Thermodynamic Genome for Materials Informatics
The SGTE Solutions database, one of the largest thermodynamic databases available to date, contains the description of 680 binary systems and has grown by ~20 systems per year over the last 30 years. Extrapolating these growth rates, it will take another 185 years until all 4371 binary systems among the 94 elements naturally occurring on earth will be described.
To accelerate the exploration of chemical compound space, GTT-Technologies uses human and machine learning to build CALPHAD databases based on ab initio calculations. Here, the major steps in aiMP (ab initio Materials Project [1,2]) development will be presented, which involve raw data curation and accuracy assessment, identification of correlations, machine learning of S298K, and finally quality control. Phase diagrams will be shown that resemble many features of phase diagrams calculated based on full thermodynamic assessments. A few cases will be discussed where the aiMP calculated phase diagram is in major disagreement with assessed diagrams and the source of error will be discussed.
Finally, it will be shown how this thermodynamic genome database can be used for materials informatics applications.
 Jain, A. et al. “The Materials Project: A materials genome approach to accelerating materials innovation,”, APL Materials, 1.1 (2013): 011002.