Porosity Local Analysis (PoLA) for nanoporous carbons. Porous volume characterization and prediction of gas adsorption isotherms

We introduce a new computational procedure called PoLA (Porosity Local Analysis), a point-by-point description of the void space in nanoporous materials that surpasses the conventional representation of pores as homogeneous regions of regular geometry (spheres, cylinders, slits). Each volume element is assigned its own porous character through the minimum distance to opposite walls ( MinD ), a quantity directly linked to the host–guest interaction potential and therefore to physisorption behaviour. We apply PoLA to a dataset of 109 atomistic carbon models and correlate the resulting V( MinD ) distributions with N 2 and H 2 adsorption isotherms at 77 K, simulated by Grand Canonical Monte Carlo. A purpose-built machine learning procedure, based on optimized neural networks, infers V( MinD ) from a nitrogen isotherm and predicts the corresponding hydrogen uptake. Validation against four commercial activated carbons (Norit Row, Maxsorb, BAX1700, CGF4) shows excellent agreement between predicted and measured H 2 isotherms up to 60 bar, demonstrating that PoLA provides both a transferable porosity descriptor and a predictive tool for adsorbent design.