Molecules containing one or more transition metal atoms often have many nearly degenerate possible electronic configurations, resulting in strong correlation between electrons. Complete Active Space (CAS) techniques (where the wave equation is exactly solved within a predefined "active space" of orbitals and electrons) are the method of choice for such systems, but the exponentially increasing computational cost of CAS methods with active space size render them out of reach for most interesting problems. We present the use of the Adaptive Sampling Configuration Interaction (ASCI) approximation to solve the CAS problem. The ASCI method iteratively improves a CI wave function by searching and selecting the most important over all configurations connected to the most important Slater determinants of a trial wave function, selecting the most important of such connected configurations via a perturbation theory based rule, and finally, exactly diagonalizing the Hamiltonian within the subspace spanned by such important configurations to generate the wave function for the next iteration. This approach, when combined with a perturbative correction for dynamic correlation, is typically adequate in attaining chemical accuracy of ~1 kcal/mol for absolute energies. This approximate CASCI solver can also be used with orbital optimization to have an approximate CASSCF method, which we call ASCI-SCF. We present the basic outline of the ASCI-SCF method, along with three applications to transition metal problems. The first is a calculation of spin gaps in bioinorganic molecules like Fe porphyrin involving large active spaces of over 40 electrons in over 40 orbitals, which can be done with the modest computational resources of a personal computer. The second is a geometry optimization of a Fe4S8 Ferrodoxin complex starting from a distorted broken-spin DFT geometry using an active space of 54 electrons in 52 orbitals. Finally, we present a benchmark of absolute energies of binary compounds between transition metals and a heteroatom converged to ~1 kJ/mol, which could be used for accessing efficacy of more approximate electronic structure approaches (such as coupled cluster methods like CCSDTQ) and to generate benchmark data for DFT functional optimization.

Quantum chemistry of strongly correlated transition metal systems with the Adaptive Sampling Configuration Interaction Self Consistent Field (ASCI-SCF) method

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.