Visualization-driven physics-informed modeling of energetics with demonstrated transfer between configurationally isomorphic substitutional alloys

Substitutional alloys pose a fundamental challenge for first principles modeling because the number of distinct chemical configurations grows combinatorially with system size. Here, we show that for the substitutional transition metal alloys considered here on rigid lattices, the configurational energetics exhibit a largely low dimensional and approximately quasi additive character. Using 1,337 density functional theory calculations for a Fe/Mn system, physics guided data visualization reveals that energy variations are governed by site dependent chemical substitutions and global magnetic effects, while structural degrees of freedom are effectively frozen. Motivated by these observations, we construct a compact, site resolved energetic model that explicitly reflects this underlying physical structure. A simple linear formulation achieves high predictive accuracy (RMSE = 1.03 × 10 −3 Ry, R 2 = 0.959) while remaining fully interpretable in terms of individual substitutional sites and magnetic contributions. Because the representation depends only on lattice topology and site identity, it can be transferred, in a zero-shot manner, to a chemically distinct but configurationally isomorphic Co/Ni system. In a strict zero-shot setting, the model preserves relative energetic ordering across chemical systems (Spearman ρ ≈ 0.832) despite large differences in absolute energy scales. Quantitative accuracy is substantially improved through few-shot calibration using a small number of additional reference calculations. These results demonstrate that visualization driven physical insight enables compact, interpretable, and transferable modeling strategy for substitutional alloys of the type studied here, offering a data efficient alternative to black box approaches