Abstract

This talk will highlight both the successes and the stumbling blocks along the way to bring material science into the 21st century. Namely, the combination of (i) high throughput experiment, (ii) massive computational resources, (iii) machine learning, and (iv) high energy beamlines is more than summative when conducted in concert. Given finite human resources (ie graduate students), these tools serve as force multipliers to enhance their capabilities. Over the last decade, we have applied these techniques in thermoelectric and photovoltaic materials. This high through-put search of known and hypothetical compounds has led to the discovery of new classes of materials for these two applications. Through the interplay of materials physics and machine learning, general principles for designing the next generation materials have emerged. The talk will conclude with discussion of the prospects for sequential learning coupled with parallelization and automation to accelerate functional materials discovery.