Artificial Intelligence-Driven Optimization of Gaussian Orbital Expansions via Evolutionary Computing: Applications to Confined Atoms and Molecules

Abstract

Two artificial intelligence techniques, genetic algorithms and differential evolution, are applied to generate Gaussian expansions (φ-nG) for both free and sphere-confined atomic orbitals. A program (UCA-GSS-GA) is developed to enable the efficient calculation of these expansions. The accuracy of the obtained expansions is analyzed, and they are used to significantly refine self-consistent reaction field models for solvent effects. The orbitals, atoms, and molecules analyzed include some of the simplest systems (1s, H, H₂, and CH₄). However, the developed program is capable of handling all types of molecules and solvents.

Keywords: Gaussian orbital expansions; confined molecules; genetic algorithms; self‐consistent reaction field; simplified box orbitals.