OCE-Perovskite-Gaps v1.1.0 — band gaps and densities of states of 1,784 halide perovskites (SIESTA-PBE/DZP), with the symmetry-equivalent copies identified and the revised analyses

Data and code supporting the manuscript Band Gaps of Halide Perovskites from an Orbital Cluster Expansion: Mechanism, Iodide Bias, and a Hybridization Channel (D. L. Azevedo, J. Phys. Chem. C, manuscript jp-2026-05221e). Version 1.1.0 accompanies the first revision of the manuscript. Everything in 1.0.0 is kept unchanged. What is new: Symmetry-equivalent copies. 543 of the 1,784 structures are copies of another under lattice symmetry, with identical labels: the mixed cells were produced by substituting sites of a 2×2×2 supercell, and different substitution patterns can be the same structure. The 253 groups are given in data/duplicatas_simetria_1784.json, with the script that regenerates them (code/r1/acha_duplicatas.py, pymatgen StructureMatcher, tolerances 10−3). The revised paper computes every number on the 1,241 distinct structures. A revised validation protocol. The ridge penalty is selected on an inner split grouped by cation family, the fit becomes deterministic, and uncertainties are given by paired bootstrap over test structures and over families. Scripts in code/r1/, outputs in results/r1/, including every main-text figure computed under the submitted protocol, with the copies removed, and under the revised protocol. The spin–orbit control of the twelve CsBX3 endmembers (Table 1 of the paper): 36 SIESTA runs with inputs, outputs and eigenvalue files, and the fully relativistic PseudoDojo set used. The equation of state of the twelve endmembers (Tables 12 and S5): 246 single-point calculations and the Birch–Murnaghan analysis. The figures of the revised paper. Correction of record. Three numbers of the submitted manuscript — the RMSE/σ of the hybridization channel alone (0.323), of the DOS features alone (0.812), and the descriptor's gain on N(EF) (−24.7%) — are not reproduced by the scripts of 1.0.0 on the data of 1.0.0, which give 0.335, 0.578 and −1.1%. All three are superseded in the revised paper. Every other figure that was rerun as a control is reproduced to the precision printed, or within one unit of its last digit (the metallicity AUC gain, 0.013 against the printed 0.012). The band gaps and densities of states are new to this work. The structure pool is shared with a separate deposit on cohesive energies (concept DOI 10.5281/zenodo.21228177), of which this is not a version.