Numerical analysis

Before the advent of modern computers numerical meth-

ods often depended on hand interpolation in large printed
tables. Since the mid 20th century, computers calculate
the required functions instead. These same interpolation
formulas nevertheless continue to be used as part of the
software algorithms for solving differential equations.

1 General introduction
The overall goal of the field of numerical analysis is the
design and analysis of techniques to give approximate but
accurate solutions to hard problems, the variety of which
is suggested by the following:

• Advanced numerical methods are essential in mak-

Babylonian clay tablet YBC 7289 (c. 1800–1600 BC) with an- ing numerical weather prediction feasible.
notations. The approximation of the square root of 2 is four
sexagesimal figures, which is about six decimal figures. 1 + • Computing the trajectory of a spacecraft requires
24/60 + 51/602 + 10/603 = 1.41421296...[1] the accurate numerical solution of a system of
ordinary differential equations.
Numerical analysis is the study of algorithms that use
• Car companies can improve the crash safety of
numerical approximation (as opposed to general symbolic
their vehicles by using computer simulations of car
manipulations) for the problems of mathematical analysis
crashes. Such simulations essentially consist of solv-
(as distinguished from discrete mathematics).
ing partial differential equations numerically.
One of the earliest mathematical writings is a Babylonian
tablet from the Yale Babylonian Collection (YBC 7289), • Hedge funds (private investment funds) use tools
which gives a sexagesimal numerical approximation of from all fields of numerical analysis to attempt to
√
2 , the length of the diagonal in a unit square. Being calculate the value of stocks and derivatives more
able to compute the sides of a triangle (and hence, being precisely than other market participants.
able to compute square roots) is extremely important, for
instance, in astronomy, carpentry and construction.[2] • Airlines use sophisticated optimization algorithms
to decide ticket prices, airplane and crew assign-
Numerical analysis continues this long tradition of practi- ments and fuel needs. Historically, such algo-
cal mathematical calculations.
√ Much like the Babylonian rithms were developed within the overlapping field
approximation of 2 , modern numerical analysis does of operations research.
not seek exact answers, because exact answers are often
impossible to obtain in practice. Instead, much of nu- • Insurance companies use numerical programs for
merical analysis is concerned with obtaining approximate actuarial analysis.
solutions while maintaining reasonable bounds on errors.
Numerical analysis naturally finds applications in all fields The rest of this section outlines several important themes
of engineering and the physical sciences, but in the 21st of numerical analysis.
century also the life sciences and even the arts have
adopted elements of scientific computations. Ordinary
differential equations appear in celestial mechanics (plan- 1.1 History
ets, stars and galaxies); numerical linear algebra is impor-
tant for data analysis; stochastic differential equations and The field of numerical analysis predates the invention of
Markov chains are essential in simulating living cells for modern computers by many centuries. Linear interpola-
medicine and biology. tion was already in use more than 2000 years ago. Many

1
2 2 GENERATION AND PROPAGATION OF ERRORS

great mathematicians of the past were preoccupied by nu- the same manner as for an iterative method.
merical analysis, as is obvious from the names of im-
portant algorithms like Newton’s method, Lagrange in-
terpolation polynomial, Gaussian elimination, or Euler’s 1.3 Discretization
method.
Furthermore, continuous problems must sometimes be
To facilitate computations by hand, large books were pro-
replaced by a discrete problem whose solution is known
duced with formulas and tables of data such as interpo-
to approximate that of the continuous problem; this pro-
lation points and function coefficients. Using these ta-
cess is called discretization. For example, the solution of
bles, often calculated out to 16 decimal places or more
a differential equation is a function. This function must
for some functions, one could look up values to plug into
be represented by a finite amount of data, for instance by
the formulas given and achieve very good numerical esti-
its value at a finite number of points at its domain, even
mates of some functions. The canonical work in the field
though this domain is a continuum.
is the NIST publication edited by Abramowitz and Ste-
gun, a 1000-plus page book of a very large number of
commonly used formulas and functions and their values
at many points. The function values are no longer very 2 Generation and propagation of
useful when a computer is available, but the large listing errors
of formulas can still be very handy.
The mechanical calculator was also developed as a tool The study of errors forms an important part of numerical
for hand computation. These calculators evolved into analysis. There are several ways in which error can be
electronic computers in the 1940s, and it was then found introduced in the solution of the problem.
that these computers were also useful for administrative
purposes. But the invention of the computer also influ-
enced the field of numerical analysis, since now longer 2.1 Round-off
and more complicated calculations could be done.
Round-off errors arise because it is impossible to repre-
sent all real numbers exactly on a machine with finite
1.2 Direct and iterative methods memory (which is what all practical digital computers
are).
Direct methods compute the solution to a problem in a
finite number of steps. These methods would give the
precise answer if they were performed in infinite pre- 2.2 Truncation and discretization error
cision arithmetic. Examples include Gaussian elimina-
tion, the QR factorization method for solving systems of Truncation errors are committed when an iterative
linear equations, and the simplex method of linear pro- method is terminated or a mathematical procedure is ap-
gramming. In practice, finite precision is used and the proximated, and the approximate solution differs from
result is an approximation of the true solution (assuming the exact solution. Similarly, discretization induces a
stability). discretization error because the solution of the discrete
problem does not coincide with the solution of the contin-
In contrast to direct methods, iterative methods are not
expected to terminate in a finite number of steps. Start- uous problem. For instance, in the iteration in the sidebar
to compute the solution of 3x3 + 4 = 28 , after 10 or so
ing from an initial guess, iterative methods form succes-
sive approximations that converge to the exact solution iterations, we conclude that the root is roughly 1.99 (for
example). We therefore have a truncation error of 0.01.
only in the limit. A convergence test, often involving the
residual, is specified in order to decide when a sufficiently Once an error is generated, it will generally propagate
accurate solution has (hopefully) been found. Even us- through the calculation. For instance, we have already
ing infinite precision arithmetic these methods would not noted that the operation + on a calculator (or a computer)
reach the solution within a finite number of steps (in gen- is inexact. It follows that a calculation of the type is even
eral). Examples include Newton’s method, the bisection more inexact.
method, and Jacobi iteration. In computational matrix What does it mean when we say that the truncation error is
algebra, iterative methods are generally needed for large created when we approximate a mathematical procedure?
problems. We know that to integrate a function exactly requires one
Iterative methods are more common than direct methods to find the sum of infinite trapezoids. But numerically
in numerical analysis. Some methods are direct in prin- one can find the sum of only finite trapezoids, and hence
ciple but are usually used as though they were not, e.g. the approximation of the mathematical procedure. Sim-
GMRES and the conjugate gradient method. For these ilarly, to differentiate a function, the differential element
methods the number of steps needed to obtain the exact approaches zero but numerically we can only choose a fi-
solution is so large that an approximation is accepted in nite value of the differential element.
 3

2.3 Numerical stability and well-posed catastrophic cancelation) has a huge effect
problems on the results, even though both functions are
equivalent, as shown below
Numerical stability is an important notion in numerical (√ √ )
analysis. An algorithm is called numerically stable if an f (x) = x x + 1 − x
√ √
error, whatever its cause, does not grow to be much larger (√ √ ) x+1+ x
during the calculation. This happens if the problem is =x x+1− x √ √
x+1+ x
well-conditioned, meaning that the solution changes by √ √
( x + 1)2 − ( x)2
only a small amount if the problem data are changed by =x √ √
a small amount. To the contrary, if a problem is ill- x+1+ x
conditioned, then any small error in the data will grow x+1−x
= x√ √
to be a large error. x+1+ x
Both the original problem and the algorithm used to 1
= x√ √
solve that problem can be well-conditioned and/or ill- x+1+ x
conditioned, and any combination is possible. x
=√ √
x+1+ x
So an algorithm that solves a well-conditioned problem
may be either numerically stable or numerically unstable. = g(x)
An art of numerical analysis is to find a stable algorithm
for solving a well-posed mathematical problem. For in- The desired value, computed using infinite pre-
stance, computing the square root of 2 (which is roughly cision, is 11.174755...
1.41421) is a well-posed problem. Many algorithms solve
this√problem by starting with an initial approximation x0 • The example is a modification of one taken from
to 2 , for instance x0 = 1.4, and then computing im- Mathew; Numerical methods using Matlab, 3rd ed.
proved guesses x1 , x2 , etc. One such method is the fa-
mous Babylonian method, which is given by xk₊₁ = xk/2
+ 1/xk. Another method, which we will call Method X,
is given by xk₊₁ = (xk2 − 2)2 + xk.[3] We have calculated
3 Areas of study
a few iterations of each scheme in table form below, with
initial guesses x0 = 1.4 and x0 = 1.42. The field of numerical analysis includes many sub-
disciplines. Some of the major ones are:
Observe that the Babylonian method converges quickly
regardless of the initial guess, whereas Method X con-
verges extremely slowly with initial guess x0 = 1.4 and 3.1 Computing values of functions
diverges for initial guess x0 = 1.42. Hence, the Baby-
lonian method is numerically stable, while Method X is One of the simplest problems is the evaluation of a func-
numerically unstable. tion at a given point. The most straightforward approach,
of just plugging in the number in the formula is some-
Numerical stability is affected by the number times not very efficient. For polynomials, a better ap-
of the significant digits the machine keeps on, proach is using the Horner scheme, since it reduces the
if we use a machine that keeps only the four necessary number of multiplications and additions. Gen-
most significant decimal digits, a good exam- erally, it is important to estimate and control round-off
ple on loss of significance is given by these two errors arising from the use of floating point arithmetic.
equivalent functions
(√ √ )
f (x) = x x + 1 − x and g(x) =
√ x √ .
x+1+ x
3.2 Interpolation, extrapolation, and re-
If we compare the results of gression
(√ √ )
f (500) = 500 501 − 500 = 500 (22.38 − 22.36) =Interpolation
500(0.02) =solves
10 the following problem: given the
value of some unknown function at a number of points,
and what value does that function have at some other point
500
g(500) = √ √ between the given points?
501 + 500
500 Extrapolation is very similar to interpolation, except that
= now we want to find the value of the unknown function at
22.38 + 22.36
500 a point which is outside the given points.
= = 11.17
44.74 Regression is also similar, but it takes into account that
by looking to the two results above, we real- the data is imprecise. Given some points, and a measure-
ize that loss of significance (caused here by ment of the value of some function at these points (with
4 4 SOFTWARE

an error), we want to determine the unknown function. The method of Lagrange multipliers can be used to re-
The least squares-method is one popular way to achieve duce optimization problems with constraints to uncon-
this. strained optimization problems.

3.3 Solving equations and systems of equa- 3.6 Evaluating integrals

tions
Main article: Numerical integration
Another fundamental problem is computing the solution
of some given equation. Two cases are commonly distin- Numerical integration, in some instances also known as
guished, depending on whether the equation is linear or numerical quadrature, asks for the value of a definite
not. For instance, the equation 2x + 5 = 3 is linear while integral. Popular methods use one of the Newton–Cotes
2x2 + 5 = 3 is not. formulas (like the midpoint rule or Simpson’s rule) or
Gaussian quadrature. These methods rely on a “divide
Much effort has been put in the development of meth-
and conquer” strategy, whereby an integral on a relatively
ods for solving systems of linear equations. Standard di-
large set is broken down into integrals on smaller sets.
rect methods, i.e., methods that use some matrix decom-
In higher dimensions, where these methods become pro-
position are Gaussian elimination, LU decomposition,
hibitively expensive in terms of computational effort, one
Cholesky decomposition for symmetric (or hermitian)
may use Monte Carlo or quasi-Monte Carlo methods (see
and positive-definite matrix, and QR decomposition for
Monte Carlo integration), or, in modestly large dimen-
non-square matrices. Iterative methods such as the
sions, the method of sparse grids.
Jacobi method, Gauss–Seidel method, successive over-
relaxation and conjugate gradient method are usually pre-
ferred for large systems. General iterative methods can be
3.7 Differential equations
developed using a matrix splitting.
Root-finding algorithms are used to solve nonlinear equa- Main articles: Numerical ordinary differential equations
tions (they are so named since a root of a function is an and Numerical partial differential equations
argument for which the function yields zero). If the func-
tion is differentiable and the derivative is known, then
Numerical analysis is also concerned with computing (in
Newton’s method is a popular choice. Linearization is
an approximate way) the solution of differential equa-
another technique for solving nonlinear equations.
tions, both ordinary differential equations and partial dif-
ferential equations.
3.4 Solving eigenvalue or singular value Partial differential equations are solved by first discretiz-
problems ing the equation, bringing it into a finite-dimensional sub-
space. This can be done by a finite element method, a
finite difference method, or (particularly in engineering)
Several important problems can be phrased in terms of
a finite volume method. The theoretical justification of
eigenvalue decompositions or singular value decompo-
these methods often involves theorems from functional
sitions. For instance, the spectral image compression
analysis. This reduces the problem to the solution of an
algorithm[4] is based on the singular value decomposition.
algebraic equation.
The corresponding tool in statistics is called principal
component analysis.

4 Software
3.5 Optimization
Main articles: List of numerical analysis software and
Main article: Mathematical optimization Comparison of numerical analysis software

Optimization problems ask for the point at which a given Since the late twentieth century, most algorithms are im-
function is maximized (or minimized). Often, the point plemented in a variety of programming languages. The
also has to satisfy some constraints. Netlib repository contains various collections of software
The field of optimization is further split in several sub- routines for numerical problems, mostly in Fortran and C.
fields, depending on the form of the objective function Commercial products implementing many different nu-
and the constraint. For instance, linear programming merical algorithms include the IMSL and NAG libraries;
deals with the case that both the objective function and a free alternative is the GNU Scientific Library.
the constraints are linear. A famous method in linear pro- There are several popular numerical computing applica-
gramming is the simplex method. tions such as MATLAB, TK Solver, S-PLUS, LabVIEW,
 5

and IDL as well as free and open source alternatives • Higham, Nicholas J. (1996). Accuracy and Stability
such as FreeMat, Scilab, GNU Octave (similar to Mat- of Numerical Algorithms (Society for Industrial and
lab), IT++ (a C++ library), R (similar to S-PLUS) and Applied Mathematics, ISBN 0-89871-355-2).
certain variants of Python. Performance varies widely:
while vector and matrix operations are usually fast, scalar • Hildebrand, F. B. (1974). Introduction to Numeri-
loops may vary in speed by more than an order of cal Analysis (2nd ed.). McGraw-Hill. ISBN 0-07-
magnitude.[5][6] 028761-9.

Many computer algebra systems such as Mathematica • Leader, Jeffery J. (2004). Numerical Analysis and
also benefit from the availability of arbitrary precision Scientific Computation. Addison Wesley. ISBN 0-
arithmetic which can provide more accurate results. 201-73499-0.
Also, any spreadsheet software can be used to solve sim- • Wilkinson, J.H. (1965). The Algebraic Eigenvalue
ple problems relating to numerical analysis. Problem (Clarendon Press).

• Kahan, W. (1972). ""A survey of error-analysis,”

in Info. Processing 71 (Proc. IFIP Congress 71
5 See also in Ljubljana), vol. 2, pp. 1214–39, North-Holland
Publishing, Amsterdam”. (examples of the impor-
• Analysis of algorithms tance of accurate arithmetic).
• Computational science • Trefethen, Lloyd N. (2006). “Numerical analysis”,
20 pages. In: Timothy Gowers and June Barrow-
• List of numerical analysis topics Green (editors), Princeton Companion of Mathemat-
ics, Princeton University Press.
• Numerical differentiation

• Numerical Recipes
8 External links
• Symbolic-numeric computation
Journals

6 Notes • Numerische Mathematik, volumes 1-66, Springer,

1959-1994 (searchable; pages are images). (En-
[1] Photograph, illustration, and description of the root(2) glish) (German)
tablet from the Yale Babylonian Collection
• Numerische Mathematik at SpringerLink, volumes
[2] The New Zealand Qualification authority specifically 1-112, Springer, 1959–2009
mentions this skill in document 13004 version 2, dated 17
October 2003 titled CARPENTRY THEORY: Demon- • SIAM Journal on Numerical Analysis, volumes 1-
strate knowledge of setting out a building 47, SIAM, 1964–2009

[3] This is a fixed point iteration for the equation√x = (x2 −

2)2 + x = f (x) , whose solutions include 2 . The it-
Online texts
√ to the right since f (x) ≥ x . √
erates always move Hence
x1 = 1.4 < 2 converges and x1 = 1.42 > 2 di- • Hazewinkel, Michiel, ed. (2001), “Numerical anal-
verges. ysis”, Encyclopedia of Mathematics, Springer, ISBN
978-1-55608-010-4
[4] The Singular Value Decomposition and Its Applications
in Image Compression • Numerical Recipes, William H. Press (free, down-
loadable previous editions)
[5] Speed comparison of various number crunching packages
• First Steps in Numerical Analysis (archived),
[6] Comparison of mathematical programs for data analysis
R.J.Hosking, S.Joe, D.C.Joyce, and J.C.Turner
Stefan Steinhaus, ScientificWeb.com
• CSEP (Computational Science Education Project),
U.S. Department of Energy
7 References
Online course material
• Golub, Gene H. and Charles F. Van Loan (1986).
Matrix Computations, Third Edition (Johns Hopkins • Numerical Methods, Stuart Dalziel University of
University Press, ISBN 0-8018-5413-X). Cambridge
6 8 EXTERNAL LINKS

• Lectures on Numerical Analysis, Dennis Deturck

and Herbert S. Wilf University of Pennsylvania
• Numerical methods, John D. Fenton University of
Karlsruhe
• Numerical Methods for Science, Technology, Engi-
neering and Mathematics, Autar Kaw University of
South Florida

• Numerical Analysis Project, John H. Mathews

California State University, Fullerton

• Numerical Methods – Online Course, Aaron

Naiman Jerusalem College of Technology

• Numerical Methods for Physicists, Anthony O’Hare

Oxford University
• Lectures in Numerical Analysis (archived), R.
Radok Mahidol University
• Introduction to Numerical Analysis for Engineering,
Henrik Schmidt Massachusetts Institute of Technol-
ogy

• Numerical Methods for time-dependent Partial Dif-

ferential Equations, J.W. Haverkort, based on a
course by P.A. Zegeling Utrecht University
• Numerical Analysis for Engineering, D. W. Harder
University of Waterloo
 7

9 Text and image sources, contributors, and licenses

9.1 Text
• Numerical analysis Source: https://en.wikipedia.org/wiki/Numerical_analysis?oldid=745764537 Contributors: Tobias Hoevekamp, The
Anome, Tarquin, Taw, Eijkhout, Michael Hardy, Dominus, Nixdorf, Zeno Gantner, Loisel, Stevan White, Jurgen~enwiki, Charles
Matthews, Dysprosia, Jitse Niesen, Taxman, Topbanana, Ldo, Rogper~enwiki, Robbot, Jaredwf, Troworld, PedroPVZ, Blainster, Bkell,
JesseW, Saforrest, Rege~enwiki, Robinh, Peter L, Tea2min, Decrypt3, Giftlite, BenFrantzDale, Lethe, Tom harrison, Fleminra, Gua-
naco, Matt Crypto, Gadfium, APH, Bumm13, AmadeusKlocker, Fintor, Kate, PhotoBox, Zowie, Rich Farmbrough, Yuval madar, El-
wikipedista~enwiki, MisterSheik, Oyz, Dungodung, Greenleaf~enwiki, Nk, Crust, Msh210, Alansohn, Arthena, Sligocki, Bugg, Aitter,
Gene Nygaard, Bookandcoffee, Beliavsky, Oleg Alexandrov, Linas, David Haslam, Decrease789, KymFarnik, Graham87, Xask Linus,
Koavf, Salix alba, TeemuN, Hlangeveld, Bubba73, Arnero, Mathbot, Nihiltres, RexNL, Carrionluggage, Windharp, NevilleDNZ, Chobot,
Turidoth, Wavelength, RussBot, Hede2000, KSmrq, Chaos, Yahya Abdal-Aziz, Cat2020, Boivie, Zzuuzz, Ninly, Closedmouth, Cedar101,
Josh3580, Dontaskme, Willtron, JLaTondre, Pred, Gesslein, Xiaojeng~enwiki, Lunch, Yvwv, Sardanaphalus, A bit iffy, JJL, SmackBot,
Tomchen~enwiki, Adam majewski, Jagged 85, Wygk, Hongooi, Timothy Clemans, Berland, Brianboonstra, Mlpkr, Drunken Pirate, The
undertow, Lambiam, Sina2, ExamplePuzzle, IronGargoyle, Asyndeton, Stephen B Streater, Ginkgo100, MystRivenExile, FatBastardInk,
Audiosmurf, Paul Matthews, Didimos, CRGreathouse, CBM, Vitolis, Requestion, Bocianski, Ntsimp, Mattisse, Thijs!bot, Oliver202,
Quantufinity, HussainAbbas, Urdutext, BigJohnHenry, Darklilac, VictorAnyakin, Nikolas Karalis, MER-C, The Transhumanist, Eagle-
Fan, Systemlover, Khalid Mahmood, SamShearman, Gwern, Jtir, Leyo, Kawautar, Salih, Gombang, Nonphixion, JonMcLoone, Policron,
Robertgreer, Skarnani, CMaes, Sheliak, JohnBlackburne, Redgecko, Stephenkirkup, Antoni Barau, Jmath666, Aither~enwiki, Wolfrock,
SieBot, Un4v41l48l3, Lightmouse, Svick, Amahoney, WikipedianMarlith, SlackerMom, Justin W Smith, EXTER7, Wikijens, Tomtzigt,
, Nostraticispeak, Unmerklich, XLinkBot, Chrismacgr, JinJian, Addbot, Wordsoup, Manuel Trujillo Berges, Fgnievinski, Ekojekoj,
Metsavend, Tide rolls, Loupeter, Legobot, Yobot, KamikazeBot, 1exec1, Kingpin13, Citation bot, Happyrabbit, Isheden, Rsmn, Char-
vest, Gtpjg, Bekus, CES1596, FrescoBot, Steve Quinn, Gaba p, Boulaur, RedBot, Meaghan, Jauhienij, David Binner, TjBot, DSP-user,
EmausBot, Jmencisom, Bethnim, Mirificium, Google Child, Christina Silverman, Darthhappyface, Shootji, Tot12, Ramiamro, Anita5192,
ClueBot NG, Satellizer, Vacation9, Snotbot, GiantReliant, 3rdiw, Widr, Helpful Pixie Bot, Enneth with a k, Ogoorcs, Solomon7968,
Brad7777, BattyBot, ChrisGualtieri, Chunliang Lyu, JYBot, Shtamy, Nm160111, Drajay1976, Purnendu Karmakar, Zhangzk, Laguer-
reLegendre, Jarash, Brtietz, Stormmeteo, Msalimi122, Msalimi1222, Linnet1979, Sasadd, Mateusz Konieczny, KasparBot, Herobeaner
and Anonymous: 196

9.2 Images
• File:Folder_Hexagonal_Icon.svg Source: https://upload.wikimedia.org/wikipedia/en/4/48/Folder_Hexagonal_Icon.svg License: Cc-by-
sa-3.0 Contributors: ? Original artist: ?
• File:Internet_map_1024.jpg Source: https://upload.wikimedia.org/wikipedia/commons/d/d2/Internet_map_1024.jpg License: CC BY
2.5 Contributors: Originally from the English Wikipedia; description page is/was here. Original artist: The Opte Project
• File:LemonadeJuly2006.JPG Source: https://upload.wikimedia.org/wikipedia/commons/2/28/LemonadeJuly2006.JPG License: CC-
BY-SA-3.0 Contributors: ? Original artist: ?
• File:Linear-regression.svg Source: https://upload.wikimedia.org/wikipedia/commons/1/12/Linear-regression.svg License: Public do-
main Contributors: Own work Original artist: Qef
• File:Portal-puzzle.svg Source: https://upload.wikimedia.org/wikipedia/en/f/fd/Portal-puzzle.svg License: Public domain Contributors: ?
Original artist: ?
• File:Schumacher_(Ferrari)_in_practice_at_USGP_2005.jpg Source: https://upload.wikimedia.org/wikipedia/commons/9/9d/
Schumacher_%28Ferrari%29_in_practice_at_USGP_2005.jpg License: CC BY-SA 2.0 Contributors: ? Original artist: ?
• File:Text_document_with_red_question_mark.svg Source: https://upload.wikimedia.org/wikipedia/commons/a/a4/Text_document_
with_red_question_mark.svg License: Public domain Contributors: Created by bdesham with Inkscape; based upon Text-x-generic.svg
from the Tango project. Original artist: Benjamin D. Esham (bdesham)
• File:Wikibooks-logo.svg Source: https://upload.wikimedia.org/wikipedia/commons/f/fa/Wikibooks-logo.svg License: CC BY-SA 3.0
Contributors: Own work Original artist: User:Bastique, User:Ramac et al.
• File:Wikiquote-logo.svg Source: https://upload.wikimedia.org/wikipedia/commons/f/fa/Wikiquote-logo.svg License: Public domain
Contributors: Own work Original artist: Rei-artur
• File:Wind-particle.png Source: https://upload.wikimedia.org/wikipedia/commons/e/ef/Wind-particle.png License: Public domain Con-
tributors: http://en.wikipedia.org/wiki/File:Wind-particle.png Original artist: Loisel
• File:Ybc7289-bw.jpg Source: https://upload.wikimedia.org/wikipedia/commons/0/0b/Ybc7289-bw.jpg License: CC BY 2.5 Contribu-
tors: Own work Original artist: Bill Casselman

9.3 Content license

• Creative Commons Attribution-Share Alike 3.0