Photochemistry is undeniably fascinating due to its ability to harness the power of light to drive chemical reactions. At its core, photochemistry explores the intricate dance between molecules and photons, offering a window into the fundamental processes that shape our world. Molecular photoswitches and photomotors stand out as these molecular entities exploit light-triggered conformational changes or rotational motions, offering a remarkable level of control at the nanoscale.
However, comprehensively understanding structure-property relationships of photochemical reactions through experiments alone is challenging, if not unfeasible, due to the ultrafast nature of light-induced processes. Hence, quantum chemical methods have become an indispensable tool in revealing photoreaction mechanisms and towards the development of design rules within chemical space. In practice, the applicability of quantum chemical calculations and dynamics simulations is largely limited due to two main reasons: A) the need for theoretical expertise and a priori knowledge of photochemical properties of the system and B) the high computational expenses associated.
To tackle these challenges, our aim is to extract the fundamental principles governing photochemical reactions through a database-supported analysis of quantum chemical results in connexion with spectroscopic data. Moreover, science-driven machine learning—relying on existing quantum chemical data such as energies, forces and couplings in excited states—is employed to predict photochemical properties, bearing the potential to greatly expedite non-adiabatic molecular dynamics simulations while upholding ab initio accuracy. To achieve these goals, we introduce the Surface Hopping Newly Invented Training Set for Excited-state Learning (SHNITSEL [1]) database, which contains computational data for a multitude of photoreactions. Specifically, we showcase example molecules from the compound class of photoswitches and photomotors undergoing light-induced E/Z isomerization and directional rotation, respectively.
Martina Hartinger