Many compelling functional materials and highly selective catalysts have been discovered that are defined by their metal-organic bonding. The rational design of de novo transition metal complexes however remains challenging. First-principles (i.e., with density functional theory, or DFT) high-throughput screening is a promising approach but is hampered by high computational cost, particularly in the brute force screening of large numbers of ligand and metal combinations. In this talk, I will outline our efforts over the past few years to develop representations for accurate and interpretable machine learning of open shell transition metal complex properties. I will describe how analysis of models and feature importance provide guidance on design rules for tailoring properties in transition metal chemistry (e.g., geometry, spin state, redox potential, and catalysis energetics). I will describe how we pair these models with new uncertainty quantification metrics to carry out multi-objective evolutionary optimization of transition metal complexes. Finally, time permitting, I will describe how these same representations also allow us to predict when simulations or electronic structure methods will succeed and fail, enabling the development of decision engines that provide autonomous control of simulation workflows when paired with evolutionary algorithms for chemical space exploration.
Back to Workshop II: Interpretable Learning in Physical Sciences