Skip to main content

Active Learning of Atomistic Surrogate Models for Rare Events

Chemistry-Physics Bldg. Room 303
Speaker(s) / Presenter(s):
Gang Seob Jung (Oak Ridge National Laboratory)

Multiscale modeling methods are typically envisioned as precise and predictive
simulation tools to solve complex science and engineering problems. However, even
conventional atomistic models are often insufficient in terms of computational efficiency
and accuracy to provide reliable information for the large-scale continuum models. In this
seminar, I will focus on method developments aimed at overcoming these critical
At the beginning of the talk, I will introduce how atomic models can help us understand
the experimental observation of crystal growth in 2D materials using empirical reactive
forcefield (FF). Although atomistic models can provide useful insight at the atomic scale,
developing reliable FFs is extremely limited due to the fixed potential expressions.
Recently, neural network (NN) potentials (or surrogate models) have emerged as a way to
overcome the long-standing limitations of empirical potentials.
I will present recent developments that integrate ab-initio level calculation (DFT and
DFTB) and a PyTorch implementation of NN potentials (TorchANI) into the LAMMPS
molecular dynamics software. I will discuss the pros and cons of NN potentials illustrated
by a simple carbon system, graphene. While NN potentials can provide higher accuracy
than other FFs, e.g., ReaxFF and AIREBO, and lower computational cost than quantum
calculations, efficient sampling or data generation arises as a critical issue.
In the end, I will present the ongoing development of active learning capabilities through
LAMMPS and NNPs as an efficient sampling method for rare events. The developed
capabilities will provide useful tools for fundamental understanding of the chemical
process and mechanistic insights into the predictive design and interpretive simulation
of materials properties and processes.

GS Jung is a Research Staff at Oak Ridge National Laboratory. His research interests are
on the multiscale modeling of materials to understand their fundamental properties from
synthesis and growth to performance and failures. Before joining ORNL, he earned his
Ph.D. in multiscale modeling for 2D materials from the Massachusetts Institute of