Bridging Density Functional Theory and Machine Learning: Predicting Formation Energies of Oxide Perovskites
Author(s): Divya T L, Chandrani Chakravorty, Shashanka, Shashank Hegde
Publication #: 2509015
Date of Publication: 30.11.2025
Country: India
Pages: 1-6
Published In: Volume 11 Issue 6 November-2025
DOI: https://doi.org/10.62970/IJIRCT.v11.i6.2509015
Abstract
Perovskite oxides with the general formula ABO3 have formed the cornerstone in creating high functional materials because of their great tunability and versatility. Their application spans solar energy, superconductivity, catalysis, and fuel cells. Thermodynamic stability, expressed through formation energy, governs their usefulness. Conventionally, Density Functional Theory (DFT) is used to compute formation energies, but it is computationally expensive. In this work, we propose a machine learning-based prediction of perovskite formation energies using descriptors derived from elemental and structural features. After feature engineering and selection, several models were evaluated. XGBoost achieved the best R2 score of 0.9586, followed closely by Random Forest. These results demonstrate the potential of ensemble-based machine learning models in accelerating perovskite discovery.
Keywords: Formation Energy, Machine Learning, Random Forest, Oxide Perovskites, XGBoost
Download/View Count: 98
Share this Article