A Modular Classical Density-Functional Framework for Gas Adsorption in Nanoporous Materials: From First-Principles Binding Energies to Kinetic Molecular Sieving

Abstract

Classical density-functional theory (cDFT) predicts fluid adsorption in nanoporous solids at a fraction of the cost of molecular simulation, yet most implementations are one-dimensional, assume a rigid host, or are hard-wired to a single fluid and interaction model. We present porecdft, a modular, host- and fluid-agnostic three-dimensional cDFT package built around a pluggable external-potential interface that accepts analytic force fields (Lennard-Jones, Morse, Gaussian-smeared-Coulomb, quadrupole-electric-field-gradient) or pretabulated grids from ab initio molecular dynamics and machine-learning interatomic potentials. The excess free energy couples White-Bear II fundamental-measure theory with Wertheim first-order thermodynamic perturbation theory (TPT-1) association and an optional elastic framework response, solved by an Anderson-accelerated fixed point. On the CO2/aluminum-formate (ALF) benchmark, the parameter-free fundamental-measure functional reproduces the 298 K isotherm to a root-mean-square error (RMSE) of 0.44 mmol g–1 and the calorimetric isosteric-heat band (25–32 kJ mol–1); the anomalous ordering N (323 K) ≈ N (298 K) ≫ N (273 K) is identified from ab initio molecular dynamics as a kinetic effect of formate-linker librations, and the CO2/N2 Ideal Adsorbed Solution Theory (IAST) selectivity of ∼4, two orders of magnitude below the experimental 350–600, establishes ALF as a kinetic molecular sieve in which transport, not equilibrium, governs separation. Applied without modification to H2 in four metalated covalent organic frameworks across five first-row transition metals, the same engine ranks cobalt highest in every framework through its broad, soft Morse well rather than its depth, a result that requires a three-dimensional treatment.