Deep neural networks ("deep learning") have emerged as a technology of choice to tackle... more Deep neural networks ("deep learning") have emerged as a technology of choice to tackle problems in natural language processing, computer vision, speech recognition and gameplay, and in just a few years has led to superhuman level performance and ushered in a new wave of "AI." Buoyed by these successes, researchers in the physical sciences have made steady progress in incorporating deep learning into their respective domains. However, such adoption brings substantial challenges that need to be recognized and confronted. Here, we discuss both opportunities and roadblocks to implementation of deep learning within materials science, focusing on the relationship between correlative nature of machine learning and causal hypothesis driven nature of physical sciences. We argue that deep learning and AI are now well positioned to revolutionize fields where causal links are known, as is the case for applications in theory. When confounding factors are frozen or change onl...
The emergence of scanning probe and electron beam imaging techniques has allowed quantitative stu... more The emergence of scanning probe and electron beam imaging techniques has allowed quantitative studies of atomic structure and minute details of electronic and vibrational structure on the level of individual atomic units. These microscopic descriptors, in turn, can be associated with local symmetry breaking phenomena, representing the stochastic manifestation of the underpinning generative physical model. Here, we explore the reconstruction of exchange integrals in the Hamiltonian for a lattice model with two competing interactions from observations of microscopic degrees of freedom and establish the uncertainties and reliability of such analysis in a broad parameter-temperature space. In contrast to other approaches, we specifically specify a loss function inherent to thermodynamic systems and utilize it to estimate uncertainty in simulated realizations of different models. As an ancillary task, we develop a machine learning approach based on histogram clustering to predict phase d...
Materials characterization and property measurements are a cornerstone of material science, provi... more Materials characterization and property measurements are a cornerstone of material science, providing feedback from synthesis to applications. Traditionally, a single sample is used to derive information on a single point in composition space, and imperfections, impurities and stochastic details of material structure are deemed irrelevant or complicating factors in analysis. Here we demonstrate that atomic-scale studies of a single nominal composition can provide information on a finite area of chemical space. This information can be used to reconstruct the material properties in a finite composition and temperature range. We develop a statistical physics-based framework that incorporates chemical and structural data to infer effective atomic interactions driving segregation in a La5/8Ca3/8MnO3 thin-film. A variational autoencoder is used to determine anomalous behaviors in the composition phase diagram. This study provides a framework for creating generative models from diverse dat...
SUMMARY Injection of CO 2 into geologic formations has been identified as a key strategy for miti... more SUMMARY Injection of CO 2 into geologic formations has been identified as a key strategy for mitigating the impact of anthropogenic emissions of CO 2 . A key aspect of this process is the prevention of leakage from the host formation by an effective cap or seal rock which has low porosity and permeability characteristics. Shales comprise the majority of cap rocks encountered in subsurface injection sites with pore sizes typically less than 100 nm and surface chemistries dominated by quartz (SiO 2 ) and clays. We still lack a fundamental understanding of the structural and dynamic behavior of CO 2 (and CO 2 -bearing aqueous fluids) in cap rock environments dominated by nanoporosity for state conditions encountered in injection systems. Even for a simple fluid such as CO 2 we have not adequately explored interfacial phenomena such as the wetting and adsorption for variable pore sizes, pore geometries and pore wall chemistry at conditions approaching and crossing into the supercritical...
Atomic-scale materials synthesis via layer deposition techniques present a unique opportunity to ... more Atomic-scale materials synthesis via layer deposition techniques present a unique opportunity to control material structures and yield systems that display unique functional properties that cannot be stabilized using traditional bulk synthetic routes. However, the deposition process itself presents a large, multidimensional space that is traditionally optimized via intuition and trial and error, slowing down progress. Here, we present an application of deep reinforcement learning to a simulated materials synthesis problem, utilizing the Stein variational policy gradient (SVPG) approach to train multiple agents to optimize a stochastic policy to yield desired functional properties. Our contributions are (1) A fully open source simulation environment for layered materials synthesis problems, utilizing a kinetic Monte-Carlo engine and implemented in the OpenAI Gym framework, (2) Extension of the Stein variational policy gradient approach to deal with both image and tabular input, and (...
Deep neural networks (‘deep learning’) have emerged as a technology of choice to tackle problems ... more Deep neural networks (‘deep learning’) have emerged as a technology of choice to tackle problems in speech recognition, computer vision, finance, etc. However, adoption of deep learning in physical domains brings substantial challenges stemming from the correlative nature of deep learning methods compared to the causal, hypothesis driven nature of modern science. We argue that the broad adoption of Bayesian methods incorporating prior knowledge, development of solutions with incorporated physical constraints and parsimonious structural descriptors and generative models, and ultimately adoption of causal models, offers a path forward for fundamental and applied research.
Understanding of structural, electrical, and gravimetric peculiarities of water vapor interaction... more Understanding of structural, electrical, and gravimetric peculiarities of water vapor interaction with ion-intercalated MXenes led to design of a multimodal humidity sensor. Neutron scattering coupled to molecular dynamics and ab initio calculations showed that a small amount of hydration results in a significant increase in the spacing between MXene layers in the presence of K and Mg intercalants between the layers. Films of K- and Mg-intercalated MXenes exhibited relative humidity (RH) detection thresholds of ∼0.8% RH and showed monotonic RH response in the 0-85% RH range. We found that MXene gravimetric response to water is 10 times faster than their electrical response, suggesting that HO-induced swelling/contraction of channels between MXene sheets results in trapping of HO molecules that act as charge-depleting dopants. The results demonstrate the use of MXenes as humidity sensors and infer potential impact of water on structural and electrical performance of MXene-based devices.
We developed the Gaussian charge-on-spring (GCOS) version of the original self-consistent field i... more We developed the Gaussian charge-on-spring (GCOS) version of the original self-consistent field implementation of the Gaussian Charge Polarizable water model and test its accuracy to represent the polarization behavior of the original model involving smeared charges and induced dipole moments. For that purpose we adapted the recently proposed multiple-particle-move (MPM) within the Gibbs and isochoric-isothermal ensembles Monte Carlo methods for the efficient simulation of polarizable fluids. We assessed the accuracy of the GCOS representation by a direct comparison of the resulting vapor-liquid phase envelope, microstructure, and relevant microscopic descriptors of water polarization along the orthobaric curve against the corresponding quantities from the actual GCP water model.
ABSTRACT Metal oxide surface protonation under hydrothermal conditions is summarized. Important c... more ABSTRACT Metal oxide surface protonation under hydrothermal conditions is summarized. Important concepts and definitions are introduced first, followed by a brief overview of experimental methods and presentation of representative results. Finally, the modeling methods that are most useful in predicting surface protonation behavior between 0 and 300oC are presented and compared.
Deep neural networks ("deep learning") have emerged as a technology of choice to tackle... more Deep neural networks ("deep learning") have emerged as a technology of choice to tackle problems in natural language processing, computer vision, speech recognition and gameplay, and in just a few years has led to superhuman level performance and ushered in a new wave of "AI." Buoyed by these successes, researchers in the physical sciences have made steady progress in incorporating deep learning into their respective domains. However, such adoption brings substantial challenges that need to be recognized and confronted. Here, we discuss both opportunities and roadblocks to implementation of deep learning within materials science, focusing on the relationship between correlative nature of machine learning and causal hypothesis driven nature of physical sciences. We argue that deep learning and AI are now well positioned to revolutionize fields where causal links are known, as is the case for applications in theory. When confounding factors are frozen or change onl...
The emergence of scanning probe and electron beam imaging techniques has allowed quantitative stu... more The emergence of scanning probe and electron beam imaging techniques has allowed quantitative studies of atomic structure and minute details of electronic and vibrational structure on the level of individual atomic units. These microscopic descriptors, in turn, can be associated with local symmetry breaking phenomena, representing the stochastic manifestation of the underpinning generative physical model. Here, we explore the reconstruction of exchange integrals in the Hamiltonian for a lattice model with two competing interactions from observations of microscopic degrees of freedom and establish the uncertainties and reliability of such analysis in a broad parameter-temperature space. In contrast to other approaches, we specifically specify a loss function inherent to thermodynamic systems and utilize it to estimate uncertainty in simulated realizations of different models. As an ancillary task, we develop a machine learning approach based on histogram clustering to predict phase d...
Materials characterization and property measurements are a cornerstone of material science, provi... more Materials characterization and property measurements are a cornerstone of material science, providing feedback from synthesis to applications. Traditionally, a single sample is used to derive information on a single point in composition space, and imperfections, impurities and stochastic details of material structure are deemed irrelevant or complicating factors in analysis. Here we demonstrate that atomic-scale studies of a single nominal composition can provide information on a finite area of chemical space. This information can be used to reconstruct the material properties in a finite composition and temperature range. We develop a statistical physics-based framework that incorporates chemical and structural data to infer effective atomic interactions driving segregation in a La5/8Ca3/8MnO3 thin-film. A variational autoencoder is used to determine anomalous behaviors in the composition phase diagram. This study provides a framework for creating generative models from diverse dat...
SUMMARY Injection of CO 2 into geologic formations has been identified as a key strategy for miti... more SUMMARY Injection of CO 2 into geologic formations has been identified as a key strategy for mitigating the impact of anthropogenic emissions of CO 2 . A key aspect of this process is the prevention of leakage from the host formation by an effective cap or seal rock which has low porosity and permeability characteristics. Shales comprise the majority of cap rocks encountered in subsurface injection sites with pore sizes typically less than 100 nm and surface chemistries dominated by quartz (SiO 2 ) and clays. We still lack a fundamental understanding of the structural and dynamic behavior of CO 2 (and CO 2 -bearing aqueous fluids) in cap rock environments dominated by nanoporosity for state conditions encountered in injection systems. Even for a simple fluid such as CO 2 we have not adequately explored interfacial phenomena such as the wetting and adsorption for variable pore sizes, pore geometries and pore wall chemistry at conditions approaching and crossing into the supercritical...
Atomic-scale materials synthesis via layer deposition techniques present a unique opportunity to ... more Atomic-scale materials synthesis via layer deposition techniques present a unique opportunity to control material structures and yield systems that display unique functional properties that cannot be stabilized using traditional bulk synthetic routes. However, the deposition process itself presents a large, multidimensional space that is traditionally optimized via intuition and trial and error, slowing down progress. Here, we present an application of deep reinforcement learning to a simulated materials synthesis problem, utilizing the Stein variational policy gradient (SVPG) approach to train multiple agents to optimize a stochastic policy to yield desired functional properties. Our contributions are (1) A fully open source simulation environment for layered materials synthesis problems, utilizing a kinetic Monte-Carlo engine and implemented in the OpenAI Gym framework, (2) Extension of the Stein variational policy gradient approach to deal with both image and tabular input, and (...
Deep neural networks (‘deep learning’) have emerged as a technology of choice to tackle problems ... more Deep neural networks (‘deep learning’) have emerged as a technology of choice to tackle problems in speech recognition, computer vision, finance, etc. However, adoption of deep learning in physical domains brings substantial challenges stemming from the correlative nature of deep learning methods compared to the causal, hypothesis driven nature of modern science. We argue that the broad adoption of Bayesian methods incorporating prior knowledge, development of solutions with incorporated physical constraints and parsimonious structural descriptors and generative models, and ultimately adoption of causal models, offers a path forward for fundamental and applied research.
Understanding of structural, electrical, and gravimetric peculiarities of water vapor interaction... more Understanding of structural, electrical, and gravimetric peculiarities of water vapor interaction with ion-intercalated MXenes led to design of a multimodal humidity sensor. Neutron scattering coupled to molecular dynamics and ab initio calculations showed that a small amount of hydration results in a significant increase in the spacing between MXene layers in the presence of K and Mg intercalants between the layers. Films of K- and Mg-intercalated MXenes exhibited relative humidity (RH) detection thresholds of ∼0.8% RH and showed monotonic RH response in the 0-85% RH range. We found that MXene gravimetric response to water is 10 times faster than their electrical response, suggesting that HO-induced swelling/contraction of channels between MXene sheets results in trapping of HO molecules that act as charge-depleting dopants. The results demonstrate the use of MXenes as humidity sensors and infer potential impact of water on structural and electrical performance of MXene-based devices.
We developed the Gaussian charge-on-spring (GCOS) version of the original self-consistent field i... more We developed the Gaussian charge-on-spring (GCOS) version of the original self-consistent field implementation of the Gaussian Charge Polarizable water model and test its accuracy to represent the polarization behavior of the original model involving smeared charges and induced dipole moments. For that purpose we adapted the recently proposed multiple-particle-move (MPM) within the Gibbs and isochoric-isothermal ensembles Monte Carlo methods for the efficient simulation of polarizable fluids. We assessed the accuracy of the GCOS representation by a direct comparison of the resulting vapor-liquid phase envelope, microstructure, and relevant microscopic descriptors of water polarization along the orthobaric curve against the corresponding quantities from the actual GCP water model.
ABSTRACT Metal oxide surface protonation under hydrothermal conditions is summarized. Important c... more ABSTRACT Metal oxide surface protonation under hydrothermal conditions is summarized. Important concepts and definitions are introduced first, followed by a brief overview of experimental methods and presentation of representative results. Finally, the modeling methods that are most useful in predicting surface protonation behavior between 0 and 300oC are presented and compared.
Uploads
Papers by Lukas Vlcek