phea-atlas: a calibrated probabilistic decision atlas for 5-component high-entropy alloys (1,086,008 systems)

Probabilistic decision atlas for five-component high-entropy alloys (HEAs). For every one of the 1,086,008 five-element subsets of a 44-element pool, the atlas reports a calibrated probability of forming a solid solution, P(HEA) in [0, 1], together with its epistemic (Gaussian-process) uncertainty, summarised over each subset's composition grid: the mean, maximum, and equimolar P(HEA), the optimal composition, and the mean propagated uncertainty. The atlas is the derived output of the phea-atlas software (linked below): a three-layer, physics-guided Bayesian expert system combining a Miedema semi-empirical prior on binary mixing enthalpies, a Gaussian-process posterior trained on experimental binary mixing enthalpies (with analytic uncertainty propagation to any multicomponent composition), and a temperature-scaled logistic outcome layer calibrated on experimental HEA outcomes. Files: atlas_phea.csv (canonical atlas, exact covariance propagation), atlas_phea_independence.csv (diagonal/independence approximation used for the appendix comparison), ATLAS_README.md / ATLAS_README.pdf (column schema and provenance), and summary JSONs (atlas_frequency.json, independence_atlas_comparison.json). See ATLAS_README for the full column description and the underlying data sources to cite. The accompanying manuscript is currently under peer review; citation details will be completed on acceptance.