Shift-Collapse Acceleration of Generalized Polarizable Reactive Molecular Dynamics for Machine Learning-Assisted Computational Synthesis of Layered Materials

Abstract

Reactive molecular dynamics is a powerful simulation method for describing chemical reactions. Here, we introduce a new generalized polarizable reactive force-field (ReaxPQ+) model to significantly improve the accuracy by accommodating the reorganization of surrounding media. The increased computation is accelerated by (1) extended Lagrangian approach to eliminate the speed-limiting charge iteration, (2) shiftcollapse computation of many-body renormalized n-tuples, which provably minimizes data transfer, (3) multithreading with roundrobin data privatization, and (4) data reordering to reduce computation and allow vectorization. The new code achieves (1) weak-scaling parallel efficiency of 0.989 for 131,072 cores, and (2) eight-fold reduction of time-to-solution (T2S) compared with the original code, on an Intel Knights Landing-based computer. The reduced T2S has for the first time allowed purely computational synthesis of atomically-thin transition metal dichalcogenide layers assisted by machine learning to discover a novel synthetic pathway.

ReaxFF MD Simulations of Peptide-Grafted Gold Nanoparticles

“…Interestingly, these results show a consistent pattern similar to that of smaller-sized (diameter < 3 nm) gold nanoparticles interacting with thiol ligands. 65,68,69 Further investigation into the accuracy of the polarization, particularly in an aqueous environment, will be reported in our future publication using an improved polarizable QEq method, called ReaxPQ+, 70 which has recently been proposed based on a core−shell charge and implemented in the open source code of RXMD (ReaxFF reactive molecular dynamics). The exterior Au atoms typically have a slightly positive charge due to the interactions of the thiol ligands with the exterior atoms.…”

Chemical vapor deposition of 2D materials: A review of modeling, simulation, and machine learning studies

“…Liu et al. used reactive MD simulations to study the CVD growth of MoS 2 from MoO 3 and S precursors ( Liu et al., 2019 ). Subsequently, the authors used a machine-learning approach involving feedforward neural networks to identify the critical reaction mechanisms.…”

“…Subsequently, the authors used a machine-learning approach involving feedforward neural networks to identify the critical reaction mechanisms. The results from the training of 36,000 simulation datapoints revealed novel growth mechanisms which turned out to be fundamental for augmenting the experimental CVD growth of MoS 2 ( Liu et al., 2019 ). To summarize this section, ML-driven predictions are less expensive and faster than the traditional ab initio calculations and hence can be effectively used to obtain required information from a vast pool of data quickly.…”

Dielectric Polymer Property Prediction Using Recurrent Neural Networks with Optimizations

“…Compared to the vast success of machine learning (ML) for predicting structure–property relationships in molecular and material research, predicting the dielectric properties of polymers (i.e., dielectric polymer genome) is still in its infancy. − Recently, computationally synthesized polymer databases have been created, which involved highly accurate computation of dielectric constants based on the new generation of first-principles-informed polarizable reactive force field methods. − These new computational databases provide an ideal testing ground for the quantitative assessment of ML models for the dielectric polymer genome within a well-controllable environment.…”