Title: A Curated Literature Database for Double and Vacancy-Ordered Halide Perovskites (A2BB'X6 / A2BX6) and Related Compound

ummary This dataset is a manually curated, source-traceable compilation of structural, electronic, mechanical, optical, thermoelectric, magnetic, and device-level (solar-cell simulation) properties for the double-perovskite family A2BB'X6 and the vacancy-ordered perovskite family A2BX6 (X = F, Cl, Br, I), together with chemically related families (double oxide perovskites, complex hydrides, magnetic fluorides). Data are extracted from peer-reviewed articles and preprints, with systematic traceability to the original source (file name, DOI). Scale 23 thematic sheets, ~1,600 data rows in total 209 unique compounds covered across the five principal property sheets (structural, electronic, elastic, optical, thermo-mechanical), plus several dedicated specialised families (Mott-insulator fluorides A2MF6, Weyl half-metal polymorphs X2RhF6, hydrides A2RuH6 and Q2LiMH6, OER electrocatalysis, SCAPS-1D/TCAD photovoltaic device simulations) Methods covered: DFT (GGA-PBE, LDA, TB-mBJ, HSE06, GW/BSE, DFT+U, DFT+DMFT) and experimental characterisation (XRD, XPS, Raman spectroscopy, UV-Vis absorption) Data model Each record is retained at the level of composition + phase + method + condition + property, rather than collapsed into a single averaged value per compound. This design preserves genuine disagreements between independent sources instead of masking them. Every value is annotated with its source and, where relevant, an explicit quality/comparability status. Known data-quality issues already documented in the database (Notes & Vérifications sheet, 54 entries) Major structural homonymies (K2RhF6/Rb2RhF6: three incompatible sources under the same nominal formula — Mott insulator vs. Weyl half-metal) Unresolved conflicts between independent sources (Cs2PtI6) Internal inconsistencies within individual source articles (text vs. table) Unit ambiguities (Cs2Au2X6: elastic constants reported in N/m instead of GPa) Intended use Screening of lead-free candidates for optoelectronic applications, comparative literature meta-analysis, theory–experiment cross-validation, and extraction of subsets suitable for machine learning (a "ML-ready" subset of 25 compounds with complete property coverage, together with computed structural descriptors, has been extracted separately). Known limitations: property coverage is heterogeneous across families (structural data are nearly complete; thermoelectric data are very sparse); computational methods are not harmonised across sources and must be filtered via the Method column before any quantitative comparison.