0 ) is not the same as the probability that the neutron will have its next collision in moderator (PF 0 ), but the two are (1) related. Let Gm be the probability that a neutron escaping from the original fuel element will traverse the line-of-sight distance of moderator separating the original (2) fuel element from other fuel elements without a collision, Gf be the probability 427 428 11 Resonance Absorption that the neutron will traverse the second fuel element without collision to reenter moderator, and so on. Then we can write     (1) (2) (2) PF 0 = P0 1 − G(1) m + Gm G f 1 − G m  (2) (2) (3)  (3) (11.46) + G(1) m Gf Gm Gf 1 − G m + · · · (n) The various Gx depend on the lattice geometry and are not the same for successive flights in moderator or fuel (i.e., not equal for n = 1 and n = 2 or n = 4 and n = 5). However, if we make the approximation that the individual flight probabilities can be replaced by an average value, Eq. (11.46) can be summed: PF 0 = P0 (1 − Gf )(1 − Gm ) 1 − Gf Gm (11.47) Gm,f can be estimated by analogy with Eq. (11.43) or heuristically as Gm,f = 1 1 + l¯m,f /λm,f or ¯ Gm,f = e−lm,f /λm,f (11.48) when l¯m and λm are the mean chord length through the moderator between fuel elements and the mean free path of neutrons in the moderator, etc. Such corrections to P0 are known as Dancoff corrections and allow PF 0 to be written PF 0 = P0 (1 − γ ) (11.49) where the factor γ accounts for the decrease in the probability that a neutron escaping a fuel element will first collide with a moderator nucleus because of the presence of other fuel elements. For fuel rods arranged in a square or hexagonal lattice structure, γ= where exp(−τ l¯s NF σsM ) 1 + (1 − τ )l¯s NF σsM √   1/2   2π 1 + VVF − 1 VVMF − 0.08 square M τ =     1/2  π 1/2  √ − 1 VVMF − 0.12 hexagonal 1 + VVMF (11.50) (11.51) 2 3 11.4 Unresolved Resonances Unlike the case of the resonances in the resolved energy region (up to a few hundred eV or less) where parameters for each individual resonance can be evaluated explicitly from the high-resolution data, the evaluations of such parameters become increasingly more difficult as the Doppler and instrument resolution widths 11.4 Unresolved Resonances become much greater than the corresponding natural width in the relatively high energy range. Under such circumstances, it is not possible to deal with the physical quantities of interest as a function of energy in great detail. Instead, it is necessary to estimate the expectation values of these quantities on the basis of statistical theory. Two types of expectation values of particular interest in reactor applications are the reaction rate of a given process, denoted by σx φ , and the average flux, denoted by φ , in the energy interval where many resonances are present. Since the NR-approximation described earlier is usually applicable in the unresolved energy range at relatively high energy, the extension of the J -integral approach is quite natural. The expectation values of interest can be expressed in terms of the population averages of an ensemble of resonance integrals with their resonance parameters determined by the known distribution functions from the statistical theory of spectra. In principle, these averages can be determined once the average resonance parameters are specified. Two types of distributions are needed to characterize the statistical behavior of the resonance parameters. According to Porter–Thomas, the partial widths are theoretically expected to exhibit a chi-squared distribution with ν degrees of freedom about their mean value x : Pν (y) dy =  (ν/2)−1 ν νy e−(νy/2) dy 2 (ν/2) 2 (11.52) where y = x / x and F (ν/2) is the gamma function of argument ν/2. The degree of freedom ν is identifiable with the number of open channels for reaction process x. The level spacing between two adjacent levels for a given spin state, D = |Ek − Ek+1 |, is characterized by the Wigner distribution of the form W (y) dy =   π π y exp − y 2 dy 2 4 (11.53) where y = D/ D . Physically, it signifies the tendency of repulsion between the adjacent levels of the same spin sequence. For the integral approach to be described, a level correlation function that specifies the probability of finding any level Ek  at a distance |Ek  − Ek | away from a given level Ek is also required. It is related to the Wigner distribution via the convolution integral equation, defined as follows:  (y) = W (y) + y W (t)(y − t) dt (11.54) 0 For Ek  belonging to a different spin sequence with respect to Ek , the levels are statistically independent, and consequently, the correlation function becomes unity. With the specification of distributions, the averages can be determined once the average parameters are provided. For elastic scattering, in which the neutron width is explicitly energy dependent, one convenient average parameter usually used is 429 430 11 Resonance Absorption the strength function. For neutrons with orbital angular momentum l, the strength function is  (0) gJ nlk , k ∈ J, l (11.55) Sl = J Dk (0) where nlk is the average reduced neutron width for given l and k, which is energy independent, and Dk is the average level spacing for the sequence in which level k occurs, with gJ = 2J + 1 2(2I + 1) (11.56) where J is the total spin of the neutron–nuclide system and I is the spin of the target nucleus. Statistical resonance parameters for some of the principal nuclides are given in Table 11.2. The single-level Breit–Wigner formula, now generalized to include spin effects, is σxk = πλ20 gJ xk (E − Ek )2 + ( nk (11.57) 2 k /2) where Fxk , nk , and k are the capture (x = γ ) or fission (x = f ) width, the neutron width, and the total width, respectively, for a resonance at Ek , and λ0 is the De Broglie neutron wavelength. Multigroup Cross Sections for Isolated Resonances Using a narrow resonance approximation for the neutron flux, φ ∼ 1/t , the effective multigroup cross section is g σxk = Eg−1 dE σxk (E)/t (E) Eg Eg−1 dE/t (E) Eg where f= p Eg  Eg−1 Eg dE t (E) = p xk Jk f Eg Nres (11.58) (11.59) In the unresolved resonance region, statistical averages over the distribution functions of Eqs. (11.52) to (11.54) are used to construct an effective multigroup cross section for process x: 0 g 1 σp  σxk = f spin states xk Jk Dk (11.60) where σp = p /Nres is the potential scattering cross section per resonance nucleus and · indicates averages over statistical distributions of both widths and level spacings. 11.4 Unresolved Resonances Table 11.2 Statistical Resonance Parameters for 235 U, 238 U, and 239 Pu 235 U 238 U 239 Pu S0 = (0.915 ± 0.5) × 10−4 S0 = (0.90 ± 0.10) × 10−4 S0 = (1.07 ± 0.1) × 10−4 S1 = (2.0 ± 0.3) × 10−4 S1 = (2.5 ± 0.5) × 10−4 S1 = (2.5 ± 0.5) × 10−4 l=0 = 0.53 ± 0.05 eV D̄obs l=0 = D l=0,1 = 20.8 ± 2.0 eV D̄obs J =1/2 l=0 = 2.3 ± 0.2 eV D̄obs D̄Jl=1 =2 = 1.23 eV D̄Jl=1 =1/2 = 11.4 ± 1.1 eV D̄Jl=0,1 =0 = 8.78 eV l=0,1 D̄Jl=0,1 =3 = DJ =4 = 1.06 eV ¯ γl=0 = 24.8 ± 5.6 meV D̄Jl=0,1 =1 = 3.12 eV D̄Jl=1 =5 = 1.18 eV νγ = 39 D̄Jl=1 =2 = 2.12 eV ¯ γl=0 = 47.9 meV σpot = 10.6 ± 0.2 barns ¯ γl=0 = 38.7 meV (0 < E < 50 eV); S0J =3 ≈ S0J =4 ≈ S0 l=0 = 65.1 meV (0 < E < 50 eV) R = (9.18 ± 0.13) × 10−13 cm νf = 4 ¯ l=0 f J =0 = 2800 meV ¯ l=0 f J =1 = 57 meV (0 < E < 100 eV) νγ = 27 νf = 2 for (1, J ) = (0, 0) σpot = 11.7 ± 0.1 barns νf = 1 for (1, J ) = (0, 1) R = (9.65 ± 0.05) × 10−13 cm νγ = 24 σpot = 10.3 ± 0.15 barns R = (9.05 ± 0.11) × 10−13 cm Source: Data from Ref. 5; used with permission of American Nuclear Society. Self-Overlap Effects The large Doppler width for high-energy neutrons and the small level spacing produce a high degree of self-overlap among the resonances for fissile isotopes, and significant but less self-overlap for the fertile isotopes. In fast reactor spectra, the self-overlap effect is not important for the fertile isotopes at operating temperatures, but does affect the temperature dependence of the Doppler effect above about 10 keV. The effect of the presence of other resonances on the effective cross section of resonance k arises from their effect on the flux, φ ∼ 1/t , and gives rise to a 431 432 11 Resonance Absorption generalization of the J function:  ψk dxk 1 ∞ ψk dxk  = 2 −∞ ψk + βk −∞ ψk + βk + k  =k (σ0k  /σ0k )ψk    ∞ ψk k  =k (σ0k  /σ0k )ψk  dxk 1  − 2 −∞ (βk + ψk )[βk + ψk + k  =k (σ0k  /σ0k )ψk  ]  ≡ Jk (ξk , βk ) + Hkk  (11.61) J ∗ (ξk , βk ) =  1 2 ∞ k  =k where ψk and ψk  are evaluated for the respective resonance parameters of the resonances at Ek and Ek  . The evaluation of the second, overlap, term is quite complicated because of the statistical average over resonance parameters and level spacings, and useful approximations have been developed (Refs. 5 and 6). The multigroup cross section then consists of a term like Eq. (11.58) plus a negative overlap correction term: 0 g 1 σp  σxk = f xk Jk Dk spin states + σp   f  spin states k xk Hkk  Dk (11.62) Overlap Effects for Different Sequences The spacings of resonances belonging to different J -spin states in the same isotope or to two different isotopes are not correlated. The most important case is the overlap of resonances in a fissile isotope by resonances in a fertile isotope. Neglecting self-overlap, for the moment, the generalized J function for a fissile isotope with a resonance sequence at energies Ek overlapped by a fertile isotope with a resonance sequence at energies Ei is Jk∗ (ξk , βk ) = 1 2  ∞ −∞ ψk dxk βk + ψk + (σ0i /σ0k )ψi (11.63) Separating the generalized J -function into the normal J -function and an overlap term, as in Eq. (11.61), and making some further approximations, it is possible to write the effective multigroup resonance cross section as 0 g 1 σp  σxk = f spin states xk Jk Dk  1− i Ji Di  (11.64) It can be shown that for a single spin state in the fissile isotope, the flux correction factor f of Eq. (11.59) can be written  f  1− k Jk Dk  1− i Ji Di  (11.65) 11.5 Multiband Treatment of Spatially Dependent Self-Shielding so that the effective multigroup cross section for a fissile isotope overlapped by a fertile isotope can be written "  0 g1 xk Jk k Jk 1− (11.66) σxk = σp Dk Dk In this approximation, the effect of the resonance overlap is compensated by the corresponding change in flux that it produces, and the parameters of the overlapping fertile isotope sequence do not appear. With respect to Eq. (11.64), the effect of the overlapping sequence i enters via the 1/t in both the numerator and denominator and, to first order, these two effects cancel. Combining the self-overlap and different sequence overlap results, the effective multigroup cross section for a fissile resonance sequence k with self-overlap and with overlap by a fertile isotope resonance sequence i is   σp spin [( xk Jk / Dk ) + k  xk Hkk  / Dk ] 0 g1 states  (11.67) σxk = 1− k Jk / Dk spin states 11.5 Multiband Treatment of Spatially Dependent Self-Shielding Spatially Dependent Self-Shielding Approximate methods for calculating effective multigroup cross sections for resonance absorbing isotopes have been discussed in Sections 4.3 and 11.2. It was found that the approximate flux to be used in evaluating the resonance integral was of the form φ(E) ∼ fss (t (E)) × M(E), where M(E) is a spectral function with an energy dependence that would exist even in the absence of the resonance absorber and fss is a self-shielding factor that depends on the energy via the dependence of the total cross section on energy [e.g., fss ∼ 1/(tres (E) + sM ) and M(E) ∼ 1/E in the narrow resonance approximation for a homogeneous mixture given by Eq. (4.65)]. This same general form persists in approximate treatments of heterogeneous resonance absorbers, as may be seen from Eqs. (11.6) and (11.10). In the approximate treatment of heterogeneous resonance absorbers discussed in Section 11.2, the self-shielding factor, fss , and hence also the resulting multigroup cross section, was implicitly assumed to be spatially independent within the resonance absorber. However, simple physical considerations suggest that the self-shielding will be much more pronounced deep within a resonance absorber than on its surface, where the neutron spectrum is dominated by neutrons entering from the adjacent moderator and, furthermore, that the self-shielding near the surface will be different for neutrons entering from the moderator than for neutrons coming from deeper within the resonance absorber. Thus, even if accurate spatially constant multigroup cross sections that preserve volume-averaged reaction rates are obtained for a heterogeneous resonance absorber (e.g., a fuel pin), 433 434 11 Resonance Absorption the spatial dependence of reaction rates within the resonance absorber will not be calculated properly, which will introduce an error into calculations of fuel depletion, fission heating distribution, and so on. Even if the spatial multigroup flux distributions within the resonance absorber are calculated with multigroup transport theory, there will remain an inaccuracy in calculating the spatial distribution of reaction rates because these spatially independent volume-averaged multigroup cross sections were used instead of spatially dependent cross sections which take into account the spatial dependence of the self-shielding. The most straightforward way to solve this problem of spatially dependent within-group self-shielding might seem to be to further subdivide the normal multigroup structure (e.g., 20 to 50 groups) that would be used in a pin-cell transport calculation (see Section 14.4) into ultrafine groups in the resonance energy region. If the ultrafine groups could be sufficiently numerous that the within-group self-shielding term was almost unity (i.e., such that the variation of the cross section within any ultrafine group was small), the ultrafine-group cross sections could be accurately calculated as described previously, and the spatial self-shielding effect on the normal multigroup level would be calculated on the ultrafine-group level. However, this procedure is impractical except for special cases because each ultrafine-group width would have to be narrow compared with the width of the resonances at that energy, resulting in an enormous number of ultrafine groups to span the resonance energy region. This approach would not work at all for the unresolved resonances, of course. Multiband Theory In the multiband method, each normal group is further subdivided, not into finer energy intervals as in the ultrafine-group method discussed above, but into intervals of the total cross section magnitude which span the variation in the total cross section within the normal group. The multiband equations are derived by an extension to the derivation of the multigroup equations. Starting with the energydependent transport equation (with scattering and fission included in a general transfer function s ),  · ∇(r, E, ) + t (r, E)(r, E, )  ∞  4π     dE  d s r, E  → E,  →   r, E  ,  = 0 (11.68) 0 the normal multigroup equations are formally derived by integrating over the energy interval Eg ≤ E ≤ Eg−1 : g  · ∇g (r, ) + t (r, )g (r, ) = G   g  =1 0 4π g  →g  d s  r,  →  g  (r, ), g = 1, . . . , G (11.69) 11.5 Multiband Treatment of Spatially Dependent Self-Shielding Fig. 11.6 The Heaviside function Hgb=3 for the third band in a four-band representation of the total cross section (Hsb=3 = 1.0 in dark energy intervals and = 0.0 elsewhere). (From Ref. 11; used with permission of CRC Press.) where  g (r, ) ≡ Eg  g t (r, ) ≡ Eg−1 Eg−1 dE g (r, E, ) dE t (r, E)(r, E, )/g (r, ) Eg  r,  →   E  Eg−1 g −1    4   dE dE  s r, E  → E,  →   r, E  ,  g  r,  , = g  →g  s Eg Eg   g, g = 1, . . . , G (11.70) The multiband equations are formally derived by a similar process, but now with each group (g) energy interval subdivided into B cross-section bands (g, b), which span the range of total cross section in group g, as depicted in Fig. 11.6. Defining a Heaviside function Hgb (E) which is unity for those energy intervals for which the total cross section is within the band tb+1 ≥ t (E) ≥ tb and zero elsewhere, the multiband equations are derived by first multiplying Eq. (11.68) by Hgb (E) and then integrating over both the energy interval Eg ≤ E ≤ Eg−1 of group g and over the total cross-section range tb+1 ≥ t (E) ≥ tb of band b: gb  · ∇gb (r, ) + t (r, )gb (r, ) = G  B   4π    r,  →  g  b r,  , g  b →gb  d s g  =1 b =1 0 g, g  = 1, . . . , G; b, b = 1, . . . , B (11.71) 435 436 11 Resonance Absorption where the multiband parameters are given by  Eg−1  tgb+1 gb (r, ) ≡ dE dt∗ (E)Hgb (E)(r, E, ) Eg  gb t (r, ) ≡  Eg−1 tgb+1 dE Eg tgb  r,  →  ≡ g  b →gb  s tgb  dt∗ (E)Hgb (E)t (r, E)(r, E, )/gb (r, )  Eg−1 dE Eg  × Eg  −1 Eg  tb +1 tb dE   tb+1 tb dt∗ (E) (11.72)     dt∗ E  Hgb (E)Hg  b E        × s r, E  → E,  →   r, E  ,  /g  b r,  with the quantity t∗ normalized such that B   Etb+1 b=1 Etb dt∗ (E) = 1 (11.73) The normal multigroup quantities are related to the multiband quantities within the different groups as g (r, ) = B  gb (r, ) b=1 g t (r, ) = B  ' gb t (r)gb (r, ) b=1 B  (11.74) gb (r, ) b=1 Evaluation of Multiband Parameters Direct evaluation of the multiband parameters from the relationships above (actually, from the relationships that result when some discrete representation of the angular dependence is invoked) is possible in principle, but these relationships may be recast into a form that can make use of existing self-shielded multigroup libraries. The definition of the normal multigroup cross section for process x as an integral over energy can be exactly transformed into an integral over total cross section: g x ≡ = Eg−1 dE x (E)(E) Eg Eg−1 dE (E) Eg = Eg−1 dE x (E)M(E)fss (E) Eg Eg−1 dE M(E)fss (E) Eg  Eg−1 tb+1 dE B dt∗ (E)Hgb (E)x (E)M(E)fss (E) b=1 tb Eg  Eg−1 tb+1 dE B dt∗ (E)Hgb (E)M(E)fss (E) b=1 tb Eg (11.75) 11.5 Multiband Treatment of Spatially Dependent Self-Shielding where use has been made of the approximate relationship φ(E) ∼ M(E)fss (E) discussed previously. Performing the integration over energy first and defining p   t∗  Eg−1 dE Hgb (E)M(E) Eg Eg−1 dE M(E) Eg ≡ x t∗  ≡ Eg−1 Eg Eg−1 Eg dE Hgb (E)M(E) (11.76) (11.77) dE M(E)Hgb (E) leads to the equivalent definition of the normal multigroup cross section: B g x tb+1 dt∗ x (t∗ )fss (t∗ )p(t∗ ) b=1 tb B tb+1 dt∗ fss (t∗ )p(t∗ ) b=1 tb = (11.78) in terms of the total cross-section probability distribution function p(t∗ ), defined such that p(t∗ )dt∗ is just the normalized probability of the total cross section being within dt∗ of t∗ within the energy interval Eg ≤ E ≤ Eg−1 . For the practical evaluation of Eq. (11.78), average values of the cross sections xb for process x in each band b are used to replace the integrals with quadratures: g B x = gb ∗ b=1 x fss (tb )Pb ∗ b=1 fss (tb )Pb B (11.79) where  Pb ≡ tb+1 tb   p t∗ dt∗ (11.80) is the band weight. The computational advantage of this approach relative to a direct quadrature approximation of the second form of Eq. (11.75) is that x (t∗ ) is generally a much smoother function than is x (E) over the resonance energy range, so it is much easier to define an appropriate quadrature. Once the total crosssection probability distribution is evaluated, integrals involving this distribution may be performed quite accurately and efficiently. Calculation of Multiband Parameters Although it would be most straightforward to choose the band structure (t ) a priori and just evaluate the Pb and the various tb , it is more common to use a moments method to calculate these multiband parameters to reproduce the results obtained using certain limiting forms for the self-shielding. Using a generalized self-shielding factor of the form fss (E) = 1 [t (E) + 0 ]n (11.81) 437 438 11 Resonance Absorption the band parameters can be calculated by requiring that the multiband expression agrees with the known results for various values of 0 and n. As an example, for two bands, there are two weighting parameters (P1 , P2 ) and for each reaction g1 g2 process x two group-band cross sections (x , x ) in each group. A normal multigroup processing code will provide the unshielded [fss = 1 = 1/(t + 0 )0 ], the totally self-shielded flux-weighted [fss = 1/(t + 0 )1 ], and sometimes the totally self-shielded current-weighted [fss = 1/(t + 0 )2 ] values of the various cross secg g g tions in group g, (x )0 , (x )1 , and (x )2 , respectively. Requiring that Eq. (11.79) yield these three values of the total cross section and realizing that P1 + P2 = 1 yields four equations from which the band parameters can be calculated. It is necessary to introduce the ordering t1 < t2 in order to obtain a unique solution, since the two bands are otherwise indistinguishable. The solutions are P1/2 = 1 ± δ, 2 t1/2 = 1 A±B (11.82) where g δ≡ 1 − A(t )1 g 2B(t )1 A≡ 1 (t )0 − (t )2 g g g 2(t )2 (t )0 − (t )1 g B2 ≡ 1 g g (t )0 (t )1 g (11.83)  g  g  g 1 − 2A t 1 + t 0 t 1 A2 g1 g2 Having thus determined (P1 , P2 , t , and t ), the group-band cross sections g1 g2 for the individual processes x, (x , x ) can then be determined by requiring that Eq. (11.79) yield the unshielded [fss = 1 = 1/(t + 0 )0 ] and the totally shielded g flux-weighted [fss = 1/(t + 0 )1 ] cross sections (x )0 and (x )1 , respectively, which yields g2/1 x  g C = x 0 ± , P1 C=  g  g  P1 /tg1 + P2 /tg2 x 0 − x 1 g1 g2 1/t − 1/t (11.84) This general procedure may be extended to more bands. In practice, it has been found that two to four bands are sufficient. The scattering transfer rate from group g  to group g in the normal multigroup g theory depends on the scattering cross section in group g  , s , and on the transfer  probability, T g →g , that a neutron scattered in group g  will have final energy in  g  →g g = s T g →g ). The transfer probability does not depend on group g (i.e., s the scattering cross section in either group g  or group g. The usual procedure for constructing the group band g  b to group band gb scattering transfer rate is g g  b to replace s with s and to assume that the transfer probability from group band g  b to group band gb is just the group g  to group g transfer probability  g g  b →gb g  b g = s T g →g Pb ). Various times the weight Pb of band b in group g (i.e., s 11.6 Resonance Cross-Section Representations extensions of this definition of the group-band scattering transfer probability have been suggested, but its calculation remains heuristic. Interface Conditions The interface–boundary conditions for multiband transport theory remain somewhat heuristic as well. Clearly, the continuity of directional group-flux condition requires that B  gb (−s, ) ≡ g (−s, ) = g (s, ) ≡ b=1 B  gb (s, ) (11.85) b=1 where ±s refers to the + and − sides of the interface at r = s. The argument that the cross sections are not correlated across an interface between dissimilar media is used to justify the distribution of directional group fluxes crossing an interface from − to + side according to the weights on the + side: g+ g+ gb (+s, ) = Pb g (−s, ) = Pb B  gb (−s, ) (11.86) b =1 11.6 Resonance Cross-Section Representations∗ R-Matrix Representation The quantum mechanical representation of reaction cross sections is given most generally by R -matrix theory, in which the reaction cross section for any incident channel c and exit channel c is generally expressed in terms of the collision matrix Ucc : σcc = πλ2 gc |δcc − Ucc |2 (11.87) where gc and δcc are the statistical factor and the Kronecker delta, respectively. The unitary property of Ucc leads to the expression of the total cross section as a linear function of Ucc :  σcc = 2πλ2 gc (l − Re{Ucc }) (11.88) σt = c The collision matrix, in turn, can be expressed in terms of the resonance parameter matrix R according to Wigner and Eisenbud: +  −1 P 1/2 Ucc = exp[−i(φc + φc )] δcc + iP I − RL 0 R c ∗ 1/2 , P cc c This section was prepared with the extensive collaboration of R. N. Hwang. (11.89) 439 440 11 Resonance Absorption Table 11.3 Momentum-Dependent Factors for Various l-States Defined at Channel Radius rc (ρ = krc ) l=0 l=1 l=2 l=3 Pl ρ ρ3 1 + ρ2 ρ5 9 + 3ρ 2 + ρ 4 225 + 45ρ 2 + 6ρ 4 + ρ 6 Sl 0 −1 1 + ρ2 −(18 + 3ρ 2 ) 9 + 3ρ 2 + ρ 4 −(675 + 90ρ 2 + 6ρ 4 ) 225 + 45ρ 2 + 6ρ 4 + ρ 6 φl ρ ρ − tan−1 ρ Factors ρ − tan−1 3ρ 3 − ρ2 ρ7 ρ − tan−1 ρ(15 − ρ 2 ) 15 − 6ρ 2 Source: Data from Ref. 12; used with permission of Institute for Nuclear Research and Nuclear Energy, Sofia. where Rcc =  γλc γλc Eλ − E (11.90) λ is a real symmetric matrix and L0cc = (Sc − Bc + iPc )δcc (11.91) The energy-independent parameters Eλ , γλc , and Bc denote the R -matrix state, reduced width amplitude and arbitrary boundary parameters, respectively. Of all parameters given above, φc , Sc , and Pc are momentum dependent for the elastic scattering channels only. φc , the hard-sphere phase shift factor, is related directly to the argument of the outgoing wavefunction at the channel radius, whereas Sc , the shift factor, and Pc , the penetration factor, reflect the real and imaginary parts of its logarithmic derivative, respectively, as defined in Table 11.3. These quantities, along with the matrix R , specify the explicit energy dependence of the cross section. It should be noted that the matrix R is primarily responsible for the sharp rise in cross section at the energy near the resonance energy, Eλ , while the other energydependent quantities are relatively smooth by comparison. All energy-independent quantities given above are, in principle, determined by the fitting of the experimental data. An alternative expression for the collision matrix is the equivalent level matrix representation derived by Wigner, which can provide more analytical insight for the discussions to follow. It is given as   Ucc = e−i(φc +φc ) δcc + i λ,μ  1/2 1/2 λc Aλμ μc (11.92) 11.6 Resonance Cross-Section Representations where the level matrix A is defined as   −1  A λc = (Eλ − E)δλμ − γλc L0c γμc , 1/2 λc = ) 2Pc γλc (11.93) c This expression provides a clearer picture of the explicit energy dependence of the collision matrix. Practical Formulations Although the formal R -matrix representation is rigorous on theoretical grounds, it is quite obvious that simplifications are required before its deployment as the basis for nuclear data evaluations and subsequent use in reactor applications. In the current ENDF/B format, four major formalisms pertinent to the treatment of the resonance absorption are allowed: the single-level Breit–Wigner (SLBW), multilevel Breit–Wigner (MLBW), Adler–Adler (AA), and Reich–Moore (RM) formalisms. These formalisms are based on approximations of the formal R-matrix theory to various degrees of sophistication. With exception of the Reich–Moore formalism, all these formalisms exhibit a similar form as a function of energy and can be considered as the consequences of various approximations of the Wigner level matrix. It is convenient to cast them into the pole expansion form in either the energy or momentum domain (k-plane). A simple generic form that is widely used in reactor physics is √  1  (x) −i E Re ρl,J,λ (energy domain) σx = E dλ − E l,J = λ   (x)  ρl,J,λ 1  −i −i Re √ −√ √ √ E 2 dλ − E dλ + E l,J λ (momentum domain) (11.94) where the superscript x denotes the type of reaction under consideration. The superscripts f , γ , and R will be used to denote fission, capture, and compound nucleus (or total resonance) cross sections, respectively. Physically, each term retains the general features of a Breit–Wigner resonance upon which the traditional resonance integral concept was based. The relationships between these pole and residue parameters and the traditional resonance parameters for three of the major formalism are tabulated in Table 11.4. The use of complex arithmetic here makes possible a direct comparison of these traditional formalisms to the rigorous pole (x) representation to be discussed later. ρl,J,λ , however, are different and depend on the approximations assumed. Single-Level Breit–Wigner Approximation (SLBW). The SLBW approximation represents the limiting case when the resonances are well separated from each other. Thus the level matrix A at a given E can be viewed as a matrix with only one element. In much of the previous discussion, the resonance integrals were also 441 442 11 Resonance Absorption Table 11.4 Poles and Residues for Traditional Formalism Formalism SLBW MLBW (x) Poles, dλ E0λ − i lλ 2 Same as above Residues, ρl,j,λ 1 √ ; x ∈ {f, γ } E 1 CgJ 2 nλ √ exp(−i2φl ); x ∈ R E Same as above if x ∈ {f, γ } 1 CgJ 2 nλ √ [exp(−i2φl ) + Wλ ]; x ∈ R E where  i nλ Wλ = (E − E  ) + i[( tλ + tλ )/2] λ λ  CgJ xλ nλ tλ /2 λ =λ Adler–Adler μλ − ivλ (x) (x) CgJ [Gλ + iHλ ]; x ∈ {f, γ } (x) CgJ exp(−i2φl )[Gλ + iH (x) ]; x ∈ R Source: Data from Ref. 12; used with permission of Institute for Nuclear Research and Nuclear Energy, Sofia. treated in this approximation. In reality, the resonance cross sections clearly cannot be taken as a disjoint set of isolated resonances in a rigorous treatment. Ambiguity can arise as to what constitutes the macroscopic cross sections in a detailed treatment of the neutron slowing-down problem over an energy span consisting of many resonances of more than one nuclide. For this reason, the single-level description used in practical applications such as that specified in the ENDF/B manual is often given in the context of Eq. (11.92) as a linear combination of Breit– Wigner terms supplemented by the tabulated pointwise smooth data so that the continuous nature of the cross sections, and thus the flux, can be preserved. Multilevel Breit–Wigner Approximation (MLBW). The MLBW approximation corresponds to the situation in which the inverse of the level matrix is taken to be diagonal. One constraint for SLBW and MLBW approximations of practical interest is that all parameters must be positive. It is worth noting that poles and residues are energy dependent, although in many applications √ they are taken to be energy independent. Otherwise, additional terms in the E-domain would result in all l > 0 sequences using SLBW and MLBW formalism. (x) It should be noted that, strictly speaking, the amplitude ρl,J,λ and the pole dλ are also energy dependent for the single-level Breit–Wigner and multilevel Breit– Wigner approximations if the explicit energy dependence of the penetration factor and the level-shift factor are considered for all l > 1 states. In reference to the 11.6 Resonance Cross-Section Representations ENDF/B manual, the amplitude of an individual resonance is proportional to the penetration factor, while the real part of dλ is identifiable as dλ = Eλ = Eλ + Sl (|Eλ |) − Sl (E) 2Pl (|Eλ |) nλ (|Eλ |) The latter is equivalent to assuming that the boundary parameter is set to be Bl = Sl (|Eλ |). Thus the rational function nature of Pl (ρ) and Sl (ρ) defined in Table 11.3 can lead to 2(l + 1) pole terms for each resonance in the momentum domain, instead of two terms. The absolute values of the additional 2l poles are generally large compared to those two, resulting from the line-shape function directly, so as to reflect the relatively smooth nature of the energy dependence of the penetration factor and the level shift factor. The inclusion of these secondary energy effects can readily be added within the context of the generalized pole representation to be described (Ref. 13). Adler–Adler Approximation (A–A). The diagonalization of the inverse level matrix A −1 leads directly to the pole expansion defined by Eq. (11.92). The Adler–Adler approximation is equivalent to the Kapur–Peierls representation, in which the poles and residues are assumed to be energy independent. In the context of the forgoing discussion, it is equivalent to assuming the energy independence of L0 of Eq. (11.91) when the inverse of A −1 is considered. The approximation is usually restricted to the s-wave sequences of the fissionable isotopes in the low-energy region, where the assumption is valid. Reich–Moore Formalism. For practical applications, the formal R -matrix representation is obviously difficult to use when many levels and channels are present. The problem has been simplified significantly by the method proposed by Reich and Moore. The only significant assumption required, in principle, is  γλc L0cc γμc = δλμ c∈γ  2 0 γλc Lc (11.95) c∈γ which utilizes the presence of the large number of capture channels and the random sign of γλc . It is consistent with the observed fact that the total capture width distribution is generally very narrow. If one partitions the collision matrix into a 2 × 2 block matrix arranged such that the upper and lower diagonal blocks consist of only noncapture and capture channels, respectively, and utilizes Eq. (11.95) as well as Wigner’s identity between the channel matrix and the level matrix, the collision matrix can be reduced to the order of m × m where m is the total number of noncapture channels. The reduced collision matrix remains in the same form except that the real matrix R is replaced by a complex matrix R  with elements  Rcc  =  λ γλc γλc Eλ − E − i λγ /2 (11.96) 443 444 11 Resonance Absorption The substitution of the reduced R -matrix into the original equation leads to the following general form of collision matrix expression for the Reich–Moore approximation in terms of the familiar notations commonly used in applications: Y )e−iϕ U (E) = e−iϕ (II + 2Y (11.97) where  −1 P 1/2 I − R L̂ L RP 1/2 = F −1 (II − iF FK Y = iP K)−1 − I  1/2  0 1/2 S P −1 R L , F = I − iŜ K = L0 (11.98) (11.99) S = S (E) − B . and Ŝ It should be noted that the traditional Reich–Moore representation currently specified by the ENDF/B manual was originally developed for applications in the relatively low energy regions. It is different from the general form given above beS is taken to be zero. The cause two additional assumptions were introduced. First, Ŝ rationale is based on the fact that lim St (E) = −l, implied by the rational functions listed in Table 11.3. Thus, by taking limE→0 Bc = −l, the quantity Ŝl = 0 and the level shift factor will not play a role in the low-energy region. Second, one elastic scattering channel is allowed in the channel matrix K . Although the assumption simplifies the computation required, the issue may still arise for nuclides with odd atomic weight, for which the multiple elastic channels still may play a role. All evaluated resonance data given in the ENDF/B-VI to date are based on these assumptions. One consequence of the Reich–Moore approximation is that the reduced collil is complex. For practical applications, sion matrix is no longer unitary because Rcc  this presents no problem since the total cross section can be preserved if the capture cross section is defined as σγ = σt −  σcc (11.100) c ∈γ / All parameters retain the physical as well as the statistical properties specified by the formal R -matrix theory. The order of the channel matrix is usually no greater than 3 × 3. Hence the method is attractive for data evaluations, and in fact, Reich– Moore parameters have become widely available in the new ENDF/B-VI data. However, unlike the other three formalisms, resonances defined by the Reich– Moore formalism can no longer be perceived in the context upon which the traditional resonance theory in reactor physics was based. The direct application of this formalism to reactor calculations not only requires the entry of excessive files of precomputed, numerically Doppler-broadened pointwise cross sections at various temperatures, but also renders useless many well-established methods based on the resonance integral concept. Hence there is strong motivation to seek remedies so that the newly released Reich–Moore parameters can be fully utilized within the framework of the existing methodologies. 11.6 Resonance Cross-Section Representations Generalization of the Pole Representation Although any given set of R -matrix parameters, including those in the Reich– Moore form, can be numerically converted into parameters of the Kapur–Peierls type, the parameters so obtained are implicitly energy dependent. With the exception of low-lying resonances of a few fissionable isotopes, such dependence is generally not negligible. Thus, from the practical point of view, the traditional pole expansion is not useful for most nuclides of interest. However, a desirable representation directly compatible with the traditional forms given by Eq. (11.94) can be derived if the pole expansion is cast into a somewhat different form. Rigorous Pole Representation. One attractive means to preserve the rigor of the R -matrix description of cross sections is to perform the pole expansion in the k-plane (or momentum domain). Such a representation is natural for the SLBW, MLBW, and Adler–Adler approximations. The theoretical justification of such a representation is based on the rationale that the collision matrix must be single valued and meromorphic in the momentum domain. Any function that exhibits such properties must be a rational function according to a well-known theorem in complex analysis. The √ rational function characteristics are quite apparent if one examines the explicit E-dependence of the collision matrix Ucc defined by Eq. (11.89), if the level matrix is expressed as the ratio of the cofactor and the determinant of its inverse. By substituting Sl and Pl into Eq. (11.89) or (11.92), the quantity Ucc is expressible in terms of a rational function of order 2(N + l), where N is the total number of resonances. This reflects the polynomial nature of the cofactor and the determinant of the inverse level matrix. Thus one obtains via partial fractions the similar pole representation for other approximations. A general expression that can be used with all cross-section representations is σt = σp +  M jj 1  −i (t) Re Rl,J,j,λ e−i2φ1 · (j )∗ √ E pλ − E l,J λ=1 j =1 (11.101) and similarly,  M jj −i 1  (x) Re Rl,J,j,λ (j )∗ √ σx = E p − E l,J λ=1 j =1 (11.102) λ (x) (j ) for the reaction cross section of process x, where Rl,J,j,λ and pλ (j )∗ are pole and residue, respectively. Note that the complex conjugate pλ is used here in order to cast the expressions into the form defined by Eq. (11.92). These equations can be viewed as the generalized pole expansion in which all parameters are truly energy independent and the energy dependence of the cross sections is specified explicitly by the rational terms alone. The indices M and jj depend on the type of resonance parameters and assumptions used to generate these pole parameters: 445 446 11 Resonance Absorption • Adler–Adler: M = N (total number of resonance); jj = 2. All pole parameters can be deduced directly via partial fractions. • SLBW and MLBW: M = N ; jj = 2 if penetration factor and level shift factor are taken to be energy independent, an assumption used in the traditional resonance integral approach. Otherwise, M = N ; jj = 2(l + 1) if all energy-dependent features are included. • Reich–Moore: M = N + l; jj = 2 for both scenarios with Ŝl (E) = 0 and Ŝl (E) = 0 if Eq. (11.98) is used. Another possible scenario is to keep the traditional expression specified by the ENDF/B manual intact; that is, let F = I in Eq. (11.98), but to introduce the level shift factor via replacing the resonance energy Eλ with Eλ = Eλ + Sl (|Eλ |) − Sl (E) 2Pl (|Eλ |) nλ (|Eλ |) the same as for the SLBW and MLBW approximations. The resulting number of poles becomes M = N ; jj = 2(l + 1). By comparing Eqs. (11.94) and (11.101), one is led to the following observations: (1) For the s-wave, both the rigorous pole representation and the traditional formalism consist of an identical number of terms with the same functional form in the momentum domain. In particular, the Adler–Adler formalism for the s-wave can be considered as the special case of the former when pλ(1) = −pλ(2) and (x) (x) Rl,J,1,λ = Rl,J,2,λ . (2) For higher angular momentum states, Eq. (11.101) consists of either 2l or 2lN more terms than those defined by Eq. (11.94). The difference, however, is only superficial. The same number of terms would have resulted if the detailed energy dependence of the penetration factor and the shift factor had been included in Eq. (11.94). Equations (11.101) and (11.102) provide the basis whereby any given set of R -matrix parameters, in principle, can be converted into pole parameters, although it may not be an easy task in practice. The recent availability of R -matrix parameters in the Reich–Moore form greatly alleviates the numerical difficulties for such a conversion process. One obvious disadvantage of this method is that two to as many as 2(l + 1) terms must be considered for each resonance if the cross section is to be evaluated in the momentum domain. This is obviously undesirable for computing efficiency, storage requirement, and its amenability to the existing codes for reactor calculations. Simplified Pole Representation. The Mxjj poles for a given l and J defined in Eqs. (11.101) and (11.102) can be divided into two distinct classes. There are 2N s-wavelike poles with sharp peaks and distinct spacings, while the remaining 2l or 2lN poles are closely spaced and are characterized by their extremely large imaginary components (or widths). In fact, the contributions of the latter to the sums 11.6 Resonance Cross-Section Representations are practically without any resonance-like fluctuations, as if they were a smooth constituent. On the other hand, the s-wavelike poles always appear in pairs with opposite signs but not necessarily with the same magnitude. These characteristics provide a valuable basis for simplification. (x) √ Let ql ( E) denote the contributions from those additional 2l or 2lN terms involving poles with giant width. Equation (11.102) can be cast into the same form as that of Humblet–Rosenfeld:  √  N   √  1  2(−i) E (x) Rl,J,1,λ − (1)∗ + sl(x) E Re σx = 2 E (pλ ) − E l J λ=1 * √ √  + ql(x) E · δl , E>0 (11.103) where (x)  sl   (x) (x) N √   Rl,J,1,λ (−i) Rl,J,2,λ (−i) E = √ − √ (2)∗ (1)∗ − E −pλ − E J λ=1 pλ (11.104) √ and δ0 = 0 and δl = 1 for l > 0. The quantity sl(x) ( E), physically signifying a ∗ ∗ measure of deviation from the Adler–Adler limit of pλ(2) = −pλ(1) , is usually not only small in magnitude but also smooth as√a function of√energy in the region (x) (x) where the calculations take place. Thus sl ( E) and ql ( E) can be construed as the energy-dependent smooth term in the Humblet–Rosenfeld representation with its energy dependence explicitly specified. Hence, for a given range of practical interest, the rigorous pole representation can be viewed as a combination of fluctuating terms, consisting of N poles with (1) Re{pλ } > 0 expressed in the energy domain consistent with the traditional formalism and two nonfluctuating (or background) terms attributed to the tails of √ outlying poles (in reference to the domain E > 0, where calculations are to take place) with negative real component and the poles with extremely large width (or (j ) |Im{pλ }|) for l > 0 states, respectively. The striking behavior of the fluctuating and nonfluctuating components have been confirmed in recent calculations for all major nuclei specified by the Reich–Moore parameters in the ENDF/B VI files. The smooth behavior of these terms clearly suggests that their energy dependence can be reproduced by other, simpler functions within the finite interval of practical interest. It is well known in numerical analysis that the rational functions are best suited to approximate a well-behaved function within√a finite range.√Hence (x) (x) the obvious choice is to set the approximate functions ŝl ( E) and q̂l ( E) to be rational functions of arbitrary order. Mathematically, they can be viewed as the (x) √ (x) √ ( E) and q̂ ( E) within doanalytic continuations of the original functions ŝ l l √ main E > 0. One attractive feature of the method proposed is that the rational functions so obtained can be again expressed in the form of a pole expansion via 447 448 11 Resonance Absorption partial fraction, that is, ŝl(x) √ NN (x) √  PMM ( E)  rλ (−i) E = √ = √ QNN ( E) λ=1 αλ∗ − E (11.105) NN (x) √   bλ (−i) E = √ ξ∗ − E λ=1 λ (11.106) q̂l(x) if NN > MM. αλ∗ and ξλ∗ are the poles of the fitted rational functions (i.e., the ratio √ √ of the two low-order polynomials) for ŝl(x) ( E) and q̂l(x) ( E), respectively, while (x) (x) rλ and ξλ are their corresponding residues. Doppler Broadening of the Generalized Pole Representation Either one of two approaches are usually taken, depending on the accuracy required. The rigorous broadening must be carried out in the momentum domain, whereas the simplified broadening is based on the approximate kernel in the energy domain. In the following discussions, the Doppler-broadened cross section based on the traditional formalism and the generalized pole representation are compared. Exact Doppler Broadening. The Maxwell–Boltzmann kernel can be expressed rigorously as √ √  S E, E  = √ √   √  ( E − E  )2 E exp − (πE)1/2 m 2m √   √ ( E + E  )2 − exp − 2m (11.107) where $ m = kT = Doppler width in momentum space A (11.108) √ √ The Doppler broadening of E  σx ( E  ) defined by Eq. (11.94) in momentum space and that defined by Eq. (11.102) lead immediately to: • Traditional representation: σx √   (x) √   √ N ρl,J,λ π E − ζλ 1  W E, T = Re E 2 m 2m  l,J λ=1 −W ∗ √ E + ζλ∗ m  (11.109) 11.6 Resonance Cross-Section Representations • Generalized pole representation: σx √  √ √ 2 N  (j )∗   E − pλ π 1  (x) E, T = Re Rl,J,λ,j W E m m  J λ=1 j =1 l √  + ŝl(x) E, T  (x)  + ql √  E δl * (11.110) (x) √ where q̂l ( E) is insensitive to Doppler broadening and ŝl(x) √ NN   (x) E, T = rk k=1 √  √ E − αk∗ π W m m (11.111) and W (z) is the complex probability integral and is directly related to the usual Doppler-broadened line shape functions via the relation  ∞ e−t dt −∞ z − t √ ψ(x, y) + iχ(x, y) = πyW (z) W (z) = i π 2 (11.112) (11.113) and z = x + iy. In the single-level limit, Eq. (11.109) is equivalent to the generalized form of the exact Doppler broadening defined by Ishiguro. Thus, except for the superficial (x) √ difference leading to the smooth term q̂l ( E)δl , Eqs. (11.109) and (11.110) have the same functional form but are characterized by different parameters. From a practical point of view, the computational requirements for these equations are expected to be comparable if the smooth term is replaced by the approximation defined by Eqs. (11.100) and (11.101). Approximate Doppler Broadening. For most of the existing codes based on the traditional formalism, the Doppler broadening is generally based on the approximate Gauss kernel defined in the energy domain   1 (Eλ − E)2 (11.114) M(Eλ − E) = √ exp E πE √ where E = 4kT E/A is the Doppler width in the energy domain. The validity of such an approximation requires the criterion E m . It has been well established that the use of the Gauss kernel in the energy domain is generally satisfactory for E > 1 eV. The Doppler-broadened cross sections become: • Traditional formalism: √    N √ (x) π 1  E − dλ σx (E, T ) = Re Eρl,J,λ W E E E l,J λ=1 (11.115) 449 450 11 Resonance Absorption • Generalized pole representation after simplification:  √   N √  π 1  E − ελ (x) σx (E, T ) = Re 2 ERl,J,1,λ W E E E l J λ=1 * √  (x)  (x) E, T + q̂l (E)δl + ŝl (11.116) where ∗ ελ = pλ(1) 2 (11.117) References 1 W. Rothenstein and M. Segev, “Unit Cell Calculations,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations I, CRC Press, Boca Raton, FL (1986). 2 J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976), Chaps. 2 and 8. 3 A. F. Henry, Nuclear-Reactor Analysis, MIT Press, Cambridge, MA (1975), Chap. 5. 4 G. I. Bell, Nuclear Reactor Theory, Van Nostrand Reinhold, New York (1970), Chap. 8. 5 H. H. Hummel and D. Okrent, Reactivity Coefficients in Large Fast Power Reactors, American Nuclear Society, La Grange Park, IL (1970). 6 R. B. Nicholson and E. A. Fischer, “The Doppler Effect in Fast Reactors,” in Advances in Nuclear Science and Technology, Academic Press, New York (1968). 7 R. N. Hwang, “Doppler Effect Calculations with Interference Corrections,” Nucl. Sci. Eng. 21, 523 (1965). 8 L. W. Nordheim, “The Doppler Coefficient,” in T. J. Thompson and J. G. Beckerley, eds., The Technology of Nuclear Reactor Safety, MIT Press, Cambridge, MA (1964). 9 A. Sauer, “Approximate Escape Probabilities,” Nucl. Sci. Eng. 16, 329 (1963). 10 L. Dresner, Resonance Absorption in Nuclear Reactors, Pergamon Press, Elmsford, NY (1960). 11 D. E. Cullen, “Nuclear Cross Section Preparation,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations I, CRC Press, Boca Raton, FL (1986). 12 R. N. Hwang, “An Overview of the Current Status of Resonance Theory in Reactor Physics Applications,” in W. Audrejtscheff and D. Elenkov, eds., Proc. 11th Int. School Nuclear Physics, Neutron Physics, and Nuclear Energy, Institute for Nuclear Research and Nuclear Energy, Sofia, Bulgaria (1993). 13 C. Jammes and R. N. Hwang, “Conversion of Single- and Multi-Level Breit–Wigner Resonance Parameters to Pole Representation Parameters,” Nucl. Sci. Eng. 134, 37 (2000). 14 A. G. Croff, ORIGEN2: A Revised and Updated Version of the Oak Ridge Isotope Generation and Depletion Code, ORNL-5621, Oak Ridge National Laboratory, Oak Ridge, TN (1980). Problems Problems 11.1. Carry through the steps indicated to derive the narrow resonance approximation flux of Eq. (11.6) and the wide resonance approximation flux of Eq. (11.10). 11.2. A fuel assembly in a reactor consists of a uniform array of fuel pins 1 cm in diameter set in parallel rows such that the center-to-center separation between adjacent rows is 3 cm in both ways. The fuel is 2.8% enriched UO2 operating at 800◦ C. The moderator is H2 O at 0.85 g/cm3 . Calculate the heterogeneous resonance integral in the narrow resonance and the wide resonance approximations for the 238 U resonance at 36.8 eV, in the isolated fuel rod approximation. 11.3. Repeat Problem 11.2 taking into account the Dancoff correction for a closely packed square lattice. 11.4. Assume that the fuel and moderator in Problem 11.2 are mixed homogeneously together. Calculate the homogeneous resonance integral for the 238 U resonance at 36.8 eV at 800◦ C. 11.5. Calculate the contribution of the 238 U resonance at 36.8 eV to the multigroup cross section of a group with Eg = 10 eV and Eg−1 = 100 eV, for Problems 11.2 to 11.4. 11.6. Repeat Problem 11.2 for D2 O moderator. Calculate the contribution to the multigroup cross section over Eg = 100 eV to Eg−1 = 100 eV. 11.7. Compare the calculated escape probability for a fuel plate immersed in water for values of λ/ ls in the range 0.1 ≤ λ/ ls ≤ 10.0, using the exact expression of Eq. (11.32) and the rational approximation of Eq. (11.43). 11.8. Derive the first-flight escape probability for neutrons created uniformly over a slab of thickness a given by Eq. (11.32). 11.9. Write a code to calculate the unresolved 238 U multigroup capture cross section of Eq. (11.62) for a group extending from Eg = 1 keV to Eg−1 = 10 keV. 11.10. Evaluate the two-band group absorption cross section for a group extending from Eg = 10 eV to Eg−1 = 100 eV for a nuclide for which the absorption and total cross sections are a1 = 0.4 cm−1 and t1 = 0.5 cm−1 from 10 ≤ E ≤ 50 eV and are a2 = 0.6 cm−1 and t2 = 0.8 cm−1 from 50 eV ≤ E ≤ 100 eV. 451 453 12 Neutron Thermalization The thermalization of neutrons is complicated, relative to neutron slowing down, by the fact that the thermal energies of the target nuclei are comparable to the neutron energies, so that a neutron may gain or lose energy in a scattering collision, and by the fact that the nuclei are generally bound in a lattice or molecular structure, which considerably complicates both the calculation of a scattering cross section and the scattering kinematics. The objectives of neutron thermalization theory are first to calculate cross sections that characterize the thermal neutron scattering and energy transfer and then to use these cross sections in calculation of the thermal neutron spectrum. In this chapter we consider some approximate models of neutron thermalization that provide useful physical insights, discuss the construction of thermal neutron scattering kernels, and then discuss the analytical and numerical calculation of the neutron thermal energy spectrum in homogeneous media and heterogeneous reactor lattices. 12.1 Double Differential Scattering Cross Section for Thermal Neutrons A quantum mechanical analysis of the scattering event in which an incident neutron interacts with an assembly of target nuclei of atomic mass A leads to an expression of the differential scattering cross section for scattering from energy E  to energy E and from direction  to direction :   b s E  → E, μ0 = 4πkT  E E 1/2   β S(α, β) exp − 2 (12.1) where μ0 =  · , σb is the scattering cross section for a neutron incident on a bound nucleus, √ E  + E − 2μ0 E  E , α≡ AkT β≡ E −E kT (12.2) 454 12 Neutron Thermalization and S(α, β) is a scattering function that depends in a complicated way on the detailed dynamics and structure of the scattering material. Hence  A+1 b = Nσb = N A  2 σf = A+1 A 2 f is the bound atom cross section, given in terms of the free atom cross section, f , and the ratio of scattering to neutron masses, A. 12.2 Neutron Scattering from a Monatomic Maxwellian Gas Differential Scattering Cross Section The simplest, but by no means simple, model of neutron thermalization is for neutrons scattering from a monatomic gas of unbound nuclei distributed in energy according to a Maxwellian distribution, for which the scattering function is   1 α2 + β 2 S(α, β) = exp − 4α 2(πα)1/2 (12.3) which yields for the differential scattering cross section 1/2         A 1 2 f E 1/2 s E → E, μ0 = 1 + A 4π E  2πkT h̄2 κ 2     h̄2 κ 2 2 A ε − × exp − 2A 2kT h̄2 κ 2 (12.4) where A is the atomic mass (amu) of the target nuclei, σf is the total scattering cross section for a neutron incident on a free nucleus, and ε ≡ E  − E, √   h̄2 κ 2 = 2m E  + E − 2μ0 E  E (12.5) Integrating Eq. (12.4) over −1 ≤ μ0 ≤ 1 and using the relationship between μ0 and (E, E  ) for elastic scattering,   $ $ E E 1 μ0 = (A + 1) − (A − 1) 2 E E yields for the zeroth Legendre moment of the scattering transfer function,   s0 E  → E    f θ 2 E E = exp − 2E  kT kT (12.6)  % $ 12.2 Neutron Scattering from a Monatomic Maxwellian Gas $ & % $ $ & E E E E −η ± erf θ +η kT kT kT kT % $ & % $ &* $ $ E E E E + erf θ −η ∓ erf θ +η kT kT kT kT × erf θ (12.7) where erf(x) is the error function and A+1 θ≡ √ , 2 A η≡ A−1 √ 2 A (12.8) The upper signs are used when E  < E, and the lower signs are used when E  > E. Cold Target Limit In the limit T → 0, Eq. (12.7) reduces to the scattering transfer function for elastic scattering from a stationary target: ⎧  2  ⎨ f (E ) (A + 1) , E < E  < E    s0 E → E = (12.9) 4A α ⎩ E 0, otherwise which was used in Chapter 10 for the treatment of neutron slowing down in the energy range well above thermal where the nuclear motion is negligible compared to the neutron motion. Free-Hydrogen (Proton) Gas Model Hydrogen, in the form of water molecules, is a dominant nuclear species for neutron thermalization in water-cooled nuclear reactors. The free-hydrogen gas model neglects the fact that hydrogen is present in molecular form and treats the thermalization of neutrons by a gas of free protons (hydrogen nuclei). For scattering from hydrogen nuclei (A = 1), the zeroth Legendre moment of the scattering energy transfer function of Eq. (12.7) simplifies to ⎧ H      %$  & f (E ) E −E E ⎪ ⎪ ⎪ exp erf , E < E ⎪  ⎨ E kT kT    (12.10) s0 E → E = %$ & ⎪ fH (E  ) ⎪ E ⎪  ⎪ erf , E >E ⎩ E kT Radkowsky Model for Scattering from H2 O The Radkowsky model uses the hydrogen gas model of Eq. (12.10) to describe the zeroth Legendre moment of the scattering transfer function, and uses     s1 E  → E = s (E )μ̄0 (E )δ E − E  (12.11) 455 456 12 Neutron Thermalization to describe the first Legendre moment. The bound-state cross section, b , is related to the free-state cross section, f , by  s = b = f  2 Amol [(A + 1)/A]2 = 4 f Amol + 1 (1 + 1/Amol )2 (12.12) where A is the mass of an atom bound in a molecule of mass Amol and has been set to unity in the last step to represent the hydrogen bound in water molecules. Application of this model is implemented by adjustment of Amol until b agrees with an experimentally measured scattering cross section, as a function of energy. Then μ0 = 2/(3Amol ) is used to calculate tr (E ) = b (E )[1 − μ0 (E )]. Heavy Gas Model In the limit of large A, the scattering transfer function of Eq. (12.7) can be expanded in powers of A−1 . When only the leading term is retained, the result is   1/2     E +E E s0 E  → E = f δ E − E  +  E A     × −δ  E − E  + kT δ  E − E  (12.13) where δ  and δ  are the first and second derivatives of the delta function with respect to x. Integrating this expression over E defines the total scattering cross section in this model:   kT (12.14) s0 (E ) = f 1 + 2AE Using Eq. (12.13) to evaluate the scatter-in integral yields  ∞     dE  s0 E  → E φ E  0 =   2f dφ(E ) d 2 φ(E ) + E kT E + φ(E ) + s0 (E )φ(E ) A dE dE 2 (12.15) when the property of the derivatives of the delta functions,  dx f (x) n dn n d f (x = a) δ(x − a) = (−1) dx n dx n (12.16) is taken into account. Substituting this expression for the scatter-in integral into the neutron balance equation  t (E )φ(E ) = 0 ∞     s0 E  → E φ E  dE  (12.17) 12.3 Thermal Neutron Scattering from Bound Nuclei yields   f d 2 φ(E ) dφ(E ) a (E )φ(E ) = 2 EkT + φ(E ) +E A dE dE 2 (12.18) which is the heavy gas model for the thermal neutron spectrum, φ(E ). 12.3 Thermal Neutron Scattering from Bound Nuclei The quantum mechanical theory for neutron scattering from a system of bound nuclei leads to an expression for the double differential scattering function for scattering from energy E  to energy E and from direction  to direction : $  ∞ E 1 ei(κ·r−εt/h̄) G(r, t) dr dt E  2π −∞ $  ∞ inc E 1 ei(κ·r−εt/h̄) Gs (r, t) dr dt + 4π h̄ E  2π −∞ (12.19)  coh  s E  → E,  →  = 4π h̄ where h̄ is the reduced Planck’s constant, h̄κ = m(ν  − ν) is the neutron momentum exchange vector, ε = E  − E is the neutron energy change, and coh and inc are the bound coherent and incoherent macroscopic cross sections. The coherent scattering takes into account the interference of neutrons scattering from different nuclei, which is important when the neutron wavelength λ (cm) = 2.86 × 10−9 /[E(eV)]1/2 is comparable with the spacing between atoms in a crystal or molecule, and the incoherent scattering takes into account the scattering of neutrons from isolated nuclei. Pair Distribution Functions and Scattering Functions The functions G(r, t) and Gs (r, t) are pair distribution functions. If a scattering target atom is at the origin r = 0 at time t = 0, then G(r, t) is the probability that an atom will be present within unit volume dr about r at time t . G(r, t) = Gs (r, t) + Gd (r, t), where Gs (r, t) is the probability that the atom present in dr about r at time t is the same atom that was present at r = 0 at time t = 0, and Gd (r, t) is the probability that a different atom is present in dr about r at time t . The integrals involving the pair distribution functions in Eq. (12.19) are defined as the scattering functions S(κ, G) = 1 2π  ∞ −∞  ei(κ·r−εt/h̄) G(r, t) dr dt with a similar definition for Ss in terms of Gs . (12.20) 457 458 12 Neutron Thermalization The principle of detailed balance requires that $ E coh S(−κ, −ε) + inc Ss (−κ, −ε) E $   E = M E , T coh S(κ, ε) + inc Ss (κ, ε) E M(E, T ) (12.21) be satisfied separately for the incoherent and coherent contributions. Recalling that M(E, T ) = √   E 2π E exp − kT (πkT )3/2 (12.22) this detailed balance requirement may be written e−ε/2kT S(κ, ε) = eε/2kT S(−κ, −ε) (12.23) with a similar requirement for Ss , which requires that both S(κ, ε) and Ss (κ, ε) be even functions of ε. In many scattering models, S(κ, ε) is a function of κ 2 , and an equivalent scattering function can be defined: S(α, β) = kT e β/2   1 α2 + β 2 S(κ, ε) = exp − 4α 2(πα)1/2 (12.24) where α and β are defined by Eq. (12.2). Using this scattering function, the double differential scattering transfer function can be represented as   s E  → E,  →  = 1 4π h̄kT $ E −β/2 e coh S(α, β) + inc Ss (α, β) E (12.25) Intermediate Scattering Functions An equivalent representation of the double differential scattering transfer function is   s E  → E,  →  $  ∞ 1 E 1 = e−iεt/h̄ χcoh (κ, t) + χinc (κ, t) dt 4π h̄ E  2π −∞ (12.26) where the intermediate scattering functions are defined:  χcoh (κ, t) ≡ eiκ·r G(r, t)dr (12.27)  χinc (κ, t) ≡ e iκ·r Gs (r, t)dr 12.3 Thermal Neutron Scattering from Bound Nuclei Incoherent Approximation The interference effects, which are contained entirely in the pair distribution function Gd , are important in elastic scattering, but are less important in inelastic scattering, particularly in liquids and polycrystalline solids. This observation leads to the incoherent approximation, obtained by setting Gd = 0 in Eq. (12.19):  coh + inc  s E  → E,  →  = 4π h̄ $ E 1 E  2π  ∞ −∞ ei(κ·r−εt/h̄) Gs (r, t) dr dt (12.28) Note that this approximation retains the coherent scattering cross section, coh . With the incoherent approximation, S(α, β) = Ss (α, β) in Eq. (12.25) and χcoh = χinc in Eq. (12.26). Gaussian Representation of Scattering In the incoherent approximation, the intermediate scattering function has a Gaussian form in many important cases:   1 2 χs (κ, t) = exp − κ (t) (12.29) 2 where   ∞    cos ωt dω ω i = − g(ω) coth  t+ 2T 2kT sinh(ω/2kT ) ω 0 (12.30) The properties of a particular moderator are represented in the frequency distribution function, g(ω). For crystals, g(ω) is a true phonon frequency spectrum. For liquids and molecules, g(ω) contains the diffusive and vibrational characteristics and may be temperature dependent. Representative frequency distribution functions are: Perfect gas: Debye crystal: Einstein crystal: Molecular liquid: 1 δ(ω) A 3ω2 g(ω) = Aθ 3 1 g(ω) = δ(ω − θ) A  1 γi δ(ω − ωi ) g(ω) = fd (ω) + A g(ω) = (12.31) i where θ = hνm /2πk ≡ h̄νm /k, νm is the maximum allowed frequency, fd (ω) describes the frequency distribution associated with diffusive motion of the molecule, and the γi δ(ω − ωi ) describe the internal vibrations with frequency ωi of the individual atoms of which the molecular fluid is composed. 459 460 12 Neutron Thermalization The corresponding scattering functions in the Gaussian representation are  ∞ 1 Ss (α, β) = dt eiβt exp −αW 2 (t) (12.32) 2π −∞ where  W (t) = 2 ∞ −∞ g(β)[cosh(β/2) − cos(kT βt/h̄)]dβ β sinh(β/2) (12.33) Measurement of the Scattering Function The scattering transfer function can be determined from Eq. (12.25), in the incoherent approximation, by measuring s (E  → E,  → ). For small values of neutron momentum transfer, κ 2 , and energy transfer, the exponential in Eq. (12.32) can be expanded to obtain a relation between the frequency distribution function and the measured Ss (α, β): f (β) = 2β sinh Ss (α, β) β lim 2 α→0 α (12.34) and noting that hω/2π = E  − E = βkT . Thus, by measuring the scattering double differential cross section for small momentum and energy transfer events, the scattering function Ss can be inferred and related to the frequency distribution. This enables the experimental determination of g(ω), which can then be extrapolated and used to calculate scattering transfer functions for larger energy and momentum transfers. Applications to Neutron Moderating Media The double-differential scattering transfer function for water has been calculated with a molecular liquid model in which the frequency distribution function is given by  1 1 δ(ω − ωi ) + 18 Ai 4 g(ω) = (12.35) i=2 where the first term represents the translational (diffusive) motion of free gas molecules, the second term represents hindered rotation (A2 = 2.32, hω/2π = 0.06 eV), and the third and fourth terms represent vibrational modes with (A3 = 5.84, hω/2π = 0.205 eV) and (A4 = 2.92, hω/2π = 0.481 eV). This Nelkin distribution function was used to evaluate the scattering function of Eq. (12.32), which was then used to evaluate the double differential scattering transfer function of Eq. (12.1). The results are compared with experimental measurements of the double differential scattering transfer function, for different incident neutron energies, in Fig. 12.1. Also shown are results calculated with the free-hydrogen gas model of Eq. (12.4). 12.3 Thermal Neutron Scattering from Bound Nuclei Fig. 12.1 Calculated and measured double differential scattering transfer functions in liquid water at various incident neutron energies. (From Ref. 3; used with permission of Wiley.) The phonon frequency spectrum for graphite, based on two slightly different models, is shown in Fig. 12.2. Specializing the incoherent approximation for the double differential scattering transfer function of Eq. (12.28) to a crystal lattice with cubic symmetry and harmonic interatomic forces yields   s E  → E,  →  $  b E 1 ∞ −iεt/h̄ = e 4π h̄ E  2 −∞  2  ∞   f (ω)e−h̄ω/2kT  −iωt h̄κ × exp − 1 dω dt e 2Am −∞ 2ω sinh(h̄ω/2kT ) (12.36) Using the Young–Koppel frequency distribution shown in Fig. 12.2 to evaluate Eq. (12.36) yields the inelastic cross section shown in Fig. 12.3. Adding to this the absorption cross section of graphite and an elastic scattering cross section calculated without making the incoherent approximation, the total calculated cross section for graphite is compared to measured values in Fig. 12.3. Here m is the neutron mass. 461 462 12 Neutron Thermalization Fig. 12.2 Phonon frequency distribution functions for graphite derived from two different models. (From Ref. 3; used with permission of Wiley.) 12.4 Calculation of the Thermal Neutron Spectra in Homogeneous Media Turning now to the calculation of the neutron energy spectrum, the neutron balance equation for thermal neutrons, neglecting leakage, is  [a (E ) + s (E )]v(E )n(E ) = ∞       dE  s0 E  → E v E  n E  (12.37) 0 The principle of detailed balance for a neutron distribution in equilibrium,       M E  , T s0 E  → E = M(E, T )s0 E → E  (12.38) where M(E, T ) is the Maxwellian neutron particle distribution at temperature T , M(E, T ) =   √ 2π E E exp − kT (πkT )3/2 (12.39) is quite important in developing solutions for the thermal neutron distribution. 12.4 Calculation of the Thermal Neutron Spectra in Homogeneous Media Fig. 12.3 Calculated and measured cross sections in graphite (GASKET and SUMMIT refer to codes). (From Ref. 3; used with permission of Wiley.) Wigner–Wilkins Proton Gas Model The zeroth Legendre moment of the scattering energy transfer function for neutron scattering from a free gas of hydrogen nuclei with a Maxwellian distribution is given by Eq. (12.10). It is convenient to define the dimensionless variable x = (E/kT )1/2 and to symmetrize the scattering transfer function: ( 1 S(x  → x) ≡ H f M(x  )  x s0 (x  → x) M(x) (12.40) 463 464 12 Neutron Thermalization In terms of the reduced density, n(x) χ(x) ≡ √ M(x) (12.41) Eq. (12.37) can be written    ∞ xs (x) a0 χ(x) = + S(x  → x)χ(x  )dx  f f 0 or more explicitly,    1 exp(−x 2 ) a0 χ(x) + x+ erf(x) + √ 2x f π  x    1 1 = 2 exp − x 2 χ(x  ) exp − (x  )2 erf(x  )dx  2 2 0      ∞ 1 2 1  2 + 2 exp x erf(x) exp − (x ) χ(x  )dx  2 2 x (12.42) (12.43) where erf(x) is the error function of argument x, 1/v absorption has been assumed, and a0 = a (v0 = 2200 m/s) is the 2200-m/s macroscopic absorption cross section. Equation (12.42) can be transformed into a second-order differential equation by defining a second-order differential operator which when applied to either erf(x) exp(x 2 /2) or exp(−x 2 /2) yields zero. Such an operator is L= d2 d + b(x) + a(x) dx dx 2 with a(x) = √ − π erf(x) , √ exp(−x 2 ) + x π erf(x) (12.44) b(x) = exp(−x 2 ) − x2 √ exp(−x 2 ) + πx erf(x) (12.45) When this operator is divided by  √  P (x) = exp −x 2 + πx erf(x) (12.46) and then applied to Eq. (12.42), the Wigner–Wilkins equation results:    1 d a0 d χ(x) V (x) + − dx P (x) dx f    4 a0 − √ χ(x) = 0 + W (x) V (x) + f π (12.47) where W (x) = x2 exp(−x 2 ) , − P (x) P 2 (x) V (x) = xs (x) (12.48) 12.4 Calculation of the Thermal Neutron Spectra in Homogeneous Media Appropriate low-energy boundary conditions can be deduced from setting x = 0 in Eq. (12.43), which leads to the low-energy boundary condition χ(0) = 0. The two solutions of Eq. (12.43) near x = 0 are a constant and a solution that varies like x; and only the latter can satisfy the boundary condition χ(0) = 0. The other boundary condition follows from the requirement that the flux take on the asymptotic 1/E ∼ 1/x 2 form from the slowing-down region at high energies in the thermal range. Defining   1 dμ(x) a0 χ(x), y(x) ≡ (12.49) μ(x) ≡ V (x) + f μ(x)P (x) dx Eq. (12.47) can be reduced to a Ricatti equation:   a0 −1 4 dy(x) − P (x)y 2 (x) = W (x) − √ V (x) + dx f π (12.50) At low energies (small x), Eq. (12.50) has a power series solution y(x)  a1 + a2 x + a3 x 3 + a4 x 4 + · · · x (12.51) Defining δ≡ √ √ π a (x) π a0  4 f 4 f (12.52) the coefficients are   4 1+δ a2 = − , 3 1 + 2δ a1 = 1, a3 = 103 + 380δ + 364δ 2 90(1 + 2δ)2 (12.53) The solution can be extended numerically to higher energies (larger x) by fitting a polynomial to values for which the power series is valid, say up to xn , to define the polynomials W (x) = K # (x − xk ), k=1 q(x) = K  W (x) dV (xk )/dx x − xk dW (xk )/dx (12.54) k=1 which can be used to extrapolate the solution to higher x > xn :  xn+1 q(x)dx y(xn+1 ) = y(yn ) + (12.55) xn These algorithms can be used as a predictor coupled with Eq. (12.50) in a predictor– corrector type of solution. The boundary condition μ(0) = 0, together with μ(x) = 0, implies that  1 =0 (12.56) lim y(x) − x→0 x 465 466 12 Neutron Thermalization Fig. 12.4 Comparison of Wigner–Wilkins and Maxwellian thermal neutron spectra for a typical PWR composition. (From Ref. 2; used with permission of Wiley.) which in turn implies that xdμ(0) exp μ(x) = dx  x  0  1 P (x )y(x ) −  dx  x   (12.57) Numerical integration of the exponent then allows the density to be constructed from √   x  4 x M(x) 1 n(x) = 3/4 y(x  )P (x  ) −  dx  (12.58) exp V (x) + a0 /f x π 0 The development can be extended to include non-1/v absorbers and leakage by the replacement a0 → a (E ) + B2 3[a (E ) + s (E )(1 − μ̄0 )] (12.59) where B characterizes a simple buckling mode. A thermal spectrum calculated for a 1/v absorber and with a thermal resonance, and matched to a 1/E slowing-down source upper boundary condition, is compared with a Maxwellian in Fig. 12.4. The spectrum hardening effects of the 1/v absorber in preferentially absorbing the lower-energy neutrons and of the 1/E slowing-down source in increasing the higher-energy neutron population are apparent, as is the flux depression in the vicinity of the resonance. Heavy Gas Model The heavy gas model given by Eq. (12.18) is a second-order differential equation for the thermal neutron flux. It is instructive to rederive that result before looking 12.4 Calculation of the Thermal Neutron Spectra in Homogeneous Media for a solution. We take advantage of the fact that the thermal neutron spectrum is expected to be similar to a Maxwellian for small absorption to look for a solution of Eq. (12.37) of the form φ(E ) = M(E )ψ(E ) (12.60) and then make use of the detailed balanced condition of Eq. (12.38) to rewrite Eq. (12.37):  ∞ t (E )M(E )ψ(E ) = M(E )     s0 E → E  ψ E  dE  (12.61) 0 Assuming that ψ is a slowly varying function of E, we make a Taylor’s series expansion  dψ(E ) 2 d 2 ψ(E )    1  + ··· + E −E ψ E  = ψ(E ) + E  − E dE 2! dE 2 (12.62) of ψ(E  ) in the scatter-in integral, to obtain a (E )M(E )ψ(E ) = M(E ) ∞  1 d n ψ(E ) An (E ) n! dEn (12.63) n=1 where the energy moments of the scattering energy transfer function are  ∞ E −E An (E ) = n   s0 E → E  dE  (12.64) 0 and where the first term in the expansion has canceled with the scattering contribution to the total cross section on the left side of the equation. This expansion is valid for any scattering transfer function, but its utility depends on rapid convergence of the Taylor series, which requires that s0 (E → E  ) is strongly peaked about E  = E (i.e., for heavy mass moderators which cannot produce a large energy change). Making a 1/A expansion of the gas scattering transfer function of Eq. (12.7) and using the result to evaluate Eq. (12.64) yields   2f 1 (2kT − E ) + O A A2  2   4f A 1 A2 (E ) = EkT + O A+1 A A2   1 An (E ) = O , n≥3 A2  A1 (E ) = A A+1 2 (12.65) If only terms through n = 2 are retained in Eq. (12.63), the resulting equation is identical to Eq. (12.18) to within a factor [A/(1 + A)]2 , which approaches unity for large A. 467 468 12 Neutron Thermalization It is convenient to rewrite Eq. (12.63) in terms of the variable x = (E/kT )1/2 : x  dn(x) d 2 n(x)  2 + 2x − 1 + (4x − )n(x) = 0 2 dx dx (12.66) where 1/v absorption has been assumed and the absorption parameter is ≡ 2Aa0 f (12.67) This equation can be solved exactly in the case of zero absorption ( = 0):   2 2 n(x) = a1 x 2 e−x + a2 x 2 e−x E1 x 2 − 1 (12.68) where E1 is the exponential integral function. Since the second term is negative at x = 0 and positive for large x, a2 must be zero. When absorption is present, Eq. (12.66) can be integrated once to obtain  x   2A dn(x) + 2 x 2 − 1 n(x) =  n(x  )dx  = q(x) (12.69) x dx f 0 where we have used the fact that all of the neutrons slowing down below x—the slowing down density q(x)—must be absorbed in the interval x < x  < 0. Integrating a second time yields an integral equation: 2 −x 2 n(x) = x e  4 √ + π  0 x 2 eu u3  u  n(u )du du   (12.70) 0 that is well suited to solution by iteration. The asymptotic form for the neutron flux φ = nv at large values of x is φ(E ) =   1/2   kT A 1 1 kT + 2 + 16 1−  + ··· 2f E 2 E 8 E (12.71) Equation (12.70) can be solved numerically to obtain the thermal neutron spectrum, φ(E ). The solution is shown in Fig. 12.5 for different values of the parameter = a0 /f . Numerical Solution Neutron scattering kernels are frequently so complicated that analytical or even semianalytical solutions are impractical, in which case direct numerical solution of the governing equation is the method of choice. A general numerical solution method, applicable to any scattering kernel, is illustrated for the proton gas model, for which Eq. (12.37) may be rewritten  ∞  xc dx  G(x  → x)N(x  ) + dx  G(x  → x)Nasym (x  ) V (x) + N(x) = 0 xc (12.72) 12.4 Calculation of the Thermal Neutron Spectra in Homogeneous Media Fig. 12.5 Neutron spectrum predicted by the heavy gas model for a 1/v absorber and different values of Ref. 2; used with permission of Wiley.) = a0 /f . (From where V (x) = xs (x) , f = a0 , f G(x  → x) = x  (x  → x) (12.73) and xc has been chosen so that the asymptotic form Nasym from the slowing-down range may be used for x  > xc . In this case, the last term in Eq. (12.72) may be written cx erf(x)/(xc + )2 , where erf(x) is the error function. Dividing the thermal energy range (0 ≤ x ≤ xc ) into I intervals and using the trapezoidal rule, the right side of Eq. (12.72) may be approximated: i−1  j =1 G(xj → xi )N(xj )j + G(xi → xi )N(xi )i + 2cxi erf(xi ) (xc + )2 (12.74) 469 470 12 Neutron Thermalization Equations (12.72) now may be solved directly by matrix inversion or by iteration. For the iterative solution, the equations are rearranged to obtain   I  1 2cxi erf(xi ) G(xj → xi )N(xj )j + N(xi ) = (V (xi ) + ) (xc + )2 j =1 i = 1, . . . , I (12.75) The iterative solution of Eqs. (12.75) proceeds by guessing N (0) (xj ), evaluating the right-hand side, calculating N (1) (xi ), and so on. A convenient starting guess is N (0) (xi ) = Nasym (xi ). It is important to enforce neutron conservation during the iteration, which is done by adjusting c. Moments Expansion Solution The continuous slowing-down, or moments expansion, methodology that was applied in Chapter 11 to the neutron slowing-down problem is also applicable to the neutron thermalization problem. For heavy elements, the development is similar to that of age theory. Defining u ≡ ln T E (12.76) and changing to the lethargy variable, Eq. (12.37) may be written  ∞ (a + s )φ(u) ≡ (u → u)φ(u ) du −∞ (12.77) where φ(u) ≡ En(E )v(E ),   (u → u) ≡ E E  → E Since φ(u) is approximately constant in the slowing-down range above thermal, φ(u ) is expanded in a Taylor’s series about u to obtain 0 1 a (u) + s (u) − ξ 0  φ(u) = − ξ 1 d 2 φ(u) 10 (−1)n 0 n 1 d n φ(u) dφ(u) + ξ 2 ξ  + ··· + + ··· 2 du 2! n! dun du (12.78) where 0 1 ξ n  ≡ (−1)n  ∞ −∞ (u − u)n (u → u)du (12.79) Noting that for energies above thermal (no upscattering) the nth term in Eq. (12.78) is of order (ξ0 )n−1 relative to ξ n  dφ/du, where ξ0 is the average logarithmic energy loss for scattering by free atoms at rest [ξ0 ≡ ξ iso = 1 + 12.4 Calculation of the Thermal Neutron Spectra in Homogeneous Media α ln α/(1 − α)]. Hence, for scattering from atoms other than hydrogen and deuterium, Eq. (12.38) can be truncated after a few terms with little loss in accuracy. Differentiating Eq. (12.38), truncating terms higher than d 2 φ/du2 , solving for ξ  d 2 φ/du2 , using this result in Eq. (12.38), and neglecting terms of order (ξ03 ) and higher yields 0 1 a (u) + s (u) − ξ 0  φ(u)  0 1 d ξ = − s (u)ξ(u) + γ (u) a (u) + s (u) − ξ 0 s + γ (u) du dφ(u) du (12.80) which can be integrated to obtain φ(u) = K(u) exp{− u     ˆ   ˆ −∞ [a (u )/(ξ(u )s (u ) + γ (u )a (u ))]du } ˆ a (u) ξ(u)s (u) + γ (u) (12.81) where  K(u) ≡ exp u  −∞ ˆ a (u )(dγ (u )/du ) + (d[ξ(u )s (u )]/du )  ˆ a (u ) ξ(u )s (u ) + γ (u )  ˆ a (u )/du )] g(u )[a (u ) + γ (u )(d  du − ˆ a (u ) ξ(u )s (u ) + γ (u ) ˆ a (u) ξ(u)s (u) + γ (u) −1 ˆ s (u)ξ(u) + γ (u)a (u) + γ (u)(d ξ  /du) 0 1 ˆ a (u) ≡ a (u) + s (u) − ξ 0   g(u) ≡ ξ(u) ≡ ξ (u) s (u) γ (u) ≡ 1 ξ 2 (u) 2 ξ (u) (12.82) The moments of the scattering kernel are given by  0 n 1 α n+1 ξ  = (−1) f (ln α)n − n(ln α)n−1 + n(n − 1)(ln α)n−2 1−α + · · · + (−1)n n! +  2μ 1 1−μ − μ(1 − α) − α(ln α)n 1+μ 2 1+μ    T̄ μ − (1 − α)α(ln 2)n−1 (2 ln α + n) +O E E2 − (1 + μ)2 4μ  (−1)n+1 n! 1−α (12.83) 471 472 12 Neutron Thermalization Table 12.1 Thermalization Parameters for Carbon Graphite Free Gas T (K) T̄ /T (K 2 )av /T 2 Bav /T 2 T̄ /T (K 2 )av /T 2 Bav /T 2 300 600 2.363 1.432 21.63 7.794 25 25/4 1 1 15/4 15/4 0 0 where μ ≡ m/M, the ratio of the masses of the neutron and the scattering atom. For scattering by unbound atoms at rest, K(u) → 1 and Eq. (12.81) is identical to the Grueling–Goertzel approximation of Chapter 10. For γ = 0, Eq. (12.81) reduces to Fermi age theory. The presence of γ = 0 accounts for upscattering in the thermal ˆ a = a + s − ξ 0  , rather than the range of energies. It is the decrease in  decrease in ξ s , that is the dominant effect of the chemical binding. For neutron thermalization by graphite, an explicit expression for the thermal spectrum is given by       1 T̄ 2 1 z ξ f Eφ(E ) = 1 − 1.1138  z + 0.6526  + 1.913 2 2 T   3  1 T̄ 3 − 0.2673  + 3.313 z 2 T    4 2  2 T̄ 1 1 T̄ + 0.08596  + 2.752  + 4.935 2 2 T T    (K 2 )av Bav 4 z + O z5 + 0.201 − 0.6204 (12.84) T2 T2 where z ≡ (T /E )1/2 ,  ≡ 2a (T )/μf , and the other parameters are defined in terms of the crystal vibration spectra for perpendicular, ρ1 (ω), and parallel, ρ2 (ω), vibrations: T̄ ≡ 13 T1 + 23 T2  1 θi ω dω Ti ≡ ωρi (ω) coth 2 0 2T  2 K av ≡ 34 T12 + T1 T2 + 2T22   1 θ1 2 2 θ2 2 Bav ≡ ω ρ1 (ω) dω + ω ρ2 (ω) dω 3 0 3 0 (12.85) where the θi are the cutoff frequencies for the respective crystal vibration modes. These parameters are given for graphite and a free carbon gas in a Maxwellian distribution in Table 12.1. 12.4 Calculation of the Thermal Neutron Spectra in Homogeneous Media For hydrogenous atoms, it is not possible to truncate Eq. (12.78) as described above for heavy mass scattering atoms. However, noting that   0 n 1 1 (12.86) ξ  = (−1)n n!f + O E2 for hydrogen, it is possible to obtain a solution φ(u) accurate to O(1/E 2 ) by neglecting terms of order 1/E 2 and higher in Eq. (12.78), which enables this equation to be written n ˆ a (u)  dφ(u) d 2 φ(u) n dφ (u) + φ(u) = − − · · · + (−1) + ··· f du dun du2 (12.87) Operating on Eq. (12.87) with 1 + d/du and integrating then yields φ(u) = 1 ˆ a (u) + f    exp − ∞ E ˆ a (E  )  dE  ˆ a (E  ) + f E    (12.88) Expanding Eq. (12.88) in inverse powers of (E/T )1/2 yields    2  T̄ 2 1 1  z+ 6  + z 2 2 T   3    1 25 1 T̄ 3 − 10  +  z 2 6 2 T 4  2       T̄ 3 T̄ 2 4 (K 2 )av 1 43 1  + + 15  + + 2 4 2 T 4 T 5 T2    1 Bav 4 z + O z5 − 2 2T  EφE = 1 − 3 (12.89) For hydrogen bound in water molecules at 293 K, the thermalization parameters are T̄ /T = 4.345, Bav /T 2 = 126.90, and (K 2 )av /T 2 = 53.63. Multigroup Calculation The thermal neutron scattering transfer function discussed in the preceding sections can be used in a multigroup calculation of the thermal neutron energy spectrum. The group-to-group scattering transfer term is defined as  g  →g = Eg−1 Eg dE Eg  −1 Eg  dE  s0 (E  → E )φ(E  ) Eg  −1 Eg  dE  φ(E  ) (12.90) Evaluation of Eq. (12.90) requires an approximation for the energy dependence of the thermal neutron flux over the energy interval Eg < E < Eg−1 . One of the approximate thermal neutron spectra above can be used for this purpose, or if the interval is sufficiently small, φ = constant can be used. 473 474 12 Neutron Thermalization The multigroup thermal neutron flux balance equation, neglecting leakage, is g a φg = G  g  →g s φg  + Sg , g = 1, . . . , G (12.91) g  =g where Sg is the slowing-down source to the upper groups in the thermal energy range. Applications to Moderators The thermal neutron flux distribution has been calculated numerically for water with various amounts of admixed cadmium absorber, using both the free gas and Nelkin models to calculate the scattering transfer cross section. Results of the calculations are compared with experiment in Fig. 12.6. The thermal neutron flux distribution has also been calculated numerically for a large graphite block poisoned with boron, using both the crystal model of Eq. (12.36) and the heavy gas model of Eq. (12.13) to evaluate the scattering transfer cross section. The results are compared with experiment in Fig. 12.7. 12.5 Calculation of Thermal Neutron Energy Spectra in Heterogeneous Lattices The transport equation for neutrons in the thermal energy region E < Eth ∼ 1 eV is  Eth  · ∇ψ(r, E, ) = dE   4π     d s r,  → , E  → E ψ r, E  ,  0 0 S(r, E) , + 4π 0 < E < Eth (12.92) where  S(r, E ) = ∞ Eth     dE  s r, E  → E φ r, E  (12.93) is the source of neutron scattering into the thermal region from the slowing-down region. With reference to Section 9.2, this equation can be converted into an integral equation for the scalar neutron flux, which for the case of isotropic scattering may be written  φ(r, E ) =  e−α(r,r ) dr 4π|r − r |2   Eth           dE s0 r, E → E φ r, E + S r, E  0 (12.94) 12.5 Calculation of Thermal Neutron Energy Spectra in Heterogeneous Lattices Fig. 12.6 Experimental and calculated neutron energy spectrum in water with cadmium poisons. (From Ref. 3; used with permission of Wiley.) Dividing the problem of interest (e.g., a fuel assembly) up into I spatial regions, integrating Eq. (12.94) over the volume Vi of region i, and defining [by analogy with Eq. (9.52)] T j →i 1 (E ) ≡ Vi   dri Vi drj Vj e−α(ri ,rj ) 4π|ri − rj |2 (12.95) leads to a coupled set of equations for the group fluxes φi in each region: φi (E ) = I  j =1  T j →i (E ) 0 Eth        dE  sj r , E  → E φj E  + Sj E  (12.96) 475 476 12 Neutron Thermalization Fig. 12.7 Experimental and calculated neutron energy spectrum in graphite at 323 K. (From Ref. 3; used with permission of Wiley.) Dividing the thermal energy range into G groups and using an appropriate differential scattering cross section and weighting spectrum to calculate g  →g sj Eg−1 Eg  −1 Eg Eg  ≡ sj (E  → E )w(E  )dE  dE Eg−1 Eg w(E  )dE  (12.97) Eqs. (12.96) can be integrated over Eg < E < Eg−1 to obtain the set of multigroup equations g φi = N  % j →i Tg G  g  =1 j =1 g  →g g  sj φj & g + Sj g = 1, . . . , G , (12.98) Following Section 9.3, define the collision probability ji g g j →i Pg ≡ Vi ti tj Tg (12.99) in terms of which Eqs. (12.98) can be written g g ti Vi φi = N  j =1 G ji Pg g  →g g  φi g  =1 sj g tj g + Sj , g = 1, . . . , G; i = 1, . . . , I (12.100) The collision probabilities can be calculated by the methods of Section 9.3. The multigroup scattering transfer cross sections can be calculated using one of the differential scattering cross sections and a plausible weighting function, as discussed in this chapter. Then the set of I × G Eqs. (12.100) can be solved for the group fluxes in each region. Such methods are widely employed for practical calculations of the thermal spectra in heterogeneous reactor fuel assemblies. 12.6 Pulsed Neutron Thermalization 12.6 Pulsed Neutron Thermalization Spatial Eigenfunction Expansion The time-dependent diffusion equation that describes the neutron flux distribution following the introduction of a pulse Q of neutrons with energy E0 at time t = 0 into a uniform but finite nonmultiplying medium is 1 ∂φ (r, E, t) − D(E )∇ 2 φ(r, E, t) + a (E )φ(r, E, t) ν ∂t  ∞      E  → E φ r, E  , t dE  − s (E )φ(r, E, t) = 0 + δ(t)Q(r)δ(E − E0 ) (12.101) Assuming that the spatial eigenfunctions satisfying ∇ 2 Gn (r) + Bn2 Gn (r) = 0 (12.102) and the physical boundary conditions form a complete set, the solution of Eq. (12.101) can be expanded:  φn (E, t)Gn (r) (12.103) φ(r, E, t) = n and the general orthogonality property  Gn (r)Gm (r)dr = δnm (12.104) can be used to reduce Eq. (12.101) to a coupled set of equations for the φn (E, t): 1 ∂φn (E, t) + D(E )Bn2 φn (E, t) + a (E )φn (E, t) v ∂t  ∞      E  → E φn E  , t dE  + δ(E − E0 )δ(t)Qn = (12.105) 0 where Qn ≡ dr Gn (r)Q(r). Energy Eigenfunctions of the Scattering Operator The scattering operator S0 defined by  ∞      E  → E φ E  dE  − s (E )φ(E ) S0 φ(E ) ≡ (12.106) 0 possesses an eigenvalue spectrum and a set of eigenfunctions in terms of which the energy dependence of the neutron spectrum may be expanded. The general eigenvalue problem is  ∞      E  → E χ E  dE  (12.107a) κχ(E ) = s (E )χ(E ) − 0 477 478 12 Neutron Thermalization or κχ(E ) = −S0 χ(E ) (12.107b) The adjoint operator S0+ defined (see Chapter 13) by  χ + (E )S0 χ(E )dE ≡  χ(E )S0+ χ + (E )dE (12.108) is S0+ χ + (E ) =  ∞     dE   E → E  χ + E  − s (E )χ + (E ) (12.109) 0 The principle of detailed balance,        E  → E M E  =  E → E  M(E ) (12.110) requires that χ(E ) = M(E )χ + (E ) (12.111) where M(E ) is the Maxwellian distribution. Thus the principle of detailed balance ensures that the lowest eigenvalue κ0 = 0 and eigenfunction χ0 (E ) = M(E ), independent of scattering model. As an example, consider the heavy gas model of Section 12.2. From Eq. (12.15),   2f d 2 χ(E ) dχ(E ) kT E + χ(E ) +E S0 χ(E ) = A dE dE 2 (12.112) and from Eq. (12.108), S0+ χ + (E ) =   2f d 2 χ + (E ) dχ + (E ) kT E + (2kT − E ) A dE dE 2 (12.113) The eigenvalues of the direct and adjoint eigenvalue problems of Eqs. (12.107) and (12.109) are identical (Chapter 13). Substitution of χ + (E ) = ∞  an E n (12.114) n=0 into the adjoint eigenvalue problem S0+ χ + (E ) + κχ + (E ) = 0 (12.115) and working use of Eq. (12.113) reveals that the eigenvalue spectrum is discrete: κn = 2f n, A n = 0, 1, 2, . . . (12.116) 12.6 Pulsed Neutron Thermalization The associated eigenfunctions are the Laguerre polynomials of order unity (1) Ln (E ) χn+ (E ) = √ n+1 (12.117) where (1) L0 (E ) = 1, (1) L1 (E ) = 2 − E, (1) L2 (E ) = 3 − 3E + 12 E 2 , . . . (12.118) These polynomials constitute a complete set, so any arbitrary function can be expanded in them. Expansion in Energy Eigenfunctions of the Scattering Operator Assuming that the function φn (E, t) can be represented as φn (E, t) ∼ φn (E )e−λn t (12.119) the homogeneous part of Eq. (12.105) reduces to the eigenvalue problem   λn 2 (12.120) − + D(E )Bn φn (E ) = S0 φn (E ) v Expanding each φn (E ) in the eigenfunctions of the scattering operator χm (E ),  Cmn χm (E ) (12.121) φn (E ) = m substituting into Eq. (12.120), multiplying by χp+ (E ), and integrating over energy yields   (12.122) −λn Vmp + Dmp Bn2 + κm δmp Cmn = 0 m where  Vmp = ∞ 0  Dmp = 0 ∞ 1 (E ) + χ χ (E )dE v m p (12.123) D(E )χm (E )χp+ (E )dE The set of Eqs. (12.122) formed by multiplying by each χp+ (E ) must simultaneously vanish, which by Cramer’s rule requires that   det −λn Vmp + Dmp Bn2 + κm δmp = 0 (12.124) This is the eigenvalue condition from which the κm are determined. The spatial harmonics n > 0 will decay more rapidly than the n = 0 modes, because λn>0 > λ0 , due to the larger Bn2 . When all the higher spatial harmonics have 479 480 12 Neutron Thermalization become negligible, the neutron pulse will decay as a series of energy harmonics of the fundamental spatial mode:  φ(r, E, t) = G0 (r) Ap φp,0 (E )e−λp,0 t (12.125) p At long times, φ(r, E, t) → G0 (r)φ0,0 (E )e−λ0,0 t (12.126) since λp,0 < λp,n for n > 0. If Eq. (12.121) is truncated at one term (i.e., only the fundamental energy eigenfunction is retained), then Eq. (12.124) yields λ0,0 = D00 2 B ≡ D̄0 B02 V00 0 (12.127) If the first two terms are retained in Eq. (12.121), then λ0,0 = D̄0 B02 − 1 V00 κ1  2 D01 − D̄0 V01 B04 (12.128) Thus measurement of the time decay of the neutron pulse yields information about the Maxwellian average diffusion coefficient, D00 . The second term in Eq. (12.128), which is known as the diffusion cooling term, depends explicitly on the thermalizing properties of the medium. References 1 W. Rothenstein and M. Segev, “Unit Cell Calculations,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations I, CRC Press, Boca Raton, FL (1986). 2 J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976), Chap. 9. 3 G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Wiley (Van Nostrand Reinhold), New York (1970), Chap. 7. 4 D. E. Parks, M. S. Nelkin, N. F. Wikner, and J. R. Beyster, Slow Neutron Scattering and Thermalization with Reactor Applications, W. A. Benjamin, New York (1970). 5 Neutron Thermalization and Reactor Spectra, STI/PUB/160, International Atomic Energy Agency, Vienna (1968). 6 I. I. Gurevich and L. V. Tarasov, Low Energy Neutron Physics, Wiley, New York (1968). 7 M. M. R. Williams, The Slowing Down and Thermalization of Neutrons, WileyInterscience, New York (1966). 8 R. J. Breen et al., “The Neutron Thermalization Problem,” in A. Radkowsky, ed., Naval Reactors Physics Handbook, U.S. Atomic Energy Commission, Washington, DC (1964). 9 K. H. Beckhurts and K. Wirtz, Neutron Physics, Springer-Verlag, Berlin (1964). 10 T.-Y. Wu and T. Ohmura, Quantum Theory of Scattering, Prentice Hall, Englewood Cliffs, NJ (1962). 11 H. C. Honeck, “Thermos, A Thermalization Transport Theory Code for Reactor Lattice Calculations,” USAEC report BNL-5826, Brookhaven National Laboratory, Upton, NY (1961). 12 J. R. Beyster, N. Corngold, H. C. Honeck, G. D. Joanou, and D. E. Problems Parks, in Third U.N. Conference on Peaceful Uses of Atomic Energy (1964), p. 258. 13 E. P. Wigner and J. E. Wilkins, “Effect of the Temperature of the Mod- erator on the Velocity Distribution of Neutrons with Numerical Calculations for Hydrogen as Moderator,” USAEC report AECD-2275 (1944). Problems 12.1. Use the proton gas model of Eqs. (12.51) to (12.53) and (12.58) to calculate the low-energy neutron flux distribution in water at 300 K. Use δsH = 38 barns, σaH2 O = 0.66 barn, and σsO = 4.2 barns. 12.2. Use the effective neutron temperature model of Eqs. (4.30) and (4.31) to calculate the thermal neutron spectrum in water at 300 K and compare with the results of Problem 12.1. 12.3. Repeat Problems 12.1 and 12.2 including a 1/v absorber with σa0 = 25 barns and Na/NH2 O = 0.1. 12.4. An H2 O-moderated reactor has a thermal flux of 2.5 × 1014 n/cm2 · s. Compute the absorption rate density in water at density 0.75 g/cm3 . 12.5. Evaluate the heavy gas model expression for the neutron flux in the limit E  kT [Eq. (12.71)] for neutron moderation in graphite at 500 K. Use σsc = 4.8 barns and σac = 0.004 barn. 12.6. Repeat the calculation of Problem 12.5 for an admixture of 1/v absorber with σa0 = 0.5 barn per carbon atom. 12.7. Write a computer code to integrate the nonlinear differential equation (12.50) describing neutron thermalization in a free proton gas. Use an energy mesh of E = 0.01 eV. Calculate the neutron spectrum in water at 300 K. Use σsH = 38 barns, σaH2 O = 0.66 barn, and σaO = 4.2 barns. 12.8. Solve the problem of neutron thermalization in a free proton gas model of water at 300 K by direct numerical solution. Compare your results with the results of Problem 12.7. 12.9. Calculate and plot the thermal energy spectrum of neutrons thermalizing in graphite and in a Maxwellian gas of carbon atoms of the same density at 300 K. Use Eq. (12.84). 12.10. Calculate and plot the thermal energy spectrum of neutrons thermalizing in water at 293 K from Eq. (12.89). 481 483 13 Perturbation and Variational Methods In many situations it is necessary to estimate the effect of numerous individual perturbations in the materials properties of the reactor on the multiplication constant or on a reaction rate in a reactor. Perturbation theory provides a means for obtaining an estimate of the change in multiplication constant or reaction rate, neglecting the effect of any change in the neutron flux distribution caused by the perturbation. Generalized perturbation estimates and variational estimates provide a means for taking into account the change in the neutron flux distribution caused by the perturbation, without actually having to calculate it, thus providing a powerful methodology for calculating reactivity coefficients and for performing sensitivity studies. Variational methods also have a much wider application in reactor physics in the development of approximations, and several of these are described. 13.1 Perturbation Theory Reactivity Estimate Multigroup Diffusion Perturbation Theory Let us return to the question of estimating the reactivity worth of a small change made in a critical reactor described by multigroup diffusion theory: g g g g −∇ · D0 (r)∇φ0 (r) + t0 (r)φ0 (r) = G  g  =1 g  →g 0 g (r)φ0 (r) + G 1 g g g χ νf (r)φ0 (r), k0  g = 1, . . . , G (13.1) g =1 Assume a change in microscopic cross section, density, or geometry such that D0 → D0 + D, 0 → 0 + . This change will produce a change in the flux φ0 → φ0 + δφ and a change in effective multiplication constant k0 → k0 + k, such that the perturbed system is described by  g   g   g  g  g −∇ · D0 + D g ∇ φ0 + δφ g + t0 + t φ0 + δφ g 484 13 Perturbation and Variational Methods = G   g  →g  g    +  g →g φ0 + δφ g + 0 g  =1 G   g    g χg νf 0 +  νf k0 + k  g =1  g  × φ0 + δφ g , g = 1, . . . , G (13.2) Equation (13.2) can, in principle, be solved to determine k using the methods described previously. However, in some applications (e.g., the calculation of reactivity coefficients associated with many different possible changes or the evaluation of the sensitivity of the multiplication constant to cross-section uncertainties) this would be impractical because of the large number of such calculations that would be involved. The objective of perturbation theory is to provide an estimate of k without requiring a calculation of the perturbed configuration (i.e., without calculating δφ). Using Eqs. (13.1) to eliminate certain terms in Eqs. (13.2), multiplying the resulting equation in each group by an arbitrary (at this point) spatially dependent function φg+ , integrating over the reactor and summing over groups, we obtain an exact expression for k:   G G    g g g  →g + dr −∇ · D0 ∇(δφ g ) + t0 δφ g − φg 0 δφ g g  =1 g=1 − G  χg  g νf 0 δφ g k0 + k   g =1  g G  g g + −∇ · D g ∇φ0 + t φ0 − − G   g  g χg  νf φ0 k0 + k  g  =1  g =1  g + −∇ · D g ∇(δφ g ) + t δφ g − − g   g →g φ0 G   g χg  νf δφ k0 + k  G    g →g δφ g  g  =1 *  g g =1  = 1 1 − k0 + k k0 k =− k0 (k + k0 )  dr G  φg+ χ g g=1  dr G  g=1 φg+ χ g G  g  =1 G  g  =1 g g νf 0 φ0 g g νf 0 φ0 (13.3) if we were willing to calculate δφ g in order to evaluate it. However, we wish to neglect δφ, which appears in two of the [·] terms on the left side. We might argue 13.1 Perturbation Theory Reactivity Estimate (for the moment) that since the third [·] term is the product of δφ and  or D, we can neglect it as being of second order in small quantities. However, we cannot make this argument for the first [·] term, which is of first order in small quantities, as is the second [·] term which we wish to evaluate to obtain the perturbation estimate of k. Thus are we motivated to choose the φg+ to cause the first [·] term in Eq. (13.3) to vanish for arbitrary δφ g . To determine the equation that must be satisfied by φg+ , it is necessary to twice integrate by parts the gradient part of the first [·] term and use the divergence theorem:     g g + g − dr φg ∇ · D0 ∇(δφ ) = − dr ∇ · φg+ D0 ∇(δφ g )  +  +  =− s  − dr D0 ∇φg+ · ∇(δφ g ) = − g   g dr ∇ · δφ g D0 ∇φg+ − ds φg+ D0 ∇(δφ g ) · ns + g    ds φg+ D0 ∇(δφ g ) · ns g s  g  dr δφ g ∇ · D0 ∇φg+ ds δφ g D0 ∇φg+ · ns g s  g  dr δφ g ∇ · D0 ∇φg+ (13.4) where ns is the outward normal unit vector to the surface of the reactor and the integrals over s are surface integrals. δφ g , which must satisfy the same boundary conditions as φ g , vanishes on the surface of the reactor, which causes the second term on the right in the final form of Eq. (13.4) to vanish. If we choose a boundary condition φg+ (rs ) = 0 (i.e., φg+ vanishes on the surface of the reactor), the first term on the right in Eq. (13.4) also vanishes. Using this result in Eq. (13.3) and interchanging the dummy g and g  indices, the vanishing of the first [·] term requires that   G  g g g dr δφ − ∇ · D0 ∇φg+ + t0 φg+ g=1 − G  g  =1 g→g  + 0 φg  − g G νf 0  k0 + k g  =1   χ g φg+ =0 (13.5) which is satisfied for arbitrary δφ g if φg+ satisfies −∇ g · D0 ∇φg+ g + t0 φg+ = G  g  =1   g→g φg+ g = 1, . . . , G + g G νf 0  k0 + k  χ g φg+ , g  =1 (13.6) and vanishes on the surface of the reactor: φg+ (rs ) = 0, g = 1, . . . , G (13.7) 485 486 13 Perturbation and Variational Methods With the function φg+ which satisfies Eqs. (13.6) and (13.7), with neglect of the third [·] term on the left, and with the approximation k0 (k0 + k) → k0 , Eq. (13.3) reduces to the perturbation theory expression for the reactivity worth:  k ρpert ≡ = k0 dr G   φg+ g g g ∇ · D g ∇φ0 − t φ0 + g=1 + G χ g   g  g  νf 0 φ0 k0   G  g  =1  g  g →g φ0 g =1  ÷ dr G  φg+ χ g G  g g  =1 g=1 g νf 0 φ0 + O(δφ) (13.8) where we indicate by O(δφ) that the neglected third [·] term in Eq. (13.3) introduces an error of order δφ (into a term which itself is of order ). 13.2 Adjoint Operators and Importance Function Adjoint Operators Equation (13.6) is mathematically adjoint to Eq. (13.1), when k0 + k → k0 , and the function φg+ is called the adjoint function. Comparing Eqs. (13.1) and (13.6) term by term identifies the direct and adjoint operators of multigroup diffusion theory, which are denoted symbolically as Direct Adjoint [D(φ)]g ≡ −∇ · D g ∇φ g g [(φ)]g ≡ t φ g [S(φ)]g ≡ G   g  →g g  D + (φ + ) g ≡ −∇ · D g ∇φg+  + (φ + ) g ≡ t φg+ + φ + S (φ ) g  =1 [F (φ)]g ≡ χ g G  g  =1 g νf φ g  F + (φ + ) g g ≡ G  g  =1   g→g φg+ g ≡ νf g G  g  =1 (13.9)  χ g φg+ The direct and adjoint operators for group diffusion and group absorption are identical; these operators are said to be self-adjoint. On the other hand, the adjoint group scattering and fission operators differ from the direct operators. Note that there is an adjoint boundary condition [Eq. (13.7)] associated with the definition of the adjoint group diffusion operator. In terms of these operators, Eq. (13.8) for the perturbation theory estimate of the reactivity worth of a change in reactor properties 13.2 Adjoint Operators and Importance Function becomes + , −[D(φ0 )]g − [(φ0 )]g + [S(φ0 )]g + (1/k0 )[F (φ0 )]g  + dr G g=1 φg [F0 (φ0 )]g + ,  + dr G g=1 φg −[A(φ0 )]g + (1/k0 )[F φ0 ]g + O(δφ) (13.10)  + dr G g=1 φg [F0 (φ0 )]g dr ρpert = ≡ G + g=1 φg It is clear from the derivation above that the adjoint operators were defined by the requirement  dr G  φg+ [B(φ)]g ≡  g=1 dr G  φ g B + (φ + ) g (13.11) g=1 where [B(φ)]g represents any one of the operators in Eq. (13.9). This definition of adjoint group operator is quite general and provides for the immediate generalization of perturbation theory to multigroup transport theory by replacement of [D(φ)]g with the appropriate transport group operator [T(φ)]g . This formalism may be generalized immediately from multigroup to energydependent diffusion or transport theory by replacing the sum over groups by an integral over energy. At this point, we introduce the notation  Bφ ≡ dr G  [B(φ)]g  or  ∞ dr g=1 dE B(φ) (13.12) 0 which allows the compact expression of the perturbation theory estimate of reactivity worth: ρpert = φ + , (F − A)φ0 + O(δφ) φ + , F0 φ0 In this notation, the definition of the adjoint operator becomes 1 0 1 0 + φ , Bφ ≡ φ, B + φ + (13.13) (13.14) Importance Interpretation of the Adjoint Function The mathematical definition of the adjoint function (in the diffusion theory approximation) is simply the function that satisfies Eq. (13.6). The physical interpretation is somewhat subtle and involves the concept of neutron importance, more specifically the importance of a neutron introduced into a reactor to a specific observable physical quantity. For example, an importance may be defined as the expected number of counts that will be produced at all subsequent times by a neutron introduced into a reactor with a given position, energy and direction or by the secondary, tertiary, etc. neutrons produced at other energies and directions as a result of fission and scattering events due to the original neutron. For a critical reactor, it is 487 488 13 Perturbation and Variational Methods conventional to define the neutron ‘importance,’ ψ + (r, , E ), as the asymptotic increase in neutron population in the reactor due to a single neutron introduced into the reactor with a given energy E, with a given direction  and at a given location r. (Actually, we need to speak of neutrons introduced within dE about E, dr about r, and d about , but we will leave this cumbersome terminology to be understood.) Neutrons introduced with a given energy and direction at a given location can (1) move to another location r + dr where the importance is different; (2) be captured, which causes the importance to become zero; (3) be scattered into a different energy E  and direction  where the importance is different; or (4) produce fission, which causes the importance of the original neutron to become zero, but which produces ν new neutrons distributed in energy E  and distributed isotropically in direction  with different importances. In a critical reactor, the importance must be conserved as the N neutrons move about and undergo these various reactions, which can be expressed as  N ψ + (r + dr, , E) − ψ + (r, , E) − a (r, E) + s (r, E) ψ + (r, , E)  + 0 + ∞ dE   0 νf (r, E) k 4π  d s (r, E → E  ,  →  )ψ + (r,  , E  ) ∞ dE   0 4π  d χ(E  )ψ + (r, , E  ) = 0 (13.15) 0 Making a Taylor’s series expansion ψ + (r + dr, , E)  ψ + (r, , E) +  · ∇ψ + (r, , E) (13.16) in Eq. (13.15) leads to the transport equation satisfied by the neutron importance:  · ∇ψ + (r, , E) − t (r, E)ψ + (r, , E)  4π  ∞     + dE  d s r, E → E  ,  →  ψ + r,  , E  0 + 0 νf (r, E) k  0 ∞ dE   4π   d χ(E  )ψ + r,  , E  = 0 (13.17) 0 The importance of neutrons leaving the reactor is zero, which provides a boundary condition for the neutron importance, ψ + (rs , , E) = 0, ns ·  > 0 (13.18) where ns is the outward normal unit vector to the surface of the reactor. Compare these equations with the neutron transport equation and surface boundary condition derived in Chapter 9: 13.3 Variational/Generalized Perturbation Reactivity Estimate − · ∇ψ(r, , E) − t (r, E)ψ(r, , E)  4π  ∞     + dE  d s r, E  → E,  →  ψ r,  , E  0 + χ(E) k 0  ∞ dE   0 4π     d νf r, E  ψ r, , E  = 0 (13.19) 0 and ψ(rs , , E) = 0, ns ·  < 0 (13.20) The neutron transport equation is based on a backward balance of neutrons among those neutrons that scattered or were produced in fission or moved from a nearby location in the immediate past (i.e., in the interval t − t to t ) and those neutrons that are undergoing absorption and scattering now (i.e., at time t ). The importance equation is based on a forward balance of the importance among those neutrons that are being absorbed or scattered now (i.e., at time t ) and the importance of those neutrons that will move to a nearby location or be scattered into a different energy and direction or produce fission neutrons with different energy and direction in the immediate future (i.e., in the interval t to t + t ). Eigenvalues of the Adjoint Equation In the foregoing development of Eq. (13.17) from physical arguments, the same effective multiplication constant was used to achieve a steady-state importance balance equation as was used to achieve a steady-state neutron balance equation. We now establish formally that the eigenvalues of the neutron balance equation (A − λF )φ = 0 (13.21) and of the adjoint equation   + A − λ+ F + φ + = 0 (13.22) are identical when the adjoint operators are related to the direct operators by Eq. (13.14). Multiplying Eq. (13.21) by φ + and integrating over space, direction, and energy, multiplying Eq. (13.22) by φ and integrating, and making use of Eq. (13.14) yields λ= φ, A+ φ + φ + , Aφ = = λ+ + φ ,Fφ φ, F + φ + (13.23) 13.3 Variational/Generalized Perturbation Reactivity Estimate In many practical applications, a perturbation to the properties of the reactor will cause a change in the neutron flux distribution which has a significant effect on the 489 490 13 Perturbation and Variational Methods reactivity worth of the perturbation (i.e., the neglected third [·] term in Eq. (13.3) is important). The perturbation theory of Section 13.1 can be extended to take into account the change in the flux distribution without actually requiring its calculation. Such extensions can be developed within the context of variational theory or simply as a heuristic extension of perturbation theory; the results are the same except for minor differences. This extended perturbation theory is widely used in reactor physics in the calculation of reactivity worths (and reaction rate ratios— next section) and for the performance of sensitivity studies. Since the variational theory is more systematic and has broader applications in reactor physics, we follow the variational development of an extended perturbation theory for estimating reactivity worths. One-Speed Diffusion Theory Consider a critical reactor described by the one-speed diffusion equation −∇ · D0 ∇φ0 + a0 φ0 − λ0 νf 0 φ0 = 0 (13.24) where, for convenience of notation, we set λ ≡ k −1 . Making use of the definition of adjoint operator given by Eq. (13.11) with G = 1, the one-speed diffusion theory adjoint equation satisfies −∇ · D0 ∇φ0+ + a0 φ0+ − λ0 νf 0 φ0+ = 0 (13.25) Thus the one-speed diffusion equation is self-adjoint and φ + = φ. Now consider perturbations D0 → D = D0 + D and 0 →  = 0 + , which cause φ0 → φex = φ0 + δφ and λ0 → λ = λ0 + λ. The perturbed system satisfies −∇ · D∇φex + a φex − λνf φex = 0 (13.26) Multiplying Eq. (13.26) by φ0+ , multiplying Eq. (13.25) by φex , integrating over volume, subtracting, and rearranging yields an exact expression for the reactivity worth of the perturbation: + , ρex φ0+ , φex = −λ = φ0+ , (λ0 (νf )φex + ∇ · D∇φex − a φex ) φ0+ , νf φex (13.27) If we used the approximation φex ≈ φ0 to evaluate Eq. (13.27), we would obtain the perturbation theory estimate of the reactivity worth of the change, in one-speed diffusion theory: + , φ + , (λ0 (νf )φ0 + ∇ · D∇φ0 − a φ0 ) ρpert φ0+ , φ0 = 0 φ0+ , νf φ0 + O(δφ) which is accurate to first order in δφ. (13.28) 13.3 Variational/Generalized Perturbation Reactivity Estimate Variational or generalized perturbation theory allows us to obtain an estimate that is accurate to second order in δφ. Note that Eq. (13.27) defines a number that is evaluated by performing integrals over space (more generally over space and energy) involving the functions φ0+ and φex . Such a function of functions is known as a functional. The idea behind variational theory is to construct an equivalent variational functional ρvar {φ + , φ, + } which has the prop+ } has the same value as the functional ρ {φ + , φ } erties: (1) ρvar {φ0+ , φex , ex ex 0 ex + if φ0 and φex are used to evaluate ρvar , and (2) ρvar {φ0+ , φ, + } evaluated with functions φ0+ and φ = φex + δφ yields a value that differs from ρex {φ0+ , φex } by O(δφ 2 , δφδ + ). In particular, ρvar {φ0+ , φ0 , + } = ρex {φ0+ , φex } + O(δφ 2 ) when φex = φ0 + δφ. We construct + ρvar φ0+ , φ, + , = φ0+ , (λ0 (νf )φ + ∇ · D∇φ − a φ) × 1− 0 φ0+ , νF φ + 1 , (−∇ · D∇φ + a φ − λνf φ) (13.29) by taking the exact functional of Eq. (13.27) and multiplying it by 1 minus a correction functional constructed by premultiplying the exact Eq. (13.26) by + and integrating over space (space and energy in general). This functional obviously satisfies the first of the properties of the variational functional above, because when φ =+ φex , the,correction functional vanishes and the first term reduces identically to ρex φ0+ , φex . Subtracting yields + , + , ρvar φ0+ , φex , + − ρvar φ0+ , φex − δφ, + =− × + φ0+ , (λ0 (νf )φex + ∇ · D∇φex − a φex )  0 φ0+ , νf φex + 1 , (−∇ · D∇δφ + a δφ − λνf δφ) + φ0+ , νf δφ  φ0+ , νf φex φ0+ , (λ0 (νf )δφ + ∇ · D∇δφ − a δφ) φ0+ , νf φex   + O δφ 2 (13.30) The explicit terms on the right in this expression will vanish for arbitrary δφ if is chosen to satisfy −∇ · D∇ = + ex + a + ex − λνf + ex −∇ · D∇φ0+ + a φ0+ − λ0 (νf )φ0+ φ0+ , (−∇ + · D∇φex + a φex − λ0 (νf )φex ) − νf φ0+ + φ0 , νf φex (13.31) + } = ρ {φ + , φ } + O(δφ 2 ). Thus evaluation of the variational so that ρvar {φ0+ , φ, ex ex 0 ex + given by functional of Eq. (13.29) using the functions φ0+ given by Eq. (13.25), ex Eq. (13.31), and any function φ = φex + δφ yields an estimate of the reactivity worth of the change which is accurate to O(δφ 2 ). 491 492 13 Perturbation and Variational Methods Unfortunately, solving Eq. (13.31) requires a knowledge of φex , avoidance of the calculation of which is the purpose of this development. If, instead of Eq. (13.31), we use the equation obtained by changing φex → φ0 −∇ · D0 ∇ = + 0 + a0 + 0 − λ0 νf 0 + 0 −∇ · D∇φ0+ − λ0 (νf )φ0+ + a φ0+ φ0+ , (−∇ · D∇φ0 + a φ0 − λ0 (νf )φ0 ) − νf φ0+ + φ0 , νf φ0 (13.32) + = it can be shown that ρvar {φ0+ , φ, 0+ } = ρex {φ0+ , φex } + O(δφ 2 , δφδ + ), where ex + + + + . Thus the variational estimate ρvar {φ0 , φ, 0 } is accurate to second order 0 +δ in the (presumably) small quantities δφ and δ . The function + is related to the flux change, δφ, caused by the perturbation. The equation satisfied by δφ is obtained by using D = D0 + D,  = 0 + , and φex = φ0 + δφ in Eq. (13.26) and making use of Eq. (13.24): −∇ · D∇(δφ) + a (δφ) − λνf (δφ) = −[−∇ · D∇φ0 + a φ0 − (λνf )φ0 ] (13.33) Comparing this equation with Eq. (13.31), it is apparent that + ∼ −δφ, since φ0 = φ0+ for one group. A similar relationship may be established for multigroup theory. Defining the variational flux correction factor + , 0 fvar φ0 , 0+ = 0+ , (−∇ · D∇φ0 + a φ0 − λ0 (νf )φ0 1 − λ(νf )φ0 ) (13.34) the variational estimate for the reactivity worth of a change in reactor properties may be written + , + , + , (13.35) ρvar φ0+ , φ0 , 0+ = ρpert φ0+ , φ0 1 − fvar φ0 , 0+ as the perturbation theory estimate times a flux correction factor. The calculations required for the variational estimate include the solution for the three spatial functions φ0+ , φ0 , and 0+ for the parameters of the critical reactor and the evaluation of the indicated spatial integrals in Eq. (13.29). The left side of Eq. (13.32) is identical with the homogeneous Eq. (13.24). However, the useful biorthogonality property 0+ , F0 φ0 = 0 can be demonstrated (Ref. 13), which assures the existence of a solution. Note that the source term on the right of Eq. (13.32) will in general be the same for all perturbations taking place within a given spatial domain, since the magnitude of the perturbations appear in the numerator and denominator, implying that the calculation of one such 0+ for each distinct spatial domain of interest will allow the evaluation of the reactivity worths of a large number of perturbations of different types and magnitudes within that spatial domain. The reactivity estimate of Eq. (13.35) has been found to be quite accurate when the change in properties is such as to produce a positive reactivity (fvar < 0) or a 13.3 Variational/Generalized Perturbation Reactivity Estimate small negative reactivity for which 0 < fvar  1. However, for large negative reactivities such that fvar ∼ 1, the (1 − fvar ) term becomes inaccurate. In such cases it is more accurate to use ρvar = ρpert [1 − fvar /(1 + fvar )], which form can be derived from consideration (Ref. 1) of the exact functional of Eq. (13.27). Thus a better variational estimate for the reactivity worth is  + ρ0 > 0  1 − fvar {φ0 , 0 }, + + + + , +, + ρvar φ0 , φ0 , 0 = ρpert φ0 , φ0 fvar {φ0 , 0 }  , ρ0 < 0 1 − 1 + fvar {φ0 , 0+ } (13.36) Other Transport Models This formalism can be generalized immediately to other representations of neutron transport (e.g., multigroup diffusion or transport theory). Let the operator A represent the transport, absorption and scattering and the operator F represent the fission. Then Eqs. (13.24) and (13.25) for the flux and adjoint in the critical reactor generalize to (A0 − λ0 F0 )φ0 = 0 (13.37)  +  A0 − λ0 F0+ φ0+ = 0 (13.38) and Eq. (13.26) for the flux in the perturbed reactor generalizes to (A − λF )φex = 0 (13.39) The exact value and perturbation theory estimate of the reactivity worth of the perturbation of Eqs. (13.27) and (13.28) become + , φ + , (λ0 F − A)φex ρex φ0+ , φex = 0 φ0+ , F φex (13.40) + , φ + , (λ0 F − A)φ0 ρpert φ0+ , φ0 = 0 φ0+ , F φ0 (13.41) Equation (13.32) for the generalized adjoint function (A0 − λ0 F0 ) + 0 = (A+ − λ0 F + )φ0+ φ0+ , (A − λ0 F )φ0 − + 0 becomes F0+ φ0+ φ0+ , F0 φ0 (13.42) The variational estimate for the reactivity worth of the perturbation is still given by Eq. (13.36), where now ρpert is given by Eq. (13.41) and the flux correction factor is given by + fvar φ0+ , φ0 , 0 + +, 0 = 0 , (A − λ0 F − λF )φ0 1 (13.43) 493 494 13 Perturbation and Variational Methods Reactivity Worth of Localized Perturbations in a Large PWR Core Model Exact, perturbation theory, and variational calculations were made of the reactivity worth of a change in the thermal group absorption cross section in a two-group model of a large (about 40 migration lengths) slab model of a PWR core. The perturbations were made in the left quarter of the core model. Small cross-section changes produced small reactivity changes that were well estimated by both perturbation and variational methods because the associated flux change was small. Larger cross-section changes, which produced larger reactivity worths and significant flux changes were poorly predicted by perturbation theory, but the variational flux correction resulted in quite accurate predictions even for flux tilts on the order of 100%. The reactivity predictions are shown in Fig. 13.1, and associated flux shapes for the unperturbed core and for the core with two of the perturbations are shown in Fig. 13.2. The unit of reactivity is pcm = 10−5 . Fig. 13.1 Reactivity worth of thermal cross-section changes over the left quarter of a slab PWR model in two-group diffusion theory: comparison of exact, perturbation theory, and variational calculations. (From Ref. 1; used with permission of American Nuclear Society.) 13.4 Variational/Generalized Perturbation Theory Estimates of Reaction Rate Ratios in Critical Reactors Fig. 13.2 Thermal flux distributions for the unperturbed and two perturbed conditions in a two-group slab PWR model. (From Ref. 1; used with permission of American Nuclear Society.) Higher-Order Variational Estimates A variational formalism (Refs. 17 and 20) has been developed for making reactivity estimates that are accurate to higher order in δφ = φex − φ0 . However, the complexity of such estimates has limited their practical application. 13.4 Variational/Generalized Perturbation Theory Estimates of Reaction Rate Ratios in Critical Reactors Consider the problem of estimating the reaction rate ratio RR{φex } = i φex j φex (13.44) in a perturbed system in which the exact flux is given by Eq. (13.39), without solving Eq. (13.39) for φex . The unperturbed critical reactor is described by Eq. (13.37). 495 496 13 Perturbation and Variational Methods As before, the operators A and F represent transport-absorption-scattering and fission, respectively, in whatever theory is chosen to describe the neutron distribution in the reactor (e.g., multigroup diffusion, multigroup Sn ). Clearly, a perturbation theory estimate RRpert {φ0 } = i φ0 j φ0 (13.45) has errors of O(δφ = φex − φ0 ). Defining the generalized adjoint function,  A+ − λF +  + R = + R, by j i − i φex j φex the variational estimate 0 + i φ0 + , RRvar φ0 , R0 1− = j φ0 (13.46) 1 + R0 , (A − (λF ))φ0 (13.47) + is calculated from Eq. (13.46) but with A → A0 , F → F0 and φex → φ0 , where R0 will have a second-order error O(δ + δφ), as may be demonstrated by evaluating  +  + , (13.48) = O δ R+ δφ RRex {φex } − RRvar φ0 , R0 Several reaction rate ratios calculated for a multigroup diffusion theory model of the spherical ZEBRA fast reactor critical assembly are given in Table 13.1. The breeding ratio is the ratio of the 238 U capture rate integrated over the region to the 239 Pu fission rate integrated over the region. The reference assembly composition is given in Table 13.2. It is clear that the flux correction provided by the variational (generalized perturbation theory) calculation is important in achieving an accurate estimate. Table 13.1 Table Perturbed Reaction Rate Ratios Ratio Reference Value Central σc28 /σf49 0.09866 0.09866 0.09866 Core breeding ratio Assembly breeding ratio 0.80040 2.1844 2.1844 Perturbation RRexact RRpert RRvar Add 0.01 at/cm3 Na 0 → 9.45 cm 0.10241 0.09866 0.10225 Increase σf49 10% Add 0.0015 at/cm3 Pu 9.45 → 22.95 cm Add 0.01 at/cm3 Na 0 → 9.45 cm Increase σf49 10% 0.08964 0.09887 0.08969 0.09866 0.08964 0.09884 0.80554 0.80040 0.80549 1.9939 2.0038 1.9937 Add 0.0015 at/cm3 Pu 9.45 → 22.95 cm 1.6034 1.6446 1.6049 Source: Data from Ref. 13; used with permission of Academic Press. 13.5 Variational/Generalized Perturbation Theory Estimates of Reaction Rates Table 13.2 Composition of Spherical Computational Model of ZEBRA Critical Assembly Core (0 → 22.95 cm) Isotope 239 Pu Blanket (22.95 → 49.95 cm) 0.00371 0.03174 0.005698 238 U 56 Fe 0.0003 0.04099 0.00477 Source: Data from Ref. 13; used with permission of Academic Press. 13.5 Variational/Generalized Perturbation Theory Estimates of Reaction Rates Many problems in reactor physics can be formulated as fixed source problems described by A0 φ0 = S (13.49) where the operator A represents transport, absorption, scattering, and if present in the particular problem, fission. Let us imagine that Eq. (13.49) has been solved for φ0 and then the reactor is perturbed, so that the flux now satisfies Aφex = S (13.50) and we wish to evaluate the reaction rate R{φex } = φex (13.51) without calculating φex . The perturbation theory estimate Rpert {φ0 } = φ0 + O(δφ) obviously is only accurate to zero order in the flux perturbation that is caused by the perturbation in the reactor properties. + , by Defining an adjoint function, φR0 + = A+ φR0 (13.52) it is easy to show that the variational estimate 0 + 1 + + , , (S − A0 φ0 ) = φ0 − φR0 Rvar φ0 , φR0 (13.53) differs from the exact calculation of the reaction rate in the perturbed reactor by a second-order term, +   + , = O δφδφR+ Rex {φex } − Rvar φ0 , φR0 + is calculated from Eq. (13.52) with A → A0 . where φR0 (13.54) 497 498 13 Perturbation and Variational Methods By making use of the definition of the adjoint operator, it follows that 1 0 1 0 1 0 φ = A+ φR+ , φ = φR+ , Aφ = φR+ S (13.55) implying that the reaction rate can also be calculated by integrating the product of the source distribution S and the generalized adjoint function φR+ over the volume of the reactor. This result suggests the interpretation of φR+ as an importance function for a source neutron to produce the reaction represented by . 13.6 Variational Theory Stationarity We have constructed variational extensions of perturbation theory by establishing functionals which when evaluated with the exact solutions of the governing equations yielded the exact value of a quantity of interest (e.g., reactivity worth, reaction rate) and which when evaluated with approximate solutions of the governing equations (or exact solutions of equations that approximated the governing equations) differed from the exact result by terms of second order in the difference between the approximate solutions and the exact solutions. In other words, terms involving first-order variations between the exact and approximate solutions vanished when the approximate solutions were used in the variational functionals. This property is described by stating that the variational functionals are stationary about the exact solutions of the governing equations (i.e., the first variations vanish), and the functions that make the variational functional stationary (by satisfying the governing equations) are known as the stationary functions. This means that the same value of the variational functional will be obtained when evaluated with two different functions that differ infinitesimally, if one of these functions exactly satisfies the governing equations (i.e., is the stationary function of the variational functional). Minimum principles of various sorts are usually represented by variational functionals, and the minimum property of the variational functional is a form of stationarity condition. However, with a minimum principle or minimum variational functional, the value of the variational functional will increase when evaluated with any function which differs sufficiently from the stationary function that δφ 2 is significant, whereas the value of a stationary variational functional may be greater or less than the stationary value when evaluated with a function that differs sufficiently from the stationary function. Roussopolos Variational Functional Consider again the variational functional of Eq. (13.53), which we now write in the more general form known as the Roussopolos functional: 1 + , 0 (13.56) Rvar φ, φR+ = φ − φR+ , Aφ − S 13.6 Variational Theory The stationarity condition is + , + , δRvar ≡ Rvar φ + δφ, φR+ + δφR+ − Rvar φ, φR+ 0 1 0 1 = δφ − δφR+ (Aφ − S) + A+ φR+ , δφ = 0 (13.57) For arbitrary and independent variations δφ and δφR+ , this requires that + = 0, δφ:  − A+ φRs δφ + : S − Aφs = 0 (13.58) where the subscript s indicates the stationary solution. When the stationary solutions are used to evaluate the functional of Eq. (13.56), the exact value φs is obtained. When approximate functions—trial functions—φ = φs + δφ and φR+ = + + δφR+ are used to evaluate the functional of Eq. (13.56), the value obtained φRs differs from the exact value by a term of order (δφδφR+ ). Schwinger Variational Functional The estimate of the reaction rate provided by Eq. (13.56) or (13.53) is obviously sensitive to the normalization of the trial functions. The stationarity of the variational functional can be used to choose the best normalization. Write χ + = c+ φR+ and χ = cφ. Substitute these trial functions into the variational functional of Eq. (13.56) and require stationarity with respect to arbitrary and independent variations δc+ and δc: 0 1 0 1 δRvar = φ δc + φR+ , (S − Acφ) δc+ − A+ c+ φR+ , φ δc = 0 (13.59) which is satisfied for arbitrary δc and δc+ only if c+ = φ , + φR , Aφ c= φR+ S φR+ , Aφ (13.60) Using these normalizations in Eq. (13.56) yields the equivalent Schwinger variational principle: , + φ φR+ S J φ, φR+ = φR+ , Aφ (13.61) the value of which is independent of the normalization of the trial functions. Rayleigh Quotient Consider the critical reactor eigenvalue problem described by the transport and adjoint equations (A − λF )φ = 0,   + A − λF + φ + = 0 (13.62) 499 500 13 Perturbation and Variational Methods The Rayleigh quotient λ{φ + , φ} = φ + , Aφ φ+, F φ (13.63) is a variational functional for the eigenvalue. The value of Eq. (13.63) when the exact solution of the first of Eqs. (13.62) is used in its evaluation is clearly the exact eigenvalue. The requirement that the first variation of the Rayleigh quotient vanish, δλ = δφ + , Aφ + φ + , Aδφ φ + , Aφ − + + φ ,Fφ φ ,Fφ δφ + , F φ + φ + , F δφ =0 φ+, F φ (13.64) for arbitrary and independent variations δφ + and δφ requires that the stationary functions φs and φs+ satisfy Eqs. (13.62). Construction of Variational Functionals Although the construction of variational functionals is usually done by trial and error, there is a systematic procedure that can guide the process. The basic idea is to add the inner product of some function φ + or + with the governing equation for φ to the quantity of interest and then use the stationarity requirement to determine the equation satisfied by φ + or + . For example, if we want to estimate a reaction rate φ and φ is determined by Aφ = S, we construct the Roussopolos functional Rvar {φ + , φ} = φ − φ + , (S − Aφ ) of Eq. (13.56), and find from the stationarity requirement that φ + must satisfy A+ φ + = . As another example, if we want to estimate the reactivity worth φ0+ , (λ0 F − A)φ / φ0+ , F φ of changes F and A leading from (A0 − λ0 F0 )φ0 = 0 to (A − λF )φ = 0, we construct ρvar {φ0+ , φ, + } = φ0+ , (λ0 F − A)φ / φ0+ , F φ [1 − + , (A − λF )φ ] of Eq. (13.29). 13.7 Variational Estimate of Intermediate Resonance Integral Consider, as an application, the elastic slowing down of neutrons in the presence of a resonance absorber and a moderator (m), which is described by  u  e(u −u) [σm + σ (u)]φ(u) = du σm φ(u ) 1 − αm u−m  u  e(u −u) du σs (u )φ(u ) + 1−α u−  u  e(u −u)  σm + du (13.65) σs (u )φ(u ) 1−α u− where σm , σ , and σs are moderator scattering cross section per atom of resonance absorber and the total and scattering microscopic cross sections of the resonance 13.7 Variational Estimate of Intermediate Resonance Integral absorber, respectively. It has been assumed that the moderator in-scatter integral can be evaluated using the asymptotic flux, which is constant in lethargy, and the constant has been chosen as unity, in writing the second form of the equation. This equation corresponds to the second of Eqs. (13.58). The quantity of physical interest is the resonance integral  I = σa (u)φ(u) du = σa φ (13.66) Using the definition of adjoint operator given by Eq. (13.14), where · now indicates an integral over lethargy, the first of Eqs. (13.58)—the adjoint equation—for this problem is  u+  e(u−u ) σs (u)φR+ (u ) = σa (u) du (13.67) [σm + σ (u)]φR+ (u) − 1−α u and the Schwinger variational functional of Eq. (13.61) becomes J {φ, φR+ } = = φ φR+ S φR+ , Aφ ∞ ∞ + 0 du σa (u)φ(u)][ 0 du φR (u)σm ] ∞ u  −u) +  (u /(1 − α)]σs (u )φ(u )} 0 du φR (u){[σm + σ (u)]φ(u) − u− du [e [ (13.68) In choosing trial functions, we recall the narrow resonance and wide resonance approximations of Chapter 4: φNR (u) = σm + σp , σm + σ (u) φWR (u) = σm σm + σa (u) (13.69) where σp is the background scattering cross section of the resonance absorber. Making similar approximations in Eq. (13.67) as were made in deriving Eqs. (13.69), we can derive approximate adjoint functions. For wide resonances, σs φR+ is approximately constant over the scattering interval and can be removed from the integral in Eq. (13.67), yielding + (u) = φWR σa (u) σm + σa (u) (13.70) In the limit of very narrow resonances the off-resonance form for σs φR+ can be used to evaluate the scattering integral to obtain + (u) = φNR σa (u) σm + σp σm σm + σ (u) (13.71) These results suggest the trial functions φλ (u) = σm + λσp σm + σa (u) + λσs (u) φκ+ (u) = σm + κσp σa (u) σm σm + σa (u) + κσs (u) (13.72) 501 502 13 Perturbation and Variational Methods which contain arbitrary constants λ and κ that are determined by using Eqs. (13.72) in the variational functional of Eq. (13.63) and requiring stationarity with respect to arbitrary and independent variations δλ and δκ, which leads to the transcendental equations 2 χκλ λ= 2 1 + χκλ , βκ = β λ 1 + 2σp (1 − Yκλ ) σm (13.73) which must be solved for χκλ and Yκλ , where βi2 = 1 + χκλ = σ0 σm + iσp γ +i n , i = λ, κ, 0, 1 2E0 (1 − α) (βκ + βλ ) (13.74) Yκλ = arctan χκλ /χκλ where the ’s are the resonance widths, and σ0 and E0 are the peak resonance cross section and the energy at which it occurs. The variational estimate of the resonance integral is + , J φλ , φκ+ = βλ + ((σm + σp )(β12 πσ0 γ /2E0 2 − β0 )/(σm + λσp )(βκ + βλ ))(1 − λ − Yκλ ) (13.75) which has been shown to provide a more accurate estimate than either the narrowresonance or wide-resonance approximations to the resonance integral for resonances of ‘intermediate’ width. 13.8 Heterogeneity Reactivity Effects As an application of the Raleigh quotient, consider a heterogeneous lattice described by collision probability integral transport theory. Equations (12.100) become  g  n g  →g  g ng g n g  Pn n μs + λνμf χ g φn , μt φn = g  n  n, n = 1, . . . , N; g, g  = 1, . . . , G and the corresponding adjoint equations are  g   ng→g    +g  ng +g ng Pnn μs + λνμf χ g φn , μt φn = g  n n, n = 1, . . . , N; g, g  = 1, . . . , G (13.76) 13.9 Variational Derivation of Approximate Equations g where n and g refer to spatial region and group, Pn n is the probability that a neutron in group g and region n has its next collision in region n, and μng is the cross section in group g and region n times the volume of region n divided by the total volume of all regions. The Rayleigh quotient of Eq. (13.63) becomes  + λ{φ , φ} =  +g  g g n g  →g g   φn gn φn μt φn − g  n Pn n μs  +g g  g n g  g  gn φn χ g  n Pn n νμf φn (13.77) This expression can be used in a number of ways. For example, approximate flux and adjoint distributions (even one based on a homogenized model) can be used as trial functions in Eq. (13.77) to obtain a more accurate estimate of the infinite multiplication factor in a heterogeneous lattice. 13.9 Variational Derivation of Approximate Equations The requirement that a variational functional be stationary about the function φs which causes the first variation of the functional to vanish is entirely equivalent to requiring that the function φs satisfy the governing equation for φ if the variational functional is constructed so that satisfaction of this governing equation is the stationarity condition. Thus the equations of reactor physics can be stated equivalently as stationary variational functionals, just as the equations of particle dynamics can be equivalently stated in terms of a Hamiltonian. For example, the statement that φs+ and φs make the Raleigh quotient of Eq. (13.63) stationary is entirely equivalent to the statement that φs and φs+ satisfy Eqs. (13.62) and the associated boundary conditions. This equivalence provides a basis for the variational derivation of approximate equations. As an example, consider a reactor described by one-speed diffusion theory in two dimensions:     ∂ ∂φ(x, y) ∂ ∂φ(x, y) − D(x, y) − D(x, y) ∂x ∂x ∂y ∂y   νf (x, y) φ(x, y) = 0 + a (x, y) − k (13.78) An equivalent variational description is the stationarity requirement for the variational functional: F {φ + , φ} =  dx dy φ + (x, y)     νf ∂ ∂φ ∂ ∂φ − D + a − φ × − D ∂x ∂x ∂y ∂y k (13.79) 503 504 13 Perturbation and Variational Methods Recalling that the one-speed diffusion equation is self-adjoint, we look for a separable solution: φ + (x, y) = φ(x, y) = φx (x)φy (y) (13.80) consisting of a known function φy (y), perhaps obtained from a one-dimensional calculation, and an unknown function φx (x). Substituting Eq. (13.80) into Eq. (13.79) and requiring stationarity with respect to arbitrary variations δφx (φy is specified and hence does not allow arbitrary variations) leads to a one-dimensional equation for the unknown φx (x):     νf x d dφx (x) − Dx (x) + Rx (x) + Dx (x)Bg2 (x) − φx (x) = 0 dx dx k (13.81) where the effective y-independent constants are defined as weighted integrals over y:  Dx (x) ≡ dy φy2 (y)D(x, y)  Dx (x)Bg2 (x) ≡ − dy φy (y)  x (x) ≡   ∂φy (y) ∂ D(x, y) ∂y ∂y (13.82) dy φy2 (y)(x, y) This procedure is referred to as variational synthesis and is described more fully in Chapter 15. Inclusion of Interface and Boundary Terms In deriving Eqs. (13.81) and (13.82) it was implicitly assumed that the known function φy (y) is continuous over all y, which limits the approximation to trial functions φy which are continuous in y. This limitation can be removed if the variational functional is modified so that stationarity requires not only satisfaction of Eq. (13.78) but also continuity of flux and current across an interface at y = yi . Stationarity of the modified functional  Fdis {φ + , φ} = dx dy φ + (x, y)         νf ∂φ ∂ ∂φ ∂ D − D + a − φ × − ∂x ∂x ∂y ∂y k  + dx φi+ (x, yi )[φ(x, yi + ε) − φ(x, yi − ε)]  +  ∂φ(x, yi + ε) dx Ji+ (x, yi ) − D(x, yi + ε) ∂y + D(x, yi − ε) ∂φ(x, yi − ε) ∂y  (13.83) 13.10 Variational Even-Parity Transport Approximations with respect to arbitrary and independent variations δφx over the volume and variations δφi+ and δJi+ on the interface at y = yi requires both that Eq. (13.78) be satisfied everywhere in the reactor except on the interface and that continuity of flux and current be satisfied at the interface: φ(x, yi + ε) = φ(x, yi − ε) − D(x, yi + ε) ∂φ(x, yi + ε) ∂φ(x, yi − ε) = −D(x, yi − ε) ∂y ∂y (13.84) Boundary terms can be included in a similar fashion, leading to variational functionals which admit trial functions that do not satisfy the boundary conditions. Inclusion of interface and boundary terms is important for the development of synthesis and nodal approximations and is discussed in greater detail in Chapter 15 as well as in Section 13.11. 13.10 Variational Even-Parity Transport Approximations Variational Principle for the Even-Parity Transport Equation The even-parity form of the transport equation introduced in Section 9.11 is convenient for the development of approximate transport equations when the scattering and source are isotropic. A variational functional for the even-parity component of the angular flux, which is self-adjoint, may be written     1 J {ψ + , φ} = dr d ( · ∇ψ + )2 + t (ψ + )2 − s φ 2 − 2φS  V   (13.85) + ds d|n · |[ψ + (rs )]2 S where the dependence on (r, ) has been suppressed, and the two integrals are over the volume V and the bounding surface S, with n being the outward normal to the surface. Note that here ψ + refers to the even component of the angular flux, not an adjoint function. Taking the variation of the functional J with respect to arbitrary but dependent (since φ depends on ψ + ) variations δψ + and δφ about some reference functions ψ0+ and φ0 yields , + , + δJ ≡ J ψ0+ + δψ + , φ + δφ − J ψ0+ , φ      1   · ∇δψ +  · ∇ψ0+ = 2 dr d (r) V  + t ψ0+ δψ + − δφ(s φ0 + S)   +2 ds S   d|ns · |ψ0+ δψ + + O (δψ + )2 , (δφ)2 505 506 13 Perturbation and Variational Methods     1 dδψ + − · ∇  · ∇ψ0+ + t ψ0+ − s φ0 − S t V        1  · ∇ψ0+ + O (δψ + )2 + 2 ds dδψ + |n · |ψ0+ + n ·  t S (13.86)  =2  dr where integration by parts and the divergence theorem have been used to obtain the final form. The requirements that the volume and surface integrals vanish for arbitrary and independent variations δψ + in the volume and on the surface are just the transport equation for the one-speed (or within-group) even-parity transport equation:   1 − · ∇  · ∇ψ0+ (r, ) + t (r)ψ0+ (r, ) − s (r)φ0 (r) − S(r) = 0 t (r) (13.87) and the vacuum boundary condition satisfied by the even-parity flux component:  · ∇ψ + (rs , ) ± t (rs )ψ + (rs , ) = 0,  · ns ≷ 0 (13.88) Ritz Procedure This is a procedure for constructing an improved approximate solution by combining several plausible approximate solutions, each of which perhaps represents some feature expected in the exact solution, that is, by approximating the evenparity flux by an expansion in known functions χi (r, ):  ψ + (r, )  ai χi (r, ) (13.89) i The general Ritz method proceeds by substituting this expansion into the variational functional describing the system of interest, Eq. (13.85) in our case, and requiring stationarity (vanishing of first variations) for arbitrary and independent variations of the combining coefficients, ai : Aa − S ] δJ {aa } = 0 = 2δaa T [A (13.90) where a is a column vector of the a i , a T is the transposed row vector, A is a matrix with elements     1 dr d ( · ∇χi )( · ∇χj ) − t χi χj Aij = t V     (13.91) − s d χi dχj + ds d|ns · |χi χj S and  Si =  dr V d χi S(r) (13.92) 13.10 Variational Even-Parity Transport Approximations Thus the requirement for stationarity of the variational principle defines the ai as the solution of Aa = S (13.93) Diffusion Approximation The diffusion approximation was shown in Chapter 9 to follow from a representation of the angular flux of the form ψ(r, )  φ(r) + 3 · J(r) (13.94) With this representation, the even-parity component of the angular flux is just the scalar flux, 1 ψ + (r, ) ≡ [ψ(r, ) + ψ(r, −)] = φ(r) 2 (13.95) Using this representation for the even-parity flux in the variational principle of Eq. (13.85) leads to dr V +  =    J {φ} = d   ds φ 2 S  dr V 1 ( · ∇φ)2 + t φ 2 − s φ 2 − 2φS t  d| · ns |   1 1 (∇φ)2 + (t − s )φ 2 − 2φS + ds φ 2 3t 2 s (13.96) Requiring stationarity with respect to arbitrary and independent variations δφ in the volume and on the surface leads to the equation  −∇ ·  1 ∇φ0 (r) + [t (r) − s (r)]φ0 (r) = S(r) 3t (r) (13.97) and the boundary condition − 2 ns · ∇φ0 (rs ) + φ0 (rs ) = 0 3t (rs ) (13.98) Equation (13.97) differs from the previous diffusion equation only by the t−1 rather than (t − μ0 s )−1 in the first term, and had we neglected anisotropic scattering (→ μ0 = 0) in Chapter 9 as we have here, the two would be identical. Equation (13.98) specifies that the flux extrapolate to zero a distance 2/3t outside the boundary, which is the same result (for isotropic scattering) that was obtained from P1 theory in Chapter 9. 507 508 13 Perturbation and Variational Methods One-Dimensional Slab Transport Equation In a slab varying from x = 0 to x = a, the variational principle of Eq. (13.85) becomes      a  1 + , dμ μ2 ∂ψ + 2 dx + t (ψ + )2 − s φ 2 − 2φS J ψ+ = t ∂x 0 −1 2  1  1 dμ dμ (13.99) + |μ|(ψ + )2 x=0 + |μ|(ψ + )2 x=a 2 −1 −1 2 Requiring stationarity with respect to arbitrary and independent variations δψ + within 0 < x < a and at x = 0 and x = a yields a one-dimensional transport equation for the even-parity flux component: −μ2   ∂ 1 ∂ + ψ (x, μ) + t (x)ψ + (x, μ) ∂x t (x) ∂x = s (x)φ(x) + S(x) (13.100) and a pair of extrapolated vacuum boundary conditions ψ + (a, μ) + 1 ∂ + ψ (a, μ) = 0 t (a) ∂x 1 ∂ + ψ (0, μ) = 0 ψ (0, μ) − t (a) ∂x (13.101) + 13.11 Boundary Perturbation Theory Consider a reactor described by the multigroup diffusion equations, which are written in operator notation as A0 (r)φ0 (r) = λ0 F0 (r)φ0 (r) (13.102) with general boundary conditions given by a0 n · ∇φ0 (rs ) + b0 φ0 (rs ) = 0 (13.103) where a0 and b0 are group-dependent operators which may vary with position on the surface rs . The adjoint equation is + + + A+ 0 (r)φ0 (r) = λ0 F0 (r)φ0 (r) (13.104) where the definition (13.11) or (13.14) of adjoint operator has been used. The double integration by parts of the spatial derivative term in the diffusion operator yields 13.11 Boundary Perturbation Theory  − V dr φ02 (∇ · D0 ∇φ0 ) =−   V dr φ0 ∇  · D0 ∇φ0+ −  s   ds n · φ0+ D0 ∇φ0 − φ0 D0 ∇φ0+ (13.105) Using the boundary condition of Eq. (13.103) to evaluate the n · ∇φ0 term in the surface integral reveals that the natural adjoint boundary condition (the condition that leads to vanishing of the surface integral) is a0 n · ∇φ0+ (rs ) + b0 φ0+ (rs ) = 0 (13.106) Now let the boundary condition be changed by perturbing b0 to b0 + b1 : a0 n · ∇φ + (rs ) + (b0 + b1 )φ + (rs ) = 0 and a0 n · ∇φ(rs ) + (b0 + b1 )φ(rs ) = 0 (13.107) where |b1 /b0 | ≡ ε  1. The perturbed flux, which must satisfy a different boundary condition and is associated with a different eigenvalue as a consequence, now satisfies A0 (r)φ(r) = λF0 (r)φ(r) (13.108) Expanding the perturbed flux and eigenvalue φ = φ0 + φ1 + φ2 + · · · (13.109) λ = λ0 + λ1 + λ2 + · · · (13.110) where the subscript indicates the order of the term with respect to the small parameter |b1 /b0 | ≡ ε  1, and substituting into Eqs. (13.107) and (13.108) results in the following hierarchy of perturbation equations and boundary conditions: • Order ε0 : A0 (r)φ0 (r) = λ0 F0 (r)φ0 (r) (13.111) a0 n · ∇φ0 (rs ) + b0 φ0 (rs ) = 0 (13.112) • Order ε1 : [A0 (r) − λ0 F0 (r)]φ1 (r) = λ1 F0 (r)φ0 (r) (13.113) a0 n · ∇φ1 (rs ) + b0 φ1 (rs ) + b1 φ0 (rs ) = 0 (13.114) • Order ε2 :   A0 (r) − λ0 F0 (r) φ2 (r) = λ1 F0 (r)φ1 (r) + λ2 F0 (r)φ0 (r) (13.115) a0 n · ∇φ2 (rs ) + b0 φ2 (rs ) + b1 φ1 (rs ) = 0 (13.116) 509 510 13 Perturbation and Variational Methods The leading-order estimate of the eigenvalue is obtained by multiplying Eq. (13.111) by φ0+ and integrating over space and summing over groups (indicated by · ): λ0 = φ0+ , A0 φ0 (13.117) φ0+ , F0 φ0 The first-order correction to the eigenvalue is obtained by multiplying Eq. (13.113) by φ0+ and integrating over space and summing over groups, integrating the derivative term by parts twice, and using the boundary conditions of Eqs. (13.106) and (13.114): λ1 = φ0+ , D0 a0−1 b1 φ0 S (13.118) φ0+ , F0 φ0 where · S indicates an integral over the surface and a sum over groups. The second-order correction to the eigenvalue is obtained by multiplying Eq. (13.115) by φ0+ and integrating over space and summing over groups, integrating the derivative term by parts twice and using the boundary conditions of Eqs. (13.106) and (13.114): λ2 = −λ1 φ0+ , F0 φ1 φ0+ , F0 φ0 + φ0+ , D0 a0−1 b1 φ1 S (13.119) φ0+ , F0 φ0 The perturbation theory estimate, through second order, is λ  λ0 + λ1 + λ2 = φ0+ , A0 φ0 + (1 − φ0+ , F0 φ1 / φ0+ , F0 φ0 ) φ0+ , D0 a0−1 b1 φ0 S φ0+ , F0 φ0 + φ0+ , D0 a0−1 b1 φ1 S (13.120) To evaluate this second-order estimate, it is necessary to solve Eqs. (13.104), (13.111), and (13.113), with the associated boundary conditions. The first two equations, for φ0+ and φ0 , and their boundary conditions are independent of the boundary perturbation b1 . Upon using Eq. (13.118), Eq. (13.113) for φ1 can be written [A0 (r) − λ0 F0 (r)]φ1 (r) = φ0+ , D0 a0−1 b1 φ0 s F0 (r)φ0 (r) φ0+ , F0 φ0 (13.121) showing that the amplitude of φ1 depends on the magnitude of the perturbation in boundary condition b1 . The first-order perturbation theory estimate is λ = λ0 + λ1 and corresponds to omitting the φ1 terms in Eq. (13.120), which obviates the necessity of calculating φ1 . References References 1 W. M. Stacey and J. A. Favorite, “Variational Reactivity Estimates,” Joint Int. Conf. Mathematical Methods and Supercomputing for Nuclear Applications I, American Nuclear Society, La Grange Park, IL (1997), pp. 900–909. 2 K. F. Laurin-Kovitz and E. E. Lewis, “Solution of the Mathematical Adjoint Equations for an Interface Current Nodal Formulation,” Nucl. Sci. Eng. 123, 369 (1996). 13 14 15 3 Y. Ronen, ed., Uncertainty Analysis, CRC Press, Boca Raton, FL (1988). 4 A. Gandini, “Generalized Perturbation Theory (GPT) Methods: A Heuristic Approach,” in J. Lewins and M. Becker, eds., Advances in Nuclear Science and Technology, Vol. 19, Plenum Press, New York (1987). 16 17 5 T. A. Taiwo and A. F. Henry, “Perturbation Theory Based on a Nodal Model,” Nucl. Sci. Eng. 92, 34 (1986). 6 M. L. Williams, “Perturbation Theory for Nuclear Reactor Analysis,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations I, CRC Press, Boca Raton, FL (1986). 18 7 F. Rahnema and G. C. Pomraning, “Boundary Perturbation Theory for Inhomogeneous Transport Equations,” Nucl. Sci. Eng. 84, 313 (1983). 19 20 8 E. W. Larsen and G. C. Pomraning, “Boundary Perturbation Theory,” Nucl. Sci. Eng. 77, 415 (1981). 9 D. G. Cacuci, “Sensitivity Theory for Nonlinear Systems,” J. Math. Phys. 22, 2794 and 2803 (1981). 10 D. G. Cacuci, C. F. Weber, E. M. Oblow, and J. H. Marable, “Sensitivity Theory for General Systems of Nonlinear Equations,” Nucl. Sci. Eng. 75, 88 (1980). 11 M. Komata, “Generalized Perturbation Theory Applicable to Reactor Boundary Changes,” Nucl. Sci. Eng. 64, 811 (1977). 12 E. Greenspan, “Developments in Perturbation Theory,” in J. Lewins and 21 22 23 24 M. Becker, eds., Advances in Nuclear Science and Technology, Vol. 9, Plenum Press, New York (1976). W. M. Stacey, Variational Methods in Nuclear Reactor Physics, Academic Press, New York (1974). W. M. Stacey, “Variational Estimates of Reactivity Worths and Reaction Rate Ratios in Critical Systems,” Nucl. Sci. Eng. 48, 444 (1972). W. M. Stacey, “Variational Estimates and Generalized Perturbation Theory for Ratios of Linear and Bilinear Functionals,” J. Math. Phys. 13, 1119 (1972). G. I. Bell and S. Glasstone, Nuclear Reactor Theory, Van Nostrand Reinhold, New York (1970), Chap. 6. J. Devooght, “Higher-Order Variational Principles and Iterative Processes,” Nucl. Sci. Eng. 41, 399 (1970). A. Gandini, “A Generalized Perturbation Method for Bilinear Functionals of the Real and Adjoint Neutron Fluxes,” J. Nucl. Energy Part A/B 21, 755 (1967). G. C. Pomraning, “The Calculation of Ratios in Critical Systems,” J. Nucl. Energy Part A/B 21, 285 (1967). G. C. Pomraning, “Generalized Variational Principles for Reactor Analysis,” Proc. Int. Conf. Utilization of Research Reactors and Mathematics and Computation, Mexico, D.F. (1966), p. 250. G. C. Pomraning, “Variational Principles for Eigenvalue Equations,” J. Math. Phys. 8, 149 (1967). G. C. Pomraning, “A Derivation of Variational Principles for Inhomogeneous Equations,” Nucl. Sci. Eng. 29, 220 (1967). J. Lewins, “A Variational Principle for Ratios in Critical Systems,” J. Nucl. Energy Part A/B 20, 141 (1966). G. C. Pomraning, “A Variational Description of Dissipative Processes,” J. Nucl. Energy Part A/B 20, 617 (1966). 511 512 13 Perturbation and Variational Methods 25 J. Lewins, Importance: The Adjoint Function, Pergamon Press, Oxford (1965). 26 L. N. Usachev, “Perturbation Theory for the Breeding Ratio and Other Number Ratios Pertaining to Various Reactor Processes,” J. Nucl. Energy Part A/B 18, 571 (1964). 27 D. S. Selengut, Variational Analysis of Multidimensional Systems, Hanford Engineering Laboratory report HW59126 (1959). 28 N. C. Francis, J. C. Stewart, L. S. Bohl, and T. J. Krieger, “Variational Solutions of the Transport Equation,” Prog. Nucl. Energy Ser. I 3, 360 (1958). Problems 13.1. Use one-speed diffusion theory and perturbation theory to estimate the reactivity worth of a 0.25% increase in the fission cross section over the left half of a critical slab reactor of 1-m thickness. 13.2. Use two-group diffusion theory perturbation theory to estimate the reactivity worth of a 0.5% change in the thermal absorption cross section of a very large core described by: group 1—D = 1.2 cm, a = 0.012 cm−1 ,  1→2 = 0.018 cm−1 , νf = 0.006 cm−1 ; group 2— D = 0.40 cm, a = 0.120 cm−1 , νf = 0.150 cm−1 . 13.3. Prove that each term in the importance equation [Eq. (13.17)] is mathematically adjoint to the corresponding term in the neutron transport equation [Eq. (13.19)]. 13.4. Derive the multigroup discrete ordinates adjoint equation for a critical reactor (a) directly from the discrete ordinates equations, and (b) by making the discrete ordinates approximation of the adjoint transport equation. 13.5. Derive an explicit expression for the perturbation theory reactivity estimate in the multigroup discrete ordinates representation of neutron transport. 13.6. Solve for the infinite medium neutron flux and adjoint energy distributions in a three-group representation: group 1—a = 0.030 cm−1 ,  1→2 = 0.060 cm−1 , νf = 0.004 cm−1 ; group 2—a = 0.031 cm−1 ,  2→3 = 0.088 cm−1 , νf = 0.018 cm−1 ; group 3—a = 0.120 cm−1 , νf = 0.180 cm−1 . 13.7. Carry through the derivation to show that ρvar {φ0+ , φ, 0+ } = ρex {φ0+ , φex } + O(δφ 2 , δφδ ), where + + + and + is obtained by solving Eq. (13.32). ex = 0 + δ 0 13.8. Consider a critical slab reactor with one-speed diffusion theory constants D = 1.0 cm, a = 0.15 cm−1 , and νf = 0.16 cm−1 . Calculate the flux correction function, + 0 , from Eq. (13.32) for a 1% change in the absorption Problems 13.9. 13.10. 13.11. 13.12. 13.13. 13.14. 13.15. 13.16. 13.17. cross section in the left one-fourth of the critical slab. (Hint: Note that 0+ is orthogonal to φ0+ = φ0 and expand 0+ in the higher harmonics of the critical reactor eigenfunctions.) Evaluate the variational/generalized perturbation reactivity estimate of Eq. (13.36) for Problem 13.8. Carry out the missing steps of the derivation in Section 13.4 + } − RRex {φex } = O(δ R+ δφ). to show that RRvar {φ0 , R0 Consider a critical bare slab reactor described by one-speed diffusion theory with D = 1.0 cm, a = 0.15 cm−1 , and νf = 0.16 cm−1 . Use Eq. (13.47) to evaluate the variational estimate for a 1% increase in absorption cross section on the absorption-to-fission rate ratio in the right one-tenth of the slab core. Use the Rayleigh quotient to estimate the effective multiplication constant for a bare cylindrical core with H /D = 1, H = 2 m and one-speed diffusion theory parameters D = 1.0 cm, a = 0.15 cm−1 , and νf = 0.16 cm−1 . In the window-shade model, a control rod bank can be represented by a 10% increase in a for Problem 13.11. Use the Rayleigh quotient to estimate the effective multiplication constant when the control rod bank is inserted halfway, using the flux and adjoint distributions calculated in Problem 13.11. Recalculate the effective multiplication constant directly (i.e., solve the two-region diffusion theory problem) for the control rod bank inserted halfway and compare with the variational estimate. Consider a uniform slab nonfissioning assembly of width 50 cm in which there is a uniform source Sf of fast neutrons in the left half. Calculate the thermal absorption rate in the right half (a) directly and (b) using the Schwinger variational estimate evaluated trial functions, obtain from an infinite medium calculation with a source 1/2Sf . Use the two-group representation: fast group—D = 2.0 cm, a = 0.006 cm−1 , and  1→2 = 0.018 cm−1 ; thermal group—D = 0.40 cm and a = 0.120 cm−1 . Repeat the derivation of Section 13.3 for multigroup diffusion theory. Discuss how the result of Eq. (13.55) could be employed to calculate the response of a localized detector to a point neutron source some distance away if the adjoint function is known in the vicinity of the source. Carry through the steps in deriving the variational synthesis approximation of Eq. (13.81). 513 514 13 Perturbation and Variational Methods 13.18. Demonstrate that stationarity of the variational functional of Eq. (13.83) requires that the diffusion equation be satisfied and that the flux and current be continuous at the interface y = yi . (Hint: Consider arbitrary and independent variations of the adjoint flux and current within the volume and on the interface.) 13.19. Derive the transport Eq. (13.100) and the associated boundary conditions of Eq. (13.101) from the stationarity of the functional of Eq. (13.99). 13.20. Consider a uniform slab reactor of thickness 2a with zero flux conditions at each boundary, which may be represented as a slab with a zero flux condition at x = 0 and a symmetry n̂ · ∇φ = 0 condition at x = a. Use boundary perturbation theory to derive an estimate for the change in eigenvalue, λ1 , that would result from replacing the symmetry condition at x = a with the condition n̂ · ∇φ + b1 φ = 0. 13.21. In a critical uniform slab reactor in one-group theory, 10% of the neutrons leak from the reactor and the other 90% are absorbed. Use perturbation theory to calculate the reactivity worth of a 5% increase in absorption cross section over the right half of the reactor. Discuss the error in this estimate due to the failure to take into account the change in flux distribution caused by the increase in absorption cross section. Would the perturbation theory estimate be expected to underpredict or overpredict the reactivity worth because of this error? Discuss how the effect of this flux change on the reactivity worth could be taken into account without actually calculating the flux change. 515 14 Homogenization Nuclear reactor cores are composed of a large number of fuel assemblies, each containing a large number of discrete fuel elements of differing composition and consisting of separate fuel and cladding regions, coolant, structural elements, burnable poisons, water channels, control rods, and so on—tens to hundreds of thousands of discrete, heterogeneous regions. On the other hand, most of the methods for calculating criticality and global flux distributions that are in use (in particular diffusion theory) are predicated on the existence of large (with respect to a mean free path) homogeneous regions. The methods employed to replace a heterogeneous lattice of materials of differing properties with an equivalent homogeneous mixture of these materials to which the previously discussed methods for the calculation of ultrafine group spectra, calculation of the diffusion of neutrons during the slowing-down process, and so on, is referred to as homogenization theory. Homogenization of a heterogeneous assembly usually proceeds in two steps: a lattice transport calculation to obtain the detailed heterogeneous flux distribution within a unit cell or fuel assembly, followed by the use of this detailed flux distribution to calculate average homogeneous cross sections for the unit cell or assembly. The general procedure that is followed in nuclear reactor analysis is to perform very detailed energy and spatial calculations on a local basis to obtain cross sections averaged over energy and spatial detail which can be used in few group global core calculations. For example, for a thermal reactor, a pin-cell transport calculation of a cell consisting of the fuel, clad, coolant, and structural in a local region may be carried out in 20 to 100 fine groups to obtain homogenized 6 to 20 intermediategroup cross sections averaged over the pin-cell geometry and the 20 to 100 fine group spectrum. Several such pin-cell calculations may be needed for a fuel assembly and the adjacent water gaps and control rods. Intermediate-group assembly transport calculations are then performed for models that represent all fuel pins, control rods, water channels, can walls, and so on, associated with a given fuel assembly. It is important that the intermediate-group assembly transport calculation uses enough groups to represent the spectral interactions among fuel pins of different composition, control rods, water channels, and so on, at the intermediategroup level. Several such intermediate-group assembly calculations may be needed for the reactor core, and a large number of such calculations may be needed to represent different operating temperatures, depletion steps, void fractions, and so 516 14 Homogenization on. The results of the intermediate-group assembly transport calculations are next averaged over the assembly spatial detail and the intermediate-group spectra to obtain two to six few-group homogenized assembly cross sections which can be used in few-group global core calculations of criticality and flux distribution. We have discussed the procedure of group collapsing to obtain few-group cross sections from fine- or intermediate-group spectra in previous chapters. Here we are interested in the spatial-averaging procedures used to obtain homogenized cross sections appropriately averaged over spatial detail, and in the procedures used to construct effective diffusion theory cross sections for regions such as control rods in which the basic assumptions of diffusion theory are not satisfied. 14.1 Equivalent Homogenized Cross Sections The general problem of homogenization can be illustrated by considering a symmetric, repeating array of fuel and moderator elements of volumes VF and VM . The average absorption cross section for the fuel–moderator unit cell is 0 1  F φF VF + aM φM VM  F + aM (VM /VF )ξ = a a cell = a φF VF + φM VM 1 + (VM /VF )ξ (14.1) where φM φF ξ≡ (14.2) is referred to as the flux disadvantage factor, and φF and φM are the average neutron fluxes in the fuel and moderator, respectively. The homogenized cell average cross section of Eq. (14.1) is equivalent in the sense that if it is multiplied by the exact average cell flux φcell = φF VF + φM VM φF VF + φM VM ≡ V F + VM Vcell (14.3) and the cell volume, the result will be the exact absorption rate in the cell: 0 a 1 φ V cell cell cell ≡ aF φF VF + aM φM VM (14.4) This type of definition can obviously be extended to a multiregion heterogeneous assembly by defining 0 a 1 = cell a1 + I I 1+ i i=2 a (Vi /V1 )ξi I i=2 (Vi /V1 )ξi φi Vi φcell ≡ i=1 ≡ I i=1 Vi I i=1 φi Vi Vcell , ξi ≡ φi φ1 (14.5) 14.2 ABH Collision Probability Method The same type of definition defines equivalent cell average fission and scattering cross sections. The appropriate definition of the cell average diffusion coefficient is less straightforward. An equivalent cell average diffusion coefficient must represent the net leakage from the cell, but that depends on the calculational method which will be employed for that purpose. We will return to this subject when we consider the detailed homogenization procedures. Thus the problem of cell homogenization reduces to the problem of determining the flux disadvantage factors, ξi , which will enable the homogenized model to predict the correct intracell reaction rates, and of determining the equivalent diffusion coefficients (or other leakage representation) which will enable the homogenized model to predict correct intercell leakage. Note that it is only necessary to know the relative value of the neutron flux in the different regions of the problem, not the absolute values, in order to calculate homogenized cross sections, which enables calculation of homogenized cross sections for local regions in a reactor before the absolute value of the flux is determined from a global calculation (which utilizes the homogenized cross sections). Calculation of the flux disadvantage factors from diffusion theory was discussed in Chapter 4. We turn now to methods that can be used when diffusion theory does not provide an adequate treatment of the heterogeneous problem, which is the usual case in a nuclear reactor. 14.2 ABH Collision Probability Method The ABH collision probability method (named after its originators—Ref. 15) found widespread use for calculation of thermal disadvantage factors before the availability of assembly transport codes (discussed in Section 14.4) and provides physical insight into the unit cell transport problem. A unit cell of fuel (F ) and moderator (M) with zero net current on the cell boundary is assumed. It is further assumed that the neutron slowing-down source is uniform in the moderator and zero in the fuel. We define PF M ≡ average probability that a neutron born uniformly and isotropically in region F will eventually be absorbed in region M PF ≡ average probability that a neutron born uniformly and isotropically in region F escapes from the fuel before being absorbed βM ≡ conditional probability that a neutron, having escaped from F into M, will then be absorbed in M A similar probability PMF can be defined for region M. As discussed in Chapter 11, there exists a reciprocity relation VF aF PF M = VM aM PMF (14.6) 517 518 14 Homogenization Since neutrons are only slowing down to thermal in the moderator, the reciprocity relation can be used to write the thermal utilization factor in terms of PF M : f ≡ PF M = VF aF aF VF φF PF M = F M VM a a VF φF + aM VM φM and to write the thermal disadvantage factor   φM 1 aF VF 1 aF VF ξ≡ −1 = = M − M φF a VM f PF M a VM (14.7) (14.8) The probability that a neutron born uniformly and isotropically in the fuel escapes into the moderator without making a collision was discussed in Chapter 11 and is given approximately by PF 0  1 1 + 4(VF tF /SF ) (14.9) where SF is the surface area of the fuel. If the neutron does not escape but has a scattering event [probability (1 − PF 0 )sF /tF ], it also has a probability PF 0 of escaping without a second collision. Continuing this line of argument, the total probability that the neutron escapes from the fuel into the moderator may be written   F 2  s F + · · · PF = PF 0 1 + (1 − PF 0 ) sF + (1 − PF 0 )2 t tF 1 PF 0 (14.10) = = 1 − (1 − PF 0 )(sF /tF ) 1 + aF (1 − PF 0 )/tF PF 0 A somewhat more accurate expression, which takes into account the nonuniform distribution of the first collisions for a cylindrical fuel rod of radius a, is *−1     F 2  aF 1 − PF 0 sF s F + aa PF = 1 + P +a 1+α F +β (14.11) PF 0 t t tF where the parameters α and β are as given in Fig. 14.1. It is apparent from the definitions that PF M = PF βM (14.12) Equation (14.7) can be rearranged to 1 1 − f − βM  M VM 1 + − 1 = aF f a VF PF f = 1 − PMF βM aM VM 1 + − aF VF PF PMF PMF (14.13) Using the reciprocity relation of Eq. (14.6) and, for the purpose of estimating βM only, approximating PF ≈ PF 0 ≈ SF /4VF aF , yields an approximation for the conditional probability: 14.2 ABH Collision Probability Method Fig. 14.1 Parameters α and β for use in calculation of ABH cylindrical escape probability. (From Ref. 11; used with permission of Wiley.) βM  4aM VM PMF SM (14.14) We approximate PMF ≈ PM (≡ probability that a neutron born in the moderator escapes from the moderator before being absorbed). We calculate PM by solving the diffusion equation in the moderator −DM ∇ 2 φM (r) + aM φM (r) = qM (14.15) where qM is the uniform slowing-down density in the moderator. The boundary conditions for Eq. (14.15) are symmetry at the cell boundary and a transport boundary condition at the fuel–moderator interface: Cell boundary: Fuel–moderator interface: ns · ∇φM = 0   1 ∇φM ns · φM = a 1 d (14.16) where ns is the unit vector normal to the surface and d is a transport parameter related to the transport mean free path in the moderator and is given in Fig. 14.2 for a cylindrical unit cell. Assuming that all neutrons diffusing from the moderator into the fuel are absorbed, PM is just the total neutron flow into the fuel from the moderator divided by the total neutron source in the moderator: −1   J M SF SF adVM a b PM = out = ns · DM ∇φM |a = + E , qM V M qM V M LM LM 2VF L2M (14.17) where a is the thickness or radius of the fuel region, b is the thickness of the moderator region associated with a fuel element, L2M = DM /aM , and E(a/LM , b/LM ) is the lattice function given in Table 3.6. 519 520 14 Homogenization Fig. 14.2 Transport boundary condition for cylinders. (From Ref. 11; used with permission of Wiley.) With these approximations for βM and PMF given by Eqs. (14.14) and (14.17), respectively, and the expression for PF given by Eq. (14.11), Eq. (14.13) can be evaluated:     F 2 * sF 1 aF 1 − PF 0 s aM VF F 1+α F +β 1+ F − at − 1= F f a VF PF 0 t t tF     ad a b M VM + −1 (14.18) − aa +E , VF LM LM 2L2M the disadvantage factor can be calculated from ξ= aF VF aM VM   1 −1 f (14.19) and the homogenized cross section can be calculated from Eq. (14.1). Although we have developed the ABH method in the context of thermal neutrons, the same general procedure can be applied to homogenize cross sections for any group of a multigroup scheme. 14.3 Blackness Theory Blackness theory refers to a class of methods for matching an approximate (e.g., diffusion theory) solution in one region to a very accurate solution of the transport 14.3 Blackness Theory equation in an adjacent region in order to obtain an effective diffusion theory cross section that will preserve the transport theory accuracy in the calculation of reaction rate. Such a procedure is required in order to treat control rods, lumped burnable poisons, and so on, within the context of multigroup diffusion theory. Consider a purely absorbing slab occupying the region xi ≤ x ≤ xi+1 . The onespeed transport equation within the absorbing slab is μ ∂ ψ(x, μ) + a ψ(x, μ) = 0 ∂x (14.20) This equation may be solved for the exiting neutron fluxes ψ + (xi+1 , μ) to the right and ψ − (xi , μ) to the left in terms of the entering fluxes from the left ψ + (xi , μ) and from the right ψ − (xi , μ), where the +/− denotes μ > 0/μ < 0:    , 1>μ>0 ψ + (xi+1 , μ) = ψ + (xi , μ) exp −a μ (14.21)    − − , −1 < μ < 0 ψ (xi , μ) = ψ (xi+1 , μ) exp a μ where  = xi+1 − xi . The incident fluxes into the purely absorbing region are assumed to have the Pl form that is consistent with a diffusion theory solution in the adjacent fuel-moderating region: ψ + (xi , μ) = φ(xi ) + 3μJ (xi ), 1>μ>0 ψ − (x −1 < μ < 0 i+1 , μ) = φ(xi+1 ) + 3μJ (xi+1 ), The currents at the surfaces of the absorbing region can be written  1 dμ μψ(xi , μ) J (xi ) ≡ −1 2  1  0 dμ dμ μψ − (xi , μ) + μ φ(xi ) + 3μJ (xi ) = 2 −1 2 0  J (xi+1 ) ≡  = 1 −1 0 −1 (14.22) (14.23) dμ μψ(xi+1 , μ) 2 dμ μ φ(xi+1 ) + 3μJ (xi+1 ) + 2  1 0 dμ μψ(xi+1 , μ) 2 Using Eqs. (14.21) to evaluate the exiting fluxes, these equations become 1 1 3 1 J (xi ) = φ(xi ) − E3 (a )φ(xi+1 ) + E4 (a )J (xi+1 ) 2 4 2 2 (14.24) 1 1 1 3 − J (xi+1 ) = φ(xi+1 ) − E3 (a )φ(xi ) − E4 (a )J (xi ) 2 4 2 2 where E3 and E4 are the exponential integral functions,    1 a  En+2 (ξ ) ≡ dμ μn exp − μ 0 (14.25) 521 522 14 Homogenization Equations (14.24) can be rearranged to define the blackness parameters J (xi ) − J (xi+1 ) 1 − 2E3 (a ) = φ(xi ) + φ(xi+1 ) 2[1 + 3E4 (a )] α(a ) ≡ 1 + 2E3 (a ) J (xi ) + J (xi+1 ) = β(a ) ≡ φ(xi ) − φ(xi+1 ) 2[1 − 3E4 (a )] (14.26) The parameter α is the ratio of the average inward current to the average flux at the surface of the absorbing slab. This quantity is used as a boundary condition for the diffusion theory calculation in the adjacent region, JM D M ∇φM =− =α φM φM (14.27) (e.g., the transport parameter d of the ABH method is d/λtr = 1/3α). This transport boundary condition was used in Chapter 3 to derive an effective diffusion theory cross section for the control rod: ac = aM M a[(a /α) + (1/LM ) coth(a/LM )] − 1 (14.28) where a is the half-thickness of the fuel–moderator region denoted by M. Since this development was for a purely absorbing slab, the results are valid at any energy, provided that the cross sections for that energy are used. For a purely absorbing slab with a spatially dependent absorption cross section, the results above are valid if the following replacement is made:  a  → xi+1 xi a (x) dx (14.29) 14.4 Fuel Assembly Transport Calculations Pin Cells A fuel assembly consists of a large number of fuel pins of differing fuel loading, enrichment, burnup, and so on, each of which is clad and surrounded by moderator and perhaps other elements, such as structure and burnable poisons, as depicted in Fig. 14.3. At this most detailed level of heterogeneity, the assembly can be considered to be made up of a large number of units cells, or pin cells, consisting of a fuel pin, cladding, surrounding moderator, and perhaps structure and burnable poison. The first step in homogenizing the fuel assembly is to homogenize each of the pin cells, by calculating the multigroup flux distribution across the fuel, clad, moderator, and so on, and using it to calculate volume-averaged cross sections for the pin cell. 14.4 Fuel Assembly Transport Calculations Fig. 14.3 Representative fuel assembly. (From Ref. 16; used with permission of American Nuclear Society.) If the pin cell can be considered to be one of a large number of identical pin cells, reflective symmetry boundary conditions can be used. However, this assumption becomes questionable in the vicinity of gaps, control pins, burnable poisons, or fuel pins of very different composition (e.g., MOX pins near UO2 pins). The influence of the surrounding environment can be introduced into the pin-cell calculation by specifying the partial inward current J − (and a zero reflection, or black boundary conditions) or the net current J = J + − J − (and a perfectly reflecting boundary condition) on the cell boundary. Wigner–Seitz Approximation If the cell associated with each pin is defined symmetrically and such that the cells fill the volume of the assembly, the pin-cell boundary will have a noncylindrical shape depending on the lattice geometry, generally square or hexagonal. Since the pin geometry is cylindrical, it is convenient to approximate the actual pin-cell geometry by an equivalent cylindrical cell that preserves moderator volume. The approximate Wigner–Seitz cell has a radius R that depends on the pin-to-pin distance p as R = p/π 1/2 for a square pitch fuel lattice and R = p(31/2 /2π)1/2 for a hexagonal pitch lattice. The change in geometry can lead to an anomalously high flux in the moderator of a cell with reflective boundary conditions because a neutron introduced into the cell traveling in the direction of a chord that does not pass through the innermost n shells before intersecting the reflecting cell boundary will never pass through these innermost n shells since spectral reflection from the cylindrical wall will result in 523 524 14 Homogenization Fig. 14.4 Reflection misrepresentation in Wigner–Seitz approximation. (From Ref. 16; used with permission of American Nuclear Society.) motion along a similar chord. On the other hand, as shown in Fig. 14.4, correct reflection from a square or hexagonal boundary will cause motion into the innermost shells. This problem can be corrected by “white reflection” in a cosine distribution with respect to the inward normal. Collision Probability Pin-Cell Model The collision probability methodology of Section 9.3 can be extended to handle the albedo (partial reflection) and incident current conditions that enable the environment to influence the pin-cell calculations. With reference to Fig. 14.5, consider a cylindrical pin-cell consisting of i annular regions. Using the notation of Section 9.3, define the probability, γ0i , that an uniformly distributed isotropic flux of neutrons at the external surface (SB ) of the pin-cell will suffer a first collision in region i before exiting across surface SB : γ0i ≡ ti Vi dri SB drB ⊃Vi d(n · )(1/4π)e−α(ri ,rB ) SB ·n<0 d(n · )(1/4π) (14.30) where  ⊃ Vi indicates those values of  that intersect the volume Vi , n is the outward unit vector to the surface SB , α(ri , rB ) is the optical distance (e.g., distance measured in mean free paths) along the chord from rB to ri , and n · /4π is the rate at which neutrons in an isotropic flux of unit strength will cross the surface at SB into the pin-cell. This probability is related to the first-flight escape probability, P0i , that a neutron introduced in volume Vi will exit the pin-cell across surface SB without a collision: P0i ≡ Vi dri SB drB ⊂Vi 4π d d(n · )(1/4π)e−α(rB ,ri ) Vi dri (1/4π) (14.31) where 1/4π is the isotropic angular flux corresponding to unit scalar flux in Vi and  ⊂ Vi indicates those values of  for which a neutron could have reached rB on a 14.4 Fuel Assembly Transport Calculations Fig. 14.5 Cylindrical pin-cell model. first flight from within volume Vi . Except for the ti , the numerators are identical, reflecting the fact that the probabilities for neutrons traveling from the surface into volume Vi without collision and traveling in the opposite direction from within Vi to the surface are identical. This allows Eq. (14.30) to be written γ0i = ti Vi P0i 0 2π SB −1 dμ 0 dφ(n · /4π) = ti Vi P0i 1 4 SB (14.32) In terms of the probability P ij /ti Vi that neutrons introduced uniformly and isotropically within volume Vi have their first collision in volume Vj , this may be written % & % & I I   4 P ij ti Vi ij 1− = ti Vi − (14.33) P γ0i = 1 ti Vi SB 4 SB j =1 j =1 A similar line of argument leads to the result that the probability, Ri , that a uniformly distributed isotropic flux of neutrons crossing the external surface (SB ) of the pin-cell will be removed (absorbed or scattered to another group) by a collision in region i before exiting across surface SB is related to the total escape probability, Pi , that a neutron introduced in volume Vi will escape (perhaps after multiple collisions) out of the pin-cell across surface SB : Ri = ri Vi Pi 1 4 SB (14.34) We now wish to construct source and current flux response functions in terms of which the flux in any one of the annular regions of the pin-cell can be constructed: φi = I  k=1 Q k X ki (β) + jex− Yi (β) (14.35) 525 526 14 Homogenization where Q k is the neutron source density in annular region k, X ki (β) is the neutron flux produced in region i by a unit neutron source density in region k, taking into account possible multiple reflections at the cell boundary with albedo β, and Y i (β) is the neutron flux produced in annular region i by unit neutron inward current across the cell boundary. The quantities X ki (0) and Yi (0) refer to the response functions above when the albedo of the region surrounding the pin-cell is zero (i.e., when there is no reflection of neutrons exiting the pin-cell back across surface SB ). The response functions X ki (β) and Y i (β) can be calculated in terms of X ki (0) and Yi (0) and the albedo, β. For a neutron incident into the pin-cell across the boundary SB , the cell has an effective albedo (1 − R), where R= I  j =1 Ri = I  ri Vi Yi (0) (14.36) i=1 is the total removal (r is the cross section for absorption plus scatter to another group) probability for a neutron incident on the cell from outside. For a cohort of incident neutrons, a fraction R is removed and a fraction (1 − R) is returned to the boundary SB . Of the (1 − R) returned to the boundary, a fraction β (the albedo of the surrounding assembly for neutrons exiting the pin-cell) is reflected back into the pin-cell. Of the fraction (1 − R)β that enter the cell for a second time, a fraction R is removed and a fraction (1 − R) return to the surface SB a second time, and so on. Thus an inward partial current of neutrons incident across SB is effectively amplified by the factor 1 + (1 − R)β + [(1 − R)β]2 + · · · = 1/[1 − (1 − R)β]. If Y i (0) is the neutron flux produced in annular region i by unit neutron inward current across the cell boundary, without taking into account reflection of exiting neutrons back into the pin-cell, the neutron flux due to a unit inward, current taking reflection into account, is Yi (β) = Yi (0) 1 − β(1 − R) (14.37) The flux X ki (β) in volume Vi due to a unit neutron source density in volume Vk is made up of two components: the flux X ki (0) due to source neutrons from volume Vk which have not been reflected from the boundary SB , and the flux due to the number of source neutrons Pk Vk from volume Vk which do reach the boundary and are reflected with albedo β. These reflected neutrons can be treated as an incoming flux, and the flux produced by it in volume Vi is found by multiplying by Yi (β). The resulting expression is X ki (β) = X ki (0) + βPk Vk Yi (0) 1 − β(1 − R) (14.38) The collision probability equations (9.54) were derived under the implicit assumption of no reflection from the external boundary (i.e., β = 0) and no incident current. Thus these equations are suitable for calculating the basic response functions X ki (0) and Yi (0) when a first collision source term to account for incident 14.4 Fuel Assembly Transport Calculations partial current density jex− is included: ti Vi φi = I   P ji j =1 ≡ I  j =1 Qj (sj + νfj )φj + tj tj  + γi0 jex−   Qj + γi0 jex− P j i cj φj + tj (14.39) The collision probabilities P j i for a cylindrical cell are given by Eqs. (9.63) to (9.65). In some applications it may be more convenient to treat the fission neutron source as a fixed source and include it in the Qj term. The quantities X ki (0) satisfy this equation with a unit source density in volume Vk only and no incident current density: ti Vi X ki (0) = I  P j i cj X kj (0) + j =1 P ki , k i = 1, . . . , I, k = 1, . . . , I (14.40) This constitutes a set of I 2 equations to be solved for the X ki (0). The quantities Yi (0) satisfy Eq. (14.39) with no volumetric source but with a unit external current density: i Vi Yi (0) = I  P j i cj Yj (0) + γi0 , i = 1, . . . , I (14.41) j =1 a set of I equations to be solved for the Yi (0). In summary, the pin-cell calculation consists of: (1) solve Eqs. (14.40) and (14.41) for the isolated pin-cell flux response functions X ki (0) and Yi (0); (2) construct the flux response functions X ki (β) and Yi (β) which take into account reflection from the surrounding medium by the albedo β from Eqs. (14.37) and (14.38); (3) calculate the flux in each annular region of the pin-cell using Eq. (14.35); and (4) construct homogenized cross sections for the cell using Eq. (14.5). Interface Current Formulation The outward partial current density from the pin-cell across surface SB consists of two components: (1) the source neutrons which are introduced within the pin-cell  and which are crossing SB for the first time ( Ii=1 Pi Vi Qi ), and (2) the incident neutrons (jex− ) which traverse the pin-cell without being removed with probability (1 − R)—and both components are reflected with probability β and constitute an inward current that may traverse the cell without removal, and so on. The total outward partial current density due to neutron sources within the pin-cell and neutrons incident on the pin-cell from the surrounding medium is + (β) = jout I + (1 − R)jex− 1 − β(1 − R) i=1 Vi Pi Qi (14.42) 527 528 14 Homogenization The inward partial current density across surface SB also has two components: (1) the source neutrons that escape from the pin-cell to reach SB for the first time  ( Ii=1 Pi Vi Qi ) and are reflected with probability β, and (2) the incident neutrons (jex− ), both of which may traverse the pin-cell without removal with probability (1 − R) to reach surface SB and be reflected with probability β, and so on. The total incident partial current density is Jin− (β) = β I + jex− 1 − β(1 − R) i=1 Vi Pi Qi (14.43) The net current density (in the outward direction) across the surface of the pin-cell is  (1 − β) Ii=1 Vi Pi Qi − Rjex− + j (β) ≡ jout (β) − jin− (β) = (14.44) 1 − β(1 − R) Multigroup Pin-Cell Collision Probabilities Model The pin-cell model above extends immediately to multigroup by making the reg g g→g (i.e., group removal cross section), placements ti → ti , ri → ti − si g ji g j i γ0i → γ0i , P → Pg and Ri → Ri , and extending certain equations to multigroup. Equations (14.40) become g ti Vi Xgki (0) = I  G ji Pg g  →g g  =1 (sj g ji + χ g νf )Xg  (0) + δj k g tj j =1 g = 1, . . . , G; i, k = 1, . . . , I , (14.45) which can be written in matrix notation as Vi ti X ki (0) = I  ji P SF X j i (0) + P j i , i, j = 1, . . . , I (14.46) j =1 and Eqs. (14.41) become g g ti Vi Yi (0) = I  G ji Pg g  →g g  =1 (sj j =1 g g + χ g νfj )Yj (0) g tj g = 1, . . . , G; i, j = 1, . . . , I g + γ0i , (14.47) which can be written in matrix notation as Vi ti Yi (0) = I  ji P SF Y j (0) + γ0i , i = 1, . . . , I (14.48) j =1 Equations (14.37) and (14.38), with the appropriate group cross probabilities, can g be used to correct the basic flux response functions Xgki (0) and Yi (0) to account 14.5 Homogenization Theory for reflection from the surface SB , and the multigroup fluxes in each region of the pin-cell can be calculated from the multigroup version of Eq. (14.35): g φi = I  g −g g Qk Xgki (β) + jex Yi (β), i = 1, . . . , I (14.49) k=1 Resonance Cross Sections Homogenized resonance cross sections are calculated at the pin-cell level using the methods discussed in Chapter 11. Full Assembly Transport Calculation Once the finest level of heterogeneity has been homogenized with a series of pincell calculations, the assembly is made up of a large number of homogeneous regions (e.g., the square pin-cells of Fig. 14.3), surrounded by structure, water gaps, control rods, other dissimilar assemblies, and so on (i.e., the assembly is still a heterogeneous medium embedded in a larger-scale heterogeneous medium, the reactor core). The next step in the homogenization process is to perform a multigroup transport calculation on the pin-cell-homogenized assembly for the purpose of obtaining average group fluxes for each homogenized pin-cell that can be used to calculate homogenized cross sections that will allow the entire assembly to be represented as a homogenized region. Any of the transport methods discussed in Chapter 9 (collision probabilities, discrete ordinates, Monte Carlo) or even diffusion theory in some cases can be used for the full assembly transport calculation. Such calculations are normally performed using reflective conditions on the assembly boundary, or more correctly on the boundary defined by the centerline of the water gap or other medium separating adjacent assemblies, thus implicitly assuming an infinite array of identical assemblies. The fact that different assemblies have different homogenized properties is taken into account in the global core calculation based on a homogenized assembly model which follows assembly homogenization. However, the fact that the adjacent assembly is dissimilar or that there is a control rod nearby or that there is significant leakage out of or into an assembly affects the assembly calculation and hence the homogenized properties of the assembly. Stratagems such as extending the boundaries for an assembly calculation into adjacent assemblies or over a larger planar region have evolved for dealing with this problem. 14.5 Homogenization Theory When used in the calculation for which they were intended, homogenized cross sections, should yield a result that is equivalent, in some sense, to the result that would have been obtained if the calculation could be performed with all the spatial 529 530 14 Homogenization detail without the need for homogenization. It is useful, in this regard, to develop homogenization procedures that would preserve the essential integral properties of a global heterogeneous transport calculation, the result of which is assumed known for the purpose of development of homogenization procedures, and then to evaluate the homogenized cross sections using an approximation to the global heterogeneous transport solution. Homogenization Considerations The neutron flux distribution and effective multiplication constant, k, can be described exactly by multigroup transport theory, which we write in the general form g ∇ · Jg (r) + t (r)φg (r) = G G  χ g  g  νf (r)φg  (r) +  g →g (r)φg  , k  g = 1, . . . , G (14.50) g =1 g=1 Imagine that we know the solution to Eq. (14.50) and wish to use it to define homogenized cross sections which when used in the solution of the homogenized transport equation ˆ tg (r)φ̂g (r) ∇ · Ĵg (r) +  = G χg  k̂ g  =1 G  g ˆ (r)φ̂g  (r) + ν f ˆ g  →g φ̂g  ,  g = 1, . . . , G (14.51) g  =1 yield the same result for certain important quantities as would be obtained if the detailed cross sections and the exact solution of Eq. (14.50) were used in their evaluation (i.e., preserves certain properties of the exact solution). The most important quantities to be preserved are the multiplication constant, k, the group reaction rates averaged over the homogenization region, and the group currents averaged over the surface of the homogenization region. Preservation of the last two quantities requires that   g ˆ xg (r)φ̂g (r) dr =  x (r)φg (r) dr (14.52)  Vi Sik Vi  Ĵg (r) · dS = Sik Jg (r) · dS (14.53) where Vi is the volume of the homogenization region i and Sik is the kth surface of the homogenization region i. Satisfaction of Eqs. (14.52) and (14.53) would also ensure preservation of k. If the homogenized cross sections are uniform over the homogenization region, an exact definition is g ˆ xgi ≡  Vi x (r)φg (r) dr Vi φ̂g (r) dr (14.54) 14.6 Equivalence Homogenization Theory and when diffusion theory is to be used in the homogenized calculation, g D̂ik ≡ − Sik Sik Jg (r) · dS ∇ φ̂g (r) · dS (14.55) The practical difficulty in using Eqs. (14.54) and (14.55), of course, is that the exact solution of the global transport equation is not known (and never will be, or we would not be bothering with homogenization) and the homogenized solution of the global diffusion equation is not known prior to solving Eq. (14.51), which requires the homogenized group constants as input. Another conceptual problem is that the integrals in Eq. (14.55) will generally be different for each surface, k, so that it is not possible to define a constant value of the homogenized diffusion coefficient which preserves the surface-averaged currents over all the surfaces. Conventional Homogenization Theory The conventional pin-cell or assembly homogenization procedure approximates the solution to the global core transport equation, φ g (r) and Jg (r), with the solug g tions, φA (r) and JA (r), to a pin-cell or assembly transport calculation, usually with g symmetry boundary conditions, n · JA (r) = 0. The numerator of Eq. (14.54) is then g evaluated using φA (r) instead of the (unavailable) exact global transport solution g φ g (r). This assembly transport solution, φA (r), is also used to evaluate the flux integral in the denominator of Eq. (14.54). A possible choice of the homogenized diffusion coefficient is g g D̂iA ≡ Vi D g (r)φA (r) dr g Vi φA (r) dr (14.56) Rather large errors have been found in calculations that employed these conventional homogenization methods when compared with exact solutions for benchmark problems. The major source of error is in the treatment of the homogenized diffusion coefficients and the imposition of continuity of flux and current continuity boundary conditions at interfaces between homogenization regions. The source of the problem is that the homogenized diffusion equation, with continuity of current and flux imposed at interfaces, lacks sufficient degrees of freedom to preserve both surface currents and reaction rates. 14.6 Equivalence Homogenization Theory It is possible to require that both the volume-integrated reaction rates and the surface-integrated currents from the heterogeneous problem be preserved in the homogenized problem [i.e., that Eqs. (14.52) and (14.53) be satisfied] if the conti- 531 532 14 Homogenization Fig. 14.6 Equivalence theory notation. nuity of flux condition is relaxed. Instead of continuity of flux, the flux interface condition − − φ̂i+ (xi+1 )fi + (xi+1 ) = φ̂i+1 (xi+1 )fi+1 (xi+1 ) (14.57) is imposed at the interface at xi+1 between homogenization regions i and i + 1, − (xi+1 ) are the homogenized fluxes in homogenization where φi+ (xi+1 ) and φi+1 regions xi ≤ x ≤ xi+1 and xi+1 ≤ x ≤ xi+2 , respectively, both evaluated at the in− (xi+1 ) refers terface xi+1 between the two, as indicated in Fig. 14.6. Similarly, fi+1 to the flux discontinuity factor at the lower (minus) interface xi+1 of the region xi+1 ≤ x ≤ xi+2 , and fi + (xi+1 ) refers to the flux discontinuity factor at the upper (plus) interface xi+1 of the region xi ≤ x ≤ xi+1 . The flux discontinuity factors on each side of the interface at xi+1 are defined by the ratios of the heterogeneous to homogeneous fluxes at this interface: fi + (xi+1 ) = φi+ (xi+1 ) φ̂i+ (xi+1 ) − (xi+1 ) = fi+1 , − φi+1 (xi+1 ) (14.58) − φ̂i+1 (xi+1 ) Equations (14.57) and (14.58) express the requirement that the heterogeneous flux is continuous at the interface and relate the homogeneous to heterogeneous fluxes at the interface. The discontinuity factors introduce additional degrees of freedom into the homogenization procedure, which permits the satisfaction of Eqs. (14.52) and (14.53). Let us now consider the implementation of equivalence theory. For the moment, we continue to assume the existence of an exact heterogeneous solution for the entire core. The evaluation of homogenized cross sections from Eq. (14.54) is straightforward. We examine implementation of the requirement of Eq. (14.53) for the homogenized multigroup diffusion equation in two dimensions: ˆ φ̂ (x, y) −∇ · D̂ij ∇ φ̂ij (x, y) +  tij ij g = G  g  =1 g g g G g      ˆ g →g φ̂ g (x, y) + χ ˆ g φ̂ g (x, y),  ν ij f ij ij ij k  g = 1, . . . , G (14.59) g =1 where the homogenized cross sections for homogenization region (i, j ) have been calculated from Eq. (14.54) and both homogenized cross sections and diffusion 14.6 Equivalence Homogenization Theory coefficients are constant within region (i, j ). Integrating this equation over the y-dimension of the homogenization region (i, j ), which is defined by xi ≤ x ≤ xt+1 and yj ≤ y ≤ yj +1 , yields  yj +1 2  yj +1 d2 g g d g g dy φ̂ij (x, y) − D̂ij dy 2 φ̂ij (x, y) −D̂ij 2 dx yj dy yj  yj +1 g ˆ tg dy φ̂ij (x, y) + yj = G   g  =1 ˆ g →g  ij  yj +1 yj g dy φ̂ij (x, y) +  yj +1 G  χg  g ˆg ν dy φ̂ij (x, y), f ij k  yj g =1 g = 1, . . . , G (14.60) Since the heterogeneous solution is assumed to be known, the heterogeneous g y-direction leakage (Lijy ) is known, in principle, and may be used to evaluate the y-direction leakage term in Eq. (14.60); that is,  yj +1 d2 g g g g L̂ijy (x) ≡ −D̂ij dy 2 φ̂ij (x, y) = Lijy (x) dy yj  yj +1 d g dy (14.61) J (x, y) = J g (x, yj +1 ) − J g (x, yj ) ≡ dy yi Furthermore, the known values of the heterogeneous currents (Jg ) at xi+1 and xi can be used as boundary conditions for the solution of Eq. (14.60) in the homogenization region (i, j ):  yj +1  yj +1 g d g −D̂ij dy φ̂ij (xi+1 , y) = Jg (xi+1 , y) dy dx yj yj (14.62)  yj +1  yj +1 g d g dy φ̂ij (xi , y) = Jg (xi , y) dy −D̂ij dx yj yj With the (assumed) known values of the heterogeneous fluxes at the interfaces and the calculated values of the homogeneous flux integrals, the discontinuity factors for region (i, j ) at the surfaces at xi+1 and at xi can be calculated as the ratio of heterogeneous-to-homogeneous flux integrals: fig+ = ˆ where ˆ g (x) ≡ i g i (xi+1 ) , g i (xi+1 )  yj +1 yj fig− = g dy φ̂ij (x, y), ˆ g i (xi ) g i (xi ) g i (x) = (14.63)  yj +1 dy φ(x, y) (14.64) yj The global heterogeneous solution will not be known, of course, so the practical implementation of the prescriptions above requires their approximation using a local heterogeneous solution for an assembly or set of assemblies, usually performed 533 534 14 Homogenization with a zero current boundary condition. It is important that the same approximate heterogeneous solution be used to evaluate the leakage term of Eq. (14.61) in Eq. (14.60), to evaluate the boundary conditions of Eq. (14.62) for Eq. (14.60), and to evaluate the numerators of the flux discontinuity factors. A similar procedure yields the flux discontinuity factors fj+ and fj−+1 for region (i, j ) at the surfaces at y = yj and y = yj +1 . The four different flux discontinuity factors for region (i, j ) will in general be different. Note that this procedure can be implemented for any arbitrary definition of the homogenized diffusion coefficient. The choice of diffusion coefficient will, of course, affect the solution for the homogeneous flux in the calculation above, hence affect the value of the computed flux discontinuity factor. A common choice for the homogenized diffusion coefficient is the simple heterogeneous flux-weighted value: g D̂ij = xi +1 y dx yjj +1 dy D g (x, y)φ g (x, y) xi xi +1 y dx yjj +1 dy φ g (x, y) xi (14.65) The calculation of flux discontinuity factors can be implemented by using assembly calculations of both the heterogeneous and homogeneous fluxes and currents. The volume integral of flux over the assembly can be normalized to be the same in both calculations. If the homogeneous assembly calculation is carried out with zero current symmetry boundary conditions, the homogeneous flux distribution is uniform within the homogenization region. Under these approximations, the flux discontinuity factor can be calculated entirely from the results of the heterogeneous assembly calculation as the ratio of the surface integral of the heterogeneous assembly flux to the volume integral of the heterogeneous flux, as may be seen by considering yj +1 yj xi+1 xi dx g dy φA (xi+1 , y) yj +1 yj g dy φA (x, y) = = g xi+1 xi dx g i+1 (xi+1 ) g yj +1 dy φ̂ij (x, y) yj g i+1 (xi+1 ) g xi ˆ ij ≡ fig+ x (14.66) where φA (x, y) is the heterogeneous flux from the assembly calculation, the common normalization of the heterogeneous and homogeneous fluxes has been used in the second step, and the uniformity of the homogeneous assembly flux with symmetry boundary conditions has been used in the third step. The discontinuity factors calculated from Eq. (14.66), referred to as assembly discontinuity factors, will be accurate for assemblies in which the net current almost vanishes over the boundaries, but will be inaccurate for conditions in which there is significant leakage across assembly interfaces; this is an area of active research. This formulation of equivalence theory is appropriate for any nodal method that uses surface-averaged fluxes [e.g., the quantities defined by Eq. (14.64)] in evaluat- 14.7 Multiscale Expansion Homogenization Theory ing node-to-node coupling. The expression for the nodal interface current on the interface at xi+1 between nodes (i, j ) and (i + 1, j ) is g + Jg(i,j )= g 2Dij Di+1j fgi+ g i − − fgi+1 g i+1 g g − xi xi+1 fgi+1 Dij /xi + fgi+ Di+1j /xi+1 (14.67) Similar expressions obtains for the other nodal interfaces. 14.7 Multiscale Expansion Homogenization Theory A more formal development of homogenization theory builds on the spatial structure typical of a nuclear reactor, a repeating array of highly heterogeneous fuel assemblies within an almost periodic (symmetric) configuration with assemblyaveraged properties that vary slowly from assembly to assembly. This suggests the introduction of two spatial scales—the fine scale of the intra-assembly heterogeneity (rf ) and the coarse scale of the global inter-assembly variation (rc )—which are treated as independent spatial variables. The multiscale homogenization theory will be illustrated with one-group diffusion theory, the governing equation for which is written with rf and rc as formally independent spatial variables:     ∂ ∂ ∂ ∂ − · D(rc , rf ) (rc , rf ) + + ∂rc ∂rf ∂rc ∂rf 1 + a (rc , rf ) (rc , rf ) − νf (rc , rf ) (rc , rf ) = 0 k (14.68) Normalized to a core average diffusion length, L, the spatial gradients are of different order: O(Ld/drc ) ∼ O(Lrf /rc d/drf ) ∼ εO(Ld/drf ), where ε ≡ rf /rc is a small parameter on the order of the ratio of the scale lengths of the intra-assembly heterogeneity to the assembly dimensions. Making flux and eigenvalue expansions in powers of the small parameter ε, (rc , rf ) =  εn n (rc , rf ), n=0 1  εn = k kn (14.69) n=0 and substituting in Eq. (14.68) yields to leading order O(ε0 ): Lrf (rc , rf ) ≡ − − ∂ ∂ · D(rc , rf ) ∂rf 1 νf (rc , rf ) k0 0 (rc , rf ) ∂rf + a (rc , rf ) 0 (rc , rf ) = 0 0 (rc , rf ) (14.70) Equation (14.70) plus the periodic (symmetry) boundary conditions on an assembly defines the detailed heterogeneous intra-assembly flux for an assembly k; there will be K such heterogeneous assembly problems, corresponding to the K different fuel assembly types in the reactor core. The dependence on rc indicated in 535 536 14 Homogenization Eq. (14.70) is a dependence on the assembly for which the calculation is made; all intra-assembly spatial dependence is represented by the rf dependence. Since no spatial gradients with respect to rc occur in Eq. (14.70), the general solution is 0 (rc , rf ) = A0 (rc )φ0 (rc , rf ) (14.71) where A0 (rc ) is an arbitrary function of the global spatial scale parameter which will be determined from a higher-order equation. The first-order O(ε1 ) equation is Lrf 1 (rc , rf ) = ∂ ∂ · D(rc , rf ) φ0 (rc , rf )A0 (rc ) ∂rc ∂rf + ∂φ0 (rc , rf ) ∂ · D(rc , rf ) A0 (rc ) ∂rf ∂rc + 1 νf (rc , rf )φ0 (rc , rf )A0 (rc ) k1 (14.72) which is an inhomogeneous equation of the same form as the homogeneous Eq. (14.70). By the Fredholm alternative theorem, Eq. (14.72) has a solution only if the right side is orthogonal to the solutions of the equation that is adjoint to Eq. (14.70). Since this equation is self-adjoint for one-group diffusion theory (it is not for multigroup diffusion theory or transport theory) with periodic boundary conditions, a solvability condition for Eq. (14.72) is  6 5  ∂A0 ∂ ∂ 1 φ0 · = φ0 , νf φ0 A0 + φ0 , D+D k1 ∂rf ∂rf ∂rc 6 5 ∂φ0 ∂D ∂ ∂φ0 A0 + φ0 , · +D · ∂rf ∂rc ∂rf ∂rc (14.73) where · indicates a spatial integral over rf within node k. Equation (14.73) provides a calculation for k1 . The solution of Eq. (14.72) consists of a solution to the homogeneous equation, which is φ0 , with an arbitrary multiplier A1 (rc ), and particular solutions corresponding to the terms on the right side: 1 (rc , rf ) = A1 (rc )φ0 (rc , rf ) +  gξ (rc , rf )nξ · ξ ∂A0 (r0 ) ∂rc + q(rc , rf )A0 (rc ) (14.74) where the particular solutions satisfy  Lrf gξ (rc , rf ) = nξ ·  ∂ ∂ ∇φ0 (rc , rf ) + D(rc , rf ) φ0 (rc , rf ) ∂rf ∂rf (14.75) 1 Lrf q(rc , rf ) = νf (rc , rf )φ0 (rc , rf ) k1 14.7 Multiscale Expansion Homogenization Theory with periodic assembly boundary conditions. There is an equation of the form of the first of Eqs. (14.75) for each coordinate direction. The second-order O(ε2 ) equation is   ∂ ∂ ∂ ∂ ∂ ∂ ·D ·D + ·D Lrf φ2 = 0+ 1 ∂rc ∂rc ∂rf ∂rc ∂rc ∂rf 1 1 + νf 1 + νf 0 (14.76) k1 k2 which has a solvability condition 5  6 5 1 1 ∂ ∂ φ0 , νf 1 + νf 0 + φ0 , ·D k1 k2 ∂rc ∂rc  6 5  ∂ ∂ ∂ ∂ ·D + ·D + φ0 , 1 =0 ∂rf ∂rc ∂rc ∂rf 6 0 (14.77) that provides a solution for k2 . Integrating Eq. (14.76) over the rf intra-assembly heterogeneous spatial scale yields the global diffusion equation with parameters averaged over the fuel assembly:   ∂ ∂ 1 1 νf − a A0 (rc ) · D A0 (rc ) + 2 ∂rc ∂rc ε k 0 1 ∂ +  · A0 (rc ) + S A0 (rc ) = 0 (14.78) ∂rc where, defining the normalization N ≡ φ0 , φ0 , the appropriate assembly-averaged homogenized nu-fission and absorption cross section are flux-adjoint weighted with the detailed intra-assembly solutions φ0 , νf φ0 N φ0 , a φ0 = N νf = a the elements of the diffusion tensor for a two-dimensional problem are 6!  5 ∂ ∂ N D11 = φ0 , Dφ0 + φ0 , D g2 + Dg2 ∂rf 1 ∂rf 1  6! 5  ∂ ∂ g1 N D12 = φ0 , D+D ∂rf 1 ∂rf 1  6! 5  ∂ ∂ g2 N D+D D21 = φ0 , ∂rf 2 ∂rf 2 6!  5  ∂g1 ∂ + Dg1 N D22 = φ0 , Dφ0 + φ0 , D ∂rf 2 ∂rf 2 φ0 , Dφ0 D33 = N D13 = D23 = D31 = D32 = 0 (14.79) (14.80) 537 538 14 Homogenization there is a source that acts like an effective fission or absorption cross section,  6 5  1 ∂ ∂ ∂ ∂ q + φ0 , νf q ·D + ·D S = φ0 , ∂rc ∂rf ∂rf ∂rc k 6 5 ∂ ∂ 1 (14.81) φ0 , D φ0 − φ0 , νf φ0 − k ∂rc ∂rc and there is a convection term (defined in Ref. 1). The source and convection terms arise because of the assembly-to-assembly variation of cross sections and diffusion coefficient. These terms, which vanish for a reactor with exactly periodic conditions associated with each assembly, account for the effect of inter-assembly leakage between adjacent assemblies, which is not accounted for in the calculation of φ0 . Thus the solution of Eqs. (14.70) and (14.75), with periodic boundary conditions, for the detailed intra-assembly flux distribution φ0 and supplementary intranodal functions gξ and q can be used to calculate flux-adjoint-weighted homogenized assembly parameters for a consistently formulated global diffusion equation (14.78). This type of multiscale procedure can also be employed to develop a global diffusion equation based on assembly homogenization with transport lattice calculations replacing Eq. (14.70). 14.8 Flux Detail Reconstruction The homogenization procedure results in homogenized cross sections that can be used for an entire fuel assembly or collections of fuel assemblies (e.g., modules) in a full core calculation. The resulting flux distribution from the full core calculation reflects the global flux distribution, but not the local detailed flux distribution. The detailed assembly or module flux calculations that were used in the homogenization process must be superimposed on the global flux distribution, and the detailed pin-cell flux distributions must be further superimposed on the assembly or module flux distributions. It is important that the assumptions used in reconstructing the detailed flux distribution be consistent, if not identical, with the assumptions made in the homogenization process. References 1 H. Zhang, Rizwan-uddin, and J. J. Dorning, “Systematic Homogenization and Self-Consistent Flux and Pin Power Reconstruction for Nodal Diffusion Methods, Part I: Diffusion Theory Based Theory,” Nucl. Sci. Eng. 121, 226 (1995); “Transport-EquationBased Systematic Homogenization Theory for Nodal Diffusion Methods with Self-Consistent Flux and Pin Power Reconstruction,” J. Transport Theory Stat. Phys. 26, 433 (1997); “A Multiple-Scales Systematic Theory for the Simultaneous Homogenization of Lattice Cells and Fuel Assemblies,” J. Transport Theory Stat. Phys. 26, 765 (1997). 2 A. Hebert et al., “A Consistent Technique for the Global Homogenization of a Pressurized Water Reactor Problems 3 4 5 6 7 8 Assembly,” Nucl. Sci. Eng. 109, 360 (1991); “Development of a Third Generation SPH Method for the Homogenization of a PWR Assembly,” Proc. Conf. Mathematical Methods and Supercomputing in Nuclear Applications, Karlsruhe, Germany (1993), p. 558; “A Consistent Technique for the Pinby-Pin Homogenization of a Pressurized Water Assembly,” Nucl. Sci. Eng. 113, 227 (1993). K. S. Smith, “Assembly Homogenization Techniques for Light Water Reactor Analysis,” Prog. Nucl. Energy 14, 303 (1986). A. Jonsson, “Control Rods and Burnable Absorber Calculations,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations III, CRC Press, Boca Raton, FL (1986). R. J. J. Stamm’ler and M. J. Abbate, Methods of Steady State Reactor Physics in Nuclear Design, Academic Press, London (1983), Chap. VII. A. Kavenoky, “The SPH Homogenization Method,” Proc. Specialist’s Mtg. Homogenization Methods in Reactor Physics, Lugano, Switzerland, 1978, IAEA-TECDOC-231, International Atomic Energy Agency, Vienna (1980). V. C. Deniz, “The Theory of Neutron Leakage in Reactor Calculations,” in Y. Ronen, ed., CRC Handbook of Nuclear Reactor Calculations II, CRC Press, Boca Raton, FL (1986), p. 409. K. Koebke, “A New Approach to Homogenization and Group Condensation,” Proc. Specialist’s Mtg. Homogenization Methods in Reactor Physics, Lugano, Switzerland, 1978, IAEATECDOC-231, International Atomic Energy Agency, Vienna (1980). 9 R. T. Chiang and J. Dorning, “A Homogenization Theory for Lattices with Burnup and Non-uniform Loadings,” Proc. Top. Mtg. Advances in Reactor Physics and Core Thermal-Hydraulics, American Nuclear Society, La Grange Park, IL (1980), p. 240. 10 E. W. Larsen, “Neutron Transport and Diffusion in Inhomogeneous Media, I,” J. Math. Phys. 16, 1421 (1975); “Neutron Transport and Diffusion in Inhomogeneous Media, II,” Nucl. Sci. Eng. 60, 357 (1976); “Neutron Drift in Heterogeneous Media,” Nucl. Sci. Eng. 65, 290 (1978). 11 J. J. Duderstadt and L. J. Hamilton, Nuclear Reactor Analysis, Wiley, New York (1976), Chap. 10. 12 A. F. Henry, Nuclear-Reactor Analysis, MIT Press, Cambridge, MA (1975), Chap. 10. 13 J. R. Askew, F. J. Fayers, and F. B. Kemshell, “A General Description of the Lattice Code WIMS,” J. Br. Nucl. Energy Soc. 5, 564 (1966). 14 C. W. Maynard, “Blackness Theory for Slabs,” in A. Radkowsky, ed., Naval Reactors Physics Handbook, U.S. Atomic Energy Commission, Washington, DC (1964), pp. 409–448. 15 A. Amouyal, P. Benoist, and J. Horowitz, “New Method of Determining the Thermal Utilization Factor in a Unit Cell,” J. Nucl. Energy 6, 79 (1957). 16 E. E. Lewis and W. F. Miller, Computational Methods of Neutron Transport, American Nuclear Society, La Grange Park, IL (1993). Problems 14.1. Carry through the detailed derivation of the ABH method. 14.2. Consider a two-region slab geometry model of a unit cell consisting of a fuel plate of thickness a = 1 cm with a moderator region of thickness b = 2 cm on each side, with zero current cell boundary conditions and a uniform slowing-down source in the moderator. The fuel is UO2 , with thermal cross sections a = 0.169 cm−1 , 539 540 14 Homogenization 14.3. 14.4. 14.5. 14.6. 14.7. 14.8. 14.9. 14.10.∗ * s = 0.372 cm−1 , and 1 − μ0 = 0.9887. The moderator is H2 O, with thermal cross sections a = 0.022 cm−1 , s = 3.45 cm−1 , and 1 − μ0 = 0.676. Use the ABH method to calculate the thermal disadvantage factor, thermal utilization and homogenized scattering, and absorption cross sections for the cell. Carry through the detailed derivation of blackness theory. A reactor assembly consists of repeating arrays of three fuel–moderator unit cells of the type described in Problem 14.2, then a 0.1-cm-thick boron plate with thermal cross sections a = 25 cm−1 , s = 0.346 cm−1 , and (1 − μ0 ) = 0.9394, and then another three fuel–moderator unit cells. Use blackness theory to calculate an effective diffusion theory cross section to represent the boron slabs in the fuel–moderator plus boron plate array. A reactor fuel assembly consists of five of the fuel–moderator boron arrays described in Problem 14.4. Use one-group diffusion theory to calculate the assembly detailed heterogeneous flux distribution. Calculate the homogenized assembly absorption and scattering cross sections and diffusion coefficient and the assembly flux discontinuity factors using equivalence theory. Construct the Wigner–Seitz cell model for a fuel pin 1 cm in diameter within a 2-cm square of moderator. Set up and solve the collision probability equations for Problem 14.6, in one-group theory. Use the fuel and moderator parameters given in Problem 14.2. Calculate the homogenized cross sections for the pin-cell model of Problem 14.7, using conventional homogenization theory. Consider a lattice made up of a repeating array of 1-cm-thick fuel plates separated by 2 cm of H2 O, as described in Problem 14.2, but with different fuel enrichments in different plates. Taking a fuel plate and 1 cm of H2 O on each side as an assembly, use diffusion theory to solve for the assembly heterogeneous flux, with zero current assembly boundary conditions. Calculate the homogenized assembly cross sections and diffusion coefficient and the assembly flux discontinuity factor, using equivalence homogenized theory. Write a one-dimensional S4 code in slab geometry and repeat Problem 14.9 using an S4 assembly heterogeneous flux. Problem 14.10 is a longer problem suitable for a take-home project. 541 15 Nodal and Synthesis Methods Even after the local fuel pin, clad, coolant, and so on, heterogeneity is replaced by a homogenized representation, a reactor core remains a highly heterogeneous medium because of the intra-assembly and assembly-to-assembly variation in fuel composition, burnable poisons, control rods, water channels, structure and so on. The mesh spacing in a conventional few-group finite-difference model of such a core is constrained by two requirements: (1) it must be sufficiently fine to represent the remaining spatial heterogeneity adequately, and (2) it must be no larger than the shortest (thermal) group diffusion length in order to avoid numerical inaccuracy. A few-group finite-difference model that could adequately describe such a core might well have 105 to 106 unknowns (the fluxes in each group at each mesh point). The direct solution of such a problem, even in diffusion theory, remains a formidable computation that was unthinkable until very recently. For calculations such as fuel burnup or transient analysis, in which many full-core spatial solutions are needed, direct few-group finite-difference solutions remain impractical. A large number of approximation methods have been developed to enable a more computationally tractable solution for the effective multiplication constant and neutron flux distribution in reactor cores. Following historical precedent, these methods can generally be classified as nodal, coarse-mesh, or synthesis methods, although the distinction among the categories may be largely a matter of perspective and sequencing of calculational steps. Nodal methods characterize the global neutron flux distribution in terms of a small number of parameters in each of several large regions, or nodes, into which the reactor core is subdivided for this purpose. Such methods generally require detailed heterogeneous intranodal flux distributions to construct homogenized parameters for each of the many nodes into which a reactor core may be divided and to calculate coupling parameters that link the average flux solutions in adjacent nodes. The global average nodal fluxes must then be combined with the intranodal heterogeneous flux solutions if a heterogeneous flux distribution is required. Coarse-mesh methods extend the numerical accuracy of conventional finitedifference methods by using higher-order approximations for the flux variation among mesh points. Like nodal methods, coarse-mesh methods generally require detailed regional heterogeneous flux distributions in order to construct homog- 542 15 Nodal and Synthesis Methods enized parameters and to combine with the coarse-mesh solution to construct a detailed heterogeneous flux solution. Synthesis methods generally combine detailed heterogeneous two-dimensional planar flux distributions by means of a one-dimensional axial calculation to obtain a global heterogeneous flux solution. Such methods do not require a previous homogenization within large regions of the core as do nodal and coarse mesh methods, but in effect perform a homogenization in constructing the parameters to be used in the axial synthesis calculation, thus ensuring a certain consistency between the homogenization and the approximate model calculation. 15.1 General Nodal Formalism Writing the multigroup neutron balance equations in the form g ∇ · Jg (r) + t (r)φg (r) G  =   g →g (r)φg  (r) + g  =1 G χg  g νf (r)φg  (r), k  g = 1, . . . , G (15.1) g =1 and integrating over the volume of node n (Fig. 15.1) yields an integral balance on node n:  n g g Lnn + tn φ̄gn Vn = G  g  =1 g  →g n φ̄g  Vn n g = 1, . . . , G, n = 1, . . . , N + G χg  g νf n φ̄gn Vn , k  g =1 (15.2) where the nodal average total, scattering, and fission cross sections are defined by expressions of the form g g tn ≡ Vn dr t (r)φg (r) Vn dr φg (r) (15.3) the average nodal flux is φ̄gn ≡ Vn φg (r) dr Vn (15.4) and the leakage between node n and adjacent node n is defined by a surface integral over the common interface:  g drs n · Jg (rs ) (15.5) Lnn ≡ rs ∈Snn To be more specific, in discussion of the leakage term, we consider a parallelepiped node of dimensions x, y, and z, as shown in Fig. 15.1. The surface integrals 15.1 General Nodal Formalism Fig. 15.1 Nodal model nomenclature. of the net x-direction current at the node boundaries at x = +x/2 and at −x/2 are defined as n Jgx± ≡ y/2 z/2 −y/2 dy −z/2 dz nx · Jg (±x/2, y, z) yz (15.6) with similar definitions for the surface integrals of net y- and z-direction currents at ±y/2 and ±z/2, respectively. Surface integrals of the outward and inward x-directed partial currents at ±x/2 are defined in terms of the partial currents di− rected to the right (J+ g ) and to the left (Jg ), respectively:   n J¯gx±  ≡ out  n J¯gx± in ≡− y/2 z/2 −y/2 dy −z/2 dz nx · J± g (±(x/2), y, z) yz y/2 z/2 −y/2 dy −z/2 dz nx · J∓ g (±(x/2), y, z) (15.7) yz with similar definitions for surface integrals of partial y- and z-direction currents at ±y/2 and ±z/2, respectively. The surface integrals of the net current are related to the surface integrals of the partial currents as the net current is related to the partial currents:  n   n  n = ± J¯gx± − J¯gx± J¯gx± out in (15.8) 543 544 15 Nodal and Synthesis Methods Using these definitions of surface integrals of the net current over the six faces bounding the node, the balance equations (15.2) can be written in the more explicit form    1  ¯n 1  ¯n 1  ¯n g n n n − J¯gx− − J¯gy− − J¯gz− J J J + + + tn φ̄gn x gx+ y gy+ z gz+ = G  g  =1 g  →g n φ̄g  n + G χg  g νf n φ̄gn , k  g = 1, . . . , G (15.9) g =1 The various nodal formulations are distinguished primarily by the methods used to evaluate the surface currents in Eq. (15.9). In diffusion theory approximation, the x-directed partial currents and the flux are related by dφg (x) 1 1 ± Jgx (x) = φg (x) ∓ D g 4 2 dx (15.10) with similar relations for the y- and z-directed partial currents. Thus the surface integrals of the flux at ±x/2 n ≡ φ̄gx± y/2 z/2 −y/2 dy −z/2 dz φ(±(x/2), y, z) yz are related to the corresponding surface integrals of the partial currents:  n   n  n φ̄gx± = 2 Jgx± + Jgx± out in (15.11) (15.12) with similar relations for the y- and z-directed partial currents at ±y/2 and ±z/2, respectively. All surface integrals with the node index n are evaluated in the limit as the surface is approached from within the nth node. As mentioned, the various nodal formulations are distinguished primarily by the methods used to evaluate the surface currents in Eq. (15.9). Two rather distinct classes of nodal methods have evolved. The first class, often referred to as conventional or simulation models, makes use of detailed calculations or reactor operating experience to evaluate the surface current integrals in terms of differences in nodeaveraged fluxes for adjacent nodes, with empirically adjusted coupling coefficients. The second class, sometimes referred to as consistently formulated models, makes use of the concept of transverse integration and of higher-order (than ordinary finitedifference) approximations to evaluate the surface integrals of the current and the internodal coupling terms in order to derive nodal equations that can be expected to converge to the exact solution in the limit of small mesh spacing. 15.2 Conventional Nodal Methods The first class of nodal models to be considered is based on relatively simple mathematical models with parameters that can be adjusted to match the results 15.2 Conventional Nodal Methods of more detailed calculation or measurement. Such methods are widely used in three-dimensional simulators, which play a key role in guiding and interpreting the operation of research and power reactors. The basis of such methods is the representation of the neutron flux or neutron fission rate within each of the many homogenized fuel assemblies by a single nodal average flux or fission rate that is coupled to the average flux or fission rate in adjacent nodes by the internodal diffusion of fast neutrons, which is represented by coupling coefficients. The reflector is usually represented by an albedo. Such methods are frequently based on 1 12 -group theory. The coupling coefficients and the reflector albedo are normally adjusted to provide agreement with more detailed calculations or measurements. The earlier versions of this class of nodal methods imposed a continuity of net current condition at interfaces:   n  xn g d φ̄gx g  n,n+1 n+1 n n,n n ¯ φ̄g − αgx φ̄g (15.13) = −Dn Jgx (xn /2) = −Dn,eff αgx 2 dx where φ̄gn is the node-averaged flux, and chose the effective diffusion coefficients and coupling parameters α to match interface net currents or nodal average fluxes from detailed planar finite-difference calculations. The sometimes unphysical nature of the solution or the strong sensitivity to the properties of both adjacent fuel assemblies of the coupling coefficients obtained by such net current-matching procedures led to the development of coupling coefficients based on matching partial currents at node interfaces:     (y/2)   (z/2) n+ xn + x ¯ ≡ , y, z = αgn,n+1 φ̄gn xn dy dz nx · Jg Jgx 2 2 (−y/2) (−z/2)      (z/2)  (y/2) x (15.14) n− xn dy dz nx · J− ≡− , y, z J¯gx g 2 2 (−y/2) (−z/2) = αgn+1,n φ̄gn+1 xn+1 The gross coupling method uses detailed finite-difference diffusion theory fluxes from a heterogeneous planar (x, y) model to calculate interface partial currents:       ∂φg ((xn /2), y) xn xn 1 1 ± xn nx · Jg , y = φg , y ∓ Dg ,y 2 4 2 2 2 ∂x (15.15) which are used to evaluate the coupling coefficients, α. For φg (xn /2, y), φ̄gn and φ̄gn+1 obtained from detailed planar calculations, Eqs. (15.14) and (15.6)—with the integral over z suppressed—are used to evaluate αgn and αgn+1 . The nodal equations (15.9) in two-dimensional geometry may be written as −αgnx +1,n xnx +1 nx +1 xnx −1 nx −1 n +1,n yny +1 ny +1 φ̄g φ̄g φ̄ − αg y − αgnx −1,n xn yn xn g n −1,n yny −1 − αg y yn n −1 φ̄g y 545 546 15 Nodal and Synthesis Methods  n,n +1 n,n −1 g + αgn,nx +1 + αg y + αgn,nx −1 + αg y + tn φ̄gn − G  g  →g n φ̄g  n g  =1 − G χg  g νf φ̄gn = 0 k  (15.16) g =1 where the node n is designated by sub- and superscripts nx and ny so that the adjacent node in the x- and y-directions may be indicated by nx ± 1 and ny ± 1, respectively [e.g., n, nx + 1 refers to the coupling between node n (nx , ny ) and the adjacent node (nx + 1, ny ) at x = +xn /2] (see Fig. 15.1). Most of the conventional nodal models do not make use of detailed planar calculations to evaluate the internodal coupling coefficients. Instead, the coupling coefficients are reinterpreted in a manner that enables intranodal collision probability methods to be used in their evaluation. The one-group version of Eq. (15.16) may be rewritten as −W nx +1,n S nx +1 − W ny +1,n S ny +1 − W nx −1,n S nx −1 − W ny −1,n S ny −1   k + W n,nx +1 + W n,ny +1 + W n,nx −1 + W n,ny −1 + n − 1 S n = 0 (15.17) k∞ as a balance among the fission neutron production rates in the various nodes, where S n ≡ νf n φ n xn yn , n k∞ ≡ νf n an (15.18) and the coupling terms are of the form W n,nx +1 = W nx +1,n = J¯xn+ ((xn /2)) νf n φ̄ n J¯xn− ((xn /2))xn (15.19) νf n φ̄ n+1 xn+1 The new coupling coefficients W n,nx +1 may be interpreted as the probability that a fission neutron born in node (nx , ny ) escapes into node (nx+1 , ny ), and so on, quantities which readily lend themselves to calculation using collision probabilities or other methods. For example, the well-known FLARE code uses ) Mn2 Mn2 +g (15.20) W n,nx +1 = (1 − g) 2xn (xn )2 where Mn2 is the migration area in node (nx , ny ) and g is an adjustable parameter. The two terms correspond to the one-group transport and diffusion kernels for leakage from a slab of thickness xn . A reformulation of the FLARE equations in 1 12 -group theory leads to W n,nx +1 = Mn2 1 2 · n · 2 xn k∞ 1 + Mn+1 /Mn (15.21) 15.3 Transverse Integrated Nodal Diffusion Theory Methods Neutron conservation for an internal node requires that Sn = 6 n  k∞ W m,n S m k (15.22) m=1 where the sum is over the six adjacent nodes. W m,n represents the probability that a neutron created from fission in node m will be absorbed in node n, since it has been assumed that a neutron escaping into an adjacent node is absorbed therein. (This assumption can be removed.) For nodes on the surface of the core, an albedo βnr is used for each surface r which faces a reflector, so that the balance equation is   6 n  k∞ n m,n m n,r n (15.23) S = W S + (1 − βnr )W S , n = 1, . . . , N k m=r Equations (15.22) and (15.23) are solved iteratively, with the eigenvalue guess updated on each iteration by using the most recently calculated S n in the neutron balance to evaluate  n S [1 − (1 − βnr )W n,r ]  n n k= n (15.24) n S /k∞ Nodal methods of the type described in this section generally require parameter adjustment to obtain agreement with more detailed calculations or measurements of power distribution, effective multiplication constant, and so on. Computations based on these nodal methods run very fast and have found widespread use in three-dimensional reactor simulators. 15.3 Transverse Integrated Nodal Diffusion Theory Methods A second class of nodal methods are those that have been formulated on the basis of integrating the three-dimensional diffusion equation over two transverse directions to obtain a one-dimensional diffusion equation, with transverse leakage terms, which can be solved within a node by approximating the dependence on the remaining spatial variable, usually with a polynomial. These methods are consistently formulated in that they reduce in the limit of small node sizes to the conventional finite-difference method for the homogenized reactor model. Transverse Integrated Equations Integration of the three-dimensional multigroup diffusion equations over the two transverse directions to obtain a one-dimensional equation in node n yields d ¯n 1 g 1 g g n (x) J (x) + Lny (x) + Lnz (x) + tn φ̄gx dx gx y z 547 548 15 Nodal and Synthesis Methods = G  g  →g n φ̄g  x (x) + n g  =1 G χg  g νf n φ̄gn x (x), k  g = 1, . . . , G (15.25) g =1 The x-dependent flux and current averaged over the transverse directions are n φ̄gx (x) ≡ y/2 z/2 n −y/2 dy −z/2 dz φg (x, y, z) (15.26) yz y/2 z/2 n −y/2 dy −z/2 dz Jg (x, y, z) n (x) ≡ J¯gx (15.27) yz and leakage terms transverse to the x-direction are      y y , z − Jg x, − ,z dz ny · Jg x, 2 2 −z/2   z/2  −1 g ∂φg (x, (y/2), z) g ∂φg (x, −(y/2), z) − Dn dz Dn = z −z/2 ∂y ∂y (15.28)       y/2 z z 1 g − Jg x, y, − dy nz · Jg x, y, Lnz (x) = y −y/2 2 2   y/2  −1 g ∂φg (x, y, (z/2)) g ∂φg (x, y, −(z/2)) dy Dn − Dn = y −y/2 ∂z ∂z (15.29) g Lny (x) = 1 z  z/2 Making the diffusion theory approximation g d n n (x) = −Dn φ̄gx (x) J¯gx dx (15.30) the multigroup diffusion theory x-direction transverse integrated equation for node n is − d g d n 1 g 1 g g n Dn φ̄gx (x) + Lny (x) + Lnz (x) + tn φ̄gx dx dx y z = G  g  →g n φ̄g  x (x) + n g  =1 G χg  g νf n φ̄gn x (x), k  g = 1, . . . , G (15.31) g =1 The node-averaged values of the group flux and transverse leakage terms are φ̄gn ≡ = 1 x  x/2 −x/2 1 xyz n dx φ̄gx (x)   x/2 −x/2 dx  y/2 −y/2 dy z/2 −z/2 dz φg (x, y, z) (15.32) 15.3 Transverse Integrated Nodal Diffusion Theory Methods g L̄ny ≡ g L̄nz ≡ 1 x 1 x  x/2 −x/2  x/2 −x/2 g n n dx Lny (x) = Jgy+ − Jgy− (15.33) g n n dx Lnz (x) = Jgz+ − Jgz− Integrating Eq. (15.25) over x and using Eqs. (15.32) and (15.33) yields the nodal balance equation (15.19). One-dimensional transverse integrated equations in the y- and z-directions are derived in a similar manner. Polynomial Expansion Methods The coarse mesh methods can obtain a higher-order accuracy than conventional finite-difference methods by expanding the x-dependence of the flux: n φ̄gx (x)  φ̄gn f0 (x) + I  n agxi fi (x), i=1 − x x ≤x≤ 2 2 (15.34) where the polynomials f0 (x) = 1, f1 (x) = x ≡ξ x    1 1 1 f2 (x) = 3ξ 2 − , f3 (x) = ξ ξ − ξ+ 4 2 2     1 1 1 f4 (x) = ξ 2 − ξ− ξ+ , ... 20 2 2 (15.35) are normalized so that the volume average of the polynomial representation of the flux is the volume average of the flux defined by Eq. (15.32):   x/2 1 1, n = 0 dx fn (x) = (15.36) 0, n > 0 x −x/2 and the surface average of the flux is equal to the surface-averaged flux defined by Eq. (15.11) at x = ±x/2:   x n n ± (15.37) = φgx± φ̄gx 2 These requirements are satisfied by polynomial expansion coefficients, n = φn n agx1 gx+ − φgx− n = φn n n agx2 gx+ − φgx− − 2φ̄g and the requirement that   x fi ± = 0, n > 2 2 on the polynomials. (15.38) (15.39) 549 550 15 Nodal and Synthesis Methods In terms of these polynomials, the outgoing x-direction surface-averaged currents at x = ±x/2 are    n   n   n  x g d n n φ̄ + Jgx+ + Jgx+ = −D Jgx+ out = Jgx+ n gx in in dx 2  g  n  Dn n 1 n 1 n n + Jgx+ + agx3 + agx4 agx1 + 3agx2 in x 2 5    n   n  x g d n n − + Jgx− = −Jgx− + Jgx− = Dn φ̄gx in in dx 2 =−  n Jgx−  out =  g  n  Dn n 1 n 1 n n + Jgx− + agx3 − agx4 agx1 − 3agx2 in x 2 5 (15.40) (15.41) with similar expressions for the y- and z-direction surface-averaged currents at ±y/2 and ±z/2, respectively. If the polynomial expansion of the x-direction flux in Eq. (15.34) is terminated at I = 2, and similarly for the y- and z-direction expansions, the transverse-integrated nodal equations are well posed in terms of node-averaged fluxes and incoming and outgoing partial currents over node boundaries (i.e., the number of equations and the number of unknowns agree). Equations (15.38) and (15.12) can be used to express Eqs. (15.40) and (15.41) in terms of node-averaged flux and partial currents at x = ±x/2:   g g  n  4Dng 8Dn 6Dn n 1+ + Jgx− − φ̄ out x x x g   g g  n   n  8Dn 4Dn 1 − − = Jgx+ + J gx− in in x x  n Jgx+ out   g g  n  4Dng 6Dn n + J − φ̄ gx+ out out x x g   g g  n   n  8Dn 4Dn 1 − − = Jgx− + J gx+ in in x x n Jgx−  (15.42) 1+ 8Dn x (15.43) with similar expressions for the y- and z-direction surface-averaged currents at ±y/2 and ±z/2, respectively. Equation (15.8) can be used to replace the currents with partial currents in the nodal balance equation (15.9) to obtain  n   n  ,  n  1 + n  + Jgx− Jgx+ out + Jgx− − Jgx+ out in in x  n   n  ,  n  1 + n  + Jgy+ out + Jgy− + Jgy− − Jgy+ out in in y +  n   n  ,  n  1 + n  Jgz+ out + Jgz− + Jgz− − Jgz+ out in in z 15.3 Transverse Integrated Nodal Diffusion Theory Methods g + tn φ̄gn = G  g  =1 g  →g n φ̄g  n + G χg  g νf n φ̄gn , k  g = 1, . . . , G (15.44) g =1 Note that this equation could be derived directly by integrating Eq. (15.1) over the node. The incoming x-direction partial currents to node n may be related to the outgoing partial currents from the adjacent node n + 1 at x/2. Using the flux discontinuity condition discussed in Chapter 14, the surface-averaged fluxes are related by n+1 n+1 n n fgx+ φgx+ = fgx− φgx−  n   n+1    n+1  n+1  n n fgx+ Jgx+ out + Jgx+ in = fgx− Jgx− out + Jgx− in (15.45) where Eq. (15.12) has been used to write the second form of the equation. For unity flux discontinuity factors, Eq. (15.45) becomes the continuity of flux condition. The surface-averaged current continuity condition n+1 n Jgx+ = Jgx−  n   n+1   n+1   n  = Jgx− − Jgx− Jgx+ out − Jgx+ in in out (15.46) may be combined with the flux discontinuity condition to obtain  n  Jgx+ in = n+1 )out 2(Jgx− n /f n+1 1 + fgx+ gx− + n /f n+1 1 − fgx+ gx−  n  J n+1 gx+ out n 1 + fgx+ /fgx− (15.47) Imposition of similar conditions at the interface with adjacent node n − 1 at −x/2 yields  n  Jgx− in = n−1 )out 2(Jgx+ n /f n−1 1 + fgx− gx+ + n /f n−1 1 − fgx− gx+  n /f n−1 1 + fgx− gx+ n Jgx−  (15.48) out Similar expressions are obtained relating the incoming y- and z-direction surfaceaveraged partial currents at ±y/2 and ±z/2, respectively, to the outgoing partial currents from the adjacent nodes in the y- and z-directions. The equations above can be derived directly from an expansion of the form φgn (x, y, z) = φ̄g + 2  i=1 n αgxi fi (x) + 2  j =1 αgyj fj (y) + 2  n αgzk fk (z) (15.49) k=1 without recourse to the transverse integration stratagem. In fact, Eqs. (15.44) follow directly from Eqs. (15.8) and (15.9), and the interface conditions of Eqs. (15.45) and (15.46) arise from other considerations. However, this transverse integration stratagem is essential for extending the formalism to higher order. For polynomial expansions with I > 2 in Eq. (15.34), the transverse integrated equations are no longer well posed in the sense of having the same number of 551 552 15 Nodal and Synthesis Methods equations and unknowns. However, weighted residuals methods can be used to develop higher-order approximations, but this requires the further approximation of higher-order leakage moments. Multiplying Eq. (15.25) by the spatial function wi (x) and integrating yields 5 6 d ¯n n n wi (x), φgxi Jgx (x) + tn dx = G  g  =1 g  →g n φg  xi n + G χg  1 g 1 g νfng  φgn xi − Lnyxi − L k  y z nzxi (15.50) g =1 where the ith spatial moment of the flux is n φ̄gxi 0 1 1 n ≡ wi (x), φ̄gx (x) ≡ x  (x/2) −(x/2) n dx wi (x)φ̄gx (x) (15.51) and the ith spatial moment of the transverse leakage is 0 1 1 g g Lnyxi ≡ wi (x), Lny (x) ≡ x  (x/2) −(x/2) g dx wi (x)Lny (x) (15.52) with a similar term for the z-direction transverse leakage. The nodal balance equation results from choosing w0 = 1 in Eq. (15.50). Numerical comparison with detailed finite-difference solutions indicates that the choices w1 (x) = f1 (x) and w2 (x) = f2 (x) yield good results. Using these two functions and integrating the first term in Eq. (15.50) by parts yields the two equations that must be solved for the higher-order flux moments: g 1 Dn n g g n Tnx + α + tn φgx1 2x (x)2 gx1 = G  g  =1 g  →g n φg  x1 n + G χg  1 g 1 g L L νfng  φgn x1 − − (15.53) k  y nyx1 z nzx1 g =1 g 3Dn 1 g g n Lnx + α n + tn φgx2 2x (x)2 gx2 = G  g  =1 g  →g n φg  x2 n + G χg  1 g 1 g L L νfng  φgn x2 − − (15.54) k  y nyx2 z nzx2 g =1 where g n n + Jgx− , Tnx ≡ Jgx+ g n n Lnx ≡ Jgx+ − Jgx− (15.55) Using w1 (x) = f1 (x) and w2 (x) = f2 (x) and Eq. (15.49) in Eq. (15.51) then yields the higher-order expansion coefficients n n n αgx3 = −120φgx1 + 10αgx1 , n n n αgx4 = −700φgx2 + 35αgx2 (15.56) 15.3 Transverse Integrated Nodal Diffusion Theory Methods Solution of Eqs. (15.54) requires further approximation for the x-dependence of the x-direction transverse leakage (and similarly for the y- and z-direction transverse leakage terms). A number of approximations have been used, but the most successful has been the quadratic approximation g g n n f1 (x) + Cgy2 f2 (x) Lny (x) = L̄ny + Cgy1 (15.57) which is assumed, for the purpose of evaluating moments of the transverse leakage, to extend over node n and the two nodes adjacent to node n in the x-direction. Use of Eq. (15.57) in Eq. (15.52) then makes it possible to evaluate the transverse leakage moments in terms of the surface-averaged leakages (thus surface-averaged partial currents) in the adjacent nodes. Combining results in the three coordinate directions leads to an interface current balance in each group of the form = P ng Q ng − L ng + R ngJ n,in J n,out g g (15.58) The column vectors J n,out and J n,in contain the six outgoing and incoming, reg g spectively, surface-averaged partial currents for the nth node. The column vector Q ng contains the node-averaged scatter-in and fission sources to group g, and L ng contains the higher-order spatial moments of the transverse leakage computed using the quadratic fit or some other approximation. The matrices P ng and R ng contain nodal coupling coefficients. A variety of iterative schemes have been devised for solving Eq. (15.58), within an outer power iteration solution procedure. Generally, the three-dimensional geometry is subdivided into a number of axial planes, and the nodes within each plane are solved (swept) a few times using the most recent values for group fluxes in nodes in the adjacent planes. The number of planar sweeps required per group generally increases with the planar average diffusion length within the group. The nodal procedure outlined above uses constant homogenized cross sections over the node. In applications where the actual cross sections vary significantly over the node, the use of constant cross sections introduces an error in calculating effects such as space-dependent internodal burnup. An extension to include loworder polynomial dependence of the cross sections over the node has been shown to lead to improved accuracy in such cases. Analytical Methods There are variants of the transverse integrated method in which an analytical solution is used in some part of the derivation of the transverse integrated nodal equations. In a variant known as the analytical nodal method the one-dimensional transverse integrated equation is integrated analytically to relate the nodal leakage in that dimension to the nodal average fluxes in the node and in the adjacent nodes in that dimension. In another variant known as the nodal Green’s function method, the one-dimensional transverse integrated equation is formally solved by the method of Green’s functions, resulting in expressions that can be used together 553 554 15 Nodal and Synthesis Methods with the polynomial expansion to evaluate coefficients. These are discussed more fully in Ref. 2. Heterogeneous Flux Reconstruction The results of the nodal calculation are global node-averaged fluxes, φ̄gn flux distributions consisting of the polynomial flux distributions φgn (x, y, z) within each node or assembly n [e.g., as constructed from Eqs. (15.34) for each direction] and nodal interface currents. These global fluxes and flux distributions are normalized to the reactor power level. To obtain a more detailed heterogeneous intra-assembly flux distribution, it is necessary to superimpose on these nodal average or smoothly varying polynomial flux distributions a detailed intranodal flux shape, Ang (x, y), usually taken from a planar assembly transport calculation: n n n g (x, y, z) = φg (x, y, z)Ag (x, y) (15.59) The simplest such procedures use an assembly calculation with symmetry boundary conditions to determine Ang (x, y) and Eq. (15.59). Improved accuracy has been obtained by using the first of Eqs. (15.59) to construct a gross intranodal flux distribution that approximates the gross intranodal flux shape from the global calculation. Use of the same intranodal flux shape for the nodal homogenization and flux reconstruction is necessary for consistency, but this is difficult to achieve in practice without an iteration among the homogenization, nodal solution, and flux reconstruction steps. 15.4 Transverse Integrated Nodal Integral Transport Theory Models Transverse Integrated Integral Transport Equations The concepts and procedures introduced in Section 15.3 can be extended to develop nodal methods based on integral transport theory. To limit the notational complexity, we discuss the development of integral transport nodal methods in two-dimensional rectangular geometry, although we note that three-dimensional models are in use for nuclear reactor analysis. Assuming that a detailed heterogeneous assembly transport calculation has been performed to produce homogenized multi-group constants that are uniform over the domain of node n (−x/2 ≤ x ≤ x/2, −y/2 ≤ y ≤ y/2), the transport equation for the multigroup neutron flux within node n in two-dimensional Cartesian geometry may be written . ∂ ∂ g μ ψgn (x, y, μ, φ) + 1 − μ2 cos φ ψgn (x, y, μ, φ) + tn ψgn (x, y, μ, φ) ∂x ∂y = 1 n S (x, y), 4π g g = 1, . . . , G (15.60) 15.4 Transverse Integrated Nodal Integral Transport Theory Models Fig. 15.2 Coordinate system for two-dimensional nodal transport model. where, for notational convenience, the group in-scatter and fission terms have been written as a source term: Sgn (x, y) =  1  2π G χg  g νf dμ dφ ψgn (x, y, μ, φ) k  −1 0 g =1 + G  g  =1 g  →g νn   1 −1 2π dμ 0 dφ ψgn (x, y, μ, φ) (15.61) isotropic scattering has been assumed, and the coordinate system is defined such that . x ≡  · nx = μ, y ≡  · ny = 1 − μ2 cos φ (15.62) The coordinate system and spatial domain of node n are depicted in Figs. 15.2 and 15.3. Fig. 15.3 Spatial domain for two-dimensional nodal model. 555 556 15 Nodal and Synthesis Methods Integrating Eq. (15.60) over −y/2 ≤ y ≤ y/2 yields the one-dimensional x-direction transverse integrated transport equation for node n: μ ∂ n g n g (x, μ, φ) + Lny (x, μ, φ) ψ (x, μ, φ) + tn ψgx ∂x gx  y/2 1 n 1 dy Sgn (x, y) ≡ S (x) = 4π −y/2 4π g (15.63) where the x-direction angular flux is n ψgx (x, μ, φ) ≡ 1 y  y/2 −y/2 dy ψgn (x, y, μ, φ) (15.64) and the transverse leakage term defining the average net neutron loss rate across the node boundaries at y = −y/2 and y = y/2 is g Lny (x, μ, φ)      . y y 1 , μ, φ − ψgn x, − , μ, φ (15.65) 1 − μ2 cos φ ψgn x, + ≡ y 2 2 Equation (15.63) can be integrated if the scattering, fission, and leakage are treated as a known source:    x g 1 1 n   g n ψgx Sg (x ) − Lny (x  , μ, φ) (x, μ > 0, φ) = dx  e−tn (x−x )/μ μ 4π −(x/2) g n,in (μ, φ)e−tn (x+(x/2))/μ , + ψgx−  (x/2) g  n dx  e−tn (x−x )/μ ψgx (x, μ < 0, φ) = − μ>0 (15.66a) μ<0 (15.66b) x   1 1 n  g  × S (x ) − Lny (x , μ, φ) μ 4π g g n,in + ψgx+ (μ, φ)e−tn (x−(x/2))/μ , where the inward-directed average angular fluxes at x = x/2 and x = −x/2 are   x n,in n ± ψgx± , μ ≶ 0, φ (15.67) (μ, φ) ≡ ψgx 2 and the outward-directed average angular fluxes at x = x/2 and x = −x/2 are   x n,out n ± ψgx± (μ, φ) ≡ ψgx , μ ≷ 0, φ (15.68) 2 The average scalar flux in the x-direction problem is  n φgx (x) =  2π dφ 0 0 1  n dμ ψgx (x, μ > 0, φ) + 0 −1  n dμ ψgx (x, μ < 0, φ) 15.4 Transverse Integrated Nodal Integral Transport Theory Models 1 2 =  x/2 −x/2    g g,iso dx  E1 tn |x − x  | Sgn (x  ) − Lny (x  )  2π g dμ −tn g,anis |x−x  |μ Lny (x  , |μ|, φ) dφ e −x/2 0 μ 0  2π  1 g n,in dμ e−tn (x+(x/2))/μ ψgx− (μ, φ) dφ + x/2 − dx   1 0  + 0 0 −1 dμ e g −tn (x−(x/2))/μ  2π 0 n,in ψgx+ (μ, φ) dφ (15.69) where the exponential integral function is   1 En (ξ ) ≡ n−2 dμ μ 0 ξ exp − μ  (15.70) and the transverse leakage has been split into an isotropic and an anisotropic component: Lny (x  , μ, φ) = g 1 g,iso  g,anis Lny (x ) + Lny (x  , μ, φ) 4π (15.71) Polynomial Expansion of Scalar Flux Following the same general procedure used to develop the diffusion theory nodal model, the scalar flux for the x-direction problem is expanded: n φgx (x) = I  n ai φgxi fi (x), I ≤2 (15.72) i=1 The expansion coefficients are normalized such that 1 1 ≡ ai x  x/2 −x/2 dx [fi (x)]2 (15.73) and the polynomials f0 = 1, x f1 (x) = , x  x f2 (x) = 3 x 2 − 1 4 (15.74) are used. The moments of the scalar flux are  n φgxi ≡ x/2 −x/2 n dx φgx (x)fi (x) n is the node-averaged scalar flux. so that φgx0 (15.75) 557 558 15 Nodal and Synthesis Methods Isotropic Component of Transverse Leakage The surface average of the isotropic component of the transverse leakage is g,iso L̄ny  x/2 1 g,iso dx Lny (x) x −x/2  n   n   n   n  = Jgy+ − Jgy+ − Jgy− − Jgy− out in in out ≡ (15.76) where the surface average of the outward and inward partial currents at +y/2 and −y/2 are  2π  1  n  n,out dφ dμ μψgy+ (μ, φ) Jgy+ out ≡ 0 0 (15.77)  2π  0  n  n,out dφ dμ μψgy− (μ, φ) Jgy− out ≡ −1 0 and   n Jgy+   in  n Jgy− in ≡  2π dφ 0  ≡ 2π −1  1 dφ 0 0 0 n,in dμ μψgy+ (μ, φ) (15.78) n,in dμ μψgy− (μ, φ) n,in respectively, with the directional neutron fluxes at +y/2 and −y/2, ψgy ± and n,out ψgy ± defined by equations similar to Eqs. (15.67) and (15.68). Double-Pn Expansion of Surface Fluxes The angular dependence of the neutron flux on the surfaces of the node is approximated by a double-P1 approximation, which allows independent linearly anisotropic distributions for the incident and exiting fluxes on a surface. In terms of the half-space polynomials, which are related to the Legendre polynomials by pn+ (ξ ) = Pn (2ξ − 1) for 1 ≥ ξ ≥ 0 and pn− (ξ ) = Pn (2ξ + 1) for 0 ≥ ξ ≥ −1, the surface-averaged inward neutron fluxes at ±x/2 are expanded:   x n,in n ± ψgx± (μ, φ) ≡ ψgx , μ ≶ 0, φ 2   3 ± ∓ 1 1 ± ∓ 3 ± ∓ p1 (y ) a0 p0 + a1x p1 (x ) + a1y ≈ 2π 2 2 2   . 3 ± 1 1 ± 3 ± 1 − μ2 cos φ C0 + C1x μ + C1y ≈ 2π 2 2 2    1  ¯n,in 1 n,in n,in n,in 4φ̄gx± 12Jgx± ± 6ψgx± + μ = ± 6Jgx± 2π 2π . 1  ¯n,in  + 3Jgx± 1 − μ2 cos φ (15.79) 2π 15.4 Transverse Integrated Nodal Integral Transport Theory Models The angular moments of the surface-averaged inward fluxes that appear in Eq. (15.79) are n,in ψ̄gx− ≡ n,in ≡ ψ̄gx+ n,in ≡ J¯gx− n,in J¯gx+  0  0  2π dφ 0  2π 0 −1  1 dφ 0 0  2π ≡ n,in ≡ J¯gy+ 1 dφ  n,in ≡ J¯gy−  2π dφ 0  0 −1 1  2π dφ 0  0  2π dφ 0 0 −1 n,in dμ ψgx− (μ, φ) n,in dμ ψgx− (μ, φ) n,in dμ μψgx− (μ, φ) (15.80) n,in dμ μψgx+ (μ, φ) . n,in dμ 1 − μ2 cos φψgy− (μ, φ) . n,in dμ 1 − μ2 cos φψgy+ (μ, φ) Using Eq. (15.79) to evaluate the integrals involving the incident fluxes in Eq. (15.69) yields  n (x) = φgx 1 n   g g,iso dx  E1 tn |x − x  | S (x ) − Lny (x  ) 2 g −x/2  x/2  1  2π g  dμ −tn g,anis |x −x  |/μ − dx  dφ Lny (x  , |μ|, φ) e −x/2 0 μ 0        x x g g n,in 4E2 tn x + + ψ̄gx− − 6E3 tn x + 2 2        x x g g n,in ¯ − 6E2 tn x + + Jgx− 12E3 tn x + 2 2        g x g x n,in 4E2 tn −x − 6E3 tn −x + ψ̄gx+ 2 2        g x g x n,in 6E2 tn −x − 12E3 tn −x (15.81) + J¯gx+ 2 2 x/2 Angular Moments of Outgoing Surface Fluxes The angular moments of the surface averaged outgoing flux and current at x/2 can be constructed from Eq. (15.66a), using Eq. (15.79) to expand the angular dependence of the incoming flux at −x/2:    2π  1 x n,out n ψ̄gx+ , μ > 0, φ ≡ dφ dμ ψgx 2 0 0     x/2 1 n  g x g,iso dx  E1 tn − x S (x ) − Lny (x  ) = 2 2 g −x/2 559 560 15 Nodal and Synthesis Methods  x/2  1 g dμ −tn ((x/2)−x  )/μ e −x/2 0 μ    g  g n,in + ψ̄gx− 4E2 tn x − 6E3 tn x    g  g n,in + J¯gx− 12E3 tn x − 6E2 tn x − dx   2π g,anis Lny (x  , μ, φ) dφ 0 (15.82)   x , μ > 0, φ 2 0 0     x/2 1 n  g x g,iso dx  E2 tn − x Sg (x ) − Lny (x  ) = 2 2 −x/2  x/2  1  2π g  g,anis dx  dμ e−tn ((x/2)−x )/μ Lny (x  , μ, φ) dφ − n,out ≡ J¯gx+   2π 1 dφ −x/2 n,in + ψ̄gx− n,in + J¯gx− n dμ μψgx 0 0    g  g 4E3 tn x − 6E4 tn x    g  g 12E4 tn x − 6E3 tn x (15.83) The angular moments of the surface-averaged outgoing flux and current at −x/2 can be constructed from Eq. (15.66b), using Eq. (15.79) to expand the angular dependence of the incoming flux at +x/2:   x n − dμ ψgx , μ < 0, φ 2 0 −1     x/2 1 n  g x g,iso + x S (x ) − Lny (x  ) dx  E1 tn = 2 2 g −x/2  0  2π  x/2 g dμ tn  g,anis e (x/2+x )/μ dx  dφ Lny (x  , μ, φ) − −x/2 −1 μ 0    g  g n,in 4E2 tn x + 6E3 tn x + ψ̄gx+    g  g n,in + J¯gx+ (15.84) 12E3 tn x + 6E2 tn x n,out ≡ ψ̄gx−   2π 0 dφ   x n − dμ μψgx , μ < 0, φ 2 0 −1     x/2 1 n  g x g,iso   +x Sg (x ) − Lny (x  ) dx E2 tn = 2 2 −x/2  0  2π  x/2 g  g,anis dx  dμ etn ((x/2)+x )/μ dφ Lny (x  , μ, φ) + n,out J¯gx− ≡   2π 0 dφ −x/2 n,in + ψ̄gx+ n,in + J¯gx+ −1   g   g 4E3 tn  + 6E4 tn x    g  g 12E4 tn x + 6E3 tn x 0 (15.85) 15.5 Transverse Integrated Nodal Discrete Ordinates Method Nodal Transport Equations These equations can be written, in terms of matrices and column vectors, in a form analogous to the diffusion theory relation of Eq. (15.58): n P g Q ng − L ng + R̃ ng ψ n,in = P̃ ψ n,out g g Q ng (15.86) L ng and are defined as for diffusion theory and represent The column vectors the fission plus in-scatter source and the transverse leakage, respectively. The colcontains outgoing surface-averaged partial currents [Eqs. (15.83) umn vector ψ n,out g and (15.85)] and half-angle integrated fluxes [Eqs. (15.82) and (15.84)]; and the colcontains incoming surface-averaged partial currents and halfumn vector ψ n,in g angle integrated fluxes [Eqs. (15.80)] for each of the six (in three dimensions) nodal surfaces. The matrices P ng and R ng contain the nodal coupling coefficients. The transverse integrated formulation allows for direct transmission of neutrons entering node n over the x-surface at −x/2 across the node to exit over the n,in n,in and Jgx+ terms in Eqs. (15.82) and (15.83)], x-surface at +x/2 [e.g., the ψgx+ but does not allow for the direct transmission of neutrons entering node n over an x-surface across the node to exit over a y- or z-surface. 15.5 Transverse Integrated Nodal Discrete Ordinates Method A nodal transport equation can also be formulated in terms of the discrete ordinates approximation. The development is similar to that of Sections 15.3 and 15.4 and we will only briefly examine how the nodal coupling equations are formulated in terms of discrete ordinates. In two-dimensional Cartesian geometry, the multigroup discrete ordinates equations with isotropic scattering within a node of constant homogenized cross section may be written μm ∂ψgm ∂x (x, y) + μm ∂ψgm ∂y g (x, y) + t ψgm (x, y) = g s Q g (x, y) φg (x, y) + 2π 2π (15.87) where ψgm (x, y) = ψg (x, y, m ) is the group flux in the ordinate direction m , μm = nx · m , and ηm = ny · m . Letting the node extend over −x/2 < x < x/2, −y/2 < y < y/2 and integrating Eq. (15.87) over −y/2 < y < y/2 leads to the transverse-integrated one-dimensional discrete ordinates equation μm m (x) ∂ψgx ∂x g m + t ψgx (x) = g g g s φg (x) Qx (x) Lmg (x) + − ≡ Sgm (x) (15.88) 2π 2π y where the y-direction transverse leakage is  (y/2) ∂ψgm (x, y) g m m dy ηm (x) − ψgy− (x) = ηm ψgy+ Lmy (x) ≡ ∂y −(y/2) m (x) = ψ m (x, y = ±y/2). and ψgy± gy (15.89) 561 562 15 Nodal and Synthesis Methods We now make a polynomial expansion of Sgm (x) within the node in the polynomials 1 3 , f3 (x) = x 3 − x, 12 20 3 3 5 5 (15.90) , f5 (x) = x 5 − x 3 + x, f4 (x) = x 4 − x 2 + 14 560 18 336 15 5 2 5 x − f6 (x) = x 6 − x 4 + 44 176 14,784 f0 (x) = 1, f1 (x) = x, f2 (x) = x 2 − integrate Eq. (15.88) over −x/2 < x < x/2 in the direction of neutron flow (i.e., from −x/2 to +x/2 for μm > 0 and in the opposite direction for μm < 0), and make use of the orthogonality property  x/2   x/2 fi (x)fj (x) dx = dx fi 2 (x) δij (15.91) −x/2 −x/2 to obtain a relation among the outward-directed fluxes at one boundary, the inwarddirected fluxes at the other boundary, and the group sources within the node m,out ψgx± =  6 g 1  m x/2 m S fi (±x)e(t (x−x/2))/μ dx gi μm −x/2 i=0 g m,in −(t x/|μ + ψgx∓ e m |) (15.92) m are the coefficients of the polynomial expansion of S m (x). where the Sgi g Equation (15.88) can be integrated over −x/2 < x  < x for μm > 0 and in the m (x) similar opposite direction for μm < 0 to obtain an expression for the flux ψgx to that given by Eq. (15.92) but with the upper limit of the integral replaced by x. This expression can expanded in polynomials, multiplied by fi , and integrated over −x/2 < x < x/2 to obtain an expression for the ith node-averaged flux expansion coefficient in terms of the inward flux at ±x/2 and the group sources within the node:   x 6 1  m x/2  m m ψgi = m Sgj dx fi (±x) dx  fj (±x  )e(t (x −x)/|μ |) μ Di −x/2 −x/2 j =0 + 1 m,in ψ Di gx∓  x/2 −x/2 g fi (±x)et (x−x/2)/|μ | , m i = 0, . . . , 6 The fluxes at the interface between nodes n − 1 and n are coupled by  m,in   m,out  , μm > 0 ψgx+ n−1 = ψgx− n  m,in   m,out  ψgx+ n−1 = ψgx− , μm < 0 n (15.93) (15.94) which enables the development of equations for solving the x-direction transverse integrated equations. A similar procedure is then applied to develop and solve the y-direction transverse integrated equations. 15.6 Finite-Element Coarse Mesh Methods 15.6 Finite-Element Coarse Mesh Methods The finite-element methodology provides a systematic procedure for developing coarse mesh equations with higher-order accuracy than the conventional finitedifference equations. In the finite-element method, the spatial (or other) dependence of the neutron flux and current are represented by a supposition of trial functions which are nonzero only within a limited range of the spatial variables. These trial functions are continuous within volumes Vi , but may be discontinuous across the interfaces between adjacent volumes. The finite-element approximation will be developed from a variational principle that admits discontinuous trial functions, but it could also be derived from a weighted residuals development. The development of finite-element approximations will be discussed for onegroup P1 and diffusion theory. The results can formally be extended to multigroup theory by replacing the total and fission cross sections with diagonal cross-section matrices, replacing the scattering cross section with the multigroup scattering matrix, and replacing fluxes and currents with column vectors of multigroup fluxes and currents. Variational Functional for the P1 Equations The volume of the reactor core may be subdivided into volumes Vi within which the trial functions for the neutron flux and current are continuous. These regions are bounded by interfaces Sk across which the trial functions may be discontinuous. A variational functional for the one-group P1 equations is , + F1 J∗ , φ ∗ , J, φ     νf ∗ t − s − φ + ∇ · J dr φ = k Vi i   ∗ −1 J · D J + ∇φ dr + nk · φk∗ [Jk+ − Jk− ] ds + + i Vi k sk  k J∗ · nk [φk+ − φk− ] ds Sk (15.95) where the first two terms are sums over the volumes within which the admissible trial functions are continuous and the last two terms are surface integrals over the interfaces between these volumes. The subscripts k+ and k− refer to limiting values as the surface k is approached from the positive and negative sides, respectively. The stationarity of this variational functional with respect to independent and arbitrary variations of the adjoint flux (φ ∗ ) and current (J∗ ) within the different volumes Vi requires that 563 564 15 Nodal and Synthesis Methods    νf δφ ∗ t − S − φ + ∇ · J dr = 0 k Vi   νf ∇ · J + t − S − φ = 0, r ∈ Vi k δF1 ∗ δφ = δφi i ⇒ δF1 ∗ δJ = δJ∗i i   (15.96) δJ∗ · [D − J + ∇φ] dr = 0 ⇒ J = −D∇φ, r ∈ Vi (15.97) (i.e., that the P1 equations are satisfied within the different volumes Vi ). The stationarity of this variational functional with respect to independent and arbitrary variations of the adjoint flux (φk∗ ) and current (J∗k ) on the interfaces between volumes Vi requires that  δF1 ∗ δφ = δφk∗ nk · [Jk+ − Jk− ] = 0 δφk∗ k Sk ⇒ nk · Jk+ = nk · Jk− δF1 ∗ δJ = δJ∗k k  Sk (15.98) δJ∗k nk [φk+ − φk− ] = 0 ⇒ φk+ = φk− (15.99) (i.e., that the normal component of the current and the flux are continuous across each interface). Thus, the requirements that the variational functional of Eq. (15.95) is stationary with respect to arbitrary and independent variations of the adjoint flux and current in each volume Vi and on each interface Sk is equivalent to the requirements that the Pl equations are satisfied within each volume Vi and that the normal component of the current and the flux are continuous across the interfaces bounding these volumes. This equivalence will now be exploited to develop finite-element approximations for the neutron flux distribution. One-Dimensional Finite-Difference Approximation Although it is not a finite-element approximation per se, it is instructive to derive variationally the conventional finite-difference approximation for a slab extending from 0 < x < a with zero flux boundary conditions. The slab is partitioned into N mesh intervals and the flux and current are expanded in piecewise constant functions φ(x) = N−1  φn Hn (x), n=1 J (x) = N−1  n=0 φ ∗ (x) = N−1  φn∗ Hn (x) (15.100) n=1 Jn Kn (x), J ∗ (x) = N−1  n=0 Jn∗ Kn (x) (15.101) 15.6 Finite-Element Coarse Mesh Methods Fig. 15.4 Trial functions for finite-difference approximation. where the Hn and Kn  1, Hn (x) = 0,  1, Kn (x) = 0, are Heaviside functions: xn − 12 hn−1 < x < xn + 12 hn otherwise xn < x < xn+1 otherwise (15.102) the domain of which is illustrated in Fig. 15.4. The volumes Vi over which the flux and adjoint flux trial functions are continuous are the mesh intervals xn −hn−1 /2 < x < xn + hn /2, and the surfaces bounding these regions are at xn − hn−1 /2 and xn + hn /2. The volumes Vi over which the current and adjoint current trial functions are continuous are the mesh intervals xn < x < xn+1 , and the surfaces bounding these regions are at xn and xn+1 . For the piecewise constant adjoint flux trial functions of Eq. (15.100), the variations on the surfaces are not independent of the variations in the volumes; that is, instead of having separate Eqs. (15.96) and (15.98), these two equations must be combined, leading in this case to  xn + 1 hn    N−1 2 νf dJ δF1 ∗  ∗ φ + δφ = δφ dx  −  − t S n δφ ∗ k dx xn − 12 hn−1 n=1 + (Jn − Jn−1 ) = N−1  n=1 δφn∗    νf n−1 1 hn−1 tn−1 − sn−1 − φ 2 k   νf n 1 φn + (Jn − Jn−1 ) = 0 + hn tn − sn − 2 k (15.103) where the materials properties denoted by the subscript n have been taken to be uniform in the interval xn ≤ x ≤ xn+1 . Similarly, the variations of the adjoint current trial functions in the volumes and on the surfaces are not independent, requir- 565 566 15 Nodal and Synthesis Methods ing that Eqs. (15.97) and (15.99) be combined to yield  xn+1    N−1 δF1 ∗  ∗ dφ −1 δJ = δJ dx D J + − φ ) + (φ n+1 n n δJ ∗ dx xn n=0 = N−1  δJn∗ hn Dn−1 Jn + (φn+1 − φn ) = 0 (15.104) n=0 Requiring that the variational functional be stationary with respect to arbitrary and independent variations δφn∗ and δJn∗ in each mesh interval yields      νf n−1 νf n 1 1 hn−1 tn−1 + sn−1 − + hn tn − sn − φn 2 k 2 k + (Jn − Jn−1 ) = 0 Jn = −Dn (15.105) φn+1 − φn hn which may be combined to obtain the standard form of the finite-difference diffusion equation.      νf n−1 νf n 1 1 hn−1 tn−1 − sn−1 − + hn tn − sn − 2 k 2 k  Dn−1 Dn+1 Dn−1 Dn φn − + φn−1 − φn+1 = 0 (15.106) + hn−1 hn hn−1 hn The volumes Vi over which the flux and adjoint flux trial functions are continuous are the mesh intervals xn − hn−1 /2 < x < xn + hn /2, and the surfaces bounding these regions are at xn − hn−1 /2 and xn + hn /2. Diffusion Theory Variational Functional We shall restrict our attention to trial functions that are continuous over the volume of the reactor, which means that the last term in the variational functional of Eq. (15.95) is identically zero. We further restrict ourselves to current and adjoint current trial functions which satisfy Fick’s law, so that the second term in Eq. (15.95) is identically zero and the current in the first and third terms may be replaced by −D∇φ to obtain the diffusion theory variational functional     νf φ − ∇ · D∇φ dr φ ∗ t − s − Fd {φ ∗ , φ} = k Vi i   φ ∗ ns (−D∇φ)k+ − (−D∇φ)k− dS + k Sk      νf φ ∗ t − s − φ + ∇φ ∗ · D∇φ dr = k Vi i (15.107) 15.6 Finite-Element Coarse Mesh Methods Fig. 15.5 Trial functions for linear finite-element approximation. The second form of this functional resulted from integrating the divergence in the first term by parts over the various volumes to obtain terms that cancel identically with the interior surface terms in the second term and vanish on the outer boundary because of the physical boundary condition. Note that this second form of the functional admits trial functions which do not identically satisfy continuity of −D∇φ · ns across interior surfaces. Linear Finite-Element Diffusion Approximation in One Dimension We consider the same problem as above, a slab reactor with zero flux boundary conditions in one-group diffusion theory. The neutron flux is expanded φ(x) = N−1  φn Hn (x), n=1 φ ∗ (x) = N−1  φn∗ Hn (x) (15.108) n=1 in tent function trial functions  xn+1 − x   , xn < x < xn+1   hn  Hn (x) = x − xn−1  , xn−1 < x < xn     hn−1 0, otherwise (15.109) depicted in Fig. 15.5. The volumes Vi over which the trial functions φ and φ ∗ of Eq. (15.108) and the vectors D∇φ and D∇φ ∗ are continuous are just the mesh intervals xn−1 < x < xn , and the surfaces are the xn . Requiring that the variational functional of Eq. (15.107) be stationary with respect to arbitrary and independent variations in all the adjoint trial functions yields     xn νf n−1 x − xn δFd ∗ ∗  δφ = δφ −  − tn−1 sn−1 n δφn∗ n hn−1 k xn−1 * n+1 n+1  Dn−1  d [φn Hn (x)] dx φn Hn (x) + × hn−1  dx  n =n−1 n =n−1 567 568 15 Nodal and Synthesis Methods  + × xn+1  xn n+1  n =n−1   νf n xn+1 − x tn − sn − hn k *  n+1 Dn  d [φn Hn (x)] dx φn Hn (x) − hn  dx n =n−1 =0 (15.110) Carrying out the integration results in a three-point coarse mesh equation for each mesh point:     νf n−1 1 1 hn−1 tn−1 − sn−1 − φn + φn−1 k 3 6     νf n 1 1 φn + φn+1 + hn tn − sn − k 3 6 + Dn Dn−1 (φn − φn−1 ) − (φn+1 − φn ) = 0 hn−1 hn (15.111) which are similar to the finite-difference equations (15.106), but with more coupling among mesh points. Numerical studies reveal that Eqs. (15.111) can achieve the same accuracy as Eqs. (15.106) with much larger mesh spacing, hn . This result is physically intuitive because the piecewise linear representation of the flux allowed by the trial functions of Eqs. (15.109) is more realistic than the step function representation allowed by the trial functions of Eqs. (15.100) and (15.101), as illustrated in Fig. 15.6. It stands to reason that higher-order polynomial trial functions should provide an even better representation of the flux and hence be more accurate. Fig. 15.6 Finite-difference (solid lines) and linear finite-element (dashed lines) representation of flux solution. 15.6 Finite-Element Coarse Mesh Methods Higher-Order Cubic Hermite Coarse-Mesh Diffusion Approximation The cubic Hermite interpolating polynomials      x − xn−1 3 x − xn−1 2   3 −2 , xn−1 ≤ x ≤ xn   hn−1 hn−1     0 2 3 Hn (x) = xn+1 − x xn+1 − x   3 −2 , xn ≤ x ≤ xn+1   hn hn  0, otherwise        x − xn−1 2 x − xn−1 3 − hn−1 , + Hn− (x) = hn−1 hn−1  0, (15.112) xn−1 ≤ x ≤ xn otherwise        xn+1 − x 2 xn+1 − x 3 hn , xn ≤ x ≤ xn+1 − Hn+ (x) = hn hn  0, otherwise are frequently used for the development of coarse-mesh finite-element approximations. These polynomials have the properties dHn− dHn+ (xn ) = (xn ) = 1 dx dx Hn− (xn ) = Hn+ (xn ) = 0 Hn0 (xn ) = (15.113) These polynomials are used to construct trial functions: φ(x) = φ ∗ (x) = N−1  n=1 N−1  φn0 Hn0 (x) + φn− (x)Hn− (x) + φn+ Hn+ (x) (15.114) φn∗0 Hn0 (x) + φn∗− Hn− (x) + φn∗+ Hn+ (x) n=1 The second property of Eq. (15.113) ensures that this trial function is continuous at the xn . Thus the variational functional of Eq. (15.107) admits these trial functions. Requiring stationarity of the variational functional with respect to arbitrary and independent variations of all the adjoint trial functions in all interior mesh intervals; that is, requiring (δFd /δφn∗0 )δφn∗0 = 0, (δFd /δφn∗− )δφn∗− = 0, and (δFd /δφn∗+ )δφn∗+ = 0 for n = 1, . . . , N − 1 yields three equations for each mesh point: 569 570 15 Nodal and Synthesis Methods −  Dn−1  0  Dn  0 0 − φn0 + φ φ − φn−1 hn n+1 hn−1 n    νf n 7 0 3 0 1 1 − φn + φn+1 − hn φn+1 + hn φn+ + hn tn − sn−1 − k 20 20 30 20   νf n−1 + hn−1 tn−1 − sn−1 − k   7 0 3 0 1 1 + × − hn−1 φn− = 0 φn + φn−1 + hn−1 φn−1 20 20 30 20  (15.115)    νf n 3  0 hn  − 0 + hn tn − sn − − φ φ − φn + + φn k 140 n+1 420 n+1    hn  −  Dn 1 0 + − φn+1 φ − φn0 + + φn+ = 0 hn 5 10 n+1      hn  −  νf n 1 0 hn tn − sn − − φn0 + − φn+1 φn+1 − φn+ k 2 60  Dn  − + φ + φn+ = 0 6 n+1 These equations are for the interior mesh intervals. The zero flux boundary con0 to vanish (or a symmetry boundary condition would dition requires φ00 and φN 0 require, for example, φ0 = φ10 ). However, additional constraints must be imposed ± . Requiring stationarity of the variational functional at the to evaluate φ0± and φN ∗± ∗± )δφN = 0 and (δFd /δφ0∗± )δφ0∗± = 0] provides external boundaries [i.e., (δFd /δφN the additional equations that are necessary to specify the problem completely. The use of cubic Hermite polynomials is found to increase the accuracy of the finite-element approximation relative to use of the linear polynomial of Eq. (15.109), but, of course, to increase the computing time because three equations per mesh point are involved instead of only one. The accuracy of which we are speaking is the error with respect to an exact solution of the homogenized problem, not with respect to an exact solution of the true heterogeneous problem. Although it seems plausible that a more accurate solution to the global homogeneous problem, when combined with a local heterogeneous solution, will yield a more accurate solution of the actual global heterogeneous problem, this is not obvious. Multidimensional Finite-Element Coarse-Mesh Methods In two dimensions, the volume of a core can be partitioned into region volumes Vi , which we refer to as elements. These elements can have a variety of shapes: triangles, quadrilaterals, tetrahedral, and so on. A finite-element approximation for the solution is represented by a linear combination of shape functions associated with each element, normally polynomials in the local coordinates within the element. A shape function has the value unity at its associated coarse mesh point and goes 15.7 Variational Discrete Ordinates Nodal Method to zero at the surface of the volumes Vi associated with that element. For example, a quadratic polynomial φ(x, y) = a1 + az y + a3 y + a4 x 2 + a5 xy + a6 y 2 (15.116) might be used to represent the flux within a triangular element. Usually, the polynomial is redefined so that the coefficients have the values of the flux at various support points throughout the element. A quadratic approximation clearly requires six support points, a linear approximation would require three support points, and so on. The value of the flux at each support point is an unknown in the resulting equations. 15.7 Variational Discrete Ordinates Nodal Method The nodal and coarse mesh calculations described in Sections 15.2 to 15.4 proceed in three distinct steps: (1) the performance of local assembly two-dimensional transport calculation and the preparation of homogenized cross sections for each node, (2) the global solution of the nodal equations for the average flux in each node, and (3) reconstruction of the detailed heterogeneous intranode fluxes. We found in considering the coarse mesh methods that the higher-order polynomials which better represented the overall flux distribution within the coarse mesh region led to more accurate solutions of the homogenized problem, which is solved with nodal or coarse mesh equations in step 2. It is possible to combine the three steps—homogenization, flux solution, detailed flux reconstruction—into a single, self-consistent procedure that uses the detailed heterogeneous assembly transport flux directly, instead of a polynomial approximation, to represent the flux distribution within the node or coarse mesh region. Since a relatively high order transport solution is needed for the heterogeneous assembly calculation, but a relatively low order transport calculation will usually suffice for the global nodal calculation, we illustrate the development of a methodology that can make use of high-order discrete ordinates heterogeneous two-dimensional assembly calculations as trial functions to develop a low-order discrete ordinates nodal calculational model. Variational Principle A variational principle for the neutron transport equation is F ψ(r, ), ψ ∗ (r, ) =      λ=1 Vλ  +  d ψ ∗ (r, ) − · ∇ψ(r, ) − t (r)ψ(r, )  dV 4π     1 d S r,  →  ψ r,  + 4π 4π  4π    d νf (r)ψ r,  571 572 15 Nodal and Synthesis Methods +    dS σλ ,ν(λ) ν(λ) d H (− · n)( · n)ψ ∗ (rλ,ν(λ) , ) 4π × ψ(rλ,ν(λ) , ) − ψ(rν(λ),λ , )  +  d H (− · n)( · n)ψ ∗ (rex , )ψ(rex , ) dS Sλ (15.117) 4π The functional F is a sum over  volumetric reactor regions (or nodes). The first term of the sum is an integration over the nodal volume Vλ and the entire solid angle (4π ). The second term of the sum is a sum over all the interior surfaces ν(λ) of node λ; this term is included to allow trial functions that are discontinuous across any surface. The notation rλ,ν(λ) refers to the limit of all the points on the surface ν(λ) as approached from within node λ; similarly, rν(λ),λ refers to those same points as approached from the node adjacent to node λ [the node on the other side of surface ν(λ)]. Each of the terms in this sum is an integral over the surface σλ,ν(λ) (formed by the points rλ,ν(λ) and the solid angle 4π ). The final term in the sum is an integral over the exterior surface Sλ of node λ (formed by the points rex ). This term is included to allow trial functions that do not satisfy vacuum boundary conditions. In the functional F , n refers to the outward unit normal vector from node λ across an interior or exterior surface, and H is the Heaviside step function. In addition, t (r), S (r,  → ), and νf (r) are the usual cross sections for removal, scattering from angle  to , and neutron production from fission, respectively. Although the functional F has been presented in a one-energy-group (or energy-independent) form, the extension of the results below to the multigroup case is straightforward. The condition (δF /δψ ∗ )δψ ∗ = 0 requires that the stationary value of the trial function ψ(r, ) be identified as the forward angular flux satisfying the Boltzmann transport equation,  · ∇ψ(r, ) + t (r)ψ(r, )         1 = d S r,  →  ψ r,  + d νf (r)ψ r,  4π 4π 4π (15.118) as well as the interface continuity and vacuum boundary conditions ψ(rλ,ν(λ) , ) − ψ(rν(λ),λ , ) = 0 (15.119) ψ(rex , ) = 0, (15.120) and ·n<0 respectively. The condition (δF /δψ)δψ = 0 requires that the stationary value of the trial function ψ ∗ (r, ) be identified as the adjoint angular flux satisfying the adjoint trans- 15.7 Variational Discrete Ordinates Nodal Method Fig. 15.7 Bounding surface notation. port equation − · ∇ψ ∗ (r, ) + t (r)ψ ∗ (r, )     ∗    1    = d S r,  →  ψ r,  + d νf (r)ψ ∗ r,  4π 4π 4π (15.121) as well as the interface continuity and vacuum boundary conditions ψ ∗ (rλ,ν(λ) , ) − ψ ∗ (rν(λ),λ , ) = 0 (15.122) ψ ∗ (rex , ) = 0, (15.123) and ·n>0 respectively. To apply the functional F to develop a nodal method, the reactor volume is first partitioned into I × J × K regions, where I , J , and K are the number of partitions along the x, y, and z coordinates, respectively, and I × J × K is equal to the  of the overall sum in F . Node ij k is bounded by the surfaces xi , xi+1 , yj , yj +1 , zk , and zk+1 as illustrated in Fig. 15.7. The nodal or volumetric domain function, designated ij k (x, y, z), is defined as  1, xi < x < xi+1 , yj < y < yj +1 , zk < z < zk+1 ij k (x, y, z) = (15.124) 0, otherwise The I × J regions in the radial (x–y) plane are called channels. The angular flux ψ(x, y, z, ) is represented in each channel ij as the product of a one-dimensional axial function gij k (z, ) and a precomputed two-dimensional planar function fij k (x, y, ) that is used over the axial domain k. The angular dependence of the axial functions gij k (z, ) is discretized into eight functions gijn k (z), one for each octant of the unit sphere, and the octant domain function is designated n (), defined as  1,  within octant n n () = (15.125) 0, otherwise 573 574 15 Nodal and Synthesis Methods Fig. 15.8 Angular geometry notation for domain functions mn () in an S8 quadrature. The angular geometry is illustrated in Fig. 15.8 with a hypothetical arrangement of the 10mn () in octant 1 (μ > 0, η > 0, and ξ > 0) for a scheme with M = 80 (corresponding to an S8 quadrature set). Note that the boundaries between adjacent domains are arbitrary. The surface area of that part of the unit sphere corresponding to region m within octant n is designated wmn , defined as   1 1 wmn ≡ d n ()mn () = d mn () (15.126) 4π 4π 4π π/2 where mn is unity within region m in octant n and otherwise zero. Thus the wmn have the same interpretation as standard discrete-ordinates weights, and 8 M/8 mn = 1. Note that the notation mn is shortened for “m within n” and n=1 m=1 w that this notation scheme requires that the M/8 subregions of octant n be ordered symmetrically with respect to those of each of the other octants, thus effectively specifying the use of a standard level-symmetric quadrature set. This requirement may be eliminated with a suitable, though possibly more confusing change of notation and is in no way limiting. As indicated in Fig. 15.9, the angle  is decomposed into its three direction cosines as follows (i, j, and k are the unit vectors along the x, y, and z axes, respectively): μ =  · i, η =  · j, ξ =·k (15.127) In addition, the azimuthal angle ω is defined (see Fig. 15.9) to be the angle between the z-axis and the projection of  onto the y–z plane; thus ) η = 1 − μ2 sin ω (15.128) ) ξ = 1 − μ2 cos ω 15.7 Variational Discrete Ordinates Nodal Method Fig. 15.9 Definition of angles. Consistent averages of the  direction cosines may be defined as follows:  1 μmn ≡ d mn ()μ octant 4πwmn n  1 d mn ()η ηmn ≡ (15.129) octant 4πwmn  n 1 d mn ()ξ ξ mn ≡ octant 4πwmn n Using the domain functions defined in Eqs. (15.124) and (15.125), the trial function used for the forward angular flux in the functional F is ψ(x, y, z, ) ≈ g(z, )f (x, y, ) = I ×J ×K  ij k (x, y, z)[gij k (z, )fij k (x, y, )] i,j,k=1 = I ×J ×K  i,j,k=1 ij k (x, y, z) 8  n=1 n ()gijn k (z) M/8  mn ()fijmn k (x, y) m=1 (15.130) The adjoint angular flux ψ ∗ (x, y, z, ) is expanded analogously. 575 576 15 Nodal and Synthesis Methods Using the trial function of Eq. (15.130) and the analogous expansion for the adjoint flux in F and requiring stationarity of the functional with respect to each of the adjoint axial functions gij∗nk (z) yields reduced equations for the forward axial functions gijn k (z), with homogenized parameters defined in terms of the precom∗mn puted nodal basis functions fijmn k (x, y) and fij k (x, y). The equation for each of n the eight axial functions gij k (z) is ξ̄ijn k dgijn k (z) dz = 8  n =1 ¯ ij k (z)gijn k (z) + B̄ijn k gijn k (z) +   8   nn n ¯ f,ij ν k (z)gij k (z) + n =1 nn n ¯ s,ij  k (z)gij k (z) n − (1 − δi1 )H (μn ) μ̄nij k (xi )gijn k (z) − μ̄n(i,i−1)j k (xi )g(i−1)j k (z) n + (1 − δi1 )H (−μn ) μ̄nij k (xi+1 )gijn k (z) − μ̄n(i,i+1)j k (xi+1 )g(i+1)j k (z) n n n n − δi1 H (μn )μ̄n1j k (xi )g1j k (z) + δiI H (−μ )μ̄Ij k (xI +1 )gIj k (z) n n n n − (1 − δj 1 )H (ηn ) η̄ij k (yj )gij k (z) − η̄i(j,j +1)k (yj )gi(j +1)k (z) n n n n + (1 − δj J )H (−ηn ) η̄ij k (yj +1 )gij k (z) − η̄i(j,j +1)k (yj +1 )gi(j +1)k (z) n n n n − δj 1 H (ηn )η̄i1k (yi )gi1k (z) + δj J H (−ηn )η̄iJ k (yJ +1 )giJ k (z) (15.131) In Eq. (15.131), the homogenized total, fission, and scattering cross sections in channel ij and axial region k for octant n are defined consistently as ¯ ijn k (z) =  xi+1 xi M/8 mn dy (x, y, z) m=1 wmn fij∗mn k (x, y)fij k (x, y) M/8 xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi (15.132) dx yj +1 yj nn ¯ f,ij  k (z) = xi+1 xi dx yj +1 yj dy f (x, y, z) xi+1 xi dx yj +1 yj M/8 M/8 m n m n mn ∗mn fij k (x, y) m=1 w fij k (x, y) m =1 w M/8 mn ∗mn mn dy m=1 w fij k (x, y)fij k (x, y) (15.133) and mn ¯ s,ij  k (z) = %  xi+1 dx xi × yj +1 yj  M/8  m=1 w dy L  (2l + 1)sl (x, y, z) l=0 mn mn Pl (μ  )fij∗mn k (x, y) 15.7 Variational Discrete Ordinates Nodal Method ×  M/8         wm n Pl (μm n )fijmk n (x, y) m =1 +2  M/8 l  (l − k)!  k=1 × (l + k)!  M/8   w m n       Plk (μm n ) cos(kωm n )fijmk n (x, y)   M/8 l  (l − k)!  k=1 × (l + k)!  M/8  m =1  w  xi+1 dx xi Plk (μmn ) cos(kωmn )fij∗mn k (x, y) m=1 m =1 +2  mn w mn Plk (μmn ) sin(kωmn )fij∗mn k (x, y) m=1       *&"   wm n Plk (μm n ) sin(kωm n )fijmk n (x, y) yj +1 dy yj M/8  mn wmn fij∗mn k (x, y)fij k (x, y) (15.134) m=1 respectively. Note that in deriving Eq. (15.134), the scattering cross section s (r,  → ) has been expanded in a Legendre polynomial of order L, and the addition theorem for spherical harmonics has been applied in the usual way. The isotropic portion of the homogenized scattering cross section has the same form as that of the fission cross section above. The transverse leakage from node ij k (i.e., the leakage in the x- and y-directions) is defined consistently as   yj +1 M/8 xi+1  ∂fijmn k (x, y) n dx dy wmn μmn fij∗mn (x, y) B̄ij k = k ∂x xi yj m=1  +   xi+1 dx xi yj  xi+1 dx xi yj +1 yj +1 yj dy M/8  w mn mn η fij∗mn k (x, y) ∂fijmn k (x, y) m=1 dy M/8  " ∂y mn wmn fij∗mn k (x, y)fij k (x, y) (15.135) m=1 The homogenized discrete ordinate for octant n is defined consistently as ξ̄ijn k = M/8 xi+1 y mn dx yjj +1 dy m=1 wmn ξ mn fij∗mn k (x, y)fij k (x, y) xi M/8 xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi (15.136) The four parameters required (for each octant) for coupling node ij k to its x-direction neighbors, nodes (i − 1)j k and (i + 1)j k, are 577 578 15 Nodal and Synthesis Methods μ̄nij k (xi ) μ̄n(i,i−1)j k (xi ) = = μ̄nij k (xi+1 ) = μ̄n(i,i−1)j k (xi+1 ) = yj +1 yj dy xi+1 xi dx yj +1 yj xi+1 xi dy dx M/8 mn μmn f ∗mn (x , y)f mn (x , y) i i ij k ij k M/8 mn ∗mn mn dy m=1 w fij k (x, y)fij k (x, y) m=1 w yj +1 yj M/8 mn mn mn ∗mn m=1 w μ fij k (xi , y)f(i−1)j k (xi , y) M/8 yj +1 mn dy m=1 wmn fij∗mn k (x, y)fij k (x, y) yj (15.137) M/8 yj +1 mn dy m=1 wmn μmn fij∗mn k (xi+1 , y)fij k (xi+1 , y) yj M/8 xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi M/8 yj +1 mn dy m=1 wmn μmn fij∗mn k (xi+1 , y)f(i+1)j k (xi+1 , y) yj M/8 xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi The four parameters required (for each octant) for coupling node ij k to its y-direction neighbors, nodes i(j − 1)k and i(j + 1)k, are M/8 xi+1 mn dy m=1 wmn ηmn fij∗mn k (x, yj )fij k (xi , yj ) xi n (y ) = η̄ij M/8 k j xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi n η̄i(j,j −1)k (yi ) = M/8 xi+1 mn dx m=1 wmn ηmn fij∗mn k (x, yj )fi(j −1)k (x, yj ) xi M/8 xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi (15.138) n (y η̄ij k j +1 ) = n η̄i(j,j +1)k (yj +1 ) = M/8 xi+1 mn dx m=1 wmn ηmn fij∗mn k (x, yj +1 )fij k (x, yj +1 ) xi  M/8 xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi M/8 xi+1 mn dx m=1 wmn ηmn fij∗mn k (x, yj +1 )fi(j +1)k (x, yj +1 ) xi  M/8 xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi Note that the arguments of the Heaviside step functions H in Eqs. (15.131) are the direction cosines μmn and ημν of Eqs. (15.129), but with the single superscript n because only the octant needs to be identified. The interface conditions that couple node ij k to its axially adjacent neighbors, nodes ij (k − 1) and ij (k + 1), are ξ̄ijn k gijn k (zk ) = ξ̄ijn (k,k−1) gijn (k−1) (zk ), ξ̄ijn k > 0 ξ̄ijn k gijn k (zk+1 ) = ξ̄ijn (k,k+1) gijn (k+1) (zk+1 ), ξ̄ijn k < 0 (15.139) where the coupling parameters are defined as M/8 xi+1 y mn dx yjj +1 dy m=1 wmn ξ mn fij∗mn k (x, y)fij (k−1) (x, y) xi n ξ̄ij (k,k−1) =  M/8 xi+1 y mn dx yjj +1 dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi (15.140) M/8 mn mn ∗mn xi+1 yj +1 mn dx dy w ξ f (x, y)f (x, y) ij k m=1 ij (k+1) x y i j ξ̄ijn (k,k+1) = M/8 xi+1 yj +1 mn dx dy m=1 wmn fij∗mn k (x, y)fij k (x, y) xi yj 15.7 Variational Discrete Ordinates Nodal Method Finally, the boundary conditions on Eqs. (15.131) are gijn 1 (z1 ) = 0, ξ̄ijn 1 > 0 gijn K (zK+1 ) = 0, ξ̄ijn K < 0 (15.141) Using the angular flux trial function of Eq. (15.130) in the usual definition of the isotropic flux specifies the heterogeneous flux reconstruction equation:  φ(x, y, z) ≡ d ψ(x, y, x, ) 4π  ≈ d 4π = 8  n ()gijn k (z) n=1 gijn k (z) n=1 = 4π 8   M/8  m=1 gijn k (z) mn ()fijmn k (x, y) m=1 fijmn k (x, y) m=1 8  M/8  M/8  d n ()mn () 4π wmn fijmn k (x, y) (15.142) m=1 Application of the Method The steps required for application of the variational nodal discrete-ordinates method are the same as those required for standard nodal methods. First, a set of fine-mesh high-order two-dimensional calculations is performed for small heterogeneous local regions in the x–y plane, such as assemblies or extended assemblies, for the purposes of homogenizing the nodes. (It is also possible to use full twodimensional planar calculations to provide nodal trial functions.) In standard nodal methods, even for full three-dimensional global calculations, only two-dimensional local calculations are performed, but the manner of axial coupling required for the nodes is rarely specified. In the variational nodal discrete-ordinates method, the local calculations are performed using the discrete-ordinates (SN ) method. The finemesh SN calculations yield the angular fluxes fijmn k (x, y), and M = N(N + 2) for a three-dimensional problem. For this method it is also necessary to calculate the adjoint angular SN fluxes fij∗mn k (x, y) in each node. There is a different homogenized cross section defined for each of the eight S2 directions; however, because of the axial symmetry obtained by the use of two-dimensional basis functions, only four ∗mn are required. The basis functions fijmn k (x, y) and fij k (x, y) are used with standard SN ordinates and weights to compute homogenized parameters in accordance with the definitions of Eqs. (15.132) to (15.138) and (15.140). The second step in standard nodal methodology is a global diffusion-theory calculation, which involves (in general, for the transverse integrated methods) three one-dimensional equations for the transverse integrated x-, y-, and z-direction fluxes. Usually, the problem is reduced to that of finding coefficients of fourthorder polynomials. In the variational nodal method, the global equations are one- 579 580 15 Nodal and Synthesis Methods dimensional (z-direction) S2 first-order differential equations, which are equivalent in accuracy to the diffusion equations. The spatial discretization to use on the z-axis is not specified; thus any method may be used, including coarse-mesh, finite-difference, high-order polynomial expansion, or other standard method. The final step in the nodal calculation for the standard and variational nodal methods is the reconstruction of heterogeneous fluxes or reaction rates from the homogeneous (global) calculation results. In the variational nodal method, the flux reconstruction is completely specified by Eq. (15.142). 15.8 Variational Principle for Multigroup Diffusion Theory A complete mathematical description of the neutron distribution, within the context of multigroup diffusion theory, is provided by a coupled set of partial differential equations for the direct and adjoint flux (and current) with associated boundary, initial, final, and continuity conditions. An equivalent variational formulation must not only have the original equations as Euler equations, but must also embody the associated boundary, initial, final, and continuity conditions, either directly or indirectly through limitations on the admissible class of trial functions. The following variational principle embodies all these conditions:  J =   tf F T φ + φ ∗T ∇ · j + φ ∗T τ φ̇ drr φ ∗T  − (1 − β)χF dt t0 V − φ ∗T M  λm χ m Cm − j ∗T · 3 trj − j ∗T · ∇φ m=1 − M  ∗ Cm βm F T φ + m=1 +  + dt t0 Sin ds n · ∗ Cm λm Cm m=1 ∗ Cm Ċm m=1  tf M  M  + ∗T * ∗T −φ S +S φ 1  ∗T j + − j −) γ φ ∗T + + (1 − γ )φ − (j  ,  j ∗T − ηjj ∗T + + (1 − η)j − (φ + − φ − )  +  drr V   2    aφ ∗T (+) + (1 − a)φ ∗T (−) τ φ(+) − φ(−) M     ∗ ∗ (−) Cm (+) − Cm (−) + bCm (+) + (1 − b)Cm m=1 *  + 15.8 Variational Principle for Multigroup Diffusion Theory  V drr φ ∗T (t0 )τ (φ(t0 ) − g 0 ) − g ∗T f τ φ(tf ) +  +  tf dt t0 M  Cm (t0 ) − hm0  * − h∗mf Cm (tf ) m=1 ds s0  ∗ (t0 ) Cm +  j s0 · n) j ∗T s0 · n (φ s0 + j  , ≡ J1 + J2 + J3 + J4 + J5 4 5 (15.143) where φ∗, φ = G × 1 column matrices of group adjoint and direct flux, respectively = G × 1 column matrices of group adjoint and direct j ∗, j current, respectively—vector quantities S ∗ , S = G × 1 column matrices of group adjoint and direct source, respectively ∗ , C = scalar adjoint and direct delayed neutron precursor Cm m densities, respectively  = G × G matrix of group removal and scattering cross sections = G × G diagonal matrix of group transport cross sec tr tion τ = G × G diagonal matrix of inverse group neutron speeds F = G × 1 column matrix of group nu-fission cross sections χ, χ m = G × 1 column matrices of prompt- and delayedfission neutron spectra, respectively λm , βm = delayed neutron precursor decay rate and precursor yield per fission, respectively The term in the first set of brackets is an integral over the time of interest, t0 ≤ t ≤ tf , and the volume of the reactor. The Euler equations for this term are the direct and adjoint flux, current, and precursor equations, which result from the requirement that the first variations of J1 with respect to each of the argu∗ , C ) vanishes. In taking the first variation of J , ment functions (φ ∗ , φ,jj ∗ ,jj , Cm m 1 integration by parts is required, which introduces certain additional terms. The requirement that these additional terms vanish, and hence that stationarity of J1 implies satisfaction of the Euler equations, imposes restrictions on the admissible class of trial functions. The purpose of the additional terms, J2 − J5 , is to remove these restrictions. If direct and adjoint flux and current trial functions that are discontinuous across an internal interface, Sin , are admitted, a term of the general form of J2 must be added to J1 in order that stationarity of the functional J12 ≡ J1 + J2 implies 581 582 15 Nodal and Synthesis Methods satisfaction of the Euler equations and flux and current continuity conditions. The subscripts indicate limiting values on the + and − sides, with respect to the unit normal vector n, of the surface Sin . γ and η are arbitrary constants. Terms of the general form of J2 have given rise to an overdetermination of interface conditions in synthesis applications. Consider, for example, the variation of J12 with respect to φ ∗ (by a variation with respect to the column vector we intend separate and independent variations with respect to each element of the column matrix): δJ12 δφ ∗T δφ ∗T  =0=   tf drr δφ dt t0 ∗T V F T φ + ∇ · j + τ φ̇  − (1 − β)χF −  + dt t0 Sin * λm χ m Cm − S m=1  tf M  ds n ·  , × (jj + − j − ) + 1 ∗T γ δφ ∗T + + (1 − γ )δφ − (15.144) 2 For completely arbitrary δφ ∗T , the first term vanishes only if the expression within the first set of braces is identically zero, which is just the condition that the neutron balance equation is satisfied. Vanishing of the second term for arbitrary δφ ∗T + and appears to lead to two current continuity conditions. However, continuity of δφ ∗T − ∗T = δφ , and in fact there is only one current conadjoint flux requires that δφ ∗T + − tinuity condition. The difficulty in synthesis applications results from the failure ∗T to impose the condition δφ ∗T + = δφ − on trial functions that are partially specified. If direct and adjoint flux and precursor trial functions which are discontinuous in time at tin are admitted, a term of the general form of J3 must be added to J1 in order that stationarity of the functional J13 = J1 + J3 implies the satisfaction of the Euler equations and flux and precursor time continuity conditions. The + and − arguments refer to times just after and just before, respectively, tin . a and b are arbitrary constants. An overdetermination problem, analogous to that discussed for J2 , has also arisen in synthesis applications of J3 . If direct flux and precursor trial functions that do not satisfy the known initial conditions g 0 and hm0 , and adjoint flux and precursor trial functions that do not satisfy known final conditions gf∗ and h∗mf , are admitted, J4 must be added to J1 in order that stationarity of the resulting variational principle implies satisfaction of the Euler equations and the appropriate initial and final conditions. Similarly, stationarity of J15 = J1 + J5 implies satisfaction of the Euler equations and the external boundary conditions φ s0 + jj s0 · n = 0, φ ∗s0 + jj ∗s0 · n = 0, even if the flux and current trial functions do not satisfy these boundary conditions identically. 15.9 Single-Channel Spatial Synthesis A general second-order variational principle for multigroup diffusion theory can also be written which admits the same extended class of trial functions as J of Eq. (15.143), and which leads to the same apparent interface overdetermination problem in synthesis applications. Using Fick’s law to relate flux and current, and integrating by parts in Eq. (15.143), leads to  F=   tf F T φ + ∇φ ∗T · D ∇φ + φ ∗T τ φ drr φ ∗T  − (1 − β)χF dt t0 V − φ ∗T M  λm χ m Cm − m=1 +  + dt t0 Sin ∗ Cm βm F T φ m=1 ∗ Cm λm Cm + m=1  tf M  M  M  ∗ Cm Ċ * ∗T ∗T −φ S +S φ m=1 1 +   ∗T D + ∇φ + + γ D − ∇φ − ds n · φ ∗T + − φ − (1 − γ )D  ,  ∗T + η∇φ ∗T + D + + (1 − η)∇φ − D − (φ + − φ − )  +   M   ∗ ∗ (+) + (1 − b)Cm (−) bCm  Cm (+) − Cm (−) *  m=1  V 2  ∗T    aφ (+) + (1 − a)φ ∗T (−) τ φ(+) − φ(−) drr V + +  3 drr φ ∗T (t0 )τ (φ(t0 ) − g 0 ) − g ∗T f τ φ(tf ) +   + − M  ∗ (t0 ) Cm  Cm (t0 ) − hm0 m=1  tf dt t0 ds  * − h∗mf Cm (tf ) 4 + s0 n · ∇φ ∗T s0 D (φ s0 ≡ F 1 + F2 + F3 + F4 + F5 D ∇φ s0 · n) − φ ∗T − D s0 D ∇φ s0 ·n  , 5 (15.145) The diffusion coefficient matrix, D = 13  −1 tr , has been introduced in Eq. (15.145). 15.9 Single-Channel Spatial Synthesis The basic idea of single-channel synthesis is illustrated by the example of a uniform reactor with a rod (or bank of rods) partially inserted, as illustrated in Fig. 15.10. A few diffusion lengths above and below the rod tip the flux solution is essentially a one-dimensional radial flux shape φrod and φunrod , respectively. In the vicinity of the rod tip, it is plausible that some mixture of the two flux shapes will describe the 583 584 15 Nodal and Synthesis Methods Fig. 15.10 Single-channel synthesis example. actual radial flux distribution. The synthesis approximation is developed by using trial functions of the form φ(x, y, z, t) = N  ψ n (x, y)ρ n (z, t) (15.146) n=1 j (x, y, z, t) = N  J nx (x, y)bb n (z, t)i + J ny (x, y)gg n (z, t)j n=1 + J nz (x, y)dd n (z, t)k (15.147) with similar expansions for the adjoint flux and current. ψ n and the J n are G × G diagonal matrices with elements given by the known group expansion functions g g ψn (x, y) and Jn (x, y), while ρ n , b n , g n , and d n are G × 1 column matrices with elements given by the corresponding unknown group expansion coefficients. (Direct and adjoint expansion functions must each be linearly independent, but similar functions may be used for direct and adjoint expansion functions.) Precursor trial functions of the form  βm T F (x, y, z, t) ψ n (x, y)πCm,n (z, t) λm N Cm (x, y, z, t) = (15.148) n=1 ∗ (x, y, z, t) = χ Tm Cm N  ∗ ψ ∗n (x, y)πCm,n (z, t) (15.149) n=1 are used, where π is a G × 1 column matrix with unit elements (i.e., a sum vector). 15.9 Single-Channel Spatial Synthesis When the variational principle J of Eq. (15.143) is required to be stationary with respect to arbitrary variations in the trial functions, which are limited to variations in the expansion coefficients because the expansion functions are fixed, equations that must be satisfied by the expansion coefficients are obtained. If the trial functions above are used throughout the reactor and at all times, then J2 and J3 are identically zero. In this case, equations valid for 0 < z < L and t > t0 are obtained from J1 and that part of J5 contributed by the vertical (side) external surface. δJ = 0, n = 1, . . . , N δJ = 0, δbb ∗T n n = 1, . . . , N δJ = 0, δgg ∗T n n = 1, . . . , N δJ = 0, δdd ∗T n n = 1, . . . , N δρ ∗T n (15.150) δJ = 0, m = 1, . . . , M; n = 1, . . . , N ∗ δCm,n  Equations (15.150) can be combined to eliminate b n , g n , and d n , leaving NG scalar equations, which can be written in matrix form as   M  ∂ ∂ M +R − A βm F m C m + S ρ + T ρ̇ = ∂z ∂z (15.151) m=1 where A , M , and R are NG × NG matrices, and F m is an NG × N matrix. R and A are radial and axial leakage matrices resulting from the elimination of b n , g n , and d n . ρ and S are NG × 1 column matrices and C m is an N × 1 column matrix. Thus G three-dimensional, time-dependent second-order PDEs (the multigroup diffusion equations) are replaced by NG one-dimensional time-dependent second-order PDEs [Eq. (15.151)]. The M three-dimensional, first-order ODEs (precursor equations) are replaced by NM one-dimensional first-order ODEs [last of Eqs. (15.150)]. Boundary conditions at the top (z = L) and bottom (z = 0) of the model are obtained by requiring stationarity of J with respect to arbitrary variations δdd ∗T n on the top and bottom surfaces: δJ δdd ∗T n (z = 0, L) = 0, n = 1, . . . , N (15.152) Initial conditions are derived by requiring stationarity of J with respect to arbi∗T and δC ∗ at t = t : trary variations δρN 0 m,n δJ δρ ∗T n (t0 ) = 0, n = 1, . . . , N (15.153) 585 586 15 Nodal and Synthesis Methods δJ = 0, C ∗T δC m,n (t0 ) m = 1, . . . , M; n = 1, . . . , N (15.154) A formally identical result could be obtained by deriving the synthesis equations from the second-order variational principle F , the only difference arising in the definition of the elements of the leakage matrices R and A in Eq. (15.151). Under certain restrictive conditions the two formulations are exactly identical. Two-dimensional static flux solutions for x–y slices through the reactor at various axial locations and/or for various conditions are normally chosen as expansion functions. For some problems different sets of expansion functions are appropriate for different axial regions, and it is convenient to use a discontinuous trial function formulation. In this case a term of the form of J2 would be included in the variational principle for each x–y planar surface at which the set of expansion functions changed. Equation (15.151) would again obtain within each axial zone, with the coefficients defined in terms of the expansion functions appropriate to that zone. Interface conditions result from the requirement that the variational principle be stationary with respect to arbitrary variations δjj ∗T · n and δφ ∗T on the interface, which results in  zn  tf    ∗T ∗T d ∗T dt dx dy n · k ηδdd ∗T n− J nz− n + J n z+ + (1 − η)δd t0 Sin n =1 × N  * (ψ n+ ρ n+ − ψ n− ρ n− ) = 0 (15.155) n=1  dt t0   tf Sin dx dy n · k N    ∗T ∗T ∗T γ δρ ∗T n + ψ n + + (1 − γ )δρ n− ψ n− n =1 × N  * (JJ nz+d n+ − J nz−d n− ) = 0 (15.156) n=1 ∗T If every δρ ∗T n + and δρ n − is assumed independent, 2N equations relating the Ndn d ∗T are obtained, and similarly the assumption that every δdd ∗T n + and δd n − is independent leads to 2N equations relating the Nρ n . Hence the system is overdetermined by a factor of 2. Several stratagems, have evolved for avoiding this difficulty. By requiring that the flux (direct and adjoint) and current (direct and adjoint) trial functions not be discontinuous at the same interface, the overdetermination ∗T d ∗T d ∗T problem disappears. In this case J ∗T nz+ ≡ J nz− and δd n+ ≡ δd n− , and so on. This technique of staggering the interfaces at which flux and current expansion functions are changed, which has been widely employed, has the disadvantage that Sin frequently corresponds to a physical interface in the reactor and it is desirable to change current and flux expansion functions at the same point. This may be accomplished, for all practical purposes, by allowing the two interfaces at which the current and flux trial functions are discontinuous to approach each other arbitrarily 15.9 Single-Channel Spatial Synthesis closely. A second strategy is to select γ and η = 0, 1. This is essentially what is done when a Lagrange multiplier principle is used in deriving the synthesis equations and the Lagrange multipliers are expanded in terms of the flux or current expansion functions on either the + or − side of the interface. Such interface conditions are not symmetric with respect to the arbitrary choice of the + and − sides of the ∗T d ∗T d ∗T interface. A third strategy consists of requiring that δρ ∗T n+ = δρ n− and δd n+ = δd n− . 1 With γ = η = 2 , the interface conditions are independent of the arbitrary choice of the + and − sides of the interface. Thus all of these stratagems for removing the overdetermination of interface conditions have certain unsatisfactory features. As mentioned in Section 15.8, overdetermined interface conditions result because of failure to impose the restrictions δφ ∗+ = δφ ∗− and δjj ∗+ = δjj ∗− , which must be satisfied if the adjoint flux and current are continuous, upon the trial functions. Although these restrictions cannot be imposed exactly, because the trial functions are partly specified, they can be imposed in an approximate manner to obtain re∗T d ∗T d ∗T lations among the variations δρ ∗T n+ and δρ n− among δd n+ and δd n+ . The require∗ ∗ ment δφ+ = δφ− becomes N  ψ ∗n+ δρ ∗n+ = n=1 N  ψ ∗n− δρ ∗n− n=1 which cannot, in general, be satisfied exactly. However, this relation can be satisfied approximately. Multiplying by an arbitrary diagonal G × G matrix ωn (x, y) and integrating over the surface Sin yields one condition relating the Nδρ ∗n− to the Nδρ ∗n+ : N   n=1 Sin  N   dx dy ωn ψ ∗n+ δρ ∗n+ = n=1 Sin  p ∗n− dx dy ωn ψ ∗n− δp If this is repeated for N different matrix functions ωn the resulting set of equations may be written A + δρ ∗+ = A − δρ ∗− where A + is an NG × NG matrix and δρ + is an NG × 1 column matrix. This equation may be solved: A+ )−1A − δρ ∗− ≡ Q δρ ∗− δρ ∗+ = (A The NG × NG matrix Q may be partitioned into N 2 diagonal G × G matrices Q n ,n in terms of which the equation above may be written δρ ∗n+ = N  n =1 δdd ∗n+ = N  n =1 Qnn δρ n − , P nn δdd ∗n − , n = 1, . . . , N n = 1, . . . , N (15.157) (15.158) where P nn is one of the N 2 diagonal G × G matrices analogous to the Q nn . 587 588 15 Nodal and Synthesis Methods Making use of Eqs. (15.157) and (15.158), Eqs. (15.155) and (15.156) each yield N interface conditions. These interface conditions have several advantages relative to those described previously. The theoretical derivation is consistent and the problem of overdetermination never arises. Flux and current trial function discontinuities are allowed at the same interface. Unfortunately, these interface conditions are not, in general, symmetric with respect to the arbitrary choice of the + and − directions. When the physical configuration of the reactor changes significantly during a transient, it may be plausible to use different sets of expansion functions over different intervals of time. For this situation a term of the form J3 would be included in the variational principle for each time interface at which the set of expansion functions are changed. Requiring stationarity of the variational principle again results in Eqs. (15.150) and (15.151) within each time interval, with the coefficients defined in terms of the expansion functions appropriate to the time interval. Similarly, the boundary conditions of Eq. (15.152) and the spatial interface conditions of Eqs. (15.155) and (15.156), with the coefficients defined in terms of the expansion functions appropriate to the time interval, and the initial conditions of Eqs. (15.153) and (15.154) are obtained again. In addition, temporal interface conditions arise from the J3 term, the variation of which is  N  n =1 V × ∗T ∗T ∗T drr aδρ ∗T n (+)ψ n (+) + (1 − a)δρ n (−)ψ n (−) τ N  ψ n (+)ρ n (+) − ψ n (−)ρ n (−) n=1 + πT M   m=1 V ∗ ∗T ∗T ∗T drr bδCm,n  (+)ψ n (+) + (1 − b)δCm,n (−)ψ n (−) βm T  F ψ n (+)π Cm,n (+) − ψ n (−)π Cm,n (−) λm N × χm * = 0 (15.159) n=1 The same type of overdetermination, for the same reason [ failure to impose the ∗ (+) = δC ∗ (−)], has arisen in connection restrictions δφ ∗ (+) = δφ ∗ (−) and δCm m with the temporal interface conditions. If each variation δρ ∗n (+) and δρ ∗n (−) is (incorrectly) assumed to be independent, the 2N conditions are obtained relat∗ (+) = δC ∗ (−). The ing the Nρn (+) to the Nρn (−), and similarly for the δCm,n m,n stratagems that have been used to remove this apparent overdetermination parallel those employed for the spatial interface problem. Staggered discontinuities, in which the direct and adjoint flux (and precursor densities) are discontinuous at alternative times have been suggested, and the inconvenience of changing expansion functions at different times has been essentially eliminated by allowing alternative times to approach each other arbitrarily closely. As in the case of the spatial interface, the overdetermination never arises if the restrictions δφ ∗ (+) = δφ ∗ (−) and ∗ (+) = δC ∗ (−) are imposed in an approximate manner. δCm m 15.10 Multichannel Spatial Synthesis Adjoint synthesis equations may be derived by an analogous development, in this case by requiring stationarity of the functional with respect to the direct expansion coefficients. Similar results are obtained except that final conditions, rather than initial conditions, are obtained. It is necessary to impose approximately the conditions δφ + = δφ − , δjj + = δjj − to obtain restrictions on the variations in the flux and current expansion coefficients at spatial interfaces, and to impose approximately the conditions δφ(+) = δφ(−) and δCm (+) = δCm (−) at temporal interfaces, to avoid an overdetermination difficulty. 15.10 Multichannel Spatial Synthesis In Section 15.9 the idea of using different trial functions (i.e., different sets of expansion functions) in different axial regions was discussed. It is also possible to use different trial functions in different planar regions (channels), a procedure referred to as multichannel synthesis. This introduces two attractive possibilities. With expansion functions obtained from two-dimensional (x–y) calculations based on a model encompassing the entire cross-sectional area of the reactor, the multichannel feature provides the additional flexibility of allowing different expansion coefficients to be used in different channels. Thus a greater range of planar flux shapes can be synthesized from a given set of expansion functions than is possible with the single-channel synthesis of Section 15.9. A second possibility, which has not been exploited, is the use of expansion functions in each channel obtained from two-dimensional (x–y) calculations based on a model encompassing only the cross-sectional area of the channel. The basic idea of multichannel synthesis can be illustrated by the simple example shown in Fig. 15.10. Let the radial dimension of the reactor model be divided into two channels, 0 ≤ r ≤ a/2 and a/2 ≤ r ≤ a. Then the flux would be constructed by separately mixing φrod and φunrod in each channel: a 2 1 (z)φ (r) + a 1 φ(r, z) = arod rod unrod (z)φunrod (r), 0≤r ≤ 2 (z)φ (r) + a 2 φ(r, z) = arod rod unrod (z)φunrod (r), a ≤r ≤a 2 The multichannel synthesis equations are derived by using separate trial functions for each channel, denoted by a superscript c, of the form φ c (x, y, z, t) = N  ψnc (x, y)ρnc (z, t) n=1 jc (x, y, z, t) = N  + n=1 c (x, y)bnc (z, t) + ψnc (x, y)Bnc (z, t) i Jnx (15.160) 589 590 15 Nodal and Synthesis Methods c + Jny (x, y)gnc (z, t) + ψnc (x, y)Gcn (z, t) j c (x, y)dnc (z, t)k + Jnz , (15.161) with similar expansions for the adjoint flux and current. The x and y components of the current are each expanded in two separate types of terms in anticipation of the frequent procedure of using Jnx = −D(∂ψn , ∂x), and so on. The second term, proportional to ψn , is included both for added flexibility and to ensure the existence of coupling between channels across an interface located where ∂ψn /∂x may be zero. For the sake of illustration, the channel structure will be taken as concentric annuli, so that the interface terms J2 are included for each vertical cylindrical surface, Sin , that separates channels. Because the derivation of initial conditions and interface conditions for axial and temporal trial function discontinuities, and the inclusion of external boundary terms, are identical to the derivation given in the preceding section, they will be omitted. Thus the multichannel synthesis equations are obtained from consideration of the stationarity properties of J12 = J1 + J2 , with J2 consisting of the terms discussed above. These equations may be written in matrix form as M c ρ c + T c ρ̇ c + cx b c + cx+b c+1 − cx−b c−1 + − c c−1 x−B + cy g c + cy+g c+1 − cy−g c−1 + − c c−1 y− G + Ac c c xB ∂dd c  βm F cm C cm − S c = 0 − ∂z + c c+1 x+B c yG + c c+1 y+G M (15.162) m=1 δJ12 =0 T δ(bb c∗ n ) ⇒  cx b c + L cx B c + K cx ρ c + K cx+ ρ c+1 − K cx− ρ c−1 = 0 δJ12 =0 T δ(gg c∗ n ) ⇒  cy g c + L cy G c + K cy ρ c + K cy+ ρ c+1 − K cy− ρ c−1 = 0 δJ12 =0 T B c∗ δ(B n ) ⇒ h cx b c + H cx B c + W cx ρ c + W cx+ ρ c+1 − W cx− ρ c−1 = 0 δJ12 =0 T Gc∗ δ(G n ) ⇒ hcy g c + H cy Gc + W cy ρ c + W cy+ ρ c+1 − W cy− ρ c−1 = 0 (15.163) (15.164) (15.165) (15.166) 15.11 Spectral Synthesis δJ12 =0 T δ(dd c∗ n ) ⇒ U cd c + V c ∂ρ c ∂z =0 (15.167) The matrices and column matrices in Eqs. (15.162) to (15.167) are of order NG × G and NG × 1, respectively, except for F cm and C cm , which are NG × N and N × 1, respectively. Equations (15.162) to (15.167) may be combined to eliminate the current expansion coefficients, resulting in the set of one-dimensional (z) time-dependent matrix differential equations M c ρ c + T c ρ̇ c + R c ρ c + R c+ ρ c+1 + R c++ ρ c+2 + R c− ρ c−1 + R c− − ρ c−2  M c  ∂ c −1 c ∂ρ U ) V (U − −A βmF cmC cm − S c = 0 ∂z ∂z c m=1 c = 1, . . . , number of channels (15.168) The matrices R + and R ++ in Eqs. (15.168) result from elimination of the currentcombining coefficients and serve to couple channel c radially to channels c + 1 and c + 2, and similarly, R − and R − − couple channel c to channels c − 1 and c − 2. This general feature of nearest-neighbor and next-nearest-neighbor coupling is characteristic of the multichannel formulation, independent of the particular choice of channel structure. Construction of the radial coupling matrices involves the evaluation of surface integrals containing normal derivative terms and considerable matrix inversion and matrix multiplication. The results are sensitive to the accuracy and consistency with which the surface integrals are evaluated, a fact that has hindered exploitation of the multichannel formalism. Moreover, the transport cross section is embedded in the matrices that lead to R + , and so on, and a change in this quantity requires that the matrix inversions and multiplications involved in the construction of R + , and so on, be repeated. These factors tend to mitigate the advantages of extra flexibility and increased accuracy inherent in the multichannel formulation. See Refs. 10 and 13 for a more detailed description. 15.11 Spectral Synthesis In previous sections the emphasis has been on synthesizing the spatial dependence of the neutron flux, and the approximations that were discussed have, in fact, found their greatest application in problems where it was important, but uneconomical, to represent the detailed spatial variation of the flux. Another class of problems exists wherein it is important, but uneconomical, to represent the spectral variation of the flux in great detail. For such problems an attempt to synthesize the detailed spectrum from a few spectral functions is appealing. The general basis of the method is a trial function expansion in each spatial region, or channel c, of the 591 592 15 Nodal and Synthesis Methods form φ c (x, y, z, t) = N  ψ cn ρnc (x, y, z, t) (15.169) J cnb cn (x, y, z, t) (15.170) n=1 j c (x, y, z, t) = N  n=1 with similar expansions for the adjoint flux and current. Here ψ n and J n are known G × 1 column matrices, and a single expansion coefficient ρn applies to all the G group components of the corresponding expansion function ψn . Because the objective of the method is to approximate the spectral dependence, it is not necessary to make an expansion of the precursor trial function. Requiring stationarity of the variational principle with respect to the adjoint expansion coefficients yields the spectral synthesis equations within each channel c: δJ1 = 0, δ(ρnc∗ ) n = 1, . . . , N (15.171) δJ1 = 0, δ(bnc∗ ) n = 1, . . . , N (15.172) These two sets of N equations may be written as two matrix equations, and combined to eliminate the current combining coefficients, M   −1 βmF cm Cm − S c = 0 M c ρ c − c ∇ ·  c K c ∇ρ c + T c ρ̇ c − (15.173) m=1 M c , c ,  c , K c , and T c are N × N matrices, while ρ c , F cm and S c are N × 1 column matrices. For each spatial interface between channels a term of the form of J2 must be added to the variational principle. The variation of such a term leads to N  N  n =1 n=1 n· + ∗ ∗T J n+b n+ − J n−b n− ) γ δρ ∗n + ψ ∗T n + + (1 − γ )δρ n − ψ n − (J , ∗T b ∗T + ηδbb ∗n +J ∗T n + + (1 − η)δb n −J n − (ψ n+ ρn+ − ψ n− ρn− ) = 0 (15.174) As was the case with the spatial synthesis, it is necessary to impose some form of ∗ and δρ ∗ , and δb∗ and δb∗ , or restriction among the allowable variations δρn+ n− n+ n− to resort to some stratagem such as staggering the interfaces at which flux and current may be discontinuous or to set γ , η = 1, 0; otherwise, the interface conditions appear to be overdetermined. A restriction among the variations arises naturally from the requirement that variations δφ ∗+ and δφ ∗− be equal to ensure continuity, 15.11 Spectral Synthesis and similarly, that δjj ∗+ = δjj ∗− . This requirement must be imposed in an approximate manner (unless N = G, in which case there is no advantage whatsoever to using spectral synthesis), and leads to N  ∗ = δρn+ n =1 Qnn δρn∗ − , δb∗n+ · n = N  Pnn δb∗n− · n n =1 ∗ and δb∗ , Eq. (15.174) can then be required Using these relations to eliminate δρn+ n+ ∗ and δb∗ , which yields to be satisfied for arbitrary and independent variations δρn− n− the proper number of interface conditions. The other strategies that have been suggested may be considered as special cases with γ , η = 0, 1 and/or Pnn = Qnn = δnn . Thus, in general, the spatial interface conditions may be written in the form N  n· + , A Tn J n+ bn+ − A Tn J n− bn− = 0, n = 1, . . . , N (15.175) n = 1, . . . , N (15.176) n=1 N  + , B Tn ψ n+ ρn+ − B Tn ψ n− ρn− = 0, n=1 Note that unless A Tn J n+ = A Tn J n− , n = 1, . . . , N (15.177) B Tn ψ n+ = B Tn ψ n− , n = 1, . . . , N (15.178) Eqs. (15.175) and (15.176) do not reduce to continuity requirements of the form n · (bn+ − bn− ) = 0 ρn+ − ρn = 0 The conventional few-group approximation is a special case of the spectral synthesis approximation in which an expansion function ψ n or J n has nonzero elements only for those groups that are to be collapsed into few-group n. Thus the result above indicates that continuity of few-group flux and normal current is generally not the proper interface condition, obtaining only under the special circumstances whereby Eqs. (15.177) and (15.178) are satisfied. If different sets of spectral expansion functions are used in different time intervals, it is necessary to include terms of the form J3 , the stationarity of which yield temporal continuity conditions on the flux expansion coefficients at the time when the expansion functions are changed. To avoid an apparent overdetermination of continuity conditions, it is necessary either to resort to some stratagem such as requiring that the adjoint and direct flux expansion functions change at different times or setting a = 0, 1, or to impose in an approximate fashion the continuity condition δφ ∗ (+) = δφ ∗ (−) to relate δρn∗ (+) and δρn∗ (−). 593 594 15 Nodal and Synthesis Methods The continuity conditions resulting from this or other derivations can be written in the form N  + , D Tn ψ n (+) ρn (+) − D Tn ψ n (−) ρn (−) , n = 1, . . . , N (15.179) n=1 where D n is an N × 1 column matrix. Thus, in general, ρn (+) = ρn (−) is not the continuity condition. Consequently, recalling that a few-group approximation is a special case of the spectral synthesis approximation, continuity of few-group fluxes at times when the expansion functions (within-group fine-group fluxes) changes is generally not the proper continuity condition and obtains only when D Tn ψ n (+) = D Tn ψ n (−), n = 1, . . . , N The synthesis approximations lack, in general, the positivity properties associated with the multigroup diffusion equations. A consequence of this is that there is no a priori assurance that the fundamental eigenvalue (the one associated with an everywhere nonnegative flux solution) is larger in absolute value than any of the harmonic eigenvalues. Most numerical iteration schemes used in the solution of the synthesis equations converge to the eigenvalue, and corresponding eigenfunction, with the largest magnitude, but it is possible that a calculation will not converge to the fundamental solution. References 1 J. A. Favorite and W. M. Stacey, “A Variational Synthesis Nodal DiscreteOrdinates Method,” Nucl. Sci. Eng. 132, 181 (1999). 2 T. M. Sutton and B. N. Aviles, “Diffusion Theory Methods for Spatial Kinetics Calculations,” Prog. Nucl. Energy 30, 119 (1996). 3 R. T. Ackroyd et al., “Foundations of Finite Element Applications to Neutron Transport,” Prog. Nucl. Energy 29, 43 (1995); “Some Recent Developments in Finite Element Methods for Neutron Transport,” Adv. Nucl. Sci. Technol. 19, 381 (1987). 4 R. D. Lawrence, “Progress in Nodal Methods for the Solution of the Neutron Diffusion and Transport Equations,” Prog. Nucl. Energy 17, 271 (1986); “Three-Dimensional Nodal Diffusion and Transport Methods for the Analysis of Fast-Reactor Critical Experiments,” Prog. Nucl. Energy 18, 101 (1986). 5 J. J. Stamm’ler and M. J. Abbate, Methods of Steady-State Reactor Physics in Nuclear Design, Academic Press, London (1983), Chap. XI. 6 N. K. Gupta, “Nodal Methods for Three-Dimensional Simulators,” Prog. Nucl. Energy 7, 127 (1981). 7 J. J. Dorning, “Modern Coarse-Mesh Methods: A Development of the 70’s,” Proc. Conf. Computational Methods in Nuclear Engineering, Williamsburg, VA, American Nuclear Society, La Grange Park, IL (1979), p. 3-1. 8 M. R. Wagner, “Current Trends in Multidimensional Static Reactor Calculations,” Proc. Conf. Computational Methods in Nuclear Engineering, Charleston, SC, CONF-750413, American Nuclear Society, La Grange Park, IL (1975), p. I-1. Problems 9 A. F. Henry, Nuclear-Reactor Analysis, MIT Press, Cambridge, MA (1975), Chap. 11; “Refinements in Accuracy of Coarse-Mesh Finite-Difference Solutions of the Group-Diffusion Equations,” Proc. Semin. Numerical Reactor Calculations, International Atomic Energy Agency, Vienna (1972), p. 447. 10 W. M. Stacey, “Flux Synthesis Methods in Reactor Physics,” Reactor Technol. 15, 210 (1972); “Variational Flux Synthesis Methods for Multigroup Neutron Diffusion Theory,” Nucl. Sci. Eng. 47, 449 (1972); “Variational Flux Synthesis Approximations,” Proc. IAEA Semin. Numerical Reactor Calculations, International Atomic Energy Agency, Vienna (1972), p. 561; Variational Methods in Nuclear Reactor Physics, Academic Press, New York (1974), Chap. 4. 11 R. Froehlich, “A Theoretical Foundation for Coarse Mesh Variational Techniques,” Proc. Int. Conf. Research on Reactor Utilization and Reactor Computation, Mexico, D. F., CNM-R-2 (1967), p. 219. 12 S. Kaplan, “Synthesis Methods in Reactor Analysis,” Adv. Nucl. Sci. Technol. 3 (1966); “Some New Methods of Flux Synthesis,” Nucl. Sci. Eng. 13, 22 (1962). 13 E. L. Wachspress et al., “Multichannel Flux Synthesis,” Nucl. Sci. Eng. 12, 381 (1962); “Variational Synthesis with Discontinuous Trial Functions,” Proc. Conf. Applications of Computational Methods to Reactor Problems, USAEC report ANL-7050, Argonne National Laboratory, Argonne, IL (1965), p. 191; “Variational Multichannel Synthesis with Discontinuous Trial Functions,” USAEC report KAPL-3095, Knolls Atomic Power Laboratory, Schenectady, NY (1965). Problems 15.1 Derive the nodal fission rate balance equations of Eq. (15.17) from the nodal flux balance equations of Eq. (15.16). 15.2 Use the rational approximation for the escape probability to calculate the coupling terms W n,n+1 for cubic nodes. 15.3 Consider a slab reactor consisting of two core regions each 50 cm thick described by the parameters given for core 1 and core 2 in Table 15.1, with zero flux conditions on both external boundaries. Solve for the exact solution in two-group diffusion theory. 15.4 Construct a two-node conventional nodal model for the slab reactor of Problem 15.1. Solve for the multiplication constant and compare with the exact result of Problem 15.3. 15.5 Derive the transverse integrated nodal diffusion equations given by Eq. (15.31) and similar equations in the y- and z-directions. 15.6 Construct a two-node transverse integrated model for the slab reactor of Problem 15.3. Solve for the multiplication constant and compare with the exact result. 15.7 Derive the elements of the matrices P ng and R ng in the interface current balance of Eq. (15.58) for nodal diffusion theory. 595 596 15 Nodal and Synthesis Methods Table 15.1 Group Constant χ νf (cm−1 ) a (cm−1 ) s1→2 (cm−1 ) D (cm) Core 1 Core 2 Group 1 Group 2 Group 1 1.0 0.0085 0.0121 0.0241 1.267 0.0 0.1851 0.121 – 0.354 1.0 0.006 0.010 0.016 1.280 Group 2 0.0 0.150 0.100 – 0.400 15.8 Derive the nodal balance Eqs. (15.44) directly by integrating the transport equation (15.1) for each group over the node. 15.9 Derive the elements of the matrices P ng and R ng in the interface current balance of Eq. (15.86) for nodal GP1 transport theory. 15.10 Construct a two-coarse-mesh finite-element model for the slab reactor of Problem 15.3. Solve for the multiplication constant and compare with the exact result. 15.11 Prove that the two forms of the variational functional Fd of Eq. (15.107) are equivalent in that the stationarity of both forms with respect to arbitrary and independent variations requires that the diffusion equation is satisfied within the volumes Vi and that the diffusion theory current is continuous across the surfaces separating adjacent volumes. 15.12 Derive a finite-element coarse-mesh approximation, based on a quadratic polynomial expansion, for the one-dimensional one-group diffusion equation. 15.13 Carry through the derivation to prove that stationarity of the variational functional of Eq. (15.143) with respect to ∗ arbitrary and independent variations in φ ∗ , j ∗ , and Cm requires that the stationary functions φ, j , and Cm satisfy the time-dependent transport equation, Fick’s law relation, and precursor balance equation. ∗ by 15.14 Derive the time-dependent equations for φ ∗ , j ∗ , and Cm requiring stationarity of the variational function of Eq. (15.143) with respect arbitrary and independent variations in φ, j , and Cm . 15.15 Construct a single-channel synthesis model for the slab reactor of Problem 15.3, but in one-group diffusion theory. Obtain the one-group constants by using the two-group constants of Problem 15.3 in an infinite-medium spectrum calculation for φ1 and φ2 , which can be used to construct effective one-group cross sections. Using the trial function φ(x) = a cos(πx/100) for the flux and adjoint flux, calculate Problems the multiplication constant and compare with the exact result of Problem 15.3. 15.15 Repeat Problem 15.15 using a two-channel synthesis model. 597 599 16 Space–Time Neutron Kinetics The discussion of reactor dynamics in Chapter 5 was based on the implicit assumption that the spatial neutron distribution remained fixed and only the total neutron population changed in time. However, when a critical reactor is perturbed locally, the spatial neutron flux distribution, as well as the total neutron population, will change, and the change in the spatial flux distribution will affect the change in the total neutron population. A very local perturbation (e.g., the withdrawal of a control rod) will obviously affect the neutron flux in the immediate vicinity of the perturbation. However, a local or regional perturbation can also affect the global neutron flux distribution (i.e., produce a flux tilt), which will, in turn, alter the reactivity and affect the global neutron population. Moreover, for a transient below prompt critical, the largest part of the neutron source is due to the decay of delayed neutron precursors, which tends to hold back a flux tilt until the delayed neutron precursor distribution also tilts. The point kinetics equations discussed in Chapter 5 can be extended to treat flux tilts and delayed neutron holdback effects by recomputing the point kinetic parameters during the course of a transient. The various methods that have been discussed for calculating the spatial distribution of the neutron flux can also be extended to calculate the space- and time-varying neutron flux distribution by adding neutron density time derivative and delayed neutron precursor source terms and appending a set of equations to calculate local delayed neutron precursor densities. The methods of stability analysis and control can also be extended to include spatial dependence, as illustrated by an analysis of xenon spatial oscillations. 16.1 Flux Tilts and Delayed Neutron Holdback Physical insight into the flux tilting and delayed neutron holdback phenomena can be obtained by considering a step local perturbation in the material composition of an initially critical reactor. In multigroup diffusion theory, the initial critical state of the reactor is described by 600 16 Space–Time Neutron Kinetics g −∇ · D g (r, t)∇φg (r, t) + t (r, t)φg (r, t) − G  = χg g  =1 G    g →g (r, t)φg  (r, t) g  =1 g νf (r, t), g = 1, . . . , G (16.1) which will be written in operator notation as A0 φ0 = M0 φ0 (16.2) where the zero subscript is used to indicate the initial critical state. Now we consider a spatially nonuniform change in materials properties which is represented by the changes A in the destruction operator and M in the fission operator so that A0 → A = A0 + A and M0 → M = M0 + M. For changes producing reactivities well below prompt critical, the prompt jump approximation may be used to describe the neutron kinetics. Making the further approximation of a single delayed neutron precursor group, the neutron kinetics is described by   0 = −A + {1 − β)M φ + λC (16.3) Ċ = βMφ − λC (16.4) Expanding about the initial critical distributions φ(r, t) = φ0 (r) + φ(r, t) C(r, t) = C0 (r) + C(r, t) ≡ (16.5) βM0 φ0 + C λ (16.6) linearizing (i.e., ignoring quadratic terms M φ, etc.), Laplace transforming, and combining the two equations results in an equation for the time dependence of the neutron flux φ in the frequency domain:         1 sβ sβ M0 φ̃(r, s) + −A + 1 − M φ0 0 = −A0 + 1 − s +λ s s +λ (16.7) Modal Eigenfunction Expansion We now expand the time-dependent flux,  an (t)ψn (r) φ(r, t) = (16.8) n=0 where the ψn are the spatial eigenfunctions of the initial critical reactor and satisfy A0 ψn = 1 M0 ψn kn (16.9) [e.g., in a uniform slab reactor of width a, ψn = sin(nπx/a)]. The corresponding adjoint eigenfunctions of the initial critical reactor are defined by A∗0 ψn∗ = 1 ∗ ∗ M ψ kn 0 n (16.10) 16.1 Flux Tilts and Delayed Neutron Holdback From the definition of the adjoint operator discussed in Chapter 13, the orthogonality property 1 0 ∗ (16.11) ψm , M0 ψn = δmn the relationship 0 ∗ 1 1 10 ∗ ψm , A0 ψn = ψ , M0 , ψn kn m (16.12) can be established, where XX indicates integration over space and summation over groups. Using the eigenfunction expansion of Eq. (16.8) in Eq. (16.7), multiplying the ∗ , integrating over space and summing over groups, and resulting equation by ψm using Eqs. (16.11) and (16.12) yields ãm (s) = ∗ , (−A + (1 − (s + λ) ψm 1−km s[(s + λ)( 1−(1−β) k m sβ s+λ )M)φ0 m ∗,M ψ )]( 1−(1−β)k ) ψm 0 m km which may be inverse Laplace transformed to obtain    βkm ρm k m −λ(1 − km )t 1− exp am (t) = 1 − km 1 − (1 − β)km 1 − (1 − β)km   ∗ −λ(1 − km )t βkm ψm , Mφ0 exp − ∗,M ψ [1 − (1 − β)km ] ψm 1 − (1 − β)km 0 m (16.13) (16.14) where ρm ≡ ∗ (−A + M)φ ψm  0 ∗,M ψ ψm 0 m (16.15) is the mth-mode reactivity. Flux Tilts If ρm = 0, a nonuniform perturbation in materials properties in a critical reactor will introduce higher harmonic eigenfunctions into the flux distribution, which becomes after the transient terms in Eq. (16.14) have died out φ(r, ∞) = 1 + a0 (∞) φ0 (r) +  ρn kn ψn (r) 1 − kn (16.16) n=1 For a uniform slab reactor in 1 12 -group diffusion theory, the results of Chapter 3 can be used to write the nth-mode eigenvalue: kn = k∞ k∞ 1 + M 2 (π/a)2 = = 2 2 2 2 1 + M Bn 1 + M [(n + 1)π/a] 1 + M 2 [(n + 1)π/a]2 (16.17) where M 2 is now the migration area and we have taken advantage of the fact that k0 = 1 to write the last form of the equation. 601 602 16 Space–Time Neutron Kinetics The amplitude of the first harmonic eigenfunction, which would be the main component of a flux tilt, depends on the magnitude of the first harmonic reactivity, ρ1 , and on the first harmonic eigenvalue separation, 1 − k1 (note that k0 = 1). Using Eq. (16.17), the 1 12 -group diffusion theory estimate for the first harmonic eigenvalue separation of a uniform slab reactor is   3(Mπ/a)2 Mπ 2 1 − k1 = ≈3 (16.18) a 1 + (2Mπ/a)2 Thus reactors that are very large in units of migration length (a/M  1) will have a small first harmonic eigenvalue separation and will be very “tilty.” Delayed Neutron Holdback As indicated by Eq. (16.14), a tilt will not occur instantaneously upon the introduction of a nonuniform step change in materials properties into a critical reactor, but will gradually build in over a time t ≈ 2 to 3τtilt , where τtilt = 1 − (1 − β)k1 > λ−1 λ(1 − k1 ) (16.19) Physically, the prompt neutrons respond essentially instantaneously (on the neutron lifetime scale) to the change in materials properties, but the delayed neutron source only gradually changes from the initial fundamental mode distribution into the asymptotic distribution. 16.2 Spatially Dependent Point Kinetics The multigroup diffusion theory approximation for the space and time dependence of the neutron flux within a nuclear reactor is described by the set of G equations 1 ∂φ g (r, t) g = ∇ · D g (r, t)∇φ g (r, t) − t (r, t)φ g (r, t) vg ∂t G     g →g (r, t)φ g (r, t) + g  =1 g + λ0 (1 − β)χp G  g  =1 + M  g g  νf (r, t)φ g (r, t) λm χm Cm (r, t), g = 1, . . . , G (16.20a) m=1 which for notational convenience we shall write in operator notation as 1 ∂φ(r, t) = −A(r, t)φ(r, t) + λ0 (1 − β)Fp (r, t)φ(r, t) v ∂t M  λm Cm (r, t) + m=1 (16.20b) 16.2 Spatially Dependent Point Kinetics The space and time dependence of the M groups of delayed neutron precursors are described by  g ∂Cm (r, t) = βm νf (r, t)φ g (r, t) − λm Cm (r, t) ∂t G (16.21a) g=1 which in operator notation becomes ∂Cm (r, t) = λ0 βm F (r, t)φ(r, t) − λm Cm (r, t), ∂t m = 1, . . . , M (16.21b) where A, F = loss and production operator, respectively φ(r, t) = neutron flux Cm (r, t) = precursor density of type m v = neutron speed χm , λm , βm = fission spectrum, decay constant, and delayed neutron fraction, respectively, for precursor type m Fp ≡ χp F = fission source for prompt neutrons (χp is the fission spectrum for prompt neutrons; Fm ≡ χm F will be the fission source for delayed neutrons from precursor group m in subsequent equations) λ0 = eigenvalue adjusted to render the system critical at time t = 0. In the multigroup form of Eqs. (16.20a) and (16.21a), φ(r, t) represents a column vector of group fluxes, and A and F are matrices. For the initial, static configuration, these equations reduce to (A0 − λ0 F0 )φ0 = 0 (16.22) For the perturbed static configuration (i.e., after the delayed neutrons reach equilibrium), these equations reduce to (Ae − λe Fe )φe = 0 (16.23) and the quantity −λ = λ0 − λe ≡ (ke − k0 )/ke k0 is called the static reactivity worth of the perturbation (k = λ−1 ). [Note that because Eq. (16.23) is an eigenvalue problem, the word static here refers only to the flux distribution, not the amplitude.] The static reactivity worth of the perturbation is ρe ≡ −λ = φ0∗ , (λ0 F − A)φe φ0∗ , F φe where the static flux adjoint function φ0∗ satisfies  ∗  A0 − λ0 F0∗ φ0∗ = 0 (16.24) (16.25) (The inner product notation , indicates an integration over volume and a sum over energy groups.) 603 604 16 Space–Time Neutron Kinetics Derivation of Point Kinetics Equations The exact space–time equations are reduced to the point reactor kinetics model by writing the flux as a product of a shape function and an amplitude function; that is, φ(r, t) = ψ(r, t)n(t) (16.26) The point kinetics equations are derived by weighting Eqs. (16.20) and (16.21) with the static adjoint flux and integrating over volume and summing over energy:  ρ(t) − β̄(t) n(t) + λm Pm (t) (t) M ṅ(t) = (16.27) m=1 Ṗm (t) = βm γm (t) n(t) − λm Pm (t), (t) m = 1, . . . , M (16.28) where the dynamic reactivity, prompt neutron generation time, and delayed neutron effectiveness are defined as φ0∗ , (λ0 F − A)ψ(r, t) φ0∗ , F ψ(r, t)  *  G G    g  g g∗ g φ0 (r)χp λ0  νf (r, t) ψ (r, t) = dr ρ(t)= g  =1 g=1  g − ∇ · D g (r, t)∇ψ g (r, t) + t (r, t)ψ g (r, t) *' G  g  →g g ∇ (r, t)ψ (r, t) − g  =1  dr G  g=1 g∗ g φ0 (r)χp G  g  =1 g  νf (r, t)ψ g (r, t) φ0∗ , F ψ(r, t) φ0∗ , v−1 ψ(r, t)  g∗ g G g g dr G g=1 φ0 (r)χp g  =1 νf (r, t)ψ (r, t) =  g∗ g g dr G g=1 φ0 (r)(1/v )ψ (r, t) (16.29) −1 (t) = (16.30) and φ ∗ , Fm ψ(r, t) γm (t) = 0∗ = φ0 , F ψ(r, t) g∗ g G g g g=1 φ0 (r)χm g  =1 νf (r, t)ψ (r, t)  g∗ g G g g dr G g=1 φ0 (r)χp g=1 νf (r, t)ψ (r, t) dr G (16.31) 16.2 Spatially Dependent Point Kinetics respectively (β̄ = γ1 β1 + · · · + γM βM , A = A − A0 , and F = F − F0 ). In principle, the point kinetics equations can be used to calculate the exact space–time neutron flux, if the correct spatial flux shape is used at all times to evaluate the parameters defined by Eqs. (16.29) to (16.31). Note that these parameters do not depend on the amplitude of the flux, only the flux distribution. In a large LWR core, the flux is slow to reach equilibrium in its perturbed static distribution, due to the holdback effect of the delayed neutrons. Thus, for the first few seconds after a perturbation, the time-dependent flux shape ψ(r, t) differs from the static perturbed flux shape φe , and the dynamic reactivity of Eq. (16.29) differs from the static reactivity of Eq. (16.24). In the standard implementation of the point kinetics method, the parameters are estimated using the initial static flux distribution φ0 . This approximation corresponds to first-order perturbation theory, and for the reactivity, it is denoted ρ0 = φ0∗ , (λ0 F − A)φ0 φ0∗ , F φ0 (16.32) This expression can be shown (Chapter 13) to be a first-order approximation of the static reactivity [i.e., ρ0 is an estimate of the difference (−λ = λ0 − λe ) of reciprocal eigenvalues for the initial and perturbed core static configurations that is accurate to first order in the flux perturbation φ = φe − φ0 (i.e., error ∼ φ)]. Adiabatic and Quasistatic Methods If parameters of Eqs. (16.29) to (16.31) calculated with the initial spatial flux shape are used throughout the transient calculation, the result is the standard point kinetics approximation of Chapter 5. If the parameters are recomputed at selected times during the transient, using a static neutron flux solution corresponding to the instantaneous conditions of the reactor, the result is an improvement to the standard point kinetics known as the adiabatic method. In the quasistatic (QS) method, the point-kinetics equations are used for the flux amplitude, but the flux shape is recomputed (at time steps t = tn ) using  A − λ0 (1 − β)Fp + =   1 ṅ 1 Sn + v n tn tn M 1 1  Sn−1 + χm λm Cm (r, tn ) vtn n(tn ) (16.33) m=1 where tn = tn − tn−1 is the shape time step. (The precursor density is computed directly from the flux history.) When the flux shape from the nth such recalculation Sn is used directly to estimate the reactivity using the inner-product definition [Eq. (16.29)], the result is ρn (t) ≡ φ0∗ , (λ0 F − A)Sn φ0∗ , F Sn (16.34) 605 606 16 Space–Time Neutron Kinetics a potentially accurate estimate of the dynamic reactivity, depending, of course, on the accuracy of the flux calculation. It is more accurate to use a flux shape interpolated from the most recent known shape Sn−1 and the best guess for the next shape Sn , where  represents the most recent (the th) calculation of Sn using Eq. (16.33). (The Sn are considered converged when the last one satisfies a normalization constraint.) Regardless of the approximate shape that is used, ρn (t) is a first-order estimate of the static reactivity corresponding to the reactor conditions at time t , and it will be referred to as such. Variational Principle for Static Reactivity A variational estimate, accurate to second order [error ∼ (φ)2 ], for the static reactivity worth of a perturbation to an altered system (i.e., a system other than the one for which φ0 and φ0∗ were calculated) is ρv,e = 0 1 0 φ0∗ , (λ F − A)S × 1 − φ0∗ , (A − λF ) − φ0∗ , F  S where the generalized adjoint function  A∗0 − λ0 F0∗ and the function  ∗ = ∗ ∗   1 , A − λ F  S (16.35) is calculated using F ∗φ∗ (A∗ − λ0 F ∗ )φ0∗ − ∗0 0 ∗ φ0 , (A − λ0 F )φ0 φ0 , F0 φ0 (16.36) is calculated using (A0 − λ0 F0 ) = (A − λ0 F )φ0 F0 φ0 − ∗ φ0∗ , (A − λ0 F )φ0 φ0 , F0 φ0 (16.37) In Eq. (16.35), the unprimed operators and eigenvalue refer to the altered system at time tn and the primed operators and eigenvalue refer to the altered (by previous changes from the initial) system plus a perturbation (i.e., A = A − A and F = F  − F ) at time t > tn . The variational functional ρv,e provides an estimate of the static reactivity worth of the perturbation, −λ = λ − λ , in the altered system. The functional ρv,e is stationary about the altered and static perturbed altered adjoint and direct eigenvalue equations, respectively [as well as being stationary about the equations for ∗ and , for which Eqs. (16.36) and (16.37) are approximations]. When ρv,e is used to estimate the reactivity for the point-kinetics method without updating the flux shapes, the initial configuration described by Eqs. (16.22) and (16.25) is considered the altered system and φ0 is used for S. In this case, ρν,e provides a second-order estimate of the static reactivity of Eq. (16.24), rather than the dynamic reactivity of Eq. (16.29). In so doing, it ignores the delayed neutron holdback effect, an omission that leads to errors in reactivity estimates and consequent errors in power calculations. When ρv,e is used to estimate the reactivity for the QS method, the configuration at the time tn of the most recent shape calculation is considered the altered system, and the Sn is used for S. In this case, ρv,e provides an estimate accurate to second 16.2 Spatially Dependent Point Kinetics order of the static reactivity worth of perturbations made since time tn . This estimate ignores the delayed neutron holdback effect. The total reactivity worth of all perturbations (and alterations) is found by adding this perturbed reactivity worth in the altered system to the best available estimate of the dynamic reactivity worth of the alteration, which is ρn [Sn (r, tn )]. Because it is necessary to use the flux shape corresponding to the altered system, it is not appropriate to use the variational static reactivity estimate with interpolated flux shapes. Variational Principle for Dynamic Reactivity To account for the delayed neutron holdback effect on the reactivity, a variational principle should be stationary about the solutions of the time-dependent diffusion and precursor equations, rather than stationary about the solution of the perturbed static diffusion equation. To this end, the following functional was constructed: ρv ψ, ψ ∗ , ξm , ξm∗ , , = ∗ ψ ∗ , (λ0 F − A)ψ ψ ∗, F ψ  5 0 1 × 1 − ψ ∗ , (A0 − λ0 F0 ) −   6 1 ∂ , A − λ0 (1 − β)Fp + ψ v ∂t 3 2 M 2  3  M   ∂ ∗ ∗ ξm , λm χm ξm − ξm , λm + + ∂t 2 + m=1 M  3* ∗ m=1 ξm∗ , λ0 βm F ψ (16.38) m=1 The usual procedure is to require that the functional be stationary with respect to arbitrary and independent variations of the trial functions over all the independent variables. However, in order to retain the time dependence of the dynamic reactivity, the integrals in ρv (indicated by , ), are only over space and energy, not time. Thus, the stationarity conditions for the functional are established by requiring that it be stationary with respect to arbitrary and independent variations of only the space and energy dependencies of the functions ∗ , ξm∗ , , ψ ∗ , ψ , and ξm . The following equations result: Aψs − λ0 (1 − β)Fp ψs − M  λm χm ξm,s + m=1 1 ∂ψs =0 v ∂t (16.39) ∂ξm,s = λ0 βm F ψs − λm ξm,s ∂t (16.40) A∗0 ψs∗ − λ0 F0∗ ψs∗ = 0 (16.41) (A0 − λ0 F0 ) s = (A − λ0 F )ψs ∗ ψs , (A − λ0 F )ψs − F ψs ∗ ψs , F ψs (16.42) 607 608 16 Space–Time Neutron Kinetics A∗ ∗ s − λ0 (1 − β)Fp∗ ∗ s − M  ∗ λ0 βm F ξm,s m=1 = (A∗ − λ0 F ∗ )ψs∗ F ∗ψ ∗ − ∗ s ∗ ψs , (A − λ0 F )ψs ψs , F ψs (16.43) and ∗ − λχmT λm ξm,s ∗ s =0 (16.44) respectively. Comparing Eqs. (16.39) and (16.20), Eqs. (16.40) and (16.21), and Eqs. (16.41) and (16.25), it is clear that ψs and ξm,s can be identified as the solutions φ(r, t) and Cm (r, t) of the exact time-dependent diffusion and precursor equations and that ψs∗ can be identified as the unperturbed static adjoint flux φ0∗ . The stationary value of ρv is ρv,s = φ0∗ , (λ0 F − A)φ(r, t) φ0∗ , F φ(r, t) (16.45) the exact, dynamic reactivity worth of a perturbation. To adapt the functional ρν for use with the QS method, we introduce as a trial function ψ(r, t) ≈ S(r, t)n(t) (16.46) and note that the best available approximation for the time derivative of the precursor density is ∂Cm (r, t) ≈ βm F S(r, t)n(t) − λm Cm (r, t) ∂t (16.47) Under these conditions [and noting that ψ ∗ = φ0∗ is available from Eq. (16.25)], the functional becomes  6 5 0 ∗ 1 φ0∗ , (λ0 F − A)S ∗ 1 ∂S 1 − G , [A − λ (1 − β)F ]S − G , ρv = 0 p φ0∗ , F S v ∂t 3* 2 5 6 M  1 1 ṅ G∗ , λm χm Cm − G∗ , S + n v n m=1 (16.48) where the quantity ∗ (r, t)n(t) has been replaced by a trial function G∗ (r, t). Note that A and F here refer to the total perturbation, not the perturbation since the most recent shape calculation, and that A = A0 + A, F = F0 + F . Using Eqs. (16.44) and (16.46) in Eq. (16.43) results in the following equation for G∗ (r, t):   A∗ − λ0 F ∗ G∗ (r, t) = (A∗ − λ0 F ∗ )φ0∗ F ∗ φ0∗ − φ0∗ , (A − λ0 F )S(r, t) φ0∗ , F S(r, t) (16.49) 16.3 Time Integration of the Spatial Neutron Flux Distribution It is computationally economical to compute the generalized adjoint function G∗ only once for a particular core configuration. In this case, the initial static configuration is used, resulting in the following approximation:   ∗ (A∗ − λ0 F ∗ )φ0∗ F ∗φ∗ − ∗0 0 A0 − λ0 F0∗ G∗ = ∗ φ0 , (A − λ0 F )S0 φ0 , F0 S0 (16.50) (any magnitude perturbation A and/or F can be used since these operators appear in both the numerator and denominator of the same term). Thus G∗ (r) differs only in amplitude from ∗ (r) of Eq. (16.36). The form of the functional represented by Eq. (16.48) is well suited for use with the QS method. In the QS method, the point-kinetics equations are used for the flux amplitude n(t), the precursor concentration densities Cm (r, t) are updated at each time step and are therefore available for use in the variational estimate, and the flux shape S(r, t) is recomputed periodically using Eq. (16.33). The variational dynamic reactivity estimate can be used with or without flux shape interpolations. It should be noted that the G∗ of Eq. (16.50) satisfies the orthogonality condition 1 0 ∗ (16.51) G , F0 S0 = 0 As a consequence, when the initial flux shape S0 is used in ρν and if the precursor density functions Cm (r, t) have the same shape as S0 , the variational estimate for dynamic reactivity reduces to the variational estimate for static reactivity, ρν,e of Eq. (16.35) [in which the second term in the square brackets disappears because of Eq. (16.25)]. The effect of this reduction is that until the flux shape is recomputed or until some other approximation is made to replace S0 , the new variational functional still ignores the delayed neutron holdback effect. Numerical tests on a large LWR model indicate that the flux shape computational effort required with the QS method can be reduced by a factor of 3 to 4 by using the variational estimate of dynamic reactivity. In addition, use of a variational reactivity estimate rather than the standard first-order estimate of static reactivity can improve the accuracy of the QS method enough that the time-consuming flux shape interpolation/recomputation procedure may not be necessary. 16.3 Time Integration of the Spatial Neutron Flux Distribution The various methods that have been discussed for calculating the spatial neutron flux distribution (finite-difference, nodal, finite-element, synthesis, etc.) can be extended to calculate the space–time neutron flux distribution by adding a neutron density time derivative, distinguishing between prompt and delayed neutron sources in the neutron balance equation and appending equations to calculate the delayed neutron precursor densities [e.g., Eqs. (16.20) and (16.21)]. Writing the group fluxes and precursor densities at every spatial point (e.g., mesh point, node) as a column vector ψ , and writing the terms of the multigroup neutron and delayed neutron precursor balance equations at each spatial point as a matrix H , the 609 610 16 Space–Time Neutron Kinetics space–time neutron kinetics equations can be written as a coupled set of ordinary differential equations ψ H ψ = ψ̇ (16.52) Explicit Integration: Forward-Difference Method The simplest approximate solution to Eq. (16.52) is obtained by a simple forwarddifference algorithm, H (p)ψ(p) ψ(p + 1) = ψ(p) + tH (16.53) where the argument p denotes the value at time tp , and t = tp+1 − tp . In terms of the multigroup diffusion equations, this algorithm is  g g φ g (p + 1) = φ g (p) + tvg ∇ · D g (p)∇φ g (p) − a (p) + s (p) φ g (p) + G  g  →g s  (p)φ g (p) g  =1 g + (1 − β)χp G  g  =1 + M  g  νf (p)φ g (p) g λm χm Cm (p) * , g = 1, . . . , G (16.54) m=1 and for the precursors,  Cm (p + 1) = Cm (p) + t βm G   g νf (p)φ g (p) − λm Cm (p) , g=1 m = 1, . . . , M (16.55) where the spatial dependence is implicit. This algorithm suffers from a problem of numerical stability, which requires the use of such small time steps that the advantage offered by the simplicity of the algorithm is usually more than offset by the large number of time steps required. The nature of this problem is seen by considering an expansion of ψ(p) in the eigenfunctions of the operator H :  an Ω n (16.56) ψ(p) = n where H Ω n = ωn Ω n (16.57) 16.3 Time Integration of the Spatial Neutron Flux Distribution Substituting Eq. (16.56) into Eq. (16.53) yields ψ(p + 1) =  an (1 + ωn t)Ω n (16.58) n The condition for numerical stability is that the fundamental mode Ω 1 grow more rapidly than the harmonics Ω n , n ≥ 2. This requires that |1 + ω1 t| > |1 + ωn t|, n≥2 (16.59) To ensure this, |ωn t| must be much less than unity. The eigenvalue problem of Eq. (16.57) is a generalization to several groups and many spatial points of the in-hour equation of Section 5.3. The magnitude of the fundamental eigenvalue is on the order of the precursor decay constant, except for highly supercritical transients, in which case small time steps must be used in any case. Numerical studies g have shown that the smallest eigenvalues can be on the order of −(vg a ), which 4 7 can be about −10 for thermal neutrons and about −10 for fast neutrons. Thus t < 10−7 may be required for stability. When the time derivative terms for the epithermal groups are assumed to vanish (a useful approximation since 1/vG  1/vg , g = G), t < 10−4 may be required. Implicit Integration: Backward-Difference Method The numerical stability problem associated with the preceding method can be all but eliminated by the backward-difference algorithm: H (p + 1)]−1 ψ(p) ψ(p + 1) = [II − tH (16.60) In terms of the precursor and multigroup diffusion equations, this algorithm is Cm (p + 1) = G  Cm (p) βm g νf (p + 1)φ g (p + 1), + 1 + λm t 1 + λm t g=1 m = 1, . . . , M (16.61) g g ∇ · D g (p + 1)∇φ g (p + 1) + a (p + 1) + s (p + 1) φ g (p + 1) + G  g  →g s  g (p + 1)φ g (p + 1) + (1 − β)χp g  =1 + g  =1 g  νf (p + 1)φ g (p + 1) M G g  λm χm βm  1  g νf (p + 1)φ g (p + 1) − g φ g (p + 1) 1 + λm t  v t m=1 =− G  1 φ g (p) − vg t g =1 M  m=1 g λm χm Cm (p) , 1 + λm t g = 1, . . . , G (16.62) 611 612 16 Space–Time Neutron Kinetics An expansion of the type of Eq. (16.56) substituted into Eq. (16.60) yields  ψ(p + 1) = an (1 − tωn )−1 Ω n (16.63) n and the condition for stability is (1 − tω1 )−1 > (1 − tωn )−1 , n≥2 (16.64) The method is unconditionally stable if 0 > Re{ω1 } > Re{ωn }, n ≥ 2. For Re{ω1 } > 0, the stability requirement is determined by the requirement that ψ(p + 1) be a positive vector, which necessitates that t < 1 ω1 (16.65) This requirement is restrictive only for large ω1 that correspond to fast transients where small time steps would be necessary in any case. The difficulty with the backward-difference method arises from the necessity of inverting a matrix at each time step. The actual matrix that must be inverted is the coefficient matrix for the left side of Eq. (16.62); the delayed neutrons can be determined directly. Thus, although much larger time steps can be taken with the implicit method than with the explicit method, the computation time needed for the matrix inversions may more than offset this advantage. The size time step used in the backward-difference method is usually limited by the effect of truncation error (of order t 2 ) upon the accuracy of the solution rather than by numerical stability. Implicit Integration: θ Method For a constant H in the interval tp ≤ t ≤ tp+1 , Eq. (16.52) has the formal solution   t 2 2 H )ψ(p) = I + tH H+ H + · · · ψ(p) ψ(p + 1) = exp(tH 2! (16.66) The algorithms of Eqs. (16.53) and (16.60) may be considered as approximations to Eq. (16.66). An improved algorithm results from the prescription M ψ(p + 1) + (H H − M )ψ(p)] ψ(p + 1) − ψ(p) = t[M (16.67) with matrix elements of M and H related by mij = θij hij (16.68) where the mij , thus the θij , are chosen so that ψ(p + 1) calculated from Eq. (16.67) agrees with ψ(p + 1) calculated from Eq. (16.66). This requires that M= 1 H ) − I ]−1 I − H [exp(tH t (16.69) 16.3 Time Integration of the Spatial Neutron Flux Distribution Assuming that H has distinct eigenvalues, it may be diagonalized by the transformation  + T HJ = Γ (16.70) J where J and J + are the modal matrices corresponding to H and H T (i.e., the columns of J and J + are the eigenvectors of H and H T , respectively), and Γ is a diagonal matrix composed of the eigenvalues of H . Thus  + T 1 MJ = I − Γ [exp(tΓ ) − I ]−1 = L J t (16.71) with L diagonal. From this it follows that  T M =JL J + (16.72) and the factors θij can be determined from θij = mij hij after the mij are found from Eq. (16.72). Because solving for the θij rigorously would entail a great deal of effort, several approximations are made in employing this method to arrive at an algorithm for solutions of the multigroup kinetics equations. The delayed neutrons are treated as sources, and thus are neglected in the determination of the θij . An average spaceindependent value of θij is calculated based on a flux square weighting procedure. The delayed neutron precursors have a separate θij . Denoting the θij associated with groups g and g  as θgg  and θij associated with the delayed neutrons as θd , the following algorithm results: Cm (p + 1) = 1 − (1 − θd )λm t tβm Cm (p) + 1 + θd λm t 1 + θd λm t  G  g × νf (p + 1)φ g (p + 1)θ1g g=1 + G   g νf (p)φ g (p)(1 − θ1g ) , m = 1, . . . , M (16.73) g=1 , + g g θgg ∇ · D g (p + 1)∇φ g (p + 1) − a (p + 1) + s (p + 1) φ g (p + 1) + G  g  →g θgg  s  (p + 1)φ g (p + 1) g  =1 g + χp (1 − β) G  g  =1 g  θgg  νf (p + 1)φ g (p + 1) − 1 φ g (p + 1) tvg 613 614 16 Space–Time Neutron Kinetics + M G g  χm λm tβm θd   g νf (p + 1)φ g (p + 1)θ1g  1 + θd λm t  g =1 m=1 + g g = −(1 − θgg ) ∇ · D (p)∇φ g (p) − a (p) + s (p) φ g (p) − G  g g  →g (1 − θgg  )s ,  (p)φ g (p) g  =1 g − (1 − β)χp G  g  =1 g  χm λm Cm (p) 1 g φ (p) − tvg 1 + θd λm t M −  (1 − θgg  )νf (p)φ g (p) g m=1 − M G g  χm θd λm tβm   g νf (p)φ g (p)(1 − θ1g  ) 1 + θd λm t  g =1 m=1 g = 1, . . . , G (16.74) In the limit θgg  , θd → 1 Eqs. (16.73) and (16.74) reduce to the backwarddifference algorithms of Eqs. (16.61) and (16.62), while Eqs. (16.73) and (16.74) reduce to the forward-difference algorithms of Eqs. (16.54) and (16.55) in the limit θgg  , θd → 0. As mentioned, a number of approximations are made in arriving at Eqs. (16.73) and (16.74), so the mathematical properties associated with Eqs. (16.67) to (16.72) are not rigorously retained by Eqs. (16.73) and (16.74). Insight into the stability properties of the θ -method can be gained by considering the situation for a constant matrix H and a constant time step t . Expanding the exact solutions of Eq. (16.52) in the eigenfunctions Ω n , of H given by Eq. (16.57), ψ(tp ) = N  an Ω n eωn tp = n=1 N  an Ω n eωn (pt) (16.75) n=1 where the expansion coefficients an are determined from the initial conditions and where ω1 > ω2 > · · · > ωN . For the same eigenfunctions to satisfy Eq. (16.67), which becomes M )−1 (II + tH H − tM M )Ω n γn Ω n = (II − tM (16.76) the eigenvalues must be related by γn = 1 + (1 − θ)ωn t 1 − θωn t (16.77) 16.3 Time Integration of the Spatial Neutron Flux Distribution The general solution for the θ -approximation of Eq. (16.67) may be written ψ(tp ) = N  p an γn Ω n (16.78) n=1 where tp = p t . Comparison with the exact solution of Eq. (16.75) indicates that p exp(ωn t) = exp(ωn pt) has been replaced by γn in the approximate solution. For p a stable θ approximation, γn > −1; otherwise, γn will oscillate and diverge as time increases. Thus, Eq. (16.77) and the eigenvalues ωn can be used to determine a maximum stable step size t . Numerical experience indicates that the algorithm of Eqs. (16.73) and (16.74) is (1) numerically stable for time steps two orders of magnitude greater than are required for stability of Eqs. (16.54) and (16.55), and (2) somewhat more accurate than the algorithm of Eqs. (16.61) and (16.62) for the same time steps. The algorithm of Eqs. (16.74) requires inversion of the same type of matrix as does the backward-difference algorithm of Eqs. (16.62), and, in addition, requires computation of θgg  and θd , although the latter computation is negligible with respect to the time required for the matrix inversion. In practice, the θ ’s are predetermined based on experience or intuition. Implicit Integration: Time-Integrated Method The delayed neutron precursor equations may, in principle, be integrated directly between tp and tp+1 : Cm (p + 1) = exp(−λm t)Cm (p)  tp +1 G  g + βm dt exp[−λm (tp+1 − t)] νf (t)φ g (t) tp (16.79) g=1 If the assumption is made that the group-fission rate at each point varies linearly in time in the interval tp ≤ t ≤ tp+1 , Eq. (16.79) yields an implicit integration algorithm for the precursors, Cm (p + 1) = exp(−λm t)Cm (p)   G βm 1 − exp(−λm t) g + νf (p)φ g (p) − exp(−λm t) λm λm t g=1 *   G 1 − exp(−λm t) g g − νf (p + 1)φ (p + 1) −1 λm t g=1 (16.80) Integration of the multigroup diffusion equation over the interval tp ≤ t ≤ tp+1 , with the assumption that all reaction rates vary linearly in that interval, results in 615 616 16 Space–Time Neutron Kinetics an implicit integration algorithm for the neutron flux, g g ∇ · D g (p + 1)∇φ g (p + 1) − a (p + 1) + s (p + 1) φ g (p + 1) G  + g  →g s  (p + 1)φ g (p + 1) g  =1  g χp + − M  m=1 G  × g  =1 =−  * M g   g 2 χm βm 1 − exp(−λm t) g −1 βm χp − χm + t λm λm t m=1 g  νf (p + 1)φ g (p + 1) − 2 φ g (p + 1) vg t M 2  g 2 χm [1 − exp(−λm t)]Cm (p) − g φ g (p) t v t  m=1 g − χp − M   g g βm χp − χm m=1 − M  m=1 × G  g  =1 − G   * g 2 χm βm 1 − exp(−λm t) − exp(−λm t) t λm λm t g  g g νf (p)φ g (p) − ∇ · D g (p)∇φ g (p) + a (p) + s (p) φ g (p) g  →g s  (p)φ g (p) (16.81) g  =1 In arriving at Eq. (16.81), integration of the precursors was treated as in Eq. (16.80) (i.e., the group-fission rate was assumed to vary linearly). Equations (16.80) and (16.81) define the time-integrated algorithm, which, like Eqs. (16.73) and (16.74), represents an attempt to reduce the truncation error associated with the simple implicit integration formulas of Eqs. (16.61) and (16.62) without materially increasing the computational time required to obtain a solution. All three implicit integration algorithms require inversion (at each time step) of roughly the same matrix. Numerical experience indicates that the θ -method and the time-integrated method yield essentially identical results, and that both methods are somewhat more accurate than the backward-difference method. Implicit Integration: GAKIN Method The mathematical properties of this method derive directly from the properties of the spatial finite-difference approximation. This approximation is θ̇ = K θ (16.82) 16.3 Time Integration of the Spatial Neutron Flux Distribution where ⎡ ⎤ ψ1 ⎢ .. ⎥ ⎢ . ⎥ ⎢ G⎥ ⎢ψ ⎥ ⎥ θ =⎢ ⎢ d1 ⎥ ⎢ ⎥ ⎢ . ⎥ ⎣ .. ⎦ dM (16.83) with ψ g and d m representing N × 1 column vectors of group fluxes and m-type precursor densities, respectively, at each of N spatial mesh points. The matrix K can be written in terms of N × N submatrices Kij : ⎡ ⎢ ⎢ K =⎢ ⎣ K 11 K 21 .. . K 12 K 22 K G+M,1 K G+M,2 K 13 K 23 ⎤ K 1,G+M K 2,G+M ⎥ ⎥ ⎥ .. ⎦ . · · · K G+M,G+M ··· ··· (16.84) The N × N matrices K ij are split, K ij = Γ ij + vi D i , 1≤i≤G (16.85) where D i represents the coupling among mesh points due to the diffusion term. By splitting K into a matrix L , which contains all the submatrices below the diagonal block; a matrix U , which contains all the submatrices above the diagonal block; and into the block diagonal matrices Γ and D , ⎡ 0 ··· 0 ··· 0 K 21 .. . 0 0 K G+M,1 K G+M,2 ⎢ L=⎢ ⎣ ⎡ 0 K 12 ⎢0 0 ⎢ U =⎢. ⎣ .. 0 ⎡ Γ 11 ⎢ 0 ⎢ ⎢ 0 ⎢ Γ =⎢ 0 ⎢ ⎢ . ⎣ .. 0 0 Γ 22 ··· ··· ··· 0⎤ 0⎥ .. ⎥ ⎦ . 0 (16.86) ⎤ · · · K 1,G+M · · · K 2,G+M ⎥ ⎥ ⎥ ⎦ K 13 K 23 ··· (16.87) 0 0 0 Γ GG ··· ··· ··· ··· K G+1,G+1 0 0 0 0 .. . K G+M,G+M ⎤ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎥ ⎦ (16.88) 617 618 16 Space–Time Neutron Kinetics ⎡ v1D 1 v2 D 2 ⎢ 0 ⎢ . ⎢ .. ⎢ ⎢ D =⎢ 0 ⎢ ⎢ 0 ⎢ . ⎣ . . 0 0 ··· 0 0 ··· ··· vG D G ··· ··· 0 ··· ··· ··· 0⎤ 0⎥ .. ⎥ .⎥ ⎥ ⎥ 0⎥ ⎥ 0⎥ .. ⎥ ⎦ . 0 (16.89) Equation (16.82) may be written L + U )θ + D θ θ̇ − Γ θ = (L (16.90) This equation may formally be integrated over the interval tp ≤ t ≤ tp+1 :  t θ(tp+1 ) = exp(tΓ )θ (tp ) +  +     L + U )θ tp + t  dt  exp t − t  Γ (L 0 t     dt  exp t − t  Γ D θ tp + t  (16.91) 0 In the first integral of Eq. (16.91), the approximation     θ tp + t  = exp ωt  θ(tp ) (16.92) is made, and the second integral is performed with the approximation     θ tp + t  = exp −ω t − t  θ (tp+1 ) (16.93) In general, ω is a diagonal matrix. Using Eqs. (16.92) and (16.93) in Eq. (16.91) results in   I − (ω − Γ )−1 I − exp (Γ − ω)t D θ(tp+1 )   L + U ) θ (tp ) (16.94) = exp(Γ t) + (ω − I )−1 exp(ωt) − exp(Γ t) (L which may be written θ (tp+1 ) = A θ(tp ) (16.95) If all the diagonal elements of ω are equal to ω1 , which is the eigenvalue of K θ n = ωn θ n (16.96) with largest real part, then from Eq. (16.90), L + U )θ 1 = (ω1I − Γ − D )θ 1 (L (16.97) From the definition of A (with ω = ω1I ) it can be shown that A θ 1 = exp(tω1 )θ 1 (16.98) 16.3 Time Integration of the Spatial Neutron Flux Distribution It can be shown that ω1 is real and simple, and that θ 1 is positive. For all real values of ω, hence for ω = ω1 , A can be shown to be nonnegative, irreducible, and primitive. From the Perron–Frobenius theorem it follows that A has a simple, real, largest eigenvalue ρ1 and a corresponding positive eigenvector. The eigenvalue ρ1 = exp(tω1 ) is seen from Eq. (16.98) to have a positive eigenvector that is the fundamental-mode solution of the kinetics equations (16.96). If it can be shown that ρ1 is the largest eigenvalue of A , Eq. (16.98) indicates that the asymptotic solution of the integration algorithm of Eq. (16.95) is the asymptotic solution of Eq. (16.82) for a step change in properties, which shows that the method is unconditionally numerically stable. The transpose matrix A T has the same properties and eigenvalue spectrum as A : A T q n = ρn q n (16.99) By the Perron–Frobenius theorem, A T has a real, simple eigenvalue, ρk , which is larger than the real part of the other eigenvalues, and the corresponding eigenvector is positive. Premultiplying Eq. (16.93) for n = k by θ T1 , premultiplying Eq. (16.98) by q T1 , and subtracting yields 0 = [exp(tω1 ) − ρ1 ]θ T1 q 1 (16.100) Because θ 1 and q 1 are positive, Eq. (16.100) is satisfied only if exp(tω1 ) = ρ1 is the real eigenvalue. Thus the method is numerically unconditionally stable. Inversion of the matrix on the left of Eq. (16.94) to obtain A can be accomplished by the inversion of GN × N matrices. In practice, an approximation to ω1 is obtained by an expression of the form ω1 = θi (tp ) 1 ln t θi (tp−1 ) (16.101) where i indicates some component or components of the θ vector, and different values of ω1 are used in different parts of the reactor (i.e., ω = ω1I ). Alternating Direction Implicit Method The implicit integration methods of previous sections all reduced to an algorithm for the neutron flux which required the inversion of a matrix at each time step. When the finite-difference spatial approximation is employed, this matrix is NG × NG, where N is the number of mesh points and G is the number of energy groups. In one-dimensional problems, the matrix to be inverted becomes block tridiagonal with G × G blocks, and inversion can be accomplished by the backwardelimination/forward-substitution method and requires the inversion of N G × G matrices. In the GAKIN method, this matrix inversion can be accomplished by inverting G N × N matrices. 619 620 16 Space–Time Neutron Kinetics However, for multidimensional problems, the matrix inversion associated with the implicit methods poses a formidable and time-consuming task. Alternative formulations of the θ and GAKIN methods have been proposed to reduce the time required for this matrix inversion. Another technique, designed to eliminate this same problem, is the alternating direction implicit (ADI) method. The basis of the ADI method is to make the algorithm implicit for one space dimension at a time and to alternate the space dimension for which the algorithm is implicit. The ideas involved are illustrated by a two-dimensional problem. The equation for the group g neutron flux can be written in the notation of Section 16.2 as     1 1 g g g ψ̇ = vg D x + Γ gg ψ g + vg D y + Γ gg ψ g 2 2 G  +  K gg  ψ g + g  =1 M  K g,G+md m (16.102) m=1 where the N × N diffusion matrix D g , which represents ∂ g ∂ ∂ g ∂ D + D ∂x ∂x ∂y ∂y g has been separated into D x , which represents ∂ g ∂ D ∂x ∂x g and D y , which represents ∂ g ∂ D ∂y ∂y For the time step tp to tp+1 , an integration algorithm which is implicit in the x-direction and explicit in the y-direction is chosen. First define g g H x ≡ vg D x + 12 Γ gg (16.103) g g H y ≡ vg D y + 12 Γ gg then the algorithm is written ψ g (p + 1) − ψ g (p)  g g = t H x (p + 1)ψ g (p + 1) + H y (p)ψ g (p) + + M  m−1 g  −1  K g,G+m (p)dd m (p) , G  g = 1, . . . , G  K gg  (p)ψ g (p) 16.3 Time Integration of the Spatial Neutron Flux Distribution or H gx (p + 1) ψ g (p + 1) I − tH H gy (p) ψ g (p) = I + tH   G M    g K gg  (p)ψ (p) + K g,G+m (p)dd m (p) , + t g  =1 g = 1, . . . , G m=1 (16.104) For the time step tp+1 to tp+2 , an algorithm that is implicit in the y-direction and in the removal, scattering, fission, and precursor terms is chosen: ψ g (p + 2) − ψ g (p + 1)  g g = t H x (p + 1)ψ g (p + 1) + H y (p + 2)ψ g (p + 2) + G   K gg  (p + 2)ψ g (p + 2) g  =1 + M   K g,G+m (p + 2)dd m (p + 2) , d = 1, . . . , G (16.105) m=1 Use, for the sake of definiteness, the implicit integration formulas of Eq. (16.61) for the precursors, d m (p + 2) = G  1 βm d m (p + 1) + F g (p + 2)ψ g (p + 2) 1 + λm t 1 + λm t g=1 (16.106) g where F G is an N × N diagonal matrix representing νf associated with each point. Using Eq. (16.106), Eq. (16.105) becomes H gy (p + 2) ψ g (p + 2) I − tH  G M    K gg  (p + 2)ψ g (p + 2) + K g,G+m (p + 2) −t g  =1 × G  m=1 βm 1 + λm t    F g (p + 2)ψ g (p + 2) g  =1 H gx (p + 1) ψ g (p + 1) = I + tH + t M  m=1 1 K g,G+m (p + 2)dd m (p + 1), 1 + λm t g = 1, . . . , G (16.107) 621 622 16 Space–Time Neutron Kinetics The solution proceeds by alternating between the algorithms of Eqs. (16.104) and (16.107). If there are N 1/2 mesh points in both the x- and y-directions, the matrices that must be inverted in order to solve Eqs. (16.104) and (16.107) can be partitioned so that, rather than inverting an NG × NG matrix, N 1/2 N 1/2 G × N 1/2 G matrices are inverted. This happens because the matrix to be inverted in Eqs. (16.104) couples mesh points only in the x-direction, and the matrix to be inverted in Eq. (16.107) couples mesh points only in the y-direction. In the case of Eq. (16.104), each of the N 1/2 G × N 1/2 G matrices can be further partitioned into G N 1/2 × N 1/2 matrices, because the neutron source terms due to fission, scattering, and precursor decay are treated explicitly in this step. More general algorithms treat these source terms implicitly in both steps. Stiffness Confinement Method The set of neutron and delayed neutron precursor equations are referred to as stiff because of the great difference in the time constants that govern the prompt neutron and precursor responses. The accuracy and stability of numerical integration methods are usually determined by the shortest time constant, the prompt neutron lifetime, which has little effect on the precursor solution. The stiffness confinement method seeks to confine the difficulty to the neutron equations by decoupling the precursor equations through the definition of dynamic frequencies: g ωφ (r, t) ≡ 1 ∂φg , φg ∂t ωcm ≡ 1 ∂Cm Cm ∂t (16.108) These definitions can be used to replace the time derivatives in the multigroup diffusion and precursor equations, which allow the latter to be formally solved and used to evaluate the precursor densities in the multigroup diffusion equations, resulting in   g G  ωφ (r, t)  g φg (r, t) +  g →g (r, t)φg  (r, t) ∇ · D g (r, t)φg (r, t) − t + vg   g + (1 − β)χp + M  m=1 g = 1, . . . , G g βm λm χm m ωc (r, t) + λm * g =1 G  g  =1 g νf (r, t)φg  (r, t) = 0, (16.109) These equations are identical to the static multigroup diffusion equations, but with modified total and fission cross sections which include the dynamic frequencies. Thus, to advance the solution in time, an estimate is made of the dynamic frequencies, Eqs. (16.109) are solved for the group fluxes, the precursors are updated, an improved guess of the dynamic frequencies is calculated using the new flux and precursor values, and the iteration is repeated until convergence. 16.3 Time Integration of the Spatial Neutron Flux Distribution Symmetric Successive Overrelaxation Method Successive over-relaxation is combined with an exponential transformation to decouple stiffness in the symmetric successive over-relaxation (SSOR) method. The matrix H is first decomposed into a lower L , a diagonal D , and an upper U matrix: H =L +D +U (16.110) The solution is then advanced iteratively over the (p + 1) time step by a forward sweep: L + D )]−1 [tp+1U ψ n (p + 1) + ψ(p)] ψ n+1/2 (p + 1) = θ[II − tp+1 (L + (1 − θ)ψ n (p + 1) (16.111) followed by a backward sweep: D + U )]−1 [tp+1L ψ n+1/2 (p + 1) + ψ(p)] ψ n+1 (p + 1) = θ[II − tp+1 (D + (1 − θ)ψ n+1/2 (p + 1) (16.112) where 1 ≤ θ ≤ 2 and n refers to the iteration number. An exponential transformation of the multigroup fluxes and the precursor densities ψ(p + 1) = exp(tp+1 ω)ψ̂(p + 1) (16.113) may be made first, using dynamic frequencies calculated from local flux and precursor values for the present and previous times: g ωφ ≡ φg (p) 1 ln , tp+1 φg (p − 1) ωm c (p + 1) ≡ 1 Cm (p) ln (16.114) tp+1 Cm (p − 1) With the transformation of Eq. (16.113), Eq. (16.52) becomes ∂ H − ω] exp(tp+1 ω)ψ̂ ψ̂ = exp(−tp+1 ω)[H ∂t (16.115) which is integrated using the over-relaxation procedure of Eqs. (16.111) and (16.112). The dynamic frequencies are estimated at the beginning of the time step from Eqs. (16.114) to determine ω0 . A global frequency correction factor ωn is computed on each iteration by considering Eqs. (16.111) and (16.112) to each advance the solution a half time step. The dynamic frequency is then corrected: Ω n = Ω 0 + ωnI where now Ω is a matrix containing the local values of the frequencies ω. (16.116) 623 624 16 Space–Time Neutron Kinetics Generalized Runge–Kutta Methods Runge–Kutta methods have long been popular for integrating ordinary differential equations, but the requirement for small time steps to achieve sufficient accuracy has limited their application in solving space-discretized space–time neutron kinetics problems. However, generalizations of these methods to allow larger time steps and increased stability (Ref. 3) have recently been applied to these problems. The Runge–Kutta method is based on an explicit time differencing of Eq. (16.52) and a linear Taylor’s series approximation: ψ(p + 1) = ψ(p) + tp+1H (p + 1)ψ(p + 1)   ∂ H ψ) (H  ψ(p) + tp+1 H (p)ψ(p) + tp+1 ∂ψ p  1 × [ψ(p + 1) − ψ(p)] tp+1 (16.117) H ψ)/∂ψ|p is the partial derivative of the left side of Eq. (16.52) where the term ∂(H with respect to the appropriate multigroup neutron flux or delayed neutron precursor density evaluated at the beginning of the time step, t = tp . The generalized Runge–Kutta methods are based on the algorithm y (p + 1) = y (p) + s  ci K i (p + 1) (16.118) i=1 for advancing the solution from tp to tp+1 , where s is the number of stages, ci are fixed expansion coefficients, and the column vectors K (p + 1) are found by solving a system of N (the number of energy groups times discrete spatial points plus the number of delayed neutron precursor groups times the number of discrete spatial points) linear equations for each of the s stages (i.e., for s different right sides for each time step):   ∂ H ψ) K i (p + 1) I − γ tp+1 (H ∂ψ p   i−1  ∂ ∗ H H K = tp+1 ψ|p + tp+1 (H ψ) γim m (p + 1) , i = 1, . . . , I ∂ψ p ∗ m=1 (16.119) where H ψ|p∗ is the evaluation of the left side of Eq. (16.52) at the intermediate points tp∗ where the solution vector is given by ψ(p∗ ) = ψ(p) + i−1  αimK m (p + 1) (16.120) m=1 where γ , γm , and αim are fixed constants. The scheme is well suited for a variable time step because it employs an embedded Runge–Kutta–Fehlberg estimate 16.4 Stability for ψ(p + 1), which provides the capability to monitor truncation error without increasing computational time. 16.4 Stability In a nuclear reactor operating at steady-state conditions, an equilibrium obtains among the interacting neutronic, thermodynamic, hydrodynamic, thermal, xenon, and so on, phenomena. The state of the reactor is defined in terms of the values of the state functions1) associated with each of these phenomena (e.g., the neutron flux, the coolant enthalpy, the coolant pressure). If a reactor is perturbed from an equilibrium state, will the ensuing state (1) remain bounded within some specified domain of the state functions, (2) return to the equilibrium state after a sufficiently long time, or (3) diverge from the equilibrium state in that one or more of the state functions takes on a shape outside a specified domain of state functions? This is the question of stability. In this section we extend the concepts of Section 5.9 to outline a theory appropriate for the stability analysis of spatially dependent reactor models. First, we consider the stability analysis of the coupled system of ordinary differential equations that results when the spatial dependence is discretized by a finite-difference, nodal, or other approximation. Then the extended Lyapunov theory for the stability analysis of the coupled partial differential equations which describe spatially continuous systems is discussed. Classical Linear Stability Analysis The finite-difference, time-synthesis, nodal, or point kinetics approximations, and the corresponding approximations to the other state function equations, may be written as a coupled set of ordinary differential equations relating the discrete state variables yi :   ẏi (t) = fi yi (t), . . . , yN (t) , i = 1, . . . , N (16.121) where, for instance, yi may be the neutron flux at node i and yI +j may be the coolant enthalpy at node j . The coupling among the equations arises because the cross sections in the neutronics equations depend on the local temperature, density, and xenon concentration, because the temperature, density, and xenon concentration depend on the local flux, and because neutron and heat diffusion and coolant transport introduces a coupling among the value of the state variables at different locations. 1) In a spatially dependent system such as a nuclear reactor, the state of the system is defined in terms of spatially dependent state functions. When the spatial dependence is discretized by one of the approximations discussed in previous sections, the state of the system is defined in terms of discrete state variables. 625 626 16 Space–Time Neutron Kinetics Equations (16.121) may be written as a vector equation,   ẏy (t) = f y (t) (16.122) where the components of the column vectors y and f are the yi and fi , respectively. The equilibrium state y e satisfies f (yy e ) = 0 (16.123) If the solution of Eq. (16.122) is expanded about y e , y (t) = y e + ŷy (t) (16.124) and the part of the right-hand side of Eq. (16.122) that is linear in ŷy is separated out, Eq. (16.122) may be written   ŷy˙ (t) = h (yy e )ŷy (t) + g y e , ŷy (t) (16.125) The matrix h has constant elements, some of which may depend on the equilibrium state. Classical linear stability analysis proceeds by ignoring the nonlinear term g in Eq. (16.125). It is readily shown that the condition for the stability of the linearized equations is that the real part of all eigenvalues of the matrix h are negative. To illustrate this, apply a permutation transformation that diagonalizes h to the linear approximation to Eq. (16.125): P = P ThP P Tŷy (t)P P P Tẏyˆ (t)P (16.126) P TP = P P T = I (16.127) since P . Then the transformed equations are decoupled: Define X (t) = P Tŷy (y)P Ẋi (t) = ωi Xi (t), i = 1, . . . , N (16.128) where ωi are the eigenvalues of h . The solutions of these equations subject to Xi (0) = Xi0 are Xi (t) = Xi0 eωi t , i = 1, . . . , N (16.129) which may be written in vector notation as X i0 X (t) = Γ (t)X (16.130) where Γ (t) = diag(exp(ωi t)). Hence P T = P Γ (t)X X i0P T ŷy (t) = P X X(t)P (16.131) 16.4 Stability If Re{ωi } < 0, limt→∞ ŷy (t) = 0 (i.e., the state of the system returns to the equilibrium state). If Re{ωi } > 0, one or more of the components of ŷ approach ∞ as t → ∞, and the system is unstable. Thus stability analysis of the linearized equations amounts to determining if the eigenvalues of the h matrix are in the left (stable) or right (unstable)-half complex plane. This determination may be accomplished most readily by Laplace transforming the linearized equation into the frequency domain and then applying one of the methods of linear control theory (e.g., Bode, Nyquist, root locus, Hurwitz) that have been developed explicitly for this purpose. This methodology was applied in the stability analyses of Chapter 5. Lyapunov’s Method The method of Lyapunov attempts to draw certain conclusions about the stability of the solution of Eq. (16.125) without any knowledge of this solution. Essential to this method is the choice of a scalar function V (ŷy ) which is a measure of a metric distance of the state y = y e + ŷy from the equilibrium state y e . Let ŷy {t, ŷy 0 ) be the solution of Eq. (16.125) for the initial condition ŷy (t = 0) = ŷy 0 . If it can be shown that V (ŷy (t, ŷy 0 )) will be small when V (ŷy 0 ) is small, then y e is a stable equilibrium state. If, in addition, it can be shown that V (ŷy (t, ŷy 0 )) approaches zero for large times, y e is an asymptotically stable equilibrium state. Define a scalar function V (ŷy ) that depends on all the state variables ŷy i and which has the following properties in some region R about the equilibrium state y e : 1. V (ŷy ) is positive definite [i.e., V (ŷy ) > 0 if ŷy = 0, V (ŷy ) = 0 if ŷy = 0]. 2. limŷy →0 V (ŷy ) = 0, limŷy →∞ V (ŷy ) = ∞. 3. V (ŷy ) is continuous in all its partial derivatives (i.e., ∂V /∂yi exist and are continuous for i = 1, . . . , N). 4. V̇ (ŷy ) evaluated along the solution of Eq. (16.125) is nonpositive; that is, V̇ (ŷy ) = M  ∂V i=1 ∂ ŷi ŷ˙ i = N  ∂V i=1 ∂ ŷi fi ≤ 0 (16.132) A scalar function V (ŷy ) satisfying properties 1 to 4 is a Lyapunov function. Three theorems based on the Lyapunov function can be stated about the equilibrium solution of Eq. (16.125). Theorem 16.1 (Stability Theorem). If a Lyapunov function exists in some region R about y e , this equilibrium state is stable for all initial perturbations in R [i.e., for all initial perturbations ŷy 0 in R, the solution of Eq. (16.125), y (t, ŷy 0 ), remains within the region R for all t > 0]. Theorem 16.2 (Asymptotic Stability Theorem). If a Lyapunov function exists in some region R about y e , and in addition V̇ evaluated along the solution of Eq. (16.125) is negative definite (V̇ < 0 if ŷy = 0, V̇ = 0 if ŷy = 0) in R, this equilibrium state is 627 628 16 Space–Time Neutron Kinetics asymptotically stable for all initial perturbations in R [i.e., for all initial perturbations ŷy 0 in R, the solution of Eq. (16.125) is ŷy (t, ŷy 0 ) = 0 after a sufficiently long time]. Theorem 16.3 (Instability Theorem). If a scalar function V (ŷy ) which has properties 1 to 3 exists in a region R, and V̇ evaluated along the solution of Eq. (16.125) does not have a definite sign, the equilibrium state y e is unstable for initial perturbations in R [i.e., for initial perturbations ŷy 0 in R, the solution of Eq. (16.125), ŷy (t, ŷy 0 ), does not remain in R for all t > 0]. Mathematical proofs of these theorems can be constructed. Rather than repeat these proofs, which may be found in the literature (e.g., Ref. 14), it is more informative to consider a topological argument. Properties 1 to 3 define a concave upward surface (the function V ) in the phase space defined by the ŷi . This surface has a minimum within the region R at ŷ1 = · · · = ŷN = 0 by property 1, and increases monotonically in value as the ŷi increase, by properties 2 and 3. Thus contours can be drawn in the hyperplane of the ŷi representing the locus of points at which V has a given value. These contours are concentric about the equilibrium state ŷi = 0, i = 1, . . . , N . Proceeding outward from this origin, the value of V associated with each contour is greater than the value associated with the previous contour. In other words, V (yy ) is a bowl in the hyperspace of the yi , with center at ŷi = 0, i = 1, . . . , N . The outward normal to those contours is N  ∂V i=1 ∂ ŷi i where i denotes the unit vector in the direction in phase space associated with the state variable ŷi . The direction in which the state of the system is moving in phase space is given by N  ẏi i = i=1 N  (16.133) fi i i=1 For stability, the direction in which the state of the system is moving must never be toward regions in which V is larger (i.e., never away from the equilibrium state): % N  ∂V i=1 ∂ ŷi & % i · N  j =1 & fj j = N  ∂V i=1 ∂ ŷi fi ≤ 0 (16.134) For asymptotic stability, the state of the system must always move toward regions in which V is smaller (i.e., always move toward the equilibrium state). Thus the inequality must always obtain in the foregoing relation. If the system can move away from the equilibrium state into regions of larger V , the ≤ is replaced by > in the foregoing relation and the equilibrium state is unstable. 16.4 Stability The Lyapunov method yields the same results obtained in the preceding section in the limit in which the nonlinear terms are small. The function V (ŷy ) = ŷy Tŷy = N  (ŷi )2 (16.135) i=1 satisfies properties 1 to 3. Making use of Eq. (16.125) yields V̇ (ŷy ) = N  ∂V i=1 ∂ ŷi ŷi = 2 N  + , ŷi ŷ˙ i = 2ŷy Tŷy˙ = 2 ŷy Thŷy + ŷy Tg (16.136) i=1 If the region R is defined such that ŷy Thŷy > ŷy Tg a sufficient condition for V̇ to be negative definite in R is that ŷy Thŷy is negative definite, a sufficient condition for which is that the eigenvalues of h have negative real parts. This is the same result obtained in the linear analysis of the preceding section. In this case, the Lyapunov method provides, in addition, the region R within which the linear analysis is valid. In applying the Lyapunov method, construction of a suitable Lyapunov function is the main consideration. Because the Lyapunov function for a system of equations is not unique, the analysis yields sufficient, but not necessary, conditions for stability. Lyapunov’s Method for Distributed Parameter Systems A more basic characterization of a reactor system is in terms of spatially distributed state functions, rather than discrete state variables. These state functions satisfy coupled partial differential equations, which may be written   ẏi (r, t) = fi y1 (r, t), . . . , yN (r, t), r , i = 1, . . . , N (16.137) where yi is a state function (e.g., neutron group flux) and fi denotes a spatially dependent operation involving scalars and spatial derivatives on the state functions. These equations may be written   ẏy (r, t) = f y (r, t), r (16.138) where y is a column vector of the yi and f is a column vector of the operations denoted by the fi . The extension of Lyapunov’s methods to systems described by state functions involves the choice of a functional that provides a measure of the distance of the vector of state functions y from a specified equilibrium state, y eq . The distance between two states y a and y b , d[yy a ,yy b ], is defined as the metric on the product state function space consisting of all possible functions of position that the component state functions can take on. 629 630 16 Space–Time Neutron Kinetics An equilibrium state y eq (r) satisfying   f y eq (r), r = 0 (16.139) is stable if, for any number ε > 0, it is possible to find a number δ > 0 such that when d[yy 0 (r),yy eq (r)] < δ then d[yy (r, t;yy 0 ),yy eq (r)] < ε for t ≥ 0 (16.140) where y (r, t;yy 0 ) is the solution of Eq. (16.138) with the initial condition y (r, 0) = y 0 (r). If in the limit of large t , the distance d[yy (r, t;yy 0 ),yy eq ] approaches zero, then y eq is asymptotically stable. Theorem 16.4 (Stability Theorem). For an equilibrium state y eq (r) to be stable, it is necessary and sufficient that in some neighborhood of y eq (r) that includes the equilibrium state there exists a functional V [yy ] with the following properties: 1. V is positive definite with respect to d[yy ,yy eq ]; that is, for any C1 > 0, there exists a C2 > 0 depending on C1 such that when d[yy ,yy eq ] > C1 , then V [yy ] > C2 for all t ≥ 0, and limd[yy ,yy eq ]→0 V [yy ] = 0. 2. V is continuous with respect to d[yy ,yy eq ]; that is, for any real ε > 0, there exists a real δ > 0 such that V [yy ] < ε for all y in the state function space for 0 < t < ∞, when d[yy 0 ,yy eq ] < δ. 3. V [yy ] evaluated along any solution y of Eq. (16.138) is nonincreasing in time for all t > 0 provided that d[yy 0 ,yy eq ] < δ0 , where δ0 is a sufficiently small positive number. Theorem 16.5 (Asymptotic Stability Theorem). If, in addition to these three conditions, V [yy ] evaluated along any solution to Eq. (16.138) approaches zero for large t , the equilibrium state is asymptotically stable. The same type of topological arguments made above in support of the theorems for the discrete representation of spatial dependence by coupled ODES are appropriate here, if the state space is generalized to a state function space. Construction of a suitable Lyapunov functional is the essential aspect of applying the theory of this section. Although the conditions cited in the theorems are necessary and sufficient for stability, the V -functional chosen may result in more restrictive stability criteria than would be obtained from another V -functional. Thus stability analyses employing Lyapunov functionals yield only sufficient conditions for stability. Control Control An intended change in the operating state of a nuclear reactor is produced by a control action (e.g., withdrawing a bank of control rods, increasing the coolant flow). The nature of the change in operating state depends on the control action, of course, and a great deal of practical experience exists on how to effect a desired change. However, in some cases the intuitive control action can exacerbate, rather than correct, a problem—the control-induced xenon spatial oscillations in the large production reactors being a good example. The methodology of control theory has found some application in nuclear reactor control, and a brief review is provided in this section. Variational Methods of Control Theory When discrete spatial approximations (e.g., nodal, finite-difference) are employed, the dynamics of a spatially dependent nuclear reactor model are described by a system of ordinary differential equations ẏi (t) = fi (y1 , . . . , yN , u1 , . . . , uR ), i = 1, . . . , N (16.141) with the initial conditions yi (t = t0 ) = yi0 , i = 0, . . . , N (16.142) The yi are the state variables (e.g., nodal neutron flux, temperature) and the ur are control variables (e.g., control rod cross section in a node). Equation (16.141) may be written more compactly by defining vector variables y , u , and f :   u(t) ẏy (t) = f y (t),u (16.143) Many problems in control may be formulated as a quest for the control vector u ∗ that causes the solution of Eq. (16.143), y ∗ , to minimize a functional:2)  J [yy ] = tf   dt F y (t), ẏy (t) (16.144) t0 This control problem may be formulated within the framework of the classical calculus of variations by treating the control variables as equivalent to the state variables. The theory of the calculus of variations is restricted to variables that are continuous in time, which limits the admissible set of control variables. 2) Functionals of this form may arise when the objective of the control program is to correct a flux perturbation in such a manner as to minimize the deviation from the nominal flux distribution, at the same time minimizing the rate of change of local flux densities. Other typical control problems are those in which the objective is to attain a given final state in a minimum time; a functional with F = 1 and an additional term that provides a measure of the deviation from the specified final state is appropriate in this case. 631 632 16 Space–Time Neutron Kinetics The system equations are treated as constraints or subsidiary conditions, and are included in the functional with Lagrange multiplier variables:   u,λ λ] = J [yy ,u tf t0   N    +  , u(t) dt F y (t), ẏy (t) + λi (t) ẏi (t) − fi y (t),u i=1 (16.145) Variations of the modified functional J  (with respect to each yi and ur ) are required to vanish at the minimum:   δJ = 0 = tf t0  N N   ∂fj ∂F ∂F dt δyi + δ ẏi + λi δ ẏi − λj δyi ∂yi ∂ ẏi ∂yi  j =1 i=1 − R  λi r=1 ∂fi δur ∂ur * (16.146) Integrating the δ ẏi terms by parts and using the initial conditions to set δyi (t0 ) = 0, this expression becomes   δJ = 0 = tf dt t0 + N   j =1 i=1 N    * N R   ∂fj ∂fi ∂F ∂ ∂F δyi − − − λ̇i − λi λi δur ∂yi ∂t ∂ ẏi ∂yi ∂ur  ∂F δyi ∂ ẏi λi + i=1 r=1 (16.147) t=tf In order that Eq. (16.147) be satisfied for arbitrary (but continuous) variations δyi and δur , it is necessary that λ̇i (t) = − N  λj (t) j =1  + N  λi (t) i=1 ∂fj u) (yy ,u ∂yi  ∂F ∂ ∂F f )− f) , (yy ,f (yy ,f ∂yi ∂t ∂ y˙i ∂fi u) = 0, (yy ,u ∂ur r = 1, . . . , R i = 1, . . . , N (16.148) (16.149) and that λi satisfy the final conditions λi (tf ) + ∂F ∂ ẏi = 0, i = 1, . . . , N (16.150) tf Equations (16.141), (16.148), and (16.149) must be solved simultaneously, subject to the initial conditions of Eqs. (16.142) and the final conditions of Eq. (16.150), for the optimal controls u∗r (t) and the optimal solutions yi∗ (t). Control In many problems, additional constraints are placed on the allowable values that may be taken on by the state variables and control variables. Constraints of the form u(t)) = 0, φm (yy (t),u m = 1, . . . , M < N or   u(t), u̇ u(t) = 0, φm y (t), ẏy (t),u m = 1, . . . , M < N may be added to the functional of Eq. (16.144) with Lagrange multiplier variables and treated in the same fashion as before. Equations for additional Lagrange multiplier variables and the additional constraint equations are included with Eqs. (16.141), (16.148), and (16.149) in this case. When integral constraints of the form  tf   u(t), u̇ u(t) = 0, m = 1, . . . , M < N dt φm y (t), ẏy (t),u t0 are present, the functional of Eq. (16.144) is modified with Lagrange multiplier constants ωm , &   tf % M tf  ˆ dt F + ωm φm = dt F̂ (16.151) J →J = t0 m=1 t0 and the derivation proceeds as before with F → F̂ . In addition to Eqs. (16.141), (16.148), and (16.149), the constraint equations and expressions for the ωm are obtained. Inequality constraints (e.g., maximum control rod shim rates) are encountered frequently. Although these can sometimes be reduced to equivalent equality constraints of one of the three types discussed, they generally constitute a class of problems that are difficult to treat within the framework of the calculus of variations. Another class of such problems is those for which the optimal control is discontinuous. Dynamic Programming An alternative treatment of the variational problem that circumvents the requirement for continuous control variables is provided by dynamic programming. Consider the problem of determining the control vector u ∗ (t) that causes the solution y ∗ (t) of Eq. (16.143) to minimize the functional of Eq. (16.144), subject to constraints on the allowable values of the control variable that may be represented by u (t) ∈ Ω(t) (16.152) To develop the dynamic programming formalism, consider the functional of Eq. (16.144) evaluated between a variable lower limit (t,yy (t)) and a fixed upper 633 634 16 Space–Time Neutron Kinetics limit (tf ,yy (tf )). Define the minimum value of this functional as S, a function of the lower, variable limit (t,yy (t)):  tf      u(t  ) dt  F y (t  ), ẏy y (t  ),u (16.153) S t,yy (t) = min u∈ t In writing Eq. (16.153), ẏy is written as an explicit function of y and u to indicate that Eq. (16.143) must be satisfied in evaluating the integrand. By definition of S, for t > 0,  t+t        u(t  ) dt  F y (t  ), ẏy y (t  ),u S t,yy (t) ≤ S t + t,yy (t + t) + t where y (t + t) and y (t) are related by Eq. (16.143); that is,     u(t) + O t 2 y (t + t) = y (t) + t f y (t),u (16.154) (16.155) For the optimal choice of u (t  ) = u ∗ in the interval t ≤ t  ≤ t + t , the equality obtains in Eq. (16.154). Approximating the integral in Eq. (16.154) by taking the integrand constant at its value at t , this equation becomes3)        f y (t),u u(t) S t,yy (t) = min S t + t,yy (t + t) + tF y (t),f u(t)∈Ω(t) (16.156) Equation (16.156) can be solved by retrograde calculation, starting with the final condition  tf   dt  F (yy , ẏy ) = 0 (16.157) S tf ,yy (tf ) = min u∈ tf In each step of the retrograde solution, the optimal manner to proceed from each possible state y(t) to time tf is computed. Thus, when the initial time is reached, the optimal control at each discrete time and the corresponding sequence of states constituting the optimal trajectory are known. Pontryagin’s Maximum Principle When a Taylor’s series expansion of the first term on the right of Eq. (16.156) is made, this equation becomes  N      ∂S  ∂S  u(t) t,yy (t) + t,yy (t) fi y (t),u 0 = min u (t)∈Ω(t) ∂t ∂yi i=1     f y (t),u u(t) + F y (t),f (16.158) 3) In Eq. (16.156), the minimization is with respect to the values of the control vector at time t . These values are assumed constant over the interval t to t + t . On the other hand, the minimization in Eq. (16.153) is with respect to the values taken on by the control vector at all times t  , t ≤ t  ≤ tf . Control Define the variables ψi (t) ≡ −  ∂S  t,yy (t) , ∂yi ψN+1 (t) ≡ − i = 1, . . . , N (16.159)  ∂S  t,yy (t) ∂t (16.160) With these definitions, Eq. (16.158) becomes   N       u(t) + F y (t),f f y (t),u u(t) 0 = min −ψN+1 (t) − ψi (t)fi y (t),u u(t)∈(t) i=1 which may be written  ψN+1 (t) + 0 = max u(t)∈(t) N        u(t) − F y (t),f f y (t),u u(t) ψi (t)fi y (t),u i=1 (16.161) This is the maximum principle of Pontryagin. When the vector u (t) takes on its optimal value, derivatives of the quantity within the square brackets with respect to t and yi must vanish, which requires that & % N N N   ∂ψN+1  ∂ψi ∂fi ∂F ∂fi ∂F , j = 1, . . . , N + fi = − ψi + + ∂yj ∂yj ∂yj ∂yj ∂ ẏi ∂yj i=1 i=1 i=1  ∂fi ∂ψN+1  ∂ψi + fi = − ψi ∂t ∂t ∂t N N i=1 i=1 Using the identities  ∂ 2S  ∂ψi dψj ∂ 2S ∂yN+1 =− fi − = fi + , dt ∂yi ∂yj ∂t∂yj ∂yj ∂yj N N i=1 i=1 j = 1, . . . , N  ∂ 2S dψN+1 ∂ 2 S  ∂ψi ∂ψN+1 =− fi − 2 = fi + dt ∂yi ∂t ∂t ∂t ∂t N N i=1 i=1 these equations become  ∂fi dψj ψi + =− dt ∂yj N i=1 % & N  ∂F ∂fi ∂F , + ∂yj ∂ ẏi ∂yj j = 1, . . . , N (16.162) i=1  ∂fi dψN+1 ψi =− dt ∂t N (16.163) i=1 Appropriate final conditions for the ψi and ψN+1 can be shown to be ψ1 (tf ) = · · · = ψN (tf ) = ψN+1 (tf ) = 0 (16.164) 635 636 16 Space–Time Neutron Kinetics Thus Eqs. (16.141), (16.161), (16.162), and (16.163) are solved simultaneously, subject to the initial and final conditions of Eqs. (16.142) and (16.164), respectively. The computational procedure for solving either the calculus of variations or maximum principle equations is generally iterative. At t = t0 , the yi are known from the initial conditions. When the maximum principle formulation is used, initial values of ψi are guessed, and the initial value of the control variables are determined from Eq. (16.161). Then the yi and ψi are calculated at t0 + t from Eqs. (16.141) and (16.162) and (16.163) and the control is found from Eq. (16.161), and so on. This procedure is repeated in small time increments until the final time tj . Then ψi (tf ) and ψN+1 (tf ) are compared with the final conditions: ψ1 (tf ) = · · · = ψN (tf ) = ψN+1 (tf ) = 0 and the initial values of ψi and ψN+1 are changed and the entire process is repeated. This is continued until a set of initial values ψi (t0 ) and ψN+1 (t0 ) are found that yield the correct final values. Variational Methods for Spatially Dependent Control Problems The basic description of the transient neutron flux and temperature distributions within a nuclear reactor is in terms of partial differential equations. It is not clear that the optimal control computed by first reducing these equations to ordinary differential equations by discretizing the spatial variable and then using the methods above is the same as would be obtained if the optimal control were determined directly from the partial differential equation description of the reactor dynamics. The variational formalism can be extended to the partial differential equation description of the reactor dynamics. The state of the system is specified in terms of state functions yi (r, t) rather than discrete state variables as previously. The function space i , consisting of all possible functions of position that the state function yi can take on, is a component function space, and the product space = 1 ⊗ 2 ⊗ · · · ⊗ N of all such component function spaces is the state function space on which the vector state function y = (y1 , . . . , yN ) is defined. Similarly, the vector control function u = (u1 , . . . , uR ) is defined on the product space of the component function spaces defined by all possible functions of position that the control functions ur can take on. The distance between two states y a and y b is defined as the metric on . Equations for nuclear reactor dynamics can be written in the form u) ẏi (r, t) = Li (r)yi (r, t) + fi (yy ,u yi (r, t0 ) = yi0 (r) yi (R, t) = 0, i = 1, . . . , N (16.165) Control where yi denotes a state function, Li contains a spatial differential operator acting on yi and fi is a spatially dependent function of y and u . The outer boundary of the reactor is denoted by R. These equations may be written in matrix form: u) ẏy (r, t) = L (r)yy (r, t) + f (yy ,u y (r, 0) = y 0 (r), (16.166) y (R, t) = 0 Many control problems may be formulated as the quest for the control vector function u for which the solution of Eq. (16.165) minimizes a functional  tf    u] dt dr F y (r, t), ẏy (r, t) (16.167) J [yy ,u t0 V The standard calculus-of-variations formulation of this problem begins by adding Eq. (16.166) to the integrand of Eq. (16.167) with a Lagrange multiplier vector function λ(r, t) = (λi , . . . , λN ):  tf    u) u, λ] = (16.168) J  [yy ,u dt dr F (yy , ẏy ) + λT ẏy − Ly − f (yy ,u t0 V The control functions ur are treated in the same fashion as the state functions yi . Next, the variation of J  is required to vanish: %  tf  N  ∂F ∂F  δJ = dt dr δyi + δ ẏi + λi δ ẏi − λi Li δyi ∂yi ∂ ẏi t0 V i=1 − N  j =1  ∂fi ∂fi λi δyj − λi δur ∂yj ∂ur R & =0 r=1 Integration by parts of the terms involving δ ẏi and Li , δyi ,4) and use of the initial conditions δyi (r, t0 ) = 0 leads to  % &  tf  N N   ∂fj ∂F ∂ ∂F +  δyi dt dt − − λ̇i − Li λi − λj δJ = ∂yi ∂t ∂ ẏi ∂yi t0 V j =1 i=1 − R   ∂fi δur + ∂ur λi r=1  + dr V N  i=1  + λi δyi t=tf tf dt t0 N   dr V N  ∂F i=1 Pi (λi ) = 0 ∂ ẏi δyi t=tf (16.169) i=1 In arriving at Eq. (16.169), the adjoint operator L+ i and the bilinear concomitant Pi are defined by the relation   dr λi Li δyi = dr δyi L+ (16.170) i λi + Pi (λi ) V V 4) Commutability of the variational operator δ and the operators ∂/∂t and Li imply an assumption of continuous variations δyi , as does the existence of the integrals involving these terms. 637 638 16 Space–Time Neutron Kinetics δJ  must vanish for arbitrary variations of yi and ur , which requires that the Lagrange multiplier functions satisfy the partial differential equations λ˙i (r, t) = −L+ i (r)λi (r, t) −  + N  λj (r, t) j =1 u) ∂fj (yy ,u ∂yi  ∂F (yy , ẏy ) ∂ ∂F (yy , ẏy ) , − ∂yi ∂t ∂ ẏi the final conditions   ∂F λi + = 0, ∂ y˙i t=tf i = 1, . . . , N i = 1, . . . , N (16.171) (16.172) and the boundary conditions   Pi λi (R, t) = 0, i = 1, . . . , N (16.173) In addition, N  i=1 λi (r, t) u) ∂fi (yy ,u = 0, ∂ur r = 1, . . . , R (16.174) must be satisfied. In this formulation, the ur , as well as the yi , are treated as continuous functions. This imposes artificial restrictions on the ur . In some problems the control is discontinuous. Dynamic Programming for Spatially Continuous Systems Proceeding as above, the dynamic programming formalism is developed by considering the minimum value of the functional of Eq. (16.167) evaluated between a fixed upper limit and a variable lower limit as a function of the lower limit:   S t,yy (r, t) = min  u ∈Ω t tf dt            u r, t dr F y r, t  , ẏy y r, t  ,u (16.175) V In writing Eq. (16.175), the dependence of the integrand upon Eq. (16.166) is shown implicitly, and any constraints on the control vector function are implied by u ∈ Ω. By definition,     S t,yy (r, t) ≤ S t + t,yy (r, t + t)  t+t           u r, t dt  dr F y r, t  , ẏy y r, t  ,u + t V Control For the optimal control, the equality obtains. Approximating the integral over time, this becomes5)      S t,yy (r, t) = min S t + t,yy (r, t + t) u (t)∈Ω(t)  + t     u(r, t) dr F y (r, t), ẏy y (r, t),u (16.176) V Equation (16.176) is the dynamic programming algorithm for the partial differential equation description of reactor dynamics. It is solved retrogressively, with the final condition   S tf ,yy (r, t) = 0 (16.177) which is apparent from the defining Eq. (16.175). Pontryagin’s Maximum Principle for a Spatially Continuous System Using a Taylor’s series expansion   S t + t,yy (r, t + t)    ∂S  = S t,yy (r, t) + t t,yy (r, t) ∂t  N     ∂S  u(r, t) + t dr t,yy (r, t) Li (r)yi (r, t) + fi y (r, t),u ∂yi V i=1 Eq. (16.176) becomes  0= min u (t)∈Ω(t) N   ∂S  t,yy (r, t) + ∂t  i=1 V dr  ∂S  t,yy (r, t) ∂yi   u(r, t) × Li (r)yi (r, t) + fi y (r, t),u     u(r, t) dr F y (r, t), ẏy y(r, t),u + (16.178) V Define the functions ψi (r, t) = −  ∂S  t,yy (r, t) , ∂yi ψN+1 (r, t) = −  ∂S  t,yy (r, t) ∂t i = 1, . . . , N (16.179) (16.180) 5) The minimization in Eq. (16.175) is with respect to the control vector function over the time interval t ≤ t  ≤ tf , whereas the minimization in Eq. (16.176) is with respect to the control vector function evaluated at time t . 639 640 16 Space–Time Neutron Kinetics Then Eq. (16.178) becomes  0 = max u (t)∈Ω(t) ψN+1 (r, t) + N   i=1 V  −   u(r, t), dr ψi (r, t) Li (r)yi (r, t) + fi y (r, t),u *    u(r, t) dr F y (r, t), ẏy y (r, t),u (16.181) V This is the extension of Pontryagin’s maximum principle to the partial differential equation description of the reactor dynamics. When the optimal u ∗ (t) is chosen, variational derivatives of the quantity within the square brackets must vanish. This leads to the boundary conditions   Pi ψi (R, t) = 0 (16.182) where Pi is the bilinear concomitant defined in Eq. (16.170), and to the equations  ∂fi dψj = −L+ ψi j ψj − dt ∂yj N  i=1  N  ∂F ∂ ∂F + + (Li yi + fi ) , ∂yj ∂ ẏi ∂yj j = 1, . . . , N (16.183) i=1  ∂ dψN+1 ψi (Li yi + fi ) =− dt ∂t N (16.184) i=1 Identities similar to those just before Eq. (16.162) have been used in arriving at these equations. Appropriate final conditions for the ψi and ψN+1 are ψ1 (tf ) = · · · = ψN (tf ) = ψN+1 (tf ) = 0 (16.185) The optimal control functions must be found by solving Eqs. (16.166) and (16.182) to (16.184). The initial conditions associated with the yi and the final conditions associated with the ψi and ψN+1 produce a system of equations that must, in general, be solved iteratively. This formulation allows discontinuous control functions and can incorporate constraints on the control functions readily, which are its principle advantages with respect to the calculus-of-variations-formulation. 16.5 Xenon Spatial Oscillations 16.5 Xenon Spatial Oscillations Xenon-135, with a thermal absorption cross section of 2.6 × 106 barns and a halflife against β-decay of 9.2 h, is produced by the fission product decay chain fission(235 U) β− 6.1% 135 Te 0.2% <1 min β− 135 I 6.7 h (n,γ ) 135 Xe β− 9.2 h The instantaneous production rate of 135 Xe depends on the 135 I concentration and hence on the local neutron flux history over the past 50 h or so. On the other hand, the destruction rate of 135 Xe depends on the instantaneous flux through the neutron absorption process and on the flux history through the 135 Xe decay process. When the flux is suddenly reduced in a reactor that has been operating at a thermal flux level >1013 n/cm2 · s, the xenon destruction rate decreases dramatically while the xenon production rate is initially unchanged, thus increasing the xenon concentration. The xenon concentration passes through a maximum and decreases to a new equilibrium value as the iodine concentration decays away to a new equilibrium value (see Section 6.2). When a flux tilt is introduced into a reactor, the xenon concentration will initially increase in the region in which the flux is reduced, and initially decrease in the region of increased flux, for similar reasons. This shift in the xenon distribution is such as to increase (decrease) the multiplication properties of the region in which the flux has increased (decreased), thus enhancing the flux tilt. After a few hours the increased xenon production due to the increasing iodine concentration in the high-flux region causes the high-flux region to have reduced multiplicative properties, and the multiplicative properties of the low-flux region increase due to the decreased xenon production associated with a decreasing iodine concentration. This decreases, and may reverse, the flux tilt. In this manner it is possible, under certain conditions, for the delayed xenon production effects to induce growing oscillations in the spatial flux distribution. Such oscillations were common in the large production reactors at Hanford and Savannah River, and measures are required to control them in most thermal power reactors. Because of the time scale of the iodine and xenon dynamics, prompt and delayed neutron dynamics may be neglected (i.e., changes in the neutron flux are assumed to occur instantaneously, and the delayed neutron precursors are assumed to be always in equilibrium). Moreover, 135 I can be assumed to be formed directly from fission. The appropriate equations are 641 642 16 Space–Time Neutron Kinetics g g ∇ · D g (r, t)∇φ g (r, t) − a (r, t) + s (r, t) + σxG X(r, t)δg,G φ g (r, t) + G  g  →g s  g (r, t)φ g (r, t) + χp g  =1 γi G  G  g  =1 g νf (r, t)φ g (r, t) = 0, g = 1, . . . , G g f (r, t)φ g (r, t) − λi I (r, t) = I˙(r, t) (16.186) (16.187) g=1 γx G  g f (r, t)φ g (r, t) + λi I (r, t) − λx X(r, t) − σxG X(r, t)φ G (r, t) = Ẋ(r, t) g=1 (16.188) In writing these equations it is assumed that the xenon absorption cross section g is zero except in the thermal group (g = G). The absorption cross section, a , does G not include xenon. The quantity σx is the microscopic absorption cross section of xenon for thermal neutrons, γ and λ denote yields and decay constants, and I and X are the iodine and xenon concentrations. Changes in the macroscopic cross sections and diffusion coefficients are due to control rod motion or temperature feedback. Linear Stability Analysis One of the features of Eqs. (16.186) to (16.188) that makes their solution by analytical methods difficult is the nonlinearity introduced by the xenon absorption term (implicit nonlinearities are also introduced by the dependence of the cross sections on the flux via the temperature feedback). Linearizing Eqs. (16.186) to (16.188) reduces their complexity but also reduces their applicability to a small region about the equilibrium point. The linearized equations are used principally for investigations of stability; that is, if a small flux tilt is introduced, will this flux tilt oscillate spatially with an amplitude that diminishes or grows in time? The linearized equations are obtained by expanding about the equilibrium point, denoted by a zero subscript: g φ g (r, t) = φ0 (r) + δφ g (r, t) I (r, t) = I0 (r) + δI (r, t) X(r, t) = X0 (r) + δX(r, t) making use of the fact that the equilibrium solutions satisfy the time-independent version of Eqs. (16.186) to (16.188), and neglecting terms that are nonlinear in δφ g and δX: g g ∇ · D g (r)∇δφ g (r, t) − a (r) + s (r) + σxG (r)X0 (r)δg,G δφ g (r, t) + G  g  =1 g  →g s  (r)δφ g (r, t) − σxG (r)φ0G (r)δX(r, t)δg,G 16.5 Xenon Spatial Oscillations g + χp G  g  =1 γi G  g  νf (r)δφ g (r, t) = 0, g = 1, . . . , G (16.189) f (r)δφ g (r, t) − λi δI (r, t) = δ I˙(r, t) g (16.190) g=1 γx G  g f (r)δφ g (r, t) + λi δI (r, t) − λx δX(r, t) g=1 − σxG (r)X0 (r)δφ G (r, t) − σxG (r)φ0G (r)δX(r, t) = δ Ẋ(r, t) (16.191) The effect of temperature feedback has been neglected momentarily in writing Eqs. (16.189) to (16.191), in that the time dependence of the cross sections has been suppressed. Feedback effects will be reintroduced later. Upon Laplace transforming the time dependence, Eqs. (16.189) to (16.191) become g g ∇ · D g (r)∇δφ g (r, p) − a (r) + s (r) + σxG (r)X0 (r)δg,G δφ g (r, p) + G  g  →g s  (r)δφ g (r, p) − σxG (r)φ0G (r)δX(r, p)δg,G g  =1 g + χp G  g  =1 γi G  g  νf (r)δφ g (r, p) = 0, g = 1, . . . , G (16.192) g f (r)δφ g (r, p) − (p + λi )δI (r, p) = −δI (r, t = 0) (16.193) g=1 γx G    g f (r)δφ g (r, p) + λi δI (r, p) − p + λx + σxG (r)φ0G (r) δX(r, p) g=1 − σxG (r)X0 (r)δφ G (r, p) = −δX(r, t = 0) (16.194) Equations (16.192) to (16.194) may be written H δyy = δyy 0 where (16.195) ⎡ ⎤ δφ 1 (r, p) .. ⎢ ⎥ ⎢ ⎥ . ⎢ ⎥ δyy (r, p) ≡ ⎢ δφ G (r, p) ⎥ , ⎢ ⎥ ⎣ δI (r, p) ⎦ δX(r, p) ⎡ ⎢ ⎢ δyy 0 ≡ ⎢ ⎢ ⎣ 0 .. . ⎤ ⎥ ⎥ ⎥ 0 ⎥ ⎦ −δI (r, t = 0) −δX(r, t = 0) and H is composed of the coefficient terms on the left side of Eqs. (16.192) to (16.194). 643 644 16 Space–Time Neutron Kinetics The solution of Eq. (16.195) is formally δyy (r, p) = H −1 (r, p)δyy 0 (r, t = 0) (16.196) Thus the solutions of Eqs. (16.192) to (16.194) are related to the initiating perturbations by a transfer function matrix, H −1 . The condition that the solutions diminish6) in time is equivalent to the condition that the poles of the transfer function (thus the roots of H ) lie in the left-half complex plane. The roots of H are the eigenvalues, p, of Eqs. (16.192) to (16.194), with a homogeneous right-hand side. These homogeneous equations are known as the p-mode equations. The p-mode equations generally have complex eigenfunctions and eigenvalues and must be calculated numerically except for the simplest geometries. Numerical determination of the p-eigenvalues requires special codes and has been successful only for slab geometries. For practical reactor models, it is necessary to resort to approximate methods to evaluate the p-eigenvalues. Two methods that have been employed successfully are the μ- and λ-mode approximations. μ-Mode Approximation The μ-mode approximation is motivated by recognition that the only manner in which Eq. (16.192) differs from a standard static diffusion theory problem is through the additional term −σxG φ0G δX in the thermal group balance equation. Using the homogeneous versions of Eqs. (16.193) and (16.194), this term may be written σxG (r)φ0G (r)δφX(r, p) = N(r, p)fG (r)δφ G (r, p) (16.197) where N(r, p) = [1 + η(r)][γx p + λi (γx + γi )]δf (r, p) − η(r)(γx + γi )(p + λi )f0 (r) [1 + (p/λx ) + η(r)][p + λi )(1 + 1/η(r)] (16.198) with η(r) ≡ σxG (r)φ0G (r) λx G  g (r)  δφ g (r, p) f δf (r, p) ≡ fG (r) δφ G (r, p) (16.199) (16.200) g=1 f0 (r) ≡ G  g (r) g  φ0 (r) f g=1 fG (r) φ0G (r) (16.201) 6) The solutions of Eqs. (16.192) to (16.194) have an oscillatory time dependence if the roots of H have an imaginary component. The requirement that these roots lie in the left-half complex plane ensures that these solutions oscillate with a diminishing amplitude. 16.5 Xenon Spatial Oscillations In applications, the quantity δf (r, p) is usually assumed equal to f0 (r). Using these definitions, the p-mode equations [homogeneous versions of Eq. (16.192) to (16.194)] may be written in the equivalent form g g ∇ · D g (r)∇δφ g (r, p) − a (r) + s (r) + σxG (r)X0 (r)δg,G δφ g (r, p) + G  g  →g s  g (r)δφ g (r, p) + χp g  =1 G  g  =1 g = N(r, p)fG (r)δφ (r, p)δg,G , g  νf (r)δφ g (r, p) g = 1, . . . , G (16.202) If N(r, p) is real, the term NfG in Eq. (16.202) is formally like a distributed poison, and Eq. (16.202) can be solved with standard multigroup diffusion theory codes. In general, N(r, p) is complex because the p-eigenvalues are complex. The essential assumption of the μ-mode approximation is that N(r, p) is real. There are two types of μ-mode approximations and they differ in the treatment of the spatial dependence of N(r, p). In the first approximation the spatial dependence is retained explicitly and N(r, p)fG (r) is treated as a distributed poison, in which case Eqs. (16.202) become the standard multigroup criticality equations. A value of p is guessed, N(r, p) is evaluated, and Eqs. (16.202) are solved for the eigenvalue k (1/k multiplies the fission term in the eigenvalue problem). This procedure is repeated until the calculated eigenvalue agrees with the known critical eigenvalue; the corresponding value of p is an approximation to the p-eigenvalue with the largest real part. An alternative μ-mode approximation (and the one that gives rise to the name μ-mode) results when N(r, p) is assumed to be spatially independent: N(r, p) = μ(p) (16.203) In this case, Eqs. (16.202) define an eigenvalue problem for the μ-eigenvalues, which can be solved, with a slight modification to the coding, by conventional multigroup diffusion theory codes. To obtain an estimate of the p-eigenvalue from the calculated μ-eigenvalue requires definitions of effective values of η̄ and f¯0 which account for the spatial dependence of these quantities. In practice, an effective η̄ is usually defined as ∗ η̄ ≡ dr φ0G (r)fG (r)η(r)φ0G (r) ∗ dr φ0G (r)fG (r)φ0G (r) (16.204) an expression that can be motivated by perturbation theory. The asterisk denotes adjoint. Temperature feedback effects are included in the calculation of μ-eigenvalues by perturbation theory. λ-Mode Approximation The λ-mode approximation begins with Eqs. (16.192) to (16.194) and expands the spatial dependence in the eigenfunctions of the neutron balance operator at the 645 646 16 Space–Time Neutron Kinetics equilibrium point (i.e., λ-modes): g g g g ∇ · D g (r)∇ψn (r) − a (r) + s (r) + σxG (r)X0 (r)δg,G ψn (r) + G  g  →g s g (r)ψn (r) + g  =1 G 1 g g g χp νf (r)ψn (r) = 0, kn  g =1 normalized such that  G  G    g g ∗  g g dr χp ψm (r) νf (r)ψn (r) = δm,n g  =1 g = 1, . . . , G (16.205) (16.206) g=1 g∗ where ψm satisfy equations adjoint to Eq. (16.205) with appropriate adjoint boundary conditions. It is convenient to treat thermal feedback explicitly in this approximation by including a power feedback term +αδf (r)fG (r)δφ G (r, p)φ0G (r) on the left side of Eq. (16.192) for group G. When the iodine is eliminated between Eqs. (16.193) and (16.194), and the flux and xenon are expanded in λ-modes, δφ g (r, p) = N  g An (p)ψn (r), g = 1, . . . , G (16.207) n=1 δX(r, p) = N  Bn (p)fG (r)ψnG (r) (16.208) n=1 the biorthogonality relation of Eq. (16.206) may be used to reduce Eqs. (16.192) to (16.194) to a set of 2N algebraic equations in the unknowns An and Bn , with inhomogeneous terms involving spatial integrals containing δX(r, t = 0) and δI (r, t = 0). These equations may be written as a transfer function relation between the inhomogeneous terms R and the column vector A (p) containing the An and Bn : H (p) · R A (p) = Ĥ (16.209) H lie in the left-half complex Again, the condition for stability is that the poles of Ĥ p-plane. When N = 1 in the expansion of Eqs. (16.207) and (16.208), Eq. (16.209) may be reduced to the scalar relation A1 (p) = Ĥ  (p) · R (16.210) Ĥ  (p) = [(p − p1 )(p − p2 )]−1 (16.211) where 16.5 Xenon Spatial Oscillations and  1/2 p1 = −pr + i c − pr2  1/2 p2 = −pr − i c − pr2 (16.212) with     λi η (γi + γx )η 1+ − γx +η − λx  1+β    η(yi + yx ) η 1− c = λi λx (1 + η) +  1+β pr = λx 2 The parameters η, , and β, which characterize the reactor in this formulation, are defined as η≡ ≡ β≡ 1 λx δf dr ψ1G∗ (r)σxG (r)fG (r)φ0G (r)ψ1G (r) (16.213) dr ψ1G∗ (r)fG (r)ψ1G (r) 1/k1 − 1/k0 − dr ψ1G∗ (r)fG (r)ψ1G (r) dr ψ1G∗ (r)α(r)fG (r)φ0G (r)ψ1G (r) dr ψ1G∗ (r)fG (r)ψ1G (r) δf (γi + γx ) dr ψ1G∗ (r)fG (r)φ0G (r)ψ1G (r) dr ψ1G∗ (r)λx X0 (r)ψ1G (r) −1 (16.214) (16.215) The quantity δf was defined previously as the ratio of the total fission rate to the thermal group fission rate and an effective spatially independent value has been assumed. The fundamental and first harmonic λ-eigenvalues are denoted by k0 and k1 , respectively. The requirement that the poles of Ĥ  (p) lie in the left-half complex p-plane (i.e., that pr > 0) defines a relationship among η,  and β. In practice, β  η has been found to be a good approximation, so that the stability requirement defines a curve in the η– phase plane, as shown in Fig. 16.1. The effect of physical parameters upon xenon spatial stability can be traced through Eqs. (16.213) and (16.214) and Fig. 16.1. The quantity  is primarily determined by the eigenvalue separation 1/k1 –1/k0 . A reactor becomes less stable when the eigenvalue separation decreases, which occurs when the dimensions are increased, when the migration length is decreased, or when the power distribution is flattened. A negative power coefficient (α < 0) increases , thus making a reactor more stable. The quantity η is proportional to the thermal flux level, φ0G . An increase in thermal flux level is generally destabilizing (increasing η), but may be stabilizing if α < 0 (increasing ); that is, for α < 0, an increase in thermal flux moves the point characterizing a given reactor in Fig. 16.1 to the right and up. It is interesting that an increase in thermal flux level can, under some circumstances, be stabilizing, although this is not generally the case. If  > γi , Fig. 16.1 predicts stability independent of the value of η. Physically,  is a measure of the reactivity required to excite the first harmonic λ-mode in 647 648 16 Space–Time Neutron Kinetics Fig. 16.1 λ-mode linear xenon stability criterion. PWR results: open square, calculated with feedback; solid square, calculated, no feedback; open triangle, inferred from experiment. Calculated transients: open circle, decaying oscillation; cross, neutral oscillation; solid circle, growing oscillation. (From Ref. 9(c); used with permission of Academic Press.) the presence of power feedback, and γi is a measure of the maximum reactivity that can be introduced by iodine decay into xenon. The parameters η and  can be evaluated using standard multigroup diffusion theory codes. A fundamental λ-mode flux and first harmonic λ-mode flux and adjoint calculations are required. The integrals in Eqs. (16.213) and (16.214) may be performed with any code that computes perturbation theory–type integrals. Computation of first harmonic flux and adjoint requires either that the problem is symmetric so that zero flux boundary conditions may be located on node lines or that the Wielandt iteration scheme be employed. Several comparisons with experiment and numerical simulation are indicated in Fig. 16.1. The location of the symbol indicates the prediction of the stability criterion, and the type symbol indicates the experimental or numerical result. At 900 effective full power hours (EFPH), Core 1 Seed 1 of the Shippingport reactor experienced planar xenon oscillations with a doubling time of 30 h. Using this doubling time and the calculated value for η, an experimental  may be inferred that agrees with the calculated  to within 3%. Core 1 Seed 4 of the Shippingport reactor was observed to be quite unstable at 893 EFPH, and to be slightly unstable at 1397 EFPH. These observations are consistent with the predictions of the stability criterion at 1050 EFPH. The finite-difference approximations to Eqs. (16.186) to (16.188) were solved numerically for a variety of two-dimensional three-group reactor models. These same reactor models were evaluated for stability with the λ-mode stability criterion. The results depicted in Fig. 16.1 indicate that the predictions of the stability criterion were generally reliable. In the analysis of this section the total power was assumed to be held constant and the effects of nonlinearities and control rod motion on the stability were ne- 16.5 Xenon Spatial Oscillations glected. Although the effects of xenon dynamics upon the total power in an uncontrolled reactor can be evaluated, most reactors can be controlled to yield a constant power output. The treatment of nonlinearities and control rod motion is discussed next. Nonlinear Stability Criterion The extended methods of Lyapunov, which were discussed in Section 16.4, are applied to derive a stability criterion which includes the nonlinear terms that were neglected in the preceding section. Employing a one-group neutronics model and retaining the prompt neutron dynamics and expanding the flux, iodine, and xenon about their equilibrium states, the equations governing the reactor dynamics may be written in matrix form as ẏy (r, t) = L (r)yy (r, t) + g (r, t) where (16.216) ⎡ ⎤ ⎡ ⎤ δφ(r, t) vσx δXδφ + αvf (δφ)2 ⎦ y (r, t) ≡ ⎣ δX(r, t) ⎦ , g (r, t) ≡ − ⎣ σx δXδφ δI (r, t) 0 ⎤ ⎡ −vσx φ0 0 (v∇ · D∇ − va + vνf − vσx X0 − αvf φ0 ) L (r) ≡ ⎣ −(λx + σx φ0 ) λi ⎦ γx f − σx X0 γi f 0 −λi (16.217) where v is the neutron speed, α is the power feedback coefficient, and the other notation is as defined previously. A Lyapunov functional may be chosen as  1 dr y T (r, t)yy (r, t) (16.218) V [yy ] = 2 R The condition for stability (asymptotic stability) in the sense of Lyapunov is that V̇ evaluated along the system trajectory defined by Eq. (16.216) is negative semidefinite (definite).  1 dr ẏy Ty + y Tẏy V̇ = 2 R     ∗ 1 T = dr y L + L y + dr g Ty 2 R R  1/2  1/2  ≤ −μ dr y Ty + dr g Tg dr y Ty (16.219) R R where μ is the smallest eigenvalue of   ∗ 1 2 L + L ϕ n = −μn ϕ n R (16.220) 649 650 16 Space–Time Neutron Kinetics Thus the condition for stability is μ≥ ( ( g Tg )1/2 T 1/2 R dr y y ) R dr (16.221) For a given reactor model and equilibrium state, characterized by μ, relation (16.221) defines the domain of perturbations for which a stable response will be obtained. For asymptotic stability, the inequality must obtain in relation (16.221). The linear eigenvalue problem, Eq. (16.220), which must be solved for μ, involves the matrix L of Eq. (16.217) and its Hermitean adjoint L∗ . The matrix operator 1 L∗ 2 (L + L ) is self-adjoint with a spectrum of real eigenvalues and a complete set of orthogonal eigenfunctions. The foregoing choice of Lyapunov functional is not unique. As a consequence, this type of analysis provides sufficient, but not necessary, conditions for stability. Control of Xenon Spatial Power Oscillations Inclusion of the control system in a stability analysis is difficult primarily because of the difficulty encountered in analytically representing the motion of discrete control rods required to maintain criticality. Control rod motion has a profound effect on the transient response to a perturbation in the equilibrium state in many cases, however, and neglect of this effect may invalidate the stability analysis completely. Variational Control Theory of Xenon Spatial Oscillations When the spatial dependence is represented by the nodal approximation, a general optimality functional may be written (for a M-node model) J [φ1 , . . . , φM , u1 , . . . , uM ] = M   tf m=1 t0 + , dt [φm (t) − Nm (t)]2 + Ku2m (t) (16.222) where φm and Nm represent the actual and the desired, respectively, timedependent fluxes in node m, um is the control in node m, and K is a constant that can be varied to influence the relative importance of the two types of terms in the optimality functional. The purpose of the control program is to find the um (t) that minimizes the optimality functional, subject to the constraints that the reactor remain critical, 0= M  Imm [φm (t) − φm (t)] + [νf m − am − σx Xm (t) − um (t)]φm (t) m =m = f1m (16.223) and the iodine and xenon dynamics equations are satisfied, I˙m (t) = γi f m φm (t) − λi Im (t) = f2m , Im (0) = Im0 (16.224) 16.5 Xenon Spatial Oscillations Ẋm (t) = γx f m φm (t) + λm Im (t) − [λx + σx φm (t)]Xm (t) = f3m Xm (0) = Xm0 , m = 1, . . . , M (16.225) The m subscript denotes node m and Imm is the internodal coupling coefficient of the type discussed in Sections 15.2 and 15.3. Equations (16.148) become 0 = 2[φm (t) − Nm (t)] − [νf m − am − σx Xm (t) − um (t)]ω1m (t) + M  Imm [ω1m (t) − ω1m (t)] − γi f m (t)ω2m (t) m =m − [γx f m − σx Xm (t)]ω3m (t) ω̇2m (t) = λi [ω2m (t) − ω3m (t)] (16.226) (16.227) ω̇3m (t) = σx φm (t)ω1m (t) + [λx + σx φm (t)]ω3m (t), m = 1, . . . , M (16.228) (The symbol ω has been used to denote the Lagrange multipliers, since λ is conventionally used to represent the decay constants.) The final conditions corresponding to Eqs. (16.150) are ω2m (tf ) = 0, ω3m (tf ) = 0, m = 1, . . . , M (16.229) Equations (16.149) are modified somewhat in this case because the optimality functional depends on the control. The more general relation is  ∂F ∂fi u) = 0, + λi (t) (yy ,u ∂ur ∂ur N r = 1, . . . , R (16.230) m = 1, . . . , M (16.231) i=1 which becomes 2Kum (t) + ω1m (t)φm (t) = 0, Equations (16.231) can be used to eliminate the um from Eqs. (16.223) and (16.226). The modified equations, plus Eqs. (16.224), (16.225), (16.227), and (16.228), constitute a set of 6M equations which, together with the initial and final conditions specified above, can be solved for the optimal flux, iodine, xenon, and Lagrange multiplier trajectories. The optimal control can then be determined from Eqs. (16.231). If no approximation is made for the spatial dependence, an equivalent optimality functional is   tf dr dt (φ(r, t) − N(r, t))2 + Ku2 (r, t) (16.232) J [φ, u] = V t0 and the constraints are 0 = ∇ · D(r)∇φ(r, t) + [νf (r) − a (r) − σx X(r, t) − u(r, t)]φ(r, t) = f1 (16.233) 651 652 16 Space–Time Neutron Kinetics I˙(r, t) = γi f (r)φ(r, t) − λi I (r, t) = f2 , I (r, 0) = I0 (r) (16.234) Ẋ(r, t) = γx f (r)φ(r, t) + λi I (r, t) − [λx + σx φ(r, t)]X(r, t) = f3 X(r, 0) = X0 (r) (16.235) Equations (16.171) become 0 = 2[φ(r, t) − N(r, t)] − ∇ · D(r)∇ω1 (r, t) − [νf (r) − a (r) − σx X(r, t) − u(r, t)]ω1 (r, t) − γi f (r)ω2 (r, t) − [γx f (r) − σx X(r, t)]ω3 (r, t) (16.236) ω̇2 (r, t) = λi [ω2 (r, t) − ω3 (r, t)] (16.237) ω̇3 (r, t) = σx φ(r, t)ω1 (r, t) + [λx + σx φ(r, t)]ω3 (r, t) (16.238) The final conditions of Eqs. (16.172) are ω̇2 (r, tf ) = ω3 (r, tf ) = 0 (16.239) and the boundary condition of Eq. (16.173) is ω1 (R, t) = 0 (16.240) Because the optimality functional contains the control functions, Eqs. (16.174) must be modified to  ∂fi ∂F u) = 0 + λi (r, t) (yy ,u ∂ur ∂ur N (16.241) i=1 which becomes 2Ku(r, t) + ω1 (r, t)φ(r, t) = 0 (16.242) 16.6 Stochastic Kinetics The evolution of the state of a nuclear reactor is essentially a stochastic process and should, in general, be described mathematically by a set of stochastic kinetics equations. For most problems in reactor physics it suffices to describe the mean value of the state variables in a deterministic manner and to ignore the stochastic aspects. However, the stochastic features of the state variables are important in the analysis of reactor startups in the presence of a weak source and underlie some experimental techniques, such as the measurement of the dispersion of the number of neutrons born in fission, the Rossi-α measurement, and the measurement and interpretation of reactor noise. The purpose of this section is to present a computationally tractable formalism for the calculation of stochastic phenomena in a spaceand energy-dependent time-varying zero-power reactor model. 16.6 Stochastic Kinetics Forward Stochastic Model The spatial domain of a reactor may be partitioned into I space cells, and the energy range of interest may be partitioned into G energy cells. Subject to this partitioning, the state of the reactor is defined by the set of numbers + , N ≡ nig cim , i = 1, . . . , I ; g = 1, . . . , G; m = 1, . . . , M where nig is the number of neutrons in space cell i and energy cell g, and cim is the number of m-type delayed neutron precursors in space cell i. Define the transition probability P (N  t  |Nt) that a reactor that was in state N  at time t  will be in state N at time t . The probability generating function for this transition probability is defined by the relation        # nig cim G N  t  |U t ≡ P N t |Nt uig υim N (16.243) igm The summation over N implies a summation over all values of nig and cim for all i, g, and m. The quantities uig and υim play the role of transform variables. The transition probability will be written   #   #     Pig N  t  |nig t P̂im N  t  |Cim t P N  t  |Nt = ig (16.244) im for mnemonic reasons. This formalism does not denote product probabilities and is used only to facilitate the distinction between states that differ only by the number of neutrons in one space-energy cell or the number of m-type precursors in one space cell. Some properties of the probability generating function that will be needed in the subsequent analysis are:      G N  t  |U t U =1 = P N  t  |Nt ≡ 1 (16.245) N     ∂G    nig P N  t  |Nt ≡ n̄ig (t) N t |U t |U =1 = ∂uig (16.246) N    ∂G    cim P N  t  |Nt ≡ c̄im (t) N t |U t)|U =1 =  ∂υim N t) ∂ 2 G(N  t  |U nig (t)(nig (t) − 1), = Wig,i  g  (t) ≡ ∂uig ∂ui  g  U =1 nig (t)ni  g  (t), Yim,i  g  (t) ≡ ∂ 2 G(N  t  |U t) ∂υim ∂ui  g  ∂ 2 G(N  t  |U t) Zim,i  m (t) ≡ ∂υim ∂υi  m U =1 U =1 = ni  g  (t)cim (t)  = cim (t)(cim (t) − 1), cim (t)ci  m (t), (16.247) = i  g ig ig = i  g  (16.248) (16.249) im = i  m im = i  m (16.250) 653 654 16 Space–Time Neutron Kinetics The notation U = 1 indicates that the expression is evaluated for all uig and υim equal to unity. The overbar denotes an expectation value, as defined explicitly in Eqs. (16.246) and (16.247). In the foregoing equations and in the subsequent development, the dependence of the expectation values at time t on the state of the reactor at time t  is implicit. By considering the events that could alter the state of the reactor during the time interval t → t + t , balance equations for the transition probability and the probability generating function may be derived. In the limit t → 0, the probability of more than one event occurring during t becomes negligible, and the balance equations can be constructed by summing over all single event probabilities. • Source neutron emission: ∂P ∂t ∂G ∂t =  s  # #  P̂i  m Sig [Pig (nig − 1) − Pig (nig )] Pi  g  ig =  s Sig [uig − 1]G ig • Capture event (includes capture by detectors): ∂P ∂t ∂G ∂t  = c #   # P̂i  m cig [(nig + 1)Pig (nig + 1) − nig Pig (nig )] Pi  g  ig  = c cig [1 − uig ] ig ∂G ∂uig • Transport event: ∂P ∂t ∂G ∂t = T  ig = T i  ig i g lii  [(nig + 1)Pig (nig + 1)Pi  g (ni  g − 1) # #  P̂i  m Pi  g  − nig Pig (nig )Pi  g (ni  g )] g lii  [ui  g − uig ] ∂G ∂uig • Scattering event: ∂P ∂t ∂G ∂t = s  #   gg  sig (nig + 1)Pig (nig + 1) Ki Pig (nig  − 1) Pi  g  g ig = s   ig #  # P̂i  m − nig Pig (nig ) Pi  g  sig  g gg  Ki uig  − uig  ∂G ∂uig 16.6 Stochastic Kinetics • Delayed neutron emission: ∂P ∂t = d   #   g λm (cim + 1)P̂im (cim + 1) χm Pig (nig − 1) Pi  g  g im − cim P̂im (cim ) ∂G ∂t = d  λm  g χm uig #  − υim g im Pi  g  #  P̂i  m ∂G ∂υim • Fission event:    ∂P = f ig (nig + 1)Pig (nig + 1) pg (νp ) ∂t f νp ig  × # #   g   P̂i  m χp Pig (nig  − νp ) Pi  g  1 − β  νp g +   βm νp g χp Pig  (nig  − νp ) · P̂im (cim − 1) g m ×  # Pi  g  #  P̂i  m # #  P̂i  m − nig Pig (nig ) Pi  g  ∂G ∂t = f  f ig g ig ×      ∂fg g  g  fg χp − β  − βm υim uig  χp − uig ∂uig   m g ∂G ∂uig The quantity −ig represents a reaction frequency per neutron, in space cell i and energy cell g, and the subscripts c, s, and f refer to capture, scattering, and  fission, respectively.7) K gg is the probability that a scattering event which occurred g g in energy cell g transfers a neutron to energy cell g  , while χp and χm are the probabilities that a neutron produced by fission and m-type precursor decay, respectively, has energy within energy cell g. The decay constant for precursor type m is λm ,  is the average ratio of the number of m-type precursors to the number of and βm prompt neutrons produced in a fission (β  = m βm ). Sig is the neutron source rate g in space cell i and energy cell g. The quantity lii  represents the frequency per neutron at which neutrons in space cell i and energy cell g will diffuse into space cell i  / (without a change in energy). The prime on the product operator, , indicates that the product is taken over all i, g, and m except those explicitly shown in the same g g 7) For example, f ig = vg f , vg = neutron speed; f = fission cross section. 655 656 16 Space–Time Neutron Kinetics term. The quantity fg is the probability generating function for pg (νp ), which is the probability distribution function for the number of prompt neutrons emitted in a fission that was caused by a neutron in energy cell g: fg (uig  ) =  n  uigig pg (νp ) (16.251) nig  A single fissionable species is assumed for simplicity. Appropriate balance equations for the transition probability, P , and its probability generating function, G, may be constructed from these terms:  ∂P ∂P    N t |Nt = ∂t ∂t S  ∂G ∂G    N t |U t = ∂t ∂t S + + ∂P ∂t + c ∂P ∂t ∂G ∂G ∂t c ∂t + T + T ∂P ∂t ∂G ∂t + s + s ∂P ∂t ∂G ∂t + d + d ∂P ∂t ∂G ∂t (16.252) f (16.253) f Means, Variances, and Covariances By differentiating Eq. (16.253) with respect to uig and vim and evaluating the resulting expressions for U = 1, equations for the mean value of the neutron and precursor distribution, respectively, are obtained [see Eqs. (16.246) and (16.247)]: ∂ n̄ig (t) = Sig (t) − [cig (t) + sig (t) + f ig (t)]n̄ig (t) ∂t + G  g  =1 + M  g g g sig (t)Ki n̄ig  (t) + χp g χm λm c̄im (t) + I  i  =1 m=1 G  g ν̄p f ig  (t)n̄ig  (t) g  =1 g li  i (t)[n̄i  g (t) − n̄ig (t)] (16.254)  g ∂ c̄im (t)  ν̄p f ig (t)n̄ig (t) = −λm c̄im (t) + βm ∂t G g=1 g = 1, . . . , G; i = 1, . . . , I ; m = 1, . . . , M g (16.255) Making use of the identities ν̄p ≡ (1 − β)ν̄ g , where ν̄ g is the average number of neutrons (prompt and delayed) per fission induced by a neutron in energy cell g, and β  = β/(1 − β), it is apparent that these are the conventional space- and energy-dependent neutron and precursor kinetics equations in the finite-difference multigroup approximation. By taking second partial derivatives of Eq. (16.253) with respect to uig and υim , and evaluating the result for U = 1, equations for the quantities defined by Eqs. (16.248) to (16.250) are derived: 16.6 Stochastic Kinetics ∂Wig,i  g  = Sig n̄i  g  + Si  g  n̄ig  − (cig + ci  g  )Wig,i  g  ∂t  g  g + lj i (Wjg,i  g  − Wig,i  g  ) + lj i  (Wjg  ,ig − Wjg,ig ) j + j  g  g  g (sig  Ki g  g  Wi  g,ig + si  g  Ki  − (sig + si  g  )Wig,i  g  +  g + χp  Wi  g  ,ig )  g  g λm χm Yim,i  g  + χm Yi  m,ig m g  ν̄p f ig  Wi  g  ,ig  g  g + χp g  ν̄p f i  g  Wig,i  g  (f ig + f i  g  )Wig,i  g  g   g + χp  g   g  f ig  νp νp − 1 n̄ig  δig,i  ,g  g  i, i  = 1, . . . , I ; g, g  = 1, . . . , G (16.256)  g ∂Yim,i  g  lj i  (Yim,jg  − Yim,i  g  ) = Si  g  c̄im − ci  g  Yim,i  g  + ∂t j +  g  +  m g + χp g  g  si  g  Ki  Yim,i  g  − si  g  Yim,i  g  − λm Yim,i  g  g  λm χm Zim,i  m + βm   g  ν̄p f i  g  Wi  g  ,ig  g  g  ν̄p f i  g  Yim,i  g  − f i  g  Yim,i  g  g  + βm  g  2 νp f i  g  n̄i  g  δi,i  g  i, i  = 1, . . . , I ; m = 1, . . . , M; g  = 1, . . . , G (16.257)  g ∂Zim,i  m  ν̄p f i  g Yim,i  g = −λm Zim,i  m − λm Zim,i  m + βm ∂t g  + βm   g g ν̄p f ig Yi  m ,ig , i, i  = 1, . . . , I ; m, m = 1, . . . , M (16.258) Equations (16.256) to (16.258) are coupled. From Eqs. (16.248) to (16.250) it is apparent that the solutions of Eqs. (16.256) to (16.258) are related to the variances and covariances of the neutron and precursor 657 658 16 Space–Time Neutron Kinetics distributions; for example, 2 σig ≡ (nig − n̄ig )2 = Wig,ig − n̄ig (n̄ig − 1) (16.259) 2 ≡ (cim − c̄im )2 = Zim,im − c̄im (c̄im − 1) σim (16.260) 2 σigm ≡ (nig − n̄ig )(cim − c̄im ) = Yim,ig − n̄ig c̄im (16.261) Correlation Functions Define the correlation functions    nig (t)ni  g  (t  ) ≡ nig ni  g  P N  t  |Nt cim (t)ni  g  (t  ) ≡ nig (t)ci  m (t  ) ≡ cim (t)ci  m (t  ) ≡ N N N N N N N N    (16.262)   cim ni  g  P N  t  |Nt (16.263)   nig ci  m P N  t  |Nt (16.264)   cim ci  m P N  t  |Nt (16.265) By differentiating Eqs. (16.262) to (16.265) with respect to t , and using Eqs. (16.245) to (16.248), (16.254), and (16.255), equations satisfied by the correlation functions may be obtained. ∂ nig (t)ni  g  (t  ) = Sig (t)n̄i  g  (t) − [cig (t) + Tig (t) + sig (t) ∂t M  g λm χm cim (t)ni  m (t  ) + f ig (t)]nig (t)ni  g  (t) + m=1 + G  g  =1 g  g sig  (t)Ki g g  (t) + χp ν̄p f ig  (t) × nig  (t)ni  g  (t  ) (16.266)  g ∂  cim (t)ni  g  (t  ) = βm ν̄p f ig (t)nig (t)ni  g  (t  ) − λm cim (t)ni  g  (t  ) ∂t g=1 (16.267) G G  ∂ g  g g g  nig (t)ci  m (t  ) = Sig (t)c̄i  m (t  ) + sig  (t)Ki (t) + χp ν̄p f ig  (t) ∂t  g =1 × nig  (t)ci  m (t) − [cig (t) + Tig (t) + sig (t) + f ig (t)]nig (t)ci  m (t  ) + M  m=1 g λm χm cim (t)ci  m (t  ) (16.268) 16.6 Stochastic Kinetics  g ∂  cim (t)ci  m (t  ) = βm ν̄p f ig (t)nig (t)ci  m (t  ) − λm cim (t)ci  m (t  ) ∂t G g=1  i, i = 1, . . . , I ; m, m = 1, . . . , M (16.269) Equations (16.266) and (16.267) are coupled, as are Eqs. (16.268) and (16.269). The operator Tig is defined by the operation Tig nig ni  g  = I  i  =1 g li  i (ni  g ni  g  − nig ni  g  ) Physical Interpretation, Applications, and Initial and Boundary Conditions If all members of a large ensemble of identical reactors are known to be in an identical state, N  , at time t  , if all reactors are operated identically subsequent to time t  , and if the state, N , of each reactor could be determined at a later time t , the number of reactors in the ensemble that would be found to have a given state, N , would approach the distribution P (N  t  |Nt). Thus n̄ig (t) is the ensemble average for the number of neutrons in space cell i and energy cell g at time t , and similarly, c̄im (t) is the ensemble average for the number of m-type precursors in space cell i at time t . Alternatively, consider a single reactor that is brought to a known state N  at a reference time t  , and subsequently operated in a given manner until a reference time t . If this procedure is repeated a large number of times, the distribution of the number of times the reactor is in a given state N at reference time t approaches P (N  t  |Nt). Consequently, n̄ig (t) and c̄im (t) are mean values of the neutron and 2 (t) and σ 2 (t) are the mean-squared deviations in precursor populations, and σig im these populations. These deviations are an indication of the uncertainty associated with the usual assumption that the actual population is equal to the mean value, the latter being predicted by conventional kinetics equations. Such considerations are important in analyzing weak-source startup problems. A set of initial conditions for Eqs. (16.254) to (16.258) may be obtained from the identity   (16.270) P N 0 t0 |Nt0 = δNN0 where the zero superscript indicates the known state at t0 . From Eqs. (16.243) and (16.245) to (16.250), the following initial conditions may be deduced: 0   # n0ig cim G N 0 t0 |U t0 = uig uim (16.271) igm n̄ig (t0 ) =  ∂G  0 N t0 |U t0 ∂uig U =1 c̄im (t0 ) =  ∂G  0 N t0 |U t0 ∂νim U =1 = n0ig (16.272) 0 = cim (16.273) 659 660 16 Space–Time Neutron Kinetics  Wig,i  g  (t0 ) =   n0ig n0ig − 1 , n0ig n0i  g  , 0 Yim,i  g  (t0 ) = n0i  g  cim   0  0 cim cim − 1 ,   Zi m ,im (t0 ) = 0 0 cim ci  m , i  g  = ig i  g  = ig (16.274) (16.275) i  m = im i  m = im (16.276) In practice, it is not possible to ascertain the “known” initial conditions. This difficulty may be circumvented by using homogeneous initial conditions and, in a subcritical system, taking the asymptotic solution of Eqs. (16.254) to (16.258) as the initial conditions for further calculations involving changes in operating conditions. Alternatively, the time-independent versions of Eqs. (16.254) to (16.258) may be solved to provide initial conditions. External boundary conditions may be treated by assuming that the space cells on the exterior of the reactor are contiguous to a fictitious external space cell in which the mean value, variance, or covariance is zero, for the purpose of evaluating the net leakage operator. This is equivalent to the familiar extrapolated boundary condition of neutron-diffusion theory. The interpretation of P (N  t  |Nt) just discussed leads to an interpretation of the correlation functions. For example, nig (t)ni  g  (t  ) is the expectation (mean) value of the product of the number of neutrons in space cell i  and energy cell g  at t  , and the number of neutrons in space cell i and energy cell g at t . When the reactor properties are time independent, the ensemble average may be replaced by an average over time in a single reactor (the ergodic theory).8) In this case, nig (t  + τ )ni  g  (t  ) is amenable to experimental measurement if the energy and space cells are chosen to conform with the detector resolution. The corresponding theoretical quantity is obtained by solving the time-independent versions of Eqs. (16.266) using the same type of external boundary treatment discussed before, and employing corrections for the detection process and counting circuit statistics. Numerical Studies Equations (16.254) to (16.258) have been solved numerically for the special case of one energy cell, one delayed neutron precursor type, and one spatial dimension, to study the characteristics of the neutron and precursor distributions under a variety of static and transient conditions. The results of these studies may be characterized in terms of the mean value of the neutron (n̄i ) and precursor (c̄i ) distributions in region i and in terms of the relative variances in the neutron and precursor distributions in region i, which are defined by the relations μi ≡ (ni − n̄i )2 Wi,i − n̄i (n̄i − 1) = n̄2i n̄2i 8) For a subcritical reactor. (16.277) 16.6 Stochastic Kinetics εi ≡ (ci − c̄i )2 Zi,i − c̄i (c̄i − 1) = c̄i2 c̄i2 (16.278) The quantities μi and εi are measurements of the relative dispersion in the neutron and precursor statistical distributions in region i. Certain general trends emerge from the numerical studies that have been performed: 1. When the reactor is subcritical, the asymptotic values of μi and εi vary from region to region, and within a given region εi < μi . 2. When the reactor is subcritical, the asymptotic values of μi and εi depend on the source level and distribution and the degree of subcriticality. In general, increasing the source level or the multiplication factor reduces μi and εi . 3. When the reactor is supercritical, μi and εi attain asymptotic values that are identical in all regions, and μi = εi . 4. When the reactor is brought from a subcritical to a supercritical configuration, μi generally decreases and εi generally increases. 5. The asymptotic value of μi and εi in a supercritical reactor is sensitive to the manner in which the reactor is brought supercritical. a. For the withdrawal of a single rod (or group of rods) between fixed limits, the more rapid the withdrawal the larger the asymptotic value of μi and εi . b. When a number of rods are to be withdrawn, each rod at the same rate, withdrawing the rods on one side of the reactor and then withdrawing the rods on the other side of the reactor results in a larger asymptotic value for μi and εi than if all the rods are withdrawn simultaneously. c. Withdrawing a rod (group of rods) from position a to position c, then reinserting it (them) to position b (a > b > c) results in a larger asymptotic value of μi and εi than if the rod (group of rods) was withdrawn at the same rate from position a to position b. 6. The time at which μi and εi obtain an asymptotic value may differ from region to region, particularly if flux tilting is significant. 7. When the reactor is brought from a subcritical to a supercritical configuration, the asymptotic value of μi and εi depends on the source level and the initial subcritical multiplication factor. 8. The more supercritical the configuration obtained before μi and εi attain their asymptotic value, the larger this asymptotic value is. 661 662 16 Space–Time Neutron Kinetics 9. For a supercritical reactor, μi and εi generally attain their asymptotic value when n̄i is of the order of 105 n/cm3 . In a subcritical reactor, the neutron fluctuations are governed by fluctuations in the neutron sources, which are the instantaneous natural and neutron-induced fission rates and delayed neutron precursor decay rates, as well as by the fluctuations of the fission, capture, and diffusion processes. The precursor fluctuations are governed by an integral of the fission fluctuations over several mean lifetimes for the precursors (τmean = λ−1 ). This integral dependence of the precursor fluctuations on the fluctuations in the fission process tends to smooth out the fluctuations in the former relative to fluctuations in the latter:  t  dt  e−λ(t−t ) βf (t, t  )n(r, t  )v, n ∼ 10 Cm (r, t) = t−nτ In a supercritical reactor, the precursor fluctuations still depend on an integral of the fission fluctuations over the last few mean precursor lifetimes. However, the major contribution to the integral now comes from times close to the upper limit of the integral. Thus the precursor fluctuations tend to depend on the instantaneous fission fluctuations. In a supercritical reactor the major source of prompt neutrons very quickly becomes the neutron-induced fission rate. Thus the neutron and precursor fluctuations are governed by fluctuations in the instantaneous fission rates, and it is plausible that these fluctuations are statistically identical.9) In a subcritical reactor in which the relative fission and the capture and diffusion probabilities vary from region to region, it is reasonable to expect the fluctuations in the neutron population to exhibit different statistical characteristics from region to region. Similarly, when the relative absorption and scattering probabilities and the fission spectrum differ for the various energy groups in a subcritical reactor, the fluctuations in the neutron populations in the different energy groups plausibly exhibit different statistical characteristics. It is interesting that in a supercritical reactor the fluctuations in the neutron population exhibit asymptotically the same statistical characteristics at all spatial positions and in all energy groups. From the numerical results, the behavior of the stochastic distribution of the neutron and precursor populations within a reactor can be deduced. In subcritical reactors the stochastic neutron distribution is spatially and energy dependent, and the stochastic precursor distribution is spatially dependent. In general, in a subcritical reactor, the stochastic neutron distribution is more disperse than the stochastic precursor distribution at the same spatial location. In a supercritical reactor, the asymptotic stochastic neutron distribution is space and energy independent and is identical to the asymptotic stochastic precursor distribution. As a reactor is brought from a subcritical to a supercritical configuration, the stochastic neutron distribution generally becomes less disperse, whereas the stochastic precursor distribution becomes more disperse. The dispersion of the asymptotic distribution in a supercritical reactor depends on the manner in which 9) Have the same relative variances. 16.6 Stochastic Kinetics the reactor attains its final configuration as well as on the multiplicative properties of the initial and final configurations and the source level. The dispersion of the asymptotic distribution is more sensitive to changes that are made to the reactor configuration when the mean neutron and precursor densities are small than to later changes made in the presence of larger mean neutron and precursor densities. Startup Analysis The essential problem of the analysis of a reactor startup is determination of the probability that the actual neutron population is within a prescribed band about the mean neutron population predicted by the deterministic kinetics equations. As a specific example, consider a startup excursion that is terminated by a power level trip actuating the scram mechanism. The scram is initiated at a finite time after the trip point is reached, during which time interval the neutron density continues to increase. If the startup procedure consists of shimming out control rods, the principal concern is that the actual neutron population is less than the mean population, in which case the neutron density at which the trip point is reached occurs later, with the reactor being more supercritical and thus on a shorter period than is predicted by the deterministic kinetics equations. Consequently, the power excursion is more severe than would be predicted deterministically. Startup analyses may be separated into two phases, stochastic and deterministic. The first phase is analyzed with stochastic kinetics, and the results are used as initial conditions, with associated probabilities, for the second phase, which is analyzed with deterministic kinetics. Feedback effects generally may be ignored during the stochastic phase. A reasonable time to switch from the stochastic to the deterministic phase is the time at which the neutron and precursor distributions obtain their asymptotic shape. This time may probably be approximated by the time at which μi and εi of Eqs. (16.277) and (16.278) attain their asymptotic value. If the neutron and precursor distributions [i.e., P (N  t  |Nts )] were known at the switchover time ts , the probability that the actual neutron and precursor densities are less than some specified values could be calculated. The asymptotic neutron and precursor distributions in a reactor with large multiplication and no feedback can be approximated by the gamma distribution, which is completely characterized by the mean and variance of the distribution (i.e., by n̄i and μi and c̄i , and εi ). Use of the gamma distribution is suggested theoretically by the fact that the stationary probability distribution of a variate in a stationary multiplicative process approaches a gamma distribution as the multiplication increases without limit, and is justified empirically by the fact that its use in conjunction with a point reactor kinetics model leads to results that are in reasonable agreement with the GODIVA weak-source transient data. The gamma distribution is F (x) dx = r r r−1 −rx dx x e (r) (16.279) 663 664 16 Space–Time Neutron Kinetics where is the gamma function, x the ratio of the actual value of the variate to the mean value of the variate, and r the ratio of the mean value of the variate to the square root of the variance. For example, x= ni , n¯i −1/2 r = μi (16.280) for the monoenergetic model. From Eq. (16.279), the probability that x <  can be computed.  Prob{x < } =  F (x) dx = 1 − 0 in (r, r) (r) (16.281) where in is the incomplete gamma function. This can be written entirely in terms of tabulated functions by using certain identities, Prob{x < } = (r)r e−r M(1, r + 1, r) r (r) (16.282) where M is the confluent hypergeometric function. Based on the results of the stochastic phase, initial conditions for the deterministic phase can be assigned from n0i = n̄i (ts ), ci0 = c̄i (ts ) (16.283) where n̄i and c̄i are the mean values of the neutron and precursor densities at the switchover time, ts . For a given value of , Eq. (16.282) yields the probability that ni (ts ) < n̄i (ts ), ci (ts ) < c̄i (ts ). References 1 J. A. Favorite and W. M. Stacey, “Variational Estimates of Point Kinetics Parameters,” Nucl. Sci. Eng. 121, 353 (1995); “Variational Estimates for Use with the Improved Quasistatic Method for Reactor Dynamics,” Nucl. Sci. Eng. 126, 282 (1997). 2 T. M. Sutton and B. N. Aviles, “Diffusion Theory Methods for Spatial Kinetics Calculations,” Prog. Nucl. Energy 30, 119 (1996). 3 P. Kaps and P. Rentrop, “Generalized Runge–Kutta Methods of Order Four with Step-Size Control for Stiff Ordinary Differential Equations,” Numer. Math. 33, 55 (1979); W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in Fortran: The Art of Scientific Computing, 4 5 6 7 2nd Ed., Cambridge University Press, Cambridge (1992). W. Werner, “Solution Methods for the Space–Time Dependent Neutron Diffusion Equation,” Adv. Nucl. Sci. Technol. 10, 313 (1977). H. L. Dodds, “Accuracy of the Quasistatic Method for Two-Dimensional Thermal Reactor Transients with Feedback,” Nucl. Sci. Eng. 59, 271 (1976). A. F. Henry, Nuclear Reactor Analysis, MIT Press, Cambridge, MA (1975), Chap. 7. D. R. Ferguson, “Multidimensional Reactor Dynamics: An Overview,” Proc. Conf. Computation Methods in Nuclear Engineering, CONF-750413, VI, 49 (1975). Problems 8 D. C. Wade and R. A. Rydin, “An Experimentally Measurable Relationship between Asymptotic Flux Tilts and Eigenvalue Separation,” in D. L. Hetrick, ed., Dynamics of Nuclear Systems, University of Arizona Press, Tuscon, AZ (1972) p. 335. 9 W. M. Stacey, “Space- and EnergyDependent Neutronics in Reactor Transient Analysis,” Reactor Technol. 14, 169 (1971); “Xenon-Induced Spatial Power Oscillations,” Reactor Technol. 13, 252 (1970); Space–Time Nuclear Reactor Kinetics, Academic Press, New York (1969). 10 K. O. Ott and D. A. Meneley, “Accuracy of the Quasistatic Treatment of Spatial Reactor Kinetics,” Nucl. Sci. Eng. 36, 402 (1969); D. A. Meneley et al., “A Kinetics Model for Fast Reactor Analysis in Two Dimensions,” in D. L. Hetrick, ed., Dynamics of Nuclear Systems, University of Arizona Press, Tuscon, AZ (1972). 11 J. Lewins and A. L. Babb, “Optimum Nuclear Reactor Control Theory,” Adv. Nucl. Sci. Technol. 4, 252 (1968). 12 A. A. Fel’dbaum, Optimal Control Systems, Academic Press, New York (1965). 13 L. S. Pontryagin, V. G. Boltyanskii, R. V. Gamknelidze, and E. F. Mishchenko, The Mathematical Theory of Optimum Processes, WileyInterscience, New York (1962). 14 J. Lasalle and S. Lefschetz, Stability by Lyapunov’s Direct Method, Academic Press, New York (1961). 15 R. Bellman, Dynamic Programming, Princeton University Press, Princeton, NJ (1957). 16 J. N. Grace, ed., “Reactor Kinetics” in Naval Reactors Physics Handbook, A. Radkowsky, ed., USAEC, Washington (1964). Problems 16.1 Estimate the relative “tiltiness” of graphite- and H2 O-moderated thermal reactors by estimating 1 − k1 as a function of slab reactor thickness over the range 1 ≤ a ≤ 5 m. Calculate the associated time constant for the tilt to take place due to delayed neutron holdback. 16.2 Derive the orthogonality property given by Eq. (16.11) and the relationship of Eq. (16.12). 16.3 Calculate and plot the delayed neutron holdback time constant τtilt as a function of the ratio of reactor thickness to migration area for a uniform slab reactor. 16.4 Derive the point-kinetics equations from the multigroup diffusion equations. Discuss the physical significance of the point-kinetics parameters. 16.5 Consider a uniform bare slab reactor in one group diffusion theory (D = 1.2 cm, a = 0.12 cm−1 , νf = 0.14 cm−1 ) that is perturbed over the left one-half of the slab by a 1% increase in absorption cross section. Calculate the critical slab thickness and the unperturbed flux distribution. Calculate the generalized adjoint function of Eq. (16.36). Calculate the first-order perturbation theory estimate, the 665 666 16 Space–Time Neutron Kinetics 16.6 16.7 16.8 16.9 16.10. 16.11. 16.12. 16.13. 16.14. variational estimate, and the exact value of the reactivity worth of the perturbation. Numerically integrate the point-kinetics equations for the transient ensuing from the perturbation in Problem 16.5, using the three different reactivity estimates. Use the prompt jump approximation and one group of delayed neutrons (λ = 0.08 s−1 , β = 0.0075). Derive a two-node kinetics model for the slab reactor of Problem 16.5. Numerically integrate the kinetics equations for the transient ensuing from the perturbation. Use the time-integrated method for the integration of one group of delayed neutron precursors and a prompt-jump approximation. Repeat Problem 16.7, but retaining the time derivative in the neutron equations and approximating it by the θ -method. Solve the problem with θ = 0, 0.5, and 1. Assume that the absorption and fission cross sections in each node of Problem 16.7 have power temperature feedback coefficients and that the temperature in each node is determined by a balance between fission heating and conductive cooling. Analyze the linear stability of the two-node model as a function of the feedback coefficient values. It is wished to linearly increase the power in node 1 of the reactor of Problems 16.5 and 16.7 by 25% and in node 2 by 50% over 10 s, by withdrawing separate control rods in nodes 1 and 2, and then to maintain constant power. Determine the time history of the change in control rod cross section in each node which will best approximate this desired power trajectory. Use the prompt-jump approximation and assume one group of delayed neutrons. Construct a Lyapunov functional for the point kinetics equations with one delayed neutron precursor group. What can you say about the stability of these equations? Consider a reactor described by the point kinetics equation with one group of delayed neutron precursors, a conductive heat removal equation, and a temperature coefficient of reactivity αT . Analyze the linear stability of this reactor model. Construct a Lyapunov function for the reactor model of Problem 16.12 and analyze the stability. Carry through the derivation of the λ-mode linear stability criterion for xenon spatial oscillations discussed in Section 16.6. Problems 16.15. Analyze the stability with respect to xenon spatial oscillations of the reactor of Problem 16.5 as a function of equilibrium flux level and power feedback coefficient. Use the λ-mode stability criterion. 16.16. Write a two-node dynamics code for one neutron energy group and one delayed neutron precursor group to solve for the time dependences of the means and variances in the neutron and precursor populations in a low-source startup problem. Use the properties (D = 1.5 cm, f = 0.008 cm−1 , ac = 0.0125 cm−1 ) and (D = 0.1 cm, f = 0.008 cm−1 , c = 0.005 cm−1 ) for two adjacent slab regions of thickness 150 cm each, the delayed neutron parameters β = 0.0075, λ = 0.088 s−1 , and the prompt neutron parameters ν̄p = 2.41, ν̄p (νp − 1) = 3.84. Calculate the startup of the reactor with a source of S = 5 × 102 s−1 in the first regions. 16.17. Calculate the probability that the actual value of the neutron flux is less than 110% of the mean value as a function of μi , the mean-squared variance in the density to the square of the mean value of the density. 667 669 Appendix A Physical Constants and Nuclear Data I. Miscellaneous Physical Constants Avogadro’s number, NA Boltzmann constant, k Electron rest mass, me Elementary charge, e Gas constant, R Neutron rest mass, mn Planck’s constant, h Proton rest mass, mp Speed of light, c 6.022045 × 1023 mol−1 1.380662 × 10−23 J/K 0.861735 × 10−4 eV/K 9.109534 × 10−31 kg 0.5110034 MeV 1.6021892 × 10−19 C 8.31441 J mol−1 /K 1.6749544 × 10−27 kg 939.5731 MeV 6.626176 × 10−34 J/Hz 1.6726485 × 10−27 kg 938.2796 MeV 2.99792458 × 108 m/s II. Some Useful Conversion Factors 1 eV 1 MeV 1 amu 1W 1 day 1 mean year 1 Ci 1K 1.6021892 × 10−19 J 106 eV 1.6605655 × 10−27 kg 931.5016 MeV 1 J/s 86,400 s 365.25 days 8766 h 3.156 × 107 s 3.7000 × 1010 disintegrations/s 8.617065 × 10−5 eV 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 2 3 4 1 Atomic No. Atomic or Mol. Wt. 1.008 18.016 20.030 4.003 6.940 9.013 25.02 10.82 12.011 14.008 16.000 19.00 20.183 22.991 24.32 26.98 28.09 30.975 32.066 35.457 39.944 39.100 40.08 Element or Compound H H2 O D2 O He Li Be BeO B C N O F Ne Na Mg Al Si P S Cl A K Ca Nuclei per Unit Volume (×10−24 ) 5.3† 0.0335‡ 0.0331‡ 2.6† 0.0463 0.1236 0.0728‡ 0.1364 0.0803 5.3† 5.3† 5.3† 2.6† 0.0254 0.0431 0.0602 0.0522 0.0354 0.0389 5.3† 2.6† 0.0134 0.0233 Density (g/cm3 ) 8.9† 1 1.10 17.8† 0.534 1.85 3.025 2.45 1.60 0.0013 0.0014 0.0017 0.0009 0.971 1.74 2.699 2.42 1.82 2.07 0.0032 0.0018 0.87 1.55 0.3386 0.676 0.884 0.8334 0.9047 0.9259 0.939 0.9394 0.9444 0.9524 0.9583 0.9649 0.9667 0.9710 0.9722 0.9754 0.9762 0.9785 0.9792 0.9810 0.9833 0.9829 0.9833 1 − µ̄0 1.000 0.948 0.570 0.425 0.268 0.209 0.173 0.171 0.158 0.136 0.120 0.102 0.0968 0.0845 0.0811 0.0723 0.0698 0.0632 0.0612 0.0561 0.0492 0.0504 0.0492 ξ 0.33 0.66 0.001 0.007 71 0.010 0.010 755 0.004 1.88 20† 0.001 <2.8 0.525 0.069 0.241 0.16 0.20 0.52 33.8 0.66 2.07 0.44 σa 38 103 13.6 0.8 1.4 7.0 6.8 4 4.8 10 4.2 3.9 2.4 4 3.6 1.4 1.7 5 1.1 16 1.5 1.5 3.0 Microscopic Cross Section (barns) σs 38 103 13.6 0.807 72.4 7.01 6.8 759 4.80 11.9 4.2 3.90 5.2 4.53 3.67 1.64 1.86 5.20 1.62 49.8 2.16 3.57 3.44 σt 1.7† 0.022 3.3† 0.02† 3.29 124‡ 73† 103 32† 9.9† 0.000 0.01† 7.3† 0.013 0.003 0.015 0.008 0.007 0.020 0.002 1.7† 0.028 0.010 a III. 2200-m/s Cross Sections for Naturally Occurring Elements [From Reactor Physics Constants, ANL-5800 (1963)] 0.002 3.45 0.449 2.1† 0.065 0.865 0.501 0.346 0.385 50† 21† 20† 6.2† 0.102 0.155 0.084 0.089 0.177 0.043 80† 3.9 0.020 0.070 Macroscopic Cross Section (cm−1 ) s 0.002 3.45 0.449 2.1† 3.35 0.865 0.501 104 0.385 60† 21† 20† 13.5† 0.115 0.158 0.099 0.097 0.184 0.063 0.003 5.6† 0.048 0.080 t 670 Appendix A. Physical Constants and Nuclear Data 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga Ge As Se Br Kr Rb Sr Yt Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn 44.96 47.90 50.95 52.01 54.94 55.85 58.94 58.71 63.54 65.38 69.72 72.60 74.91 78.96 79.916 83.80 85.48 87.63 88.92 91.22 92.91 95.95 98.0 101.1 102.91 106.4 107.88 112.41 114.82 118.70 2.5 4.5 5.96 7.1 7.2 7.86 8.9 8.90 8.94 7.14 5.91 5.36 5.73 4.8 3.12 0.0037 1.53 2.54 5.51 6.4 8.4 10.2 – 12.2 12.5 12.16 10.5 8.65 7.28 6.5 0.0335 0.0566 0.0704 0.0822 0.0789 0.0848 0.0910 0.0913 0.0848 0.0658 0.0511 0.0445 0.0461 0.0366 0.0235 2.6† 0.0108 0.0175 0.0373 0.0423 0.0545 0.0640 – 0.0727 0.0732 0.0689 0.0586 0.0464 0.0382 0.0330 0.9852 0.9861 0.9869 0.9872 0.9878 0.9881 0.9887 0.9887 0.9896 0.9897 0.9925 0.9909 0.9911 0.9916 0.9917 0.9921 0.9922 0.9925 0.9925 0.9927 0.9928 0.9931 0.9932 0.9934 0.9935 0.9937 0.9938 0.9940 0.9942 0.9944 0.0438 0.0411 0.0387 0.0385 0.0359 0.0353 0.0335 0.0335 0.0309 0.0304 0.0283 0.0271 0.0264 0.0251 0.0247 0.0236 0.0233 0.0226 0.0223 0.0218 0.0214 0.0207 0.0203 0.0197 0.0193 0.0187 0.0184 0.0178 0.0173 0.0167 24 5.8 5 3.1 13.2 2.62 38 4.6 3.85 1.10 2.80 2.45 4.3 12.3 6.7 31 0.73 1.21 1.313 0.185 1.16 2.70 22 2.56 149 8 63 2450 191 0.625 24 4 5 3 2.3 11 7 17.5 7.2 3.6 4 3 6 11 6 7.2 12 10 4.3 8 5 7 – 6 5 3.6 6 7 2.2 4 48 9.8 10.0 6.1 15.5 13.6 45 22.1 11.05 4.70 6.80 5.45 10.3 23.3 12.7 38.2 12.7 11.2 4.3 8.2 6.16 9.70 – 8.56 154 11.6 69 2457 193 4.6 0.804 0.328 0.352 0.255 1.04 0.222 3.46 0.420 0.0326 0.072 0.143 0.109 0.198 0.450 0.157 81† 0.008 0.021 0.049 0.008 0.063 0.173 – 0.186 10.9 0.551 3.69 114 7.30 0.021 0.804 0.226 0.352 0.247 0.181 0.933 0.637 1.60 0.611 0.237 0.204 0.134 0.277 0.403 0.141 19† 0.130 0.175 0.112 0.338 0.273 0.448 – 0.436 0.366 0.248 0.352 0.325 0.084 0.132 1.61 0.555 0.704 0.501 1.22 1.15 4.10 2.02 0.937 0.309 0.347 0.243 0.475 0.853 0.298 99† 0.138 0.195 0.160 0.347 0.336 0.621 – 0.622 11.3 0.799 4.04 114 7.37 0.152 (Continued) Appendix A. Physical Constants and Nuclear Data 671 67 68 69 70 64 65 66 63 51 52 53 54 55 56 57 58 59 60 61 62 Atomic No. Atomic or Mol. Wt. 121.76 127.61 126.91 131.30 132.91 137.36 138.92 140.13 140.92 144.27 145.0 150.35 348.70 152.0 352.00 167.26 158.93 162.51 372.92 164.94 167.27 168.94 173.04 Element or Compound Sb Te I Xe Cs Ba La Ce Pr Nd Pm Sm Sm2 O3 Eu Eu2 O3 Gd Tb Dy Dy2 O3 Ho Er Tm Yb III. (Continued) 6.69 6.24 4.93 0.0059 1.873 3.5 6.19 6.78 6.78 6.95 – 7.7 7.43 5.22 7.42 7.95 8.33 8.56 7.81 8.76 9.16 9.35 7.01 Density (g/cm3 ) 0.0331 0.0295 0.0234 2.7† 0.0085 0.0154 0.0268 0.0292 0.0290 0.0290 – 0.0309 0.0128‡ 0.0207 0.0127‡ 0.0305 0.0316 0.0317 0.0126‡ 0.0320 0.0330 0.0333 0.0244 Nuclei per Unit Volume (×10−24 ) 0.9945 0.9948 0.9948 0.9949 0.9950 0.9951 0.9952 0.9952 0.9953 0.9954 0.9954 0.9956 0.974 0.9956 0.978 0.9958 0.9958 0.9959 0.993 0.9960 0.9960 0.9961 0.9961 1 − µ̄0 0.0163 0.0155 0.0157 0.0152 0.0150 0.0145 0.0143 0.0142 0.0141 0.0138 0.0137 0.0133 0.076 0.0131 0.063 0.0127 0.0125 0.0122 0.019 0.0121 0.0119 0.0118 0.0115 ξ 5.7 4.7 7.0 35 28 1.2 8.9 0.73 11.3 46 60 5600 16,500 4300 8740 46,000 46 950 2200 65 173 127 37 σa 4.3 5 3.6 4.3 20 8 15 9 4 16 – 5 22.6 8 30.2 – – 100 214 – 15 7 12 Microscopic Cross Section (barns) σs 10.0 9.7 10.6 39.3 48 9.2 24 9.7 15.3 62 – 5605 16,500 4308 8770 – – 1050 2414 – 188 134 49 σt 0.189 0.139 0.164 95† 0.238 0.018 0.239 0.021 0.328 1.33 – 173 211 89.0 111 1403 1.45 30.1 27.7 2.08 5.71 4.23 0.903 a 0.142 0.148 0.084 12† 0.170 0.123 0.403 0.263 0.116 0.464 – 0.155 0.289 0.166 0.383 – – 3.17 2.7 – 0.495 0.233 0.293 Macroscopic Cross Section (cm−1 ) s 0.331 0.286 0.248 0.001 0.408 0.142 0.642 0.283 0.444 1.79 – 173 211 89.2 111 – – 33.3 30.4 – 6.20 4.46 1.20 t 672 Appendix A. Physical Constants and Nuclear Data Lu Hf Ta W Re OS Ir Pt Au Hg Ti Pb Bi Po At Rn Fr Ra Ac Th Pa U UO2 Np Pu Am 174.99 178.5 180.95 183.86 186.22 190.2 192.2 195.09 197.0 200.61 204.39 207.21 209.0 210.0 211.0 222.0 223.0 226.05 227.0 232.05 231.0 238.07 270.07 237.0 239.0 242.0 9.74 13.3 16.6 19.3 20.53 22.48 22.42 21.37 19.32 13.55 11.85 11.35 9.747 9.24 – 0.0097 – 5 – 11.3 15.4 18.9 10 – 19.74 – † Value has been multiplied by 105 . ‡ Molecules/cm3 . 93 94 95 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 0.0335 0.0449 0.0553 0.0632 0.0664 0.0712 0.0703 0.0660 0.0591 0.0407 0.0349 0.0330 0.0281 0.0265 – 2.6† – 0.0133 – 0.0293 0.0402 0.04783 0.0223‡ – 0.0498 – 0.9962 0.9963 0.9963 0.9964 0.9964 0.9965 0.9965 0.9966 0.9966 0.9967 0.9967 0.9968 0.9968 0.9968 0.9968 0.9970 0.9980 0.9971 0.9971 0.9971 0.9971 0.9972 0.9887 0.9972 0.9972 0.9973 0.0114 0.0112 0.0110 0.0108 0.0107 0.0105 0.0104 0.0102 0.0101 0.0099 0.0098 0.0096 0.0095 0.0095 0.0094 0.0090 0.0089 0.0088 0.0088 0.0086 0.0086 0.0084 0.036 0.0084 0.0083 0.0082 112 105 21 19.2 86 15.3 440 8.8 98.8 380 3.4 0.170 0.034 – – 0.7 – 20 510 7.56 200 7.68 7.6 170 1026 8.000 – 8 5 5 14 11 – 10 9.3 20 14 11 9 – – – – – – 12.6 – 8.3 16.7 – 9.6 – – 113 26 24.2 100 26.3 – 18.8 107.3 400 17.4 11.2 9 – – – – – – 20.2 – 16.0 24.3 – 1036 – 3.75 4.71 1.16 1.21 5.71 1.09 30.9 0.581 5.79 15.5 0.119 0.006 0.001 – – – – 0.266 – 0.222 8.04 0.367 0.169 – 51.1 – – 0.0359 0.277 0.316 0.930 0.783 – 0.660 0.550 0.814 0.489 0.363 0.253 – – – – – – 0.369 – 0.397 0.372 – 0.478 – – 5.07 1.44 1.53 6.64 1.87 – 1.24 6.34 16.3 0.607 0.369 0.256 – – – – – – 0.592 – 0.765 0.542 – 51.6 – Appendix A. Physical Constants and Nuclear Data 673 674 Appendix A. Physical Constants and Nuclear Data IV. 2200-m/s Cross Sections of Special Interest 10 B: 11 B: 135 Xe: 233 U: 235 U: 238 U: 239 Pu: 240 Pu: 241 Pu: 242 Pu: σa = 3837b σa = 0.005 σa = 2.7 × 106 σγ = 49 σγ = 101 σγ = 2.73 σγ = 274 σγ = 286 σγ = 425 σγ = 30 σf = 524 σf = 577 σf = 741 σf = 0.03 σf = 950 σf < 0.2 This appendix is adapted by permission of John Wiley & Sons from James J. Duderstadt and Louis J. Hamilton, Nuclear Reactor Analysis, copyright © 1976 by John Wiley & Sons, Inc. 675 Appendix B Some Useful Mathematical Formulas (1) Solution of First-Order Linear Differential Equations: df + a(x)f (x) = g(x) dx  x   dx  eA(x ) g(x  ) + C , f (x) = e−A(x) (B.1)  A(x) = x dx  a(x  ) (B.2) (2) Differentiation of a Definite Integral: d dx  a(x) b(x) dx  F (x, x  ) = F (x, a) da db − F (x, b) + dx dx  a(x) b(x) dx  ∂F (x, x  ) ∂x (B.3) (3) Representation of Laplacian ∇ 2 in Various Coordinate Systems: (a) Cartesian: ∇2 = ∂2 ∂2 ∂2 + 2+ 2 2 ∂x ∂y ∂z (B.4) (b) Cylindrical: ∇2 = 1 ∂2 1 ∂ ∂ ∂2 r + 2 2+ 2 r ∂r ∂r r ∂θ ∂z (B.5) 676 Appendix B. Some Useful Mathematical Formulas (c) Spherical:   ∂ 1 ∂ 1 ∂ 2 ∂ r + sin θ ∂θ r 2 ∂r ∂r r 2 sin θ ∂θ ∇2 = + (B.6) ∂2 2 ∂φ 2 2 r sin θ 1 (4) Gauss’ Divergence Theorem:   d 3 r ∇ · A = dS ês · A V (B.7) S where ês is the unit vector normal to the surface element dS. (5) Green’s Theorem:    d 3 r ∇φ · ∇ψ = dS φ ês · ∇ψ − d 3 r φ∇ 2 ψ    2  3 2 d r φ∇ ψ − ψ∇ φ = dS ês · (φ∇ψ − ψ∇φ) (B.8) (B.9) (6) Taylor Series Expansion: f (x) = f (x0 ) + (x − x0 )f  (x0 ) + (x − x0 )2  f (x0 ) + · · · 2! (B.10) (7) Fourier Series Expansion: f (x) = ∞  ∞ an sin n=1  nπx 1 nπx + b0 + bn cos l 2 l (B.11) n=1 where 1 an ≡ l  nπx  dx f (x ) sin , l −l l   1 bn ≡ l  l −l dx  f (x  ) cos nπx  l (B.12) This appendix is reprinted by permission of John Wiley & Sons from James J. Duderstadt and Louis J. Hamilton, Nuclear Reactor Analysis, copyright © 1976 by John Wiley & Sons, Inc. 677 Appendix C Step Functions, Delta Functions, and Other Functions C.1 Introduction Consider the discontinuous function !(x) defined by the properties  !(x) = x<0 x≥0 0, 1, (C.1) !(x) is the unit “step function” introduced by Heaviside in his development of operational calculus (now known as integral transform analysis). One can perform numerous operations on !(x). In particular in can be integrated to yield the ramp function  η(x) = x −∞ dx  !(x  ) =  0, x < 0 x, x ≥ 0 (C.2) Let’s try something a bit more unusual by taking the derivative of !(x). Clearly this is ridiculous, because this derivative, call it δ(x), is undefined at x = 0 because !(x) is discontinuous at this point: δ(x) = ! (x) = lim  ε→0   !(x + ε) − !(x) 0, x = 0 = ∞, x =0 ε (C.3) Nevertheless Dirac, Heaviside, and others have made very good use of this strange “function.” To be more specific, the Dirac δ-function, δ(x), has the properties  δ(x − x0 ) = 0, x = x0 ∞, x = x0 ,  ∞ −∞ dx δ(x − x0 ) = 1 (C.4) 678 Appendix C. Step Functions, Delta Functions, and Other Functions In a sense, it resembles a generalization of the Kronecker δ-function  δmn = m = n m=n 0, 1, The most useful property of the Dirac δ-function occurs when it is integrated along with a well-behaved function, say f (x):  dx f (x)δ(x − x0 ) = f (x0 ) (C.5) This property not only is very interesting, but extremely useful in mathematical physics. Unfortunately the proof of this property—and, indeed, all of the theory of such generalized functions—requires a rather potent dose of mathematics. [Such generalized functions are really not functions at all, but rather a class of linear functionals called “distributions” defined on some set of suitable test functions (which are “infinitely differentiable with compact support”).] Fortunately one does not need all of this high-powered mathematics in order to use δ-functions. Only a knowledge of their properties is necessary. C.2 Properties of the Dirac δ-Function A. Alternative Representations δ(x − x0 ) = sin λ(x − x0 ) 1 lim π λ→∞ (x − x0 ) (C.6) δ(x − x0 ) = ε 1 lim π ε→0+ (x − x0 )2 + ε2 (C.7) B. Properties δ(x) = δ(−x) δ(ax) = (C.8) 1 δ(x), |a| δ[g(x)] =  n a = 0 1 |g  (x n )| δ(x − xn ) (C.9) [g(xn ) = 0, g  (xn ) = 0] (C.10) xδ(x) = 0 (C.11) f (x)δ(x − a) = f (a)δ(x − a) (C.12) C.2 Properties of the Dirac δ-Function  δ(x − y)δ(y − a) dy = δ(x − a) δ(x) = 1 2π  ∞ −∞ dk eikx (C.13) (C.14) Actually these properties only make sense when inserted in an integral. For example, property (C.8) really should be interpreted as   dx f (x)δ(x) = dx f (x)δ(−x) = f (0) (C.15) C. Derivatives One can differentiate a δ-function as many times as one wishes. The mth derivative is defined by  ∞ d mf δ [m] (x − a)f (x) dx = (−1)m m (C.16) dx x=a −∞ One can show δ [m] (x) = (−1)m δ [m] (−x)  δ [m] (x − y)δ [n] (y − a) dy = δ [m+n] (x − a) (C.17) x m+1 δ [m] (x) = 0 (C.19) (C.18) Perhaps of more direct use is the application of these properties to the first derivative  ∞ δ  (x)f (x) dx = −f  (0) (C.20) −∞ δ  (x) = −δ  (−x)  δ  (x − y)δ(y − a) dy = δ  (x − a) (C.21) xδ  (x) = −δ(x) (C.23) (C.22) One can generalize the concept of a δ-function to several dimensions. For example, we would define the three-dimensional δ-function by  d 3 r  δ(r − r)f (r ) = f (r) (C.24) Note that we could write this in Cartesian coordinates as δ(r − r ) = δ(x − x  )δ(y − y  )δ(z − z ) (C.25) Such multidimensional δ-functions are of very considerable use in vector calculus. More detailed discussions of the Dirac δ-function and its relatives are found in the following references. 679 680 Appendix C. Step Functions, Delta Functions, and Other Functions References 1 J. W. Dettman, Mathematical Methods in Physics and Engineering, 2nd Edition, McGraw-Hill, New York (1969). 2 M. J. Lighthill, Fourier Analysis and Generalized Functions, Cambridge University Press, Cambridge (1959). 3 A. Messiah, Quantum Mechanics, Vol. I, Wiley, New York (1965), pp. 468–470. This appendix is reprinted by permission of John Wiley & Sons from James J. Duderstadt and Louis J. Hamilton, Nuclear Reactor Analysis, copyright © 1976 by John Wiley & Sons, Inc. 681 Appendix D Some Properties of Special Functions (1) Legendre Functions: (a) Defining equation: (1 − x 2 )f  − 2xf  + l(l + 1)f = 0, l = integer (D.1) (b) Representation: Pl (x) = l 1 dl  2 x −1 l l 2 l! dx (D.2) (c) Properties: P0 (x) = 1, P1 (x) = x, P2 (x) =  1 2 3x − 1 2 1 P3 = (5x 3 − 3x), . . . 2  +1 2 Pl (x)Pl  (x) dx = δll  2l +1 −1 (D.3) (D.4) (d) Recurrence relations:  Pl+1 (x) − xPl (x) = (l + 1)Pl (x) (D.5) (l + 1)Pl+1 (x) − (2l + 1)xPl (x) + lPl−1 (x) = 0 (D.6) (2) Associated Legendre Polynomials: (a) Defining equation:     2  1 − x f − 2xf + l(l + 1) −  m2 f =0 1 − x2 (D.7) (b) Representation:  (m/2) d m Pl (x) Plm (x) = 1 − x 2 dx m (D.8) 682 Appendix D Some Properties of Special Functions (c) Spherical harmonics:  Ylm () = (2l + 1)(l − m)! 4π(l + m)! (1/2) Pl (cos θ)eimφ (d) Properties:  ∗ d Ylm ()Yl  m () = δll  δmm (D.9) (D.10) 4π Pl ( ·  ) = l  4π ∗ Ylm ()Ylm ( ) (2l + 1) (D.11) m=−l (3) Bessel Functions: (a) Defining equation:   x 2 f  + xf  + x 2 − n2 f = 0 (D.12) (b) Solution: Jn (x), Bessel function of first kind Yn (x), Bessel function of second kind (c) Representation: Jn (x) = ∞  k=0  n+2k (−1)k x (k + 1) (k + n + 1) 2 Jn (x) cos(nπ) − J−n (x) Yn (x) = sin nπ (D.13) (d) Hankel functions: Hn(1) (x) = Jn (x) + iYn (x) (D.14) Hn(2) (x) = Jn (x) − iYn (x) (D.15) (4) Modified Bessel Functions: (a) Defining equation:   x 2 f  + xf  − x 2 + n2 f = 0 (D.16) (b) Solution: In (x), modified Bessel function of first kind Kn (x), modified Bessel function of second kind (c) Representation: In (x) = i −n Jn (ix) = i n Jn (−ix) Kn (x) = π n+1 (1) π Hn (ix) = i −n−1 Hn(2) (−ix) i 2 2 (D.17) Appendix D Some Properties of Special Functions (5) Useful Expansions of Bessel Functions for small x: x2 x4 x6 + − + ··· 4 64 2304 J0 (x) = 1 − x x3 x5 − + − ··· 2 16 384   2 x γ + ln J0 (x) + Y0 (x) = π 2   2 x γ + ln J1 (x) − Y1 (x) = π 2 (D.18) J1 (x) = I0 (x) = 1 + (D.19)  x2 + · · · , γ ≡ 0.577216 4  1 x − + ··· x 4 x2 x4 x6 + + + ··· 4 64 2304 x x3 x5 + + + ··· 2 16 384   x x 2 3x 4 I0 (x) + + + ··· K0 (x) = − γ + ln 2 4 128   x 1 x 5x 3 I1 (x) + − − + ··· K1 (x) = γ + ln 2 x 4 64 (a) Asymptotic expansions for large x:   1 ex I0 (x) = √ 1+ + ··· 8x 2πx $ K1 (x) = (D.23) (D.24) (D.25) (D.26)   3 ex 1− + ··· I1 (x) = √ 8x 2πx $ (D.21) (D.22) I1 (x) = K0 (x) = (D.20) (D.27)   1 π −x 1− e + ··· 2x 8x (D.28)   3 π −x 1+ e + ··· 2x 8x (D.29) (b) Recurrence relations: xJn = nJn − xJn+1 = −nJn + xJn−1 (D.30) 2nJn = xJn−1 + xJn+1 (D.31) xIn = nIn + xIn+1 = −nIn + xIn−1 (D.32) xKn = nKn − xKn+1 = −nKn − xKn−1 (D.33) J0 = −J1 , (D.34) Y0 = −Y1 , I0 = I1 , K0 = −K1 683 684 Appendix D Some Properties of Special Functions (c) Integrals:   n n x n Yn−1 (x) dx = x n Yn x Jn−1 (x) dx = x Jn ,  (D.35)  n n x n Kn−1 dx = −x n Kn x In−1 dx = x In , (6) Gamma Function: (a) Definition:  ∞ (z) = dt e−t t z−1 (D.36) (D.37) 0 (b) Properties: (z + 1) = z (z) √ (0) = ∞, (1/2) = π , (1) = 1, (n) = (n − 1)! ..., (D.38) (7) Error Function: (a) Definition: 2 erf(x) = √ π  x dt e−t 2 (D.39) 0 (b) Complementary error function: 2 erfc(x) = 1 − erf(x) = √ π  ∞ dt e−t 2 (D.40) 0 (8) Exponential Integrals: (a) Definition:  ∞ En (x) = dt 1 e−xt , tn  ∞ E1 (x) = dt 1 e−xt = t  ∞ dt x e−t t (D.41) (b) Properties: E0 (x) = e−x x (D.42) En (x) = −En−1 (x) En (x) = (D.43) 1 [e−x − xEn−1 (x)], n−1 E1 (x) = −γ − ln x − ∞  (−1)n x n n=1 nn! n>1 (D.44) (D.45) References References 1 M. Abramowitz and I. Stegun (Eds.), Handbook of Mathematical Functions, Dover, New York (1965). 2 H. Margenau and G. M. Murphy, The Mathematics of Physics and Chemistry, 2nd Ed., Vol. I, Van Nostrand, Princeton, NJ (1956). 3 I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products, 4th Ed., Academic Press, New York (1965). 4 P. M. Morse and H. Feshbach, Methods of Theoretical Physics, Vols. I and II, McGraw-Hill, New York (1953). This appendix is reprinted by permission of John Wiley & Sons from James J. Duderstadt and Louis J. Hamilton, Nuclear Reactor Analysis, copyright © 1976 by John Wiley & Sons, Inc. 685 687 Appendix E Introduction to Matrices and Matrix Algebra E.1 Some Definitions One defines a matrix of order (m × n) to be a rectangular array of m rows and n columns ⎡a 11 ⎢ ⎢a A = ⎢ 21 ⎣ .. . am1 a12 ··· a13 .. . aij ··· ··· ··· a1n ⎤ .. ⎥ . ⎥ .. ⎥ ⎦ . amn (E.1) The matrix elements aij will be identified by subscripts denoting their row i and column j . If the matrix has the same number of rows as columns, it is said to be a square matrix; for example, ⎛ a11 A = ⎝ a21 a31 a12 a22 a32 ⎞ a13 a23 ⎠ a33 (E.2) A diagonal matrix has nonzero elements only along its main diagonal: ⎡ a11 A=⎣ 0 0 0 a22 0 ⎤ 0 0 ⎦ a33 (E.3) A tridiagonal matrix would have nonzero elements only along its central three diagonals: ⎡ a11 ⎢ a21 ⎢ A=⎢ 0 ⎣ .. . a12 a22 a32 .. . 0 a23 a33 .. . 0 0 a34 ⎤ ··· ···⎥ ⎥ ···⎥ ⎦ (E.4) 688 Appendix E. Introduction to Matrices and Matrix Algebra The unit matrix is the diagonal matrix with elements aij = 1, i = j : 1 0 I = 0 .. . 0 ··· 0 ···  1 ··· .. . 0 1 0 .. . (E.5) For two matrices to be equal, each of their matrix elements must be equal:     b11 b12 b13 · · · a11 a12 a13 · · ·   b21 b22 · · ·   A =  a21 a22 · · · = =B .. .. .. . . . (E.6) The transpose of a matrix is obtained by interchanging its rows and columns: AT ij = A (E.7) ji or a 11  a A T =  21  a31 .. . a12 .. . a13 · · · T  a11    a12   =  a13  .. . a21 .. . a31 ···     (E.8) The determinant of a matrix is formed by taking the determinant of the elements of the matrix: a11 A ≡ |A A| = detA a21 a31 .. . a12 .. . a13 ··· (E.9) Of course, the determinant of a matrix is a scalar—that is, just a number. One defines the cofactor of a square matrix for an element aij by deleting the ith row and j th column, calculating the determinant of the remaining array, and multiplying by (−1)i+j :   a11 a12 a13 a14 · · ·  a21 a22 a23 a24 · · ·      A)23 = cof  a31 a32 a33 a34 · · ·  (cofA a   41 a42 a43 a44 · · ·  .. .. .. .. . . . . = (−1)i+j a11 a31 a41 .. . a12 a32 a42 .. . a14 a34 a44 .. . ··· ··· ··· (E.10) E.2 Matrix Algebra We can construct the adjoint or Hermitean conjugate of a matrix by complexconjugating each of its elements and then transposing as  T A† = A∗ or   (aij )† = aj∗i (E.11) For example,  A† = a11 a21 a12 a22 †  = ∗ a11 ∗ a21 ∗ a12 ∗ a22 T  = ∗ a11 ∗ a12 ∗ a21 ∗ a22  A) = 0, then the matrix A is said to If the determinant of a matrix vanishes, det(A A) = 0, the matrix is said to be nonsingular. be singular. If det(A E.2 Matrix Algebra Two matrices of the same order may be added by adding their corresponding elements (the same holds for subtraction):     a11 a12 · · · b11 b12 · · ·    b21  A + B =  a21 +  .. .. . .   a11 + b11 a12 + b12 · · ·   (E.12) =  a21 + b21  .. . In order for matrix multiplication to be possible, the number of columns of the first matrix must equal the number of rows of the second matrix. One then calculates the matrix elements of C = A · B as cij = n  (E.13) aik bkj k=1 or more explicitly   a11 a12 −−a−13− · · · b11 − −  a21 a22 a23 ·−·−  b21 −→ − ·   A ·B = a   31 a32 a33 · · ·   b31 .. .. .. . . .   a11 b11 + a12 b21 + · · ·  a21 b12 + a22 b22 + · · ·  =  .. . b12 b22 b32 .. . b13 b23 b33  ··· ···  ···  (E.14) Notice that matrix multiplication is not commutative—that is, A · B = B · A in general. 689 690 Appendix E. Introduction to Matrices and Matrix Algebra A very important matrix concept is the inverse of a square matrix, A −1 , which is defined by the relation A −1 · A = A · A −1 = I (E.15) The inverse can be calculated as 1 A )T (cofA A| |A A −1 = (E.16) For example, consider   2 1 A= −1 1 Then A| = 3 |A while  T A) = (cofA 1 −1 1 2 T  = 1 −1 1 2  Hence A −1 = 1 3  1 1 −1 2  = 1 3 1 3 − 13  2 3 A) = 0, then it has no inverse. Notice that if a matrix is singular, that is, det(A This appendix is reprinted by permission of John Wiley & Sons from James J. Duderstadt and Louis J. Hamilton, Nuclear Reactor Analysis, copyright © 1976 by John Wiley & Sons, Inc. 691 Appendix F Introduction to Laplace Transforms F.1 Motivation Differential equations play a central role in the description of most scientific phenomena. Moreover, in many cases these phenomena can be approximately described by a particularly simple type of differential equation—namely, those with constant coefficients. In this Appendix we will try to develop one of the most powerful tools for solving such equations: the application of integral transforms, and more specifically, the use of Laplace transforms to solve differential equations. The analogy between the use of transform methods to solve differential equations and the use of logarithms to simplify arithmetic operations is quite striking. Suppose we wish to multiply two complicated numbers a and b together. Then an easy way to do this is to use logarithms a −→ log a a × b −→ “Transform” −→ log a + log b −→ “Invert” −→ e(log a+log b) =a×b That is, by first taking logs we have simplified the original problem, reducing it to a simple sum. This is essentially the idea behind integral transform techniques. Suppose we symbolically represent the transform operation on a function as f (t) −→ f˜(s) Then the idea is to transform the differential equation of interest df + · · · −→ Transform −→ s f˜(s) + · · · −→ Invert −→ f (t) dt In this manner, the integral transform can be used to convert this differential equation into a simpler problem (frequently an algebraic equation) that can then be solved rather easily for the transformed solution. We then must somehow “invert” the transform to obtain the actual solution of interest. 692 Appendix F. Introduction to Laplace Transforms Example. Consider the very simple ordinary differential equation (familiar from prompt neutron reactor kinetics)   dn ρ − n(t) = 0, dt  n(0) = n0 (F.1) Now define the Laplace transform of n(t) as  ∞ ñ(s) = dt e−st n(t) ≡ L{n} (F.2) 0 To transform the ordinary differential equation (F.1), multiply by e−st and integrate over t   ∞  ∞ dn ρ dt e−st dt e−st n(t) = 0 − dt  0 0 or using integration by parts s ñ(s) − n(0) − (ρ/)ñ(s) = 0 but this is now just an algebraic equation which can be easily solved for ñ(s) = n0 s − (ρ/) (F.3) We must now “invert” ñ(s) to find n(t) = L−1 {ñ(s)} (F.4) By noting that + , L e−at =  ∞ dt e−st e−at = 0  1 1 ⇒ L−1 s +a s +a = e−at we find n(t) = n0 L−1  1 = n0 exp[(ρ/)t] s − (ρ/) (F.5) Example. Integral transforms can also be applied to the solution of partial differential equations. Consider, for example, the initial value problem for a nonmultiplying slab in one-speed diffusion theory 1 ∂φ ∂ 2φ = D 2 − a φ(x, t) υ ∂t ∂x Initial condition: φ(x, 0) = φ0 (x) Boundary condition: φ(0, t) = φ(l, t) = 0 (F.6) F.1 Motivation Define the Laplace transform of φ(x, t) with respect to t by  ∞ φ̃(x, s) = dt e−st φ(x, t) (F.7) 0 Now multiplying (F.6) by e−st and integrating over all times t , we find the transformed partial differential equation becomes d 2 φ̃ 1 s φ̃(x, s) − φ(x, 0) = D 2 − a φ̃(x, s) υ dx Since the boundary conditions also depend on time, we must transform them to find: φ̃(0, s) = φ̃(l, s) = 0 Hence if we regard s only as a parameter, the application of Laplace transforms has reduced our original partial differential equation (F.6) to an inhomogeneous ordinary differential equation in x D   d 2 φ̃ s −  + φ̃(x, s) = φ0 (x) a υ dx 2 Boundary condition: φ̃(0, s) = φ̃(l, s) = 0 (F.8) We can now solve this in any of the standard ways (e.g., eigenfunction expansions or Green’s functions) to find φ̃(x, s), and then invert to find + , φ(x, t) = L−1 φ̃(x, s) (F.9) Hence as should be apparent from these simple examples, Laplace transforms can be used to greatly simplify the solution of differential equations by: (a) transforming the original differential equation, (b) solving the transformed equation (which is now presumably a simpler equation such as an algebraic equation or ordinary differential equation) for the transformed solution, and (c) finally inverting the transformed solution to obtain the desired solution of the original equation. It is usually a straightforward task to complete the first two steps. The final step, that of inversion, can frequently be accomplished in a “cookbook” fashion by merely looking up the inverse in a table of Laplace transforms that some other fellow has had to work out. The general theory of how to perform such inversions from scratch is important, however, since the inverses of many of the functions one encounters in practice are not tabulated. However since it is heavily steeped in the theory of functions of a complex variable, we will avoid a detailed discussion of Laplace transform inversion via contour integration here and simply refer the reader to one of several standard texts (see Refs. 1–3). 693 694 Appendix F. Introduction to Laplace Transforms F.2 “Cookbook” Laplace transforms We will now set up the recipes for solving differential equations with Laplace transforms. First we must determine just what types of equations we can consider: (a) This can be any linear differential equation (ordinary or partial) in which the variable to be transformed runs from 0 to ∞. (Such as an initial value problem in time or a half-space problem in space.) (b) We will further restrict ourselves to the study of differential equations with constant coefficients (i.e., the coefficients in the equation do not depend on the variable to which we are applying the transform). This restriction can sometimes be relaxed; however we will not consider the more general problem of differential equations with variable coefficients here. We will define the Laplace transform of a function f (t) by f˜(s) =  ∞ dt e−st f (t) (F.10) 0 There are of course some restrictions on the type of function f (t) and the ranges of values of s for which this integral will be properly defined, but let’s not worry about details at this stage of the game. The general scheme for transforming the differential equation we are interested in solving is the same as before—namely, multiply by e−st and integrate over all t , using liberal integration by parts. One then solves the resulting transformed equation and attempts to invert the solution. To facilitate in the preparation of a table of Laplace transforms (a cookbook), one merely takes the transforms of as many different functions as possible. Several useful transforms of general functions are (see Refs. 4, 5): • Derivatives:  df = s f˜(s) − f (0) (F.11) L dt Recall that we obtained this by integration by parts. Further integration by parts yields  n d f = s n f˜(s) − s n−1 f (0) − s n−2 f  (0) − · · · − f [n−1] (0) L dt n • Integration:  t 1 L dt  f (t  ) = f˜(s) s 0 (F.12) (F.13) F.2 “Cookbook” Laplace transforms Proof:  t L   dt  f t  =  0 ∞ dt e−st 0 = − e−st s  t   dt  f t  0  t   dt  f t  0 ∞ 0 + 1 s  ∞ dt e−st f (t) 0 1 = f˜(s) s • Differentiation by s: L{tf (t)} = − d f˜ ds (F.14) Proof: d f˜ = ds  ∞ dt f (t) 0 d  −st  e =− ds  ∞ dt e−st [tf (t)] 0 • Complex translation: L{eat f (t)} = f˜(s − a) (F.15) Proof:  ∞ dt eat e−st f (t) =  0 ∞ dt e−(s−a)t f (t) = f˜(s − a) 0 • Real translation: L{f (t − a)!(t − a)} = e−as f˜(s) (F.16) where !(t) is the step function,  !(t) = 1, t ≥ 0 0, t < 0 Several examples of more specific transform pairs are presented in Table F.1. Several other very useful relations (see Refs. 4, 5) are: • Convolution theorem:  t L dτ f (t − τ )g(τ ) = t˜(s) = g̃(s) (F.17) 0 (This result is useful for relating the inverse of the product of two transformed functions.) • Initial value theorem: lim f (t) = lim s f˜(s) t→0 s→∞ (F.18) 695 696 Appendix F. Introduction to Laplace Transforms Table F.1 f (t) f˜(s) 1 1 s 1 s+a e−at δ(t) 1 δ(t − t1 ) e−st1 tn t n−1 e−at (n−1)! n! s n+1 1 (s+a)n (e−bt −e−at ) a−b 1 (s+a)(s+b) (be−bt −ae−at ) (b−a) sin at cos at t sin at t cos at !(t) s (s+a)(s+b) a (s 2 +a 2 ) s (s 2 +a 2 ) 2as (s 2 +a 2 )2 (s 2 −a 2 ) (s 2 +a 2 )2 1 s • Final value theorem: lim f (t) = lim s f˜(s) t→∞ s→0 (F.19) There are a number of reasonably complete tables of such transform pairs (see Refs. 4, 5). After obtaining the transformed solution, one can then turn to such tables in an effort to locate the desired inverse. However in many cases it will be necessary to proceed with a direct inversion calculation. References 1 P. M. Morse and H. Feshbach, Methods of Theoretical Physics, Vol. 1, McGraw Hill, New York (1953), Chapter 4. 2 W. Kaplan, Operational Methods for Linear Systems, Addison-Wesley, Reading, MA (1962). 3 H. S. Carslaw and J. C. Jaegar, Operational Methods in Applied Mathematics, Dover, New York (1948). 4 P. A. McCollum and B. F. Brown, Laplace Transform Tables and Theorems, Holt, Rinehart, and Winston, New York (1965). 5 F. E. Nixon, Handbook of Laplace Transforms, Prentice-Hall, Englewood Cliffs, NJ (1960). This appendix is reprinted by permission of John Wiley & Sons from James J. Duderstadt and Louis J. Hamilton, Nuclear Reactor Analysis, copyright © 1976 by John Wiley & Sons, Inc. 697 Index ABH method, see Homogenization Absorption, 26 Absorption probability, 87, 314 Actinides, see Transuranics Adiabatic method, see Point kinetics Acelerator transmutation of waste reactor, 241 Adjoint eigenvalue, 489 function, 486, 491, 496, 498, 501, 508, 563, 567, 571, 580, 588, 603 generalized adjoint function , 606 operator, 486 Albedo boundary condition, 52 diffusion theory, 52 Asymptotic period, 149 measurement, 156 Asymptotic shape, 58 Bare reactors, see Diffusion theory Barn, 6 Bessel functions, 682 Beta decay, see Radioactive decay Bethe–Tait model, 188 Bickley function, 318 Binding energy, 3 Blackness theory, see Homogenization Breeding ratio, see Fuel composition Boltzmann equation, 307 Boundary and interface conditions albedo, see Albedo diffusion theory, see Diffusion theory extrapolated, see Extrapolation distance boundary condition Mark, see Mark boundary conditions Marshak, see Marshak boundary conditions transport theory, see Neutron transport theory Breit–Wigner resonance scattering cross section, 14, 20, 430 Buckling geometric, 57, 59, 147 material, 58 Burnable poison, 208, 250, 253 Cadmium ratio, 72 Capture, 13 Capture-to-fission ratio, α, 33 Center-of-mass system, 27 Central limit theorem, 374 Closing nuclear fuel cycle, 244 Collision probabilities method ABH method, 517 collision probability, 88, 320 collision probability annular geometry, 322 collision probability slab geometry, 320 collision probability two dimensions, 320 pin-cell model, 524, 528 reciprocity, 320 thermalization in heterogeneous lattices, 474 transmission probabilities, 320 Compound nucleus, 5 formation, 13 Control rod cross sections, effective diffusion theory, 73 follower, 257, 297 scram, 250, 257, 285, 295, 297 windowshade model, 76 Control theory dynamic programming, 633 Pontryagin’s maximum principle, 635 variational, 631 698 Index Conversion/breeding ratios, 219. See also Fuel composition Correlation methods, 179 Coupling coefficients, 85 Criticality critical, 37, 58 delayed critical, 151 prompt critical, 153 subcritical, 37, 58, 151 supercritical, 37, 58 superprompt-critical, 152 Criticality condition bare homogeneous reactor, 58, 60 Monte Carlo, 378 power iteration, 80, 138 reflected homogeneous reactor, 65 reflected slab, 64 rodded cylindrical reactor, 77 two region, two-group reactor, 132 Criticality minimum volume, 61 Criticality safety, see Nuclear reactor analysis Cross sections 2200 m/s values, 110, 670 absorption, 45, 670 capture, 14, 24, 201, 670 definition, 5 elastic scattering, 20, 24, 670 evaluated, see Evaluated nuclear data files fission, 5, 24, 201, 670 for important nuclides, 26 low-energy summary, 24 macroscopic, 24 spectrum-averaged, 24, 25, 63 total, 24, 670 transport, see Transport cross section units, 5 Cross spectral density, 180 Current net current, 45, 310 partial current, 45, 310 Delayed critical, 151 Delayed neutrons decay constants, 143 holdback, 599 kernel, 155 neutron kinetics effects, 39, 150 precursor, 147 yield, 143 Densities, elements and reactor materials, 670 Depletion model, 219 Detailed balance principle, 109, 458, 467 Diffusion coefficient, 45, 342 directional, 397 multigroup, 128, 398 Diffusion cooling, 480 Diffusion length, 48, 53, 56 Diffusion parameters, 56 Diffusion theory applicability, 47 bare homogeneous reactor, 57 boundary and interface conditions, 46, 94 derivation, 43, 342 directional, 397 kernels, 50 lethargy-dependent, 396 multigroup theory, 127, 398, 599 nonmultiplying media solutions, 48 numerical solution, 77, 137 one-dimensional geometry, 93, 347 reflected reactors, 62, 134 two-region reactors, 130 Dirac δ-function, 677 Discrete ordinates methods acceleration of convergence, 362 cylindrical and spherical geometries, 359 diamond difference scheme, 358, 360, 367 equivalence with PL equations, 95, 356 level-symmetric quadrature, 365, 574 nodal, 571 ordinates and quadratures, multidimensional, 363 ordinates and quadratures, PL and D-PL , 95, 355 ordinates and quadratures, SN , 365 slab geometry, 94, 354 spatial finite differencing and iteration, SN method in 2D Cartesian geometry, 366 spatial finite differencing and iteration, slab geometry, 357 spatial mesh size limitations, 358 sweeping mesh grid, 360, 367 Doppler broadening, 119. See also Resonance and Reactivity Dynamic programming, 638 Eigenvalue separation, 602, 647 Elastic scattering average cosine of scattering angle, 389 average logarithmic energy loss, 30, 389 cross sections, 22, 670 Index energy–angle correlation, 28, 373 kernel, 29, 102, 386 kinematics, 27, 385 Legendre moments of transfer function, see Legendre moments of elastic scattering transfer function moderating ratio, 30 potential, 20 relation between CM and lab scattering angles, 28 resonance, 20 transfer function, 386 Emergency core cooling, 285, 295 Energy release from fission, 12 Error function, 684 Escape probability, 87 Eta (number of neutrons per absorption in fuel), 34 Equivalence theory, see Homogenization Escape probability, see Integral transport theory; Interface current methods; Resonance Evaluated nuclear data files, 24, 115 Even-parity transport theory, see Neutron transport theory Excitation energy for fission, 4 Exoergic reactions, 286 Exponential integral function, 684 Extrapolation distance boundary condition, 47, 342 Fermi age, see Neutron slowing down Fertile isotopes, 198 Few group approximations, 115, 128 Fick’s law, 45, 342, 353, 397 Finite difference equations diamond difference relation, 358, 360, 367 diffusion equation, one-dimensional slab, 78 diffusion equation, two-dimensional Cartesian, 80 discrete ordinates, rectangle, 366 discrete ordinates, slab, 357 discrete ordinates, sphere, 360 limitations on mesh spacing, 358 Finite element methods cubic Hermite approximation, 569 finite-difference approximation, 564 linear approximation, 567 First collision source, see Integral transport theory Fissile isotopes, 5 Fission cross sections, 6, 201, 670 energy release, 12 fast, 34 neutron chain fission reaction, see Neutron chain fission reaction neutron yield, 8 probability per neutron absorbed, 236 process, 4 products, 8, 197 products significant in accidents, 284 spectrum, 11 spontaneous, 4, 201, 202 threshold, 4 yields, 10 Flux, scalar, 310 Flux tilts, 599, 641 Four-factor formula, 37 Flux disadvantage factor, see Thermal disadvantage factor Fuel assemblies, 68, 251, 254, 257, 259, 261– 263, 522 Fuel burnup composition changes, 205 depletion model, 219 energy extraction, 235 fission products, see Fission products in-core fuel management, 210 reactivity changes, see Reactivity transmutation-decay chains, see Transmutation-decay chains units, 205 Fuel composition discharged UO2 , 205 equilibrium distribution in recycled fuel, 236 fertile-to-fissile conversion and breeding, 217 plutonium buildup, 206 power distribution, 209 reactor-grade uranium and plutonium, 233 recycled LWR fuel, 221 recycled plutonium physics differences, 207, 225 recycled uranium physics differences, 224 weapons-grade uranium and plutonium, 233 Fuel lumping, 37 Fuel recycling, see Fuel composition Fuel reprocessing, 221 699 700 Index Gamma function, 684 Gauss’ divergence theorem, 676 Gauss–Seidel, 82 Generalized perturbation theory, see Variational methods Gaussian elimination, 79 Green’s theorem, 676 Group collapsing, 117, 395, 403 Hazard index, 226. See also Radioactive waste Heterogeneity, see Homogenization High level waste repository decay heat, 243 isotopes, 242 Homogenization ABH method, 517 blackness theory, 520 collision probabilities pin-cell model, 524 conventional theory, 531 cross sections, equivalent homogeneous, 69, 516 diffusion theory, 67 diffusion theory lattice functions F and E, 70 equivalence theory, 531 flux discontinuity factor, 532 flux reconstruction, 538, 554 flux (thermal) disadvantage factor, 69, 516, 518. See also Self-shielding interface current pin-cell method, 527 multiscale expansion theory, 535 pin-cell model, 522, 528 resonance cross sections, 423 spatial self-shielding, see Self-shielding transport boundary conditions, 520, 522 Wigner–Seitz cell, 523 Importance function, 145, 375, 487. See also Adjoint function Infinite multiplication constant, k∞ , 37, 113 Inhour equation, 149 Integral transport theory absorption probability, 314 anisotropic plane source, 312 distributed volumetric scattering and fission sources, 315 escape probability, 314 first-collision source, 315 isotropic line source, 317 isotropic plane source, 311 isotropic point source, 311 probability of traveling a distance t from a line source, 318 scattering and fission, inclusion of, 315 transmission probability, 314. See also Transmission probability Interface current methods boundary conditions, 329 emergent currents, 326, 327, 331 escape probabilities in slab geometry, 328 escape probabilities in two-dimensional geometries, 333, 335 escape probabilities rational approximations, 337 pin-cell model, 527 reflection probability in slab geometry, 328 response matrix, 329 transmission probabilities in slab geometry, 328 transmission probabilities in twodimensional geometries, 333 Iteration methods acceleration of convergence, 362, 369 alternating direction implicit, 619 forward elimination/backward substitution (Gauss elimination), 79 power, for criticality problems, 79, 82, 363, 378, 413 scattering, for discrete ordinates equations, 358, 413 successive over-relaxation, 82, 623 successive relaxation (Gauss–Seidel), 82, 137 sweeping over mesh points for onedimensional discrete ordinates, 360 sweeping over mesh points for twodimensional discrete ordinates, 367 J (ξ, β) resonance function, 125 Lagrange multiplier, 632 Laguerre polynomials, 479 Laplace transforms, 691 Laplacian representation, 675 Legendre moments of elastic scattering transfer function anisotropic scattering in CM, 389 definition, 387 isotropic scattering in CM, 388, 389 Legendre polynomials associated Legendre functions, 339, 681 definition and properties, 338, 681 Index half-angle Legendre polynomials, 347 Lethargy, 385 Loss of coolant accident, see Reactor safety Loss of flow accident, see Reactor safety Lyapunov’s method for stability analysis, 627, 629, 649 Macroscopic cross section, 44 Mark boundary conditions, see Spherical harmonics Marshak boundary condition, see Spherical harmonics Mass defect, 3 Matrix algebra, 687 Maxwellian distribution, 109 Maxwellian energy distribution, 19 Mean chord length, 426 Mean free path, 425 Mesh spacing limit, 83 Migration length, 55 Minimum critical volume, 61 Mixed oxide fuel, 222, 233, 239 Moderator properties, 30 Moderating ratio, see Elastic scattering Monte Carlo methods absorption weighting, 377 analog simulation of neutron transport, 372 correlated sampling, 378 criticality problems, 378 cumulative probability distribution functions, 371 exponential transformation, 376 flux and current estimates, 377 forced collisions, 376 importance sampling, 375 probability distribution functions, 371 Russian roulette, 377 splitting, 377 statistical estimation, 373 variance reduction, 375 Multigroup theory, 127 collision probabilities for thermalization, 476 cross-section definition, 113, 403, 413 cross-section preparation, 115 diffusion theory, 398, 599 few group constants, 117, 395 few group solutions, infinite medium, 114 mathematical properties, 113 one-and-one-half-group diffusion theory, 129 perturbation diffusion theory, 168, 483 pin-cell collision probabilities model, 528 resonance cross sections, see Resonance two-group diffusion theory, 128, 130, 134 Multiplication constant, keff , 37, 60, 129, 137 Neutron balance schematic, 36 Neutron chain fission reaction criticality, 37 delayed neutron effect on, 38 effect of fuel lumping, 37 effective multiplication constant, 37 neutron balance in a thermal reactor, 34 process, 33 prompt neutron dynamics, 38 resonance escape, 36. See also Resonance source multiplication, 39 utilization, 34 Neutron diffraction, 24 Neutron emission, 19 Neutron energy distribution fission energy range analytical solution, 101 multigroup calculation, 111 resonances, 123 slowing-down range analytical solutions, 102 spectra in UO2 and MOX fuel cells, 222 spectra typical for LMFBR and LWR, 41 thermal range analytical solutions, 108 Neutron lifetime, 61 Neutron scalar flux, 44 Neutron slowing down average cosine of scattering angle, 389 average lethargy increase, 389 B1 theory, 394 consistent P1 approximation, 405 continuous slowing down theory, 400 diffusion theory, 127, 397 discrete ordinates, 411 elastic scattering kernel, 386 Fermi age, 107 hydrogen, 103 isotropic CM scattering, 388 Legendre moments, see Legendre moments of elastic scattering transfer function P1 theory, 390 Pl continuous slowing, 407, 410 701 702 Index slowing down density, see Neutron slowing down density weak absorption, 106 without absorption, 104 Neutron slowing down density age approximation, 404 anisotropic scattering, 407 definition, 105, 400 extended age approximation, 405 Grueling–Goertzel approximation, 406 hydrogen, 403 scattering resonances, 409 Selengut–Goertzel approximation, 405 weak absorption, 106 Neutron sources accelerator-spallation, 273 tokamak D–T fusion, 273 Neutron temperature, 109 Neutron thermalization collision probability methods for heterogeneous lattices, 474 differential scattering cross section, 453, 457 effective neutron temperature, 110 energy eigenfunctions of scattering operator, 477 free-hydrogen model, 455 Gaussian representation, 459 heavy gas model, 456, 466 incoherent approximation, 459 intermediate scattering function, 458 measurement of scattering functions, 460 moments expansion, 470 monatomic Maxwellian gas, 454 multigroup calculation, 473 numerical solution, 468 pair distribution function, 457 pulsed neutron, 477 Radkowsky model, 455 scattering function, 457 spatial eigenfunction expansion, 477 thermalization parameters for carbon, 472 Wigner–Wilkins model, 463 Neutron transport equation Boltzmann, 90 integral, 88 Neutron transport theory boundary conditions, 310 collision probabilities, see Collision probabilities methods current, 87, 310 discrete ordinates, see Discrete ordinates methods equation, 305 even-parity, 369, 505 integral, see Integral transport theory interface current, see Interface current methods Monte Carlo, see Monte Carlo methods partial current, 310 scalar flux, 310 spherical harmonics, see Spherical harmonics methods streaming operator in various geometries, 309 Neutron wavelength, 20, 430 Nodal methods, 83 conventional methods, 544 double-Pn expansion, 558 formalism, 542 gross coupling, 545 polynomial expansion, 549, 557 transverse integrated diffusion theory methods, 547 transverse integrated transport theory models, 554 transverse integrated discrete ordinates methods, 561 transverse leakage, 548, 556, 561 variational discrete ordinates methods, 571 Noise analysis, 181 Nonleakage probability, 34, 60, 166 Nu (number of neutrons per fission), 11 Nuclear reactor analysis core operating data, 279 criticality and flux distribution, 276 criticality safety, 279 fuel cycle, 277 homogenized cross sections, 275. See also Homogenization safety, see Reactor safety transient, 278 Nuclear reactors advanced, 269 advanced gas-cooled reactor AGR, 260 boiling water reactor BWR, 250, 299 characteristics of power reactors, 265 classification by coolant, 41 classification by neutron spectrum, 40 high-temperature gas-cooled reactor HTGR, 260 integral fast reactor IFR, 272, 300 Index light water breeder reactor LWBR, 265 liquid-metal fast breeder reactor LMFBR, 261 MAGNOX, 260 molten salt breeder reactor MSBR, 265 pebble bed reactor, 265 pressure tube graphite-moderated reactor RBMK, 258 pressure tube heavy water reactor CANDU, 255 pressurized water reactor PWR (AP-600, PIUS), 249, 299 representative parameters, 266 Nuclear reactors, advanced advanced boiling water reactor (ABWR), 266 advanced liquid metal reactor (ALMR), 264 advanced pressure tube reactor, 268 advanced pressurized water reactor (APWR, EPR, AP-600, AP-1000, APR1400), 267 gas-cooled fast reactor (GFR), 270 generation-IV reactors (GEN-IV), 269 integral fast reactor (IFR), 264, 300 lead-cooled fast reactor (LFR), 271 modular high-temperature gas-cooled reactor (GT-MHR), 268 molten salt reactor (MSR), 271 pebble bed modular reactor (PBMR), 268 sodium-cooled fast reactor (SFR), 272 sub-critical reactors, 273 super-critical water reactor (SCWR), 272 very high temperature reactor (VHTR), 272 Nuclear stability, 4 Nuclides, 3 Number of fission neutrons, η, 33 ODE solution, 675 Optical path length, 311 Orthogonality conditions associated Legendre functions, 345 half-range Legendre polynomials, 347 Legendre polynomials, 338 reactor eigenfunctions (λ-modes), 600 spherical harmonics, 351 Perturbation theory adjoint function, see Adjoint boundary, 508 generalized, see Variational methods multigroup diffusion theory, 168, 483 reactivity worth, 169, 486, 490 samarium reactivity worth, 212 xenon reactivity worth, 216 Photoneutrons, 146 Physical constants, 473 Plutonium buildup, 206 composition in spent UO2 fuel, 207 composition—reactor-grade, 233 composition—weapons-grade, 233 concentrations in recycled PWR fuel, 221 physics differences between weapons- and reactor-grade, 234 recycle physics effects, 225 Point kinetics adiabatic method, 605 approximate solutions for fast excursions, 186 approximate solutions with feedback, 183 approximate solutions without feedback, 150 derivation of equations, 602 equations, 147 quasistatic method, 605 transfer functions, see Transfer functions Poison burnable, see Burnable poison control rods, see Control rods fission products, see Fuel burnup samarium, see Samarium soluble, see Soluble poison xenon, see Xenon Pontryagin’s maximum principle, 639 Power autocorrelation function, 181 Power coefficients, 178 Power distribution fuel burnup, 210 peaking, 73 thermal hydraulics, 280 xenon spatial oscillations, see Xenon spatial oscillations Power iteration, see Iteration methods Power peaking, see Power distribution Prompt jump, 151, 152 Prompt jump approximation, 153, 184 Prompt neutron generation time, 147 Prompt neutron lifetime, 38 Pulsed neutron measurement, 157 PUREX separation technology, 239 703 704 Index PWR typical composition and cross sections, 63, 140 Pyrometallurgical separation technology, 239 Quasi-static method, see Point kinetics Radiative capture, 13 Radioactive decay, 8, 19, 39, 143, 198, 209, 211, 213, 217, 226, 239, 283, 641 Radioactive waste cancer dose per Curie in spent fuel, 230 hazard potential, 226 radioactivity of LWR and LMFBR spent fuel, 227 radiotoxic inventory decay of spent fuel, 240 risk factor, 226 toxicity factor, 230 Reactivity autocorrelation function, 180 control rod worth, see Control rod definition, 147, 604 feedback, 161 fuel burnup penalty, 206 measurement of, 149, 156 penalty, 208 perturbation estimate, see Perturbation theory samarium worth, 212 spectral density, 180 temperature defect, 167 variational estimate, see Variational methods xenon worth, 215, 216 Reactivity coefficients delay time constants, 178 Doppler, 161, 162, 169 expansion, 164, 170 fuel bowing, 171 fuel motion, 170 nonleakage, 166 power, 178 representative values, 166, 171 sodium void, 169 temperature, 162, 170 thermal utilization, 165 Reactivity control BWRs, 250 CANDUs, 257 gas-cooled reactors, 260 LMFBRs, 261 PWRs, 249 RBMKs, 259 Reactor accidents anticipated transients without scram, 288 Chernobyl, 297 energy sources, 285 loss of coolant, 287, 295 loss of flow, 287 loss of heat sink, 287, 294 predicted frequency of fatality, 293 reactivity insertion, 287, 297 Three Mile Island, 294 Reactor noise, see Noise analysis Reactor safety accidents, see Reactor accidents analysis, see Reactor safety analysis defense in depth, 285 multiple barriers, 283 passive, 299 passive safety demonstration, 300 radionuclides of concern, 283 risks, 291 Reactor safety analysis event tree, 289 fault tree, 289 probabilistic risk assessment, 288 radiological assessment, 291 Reactor startup analysis, 663 Reflected reactors, see Diffusion theory Reflector savings, 64 Resonance Adler–Adler approximation, 443 Breit–Wigner, multilevel formula, 442 Breit–Wigner, single-level formula, 14, 430, 441 cross sections, 6, 117, 417 Dancoff correction, 428 Doppler broadening, 119, 127 equivalence relations, 422 escape probability, 36, 122 escape probability, closely packed lattices, 427 escape probability, isolated fuel element, 425 heterogeneous fuel–moderator cell, 415 heterogeneous resonance escape probability, 423 homogenized resonance cross section, 423 infinite dilution resonance integral, 422 integral, 122, 419 intermediate resonance approximation, 424, 500 Index J (ξ, β) function, 125 multiband theory, 433 multigroup cross sections, 122, 423, 430, 432 narrow resonance approximation, 123, 419 overlap of different species, 432 pole representation, 445 Porter–Thomas distribution, 429 practical width, 122 R-matrix representation, 439 rational approximation, 427 reciprocity, 418 Reich–Moore formalism, 443 Sauer rational approximation, 427 self-overlap effects, 431 self-shielding, 415, 433 statistical resonance parameters, 431 strength function, 430 unresolved resonances, 428 wide resonance approximation, 123, 420 Response matrix, 329 Rod drop measurement, 157 Rod oscillator measurement, 158, 179 Rossi-α measurement, 159 Samarium, 211 Sauer rational approximation, 427 Self-shielding resonance, 103, 415, 422, 433 spatial, 65, 433 Separation of variables, 54, 58 Soluble poison, 208, 250 Source jerk measurement, 157 Space-dependent nuclear reactor kinetics delayed flux tilts, 601 direct time-Integration, see Time integration methods dynamic programming, 638 linear analysis, 642 Lyapunov’s method for nonlinear stability analysis, 629, 649 modal eigenfunction expansion, 600 Pontryagin’s maximum principle, 639 stochastic, see Stochastic kinetics variational control theory, 636 xenon spatial oscillations, see Xenon spatial oscillations Spent nuclear fuel, 238 Spherical harmonics methods associated Legendre functions, see Legendre polynomials boundary and interface conditions, PL theory, 91, 340 boundary and interface conditions, D-PL theory, 349 diffusion equations in one-dimensional geometry, 93, 347 diffusion theory, from P1 theory, 342 diffusion theory, in multidimensional geometries, 353 double-PL theory, 348 extrapolated boundary condition, 342 half angle Legendre polynomials, see Legendre polynomials Legendre polynomials, see Legendre polynomials Mark boundary conditions, 341 Marshak boundary conditions, 340, 343, 399 multidimensional geometry, 350 PL equations in slab geometry, 91, 339 PL equations in spherical and cylindrical geometries, 344 simplified PL theory, 343 spherical harmonic functions, 350, 682 Stability criteria, 175, 178 feedback delay, 178 instability conditions for two-temperature model, 176 linear analysis, 625 Lyapunov method, 627 threshold power level, 174 transfer function analysis, 171 xenon spatial oscillations, see Xenon spatial oscillations Stochastic kinetics correlation functions, 658 forward stochastic model, 653 means, variances, and covariances, 656 reactor startup analysis, 663 transition probability, 653 transition probability generating function, 653 Synthesis methods formalism, 502 multichannel, 589 single-channel, 583 spectral, 591 Temperature defect, see Reactivity Thermal disadvantage factor, 65, 520 705 706 Index Thermal-hydraulics interaction with reactor physics, 280 reactor safety, 285 reactor stability, 172 Thermal utilization, 34, 72, 165, 518 Thorium fuel cycle, 200 Time eigenvalues, 58 Time integration methods alternating direction implicit, 619 explicit forward-difference, 610 implicit backward-difference, 611 implicit GAKIN, 616 implicit θ , 612 implicit time-integrated, 615 Runge–Kutta, generalized, 624 stiffness confinement, 622 symmetric successive overrelaxation, 623 Transfer functions measurement, 179, 182 phase angle, 158 with feedback, 171, 182 zero-power, 155, 159 Transmission probability, 87, 88, 305, 314. See also Integral transport theory and Interface current methods Transmutation–decay chains cross sections and decay data, 201 fission products, 203 fuel, 199 Transmutation of spent nuclear fuel, 237 Transport boundary condition, 74, 519, 522 Transport cross section, 45, 93, 342, 395 Transuranics cancer dose per Curie in spent fuel, 230 equilibrium distribution in continuously recycled fuel, 236 probability of fission per neutron, 237 risk factor in spent fuel, 233 transmutation, 237 Transverse leakage, 577 TRISO fuel particles, 268, 270 Unit conversion, 669 Uranium composition natural, 233 composition reactor-grade, 233 composition weapons-grade, 234 fuel cycle, 199 physics effects of recycle, 224 resource utilization, 235 Variational methods collision probability theory, 502 construction of variational functionals, 500 control theory, 631, 636 diffusion theory, 583 discontinuous trial function, 575 discontinuous trial functions, 563, 565, 569, 589 dynamic reactivity, 607 even-parity transport theory, 505 flux correction factor, 492 functional, 491 functional admitting discontinuous trial functions, 504, 563, 566, 572, 580, 582 heterogeneity reactivity, 502 interface and boundary terms, 504 intermediate resonance integral, 500 multigroup diffusion theory, 580 P1 equations, 563, 580 Rayleigh quotient, 499, 503 reaction rate ratios, 495 reaction rates, 497 reactivity worth, 490, 492 Ritz procedure, 506 Roussopolos functional, 498 Schwinger functional, 499, 501 static reactivity, 490, 606 stationarity, 498 synthesis, 504. See also Synthesis transport equation, 571 trial functions, 499, 501, 504, 505, 584, 592, 608 Weapons grade plutonium and uranium, see Fuel composition Wigner rational approximation, 427 Wigner–Seitz approximation, see Homogenization Xenon, 213, 641 Xenon spatial oscillations λ-mode stability analysis, 645 linear stability analysis, 642 μ-mode stability analysis, 644 nonlinear stability criterion, 649 variational control, 650 ZEBRA composition, 497 Related Titles Lilley, J. S. Nuclear Physics Principles and Applications 412 pages 2001 Hardcover ISBN-13: 978-0-471-97935-7 Thomas, A. W., Weise, W. The Structure of the Nucleon 303 pages with 94 figures and 9 tables 2001 Hardcover ISBN-13: 978-3-527-40297-7 Turner, J. E. Atoms, Radiation, and Radiation Protection 575 pages 1995 Hardcover ISBN-13: 978-0-471-59581-6 Griffiths, D. Introduction to Elementary Particles 399 pages with 100 figures 1987 Hardcover ISBN-13: 978-0-471-60386-3