Answers are grounded in this paper and cite their sources.
High-entropy alloys (HEAs) exhibit excellent comprehensive properties. Nevertheless, their complex multicomponent compositions and prevalent chemical short-range order (SRO) impose substantial obstacles to conventional trial-and-error-based development. This review systematically compares mainstream first-principles modeling approaches for disordered alloys, including the virtual crystal approximation (VCA), special quasirandom structures (SQS), coherent potential approximation (CPA), and hybrid Monte Carlo/molecular dynamics (MC/MD) simulations. We further summarize their applications in investigating the phase stability, plastic deformation, and functional performances of HEAs, such as catalytic behavior and hydrogen storage capacity. This work also classifies machine learning interatomic potentials (MLIPs) into three generations: handcrafted descriptor-based potentials, system-specific E(3)-equivariant graph neural network potentials (represented by NequIP and Allegro), and universal pretrained potentials (exemplified by CHGNet and MACE). The respective advantages and limitations of these potentials for large-scale atomistic simulations concerning SRO and defects are elaborated. Furthermore, integrated alloy design strategies combining the CALPHAD method, active learning, and interpretable machine learning are introduced. Finally, critical bottlenecks, including dataset heterogeneity, insufficient model transferability, and low model interpretability, are identified. Future perspectives involving universal material foundation models and autonomous closed-loop discovery of alloys are also put forward. This review provides a systematic framework for the data-driven rational design of multi-principal element alloys.
No takes yet. Share an insight, caveat, or question.
Lv et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: