Predictions of Transition Metal Thermochemistry and Singlet-Triplet Gaps for Organic Diradicals via Quantum Monte Carlo
James Shee, Benjamin Rudshteyn, Evan J. Arthur, Shiwei Zhang, David R .Reichman, Richard A. Friesner
Molecules which exhibit strong electron correlations are relevant to a myriad of chemical systems, and present a challenge for most traditional electronic structure methods. Our GPU-accelerated implementation of the phaseless variant of Auxiliary-Field Quantum Monte Carlo (ph-AFQMC), combined with a novel sampling approach, has enabled us to perform routine calculations with benchmark-quality accuracy surpassing that of Coupled Cluster (CC) methods for many strongly correlated systems. First, we compute the ionization potentials of the 3d transition metal atoms, and proceed to calculate the bond-dissociation energies of over 40 transition metal containing diatomics. Second, we calculate the singlet-triplet gaps for a set of 13 small organic diradicals, in which the singlet state wavefunction is characterized by more than one electronic configuration and thus is difficult to describe with single-reference methods. Chemical accuracy is easily achieved using ph-AFQMC with moderate CASSCF trial wavefunctions. In addition, we propose a scalable protocol that will enable accurate ph-AFQMC calculations of the spin gap for larger molecular systems.