Computational modelling plays a central role in molecular crystal discovery, fundamental to a wide range of applications, including pharmaceuticals and renewable energy. However, a reliable description of these systems requires both a high-accuracy description of the potential energy surface and a fully anharmonic quantum description of the nuclear motion. [1,2] In this talk, I will first show that reliable lattice energies can be obtained with quantum diffusion Monte Carlo. [3,4] Then, I will demonstrate the generation of machine learning interatomic potentials capable of describing molecular crystals at finite temperature and pressure with sub-chemical accuracy, using as few as ∼ 200 data structures, an order of magnitude improvement over the current state-of-the-art. [5] Our
models successfully reproduce experiments for a diverse range of molecular crystals and open up the prospects of reliable modeling for drug discovery and beyond.
[1] G. J. O. Beran, Chem. Rev. 2016, 116, 9, 5567–5613 (2016)
[2] V. Kapil and E. A. Engel, Proc. Natl. Acad. Sci. U.S.A. 119 (6) e2111769119 (2022)
[3] F. Della Pia, A. Zen, D. Alfè, A. Michaelides, J. Chem. Phys. 157, 134701 (2022)
[4] F. Della Pia, A. Zen, D. Alfè, A. Michaelides, Phys. Rev. Lett. 133, 046401 (2024)
[5] F. Della Pia, B. X. Shi, V. Kapil, A. Zen, D. Alfè, A. Michaelides, arXiv:2502.15530 (2025)