Curated EMTO-CPA elastic-constant data of Fe–Cr–Ni–M alloys, transcription records and analysis scripts

Abstract

Data and code accompanying the manuscript “Limits of cross-study transfer of composition-based surrogates for first-principles elastic constants of austenitic Fe–Cr–Ni alloys: a two-study benchmark” (H. Aourag). The record contains a curated data set of 48 austenitic alloys transcribed from two published EMTO-CPA (exact muffin-tin orbitals, coherent potential approximation, disordered local moments) studies: 33 Fe–20Cr–20Ni–xM alloys (M = Al, Co, Cu, Mo, Nb, Ti, V, W), from Al-Zoubi, Metals 9 (2019) 792, https://doi.org/10.3390/met9070792 15 Fe–20Cr–8Ni–xX alloys (X = Mo, Nb, Ta, Ti, V, W, Zr), from Dou, Luo and Zhang, Metals 9 (2019) 145, https://doi.org/10.3390/met9020145 For every alloy the file gives: the composition and four concentration-weighted descriptors (valence electron count, metallic radius, Pauling electronegativity, solute concentration); the lattice parameter and its origin (empirical or computed); the single-crystal elastic constants C11, C12, C′ and C44 and the polycrystalline B, G, E, B/G and Poisson ratio; the source table and row of each value. A transcription protocol and consistency checks (Voigt–Reuss–Hill reproduction of B and G) are included. The record also contains reference experimental and calculated values near the two base compositions, group-level validation errors (Table S3), and Python scripts that regenerate every number, table and figure of the manuscript. These cover nested leave-one-solute-out validation of ridge, PLS, kernel ridge and random-forest models, cross-study transfer tests, bootstrap uncertainty and the figures. A requirements file lists the software versions; the random seed is 0.