"The Evolution of Discovery: Mapping Chemistry’s Family Tree"

Abstract:

Do our materials have ancestors? Over generations of research, inorganic and materials chemists have re-imagined familiar scaffolds into new forms by changing substituents or replacing one element with another, each modification carrying the trace of earlier discoveries. Yet, unlike biological species, chemistry has no family tree to reveal how one idea led to another. In this talk, I will describe our efforts to build that tree and to ask a bold question: if we teach machines (materials) chemistry’s past, can they foresee its future? While such a question could be asked and answered in many areas of materials, we focus on transition-metal complexes. These complexes provide an ideal testing ground for this question because they have been systematically studied and crystallized to obtain precise structures over the past several decades, Additionally, these materials, metals with partially filled d- and f-orbitals interact delicately with organic ligands, producing a vast diversity of electronic structures and reactivities. I will present our efforts, both challenges and successes, in building this map and describe our early efforts to train generative models to predict the "future" of chemistry. Time permitting, I will also talk about our related efforts in learning on literature data and curating large datasets of both transition metal complexes and related metal-organic frameworks.