Abstract

Recently, we optimized small (vDZP), deeply contracted AO basis sets in molecular DFT calculations using standard ECPs for all elements up to radon¹. This strategy is further- more applied to a minimal set of AOs which — as a totally new ingredient — is made adaptive, i.e., radially different for symmetry distinct atoms in a molecule. The ”breathing” of the AOs in the molecular environment is parameterized efficiently by on-the-fly computed effective atomic charges (obtained by a new EEQ charge model) and coordination numbers. This so-called q-vSZP set² provides in typical DFT applications results of about or better than DZ quality. It forms the basis of our third-generation tight-binding model g-xTB (g=general). This includes non-local Fock-exchange as well as other new, many-center Hamiltonian terms (e.g., atomic correction potentials, ACP ). It aims at general purpose applicability in chemistry and more closely approaches DFT accuracy (actually ωB97M-V/aTZ³) than previous semi-empirical methods at only slightly increased computational cost (factor of 1.5 compared to GFN2 -xTB). It will be consistently available for all elements Z=1-103 with f-electrons included for lanathanides/actinides. The talk describes key improvements of the underlying TB theory as well as extensive benchmarking on a wide range of standard thermochemistry sets. [1] M. Müller, A. Hansen, S. Grimme, J. Chem. Phys. 158 (2023), 014103 [2] M. Müller, A. Hansen, S. Grimme, J. Chem. Phys. 159 (2023), 164108. Revision: JPC A , doi:10.1021/acs.jpca.4c06989 [3] N. Mardirossian and M. Head-Gordon, J. Chem. Phys. 144 (2016), 214110