Matthew McDermott1,2,Brennan McBride3,James Neilson3,Kristin Persson1,2
Lawrence Berkeley National Laboratory1,University of California, Berkeley2,Colorado State University3
Matthew McDermott1,2,Brennan McBride3,James Neilson3,Kristin Persson1,2
Lawrence Berkeley National Laboratory1,University of California, Berkeley2,Colorado State University3
Many materials that are predicted to possess important technological properties, such as multivalent battery cathodes (e.g., MgMo<sub>3</sub>P<sub>3</sub>O<sub>13</sub>), ferroelectrics (e.g., perovskite SnTiO<sub>3</sub>), etc., have yet to be successfully synthesized. The development of synthesis routes towards theoretically-predicted materials is often laborious, consisting of an iterative approach relying on a great degree of "chemical intuition". What is missing is a paradigm much like what currently exists for computational materials discovery -- a corresponding materials "synthesis by design”. Recently, we published two studies on the development of tools that can enable synthesis by design: 1) a graph-based network approach for reaction pathway prediction in solid-state ceramic synthesis, and 2) hyperdimensional chemical potential diagrams.<br/>In this presentation, we will derive the use of a new selectivity metric for solid-state synthesis from a simple theoretical picture of the chemical potential difference at reacting solid-solid interfaces, enabled by the use of hyperdimensional chemical potential diagrams. We will illustrate the inherent trade-off between a solid-state reaction’s thermodynamic driving force and its selectivity, and show how the chemical potential difference metric is a necessary complement to using reaction energy in designing solid-state synthesis routes. Using our previously developed reaction network approach, we will show how the incorporation of a selectivity metric in a computational synthesis prediction workflow can enhance the likelihood that predicted synthesis routes will be experimentally successful. We will present synthesis predictions to several technologically important materials, including BiFeO<sub>3</sub>, MgCr<sub>2</sub>S<sub>4</sub>, and SnTiO<sub>3</sub>, and compare them both with results from previous literature, as well as new results from ongoing attempts to experimentally synthesize these materials using our predictions.