Microstructure engineering of silica-derived membranes and their applications in molecular separation

Abstract

Silica-derived membranes have gained considerable attention as highly promising molecular separation materials owing to their unique molecular sieve-like porous structure, tunable chemical properties, and outstanding chemical stability. Their potential has already been demonstrated in industrial applications, particularly in solvent dehydration processes, highlighting their commercial viability. This review provides a detailed overview of recent advancements in silica-derived membrane technology for molecular separation. First, we discuss the sol–gel process employed in the fabrication of amorphous silica membranes, with a focus on the resulting structural characteristics and the factors influencing membrane performance. Then, we outline various microstructural engineering strategies designed to enhance hydrothermal stability and separation efficiency under challenging conditions. Additionally, we highlight cutting-edge silica-derived membranes that exhibit outstanding separation performance in gas separation, pervaporation, and reverse osmosis. Finally, we provide insights into the development of silica-derived membranes, emphasizing ongoing challenges and opportunities for further innovation in this field. These advancements are expected to drive the continued evolution of silica membranes for broader applications.

Graphical Abstract

amorphous silica, membrane separation, microstructure engineering

1. Introduction

Molecular-level separation and purification pose significant challenges across the petrochemical, textile, food, and pharmaceutical industries. Technology like distillation, which is commonly used in this industrial process, accounts for approximately 10% to 15% of global energy consumption.¹ Membrane-based technologies have gained increasing attention because of their intrinsic advantages of high energy efficiency, easy operation, and environmental friendliness.^2–4 The key to the development of membrane separation technology relies on the advancement of membrane materials with outstanding separation capabilities. Certain membrane properties are determined by the specific requirements of their intended application. However, in general, a membrane should⁵: (1) possess a precisely controlled pore size with a narrow pore size distribution, (2) be free from defects that may compromise the separation selectivity, (3) be sufficiently thin to enable high permeation fluxes and support cost-effective production, and (4) remain stable under the specific physical and chemical conditions. Polymeric membranes are predominantly used in industrial applications because of their cost-effectiveness and excellent processability.⁶ However, the free volumes in polymers are randomly distributed, influenced by their chemical structure, and the thermal vibration of chain segments, which results in a trade-off between permeability and selectivity, i.e. polymers with high gas permeability tend to have limited size-sieving capabilities.⁷ Furthermore, polymers generally exhibit lower chemical and thermal stability than inorganic materials.

Silica membranes, a key category of inorganic membranes, have garnered considerable attention owing to their distinctive porous structures and exceptional separation capabilities.^8–10 Silica is characterized by the unique connectivity of its elemental units, namely, SiO₄ tetrahedra, which can form a variety of amorphous or crystalline solids. Although silica exists in multiple crystalline forms, amorphous silica has been extensively investigated for the fabrication of microporous membranes because its structural properties can limit the defects formation. Compared with other metal oxides (e.g. TiO₂ and ZrO₂), silica possess a higher degree of flexibility, making it more suitable for the development of ultramicro/nanoporous membranes with pore sizes smaller than 1 nm, thus enabling effective separation of small molecules like H₂.^10,11 Furthermore, silica membranes exhibit superior thermal stability compared with organic and hybrid materials, and their highly interconnected porous structure within a rigid oxide framework allows for an optimal balance between permeability and selectivity. Despite these advantages, the inherent disorder in amorphous silica structures, along with the presence of (Si–O)_xSi–OH defects, leads to poor hydrothermal stability, with their structure undergoing hydrolysis and densification at elevated temperatures and humidity, posing a significant challenge to their practical industrial applications.^12,13 Thus, strategies for enhancing the separation performance and hydrothermal stability of silica membranes are essential for their further development.

Membrane performance is influenced by both permselectivity (separation factor) and permeability. Microporous membranes typically exhibit relatively low permeability; thus, the selective layer must be sufficiently thin to achieve the desired fluxes. Various techniques, such as the sol–gel process and chemical vapor deposition, have been employed to produce homogeneous, defect-free amorphous silica layers on porous substrates.^14–16 Among these, the sol–gel process is the most widely studied and versatile method for synthesizing amorphous silica, owing to its feasibility and cost-effectiveness. The sol–gel technique enables thin-layer fabrication from a broad range of precursors and holds the potential to be integrated with additional reactions to precisely control the membrane structure.^17,18 On this basis, this review focuses on the microstructural engineering of silica membranes via the sol–gel process and their applications in molecular separation. Additionally, the significance of surface properties on the membrane separation performance in various systems is discussed, along with perspectives on future development.

2. Sol–gel fabrication and physicochemical properties of amorphous silica membranes

The term “sol–gel” was first coined to describe silica sols by Graham in 1864. Since then, extensive literature has documented the advancements of the sol–gel process, particularly in the synthesis of silica-based materials.^19,20 The sol–gel method is the most established technique for the fabrication of silica membranes, primarily because of the wide range of available silica precursors and the ability to control the sol properties for optimized separation characteristics. Moreover, silica-derived membranes fabricated using the sol–gel process have been successfully commercialized for the dehydration of organic solvents, highlighting the promising application potential of this approach.²¹ This section provides an overview of the sol–gel process for silica membrane fabrication.

The sol–gel membrane fabrication method typically consists of 3 main steps, as shown in Fig. 1a: sol preparation, coating, and calcination. The sol–gel process is typically conducted in an organic cosolvent, where hydrolysis (sol) and polycondensation (gel) occur either simultaneously or sequentially, resulting in the release of water and/or alcohol, as illustrated in Equations 1a–c²³:

$$\equiv \mathrm{Si}\text{–}\mathrm{OR}+{\mathrm{H}}_2\mathrm{O}\leftrightarrow \equiv \mathrm{Si}\text{–}\mathrm{OH}+\mathrm{R}\text{–}\mathrm{OH}$$

(1a)

$$\equiv \mathrm{Si}\text{–}\mathrm{OR}+\equiv \mathrm{Si}\text{–}\mathrm{OH}\leftrightarrow \equiv \mathrm{Si}\text{–}\mathrm{O}\text{–}\mathrm{Si}\equiv +\mathrm{R}\text{–}\mathrm{OH}$$

(1b)

$$\equiv \mathrm{Si}\text{–}\mathrm{OH}+\equiv \mathrm{Si}\text{–}\mathrm{OH}\leftrightarrow \equiv \mathrm{Si}\text{–}\mathrm{O}\text{–}\mathrm{Si}\equiv +{\mathrm{H}}_2\mathrm{O}$$

(1c)

Fig. 1.

a) Membrane fabrication procedure by sol–gel process; b) principle of sol–gel reaction for membrane fabrication. Adapted with permission from Ref.²²

Alkoxides (−OR) are initially hydrolyzed to form silanol (Si–OH) and alcohol (Equation 1a). These intermediate species subsequently condense to generate the Si–O–Si network structure (Equation 1b,c). The gelation process is inherently slow due to the low polarity of the Si–O bond in silicon alkoxides. Thus, acid/base catalysts can be introduced to the sol–gel system to promote the hydrolysis/condensation reactions, as illustrated in Fig. 1b. Basic catalysts accelerate condensation, resulting in the formation of highly condensed species and large agglomerates, which are ideal for the fabrication of mesoporous membranes for liquid separation. By contrast, acidic catalysts promote hydrolysis, facilitating the development of microporous membranes with long-chain networks for gas separation.^24,25 After the preparation of the sols, the subsequent steps, including coating the sols onto the support, gelation, drying, and thermal treatment, are essential for the fabrication of silica membranes. In addition to the sol composition, other factors, such as the reaction temperature and calcination conditions, play a crucial role in determining the network structure of the silica membranes.¹⁷

Figure 2a displayed the cross-section transmission electron microscopy image of the representative structure of the sol–gel-derived silica membrane. Amorphous silica membranes are commonly prepared on macroporous ceramic supports. To facilitate the formation of a thin defect-free separation layer, an intermediate layer was first coated onto the support to create mesopores. Subsequently, silica sol was coated on the intermediate layer. This process effectively prevented small silica sols from penetrating into the support, thus ensuring the formation of a high-quality and integrity top layer. Typically, the intermediate layer is hydrophilic, containing −OH groups, which enable covalent bonding with the silica separation layer. Recently, hydrophobic intermediate layers, interlayer-free, and layered coatings have been applied on polymer substrates for fabricating silica membranes.^28–30 These approaches aim to mitigate water condensation during gas permeation, reduce permeation resistance, and lower the costs associated with industrial-scale applications.

Fig. 2.

a) Cross-sectional TEM image of silica membranes. b) Structure of amorphous silica, adapted with permission from Ref.²⁶ c) Pore size distributions of amorphous silica and crystal silica structure, adapted with permission from Ref.²⁷ d) Rearrangement of silica network structure under hydrothermal conditions.

The structure of amorphous silica (Fig. 2b) was less dense compared with its crystalline counterpart. Amorphous silica membranes prepared by tetraethoxysilane (TEOS) exhibit network exhibits pore sizes in the range of 0.1 to 0.4 nm,²⁷ as shown in Fig. 2c. Such networks facilitate the permeation of gases with small molecular sizes, such as He (0.26 nm) and H₂ (0.289 nm), while rejecting larger molecules. Therefore, the silica membranes demonstrate superior H₂ permeance (higher than 10⁻⁶ mol m⁻² s⁻¹ Pa⁻¹) and high H₂ permselectivity (hundreds to thousands).^31,32 However, the pore size of silica membrane is too small for separating gases with larger molecular sizes, such as CO₂ (CO₂/N₂, CO₂/CH₄) and hydrocarbons (C₂H₄/C₂H₆, C₃H₆/C₃H₈). Moreover, compared with crystalline silica, the siloxane bonds (−Si–O–Si−) and silanol (Si–OH) groups in amorphous silica are relatively flexible. Upon calcination, particularly under hydrothermal conditions, densification occurs because of the random condensation of silanol groups,^11,33 as depicted in Fig. 2d. Consequently, 2 critical topics of research for silica membranes are: (1) the precise control of the microstructure for the target separation process and (2) enhancing the stability, especially hydrothermal stability, of the amorphous silica network structure.

Various structural regulation strategies have been proposed to optimize the performance of silica membranes, as summarized in Fig. 3. These strategies can generally be classified into 2 categories: (1) controlling the pore size of the membrane or enhancing its stability by introducing additional structures or utilizing specialized post-treatment methods, and (2) improving the separation performance by incorporating functional groups or components that interact specifically with the target separation species. In the following section, we focus on recent advancements in microstructural engineering technologies for silica membranes.

Fig. 3.

Overview of microstructure engineering technologies for silica membranes.

3. Microstructure and stability regulation strategies of silica membranes

3.1 Metal and fluorine doping

The flexible structure of silica facilitates its doping with various species. Metals such as Ni and Co were incorporated into the silica network via hydrolysis and condensation of TEOS during the sol–gel process. The metal doping strategy significantly enhances the hydrothermal stability of silica-derived membranes, maintaining their H₂ separation performance at temperatures between 500 and 600 °C and under steam partial pressures of 300 to 400 kPa for more than 150 h.^34,35 Although the exact mechanism of cation doping within the silica structure is not fully understood, it has been reported that the presence of doped cations reduces the thermally induced molecular motion in the silica matrix, which in turn improves its hydrothermal stability.

The network structure of silica-based membranes comprises siloxane bonds (−Si–O–Si−) and Si–OH groups. During calcination processes, the condensation of Si–OH groups leads to densification, making it challenging to precisely control the pore size at the sub-nanometer level. Therefore, controlling the density of Si–OH groups, which contribute to the condensation reaction, is crucial for precisely regulating the network size. In glass chemistry, the fluorine (F) is readily incorporated into the silica network structure and bonds with Si atoms to reduce the Si–OH group density. It has been reported that the highly electronegative F, when introduced into the silica network, attracts the shared electron pairs of the Si–O bonds toward the Si side, potentially increasing the Si–O–Si bond angle.^36,37 As shown in Fig. 4a, the Si–OH group density in F–SiO₂ (derived from triethoxyfluorosilane, TEFS) was lower than that in conventional silica membranes derived from TEOS because of the formation of Si–F bonds during the hydrolysis and condensation reactions, which led to minimal changes in the pore structure during the calcination process.³⁸ Calculated from the force constant model for the vibrational properties of the FTIR spectra (Fig. 4b),³⁹ the Si–O–Si bond angle increased with higher F concentration. For instance, in the F–SiO₂ gel with a F/Si ratio of 2/8, the Si–O–Si bond angle reached 145°, which is approximately 10° larger than that of pure SiO₂. This provides strong evidence for the structural changes in SiO₂ upon F doping, which result in the formation of a more expanded network structure.

Fig. 4.

a) Si–OH/Si–O–Si peak area ratios for TEOS and TEFS powders as a function of calcination temperature, adapted with permission from Ref.³⁸ b) Si–O–Si bond angles of SiO₂ and F–SiO₂ gels calcined at 350 °C as a function of the F concentration, adapted with permission from Ref.³⁹ Pair distribution functions of c) TEOS and d) TEFS powders before and after calcination at 350 °C, adapted with permission from Ref.³⁸ e) XRD pattern and f) N₂ adsorption/desorption isotherms of BTPA and metal-doped BTPA xerogel powders calcined at 250 °C, adapted with permission from Ref.⁴⁰

The pair distribution function G(r) for TEOS and TEFS powders before and after calcination at 350 °C (Fig. 4c,d) further reveals the enhanced stability of SiO₂ after F doping.³⁸ The peaks at 1.6, 2.6, and 3.1 Å correspond to the closest Si–O/Si–F, O–O, and Si–Si correlations, respectively, while the peaks between 3.5 and 6.0 Å correspond to the second closest Si–O/Si–F or Si–Si correlations. Distinct intermediate-range structural changes were observed in TEOS-derived samples before and after calcination at 350 °C, with a noticeable decrease in the O–O and Si–Si peaks, indicating a more disordered structure. Furthermore, the peak around 4.1 Å exhibited significant changes, highlighting structural alterations even at the relatively low calcination temperature of 350 °C. In contrast, the G(r) of TEFS remained unchanged after calcination, underscoring the enhanced thermal stability achieved through F doping. This can be attributed to the replacement of Si–OH by Si–F bonds. Recent studies have shown that even after calcination at 750 °C, the network structure in F–SiO₂ remains virtually intact, with significantly improved structural stability compared with silica membranes without F doping. Recent studies have shown that even after calcination at 750 °C, the network structure in F–SiO₂ remains virtually intact, with significantly improved structural stability compared with silica membranes without F doping.⁴¹

Recently, a novel strategy for fabricating microporous silica-based coordination polymers by doping metal ions into silica membranes containing amine groups.^40,42–45 The electron donor pair in the amine groups interacts with the electron-accepting “d” orbitals of the transition metal, which promotes the coordination reaction, resulting in the formation of an ordered microporous network with a crosslinked structure. This strategy provides a promising alternative for controlling the pore size of silica membranes. For example, an intriguing study explored the impact of metal doping with varying coordination strengths on the network structure of silica precursor.⁴⁰ Bis[3−(trimethoxysilyl)propyl]amine (BTPA) was chosen as the silica precursor, and metal dopants such as AgNO₃, Cu(NO₃)₂·3H₂O, and Ni(NO₃)₂·6H₂O served as metal dopants. The metal coordination affinity of the amine groups on BTPA followed the order Ni–BTPA > Cu–BTPA > Ag–BTPA. The X–ray diffraction (XRD) analysis (Fig. 4e) revealed that stronger metal coordination resulted in a more amorphous silica structure, which is beneficial for membrane fabrication. By contrast, Ag, with lower coordination affinity, formed nanoparticles phase, as evidenced by its sharp crystalline peaks. The N₂ adsorption/desorption results (Fig. 4f) indicate the presence of both microporous and mesoporous structures in the BTPA doped with transition metal ions. Notably, samples with stronger metal coordination exhibited higher N₂ adsorption at the initial stage, indicating a greater concentration of micropores, as the strong metal affinity promotes crosslinking via amine bridges or coordination with neutral Lewis bases. These interactions foster the formation of 1-, 2-, or 3-dimensional arrangements, while the increased rigidity in the BTPA structure contributes to a more porous network.

3.2 Spacer technology

The “spacer technique” proposed by our group has been extensively validated as an effective method for controlling membrane pore size using bridge-type silsesquioxanes (bridge-type organosilica) with Si–R–Si units (where R represents the organic linking groups). The broad options of organosilica precursors with diverse organic bridging units allow the fine-tuning of the fundamental structural unit of the silica membrane. Furthermore, the sol–gel process offers the significant advantage of low-temperature processing, enabling the preservation of organic functional groups after hydrolysis and condensation, which can effectively alter the network structure. To rationally design high-performance silica membranes, it is essential to understand the properties of membranes fabricated using various organosilica precursors.

The pore structure of bridged organosilica membranes is significantly influenced by the number of carbon atoms connecting 2 Si atoms.⁴⁶ The relationship between the number of carbon atoms in organosilica precursors and the H₂/N₂ permeance ratio at 200 °C, as well as the microporous structure, is shown in Fig. 5a,b. The molecular sieving properties, as indicated by the H₂/N₂ permeance ratio, decreased as the number of carbon atoms increased. This can be attributed to the expansion of the network size. The bis(triethoxysilyl)methane (BTESM) and bis(triethoxysilyl)ethane (BTESE) powders with smaller carbon numbers exhibited more porous structures, evidenced by the higher N₂ adsorption amount at low relative pressures and type-I adsorption/desorption isotherms in N₂ adsorption measurements. By contrast, bis(triethoxysilyl)propane (BTESP), bis(trimethoxysilyl)hexane (BTMSH), and bis(triethoxysilyl)octane (BTESO) powders showed minimal N₂ adsorption, indicating a nonporous structure. It was hypothesized that increasing the number of carbon atoms in the organosilica precursor enhanced its flexibility and inhibited the formation of a porous structure. By contrast, bis(triethoxysilyl)benzene (BTESB) with a rigid phenyl group, was able to form a porous structure similar to those of BTESM and BTESE, as shown in Fig. 5c. Overall, the network size can be controlled at the sub-nanometer level by adjusting the number of carbon atoms in the organosilica precursor. However, when the number of carbon atoms between the Si atoms exceeds 3 (Si–C₃H₆−Si), the increased flexibility of the Si–R–Si linkage leads to more pronounced polymeric behavior. Consequently, precise control of the pore structure becomes challenging, which is unfavorable for the development of membranes with high permeance. Moreover, the presence of organic groups within the organosilica networks mitigated the degradation of the Si–O–Si networks and improved hydrothermal stability. For example, the gas permeance of BTESE membranes remained consistent after exposure to 3 kPa steam at 300 °C for 90 h.⁴⁷

Fig. 5.

a) Influence of alkyl structure on the H₂ separation performance of organosilica membranes and b) N₂ adsorption isotherm at 77 K for organosilica powders with various bridge units after calcination at 350 °C, adapted with permission from Ref.⁴⁶ c) Schematic depiction of the effect of alkyl chain length in organosilica precursors on the resulting network structures.

3.3 Template method

The template technique was first proposed by Raman and Brinker in 1995,⁴⁸ where they developed the molecular sieving membrane by the co-hydrolysis and condensation of TEOS and methyltriethoxysilane (MTES), a pendant-type alkoxysilane (pendant-type organosilica). The membrane fabrication process involved the calcination of the coated layer under N₂, followed by air calcination to remove organic functional groups and adjust the pore sizes. Furthermore, octyl-, dodecyl-, and octadecyl-triethoxysilanes were also co-hydrolyzed and condensed with TEOS. Following the removal of the organic template, the sample exhibited increased porosity, and large pore contraction led to enhanced molecular sieving performance.⁴⁹ Silica, being naturally hydrophilic, often experienced decreased permeance due to adsorbed water; however, the introduction of an organic group enhances the hydrophobicity of silica membranes.^50,51 For instance, hydrophobic methylated-SiO₂ membranes can be prepared by the co-hydrolysis and condensation of MTES and TEOS without calcination in air.

To tailor the hydrophobicity and porosity, various studies have reported carbonized template molecular sieve silica (CTMSS) by calcining membranes in a vacuum or inert atmosphere.^52,53 For example, Duke et al.⁵⁴ fabricated the CTMSS membrane by carbonizing the hexyl trimethyl ammonium bromide C6 surfactant that was introduced into the membranes during the sol–gel process.^54,55 As the template decomposed, the porosity of the obtained CTMSS increased (Fig. 6a) compared with the nontemplated molecular sieve silica (MSS), which was prepared using the same method in the absence of surfactants. Moreover, the CTMSS demonstrated significantly improved water absorption stability over multiple runs, in stark contrast to the decreased water absorption observed in MSS (Fig. 6b). This clearly emphasizes the substantial enhancement in hydrothermal stability achieved through the carbonized template method. As shown in Fig. 6c, the presence of carbon within the silica matrix plays a crucial role in preserving the silica pore structure under hydrothermal conditions. Carbon acts as a protective shield, preventing disruption of the silica framework and inhibiting the movement of the silica network during hydrolytic attack, effectively safeguarding the integrity of the micropores and preventing collapse.⁵⁶

Fig. 6.

a) N₂ adsorption isotherms for MSS and CTMSS, adapted with permission from Ref.⁵⁴ b) H₂O adsorption on MSS and CTMSS in repeated analyses after degassing at 200 °C, adapted with permission from Ref.⁵⁵ c) Schematic of mechanism for stabilization of silica pore structure by the carbon template, adapted with permission from Ref.⁵⁶

3.4 Composite ceramic

Silica-based composite membranes integrated with metal oxides such as titania (TiO₂) and zirconia (ZrO₂) through the sol–gel method provide a simple yet effective approach to leverage the inherent processing flexibility of silica while enhancing its resistance to aggressive thermal and chemical environments. Among these, the doping of ZrO₂ has been extensively investigated.^57–59 Under relatively mild thermal or hydrothermal conditions, the stability of SiO₂−ZrO₂ composite membranes at 500 °C could be enhanced as the ZrO₂ content increased.⁶⁰ However, the stability of the equimolar SiO₂−ZrO₂ declined at temperatures exceeding 550 °C, where the aggregation of tetragonal ZrO₂ led to the densification of the composite structure. Additionally, the presence of steam accelerated the degradation of the network structure of the SiO₂−ZrO₂ composite.^58,60,61

More recent studies have reported that doping yttrium into SiO₂−ZrO₂ composites (Y–SiO₂−ZrO₂) can significantly enhance their thermal stability and maintain an amorphous structure up to 750 °C.^62,63 This approach contrasts with the widely reported use of yttria in yttria-stabilized zirconia, in which yttria stabilizes zirconia in its cubic phase to preserve its fluorite structure.⁶⁴ In the case of Y–SiO₂−ZrO₂, Y³⁺ ions stabilize ZrO₂ in its amorphous phase. Fig. 7a shows the attenuated total reflection Fourier transform infrared (ATR–FTIR) spectra of SiO₂−ZrO₂ and Y–SiO₂−ZrO₂, yttrium doping results in an intensified peak corresponding to the −Si–O–Zr− bond in the 750 to 1250 cm⁻¹ range. This suggests that yttrium effectively suppresses the aggregation of the ZrO₂ phase, facilitating the formation of a more uniform SiO₂−ZrO₂ composite ceramic structure and enhancing its structural stability. Furthermore, the N₂ adsorption/desorption isotherms in Fig. 7b highlight the impact of yttrium on the microstructural characteristics. Notably, the Y–SiO₂−ZrO₂ exhibits a steeper adsorption curve than the SiO₂−ZrO₂ during the initial stage of the test, indicating a higher degree of microporosity. Typically, the incorporation of ZrO₂ into SiO₂ decreases the porosity due to the inevitable aggregation of ZrO₂ phases. The introduction of Y prevented the aggregation of ZrO₂ phases and mitigated the densification of the SiO₂−ZrO₂ network structure during calcination. However, as the Y doping concentration increased, a reduction in N₂ adsorption was observed in the initial of the test, suggesting a decrease in micropore volume. This is likely due to the higher Y content, which may lead to the formation of a distinct yttria phase.

Fig. 7.

a) ATR–FTIR spectra and b) N₂ adsorption/desorption isotherms at 77 K of Y–SiO₂−ZrO₂ powders with varying yttrium concentrations prepared at 550 °C, adapted with permission from Ref.⁶² c) XRD patterns and d) AFM particle size distribution of Y–SiO₂−ZrO₂ powders calcined at 550 °C before/after steam treatment, adapted with permission from Ref.⁶³ HRTEM and SAED images of d) SiO₂−ZrO₂ and e) Y–SiO₂−ZrO₂ calcined at 850 °C, adapted with permission from Ref.⁶²

As noted, the presence of water vapor imposes more stringent stability requirements on the silica. XRD analysis (Fig. 7c) shows that Y–SiO₂−ZrO₂ remains amorphous after steam treatment at 700 °C. Although tetragonal ZrO₂ peaks appeared in the XRD pattern after steam treatment at 800 °C, only minor changes in particle size distribution were observed in the atomic force microscopy (AFM) images (Fig. 7d). High-resolution TEM (HRTEM) analysis of yttrium-doped and undoped samples calcined at 850 °C further reveals notable differences in stability (Fig. 7e,d). A high density of lattice clusters, together with distinct rings and dots in selected area electron diffraction (SAED) patterns, were observed in SiO₂−ZrO₂, indicative of aggregated high-crystallinity t–ZrO₂. By contrast, the less dense lattice pattern in Y–SiO₂−ZrO₂ suggests a reduced aggregation of the t–ZrO₂ crystals. Moreover, the halo- and ring-SAED patterns of Y–SiO₂−ZrO₂ reflect a lower crystalline intensity. The improved hydrothermal stability observed upon yttrium doping may result from the larger cationic radius of Y³⁺ (1.06 Å) compared with Zr⁴⁺ (0.87 Å), which likely reduces the nucleation site density and increases the activation energy for ZrO₂ nanocrystal aggregation. Therefore, the gas permeance of Y–SiO₂−ZrO₂ membrane remains almost unchanged within 20 h under 700 °C and 90 kPa steam.⁶⁵

3.5 Organic chelating ligands-modified composite ceramic

The incorporation of metal oxides into silica significantly enhances its chemical and hydrothermal stability. However, several challenges need to be addressed: (1) transition metal alkoxides exhibit high reactivity during hydrolysis and condensation, leading to undesirable agglomeration and phase segregation in composite ceramics, thereby diminishing some of the benefits of this approach; (2) the addition of transition metal oxides reduces the microporosity of silica, which may adversely affect the separation efficiency. Thus, it is crucial to control the hydrolysis and condensation rates of transition metal alkoxides while tailoring the pore structures of composite ceramics to meet the requirements of target separation applications. In the seminal work by Sanchez et al.⁶⁶ on the chemical modification and sol–gel chemistry of transition metal alkoxides, positively charged metal atoms tended to enhance their coordination number by utilizing vacant orbitals to accept electrons from nucleophilic ligands. This interaction consumes alkoxy groups and is facilitated by modifying the alkoxides with complexing agents, typically hydroxylated nucleophilic ligands. Consequently, chelated and hydrolytically stable transition metal oxide precursors can be obtained.

Furthermore, the application of organic chelating ligands (OCLs) with hydroxylated nucleophilic groups has been investigated to modify transition metal alkoxides and fabricate homogeneous composite ceramic membranes. As illustrated in Fig. 8a, the OCLs modification enables precise control over the rates of hydrolysis and condensation of transition metal alkoxides. The modified alkoxides were then co-hydrolyzed and polycondensed with silica precursors, resulting in a homogeneous composite ceramic structure.⁶⁷ For example, Sulaiman et al. modified zirconium (IV) tert-butoxide (ZrTB) with acetylacetone (ACAC) and prepared a SiO₂−ZrO₂−ACAC composite ceramic via co-hydrolysis and polycondensation with TEOS.⁶⁸ Upon calcination at 700 °C, the XRD patterns of the unmodified SiO₂−ZrO₂ displayed the tetragonal ZrO₂ peaks, which suggested the aggregation of the ZrO₂ phase. By contrast, the SiO₂−ZrO₂−ACAC composite remained amorphous under the same conditions, indicating that the modification facilitated the formation of a more homogeneous ceramic composite structure.

Fig. 8.

a) Preparation of OCL-modified composite ceramic networks, adapted with permission from Ref.⁶⁷ b) XRD patterns of pure SiO₂−ZrO₂ and SiO₂−ZrO₂−ACAC powders before/after calcination at 700 °C in N₂ adapted with permission from Ref.⁶⁸ c) Illustration of network modification and network formation in OCLs-modified ceramics, adapted with permission from Ref.⁶⁷ d) Illustration of the network formation for noncrosslinked and crosslinked MAPTMS–SZ–LIGAND materials and e) their N₂ adsorption/desorption isotherms at 77 K, adapted with permission from Ref.⁶⁹

In addition to controlling the hydrolysis rate of transition metal alkoxides, the incorporation of OCLs offers the potential to regulate the network architecture and pore characteristics of the composite ceramics, as depicted in Fig. 8c. When OCLs contain a nonreactive group, it exerts a modifying effect on the network structure.⁷⁰ In this scenario, the ligand serves as a site for physical or chemical manipulation. In contrast, when OCLs possess reactive groups capable of polymerization or copolymerization, they can contribute to the formation of porous polymer composites. In such cases, the ligand served as a network-forming agent, increasing the microporosity of the network.⁶⁷ As illustrated in Fig. 8d, the silica precursor 3-methacryloxypropyltriethoxysilane (MAPTMS) with unsaturated C=C bonds can undergo crosslinking reactions, either with itself or with additional C=C bonds in the presence of a radical initiator, leading to the formation of crosslinked structures.⁶⁹ Notably, the N₂ adsorption across the entire pressure range was significantly higher for the crosslinked samples than their noncrosslinked counterparts, indicating increased porosity. Furthermore, when ZrTB was modified with allyl acetoacetate, which contains a reactive C=C bond, the N₂ adsorption isotherm of the resulting composite was steeper than that of the ACAC-modified sample at a low relative pressure. This suggests that OCLs with crosslinking functionalities contribute to the development of network structures.

3.6 Carbon contained composite ceramic

Carbon molecular sieve (CMS) membranes derived from the pyrolysis of polymer precursors have demonstrated excellent separation performance and stability in various applications.^71–73 OCLs modification provides opportunities to integrate CMS membranes with ceramic membranes via the pyrolysis of the OCL moieties in the homogeneous composite ceramic network.

For example, carbon-contained composite ceramic membranes (C–SiO₂−ZrO₂) were fabricated by the pyrolysis of a SiO₂−ZrO₂−ACAC (SZa4) composite, which was derived from a network-modifying OCLs (ACAC),^74,75 in an inert atmosphere. The proposed mechanism for pyrolysis and network formation is depicted in Fig. 9a. The OCLs decompose under a high-temperature inert atmosphere, releasing volatile byproducts and leaving behind carbon nanoparticles in the intra-network spaces. The C–SiO₂−ZrO₂ (CSZ550) membrane carbonized at 550 °C demonstrates unique CO₂ adsorption property, as illustrated by the CO₂ adsorption/desorption isotherms in Fig. 9b. The isotherms of SZa4 at 25 °C and 35 °C are fully reversible, indicating no significant interaction between CO₂ and the adsorbent. By contrast, the adsorption isotherm of CSZ550 displays considerable hysteresis, suggesting that carbonization induces a special CO₂ affinity. The observed hysteresis, particularly at temperatures above the critical temperature of CO₂ (31 °C), excludes capillary condensation as a cause and points to the structural rearrangement as the primary factor. This behavior can be attributed to the graphitic properties of the carbon nanoparticles in C–SiO₂−ZrO₂, which exhibit weak van der Waals interactions due to π–π stacking and sp²-hybridized C–C σ–bonding in the planar direction.⁷⁷ In CSZ550 with a Si/Zr molar ratio of 5/5, the residual carbon enrichment led to specific interactions between the delocalized electrons and polar CO₂ molecules. Besides, the free carbon generated during pyrolysis was effectively incorporated into the SiO₂−ZrO₂ matrix without forming a distinct phase, which is beneficial to the structural integrity post-carbonization and is critical for the membrane separation performance.

Fig. 9.

a) Pyrolysis process of SiO₂−ZrO₂−ACAC to C–SiO₂−ZrO₂ and b) CO₂ adsorption/desorption isotherms of SZa4 and CSZ550 xerogels at 25 and 35 °C, adapted with permission from Ref.⁷⁴ c) Schematic illustration for the impact of organic structure on the distribution patterns of obtained free carbon and d) HRTEM images and SAED patterns of APTES–ZrTB and APTES–ZrTB–GA xerogels calcined at 550 °C, adapted with permission from Ref.⁷⁶

Since the free carbon in C–SiO₂−ZrO₂ is derived from the organic components within the network, its final state is influenced by the OCLs precursors. Previous studies demonstrated that a network-forming OCL (benzoxazine) can modify ZrTB and subsequently undergo co-hydrolysis and polycondensation with the Si precursor vinyltrimethoxysilane (VTMS). The mixture is then thermally cured at temperatures of 90 to 200 °C in the presence of the radical initiator dibenzoyl peroxide (BzO₂) to produce a SiO₂−ZrO₂−polybenzoxazine resin.^78,79 This process enhances the subsequent carbonization process to temperatures ranging from 700 to 800 °C. Additionally, benzoxazine facilitates the formation of a continuous carbonized structure rather than isolated free carbon nanoparticles. Notably, the sp² carbon generated during carbonization plays a protective role in maintaining the integrity of the network structure. The C–SiO₂−ZrO₂ membrane, fabricated at 750 °C, exhibited exceptional hydrothermal stability, retaining stable gas permeance for 80 h at 500 °C and under pressure ranging from 90 to 150 kPa.

Our recent investigation revealed that the distribution pattern of free carbon in SiO₂−ZrO₂ was influenced by the rigidity of the organic moieties in the OCLs precursors, as shown in Fig. 9c.⁷⁶ To promote the formation of crosslinked organic moieties, glyoxylic acid (GA) was employed as an OCL to modify ZrTB, followed by co-hydrolysis and polycondensation with 3-aminopropyltriethoxysilane (APTES). During this process, the aldehyde groups in GA react with the amine groups in APTES via aldimine condensation, resulting in a composite ceramic (APTES–ZrTB–GA) containing a rigid, crosslinked organic precursor. The HRTEM images (Fig. 9d) clearly show free carbon particles smaller than 1 nm in the sample carbonized at 550 °C (APTES–ZrTB–GA–550 °C and APTES–ZrTB–550 °C). The free carbon derived from the crosslinked precursor exhibited a more uniform distribution than that derived from the noncrosslinked APTES–ZrTB. The crosslinked precursor exhibits increased rigidity and reduced mobility at elevated temperatures, promoting the formation of homogeneously distributed free carbon during pyrolysis. Homogenous free carbon in the network provides enhanced protection to the ceramic matrix, as evidenced by the amorphous structure of APTES–ZrTB–GA–550 °C observed in the SAED results. Typically, with the higher carbonization temperature, the carbon from the organic component transited more to graphite-like sp² hybridization together with a densified SiO₂−ZrO₂ network structure. A 550 °C carbonization temperature leads to enhanced microporosity without network densification. Moreover, the uniformly distributed free carbon in APTES–ZrTB–GA–550 °C further promoted the development of micropores and minimized defect formation.

4. Gas and liquid separation properties of silica-derived membranes

Since the emergence of silica membranes in the 1990s, their applications have spanned gas separation as well as liquid separation processes, such as pervaporation (PV), reverse osmosis (RO), and nanofiltration. Table 1 briefly outlines the overall separation system, the target application of membrane separation, and corresponding applicable regulation strategies for silica-derived membranes. In the domain of gas separation, silica membranes have garnered attention not only for H₂ separation but also for CO₂ capture and olefin/paraffin separation. Additionally, there has been growing interest in the use of silica membranes for organic solvent recovery and water desalination over the past decade. The key performance metrics for these applications include permeance, selectivity, and operational stability. Economic simulations have highlighted that, beyond a certain level of selectivity, higher permeance is more crucial for practical application.^80,81 Stability is directly linked to the lifespan of a membrane, and improved stability translates to lower operational costs. This section focuses on the representative applications of silica membranes in various separation processes, along with the associated challenges.

Table 1.

Overview of representative practical applications via membranes separation and applicable regulation strategies for silica-derived membranes.

Separation systems	Separate molecules	Target applications	Applicable regulation strategies for silica-derived membranes
Gas separation	He/Ne He/H2 He/N2 He/CH4	He separation and purification from natural gas	Conventional dense silica
	H2/H2O, CH4, CO H2/NH3, N2 H2/hydrocarbon	H2 recovery and purification	Metal doping Carbon contained composite ceramic
	CO2/N2 CO2/CH4	Carbon capture, utilization, and storage CH4 purification from bio- and natural gas	Fluorine doping Spacer technology Template method
	O2/N2, Ar	Air separation for oxygen enrichment	Spacer technology
	Propylene/propane	Purification of olefin raw materials for the petrochemical industry	Spacer technology Template method Carbon contained composite ceramic
Pervaporation/Vapor permeation	H2O/alcohol H2O/acetic acid	Dehydration of organic solvent and separation of azeotropes	Spacer technology Carbon contained composite ceramic
	MeOH/toluene MeOH/DMC MeOH/MTBE MeOH/EA	Methanol recovery from organic solvent mixtures	Metal coordination Spacer technology
	Aromatic/alkane	Separation and purification of aromatic compounds from petroleum feedstocks	Spacer technology Organosilica hybridization
Reverse osmosis	H2O/salt	Water desalination in harsh environments	Spacer technology
	Polar/nonpolar organic solvent Aromatic/alkane solvent Isomer mixture	Recovery of organic solvents	Spacer technology Organosilica hybridization
Nanofiltration	H2O/bacteria, algal algae, and organic matter H2O/heavy metal H2O/oil H2O/dye H2O/ion H2O/salt	Groundwater and surface water purification, wastewater reuse, and water desalination	Composite ceramic
	Small/large hydrocarbon Organic solvent/dye	Organic solvent separation and purification	Spacer technology Composite ceramic

EA, Ethyl acetate; DMC, Dimethyl carbonate; MTBE, Methyl tert-butyl ether.

4.1 Gas separation

Silica membranes have gained significant attention for gas separation applications because of their versatile and resilient porous structures, which enable precise selective molecular separation. The subsequent sections provide an overview of the representative advancements in silica membranes for gas separation processes.

4.1.1 H₂ separation

H₂ separation is one of the most promising applications of silica membranes because of their ideal pore size (below 0.4 nm), which provides excellent permselectivity for H₂. Steam reforming of methane is the predominant industrial method for H₂ production,^82,83 which involves the endothermic reaction between methane and water vapor (Equation 2a) within a temperature range of 500 to 700 °C. Concurrently, a water-gas shift reaction (Equation 2b) occurs, promoting additional H₂ production while generating CO₂ as a byproduct. In this context, to effectively separate the target product H₂ (0.289 nm) from CO₂ (0.33 nm), CO (0.376 nm), and CH₄ (0.38 nm), the membrane must feature small and uniform pore sizes due to the relatively narrow molecular size differences. Additionally, the membrane must demonstrate excellent hydrothermal stability in the presence of H₂O, as shown in Fig. 10a.

$$\mathrm{C}{\mathrm{H}}_4+{\mathrm{H}}_2\mathrm{O}\leftrightarrow \mathrm{CO}+3{\mathrm{H}}_2$$

(2a)

$$\mathrm{CO}+{\mathrm{H}}_2\mathrm{O}\leftrightarrow \mathrm{C}{\mathrm{O}}_2+{\mathrm{H}}_2$$

(2b)

Fig. 10.

a) Schematic of membrane reactor for steam reforming. b) Single-gas permeance versus molecular size at 300 °C for C–SiO₂−ZrO₂ membranes before and after hydrothermal stability tests at 500 °C, adapted with permission from Ref.⁷⁵ c) Protective effect of free carbon on silica network pores, adapted with permission from Ref.⁵⁵ d) Comparison of H₂/CH₄ separation performance of APTES–ZrTB–GA-derived membranes with that of state-of-the-art membranes, adapted with permission from Ref.⁷⁶ e) Schematic of the membrane reactor for C₃H₈ dehydrogenation. f) Evolution of network structure in cured and uncured PCS, f) gas permeance versus molecular size at 200 °C for PCS M250 membrane before and after air treatment, and g) permeance and selectivity of PCS M250 to 4 membrane as a function of test time at 500 °C, adapted with permission from Ref.⁸⁴

Organosilica membranes demonstrate good H₂ separation performance. For example, BTESE membranes demonstrate H₂ permeances ranging from 10⁻⁵ to 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹ with H₂/CH₄ selectivities ranging from 20 to 90.^47,85,86 While BTESE membranes maintain stable gas permeance after steam exposure at 300 °C,⁸⁷ the decomposition of organic components at elevated temperatures will lead to degradation in membrane performance. Considering the smaller pore size and enhanced hydrothermal stability, metal-doped silica and C–SiO₂−ZrO₂ membranes are more suitable for H₂ separation in methane steam reforming processes. For instance, Co-doped silica membranes demonstrate a stable and exceptional H₂ permeance of 4.00 × 10⁶ mol m⁻² s⁻¹ Pa⁻¹ and a H₂/N₂ selectivity of 730, even under a steam pressure of 300 kPa at 500 °C.³⁴

The C–SiO₂−ZrO₂ membranes, derived from the carbonization of SiO₂−ZrO₂−OCL composite ceramic membranes, demonstrate excellent separation performance and hydrothermal stability for H₂ separation. For instance, a C–SiO₂−ZrO₂ membrane derived from SiO₂−ZrO₂−ACAC with a Si/Zr ratio of 9/1 exhibits a high H₂ permeance of 1.6 × 10⁻⁶ mol m⁻² s⁻¹ Pa⁻¹ and a H₂/CH₄ permeance ratio of 148 at 300 °C.⁷⁵ A previous study assessed the hydrothermal stability of a C–SiO₂−ZrO₂ membrane derived from SiO₂−ZrO₂−polybenzoxazine. As shown in Fig. 10b, H₂ permeance values at 300 °C (2.8 × 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹) remained stable after hydrothermal testing under steam partial pressures of 90 kPa (3.8 × 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹) and 150 kPa (3.8 × 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹). Notably, even under a more challenging steam pressure (150 kPa), the membrane retained an impressive H₂/CH₄ selectivity of 439. The outstanding hydrothermal stability of the C–SiO₂−ZrO₂ membrane can be ascribed to the free carbon formed during thermal annealing, which acts as a protective layer (Fig. 10c), shields the network structure and hinders the rehydration of Si–O–Zr bonds to hydroxyl groups. Thus, the re-condensation of terminal hydroxyl groups is mitigated under elevated temperatures and humid environments, thereby restricting inter-pore condensation.⁷⁹ The high porosity and uniform pore size distribution of the C–SiO₂−ZrO₂ membrane resulted in excellent H₂ separation performance. As shown in Fig. 10d, the C–SiO₂−ZrO₂ membrane derived from APTES–ZrTB–with a cross-linked structure at 550 °C exhibited H₂ permeance of 4.0 to 4.4 × 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹ and H₂/CH₄ selectivity of 100 to 120, significantly outperforming typical polymer membranes (H₂ permeance ∼10⁻⁸ mol m⁻² s⁻¹ Pa⁻¹, H₂/CH₄ selectivity 10 to 40).^88–90 Compared with zeolite membranes (H₂ permeance >10⁻⁶ mol m⁻² s⁻¹ Pa⁻¹, H₂/CH₄ < 100)^91–93 and CMS membranes (H₂ permeance 10⁻⁸ to 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹, H₂/CH₄ > 170),^94–96 the APTES–ZrTB–GA–550 °C membrane overcomes the trade-off effect, achieving a balanced enhancement in H₂ permeance and selectivity. The outstanding separation performance and exceptional hydrothermal stability collectively underscore the significant potential of C–SiO₂−ZrO₂ membranes for practical applications.

Propane (C₃H₈) dehydrogenation presents another promising route for H₂ production that simultaneously generates propylene, as shown in Fig. 10e. This reaction is thermodynamically favored at elevated temperatures and low pressures, and the removal of the produced H₂ from the reaction stream (Equation 3a) via membrane separation can significantly improve the reaction conversion. Under high-temperature conditions, undesirable side reactions, such as coke formation (Equation 3b), can deactivate the catalysts and diminish membrane performance.^97,98 To address this, catalysts and membranes will be regenerated using oxidative agents to burn off the deposited coke. Hence, the thermal and oxidation resistance of the membranes is crucial for maintaining their long-term performance. Furthermore, the separation of the smaller H₂ (0.289 nm) from the larger C₃H₆ (0.468 nm) and C₃H₈ (0.506 nm) is targeted in this process. Given the significant molecular size disparity, membranes with a loose structure can achieve higher H₂ permeance while maintaining effective separation performance.

$${\mathrm{C}}_3{\mathrm{H}}_8\leftrightarrow {\mathrm{C}}_3{\mathrm{H}}_6+{\mathrm{H}}_2$$

(3a)

$${\mathrm{C}}_3{\mathrm{H}}_8\leftrightarrow 3\mathrm{C}+4{\mathrm{H}}_2$$

(3b)

To improve the oxidative stability of silica-derived membranes, a structural design strategy was developed based on organic crosslinking reactions. VTMS, containing Si–CH=CH₂ groups, and triethoxysilane (TRIES), containing Si–H groups, were co-hydrolyzed and condensed to form a silsesquioxane polymer with a Si–O–Si backbone. Subsequently, 1,1,3,3-tetramethyldisiloxane (TMDSO) was incorporated into the network via hydrosilylation, enabling precise control of the network structure through crosslinking of the CH₃ groups. The resulting membrane features an oxidation-resistant SiOC structure, demonstrating excellent performance in separating H₂ from larger molecules (H₂/SF₆ = 2900). Additionally, the gas permeation properties of the SiOC membranes remained almost unchanged after air treatment at 500 °C for 12 h.⁹⁹ To further optimize the pore size to improve H₂/C₃H₈ separation and enhance the oxidative stability of silica membranes. Wang et al.⁸⁴ proposed an oxidative cross-linking process. In detail, polycarbosilane (PCS) undergoes thermal oxidation curing in air to induce cross-linking, forming a polymeric structure, which is then pyrolyzed in an inert gas atmosphere to facilitate the polymer-to-ceramic transformation. The air-curing process allows precise control over the elemental composition and microstructure of the resulting ceramic material. The most promising PCS-derived membrane was cured at 250 °C, yielding a stable and homogeneous microporous structure, followed by pyrolysis at 750 °C, where stable O–Si–C and Si–C bonds were present in the membrane. The obtained membrane exhibited not only excellent H₂ separation properties, with a H₂ permeance of 1 to 2 × 10⁻⁶ mol m⁻² s⁻¹ Pa⁻¹ at 200 °C, H₂/N₂ selectivity of 24, and H₂/C₃H₈ selectivity of 654, but also remarkable thermal stability and oxidation resistance. After air treatment at 500 °C, the membrane showed a slight increase in gas permeance, while the selectivity remained largely unchanged (Fig. 10f). Furthermore, the permeance and selectivity for an equimolar H₂/C₃H₈ mixture at 500 °C were nearly identical to those observed in single-gas conditions (Fig. 10g) suggesting the significant potential for practical applications.

4.1.2 CO₂ separation

CO₂ separation is a critical process with significant industrial and environmental implications, particularly in applications such as CO₂ removal from natural and flue gases.^100,101 Unlike H₂ separation, membranes for CO₂ separation require larger pore sizes. Additionally, the molecular size differences between gases in typical CO₂ separation systems, such as CO₂/CH₄ and CO₂/N₂ are relatively small (0.33, 0.364, and 0.38 nm for CO₂, N₂, and CH₄, respectively), thereby demanding precise control over the membrane pore structure and surface characteristics to achieve effective separation.

For CO₂/CH₄ systems, which exhibit moderate molecular size differences, the membrane separation performance can be effectively enhanced through structural modifications. Doping F in the silica network has been demonstrated to be effective in enhancing the CO₂/CH₄ separation efficiency. F can be doped by introducing ammonium fluoride (NH₄F) or using alkoxysilane with a Si–F bond during the sol–gel process.^38,41 The −F group substitutes for the −OH group, the generated Si–F bonds reduce the condensation degree and increase the bond angle of the Si–O bond, thereby resulting in a more open network structure. The reduced silanol content mitigates densification at elevated temperatures in the presence of steam.^102,103 As depicted in Fig. 11a, compared with the TEOS-derived pure SiO₂ membranes, F–SiO₂ membranes exhibited enhanced gas permeance, which increased with the amount of F doping.⁴¹ The F–SiO₂ membrane with an optimal F/Si ratio of 1/9 (Fig. 11b) demonstrated high CO₂ permeance (4.1 × 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹) and outstanding CO₂/CH₄ selectivity (approximately 300 at 35 °C). Compared with CMS membranes, F–SiO₂ membranes exhibit superior CO₂ permeance and CO₂/CH₄ selectivity, with a performance comparable to that of DDR-type zeolite membranes.³²

Fig. 11.

a) Molecular size dependence of gas permeance at 300 °C and b) comparation of, CO₂/CH₄ permeance ratio at 35 °C for F–SiO₂ membranes with state-of-the-art membranes, adapted with permission from Ref.⁴¹ c) CO₂ adsorption/desorption isotherms at 35 °C for amine-functionalized silica xerogels and d) CO₂ separation performance of amine-functionalized silica membranes at 35 °C, adapted with permission from Ref.¹⁰⁴

Organosilica precursors-derived silica membranes have also been applied in CO₂ separation. For instance, the BTESE membrane synthesized via the sol–gel process exhibits moderate CO₂ permeance of (1 × 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹) along with satisfactory CO₂/CH₄ selectivity (90 at 50 °C).¹⁰⁵ Research by Castricum et al.¹⁰⁶ indicated that the membranes fabricated from organosilica precursors containing long, flexible organic chains exhibited reduced CO₂ permeance. By contrast, precursors containing rigid organic groups, such as benzene rings and biphenyls, tended to enhance CO₂ permeability but simultaneously decrease CO₂/CH₄ selectivity. In comparison to CMS membranes, F–SiO₂ membranes exhibit superior permeability and selectivity, with performance on par with DDR-type zeolite membranes.

For CO₂/N₂ separation, the similar molecular sizes of CO₂ and N₂ require precise control of the membrane pore size to achieve effective separation. Regulating surface properties, such as by incorporating CO₂-philic amine groups, offers an effective strategy to enhance membrane separation performance.^107–110 Yu et al.¹⁰⁴ comprehensively investigated the influence of various amine group types on the CO₂ separation performance of silica membranes. Figure 11c presents the CO₂ adsorption/desorption isotherms at 35 °C for silica xerogel powders containing different amine groups. It is obvious that silica with unhindered amines exhibited irreversible adsorption/desorption isotherms primarily governed by Langmuir behavior, accompanied by distinct hysteresis loops, likely owing to the formation of stable carbamates upon CO₂ adsorption. By contrast, the samples with tertiary and sterically hindered amines displayed near-reversible Henry-dominated isotherms due to moderate interactions with CO₂.^111–114 Among these, silica with a second amine demonstrated a significantly higher CO₂ adsorption capacity, likely owing to its stronger basicity.¹¹⁵ Figure 10d illustrates the trade-off between CO₂ permeance and CO₂/N₂ permselectivity for various amine-functionalized silica membranes. Membranes with hindered amines exhibited superior CO₂ separation performance compared with those with unhindered amines or quaternary ammonium groups. Notably, the membranes with hindered amines surpassed the Robeson upper bound, which is typically observed in polymeric membranes. For instance, the most permeable tertiary amine-containing silica (TA–Si) membrane exhibited a CO₂ permeance of 1.7 × 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹ with a CO₂/N₂ selectivity of 21. These findings suggest that the types of amines exhibit a crucial impact on CO₂ transport properties. The reactivity of amine groups can determine their interactions with CO₂ molecules and modulate CO₂ mobility within the amine-functionalized membranes. For the membranes with unhindered amines (high affinity), the formation of stable carbamates or strong CO₂-amine interactions with high CO₂ binding energies leads to dual-mode sorption, potentially limiting CO₂ transport. By contrast, membranes with sterically hindered amines exhibited reversible CO₂ adsorption/desorption, providing a favorable balance between efficient CO₂ uptake and limited diffusion/desorption. This intermediate-to-low CO₂ binding energy facilitates CO₂ transport through these membranes.

4.1.3 Propylene/propane separation

The separation of propylene/propane (C₃H₆/C₃H₈) is a critical process in the petrochemical industry, traditionally achieved using distillation columns with more than 100 stages and high reflux ratios, resulting in significant energy consumption. By contrast, the energy required for membrane-based technologies has been projected to be only 1/3 or less of that of conventional distillation methods.^1,116–118

The bis(triethoxysilyl)acetylene (BTESA) membranes exhibited enhanced C₃H₆ permeance, ranging from 1.0 to 2.0 × 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹, with a C₃H₆/C₃H₈ permeance ratio of 11 to 14 at 200 °C. This superior separation performance can be attributed to the rigid acetylene linkages, which create a highly accessible pore structure within the membrane network.¹¹⁹ To achieve an enhanced C₃H₆/C₃H₈ separation performance, a co-condensation strategy was employed for the fabrication of composite organosilica membranes with tailored pore structures. For example, a molecularly engineered silica membrane was synthesized by co-condensing BTESA and BTESB precursors through the sol–gel process,¹²⁰ The pore size can be tuned by steric effects and the alignment of phenyl groups in BTESB. The composite membrane with a BTESA/BTESB molar ratio of 9:1 (denoted BTESAB 9:1) demonstrated a high C₃H₆ permeance of 4.5 × 10⁻⁸ mol m⁻² s⁻¹ Pa⁻¹ and a C₃H₆/C₃H₈ selectivity of 33 at 50 °C for an equimolar C₃H₆/C₃H₈ mixture. Simultaneously, a comparative analysis was conducted to investigate the relationship between the pore size of the silica membranes (determined using the modified gas translation [GT] model^121,122) and their C₃H₆/C₃H₈ separation performance. As the pore size of the silica membrane increased, the permeation resistance decreased, leading to a higher C₃H₆ permeance. However, the C₃H₆/C₃H₈ permeance ratio initially increased and then decreased with increasing pore size (Fig. 12a,b). The maximum C₃H₆/C₃H₈ permeance ratio was achieved when the pore size was between 0.45 and 0.52 nm. As shown in Fig. 12c, membranes with excessively small pores cannot provide sufficient permeation channels, resulting in low selectivity, as both C₃H₆ and C₃H₈ molecules permeate through the defects. Conversely, membranes with excessively large pores exhibit high gas permeance; however, they lack effective molecular sieving, allowing both C₃H₆ and C₃H₈ to permeate easily. The optimal pore size for C₃H₆/C₃H₈ separation, as determined by the modified GT model, ranges from 0.45 to 0.52 nm, which is between the molecular sizes of C₃H₆ (0.468 nm) and C₃H₈ (0.506 nm), enabling effective their separation.

Fig. 12.

Relationships between pore size and single-gas permeation properties, including a) C₃H₆ permeance and b) C₃H₆/C₃H₈ permeance ratio at 200 °C, for various silica membranes, and c) schematic representation of C₃H₆/C₃H₈ separation utilizing silica membranes with varying pore sizes, adapted with permission from Ref.¹²⁰ d) Diagram illustration and TEM images showing the effect of the calcination temperature on the SiO₂−Co–(acac)₃ network structure and e) trade-off between C₃H₆/C₃H₈ selectivity and C₃H₆ permeance for SiO₂−Co−(acac)₃ membranes, compared with state-of-the-art membranes, adapted with permission from Ref.¹²³

A recent study highlighted the superior performance of Si-based, carbon-contained composite ceramic membranes for C₃H₆/C₃H₈ separation. In this study, Soma et al.¹²³ fabricated homogeneous SiO₂−Co−(acac)₃ composite membranes using TEOS and cobalt acetylacetonate (Co−(acac)₃) as precursors via a one-pot sol–gel method coupled with copolymerization. The pore sizes of the membranes were tailored via calcination at various temperatures, as shown in Fig. 12d. Carbonization at 550 °C yielded a C–SiO₂−Co membrane with a loose structure, ideal for C₃H₆/C₃H₈ separation, and exhibited a C₃H₆ permeance of 4.0 × 10⁻⁸ mol m⁻² s⁻¹ Pa⁻¹ and high selectivity of 46 (Fig. 12e), which surpassed those of polyimide and polymer of intrinsic microporosity membranes.^124–126 Notably, the membrane pore size after carbonization at 550 °C was calculated to be 0.52 nm using the modified GT model. It is within the optimal range for C₃H₆/C₃H₈ separation (0.45 to 0.52 nm), which could also be a reason for the outstanding selectivity.

4.2 Pervaporation

The PV process has been extensively investigated for the separation of azeotropic mixtures, thermally sensitive compounds, and mixtures with closely matched boiling points. This process requires only the latent heat for the evaporation of a small fraction of the feed solution (primarily water) on the permeate side. This characteristic contributes to its superior energy efficiency compared with distillation, which necessitates the vaporization of the entire feed.^127,128 Silica membranes with robust inorganic frameworks demonstrate remarkable resistance to swelling and maintain robust molecular sieving properties, highlighting their potential applications in PV-based separation.

Dehydration of organic solvents is one of the most thoroughly explored applications of PV processes. Silica membranes derived from organosilica precursors demonstrate superior performance and robust resistance to hydrothermal degradation in this context. In addition, their porous structures can be precisely controlled to optimize the separation performance.^129–132 As shown in Fig. 13a, the BTESE membranes with a loose network structure demonstrated water permeance exceeding 10⁻⁶ mol m⁻² s⁻¹ Pa⁻¹, with water/acetic acid permeance ratios ranging from 100 to 1000. This selectivity is lower than that of zeolite membranes with uniform pore sizes, which have been commercially employed in PV dehydration processes and typically exhibit water permeance of 10⁻⁷ mol m⁻² s⁻¹ Pa⁻¹ along with water/acetic acid permeance ratios ranging from hundreds to thousands.^137–141 However, BTESE membranes offer a distinct advantage in terms of permeance, which is desirable for practical applications.¹³³

Fig. 13.

a) Comparison of PV dehydration performance for water/acetic acid mixtures of BTESE membranes with that of state-of-the-art membranes, adapted with permission from ref.¹³³ b) Long-term PV performance of an MTES-BTESE membrane in 5 wt.% water/n-butanol at 150 °C, adapted with permission from Ref.¹³⁴ c) Pictures of water droplets on the BTESE/RTES membrane surface and d) permeate concentration as a function of the number of C atoms in the R-group in PV for butanol/water at 95/5 and 5/95 wt.%, adapted with permission from Ref.¹³⁵ e) Schematic of the evolution of the microstructures of BTESBP by co-condensation of PhTES and f) PV separation performance of BTESBP/PhTES-2 membrane for various aromatic/aliphatic mixtures with 50 wt.% toluene in feed at 50 °C, adapted with permission from Ref.¹³⁶

The microstructure of the silica membrane and its PV performance can be precisely engineered by regulating the parameters of the sol–gel process and selecting different organosilica precursors. For instance, Moriyama et al.¹⁴² fabricated BTESE membranes with pore sizes of 0.44 and 0.54 nm by adjusting the acid ratio from 0.01 to 1 during the sol–gel process. Consequently, the water permeance was enhanced from 1.1 × 10⁻⁸ mol m⁻² s⁻¹ Pa⁻¹ to 2.2 × 10⁻⁸ mol m⁻² s⁻¹ Pa⁻¹. However, at 50 °C, the selectivity for the water/IPA (10 wt.%/90 wt.%) mixture significantly declined from 4,600 to 520. Furthermore, using BTESM precursor with shorter organic chains tends to reduce the pore sizes of the membrane and improves the separation efficiency of water/alcohol mixtures.¹²⁹ In organosilica-derived membranes, organic functional groups effectively shield the siloxane bonds from water, thereby significantly enhancing membrane stability. Castricum et al.¹³⁴ developed a hybrid silica membrane via the co-hydrolysis and polycondensation of BTESE and MTES. As illustrated in Fig. 13b, the hybrid silica membrane demonstrated remarkable stability, maintaining its integrity for up to 500 d during n-butanol (5 wt.% water) dehydration at 150 °C. Throughout the process, the membrane selectivity remained unchanged, with only a gradual decline in flux. The hybrid silica membranes demonstrated superior performance stability compared with both pure silica and MTES membranes, which was attributed to the incorporation of the bridged organosilica precursor (BTESE), which facilitated the formation of a more integrated and robust network structure.

The above applications target the dehydration of aqueous organic solutions by selectively removing water. However, membrane properties can be further regulated by incorporating alternative precursors to achieve alcohol selectivity.¹⁰⁶ Notably, the integration of both bridging and terminating hydrocarbon groups yielded promising results. Paradis et al.¹³⁵ incorporated various R-triethoxysilane (RTES), where the organic tail (R) ranged from C1 to C10 alkyl groups, into BTESE. The surface properties of the resulting BTESE/RTES membranes were successfully tuned from hydrophilic to hydrophobic when the number of carbon atoms in the alkyl chain increased, as evidenced by the water contact angle results (Fig. 13c). With the increased carbon chain length, the membrane transitioned from selectively permeating water to preferentially permeating butanol, even though the molecular size of butanol (0.51 nm) is larger than that of water (0.2955 nm), as indicated in Fig. 13d.

The surface properties exert a critical impact on the membrane permeation properties. In a recent study, the hybridization of organosilica precursors containing phenyl groups was used to fabricate hybrid silica membranes for the selective transport of aromatic compounds.¹³⁶ Briefly, the hybrid membranes were synthesized via the co-condensation of biphenyl-bridged 4,4′-bis(triethoxysilyl)-1,1′-biphenyl (BTESBP) and phenyltriethoxysilane (PhTES) with a phenyl pendant group, as depicted in Fig. 13e. BTESBP serves as the primary structural element and contributes to the formation of abundant open channels. PhTES is introduced to precisely modulate the membrane pore size at the sub-nanometer scale owing to its pore-filling effect.¹⁴³ In addition, PhTES enhances the accessibility of biphenyl groups to aromatic molecules by preventing the formation of lamellar structures and the π–π interactions between the aromatic groups, which significantly improve the adsorption affinity for aromatic compounds on the membrane. As a result, the membrane with an optimal BTESBP/PhTES mass ratio of 98/2 (BTESBP/PhTES-2 membrane), calcined at 450 °C, demonstrated a superior PV separation performance for aromatic/aliphatic mixtures, as shown in Fig. 13f. For instance, the total permeation flux exceeded 1 kg m⁻² h⁻¹, coupled with a separation factor of 8 to 10 for a 50 wt.% toluene/n-heptane mixture.

4.3 Reverse osmosis

RO is widely employed in water purification, particularly for desalinating saline water, as it relies on pressure as the driving force for water permeation without involving any phase change. Silica membranes offer excellent chemical resistance and tunable pore sizes (ranging from 3 to 5 Å), making them well-suited for selectively permeating water (with a molecular diameter of 2.6 Å) while effectively rejecting hydrated ions such as Na⁺ (7.2 Å) and Cl⁻ (6.6 Å).^144–146 Silica membranes derived from organosilica precursors also exhibit promising RO desalination performance, outstanding hydrothermal stability, and resilience to chlorine exposure. For instance, BTESE-based membranes demonstrated NaCl rejection exceeding 98% and exceptional stability under prolonged chlorine exposure at a concentration of 35,000 ppm h.¹⁴⁷ Nevertheless, the permeability of silica membranes is generally 1 to 2 orders of magnitude lower than that of commercial polyamide membranes (10⁻¹² to 10⁻¹¹ m³ m⁻² s⁻¹ Pa⁻¹) because of their smaller rigid pores, more pronounced hydrophobicity, and thicker separation layers.¹⁴⁸

A water-permeable network with improved porous and hydrophilic properties can be obtained by employing organosilica precursors containing rigid, hydrophilic bridging moieties. However, the NaCl rejection slightly decreased compared with BTESE membranes.^149–152 Xu et al.¹⁵³ synthesized membranes from organosilica precursors with similar frameworks but varying polaritiesy and rigidity of bridging groups, i.e. BTESE, 1,2-bis(triethoxysilyl)ethylene (BTESEthy), and bis(triethoxysilyl)acetylene (BTESA), which are bridged by ethane (−CH₂−CH₂−), ethylene (−CH=CH−), and acetylene (−C≡C−), respectively. Figure 14a and Table 2 show positron annihilation lifetime (PAL) results and average lifetimes (τPs) for the ortho-Positronium (o-Ps) of BTESE, BTESEthy, and BTESA. The maximum o-Ps values for the BTESE and BTESEthy powders were larger than 20 ns, indicating the presence of both micropores and mesopores. In contrast, BTESA only had micropores, suggesting its more uniform pore structure. Moreover, the calculated average micropore sizes were 0.41 nm for BTESE and BTESEthy, and 0.53 nm for BTESA. Quantum chemical calculations (Fig. 14b) reveal that the C–C distance increases from BTESE to BTESA, indicating a more open-pore structure. Notably, the C–H bonds in the BTESE and BTESEthy pores increase the permeation resistance. The electrostatic potentials (ESPs) of the BTESA and BTESEthy showed higher potential contrast than that of BTESE inside the network rings, which enhanced the water affinity through hydrogen bonding and/or dipole–dipole interactions.

Fig. 14.

a) PAL for the BTESE, BTESEthy, and BTESA powders, b) ESPs for models of BTESE, BTESEthy, and BTESA networks, and c) trade-off of water permeability versus salt rejection for the organosilica-derived RO membranes, adapted with permission from Ref.¹⁵³ d) Schematic of in situ carboxyl functionalization of BTESEthy, e) ESPs mapped for BTESEthy and BTESEthy-MSA, f) RO desalination performance of BTESEthy and BTESEthy-MSA membranes at 25 °C, and g) comparison of RO desalination performance with that of state-of-the-art membranes, adapted with permission from Ref.¹⁸

Table 2.

Average lifetimes of o-Ps for the BTESE, BTESEthy, and BTESA powders.

Sample	τPs−1 [ns]	τPs−2 [ns]	τPs−3 [ns]
BTESE	1.24	7.16	20.2
BTESEthy	1.19	6.44	23.8
BTESA	1.03	4.31	12.0

Adapted with permission from Ref.¹⁵³

The larger effective pore size in the acetylene-bridged organosilica network of the BTESA membrane yielded a higher molecular weight cutoff (MWCO) of ∼100 Da, comparable to that of the commercial polyamide SW30HR membrane,¹⁵⁴ in contrast to the ∼50 Da MWCO of the BTESE and BTESEthy membranes. An increase in the degree of unsaturation of the bridging carbon bonds enhanced their polarity and rigidity, promoting the formation of a hydrophilic and open network. Consequently, the water permeability of the BTESA membrane reached 8.5 × 10⁻¹³ m³ m⁻² s⁻¹ Pa⁻¹, which is approximately 8 times higher than that of the BTESE membrane (Fig. 14c).

In addition to selecting organosilica precursors with various bridging units, modifications to the existing precursor structure are anticipated to facilitate the synergistic enhancement of both water permeability and salt rejection in the RO process. For instance, a BTESEthy-MSA organosilica precursor was synthesized via in situ carboxylation via a thiol-ene click reaction between the BTESEthy double bonds and mercaptosuccinic acid (MSA),¹⁸ as shown in Fig. 14d. Quantum chemical calculations were conducted to analyze the ESPs and microstructures of the BTESEthy and BTESEthy-MSA networks, as depicted in Fig. 14e. The distance between neighboring carbon atoms in −C=C− in the BTESEthy-MSA network was shorter than that in BTESEthy, suggesting a more compact network structure. This is due to the transformation of sp² to sp³ hybridization at the reaction site (C1), which reduces the membrane pore size. Furthermore, the introduction of −COOH groups enhances the negative potential around the C1 atoms, which tends to strengthen the interactions between the membrane network and cations, such as Na⁺ and H₃O⁺. The BTESEthy-MSA membranes exhibited twice the permeability of the BTESEthy membranes with a modest increase in water/salt selectivity during the RO desalination process (Fig. 14f). Additionally, the carboxyl-functionalized organosilica membrane demonstrated excellent durability over a continuous 200 h RO desalination cycle. As illustrated in Fig. 14g, compared with the MFI-type zeolite (ZSM-5 and silicalite) membranes, which have been widely explored and demonstrated excellent performance in RO desalination,^155,156 both BTESEthy and BTESEthy-MSA membranes exhibited comparable NaCl rejection (95% to 99%) while demonstrating significantly higher water permeability (10⁻¹³ m³ m⁻² s⁻¹ Pa⁻¹). Notably, the BTESEthy-MSA membrane exhibited superior permeability, which was attributed to its enhanced hydrophilicity, highlighting its promising prospects for applications.

5. Conclusion and perspective

This review highlights significant advancements in silica membranes, with a particular emphasis on those derived from the sol–gel process, which have been successfully commercialized in membrane fabrication for solvent dehydration applications. The focus is placed on innovative microstructural engineering strategies designed to achieve preferable membrane structures and surface properties, leading to improved separation performance and hydrothermal stability. Various strategies, including spacer and template technologies using organosilica precursors, ion doping, composite ceramic incorporation with metal oxides, and carbon contained ceramic composites, have demonstrated substantial improvements in both membrane performance and stability. Furthermore, this review discusses the diverse applications of silica-derived membranes in gas- and liquid-phase separation.

We anticipate that silica membranes will continue to thrive in practical applications; however, some challenges need to be considered for their further advancement. Current research on silica-membrane structures relies predominantly on dry xerogel powders produced using the same preparation processes as those used for membrane fabrication. Therefore, the development of advanced in-situ characterization techniques is crucial for accurately analyzing the real structure of silica membranes. Additionally, while machine learning and theoretical simulations offer valuable insights into membrane design, the amorphous nature of silica complicates the development of reliable models. Therefore, robust amorphous silica models and theoretical simulations of mass transfer processes are urgently required. Finally, achieving precise control over the pore size for specific applications, along with improving the thermal stability, remains an ongoing challenge. Notably, the carbon contained ceramic composite membranes recently explored by our group have demonstrated great potential in these areas; therefore, their broader applicability in various separation systems warrants further investigation.

Acknowledgments

X.N. sincerely acknowledges the financial support from the China Scholarship Council (CSC, no. 202307040023).

Funding

None declared.

Data availability

All relevant data in this review are available from the references.

Masakoto Kanezashi

Masakoto Kanezashi obtained his PhD degree from Hiroshima University (Japan) in 2005. He subsequently joined Arizona State University (USA) as a postdoctoral research fellow before returning to Hiroshima University in 2007 as an assistant professor. He was promoted to professor in 2022 and currently leads the Separation Technology Laboratory. Prof. Kanezashi is a recognized expert in the development of membranes derived from porous materials, with a focus on evaluating their separation properties for various applications, ranging from gas to liquid separations. His research primarily centers on the control of microporous structures, particularly at the sub-nano level, with an emphasis on pore size and adsorption properties using amorphous silica.

Xinpu Niu

Xinpu Niu earned his bachelor's degree in Chemical Engineering and Technology in 2020 and his master's degree in Chemical Technology in 2023 from Zhengzhou University (China). He is currently a Ph.D. student at Hiroshima University (Japan). His research focuses on the regulation of microstructure and surface properties of silica-based ceramic membranes and their applications in molecular separation.

Author notes

Conflict of interest statement. None declared.