Simulating Ultrafast Transient Absorption Spectra from First Principles using a Time-Dependent Configuration Interaction Probe

TL;DR

This paper introduces a first-principles simulation method for ultrafast transient absorption spectra using time-dependent configuration interaction, enabling detailed interpretation of experimental TAS data with reduced computational cost.

Contribution

The authors develop a robust, efficient approach combining TD-CASCI with nonadiabatic dynamics to simulate TAS signals directly from electronic structure calculations, improving interpretability and computational efficiency.

Findings

Successfully simulated TAS of HAN molecule matching experimental data.

Decomposed spectra to identify excited-state proton transfer and relaxation pathways.

Achieved significant reduction in computational cost compared to previous methods.

Abstract

Transient absorption spectroscopy (TAS) is among the most common ultrafast photochemical experiments, but its interpretation remains challenging. In this work, we present an efficient and robust method for simulating TAS signals from first principles. Excited-state absorption and stimulated emission (SE) signals are computed using time-dependent complete active space configuration interaction (TD-CASCI) simulations, leveraging the robustness of time-domain simulation to minimize electronic structure failure. We demonstrate our approach by simulating the TAS signal of 1$^\prime$-hydroxy-2$^\prime$-acetonapthone (HAN) from ab initio multiple spawning nonadiabatic molecular dynamics simulations. Our results are compared to gas-phase TAS data recorded from both jet-cooled ($T\sim$ 40 K) and hot ($\sim$ 403 K) molecules via cavity-enhanced transient absorption spectroscopy (CE-TAS). Decomposition of the computed spectrum allows us to assign a rise in the SE signal to excited-state proton transfer and the ultimate decay of the signal to relaxation through a twisted conical intersection. The total cost of computing the observable signal ($\sim$1700 graphics processing unit hours for $\sim$4 ns of electron dynamics) was markedly less than that of the ab initio multiple spawning calculations used to compute the underlying nonadiabatic dynamics.