Data Mining-Aided Crystal Engineering for the Design of Transparent Conducting Oxides
- PDF / 394,978 Bytes
- 7 Pages / 432 x 648 pts Page_size
- 77 Downloads / 191 Views
Data Mining-Aided Crystal Engineering for the Design of Transparent Conducting Oxides Changwon Suh, Kwiseon Kim, Joseph J. Berry, Jinsuk Lee, and Wesley B. Jones National Renewable Energy Laboratory, 1617 Cole Blvd., Golden, CO 80401, U.S.A. ABSTRACT The purpose of this paper is to accelerate the pace of material discovery processes by systematically visualizing the huge search space that conventionally needs to be explored. To this end, we demonstrate not only the use of empirical- or crystal chemistry-based physical intuition for decision-making, but also to utilize knowledge-based data mining methodologies in the context of finding p-type delafossite transparent conducting oxides (TCOs). We report on examples using high-dimensional visualizations such as radial visualization combined with machine learning algorithms such as k-nearest neighbor algorithm (k-NN) to better define and visualize the search space (i.e. structure maps) of functional materials design. The vital role of search space generated from these approaches is discussed in the context of crystal chemistry of delafossite crystal structure. INTRODUCTION Crystal structure of materials is closely linked with its final property [1]. In this regard, the ability to understand structural factors governing desired properties is critical in better designing functional materials. However, one of the current challenges in finding functional materials arises from the lack of tools to explore the huge search space. A good example is the discovery process for advanced TCOs due to the extremely huge search space from the many possible combinations from the periodic table to meet the TCO requirements. There are two main approaches to handle huge search space. One is a combinatorial highthroughput synthesis and materials informatics to synthesize and interpret composition spreads, respectively [2, 3]. The other approach is to directly define the search space for TCOs such as structure mappings based on the concept of crystal chemistry [4]. While the former have been modernized successfully, the latter is still considered as a classical tool for identifying search space of materials. An example of the latter for designing new TCO includes identifying the role of the cations by Shannon et al in the 1970’s to the phase stability, chemical bonding, and transport properties [5, 6]. The starting point of their approach was a classical bivariate structure field map consisting of ionic radii of A and B sites of the delafossite ABO2 (ex. A=Cu, Ag, Pd, Pt; B=Co, Cr, Fe, Ga, In) structure [4], which were successfully revisited by Marquardt et al. later for exploring p-type TCOs [7]. This delafossite structure map was again noteworthy to the TCO community in the 2000s because only a few p-type TCOs such as Cu2O have been developed so far [8], and there is much that still needs to be explored. Nevertheless, the approaches of the latter are inconclusive to elucidate interrelationships between structural factors and electrical/optical properties because two structural parameters us
Data Loading...
