Elucidating CO₂ Dynamics in High-Entropy MOF-74 via Machine Learning Interatomic Potentials

Klichchupong Dabsamut^1*, Ching-Ming Wei¹

¹Institute of Atomic and Molecular Sciences, Academia Sinica, Taipei, Taiwan

* Presenter:Klichchupong Dabsamut, email:kdabsamut@as.edu.tw

High entropy variants of MOF-74 have demonstrated enhanced CO₂ adsorption performance, yet it remains unclear whether homogenous metal mixing alone produces intrinsic enhancements in transport beyond the behavior of single metal parents. We develop a machine learning interatomic potential (MLIP) to study CO₂ diffusion in the Mg, Co, Cu, Ni, and Zn variants of MOF-74, the equimolar high-entropy composition, and M-rich (2:1:1:1:1 mixture) condition. Using molecular dynamics driven by MLIP at full CO₂ loading, we determine the axial diffusion along the channels, with single-metal values at room temperature spanning 0.322 to 1.211×10⁻⁸ m²/s, with Mg the slowest and Cu the fastest. For the equimolar high-entropy framework, the axial diffusivity at saturation agrees with the composition-weighted average of the parents to within a few percent and remains inside the range defined by them. The M-rich mixtures follow the same pattern, which indicates that compositional disorder does not create a new transport regime under these conditions. Temperature-dependent simulations from 300 to 500 K follow Arrhenius behavior, with parent activation energies covering 0.044 - 0.099 eV and prefactors ~ 6 - 15×10⁻⁸ m²/s; the high-entropy case sits near the mean (E_a= 0.054 eV, D₀= 7.33×10⁻⁸ m²/s) and tracks the parent average across the full window. Taken together, these results show that, under dry, defect-free, fully saturated conditions, CO₂ transport in MOF-74 is additive rather than synergistic, providing a clear theoretical baseline for designing compositionally complex frameworks.

Keywords: MOF-74, High-entropy MOFs, CO₂ transport, Machine learning, CO₂ diffusion