Lessons Learned in Combining Computational and Experimental Materials Discovery—A P-Type Transparent Conductor Case Study

High-throughput computational material screenings are a promising approach to rapidly narrow down a wide material space to identify promising new compounds for energy applications and devices. Since the launch of the Materials Genome Initiative in 2011 and the establishment of computational materials databases, targeted searches have yielded a plethora of new candidates, some of which have been experimentally verified. However, many synthesis and device attempts are less successful, and these are usually not published. Thus, there is still a significant divide between transforming predictions into results that are actually achievable in the lab, and an even greater lag in scaling predicted materials into functional devices.<br/>This talk will synergize high-throughput computation, combinatorial experimental methodologies, and device integration, highlighting key challenges faced in scaling up materials discovery and proposing strategies to address these challenges. We will focus on a case study to discover and optimize new p-type transparent conductors (TCs); a high-quality p-type TC does not yet exist, but could enable advances in a wide range of photovoltaic applications, among others.<br/>First, computational approaches to materials discovery will be discussed. We define and evaluate descriptors currently used in computational screenings for TCs, showing several to have a remarkably high degree of false negatives and positives. These metrics have been used by our group and other researchers to perform various tiered computational searches, and to identify promising p-type TC compounds that have not previously been grown.<br/>Second, we discuss the synthesis of these predicted semiconductors as thin films and focus on a few ternary case studies. Our approach uses combinatorial sputter deposition, which allows for tunability of properties, followed by characterization using both combinatorial and in-depth techniques. Some successes will be discussed and also challenges encountered in the laboratory. The next step is incorporating these new p-type TC materials into solar cell devices. We have fabricated both silicon heterojunction and cadmium telluride solar cells using a novel p-type contact to explore the challenges bringing predicted materials to actual highly-optimized device stacks. This understanding, coupled with simulations, can lead to the development of additional new computational descriptors and screening criteria.<br/>Ultimately, from this combined computational-experimental materials discovery framework in which each phase continues to inform subsequent phases, we outline a set of “lessons learned.” These lessons include the need to (1) thoroughly benchmark descriptors and screening methods, to (2) reframe screenings a multi-objective optimization problem to assess how descriptors are actually coupled together, and to (3) screen for tunability in computational predictions and assess the range of experimentally achievable properties within a given material. This talk highlights grand challenges in bringing predicted materials into real-world devices, and provides insight to inform future strategies for materials discovery in TCs and beyond.