PCS2-Based Schemes for Geometry Optimization and Frequency Calculations in Criegee Intermediate-Water Reactions.

The accurate prediction of thermochemical and kinetic parameters for Criegee intermediate reactions with water is essential for atmospheric modeling, yet the unusual electronic structure of these species poses significant challenges to standard quantum chemical methods. Here, we present a benchmark study validating new approaches for the reactions of formaldehyde oxide (CH2OO), syn-CH3CHOO, and anti-CH3CHOO with water, where geometry optimizations and harmonic frequency calculations are performed at a computationally affordable level while accurate energies are obtained via the high-level W2X composite method. In particular, we assess the Pisa Composite Scheme (PCS2), which combines explicitly correlated coupled-cluster theory [CCSD(T)-F12b] with core-valence correlation corrections; an analytical-gradient variant, in which the CABS singles correction is replaced by a Hartree-Fock level basis set incompleteness term, provides equivalent accuracy while enabling more efficient harmonic frequency calculations. Both formulations show excellent agreement with CCSD(T)-F12a/cc-pVTZ-F12 reference geometries (bond length RMSD < 0.001 Å) and reproduce activation enthalpies within 0.12 kcal/mol, demonstrating that basis set convergence is achieved at the double-ζ F12 level. This systematic benchmark was enabled by an automated, black-box workflow that minimizes the number of optimization iterations even for weakly bound hydrated complexes, allowing efficient exploration of multiple low-level/high-level combinations. The results reveal that the choice of the low-level method is far from trivial: TST rate constants show that geometry and zero-point energy differences can propagate into rate-constant errors of 30-60% for systems with very low activation enthalpies. This validation establishes dual-level PCS-based approaches as reliable tools for atmospheric chemistry applications involving Criegee intermediates, while highlighting the importance of careful low-level method selection.