S.K. Ethirajan, A. Kulkarni
University of California, Davis,
United States
Keywords: Enhanced Sampling, Diffusion, Metal Organic Frameworks, Uncertainty Quantification, Active Learning, Machine Learning
Summary:
Understanding how small molecules diffuse through metal–organic frameworks (MOFs) is critical for applications in gas separation, catalysis, and chemical sensing. In many MOFs, transport is governed by rare, activated events that are difficult to capture with conventional atomistic simulations. As a result, important diffusion pathways may remain hidden, limiting mechanistic understanding and predictive design. Here, we investigate temperature-dependent diffusion of imidazole in the SALEM-2 MOF using machine-learning interatomic potentials that remain reliable under strongly distorted, off-equilibrium conditions while retaining first-principles fidelity.[1] By employing rare-event sampling to access long simulation timescales, we resolve activated transport mechanisms over a wide temperature range (300–700 K). Diffusion through the conventional six-membered ring (6-MR) windows constitutes the dominant low-barrier pathway. In addition, we uncover a previously unreported diffusion mechanism through the narrower four-membered ring (4-MR) windows. At lower temperatures, 4-MR transport proceeds via cooperative framework flexibility and transient ring opening. At elevated temperatures, a distinct node-assisted pathway emerges that involves temporary Zn–N bond dissociation. From a thermodynamic perspective, these behaviors reflect access to higher–free-energy framework configurations that become increasingly populated at elevated temperatures, leading to a qualitative change in the dominant diffusion mechanism. These findings highlight limitations of classical force fields and static structural descriptors. Notably, the framework flexibility and transient Zn–N bond dissociation identified here are also key atomistic motifs implicated in solvent-assisted ligand exchange (SALE), suggesting that SALEM-2 provides a useful mechanistic platform for understanding how dynamic framework behavior enables post-synthetic modification. More broadly, this work demonstrates how advanced atomistic modeling can uncover hidden transport pathways relevant to the rational design of MOFs for separation and catalytic applications under realistic operating conditions. Reference: 1. Ethirajan, S. K.; Kulkarni, A. R. Modeling Diffusion in Metal–Organic Frameworks Using On-the-Fly Probability Enhanced Sampling-Based Machine Learning Potentials. J. Chem. Theory Comput. 2025, 21 (21), 11197–11209. https://doi.org/10.1021/acs.jctc.5c01191