The growing demand for energy-efficient transportation systems has intensified the need for structural materials that combine low density, high strength, and environmental responsibility. High-entropy alloys (HEAs), owing to their vast compositional flexibility and tunable mechanical properties, are promising candidates for nextgeneration lightweight structural applications. However, systematic experimental exploration of their expansive compositional design space remains time-consuming and resource-intensive. In this study, a physics-guided surrogate modelling framework is developed for sustainability-aware screening of HEA compositions under datalimited conditions. Starting from 32 experimentally reported alloys spanning FCC-, BCC-, and multiphase systems, a deterministic descriptor-driven strategy was employed to generate a physically consistent synthetic dataset across a 14-element compositional space. Key thermodynamic and atomic-scale descriptors including valence electron concentration (VEC), atomic size mismatch (δ), configurational entropy (ΔS mix), electronegativity deviation (χₛtd), mixing enthalpy proxies, and density-related features were incorporated to preserve metallurgical coherence. An XGBoost regression model trained on this physics-constrained dataset achieved strong internal consistency (R 2 ≈ 0. 99) under controlled noise conditions, reflecting accurate reconstruction of the embedded descriptor–property relationships. Validation against an independent experimental literature dataset (N = 58 alloys) yielded R 2 = 0. 81, indicating physically meaningful transferability despite realworld microstructural and processing variability not explicitly captured by composition-based descriptors. Feature-importance and SHAP analyses consistently identified VEC, atomic size mismatch, and density-related terms as dominant contributors to yield strength, aligning with established solid-solution strengthening mechanisms. To extend the framework beyond mechanical optimization, a sustainability index based on elemental abundance, toxicity, and resource criticality was integrated into a composite eco-performance metric. The results demonstrate that strength-to-weight efficiency and environmental responsibility can be jointly optimized within the explored compositional domain. Overall, this work establishes a transparent and reproducible foundation for physics-informed, sustainability-aware HEA screening, positioning surrogate modelling as a structured compositional pre-screening tool to accelerate data-driven alloy design while maintaining alignment with metallurgical principles and sustainability objectives.
Aswin Karkadakattil (Thu,) studied this question.