Understanding toxic gas adsorption in MOFs via high-throughput computational screening and machine learing

Determining the best metal organic frameworks (MOFs) for a specific application is getting harder while thousands of them have been experimentally synthesized and hundreds of thousands of them have been computationally generated. Increasing demand for more efficient gas separation systems also brings about a more complicated, time consuming, and dangerous experiment environment. In this thesis, we focused on high-throughput computational screening of hybrid (QMOF) and experimental (CoRE MOF) MOF data sets for several separation and storage applications to overcome these limitations. Gas uptakes of MOFs were computed configurational biased Monte Carlo (CBMC), and grand canonical Monte Carlo (GCMC) simulations, and self-diffusivity of molecules in MOF structures was determined using molecular dynamics (MD) simulations at several conditions. Various adsorbent performance evaluation metrics, such as selectivity, working capacity, adsorbent performance score, and percent regenerability, were used to identify the best adsorbent candidates. In the first part, we examined volatile organic compounds (VOCs) capture from air, and our results showed that more than one-third of our MOFs have higher (butane) C4H10 selectivities than commercial zeolite MFI. The top five MOFs have C4H10 selectivities between 6.3"×" 103-9"×" 103 (3.8"×" 103-5"×" 103) at 1 bar (10 bar). Analysis of the structure-performance relations demonstrated that MOFs with mediocre porosity (0.4-0.6) and narrow pore sizes (6-9 Å) tend to have high C4H10 selectivities. Radial distribution function analyses of the top materials revealed that C4H10 molecules predominantly localize near the organic linkers of the MOFs. In the second part, we focused on (propane) C3H8 capture with MOFs and our results demonstrate that (vacuum-temperature swing adsorption) VTSA is the most effective process for many MOFs offering high regenerability (>90%), exceptional C3H8 selectivity (>7×103), and high C2H6+ C3H8 selectivity (>100). Top-performing MOFs are characterized by narrow pores (<10 Å), low porosities (<0.7), aromatic ring linkers, and alumina or zinc nodes. These MOFs outperform commercial zeolite MFI for air separation and surpass several commercial MOFs for natural gas streams. In the last two parts, we showed that CO uptakes (self-diffusivities) of CoRE MOFs and hMOFs range from 0.02 to 2.28 mol/kg (1.1×10-6 to 2.5×10-3 cm2/s) and 0.45 to 3.06 mol/kg (2.6×10-7 to 3.6×10-3 cm2/s), respectively, at 1 bar and 298 K. At low pressures (0.1-1 bar), Henry's constant for CO (KH,CO) is the primary determinant of performance, whereas at higher pressure (10 bar), structural factors like surface area (Sacc) and porosity (ϕ) become more significant. For CO diffusivity, results reveal that heat of adsorption (Q0st,CO) is the most important feature. Our analysis identified the top-performing adsorbents, revealing that MOFs with the highest CO uptakes typically feature narrow pores (4.5-7.2 Å), aromatic rings, carboxylic acids, halogens, and rare metals such as Li. Contrary to the uptake results, MOFs with high PLD values (>12 Å) and larger pore volume (>1.2 cm3/g) tend to have higher CO diffusivity values, DCO >10-4 cm2/s. Our results demonstrate that the utilization of machine learning algorithms and HTCS methodology with molecular simulation techniques exhibit reliable and rapid guidance to feature experimental and computational studies.