Excited-State Absorption: Reference Oscillator Strengths, Wave Function, and TDDFT Benchmarks

Excited-state absorption (ESA) corresponds to the transition between two electronic excited states and is a fundamental process for probing and understanding light-matter interactions. Accurate modeling of ESA is indeed often required to interpret time-resolved experiments. In this contribution, we present a dataset of 53 ESA oscillator strengths in three different gauges and the associated vertical transition energies between 71 excited states of 21 small- and medium-sized molecules from the QUEST database. In a few cases, we additionally investigated the effect of geometry relaxation on excited-state geometries. The reference values were obtained within the quadratic response (QR) CC3 formalism using eight different Dunning basis sets. We found that the d-<i>aug</i>-cc-pVTZ basis set is always adequate while its more compact double-ζ counterpart, d-<i>aug</i>-cc-pVDZ, performs well in most cases. These QR-CC3 data allow us to assess the performance of QR-TDDFT, with and without applying the Tamm-Dancoff approximation, using a panel of global and range-separated hybrids (B3LYP, BH&HLYP, CAM-B3LYP, LC-BLYP33, and LC-BLYP47), as well as several lower-order wave function methods, i.e., QR-CCSD, QR-CC2, EOM-CCSD, ISR-ADC(2), and ISR-ADC(3). We show that QR-TDDFT delivers acceptable errors for ESA oscillator strengths with CAM-B3LYP showing particular promise, especially for the largest molecules of our set, and in the Franck-Condon (FC) region. We also find that ISR-ADC(3) exhibits excellent performance in this region. When using excited-state optimal geometries, the relative performance of wave function-based approaches remains consistent with trends observed in the Franck-Condon region. However, for TD(A)-DFT, the accuracy varies more significantly, as the performance of different exchange-correlation functionals significantly depends on the chosen geometry.