(427c) Breaking the Rules: Open Shell Systems Break Strong Scaling Relations and Allow Discovery of Materials with Improved Multiple Property Targeting

Computational high-throughput virtual screening (HTVS) with first-principles density functional theory (DFT) can play a valuable role in unearthing design rules for scalable, industrially viable synthetic analogues that preserve selectivity and activity observed only in enzymes. A number of enzymes (e.g., methane monooxygenase) display activity for functionalization of inert C-H bonds, demonstrating their utility as potent, yet selective oxidants. Single-site catalysts represent the most promising synthetic analogues to these few-site enzymes, often enabling atom-economy, tunability, and selectivity not possible with bulk heterogeneous catalysts. Single-site catalysts with 3d transition-metals can access a range of spin- and oxidation-states. Designing the accessible spin and oxidation states is thus an important part of understanding their catalyst reaction energetics. We demonstrate our developments on HTVS for single-site light alkane oxidation catalyst energy landscapes, which quantify how spin- and oxidation-state distort the light-alkane oxidation energy landscape and complicate the process of catalyst screening. Through our expanded understanding of the open-shell alkane oxidation energy landscape, we uncover new hypotheses for puzzling experimental activity trends for open shell transition metal-oxo systems. Simultaneously, we decouple the impact of distortion from the impact of spin-state, demonstrating that metal distortion can drastically impact thermodynamics in open-shell species, while having little impact on their closed-shell equivalents, thus demonstrating the distinct behavior of open shell compounds relative to their closed shell counterparts. Lastly, we harness the scatter in the energy landscape of open-shell single-site catalysts, moving beyond linear free energy relationships (LFERs) to find catalysts that would not be uncovered by energetic descriptor-based screening. To do this, we employ independently trained models on different reaction energy steps, and use uncertainty-quantified active learning to develop robust models and screen large design spaces derived from experimental macrocycles.