Do New MOFs Perform Better for CO₂ Capture and H₂ Purification? Computational Screening of the Updated MOF Database

Abstract

High-throughput computational screening of metal organic frameworks (MOFs) enables the discovery of new promising materials for CO₂ capture and H₂ purification. The number of synthesized MOFs is increasing very rapidly, and computation-ready, experimental MOF databases are being updated. Screening the most recent MOF database is essential to identify the best performing materials among several thousands. In this work, we performed molecular simulations of the most recent MOF database and described both the adsorbent and membrane-based separation performances of 10 221 MOFs for CO₂ capture and H₂ purification. The best materials identified for pressure swing adsorption, vacuum swing adsorption, and temperature swing adsorption processes outperformed commercial zeolites and previously studied MOFs in terms of CO₂ selectivity and adsorbent performance score. We then discussed the applicability of Ideal Adsorbed Solution Theory (IAST), effects of inaccessible local pores and catenation in the frameworks and the presence of impurities in CO₂/H₂ mixture on the adsorbent performance metrics of MOFs. Very large numbers of MOF membranes were found to outperform traditional polymer and porous membranes in terms of H₂ permeability. Our results show that MOFs that are recently added into the updated MOF database have higher CO₂/H₂ separation potentials than the previously reported MOFs. MOFs with small pores were identified as potential adsorbents for selective capture of CO₂ from H₂, whereas MOFs with high porosities were the promising membranes for selective separation of H₂ from CO₂. This study reveals the importance of enriching the number of MOFs in high-throughput computational screening studies for the discovery of new promising materials for CO₂/H₂ separation.

1. Introduction

Carbon dioxide (CO₂) has to be efficiently captured from CO₂/H₂ mixtures in a precombustion plant to reduce CO₂ emissions and to obtain purified H₂.¹ Precombustion gas capture in an integrated gasification combined cycle (IGCC) is particularly efficient due to (i) availability of concentrated CO₂ (15–40 wt %) at high pressures (up to 40 bar),^2,3 (ii) high power generation diversity through different ranks of coal,⁴ and (iii) the reduced complexity of IGCC plant design compared to pulverized coal plants⁵ which enables modular integration of new separation and purification units. Adsorption-based CO₂ separation offers low energy consumption by the appropriate selection of adsorbent materials. For example, zeolites and activated carbons⁶ are widely studied to sequester CO₂ from precombustion plants to replace the energy-intensive amine scrubbing process,⁷ but these porous materials generally suffer from low to mediocre CO₂ working capacities. Membrane technology can be also used as an alternative method for CO₂ capture. Among different types of membranes, polymers and zeolites are widely studied. Polymers either have low gas permeability or low selectivity in addition to the plasticization issue.^8,9 Although zeolites are selective membranes with high thermal stabilities,¹⁰ they suffer from tunability and limited material diversity due to availability of 232 framework types.¹¹ Therefore, development of highly selective and efficient adsorbent and membrane materials for CO₂ capture and H₂ purification is still a challenge.

Metal organic frameworks (MOFs), porous materials formed by combination of metal nodes and organic moieties, offer a rich variety of morphologies with a wide range of porosities.¹² MOFs can be promising materials as adsorbents and membranes with their high surface areas (up to 10 500 m²/g),¹³ tailorable topologies,¹⁴ tunable functional groups,¹⁵ and good mechanical and thermal stabilities.¹⁶ Several reviews focused on MOFs’ potential as adsorbents and membranes.¹⁷⁻²⁰ There is a need to identify the best performing MOFs within several thousands of synthesized materials, but conducting gas separation experiments for each MOF structure is not practical due to the very high number of existing materials. High-throughput computational screening studies that can accurately assess gas separation performances of MOFs have recently gained importance. Great majority of these studies focused on postcombustion CO₂ capture with MOFs, mainly separation of CO₂/CH₄^21,22 and CO₂/N₂,^23,24 whereas the number of studies on precombustion CO₂ capture, specifically separation of CO₂/H₂ mixture, is very limited as discussed below.

Adsorbent and membrane performances of 41 825 hypothetical MOFs (hMOFs) generated based on eight common MOF structures were examined for CO₂/H₂ separation using molecular simulations.²⁵ Functionalized Mg-MOF-74 derivative achieved the highest CO₂/H₂ selectivity (72) at 1 bar, 298 K. A larger number of hMOFs, 51 163, were screened and ranked based on adsorption selectivity and working capacity for CO₂/H₂ gas mixture.²⁶ One of the top performing hMOFs having a high CO₂ working capacity (3.80 mol/kg) and CO₂/H₂ selectivity (60) at 20 bar, 313 K was also experimentally synthesized.²⁶ Grand canonical Monte Carlo (GCMC) simulations were performed for CO₂/H₂:40/60 mixture in 358 400 hMOFs and CO₂/H₂ selectivities up to 300 and CO₂ working capacities up to 35 mol/kg were reported at 40 bar, 313 K.²⁷ These studies focused on hypothetical MOF structures, however performing molecular simulations on experimentally synthesized, readily available MOF structures is also important since there are already >88 000 existing MOFs.²⁸ 531 real MOFs were screened and the top MOF was identified to have CO₂ working capacity of 6.0 mol/kg and CO₂/H₂ selectivity of 83 for CO₂/H₂ separation at 20 bar, 313 K.²⁶ Our group²⁹ recently screened 3857 experimentally synthesized MOFs obtained from the Cambridge Structural Database (CSD) and identified the top MOF adsorbents and membranes outperforming zeolites and polymers due to their high CO₂/H₂ selectivities. Many more MOFs have been deposited into the CSD since then and computation-ready MOF database has been updated. There can be new MOFs having higher potential for CO₂ capture and H₂ purification than the previously studied structures. Therefore, computational screening of the most up to date MOF database is essential to uncover the potential of newly reported MOFs and to renew the upper limits of MOFs in adsorption-based CO₂/H₂ and membrane-based H₂/CO₂ separations.

In this work, the breadth and depth of our previous work²⁹ were significantly extended on CO₂ capture and H₂ purification performance assessment of MOFs by almost tripling the number of studied materials from 3857 to 10 221.²⁹ GCMC simulations were conducted to compute CO₂/H₂:15/85 mixture adsorption in all MOFs for three different adsorption-based processes, pressure swing adsorption (PSA), vacuum swing adsorption (VSA), and temperature swing adsorption (TSA). Materials performance evaluation metrics including working capacity, selectivity, adsorbent performance score, regenerability, and separation potential were calculated for each MOF and the best MOFs were selected based on the combination of these metrics. The top performing MOFs were benchmarked against conventional adsorbents such as zeolites. We then focused on four important aspects in the field of gas adsorption/separation in MOFs: (i) the applicability of Ideal Adsorbed Solution Theory (IAST) for predicting CO₂/H₂ mixture adsorption for the top performing MOFs, (ii) the impact of inaccessible local pores of MOFs on CO₂ and H₂ adsorption, (iii) the effect of catenation in MOFs on the calculated adsorbent performance metrics, and (iv) the influence of presence of impurities in CO₂/H₂ mixture on the predicted separation performances of the top MOF adsorbents. These analyses provided molecular-level insights into the ideal gas adsorption phenomena, the consequences of catenated structures and inaccessible local pores of MOFs on simulated gas separation performances of MOFs. We then examined the potential of MOF membranes by performing computationally demanding molecular dynamics (MD) simulations to describe permeability and selectivity of MOF membranes and compared them with polymers. Studying a large variety and number of MOFs gave us the opportunity to investigate the structure-performance relations of MOFs in adsorption-based CO₂/H₂ and membrane-based H₂/CO₂ separation. Our results will direct the selection of not only the best MOF adsorbents for CO₂ capture using PSA, VSA, and TSA processes but also the top MOF membranes for H₂/CO₂ separation.

2. Computational Methods

2.1. MOF Database and Calculation of Structural Properties

We used 70 551 nondisordered MOF structures from the most recent MOF subset available in the CSD (version 5.40).²⁸ We cleaned the solvents of structures using the Python code available with the CSD MOF subset.³⁰ Structural properties of MOFs, accessible surface area (SA), nonaccessible surface area (NaSA), the largest cavity diameter (LCD), pore limiting diameter (PLD), porosity (ϕ), dimensionality (1–2–3D) were calculated using the Zeo++ software version 0.3.0.³¹ To calculate the accessible surface area, we used a N₂-sized (3.70 Å) probe whereas He-sized (2.40 Å) probe was used for the porosity calculations. For MOFs having PLDs between 3.30 and 3.70 Å, which are inaccessible by a N₂-sized probe, a CO₂-sized (3.30 Å) probe was used to compute SA. MOFs exhibiting zero SA and a small number of materials (58) for which the partial charge assignment did not convergence were omitted. As a result, molecular simulations were performed for 10 221 different MOFs. Refcodes of 10 221 MOFs and all details of GCMC and MD simulations with potential parameters were given in Tables S1 and S2 of the Supporting Information (SI).

2.2. Ideal Adsorbed Solution Theory (IAST)

IAST is commonly used to predict mixture adsorption data from the single-component gas adsorption data.³² We tested the reliability of using IAST on the selected top MOFs at PSA, VSA, and TSA conditions. Single-gas adsorption isotherms of CO₂ and H₂ were parametrized with dual-site Langmuir (DSLangmuir) models by considering the gas adsorption amount at the Henry’s law regime as suggested in the literature.³² Details of applying IAST together with fitting parameters, equations, and comparison of the Root Mean Square Errors (RMSE) for the Langmuir and DSLangmuir isotherms of CO₂ and H₂ were provided in Tables S3–S5 of the SI. Adsorption isotherms of four MOFs having the lowest and highest R² values for the fits were also provided in Figure S1. Gas uptakes and selectivities computed from IAST were then compared with the corresponding values obtained from GCMC simulations of CO₂/H₂:15/85 mixture at 0.1, 1, 2, 4, 6, 8, and 10 bar, 298 K.

2.3. Methodology to Block Inaccessible Local Pores in GCMC Simulations

Blocking inaccessible pores of porous materials in molecular simulations can alter gas adsorption when local pore size entrance is smaller than the adsorbate size. Inaccessible pores of zeolites were blocked to avoid erroneous gas adsorption in these regions.^33,34 It was recently shown that MOFs may have at least 1.5-times higher CH₄ adsorption at infinite dilution when their inaccessible pores were blocked compared to their unblocked counterparts.³⁵ For MOFs having NaSA, including blocking pockets in molecular simulations is important since they may change the calculated adsorbent performance metrics. To identify the MOFs with local inaccessible pores, we adopted a geometrical approach and first calculated NaSA values using Zeo++.³¹ The probe radius of H₂ (1.48 Å) and CO₂ (1.65 Å) were used for assigning blocking spheres. All MOFs were prescreened and 386 MOFs with NaSA > 100 m²/g were shortlisted. 170 out these MOFs were computed to have NaSA > 100 m²/g when N₂-sized probe was used but they had zero NaSA when CO₂ or H₂ were used as probes. For the remaining 216 MOFs, SA (NaSA) was calculated as 82.8–2486 m²/g (100.6–915.7 m²/g). 82 out of 216 MOFs were computed to have nonzero NaSA with only CO₂-sized probe, and 134 out of 216 MOFs were calculated to have nonzero NaSA with both H₂ and CO₂-sized probes.

2.4. Calculation of Adsorbent and Membrane Performance Metrics

To simulate PSA (VSA) processes, adsorption and desorption conditions were considered as 10 (1) and 1 (0.1) bar, respectively, at 298 K. For the TSA process, adsorption and desorption temperatures were considered as 298 and 473 K, respectively, at 1 bar. We set the temperatures and pressures for all processes based on the industrial operation conditions reported in the literature.³ We examined adsorption-based separation performances of MOFs at PSA, VSA, and TSA conditions using five different metrics, selectivity, working capacity, regenerability, adsorbent performance score, and separation potential. Calculations of all these metrics were demonstrated in Table S6. Briefly, adsorption selectivity reflects the affinity of a material toward the more strongly adsorbed gas. Working capacity is the extent of CO₂ that can be captured through each adsorbent regeneration cycle. Adsorbent performance score (APS) merges adsorption selectivity and working capacity to determine the best performing MOFs. Regenerability (R%) is used to determine how much CO₂ can be regenerated at each adsorbent regeneration cycle. Separation potential can be used to calculate the achievable gas uptakes in a simple recovery operation of a fixed bed adsorber. The uptakes of pure CO₂ and H₂ in their mixture obtained from this metric share the same assessment about the performance of MOFs with the breakthrough simulations. Gas permeability and selectivity were computed to reveal the membrane-based gas separation performances of MOFs. We noted that high gas permeability and high gas selectivity are desired to decrease the operation costs and achieve high product purity. Gas permeability was calculated as the product of the Henry’s constant and self-diffusivity computed at infinite dilution. Mixture gas permeability was calculated by multiplying the adsorbed gas loading and self-diffusivity divided by feed partial pressure of the gas at 1 and 10 bar, 298 K. Membrane selectivity was estimated by taking the ratio of gas permeabilities.

3. Results and Discussion

3.1. MOF Adsorbents

Selectivities and CO₂ working capacities of MOFs for PSA, VSA, and TSA processes were demonstrated in Figure 1. CO₂/H₂ selectivities of MOFs were computed as 4.18–4.24 × 10⁴ for PSA (Figure 1(a)), 1.20–1.19 × 10⁵ for VSA (Figure 1(b)) and TSA processes (Figure 1(c)). CO₂ working capacities were calculated as 0.02–11.66 mol/kg, 1.62 × 10^–3-7.05 mol/kg, 1.30 × 10^–4-16.11 mol/kg at PSA, VSA, and TSA conditions, respectively. In our previous work,²⁹ we showed that MOFs having high CO₂ selectivities (>5000) suffer from low CO₂ working capacities (<2 mol/kg) at PSA and VSA conditions. In this work, we significantly increased the number of MOFs studied from 3857 to 10 221 and showed that there are new MOFs offering both high selectivities (>5000) and high CO₂ working capacities (>2 mol/kg) at PSA and VSA conditions. This is an important result showing that enriching the number and variety of materials studied in a computational high-throughput screening study can lead to identification of new promising adsorbent materials. Most of the new MOFs that we studied (orange dots in Figure 1) have higher CO₂ selectivity (>10⁴) and higher working capacity (2–4 mol/kg) for PSA and VSA processes compared to the MOFs distinguished in our previous work²⁹ (black dots in Figure 1). As the selectivity increases, working capacity generally decreases for PSA and VSA processes. Conversely, MOFs can have both high CO₂ selectivity and high working capacity at TSA condition since most of the adsorbed CO₂ can easily desorb at high temperatures due to the weak intermolecular interactions. Overall, the upper limits of CO₂ selectivity and working capacity of MOFs increased by up to 15-times and 1.3-times, respectively, compared to the limits defined in our previous study.²⁹ Therefore, one can conclude that the MOFs that are reported in the updated CSD MOF database have better CO₂ selectivities.

Figure 1.

Adsorption selectivity, working capacity, and APS of MOFs for separation of CO₂/H₂:15/85 mixture at (a) PSA, (b) VSA, and (c) TSA conditions. Orange points represent the data of new 6364 MOFs studied at PSA and VSA conditions and 10 221 MOFs at TSA conditions, whereas black points represent data of 3857 MOFs from our previous study.²⁹

APS has been used to identify materials having both high selectivity and high working capacity for various gas separations such as CO₂/H₂,²⁶ CO₂/N₂,²² and CH₄/H₂.³⁶ We computed APSs of MOFs for each process and provided arbitrary limits as shown by blue and red curves in Figure 1 to illustrate the promising materials. MOFs having mediocre CO₂/H₂ selectivity (100–1000) and CO₂ working capacity (0.4–4 mol/kg) have APSs between 400 and 2000 mol/kg. 765, 1181, and 1443 MOFs out of 10 221 were identified as promising since they exceed the red curve at PSA, VSA, and TSA conditions, respectively. Most of the MOFs identified in this work outperformed zeolites in metrics such as CO₂ selectivity, working capacity, and APS for CO₂/H₂ separation. For example, 1762 MOFs provided higher CO₂/H₂ selectivity (550–4.09 × 10⁵) at 10 bar, 298 K than NaX (1510) and NaY (550) at 10 bar, 300 K.³⁷ Similarly, 5365 and 688 MOFs exhibited higher CO₂ working capacities (1.95–11.66 mol/kg at PSA and 2.03–7.05 mol/kg at VSA conditions), outperforming 10-membered ring zeolites (1.95 mol/kg at PSA and 2.03 mol/kg at VSA).³⁸ Many MOFs having APS > 2000 mol/kg outperformed NaX (29.60 mol/kg) and NaY (27.10 mol/kg) at PSA condition.³⁷

R% is an important metric for adsorption-based gas separations to decrease the operating costs for the replacement of adsorbents. Identifying selective adsorbents with R% > 85% is critical for an efficient and economic separation. We examined relations between APS and R% of MOFs in Figure 2. Results showed that APSs of the regenerable MOFs (R% > 85%) for PSA and VSA processes as shown in Figure 2(a and b), respectively, vary significantly. For example, a MOF with high APS may have low R%. Compared to PSA and VSA, there are a remarkable number of MOFs (2682) having both APS > 400 mol/kg and R% > 85% at TSA process as shown in Figure 2(c). This can be attributed to dramatic decreases in CO₂ adsorption of MOFs at 473 K due to the weak interactions between CO₂ and MOFs at high temperatures, which results in easy regeneration. ΔQ was calculated using volumetric gas uptakes of MOFs, different than APS which considers the gravimetric gas uptakes of MOFs. Thus, MOFs with high density will probably have higher ΔQ values. Figure S2 shows the relations between CO₂/H₂ selectivity and volumetric CO₂ working capacity of MOFs, color-coded with ΔQ values. We note that ΔQ values of MOFs for VSA do not differ from those obtained for TSA process due to the same adsorption conditions. Results showed that MOFs have generally higher ΔQ at PSA conditions due to their higher volumetric CO₂ uptakes at 10 bar compared to those for VSA and TSA processes at 1 bar. Krishna et al.³⁹ reported that zeolite 13X and MgMOF-74 exhibit high ΔQ of 36 and 52 mol/L, respectively, for CO₂/H₂:20/80 mixture at adsorption (desorption) pressure of 54 (14) bar. Our results suggest that at much lower pressures, many MOFs offer higher ΔQ than zeolite 13X and MgMOF-74, indicating a higher amount of H₂ recovery in a fixed bed unit.

Figure 2.

R% and APS of MOFs computed for separation of CO₂/H₂:15/85 mixture at (a) PSA, (b) VSA, and (c) TSA conditions. Orange points represent the data of new 6364 MOFs at PSA and VSA conditions and 10 221 MOFs at TSA conditions, whereas black points represent data of 3857 MOFs from our previous study.²⁹

We distinguished the best 10 MOFs exhibiting the highest APSs among the MOFs having R% > 85% and listed them in Table S7. We would like to note that we used a generic Universal Force Field (UFF)⁴⁰ to identify the top MOFs. Several studies examined the effect of the force field selection on the results of high-throughput computational screening of MOFs for CO₂ separations and showed that ranking of MOFs based on different force fields is similar, suggesting that UFF can be used to identify the promising candidates for adsorption-based CO₂ separations.^41,42 As shown in Table S7, DABWUA having a CO₂/H₂ selectivity of 333, was the best MOF at PSA condition due to its narrow pores providing strong CO₂–MOF interactions.²⁹ FAQXIF was the best MOF for VSA process with a very high selectivity of 3703. We note that four of the top 10 MOFs at VSA condition, AFEJOK, DATKIU, GUKYUI, and OJASOJ, have extra framework ions which enhance the electrostatic interactions between CO₂ molecules and adsorbents, resulting in high selectivities. To understand the impact of extra framework ions, we turned-off the Coulombic interactions between CO₂ and MOFs in molecular simulations and recalculated selectivities. Selectivities of AFEJOK, DATKIU, GUKYUI, and OJASOJ significantly decreased (from 1262 to 165; 1474 to 40; 1809 to 118; and 1949 to 52, respectively) once the electrostatic interactions were turned-off. JOVLAH10 was determined as the best adsorbent for TSA process with a very high selectivity. The top MOFs identified at TSA condition have exceptionally high selectivities (1.21 × 10⁴ to 9.91 × 10⁴) and APSs (1.01 × 10⁵ to 3.07 × 10⁵ mol/kg) and all these MOFs have very narrow PLDs < 4.91 Å, exposed metal sites, and/or halogen functional groups which enhance CO₂–MOF electrostatic interactions. Figures S3 and S4 show the snapshots of the top three MOFs obtained from GCMC simulations at PSA, VSA, and TSA conditions. As shown in Figure S3, exposed metal sites of DABWUA and FAQXIF and narrow pores of DABWUA and JOVLAH10 provided favorable adsorption sites for CO₂. Consequently, adsorption of H₂ was hindered due to the strong preference of CO₂ at adsorption and desorption conditions as shown in Figure S4. Overall, we conclude that TSA is the best process option for CO₂/H₂ separation when APS is considered as the main performance metric for MOFs having R% > 85%.

3.1.1. Applicability of IAST

IAST has been widely used to predict the multicomponent equilibria of gas mixtures in various MOFs such as CO₂/CH₄ in IRMOF-1,⁴³ CO₂/N₂ in M-MOF-74,⁴⁴ CH₄/H₂ in Zn(bdc)(ted)_0.5,⁴⁵ and CO₂/H₂ in Cu-BTC.⁴⁶ However, it was shown to fail when (i) there is a strong difference between the adsorption strength of gases such as CO₂ and H₂ where strong electrostatic interactions favor CO₂ adsorption, (ii) the gas species prefer to adsorb on different adsorption sites in a MOF, and (iii) the accessible surface area for gas species is very different.^32,47 We tested IAST for the best performing three MOFs for PSA, VSA, and TSA processes. Figure 3 compares GCMC and IAST results for (a) CO₂ uptake, (b) H₂ uptake, and (c) CO₂/H₂ selectivity computed at 0.1, 1, 2, 4, 6, 8, and 10 bar. Adsorbed CO₂ and H₂ amounts in QATPUX01 and GEJRIW at these pressures obtained from GCMC and IAST were found to be within 10% difference. Therefore, a good agreement was observed between selectivities predicted by IAST and GCMC for these MOFs. The highest deviation between IAST-predicted and simulated selectivities was found for LACDEY which was attributed to the significant difference in H₂ uptakes obtained from IAST and GCMC. LACDEY has a small surface area (300 m²/g) with narrow pore apertures (4.40 × 3.85 Ų) and open metal sites, resulting in a structural heterogeneity in terms of local adsorption strength.³² As a result of the large differences in the loadings between CO₂ and H₂ molecules, CO₂/H₂ selectivities computed from GCMC simulations were much higher than those obtained from IAST. Similarly, selectivities of DABWUA, DATKIU, EKEJOT, and TENLAC were underestimated by IAST, which can be explained by the heterogeneous adsorption sites available in these MOFs. For example, TENLAC has Cl-functionalization that acts as strong adsorption sites for CO₂. Similarly, DATKIU has free tetrafluoroborate (BF₄^–) ions which strongly enhances CO₂ adsorption as we discussed above. DABWUA and EKEJOT have open metal sites, indicating the energetic heterogeneity for their adsorption surface. Due to the presence of heterogeneous adsorption sites in these four MOFs, CO₂ adsorption was underpredicted, whereas H₂ adsorption was overpredicted by IAST, resulting in underestimated CO₂/H₂ selectivity compared to GCMC. Although JOVLAH10 has open metal sites, its selectivity was overpredicted by IAST, which may be attributed to the difference in the accessible surface areas calculated by a CO₂-sized probe of 527.70 m²/g and a H₂-sized probe of 810.46 m²/g. CO₂ adsorption was overpredicted, whereas H₂ adsorption was underpredicted by IAST, causing higher CO₂/H₂ adsorption selectivities compared to those obtained from GCMC simulations. These results showed that the top MOFs identified in this work commonly have heterogeneous adsorption sites and therefore, their CO₂/H₂ selectivities were mostly underestimated by IAST. This highlights the importance of computing selectivities by performing mixture adsorption simulations because if IAST was used to calculate selectivities, the top promising MOFs would be overlooked.

Figure 3.

Comparison between GCMC simulation results and IAST predictions for (a) CO₂ adsorption, (b) H₂ adsorption, and (c) CO₂/H₂ selectivity at 0.1, 1, 2, 4, 6, 8, 10 bar, and 298 K. Black dashed lines are to guide the eye, orange dashed lines represent 10% deviation from parity line.

3.1.2. Blocking Inaccessible Pores of MOFs

To understand the impact of accessibility of pores on the calculated selectivities and working capacities of MOFs, we isolated the inaccessible pores considering two cases: (i) the pores that are only inaccessible to CO₂ were blocked; and (ii) the pores that are inaccessible to both CO₂ and H₂ were blocked. Figure 4 shows selectivities and working capacities of 216 MOFs having NaSAs obtained from the simulations at (i) and (ii) conditions for PSA, VSA, and TSA processes. Blocking inaccessible pores resulted in much lower CO₂/H₂ selectivities and CO₂ working capacities compared to those obtained in the presence of inaccessible pores. Similar results were also discussed by Zhang et al.⁴⁸ where the inaccessible cages were found to provide a strong surface potential overlap between gas molecules and the framework. Since CO₂ molecule has a quadrupole moment, it provides stronger LJ interactions with the framework than H₂. Therefore, when the inaccessible pores of MOFs were blocked, the decrease in CO₂ uptake was found to be more pronounced compared to H₂, resulting in lower CO₂/H₂ selectivities and CO₂ working capacities. The influence of blocking was found to be more obvious for MOFs with low SA and large NaSA. For example, the largest difference between the selectivities obtained from simulations at (i) and (ii) conditions was observed for SAZPIV. Its selectivity significantly decreased from 434 to 21.5 at PSA, 1043 to 25 at VSA, and 1483 to 2.8 at TSA conditions when inaccessible pores were blocked for CO₂. This decrease was attributed to its low accessible SA (428.30 m²/g) and quite large NaSA (152.72 m²/g) which significantly contributed to CO₂ adsorption. The largest difference between the CO₂ working capacities obtained from simulations at (i) and (ii) conditions was observed for FECPIM for all processes. In this MOF, coordinated Na ions blocked the entrance of both CO₂ and H₂ to a local pore, resulting in a large NaSA of 553.44 m²/g. When the inaccessible pores of FECPIM were blocked, its working capacity decreased. Figure S5 shows how R% values vary in MOFs at (i) and (ii) conditions for PSA, VSA, and TSA applications. Blocking the inaccessible pores of MOFs commonly resulted in much higher R% values. Since the decrease in CO₂ uptakes was more notable than the decrease in CO₂ working capacities, R% values of MOFs increased when the inaccessible pores were blocked. Finally, we note that none of 216 MOFs that we studied in this section were identified as the top performers for CO₂/H₂ separation due to their low APSs. We computed selectivities, working capacities and APSs of these materials for PSA, VSA, and TSA processes by performing simulations with blocked pores and showed the results in Figure S6. When the isolated pores were blocked, selectivities, working capacities, and APSs decreased in most of the MOFs.

Figure 4.

Effect of blocking pockets of MOFs on (a, b) CO₂/H₂ selectivities, (c–e) CO₂ working capacities at PSA, VSA, and TSA conditions.

3.1.3. Catenated MOFs

It has been shown that catenated IRMOFs outperform their noncatenated counterparts in CO₂/H₂ selectivity due to strong confinement of CO₂ molecules in the catenated MOFs.⁴⁹ We identified 29 catenated MOFs and their noncatenated pairs in our data set and compared their selectivities, working capacities, and APSs at PSA, VSA, and TSA conditions in Figure 5. We note that among these 29 catenated MOFs, three MOFs, PCN6′, MOF14, and PCNHTB have their catenated pairs in our data set. We obtained the crystal structures of catenated counterparts of the remaining 26 MOFs from the hypothetical data set reported by Sezginel et al.⁵⁰ As shown in Figure 5, catenated MOFs have generally higher CO₂/H₂ selectivities due to the enhancement in MOF–CO₂ interactions triggered by the strong confinement of CO₂ within a much lower adsorption surface. Similar results were also discussed in a computational study⁵¹ where six catenated MOFs exhibited higher CO₂/H₂ selectivity compared to their noncatenated pairs up to 20 bar at 298 K, which was attributed to the creation of new adsorption sites for CO₂. Although catenation reduces pore sizes and surface areas of MOFs, CO₂ working capacities of catenated MOFs were found to be higher compared to their noncatenated counterparts. R% values of catenated MOFs and their noncatenated pairs did not significantly change. We note that noncatenated structures of these 29 MOFs have APS < 400 mol/kg, thus they are not identified as the top performing MOFs. If we used hypothetical catenated structures of these 26 MOFs in simulations to assess their CO₂/H₂ separation performance, then some of these MOFs may be promising adsorbents. For example, 11 hypothetical catenated MOFs exhibited APS > 400 mol/kg at PSA condition, whereas only one hypothetical catenated MOF provided APS > 2000 mol/kg at VSA and TSA conditions. The effect of catenation on the performance of experimentally synthesized MOFs such as MOF14 and PCN6′ was found to be insignificant since their APSs were low after catenation. However, catenated hypothetical structures exhibit higher selectivities, working capacities, and APSs compared to their noncatenated counter parts as shown in Figure S7. The enhancement in APSs of these MOFs was attributed to the decrease in their PLDs which improved CO₂–framework interactions. These results suggest that catenation can be a promising design strategy to improve APSs of MOFs without changing their R%.

Figure 5.

Comparison of adsorption selectivities, regenerabilities and working capacities of catenated MOFs with those of their noncatenated pairs at (a) PSA, (b) VSA, and (c) TSA conditions.

3.1.4. Purity of CO₂/H₂ Mixture

CO₂ capture and H₂ purification occur within the presence of impurities such as CO, CH₄, and H₂O in the precombustion process plants.⁵² Therefore, we studied the effect of impurities in the bulk gas mixture on the performances of the top 20 MOFs at PSA and VSA conditions. GCMC simulations were conducted for binary (CO₂/H₂:15/85) and quinary (CO₂/H₂/CH₄/CO/H₂O: 15/75/5/5/0.1) mixture to compute APS and R% and results are given in Table S8. To examine the water affinity of MOFs, we also computed the Henry’s constants for CO₂ (K_CO2⁰) and H₂O (K_H2O). Results showed that APSs of the top performing MOFs significantly decrease in the presence of impurities in the bulk gas mixture. For example, APS of the top performing MOF identified for PSA process, DABWUA, decreased from 2463 to 1065 mol/kg. CO₂ uptakes of MOFs decreased at PSA and VSA conditions due to high number of adsorbed H₂O molecules within the pores of MOFs in the quinary mixture, resulting in lower selectivity and working capacity compared to those obtained from the binary mixture simulations. If K_H2O⁰/K_CO2 was computed to be much larger than 1, then the MOF has a strong affinity toward H₂O, indicating remarkable changes in the calculated adsorbent metrics. For example, APS of the ninth best performing MOF, MEKKUL, at PSA condition significantly decreased from 1831 to 337 mol/kg due to its strong affinity toward H₂O with K_H2O⁰/K_CO2 value of 836. The strong affinity of MEKKUL toward H₂O can be attributed to its open Co metal centers. Similar results were also discussed in the literature⁵³ where presence of H₂O as the third component in CO₂/CH₄ and CO₂/N₂ mixtures was shown to remarkably decrease the performance metrics of MOFs having strong affinity toward water due to presence of specific functional groups. Other impurities such as CO and CH₄ were found to have less impact on the calculated metrics compared to H₂O. We generally did not observe significant changes in R% of the top MOFs in the presence of impurities except for the MOFs having high K_H2O⁰/K_CO2 ratios. Thus, we infer that calculated APSs of hydrophilic MOFs generally decrease in the presence of impurities in CO₂/H₂ mixture.

3.2. MOF Membranes

We so far focused on adsorption-based CO₂/H₂ separation and now we focus on membrane-based H₂/CO₂ separation. In Figure 6(a), adsorption, diffusion, and membrane selectivities of MOFs calculated at infinite dilution were compared where orange, green, and blue points represent MOFs having low, high, and very high diffusion selectivities toward H₂, respectively. In terms of diffusion, MOFs are H₂ selective since H₂ is a smaller and lighter (2.96 Å) molecule than CO₂ (3.30 Å). H₂ selective MOF membranes with high H₂ permeabilities are desired for efficient H₂ separation. 452 MOFs were calculated as H₂ selective membranes (1 < S_mem,H2/CO2⁰ < 6.34) in this study. Figure 6(b) shows H₂/CO₂ selectivities and H₂ permeabilities of 10 221 MOF membranes with the Robeson’s upper bound,⁵⁴ designed for polymer membranes to demonstrate the trade-off between their permeability and selectivity. This bound represents the intrinsic permeability and selectivity of the dense polymer membranes.⁵⁵ For this reason, membrane community has aimed to enhance gas diffusivity by increasing the gas solubility of polymers to obtain high-performance polymer membranes for H₂/CO₂ separation.⁵⁶ However, industrially available polymer membranes are still below the upper bound. However, we found that MOFs, having a wide range of structural and textural properties, exhibit extremely high H₂ permeabilities (>10⁵ Barrer) to overcome that trade-off. In Figure 6(b), S_mem,H2/CO2 of MOFs was found to be between 2.34 × 10^–4 and 6.34, whereas P_H2⁰ was calculated as 1.08 × 10³ to 4.38 × 10⁶ Barrers. 4234 MOFs exhibited high H₂ permeabilities (P_H2 > 10⁵ Barrer) and these materials can decrease the size and cost of membranes for H₂/CO₂ separation. Finally, we note that the number of MOFs (1854) surpassing the upper bound significantly increased compared to our previous work (899),²⁹ indicating that new MOF structures in the updated database have a strong potential as membranes for H₂/CO₂ separation.

Figure 6.

(a) Adsorption, diffusion, and membrane selectivities of 10 221 MOFs computed at infinite dilution. (b) H₂/CO₂ membrane selectivities and H₂ permeabilities of 10 221 MOF membranes. Black solid line represents the Robeson’s upper bound for H₂/CO₂ separation, whereas the black dashed line indicates S_mem,H2/CO2⁰ = 1.

We identified the top 10 promising MOF membranes having S_mem,H2/CO2⁰ > 3.80 and P_H2 > 10⁵ Barrer and listed their separation performances along with structural properties in Table S9. The best performing MOF membranes generally exhibited CO₂/H₂ adsorption selectivities <10 and H₂/CO₂ diffusion selectivities >20. Since the top MOF membranes have high porosities (ϕ > 0.80) and large pore sizes (LCD > 20 Å), H₂ diffused fast within the pores of these MOFs. New MOFs taken from the updated data set exhibit larger pore sizes (LCD > 20 Å) than the previously reported MOFs as shown in Figure S8. Therefore, new MOFs exhibited much higher diffusion selectivities toward H₂, leading to higher membrane selectivities toward H₂ compared to the previously reported MOFs. The number of H₂ selective MOF membranes (397) having S_mem,H2/CO2⁰ > 1 and P_H2 > 10⁵ Barrer is much higher than the previously identified MOF membranes (198) having S_mem,H2/CO2⁰ > 1 and P_H2 > 10⁵ Barrer, showing the potential of new MOFs as membranes for H₂/CO₂ separation.

Performing MD simulations for gas mixtures is computationally demanding if a very weakly adsorbed gas such as H₂ and a strongly adsorbed gas such as CO₂ are present within the pores of MOFs. To accurately compute mixture gas diffusivities at the desired condition, simulation cell size should be enlarged which significantly increases the computational time. For this reason, we carried out mixture MD simulations only for the best performing 10 MOFs and compared the gas permeabilities obtained from these simulations with those calculated at infinite dilution in Figure S9. CO₂ and H₂ permeabilities calculated at infinite dilution agreed reasonably with the permeabilities calculated for the binary gas mixture at both 1 and 10 bar, 298 K. H₂/CO₂ diffusion selectivities calculated at 1 bar were found to be generally higher than those computed at 10 bar. This was attributed to the higher CO₂ uptake at 10 bar, which hinders the transport of H₂, leading to a smaller diffusion coefficient for H₂ in comparison to that obtained at 1 bar. As a result, mixture selectivities of MOF membranes calculated at 1 bar were found to be generally similar to those calculated at infinite dilution whereas mixture selectivities calculated at 10 bar were generally lower than those calculated at infinite dilution as indicated in the insets of Figure S9(a,b). Thereby, infinite dilution simulations can give accurate predictions for H₂ permeability and H₂/CO₂ selectivity of the top performing MOF membranes. We also investigated the influence of framework flexibility on the best performing MOF membrane, FOTNIN, at 1 bar, 298 K for binary mixture separation using the snapshot method.⁵⁷ Adsorption and diffusion behavior and structural features for 10 flexible frameworks such as PLD, LCD, SA, and porosity were provided in Table S10. Binary mixture selectivity of flexible FOTNIN membrane was computed to range from 2.46 to 4.52 and its mixture H₂ permeability ranged from 9.84 × 10⁴-1.24 × 10⁵ Barrers. These results were similar to those calculated using rigid framework (S_mem,H2/CO2^mix: 4.17; P_H2: 1.58 × 10⁵ Barrer), indicating that rigid framework simulations can give accurate results for H₂ permeability and H₂/CO₂ selectivity of the top performing MOF membrane. We also compared separation performances of the top MOF membranes with zeolite and graphene oxide (GO) membranes. NaX⁵⁸ and LTA⁵⁹ were reported to have H₂ permeabilities of 195.41 and 125.37 Barrers and H₂/CO₂ selectivities of 3.6 and 7.6 at 1 bar, respectively. Most of the top MOF membranes identified in this work exhibited much higher H₂ permeability ranging between 4.87 × 10⁵ to 1.90 × 10⁶ (1.52 × 10⁵-1.58 × 10⁶) Barrers and almost similar H₂ selectivity between 3.52 and 5.26 (2.71–4.43) at 1 bar (10 bar). GO membrane⁶⁰ exhibited a high H₂ permeability of 5.07 × 10³ Barrer and a high H₂/CO₂ selectivity of 48 at 1 bar, 298 K. The best MOF membranes listed in Table S9 can still outperform GO membrane in terms of H₂ permeability. JUC-150⁶¹ and ZIF-8⁶² membranes were reported to exhibit H₂ permeabilities of 1.64 × 10⁴ and 4.48 × 10³ Barrers and H₂/CO₂ selectivities of 38.7 and 3.5 at 1 bar, 298 K, respectively. Due to very narrow pores of these MOF membranes, they exhibited 2 orders of magnitude lower H₂ permeabilities compared to the top MOF membranes we identified in Table S9. Overall, we can conclude that the top MOF membranes identified in this work can outperform traditional porous membranes in terms of H₂ permeability.

3.3. Structure-Performance Relations

We finally studied structure-performance relations of MOF adsorbents and membranes. Figure 7(a,b) shows the relations between porosity, LCD, and adsorption selectivity of MOFs having 1D, 2D, and 3D structures. Among 10 221 MOFs, 5046 MOFs were found to have either 1D or 2D channels, whereas 5175 MOFs exhibited 3D channels. As expected, MOFs with small LCDs (5–15 Å) generally exhibited high CO₂/H₂ adsorption selectivities (>1000), regardless of dimensionality due to the strong confinement of gas molecules within small pores. Figure 7(c,d) shows that the majority of the top MOF membranes have large LCDs (>20 Å), 3D channels, and high porosities (>0.80). Weakly adsorbed H₂ molecules in these MOFs diffused much faster than the strongly adsorbed CO₂ molecules, resulting in high diffusion selectivities toward H₂. Since MOFs with high porosities and LCDs generally exhibited very low adsorption selectivities <100, their diffusion selectivities toward H₂ governed their membrane selectivities. As a result, MOFs with large pores (>20 Å) and high porosities (>0.80) were identified to be promising for membrane-based H₂/CO₂ separation.

Figure 7.

Relations between porosities, pore sizes, and CO₂/H₂ adsorption selectivities of (a) 1D-2D MOFs and (b) 3D MOFs. Relations between porosities, pore sizes, and H₂/CO₂ membrane selectivities of (c) 1D-2D MOFs and (d) 3D MOFs. Size of the bubbles represents selectivities of MOFs. Red, yellow, blue, and black regions were scaled with a factor of 4.0 × 10^–2, 2.4 × 10^–2, 2.4 × 10^–3 and 2.4 × 10^–4, respectively in (a) and (b); scaling factor of 25 was used in (c) and (d) to have a clear representation.

We also investigated the influence of structural properties, PLD, LCD, SA, and porosity on the separation performance of the top 100 MOF adsorbents at PSA, VSA, and TSA conditions and the top 100 MOF membranes identified in our previous study²⁹ in Figure 8(a) and in our current study in Figure 8(b). When the number of MOFs increased from 3857 to 10 221, the distributions of most of the textural properties of MOFs exhibited similar behavior for PSA and VSA processes. The top MOFs at TSA condition showed similar PLDs and porosities compared to those at VSA, whereas the top MOFs at TSA had lower LCDs and porosities compared to MOFs at PSA condition. MOFs with small pore sizes (3.30–5.00 Å) and low porosities (0.30–0.60) are the best adsorbent candidates for CO₂/H₂ separation, whereas MOFs having high porosities (0.60–1.0) and large LCD (>15 Å) have high performance as membranes for H₂/CO₂ separation. We also note that MOFs with high PLDs (15–31 Å) are more suitable candidates as membranes in comparison to the previously studied MOFs.²⁹ These structure–performance relationships can be a useful guidance not only for the selection of MOFs but also for the future synthesis of new MOFs to achieve high-performance for CO₂ capture and H₂ purification.

Figure 8.

Effect of PLD, LCD, SA, and ϕ on the separation performance of the % of the top 100 MOF adsorbents at PSA, VSA, and TSA conditions and the % of the top 100 MOF membranes among (a) 3857 MOFs previously studied²⁹ and (b) 10 221 MOFs studied in this work.

4. Conclusions

In this work, we computationally assessed the performance of 10 221 MOF adsorbents and membranes for CO₂ capture and H₂ purification using GCMC and MD simulations. Many MOFs outperformed benchmark zeolites in performance metrics such as CO₂ selectivity, working capacity, APS, and ΔQ. CO₂/H₂ adsorption selectivities of the top performing MOFs were generally underestimated by IAST due to their heterogeneous adsorption sites, underlying the importance of performing mixture GCMC simulations. Blocking of the pores that are inaccessible to CO₂ and H₂ resulted in decreases in CO₂ working capacities and selectivities but increases in R% of MOFs. Results also showed that catenation can significantly improve APSs of MOFs without changing their R%, suggesting that catenation can be a useful design strategy for experiments to obtain high-performance MOF adsorbents for CO₂/H₂ separation. We found that APSs of the top MOFs generally decreased, whereas their R% remained same in the presence of impurities in CO₂/H₂ mixture. H₂/CO₂ selectivities and H₂ permeabilities of 10 221 MOF membranes were computed and a large number of MOF membranes surpassed the upper bound, outperforming polymer membranes. The upper limits of H₂ permeability and H₂/CO₂ selectivity of MOF membranes were computed as 4.38 × 10⁶ Barrer and 6.34, respectively. We finally investigated the influence of structural features of MOFs on their adsorbent and membrane performances. MOFs with small pores (3.30–5.00 Å) and low porosities (0.30–0.60) generally exhibited high adsorption selectivities for CO₂ (>1000), whereas MOFs with large pores (>20 Å) and high porosities (>0.80) exhibited high membrane selectivities for H₂ (3.80–6.34). We showed that the top MOFs identified for adsorption-based CO₂/H₂ separation have narrow pores, exposed metal sites and halogen functional groups which provide favorable adsorption sites for CO₂ molecules due to the strong electrostatic interactions between CO₂ and the MOFs’ atoms. However, the top MOF membranes identified for H₂/CO₂ separation have large pores enabling fast and efficient H₂ diffusion which is desired for efficient H₂ purification. We note that all the MOFs we studied including the top ones have been already experimentally synthesized. These results will be useful to guide the future experiments to high-performance MOFs both for adsorption-based CO₂/H₂ separation and membrane-based H₂/CO₂ separation.

We finally close by addressing the question raised in the title of this work: Our results showed that new MOFs recently added into the CSD MOF database have a great potential as adsorbents and membranes for CO₂ capture and H₂ purification. When we previously studied 3857 MOFs,²⁹ only 2% and 1% of these MOFs were found to offer APS > 2000 mol/kg at VSA and PSA conditions, respectively. In this work, we examined 10 221 MOFs and 12% of these MOFs at VSA and 8% at PSA condition were able to exceed the desired APS limit. Similarly, 17% of 3857 MOFs were previously calculated to have P_H2⁰ > 10⁵ Barrer²⁹ and 41% of 10 221 MOF membranes were found to have P_H2 > 10⁵ Barrer in this work. These results underline the importance of investigating the updated material databases for the discovery of new promising adsorbents and membranes.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsami.0c12330.

• Refcodes of 10 221 MOFs; details of molecular simulations; details of applying IAST together with fitting parameters, equations and comparison of the RMSEs for the Langmuir and DSLangmuir isotherms of CO₂ and H₂; performance evaluation metrics for MOFs; relation between adsorption selectivity, working capacity and separation potential; selectivities and working capacities of MOFs; snapshots of favorable adsorption sites of the top 3 MOFs; snapshots of GCMC simulations for the top 3 MOFs; effect of blocking pockets on regenerabilities; effect of blocking pockets on selectivities and working capacities of MOFs; effect of catenation on adsorption selectivity and working capacity of MOFs; separation performances of the top 10 MOFs for PSA, VSA, and TSA processes; separation performances of the top 10 MOFs in binary and quinary mixtures; separation performances of the top 10 H₂ selective MOF membranes; comparison of PLDs and LCDs of MOFs; comparison of gas permeabilities calculated at infinite dilution with the permeabilities calculated for binary mixture at 1 and 10 bar, effect of flexibility on the performance of the top MOF membrane (XLSX)

• MD movie of the top MOF membrane (MP4)

The authors declare no competing financial interest.