Time-Dependent Density Functional Theory (TDDFT) in the linear-response regime has become the workhorse of computational photochemistry in large systems due to its high accuracy and low computational cost. Although TDDFT performs well for valence excitations, it systematically underestimates charge-transfer excitation energies, especially in large systems with explicit solvent. As a result, the TDDFT spectrum of solvated chromophores is often riddled with hundreds of spurious charge-transfer states, dramatically increasing the computational cost due to the large number of roots required to reveal the true bright states. Artificially lowering the manifold of charge-transfer states also introduces coupling between the spurious states and bright excitations, depleting predicted intensities of the major spectral features. Moreover, when charge-transfer-to-solvent (CTTS) states are physically relevant, it can be difficult to discern which CTTS states are physical and which are spurious. In this work, we introduce an efficient screening procedure based on diabatization that isolates physically important CTTS states from those that are spurious. Our protocol introduces "natural charge-transfer orbitals" (NCTOs), which provide a compact basis for describing physically relevant CTTS excitations. When projected onto the space of solvent virtual orbitals, the NCTOs also provide a criterion for automated solvent selection in the context of projection-based embedding. Following this prescription, solvent molecules with significant participation in the CTTS acceptor process can be cleanly separated from the remainder of the "spectator" solvent molecules and treated at a higher level of theory than their ambivolent counterparts. We show that this procedure prunes the spectrum of hydrated acrolein of spurious charge-transfer states, and that the partition defined by natural charge-transfer analysis is stable with respect to basis set and functional choice, implying that in most cases the preliminary low-level supersystem calculation used in projection-based embedding can be expected to yield a subsystem partitioning that is faithful to the higher-level results. Finally, we apply natural charge-transfer analysis in conjunction with ab initio molecular dynamics to compute the CTTS spectrum of I-(H2O)96, achieving quantitative agreement with experiment without encountering spurious CTTS states.

Revealing the intrinsic spectra of chromophores in bulk solution with natural charge-transfer analysis

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.