Accelerated data-driven accurate band gap prediction

The application of machine learning (ML) to develop structural and functional materials is relatively new, however, has great potential to accelerate the discovery of materials for a desired application. A foremost requirement in designing predictive models is the availability of high quality scientific data in the form of shared materials databases. Continuous update and establishment of new databases are crucial for targeted applications. In this regard, Materials Theory and Simulations Group at Materials Research Centre has developed an open-access online repositoryof functional materials. The database is called aNANt and hosts the structures and electronic properties of more than 10000 functionalized MXene. It is expected to grow to host 25000 functionalized MXene within few months. Using information from this database, we have recently developed a ML model to predict the accurate band gap of functionalized MXene. Standard local and non-local density functionals based approach severely underestimates the band gap of a material. In order to correct the band gap, many body perturbation theory based approach such as GW method is employed. However, the GW methods are computationally very expensive and could take several years to calculate the band gap of the functionalized MXene available in aNANt. Our ML based model can predict the accurate band gaps of more than 10000 MXene in matter of minutes.

Band gap predictions of the MXenes: (a) two prominent phases of the MXene, (b) importance of the features for prediction, (c) their correlations to true (GW) band gaps, and (d) scatter plots showing the predicted versus true band gaps in the cases of primary and non-linear compound features for Gaussian Process (GPR) and bootstrap aggregated (bagging) regression models with train/test R²and rmse values.

References

(i) aNANt: a functional materials database. URL: http://anant.mrc.iisc.ac.in/

(ii) Rajan, A. C.; Mishra, A.; Satsangi, S.; Vaish, R.; Mizuseki, H.; Lee, K. R.; Singh, A. K. Chem. Mater.2018 In press. URL: https://pubs.acs.org/doi/10.1021/acs.chemmater.8b00686

Research Group URL: http://mrc.iisc.ac.in/abhishek/