Enhancing the Computational Efficiency of the DoNOF Program through a New Orbital Sorting Scheme

• DOI: 10.48550/arxiv.2502.01786
• Publication Date: 2025-02-03
• Authors: Élodie Boutou; Juan Felipe Huan ...; José M. Mercero; Mario Piris
• Keywords: Chemical Physics (physics.chem-ph); Physics - Chemical Physics; FOS: Physical sciences; Computational Physics (physics.comp-ph); Physics - Computational Physics

This work presents a novel approach to distribute orbitals into subspaces within electron-pairing-based natural orbital functionals (NOFs). modifies the coupling between weakly and strongly occupied by applying an alternating sorting strategy. In contrast previous that enforced electron pairing of contiguous orbitals, new provides greater flexibility, enabling calculation scheme where size can be gradually expanded. As consequence, one start using only (perfect pairing) progressively enlarge their incorporating more (extended up maximum allowed basis set. this way, alternate allows solving first simpler problem with small leverage its solution for intensive larger subspaces, thereby reducing overall computational cost improving convergence, as we observed in DoNOF program. The efficiency provided has been validated through benchmark calculations H2O, H2O2, NH3. particular, compared three strategies: i) directly largest (one-shot strategy), was usually done before work, ii) starting perfect stepwise increasing number until reaching (incremental iii) transitioning subspace (two-step strategy). Our results show two-step emerges most effective strategy, achieving lowest while maintaining high accuracy.