Machine Learning

Machine learning has emerged as a promising tool for prediction of the known and unknown properties of materials and discovery of new materials as well. The reliability of model developed using machine learning depends on the access to high-quality data. The high-quality data involves well-optimized structures, accurately calculated various properties of materials. In this regard, a functional materials database named “aNANt” is established. aNANt is first such computational materials database from India, it currently holds around 15,000 structures and expected to group up to 25,000 in month’s time. Other than the predicted properties and structure we intend to add tools for theoretical analysis as well. We highly encourage other research groups from India to reach us for further improving the strength of the database. The link to database is (

Bandgap prediction of MXenes using Machine Learning. (A) left and right are BB’ (T at the top of both M/M’) and CB (one T at the top of M, other at C) phase of MM’XTT’ MXene. (B) Importance of the feature involved in the prediction of the band gap of MXene by Relief algorithm, most of them are elemental properties of constituent elements extracted from the chemical repository. (C) Correlation between the target property, i.e., GW band gap with features involved in prediction. The correlation was found to be the lowest with PBE band gap. Since, PBE bandgap was not taken as one of the features, the model makes prediction over correction. (D) Scatter plot of predicted and true value for primary and compound (generated from mathematical operation )features, GPR model with eight primary feature reported the minimum error (rmse) of 0.14 eV over other methods.


  1. A. C. Rajan, A. Mishra, S. Satsangi, R. Vaish, H. Mizuseki, K. R. Lee and A. K. Singh. Machine Learning-Assisted Accurate Band Gap Predictions of Functionalized MXene. Chem. Mater. 30, 4031 (2018)
  2. aNANt: A functional materials database (
Close Menu