(88a) Accelerating discovery with computational chemistry in challenging materials spaces

Over the past 15 years, computational chemistry modeling with first-principles density functional theory has been transformed by advances in algorithms and computing power. More recently, acceleration via machine learning has led to orders of magnitude changes in scales of materials that can be studied. Nevertheless, challenges remain in identifying which first-principles model is sufficiently accurate for property prediction needed for the accelerated screening of materials. In few areas is this more challenging than in the open shell first-row transition metals that are ubiquitous in catalysis and functional materials. I will describe our efforts to build machine learning models that overcome limitations in first-principles modeling by either building models that guide method selection or leveraging consensus from functionals to build more robust workflows for the discovery of open shell transition metal complexes with light-harvesting and catalysis applications. I will also describe how we bypass direct simulation by building machine learning models to predict experimental measures of stability with application to metal-organic frameworks and use these models to guide de novo materials selection. Beyond these advances, I will describe challenges that remain in computationally guided materials discovery.