Skip to main content
OSTI.GOVU.S. Department of Energy Office of Scientific and Technical Information
U.S. Department of Energy
Office of Scientific and Technical Information
 

Quasi-Classical Trajectory Calculation of Rate Constants Using an Ab Initio Trained Machine Learning Model (aML-MD) with Multifidelity Data

Journal Article · · Journal of Physical Chemistry. A, Molecules, Spectroscopy, Kinetics, Environment, and General Theory
 [1];  [1];  [2];  [2];  [1]
  1. Princeton Univ., NJ (United States)
  2. Argonne National Laboratory (ANL), Argonne, IL (United States)
+ Show Author Affiliations

Machine learning (ML) provides a great opportunity for the construction of models with improved accuracy in classical molecular dynamics (MD). However, the accuracy of a ML trained model is limited by the quality and quantity of the training data. Generating large sets of accurate ab initio training data can require significant computational resources. Furthermore, inconsistent or incompatible data with different accuracies obtained using different methods may lead to biased or unreliable ML models that do not accurately represent the underlying physics. Recently, transfer learning showed its potential for avoiding these problems as well as for improving the accuracy, efficiency, and generalization of ML models using multifidelity data. In this work, ab initio trained ML-based MD (aML-MD) models are developed through transfer learning using DFT and multireference data from multiple sources with varying accuracy within the Deep Potential MD framework. Further, the accuracy of the force field is demonstrated by calculating rate constants for the H + HO2 → H2 + 3O2 reaction using quasi-classical trajectories. We show that the aML-MD model with transfer learning can accurately predict the rate constants while reducing the computational cost by more than five times compared to the use of more expensive quantum chemistry training data sets. Hence, the aML-MD model with transfer learning shows great potential in using multifidelity data to reduce the computational cost involved in generating the training set for these potentials.

Research Organization:
Argonne National Laboratory (ANL), Argonne, IL (United States)
Sponsoring Organization:
USDOE Office of Science (SC), Basic Energy Sciences (BES). Chemical Sciences, Geosciences & Biosciences Division (CSGB)
Grant/Contract Number:
AC02-06CH11357; SC0023357; SC0021135
OSTI ID:
2502016
Journal Information:
Journal of Physical Chemistry. A, Molecules, Spectroscopy, Kinetics, Environment, and General Theory, Journal Name: Journal of Physical Chemistry. A, Molecules, Spectroscopy, Kinetics, Environment, and General Theory Journal Issue: 17 Vol. 128; ISSN 1089-5639
Publisher:
American Chemical SocietyCopyright Statement
Country of Publication:
United States
Language:
English

References (46)

Benchmarking boron cluster calculations: Establishing reliable geometrical and energetic references for Bn (n = 1–4) journal September 2023
Vectorization of the general Monte Carlo classical trajectory program VENUS journal October 1991
Semiclassical dynamics based on quantum trajectories journal October 2002
Ab initio molecular dynamics benchmarking study of machine-learned potential energy surfaces for the HBr+ + HCl reaction journal June 2023
Reactive molecular dynamics simulation and chemical kinetic modeling of pyrolysis and combustion of n-dodecane journal February 2011
DeePMD-kit: A deep learning package for many-body potential energy representation and molecular dynamics journal July 2018
86 PFLOPS Deep Potential Molecular Dynamics simulation of 100 million atoms with ab initio accuracy journal February 2021
LAMMPS - a flexible simulation tool for particle-based materials modeling at the atomic, meso, and continuum scales journal February 2022
Classical and reactive molecular dynamics: Principles and applications in combustion and energy systems journal July 2023
Machine learning approaches for analyzing and enhancing molecular dynamics simulations journal April 2020
Machine Learning Force Fields journal March 2021
Density Functionals for Hydrogen Storage: Defining the H2Bind275 Test Set with Ab Initio Benchmarks and Assessment of 55 Functionals journal June 2020
TorchMD: A Deep Learning Framework for Molecular Simulations journal March 2021
Benchmarking the Performance of the ReaxFF Reactive Force Field on Hydrogen Combustion Systems journal June 2020
Machine Learning for Accurate Force Calculations in Molecular Dynamics Simulations journal July 2020
Theoretical Investigations of Rate Coefficients for H + O 3 and HO 2 + O Reactions on a Full-Dimensional Potential Energy Surface journal July 2020
A Full-Dimensional Potential Energy Surface and Dynamics of the Multichannel Reaction between H and HO2 journal February 2021
A Gaussian Process Based Δ-Machine Learning Approach to Reactive Potential Energy Surfaces journal October 2023
Ab Initio Computations and Active Thermochemical Tables Hand in Hand: Heats of Formation of Core Combustion Species journal August 2017
Machine-Learning Based Stacked Ensemble Model for Accurate Analysis of Molecular Dynamics Simulations journal May 2019
Machine Learning Force Fields: Construction, Validation, and Outlook journal December 2016
Exploring Complex Reaction Networks Using Neural Network-Based Molecular Dynamics Simulation journal May 2022
Permutation-Invariant-Polynomial Neural-Network-Based Δ-Machine Learning Approach: A Case for the HO2 Self-Reaction and Its Dynamics Study journal May 2022
Ring Polymer Molecular Dynamics Approach to Quantum Dissociative Chemisorption Rates journal August 2023
Activity, Selectivity, and Durability of Ruthenium Nanoparticle Catalysts for Ammonia Synthesis by Reactive Molecular Dynamics Simulation: The Size Effect journal July 2018
ReaxFF:  A Reactive Force Field for Hydrocarbons journal October 2001
Complex reaction processes in combustion unraveled by neural network-based molecular dynamics simulation journal November 2020
Neural network reactive force field for C, H, N, and O systems journal January 2021
Accurate energy barriers for catalytic reaction pathways: an automatic training protocol for machine learning force fields journal October 2023
An accurate multi-channel multi-reference full-dimensional global potential energy surface for the lowest triplet state of H 2 O 2 journal January 2016
Intersystem crossing processes in the 2CzPN emitter: a DFT/MRCI study including vibrational spin–orbit interactions journal January 2021
A transferable active-learning strategy for reactive molecular force fields journal January 2021
Q-Band relaxation in chlorophyll: new insights from multireference quantum dynamics journal January 2022
The neural network based Δ-machine learning approach efficiently brings the DFT potential energy surface to the CCSD(T) quality: a case for the OH + CH3OH reaction journal January 2023
Semiclassical calculation of cumulative reaction probabilities journal January 1999
Long-range corrected hybrid density functionals with damped atom–atom dispersion corrections journal January 2008
ωB97X-V: A 10-parameter, range-separated hybrid, generalized gradient approximation density functional with nonlocal correlation, designed by a survival-of-the-fittest strategy journal January 2014
NewtonNet: A Newtonian message passing network for deep learning of interatomic potentials and forces journal January 2022
Permutation invariant polynomial neural network approach to fitting potential energy surfaces journal August 2013
Potential energy surfaces for high-energy N + O2 collisions journal February 2021
DeePMD-kit v2: A software package for deep potential models journal August 2023
Permutationally invariant potential energy surfaces in high dimensionality journal October 2009
A generalizable machine learning potential of Ag–Au nanoalloys and its application to surface reconstruction, segregation and diffusion journal December 2021
Active learning of uniformly accurate interatomic potentials for materials simulation journal February 2019
Machine Learning for Molecular Simulation journal April 2020
Inefficient intramolecular vibrational energy redistribution for the H + HO2 reaction and negative internal energy dependence for its rate constant journal October 2022

Similar Records

Active Learning Framework
Software · Tue Jan 03 19:00:00 EST 2023 · OSTI ID:code-115537
Machine Learning Classical Interatomic Potentials for Molecular Dynamics from First-Principles Training Data
Journal Article · Thu Mar 28 00:00:00 EDT 2019 · Journal of Physical Chemistry. C · OSTI ID:1510275
Machine Learning Classical Interatomic Potentials for Molecular Dynamics from First-Principles Training Data
Journal Article · Wed Jan 02 23:00:00 EST 2019 · Journal of Physical Chemistry. C · OSTI ID:1542944

Related Subjects

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CHEMISTRY
Computational chemistry
Energy
Kinetic parameters
Mathematical methods
Molecular mechanics