Electronic structure of methyl radical photodissociation

Abstract The calculation of accurate excitation energies and potential energy surfaces of photochemical reactions is a major challenge of current quantum chemistry, especially when one wants to keep the cost low enough to make detailed dynamical simulations affordable. Methods that might be affordable for large complex molecules can be tested against benchmark results for smaller molecules, where accurate benchmarks can be available. Here we report such testing for the excitation and dissociation of the methyl radical, yielding both singlet and triplet CH 2 . The emphasis is on multistate pair density functional theory using compressed-multistate (CMS-PDFT) and linearized (L-PDFT) formulations. We also consider the less affordable XMS-CASPT2 method with the same state-averaged-complete-active-space (SA-CASSCF) reference wave functions, which has seven active electrons in 10 active orbitals. The calculations use state averaging over seven states and a model space that spans the seven lowest SA-CASSCF eigenvectors. We study three on-top density functionals: tPBE, tPBE0, and MC23. Vertical excitation energies, adiabatic excitation energies, and dissociation energies, along with cuts through the potential surfaces along the dissociation coordinate, were computed with the (7, 10) active space. XMS-CASPT2 and L-PDFT with the MC23 functional show consistent and reliable performance for excitation energies, closely reproducing benchmark values, and producing smooth, physically reasonable potential energy surfaces essential for nonadiabatic dynamics simulations, but they are less accurate for bond energies. The L-PDFT calculations with the tPBE functional are more accurate for dissociation energies, but less accurate for excitation energies.