Free energy calculations enable the quantitative understanding of physicochemical phenomena in material science, chemistry, and physics. Nevertheless, free energy methods are typically faced with computational efficiency issues, which limit their applicability in large-scale, high-throughput applications. One such application is the computational prediction of polymorphism, where the relative stabilities of tens to hundreds of putative polymorphs need to be evaluated to provide rational, physics-based prediction. A source of such limitations is that interesting metastable states, representing i.e. putative polymorphs, are usually characterized by nonoverlapping configurational Boltzmann distributions, and thus, computing free energy differences between them requires sampling intermediate states characterized by high free energies and low probabilities.
In this seminar, I will discuss how machine learning techniques informed only by locally ergodic molecular dynamics simulations can provide a blueprint to boost large-scale studies of the relative thermodynamic stability of polymorphs of molecular crystals. In particular, we propose a combination of normalizing flow models, and low variance free energy estimators1,2 to efficiently compute the anharmonic free energy of molecular polymorphs of conformationally complex organic molecules as a function of thermodynamic parameters as a function of Temperature and pressure.3,4
References
1. Jarzynski, C., (2002). Physical Review E, 65(4), p.046122.
2. Olehnovics, E., Liu, Y. M., Mehio, N., Sheikh, A. Y., Shirts, M. R., & Salvalaglio, M. (2024). Assessing the accuracy and efficiency of free energy differences obtained from reweighted flow-based probabilistic generative models. Journal of Chemical Theory and Computation, 20(14), 5913-5922.
3. Olehnovics, Edgar, et al. "Accurate Lattice Free Energies of Packing Polymorphs from Probabilistic Generative Models." Journal of Chemical Theory and Computation 21.5 (2025): 2244-2255.
4. Olehnovics, E., Liu, Y. M., Mehio, N., Sheikh, A. Y., Shirts, M., & Salvalaglio, M. (2025). Lattice free energies of molecular crystals using normalizing flow.