Composite Gradient Schemes for Accurate Molecular Geometries: A Flexible Interface and Hierarchical Optimization Workflow
Luigi Crisci, Federico Lazzari, Vincenzo Barone
Abstract
Accurate molecular geometries are crucial for predicting molecular properties and interpreting experimental spectra. While composite post-Hartree-Fock methods routinely yield chemically accurate energies, their application to the structural determination of medium-sized molecules remains limited. In this work, we build upon the Pisa Composite Schemes and recent advances in composite-gradient techniques to introduce a flexible, fully automated framework for geometry optimization. Our workflow integrates the robust internal-coordinate optimizer and efficient evaluation of analytical derivatives for hybrid and double-hybrid density functionals in Gaussian with high-accuracy post-Hartree-Fock models implemented in Molpro. It supports hierarchical refinement─starting from crude initial guesses, advancing through hybrid and double-hybrid DFT, and culminating in wave function-based composite methods─enabling both gradient-based and geometry-based composite schemes. To extend applicability to larger systems, we incorporate a multilayer PCS2/DFT strategy within the same platform. Extensive validation against benchmark equilibrium geometries shows that the proposed methodology consistently yields accurate structures across a wide range of molecular systems.