For most chemical systems, single-reference quantum chemistry methods provide adequate accuracy at reasonable cost. However, when strong multireference character emerges, these same methods may fail unexpectedly and catastrophically. Robust multireference methods, by contrast, tend to be more expensive computationally and often require significant non-black-box user insight, so they are mainly applicable to small systems or for extremely costly (in both human and computer time) simulations of high-value targets. Can we find methods that are more robust than traditional single-reference approaches, easier to use and more affordable than traditional multireference methods, and have error bounds that reassure us about the accuracy (or warn us about the inaccuracy) of our calculations? We'll present a new way to build, and test, flexible wavefunction ansatze that allow low-cost reasonably accuracy calculations to be performed, together with error bounds. The biggest challenge we are facing is the numerical optimization of these, necessarily highly nonlinear, wavefunction parameterizations, but those (more numerical than mathematical) difficulties can be mitigated by making more intelligent initial guesses for the wavefunction parameters. In many cases, extremely multi-reference systems modelled at very modest computational cost.

New wavefunction forms for describing multireference systems: Why, how, and what next?

ACS Fall 2021

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