The prediction of the lattice constant of binary body centered cubic crystals is performed in terms of first principle calculations and machine learning. In particular, 1541 binary body centered cubic crystals are calculated using density functional theory. Results from first principle calculations, corresponding information from periodic table, and mathematically tailored data are stored as a dataset. Data mining reveals seven descriptors which are key to determining the lattice constant where the contribution of descriptors is also discussed and visualized. Support vector regression (SVR) technique is implemented to train the data where the predicted lattice constants have the mean score of 83.6% accuracy via cross-validation and maximum error of 4% when compared to experimentally determined lattice constants. In addition, trained SVR is successful in predicting material combinations from a desired lattice constant. Thus, a set of descriptors for determining the lattice constant is identified and can be used as a base descriptor for lattice constants of further complex crystals. This would allow for the acceleration of the search for lattice constants of desired atomic compositions as well as the prediction of new materials based on a specified lattice constant.
|Journal||Journal of Chemical Physics|
|Publication status||Published - 28 May 2017|
ASJC Scopus subject areas
- Physics and Astronomy(all)
- Physical and Theoretical Chemistry
FingerprintDive into the research topics of 'Descriptors for predicting the lattice constant of body centered cubic crystal'. Together they form a unique fingerprint.
High Performance Computing (HPC) Facility
Steven Chapman (Manager)University of Bath