arXiv subjects

Andrew Rohskopf

Publications and source records attributed to Andrew Rohskopf.

12 recordsLinked to original sources

arXiv2024

Materials Learning Algorithms (MALA): Scalable Machine Learning for Electronic Structure Calculations in Large-Scale Atomistic Simulations

We present the Materials Learning Algorithms (MALA) package, a scalable machine learning framework designed to accelerate density functional theory (DFT) calculations suitable for large-scale atomistic simulations. Using local descriptors of the atomic environment, MALA models efficiently predict key electronic observables, including local density of states, electronic density, density of states, and total energy. The package integrates data sampling, model training and scalable inference into a unified library, while ensuring compatibility with standard DFT and molecular dynamics codes. We demonstrate MALA's capabilities with examples including boron clusters, aluminum across its solid-liquid phase boundary, and predicting the electronic structure of a stacking fault in a large beryllium slab. Scaling analyses reveal MALA's computational efficiency and identify bottlenecks for future optimization. With its ability to model electronic structures at scales far beyond standard DFT, MALA is well suited for modeling complex material systems, making it a versatile tool for advanced materials research.

cond-mat.mtrl-sci↗

arXiv2023

Probing Iron in Earth's Core With Molecular-Spin Dynamics

Dynamic compression of iron to Earth-core conditions is one of the few ways to gather important elastic and transport properties needed to uncover key mechanisms surrounding the geodynamo effect. Herein a new machine-learned ab-initio derived molecular-spin dynamics (MSD) methodology with explicit treatment for longitudinal spin-fluctuations is utilized to probe the dynamic phase-diagram of iron. This framework uniquely enables an accurate resolution of the phase-transition kinetics and Earth-core elastic properties, as highlighted by compressional wave velocity and adiabatic bulk moduli measurements. In addition, a unique coupling of MSD with time-dependent density functional theory enables gauging electronic transport properties, critically important for resolving geodynamo dynamics.

cond-mat.mtrl-sci↗

arXiv2023

Shadow molecular dynamics and atomic cluster expansions for flexible charge models

A shadow molecular dynamics scheme for flexible charge models is presented, where the shadow Born-Oppenheimer potential is derived from a coarse-grained approximation of range-separated density functional theory. The interatomic potential, including the atomic electronegativities and the charge-independent short-range part of the potential and force terms, are modeled by the linear atomic cluster expansion (ACE), which provides a computationally efficient alternative to many machine learning methods. The shadow molecular dynamics scheme is based on extended Lagrangian (XL) Born-Oppenheimer molecular dynamics (BOMD) [Eur. Phys. J. B 94, 164 (2021)]. XL-BOMD provides a stable dynamics, while avoiding the costly computational overhead associated with solving an all-to-all system of equations, which normally is required to determine the relaxed electronic ground state prior to each force evaluation. To demonstrate the proposed shadow molecular dynamics scheme for flexible charge models using the atomic cluster expansion, we emulate the dynamics generated from self-consistent charge density functional tight-binding (SCC-DFTB) theory using a second-order charge equilibration (QEq) model. The charge-independent potentials and electronegativities of the QEq model are trained for a supercell of uranium oxide (UO2) and a molecular system of liquid water. The combined ACE + XL-QEq dynamics are stable over a wide range of temperatures both for the oxide and the molecular systems, and provide a precise sampling of the Born-Oppenheimer potential energy surfaces. Accurate ground Coulomb energies are produced by the ACE-based electronegativity model during an NVE simulation of UO2, predicted to be within 1 meV of those from SCC-DFTB on average during comparable simulations.

physics.chem-ph↗

arXiv2023

Exploring Model Complexity in Machine Learned Potentials for Simulated Properties

Machine learning (ML) enables the development of interatomic potentials that promise the accuracy of first principles methods while retaining the low cost and parallel efficiency of empirical potentials. While ML potentials traditionally use atom-centered descriptors as inputs, different models such as linear regression and neural networks can map these descriptors to atomic energies and forces. This begs the question: what is the improvement in accuracy due to model complexity irrespective of choice of descriptors? We curate three datasets to investigate this question in terms of ab initio energy and force errors: (1) solid and liquid silicon, (2) gallium nitride, and (3) the superionic conductor LGPS. We further investigate how these errors affect simulated properties with these models and verify if the improvement in fitting errors corresponds to measurable improvement in property prediction. Since linear and nonlinear regression models have different advantages and disadvantages, the results presented herein help researchers choose models for their particular application. By assessing different models, we observe correlations between fitting quantity (e.g. atomic force) error and simulated property error with respect to ab initio values. Such observations can be repeated by other researchers to determine the level of accuracy, and hence model complexity, needed for their particular systems of interest.

cond-mat.mtrl-sci↗

arXiv2022

Proper Orthogonal Descriptors for Efficient and Accurate Interatomic Potentials

We present the proper orthogonal descriptors for efficient and accuracy representation of the potential energy surface. The potential energy surface is represented as a many-body expansion of parametrized potentials in which the potentials are functions of atom positions and parameters. The Karhunen-Loève (KL) expansion is employed to decompose the parametrized potentials into a set of proper orthogonal descriptors (PODs). Because of the rapid convergence of the KL expansion, relevant snapshots can be sampled exhaustively to represent the atomic neighborhood environment accurately with a small number of descriptors. The proper orthogonal descriptors are used to develop interatomic potentials by using a linear expansion of the descriptors and determining the expansion coefficients from a weighted least-squares regression against a density functional theory (DFT) training set. We present a comprehensive evaluation of the POD potentials on previously published DFT data sets comprising Li, Mo, Cu, Ni, Si, Ge, and Ta elements. The data sets represent a diverse pool of metals, transition metals, and semiconductors. The accuracy of the POD potentials are comparable to that of state-of-the-art machine learning potentials such as the spectral neighbor analysis potential (SNAP) and the atomic cluster expansion (ACE).

cond-mat.mtrl-sci↗

arXiv2021

Thermophotovoltaic Efficiency of 40%

We report the fabrication and measurement of thermophotovoltaic (TPV) cells with efficiencies of >40%, which is a record high TPV efficiency and the first experimental demonstration of the efficiency of high-bandgap tandem TPV cells. TPV efficiency was determined by simultaneous measurement of electric power output and heat dissipation from the device via calorimetry. The TPV cells are two-junction devices comprising high-quality III-V materials with band gaps between 1.0 and 1.4 eV that are optimized for high emitter temperatures of 1900-2400°C. The cells exploit the concept of band-edge spectral filtering to obtain high efficiency, using high-reflectivity back surface reflectors to reject unusable sub-bandgap radiation back to the emitter. A 1.4/1.2 eV device reached a maximum efficiency of (41.1 +/- 1)% operating at a power density of 2.39 W/cm2 under an irradiance of 30.4 W/cm2 and emitter temperature of 2400°C. A 1.2/1.0 device reached a maximum efficiency of (39.3 +/- 1)% operating at a power density of 1.8 W/cm2 under an irradiance of 20.1 W/cm2 and emitter temperature of 2127°C. These cells can be integrated into a TPV system for thermal energy grid storage (TEGS) to enable dispatchable renewable energy. These new TPV cells enable a pathway for TEGS to reach sufficiently high efficiency and sufficiently low cost to enable full decarbonization of the grid. Furthermore, the high demonstrated efficiency also gives TPV the potential to compete with turbine-based heat engines for large-scale power production with respect to both cost and performance, thereby enabling possible usage in natural gas or hydrogen-fueled electricity production.

physics.app-ph↗

arXiv2021

A Computational Method for Studying Vibrational Mode Dynamics

The traditional picture of heat transfer in solids by atomic vibrations, also known as phonons, involves phonons scattering with each other like gas particles and is commonly referred to as the phonon gas model (PGM). This physical picture accounts for interactions among propagating (i.e., plane wave modulated) vibrational modes in an ideal crystal, but it becomes problematic when describing non-propagating modes arising in realistic non-idealized systems. Here, we introduce a more general formalism for studying phonon transport, which involves projection of the interatomic interactions themselves (i.e., not just the atom motion), onto the normal modes of the system. This shows, for the first time, how energy is exchanged between modes in real-time during molecular dynamics (MD) simulations, as opposed to other MD methods which use inferences based on correlations, or other time averaged schemes that do not preserve specific features in the real-time dynamics. Applying this formalism to the example case of modes interacting in a superlattice, we illustrate a new perspective on how phonon transport occurs, whereby individual normal modes share energy through specific channels of interaction with other modes. We also highlight that while a myriad of interaction pathways exist, only a tiny fraction of these pathways actually transfer significant amounts of energy, which is surprising. The approach allows for the prediction and simulation of these mode/phonon interactions, thus unveiling the real-time dynamics of phonon behavior and advancing our ability to understanding and engineer phonon transport.

cond-mat.mtrl-sci↗

arXiv2021

Development of Interatomic Potentials to Model the Interfacial Heat Transport of Ge/GaAs

Molecular dynamics simulations provide a versatile framework to study interfacial heat transport, but their accuracy remains limited by the accuracy of available interatomic potentials. In the past, researchers have adopted the use of analytic potentials and simple mixing rules to model interfacial systems, with minimal justification for their use. On the other hand, contemporary machine learned interatomic potentials have greater complexity, but have not seen rigorous validation of interfacial heat transport properties. Moreover, when fitting to ab initio data, it is not known whether interface systems small enough to be tractable for density functional theory calculations can produce reasonable interatomic force constants. These and related questions are studied herein using a model Ge/GaAs system, with a particular focus on the harmonic force constants (IFC2s) of the interface. The ab initio IFC2s are shown to recover near bulk-like values ~ 1-2 nm away from the interface, while also exhibiting a complex relationship across the interface that likely precludes any successful application of mixing rules. Two different spectral neighborhood analysis potentials (SNAP) are developed to model the interface. One is fit to the total forces, while the other is only used to describe the anharmonicity, with a Taylor expansion used to describe the harmonic portion of the potential. Each potential, along with their merits and issues are compared and discussed, which provides important insights for future work.

cond-mat.mtrl-sci↗

arXiv2021

Modeling interfacial phonon transport with normal mode dynamics

Traditional theories of interfacial heat transfer by atomic vibrations, also known as phonons, do not explain how vibrational mode interactions contribute to interface conductance. Traditional methods also use the concept of phonons as particles or waves which propagate and transmit energy through interfaces; such methods are therefore inapplicable to realistic non-crystalline systems where phonons are not propagating. Here we introduce a more general formalism of interfacial phonon transport, rigorously derived from interatomic interactions projected onto the normal modes of the system, showing for the first time how interactions between vibrational modes contribute to thermal interface conductance. This new physical picture is based on the concept of forces and energy exchange between modes, regardless of their propagating or non-propagating character, thus providing a general formalism to describe phonon transport in all solids regardless of disorder.

cond-mat.mtrl-sci↗

arXiv2020

Understanding Phonon Transport Properties Using Classical Molecular Dynamics Simulations

Predictive modeling of the phonon/thermal transport properties of materials is vital to rational design for a diverse spectrum of engineering applications. Classical Molecular Dynamics (MD) simulations serve as a tool to simulate the time evolution of the atomic level system dynamics and enable calculation of thermal transport properties for a wide range of materials, from perfect periodic crystals to systems with strong structural and compositional disorder, as well as their interfaces. Although MD does not intrinsically rely on a plane wave-like phonon description, when coupled with lattice dynamics calculations, it can give insights to the vibrational mode level contributions to thermal transport, which includes plane-wave like modes as well as others, rendering the approach versatile and powerful. On the other hand, several deficiencies including the lack of vibrationally accurate interatomic potentials and the inability to rigorously include the quantum nature of phonons prohibit the widespread applicability and reliability of Molecular Dynamics simulations. This article provides a comprehensive review of classical Molecular Dynamics based formalisms for extracting thermal transport properties: thermal conductivity and thermal interfacial conductance and the effects of various structural, compositional, and chemical parameters on these properties. Here, we highlight unusual property predictions, and emphasize on the needs and strategies for developing accurate interatomic potentials and rigorous quantum correction schemes.

cond-mat.mtrl-sci↗

arXiv2019

Argon Simulations with EM3, a New Modular Molecular Mechanics Program

Argon molecular dynamics (MD) simulations are performed with a newly developed MD program, Easy M(1)odular M(2)olecular M(3)echanics (EM3). The program was developed in an object-oriented fashion containing classes for each critical part of a functioning MD program. An organizational scheme for a general molecular mechanics program is therefore presented, along with the framework of the EM3 program. With the modular nature and open-source availability, EM3 can serve as a learning tool for newcomers to molecular simulations and code organization via object-oriented programming. Validations of the code are presented in comparison with Monte Carlo (MC) simulations of liquid argon at different densities and temperatures. A calculation of the self-diffusion coefficient for liquid argon is also performed, exhibiting the extendibility of EM3. This report comes packaged with the EM3 source code and examples, located at https://github.com/rohskopf/em3.

physics.comp-ph↗

arXiv2016

Phonon Optimized Potentials

Molecular dynamics (MD) simulations have been extensively used to study phonons and gain insight, but direct comparisons to experimental data are often difficult, due to a lack of empirical interatomic potentials (EIPs) for different systems. As a result, this issue has become a major barrier to realizing the promise associated with advanced atomistic level modeling techniques. Here, we present a general method for specifically optimizing EIPs from ab initio inputs for the study of phonon transport properties, thereby resulting in phonon optimized potentials (POPs). The method uses a genetic algorithm (GA) to directly fit to the key properties that determine whether or not the atomic level dynamics and most notably the phonon transport are described properly.

cond-mat.mtrl-sci↗