The Department of Physics is delighted to welcome Dr Sergei Tretiak (Los Alamos National Laboratory) to give the seventh Physics Department Colloquium of Semester 2 2023/24. Please join us to listen to Dr Sergei Tretiak's seminar titled 'Machine learning in chemistry: reactive force fields and beyond'.
A reception will be held directly after the seminar, where tea and coffee will be provided.
The seminar is open to anyone from the university, students are encouraged to attend.
Title
Machine learning in chemistry: reactive force fields and beyond
Abstract
Machine learning (ML) became a premier tool for modeling chemical processes and materials properties. ML interatomic potentials have become an efficient alternative to computationally expensive quantum chemistry simulations. In the case of reactive chemistry designing high-quality training data sets is crucial to overall model accuracy. To address this challenge, we develop a general reactive ML interatomic potential through unbiased active learning with an atomic configuration sampler inspired by nanoreactor molecular dynamics. The resulting model is then applied to study five distinct condensed-phase reactive chemistry systems: carbon solid-phase nucleation, graphene ring formation from acetylene, biofuel additives, combustion of methane and the spontaneous formation of glycine from early-earth small molecules. In all cases, the results closely match experiment and/or previous studies using traditional model chemistry methods. Importantly, the model does not need to be refit for each application, enabling high throughput in silico reactive chemistry experimentation. Active learning can be further boosted with uncertainty driven dynamics that can rapidly discover configurations tot meaningfully augment the training data set. This approach modifies the potential energy surface used in molecular dynamics simulations to favor regions of configuration space for which there is large model uncertainty. Finally, a training procedure based on Iterative Boltzmann Inversion suggests a practical framework of incorporating experimental data into ML models to improve accuracy of molecular dynamics simulations. Altogether, explosive growth of user-friendly ML frameworks, designed for chemistry, demonstrates that the field is evolving towards physics-based models augmented by data science.
References:
N. Fedik, R. Zubatyuk, N. Lubbers, J. S. Smith, B. Nebgen, R. Messerly, Y. W. Li, M. Kulichenko, A. I. Boldyrev, K. Barros, O. Isayev, and S. Tretiak, Nature Rev. Chem. 6, 653 (2022).
S. Zhang, M. Z. Makos, R. B. Jadrich, E. Kraka, B. T. Nebgen, S. Tretiak, O. Isayev, N. Lubbers, R. A. Messerly, and J. S. Smith, “Exploring the frontiers of chemistry with a general reactive machine learning potential,” Nature Chem. (2024, in press) https://chemrxiv.org/engage/chemrxiv/article-details/6362d132ca86b84c77ce166c
M. Kulichenko, K. Barros, N. Lubbers, Y. W. Li, R. Messerly, S. Tretiak, J. S. Smith, and B. Nebgen, Nature Comp. Sci., 3, 230 (2023).
S. Matin, A. Allen, J. S. Smith, N. Lubbers, R. B. Jadrich, R. A. Messerly, B. T. Nebgen, Y. W. Li, S. Tretiak, K. Barros, “Machine learning potentials with iterative Boltzmann Inversion: training to experiment,” (2024, submitted) https://doi.org/10.48550/arXiv.2307.04712.