Atomic Confinement Potentials (2505.09540v1)

Abstract: Motivated by the development of reusable software for computational chemistry [Lehtola, J. Chem. Phys. 159, 180901 (2023)], we aim to establish novel open source infrastructure for calculations with numerical atomic orbital (NAO) basis sets. Soft confinement potentials are typically used in this context to force the NAO radial basis functions {\psi}nl (r) to vanish smoothly in increasing r and to generate localized unoccupied states; we review such potentials in this work as a follow-up to our recent study on hard-wall confinement [{\AA}str\"om and Lehtola, J. Phys. Chem. A 129, 2791 (2025)]. We also note that in addition to their use in NAO generation, confinement potentials are also employed to simulate environmental effects in other research areas, such as studies of (i) atoms in solids, (ii) quantum dots, and (iii) high-pressure chemistry. As in our earlier work, we perform fully numerical density functional theory (DFT) calculations with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional, and employ spherically averaged, spin-restricted densities, as is usual in NAO studies. Our calculations employ the HelFEM program, which implements the finite element method (FEM), yielding variational energies and enabling the use of various boundary conditions. We consider four families of potentials to study the Mg and Ca atoms, which are textbook examples of extended electronic structures. We show that the resulting ground-state orbitals are surprisingly insensitive to the employed form of the confinement potential, and that the orbitals decay quickly under confinement. We study increasingly steep potentials and point out how they approach the hard-wall limit in a systematic and smooth manner. Finally, we assess NAO basis set truncation errors arising from various parameter choices for the singular potentials that are now broadly used in the NAO literature.