Predicting solid-state synthesis recipes for computationally-designed materials

DFT is widely used to predict structure-property relationships for materials design, but the Schrodinger equation provides little guidance on how to actually synthesize newly-predicted materials. Here, I will show how we can guide solid-state synthesis planning using information that is largely-available in high-throughput materials databases like the Materials Project. First, I will discuss a conceptual strategy to navigate high-dimensional convex hulls in the search of reactive precursors for more-efficient materials syntheses. Using this strategy, we design novel precursors for 32 target quaternary oxide materials, and validate with a high-throughput robotic synthesis laboratory that our DFT-guided precursors are substantially more successful at synthesizing complex multicomponent oxides than traditional precursors. Next, I will show that the onset temperature of a solid-state reaction derives from the extension of liquidus curves into the metastable region of a Temperature-Composition phase diagram. In order to predict this metastable liquidus curve, I will present a strategy to train CALPHAD models on DFT convex hulls and ASM phase diagrams, which enables us to rapidly estimate the high-temperature liquidus curves of phase diagrams at minimal computational cost.