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 D60: Emerging Trends in Molecular Dynamics Simulations and Machine Learning IFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP Chair: Priya Vashishta, University of Southern California Room: Room 419 |
Monday, March 6, 2023 3:00PM - 3:36PM |
D60.00001: Exascale Simulations of Quantum Materials Guided by AI and Quantum Computing Invited Speaker: Aiichiro Nakano Computing landscape is evolving rapidly. Exascale computers have arrived, and quantum supremacy has been demonstrated for several problems, while artificial intelligence (AI) is transforming every aspect of science and engineering. Atomistic simulations at the exa-quantum-AI nexus are revolutionizing quantum materials research. I will describe research and education on atomically thin two-dimensional and other materials using our AI and quantum-computing enabled exascale materials simulator (AIQ-XMaS). Specifically, I will describe (1) self-assembly of layered material metastructures for scalable and robust manufacturing of quantum emitters for future quantum information science and technology; and (2) excited-state neural-network quantum molecular dynamics (NNQMD) trained by first-principles nonadiabatic quantum molecular dynamics (NAQMD) to prove the exciting concept of picosecond optical, electrical and mechanical control of symmetric breaking in topological ferroelectric skyrmion and skyrmionium for emerging ultralow-power polar topotronics. |
Monday, March 6, 2023 3:36PM - 3:48PM |
D60.00002: Large scale simulations of soft materials with equivariant deep learning potentials Simon L Batzner, Albert Musaelian, Boris Kozinsky Access to an accurate and computationally efficient description of the energy and atomic forces of many-atom systems is a long-standing goal in the natural sciences. Neural message passing potentials have emerged as the leading paradigm toward this goal over the past years. However, their propagation mechanics makes parallel computation difficult and inherently limits the length scales they can model. Strictly local descriptor-based methods have been scaled to massive systems, but currently lack the accuracy observed in message passing approaches. We have recently introduced the Allegro model, a fully local equivariant deep learning interatomic potential that is massively parallelizable while retaining the high accuracy of equivariant message passing potentials. Allegro obtains state-of-the-art accuracy on a series of benchmarks and has been demonstrated to recover structural and kinetic properties of an amorphous phosphate electrolyte in great agreement with AIMD. Exploiting the strict locality, the method finally has been scaled to a simulation of >100 million atoms. Here, we will show how we can leverage the unique combination of scale and accuracy of Allegro to study complex, soft materials using large-scale Molecular Dynamics simulations. |
Monday, March 6, 2023 3:48PM - 4:00PM |
D60.00003: Multi-Scale Neural Network Molecular Dynamics Simulations for Polar Topology Control in Next Generation Ferroelectric Materials. Thomas M Linker, Ken-ichi Nomura, Shogo Fukushima, Rajiv K Kalia, Aravind Krishnamoorthy, Aiichiro Nakano, Kohei Shimamura, Fuyuki Shimojo, Priya Vashishta Quantum simulations methods such as quantum molecular dynamics offer highly accurate understandings of material vibrational and structural properties, but are seriously limited in the length and time scales they are able to investigate due to the large computational cost. Over the recent years the incorporation of machine-learning based inter-atomic potentials trained on quantum data has opened the possibility to study much larger spatiotemporal scales at quantum accuracy due to orders of magnitude reduction in computation cost of the machine-learning model. In this talk we will focus on the use of machine learning based molecular dynamics in the rapidly developing field of polar “topotronics”, or the control of polar topologies in ferroelectric materials for development of next generation topological based electronics. We will discuss the use of multi-scale neural network quantum molecular dynamics simulations to explore optical, mechanical, and electrical control of ferroelectric materials. The demonstrated multi-scale quantum simulation and machine learning frameworks are not limited to ferroelectrics and offer an exciting new avenue for exploring quantum material dynamics |
Monday, March 6, 2023 4:00PM - 4:12PM |
D60.00004: Learning the committor probability using data-driven path collective variables Arthur France-Lanord, Hadrien Vroylandt, Fabio Pietrucci, Benjamin Rotenberg, A. Marco Saitta, Mathieu Salanne We have developed a new method aimed at predicting the committor probability for a system of two metastable states, focusing on a data-driven generalization of path collective variables. In this approach, we perform kernel ridge regression of the committor within a first projection of the Cartesian coordinates onto a subspace formed by collective variables (CVs). This subspace can be arbitrarily formed, e.g. by selecting a list of typical scalar CVs, or by using more abstract, high-dimensional CVs. The obtained collective variable is a one-dimensional estimatior of the committor, which makes for easy comparison of CV subspaces. This CV can subsequently be used for molecular dynamics or Monte Carlo simulations, with optional biasing. We apply this methodology to the well-known problem of ion pairing in water, with a focus on LiF, and show that better estimators of the committor can be obtained by taking into account information related to the solvent compared to the only interionic distance. |
Monday, March 6, 2023 4:12PM - 4:24PM |
D60.00005: Billions of Atoms with Machine Learning Interatomic Potentials: Application to Direct Heterogeneous Reactive Dynamics Anders Johansson, Yu Xie, Cameron J Owen, Jin Soo Lim, Lixin Sun, Jonathan P Vandermause, Boris Kozinsky Machine learning interatomic potentials (MLIPs) have become a prevalent approach to bridging the gap between slow-but-accurate ab initio calculations and fast-but-inaccurate empirical potentials for molecular dynamics. Among MLIPs, there is a pareto front of models with different tradeoffs between accuracy and speed. The FLARE interatomic potential aims to push the boundary of scalability and performance, while maintaining sufficient accuracy to study complex, reactive systems. |
Monday, March 6, 2023 4:24PM - 4:36PM |
D60.00006: Machine learning based force-fields for strongly anharmonic materials Martin Callsen, Mei-Yin Chou Computing materials properties at finite temperatures, such as thermal or electric conductivity, poses a particular challenge for anharmonic materials because the commonly used phonon-based perturbation theories break down and going beyond the harmonic approximation is costly and tedious. This applies to some extent also to other materials, since at finite temperatures the structure and the symmetry are thermodynamic averages of energetically similar, lower symmetry structures and therefore anharmonic effects are more common than generally assumed. Alternatively, sufficiently long ab initio molecular dynamics trajectories would contain the required information to calculate the abovementioned properties, however the computational cost of is in most cases prohibitively high. Recently, machine learning based force-fields have made obtaining said the trajectories feasible. However, those methods come with the intrinsic problem of choosing appropriate training data, since encountering a not well-represented structure during the molecular dynamics will lead to failure of the calculation, which is an issue for anharmonic materials in particular. |
Monday, March 6, 2023 4:36PM - 5:12PM |
D60.00007: DFT aided machine learning interatomic potentials for realistic simulations of low dimensional system Invited Speaker: Duy Le Challenges in atomistic simulations of materials include limitations in time and length scales and lack of accurate interatomic forcefields. In this talk, we will discuss the application of machine learning to develop accurate interatomic forcefields for reducing computational cost and accessing time and length-scale relevant to experiments in atomistic simulations of materials through two examples: the semiconducting-metallic transition of the Si(100) surface, and heat transport through grain boundaries in hexagonal boron nitride. Artificial Neural network (ANN) interatomic potentials are trained using large databases of structure-energy relationship of small representative systems obtained from ab initio molecular dynamic simulations, based on density functional theory (DFT). The trained ANN potentials produce potential energies that are in good agreement with those from DFT. For Si(001), results from our molecular dynamics (MD) simulations using ANN potential show that the asymmetric, buckled structure of Si dimer exists in the temperature range 300K to 900K, but the increased dimer flipping rate leads to more time in the symmetric configuration making the surface metallic. Moreover, our simulations indicate persistence of Si adatoms, formed by the breakage of Si-dimers, above 800K, that can move around on the surface, which explain the jump in metallicity of the surface around this temperature, as seen in experiments [1]. For h-BN, non-equilibrium MD simulations using ANN potential are performed to study thermal conductivity of h-BN. We show that thermal properties calculated by our simulations for pristine h-BN agree well with experimental data [2] and that grain-boundaries in h-BN hinders heat transport through h-BN, increasing its thermal resistant. |
Monday, March 6, 2023 5:12PM - 5:24PM |
D60.00008: Decoding the Hydrogen Bond Network of Water in Carbon Nanotubes with Atomistic Simulations Marcos Calegari Andrade, Tuan Anh Pham Improved understanding of confined aqueous solutions is critical for a wide range of applications, from water desalination to energy storage. However, probing the structure of confined water with experimental techniques remains a significant challenge. In this work, we use large-scale molecular dynamics simulations with a machine learning potential to compute the infrared (IR) spectrum of water in carbon nanotubes (CNTs). The theoretical spectra are then combined with existing experiments to elucidate how and when the hydrogen bond network of water is altered by confinement. We confirm that water undergoes an order-disorder transition within a CNT diameter of approximately 1.2 nm. For wider CNTs, confinement imposes a monotonic disruptive effect on the hydrogen bond network of water, which decays slowly with CNT diameter. In sharp contrast, confinement in narrower CNT pores affects water structure in a complex and non-linear fashion. Our study reveals peculiar hydrogen-bond network of confined water and it provides new interpretation for infrared spectroscopic measurements. This work also offers a general platform to simulate water in CNTs with quantum accuracy at time and length scales beyond-reach of conventional quantum-mechanical approaches. |
Monday, March 6, 2023 5:24PM - 5:36PM |
D60.00009: Kibble-Zurek Scaling Study of Phase Transition in Barium Titanate (BaTiO3) Nitish Baradwaj, Aravind Krishnamoorthy, Ken-ichi Nomurra, Aiichiro Nakano, Rajiv K Kalia, Priya Vashishta Far-from-equilibrium phenomena are some of the grand challenges of modern material science. Perovskite materials and their defect structures play an important role in a variety of solid-state devices. The understanding of non-equilibrium physics in the vicinity of symmetry-breaking phase transformations has progressed on many fronts, from scaling properties of dynamical correlation functions to the Kibble-Zurek (KZ) scaling behavior. By quenching the bulk single crystals of BaTiO3 at different rates through the ferroelectric phase transition, we see that defect density and the domain size distribution follows the KZ scaling law. Simulated Diffuse Neutron Scattering studies were also carried out to further study the scaling behavior. |
Monday, March 6, 2023 5:36PM - 5:48PM |
D60.00010: Indicator configuration: An information-matching method of data reduction for training interatomic potential Yonatan Kurniawan, Mark K Transtrum, Cody L Petrie, Dylan B Bailey Interatomic Potentials (IPs) are often trained to by fitting the IP parameters to the energies, atomic forces, or similar quantities for many atomic configurations. Typically, these training quantities are obtained from DFT calculations, and collecting data from enough unique configurations to constrain all of the IP parameters is computationally expensive. A critical problem is identifying when the training data is sufficient to constrain the predictions of the IP for material properties of interest. We present an information-matching method for selecting a minimal set of configurations, i.e., indicator configurations, that constrain the predictions of an IP for target material properties. Central to our analysis is the Fisher Information Matrix (FIM), that quantifies how much information data carries about the parameters of an IP. We calculate the FIM for the target quantities of interest and for, e.g., the energy and forces of each candidate configuration. Then, we down-select from these candidate configurations so that their combined FIM matches that of the quantities of interest, i.e., the indicator configurations are those whose information content is the same as the target predictions. We demonstrate this method on the Stillinger--Weber potential for several systems and target materials properties. In addition to improving the efficiency of the data-generation process, the indicator configurations reveal the physics and mechanisms relevant to the materials properties of interest. |
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