First principles methodologies have grown in accuracy and applicability to the point where large databases can be built, shared, and analyzed with the goal of predicting novel compositions, optimizing functional properties, and discovering unexpected relationships between the data. In order to successfully perform these tasks, however, specific property descriptors to link computable quantities with functionalities of interest must be defined.
The story in this talk is organized around the concept of descriptors as a way to provide a structure for scientists’ intuition and guide the discovery process.