Design and Evaluation of Deep Learning Models for Stability of Heusler Alloys

Computationally efficient methods for materials investigation, whether from first principles or deep learning models, have aided and accelerated the process of discovering and designing new materials. Here, we focus on a deep-learning-based approach, made possible by open and easily searchable materials databases like the Materials Project [1, 2]. Once appropriate data has been obtained and processed, one is presented with a broad array of choices in building the actual model. In this presentation, we discuss an approach to building and refining a deep learning model architecture for use in predicting the stability and formation energies of intermetallic compounds with the Heusler structure. This includes evaluating the strength of the correlation of the variables in order to determine which are most important. Additionally, the effects of altering the complexity of the model on its predictive power were also evaluated. The model and the results were compared with previous models found in the literature.

[1] Jain, A., Ong, S. P., Hautier, G. et al. Commentary: The Materials Project: A materials genome approach to accelerating materials innovation. APL Mater. 1, 011002 (2013).

[2] Horton, M.K., Huck, P., Yang, R.X. et al. Accelerated data-driven materials science with the Materials Project. Nat. Mater. 24, 1522–1532 (2025). https://doi.org/10.1038/s41563-025-02272-0.