2.5.1. Flat Sheet Membranes
An extensive comparison of the performance of polymeric membranes based on different materials and the use of a wide variety of fillers can be found elsewhere; therefore, this review will focus mainly on alternative materials, methods and procedures that have contributed to the improvement of the hydrogen recovery results already discussed in the previous section [
25,
33,
35,
40,
75,
81,
82]. Matrimid
® MMMs based on MOFs, such as Cu-414′-BPY-HFS, MIL-53, NH
2-MIL-53(Al), NH
2-MIL-101(Al), TKL-107 and FeBTC, led to a maximum hydrogen permeability of 20.3 Barrer and H
2/CO
2 selectivity of 2.6, which are comparable to the results already obtained for pristine Matrimid
®.
According to the results included in
Figure 8 and
Table 5 corresponding to the modification of Matrimid/ZIF membranes, it is possible to achieve better performance and yield from Matrimid
® membranes. The combination of ZIF, annealing procedures and blending with other polymers or the selection of optimal operating conditions could guide the process towards exceptional permeation properties. In general terms, the increase in H
2/CO
2 and H
2/N
2 selectivities occurred, maintaining or lowering hydrogen permeability, whereas higher hydrogen permeability took place at a lower H
2/CO
2 selectivity than that showed in
Figure 6.
Knebel et al. (2018) reported that the combination of ZIF-8 and ZIF-67 in Matrimid
® MMMs caused a hydrogen recovery that was in the Robeson upper bound (2008) (
Figure 8) with a hydrogen/carbon dioxide real selectivity of 5.3 and a H
2 permeability of 94 Barrer at room temperature [
83]. Meanwhile, when the operating temperature was elevated till 150 °C, H
2/CO
2 selectivity was 7.2 with a hydrogen permeability of 237 Barrer, achieving an attractive zone above the Robeson upper bound (2008) with a compromise between selectivity and permeability. When layers of ZIF-67 and ZIF-67 on the ZIF-8 layer, and ZIF-8 on the ZIF-67 layer, were prepared supported by alumina, H
2/CO
2 selectivity was up to 13.2, but H
2/N
2 and H
2/CH
4 selectivities were in fact low compared to those obtained with pristine Matrimid
®. It is likely that the sorption of carbon dioxide is more relevant than diffusion selectivity and causes similar values of selectivity independently of the gas.
Table 5.
Main bibliographical data in terms of pure gas H2, N2, CO2 and CH4 permeabilities and selectivities through modified Matrimid®/ZIF MMMs in which polymer has been substituted or blended, the MMM has been chemically modified and/or the filler has been substituted or functionalized.
Table 5.
Main bibliographical data in terms of pure gas H2, N2, CO2 and CH4 permeabilities and selectivities through modified Matrimid®/ZIF MMMs in which polymer has been substituted or blended, the MMM has been chemically modified and/or the filler has been substituted or functionalized.
Ref. | T (°C) | ΔP (Bar) | PH2 (Barrer) | PN2 (Barrer) | PCO2 (Barrer) | PCH4 (Barrer) | αH2/N2 | αH2/CO2 | αH2/CH4 | Polymer | Modification |
---|
Carter (2017) [48] | 35 | 3.0 | 30.3 | 0.70 | 9.5 | 0.32 | 43.3 | 3.2 | 94.7 | Matrimid® | pristine Matrimid |
34.0 | 0.73 | 10.5 | 0.79 | 46.6 | 3.2 | 43.0 | Silicalite calcined (10 wt.%) |
28.3 | 0.36 | 9.5 | 0.30 | 78.6 | 3.0 | 94.3 | Silicalite uncalcined (10 wt.%) |
40.2 | 1.19 | 12.5 | 1.34 | 33.8 | 3.2 | 30.0 | SAPO-34 calcined (10 wt.%) |
25.2 | 0.29 | 7.6 | 0.24 | 86.9 | 3.3 | 105.0 | SAPO-34 uncalcined (10 wt.%) |
Diestel (2015) [46] | 25 | 0.2 | 19.0 | | 2.0 | | | 9.5 | | Matrimid® | ZIF-90 + ethylendiamine |
Esposito (2019) [63] | 25 | | 328 | 6.83 | 198 | 9.14 | 48.0 | 1.7 | 35.9 | Matrimid®/PIM | |
| 1630 | 62.8 | 1380 | 77.6 | 26.0 | 1.2 | 21.0 | PIM | |
Ghanem (2020) [84] | 35 | 2.0 | 4.3 | 0.05 | 1.4 | 0.04 | 87.1 | 3.1 | 108.9 | Commercial polyimide from Alfa Aesar | d-PI |
11.2 | 0.12 | 3.2 | 0.10 | 96.9 | 3.5 | 110.2 | 5 wt.% ZIF-302 d-PI |
386.1 | 12.9 | 207.3 | 12.3 | 29.9 | 1.9 | 31.4 | s-PI |
469.2 | 7.5 | 186.0 | 11.1 | 62.6 | 2.5 | 42.3 | 5 wt.% ZIF-302 s-PI |
Hosseini (2007) [56] | 35 | | 32.2 | 0.36 | 8.3 | 0.28 | 89.4 | 3.9 | 115.8 | Matrimid® | 20 wt.% MgO untreated |
25.3 | 0.32 | 7.4 | 0.25 | 79.3 | 3.4 | 103.3 | 20 wt.% MgO, 240 °C (12 h) |
37.6 | 0.50 | 10.8 | 0.39 | 74.8 | 3.5 | 96.7 | 20 wt.% MgO, 350 °C (1 h) |
41.1 | 0.52 | 11.6 | 0.21 | 79.0 | 3.5 | 199.5 | 20 wt.% MgO, 350 °C (0.5 h) |
19.8 | 0.18 | 5.1 | 0.13 | 108.2 | 3.9 | 152.3 | 20 wt.% MgO, silver treatment 2 days |
22.5 | 0.17 | 5.1 | 0.12 | 130.1 | 4.5 | 186.0 | 20 wt.% MgO, silver treatment 5 days |
22.7 | 0.16 | 4.3 | 0.10 | 146.5 | 5.3 | 222.5 | 20 wt.% MgO, silver treatment 10 days |
Hosseini (2018) [45] | 35 | H2: 3.5 Other gases: 10.0 | 5.5 | 0.021 | 0.6 | 0.001 | 260.5 | 9.4 | 5500.0 | Matrimid®/PBI (25/75 wt.%) | |
4.0 | 0.014 | 0.3 | 0.016 | 288.6 | 13.1 | 253.2 | p-xylene dichloride |
3.6 | 0.013 | 0.1 | 0.003 | 271.2 | 26.1 | 1200.0 | p-xylene diamine |
Knebel (2018) [83] | 25 | 0.5 | | | | | 8.9 | 6.5 | 5.5 | Ceramic support of α-Al2O3 | ZIF-67 |
| | | | 10.4 | 12.9 | 11.4 | ZIF-67 on ZIF-8 |
| | | | 9.3 | 13.2 | 11.1 | ZIF-8 on ZIF-67 |
94.0 | | 17.7 | | | 5.3 | | Matrimid® | ZIF-8 and ZIF-67 |
150 | 237.0 | | 32.5 | | | 7.3 | |
Mei (2020) [85] | 30 | 4.0 | 23.3 | | 2.5 | | | | 9.3 | Polysulfone | 10 wt.% ZIF-8 with PDA coating |
Mirzaei (2020) [44] | 25 | 5.0 | 68.9 | 0.51 | 13.6 | 0.34 | 135.9 | 5.1 | 201.1 | Matrimid® | 20 wt.% Pd@ZIF-8 |
Mundstock (2017) [86] | 20 | | 17.0 | | 5.7 | | | 3.0 | | Matrimid® supported over Al2O3 | |
| 50.8 | | 12.9 | | | 4.0 | | NaX |
1.0 | 21.2 | | 4.5 | | | 4.8 | | PbX |
| 29.3 | | 5.7 | | | 5.2 | | CuX |
| 26.0 | | 4.8 | | | 5.6 | | NiX |
| 23.0 | | 4.2 | | | 5.6 | | Cox |
Perez (2009) [87] | 35 | 2.0 | 29.9 | 0.28 | 11.1 | 0.22 | 106.8 | 2.7 | 135.9 | Matrimid® | 10 wt.% MOF-5 |
38.3 | 0.40 | 13.8 | 0.34 | 95.8 | 2.8 | 112.6 | 20 wt.% MOF-5 |
53.8 | 0.52 | 20.2 | 0.45 | 103.5 | 2.7 | 119.6 | 30 wt.% MOF-5 |
Sánchez-Laínez (2018) [88] | 35 | 2.0 | | | | | | 3.8 | | Polyamide on P84® support | ZIF-8 (0%w/v) |
180 | 2.0 | | | | | | 7.9 | |
250 | 2.0 | | | | | | 8.4 | |
35 | 2.0 | | | | | | 4.4 | | ZIF-8 (0.2%w/v) |
180 | 2.0 | | | | | | 9.2 | |
250 | 2.0 | | | | | | 11.5 | |
35 | 2.0 | | | | | | 9.0 | | ZIF-8 (0.4%w/v) |
180 | 2.0 | | | | | | 14.6 | |
250 | 2.0 | | | | | | 13.4 | |
180 | 2.0 | | | | | | 7.2 | | ZIF-8 (0.8%w/v) |
Weigelt (2018) [64] | 30 | 1 | 39.0 | 0.44 | 14.5 | 0.43 | 88.6 | 2.7 | 90.7 | Matrimid® | 8% Activated Carbon |
63.8 | 0.81 | 25.6 | 0.67 | 78.8 | 2.5 | 95.2 | | 31% AC |
101 | 1.5 | 39.5 | 1.06 | 67.3 | 2.6 | 95.3 | | 44% AC |
180 | 2.8 | 66.7 | 2.25 | 64.3 | 2.7 | 80.0 | | 50% AC |
Yang (2011) [89] | 35 | 7.1 | 3.7 | | 0.4 | | | 8.7 | | PBI | pristine PBI |
7.7 | | 0.6 | | | 12.9 | | 10 wt.% ZIF-7 |
15.4 | | 1.3 | | | 11.9 | | 25 wt.% ZIF-7 |
26.2 | | 1.8 | | | 14.9 | | 50 wt.% ZIF-7 50 wt.% ZIF-7 |
180 | 440.0 | 25.4 | 14.6 |
Yang (2012) [90] | 35 | 3.5 | 3.7 | | 0.4 | | | 8.6 | | PBI | pristine PBI |
28.5 | | 2.2 | | | 13.0 | | 15 wt.% ZIF-8 |
1750 | | 426.6 | | | 4.1 | | 60 wt.% ZIF-8 |
26.2 | | 1.8 | | | 14.6 | | ZIF-7 |
Yang (2013) [91] | 35 | 3.5 | 4.1 | | 0.5 | | | 7.1 | | PBI | pristine PBI |
82.5 | | 6.9 | | | 6.8 | | 30 wt.% ZIF-8 |
1612.8 | | 397.6 | | | 2.8 | | 60 wt.% ZIF-8 |
230 | | 470.0 | | 17.9 | | | 26.3 | | 30 wt.% ZIF-8 |
| 2015.0 | | 163.8 | | | 12.3 | | 60 wt.% ZIF-8 |
Yang (2013) [92] | 35 | 3.5 | 12.7 | | 0.9 | | | 14.6 | | PBI | 10 wt.% ZIF-90 |
18.3 | | 0.9 | | | 20.6 | | 25 wt.% ZIF-90 |
24.5 | | 1.0 | | | 25.0 | | 45 wt.% ZIF-90 |
Yáñez (2020) [69] | 35 | 5.5 | 8.4 | 0.03 | 2.2 | 0.05 | 280.0 | 3.8 | 168.0 | PEI ULTEM® 1000B | |
11.3 | 0.09 | 4.4 | 0.20 | 132.4 | 2.6 | 56.3 | PES ULTRASON® E | |
0.6 | 0.002 | 0.3 | 0.001 | 322.5 | 2.2 | 645.0 | PBI Celazole® | |
Zhang (2008) [49] | 25 | 2.0 | 17.5 | 0.22 | 7.3 | 0.21 | 79.6 | 2.4 | 83.3 | Matrimid® | pristine Matrimid |
2.0 | 19.8 | 0.14 | 8.3 | 0.12 | 141.3 | 2.4 | 164.8 | 10 wt.% Meso-ZSM-5 |
1.5 | 19.6 | 0.14 | 8.5 | 0.13 | 139.7 | 2.3 | 150.5 | 10 wt.% Meso-ZSM-5 |
2.0 | 22.2 | 0.170 | 8.7 | 0.130 | 130.8 | 2.6 | 171.0 | 20 wt.% Meso-ZSM-5 |
35.4 | 0.31 | 14.6 | 0.26 | 114.1 | 2.4 | 136.0 | 30 wt.% Meso-ZSM-5 |
36.3 | 0.62 | 15.4 | 0.56 | 58.6 | 2.4 | 64.8 | 10 wt.% Meso-ZSM-5 (uncalcined) |
22.0 | 0.34 | 9.0 | 0.30 | 64.8 | 2.4 | 73.5 | 10 wt.% ZSM-5 |
23.1 | 0.30 | 9.4 | 0.28 | 77.1 | 2.5 | 82.6 | 10 wt.% MCM-48 |
Zhao (2008) [65,93] | 35 | 1.0 | 3.8 | 0.16 | 7.5 | 0.35 | 23.7 | 0.5 | 10.8 | Matrimid® | 1:0.2 PPG/PEG/PPGDA |
10.0 | 1.13 | 59.2 | 3.36 | 8.9 | 0.2 | 3.0 | 1:0.5 PPG/PEG/PPGDA |
15.8 | 2.19 | 115.8 | 6.80 | 7.2 | 0.1 | 2.3 | 1:1 PPG/PEG/PPGDA |
The functionalization of ZIF nanoparticles has arisen as a potential solution to improve hydrogen recovery, as pore aperture size can be controlled and/or active catalytic sites can be created, enhancing the gas separation performance [
94]. Palladium has gained great attention in hydrogen recovery owing to its high catalytic activity and high hydrogen sorption capacity, which facilitates hydrogen transport through the membrane [
44,
95,
96]. Some advantages of Pd NPs are the reasonable cost; high hydrogen permeability and selectivity; predictable behavior over a long period of time; or resistance to poisoning by H
2S, Cl
2, CO, etc. Nevertheless, some disadvantages accompanying Pd NP use in MMMs are associated with their potential aggregation, which can deteriorate the mechanical properties and separation parameters. To avoid Pd NP aggregation, the encapsulation of Pd in ZIF has been proposed. According to Mirzaei et al. (2020), membranes made of Pd@ZIF-8 displayed, in SEM analysis, a relatively uniform dispersion in Matrimid [
44]. When adding Pd NPs to the ZIF-8 structure, the permeability of H
2 (68.9 Barrer,
Table 5) was higher than that working with a Matrimid
®/ZIF-8 MMM (44.7 Barrer), increasing H
2/CO
2 ideal selectivity from 3.3 to 5.1, maintaining the permeabilities of CO
2, CH
4 and N
2 constant. This increase was connected to the mechanisms of adsorption, dissociation, association and desorption observed in hydrogen over palladium membranes and the ability of Pd for blocking ZIF-8 pores.
Similar to ZIF functionalization, other fillers, such as zeolite faujasite (FAU), can be functionalized by a wide range of metals. By using alumina supports modified with polydopamine (PDA) and 3-aminopropyltriethoxysilane (APTES), the homogeneous growth in the Matrimid
®/FAU membrane can be achieved [
86]. Usually FAU is synthesized as NaX, but after an ion exchange step with a cobalt, nickel, copper or lead salt, CoX, NiX, CuX or PbX, respectively, are obtained. The addition of Matrimid to form Al
2O
3/Matrimid
®/NaX led to a real H
2/CO
2 selectivity of 4.0 (
Table 5) higher than that obtained for Al
2O
3/Matrimid
® (3.0), and this is justified by the interaction of FAU with polar molecules due to its strong electrostatic potential and the accessibility to sodium sites. However, this selectivity (4.0) is lower than that obtained without the addition of Matrimid
® (10.3), a phenomenon explained by the weak Matrimid
®–NaX interaction, the sealing of existing cracks and pinholes and the widening of some of them, as can be inferred from the lower hydrogen and carbon dioxide permeances. Comparing NaX, CoX, NiX, CuX and PbX, the highest selectivity (5.6) corresponded to NIX and CoX, where Ni
2+ and Co
2+ had the highest ionic potential, confirming a higher ion–CO
2 interaction. Although H
2/CO
2 selectivity and the corresponding permeabilities are within the results obtained employing Matrimid
®/ZIF MMMs at room temperature, the performance is still deficient, as the prior upper bound is not reached. Matrimid
®/AC (activated carbon) MMMs allowed the increase in hydrogen permeability from 31.6 (pristine polymer) to 180 Barrer (50% AC), whereas H
2/CO
2 selectivity remained at around 2.7 (below the prior Robeson plot) [
64]. Nonetheless, H
2/N
2 and H
2/CH
4 separation almost reached the Robeson upper bound (
Figure 8). Good compatibility of these materials was observed by SEM, where the fillers were covered by a thin layer of polymer without displaying defects. The interaction between Matrimid
® and activated carbon could be enhanced by modifying carbon particles.
The usage of a linker in Matrimid
®/ZIF MMMs has been adopted as a strategy to further improve hydrogen recovery. The covalent binding of ZIF-90 with ethylenediamine and the mixture of this dispersion with Matrimid
® solution led to a H
2/CO
2 ideal selectivity of 9.5, with a hydrogen and carbon dioxide permeability of 19.0 and 2.0 Barrer, respectively (
Table 5) [
46]. The addition of the linker meant an improvement of the selectivity compared to Matrimid
®/ZIF MMM, yet a diminishment in H
2 and CO
2 permeabilities (
Table 4), surpassing the prior Robeson upper bound (1991). These results could be explained by the elimination in the ZIF-90 crystal surface of the linker distortion of the carboxyaldehyde imidazolate molecules. Considering these advances, it would be interesting to analyze the influence of annealing and permeation temperature. Other linkers, such as poly (propylene glycol) block poly (ethylene glycol) block poly (propylene glycol) diamine (PPG/PEG/PPGDA), suppress the hydrogen transport through the membrane, enriching the permeate stream in carbon dioxide and rendering its composition poor in the retentate (
Figure 8) [
65,
93]. This phenomenon is explained considering that the rubber phase controls the gas permeability instead of the glassy Matrimid
®.
In recent years, several studies have addressed the options of combining Matrimid with polybenzimidazole (PBI), a polymer that has interesting properties. For this reason, this review describes the characteristics and performance of some membranes based on PBI, and later the studies that report the combination of Matrimid with PBI. Polybenzimidazole (PBI) is a polymeric material selective towards hydrogen that displays good thermal stability [
90]. Nevertheless, PBI possesses brittleness and low gas permeance, even in hollow fiber configuration, as a consequence of the rigid polymeric backbone and the high-density chain packing, meaning that it is very difficult to obtain ultrathin PBI membranes [
92]. Taking into consideration the literature, Yang et al. (2011) listed six possible modifications that contribute to the enhancement of PBI permeation properties: blending with other polymers, varying the acid moiety during PBI synthesis, crosslinking procedures, thermal rearranging, N-substitution modification and incorporation of inorganic silica NPs [
89]. According to the results already published, polymer blending, the incorporation of NPs and crosslinking appear to be feasible options for using PBI in the recovery of hydrogen.
Yang et al. (2011) reported that during the fabrication of PBI/ZIF-7 MMM, good dispersion of ZIF-7 in PBI was obtained up to PBI:ZIF-7 50:50 wt.%, but particle agglomeration was recognized [
89]. The addition of ZIF-7 and the increase in the wt.% resulted in an improvement of both ideal and real H
2/CO
2 selectivity from 8.7 and 7.1 to 14.9 and 7.2, respectively, with maximum hydrogen permeabilities of 26.2 (single gas) and 13.3 Barrer (mixed gas) at 35 °C (
Figure 8 and
Table 5). These results are interesting, considering the H
2 permeability–selectivity relation that surpasses the Robeson upper bound (2008), especially when the operating temperature increased to 180 °C. The increase in hydrogen and carbon dioxide permeabilities was produced by the higher free volume, but, owing to the rigidity of the chains and ZIF-7 accessible cavity diameter (3.0 Å), hydrogen permeability was favored and, consequently, H
2/CO
2 selectivity was enhanced. As observed for Matrimid
®/ZIF, hydrogen and carbon dioxide permeabilities and H
2/CO
2 selectivity increased with ZIF-7 content in the polymer matrix.
When PBI was mixed with ZIF-8, H
2/CO
2 selectivities up to 14.6 were obtained and a maximum hydrogen permeability of 1750 Barrer was reached at 35 °C with a membrane containing 60 wt.% ZIF-8 [
90,
91]. Although permeability continuously increased with ZIF-8 content, the maximum selectivity corresponded to 30 wt.%. The combination PBI/ZIF-8 allowed the Robeson upper bound (2008) to be surpassed, but due to the higher accessible cavity of ZIF-8 (3.4 Å), the ideal selectivity was lower compared to PBI/ZIF-7 MMM, albeit higher hydrogen permeabilities could be obtained (
Figure 8). Permeation tests at a high temperature (230 °C) resulted in selectivities up to 26.3 (30 wt.%,
Table 5) and a maximum hydrogen permeability of 2015 Barrer. A diminishment in mixed gas permeability and selectivity was observed at 35 and 230 °C, but always maintaining the results above the Robeson upper bound (2008), thus exposing the gas transport competition between hydrogen and carbon dioxide [
91]. The addition of CO to the mixed gas feed stream did not affect H
2/CO
2 selectivity, but hydrogen permeability slightly decreased in 70:30 PBI/ZIF-8. In the 40:60 PBI/ZIF-8 MMM, both permeabilities decreased as a consequence of the competitive sorption and the likely CO pore blocking of the ZIF-8 cavity (CO kinetic diameter 3.76 Å versus cavity diameter 3.4 Å). The assessment of the water effect at different temperatures on the PBI/ZIF-8 MMM displayed a plateau as a result of the thermal stability of the membrane, the water condensation limit at high temperatures and the smaller kinetic diameter of water. The comparison between Matrimid
®/ZIF-8 (
Table 4) and PBI/ZIF-8 (
Table 5) usage displays a better performance of the latter in terms of permeabilities and selectivities when they are compared using the same operating temperature and ZIF-8 load.
By using ZIF-90 as a filler, PBI/ZIF-90 exhibited hydrogen permeability and H
2/CO
2 ideal selectivity up to 24.5 Barrer and 25.0, respectively, at 55:45 PBI:ZIF-90 and 35 °C (
Figure 8 and
Table 5) [
92]. According to the data, similarly, as explained in Matrimid
®/ZIF and PBI/ZIF, the higher ZIF-90 content resulted in a better separation performance. ZIF-90 allowed the obtention of higher selectivities at 35 °C compared to ZIF-7 and ZIF-8, which was attributed to the more acidic aldehyde group in ZIF-90 that reinforced the interaction with the alkali N-3 atom on the PBI imidazole ring. Simultaneous transport studies of hydrogen and carbon dioxide showed a diminishment in permeabilities and selectivity compared to ideal gases owing to the competitive sorption, but at 180 °C, hydrogen permeability reached 226.9 Barrer with a selectivity of 13.3.
Taking into consideration the advantages and disadvantages of both Matrimid
® and PBI, the performance of the polymeric blending of both materials was assessed by Hosseini et al. (2008) [
45]. Miscibility at the molecular level was confirmed by TGA and DSC at various ranges of compositions of the constituents, and it was explained considering the intermolecular interaction by hydrogen bonding between the N–H group of PBI and C=O group of Matrimid
®. Higher PBI contents resulted in lower permeabilities but higher H
2/CO
2, H
2/N
2 and H
2/CH
4 selectivities, reaching selectivities of 9.4, 260.5 and 5500, respectively (
Figure 8 and
Table 5), and this was associated with the free volume provided by Matrimid
® (0.268, pure) and PBI (0.116, pure). In spite of the similar structure, the lateral oxygen and methyl groups of Matrimid
® avoid polymeric chains to be closer. When both polymers are blended, hydrogen bonding limits the mobility of the polymer chains and decreases the interstitial distances. Although the obtained results are promising, and selectivities are higher than those corresponding to pristine Matrimid
®, Matrimid
®/ZIF or even some pristine PBI and PBI/ZIF, they are still beneath the Robeson upper bound (2008). The chemical modification of Matrimid
® carboxyl and PBI N–H groups was also proposed by Hosseini et al. (2008) to enhance Matrimid
®/PBI membrane performance [
45]. A crosslinking procedure was developed evaluating
p-xylene dichloride and
p-xylene diamine that reacts with PBI N–H groups and Matrimid
® amide groups, respectively. The modification process resulted in a sharp decrease in the permeabilities of all gases but selectivities of hydrogen improved, especially at prolonged treatment times. Gas transport performance was better by chemical modification employing
p-xylene dichloride and confirmed by XRD analysis in which the interstitial space shifted from the original value.
Yáñez et al. (2021) reported that working at near-ambient temperatures (35 °C), ULTEM
® 1000B polyetherimide (PEI) and Ultrason
® E polyethersulfone (PES) exhibited better performance than Celazole
® PBI (
Table 5) in both pure and mixed gas permeation tests, surpassing the prior Robeson upper bound (1991) in the case of H
2/N
2 and H
2/CH
4 mixtures, but far from the performance displayed by Matrimid
®/ZIF and PBI/ZIF. No competitive sorption among hydrogen and nitrogen, carbon monoxide or methane was described apart from when carbon dioxide was present [
69]. Other polyimides and modified polysulfones, even incorporating ZIF-302 and ZIF-8, respectively, displayed lower selectivities and permeabilities than those described for pristine Matrimid
® and PBI (
Figure 8 and
Table 5) [
84,
85,
97]. However, by applying the proper preparation conditions, high hydrogen permeabilities could be achieved by the polyimide membrane, reaching selectivities of H
2/N
2 and H
2/CH
4 that surpass the Robeson upper bound (2008), although H
2/CO
2 selectivity was lower than 2.5, compromising the final target of separating hydrogen from carbon dioxide [
84]. Polyimide membranes based on 3,3′,4,4′-Benzophenonetetracarboxylic dianhydride monomer displayed attractive selectivities up to 9.8 but lowering permeability down to 1.7 Barrer [
98]. The employment of polyamide on P84
® polyimide support allowed the obtention of H
2/CO
2 selectivities up to 8.4 and hydrogen permeances around 988 GPU, whereas P84
®/ZIF-8 nanocomposite membranes reached a selectivity of 18.1 at 180 °C, 0.4%(
w/v) ZIF-8 and 6 bar without transmembrane total pressure differences [
88]. The polymer of intrinsic microporosity PIM-EA(H2)-TB enabled the achievement of H
2 permeability of 1630 Barrer, but H
2/CO
2 and H
2/N
2 selectivities remained under 1.5 and 26, respectively [
63]. Its blending with Matrimid
® resulted in 328 Barrer for hydrogen but poor selectivities (1.7 and 48, respectively). Additionally, in both the pure polymer and blend, the prior Robeson line was surpassed, and the upper one was reached for H
2/CH
4 separation (
Figure 8).
Finally, it is worthwhile to compare how other fillers combined with Matrimid
® performed hydrogen recovery. Silicate, SAPO-34, MgO, ZSM-5 and MOF-5 are some of the fillers that improved pristine Matrimid
® effectiveness. Silicate and SAPO-34 are microporous materials with pore sizes equal to 5.5 and 3.8 Å, respectively, higher than the corresponding to ZIF-8 (3.4 Å) [
48]. Matrimid
®/silicate and Matrimid
®/SAPO-34 MMMs with uncalcined or calcined fillers, according to SEM analysis, seemed to be homogenously clustered (silicate) and dispersed (SAPO-34), whereas two different distribution patterns were detected for ZIF-8: clustering and homogeneous dispersion. Gas transport through Matrimid
®/uncalcined filler MMMs occurred only through interfacial voids between Matrimid and the filler, and the Matrimid
® free volume caused by the chain mobility, leading to permeabilities lower than those obtained in pristine Matrimid
®, but higher than the calcined ones (
Table 4 and
Table 5). In terms of selectivity, H
2/CO
2 selectivity slightly decreased for uncalcined and calcined silicate and SAPO-34; H
2/N
2 selectivity slightly increased for calcined fillers but drastically decreased for uncalcined fillers; and H
2/CH
4 selectivity diminished gradually by using calcined fillers and sharply when uncalcined fillers were selected. Both materials led to a recovery performance poorer than ZIF-8 that was explained considering the SEM analysis in which a better integration of ZIF-8 in the MMM was observed. The reduction in gas permeability in uncalcined fillers was related to their nonporous nature, diminishing the availability of Matrimid
® free volume.
According to Hosseini et al. (2007), the employment as a filler of MgO resulted in an homogeneous dispersion of MgO in Matrimid
® and contributed to the enhancement of H
2/CO
2 selectivity and gas permeabilities, especially at the highest MgO load (40 wt.%), albeit H
2/N
2 and H
2/CH
4 selectivities suffered a diminishment [
56]. This increase, like other MMMs already described, was associated with the interfacial microvoids between the polymer and the filler. Nevertheless, the large size of MgO pores (40 Å) hindered the selectivity towards specific gas molecules. Two strategies were followed to improve Matrimid
®/MgO results: heat treatment and silver treatment. By heat treatment, two behaviors were described, i.e., at 240 °C, there was a diminishment in permeabilities and selectivities, whereas at 350 °C, the permeabilities increased (
Table 5). This was attributed to the arrangement of the polymer chains, diminishing the free volume below T
g and increasing above
Tg due to the higher mobility. A positive influence of the rapid cooling once temperature was applied for enough time was also detected, owing to the instantaneous freezing of free volume. MMMs with MgO treated with silver and annealed at 150 °C displayed worse permeabilities than Matrimid
®/MgO (20 wt.%) but higher selectivity towards hydrogen, especially enlarging the treatment time. This behavior was explained by the size exclusion mechanism caused by silver ions. Despite the improvements achieved with MgO and its modification, the results were still below the prior Robeson upper bound (2008) and, therefore, the performance was worse than that obtained by using ZIF-based MMMs (
Figure 8).
On the other hand, ZSM-5 and MOF-5 constitute examples of fillers that contribute to better hydrogen recovery from gas streams containing nitrogen and methane but did not have any effect on, or even worsened, the hydrogen/carbon dioxide separation (
Table 5) [
49,
87]. Increasing ZSM-5 content resulted in a higher chain rigidification and due to the narrow pore size, the gas diffusion slowed, and thus, permeabilities diminished. No competitive transport was detected when working with gas mixtures (Y. Zhang et al., 2008). However, the opposite phenomena occurred with MOF-5 where higher loads led to higher permeabilities, while selectivities remained constant, which could be explained by the porosity of the MOF-5 [
49,
87].
From all the information gathered here, the enormous efforts made and the need for a parameter study to further enhance the performance of Matrimid and PBI membranes can be highlighted, but undoubtedly, both have shown magnificent separation properties, providing a new generation of material combinations whose yield is above the Robeson upper bound stablished in 2008.
2.5.2. Hollow Fiber Membranes
Although more information can be found for H
2/N
2, O
2/N
2, CO
2/N
2 and CO
2/CH
4 gas separation by polymeric HFMs, the study of H
2/CO
2 is still limited, perhaps owing to the relatively poor H
2/CO
2 selectivity already explained and caused by diffusion–solubility competition and the possible CO
2-induced plasticization [
99]. Consequently, different strategies have been proposed to improve the membrane performance and will be analyzed in this section. As previously discussed, the combination of Matrimid
® with PBI may achieve better gas separation properties, exploiting the advantages of both polymers and overcoming the drawbacks attributed to their use.
During the preparation of the Matrimid
®/PBI blend HFM, despite the appropriate cross-sectional circularity and concentricity, a nonporous structure was obtained [
68,
100]. The addition of a nonsolvent, such as methanol, led to a sponge-like structure with a small pore size and an extremely small inner fiber diameter. Furthermore, the mechanical stability was poor, identifying its brittleness when dense films were fabricated using Matrimid
®/PBI blend.
In this sense, an attractive alternative reported by Hosseini et al. (2010) is the synthesis of dual-layer hollow fibers using polysulfone as an inner-layer that confers a good mechanical support and an asymmetric structure that decreases the gas transport resistance [
99]. The outer-layer blend is then composed of a Matrimid
®/PBI blend, and the HFM is prepared by co-extrusion using the dry-jet wet-quench spinning process depicted in
Figure 3. SEM analysis displayed two different morphologies in every HF corresponding to the asymmetric structure in the outer layer formed by spongy-like cells surrounded by a thin-selective layer without macrovoids, and the thicker inner layer presenting open cell pores and finger-like macrovoids [
99,
101]. The low failure and structural collapse susceptibility was ensured by the absence of delamination between the inner and outer layers, which was justified by the good miscibility at the molecular level between Matrimid
® and PBI (formation of hydrogen bonds) and the compatibility of the inner and outer dopes in terms of solvent interdiffusion and similar solubility parameters.
In the study reported by Hosseini et al. (2020), higher hydrogen permeances and H
2/CO
2 selectivities were obtained compared to pure Matrimid
® HFMs (
Figure 9,
Table 3 and
Table 6) [
99]. It was found that the highest hydrogen permeance occurred without air gaps (43.2 GPU) in the membrane directly coagulated by water, producing the sudden polymer chain contraction without any possibility of configuration change and resulting in the presence of a larger amount of fine pores, higher free volume and, therefore, lower selectivity. Introducing an air gap and enlarging its length led to an increase in H
2 and CO
2 and a diminishment in CH
4 permeances, and higher H
2/CO
2 and H
2/CH
4 selectivities, caused by gravity-induced chain orientations that complicate the transport of bigger molecules controlling the gas transport by the diffusion selectivity. Sample IDs C and D in the comments column in
Table 6 for the results of Hosseini et al. (2010) display the transport properties effect of allowing the free fall of the nascent fiber or spinning it with elongational draw. Elongational draw resulted in an enhanced gas separation, except for H
2/CO
2 in the coated HFM. According to the data presented by Hosseini et al. (2010), elongational drawing had a major influence on the fiber microstructure compared to air gap modification. Higher elongational draw ratios are usually associated with a higher productivity and lower HF diameter, as well as the modification of the separation performance and membrane morphology [
102].
The application of a silicon rubber coating allowed the sealing of defects in the dense-selective layer and, as a consequence, the obtention of higher selectivities (
Table 6) despite the slight decrease in permeance values, providing better performance than pristine Matrimid
® HFMs [
99]. The chemical crosslinking modification of the HFM with
p-xylylene diamine caused a sharp decline in permeance and H
2/CH
4 selectivity but positively affected H
2/CO
2 selectivity, especially at short treatment times (14.3 versus 6.8 in unmodified HFM) [
99]. These results clearly demonstrated the promotion of diffusion selectivity after the chemical modification but mainly penalizing free volumes and interstitial distances between polymer chains. On the other hand, Lau et al. [
101] reported that the vapor phase modification of PBI-Matrimid
®/polysulfone with ethylenediamine converted imide groups into amides, reducing the
d-space from 5.17 to 5.07 Å, causing a diminishment in permeance values (hydrogen permeance was reduced to the half of the original) and an increase in H
2/CO
2 selectivity (almost doubled).
The increase in the outer-layer dope flow rate led to a decrease in permeance values and H2/CO2 selectivity but an increase in H2/CH4 and CO2/CH4 selectivities, which is explained considering the obtention of thicker membranes with larger gas transport resistance.
Long-term aging studies in Matrimid
® double skin layer HFMs displayed a diminishment in the gas permeance after 30 months as a consequence of polymer chain relaxation in thin films and the densification of the polymer matrix, which causes lower free volume availability, eliciting a size-sieving effect in molecules such as H
2/N
2 that enhances selectivity [
103]. This phenomenon was observed either with or without ethylendiamine crosslinking.
Table 6.
Main bibliographical data in terms of pure gas H2, N2, CO2 and CH4 permeance (Pe) and selectivities through modified Matrimid® HFMs in which polymer has been substituted or blended, the HFM has been chemically modified and/or the filler has been substituted or functionalized.
Table 6.
Main bibliographical data in terms of pure gas H2, N2, CO2 and CH4 permeance (Pe) and selectivities through modified Matrimid® HFMs in which polymer has been substituted or blended, the HFM has been chemically modified and/or the filler has been substituted or functionalized.
Ref. | T (°C) | ΔP (Bar) | PeH2 (GPU) | PeN2 (GPU) | PeCO2 (GPU) | PeCH4 (GPU) | αH2/N2 | αH2/CO2 | αH2/CH4 | Polymer or Ceramic Material | Comments |
---|
Berchtold (2016) [104] | 250 | 1.4 | 118 | | 4.9 | | | 24.0 | | PBI/polysulfone | Feed pressure influence |
6.9 | 110 | 4.8 | 23.0 |
10.3 | 120 | 5.2 | 23.0 |
13.7 | 120 | 5.7 | 21.0 |
225 | | 92 | 4.1 | 22.4 | Temperature influence |
250 | 116 | 5.3 | 22.0 |
300 | 198 | 10.5 | 18.8 |
350 | 285 | 15.9 | 17.9 |
Dahe (2019) [105] | 250 | 1.4 | 9.7 | 0.4 | 0.6 | | 24.1 | 17.1 | | PBI | HFM-1 21.3% PBI (acetone); % outer coagulant 0.5 v.% water (acetone) |
21.0 | 1.1 | 1.5 | 18.4 | 14.0 | HFM-1 20.0% PBI (acetone); % outer coagulant 2.0 v.% water (acetone) |
7.6 | 0.1 | 0.3 | 62.0 | 22.4 | HFM-1 21.5% PBI (acetone/ethanol 15/85); % outer coagulant 2.0 v.% water (acetone) |
Etxebarría (2020) [106] | 150 | 7.0 | 65 | | 3.7 | | | 17.6 | | PBI | no fillers |
107 | 6.6 | 16.1 | 10 wt.% ZIF-8 |
Hosseini (2010) [99] | 35 | H2: 3.5 other gases: 10 | 43.2 | | 7.3 | 1.46 | | 5.9 | 29.6 | Matrimid®/PBI | A before silicone rubber coating |
30.3 | 4.9 | 3.54 | 6.2 | 8.6 | C before silicone rubber coating |
36.5 | 5.5 | 2.13 | 6.7 | 17.2 | X before silicone rubber coating |
38.7 | 5.7 | 1.85 | 6.8 | 20.9 | Y before silicone rubber coating |
31.6 | 4.4 | 0.22 | 7.2 | 141.5 | A after silicone rubber coating |
17.8 | 2.0 | 0.20 | 9.0 | 89.6 | C After silicone rubber coating |
26.5 | 2.5 | 0.27 | 10.6 | 96.9 | X After silicone rubber coating |
29.3 | 2.6 | 0.33 | 11.1 | 89.2 | Y After silicone rubber coating |
39.0 | 5.8 | 0.53 | 6.8 | 74.0 | D before silicone rubber coating |
32.7 | 4.8 | 0.12 | 6.8 | 284.0 | D after silicone rubber coating |
22.1 | 4.2 | 0.09 | 5.2 | 245.2 | B before silicone rubber coating |
18.9 | 3.0 | 0.09 | 6.4 | 222.2 | B after silicone rubber coating |
6.1 | 0.42 | 0.19 | 14.5 | 32.6 | Y crosslinking 0.5 s |
5.1 | 0.37 | 0.17 | 13.9 | 29.7 | Y crosslinking 1.0 min |
0.6 | 0.06 | 0.04 | 9.2 | 16.1 | Y crosslinking 5.0 min |
Kumbharkar (2011) [107] | 100 | 5–8 | 0.3 | | 0.046 | | | 7.2 | | PBI | |
200 | 0.6 | 0.048 | 12.9 |
300 | 1.0 | 0.046 | 21.5 |
400 | 2.6 | 0.096 | 27.1 |
Lau (2010) [101] | 35 | 1.4 | 72.6 | | 42.97 | | | 1.7 | | 6FDA-NDA/PES dual layer | Original |
12.1 | 4.05 | 3.0 | Vapor phase modification (VPM) Method A 2 min |
3.4 | 0.10 | 34.8 | VPM Method A 5 min |
27.7 | 6.88 | 4.0 | Matrimid®/PBI | Original |
18.6 | 3.42 | 5.4 | VPM Method A 2 min |
11.9 | 1.56 | 7.6 | VPM Method A 5 min |
7.1 | 1.03 | 6.9 | Torlon® | Original |
1.6 | 0.16 | 10.4 | VPM Method A 2 min |
0.1 | 0.03 | 4.8 | VPM Method A 5 min |
15.4 | 4.13 | 3.7 | 6FDA-NDA/PES dual layer | VPM Method B 2 min |
4.4 | 0.13 | 35.5 | VPM Method B 5 min |
21.7 | 3.77 | 5.8 | Matrimid®/PBI | VPM Method B 2 min |
13.8 | 1.77 | 7.8 | VPM Method B 5 min |
1.3 | 0.12 | 11.0 | Torlon® | VPM Method B 2 min |
1.0 | 0.16 | 6.4 | VPM Method B 5 min |
Naderi (2019) [108] | 25 | 7.0 | 2.36 | | 0.46 | | | 5.1 | | Dual layer Inner layer: polysulfone Outer layer: Polyphenylsulfone/PBI | HSP-0: PBI/DMAc/LiCl 22/79.8/1.2 (wt.%). Before silicon rubber coating |
5.50 | 1.22 | 4.5 | HSP-5: (PBI/sPPSU 95:5)/DMAc/LiCl 22/79.8/1.2 (wt.%). Before silicon rubber coating |
7.52 | 1.75 | 4.3 | HSP-10: (PBI/sPPSU 90:10)/DMAc/LiCl 22/79.8/1.2 (wt.%). Before silicon rubber coating |
8.78 | 2.53 | 3.5 | HSP-20: (PBI/sPPSU 80:20)/DMAc/LiCl 22/79.8/1.2 (wt.%). Before silicon rubber coating |
1.54 | 0.25 | 6.2 | HSP-0 after silicon rubber coating |
3.39 | 0.74 | 4.6 | HSP-5 after silicon rubber coating |
6.14 | 1.42 | 4.3 | HSP-10 after silicon rubber coating |
7.44 | 2.14 | 3.5 | HSP-20 after silicon rubber coating |
7.6 | 1.4 | 5.5 | HSP-10-40 thermal treatment 40 °C |
7.8 | 1.3 | 6.2 | HSP-10-80 thermal treatment 80 °C |
7.6 | 1.1 | 6.8 | HSP-10-120 thermal treatment 120 °C |
5.0 | 0.7 | 7.3 | HSP-10-120 chemical crosslinking 3% DBX |
3.4 | 0.5 | 6.6 | HSP-10-120 chemical crosslinking 6% DBX |
30 | 14.0 | 13.8 | 2.4 | 5.8 | Mixed gas. HSP-10-120-30 |
60 | 26.1 | 4.4 | 5.9 | Mixed gas. HSP-10-120-60 |
90 | 35.6 | 5.7 | 6.3 | Mixed gas. HSP-10-120-90 |
30 | 6.4 | 1.1 | 6.1 | Mixed gas. HSP-10-3%DBX-120-30 |
60 | 11.3 | 1.5 | 7.4 | Mixed gas. HSP-10-3%DBX-120-60 |
90 | 16.7 | 1.7 | 9.7 | Mixed gas. HSP-10-3%DBX-120-90 |
180 | 32.1 | 2.2 | 14.9 | Mixed gas. HSP-10-3%DBX-120-180 |
Pan (2012) [109] | 22 | 1.0 | 4598 | 418 | 1194 | 358 | 11.0 | 3.9 | 12.8 | ytria-stabilized zirconia | ZIF-8 |
Singh (2014) [110] | 250 | | 540.0 | 9.3 | 28.4 | | | 58.0 | 19.0 | PBI | |
150.0 | 1.3 | 5.8 | | 120.0 | 26.0 |
Villalobos (2018) [111] | 35 | | 0.05 | | 0.01 | | | 4.8 | | PBI | Pristine |
45 | 0.07 | 0.01 | 5.0 |
60 | 0.09 | 0.02 | 5.3 |
22 | 29.0 | 4.14 | 7.0 | 0.05 M Pd NPs |
35 | 34.0 | 4.47 | 7.6 |
45 | 40.0 | 4.71 | 8.5 |
60 | 80.0 | 8.00 | 10.0 |
22 | 0.55 | 0.06 | 9.0 | 0.1 M Pd NPs |
35 | 1.0 | 0.12 | 8.5 |
45 | 1.0 | 0.12 | 8.3 |
60 | 1.65 | 0.21 | 8.0 |
Wang (2016) [112] | 20 | 2.5 | 2493.3 | 886.8 | 343.4 | | 2.8 | 7.3 | | Silicon nitride ceramic | ZIF-8 |
Yang (2012) [90] | 25 | 3.5 | 1.3 | | 0.3 | | | 5.0 | | Dual layer: inner Matrimid®; outer PBI/ZIF-8 | PZM00-MA 0% ZIF-8. Solvent-exchange: methanol. Single gas |
0.8 | 0.1 | | | 6.2 | | PZM00-MB 0% ZIF-8. Solvent-exchange: methanol. Single gas |
0.8 | 0.1 | | | 7.0 | | PZM00-MC 0% ZIF-8. Solvent-exchange: methanol. Single gas |
1.7 | 0.2 | | | 7.7 | | PZM00-IA: 0% ZIF-8. Solvent-exchange: isopropanol. Single gas |
2.1 | 0.3 | | | 6.2 | | PZM00-IB: 0% ZIF-8. Solvent-exchange: isopropanol. Single gas |
1.8 | 0.2 | | | 8.2 | | PZM00-IB: 0% ZIF-8. Solvent-exchange: isopropanol. Single gas |
6.6 | 1.7 | | | 3.9 | | PZM10-MA 10% ZIF-8. Solvent-exchange: methanol. Single gas |
0.9 | 0.1 | | | 6.6 | | PZM10-MB 10% ZIF-8. Solvent-exchange: methanol. Single gas |
1.5 | 0.4 | | | 3.8 | | PZM10-MC 10% ZIF-8. Solvent-exchange: methanol. Single gas |
13.3 | 2.1 | | | 6.3 | | PZM10-IA: 10% ZIF-8. Solvent-exchange: isopropanol. Single gas |
8.9 | 0.9 | | | 9.5 | | PZM10-IB: 10% ZIF-8. Solvent-exchange: isopropanol. Single gas |
13.2 | 2.4 | | | 5.5 | | PZM10-IB: 10% ZIF-8. Solvent-exchange: isopropanol. Single gas |
8.9 | 3.7 | | | 2.4 | | PZM20-MA 20% ZIF-8. Solvent-exchange: methanol. Single gas |
21.0 | 4.6 | | | 4.6 | | PZM20-MB 20% ZIF-8. Solvent-exchange: methanol. Single gas |
57.4 | 12.4 | | | 4.6 | | PZM20-MC 20% ZIF-8. Solvent-exchange: methanol. Single gas |
28.3 | 8.2 | | | 3.5 | | PZM20-IA: 20% ZIF-8. Solvent-exchange: isopropanol. Single gas |
32.2 | 6.4 | | | 5.0 | | PZM20-IB: 20% ZIF-8. Solvent-exchange: isopropanol. Single gas |
66.8 | 14.5 | | | 4.6 | | PZM20-IB: 20% ZIF-8. Solvent-exchange: isopropanol. Single gas |
36.0 | 21.5 | | | 1.7 | | PZM33-MA 33% ZIF-8. Solvent-exchange: methanol. Single gas |
248.9 | 77.5 | | | 3.2 | | PZM33-MB 33% ZIF-8. Solvent-exchange: methanol. Single gas |
497.6 | 152.4 | | | 3.3 | | PZM33-MC 33% ZIF-8. Solvent-exchange: methanol. Single gas |
22.7 | 7.6 | | | 3.0 | | PZM33-IA: 33% ZIF-8. Solvent-exchange: isopropanol. Single gas |
34.9 | 8.7 | | | 4.0 | | PZM33-IB: 33% ZIF-8. Solvent-exchange: isopropanol. Single gas |
32.0 | 5.8 | | | 5.5 | | PZM33-IB: 33% ZIF-8. Solvent-exchange: isopropanol. Single gas |
25 | 6.0 | 3.0 | 0.6 | | | 4.8 | | PZM10-IB, 10% ZIF-8. Mixed gas |
35 | 5.0 | 0.9 | 5.8 |
50 | 8.0 | 1.0 | 8.0 |
80 | 12.0 | 1.4 | 8.5 |
120 | 22.0 | 2.1 | 10.7 |
145 | 37.0 | 3.1 | 11.8 |
180 | 45.0 | 3.7 | 12.2 |
25 | 26.0 | 14.4 | 1.8 | PZM20-IB 20% ZIF-8. Mixed gas |
35 | 30.0 | 15.0 | 2.0 |
50 | 40.0 | 16.0 | 2.5 |
80 | 58.0 | 14.5 | 4.0 |
120 | 76.0 | 13.6 | 5.6 |
145 | 99.0 | 15.2 | 6.5 |
180 | 123.0 | 14.8 | 8.3 |
25 | 36.0 | 16.4 | 2.2 | PZM33-IB 33% ZIF-8. Mixed gas |
35 | 34.0 | 14.8 | 2.3 |
50 | 40.0 | 13.3 | 3.0 |
80 | 65.0 | 14.8 | 4.4 |
120 | 100.0 | 17.5 | 5.7 |
145 | 145.0 | 20.7 | 7.0 |
180 | 201.0 | 25.8 | 7.8 |
Zhu (2018) [113] | 35.0 | 5.0 | 63.3 | 0.5 | 12.2 | | 132.0 | 5.2 | | | Pure |
172.2 | 1.8 | 36.5 | 94.1 | 4.7 | Ultem® polyetherimide | 15% MIL-53 |
127.1 | 0.9 | 31.4 | 144.5 | 4.1 | | 15% S-MIL-53 |
Although no permeation results were provided, Li et al. (2004) presented an interesting and thorough analysis of co-extruded Matrimid
®/PES dual-layer HFMs and the main variables affecting the HFM morphology and performance [
102]. As explained for Matrimid
®/PBI/polysulfone, Matrimid
®/PES did not present delamination at the interface of the dual layer, displaying, in SEM analysis, an outer dense layer and an inner porous surface. The dense macrovoid-free layer formed by Matrimid
® remained, even when changing the inner-layer dope composition and spinning conditions, as occurred in the study of David (2012) [
100]. This phenomenon has been associated with the high viscosity and poor fluidity of Matrimid
® when in contact with nonsolvents, and the structure-transition thickness, defining the critical structure-transition thickness as the value that displays a change from a finger-like (above critical thickness) to a sponge-like structure (below critical thickness). The value of this variable depends on certain parameters, such as dope and membrane formulation and the membrane materials used.
A good tradeoff between hydrogen permeation (2493 GPU) and H
2/CO
2 selectivity (7.3) was obtained by a continuous ZIF-8 membrane on the outer surface of silicon nitride HF (
Figure 9 and
Table 6), resulting in one of the highest permeances with a selectivity higher than those presented for pristine Matrimid
® dense and HF membranes, but not for nitrogen (H
2/N
2 selectivity equals to 2.8) [
112]. These results were attributed to the blocking effect caused by the adsorbed CO
2 that stablish cooperative interactions with other carbon dioxide molecules, but not with hydrogen or nitrogen.
In HFMs fabricated with PBI, a clear influence of bore fluid composition was reported by Kumbharkar et al. (2011) [
107]. When bore fluid was composed by a water/
N,
N-dimethylacetamide (DMAc) ratio of 10/80 (wt.%), carbon dioxide was not detected in the permeate stream, and very low hydrogen permeance (1.4 GPU,
Figure 9 and
Table 6) was measured as a consequence of the sublayer resistance caused by the skin layer formation on the inner surface of the HF. It could be overcome by increasing the DMAc ratio up to 50 wt.%, increasing from 0.3 and 7.2 (100 °C) to 2.6 and 27.1 (400 °C) hydrogen permeance (GPU) and H
2/CO
2 selectivity, respectively. This selectivity entails one of the highest gathered during the elaboration of this review, and it is a consequence of the rigidity and the defect-free HFM synthesized, but the main challenge is being faced with the low permeance value. Furthermore, considering the thickness measured from SEM analysis, the permeability increased up to 22.9 Barrer (selectivity 27.1), which places this HFM in the separation attractive area, above the Robeson upper bound (2008), improving the performance observed for many flat sheet membranes. For a matter of comparison,
Figure 10 includes the permeabilities described in previous figures for flat sheet membranes and those corresponding to HFMs when authors provided approximated thickness from SEM and TEM images.
Consequently, higher hydrogen permeances are required and implied some strategies, such as the minimization of the dense selective layer thickness, minimization of gas phase resistance and development of high porous inner surfaces. In order to obtain high selectivities, the selective layer needs to be defect free, and macrovoids in the support layer need to be minimized. In an interesting study by Singh et al. (2014), the authors prepared a HFM with PBI selective layers between 160 and 2180 nm, after defect sealing, and subsequently carried out permeation tests operating at syngas temperatures (250 °C) close to water–gas shift reactors [
110]. Attractive hydrogen permeances higher than 500 GPU and H
2/CO
2 and H
2/N
2 selectivities equal to 19 and 58, respectively, were obtained for thinner selective layers (
Figure 9 and
Table 6). Membranes with a thicker selective layer yield a hydrogen permeance value and H
2/CO
2 and H
2/N
2 selectivities of 150 GPU, 26 and 120, respectively. In a more recent work from the same research group, a nonsolvent chemistry sensibility assessment was developed taking into consideration the solvent/nonsolvent solubility parameters, which include the contribution of dispersion, polar and H-bonding forces, and solvent/nonsolvent diffusion [
105]. SEM images of PBI HFMs fabricated using PBI/LiCl/DMAc displayed two different microstructures due to the diffusion features—nonsolvent dope solution: highly porous membranes were obtained using as nonsolvent acetone, ethanol, isopropanol and methanol, whereas dense HFMs were obtained by employing butyl acetate, ethyl acetate, hexane, toluene, water and xylene. From the aforementioned nonsolvents, a desired inner porous microstructure was achieved by using methanol, ethanol and acetone. The evaluation of the performance modifying the composition of acetone as bore fluid by adding ethanol and the effect of employing a water–acetone mixture as outer coagulant was conducted by obtaining hydrogen permeances and H
2/CO
2 and H
2/N
2 selectivities in the range of 7.6–21.0 GPU, 17.1–22.4 and 18.4–62.0, respectively, confirming the good performance of PBI HFMs (
Figure 9 and
Table 6).
As a consequence of the extensive research effort to obtain the best PBI HFM output, Berchtold et al. (2018), who authored the research studies of PBI HFMs analyzed here, patented a method for producing asymmetric hollow fiber membranes whose main results, taking hydrogen recovery from a syngas stream as an example, are summarized in
Figure 9 and
Table 6. Hydrogen permeances and H
2/CO
2 selectivity up to 285 GPU and 24, respectively, led these membranes to the best tradeoff and stresses the potentiality of polymeric membranes in the field of gas separation.
An additional step would consist of the addition of fillers to the PBI HFM. Considering that the selective layer thickness is normally below a micrometer, the dimension of the fillers must be smaller, with the aim of avoiding the presence of defects and particle penetration [
90]. In the work developed by Yang et al. (2012), a HFM was prepared selecting Matrimid
® as the inner layer material and PBI/ZIF-8 as the outer layer material, co-extruded through a triple-orifice spinneret by a dry-jet/wet-quench spinning process. Material selection and disposition were considered, taking into account several factors: the brittleness of PBI membranes synthesized by a nonsolvent phase-inversion procedure that may be overcome by selecting a strong material as the inner layer; the gas transport resistance of the inner layer may be diminished by selecting a support layer material with high permeability; the employment of a polymeric material in the support layer counteracts the relative high cost of PBI/ZIF-8; the use of miscible polymers avoids the delamination phenomenon; higher permeabilities may be obtained in a single-step extrusion [
38,
90,
99]. The employment of isopropanol in the post-treatment solvent exchange led to a better performance than methanol (
Figure 9 and
Table 6) [
90]. The analysis of the ZIF-8 composition effect displayed maximum H
2/CO
2 selectivities for 10 wt.% (9.5), different to that obtained for flat sheet membranes (13.0) (
Table 5 and
Table 6) for 15 wt.%. The results were justified considering the intercalation of filler NPs, the defects induced during the spinning of highly concentrated solutions and the possible formation of an interface between both layers. Despite the good performance of several HFMs prepared by these authors compared to pristine PBI/Matrimid
® and other HFMs from the bibliography, depicted in
Figure 9, the effectiveness is not as good as expected considering the addition of selective fillers. However, it should be considered that the HFM synthesis process was not the optimum process as posterior research studies showed; therefore, it is likely that better performances could be obtained from the best available synthesis method. From the comparison between single gas and H
2:CO
2 50:50 experiments, competitive gas transport was confirmed, displaying a decrease in hydrogen permeance and H
2/CO
2 selectivity.
Villalobos et al. (2018) suggested that the incorporation of palladium into a PBI matrix for the extrusion of a HFM could enhance the inherent mechanical instability of Pd and the low permeability of PBI [
111]. The nanometric size of Pd and the buffering by the polymer reduce the stress. Spinning process optimization was carried out by using Cu
2+ that is cheaper than Pd and also leads to stable complexes with PBI imidazole groups. With the goal of providing an adequate dispersion of Pd in PBI, the reduction of Pd ions to NPs was carried out by immersing the HF in a freshly prepared NaBH
4 solution. From TEM analysis, the agglomeration of Pd nanoparticles was confirmed, especially in the region close to the surface. Nevertheless, good H
2/CO
2 selectivity was reported (10) with a corresponding hydrogen permeance of 80 GPU (
Figure 9 and
Table 6), displaying a performance far from the optimized PBI HFM but with a good tradeoff permeance/selectivity, at least taking into account that results were obtained at low temperatures. Considering the dense layer thickness of these membranes, for the best-performance HFM (selectivity of 10), the permeability reached 176 Barrer, surpassing the Robeson upper line (2008).
Etxebarría-Benavides et al. (2020) directly worked with PBI/ZIF-8 membranes. For comparison, the optimal PBI content in dope solution was decreased for MMHFM preparation to obtain similar viscosities in the pristine PBI and PBI/ZIF-8 mixture [
106]. From SEM analysis, a densified outer layer structure, a good circularity, concentricity between the inner and outer diameter and a porous substructure presenting small pores and finger-like macrovoids were attributed to pure and PBI/ZIF-8 (10 wt.%). The addition of ZIF-8 resulted in an increase in the H
2 permeance value from 65 to 107 GPU, with H
2/CO
2 ideal selectivity incurring a slight decrease from 17.6 to 16.1 at 150 °C (
Table 6). Hydrogen permeance and selectivity obtained for H
2/CO
2 50/50 vol.% gas mixtures were lower (around 80 GPU and 13, respectively), yet higher than those obtained by Yang et al. (2012) for the same temperature and polymer/ZIF ratio, but with Matrimid
® as the inner layer and tested with mixed gas (37 GPU and 11.8). Consequently, the competitive sorption phenomenon can be assumed as the mixed gas permeation deviation displayed. Taking into account the preferential adsorption sites identified by simulation, for H
2, the highest adsorption energy is located on top of the 2-methylimidazolate ring over the C=C bond (8.6 kJ/mol), and the second adsorption site corresponds to the center of the channel of the Zn (6.2 kJ/mol), whereas CO
2 at low loading primarily occupies a site proximal to the C=C bond of 2-methylimidazolate (preferential compared to metal cluster), and at high loading occupies the site near the aperture and (at lower significance) the Zn cage center [
76,
114]. Therefore, the higher loading of carbon dioxide leads to its adsorption in the three sites, reducing the availability of adsorption sites and, therefore, reducing the permeances of both gases.
The features and performance of a dual layer HFM formed by a polysulfone inner layer and polyphenylsulfone (PPSU)/PBI outer layer were assessed in a recent work by Naderi et al. [
108], including the influence of post-treatment procedures, such as silicone rubber coating, thermal treatment and chemical modification. XPS and FTIR analysis confirmed the ionic interaction and covalent crosslinking of PBI and PPSU, whereas SEM analysis confirmed a highly porous inner layer formed by finger-like macrovoids and an ultrathin skin layer. During permeation tests, with the introduction of and increase in PPSU composition in the outer layer, hydrogen and carbon dioxide permeation values increased, while selectivity diminished (
Table 6) as a consequence of the increase in the chain–chain distance, i.e., the
d-spacing of the outer layer polymer chains increased from 4.7 (0 wt.% PPSU) to 5.2 Å (20 wt.%). Silicone rubber coating results displayed an increase in selectivity but a permeance diminishment in the HFM without PPSU, while the HFM with PPSU did not show selectivity differences, but permeance was reduced. By the annealing procedure, ionic interaction between the sulfonic acid groups of PPSU and the imine groups of PBI was enhanced, and consequently, the selectivity increased from 4.3 (without thermal treatment) to 6.8 (120 °C), whereas hydrogen permeance remained almost constant (7.52 and 7.60 GPU, respectively), and carbon dioxide permeance decreased (from 1.71 to 1.10 GPU). Naderi et al. (2019) reported in the same study that crosslinking with 3 wt.%
α,
α′-dibromo-
p-xylene resulted in a selectivity increase (7.34 compared to 6.80), but it decreased when using 6 wt.%; in both situations, hydrogen permeance dropped [
108]. An increase in crosslinker loading may decrease the number of SO
3H groups of PPSU that ionically crosslink with N sites of PBI; therefore, there is more SO
3H group availability to interact with carbon dioxide. However, an excessive crosslinker content may cause the formation of polymer chain that is too tight. After all these treatments, considering the best-performance HFM (with PPSU, uncoated, annealed at 120 °C and crosslinked by a 3 wt.% of crosslinker), CO
2-induced plasticization was avoided up to 20 bar. When working with binary gas mixtures of 50/50, H
2/CO
2 selectivities increased with permeation temperature and with the crosslinking procedure, achieving a hydrogen permeance value of 32.1 GPU and a H
2/CO
2 selectivity of 14.9 with a 3 wt.% crosslinker and permeation temperature of 180 °C. This value occupies a good position in the selectivity–permeability plot (
Figure 9), especially taking into account that it was obtained from a gas mixture, which could be interesting for application to steam reforming and syngas stream gas separation, depending on purity requirements.
It is worth mentioning that almost all the results presented for PBI and PBI-based membranes were obtained at high temperatures, mainly considering their application in gas separation in syngas processes, which, to an extent, invalidates a correct interpretation and comparison with Matrimid® and Matrimid®-based membrane performance, because, as explained thoroughly throughout this manuscript, temperature has a positive influence, in general, on hydrogen permeance and selectivity towards hydrogen.
Lau et al. (2010) evaluated, together with the Matrimid
®/PBI/polysulfone dual layer HF already analyzed, the influence of the vapor phase modification of 6FDA-NDA/PES (where polyimide was synthesized from the monomers 4,4′-(hexafluoroisopropylidene) diphthalic anhydride and 1,5-napthalenediamine) dual layer HF and Torlon
® single layer [
101]. In all of them, the imide groups are converted into amide due to the strong nucleophilicity of the ethylenediamine used in the vapor phase modification. In the 6FDA-NDA membrane,
d-spacing (5.08 Å) reduced to 4.93 Å; in Torlon
®,
d-spacing was reduced from 4.35 to 4.30 Å. As occurred with PBI-Matrimid
®/polysulfone, selectivity increased for 6FDA-NDA (from 1.69 to 34.80 or 35.52, depending on the modification method applied), although hydrogen permeance (72.59 GPU) decreased to 4.44–4.02 GPU. Nevertheless, after 1 min of treatment, Torlon
® displayed almost the same selectivity but reduced the permeability. Undoubtedly, 6FDA-NDA modification achieved the highest selectivities presented in this review for HFMs (
Figure 9), although permeance values should be improved to obtain a good tradeoff. Considering the thickness values provided by the authors, hydrogen permeabilities achieved for modified 6FDA-NDA, Matrimid
®/PBI and Torlon
® would be 20.7, 106.3 and 3.6 Barrer with selectivity values of 35.5, 7.8 and 6.4, respectively, surpassing the Robeson upper bound (2008) and even the performance of enhanced dense membranes (
Figure 10).
Recently, Zhu et al. (2018) reported that the incorporation of lab-synthesized S-MIL-53 functionalized by aminosilane grafting and subsequently incorporated Ultem
® allowed the fabrication of an MMHFM with an increase in plasticization resistance and permeance values [
113]. SEM analysis displayed that more and larger finger-like pores occurred in the outer region when increasing the S-MIL-53 content, but good filler–polymer adhesion was found. When the filler content was increased to 20 wt.%, particle agglomeration was observed, and the finger-like pores extended to the outer edge. Compared to pure Ultem
®, the addition of 15 wt.% of MIL-53 resulted in an increase in hydrogen permeance (from 63.3 to 172.2 GPU), but a decrease in H
2/N
2 selectivity (from 132 to 94.1) and in H
2/CO
2 selectivity (from 5.2 to 4.7) (
Table 6). Modified MIL-53 (S-MIL-53) provided lower hydrogen permeance (127.1 GPU) compared to the unmodified one, yet higher than the pure polymer, and higher H
2/N
2 selectivity (144.5), although H
2/CO
2 selectivity worsened (4.1), rendering these procedures unattractive for application in hydrogen recovery.
Pan et al. (2012) describe a procedure for obtaining ZIF-8 membranes supported in a ceramic yttria-stabilized zirconia HFM by a seeded growth method, which resulted in a sandwich-like structure with fingers initiating from both inner and outer sides [
109]. Hydrogen permeance of 4598 GPU was reached with a H
2/N
2 selectivity of 11.0, H
2/CO
2 selectivity of 3.9 and H
2/CH
4 of 12.8 (
Table 6). Despite the high permeation value (
Figure 9) and the corresponding permeability (
Figure 10), the selectivity is rather low compared to almost all the membrane performances presented in this review, although it surpasses the Robeson upper bound (2008). For binary mixtures, the competition between each gas mixture component resulted in a reduction in permeance and selectivity towards hydrogen.