4.1. Zeolite Membranes
The first zeolites were found by the Swedish mineralogist Cronstedt in 1756, and he named the new materials “zeolites”, which means boiling stones in Greek. After centuries of development, 229 unique types of zeolite frameworks have been identified, and 40 of them are naturally available minerals on the earth [
96,
97]. The general formula of these aluminosilicate crystalline minerals is (SiO
2)
x(AlO
2)
yzMO
xaH
2O, from which we can see that the main components of zeolites are silicate, aluminium and other metal oxide elements, such as sodium and titanium [
98].
Zeolites is an important material in the chemical engineering industry because of its unique regular pore structures and robustness under harsh conditions. Currently, zeolite is extensively employed in commercial catalysts or catalyst hosts for noble metals. In addition, zeolite is a promising adsorbentand has great potential for hydrogen storage, etc. [
99,
100]. Zeolite is recognized as an ideal membrane material for gas separation because some zeolites have the unique pore aperture sizes (0.3–1.3 nm) that are comparable to the scale of gas molecules [
101,
102]. Zeolite membranes are primarily fabricated through hydrothermal reactions within an autoclave reactor in a conventional oven or through microwave-assisted heating. There are two main strategies for zeolite membrane fabrication, in situ growth and seeded secondary growth. During in situ growth, the crystallization solution with or without organic templates is autoclaved with the porous support in a reactor. Then, the obtained green membrane is sintered to remove the organic templates to obtain the final zeolite membrane. In seeded secondary growth, nano-size zeolite particles are seeded onto the support prior to the secondary growth, which is subsequently carried out by immersing the seeded support with the fabrication solutions in the autoclave [
103,
104]. The bad news about zeolite membranes for gas separation is that although the subject has been investigated for decades, less than 20 zeolites have been successfully fabricated in membrane form, and even more unfortunately, only one zeolite membrane, NaA has been commercialized for industrial dehydration applications at this time [
105]. The number of publications on different zeolite membranes, as shown in
Figure 6, show that the most popular zeolite topologies for membranes are MFI (ZSM-5 and Silicate-1) and LTA, as these two topologies represent over half of the total publications on zeolite membranes [
103].
In the application of hydrogen separation, several types of zeolite membranes, including MFI, LTA, SAPO-34, DDR, etc., have been reported in the last decade.
Table 5 compares the kinetic diameters of several gas molecules and the aperture size of some zeolite topologies. From
Table 5, we can see that the pore sizes of these zeolites are larger than the kinetic diameters of the gases in syngas. Hence, all these zeolite membranes have relatively low hydrogen selectivity under the Knudsen diffusion mechanism, and their performance is unfavourable for hydrogen separation and purification. Therefore, a modification process for zeolite membranes is necessary to achieve acceptable hydrogen selectivity.
Chemical vapour decomposition (CVD) and chemical cracking decomposition (CCD) are widely used to improve the performance of the ZSM-5 membrane, as these modification methods can result in an impressive improvement in H
2/CO
2 selectivity [
106,
107,
108,
109]. Lin et al. post-treated ZSM-5 membranes in two steps. First, CVD was carried out to compensate for the defects by sintering tetraethyl orthosilicate (TEOS) into the inter-crystalline gaps. Subsequently, CCD was used to introduce methyldiethoxysilane (MDES) and simultaneously to sinter the MDES to SiO
2 in the pores to reduce the ZSM-5 pore size. The H
2/CO
2 selectivity of the ZSM-5 membrane was significantly improved from 4.3 to 4.92 after CVD and further to 25.3 after CCD. However, the hydrogen permeation decreased by more than one order of magnitude compared to the as-fabricated ZSM-5 membrane [
106]. In addition, that group also prepared a tubular ZSM-5 membrane with the same modification methods and used the modified ZSM-5 membrane to construct a membrane reactor for the water gas shift reaction. This membrane reactor showed good thermal stability during long-term operation at high temperature and in a humid environment [
106]. At almost the same time, Dong et al. reported similar work using a modified MFI membrane for a membrane reactor in high-temperature WGSR [
108], while Xu et al. used a modified ZSM-5 membrane-based reactor for low-temperature WGSR [
109].
In addition to MFI membranes, SAPO-34 membranes have also been extensively studied in recent years. The H
2/N
2 and H
2/CH
4 selectivity of SAPO-34 membranes were reported to be 25 and 7.4, respectively, with H
2 permeance of 2.4 × 10
−8 mol/s·m
2·Pa at 27 °C and 270 KPa. However, the H
2/CO
2 selectivity was very poor, only 1.3, under the same conditions [
110]. Interestingly, subsequent work from Noble’s group found that the SAPO-34 membrane had CO
2/H
2 selectivity greater than 100 at −20 °C due to enhanced CO
2 adsorption and inhibited H
2 adsorption at low temperature. However, the CO
2 permeation was still at a relatively low level of approximately 10
−8 mol/s·m
2·Pa [
111]. The NaA membrane was fabricated by a microwave-assisted heating method and tested for hydrogen separation from hydrocarbons. The result showed that the NaA membrane had H
2/N
2 and H
2/n-C
4H
10 selectivity of 3.18 and 11.8, respectively, with a very promising H
2 permeance of 2.13 × 10
−6 mol/s·m
2·Pa under the testing conditions [
112]. DDR (Deca-Dodecasil3R) is a highly siliceous zeolite with a small aperture pore size of 0.36 × 0.44 nm, as shown in
Table 5. Early work reported that a DDR membrane on a porous α-Al
2O
3 support had a low H
2 permeance of 1.1 × 10
−10 mol/s·m
2·Pa with H
2/CO and H
2/CO
2 selectivity of 11 and 9, respectively, at 500 °C [
113]. However, Dong et al. recently published their work on a DDR membrane modified by the liquid phase chemical deposition of tetramethoxysilane (TMOS), and their membrane showed a good CO
2/CH
4 selectivity of 92 and a CO
2 permeance of 2.1 × 10
−7 mol/s·m
2·Pa at room temperature. However, the H
2 permeation was slower than that of CO
2 after the modification due to the strong CO
2 adsorption, which is not favourable from a hydrogen separation perspective [
114]. A P-type zeolite GIS membrane was in situ fabricated on a porous α-Al
2O
3 support and tested for gas separation in Lin’s group. The GIS membrane showed a H
2/SF
6 selectivity of 102, which is far better than the Knudsen diffusion separation selectivity, and a H
2/Ar selectivity of 5.29 with a H
2 permeance of 5.71 × 10
−7 mol/s·m
2·Pa at room temperature. In addition, they reported that the GIS membrane was more stable in a humid environment during a phase transformation process at moderate temperature [
104]. An AlPO
4 membrane was prepared by Caro’s group and had an acceptable H
2/CO
2 selectivity of 11, which is greater than that of the other reported zeolite membranes because of the low CO
2 adsorption on the cation-free LTA zeolite membrane surface. In addition, the AlPO
4 membrane provided a H
2/CH
4 selectivity of 146 with a H
2 permeance of 2.63 × 10
−7 mol/s·m
2·Pa through a molecular sieving mechanism [
115].
Generally, the primary benefits of zeolite membranes for hydrogen separation and purification are their hydrothermal stability under the high-temperature and humid conditions that accompany hydrogen production reactions, such as biomass processing. In addition, their moderate to high gas permeability compared to dense metal and polymeric membranes could be another considerable factor in gas separation applications. However, the application progress of zeolite membranes for gas separation is much slower than predicted, as mentioned above [
105]. This delay may be attributed to the difficulty of scaling up zeolite membrane processing, including both the membrane manufacturing and the membrane maintenance on site. Furthermore, the cost of zeolite membranes is relatively high due to the low reproducibility of defect-free zeolite membranes. From the perspective of hydrogen separation in biomass processing, most zeolite membranes are not able to provide satisfactory H
2 selectivity against other syngas components, such as CO
2, CO, etc., because the aperture size of most zeolites is larger than the gas molecule cutoff sizes. Although post-treatment processing could greatly enhance the selective performance, these modification processes significantly increase the expense and further reduce the gas permeation of the zeolite membranes.
4.2. MOF Membranes
Metal-organic frameworks (MOFs) are a class of porous crystalline materials netted with metals or metal clusters as the vertex nodes linked by organic linkers. Apart from its high surface area and structural robustness, the most distinguishing feature of MOFs is their topology diversity with tunable pore sizes. This diversity occurs because their structures can be tailored by assembly from combinations of different metal nodes and organic linkers [
116,
117,
118]. Since the first prototype MOF-5 was introduced, a variety of MOF materials have been developed. Recently, some of them have been successfully commercialized for gas delivery and food storage. However, the mechanical strength and structural stability of MOF materials are not as good as those of zeolites, because of their coordination network structures with the organic linkers. Herein, ZIF is a subclass of MOF materials consisting of transition metals linked by imidazolate bridging linkers. In ZIF structures, the metal—imidazolate—metal bond has an angle of 145°, which is the same as the Si (Al)-
O-Si (Al) angle in zeolite structure. Therefore, ZIF materials were found to be more stable than other MOFs due to their zeolitic structures. To date, MOF and ZIF materials have been employed in various application areas, such as separation and adsorption, catalysis, drug delivery, sensors, etc., because of their adjustable porous structures and high surface areas [
119,
120,
121,
122].
An advantage of MOF membrane research is that the pre-developed facilities for zeolite membrane fabrication could be easily reproduced for preparing MOF membranes [
104]. The primary synthesis methods for MOF membranes are also in situ growth and seeded secondary growth, and the other fabrication strategies applied for zeolite membranes, for example, microwave-assisted heating fabrication and post-treatment modification methods, could also be applied to MOF membrane fabrications [
123]. However, the synthesis conditions of MOF membranes are milder than those of zeolite membranes, and the time required is shorter. For instance, ZIF-8 membrane can easily be prepared in aqueous solution at room temperature in 6 h [
124]. The similarity of fabrication to that of zeolite membranes reduced the barrier to initiating MOF membrane research and could account for the blooming development of MOF membranes in recent years.
In gas separation applications, MOF membranes, especially ZIF membranes, have attracted extensive attention in the last decade. Several studies have been published reviewing ZIF membranes for gas separation and comparing their performance with that of zeolite membranes [
105,
123,
125,
126,
127]. The publication numbers on ZIF membranes for gas separation have been organized by Koros et al. as shown in
Figure 7, where we can see that the ZIF-8 membrane is the most popular, accounting for over 70% of the total ZIF membrane publications, and almost half of these ZIF membrane publications aimed at hydrogen separation [
125]. The pore sizes of several representative MOF materials for hydrogen separation membranes are illustrated in
Table 5.
ZIF-8 was reported to have the aperture pore size of 3.4 Å, which was larger only than the kinetic diameters of hydrogen and carbon dioxide. This unique pore size led to the expectation that ZIF-8 membranes could be used to separate hydrogen from other, larger molecular gases, such as CO and CH
4. Caro’s group reported their first ZIF-8 membrane for gas separation in 2009. That ZIF-8 membrane showed a hydrogen permeance of 6.04 × 10
−8 mol/s·m
2·Pa in a single-gas permeation test at room temperature, with H
2/CO
2 and H
2/CH
4 separation factors of 4.54 and 12.58, respectively. In their study, the hydrogen permeance was relatively low due to the thick ZIF-8 membrane layer, greater than 30 µm thick. In addition, they claimed that the large CH
4 molecule (kinetic diameter of 3.8 Å) could permeate through the membrane because the ZIF-8 structure was more flexible than its natural network during gas permeation, thus resulting in a molecular cutoff of the ZIF-8 membrane that was larger than its ideal pore size [
128]. Since then, many studies have been undertaken to investigate the experimental parameters for controlling the MOF and ZIF crystallization process and, in turn, the membrane morphologies [
129,
130]. At the same time, studies were also carried out to investigate the gas permeation transportation mechanism through MOF membranes. Koros’s group proposed that the effective aperture size of ZIF-8 was 4.0–4.2 Å, a conclusion drawn from its sharp cutoff separation performance for propylene (4.0 Å) and propane (4.2 Å) [
131]. Lai et al. developed a rapid synthesis method for ZIF-8 nanocrystals in an aqueous system, and subsequently, a ZIF-8 membrane was also fabricated from an aqueous system at room temperature in their group [
132,
133]. The fabricated membrane was tested for propylene and propane separation for the first time, and a very promising application opportunity for ZIF-8 membrane was found because the propylene/propane separation factor was greater than 50 [
124]. Later, ZIF-8 membranes began to be fabricated on hollow fibre supports, and the thickness of the membrane layer was further reduced to approximately 1 µm. Additionally, the hydrogen permeance was significantly improved to 15.4 × 10
−7 mol/s·m
2·Pa at room temperature, with H
2/CO
2 and H
2/CH
4 separation factors of 3.85 and 11, respectively [
134]. Recently, an attempt was made to grow a ZIF-8 membrane on hollow fibre polymeric supports to reduce its fabrication expense by blending the ZIF-8 seeds into the polymer support layers followed by growing the membrane via secondary growth [
135].
Compared with ZIF-8 membranes, ZIF-7 and ZIF-22 membranes could provide better H
2/CO
2 separation factors because of their pore size of 3.0 Å, which lies between the kinetic diameters of hydrogen and carbon dioxide, as shown in
Table 5. Li et al. fabricated a ZIF-7 membrane on an alumina plate support through the secondary growth method. That ZIF-7 membrane had H
2/CO
2 and H
2/N
2 separation factors of 13.0 and 20.7, respectively, in a single-gas permeation test at 220 °C, with H
2 permeance of 4.55 × 10
−8 mol/s·m
2·Pa. In addition, the ZIF-7 membrane structure was found to be stable after operation under simulated WGSR conditions at 220 °C for 50 h [
136]. Additionally, the ZIF-22 membrane fabricated in Caro’s group showed H
2/CO
2 and H
2/N
2 separation factors of 8.5 and 7.1, respectively, at 323 K, with hydrogen permeance of 2.02 × 10
−7 mol/s·m
2·Pa [
137]. Caro’s group also developed a steam-stable ZIF-90 membrane with a thickness of 20 µm on an alumina support for hydrogen separation [
138]. The ZIF-90 membrane had a hydrogen permeance of 2.5 × 10
−7 mol/s·m
2·Pa at 473 K, with H
2/CO
2 and H
2/N
2 separation factors of 7.2 and 12.6, respectively. In addition, the membrane was tested under a steam environment for 24 h, and the H
2/CH
4 separation factor was approximately 15 and was kept stable during long-term operation. MOF-5 and HKUST-1 membranes were also fabricated. However, they had poor hydrogen separation with only Knudsen diffusion selectivity due to the larger pore aperture sizes in these two framework structures [
139,
140].
In conclusion, MOF membranes are still in a period of fast development, and more emerging structures are being identified at this time. The structural diversity and tunable pore size are the most attractive properties making MOF membranes a promising candidate for many potential gas separations. As a result, the gas separation performance of MOF membranes is dramatically greater compared with that of other membranes, such as polymer and dense metal membranes. As shown in
Figure 8, the H
2/CO
2 separation performance of all the ZIF membranes was predicted to lie in the right upper side of the Roberson Upper Boundary [
119]. In addition, the large-scale manufacturing of MOF membranes is not considered difficult because MOF membranes could be fabricated under the relatively mild conditions of an aqueous system at ambient temperature and pressure. However, MOF membranes also face some critical challenges before they can be used in practical industry processes, such as biomass processing for hydrogen production. First, MOF membranes suffer from low hydrothermal stability at high temperatures and in humid environments. Although the MOF and ZIF materials were initially considered to be thermally and chemically stable in various environments [
122], recent studies have noted that the stability of these MOF materials and their membranes was not as good as we expected, especially under harsh reaction conditions for long-term operation [
141,
142,
143]. Second, although MOF membranes possess acceptable overall hydrogen separation performance, the gas permeances are still in the range of 10
−8–10
−6 mol/s·m
2·Pa, which is still relatively low compared with industrial demands. To address this low permeance limitation, efforts have been made to fabricate an ultra-thin membrane with a selective layer several nanometers thick. Yang et al. recently prepared a ZIF-7 membrane with nanometre thickness by using 1-nanometer sheets as the building block, and the membrane showed ultra-high hydrogen permeability and enhanced H
2/CO
2 selectivity [
144]. Another challenge is to improve the reproducibility of the gas separation performance of the fabricated MOF membranes. The low reproducibility of these well-integrated crystalline membranes (MOF and zeolite) is the universal challenge from both the research laboratory and manufacturing perspectives. Therefore, the development of reliable fabrication processes with high product reproducibility is another prerequisite for the industrialization of these membranes.
4.3. Other Microporous Membranes
The silica membrane is the most famous amorphous microporous membrane and has been investigated for hydrogen purification for many years. One advantage is that the pore structure of the silica membrane is controllable, and gas separation performance can correspondingly be adjusted [
145]. Sol-gel and chemical vapour deposition are the two main methods for silica membrane fabrication. Silica membranes prepared by sol-gel methods often have relatively low selectivity but good permeability, and the pore size of the membrane can easily be adjusted by changing the fabrication conditions. However, this sol-gel process has low reproducibility, and it is difficult to achieve consistent performance [
146]. In contrast, silica membranes fabricated from CVD can provide high hydrogen selectivity but correspondingly low permeability. However, the cost of the CVD method is high, and the method is difficult to scale up for commercial manufacturing [
145]. Verweij et al. reported their work on high-selectivity high-flux silica membrane fabrication. Their silica membrane was fabricated from optimized conditions. and it possessed hydrogen permeance of 5 × 10
−6 mol/s·m
2·Pa at 200 °C in single-gas permeation testing, coupled with H
2/CO
2 and H
2/CH
4 selectivity of 71 and over 4000, respectively [
147]. The most crucial concern in employing silica membranes is the instability at high temperature in the presence of water, because such membranes lose permeability and selectivity rapidly after exposure to high-temperature steam due to densification of the pore structures [
94]. To date, the most efficient way to overcome this instability is to dope a metal or metal oxide into a silica membrane. Tsuru et al. reported a cobalt-doped silica membrane with hydrogen permeance of 1.8 × 10
−7 mol/s·m
2·Pa and H
2/N
2 selectivity of approximately 730 at 500 °C. This membrane performed stable gas separation for 60 h at 500 °C in the presence of steam [
148]. Moreover, silica-based membrane reactors have been studied for the low-temperature water gas shift reaction, and the utilization of this membrane reactor was able to shift the limited reaction equilibrium and correspondingly significantly improve the CO conversion and the hydrogen recovery [
149,
150]. However, the low hydrothermal stability of silica membranes also seriously strained the operation duration of these membrane reactors.
In addition, amorphous metal oxide membranes and carbon membranes have also been developed for hydrogen separation, and both of these membrane types presented some advantages for hydrogen separation. For example, higher H
2 selectivity has been observed for carbon membranes because of the molecular sieving mechanism, while metal oxide membranes usually provide high permeance. Correspondingly, the weaknesses of these membranes are very obvious, namely, low permeability for carbon membranes and low selectivity for metal oxide membranes. These shortages critically hinder the further application of these membranes in the industry [
94,
151].