Abstract
<jats:p>Pore size, pore shape, and framework density are among the most direct chemical handles available for tuning a metal–organic framework's (MOF's) affinity for CO2 over N2 and H2O, yet these structural descriptors are rarely used systematically to narrow a large structural library before committing to expensive Grand Canonical Monte Carlo (GCMC) simulation of co-adsorption. We present the GulfMOF library, a curated set of 12,597 MOF structures drawn predominantly from the CoRE MOF 2019 and QMOF repositories and retained after requiring a pore limiting diameter (PLD) of at least 3.3 Å — the kinetic diameter of CO2 — so that every framework in the library is CO2-accessible by construction. For each framework we compute four pore-geometry and packing descriptors — PLD, largest cavity diameter (LCD), framework density, and void fraction (available for 34.6% of the set, reflecting the mixed provenance of the source repositories) — and characterize how they are distributed across the library. A full elemental survey of the underlying CIF files shows the index spans 75 unique chemical elements, including the lanthanide series, common transition metals, and alkali/alkaline-earth metals. Building on these descriptors, we propose a transparent, editable geometric pre-screening protocol intended to prioritize a computationally tractable shortlist ahead of full GCMC evaluation of CO2/N2/H2O co-adsorption under hot, humid operating conditions. Structures below a framework density of 0.3 g/cm3 (1.1% of the set) are excluded under a documented, literature-motivated — but not independently validated — stability heuristic; the remainder are ranked by a composite score combining an LCD target-range term (5–15 Å) with a min–max-normalized inverse-density term, with both weights left editable. We report the ten highest-ranked structures as a representative example and release the full ranked list, the geometric descriptors for all 12,597 structures, and the underlying scoring spreadsheet. We are explicit throughout that this ranking is a first-pass geometric filter and not a substitute for GCMC: simulation-based validation of the resulting shortlist is in progress and will be reported separately.</jats:p>