Functional Materials by Design – Data and Automated Labs

The design of electronic and thermal materials, both purely inorganic compounds as well as inorganic-organic hybrids is a difficult challenge due to the large state space in the Structure-Process-Property-Performance paradigm. The process of Materials-by-Design constitutes multiple steps including (A) invertible representations of structure, followed by (B) creation of a database of out-of-equilibrium functional properties, and finally (C) Machine Learning models. Actual synthesis of new materials and composites requires process parameter tuning, where Bayesian Optimization proves fruitful. I will describe our efforts to design new inorganic thermoelectric materials, predicting their performance via machine learning models, and end with the description of a rapid synthesis technique producing pure phase high performing thermolectric materials. I will share some of the new materials that have been generated through a Wyckoff-based crystal representation and validated through our rapid synthesis technique.