(411b) Actively Learning Robust High-Complexity Force Fields

Machine Learning has gained significant traction in the model development community due to its versatility and potential to considerably decrease the human effort required to generate complex, high-fidelity models. Here, we present the Chebyshev Interaction Model for Efficient Simulation (ChIMES), a new machine-learned force field and development approach that has enabled simulation of complex phenomena including condensed-phase reaction-driven phase separation events. ChIMES models are comprised of linear combinations of Chebyshev polynomials explicitly describing many-body interactions and are actively learned to short Kohn-Sham density functional theory (DFT) trajectories. Resulting force fields retain much of the accuracy of DFT while decreasing computational effort by orders of magnitude, and consequently, can be viewed as a proxy for DFT dynamics on large scales.

This work was performed under the auspices of the U.S. Department of Energy by Lawrence Livermore National Laboratory under Contract DE-AC52-07NA27344. LLNL-ABS-771605.