Data Mining the Materials Genome Suggests Flat Bands Occur More Often in Hexagonal Superconductors than Non-Superconductors

Abstract

If Van Hove singularities (VHs’s) or Flat Bands (FB’s) near the fermi energy of a solid are important or even necessary conditions for superconductivity, and if there were a method to automate the detection of these electronic structure features, then scientists could data mine for potential superconductors by screening electronic structures for VHs’s and FB’s. To evaluate the incidence of VHs’s and FB’s in superconductors and non-superconductors, and to assess the structural and compositional characteristics of superconductors with VHs’s and FB’s, I constructed a 17,375- superconductor database with properties accumulated from online materials repositories. I developed a function to identify the presence of extrema in the density of states (DOS) of materials. I executed this function on both the entries in my database and samplings of non-superconductors. In contrast with metallic non-superconductors of the same crystal system, the set of metallic, hexagonal polymorphs of superconductors in my database exhibited 33% more structures with DOS peaks within 0.1 eV of Efermi. I report Nb, Fe, and B as the most frequent elements in the compositions of superconductors containing VHs’s/FB’s, occurring in 7.9%, 8.5%, and 8.9% of VHs/FB-containing superconductors in my database, respectively. My findings suggests the existence of a Van Hove singularity or Flat Band directly at Efermi may be a good screening parameter for hexagonal superconductors, as well as for superconductors containing Nb, Fe, or B. Based on the limitations I encounter in my research, I propose that the current standards of materials data analytics fail to provide sufficiently high quality information. I argue that a paradigm shift in the organization, acquisition, and experimental treatment of digital materials resources is crucial to cracking the Materials Genome.