Ab initio Quantum Chemistry Methods for Modeling Molecular Excited States Beyond Configuration Interaction Singles

Electron transfer and energy transfer play a central role in photo-induced excited state chemical dynamics and are critical for understanding the fundamental processes in photosynthesis. Understanding electron and energy transfer at the molecular level is essential, since they must compete with deactivation processes back to the molecular ground state-- and deactivation releases any captured energies as wasted heat. Modeling electronic relaxation process is very challenging, however, for 2 reasons: i) Obtaining accurate potential energy surfaces (PESs) by solving the electronic Hamiltonian (only) is nontrivial, since all electrons are coupled together, which is essentially a many-body problem. It is even more difficult in the context of photochemistry, where the relevant molecules are typically big; ii) The Born-Oppenheimer Approximation of separating electronic and nuclear motion may be invalid, and thus one has to model nonadiabatic dynamics. This thesis is focused on the first problem above, i.e. solving the electronic Hamiltonian, where there is currently a lack of effective ab initio quantum chemistry methods, especially in the presence of charge transfer (CT) states. Historically Configuration Interaction Singles (CIS) has been the standard method for modeling electronic excited states with qualitatively correct wavefunctions, but CIS is highly biased against charge transfer states-- which are very important for modeling photo-induced relaxation. Nevertheless, in this thesis, CIS proves to be a good starting point for improved ab initio quantum chemistry methods, that build in the correct molecular orbital optimization. These algorithms are labeled as: i) Orbital Optimized Configuration Interaction Singles (OO-CIS), ii) Variational Orbital Adapted Configuration Interaction Singles (VOA-CIS), and iii) Fully Variational Orbital Adapted Configuration Interaction Singles (FVOA-CIS). Each of the three algorithms above represents an improvement upon its predecessor. i) OOCIS is able to recover perturbative corrections for CT states; ii) its variational extension VOA-CIS proves to be very effective for constructing globally smooth adiabatic PESs even with CT states; and iii) because it is fully variational, FVOA-CIS PESs are so smooth that it should allow analytic gradients. We believe these approaches will be widely used for future accurate electronic structure calculations.