Analytical Nuclear Gradients of Density-Fitted Dirac-Fock Theory with a 2-Spinor Basis

An efficient algorithm is presented for evaluating the analytical nuclear gradients of density-fitted four-component relativistic Dirac-Fock theory as an initial step toward realizing large-scale geometry optimization of heavy-element complexes. Our algorithm employs kinetically balanced 2-spinor basis functions for the small components. The computational cost of nuclear gradient evaluation is found to be smaller than that of a Dirac-Fock self-consistent iteration. Timing data are presented for Ir(ppy)(3) (61 atoms) using a double-zeta basis set.