(641b) Independent Atom Ansatz of Density Functional Theory for Interpretable, Low-Cost Calculations of Potential Energy Surfaces

The accelerated discovery of new materials, molecules, and chemical reactions through computational techniques necessitates low-cost, interpretable methods for computing potential energy surfaces. To address the high computational demands of ab initio methods based on the independent electron ansatz, we introduce the independent atom (IA) ansatz within density functional theory (DFT). The new ansatz represents the ground-state electron density using self-consistent, atom-localized orbitals, enabling partial cancellation of nuclear-electron and electron-electron inter-atomic interactions. This makes it possible to derive analytic binding energy expressions in the large separation limit and extrapolate them to chemical bonding distances, while retaining their accuracy.

The IA ansatz underpins the nonempirical tight binding theory (NTB), which is parameter-free, describes bond dissociation to free atoms correctly, and incorporates the energy decomposition and charge analyses at no additional cost. NTB predicts the bond energy, bond length, and the vibrational wavenumber of an H₂ molecule to be -4.56 eV, 0.743 Å, and 4385 cm^-1, respectively, using no parameters, closely matching experimental values of -4.75 eV, 0.741 Å, and 4401 cm^-1. The NTB theory has been extended to period-2 diatomics, demonstrating superior accuracy compared to standard Kohn-Sham DFT methods (PBE and SCAN functionals) using at most two parameters: the 2s and 2p exponents of the STO-6G basis set. Ongoing work includes catalytic dissociation of H₂ on Al and Pt metals. We anticipate that the independent atom ansatz will enable rapid computational characterization of a wide array of materials and reaction mechanisms, accelerating discoveries in chemistry and catalysis.