Accurate Prediction of Photoinduced Electron Transfer Timescales with Nonadiabatic Microcanonical Rate Theory

Photoinduced electron transfer (PET) lies at the heart of energy conversion from light into chemical reactions. It governs a variety of biological processes including DNA damage and repair and biological photosynthesis. Quantifying PET rates and optimizing them is also crucial for selective photoredox catalysis. However, commonly used rate theories break down for PET operating in the strong coupling and nonequilibrium regimes, while excited-state dynamics simulations are computationally demanding and require complex analysis to extract PET times. Here, we employed surface-hopping nonadiabatic excited-state dynamics simulations and statistical rate theories to characterize ultrafast PET in a dimer of stacked adenine nucleobases. We show that the widely used classical Marcus Theory and Fermi's Golden Rule fail to describe ultrafast PET or even reproduce qualitative rate trends. Instead, we propose a chromophore-localized variant of nonadiabatic Rice-Rampsperger-Kassel-Marcus (NA-RRKM) theory, which yields PET timescales that are in excellent agreement with excited-state dynamics simulations.