We simulate the photodynamics of gas-phase cyclobutanone excited to the S2 state using fewest switches surface hopping (FSSH) dynamics powered by time-dependent density functional theory (TDDFT). We predict a total photoproduct yield of 8%, with a C3:C2 product ratio of 0 trajectories to 8 trajectories. One primary S2 → S1 conical intersection is identified involving the compression of an α-carbon–carbon–hydrogen bond angle. Excited state lifetimes computed with respect to electronic state populations were found to be 3.96 ps (S2 → S1) and 498 fs (S1 → S0). We also generate time-resolved difference pair distribution functions (ΔPDFs) from our TDDFT-FSSH dynamics results in order to generate direct comparisons with ultrafast electron diffraction experiment observables. Global and target analysis of time-resolved ΔPDFs produced a distinct set of lifetimes: (i) a 0.548 ps decay and (ii) a 1.69 ps decay, both resembling the S2 minimum, as well as (iii) a long decay that resembles the S1 minimum geometry and the fully separated C2 products. Finally, we contextualize our results by considering the impact of the most likely sources of significant errors.
Time-Dependent Density Functional Theory (TDDFT) is a workhorse computational tool for predicting nonlinear optical properties from first principles. However, its computational cost often limits its practicality for large molecules. We evaluate nonlinear response properties using our recently developed semiempirical model, TDDFT-ris (ris), and its extension, TDDFT-risp (risp), which use minimal auxiliary basis to approximate electron-repulsion integrals. Against full TDDFT with def2-SVPD calculations, ris and risp accurately reproduce (hyper)polarizabilities for a small fraction of the computational cost. For example, ris has a mean relative error of 8% when computing mean polarizabilities, which is reduced to 2% by risp. Furthermore, the strong physical justification for the risp method allows us to cure systematic failures in monopole-based semiempirical models (like ris) when computing nonlinear properties of planar molecules. Thus, ris and risp are powerful and inexpensive tools for predicting nonlinear optical properties either qualitatively or nearly quantitatively.
17 June 2026 in The Journal of Physical Chemistry C
Time-Dependent Density Functional Theory (TDDFT) is a workhorse computational tool for predicting nonlinear optical properties from first principles. However, its computational cost often limits its practicality for large molecules. We evaluate nonlinear response properties using our recently developed semiempirical model, TDDFT-ris (ris), and its extension, TDDFT-risp (risp), which use minimal auxiliary basis to approximate electron-repulsion integrals. Against full TDDFT with def2-SVPD calculations, ris and risp accurately reproduce (hyper)polarizabilities for a small fraction of the computational cost. For example, ris has a mean relative error of 8% when computing mean polarizabilities, which is reduced to 2% by risp. Furthermore, the strong physical justification for the risp method allows us to cure systematic failures in monopole-based semiempirical models (like ris) when computing nonlinear properties of planar molecules. Thus, ris and risp are powerful and inexpensive tools for predicting nonlinear optical properties either qualitatively or nearly quantitatively.