Bulletin of the American Physical Society
APS March Meeting 2023
Volume 68, Number 3
Las Vegas, Nevada (March 5-10)
Virtual (March 20-22); Time Zone: Pacific Time
Session K62: Emerging Trends in Molecular Dynamics Simulations and Machine Learning IIIFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP Chair: Ken-ichi Nomura, University of Southern California Room: Room 417 |
Tuesday, March 7, 2023 3:00PM - 3:36PM |
K62.00001: Reinforcement Learning Agent for autonomous predictive material synthesis and transport pathways Invited Speaker: ankit mishra Predictive material synthesis strategies for promising materials are time consuming using current experimental and computational techniques. Similarly, various transport related processes such as diffusion of small molecules through porous media is a challenging application to study due to combinatorically high pathways possible. We use offline and online Reinforcement Learning (RL) to find optimal time series sequence facilitating the synthesis of desired materials and transport through porous media. In this study, we will demonstrate various RL strategies such as Deep Q Networks (DQN), Policy Based Learning (REINFORCE) and Tree Search Based Methods (MCTS) to solve these challenging problems applicable to a wide range of disciplines. Offline RL is used for predicting optimal material synthesis route for a chemical vapor deposition and physical vapor deposition process. Further, online RL is used in conjunction with reactive molecular dynamics to study transport of molecules through porous media in real time. |
Tuesday, March 7, 2023 3:36PM - 3:48PM |
K62.00002: Judicious Curation of DFT Machine Learning Datasets for Accurate, Flexible, and Transferrable Atomistic Potentials for Elemental Systems and Metal Oxides Wissam A Saidi, Christopher M Andolina, Pandu Wisesa Machine learning techniques have accelerated material discovery and expanded the impact of atomistic potentials by leveraging the predictive accuracy of density functional theory (DFT) to simulations with numbers of atoms and timescales unsuitable for DFT. We focus on the initial curation and subsequent systematic expansion of these training data for the machine-learning deep neural-network potentials (MLPs) training. Compact datasets were generated for element systems across the periodic table and transition metal oxides using clear and concise criteria applicable to most elemental systems or transition metal oxides. MLPs are validated after each iteration to gauge the impact of the new data and MPL accuracy by comparing calculated material properties with DFT reference values. Remarkably, we observe good transferability for material properties not included in the training (e.g., melting points). Furthermore, we explore the systematic dataset expansion to describe oxide surface energies and thermal expansion accurately. These MLPs effectively represent materials trained with small configurational datasets; they are transferrable and flexible, providing a launch point for the broad community to adapt and expand these data and their material discovery and optimization applications. |
Tuesday, March 7, 2023 3:48PM - 4:00PM |
K62.00003: Probing Thermomechanical Properties of Two-dimensional van der Waals Architectures Using Surface Acoustic Waves Anikeya Aditya, Nitish Baradwaj, ankit mishra, Ken-ichi Nomura, Aiichiro Nakano, Priya Vashishta, Rajiv K Kalia Surface acoustic waves (SAWs) propagate along material surfaces or at solid-air, solid-liquid, solid-solid, and solid-vacuum interfaces. SAWs are confined to surfaces within the depth of one wavelength, and the SAW velocity depends on the elastic properties of the material. Hence SAWs can be used to determine thermomechanical properties like Young’s modulus and thermal conductivity of 2D materials. Inspired by a recent electron microscopy experiment [1], we use molecular dynamics (MD) simulations to investigate the effect of SAWs on the thermomechanical behavior of MoS2 mono- and bilayers. The MD results for Young’s modulus and thermal conductivity of 2D MoS2 are in excellent agreement with experiments and density functional theory calculations. We find nanopores have a dramatic effect on the thermal conductivity, which drops by an order of magnitude in a nanoporous MoS2 monolayer. We also examine the effect of SAWs on Moiré patterns between 2D van der Waals materials. |
Tuesday, March 7, 2023 4:00PM - 4:12PM |
K62.00004: An unsupervised data mining methodology for analysis of molecular dynamics sampling of local coordination Fabrice Roncoroni, Ana Sanz Matias, Siddharth Sundararaman, David Prendergast Molecular dynamics (MD) simulations present a data-mining challenge, given that they can generate a considerable amount of data but often rely on limited or biased human interpretation to examine their information content. By not asking the right questions of MD data we may miss critical information hidden within it. We combine dimensionality reduction (UMAP) and unsupervised hierarchical clustering (HDBSCAN) to quantitatively characterize the coordination environment of chemical species within MD data. By focusing on local coordination, we significantly reduce the amount of data to be analyzed by extracting all distinct molecular formulas within a given coordination sphere. We then efficiently combine UMAP and HDBSCAN with alignment or shape-matching algorithms to classify these formulas into distinct structural isomer families. The outcome is a quantitative mapping of the multiple coordination environments present in the MD data. The method was employed to reveal details of cation coordination in electrolytes based on molecular liquids and polymers. |
Tuesday, March 7, 2023 4:12PM - 4:24PM |
K62.00005: Evaluating robustness of machine learned force fields with enhanced sampling methods Gustavo R Perez Lemus, Juan J De Pablo, Pablo Zubieta, Yezhi Jin Molecular Dynamics simulations are increasing its use and applications in materials science and engineering. For that, Machine Learning force fields have emerged as a useful tool to have ab initio accuracy in energy and forces while keeping the speed of a classical simulation. However, validation for such potentials are not sufficient to demonstrate the effectiveness of the simulation as the evaluation only involves the force and energy accuracies while overlooking the stability of systems. Here we explore the robustness of different machine learned force fields for different systems and machine learning models such as DeePMD, Graph Neural Network force fields, Gaussian Approximation Potential (GAP) by performing molecular dynamics simulations and evaluating free energy landscapes as function of appropriate collective variables. The simulations are performed with PySAGES, a python library for advanced sampling simulations, using either ASE or JAX-MD as backends. We showcase how some models are susceptible to exhibit undesired behavior such as bonds breaking, while others maintain the most relevant set of features that are needed to take advantage of them as widely used force fields. We show how to use PySAGES as a tool for quickly evaluating the soundness of ML force fields. |
Tuesday, March 7, 2023 4:24PM - 4:36PM |
K62.00006: Structure and Dielectric Properties of Water and Aqueous Solutions Using Neural Network Quantum Molecular Dynamics RURU MA, Aravind Krishnamurthy, Nitish Baradwaj, Ken-ichi Nomura, Kohei Shimamura, Pankaj Rajak, Fuyuki Shimojo, Aiichiro Nakano, Rajiv K Kalia, Priya Vashishta The excellent solvation properties of water are responsible for its role in life-sustaining biological processes and for its importance to several technological applications. The structure and dynamics of aqueous solutions are highly complex, composed of transient hydrogen bonding and continuously reorganized solvation shells, which are difficult to characterize experimentally. Further, the dynamical response of these systems that are dominated by the restructuring of the hydrogen bond network is still unknown. In this study, we use neural network quantum molecular dynamics (NNQMD) to capture quantum-mechanically accurate molecular configurations and evolution of the hydrogen bond network in an aqueous solution of LiOH as a function of concentration. We further probe the dynamic response of LiOH solutions by quantifying the dielectric constant, , using the variance of the dipole moment along a long molecular dynamics trajectory in the canonical ensemble. The polarization fluctuations are computed with quantum accuracy using a secondary neural network that uses Wannier Functions to encode many-body polarization effects. |
Tuesday, March 7, 2023 4:36PM - 5:12PM |
K62.00007: HubbardNet: efficient predictions of the Bose-Hubbard model spectrum with deep neural networks Invited Speaker: Ziyan Zhu We present a deep neural network (DNN)-based model, the HubbardNet, to variationally solve for the ground state and excited state wavefunctions of the one-dimensional and two-dimensional Bose-Hubbard model on a square lattice. Using this model, we obtain the Bose-Hubbard energy spectrum as an analytic function of the Coulomb parameter, U, and the total number of particles, N, from a single training, bypassing the need to solve a new hamiltonian for each different input. We show that the DNN-parametrized solutions have excellent agreement with exact diagonalization while outperforming exact diagonalization in terms of computational scaling, suggesting that our model is promising for efficient, accurate computation of exact phase diagrams of many-body lattice hamiltonians. |
Tuesday, March 7, 2023 5:12PM - 5:24PM |
K62.00008: Ab initio-based deep potential simulation of 2D confined water Iman Ahmadabadi, Marcos Calegari Andrade, Pablo M Piaggi, Roberto Car Due to the wide range of applications in technological and living systems, the structure and dynamics of water under two-dimensional (2D) confinement have been the focus of several experimental and theoretical studies. In particular, the confinement of water in graphite nanochannels has received considerable attention due to its ability to provide a fundamental understanding of the behavior of water at interfaces. However, from a computational point of view, such a study remains challenging due to the very high cost of calculating physical observables that can be compared against experiments. Here, we apply a first-principles-based deep neural network potential (DP) that enables us to simulate water confined between graphite walls with quantum accuracy, and over time and length scales well beyond the reach of conventional first-principles methods. From the DP molecular dynamics simulation, we calculate the infrared spectra. In addition, it allows us to calculate the structure factors, which along with other structural and dynamical properties, provide insight into water under 2D confinement. |
Tuesday, March 7, 2023 5:24PM - 5:36PM |
K62.00009: Thermodynamics and Phase Behavior of Alkali Metal Mixture Using Ab-initio-based Machine Learning Interatomic Potentials Ayu Irie, Akihide Koura, Kohei Shimamura, Fuyuki Shimojo Alkali metal mixtures have characteristic composition dependence of melting points. It is experimentally reported that they have lower melting points than pure alkali metals and have lowest melting points at the different concentrations depending on the combination of alkali metals. |
Tuesday, March 7, 2023 5:36PM - 5:48PM Author not Attending |
K62.00010: Physically and chemically inspired kernel-based neural network for constructing accurate and efficient machine learning force fields for hundreds of atoms. Igor Poltavskyi, Anton Charkin-Gorbulin, Artem Kokorin, Alexandre Tkatchenko, Grgory Cordeiro Fonseca Accurate and efficient machine learning (ML) models have become a game-changing element in modern (bio)chemistry and materials science during the last decade. Among other applications, the ability to run ab initio level atomistic molecular dynamics for nanoseconds enabled an understanding of multiple challenging processes observed by experiments. The kernel-based methods and artificial neural networks are the two primary techniques for constructing ML models, each having its strengths and weaknesses. In our work, we combined the advantages of both approaches in a single ML architecture capable of accurately reconstructing force fields (FF) for systems of varying complexity from small (MD17 dataset [1]) and intermediate-sized (MD22 dataset [2]) molecules and molecular complexes to large-scale periodic systems with hundreds of atoms per unit cell (MAPbI3). The proposed MLFFs efficiently capture the multiscale nature of interatomic interactions and the indistinguishability of atoms with identical chemical environments. Moreover, the architecture can be equally used as a global ML model providing the ultimate accuracy and in a localized version enabling transferability. |
Tuesday, March 7, 2023 5:48PM - 6:00PM |
K62.00011: Bonded Potential Dynamics in Chemically-Specific Coarse-Grained Models of Polymers Frederick R Phelan, Lilian C Johnson In recent work [J. Chem. Phys. 154,084114 (2021)], we studied a coarse-grain (CG) model that aims to preserve both chemical specificity and dynamics of a polymer melt. We used iterative Boltzmann inversion (IBI) to parameterize the conservative potential and Langevin dynamics as a means to recover the all-atom (AA) dynamics. Here we further our analysis of CG models by examining the influence of bonded potential dynamics on CG diffusion. In IBI, CG potentials are parametrized such that the structural distributions match those of the AA reference system. However, after correcting CG diffusion via a Langevin friction factor, the dynamics of the CG potentials are still much faster than the AA target potential dynamics as measured by autocorrelation functions. To correct this, rescaling factors are computed and applied to the CG system. We show that rescaling to match the autocorrelation functions of the distributions brings about a better match between CG and AA dynamic properties. The effect of the refined potentials on property predictions both with both bonded potential correction and Langevin dynamics are analyzed. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700