Searching for New Material Properties Using Artificial Intelligence

Superconductors, materials that lack electrical resistance below a specific temperature, possess a range of unique properties that allow for their application to scientific and technological advancements. However, the search for properties that could identify superconductors is computationally expensive and time consuming. Using a dataset comprised of materials based on the 2-dimentional monolayer structure A₂B₂X₆, we examine the role of the X site to determine if formation energy can be predicted by a known list of material descriptors of superconductive properties through the implementation of neural networks. This data-driven approach has revealed the importance of understanding the best descriptors for determining formation energy in superconductive materials, such as the importance of electron affinity in defining a material’s formation energy. Our approach can provide insight into future research based on discovering the material properties of superconductive materials.