Materials informatics as a computational infrastructure for materials discovery

One of the critical barriers in transitioning new materials development into engineering practice is the uncertainty associated with the data that is used in the materials design process. While constitutive modeling strategies still form the foundations for computational materials design, uncertainty and incomplete information still pervade. Often the available data space is limited, and the challenge of using that data to develop computational studies for exploring new materials is made difficult. In this paper we describe how the application of information entropy metrics coupled with the tracking of the statistics of data evolution can aid in identifying key structure-property relationships in materials. This information in turn can aid in identifying new materials chemistries with targeted properties narrowed down from a large and sparse chemical search space. A brief example in the discovery of new solid-state electrolytes for fuel cell examples is given as a template for this informatics-aided computational infrastructure.