Multiconfigurational Second-Order Perturbation Theory Restricted Active Space (RASPT2) Studies on Mononuclear First-Row Transition-Metal Systems

A series of model transition-metal complexes, CrF(6), ferrocene, Cr(CO)(6), ferrous porphin, cobalt corrole, and FeO/FeO(-), have been studied using second-order perturbation theory based on a restricted active space self-consistent field reference wave function (RASPT2). Several important properties (structures, relative energies of different structural minima, binding energies, spin state energetics, and electronic excitation energies) were investigated. A systematic investigation was performed on the effect of: (a) the size and composition of the global RAS space, (b) different (RAS1/RAS2/RAS3) subpartitions of the global RAS space, and (c) different excitation levels (out of RAS1/into RAS3) within the RAS space. Calculations with active spaces, including up to 35 orbitals, are presented. The results obtained with smaller active spaces (up to 16 orbitals) were compared to previous and current results obtained with a complete active space self-consistent field reference wave function (CASPT2). Higly accurate RASPT2 results were obtained for the heterolytic binding energy of ferrocene and for the electronic spectrum of Cr(CO)(6), with errors within chemical accuracy. For ferrous porphyrin the intermediate spin (3)A(2g) ground state is (for the first time with a wave function-based method) correctly predicted, while its high magnetic moment (4.4 mu(B)) is attributed to spin-orbit coupling with very close-lying (5)A(1g) and (3)E(g) states. The toughest case met in this work is cobalt corrole, for which we studied the relative energy of several low-lying Co (II)-corrole pi radical states with respect to the Co(III) ground state. Very large RAS spaces (25-33 orbitals) are required for this system, making compromises on the size of RAS2 and/or the excitation level unavoidable, thus increasing the uncertainty of the RASPT2 results by 0.1-0.2 eV. Still, also for this system, the RASPT2 method is shown to provide distinct improvements over CASPT2, by overcoming the strict limitations in the size of the active space inherent to the latter method.