Advanced Journal of Environmental Science and Technology

ISSN 2756-3251

Advanced Journal of Environmental Science and Technology | Vol. 17 (10), pp. 120-134, October, 2026
DOI: 10.46882/2026.AJEST.171009

Research Article

Machine-Learning-Guided Design of Metal-Organic Framework Membranes for Selective Post-Combustion Carbon Dioxide/Nitrogen Separation

Aravind Subramanian¹*

¹Department of Chemical Engineering, Indian Institute of Science, Bengaluru, India

Corresponding Author: Aravind Subramanian — [email protected]

Abstract

Screening thousands of candidate metal-organic frameworks (MOFs) for mixed-matrix gas separation membranes is computationally prohibitive using quantum chemical simulations alone. We trained a graph neural network on grand canonical Monte Carlo adsorption datasets to predict CO₂/N₂ selectivity and permeability across 14,500 hypothetical MOFs, experimentally fabricating the top-ranked fluorinated aluminum MOF (Al-FFIVE-1-Ni) in a Pebax-1657 polymer matrix. At 15 wt% filler loading under humid flue gas conditions (15% CO₂, 85% N₂, 35 °C), the mixed-matrix membrane achieved a CO₂ permeability of 642 Barrer and a CO₂/N₂ selectivity of 68.4. Performance surpassed the 2008 Robeson upper bound without plasticization over 300 hours of continuous testing.

Keywords: carbon capture membranes; metal-organic frameworks; machine learning screening; mixed-matrix membranes; flue gas separation; Robeson upper bound

Manuscript Timeline: Received 11 September, 2026; Revised 14 October, 2026; Accepted 27 October, 2026; Published 31 October, 2026.

Citation: Subramanian, A. (2026). Machine-Learning-Guided Design of Metal-Organic Framework Membranes for Selective Post-Combustion Carbon Dioxide/Nitrogen Separation. Advanced Journal of Environmental Science and Technology, 17(10), pp. 120-134.