Multireference (MR) electronic structure methods are indispensable in studies of strongly correlated electronic systems. Unfortunately, they are not black-box routines and require the manual specification of an “active space”. This is the set of orbitals used to build the most important electron configurations. But this selection has long been an art-form, and this prevented MR methods, unlike the DFT-based approaches, from widespread application, because only experienced researchers in MR methods could confidently construct these spaces. The active space selection was traditionally performed manually based on chemical intuition and detailed inspection of orbitals, and therefore was impractical in calculations of multidimensional potential energy surfaces in studies of different reaction paths. To alleviate this situation, I developed automated and systematic active space selection algorithms. By using linear algebraic transformations combined with density embedding theory and molecular orbital theory, I pioneered the development of two automated procedures for the construction of active orbitals—one targeted at transition metals and one at large <i>π</i>-conjugated systems. These techniques make MR calculations easier to execute, easier to reproduce by any users, even inexperienced ones, and simplifies the determination of the active space required for accurate results. This, for example, allowed me to study multidimensional potential energy surfaces for different electronic states at different conformations of the active site in the BLUF photoreceptor proteins, which regulates physiological responses to light in living organisms, and to identify the state responsible for the double proton transfer in their photocycle.<br/>

Turning multiconfigurational methods into “black-box” routines for an accurate description of strongly correlated enzymatic systems

2019 Fall National Meeting

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