One of the most basic approaches to problem solving is to conceptualize the problem at different abstraction levels and translate from one abstraction level to the others easily, i.e., deal with them hierarchically. This concept is especially applicable to the field of novel materials discovery, wherein large candidate domains can be quickly and efficiently explored by hierarchically discarding irrelevant candidates. In this tutorial, we illustrate this approach using the example of wide band gap oxide perovskites. We will sequentially search a very large domain space of single and double oxide perovskites to identify candidates that are likely to be formable, thermodynamically stable, exhibit insulator nature and have a wide band gap. To this end, we will build four machine learning (ML) models: three classification and one regression model using experimental and DFT-calculated training data. The tutorial will discuss best practices for building ML models, commonly encountered pitfalls and how best to avoid them.

Researchers should cite this work as follows:

• 1. Talapatra, A., Uberuaga, B. P., Stanek, C. R., & Pilania, G. (2021). A Machine Learning Approach for the Prediction of Formability and Thermodynamic Stability of Single and Double Perovskite Oxides. Chemistry of Materials, 33(3), 845-858.

• Anjana Talapatra (2021), "ML-aided High-throughput screening for Novel Oxide Perovskite Discovery," https://nanohub.org/resources/perovmldis. (DOI: 10.21981/Q9CC-P934).

  BibTex | EndNote

1. data-driven materials discovery
2. density functional theory (DFT)
3. machine learning
4. material discovery
5. materials science
6. NCN Group - Machine Learning Hands-on
7. oxides
8. perovskites
9. Perovskites Discovery Tool (perovMLdis)
10. wide band gap