Active learning accelerates ab initio molecular dynamics on reactive energy surfaces
Name
active-learning-accelerates-ab-initio-molecular-dynamics-on-pericyclic-reactive-energy-surfaces.pdf
Description
Submitted version
Size
12.66 MB
Format
Adobe PDF
Checksum (MD5)
81f9904f80b1821fdf7c7d5082306d6c
Author(s)
Ang, Shi Jun
Wang, Wujie
Schwalbe-Koda, Daniel
Axelrod, Simon
Gómez-Bombarelli, Rafael
Date Issued
2021
Journal
Chem
Publisher
Elsevier BV
Citation
Ang, Shi Jun, Wang, Wujie, Schwalbe-Koda, Daniel, Axelrod, Simon and Gómez-Bombarelli, Rafael. 2021. "Active learning accelerates ab initio molecular dynamics on reactive energy surfaces." Chem, 7 (3).
Version
Original manuscript
Abstract
© 2020 Elsevier Inc. Through autonomous data acquisition and machine learning, we demonstrate that our neural-network-based reactive force fields allow us to study the dynamical effects of several pericyclic reactions and to predict solvent effects on periselectivity. Our method is over 2,000 times faster than the traditional density functional theory approach, and its accuracy matches the parent quantum mechanical method. Given the efficiency of our machine learning framework, we envisage its applicability in studying larger reactive systems with a higher complexity.
MIT Department
Massachusetts Institute of Technology. Department of Materials Science and Engineering
Terms of Use
Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use.
Persistent DSpace Link
DOI of Published Version
https://doi.org/10.1016/J.CHEMPR.2020.12.009
