Bulletin of the American Physical Society
APS March Meeting 2019
Volume 64, Number 2
Monday–Friday, March 4–8, 2019; Boston, Massachusetts
Session P21: Emerging Trends in Molecular Dynamics Simulations and Data Analytics IIIFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP Chair: Roberto Car, Princeton University Room: BCEC 157B |
Wednesday, March 6, 2019 2:30PM - 3:06PM |
P21.00001: Deep Learning for Multi-Scale Molecular Modelling Invited Speaker: Weinan E Deep learning has emerged as a promising tool for a variety of applications |
Wednesday, March 6, 2019 3:06PM - 3:42PM |
P21.00002: Predictive Atomistic Simulations of Materials using SNAP Data-Driven Potentials Invited Speaker: Aidan Thompson Molecular dynamics (MD) is a powerful materials simulation method whose accuracy is limited by the interatomic potential (IAP). In many materials science applications suitably accurate potentials simply do not exist. SNAP is an automated methodology for generating application-specific IAPs using large and diverse datasets of quantum electronic structure calculations. The SNAP IAP is formulated in terms of a set of general four-body geometric invariants that characterize the local neighborhood of each atom. This approach has been used to develop potentials for diverse applications, including metal plasticity (Ta), defects in III-V semiconductors (InP), and fusion energy materials (W/Be/He/H). In each case, the SNAP IAP is fit to density functional theory calculations of energy, force, and stress for many small configurations of atoms. Cross-validation analysis and evaluation on test problems are used to further improve IAP fidelity and robustness. Varying the number of geometric descriptors allows a continuous tradeoff between computational cost and accuracy. The resultant potentials enable high-fidelity MD simulations of these materials, providing insight into their behavior on lengthscales and timescales unreachable by other methods. The relatively large per-atom computational cost of SNAP is offset by combining LAMMPS' spatial parallel algorithms with Kokkos-based hierarchical multithreading, enabling the efficient use of large CPU and GPU clusters, allocating only a few atoms to each node. Recent extensions of the SNAP approach include multi-element geometric descriptors and the use of higher-order terms. |
Wednesday, March 6, 2019 3:42PM - 3:54PM |
P21.00003: Machine learning of reaction pathways in chemical vapor deposition of MoS2 monolayers Aravind Krishnamoorthy, Pankaj Rajak, Aiichiro Nakano, Rajiv Kalia, Priya Vashishta Scalable synthesis of two dimensional (2D) materials is a major bottleneck to more widespread adoption of layered material-based devices. Chemical vapor deposition (CVD) has emerged as a viable method for large-scale synthesis of 2D materials. However, neither experiment nor theory has been able to decipher mechanisms and selection rules, or make predictions of optimized growth parameters. Experimental challenges stem from the use of probes like TEM to characterize CVD growth reactions in situ under elevated temperatures and pressures. Computational synthesis, which simulates CVD processes using reactive molecular dynamics methods provides the atomistic resolution necessary for the deduction of reaction mechanisms. Here we use neural networks trained on trajectories from several hundred simulations of computational synthesis of MoS2 monolayers to uncover the dependence of product stoichiometry, crystallinity and phase distribution on reaction parameters like temperature, sulfur and hydrogen partial pressures, thus paving the way for rational design of CVD synthesis techniques. |
Wednesday, March 6, 2019 3:54PM - 4:06PM |
P21.00004: Development of artificial neural network potential for hexagonal boron nitride with and without defects Talat S. Rahman, Duy Le Recently, defects in hexagonal boron nitride (h-BN) have been shown to play an important role in determining its novel chemical and optical properties, which may have a variety of possible technological applications. Characterization of the structure and dynamics of these defects would be most facilitated by the availability of accurate interatomic potentials that enable large length and time scale simulations. In this work, we will summarize our development of an artificial neural network (ANN) potential for h-BN with and without defects. The ANN potential was trained by using about 70000 data points obtained from ab initio molecular dynamic simulations of point defect on a (6x6) h-BN layer. The trained ANN potential is capable of producing system energetics in agreement with that obtained from density functional theory (within few meV per atom for both training and validation). The structure and dynamics of defects and grain boundaries in h-BN using the ANN potential will be presented and results compared with available experimental data. |
Wednesday, March 6, 2019 4:06PM - 4:18PM |
P21.00005: Active-learning strategy for the development of application-specific machine-learning force fields Huan Tran, Rohit Batra, James Chapman, Chiho Kim, Anand Chandrashekaran, Ramamurthy Ramprasad Emerging data-driven approaches in materials science have triggered the development of a plethora of machine-learning (ML) force fields (FFs). In practice, they are constructed by training a statistical model on a reference database to predict potential energy or atomic forces. While most of the FFs can accurately recover the reference data, some of them are becoming useful for actual molecular dynamics simulations. In this work, we develop a simple active-learning strategy for the development of ML FFs targeted at specific simulations (applications). The strategy involves (1) preparing and fingerprinting a diverse reference database of atomic configurations and forces, (2) generating a pool of ML FFs by learning the reference data, (3) validating the FFs against a series of targeted applications, and (4) selectively and recursively improving the FFs that remain unsuitable for a given application while keeping their performance on other applications uncompromised. We demonstrate this strategy by developing a series Al and Cu ML FFs that can simultaneously be used for various applications, including (elastic) stress/strain analysis, stacking-fault energy calculations, and melting simulations. This strategy is generalizable, i.e., it may be used for other materials as well. |
Wednesday, March 6, 2019 4:18PM - 4:30PM |
P21.00006: Deep Generative Model of Interfacial Structures in Phase Transformation of
an MoWSe2 Monolayer Pankaj Rajak, Aravind Krishnamoorthy, Aiichiro Nakano, Rajiv Kalia, Priya Vashishta Optical and electrical properties of two-dimensional layered materials can be tuned by mechanical straining, which induces transformations from semiconducting to metallic phases. We use deep generative variational autoencoder (VAE) model, trained by molecular dynamics simulation data of dynamic fracture in an MoWSe2 monolayer, to predict transition pathways consisting of novel intermediate structures (a and b) between the semiconducting (2H) and metallic (1T) phases. In addition, a conditional variational autoencoder (CVAE) is used to generate intermediate structures such as a or b , and defects. Structures synthesized from VAE and CVAE are validated by quantum simulations based on density functional theory. Quantum simulations show that structures generated by VAE and CVAE are stable and can be used for nanoelectronics applications. |
Wednesday, March 6, 2019 4:30PM - 4:42PM |
P21.00007: Magnetism and superconductivity in amorphous carobn Yuki Sakai, James Chelikowsky, Marvin L Cohen We study magnetic and superconducting properties of amorphous carbon based on molecular dynamics simulations. First we use spin constrained first-principles simulations to obtain amorphous carbon structures with a desired magnetization. We show that the existence of sp2-like threefold coordinated carbon atoms plays an important role in causing magnetism in amorphous carbon. We predict detailed geometries of threefold carbon atoms that induce the magnetic order in amorphous carbon. We also consider the effect of boron doping on superconducting properties of amorphous carbon. By considering amorphous structures with various sp2:sp3 ratios, we find that sp3-hybridized atoms are necessary for high superconducting transition temperature in contrast to the magnetism. |
Wednesday, March 6, 2019 4:42PM - 4:54PM |
P21.00008: Magnetostriction and Long-Range Interactions in Coupled Spin and Lattice Dynamics Julien Tranchida, Mitchell A Wood, Attila Cangi, Stan G. Moore, Pascal Thibaudeau, Steven James Plimpton, Aidan Thompson A scalable and symplectic algorithm for coupled spin dynamics and molecular dynamics was |
Wednesday, March 6, 2019 4:54PM - 5:06PM |
P21.00009: Employing autoencoders for configuration space sampling: Application to small molecules. Igor Poltavskyi, Alexandre Tkatchenko The behavior of molecular systems in different equilibrium physical processes or chemical reactions is governed by the free energy (FE). Calculations of FE require a thorough sampling of configuration space for given external conditions. State-of-the-art sampling techniques, based on molecular dynamics, are formally applicable to systems of arbitrary size. However, in practice, they suffer from the curse of dimensionality and the limitation of the time step by the fastest process present in the system. These make FE calculations for complex molecules, where the entropy effects are of utmost importance, extremely challenging and computationally demanding. Here we propose to use a machine-learning approach based on autoencoders to generate new sampling configurations. By transition to the feature space, we effectively decrease the dimensionality of the problem and resolve the time step limitation. Training autoencoders cost only a small fraction of statistically converged molecular dynamics simulations, paving the way to efficient calculations of thermodynamic properties for complex molecular systems. |
Wednesday, March 6, 2019 5:06PM - 5:18PM |
P21.00010: Atomistic mechanisms of phase transitions from Machine Learning Rodrigo Freitas, Evan Reed Identifying and characterizing atomic mechanisms of phase changes is of fundamental importance in the study of the kinetics of the nucleation and growth process. This talk will describe a new approach in extracting information from atomistic simulations of phase transitions using Machine Learning (ML) methods. In this approach the local neighborhood of atoms is characterized in terms of symmetry functions that are used as input to a ML algorithm trained to identify atomic rearrangements leading to structural transformations. The application of the method is illustrated using Molecular Dynamics simulations of crystallization from the liquid or amorphous phase. We also discuss how meaningful physical properties can be extracted from the output of the ML algorithm. |
Wednesday, March 6, 2019 5:18PM - 5:30PM |
P21.00011: Efficient Training of Neural-Network Interatomic Potentials with Atomic Forces Simon Batzner, Boris Kozinsky Neural-Network Interatomic Potentials have emerged as a promising method to accelerate time and length scales in Molecular Dynamics simulations of condensed systems. In the learning process, models are usually trained to a set of reference energies from electronic-structure calculations. In addition to total energies, local atomic forces are often also available by the Hellmann–Feynman theorem. Including atomic forces of quantum-mechanical accuracy in the training process can greatly enhance fidelity of the learned potential energy surface as it provides a wealth of additional information about its shape. However, this training scheme often comes at a much greater computational expense. In this talk, we address the speed and fidelity of integrating forces into the training process and examine strategies for rapid training of neural-network potentials for complex systems involving large training sets. |
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