AI/HPC Seminar series

Machine-learning driven atomistic and coarse-grained simulations for condensed phase

by Ganesh Sivaraman (DSL)


Atomistic simulations based on density functional theory (DFT) have played a critical role in understanding the links between structures and properties of condensed phase.  However, DFT calculations are  infeasible for simulating large scale complex systems. Machine learning can bridge this gap by learning from DFT data, while providing near DFT accuracy at a reduced computational cost. In this talk, I will provide an  overview of the ML driven atomistic modeling. I will discuss the implication of the extending the DFT accuracy to larger systems sizes (> 1000 atoms) and longer time scales (i.e. ns )[1] . I will end the talk, with a discussion on  a GPU accelerated  coarse-grained DFT predictions via deep kernel learning [2] .

[1] Sivaraman et al., Phys. Rev. Lett., 126, 156002 (2021)
[2] Sivaraman et al., J. Chem. Theory Comput., 18, 2, 1129 (2022)

Organized by

Vangelis Kourlitis (HEP), Tanwi Mallick (MCS), Krishnan Raghavan (MCS)