From Energies to Geometries: Development and Validation of a Robust yet Accurate Tool for Explicitly Correlated Structural Optimizations at an Affordable Cost
Luigi Crisci, Mihály Kállay, Vincenzo Barone
Abstract
We present a new computational tool that integrates a robust geometry optimization module with external electronic-structure engines, enabling accurate and efficient optimizations at high levels of theory. Electronic energies and, when available, analytical gradients are directly imported, while numerical gradients are internally computed through an optimized parallel implementation with automatic symmetry reduction and granular resource control. This design ensures flexibility, stability, and full compatibility with composite post-Hartree-Fock frameworks. A key innovation is the extension of frozen natural orbital (FNO) techniques, previously employed to reduce the cost of correlated energy evaluations, to equilibrium geometry optimizations. The resulting FNO-based schemes markedly expand the size of systems that can be treated with near-spectroscopic accuracy. Extensive benchmarks demonstrate that both additive ("geometry") and composite-gradient formulations deliver stable and reliable convergence across a wide range of molecular systems. Overall, this development bridges energetic and structural focal-point analyses, paving the way for next-generation composite simulations integrating structural, thermochemical, and spectroscopic predictions.