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 Q60: Emerging Trends in Molecular Dynamics Simulations and Machine Learning VFocus
|
Hide Abstracts |
Sponsoring Units: DCOMP DSOFT GDS DPOLY Chair: Aravind Krishnamoorthy, University of Southern California Room: Room 419 |
Wednesday, March 8, 2023 3:00PM - 3:36PM |
Q60.00001: Predicting phase preferences of transition metal dichalcogenides using machine learning techniques Invited Speaker: Pratibha Dev Unlike most other 2D materials, transition metal dichalcogenides (TMDs) can adopt several phases that possess dramatically different electronic structure properties. One of the natural questions that arises is: What dictates the observed phase preference of TMDs? In this talk, I will discuss our recent work [Phys. Rev. Materials 6, 094007 (2022)], which combines high-throughput quantum mechanical computations with machine learning algorithms to address this old problem. Our analysis not only rediscovers known physicochemical attributes considered by earlier researchers, but goes beyond these attributes to discover other factors that were not previously known to influence the structural preferences displayed by TMDs. This work demonstrates how machine learning can be used to tackle old problems in Condensed Matter Physics. |
Wednesday, March 8, 2023 3:36PM - 3:48PM |
Q60.00002: Surface doping of MoO3-x on hydrogenated diamond Liqiu Yang, Thomas M Linker, Aravind Krishnamoorthy, Ken-ichi Nomura, Rajiv K Kalia, Aiichiro Nakano, Priya Vashishta Surface doping is reported to be a potential solution for diamond, which is known hard to be doped traditionally for high power electronics applications. While MoO3 was found to be an effective surface electron acceptor for hydrogen-terminated diamond with negative electron affinity, the effects of commonly existing oxygen vacancies remain elusive. We performed reactive molecular dynamics simulations to study the deposition of MoO3-x on hydrogenated diamond (111) surface and used first-principles calculations based on density functional theory to investigate the change transfer and electronic structures. Shift of the electronic band alignment is observed. Bader charge calculations show that MoO3-x are effective surface electron acceptor materials, where more electrons are transferred with increased O stoichiometry. The charge density difference after the deposition is used to characterize the spatial extent of doped holes. |
Wednesday, March 8, 2023 3:48PM - 4:00PM |
Q60.00003: Supercharging semi-empirical Quantum Chemistry with Machine Learning Martin Stoehr, Todd J Martinez Understanding the mechanisms that underpin chemical and biological processes under realistic conditions is crucial for the development of novel pharmaceutical and technological applications. Over the past years, machine learning (ML) has revolutionized our approach to gain such insights promising to bypass the prohibitive costs of ab initio calculations. However, purely data-driven ML models often suffer from limited transferability as well as reduced access to consistent molecular properties and chemical insights. Another computationally-efficient alternative is semi-empirical quantum chemistry (SEQC), which constructs effective, reduced-order Hamiltonians from parametric interaction models. While offering seamless access to electronic properties, SEQC can show poor accuracy outside the domain of its usual fixed, element-wise parametrization. Here, we present an extended SEQC formalism dynamically parametrized via ML. This hybrid approach provides an accurate description of organic molecules outperforming fixed-parameter SEQC as well as standard ML in terms of accuracy, transferability, and data efficiency. The hybrid framework allows us to avoid deep ML architectures without loss of performance, which together with the SEQC structure offers a high degree of interpretability. Combining computational efficiency, transferability, and scalability, hybrid SEQC/ML paves the way to an accurate understanding of chemical processes at practically relevant length and time scales. |
Wednesday, March 8, 2023 4:00PM - 4:12PM Author not Attending |
Q60.00004: Efficient calculation of χ parameters for polymer interactions from simulation Kevin Shen, Glenn H Fredrickson, M. Scott Shell, My Nguyen, Charles Li, Dan Sun, Nick Sherck, Paul R Irving, Venkatraghavan Ganesan The χ-parameter is a classic parameter used to describe the thermodynamics of polymeric systems, and its accurate determination is a bottleneck for applying coarse grained theoretical and simulation methods to real chemical systems. Usually, the χ parameter is characterized in the long chain length limit using the Random Phase Approximation, where fluctuations contributions are expected to be small, and much theory has been developed to characterize fluctuation effects associated with finite chain lengths. Unfortunately, to date the chain lengths required to accurately determine the χ parameter are only accessible to experiment and simplified, coarse-grained theoretical models of polymers. As such, there is much active research to develop alternative methods of determining χ in atomistically-detailed models, such as via free energy calculations. In this talk, we present impacts of chain length on estimates of the χ parameter from atomistic simulations. Using insights from the renormalized one-loop theory for diblock copolymers, we demonstrate an approach where we can accurately determine the χ parameter from short-chain simulations, and robustly extrapolate to the long chain limit. We demonstrate our approach an an array of real chemical systems, and compare to experimental and prior computational estimates of the χ-parameter. Our approach greatly reduces the chain lengths and computational cost required to accurately determine χ from simulations, and sets the stage for future, massive computational screening and determination of χ-parameters. |
Wednesday, March 8, 2023 4:12PM - 4:24PM |
Q60.00005: First-principles path-integral molecular dynamics study of ferroelectricity and isotope effects in KDP crystals with deep neural networks Bingjia Yang, Pinchen Xie, Roberto Car We study the ferroelectric phase transition of KDP and DKDP with all-atom path-integral molecular dynamics (PIMD) based on a neural network potential energy model trained on density functional theory with SCAN approximation. The effective mass of the proton/deuteron used in PIMD is determined by fitting the experimental H/D off-centering, to correct for the intrinsic error of SCAN. Then, a series of calculated geometric isotope effects including the Ubbelohde effects are in satisfactory agreement with the experiments. |
Wednesday, March 8, 2023 4:24PM - 4:36PM |
Q60.00006: Sobolev Sampling of Free Energy Landscapes Pablo Zubieta, Juan J De Pablo We present a family of fast sampling methods for classical and first principle molecular simulations of systems having rugged free energy landscapes. The methods represent a general strategy consisting of adjusting a model for the free energy as a function of one- or more collective variables as a simulation proceeds. Such a model is gradually built as a system evolves through phase space from both the frequency of visits to distinct states and generalized force estimates corresponding to such states. A common feature of the methods is that the underlying functional models and their gradients are easily expressed in terms of the same set of parameters, thereby providing faster and more effective fitting of the model from simulation data than other available sampling techniques. They also eliminate the need to train simultaneously more than one neural network as in the Combined Force-Frequency Sampling Method [1], while retaining the advantage of generating smooth and continuous functional estimates that enable biasing outside the support grid. Implementation of the methods is simple and, more importantly, they are found to provide gains in computational efficiency over existing approaches. |
Wednesday, March 8, 2023 4:36PM - 4:48PM |
Q60.00007: Modified metal-assisted exfoliation for low disorder, large monolayer devices Yangchen He, Daniel Rhodes Exfoliating two-dimensional(2D) materials down to monolayers is of great significance for fundamental exploration of emerging physical phenomenon because of their unique band structure and flexible control with electrostatic gate tuning. Mechanical exfoliation, which starts with the first isolation of graphene with scotch tape, provides high quality 2D material single crystals but usually suffers from insufficient monolayer yield and lateral size. Here, we report a facile modified metal-assisted mechanical exfoliation method to produce high quality 2D material monolayer arrays on bare Si chip without exposing to water or solution based etching process. The resulting transition metal dichalcogenide (TMD) monolayers can be assembled by standard dry transfer techniques into BN encapsulated devices with low disorder, which allows for further investigation of their intrinsic properties with tunable dual gates. This approach provides a simple and efficient way to produce high quality, dry transfer assembly compatible 2D material arrays with intrinsic properties, and thus accelerates potential fundamental research on properties and development of proof-of concept devices. |
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