Deep Learning Functionals based on Møller-Plesset Adiabatic Connection for Non-Covalent Interactions

Heng Zhao, Kim J. Daas, Elias Polak, Eduardo Fabiano, Stefan Vuckovic

The modeling of non-covalent interactions (NCIs) is crucial in many areas of chemistry and material science, as these interactions often govern the structure, stability, and function of complex molecular systems. To improve pure quantum chemical simulations of NCIs, we propose an interpolation method along the Møller–Plesset adiabatic connection (MP AC), which approximates the correlation energy by combining MP2 at small coupling strengths and the strong-coupling limit of the MP AC. By leveraging deep learning techniques, we obtain models ensuring size-consistency and the accurate capture of NCIs, which particularly shines for pi-pi stacking dominated systems. While our models have the same cost as double hybrids, they offer major improvements over double hybrids for noncovalent interactions.

1. T. J. Daas, J. Grossi, S. Vuckovic, Z. H. Musslimani, D. P. Kooi, M. Seidl, K.J.H Giesbertz, P. Gori-Giorgi, J. Chem. Phys. 2020, 153, 214112.
2. T. J. Daas, E. Fabiano, F. Della Sala, P. Gori-Giorgi, S. Vuckovic, J. Phys. Chem. Lett. 2021, 12, 4867.
3. K. J. Daas, D. P. Kooi, N. C. Peters, E. Fabiano, F. Della Sala, P. Gori-Giorgi, Stefan Vuckovic, J. Phys. Chem. Lett. 2023, 14, 8448.

 Heng Zhao