Laura Gagliardi
Automating Multireference Methods for Ab-initio Dynamics Simulations
Tue, Feb. 25, 2025, 4:30pm
Taylor Auditorium, Frick Lab, B02
Host: Will Jacobs
Multimetallic molecules often exhibit strong electron correlation, requiring a multiconfigurational treatment of the wave function. The localized active space self-consistent field (LASSCF) method provides a practical alternative to complete active space (CAS) approaches by factorizing the way function into localized fragments, with inter-fragment correlation reintroduced via LAS state interaction (LASSI). [1] However, optimal strategies for defining LAS fragments and LASSI model spaces remain an open challenge. In this talk, I will present out latest efforts to automate LASSI and its application to multi metallic complexes. [2]
Post-CASSCF methods such as multiconfiguration pair-density functional theory (MC-PDFT)[3] provide efficient electronic correlation recovery but face challenges due to the dependence on active space selection. This dependence creates difficulties in achieving a consistent description of molecular properties across nuclear configurations, which hinders the training of machine learning potentials (MLPs) using multireference methods. To address this, we introduce the weighted active space across geometries.
By integrating WASP and on-the-fly probability enhanced sampling method, we construct an MLP trained on MC-PDFT energetics, marking the first demonstration of an MLP trained on multi reference electronic structure calculations. We tested our method on the ab-initio reaction dynamics of the TiC+ catalyzed C–H activation of methane, a challenging re4action fort Kohn-Sham DFT because of the multi reference character of the species involved. This framework enables accurate simulations of rare events beyond the reach of classical or other ab-initio methods, opening new possibilities for studying complex reactive processes.
[1] M. R. Hermes, R Pandharker. L. Gagliardi, Variational Localized Active Space Self-Consistent Field Menthod, J. Chem. Theory Compute. 2020, 16, 4923-4937.
[2] V. Agarawal, D. S. King, M. R. Her,es, L. Gagliardi, Automatic State Interaction with Large Localized Active Spaces for Multimetallic Systems, J. Chem. Theory Comput, 2024 20, 4654-4662.
[3] M. Hennefarth, M. Hermes, D. Truhlar, L. Gagliardi, Linearized Pair Density Functional Theory, J. Chem, Theory Comput., 2023, 3172-3183.
[4] A Seal, S. Perego, M. R. Hennefarth, U. Raucci, L. Bonati, M. Parrinello, L. Gagliardi, Weighted Active Space Selection for Multi-Reference Machine-Learned Potentials for Ab-initio Dynamics Simulations for Catalysis, in preparation, 2025.
Sponsorship of an event does not constitute institutional endorsement of external speakers or views presented.