UC Berkeley

Chemistry is teeming with light-driven processes such as photosynthesis and molecular switches, where electronically excited states are paramount to understanding the underlying mechanisms at play. Nevertheless, the appropriate treatment of effects such as strong electron correlation and orbital relaxations are beyond the simplifying assumptions of linear response theory and remains a challenge for theoretical modeling, consequently hampering the theory’s ability to provide meaningful predictions and inform experiment. In this thesis, we present a novel approach with the flexibility and precision to overcome these obstacles.

To contextualize this work, we first discuss current excited state ansatzes and optimization methods, demonstrating their strengths and weaknesses in handling the challenges of modeling multi-reference systems like excited states. Using a generalized variational principle, we construct a fully excited-state-specific wave function optimization algorithm and demonstrate both its resilience to common optimization problems such as root flipping and its ability to locate and tightly converge excited state stationary points. Exhibiting both accuracy and reliability in calculating states with strongly correlated character and strong orbital relaxations, this approach is capable of providing state-specific insight in scenarios which elude existing methods, including some core, charge transfer and double excitations. We expand our study to explore the effect of these methodological improvements in comparison to a prominent traditional approach for geometry relaxations and potential energy surfaces for a photochemically relevant system, helping to elucidate the complexity of excited state energetics. The realization of this novel method constitutes significant steps forward for state-specific multi-reference approaches for the precise, accurate and robust modeling of electronically excited states.