Accurate, Affordable and Unsupervised: Analytical F12 Gradients Driven by Generalized Internal Coordinates

Abstract

Accurate yet affordable quantum-chemical predictions are essential in several fields of molecular sciences, such as high-resolution rotational and vibrational spectroscopy. Despite major methodological and technological advances, benchmark-level methods remain prohibitive for molecules of realistic size. We present a modular implementation of the Pisa Composite Schemes (PCS), where analytical gradients ensure robust optimizations and efficient frequency calculations. PCS variants are combined hierarchically to reduce the iterations required in geometry optimizations with the most accurate and costly model, while multilayer ONIOM descriptions extend applicability by treating chemically central regions at high accuracy and peripheral fragments with lower-cost DFT variants. This strategy achieves near-spectroscopic accuracy for medium-sized molecules, including challenging cases such as CN-substituted PAHs and transition states. Freely available and user-friendly, the platform paves the way for routine, spectroscopically accurate simulations of complex systems with quantified uncertainty.