Crystal Growth & Design
Leonid Zhukov Roman Eremin Innokentiy Humonen Pavel Zolotarev Inna Medrish Semen Bydennyy

Hybrid DFT/Data-Driven Approach for Searching for New Quasicrystal Approximants in Sc-X (X = Rh, Pd, Ir, Pt) Systems


Intermetallic compounds formed by two or more metals are characterized by wide structural diversity. The design of complex intermetallics, such as quasicrystals or their approximants, is a challenging scientific problem. We present a hybrid computational approach for searching for new stable 1/1 Mackay-type quasicrystal approximants in Sc-rich intermetallics. For the Sc-Rh, Sc-Pd, Sc-Ir, and Sc-Pt systems, we developed a routine for the generation of simplified composition/configuration spaces that contain up to 2107 configurations each. Using the density functional theory (DFT) for the evaluation of their thermodynamic properties, we rationalized extended search spaces consisting of 3942 configurations each. Random Forest and graph neural network (GNN) regression models were trained on the simplified sets and utilized as inference machines for enhanced search spaces. Among the systems studied, more than 20 potentially new Sc-Pt and Sc-Pd approximants were predicted and confirmed within DFT calculations. A comparative analysis shows that GNN performance depends on the target energy type─relaxed energy, formation energy, and energy above the convex hull. Nevertheless, GNN-based solutions may provide a higher recall of predictions depending on architecture modifications and target energy chosen.