Unifying CO2 diffusion mechanisms in diamine-functionalized metal–organic frameworks via quantum-accurate machine learning dynamics

Diamine-functionalized Mg 2 (dobpdc) metal–organic frameworks (MOFs) (dobpdc 4 − = 4,4’-dioxidobiphenyl-3,3’-dicarboxylate) exhibit exceptional CO 2 capture, yet their CO 2 diffusion behavior has remained difficult to rationalize across pore sizes, loadings, and functionalization. Here we resolve this gap by combining diffusion measurements with quantum-accurate machine-learning-potential molecular dynamics (MLP-MD) to establish a unified, loading-dependent mechanism for CO 2 transport. We focus on bare and functionalized Mg frameworks featuring strong open-metal-site interactions and well-established cooperative CO 2 adsorption chemistry. Our MLPs reproduce density functional theory energetics and forces, enabling converged diffusion coefficients and microscopic trajectories. By comparing Mg 2 (dobdc), Mg 2 (dobpdc), and Mg 2 (dotpdc), we show that CO 2 is transiently trapped near open Mg sites at low uptake, producing suppressed diffusivity in both MLP-MD simulations and diffusion measurements. With increasing uptake, the progressive saturation of these strong adsorption sites screens their trapping potential, leading to a rise in diffusion coefficients for Mg 2 (dobpdc) and Mg 2 (dotpdc), quantitatively consistent with experiments. Diamine functionalization further accelerates CO 2 transport by limiting access to open Mg sites and introducing cooperative CO 2 –diamine interactions. Finally, the same MLPs accurately predict CO 2 adsorption isotherms via grand canonical Monte Carlo. Together, the experimentally validated MLP-MD framework reveals how pore size and diamine functionalization synergistically govern CO 2 diffusion, providing design rules for high-capacity MOFs with improved mass transport.