(183h) Modeling the Reactivity of Open Metal Sites in Metal Organic Frameworks for Olefin Paraffin Separation

Metal Organic Frameworks (MOFs) are a class of porous material, in which metal nodes are connected using organic linkers. The properties of MOFs are highly tunable due to the combination of nodes and linkers available. Additionally, individual metal atoms on the nodes can be exchanged and exposed to create tailored open metal sites, thus resulting in a well-defined binding site immobilized on a support with controllable pore size. In this work, we investigate the reactivity of this site in a combination of density functional theory (DFT) calculations and experimental validation, focusing on the adsorption energy of gases in metal-exchanged MOFs. Here, we focus on the application of open metal sites in MOFs in the separation of ethane/ethylene and propane/propylene via selective adsorption of alkenes over alkanes.

We have calculated the binding energies for a variety of adsorbates (C₂H₄, C₂H₆, C₃H₆, C₃H₈, CO₂, CH₄, H₂, N₂, CO) in order to model the reactivity of the open metal site. Since DFT calculations of adsorption on metal centers are susceptible to self-interaction errors (SIE), we have compared different levels of theory to estimate the importance of self-interaction correction. We have also applied uncertainty quantification to show the variation in calculated binding energy due to the choice of exchange-correlation (XC) functional. We made use of multiple DFT codes in our work. Some calculations were performed using the CP2K code, with the PBE functional to describe the exchange correlation energy. Grimme’s D3 dispersion correction was used to account for van der Waals interactions. For uncertainty quantification, we used the XC-ensemble of the BEEF-vdW[1] functional implemented in the GPAW code. In addition to the periodic structure, we have also constructed a cluster analog around the binding site. This cluster was able to reproduce binding energies calculated in the MOF for both PBE and BEEF-vdW. We have also performed B3LYP calculations on these clusters using the Gaussian code, and SCAN calculations using VASP.

We have synthesized samples with varying concentrations of open metal sites and modeled the effects of the extent of metal substitution. Results from both DFT calculations and temperature program desorption experiments have shown that the binding energies of adsorbates are independent of metal loading and adsorbate coverage. While, the B3LYP and BEEF-vdW results agreed well with experimental measurements, the other functionals consistently over-predict (by ~50%) the binding. For example, in the case of CO binding, PBE predicted -1.76 eV, while BEEF-vdW, B3LYP (cluster), and experimental values were -1.12 ± 0.19 eV, -1.16 eV, and -1.06 ± 0.06 eV, respectively. Furthermore, the difference between PBE and other computational estimates and experimental values is much larger than the range of errors given by the BEEF-ensemble. This suggests that PBE overbinds due to errors in the electron density[2] caused by inherent shortcomings in DFT (e.g., SIE), and can cannot be removed by simply perturbing the functional form acting on this electron density. Therefore, great care must be taken in selecting the correct method for describing this system. Finally, based on these computational-experimental investigations, we identified a MOF that shows strong affinity towards double bonds (0.6-0.7 eV) and hence a significant separation would occur for the olefin/paraffin pairs. This prediction is validated and highly selective olefin/paraffin separation was achieved in column breakthrough experiments.

1. Wellendorff, J.; Lundgaard, K.; Møgelhøj, A.; Petzold, V.; Landis, D.; Nørskov, J.; Bligaard, T.; Jacobsen, K.; Phys. Rev. B 85, 235149 (2012).
2. Medvedev, M. G.; Bushmarinov, I. S.; Sun, J.; Perdew, J. P.; Lyssenko, K. A.; Science, 355, 49-52 (2017).