In recent years, machine learning methods have entered the research field of photochemistry and demonstrated their applicability to support excited state predictions and decipher fundamental mechanisms underlying light-matter interactions. However, not many studies have gone beyond small model systems, and a major problem remains the lack of accurate reference data for excited states . In this talk, I will show how combining data from different quantum chemistry methods and incorporating underlying physics into machine learning architectures can help to overcome the sparse data problem and enable more accurate and data-efficient models [2,3]. These techniques will be used to enable photodynamics simulations of tyrosine at experimentally relevant time scales. In this way, unexpected reaction mechanisms are discovered that provide new insights into the photochemistry of biological systems .
 J. Westermayr, P. Marquetand Chem. Rev. 121(16), 9873-9926 (2021).
 J. Westermayr, et al. J. Phys. Chem. Lett. 11(10) 3828–3834 (2020).
 J. Westermayr, et al. Chem. Sci. 10, 8100-8107 (2019).
 J. Westermayr, et al. Nat. Chem. (accepted, 2022), arXiv:2108.04373.
Back to Workshop IV: Monte Carlo and Machine Learning Approaches in Quantum Mechanics