Speaker: Johannes Hoja, University of Graz

Gaining quantitative insight into the dynamics and kinetics of molecular crystals is vital for drug development and for advancing solvent-free, sustainable synthesis methods that rely on solid-state reactions. However, this remains a difficult task as molecular crystals often exhibit multiple polymorphs that differ by only a few kJ/mol in their Gibbs free energies. Thus, a reliable description of these polymorphs requires very accurate calculations to correctly capture their relative stabilities [1]. Therefore, we have developed a multimer embedding approach [2], which approximates expensive periodic PBE0+MBD calculations by embedding multimers into lower-cost PBE+MBD calculations, thereby significantly reducing computational cost and memory requirements. This method reproduces lattice energies within 1 kJ/mol, unit cell volumes within 1 %, and harmonic vibrational free energies within 1 kJ/mol of the canonical periodic PBE0+MBD results and provided very accurate results in a recent crystal structure prediction blind test [3]. Building on this foundation, we utilize solid-state nudged elastic band (NEB) simulations to explore energy barriers and transition pathways of molecular crystal transformations, covering interconversions between polymorphs and a single-crystal Diels-Alder reaction. We showcase a novel interpolation method for automatically generating viable initial pathways that avoids atomic collisions, enabling NEB calculations for crystals containing larger and quite flexible molecules [4]. To accelerate these calculations, we further demonstrate the use and accuracy of machine-learned force fields trained on system-specific DFT data. Collectively, these advances bring us closer to a comprehensive, computationally efficient picture of polymorph transformation pathways and reaction mechanisms in molecular crystals.

[1] Hoja, J.; Ko, H.-Y.; Neumann, M.A.; Car, R.; DiStasio, R.A.; Tkatchenko, A. Sci. Adv. 2019, 5, eaau3338.
[2] Hoja, J.; List, A.; Boese, A.D. J. Chem. Theory Comput. 2024, 20, 357-367.
[3] Hunnisett, L.M., et al. Acta Cryst. B 2024, 80, 548-574.
[4] Goncharova, N.; Hoja, J. arXiv preprint 2025, arXiv:2410.10506.