Automatic Orbital\nPair Selection for Multilevel Local\nCoupled-Cluster Based on Orbital Maps

We present an automatic,\norbital-map based orbital-pair\nselection\nscheme for multilevel local coupled-cluster approaches that exploits\nthe locality of chemical reactions by focusing on the part of the\nmolecule directly involved in the reaction. The previously introduced\npair-selected multilevel extension to domain-based local pair natural\norbital coupled-cluster with singles, doubles, and semicanonical perturbative\ntriples [DLPNO–CCSD(T<sub>0</sub>)] partitions the orbital\npairs according to relative changes in pair correlation energies [Bensberg, M.; Neugebauer, J. Orbital\npair selection for relative energies in the domain-based local pair\nnatural orbital coupled-cluster method. J.\nChem. Phys. 2022, 157, 064102. 10.1063/5.0100010]. To this end, maps between localized orbitals are required\nwhich in turn require maps between the atoms of structures along reaction\npaths. So far, these atom maps have been manually determined, which\ncan be a (human) time-consuming procedure. Here, we present an automatic\natom mapping algorithm based on the principle of minimum chemical\ndistance that incorporates orientation dependence through dihedral\nangles. A similar strategy is then introduced to obtain orbital maps,\nwhich proves advantageous over the previously used direct orbital\nselection. Along with a modified orbital pair prescreening, this results\nin an improved variant of the pair-selected multilevel DLPNO–CCSD(T<sub>0</sub>) method. The performance of this approach is demonstrated\nfor various reaction types showing a significant efficiency gain and\naccurate results due to beneficial, systematic error cancellation.\nThe presented method operates in a black-box manner due to its fully\nautomatized algorithms with only the need to specify a single target-accuracy\nparameter. Additionally, we demonstrate that basis set extrapolation\ntechniques can be applied. In this context, the approach shows deficiencies\nfor the use of large basis sets, especially with diffuse basis functions,\nwhich can be traced back to the semicanonical triples correction.