arXiv subjects

Boris Kozinsky

Publications and source records attributed to Boris Kozinsky.

At least 73 records · Page 4Linked to original sources

arXiv2021

Spectral denoising for unsupervised analysis of correlated ionic transport

Computation of correlated ionic transport properties from molecular dynamics in the Green-Kubo formalism is expensive as one cannot rely on the affordable mean square displacement approach. We use spectral decomposition of the short-time ionic displacement covariance to learn a set of diffusion eigenmodes that encode the correlation structure and form a basis for analyzing the ionic trajectories. This allows to systematically reduce the uncertainty and accelerate computations of ionic conductivity in systems with a steady-state correlation structure. We provide mathematical and numerical proofs of the method's robustness, and demonstrate it on realistic electrolyte materials.

physics.comp-ph↗

arXiv2021

Active learning of reactive Bayesian force fields: Application to heterogeneous hydrogen-platinum catalysis dynamics

Accurate modeling of chemically reactive systems has traditionally relied on either expensive ab initio approaches or flexible bond-order force fields such as ReaxFF that require considerable time, effort, and expertise to parameterize. Here, we introduce FLARE++, a Bayesian active learning method for training reactive many-body force fields on the fly during molecular dynamics (MD) simulations. During the automated training loop, the predictive uncertainties of a sparse Gaussian process (SGP) force field are evaluated at each timestep of an MD simulation to determine whether additional ab initio data are needed. Once trained, the SGP is mapped onto an equivalent and much faster model that is polynomial in the local environment descriptors and whose prediction cost is independent of the training set size. We apply our method to a canonical reactive system in the field of heterogeneous catalysis, hydrogen splitting and recombination on a platinum (111) surface, obtaining a trained model within three days of wall time that is twice as fast as a recent Pt/H ReaxFF force field and considerably more accurate. Our method is fully open source and is expected to reduce the time and effort required to train fast and accurate reactive force fields for complex systems.

cond-mat.mtrl-sci↗

arXiv2021

Interband tunneling effects on materials transport properties using the first principles Wigner distribution

Electronic transport in narrow gap semiconductors is characterized by spontaneous vertical transitions between carriers in the valence and conduction bands, a phenomenon also known as Zener tunneling. However, this effect is not captured by existing models based on the Boltzmann transport equation. In this work, we propose a new fully first principles model for electronic transport using the Wigner distribution function and implement it to solve the equations of motion for electrons. The formalism generalizes the Boltzmann equation to materials with strong interband coupling and include transport contributions from off-diagonal components of the charge current operator. We illustrate the method with a study of Bi$_2$Se$_3$, showing that interband tunneling dominates the electron transport dynamics at experimentally relevant small doping concentrations, a behavior that is likely shared with other semiconductors, including topological insulators. Surprisingly, Zener tunneling occurs also between band subvalleys separated by energy much larger than the band gap.

cond-mat.mtrl-sci↗

arXiv2021

Anomalous thermoelectric transport phenomena from interband electron-phonon scattering

The Seebeck coefficient and electrical conductivity are two critical quantities to optimize simultaneously in designing thermoelectric materials, and they are determined by the dynamics of carrier scattering. We uncover a new regime where the co-existence at the Fermi level of multiple bands with different effective masses leads to strongly energy-dependent carrier lifetimes due to intrinsic electron-phonon scattering. In this anomalous regime, electrical conductivity decreases with carrier concentration, Seebeck coefficient reverses sign even at high doping, and power factor exhibits an unusual second peak. We discuss the origin and magnitude of this effect using first-principles Boltzmann transport calculations and simplified models. We also identify general design rules for using this paradigm to engineer enhanced performance in thermoelectric materials.

cond-mat.mtrl-sci↗

arXiv2021

Bayesian Force Fields from Active Learning for Simulation of Inter-Dimensional Transformation of Stanene

We present a way to dramatically accelerate Gaussian process models for interatomic force fields based on many-body kernels by mapping both forces and uncertainties onto functions of low-dimensional features. This allows for automated active learning of models combining near-quantum accuracy, built-in uncertainty, and constant cost of evaluation that is comparable to classical analytical models, capable of simulating millions of atoms. Using this approach, we perform large scale molecular dynamics simulations of the stability of the stanene monolayer. We discover an unusual phase transformation mechanism of 2D stanene, where ripples lead to nucleation of bilayer defects, densification into a disordered multilayer structure, followed by formation of bulk liquid at high temperature or nucleation and growth of the 3D bcc crystal at low temperature. The presented method opens possibilities for rapid development of fast accurate uncertainty-aware models for simulating long-time large-scale dynamics of complex materials.

physics.comp-ph↗

arXiv2020

Multitask machine learning of collective variables for enhanced sampling of rare events

Computing accurate reaction rates is a central challenge in computational chemistry and biology because of the high cost of free energy estimation with unbiased molecular dynamics. In this work, a data-driven machine learning algorithm is devised to learn collective variables with a multitask neural network, where a common upstream part reduces the high dimensionality of atomic configurations to a low dimensional latent space, and separate downstream parts map the latent space to predictions of basin class labels and potential energies. The resulting latent space is shown to be an effective low-dimensional representation, capturing the reaction progress and guiding effective umbrella sampling to obtain accurate free energy landscapes. This approach is successfully applied to model systems including a 5D Müller Brown model, a 5D three-well model, and alanine dipeptide in vacuum. This approach enables automated dimensionality reduction for energy controlled reactions in complex systems, offers a unified framework that can be trained with limited data, and outperforms single-task learning approaches, including autoencoders.

physics.comp-ph↗

arXiv2020

Electron-phonon drag enhancement of transport properties from fully coupled \textit{ab initio} Boltzmann formalism

We present a combined treatment of the non-equilibrium dynamics and transport of electrons and phonons by carrying out \textit{ab initio} calculations of the fully coupled electron and phonon Boltzmann transport equations. We find that the presence of mutual drag between the two carriers causes the thermopower to be enhanced and dominated by the transport of phonons, rather than electrons as in the traditional semiconductor picture. Drag also strongly boosts the intrinsic electron mobility, thermal conductivity and the Lorenz number. Impurity scattering is seen to suppress the drag-enhancement of the thermal and electrical conductivities, while having weak effects on the enhancement of the Lorenz number and thermopower. We demonstrate these effects in \textit{n}-doped 3C-SiC at room temperature, and explain their origins. This work establishes the roles of microscopic scattering mechanisms in the emergence of strong drag effects in the transport of the interacting electron-phonon gas.

cond-mat.mtrl-sci↗

arXiv2020

Accurate and scalable multi-element graph neural network force field and molecular dynamics with direct force architecture

Recently, machine learning (ML) has been used to address the computational cost that has been limiting ab initio molecular dynamics (AIMD). Here, we present GNNFF, a graph neural network framework to directly predict atomic forces from automatically extracted features of the local atomic environment that are translationally-invariant, but rotationally-covariant to the coordinate of the atoms. We demonstrate that GNNFF not only achieves high performance in terms of force prediction accuracy and computational speed on various materials systems, but also accurately predicts the forces of a large MD system after being trained on forces obtained from a smaller system. Finally, we use our framework to perform an MD simulation of Li7P3S11, a superionic conductor, and show that resulting Li diffusion coefficient is within 14% of that obtained directly from AIMD. The high performance exhibited by GNNFF can be easily generalized to study atomistic level dynamics of other material systems.

physics.comp-ph↗

arXiv2020

AiiDA 1.0, a scalable computational infrastructure for automated reproducible workflows and data provenance

The ever-growing availability of computing power and the sustained development of advanced computational methods have contributed much to recent scientific progress. These developments present new challenges driven by the sheer amount of calculations and data to manage. Next-generation exascale supercomputers will harden these challenges, such that automated and scalable solutions become crucial. In recent years, we have been developing AiiDA (http://www.aiida.net), a robust open-source high-throughput infrastructure addressing the challenges arising from the needs of automated workflow management and data provenance recording. Here, we introduce developments and capabilities required to reach sustained performance, with AiiDA supporting throughputs of tens of thousands processes/hour, while automatically preserving and storing the full data provenance in a relational database making it queryable and traversable, thus enabling high-performance data analytics. AiiDA's workflow language provides advanced automation, error handling features and a flexible plugin model to allow interfacing with any simulation software. The associated plugin registry enables seamless sharing of extensions, empowering a vibrant user community dedicated to making simulations more robust, user-friendly and reproducible.

cs.DC↗

arXiv2019

On-the-Fly Active Learning of Interpretable Bayesian Force Fields for Atomistic Rare Events

Machine learned force fields typically require manual construction of training sets consisting of thousands of first principles calculations, which can result in low training efficiency and unpredictable errors when applied to structures not represented in the training set of the model. This severely limits the practical application of these models in systems with dynamics governed by important rare events, such as chemical reactions and diffusion. We present an adaptive Bayesian inference method for automating the training of interpretable, low-dimensional, and multi-element interatomic force fields using structures drawn on the fly from molecular dynamics simulations. Within an active learning framework, the internal uncertainty of a Gaussian process regression model is used to decide whether to accept the model prediction or to perform a first principles calculation to augment the training set of the model. The method is applied to a range of single- and multi-element systems and shown to achieve a favorable balance of accuracy and computational efficiency, while requiring a minimal amount of ab initio training data. We provide a fully open-source implementation of our method, as well as a procedure to map trained models to computationally efficient tabulated force fields.

physics.comp-ph↗

arXiv2019

Charge density and redox potential of LiNiO2 using ab initio diffusion quantum Monte Carlo

Electronic structure of layered LiNiO2 has been controversial despite numerous theoretical and experimental reports regarding its nature. We investigate the charge densities, lithium intercalation potentials and Li diffusion barrier energies of LixNiO2 (0.0 < x < 1.0) system using a truly ab-initio method, diffusion quantum Monte Carlo (DMC). We compare the charge densities from DMC and density functional theory (DFT) and show that local and semi-local DFT functionals yield spin polarization densities with incorrect sign on the oxygen atoms. SCAN functional and Hubbard-U correction improves the polarization density around Ni and O atoms, resulting in smaller deviations from the DMC densities. DMC accurately captures the p-d hybridization between the Ni-O atoms, yielding accurate lithium intercalation voltages, polarization densities and reaction barriers.

cond-mat.mtrl-sci↗

arXiv2019

Unsupervised landmark analysis for jump detection in molecular dynamics simulations

Molecular dynamics is a versatile and powerful method to study diffusion in solid-state ionic conductors, requiring minimal prior knowledge of equilibrium or transition states of the system's free energy surface. However, the analysis of trajectories for relevant but rare events, such as a jump of the diffusing mobile ion, is still rather cumbersome, requiring prior knowledge of the diffusive process in order to get meaningful results. In this work, we present a novel approach to detect the relevant events in a diffusive system without assuming prior information regarding the underlying process. We start from a projection of the atomic coordinates into a landmark basis to identify the dominant features in a mobile ion's environment. Subsequent clustering in landmark space enables a discretization of any trajectory into a sequence of distinct states. As a final step, the use of the smooth overlap of atomic positions descriptor allows distinguishing between different environments in a straightforward way. We apply this algorithm to ten Li-ionic systems and conduct in-depth analyses of cubic Li$_{7}$La$_{3}$Zr$_{2}$O$_{12}$, tetragonal Li$_{10}$GeP$_{2}$S$_{12}$, and the $β$-eucryptite LiAlSiO$_{4}$. We compare our results to existing methods, underscoring strong points, weaknesses, and insights into the diffusive behavior of the ionic conduction in the materials investigated.

cond-mat.mtrl-sci↗

arXiv2019

Fast Neural Network Approach for Direct Covariant Forces Prediction in Complex Multi-Element Extended Systems

Neural network force field (NNFF) is a method for performing regression on atomic structure-force relationships, bypassing expensive quantum mechanics calculation which prevents the execution of long ab-initio quality molecular dynamics simulations. However, most NNFF methods for complex multi-element atomic systems indirectly predict atomic force vectors by exploiting just atomic structure rotation-invariant features and the network-feature spatial derivatives which are computationally expensive. We develop a staggered NNFF architecture exploiting both rotation-invariant and covariant features separately to directly predict atomic force vectors without using spatial derivatives, thereby reducing expensive structural feature calculation by ~180-480x. This acceleration enables us to develop NNFF which directly predicts atomic forces in complex ternary and quaternary-element extended systems comprised of long polymer chains, amorphous oxide, and surface chemical reactions. The staggered rotation-invariant-covariant architecture described here can also directly predict complex covariant vector outputs from local physical structures in domains beyond computational material science.

physics.comp-ph↗

arXiv2018

Accelerated screening of thermoelectric materials by first-principles computations of electron-phonon scattering

Recent discovery of new materials for thermoelectric energy conversion is enabled by efficient prediction of materials' performance from first-principles, without empirically fitted parameters. The novel simplified approach for computing electronic transport properties is described, which achieves good accuracy and transferability while greatly reducing complexity and computation cost compared to the existing methods. The first-principles calculations of the electron-phonon coupling demonstrate that the energy dependence of the electron relaxation time varies significantly with chemical composition and carrier concentration, suggesting that it is necessary to go beyond the commonly used approximations to screen and optimize materials' composition, carrier concentration, and microstructure. The new method is verified using high accuracy computations and validated with experimental data before applying it to screen and discover promising compositions in the space of half-Heusler alloys. By analyzing data trends the effective electron mass is identified as the single best general descriptor determining material's performance. The Lorenz number is computed from first principles and the universality of the Wiedemann-Franz law in thermoelectrics is discussed.

cond-mat.mtrl-sci↗

arXiv2015

Electron-phonon-averaged approximation for first-principles computations of electron relaxation times and transport properties in semiconductor materials

We present a simple and efficient approximation to the electron-phonon scattering rate suitable for high-throughput screening of candidate materials for thermoelectric devices, based on electronic transport. The method is applied to calculate the electronic transport coefficients of half-Heusler compounds, showing agreement with experimental data. By directly computing electrical and the electronic part of the thermal conductivities, we find deviations from the Wiedemann-Franz law in these compounds at high temperatures and low carrier concentrations.

cond-mat.mtrl-sci↗

arXiv2015

AiiDA: Automated Interactive Infrastructure and Database for Computational Science

Computational science has seen in the last decades a spectacular rise in the scope, breadth, and depth of its efforts. Notwithstanding this prevalence and impact, it is often still performed using the renaissance model of individual artisans gathered in a workshop, under the guidance of an established practitioner. Great benefits could follow instead from adopting concepts and tools coming from computer science to manage, preserve, and share these computational efforts. We illustrate here our paradigm sustaining such vision, based around the four pillars of Automation, Data, Environment, and Sharing. We then discuss its implementation in the open-source AiiDA platform (http://www.aiida.net), that has been tuned first to the demands of computational materials science. AiiDA's design is based on directed acyclic graphs to track the provenance of data and calculations, and ensure preservation and searchability. Remote computational resources are managed transparently, and automation is coupled with data storage to ensure reproducibility. Last, complex sequences of calculations can be encoded into scientific workflows. We believe that AiiDA's design and its sharing capabilities will encourage the creation of social ecosystems to disseminate codes, data, and scientific workflows.

physics.comp-ph↗

arXiv2015

Effects of sublattice symmetry and frustration on ionic transport in garnet solid electrolytes

We use rigorous group-theoretic techniques and molecular dynamics to investigate the connection between structural symmetry and ionic conductivity in the garnet family of solid Li-ion electrolytes. We identify new ordered phases and order-disorder phase transitions that are relevant for conductivity optimization. Ionic transport in this materials family is controlled by the frustration of the Li sublattice caused by incommensurability with the host structure at non-integer Li concentrations, while ordered phases explain regions of sharply lower conductivity. Disorder is therefore predicted to be optimal for ionic transport in this and other conductor families with strong Li interaction.

cond-mat.mtrl-sci↗

arXiv2014

Insights and challenges of applying the $GW$ method to transition metal oxides

The ab initio $GW$ method is considered as the most accurate approach for calculating the band gaps of semiconductors and insulators. Yet its application to transition metal oxides (TMOs) has been hindered by the failure of traditional approximations developed for conventional semiconductors. In this work, we examine the effects of these approximations on the values of band gaps for ZnO, Cu$_2$O, and TiO$_2$. In particular, we explore the origin of the differences between the two widely used plasmon-pole models. Based on the comparison of our results with the experimental data and previously published calculations, we discuss which approximations are suitable for TMOs and why.

cond-mat.mtrl-sci↗