Combinatorial Materials Design through Database Science
- PDF / 200,825 Bytes
- 9 Pages / 612 x 792 pts (letter) Page_size
- 49 Downloads / 221 Views
JJ9.23.1
Combinatorial Materials Design through Database Science Changwon Suh, Arun Rajagopalan, Xiang Li and Krishna Rajan Department of Materials Science and Engineering Combinatorial Materials Science and Materials Informatics Laboratory Rensselaer Polytechnic Institute, Troy NY 12180-3590 Email: [email protected] URL: http://www.rpi.edu/~rajank/materialsdiscovery and http://cosmic.rpi.edu
ABSTRACT Large scale materials databases have been traditionally used for search and retrieval of experimental and theoretical data. In this paper, three different cases are used to illustrate applications of statistical techniques in databases that extend beyond searching. A complete large scale database of molten salts is visualized for pattern seeking. In the second case, a large virtual combinatorial library of chalcopyrite semiconductors is developed from a small experimental and theoretical dataset. This involves selecting statistically appropriate parameters based on the physics of the materials. In the third case, ‘secondary’ descriptors are developed for a zeolites database to better understand the topology of mesoporous structures and as a materials design tool. These examples serve to demonstrate how databases can be used to identify important combinations of parameters relevant to combinatorial experimentation. INTRODUCTION The development of large and diverse materials databases and creative application of data mining tools is essential for identification of new structure-property correlations and accelerated materials design. In this paper, we use three different statistical approaches on different databases to explore the following broad issues. • Pattern seeking through compression and visualization of complete databases. • Creation of a large virtual combinatorial library for materials design from a seed experimental and theoretical materials library. • Development of statistically derived descriptors for enhancement of standard materials databases for specific applications (such as materials design). As a case study for the first issue, a principal components technique is applied to a classic molten salts database. As for the second issue, chalcopyrite semiconductors are used as a testbed for combinatorial materials design. In it we show how a large library for band gap engineering can be built from a relatively small dataset. The third issue is illustrated by developing secondary descriptors for zeolite topology and using it as a design guide for zeolite frameworks. In this section we will first introduce the nature of different databases in these examples followed by a discussion of how combinatorial parameters can be extracted from these databases in each case study.
JJ9.23.2
RESULTS AND DISCUSSION
i.
Visualization of multivariate data - molten salts database
The molten salts database of Janz et al [1,2] is a excellent example of a rigorous, experimentally derived database. The multivariate nature of the data means that it is impossible to look at the entire database in a single plot. However we can st
Data Loading...
