Finding a Needle in a Haystack: Success stories of Data Mining and Machine Learning for Electronic Materials Selection

Summary form only given, as follows. The complete presentation was not made available for publication as part of the conference proceedings. dentification of new materials is the bottleneck for a number of key challenges in electronics, including new interconnect materials, integration of 2D materials, and ferroelectric materials. The challenge is made greater by the need for these new materials to exhibit favorable chemical interactions at the interfaces with adjacent materials. I will describe our efforts to combine the curation and utilization of available experimental data with electronic structure calculations to screen tens of thousands of candidate materials for these applications, taking into account the desirable bulk properties and the interfacial chemistry. Another grand challenge is the identification of useful materials that have yet to be synthesized and characterized. I will further discuss our efforts to develop physics based machine learning to elucidate the full spectrum of electronic materials that are likely to be synthesizable and chemically stable, but have yet to be synthesized in the lab. In particular, we find that there are likely to be an additional 1000 2D materials that have yet to be synthesized, adding to the approximately 1500 2D materials that are already known to exist. I will discuss the predicted spectrum of properties of these materials.