Cigdem Altintas‡
a,
Gokay Avci‡a,
Hilal Daglar‡a,
Ezgi Gulcay-Ozcan‡b,
Ilknur Erucar*c and
Seda Keskin*a
aDepartment of Chemical and Biological Engineering, Koc University, Rumelifeneri Yolu, Sariyer, Istanbul, 34450, Turkey. E-mail: skeskin@ku.edu.tr; Tel: +90 212 338 1362
bDepartment of Mechanical Engineering, Faculty of Engineering, Ozyegin University, Cekmekoy, Istanbul, 34794, Turkey
cDepartment of Natural and Mathematical Sciences, Faculty of Engineering, Ozyegin University, Cekmekoy, Istanbul, 34794, Turkey. E-mail: ilknur.erucar@ozyegin.edu.tr; Tel: +90 216 564 9297
First published on 15th March 2018
Design of new membranes having high H2/CH4 selectivity and high H2 permeability is strongly desired to reduce the energy demand for H2 production. Metal organic frameworks (MOFs) offer a great promise for membrane-based gas separations due to their tunable physical and chemical properties. We performed a high-throughput computational screening study to examine membrane-based H2/CH4 separation potentials of 4240 MOFs. Grand canonical Monte Carlo (GCMC) and molecular dynamics (MD) simulations were used to compute adsorption and diffusion of H2 and CH4 in MOFs. Simulation results were then used to predict adsorption selectivity, diffusion selectivity, gas permeability and membrane selectivity of MOFs. A large number of MOF membranes was found to outperform traditional polymer and zeolite membranes by exceeding the Robeson's upper bound for selective separation of H2 from CH4. Structure–performance analysis was carried out to understand the relations between MOF membranes' selectivities and their pore sizes, surface areas, porosities, densities, lattice systems, and metal types. Results showed that MOFs with pore limiting diameters between 3.8 and 6 Å, the largest cavity diameters between 6 and 12 Å, surface areas less than 1000 m2 g−1, porosities between 0.5 and 0.75, and densities between 1 and 1.5 g cm−3 are the most promising membranes leading to H2 selectivities >10 and H2 permeabilities >104 Barrer. Our results suggest that monoclinic MOFs having copper metals are the best membrane candidates for H2/CH4 separations. This study represents the first high-throughput computational screening of the most recent MOF database for membrane-based H2/CH4 separation and microscopic insight provided from molecular simulations will be highly useful for the future design of new MOFs having extraordinarily high H2 selectivities.
Continuous operation of membranes generates significant advantages compared to the more complex modes of operation needed for adsorption-based gas separation methods. Although a large number of experimental studies investigated adsorption-based gas separations with MOFs,11 the number of studies on membrane-based gas separations using MOFs is limited due to the experimental challenges in making defect-free membranes from new crystalline materials.12 We reviewed the literature on MOF membranes and showed that only 29 different types of MOF membranes were experimentally fabricated and tested for gas separations.13 These MOF membranes were reported to have very good gas separation performances, higher selectivities and higher permeabilities, compared to the traditional polymer and zeolite membranes.14 The number of synthesized MOF membranes has been increasing but most of the membranes were fabricated using the same materials, generally prototype MOFs, such as MOF-5,15–19 CuBTC,20–22 ZIF-8.23–25 Considering the fact that the number of available MOFs has already reached to several thousands,26 it is highly possible that there are many other materials with better gas separation performances but these MOFs have not been fabricated and tested as membranes yet. The large materials space creates an opportunity for the discovery of highly useful membrane candidates for target gas separations. On the other hand, it is not possible to fabricate and test thousands of MOF membranes using purely experimental manners at the lab scale.
Computational methods, specifically molecular simulations, play an important role in screening large number of MOFs for a target application in a time effective manner. Potential of molecular simulations in choosing the best MOFs for a given gas separation has been shown in several studies.27 These simulations also provide molecular-level insights that can guide the experiments for design of new materials with better separation performances. Although molecular simulations have been widely used to examine CO2/CH4 and CO2/N2 separation performances of MOFs,27–30 H2/CH4 separation is rarely investigated as we will discuss below. H2/CH4 separation is industrially important in the process of purification of synthesis gas obtained from steam reforming of natural gas. Hydrogen is an important chemical in many applications and its production depends on the decomposition of CH4. Therefore, separation of H2 from CH4 with high selectivity and high permeability is strongly desired. Several different types of membrane materials such as zeolites31 and polymers32 have been studied for membrane-based separation of H2/CH4 mixtures in the past. However, these membranes could not meet the requirements of high H2 selectivity and high H2 permeability.33
Due to the tunability of pore sizes and shapes by judicious selection of metal clusters and organic linkers, MOFs offer great promise for efficient H2/CH4 separations. A small number of MOFs, maximum of 20, was studied using molecular simulations for membrane-based H2/CH4 separations in the literature.27,34–38 Haldoupis et al.39 performed the first large-scale molecular simulation study and used a geometric approach to calculate the ideal H2/CH4 selectivity of 143 MOF membranes at infinite dilution condition. They showed that many MOFs have high H2 permeabilities (>105 Barrer) relative to the well-known polymers in addition to high H2/CH4 selectivities in the range of 1–100. Our research group performed molecular simulations for equimolar H2/CH4 mixtures to examine the membrane-based separation performances of 172 MOFs under industrial operating conditions, 10 bar and 298 K.40 Adsorption data obtained from grand canonical Monte Carlo (GCMC) and diffusion data obtained from molecular dynamics (MD) simulations were used to predict gas permeabilities and selectivities of MOF membranes. This was the largest number of MOFs for which MD simulations were performed for H2/CH4 mixtures in the literature. Results showed that only a small number of MOF membranes is H2 selective and diffusion selectivity for H2 dominates the adsorption selectivity for CH4 in these MOFs. As can be seen from this literature review, there is no high-throughput computational study that examines the membrane-based H2/CH4 separation potential of all existing MOFs in the literature.
The most complete collection of MOFs maintained by the Cambridge Structural Database (CSD) was recently reported.41 We recently screened this database for adsorption-based separation of CH4/H2 mixtures and reported several adsorbent evaluation metrics of MOFs such as adsorption selectivity, working capacity, adsorbent performance score, sorbent selection parameter, and regenerability.42 However, this MOF database has not been screened for any membrane-based gas separation application to date to the best of our knowledge. This type of study is strongly needed not only to see if MOF membranes can replace traditional membranes but also to provide insights into the influence of pore size, pore topology and chemistry of MOFs on the membranes' performances. In this work, we performed the first high-throughput molecular simulation study in the literature to assess the potential of the most complete collection of MOFs for membrane-based H2/CH4 separation. Combining GCMC and MD simulations, we computed adsorption and diffusion coefficients of H2 and CH4 in all MOFs. This data was then used to estimate adsorption selectivity, diffusion selectivity and membrane selectivity of MOFs and relations between these selectivities were discussed to understand the individual effects of adsorption and diffusion on the membranes' performances. H2/CH4 selectivities and H2 permeabilities of 4240 different MOF membranes were calculated and compared with traditional membranes, such as zeolites and polymers, to reveal the membrane-based H2/CH4 separation potential of the MOF structures. The most promising MOF membranes that are above the Robeson's upper bound43 were identified. Mixture GCMC and MD simulations were also performed for the top ten most promising MOF membranes to evaluate their performances for separation of equimolar H2/CH4 mixtures under practical operating conditions. Results were compared with the predictions of molecular simulations performed at infinite dilution loadings of single-component gases. The accuracy of the high-throughput screening of MOFs based on single-component adsorption and diffusion data was discussed in detail. We finally examined the relations between structural properties of MOFs such as pore sizes, porosities, surface areas, densities, lattice structure types, metal types and their membrane selectivities and gas permeabilities to provide quantitative structure–performance relationships that can serve as a map for the development of new MOF membranes with better H2/CH4 separation performances. This type of fundamental understanding into structure–performance relations of MOF membranes that are capable of achieving energy efficient H2/CH4 separations has the potential to allow revolutionary changes in practical gas separation applications. Therefore, results presented in this work, identification of the high performance MOF membranes and microscopic insights provided for the structure–performance relations, will motivate a large number of researchers working on design and development of MOF membranes for various gas separations.
The Henry's constants (K0) of gas molecules were initially calculated at the limit of zero-coverage (infinite dilution) using 105 moves of the Widom particle insertion method47 by performing Monte Carlo simulations. We also performed GCMC simulations for equimolar H2/CH4 mixtures at 1 bar and 298 K. In mixture GCMC simulations, four different types of moves including translation, reinsertion, swap of a molecule and identity exchange of molecules were considered. The Lorentz–Berthelot mixing rules were employed and the Peng–Robinson equation of state was used to convert the pressure to the corresponding fugacity. The cut-off distance for truncation of the intermolecular interactions was set to 13 Å. The simulation cell lengths were increased to at least 26 Å along each dimension and periodic boundary conditions were applied in all simulations. For each MOF, simulations were carried out for 10000 cycles with the first 5000 cycles for initialization and the last 5000 cycles for taking ensemble averages.
We then performed MD simulations to compute single component self-diffusivities (D0) of H2 and CH4 in MOFs' pores at 298 K and infinite dilution. Following the literature,29 we switched off the gas–gas intermolecular interactions and inserted 30 gas molecules into each MOF to represent the infinite dilution condition and to obtain high accuracy in the simulations. The self-diffusion coefficients were calculated by using the slope of the mean square displacement of gas molecules. For each MOF, MD simulations were carried out for 106 cycles in the NVT ensemble using a time step of 1 fs. We used 1000 initialization cycles and 10000 equilibration cycles. The Nosé–Hoover thermostat47 was used in NVT-MD simulations. We also performed mixture MD simulations and the initial states of these simulations were created with the appropriate loadings determined from the GCMC simulations performed at 1 bar and 298 K. Both gas–gas and gas–framework interactions were considered. Mixture MD simulations were carried out using 106 cycles in the NVT ensemble with a time step of 1 fs. At least 10 trajectories were used to compute self-diffusivities of each component. More details of these simulations can be found in the literature.47,57
MOFs were assumed to be rigid in their reported crystallographic structures in simulations. This assumption has been used in all large-scale molecular simulation studies of MOFs to save significant computational time. We recently showed that if the MOF has large pore sizes, framework flexibility has a negligible effect on the gas permeability and H2/CH4 selectivity of MOF membranes.40 Flexibility was found to affect only gas permeability without changing the selectivity of the MOF membranes which have narrow pore sizes. In this work, we only considered MOFs with pore sizes larger than the kinetic diameters of both gas molecules, therefore flexibility is expected to have a negligible effect on the results.
Formula | |
---|---|
a K 0: Henry's constant at infinite dilution. D0: self-diffusivity at infinite dilution. C: loading of the gas species in the mixture. y: composition of the gas species in the bulk phase. Dmix: self-diffusivity of gas species in the mixture. i, j: gas species and i/j represents the selectivity of i over j. | |
Permeability | P 0i = K0i × D0i |
Adsorption selectivity | |
Diffusion selectivity | |
Membrane selectivity | S 0mem,i/j = S0ads,i/j × S0diff,i/j |
Mixture adsorption selectivity | |
Mixture diffusion selectivity | |
Mixture membrane selectivity | S mixmem,i/j = Smixads,i/j × Smixdiff,i/j |
Fig. 1 Comparison of our results and experiments for single-component and mixture gas permeances through IRMOF-1, Ni-MOF-74, ZIF-8, ZIF-69, ZIF-78, ZIF-90 and ZIF-95 membranes. Gas permeance measurement conditions of MOF membranes and related experimental references can be seen in Table S1.† |
The membrane selectivities of MOFs computed at infinite dilution change from 1.4 × 10−3 to 97. The majority of the MOFs (3777) was identified to be CH4 selective as shown in Fig. 2. The maximum CH4 selectivity of MOF membranes was predicted to be 713. In these MOFs, high adsorption selectivity towards CH4 dominates the diffusion selectivity towards H2. We calculated selectivity of 68 MOFs to be around unity, , which means they do not have a preference for CH4 or H2, therefore they cannot be used as selective membranes. A smaller number of MOFs, 395, was found to be H2 selective. In these MOFs, high diffusion selectivity towards H2 overcompensates the adsorption selectivity towards CH4 and the highest H2 selectivity of the MOF membrane was estimated to be 97. Overall, Fig. 2 represents that the most H2 selective MOF membranes are the ones that have high diffusion selectivities as shown in red colors.
Separation performances of MOF membranes were compared with those of traditional polymer and zeolite membranes to assess the potential of MOFs in H2/CH4 separations. Polymeric membranes have been reported to be selective for H2 over CH4 and Robeson43 described an upper bound for these membranes. Several polyimide membranes establish the upper bound at the low H2 permeability end whereas poly(trimethylsilylpropyne) membranes are located at the high H2 permeability end of the upper bound. Fig. 3 shows H2 permeability and H2/CH4 selectivities of MOF membranes together with the Robeson's upper bound. The H2 permeabilities of MOF membranes were calculated to have a very wide range from 0.372 to 1.67 × 106 Barrer. Most of MOFs (4162) exhibit H2 permeabilities in the range of 104–106 Barrer as shown with blue points in Fig. 3. At that point it is important to highlight the importance of studying the entire MOF database rather than focusing on a small number of MOFs having similar physical and chemical properties. In our previous work,40 by studying 172 MOFs, we concluded that H2 permeabilities of MOFs are generally in the range of 103–105 Barrer. In this work, by examining the most recent and complete collection of the MOF database that spans a large variety in pore sizes and chemical topologies, we showed that H2 permeabilities of many MOFs are actually very large, >105 Barrer. Since these H2 permeabilities are significantly higher than the permeabilities of polymeric membranes, we extrapolated the Robeson's upper bound with a dashed line in Fig. 3. High gas permeabilities of MOFs can be attributed to their high porosities as we will discuss below. 1545 MOFs were found to exceed the upper bound either due to their high H2 permeabilities or high H2 selectivities or a combination of these two. In order to describe the most promising membrane materials, we specifically focused on the MOFs that have and Barrer, which are shown by red points in Fig. 3. Gas diffusivities, gas permeabilities, adsorption, diffusion and membrane selectivities of the top ten MOF membranes are all listed in Table 2 in addition to their structural properties.
MOF | LCD (Å) | PLD (Å) | ϕ | (cm2 s−1) | (cm2 s−1) | (Barrer) | (Barrer) | |||
---|---|---|---|---|---|---|---|---|---|---|
MEFMEQ | 5.24 | 4.26 | 0.52 | 2.98 × 10−7 | 2.77 × 10−4 | 1.16 × 102 | 1.13 × 104 | 9.60 | 930.57 | 96.95 |
OGAJEN | 6.47 | 3.96 | 0.55 | 3.74 × 10−8 | 4.36 × 10−4 | 2.15 × 103 | 6.02 × 104 | 416.16 | 11650.87 | 28.00 |
OGAMOA | 6.37 | 3.80 | 0.54 | 2.58 × 10−8 | 3.51 × 10−4 | 2.33 × 103 | 5.06 × 104 | 626.72 | 13599.93 | 21.70 |
OGALEP | 6.38 | 3.80 | 0.54 | 2.78 × 10−8 | 3.49 × 10−4 | 2.38 × 103 | 4.96 × 104 | 603.51 | 12558.17 | 20.81 |
OGALUF | 6.38 | 3.84 | 0.54 | 3.35 × 10−8 | 3.68 × 10−4 | 2.56 × 103 | 5.17 × 104 | 545.05 | 10995.69 | 20.17 |
OGAKUE | 6.32 | 3.81 | 0.54 | 2.82 × 10−8 | 3.63 × 10−4 | 3.09 × 103 | 5.37 × 104 | 743.25 | 12896.03 | 17.35 |
PIZHOX | 6.01 | 4.81 | 0.65 | 5.58 × 10−7 | 2.36 × 10−4 | 1.01 × 103 | 1.70 × 104 | 24.99 | 422.00 | 16.88 |
OGAJAJ | 6.24 | 3.82 | 0.54 | 3.22 × 10−8 | 2.72 × 10−4 | 2.87 × 103 | 3.89 × 104 | 622.61 | 8438.47 | 13.55 |
HIFVUO | 7.50 | 5.98 | 0.71 | 4.50 × 10−6 | 4.65 × 10−4 | 1.84 × 103 | 2.18 × 104 | 8.67 | 103.15 | 11.90 |
QONKOV | 5.98 | 4.81 | 0.64 | 1.01 × 10−6 | 2.74 × 10−4 | 1.80 × 103 | 2.04 × 104 | 24.09 | 272.50 | 11.31 |
The most promising MOF membrane in Table 2 is MEFMEQ. This MOF offers both high , 1.1 × 104 Barrer and high , 97. The high membrane selectivity is due to its large diffusion selectivity towards H2 (931) that dominates the weak adsorption selectivity towards CH4 (9.6). Similarly, the second most promising MOF, OGAJEN, has a high (28) due to its high diffusion selectivity for H2 (11651). Although adsorption strongly favored CH4 in this MOF with a selectivity of 416, very fast diffusion of H2 (4.4 × 10−4 cm2 s−1) compared to CH4 (3.7 × 10−8 cm2 s−1) dominated the adsorption selectivity. In fact, Table 2 shows that all promising MOF membranes have very high (>100). These MOFs have narrow PLDs, generally around 3.8–4.8 Å, close to the kinetic diameter of CH4. Strongly confined CH4 molecules diffuse slowly, in the order of 10−6–10−8 cm2 s−1, in these MOFs whereas H2 diffusion is at least two orders of magnitude higher, 10−4 cm2 s−1. As a result of this difference in diffusivities, MOFs become H2 selective membranes. Relations between selectivities and structural properties of MOFs will be discussed in detail below.
We also compared the MOF membranes with zeolite membranes. Similar to polymeric membranes, zeolites CHA, LTA and ITQ-29 (which is the pure silica version of LTA, in other words, the same structure) are H2 selective membranes because they have narrow windows which are not accessible for CH4 molecules. H2 selectivities of these zeolite membranes were reported as ∼89, 17, 125 and their H2 permeabilities were reported as 4.57 × 104, 5.8 × 103 and 8.21 × 103 Barrer, respectively.58 MOFs offer similar H2 selectivities and higher H2 permeabilities than these zeolites. Zeolite MFI, which has larger pore windows compared to other zeolites, was reported to exhibit high CH4 permeability (∼105 Barrer) and it is a CH4 selective membrane with a selectivity of 13. We identified 795 MOF membranes that are much more CH4 selective ( in the range of 13–713) and much more CH4 permeable ( in the range of 1.0 × 105–1.2 × 108 Barrer) than MFI membrane. Carbon nanotube membranes (CNT) were shown to exhibit very large CH4 permeabilities (2 × 107 Barrer) and good CH4 selectivities (28) due to the very fast diffusion of gas molecules (0.1–1 cm2 s−1) which was attributed to the smoothness of the potential energy surface.59 17 MOFs were identified to outperform CNT membranes in terms of CH4 selectivities (69–713) and CH4 permeabilities (∼107–108 Barrer) in this work. This comparison shows that several MOFs can replace or even outperform zeolite and polymer membranes in H2/CH4 separations.
Results obtained so far showed that there are many promising MOFs that offer high permeability and selectivity for efficient separation of H2 from CH4. However, these MOFs have not been fabricated as membranes so far. In other words, MOFs that we identified as top promising membranes have not been fabricated as membranes yet, therefore it is not possible to make a direct comparison between experiments and simulations for the top promising MOF membranes. We already showed the good agreement between our simulations and experimental measurements for the performance of several different types of fabricated, prototype MOF membranes in Fig. 1 and we believe that these are the direct evidence of the good predicting power of our calculation methodology. Membrane studies in the literature generally focused on well-known MOFs. For example, several different research groups18,19 fabricated IRMOF-1 (also known as MOF-5) membranes and reported its H2/CH4 selectivity as 2–3 and H2 permeance (permeability divided by membrane thickness) as 8–47 × 10−7 mol m−2 s−1 Pa−1 based on single-component gas measurements. By studying the entire MOF database, we found that there are 21 MOFs that possess H2/CH4 selectivity >2 and H2 permeance >50 × 10−7 mol m−2 s−1 Pa−1 assuming 25 μm membrane thickness. There are 39 MOFs that exhibit H2/CH4 selectivity around 2, similar to MOF-5. It is important to note that selectivity and permeance of the MOF-5, SAHYIK in the CSD, was predicted as 1.39 and 44 × 10−7 mol m−2 s−1 Pa−1 in this work which agree well with the experimentally reported values.40,56 These results highlight the importance of computational screening of the entire MOF database to direct the experimental studies to more promising membrane materials. At that point, we would like to note that one of the most widely fabricated MOF membrane, ZIF-8, did not appear in our discussion because we limited our focus with MOFs having limiting pore sizes greater than 3.8 Å so that both CH4 and H2 molecules can pass through the membranes' pores. Since the small pore size of ZIF-8 was computed as 3.4 Å, this MOF was not included in the 4240 MOFs that we examined.
We then examined relations between gas diffusivities and PLDs of MOFs. Fig. 5(a) and (b) show that both increase as the PLD and ϕ of MOFs increase. is generally at least one order of magnitude larger than in MOFs having small PLDs (<5 Å) since smaller H2 molecules diffuse faster than the bulkier CH4 molecules in narrow pores of MOFs. Diffusivities get close to each other in MOFs having large pore sizes (>10 Å) since both gases can easily diffuse regardless of the molecular size. As a result of these, the highest is located at PLDs < 5 Å and ϕ < 0.75 as shown in Fig. 5(c). As the PLD and ϕ increase, generally decreases since both molecules can readily diffuse regardless of their molecular sizes. At that point it is important to note that diffusion rates of gases and hence S0diff do not solely depend on the kinetic diameters of gas molecules and pore sizes of MOFs. At the narrow PLD region, show a large variety of values from 10−8 to 10−3 cm2 s−1 indicating that diffusion is not only determined by the pore sizes but also interactions between gas molecules and MOFs play an important role. Finally, Fig. S4† shows that D0 of both gases generally increases (decreases) with increased SA and ϕ (ρ). This is expected because as the porosity increases, framework density decreases and gas molecules easily diffuse in less confined spaces. Since both diffusivities have similar trends with the SA and ρ, , which was calculated as the ratio of diffusivity of H2 to that of CH4, is almost independent from them. According to Fig. S5,†S0diff does not show an obvious dependency to SA, ϕ, and ρ. Relations between P0 of gases, PLD and ϕ of MOFs are shown in Fig. 6. The trend of P0 and PLD shown in Fig. 6(a) is very similar to the one between D0 and PLD given in Fig. 5(a), since increases with ϕ, also increases. Fig. S6† represents that is positively (negatively) correlated with SA (ρ) due to the same reasons we discussed for . As a result of these, narrow pore sizes and smaller porosities lead to higher as shown in Fig. 6(c). also shows an increasing (decreasing) trend with respect to SA and ϕ (ρ) as shown in Fig. S7.† We defined quantitative limits for the structural properties of MOFs that lead to high performance membranes in the next section. It is also important to note that these structure–performance relations have some similarities with the ones reported in a molecular simulation study which examined the CO2/CH4 separation performances of hypothetical MOF membranes that have PLDs between 3 and 4 Å.29 It was shown that diffusivities of CO2 and CH4 increase with increasing PLDs of hypothetical MOFs and the highest diffusion selectivities for CO2/CH4 were obtained at the low PLD region, between 3.0 and 3.2 Å. In our work, we considered real, synthesized MOF database with PLDs between 3.8 and 31 Å and showed that the highest diffusion selectivities for H2/CH4 separation was also found to be at the low PLD region, <5 Å.
Fig. 5 Self diffusivities of (a) CH4 (b) H2 as a function of PLDs of MOFs. (c) Diffusion selectivities of MOFs for H2/CH4 as a function of PLD and porosity. |
Fig. 6 Gas permeabilities as a function of (a) PLD (b) porosity of MOFs. (c) Membrane selectivities of MOFs for H2/CH4 as a function of PLD and porosity. |
Fig. 7 Effects of structural properties on the performance of MOF membranes. Numbers on the circles represent the number of MOFs. The outer circle represents all MOFs considered in this study (4240 MOFs), the middle circle represents the top 100 H2 selective MOFs having Barrer and the inner circle represents the top 10 most promising MOFs that we discussed in Table 2, the ones having and Barrer. |
We also examined the structural similarities of the top ten most promising MOFs and results are shown in Fig. S8† together with the 2 × 2 × 2 unit cell representations of these MOFs. Structural similarity between two MOFs is represented with an index. Green colors correspond to the most similar materials, the ones having similarity index >0.8. Light green colors represent highly similar structures with indexes between 0.7 and 0.8, yellow colors show the slightly similar structures with indexes between 0.5 and 0.7 and orange colors correspond to least similar structures with indexes between 0.3 and 0.5. Red colors show the dissimilar MOFs in term of their analyzed ray trace representation resulting in an index of 0.0–0.3. Fig. S8† shows that MEFMEQ and 6 OGA-materials have high similarities with indexes greater than 0.8. PIZHOX and QONKOV also showed similarities with other top MOFs with an index of 0.65 whereas the index between those two structures was 0.83 indicating a high similarity. HIFVUO has the least similarity with other MOFs giving indexes around 0.5. There is no red color in Fig. S8† which means that the top ten MOF membranes have some common structural features which is also observable from the unit cell representations.
As explained in the screening strategy described in Section 2.3, the top ten most promising MOFs were identified by performing single-component simulations at infinite dilution. We also performed equimolar mixture GCMC and MD simulations at 1 bar to unlock the mixture separation performances of the top promising MOF membranes under realistic operating conditions. Gas diffusivities, permeabilities, and all three selectivities computed using single-component data at infinite dilution were compared with the ones computed for equimolar mixtures in Table 3. Results showed that mixture adsorption selectivities are similar to the ones computed at infinite dilution for 4 MOFs, HIFVUO, MEFMEQ, PIZHOX and QONKOV. These are the MOFs that have relatively low adsorption selectivities, <25. On the other hand, the MOFs (OGA-materials) which exhibit very high adsorption selectivities at infinite dilution, ∼500, were found to have significantly lower mixture adsorption selectivities, ∼90. Fig. S9† compares computed at 1 bar with for the entire MOF database and results show that infinite dilution adsorption selectivities significantly overestimate mixture adsorption selectivities especially if the MOF is highly CH4 selective (>50). For 4 MOFs which have similar and (MEFMEQ, PIZHOX, HIFVUO and QONKOV), mixture gas diffusivities were found to be very similar to the ones computed at infinite dilution. As a result, membrane selectivity computed for mixture was found be close to the membrane selectivity computed at infinite dilution . On the other hand, for MOFs that have significant differences between and , mixture diffusivities were computed to be different than the single-component diffusivities. Table 3 shows that is almost one order of magnitude larger than whereas and are almost same. The increase in the diffusion rate of CH4 can be attributed to the multi-component mixture effects60 and collaborative interactions between many adsorbed CH4 molecules, which are missing in the single-component case. It is known that fast-diffusing gas molecules (in our case H2) can fasten the slow-diffusing gas molecules (in our case CH4) in the MOFs' pores due to the multi-component mixture effects.36 As a result, becomes much lower than . Overall, simulations performed for equimolar mixtures at 1 bar give higher adsorption selectivities but lower diffusion selectivities for H2 than the simulations performed at infinite dilution. As a result, H2 selectivities of MOF membranes, which were computed as the multiplication of adsorption and diffusion selectivities, become similar. Comparison of the single-component and mixture calculations for the H2 permeability and H2/CH4 selectivities of the top promising MOF membranes is also given in Fig. S10.† 9 out of 10 top performing MOFs analyzed using mixture simulations still satisfy the two performance criteria, and Barrer, used to identify the best MOF membranes. This result shows that the high-throughput computational screening strategy which utilizes gas loadings and diffusivities computed at infinite dilution can be used to make quick and accurate estimations for the mixture separation performances of MOF membranes to identify the most promising candidates.
MOF | (cm2 s−1) | (cm2 s−1) | (cm2 s−1) | (cm2 s−1) | ||||||
---|---|---|---|---|---|---|---|---|---|---|
MEFMEQ | 9.60 | 10.01 | 930.57 | 526.66 | 96.95 | 52.60 | 2.98 × 10−7 | 3.86 × 10−7 | 2.77 × 10−4 | 2.04 × 10−4 |
OGAJEN | 416.16 | 86.74 | 11650.87 | 1700.71 | 28.00 | 19.61 | 3.74 × 10−8 | 2.63 × 10−7 | 4.36 × 10−4 | 4.47 × 10−4 |
OGAMOA | 626.72 | 90.95 | 13599.93 | 1527.64 | 21.70 | 16.80 | 2.58 × 10−8 | 2.30 × 10−7 | 3.51 × 10−4 | 3.51 × 10−4 |
OGALEP | 603.51 | 91.19 | 12558.17 | 1408.20 | 20.81 | 15.44 | 2.78 × 10−8 | 2.74 × 10−7 | 3.49 × 10−4 | 3.86 × 10−4 |
OGALUF | 545.05 | 93.86 | 10995.69 | 2068.35 | 20.17 | 22.04 | 3.35 × 10−8 | 2.30 × 10−7 | 3.68 × 10−4 | 4.76 × 10−4 |
OGAKUE | 743.25 | 96.23 | 12896.03 | 1969.58 | 17.35 | 20.47 | 2.82 × 10−8 | 1.96 × 10−7 | 3.63 × 10−4 | 3.85 × 10−4 |
PIZHOX | 24.99 | 23.72 | 422.00 | 567.13 | 16.88 | 23.91 | 5.58 × 10−7 | 4.70 × 10−7 | 2.36 × 10−4 | 2.66 × 10−4 |
OGAJAJ | 622.61 | 97.36 | 8438.47 | 1800.22 | 13.55 | 18.49 | 3.22 × 10−8 | 1.77 × 10−7 | 2.72 × 10−4 | 3.19 × 10−4 |
HIFVUO | 8.67 | 8.31 | 103.15 | 171.90 | 11.90 | 20.68 | 4.50 × 10−6 | 3.76 × 10−6 | 4.65 × 10−4 | 6.44 × 10−4 |
QONKOV | 24.09 | 22.93 | 272.50 | 251.49 | 11.31 | 10.97 | 1.01 × 10−6 | 9.49 × 10−7 | 2.74 × 10−4 | 2.39 × 10−4 |
Finally, it is important to discuss the advancement of this work compared to our previous study on MOF membranes.40 In our previous work, we performed mixture GCMC and MD simulations to predict membrane performances of 172 types of MOFs. In this work, we described a high-throughput computational screening strategy which utilizes gas loadings and diffusivities computed at infinite dilution to make accurate estimations for the mixture separation performances of MOF membranes. Since we screened a very large number of MOF membranes, we were able to show clear structure–performance results that were not present when we studied only hundreds of MOFs. Therefore, we advanced the methodology to save significant computational time and also provided solid insights into structure–performance relations of MOFs by studying the whole MOF database.
Footnotes |
† Electronic supplementary information (ESI) available: Supporting data associated with this article can be found in the online version: comparison of our results and experiments for single-component and mixture gas permeances through various MOF membranes; relations between Henry's constants, diffusivities, permeabilities, adsorption selectivities, diffusion selectivities, membrane selectivities and PLD, LCD, SA, ϕ, ρ of MOFs; similarity indexes and unit cell representations of the top ten promising MOFs; comparison of with ; comparison of the single-component and mixture calculations for the permeability and selectivities of the top promising MOF membranes. See DOI: 10.1039/c8ta01547c |
‡ Equal contribution. |
This journal is © The Royal Society of Chemistry 2018 |