The Use of Modern Hybrid Membranes for CO₂ Separation from Synthetic and Industrial Gas Mixtures in Light of the Energy Transition

Abstract

The global energy transition and the implementation of carbon capture, utilization, and storage (CCUS) strategies require energy-efficient and scalable CO₂ separation technologies. Mixed-matrix membranes (MMMs), combining polymer matrices with functional inorganic or hybrid nanofillers, have emerged as advanced separation platforms capable of surpassing the conventional permeability–selectivity trade-off observed in neat polymer membranes. This review critically evaluates recent developments in modern hybrid membranes for CO₂ separation from synthetic and industrial gas mixtures, including CO₂/N₂ (flue gas), CO₂/CH₄ (natural gas and biogas upgrading), and syngas systems. Particular emphasis is placed on MMMs incorporating covalent organic frameworks (COFs), metal–organic frameworks (MOFs), graphene oxide (GO), MXenes, transition metal dichalcogenides (TMDs), carbon nanotubes (CNTs), g-C₃N₄, layered double hydroxides (LDH), zeolites, metal oxides, and magnetic nanoparticles. Reported performance ranges include CO₂ permeability (P_CO2) typically between 100 and 800 Barrer, CO₂/N₂ selectivity up to 319, and CO₂/CH₄ selectivity up to 249, depending on filler chemistry, loading, and interfacial compatibility. The mechanisms governing gas transport—molecular sieving, selective adsorption, facilitated transport, and diffusion-pathway engineering—are systematically discussed. Key challenges addressed include filler dispersion, polymer–filler interfacial defects, physical aging, moisture sensitivity, oxidation (particularly in MXenes), and scalability toward industrial membrane modules. Future perspectives focus on sub-nanometer pore engineering, surface functionalization to enhance CO₂ affinity, controlled alignment of 2D nanosheets to promote directional transport, multifunctional core–shell and hollow structures, and the integration of computational modeling and machine learning for accelerated material design. Modern hybrid MMMs are identified as strategically important materials enabling high-efficiency CO₂ separation processes aligned with decarbonization and energy transition objectives.

1. Introduction

The continuous growth of industrial activity and global energy demand has led to a substantial increase in greenhouse gas emissions, among which carbon dioxide (CO₂) is the most dominant contributor to climate change [1]. The excessive release of CO₂, primarily originating from fossil fuel combustion in power generation, transportation, and industrial processes, has intensified the greenhouse effect and accelerated global warming, resulting in rising average temperatures, extreme weather events, melting glaciers, and ecosystem degradation [2,3]. Consequently, the development of efficient CO₂ capture and separation technologies has become an urgent environmental and technological priority. Several technologies have been developed for CO₂ capture, including chemical absorption, physical adsorption, cryogenic separation, and membrane-based separation [4]. Absorption processes, particularly amine scrubbing, are currently the most mature and widely implemented at industrial scale; however, they suffer from high energy consumption, solvent degradation, corrosion issues, and substantial operational costs [5]. Adsorption-based systems often involve complex regeneration steps and batch operation, while cryogenic separation requires extremely high energy input due to refrigeration demands, limiting its economic feasibility [6]. In contrast, membrane separation has emerged as a promising alternative owing to its operational simplicity, modularity, low energy demand, small footprint, and environmental compatibility [7,8]. Membrane-based gas separation relies on differences in permeability and diffusivity of gas components through a selective barrier. Compared with conventional separation methods, membranes enable continuous operation, linear scalability, and facile integration into existing industrial processes [9]. Based on material composition, CO₂ separation membranes are generally classified into inorganic membranes, polymer membranes, and composite membranes [10,11]. Inorganic membranes exhibit excellent thermal and chemical stability as well as high gas permeability; nevertheless, their widespread application is constrained by high fabrication costs, brittleness, and poor processability [12]. Polymeric membranes, on the other hand, are cost-effective and easy to manufacture, but their separation performance is fundamentally limited by the permeability–selectivity trade-off, commonly described by the Robeson upper bound (2008) [13]. Additionally, polymeric membranes are prone to physical aging and plasticization in the presence of CO₂ and other condensable gases, leading to long-term performance degradation [8]. To overcome these limitations, mixed matrix membranes (MMMs) have been extensively investigated as an advanced membrane architecture. MMMs combine the mechanical flexibility and processability of polymers with the superior transport properties of inorganic or hybrid fillers, offering a viable strategy to surpass the conventional trade-off effect [14]. However, the separation performance of MMMs strongly depends on filler characteristics such as morphology, thickness, surface chemistry, dispersion uniformity, and interfacial compatibility with the polymer matrix [15]. Poor interfacial adhesion may result in non-selective voids or agglomeration, adversely affecting membrane performance and mechanical integrity [16]. In recent years, various nanomaterials have attracted increasing attention as highly efficient fillers for MMMs [17]. Owing to their atomic-scale thickness, large lateral dimensions, and high aspect ratios, 2D materials can provide fast and selective gas transport pathways while minimizing mass transfer resistance [18]. Commonly explored fillers include zeolites, graphene and its derivatives, MXenes, nanocarbons and carbon nanotubes (CNTs), graphitic carbon nitride (g-C₃N₄), layered double hydroxides (LDHs), metal–organic frameworks (MOFs), magnetic nanofillers, covalent organic frameworks (COFs), oxide nanoparticles, layered double hydroxides, and transition metal dichalcogenides (TMDs) [19,20,21,22,23,24,25,26]. MMMs incorporating these materials have demonstrated significantly enhanced CO₂ permeability and selectivity, showing strong potential to exceed the Robeson upper bound [27]. Despite these advantages, challenges remain in the large-scale fabrication of hybrid membranes, including nanosheet restacking, filler alignment control, long-term stability, and reproducibility [28]. Addressing these issues requires advances in materials design, surface functionalization, membrane processing techniques, and predictive modeling approaches, including machine learning-assisted membrane optimization [29].

This review begins by outlining the fundamental mechanisms of membrane-based gas separation, with particular attention paid to different sources of CO₂ and the associated separation conditions. Subsequently, the main technologies for CO₂ separation are systematically discussed, followed by an in-depth analysis of separation parameters and transport mechanisms governing gas permeation in composite membranes, including key criteria for membrane material selection. A comprehensive overview of polymeric, inorganic, and hybrid membranes used for CO₂ separation is then presented. Hybrid membranes are classified and described according to the type of incorporated fillers, such as zeolites, graphene and graphene oxide, nanocarbons and carbon nanotubes (CNTs), magnetic nanofillers, metal–organic frameworks (MOFs), MXene-based fillers, g-C₃N₄-based materials, oxide nanoparticles, layered double hydroxides (LDHs), and transition metal dichalcogenides (TMDs). Particular emphasis is placed on recent advances in nanomaterial-based composite membranes for CO₂ separation, highlighting their advantages, limitations, and emerging performance trends. In addition, issues related to membrane module configurations, process design, and industrial applications of membrane-based CO₂ separation are critically addressed. The review also explores the growing role of big data and machine learning approaches in the design, optimization, and performance prediction of CO₂ separation membranes. Finally, the remaining challenges and future development directions toward industrially viable and scalable CO₂ separation membranes are critically analyzed.

2. Different Sources of CO₂ and Associated Separation Conditions

At the beginning of the 20th century, coal was the fundamental energy source powering heat production, electricity generation, and transportation. It shaped industrial development in Europe, the United States, China, India, and many other regions. Its role in economic growth, technological progress, and improved living conditions was undeniable. However, long-term dependence on coal resulted in significant environmental impacts, particularly high emissions of CO₂ and other pollutants. As climate awareness increased, the need for a global energy transition—a systemic shift towards low-emission and sustainable energy sources—became unavoidable [30]. Throughout the last decades, many countries have reduced the share of coal in their energy mixes. European Union states such as Portugal, Germany, and the Czech Republic show clear long-term declines in coal consumption. Meanwhile, countries including China and India continue to rely on coal for over 50% of primary energy, driven by rapidly growing demand and economic expansion. Because China alone accounts for more than half of global coal consumption, global decarbonization efforts cannot be fully successful without low-carbon solutions deployed in coal-dependent regions [31]. The transition toward climate neutrality—targeted for 2050 in the EU and USA, 2060 in China, and 2070 in India—requires two parallel approaches. The first path is the rapid expansion of renewable energy sources (RES), nuclear power, and, where feasible, natural gas as a transitional fuel. This is the leading strategy in most EU countries, although it remains challenging for states still strongly dependent on coal for electricity generation [32]. The second path involves integrating clean coal technologies (CCT) and advanced emission-reduction systems to minimize environmental impacts where coal continues to play a key role in ensuring energy security. In the context of global energy transition, CCTs are part of a broader portfolio of clean technologies that support controlled decarbonization, stabilize power systems, and help countries meet international climate goals such as the Paris Agreement, the UN Sustainable Development Goals (particularly SDG 7), and the European Green Deal [33,34,35]. Interest in CCT and low-carbon technologies has grown dramatically over the last two decades, especially in countries where coal remains essential. Despite this, combustion inevitably produces CO₂, SO_x, NO_x, and particulate matter—making it critical to incorporate advanced CO₂ mitigation systems. These include pre-combustion and post-combustion capture, oxy-fuel combustion, gas purification, and increasingly, membrane-based and hybrid separation technologies designed to isolate CO₂ efficiently and with lower energy consumption [36,37,38]. In the broader framework of the energy transition, clean technologies—whether tied to fossil fuels, renewable energy systems, or industrial decarbonization—are indispensable tools for reducing global CO₂ emissions. They bridge the gap between today’s fossil-dependent energy systems and the fully decarbonized structures envisioned for mid-century. For coal-reliant countries, they offer a realistic pathway to stability, energy security, and gradual but effective reduction in greenhouse gas emissions. Carbon dioxide emissions originating from fossil fuel-based energy production remain the dominant contributor to anthropogenic greenhouse gas releases (

Table 1. Overview of major CO₂ sources and separation conditions.

CO2 Source/Sector	Typical CO2 Concentration	Separation Route/Context	Key Advantages	Main Limitations	Key Challenges & Research Needs
Fossil fuel power generation	4–15% (post-combustion flue gas)	Post-combustion capture	Retrofit compatibility; flexible operation; mature infrastructure	Low CO2 partial pressure; large gas volumes; high energy penalty	Development of high-permeability/selective membranes; reduction in capture energy
	5–25% (syngas)	Pre-combustion (IGCC)	High CO2 concentration; elevated pressure; strong driving force	High capital cost; complex system design	Cost reduction; simplified gasification systems
	>90%	Oxy-fuel combustion	Produces high-purity CO2 directly	Very energy-intensive oxygen separation	Energy-efficient air separation units
Cement industry	14–33%	Post-combustion/process emissions	Large point sources; continuous operation	CO2 from both fuel combustion and calcination	Capture under harsh conditions; process integration
Iron & steel industry	20–27%	Post-combustion/process gas	Concentrated emission points	Carbon-based reduction inherently emits CO2	Alternative reduction pathways; CCS integration
Transport—road	Low, dispersed	Indirect (via electrification)	Emission shift to centralized sources	Depends on grid decarbonization	Clean electricity generation
Transport—maritime	5–15%	Onboard CCS	Enables decarbonization of long-distance shipping	Space, weight, and energy constraints	Compact, low-energy capture systems
Natural gas industry	4–20%	Pre-processing separation	Largest industrial CO2 separation application; high pressure	Handling large CO2 volumes	CCS chain integration; cost-effective membranes
Biogas upgrading	30–50%	CH4 enrichment	Renewable energy source; potential negative emissions	Variable gas composition	CO2 utilization/storage; robust separation materials
Small/distributed sources (agriculture, waste)	Variable	Often overlooked	Large cumulative mitigation potential	Low concentration; decentralized nature	Modular, low-cost capture technologies

). Consequently, the deployment of carbon capture and storage (CCS) technologies within this sector has been widely recognized by the scientific and engineering communities as one of the most effective pathways to achieving internationally agreed climate mitigation targets [39]. Over the past two decades, extensive efforts by academic researchers, industrial consortia, and governmental agencies have focused on implementing CCS in large point-source emitters, including coal- and gas-fired power plants, cement kilns, iron and steel production facilities, and upstream oil and gas operations.

Despite these efforts, climate and energy researchers have emphasized that the global pace of CCS deployment remains insufficient relative to the scale of emissions reductions required [40]. Several industrial sectors are only partially decarbonized, while others—particularly those characterized by distributed or small-scale emission points—remain largely unaddressed. This mismatch highlights the necessity of tailoring CO₂ separation strategies to the specific composition, pressure, temperature, and impurity profile of each emission source, so that the inherent advantages of individual capture technologies can be translated into both economic and technical benefits [41]. Importantly, enabling CO₂ separation from currently overlooked sources, such as agricultural processes and transportation-related emissions, is increasingly viewed as a critical step toward comprehensive decarbonization. Among all emission categories, power generation accounts for the largest share of global CO₂ output. In this sector, three principal capture routes have been established by the CCS research community—oxy-fuel combustion, pre-combustion capture, and post-combustion capture—which differ fundamentally in the stage at which CO₂ is separated from the process stream [11]. Each route imposes distinct separation conditions and presents specific trade-offs. Oxy-fuel combustion relies on burning fossil fuels in a near-pure oxygen atmosphere, producing an exhaust stream composed primarily of CO₂ and water vapor. This approach enables the generation of a high-purity CO₂ stream with minimal downstream separation requirements. However, studies in air separation and process integration have consistently shown that the energy penalty associated with large-scale oxygen production remains a major drawback [42]. Demonstration projects such as Aker Solutions’ ZEUS concept in Norway—where natural gas is combusted with pure oxygen in a subsea environment and the resulting CO₂ is directly reinjected—illustrate the technical feasibility of this route but also underscore its dependence on energy-intensive oxygen separation units [43]. As CO₂ purification is largely unnecessary in this case, oxy-fuel systems are often excluded from comparative membrane-based separation analyses. Pre-combustion capture, extensively studied within the integrated gasification combined cycle (IGCC) framework, involves converting carbonaceous fuels into synthesis gas composed mainly of CO, H₂, and CO₂. After water–gas shift conversion, CO₂ is separated from a hydrogen-rich stream, which is subsequently used for power generation or chemical synthesis [44]. Researchers have highlighted several advantages of this route, including high CO₂ concentrations (typically 30–45%) and elevated operating pressures, which provide a strong thermodynamic driving force for separation and reduce the specific energy demand. Nevertheless, the complexity and high capital investment associated with IGCC infrastructure have limited its large-scale deployment. Recent work on bio-derived hydrogen suggests that biomass-based feedstocks may offer similar separation advantages under milder conditions, although supply chain and scalability challenges remain. Post-combustion capture is currently the most widely implemented CCS route, particularly because of its compatibility with existing power plants and industrial facilities. In this approach, CO₂ is separated from flue gases after combustion, making it attractive for retrofitting purposes in sectors such as cement manufacturing, steel production, and petroleum refining. However, researchers consistently point out that the low partial pressure of CO₂ and the large volumetric flow rates of flue gas—typically at near-atmospheric pressure—result in a limited driving force for separation, which translates into higher energy consumption and larger equipment sizes. Beyond power generation, cement production represents one of the most CO₂-intensive industrial processes due to both fuel combustion and the calcination reaction (CaCO₃ → CaO + CO₂). Despite incremental efficiency improvements, materials scientists and process engineers agree that cement manufacturing remains the largest industrial CO₂ emitter after the power sector [45]. Similarly, the iron and steel industry generates substantial emissions because traditional blast furnace and basic oxygen furnace routes rely heavily on coke as both a fuel and a reducing agent. As long as carbon-based reductants are used to remove oxygen from iron ore, the potential for deep emissions reductions remains constrained [46]. The transport sector also constitutes a growing source of CO₂ emissions. While electrification of road transport can shift emissions toward centralized power generation, researchers note that this strategy is far less practical for maritime transport due to the current limitations of high-capacity batteries and alternative carbon-free fuels. Consequently, onboard CCS has emerged as a critical option for decarbonizing long-distance shipping. Reports from the International Maritime Organization emphasize that CO₂ capture technologies must be integrated into vessels to achieve carbon-neutral operation and extend the service life of existing fossil-fuel-powered fleets [47,48]. However, challenges related to space constraints, energy efficiency, and operational reliability remain unresolved. From an industrial separation perspective, natural gas sweetening represents the largest existing application of CO₂ removal. CO₂ must be extracted from raw natural gas streams to meet pipeline specifications and calorific value requirements, a process extensively studied by chemical engineers and separation technologists [10]. Given the large volumes involved, the integration of captured CO₂ from gas sweetening into broader CCS value chains is increasingly regarded as both necessary and economically attractive. Biogas upgrading constitutes another important CO₂ separation application, aimed at increasing methane content to produce pipeline-quality or vehicle-grade fuel. Depending on the feedstock, biogas typically contains 30–50% CO₂ and 35–70% CH₄ [12]. Energy and environmental researchers have highlighted that biogas upgrading offers a unique opportunity for negative emissions if the separated CO₂ is captured and permanently stored or utilized, thereby coupling renewable energy production with CCS.

Overall, the diversity of CO₂ sources and separation conditions underscores the need for flexible, source-specific capture technologies. While significant progress has been achieved, ongoing challenges—including energy efficiency, cost reduction, scalability, and integration with existing infrastructure—continue to motivate research into advanced materials and process designs, particularly membrane-based and hybrid separation systems.

3. Technologies for CO₂ Separation

Carbon dioxide separation has been the subject of extensive research over recent decades, leading to the development of several mature and emerging technologies, including chemical absorption, adsorption, membrane-based separation, and cryogenic processes. Owing to the wide diversity of CO₂ sources—ranging from high-pressure, high-concentration streams to dilute atmospheric emissions—no single separation technology can be universally applied. As highlighted by multiple CCS-focused research groups, including those led by Buckingham and Wang, the selection of an appropriate CO₂ capture technology must be guided by detailed techno-economic assessments and, in many cases, by the integration of complementary separation methods to overcome the limitations of individual approaches [41,49].

Among existing technologies, chemical absorption remains the most commercially established option, particularly for post-combustion CO₂ capture. Seminal and ongoing studies by Asif and co-workers have demonstrated that aqueous amine-based solvents, such as monoethanolamine (MEA), provide high CO₂ selectivity and reliable capture efficiency even at low CO₂ partial pressures [49,50]. Nevertheless, the widespread deployment of chemical absorption is constrained by several inherent drawbacks. These systems require large absorber and stripper columns due to slow mass transfer kinetics, resulting in high capital costs. Furthermore, solvent regeneration imposes a substantial energy penalty, while solvent degradation, corrosion, and solvent loss pose environmental and operational challenges [41,50]. Despite continuous improvements in solvent chemistry and process integration, these fundamental disadvantages remain difficult to fully eliminate. Adsorption-based CO₂ separation, extensively investigated by research groups led by Kolle, He and Pullumbi, relies on selective gas–solid interactions using porous sorbents such as zeolites, activated carbons, and metal–organic frameworks [51,52,53]. Pressure swing adsorption (PSA) and temperature swing adsorption (TSA) are the most common operational modes. Adsorption systems offer several advantages over absorption, including reduced corrosion, mechanical simplicity, and the capability to achieve ultra-high gas purities (e.g., >99.99%), which are difficult to obtain using liquid solvents [53,54]. However, adsorption-based CO₂ capture is limited by relatively low working capacities, high sensitivity to moisture, and the need for extensive gas pretreatment. Additionally, frequent regeneration cycles increase energy consumption and reduce overall process efficiency, particularly for low-CO₂-concentration streams [52]. A specific application of sorption-based technology is direct air capture (DAC), which has gained significant attention as a pathway toward achieving negative CO₂ emissions. Research led by McQueen et al., demonstrates that DAC systems typically rely on chemisorption due to the extremely low atmospheric CO₂ concentration (~400 ppm). While DAC provides a unique solution for addressing diffuse emissions from sectors such as agriculture and transportation, it remains technologically and economically challenging. Key unresolved issues include sorbent durability, regeneration energy demand, and large-scale deployment feasibility [55].

Cryogenic separation is a less frequently applied but technically viable CO₂ separation method. This approach involves staged compression, cooling, expansion, and phase separation, enabling CO₂ liquefaction or solidification. Cryogenic processes have been investigated by several process engineering groups, particularly for gas streams with high CO₂ concentrations. However, due to the substantial energy demand associated with compression and deep cooling, cryogenic separation is generally considered economically unfavorable as a standalone technology. Consequently, it is more commonly proposed as a downstream polishing or liquefaction step in hybrid separation schemes, such as membrane–cryogenic or adsorption–cryogenic processes [56]. In contrast, membrane-based CO₂ separation has emerged as one of the most intensively studied technologies over the past two decades. Extensive contributions from the research groups of Robeson, Freeman, Koros, Bernardo, and Baker have established membranes as a promising alternative for CO₂ capture and gas separation applications [57,58]. Membrane processes offer several intrinsic advantages, including compact and lightweight system design, high modularity, ease of scale-up, and continuous operation without the need for chemical additives. Unlike absorption and adsorption processes, membrane separation does not require energy-intensive regeneration steps, making it an environmentally benign and operationally simple technology [58]. From an industrial perspective, membrane systems are particularly attractive for small- to medium-scale applications due to their relatively low capital and operating costs and their suitability for decentralized deployment. However, despite these advantages, membrane-based CO₂ separation still faces significant challenges that hinder large-scale industrial adoption. As emphasized in several critical reviews, the technology readiness level (TRL) of most CO₂-selective membranes remains below 5, indicating a substantial gap between laboratory-scale research and commercial implementation [56,57]. Key issues include the permeability–selectivity trade-off described by Robeson, long-term material stability, plasticization, fouling, and performance degradation under realistic operating conditions.

To address these challenges, hybrid and integrated separation technologies have been proposed, combining membranes with other separation principles to exploit synergistic effects. Examples include membrane–absorption systems and membrane-assisted DAC concepts, which have been explored by multiple research groups [59,60,61,62,63]. While these hybrid approaches show promise in overcoming individual process limitations, they introduce additional complexity and are still at an early stage of development. Consequently, they remain outside the scope of the present work.

4. Separation Parameters and Transport Mechanisms in Composite Membranes

Gas transport through membranes is governed by several distinct mechanisms, which depend primarily on membrane morphology, pore size distribution, and the physicochemical properties of the permeating gases. For porous membranes, the dominant transport mechanisms include convective (viscous) flow, Knudsen diffusion, and molecular sieving, whereas dense membranes operate mainly via the solution–diffusion mechanism (Figure 1). In certain systems, particularly CO₂-selective membranes, facilitated transport may additionally contribute to gas permeation.

4.1. Transport Mechanisms in Porous Membranes

In porous membranes and porous supports of composite membranes, gas transport mechanisms depend strongly on pore size. When pores are much larger than the mean free path of gas molecules, transport occurs via non-selective convective flow. As pore sizes decrease below the mean free path, Knudsen diffusion becomes dominant, with selectivity determined by molecular weight differences rather than chemical affinity. Because Knudsen selectivity is generally insufficient for CO₂ separation, this mechanism is of limited practical relevance. When pore dimensions approach molecular sizes, molecular sieving becomes the prevailing mechanism. In this case, separation is achieved based on size exclusion, allowing smaller molecules to permeate while larger ones are rejected. Microporous materials such as metal–organic frameworks (MOFs), covalent organic frameworks (COFs), and graphene-based fillers are particularly effective in enabling molecular sieving when incorporated into mixed-matrix or composite membranes. When the pore diameter of a membrane is significantly larger than the mean free path of gas molecules (typically >0.1 μm), gas transport occurs via convective diffusion (viscous flow). In this regime, gas molecules primarily collide with each other rather than with pore walls, leading to non-selective transport driven by pressure gradients. Consequently, convective diffusion does not provide meaningful selectivity for gas separation and is unsuitable for CO₂-selective membranes [64].

As the pore size decreases below the mean free path of gas molecules (approximately 0.01–0.1 μm), Knudsen diffusion becomes the dominant mechanism. In this case, gas molecules collide more frequently with the pore walls than with other gas molecules. The Knudsen diffusion coefficient for gas i is given by [65]:

$$D_{K,i}={\displaystyle \frac{1}{3}}{d}_p{\left({\displaystyle \frac{8RT}{\pi M_i}}\right)}^{1/2}$$

(1)

where $d_p$ is the pore diameter, R is the universal gas constant, T is the absolute temperature, and $M_{i }$ is the molar mass of gas i. The ideal Knudsen selectivity between gases i and j depends solely on their molar masses:

$${\alpha }_{ij}^K={\left({\displaystyle \frac{M_j}{M_i}}\right)}^{1/2}$$

(2)

Because CO₂, N₂, and CH₄ have relatively similar molecular weights, Knudsen diffusion provides only limited selectivity and is therefore inadequate for efficient CO₂ separation.

When membrane pore sizes approach the kinetic diameters of gas molecules (typically 0.3–1 nm), molecular sieving becomes operative. In this regime, only molecules smaller than the effective pore size can permeate through the membrane, while larger molecules are sterically excluded. Molecular sieving offers high selectivity and is the most effective porous-membrane mechanism for CO₂ separation, particularly in rigid inorganic membranes and certain carbon-based membranes [66].

In addition, surface diffusion may contribute to gas transport when condensable gases such as CO₂ adsorb onto pore walls and migrate between adsorption sites. This mechanism becomes significant in microporous membranes with strong gas–surface interactions [65].

4.2. Gas Transport in Dense Polymeric Membranes: Solution–Diffusion Mechanism

The performance of gas separation membranes is commonly assessed using two fundamental parameters: permeability (or permeance) and selectivity. These parameters quantify the ability of a membrane to transport a given gas species and to discriminate between different components in a gas mixture [57,67]. The permeability of a gas is defined as the product of its diffusivity and solubility within the membrane material. Diffusivity is largely governed by the kinetic diameter of gas molecules and the available free volume within the polymer matrix, while solubility depends on gas condensability and affinity toward the polymer. CO₂, characterized by a relatively high critical temperature and moderate kinetic diameter, typically exhibits higher solubility than N₂ and CH₄, enabling effective CO₂ separation via the solution–diffusion mechanism. However, separating CO₂ from H₂ is more challenging due to the high diffusivity of H₂, despite its low solubility. As membrane pore sizes decrease further and approach the scale of polymer segmental motion, membranes transition from porous to dense structures. In dense polymeric membranes, gas transport is predominantly governed by the solution–diffusion mechanism, which consists of three sequential steps: (i) sorption of gas molecules at the feed side, (ii) diffusion through the polymer matrix, and (iii) desorption at the permeate side [64]. The performance of gas separation membranes is quantitatively evaluated using several key parameters that describe gas transport and separation efficiency. These parameters are derived from the fundamental relationships governing mass transfer through membranes and are applicable to dense, asymmetric, and composite membrane systems [58,67].

The gas flux of component i (J_i) represents the amount of gas permeating through a unit membrane area per unit time and is defined as [58]:

$$J_i={\displaystyle \frac{n_i}{A\cdot t}}$$

(3)

where n_i is the amount of gas permeated (mol), A is the effective membrane area (m²), and t is the permeation time (s).

The steady-state flux of gas species i through a dense membrane can be expressed as:

$$J_i={\displaystyle \frac{D_i}{l}}\left(C_{i,f}-C_{i,p}\right)$$

(4)

and

$$J={\displaystyle \frac{Q_{STP}}{A}}$$

(5)

where $D_i$ is the diffusion coefficient, l is the membrane thickness, and $C_{i,f}$ and $C_{i,p}$ are the gas concentrations on the feed and permeate sides, respectively, Q_STP is a flow rate at standard conditions [cm³_STP/s].

For membrane separation processes driven by a partial pressure difference, the permeance (Π_i) is defined as the pressure-normalized gas flux:

$${\Pi }_i={\displaystyle \frac{J_i}{\Delta p_i}}$$

(6)

where Δp_i is the partial pressure difference of gas i across the membrane (Pa). Permeance is commonly reported in gas permeation units (GPU), particularly for asymmetric and composite membranes where the selective layer thickness cannot be precisely determined [67].

The intrinsic transport property of a membrane material is expressed by permeability (P_i), which accounts for the membrane thickness [57]:

$$P_i={\Pi }_i\cdot l={\displaystyle \frac{J_i\cdot l}{\Delta p_i}}$$

(7)

where l is the thickness of the selective layer (m). Permeability is typically reported in Barrer and enables direct comparison between different membrane materials.

For dense, homogeneous membranes with a well-defined thickness, gas permeability is the preferred descriptor of transport performance, as it reflects the intrinsic material property of the membrane [57]. In contrast, for asymmetric or composite membranes, where the effective thickness of the selective layer may be difficult to determine, permeance is more commonly used. The two quantities are related through the thickness of the selective layer, emphasizing the strong influence of membrane morphology on separation performance.

Selectivity provides a measure of separation efficiency and is most often defined as the ratio of permeabilities of two pure gases (ideal selectivity). In practical applications, especially for multilayer composite membranes or membrane modules, separation performance is better represented by the separation factor, which compares the compositions of the permeate and feed streams [57,67,68].

Membrane selectivity is a measure of the ability of a membrane to separate two gas species. The ideal selectivity (α*_ij) is defined as the ratio of pure-gas permeabilities [57]:

$${\alpha }_{ij}^{\ast }={\displaystyle \frac{P_i}{P_j}}$$

(8)

where i and j denote two different gas components. For gas mixtures and membrane modules, the separation factor (α_ij) provides a more realistic measure of separation performance [67]:

$${\alpha }_{ij}={\displaystyle \frac{\left(y_i/y_j\right)}{\left(x_i/x_j\right)}}$$

(9)

where x_i and x_j are the mole fractions of components i and j in the feed stream, and y_i and y_j are their mole fractions in the permeate stream.

In dense polymeric membranes, gas transport is predominantly governed by the solution–diffusion mechanism [57,67]. According to this model, permeability is the product of the gas diffusivity (D_i) and solubility (S_i) in the membrane [57]:

$$P_i=D_i\cdot S_i$$

(10)

The diffusivity reflects the mobility of gas molecules within the polymer matrix and depends primarily on the kinetic diameter of the gas and the free volume of the membrane. The solubility describes the affinity of the gas toward the membrane material and is related to the gas condensability and polymer–gas interactions [69].

Here, diffusivity depends primarily on the kinetic diameter of the gas and the free volume of the polymer, whereas solubility is governed by gas condensability and polymer–gas interactions. Polar and condensable gases such as CO₂ typically exhibit higher solubility in polymer matrices than non-polar gases [70].

The ideal selectivity between gases i and j is defined as [59]:

$${\alpha }_{ij}={\displaystyle \frac{P_i}{P_j}}={\displaystyle \frac{D_i}{D_j}}\cdot {\displaystyle \frac{S_i}{S_j}}$$

(11)

This expression highlights that membrane permselectivity arises from contributions of both diffusivity selectivity and solubility selectivity, a principle particularly relevant for polymeric and mixed-matrix membranes (MMMs). In composite and mixed-matrix membranes, overall separation performance results from the combined contributions of diffusivity selectivity and solubility selectivity. The incorporation of porous fillers, reactive carriers, or nanostructured materials can introduce new transport pathways and enhance CO₂ separation beyond the limits of conventional polymer membranes. However, membrane morphology, interfacial compatibility, and transport mechanisms must be carefully optimized to fully exploit these advantages.

4.3. Facilitated Transport Mechanisms

Beyond the classical solution–diffusion mechanism, CO₂ transport in polymeric and composite membranes can be significantly enhanced through facilitated transport when the membrane incorporates reactive carrier sites (

Table 2. Comparison of Facilitated Transport Carrier Systems.

Type of Carrier	Description	Advantages	Limitations/Challenges	Common Filler Systems
Fixed carriers	Carrier groups covalently bound to polymer chains; immobilized amines, IL fragments, metal complexes.	High structural stability; no leaching; consistent long-term selectivity.	Limited mobility reduces carrier-mediated flux; potential steric constraints.	Amine-grafted polymers, IL-functionalized backbones, immobilized metal complexes [72,74].
Mobile carriers	Free-diffusing carriers dispersed within the membrane (amines, ILs, task-specific ILs).	High mobility → higher CO2 transport rates; chemically tunable.	Risk of leaching; phase separation; lower long-term stability depending on polymer compatibility.	Supported ionic liquid membranes, polymer–IL blends, mobile amine systems [72,73].
Semi-mobile carriers	Carriers constrained but not fully immobilized by fillers (nanoparticles, MOFs, IL-infused particles).	Balance of mobility and stability; reduced leaching; enhanced interfacial transport.	Diffusion limitations due to filler density; performance highly dependent on dispersion quality.	Amine-functionalized silica, IL-infused nanofillers, reactive MOFs [74,75].

). In such systems, CO₂ undergoes reversible chemical interactions with functional groups—most commonly amines—embedded within the membrane matrix, creating an additional carrier-mediated transport pathway that improves both CO₂ permeability and selectivity [71,72]. Facilitated transport membranes may contain fixed, mobile, or semi-mobile carriers, whose effectiveness and stability depend strongly on the filler system and polymer–filler interactions [72]. The enhancement is particularly pronounced at low CO₂ partial pressures, where reversible chemical reactions compensate for the reduced driving force, while at elevated pressures carrier saturation may limit further performance gains [72,73]. In membranes containing reactive carrier sites such as amine-functionalized polymers, ionic liquids (ILs), task-specific ILs, or metal-complex carriers, CO₂ forms transient complexes (e.g., carbamates or bicarbonates) that migrate across the membrane and dissociate on the permeate side, thereby increasing selective transport [72]. The total CO₂ flux can be described as the sum of physical diffusion and carrier-mediated transport according to

$$J_{C{O}_2}={\displaystyle \frac{D_{C{O}_2}}{l}}\Delta C_{C{O}_2}+{\displaystyle \frac{D_{C{O}_2–C}}{l}}\Delta C_{C{O}_2–C}$$

(12)

where $D_{C{O}_2}$ and $D_{C{O}_2–C}$ represent the diffusivities of physically dissolved CO₂ and the CO₂–carrier complex, respectively, l is the membrane thickness, and $\Delta C$ denotes the relevant concentration gradients [72]. Fixed carrier systems rely on covalently tethered functional groups or immobilized IL fragments and metal complexes, offering high structural stability and resistance to leaching but limiting transport enhancement due to restricted carrier mobility [72,74]. Mobile carrier systems, based on freely diffusing amines or ILs dispersed within the polymer matrix, can provide higher carrier-mediated fluxes and tunable CO₂ affinity, yet may suffer from leaching, phase separation, or long-term instability depending on polymer compatibility [72,73]. Semi-mobile or hybrid systems, such as IL-infused nanoparticles, amine-functionalized silica, or MOFs with reactive sites, attempt to balance mobility and stability by partially constraining the carriers; however, their performance is highly sensitive to filler dispersion, interfacial morphology, and diffusion resistance within densely loaded filler domains [74,75]. Despite their advantages, facilitated transport membranes face important limitations, including carrier saturation at high CO₂ partial pressures when reaction kinetics exceed diffusion rates [75], diffusion constraints in filler-rich regions [73], humidity sensitivity of primary and secondary amines affecting reaction equilibria and transport pathways [72], and thermal or chemical stability issues, particularly for mobile carriers such as ILs or dissolved amines that may reorganize under operating conditions [74].

4.4. Gas Transport Models for Inorganic–Organic Hybrid Membranes

Gas permeation through inorganic–organic hybrid membranes, commonly referred to as mixed matrix membranes (MMMs), represents a highly complex transport problem due to the coexistence of phases exhibiting fundamentally different physicochemical properties. Accurate prediction of gas permeability and selectivity in such systems requires theoretical models capable of capturing the combined influence of filler content, phase morphology, interfacial interactions, and transport pathway geometry. Consequently, numerous analytical and semi-empirical models have been developed over the years to reduce experimental effort and to support the rational design of MMMs with tailored separation performance [76].

Early theoretical descriptions of gas transport in MMMs assumed an idealized two-phase structure consisting of a continuous polymer matrix and a uniformly dispersed inorganic filler phase. A key assumption of these models is a defect-free polymer–particle interface, implying perfect adhesion and continuity between the two phases. Under such conditions, gas transport can be effectively described by the classical Maxwell model, originally derived for thermal and electrical conductivity in heterogeneous media and later adapted to membrane transport phenomena [77,78]. The Maxwell approach expresses the relative permeability of a mixed matrix membrane as a function of the permeability ratio between the dispersed and continuous phases, as well as the volume fraction of the filler. The relative permeability $P_r$ is defined as:

$$P_r={\displaystyle \frac{P}{P_m}}={\displaystyle \frac{2(1-\Phi )+(1+2\Phi ){\lambda }_{dm}}{(2+\Phi )+(1-\Phi ){\lambda }_{dm}}}$$

(13)

where P is the effective gas permeability of the MMM, $P_m$ is the permeability of the polymer matrix, Φ denotes the volume fraction of the filler particles, and ${\lambda }_{dm}={\displaystyle \frac{P_d}{P_m}}$ is the permeability ratio of the dispersed phase ($P_d$) to the polymer matrix.

Despite its simplicity and analytical convenience, the Maxwell model is strictly valid only for dilute suspensions of spherical filler particles, typically for Φ < 0.2. This limitation arises from the assumption that individual filler particles do not interact and that the transport streamlines around each particle are unaffected by neighboring inclusions. Moreover, the model neglects particle size distribution, shape anisotropy, aggregation effects, and interfacial defects, which significantly restrict its applicability to real MMM systems [77,79].

In practice, achieving a perfectly bonded polymer–filler interface is extremely challenging. The incorporation of inorganic particles often alters the structure and mobility of polymer chains in the interfacial region, giving rise to non-ideal morphologies. Depending on the nature of polymer–filler interactions, several characteristic morphologies can be distinguished, including ideal dispersion, interfacial void formation, polymer chain rigidification, pore blocking, and particle agglomeration combined with pore obstruction [80,81].

Interfacial voids generally lead to an increase in gas permeability, particularly at low filler loadings, but do not substantially improve selectivity. At higher filler contents, poor particle dispersion and agglomeration may result in large voids surrounding particle clusters, causing a pronounced decrease in selectivity. In contrast, polymer rigidification near the filler surface typically reduces permeability while enhancing selectivity. Pore blocking effects, on the other hand, usually decrease permeability but may increase selectivity relative to the neat polymer matrix [79].

To extend permeability predictions beyond the dilute filler regime, the Bruggeman effective medium model has been proposed. This model accounts for higher filler volume fractions by treating both phases symmetrically and is expressed as [77]:

$$P_r^{1/3}\left({\displaystyle \frac{{\lambda }_{dm}-1}{{\lambda }_{dm}-P_r}}\right)=(1-\Phi {)}^{-1}$$

(14)

Unlike the Maxwell equation, the Bruggeman model can be applied over a broader range of Φ, although it requires numerical solution for $P_r$. Nevertheless, it still neglects particle shape, size distribution, aggregation, and interfacial resistance, limiting its predictive accuracy for membranes with complex morphologies.

Experimental validation of these classical models has shown that both Maxwell and Bruggeman approaches can qualitatively capture permeability and selectivity trends at low filler loadings. However, for Φ > 0.2, significant discrepancies between theoretical predictions and experimental data are commonly observed, reflected in increased calculation errors [81].

In recent years, increasing attention has been devoted to MMMs containing non-spherical fillers, particularly tubular structures such as carbon nanotubes. Since classical composite models were originally developed for spherical inclusions, their applicability to nanotube-based systems is limited. One of the first models adapted for elongated fillers is the Hamilton–Crosser (HC) model, which incorporates a shape factor and is given by [82]:

$${\displaystyle \frac{P_{eff}}{P_m}}={\displaystyle \frac{5P_m+P_f+5\varphi (P_f-P_m)}{5P_m+P_f-\varphi (P_f-P_m)}}$$

(15)

where $P_{eff }$ is the effective permeability of the MMM, $P_f$ is the permeability of the filler phase, and $\varphi$ is the filler volume fraction.

A more advanced approach for tubular fillers was proposed by Kang, Jones, and Nair (KJN), who developed a model based on a parallel–series resistance analogy that explicitly accounts for nanotube aspect ratio and orientation [83]:

$${\displaystyle \frac{P_{eff}}{P_m}}={\left[\left(1−{\displaystyle \frac{\cos \theta }{\mathrm{c}\mathrm{o}\mathrm{s}\theta +\frac{1}{\alpha }\mathrm{s}\mathrm{i}\mathrm{n}\theta }}\varphi \right)+{\displaystyle \frac{P_m}{P_f}}\left({\displaystyle \frac{1}{\mathrm{c}\mathrm{o}\mathrm{s}\theta +\frac{1}{\alpha }\mathrm{s}\mathrm{i}\mathrm{n}\theta }}\right)\varphi \right]}^{-1}$$

(16)

where $\alpha =L/d$ is the aspect ratio of the nanotubes, with L and d denoting their length and diameter, respectively, and θ is the orientation angle relative to the direction of gas transport.

Although the HC and KJN models improve the description of filler geometry, they still assume a perfectly bonded, defect-free polymer–filler interface. In real MMMs, however, interfacial voids and additional resistance to mass transfer are unavoidable due to differences in surface chemistry between the polymer matrix and inorganic fillers. As a result, MMMs should be treated as three-phase systems consisting of the polymer matrix, filler particles, and an interfacial region.

To address this issue, Chehrazi et al. proposed a three-phase model that introduces the concepts of interfacial thickness ($a_{int}$) and interfacial permeation resistance [38,84,85,86]. Within this framework, the effective permeability of nanotube-based MMMs is described as:

$${\displaystyle \frac{P_{eff}}{P_m}}={\displaystyle \frac{3+\varphi \left[\frac{2(d/(a_{int}-1\left)\right)}{d/(a_{int}+1)}+\frac{\left(P_{NT}/P_m\right)}{1+(2a_{int}/L)(P_{NT}/P_m)}+\frac{L}{a_{int}}-1\right]}{3-\varphi \left(\frac{2(d/(a_{int}-1\left)\right)}{d/a_{int}+1}\right)}}$$

(17)

where $P_{NT}$ is the permeability of the nanotube, and the remaining parameters retain their usual meanings.

The predictive performance of theoretical models is commonly evaluated using the average absolute relative error (AARE), defined as [38,85,86]:

$$\%AARE={\displaystyle \frac{100}{NDP}}\sum _{i=1}^{NDP}\mid {\displaystyle \frac{P_i^{pred}-P_i^{exp}}{P_i^{exp}}}\mid$$

(18)

where NDP is the number of experimental data points, and $P_i^{pred}$ and $P_i^{exp}$ are the predicted and experimental permeability values, respectively.

Comparative analyses demonstrate that three-phase models incorporating interfacial effects provide significantly improved agreement with experimental data compared to classical two-phase approaches, particularly for membranes with higher filler loadings and complex morphologies. These findings confirm that interfacial resistance, filler geometry, and dispersion state play a decisive role in governing gas transport through modern mixed matrix membranes.

In summary, while classical models such as Maxwell and Bruggeman remain valuable tools for preliminary membrane screening, advanced models that explicitly consider non-ideal morphology, filler anisotropy, and interfacial phenomena are essential for accurately describing gas transport in contemporary MMMs and for guiding the design of high-performance membranes for gas separation applications.

Beyond the classical Maxwell, Bruggeman, and nanotube-oriented models, a number of additional theoretical frameworks have been proposed to more accurately describe gas transport in mixed matrix membranes, particularly in systems characterized by high filler loadings, non-spherical inclusions, or non-ideal polymer–filler interfaces. These models aim to capture physical effects such as filler packing limitations, percolation phenomena, interfacial resistance, and redistribution of polymer free volume, which are not adequately represented in ideal two-phase theories.

One important extension of the Maxwell formalism is the Lewis–Nielsen model, which introduces corrections related to filler geometry and maximum packing constraints. In this approach, the effective permeability of an MMM is expressed as [87]:

$${\displaystyle \frac{P_{eff}}{P_m}}={\displaystyle \frac{1+A\Phi }{1-B\Phi }}$$

(19)

where $P_{eff }$ and $P_m$ are the permeability coefficients of the mixed matrix membrane and the polymer matrix, respectively, and Φ denotes the filler volume fraction. The parameters A and B depend on the permeability ratio of the filler to the polymer and on the maximum packing fraction ${\Phi }_m$, which reflects particle shape and packing efficiency. By incorporating ${\Phi }_m$, the Lewis–Nielsen model improves permeability predictions at moderate and relatively high filler loadings. Nevertheless, the model still assumes ideal interfacial contact and homogeneous filler dispersion, which limits its applicability to membranes with defect-free morphology.

For MMMs containing highly permeable fillers, such as MOFs, capable of forming interconnected transport pathways (continuous “gas-transport highways”), percolation-based models have been introduced. In such systems, gas permeability may increase abruptly once the filler concentration exceeds a critical percolation threshold ${\Phi }_c$. This behavior is commonly described using a power-law relationship [88]:

$$P_{eff}=P_m+C(\Phi -{\Phi }_c{)}^t\mathrm{for} \Phi >{\Phi }_c$$

(20)

where C is a fitting constant and t is the critical exponent related to the dimensionality of the percolating network. This model successfully explains the strong enhancement in CO₂ permeability (up to 6.6-fold relative to neat polymer) observed in phase-separated MOF MMMs [89]. Percolation models are particularly relevant for membranes containing carbon nanotubes, graphene, or other high-aspect-ratio fillers, where continuous diffusion pathways can develop. Although these models successfully capture sharp permeability enhancements above the percolation threshold, they provide limited insight into selectivity changes and are generally less accurate below ${\Phi }_c$.

Another class of models is based on free volume theory, which assumes that gas transport in polymers is governed by the availability and distribution of free volume elements. Within this framework, gas permeability is related to the fractional free volume (FFV) according to [90,91]:

$$P=Aexp\left(-{\displaystyle \frac{B}{FFV}}\right)$$

(21)

where A and B are gas-specific constants. In MMMs, the incorporation of inorganic fillers can modify the polymer free volume by disrupting chain packing or inducing rigidified interfacial regions. Hybrid models combining free volume concepts with composite transport equations have been shown to successfully describe permeability trends in glassy polymer MMMs; however, they often require additional experimental input, such as density measurements or positron annihilation lifetime spectroscopy data.

Resistance-based models have also been applied to describe gas transport in MMMs, particularly for membranes with anisotropic structures or preferential filler orientation. In resistance-in-series models, the overall transport resistance is assumed to be the sum of individual resistances of each phase:

$${\displaystyle \frac{1}{P_{eff}}}=\sum _i{\displaystyle \frac{l_i}{P_i}}$$

(22)

where $l_i$ and $P_{i }$ represent the thickness fraction and permeability of the i-th phase, respectively. Conversely, resistance-in-parallel models assume independent transport pathways and describe the effective permeability as:

$$P_{eff}=\sum _i{\Phi }_i{P}_i$$

(23)

where ${\Phi }_i$ is the volume fraction of each phase. Although these approaches are mathematically straightforward, they often oversimplify real MMM morphologies and neglect important effects such as interfacial resistance, tortuosity, and polymer chain rigidification [57,67].

More advanced numerical techniques, including finite element modeling and Monte Carlo simulations, explicitly represent membrane microstructure and filler distribution. These methods solve the local mass transport equation:

$$\nabla \cdot \left(P\left(\mathbf{r}\right)\nabla c\left(\mathbf{r}\right)\right)=0$$

(24)

where $P\left(\mathbf{r}\right)$ is the spatially dependent permeability and $c\left(\mathbf{r}\right)$ is the penetrant concentration. Such approaches provide detailed insight into local flux heterogeneities and interfacial transport phenomena but are computationally demanding and therefore less suitable for rapid membrane screening or parametric studies [92,93].

Three-dimensional simulations for MMMs containing impermeable cuboid nanoparticles have revealed significant deviations from Maxwell predictions. A regression-based permeability model was developed in

$$P_r=a_0+a_1 \left({\displaystyle \frac{t}{L}}\right)+a_2 \left({\displaystyle \frac{A_p}{A_m}}\right)$$

(25)

where $t/L$ captures particle shape and ${\displaystyle \frac{A_p}{A_m}}$ captures the projected area ratio. This model performs well for a wide range of aspect ratios and volume fractions [94].

Recent resistance-based models developments incorporate interfacial resistance, which dominates in MMMs containing rigid porous fillers such as MOFs or mesoporous silica [95]. The updated form is:

$${\displaystyle \frac{1}{P_{\mathrm{e}\mathrm{f}\mathrm{f}}}}={\displaystyle \frac{l_m}{P_m}}+{\displaystyle \frac{l_f}{P_f}}+{\displaystyle \frac{R_{\mathrm{i}\mathrm{n}\mathrm{t}}}{l_{\mathrm{t}\mathrm{o}\mathrm{t}}}}.$$

(26)

This formulation captures interfacial effects absent from classical two-phase models.

In hydrogen-separation MMMs [96], polymer–filler interactions reduce local free volume, lowering permeability. Free-volume permeability is described as [21] with:

$$FF{V}_{\mathrm{e}\mathrm{f}\mathrm{f}}=(1-\Phi )FF{V}_m+\Phi FF{V}_f-{\delta }_{\mathrm{r}\mathrm{i}\mathrm{g}},$$

(27)

where the rigidification correction ${\delta }_{\mathrm{r}\mathrm{i}\mathrm{g}}$ captures reduced chain mobility near filler surfaces.

Recent process-level models [97] connect membrane-scale diffusion with module-scale operation by using dimensionless variables

$$J_i^{\ast }=-\nabla c_i^{\ast } c_i^{\ast }={\displaystyle \frac{c_i}{c_{i,0}}}$$

(28)

and:

$$P_{\mathrm{e}\mathrm{f}\mathrm{f}}^{\ast }=f(\Phi ,{\lambda }_{dm},Sh,Pe),$$

(29)

where Sherwood and Peclet numbers incorporate mass-transfer resistances external to the membrane.

Overall, the wide range of available transport models reflects the intrinsic complexity of gas permeation in mixed matrix membranes. Each model captures selected physical aspects of transport, yet none can universally describe all MMM systems. Consequently, the most reliable strategy for predicting gas transport behavior and guiding the rational design of high-performance MMMs involves a combination of analytical modeling, experimental validation, and advanced numerical simulations.

5. Membranes for CO₂ Separation

In recent years, membrane-based technologies have been increasingly implemented in gas separation processes. This trend is driven by the need to reduce both energy consumption and the environmental impact of industrial operations, particularly under conditions of volatile energy prices. Membrane gas separation (GS) is therefore regarded as a promising alternative or complement to conventional separation methods. Depending on their composition and structure, membranes applied for CO₂ separation are generally classified into inorganic membranes, polymeric membranes, and mixed matrix membranes (MMMs) (

Table 3. Comparison of membrane types for CO₂ separation.

Membrane Type	Dominant Transport Mechanism	Key Advantages	Main Limitations	Typical Applications/ Remarks
Conventional polymeric membranes	Solution–diffusion	Mature and commercially established technology Low fabrication cost and good processability Easy fabrication into TFC and hollow-fiber modules	Permeability–selectivity trade-off (Robeson upper bound) CO2-induced plasticization and swelling Physical aging, especially in thin films	Natural gas sweetening, limited post-combustion capture; dominant in current industrial use
CO2-philic polymeric membranes	Solution–diffusion with enhanced CO2 solubility	High CO2 permeability due to strong CO2–polymer interactions Good performance under humid conditions Tunable chemistry (PEO, ILs, DES, PILs)	Mechanical weakness or crystallization (PEO-based) Stability issues for IL/DES-containing membranes Challenges in fabricating ultrathin selective layers	Post-combustion capture; hybrid and MMM concepts; mostly laboratory and pilot scale
High-free-volume polymers (PIMs, TR, TB, modified PIs)	Solution–diffusion with size-selective diffusion	Extremely high CO2 permeability Tunable microporosity Potential to exceed updated upper bounds	Moderate selectivity for some gas pairs Severe physical aging Limited long-term stability in TFC form	Advanced MMMs; research-driven development; limited industrial implementation
Facilitated transport membranes (FTMs)	Reaction–diffusion via fixed or mobile carriers	Very high CO2 permeability and selectivity Can surpass Robeson upper bound Excellent performance at low CO2 partial pressure	Strong dependence on humidity Carrier saturation at high pressure Stability and carrier loss concerns	Highly promising for post-combustion CO2 capture (3–15% CO2, near ambient pressure)
Carbon molecular sieve (CMS) membranes	Molecular sieving	High CO2 selectivity Good resistance to plasticization Suitable for high-pressure operation	Brittleness and difficult module fabrication Physical aging and pore shrinkage Performance loss in humid streams	Natural gas sweetening, H2/CO2 separation; niche and emerging industrial use
Zeolite membranes	Molecular sieving + adsorption	Very high CO2 selectivity Well-defined crystalline pores	Defect sensitivity Complex fabrication and scale-up Limited tolerance to impurities	Mainly laboratory-scale demonstrations
MOF membranes	Molecular sieving + adsorption–diffusion	Highly tunable pore size and chemistry Strong CO2 affinity	Moisture and stability issues Difficult large-area fabrication	Proof-of-concept CO2 separations; low TRL
Metallic Membranes (Pd-based)	Proton-conductive transport (H2-selective)	Near-infinite H2 selectivity High thermal stability	Very high cost Susceptible to poisoning Not CO2-selective	H2/CO2 separation, membrane reactors for CCS

).

5.1. Key Criteria for Membrane Material Selection

The growing interest in membrane-based gas separation technologies has been driven by the need to reduce energy consumption and environmental impact in industrial processes. Compared to conventional separation methods, membrane gas separation (GS) offers lower energy demand, modular design, and reduced operational footprint, making it an attractive option for CO₂ capture under increasingly stringent climate regulations. The selection of membrane materials for CO₂ separation is primarily governed by their intrinsic separation performance, particularly CO₂ permeance and selectivity. High CO₂ permeability is essential to achieve elevated gas fluxes and minimize membrane area requirements, while adequate selectivity is necessary to reach the desired CO₂ purity. However, membrane performance strongly depends on process-specific parameters, including feed gas composition, presence of impurities, operating pressure, temperature, and flow conditions. Consequently, membrane material selection must often be tailored to individual applications. Beyond separation efficiency, practical considerations such as processability, scalability, mechanical robustness, and cost play a decisive role in determining the industrial viability of membrane materials. Polymeric membranes are generally solution-processable and economically favorable, whereas inorganic membranes often require complex fabrication routes and higher production costs. Materials that cannot be fabricated into thin, defect-free selective layers—such as certain laboratory-scale thick films or poorly processable materials—are unlikely to meet industrial upscaling requirements [98,99].

Resistance to impurities present in CO₂-containing gas streams is another critical factor. Flue gases may contain SO_x, NO_x, and water vapor, while natural gas streams often include H₂S and heavy hydrocarbons. These components can adversely affect membrane performance through competitive sorption, irreversible chemical reactions, or membrane swelling. In polymeric membranes, CO₂-induced plasticization remains a major challenge, particularly at high pressures, where excessive swelling increases gas permeability indiscriminately and reduces selectivity [100]. Thin selective layers are especially vulnerable due to enhanced polymer chain mobility.

Membrane aging, particularly physical aging in glassy polymers, further complicates long-term operation. The gradual loss of excess free volume leads to declining permeability over time, with thin membranes and high-free-volume polymers being most affected [7,101]. Since industrial membrane modules are expected to operate reliably for several years, long-term stability testing under realistic conditions is indispensable during material development.

5.2. Polymer Membranes for CO₂ Separation

Polymer membranes are the most mature and widely used materials for gas separation, representing over 90% of commercial applications due to good processability, mechanical flexibility, and low cost. Commercial CO₂ membranes are mainly based on cellulose acetate (CA), polysulfone (PSf), polyethersulfone (PES), and polyimide (PI) [102]. They can be easily fabricated into thin-film composite (TFC) structures, scaled up, and integrated into existing modules. However, their performance is limited by the permeability–selectivity trade-off described by the Robeson upper bound (2008) [69], and most commercial membranes remain below current upper-bound limits. A key drawback is CO₂-induced plasticization and swelling, which reduce mechanical stability and separation efficiency, especially for CO₂/N₂ and CO₂/CH₄ separations [99]. These limitations have driven the development of advanced polymer chemistries and hybrid systems such as mixed matrix membranes (MMMs) [29].

5.2.1. CO₂-Philic Polymer Membranes

CO₂-philic polymers utilize the high condensability and quadrupole moment of CO₂, leading to greater solubility than other light gases. Incorporation of polar groups (e.g., ethylene oxide units, ionic moieties) enhances CO₂–polymer interactions. Poly(ethylene oxide) (PEO)-based polymers are key examples, showing strong dipole–quadrupole interactions with CO₂ [103,104]. PEO membranes offer high CO₂ permeability and good selectivity, especially under humid conditions, but suffer from low mechanical strength (low molecular weight) and high crystallinity (high molecular weight). These drawbacks have been addressed via block copolymerization, crosslinking, grafting, and hybrid membrane approaches [105,106,107,108]. Commercial examples such as Pebax^® (Arkema, Colombes, France) and PolyActive™ (Covestro (formerly DSM Engineering Materials), Leverkusen, Germany) combine PEO segments with mechanically robust blocks [109,110]. Ionic liquids (ILs) and deep eutectic solvents (DES) are also used as CO₂-selective additives due to their strong CO₂ affinity [111,112,113,114,115,116]. Although these systems can achieve excellent separation performance, long-term stability and mechanical strength remain challenges, often reducing gas transport when improved.

5.2.2. High-Free-Volume Polymers

Glassy polymers with high fractional free volume (FFV) are key materials for CO₂ separation, including polymers of intrinsic microporosity (PIMs), Tröger’s base (TB) polymers, thermally rearranged (TR) polymers, and modified polyimides [117]. PIMs, introduced by Budd and McKeown, possess rigid, contorted backbones forming interconnected microporous networks. They offer very high CO₂ permeability (>6000 Barrer) and tunable functionality [118], but show moderate selectivity and significant physical aging, particularly in thin films, limiting long-term use [119]. TR polymers provide tunable pore sizes, good plasticization resistance, high CO₂ permeability, and moderate selectivity [120,121,122], yet poor solubility, high processing temperatures, and high costs hinder commercialization. Perfluoropolymers (e.g., Teflon™ AF (Chemours, Wilmington, DE, USA), Cytop™ (AGC Inc., Tokyo, Japan)) exhibit very high CO₂ permeability due to large free volume but have low selectivity and solvent processing issues [123,124]. Hybrid systems combining high-FFV polymers with facilitated transport layers have recently shown promise in overcoming these drawbacks [125].

5.2.3. Facilitated Transport Membranes

Facilitated transport membranes (FTMs) are a leading concept for CO₂ separation, exploiting its acidic, reactive nature. Unlike solution–diffusion membranes, FTMs use chemical carriers that reversibly react with CO₂, creating an additional transport pathway that enhances permeability and selectivity [126,127]. Two configurations exist: fixed-site carriers (FSCs), covalently bound to the polymer backbone, and mobile carriers dispersed in the matrix. Common carriers include primary, secondary, and tertiary amines, amino acid salts, ionic liquids, and enzyme-mimicking compounds [126]. CO₂ transport occurs via reversible complex formation and diffusion, described by reaction–diffusion models [127]. FTMs strongly depend on water, which participates in CO₂–amine reactions. Early studies on polyvinylamine (PVAm) showed low CO₂ permeance despite high amine density [128], but later work confirmed that membrane hydration activates facilitated transport by enhancing reaction kinetics and carrier mobility [129]. Thus, FTMs typically use hydrophilic polymers and operate optimally under humid conditions, with performance closely correlated to water content [129,130]. Under full humidity, PVAm membranes exceed the Robeson upper bound (2008), combining very high CO₂ permeability with excellent CO₂/N₂ selectivity [61]. Mechanical stability and performance have been improved by blending PVAm with PVA or adding nanofillers (nanocellulose, CNTs, GO, TiO₂) [131,132], as well as by chemical modification into tailored copolymers for better long-term stability [133]. Optimized systems report CO₂/N₂ selectivity approaching 1000 with CO₂ permeance > 2400 GPU (P_CO2 = 588 Barrer) [134], and metal-doped tertiary amine FTMs reach ~9000 GPU (P_CO2 = 1600 Barerr) with moderate CO₂/N₂ selectivity around 40 [135]. Other amine polymers (poly(allylamine), PEI, PAMAM) have been used as fixed-site carriers [136,137], while tertiary amines improve selectivity via steric effects [138,139]. Mobile carriers (amino acid salts, ionic liquids) in hydrophilic matrices such as PVA or PVP enhance permeability but may plasticize the polymer and reduce mechanical stability [140,141,142,143]. Immobilizing carriers on functionalized nanofillers improves stability and retention [143]. Limitations include carrier saturation at high CO₂ partial pressures, reducing efficiency and suitability for high-pressure separations [144]. Increasing membrane thickness or adding carriers/nanofillers can partially mitigate this [131,145]. FTMs are especially suitable for post-combustion CO₂ capture (3–15% CO₂, near-atmospheric pressure), where they outperform conventional polymer membranes in permeability and selectivity [146].

5.3. Inorganic Membranes

Inorganic membranes constitute an important class of materials for CO₂ separation, particularly in applications requiring high thermal, chemical, and mechanical stability. Compared with polymeric membranes, inorganic systems generally tolerate harsher operating environments, including elevated temperatures, high pressures, and chemically aggressive gas streams [147]. These properties make them attractive for specialized industrial separations where polymeric membranes may fail. Gas transport in inorganic membranes is mainly governed by molecular sieving, with possible contributions from surface diffusion and adsorption–diffusion, depending on material and conditions. High CO₂ selectivity is achieved when pore sizes are precisely tuned between the kinetic diameters of gas pairs, as in zeolite, MOF, and carbon molecular sieve (CMS) membranes [147,148,149]. Zeolite and MOF membranes provide ordered pores and defined adsorption sites for selective CO₂ transport. CMS membranes, produced by carbonizing polymer precursors (e.g., polyimides, PPO, cellulose derivatives), form rigid microporous carbon networks with strong molecular sieving properties [149,150].

5.3.1. Carbon Molecular Sieve Membranes

CMS membranes are widely studied for CO₂ separations. Hollow-fiber CMS membranes from cellulose acetate or polyimides show excellent performance in biogas upgrading, natural gas sweetening, and H₂ purification, with high CO₂ selectivity, plasticization resistance, and suitability for high pressure [149]. Limitations include brittleness, physical aging (pore shrinkage and permeability loss), and moisture sensitivity—water adsorption can significantly reduce CO₂ permeance above ~30% RH [150]. Improvements include ceramic-supported CMS membranes (≈220 GPU CO₂ permeance, 27.5 Barrer, CO₂/CH₄ selectivity ~30) for high-pressure gas processing [151], CMS from Tröger’s base polymers with strong CO₂/H₂ separation [152], and low-temperature carbonized CMS with improved flexibility and scalability [152,153].

5.3.2. Zeolite and MOF Membranes

Zeolite membranes, with crystalline and uniform pores, can achieve CO₂/CH₄ selectivity >250 at ~1 μm selective layer thickness [154]. However, large-scale use is limited by fabrication complexity, defect sensitivity, reproducibility, and cost. MOF membranes offer greater tunability of pore size and functionality via ligand/metal design. Despite promising CO₂ separation results, long-term stability—especially under moisture and contaminants—remains a key barrier to industrialization [148,155,156].

5.3.3. Metallic Membranes

Metallic (mainly Pd-based) membranes are used primarily for H₂/CO₂ separation. They exhibit near-infinite hydrogen selectivity via proton-conductive transport through the metal lattice [157], making them suitable for hydrogen purification and membrane reactors with integrated CO₂ capture [158]. High Pd cost and susceptibility to sulfur/carbon poisoning limit broader use [159]. Hybrid systems embedding Pd nanoparticles in polymers aim to reduce cost and enhance H₂/CO₂ separation [160]. Inorganic membranes offer high thermal/chemical stability, strong molecular sieving selectivity, and suitability for harsh conditions. However, mechanical fragility, aging, moisture sensitivity, high costs, and challenges in producing large, defect-free membranes limit commercialization. Consequently, polymeric and mixed-matrix membranes dominate CO₂ separation industrially due to better scalability and integration. Future progress requires advances in manufacturability, durability, and system-level integration alongside performance improvements.

5.4. Mixed Matrix and Hybrid Membranes for CO₂ Separation

Mixed matrix membranes (MMMs), also called hybrid or nanocomposite membranes, are extensively studied for CO₂ separation. Since the classification by Okumus et al. [161], MMMs are grouped into liquid–polymer, solid–polymer, and solid–liquid–polymer systems, with solid–polymer MMMs most investigated due to better mechanical stability and compatibility with conventional processing [162,163]. They combine high-performance fillers with scalable polymer matrices. Early used fillers included CMS, silica, and zeolites. CMS-based MMMs offer rigid microporosity and plasticization resistance [80,150,151,153], while zeolites provide defined pores and strong CO₂ adsorption but often suffer from interfacial defects [163]. Recent work focuses on advanced fillers such as CNTs, MOFs, GO, COFs, PAFs, and other nanostructures [164,165,166,167]. CNTs enhance permeability and strength but require functionalization for selectivity [164,166,168]. MOFs (e.g., IRMOF-1, MIL-100, UiO-66, ZIFs, HKUST-1) are widely used due to high surface area and tunable functionality, often improving CO₂ permeability and sometimes selectivity [169,170,171,172,173,174,175,176], though stability under humid/acidic conditions remains a concern [177]. Fillers can also be classified by dimensionality: 0D nanoparticles (e.g., silica, TiO₂) increase free volume [137,178]; 1D fillers (CNTs, nanocellulose) enhance anisotropic transport and mechanics; 2D nanosheets (graphene, GO, COFs) create tortuous, selective pathways [155,179,180]; and 3D porous fillers (MOFs, zeolites) combine adsorption and diffusion selectivity [167,181]. MMM performance strongly depends on the polymer–filler interface. Poor compatibility causes non-selective microvoids (“sieve-in-a-cage”) that raise permeability without improving selectivity [177]. Other issues include polymer rigidification and pore blockage [182,183]. To overcome these challenges some solutions were proposed like surface functionalization, polymer modification, and interfacial engineering, often introducing amine, hydroxyl, or ionic groups to enhance CO₂ affinity [166,184]. While three-phase MMMs using CO₂-philic liquids (ionic liquids, DES) further enhance transport [106,183], they risk phase separation. Most MMMs are thick dense films (tens to >100 µm), limiting industrial use. Reducing thickness while maintaining filler dispersion is challenging, prompting development of hybrid thin-film composite (TFC) membranes. Promising TFC systems include nanocellulose and functionalized GO, with 2D fillers particularly suited for ultrathin layers [185]. Novel strategies, such as soluble porous organic cages or solid–solvent processing, enable ultrathin MMMs with high filler loading and simultaneous gains in permeability and selectivity [186]. Overall, MMMs offer a versatile route to overcome the permeability–selectivity trade-off of polymers. However, interfacial compatibility, long-term stability, membrane thinning, and scalable fabrication remain key challenges. Further advances in materials design and processing are essential for industrial CO₂ separation applications. A more detailed discussion of the structure–property relationships, and practical limitations associated with individual filler classes will be provided in the following sections, which focus separately on specific types of fillers used in mixed matrix and hybrid membranes.

5.4.1. Zeolite Filler

Zeolites are crystalline aluminosilicates with uniform micropores (~3–15 Å) and high thermal and chemical stability [187,188]. Owing to their molecular sieving properties and strong CO₂ affinity, they were among the first inorganic fillers used in mixed matrix membranes (MMMs) and remain extensively studied for CO₂ separation [70,81]. Their performance enhancement is mainly attributed to preferential CO₂ adsorption, polymer–filler interactions, increased fractional free volume, and selective diffusion pathways (

Table 4. Comparative overview of zeolite-based mixed matrix membranes for CO₂ separation.

Zeolite Type	Polymer Matrix	Filler Loading [wt.%]	Conditions	PCO2 [Barrer]	Selectivity α*	Key Features	Advantages	Limitations	Ref.
Zeolite 13×	PEBAX	1	p = 14 bar, T = 25 °C	194.1	CO2/N2: 56.5, CO2/CH4: 56	Deposited on PSf/PE	Improved permeability and selectivity at low loading	Potential interfacial defects at higher loadings	[191]
Zeolite T	6FDA-durene	1	p = 3.5 bar, T = 30 °C	843.6	CO2/CH4: 19.1	Neat filler	High permeability, plasticization-resistant up to 20 bar	Limited adaptability to other gas pairs	[192]
Zeolite 4A	PVAc	25	p = 0.1 MPa, T = 30 °C	2.41	CO2/N2: 100.5	Unmodified	Enhanced selectivity	Reduced permeance	[193]
Zeolite 4A	PVAc	50	p = 440 psi, T = 35 °C	11.4	CO2/CH4: 25	High loading	Maintains selectivity at high filler content	High pressure needed; scalability concerns	[194]
LNZ-25 (Li/Na-ZSM-25)	Matrimid® 5218	5	p = 5 bar, T = 35 °C	12	CO2/CH4: 169	Partially lithiated	Very high selectivity, plasticization resistant	Low CO2 permeability	[195]
NaY + PEG-600	Pebax	30	p = 0.15 MPa, T = 35 °C	172.6	CO2/N2: 107.9	Tertiary-component MMM	High selectivity, CO2 enrichment from 15% to 96.7%	Complex fabrication	[196]
Hierarchical Zeolite 5A	Carbonized Matrimid® 5218	30	p = 1 bar, T = 35 °C	450	CO2/CH4: 19.3	Micropores + mesopores, thermally carbonized	Reduced transport resistance, improved free volume	Moderate selectivity	[197]
Zeolite 13× (NH2-silanized)	6FDA-Durene	15	p = 0.2 MPa, T = 25 °C	887	CO2/N2: 25.3	Surface functionalization	Enhanced filler–polymer compatibility	Functionalization adds complexity	[198]
EMC-2 (NH2-silanized)	6FDA-ODA	25	p = 150 psi, T = 35 °C	40.9	CO2/CH4: 80.2	Crosslinked	High selectivity	Moderate permeability	[199]
NaY (NH2-silanized)	Matrimid® 5218	15	p = 2 bar, T = 35 °C	9.7	CO2/CH4: 57.1	Crosslinked with APDEMS	Improved filler dispersion, increased selectivity	Low permeability	[200]
Zeolite 3A (NH2-silanized)	PSf	40	p = 12 bar, T = 25 °C	4.22 GPU	H2/CO2: 7.12	Crosslinked with APTMS	High H2 selectivity	Moderate CO2 performance	[201]
MCM-41 (NH2-silanized)	PSf	30	p = 10 bar, T = 25 °C	9.13	CO2/N2: 32.97, CO2/CH4: 31.48	Crosslinked	High ideal selectivity	Complex preparation	[202]
SAPO-34 + IL	PSf	5	p = 3.5 bar, T = 30 °C	7.19 GPU	CO2/N2: 44.9	Ionic liquid added	Seals interfacial defects, enhances selectivity	Low permeability	[203]
ZSM-5 + IL	6FDA-TeMPD	15	p = 75 mmHg, T = 35 °C	142	CO2/N2: 32.6, CO2/H2: 25.6	Ionic liquid	Improved interfacial adhesion	Moderate selectivity	[204]
SAPO-34 + PIL–RTIL	–	25–30	p = 40 bar, T = 25 °C	202–260	CO2/CH4: 43–90	Poly(ionic) liquid + RTIL	High CO2 solubility, mechanical stability	Complex multicomponent system	[205,206]

) [189]. Low zeolite loadings have been shown to significantly improve separation performance. Yu et al. reported a CO₂ permeability of 66.9 Barrer and CO₂/N₂ selectivity of 9.6 for PES-g-PEG/MFI MMMs at 35 °C, attributed to improved polymer–filler interactions and mesoporosity [190]. Asghari et al. demonstrated that 1 wt.% zeolite 13× in PEBAX increased CO₂ permeability by ~200% and enhanced CO₂/N₂ and CO₂/CH₄ selectivity by more than one order of magnitude [191]. Similarly, Jusoh et al. achieved a high CO₂ permeability of 843.6 Barrer with CO₂/CH₄ selectivity of 19.1 in 6FDA-durene MMMs containing 1 wt.% zeolite T, together with improved resistance to plasticization up to 20 bar [192]. At higher filler contents, increased selectivity is often accompanied by reduced permeability due to diffusion resistance. Ahmad et al. observed a 75% increase in CO₂/N₂ selectivity for PVAc membranes containing 25 wt.% zeolite 4A, with a concomitant decrease in permeance [193]. However, Adams et al. reported that MMMs with up to 50 vol.% zeolite 4A retained enhanced CO₂/CH₄ selectivity at 440 psi when interfacial defects were effectively suppressed [194]. Tailored zeolite chemistry can further enhance molecular sieving, as shown by Zhao et al., who achieved a CO₂/CH₄ selectivity of 169 using partially lithiated Na-ZSM-25 in Matrimid^® [195]. Zhang et al. developed a ternary Pebax/PEG/NaY MMM enabling CO₂/N₂ selectivity of 107.9 and enrichment of CO₂ from 15% to 96.7% in a single step [196]. Despite these advantages, zeolite-based MMMs face challenges related to poor compatibility with hydrophobic polymers, leading to non-selective interfacial voids or microcracks at higher loadings, as well as limited structural tunability compared to MOFs [197]. Surface modification strategies have therefore been widely adopted. Mashhadikahn et al. improved adhesion via aminosilane-modified zeolite 13×, achieving 887 Barrer CO₂ permeability and CO₂/N₂ selectivity of 25.3 in 6FDA-durene MMMs [198]. Chen et al. reported a CO₂/CH₄ selectivity of 80.2 using APTES-functionalized EMC-2 in a 6FDA-ODA matrix [199], while Amooghin et al. enhanced CO₂/CH₄ selectivity by 57% through silane crosslinking in Matrimid^®/NaY MMMs [200], with similar trends in related systems [201,202]. Interfacial sealing with ionic liquids has also proven effective. Ahmad et al. increased CO₂/N₂ selectivity to 44.9 and CO₂ permeance to 7.19 GPU in PSf/SAPO-34 MMMs using [emim][Tf₂N] [203], while SAPO-34/PIL–RTIL systems achieved CO₂/CH₄ selectivity up to 90 with stable operation up to 40 bar [204,205,206].

Overall, zeolite-based MMMs offer well-defined microporosity, strong CO₂ affinity, and excellent stability, resulting in α_CO2/N2 ≈ 33–108, α_CO2/CH4 ≈ 19–169, and CO₂ permeabilities of ~2–887 Barrer. However, interfacial defects, pore blockage, limited framework flexibility, and scalability remain key limitations. Advances in nanozeolites, hierarchical structures, and surface-functionalized fillers are essential to improve adhesion and expand industrial applicability.

5.4.2. Graphene- and Graphene Oxide-Based Fillers

Graphene-derived materials, particularly graphene oxide (GO), have emerged as important fillers for mixed matrix membranes (MMMs) (Table 5) due to their two-dimensional morphology, high aspect ratio, and tunable surface chemistry [207,208]. GO consists of graphene sheets decorated with oxygen-containing functional groups, which enlarge interlayer spacing and provide interaction sites for polymers and gas molecules, making GO attractive for CO₂ separation. In MMMs, GO primarily increases transport tortuosity, where impermeable basal planes hinder N₂ and CH₄, while interlayer galleries and interfacial regions preferentially facilitate CO₂ transport. However, excessive loading or restacking can severely suppress permeability. Ha et al. demonstrated that incorporation of 8 wt.% GO into PDMS reduced permeability of all tested gases by over 99.9%, despite increasing CO₂/N₂ and CO₂/CH₄ selectivity by more than two orders of magnitude, highlighting the inherent permeability–selectivity trade-off of GO fillers [207]. At low loadings, GO can simultaneously enhance permeability and selectivity. Shin et al. reported that only 0.3 wt.% GO in Pebax/PEG increased CO₂ permeability to ~650 Barrer with CO₂/N₂ selectivity of 55.8, while improving resistance to CO₂-induced plasticization up to 9 bar and maintaining stable mixed-gas performance for nearly 100 days [209]. GO has also been shown to mitigate physical aging in high-free-volume polymers; Luque-Alled et al. grafted PIM-1 chains onto APTS-functionalized GO, producing crosslinked membranes that retained initial CO₂ permeability for over 150 days, albeit with increased synthesis complexity [210]. Chemical functionalization is widely employed to enhance CO₂ affinity. Zhang et al. incorporated aminosilane-modified GO into Pebax, achieving CO₂ permeability of 172 Barrer under dry conditions and as high as 934.3 Barrer under humid conditions, with CO₂/CH₄ and CO₂/N₂ selectivities of 40.9 and 71.1, respectively, demonstrating strong synergy between GO and facilitated transport [211]. Structural modification has also been explored to overcome permeability limitations. Dong et al. introduced porous reduced graphene oxide (PRG) into Pebax 1657, combining narrow interlayer gaps for molecular sieving with mesopores for accelerated transport, yielding CO₂/N₂ selectivity of 104 at moderate permeability [212]. He et al. further enhanced performance using a layer-by-layer Pebax membrane incorporating porous GO and a CO₂-philic hierarchical porous polymer, achieving 232.7 Barrer and CO₂/N₂ selectivity of 80.7 [213]. Synergistic effects between GO and other nanofillers have also been reported. Li et al. combined multi-walled carbon nanotubes and GO in Matrimid^®, where CNTs provided fast diffusion pathways and GO contributed molecular sieving, resulting in increases of 331%, 149%, and 147% in CO₂ permeability, CO₂/CH₄ selectivity, and CO₂/N₂ selectivity, respectively [214]. Despite these advances, challenges remain, including aggregation at higher loadings, sensitivity to humidity-induced interlayer swelling, possible non-selective interfacial voids, and mechanical weakening due to excessive functionalization. Although sulfonated polymer brush grafting has been shown by Dong et al. to improve compatibility and CO₂ selectivity, such approaches increase processing complexity and cost [212]. Overall, GO-based MMMs exhibit high CO₂ affinity, good mechanical stability, and improved aging resistance, achieving α_CO2/N2 ≈ 24–104, α_CO2/CH4 ≈ 15–41, and CO₂ permeability in the range of ~27–5235 Barrer. However, intrinsic impermeability, restacking tendencies, and processing sensitivity remain key barriers, necessitating further development of controlled interlayer architectures, layer-by-layer assembly, and targeted chemical functionalization to ensure long-term stability and industrial relevance.

5.4.3. Carbon Nanotube and Nanocarbon Fillers

Since their discovery by Iijima et al. in the early 1990s, carbon nanotubes (CNTs) have attracted considerable interest as one-dimensional nanocarbon fillers due to their exceptional mechanical, thermal, and transport properties [168,215,216]. CNTs consist of rolled graphene sheets and are classified into single-walled CNTs (SWCNTs) and multi-walled CNTs (MWCNTs), the latter comprising concentric graphene layers spaced by ~0.34–0.35 nm. Their high aspect ratio, mechanical robustness, and chemical stability make CNTs attractive fillers for mixed matrix membranes (MMMs) for gas separation (Figure 2, Table 5) [168,215,216]. From a transport perspective, CNTs provide atomically smooth, graphitic inner channels that enable extremely fast gas diffusion. Kim and Skoulidas et al. demonstrated through simulations and experiments that gas diffusivities inside CNTs are several orders of magnitude higher than in conventional microporous materials, giving rise to the so-called “nanotube highway effect” [215,216]. Consequently, CNT-based MMMs (Table 4) often exhibit substantial permeability enhancement and partial mitigation of the Robeson trade-off. In addition, CNT incorporation can improve tensile strength, modulus, and resistance to plasticization, which is critical for industrial operation [217,218]. However, CNT performance strongly depends on dispersion, alignment, and interfacial adhesion. Pristine CNTs are hydrophobic and chemically inert, often leading to poor compatibility with polar polymers and the formation of non-selective interfacial voids [215]. Kim et al. reported increased CO₂ permeability but reduced CO₂/CH₄ selectivity in polysulfone MMMs containing alkyl-modified SWCNTs, attributing the selectivity loss to interfacial defects [215]. These results demonstrate that permeability gains alone do not guarantee improved separation performance. To address compatibility issues, multiple functionalization strategies have been developed. Ahmad et al. fabricated β-cyclodextrin-functionalized MWCNT/cellulose acetate membranes, achieving simultaneous improvements in CO₂ permeability and CO₂/N₂ selectivity at only 0.1 wt.% filler loading [168]. Zhao et al. incorporated amine-modified MWCNTs into Pebax^® 1657, obtaining a threefold increase in CO₂ permeability while maintaining CO₂/N₂, CO₂/CH₄, and CO₂/H₂ selectivity due to enhanced CO₂ affinity and polymer chain mobility [217]. Zhang et al. further demonstrated synergistic effects by coating MWCNTs with a PNIPAM hydrogel layer, yielding CO₂ permeabilities up to 567 Barrer and CO₂/N₂ selectivities approaching 70 under humid conditions [218]. Compared with CNTs, two-dimensional graphene oxide (GO) nanosheets primarily enhance selectivity rather than permeability. Due to their intrinsic impermeability and high electron density, pristine GO sheets often increase diffusion resistance at high loadings. Ha et al. observed a drastic reduction in gas permeability in PDMS/GO membranes, accompanied by over two orders of magnitude increase in CO₂/N₂ and CO₂/CH₄ selectivity [207]. At low loadings or with appropriate functionalization, however, GO can improve CO₂ solubility without severe permeability loss. Shin et al. reported that 0.3 wt.% GO in Pebax^®/PEG membranes achieved ~650 Barrer CO₂ permeability and CO₂/N₂ selectivity of 55.8, with excellent resistance to plasticization and long-term stability [209]. Further enhancements were achieved using amino-functionalized GO [199] and porous reduced GO architectures combining molecular sieving and fast transport [211,213]. Zero-dimensional carbon quantum dots (CQDs) represent an additional nanocarbon filler class. Shi et al. showed that graphitic CQDs at ultralow loadings (0.05 wt.%) effectively tuned Pebax^® microphase separation, enhancing CO₂ permeability and selectivity, whereas polymer-like CQDs required higher loadings [219]. Hybrid nanocarbon systems have also been explored. Li et al. demonstrated that Matrimid^® MMMs containing both MWCNTs and GO exhibited simultaneous improvements in CO₂ permeability (331%), CO₂/CH₄ selectivity (149%), and CO₂/N₂ selectivity (147%), confirming a synergistic effect between CNT fast-transport channels and GO-induced molecular sieving [214,220]. Overall, CNT-based MMMs exploit rapid nanotube-mediated transport and mechanical reinforcement, achieving α_CO2/N2 ≈ 22–81, α_CO2/CH4 ≈ 16–85, and CO₂ permeability of ~5–742 Barrer. Nevertheless, dispersion challenges, interfacial defects, and high filler loadings may impair processability and membrane integrity. Current research therefore emphasizes chemical functionalization and alignment strategies to enhance CO₂ affinity and promote efficient anisotropic transport.

Table 5. Comparison of MMMs based on Carbon Nanotube (CNT), graphene- and graphene oxide-based fillers applied for CO₂ separation.

Filler Type	Polymer Matrix	Filler Loading [wt.%]	Conditions	Permeability (Barrer)	Selectivity α*	Advantages	Limitations	Ref.
GO nanosheet	PDMS	8	p = 10 bar, T = 35 °C	PCO2 = 27.7	CO2/N2 = 24	Extremely high aspect ratio; strong selectivity enhancement	Basal planes impermeable to gases; severe permeability loss due to restacking	[207]
GO nanosheet	Pebax, PEG	0.3	p = 1 bar, T = 35 °C	PCO2 = 650	CO2/N2 = 55.8	Tunable chemistry; improved CO2 affinity; enhances plasticization resistance	Aggregation and restacking; humidity sensitivity; limited permeability at high loadings	[209]
GO	PIM-1	5	p = 1.5 bar, T = 25 °C	PCO2 = 5235 PCH4 = 359	CO2/CH4 = 14.9	Strong aging suppression; long-term stability	Covalent grafting increases cost and complexity	[210]
Amino-GO nanosheet	Pebax	0.9	p = 2 bar, T = 35 °C	PCO2 = 934.3	CO2/N2 = 71.1 CO2/CH4 = 40.9	High permeability and selectivity under humid conditions; improved dispersion	Excessive functionalization may weaken mechanical stability	[211]
Porous Reduced GO	Pebax 1657	5	p = 2 bar, T = 30 °C	PCO2 = 119	CO2/N2 = 104	Balanced permeability–selectivity trade-off; reduced restacking	Complex synthesis; precise reduction control required	[207]
Porous Graphene Oxide (PGO)	Pebax	2	p = 1 bar, T = 25 °C	PCO2 = 232.7	CO2/N2 = 80.7	High selectivity with moderate permeability; structural stability	Multistep fabrication; interlayer control critical	[213]
GO + MWCNT Hybrid	Matrimid®	5	p = 2 bar, T = 30 °C	PCO2 = 38.07 PCH4 = 0.45 PN2 = 0.47	CO2/N2 = 81.0 CO2/CH4 = 84.6	Synergistic permeability and selectivity enhancement	Filler–filler dispersion complexity	[214]
SWCNTs	Polysulfone (PSf)	5	p = 4 bar, T = 35 °C	PCO2 = 5.12 PCH4 = 0.27 PN2 = 0.23	CO2/N2 = 22.26 CO2/CH4 = 18.82	Extremely fast gas diffusion due to near-frictionless transport	High cost; aggregation; difficult large-scale dispersion	[215],
Pristine MWCNTs	Cellulose acetate (CA)	0.1	p = 3 bar, T = 35 °C	PCO2 = 741.67 PN2 = 18.46	CO2/N2 = 40.17	Better availability; mechanical reinforcement	Poor interfacial compatibility; risk of non-selective voids	[168]
Amino- functionalized MWCNTs (MWCNT–NH2)	Pebax® 1657	33	p = 7 bar, T = 35 °C	PCO2 = 361	CO2/N2 = 52 CO2/CH4 = 16	Improved polymer–filler adhesion; enhanced CO2 affinity	Excessive loading may reduce selectivity	[217]
Hydrogel-wrapped MWCNTs	Pebax® 1657	5	p = 2 bar, T = 25 °C	PCO2 = 567	CO2/N2 = 70 CO2/CH4 = 35	Synergistic effect: CNT fast transport + hydrogel CO2-philicity	Performance depends strongly on humidity	[218]

Table 5. Comparison of MMMs based on Carbon Nanotube (CNT), graphene- and graphene oxide-based fillers applied for CO₂ separation.

Filler Type	Polymer Matrix	Filler Loading [wt.%]	Conditions	Permeability (Barrer)	Selectivity α*	Advantages	Limitations	Ref.
GO nanosheet	PDMS	8	p = 10 bar, T = 35 °C	PCO2 = 27.7	CO2/N2 = 24	Extremely high aspect ratio; strong selectivity enhancement	Basal planes impermeable to gases; severe permeability loss due to restacking	[207]
GO nanosheet	Pebax, PEG	0.3	p = 1 bar, T = 35 °C	PCO2 = 650	CO2/N2 = 55.8	Tunable chemistry; improved CO2 affinity; enhances plasticization resistance	Aggregation and restacking; humidity sensitivity; limited permeability at high loadings	[209]
GO	PIM-1	5	p = 1.5 bar, T = 25 °C	PCO2 = 5235 PCH4 = 359	CO2/CH4 = 14.9	Strong aging suppression; long-term stability	Covalent grafting increases cost and complexity	[210]
Amino-GO nanosheet	Pebax	0.9	p = 2 bar, T = 35 °C	PCO2 = 934.3	CO2/N2 = 71.1 CO2/CH4 = 40.9	High permeability and selectivity under humid conditions; improved dispersion	Excessive functionalization may weaken mechanical stability	[211]
Porous Reduced GO	Pebax 1657	5	p = 2 bar, T = 30 °C	PCO2 = 119	CO2/N2 = 104	Balanced permeability–selectivity trade-off; reduced restacking	Complex synthesis; precise reduction control required	[207]
Porous Graphene Oxide (PGO)	Pebax	2	p = 1 bar, T = 25 °C	PCO2 = 232.7	CO2/N2 = 80.7	High selectivity with moderate permeability; structural stability	Multistep fabrication; interlayer control critical	[213]
GO + MWCNT Hybrid	Matrimid®	5	p = 2 bar, T = 30 °C	PCO2 = 38.07 PCH4 = 0.45 PN2 = 0.47	CO2/N2 = 81.0 CO2/CH4 = 84.6	Synergistic permeability and selectivity enhancement	Filler–filler dispersion complexity	[214]
SWCNTs	Polysulfone (PSf)	5	p = 4 bar, T = 35 °C	PCO2 = 5.12 PCH4 = 0.27 PN2 = 0.23	CO2/N2 = 22.26 CO2/CH4 = 18.82	Extremely fast gas diffusion due to near-frictionless transport	High cost; aggregation; difficult large-scale dispersion	[215],
Pristine MWCNTs	Cellulose acetate (CA)	0.1	p = 3 bar, T = 35 °C	PCO2 = 741.67 PN2 = 18.46	CO2/N2 = 40.17	Better availability; mechanical reinforcement	Poor interfacial compatibility; risk of non-selective voids	[168]
Amino- functionalized MWCNTs (MWCNT–NH2)	Pebax® 1657	33	p = 7 bar, T = 35 °C	PCO2 = 361	CO2/N2 = 52 CO2/CH4 = 16	Improved polymer–filler adhesion; enhanced CO2 affinity	Excessive loading may reduce selectivity	[217]
Hydrogel-wrapped MWCNTs	Pebax® 1657	5	p = 2 bar, T = 25 °C	PCO2 = 567	CO2/N2 = 70 CO2/CH4 = 35	Synergistic effect: CNT fast transport + hydrogel CO2-philicity	Performance depends strongly on humidity	[218]

5.4.4. Magnetic Nanofillers

A promising subgroup of CNT-based fillers includes magnetically modified nanostructures such as carbon-encapsulated magnetic nanoparticles (CEMNPs) and iron-filled multiwalled carbon nanotubes (Fe@MWCNTs) [38,221,222,223,224,225,226,227,228,229]. In these hybrids, magnetic phases (metallic iron or iron oxides) are encapsulated within or attached to CNT frameworks, providing magnetic responsiveness while retaining the intrinsic transport and mechanical advantages of CNTs. Although initially explored for catalysis, sensing, and biomedical applications, these materials have recently attracted attention for gas separation membranes [221,227,228,229]. From a gas-transport perspective, magnetic nanofillers introduce additional control mechanisms absent in conventional MMMs. Local magnetic fields generated by iron-based domains can influence gas trajectories, residence times, and local solubility. While these effects are most pronounced for paramagnetic gases, several studies indicate that CO₂ transport is also affected due to its high quadrupole moment and stronger interactions with magnetically active regions [221,222,226,230,231]. Rybak et al. demonstrated that magnetic hybrid membranes exhibit altered gas permeability and selectivity compared to non-magnetic counterparts, confirming the functional role of magnetic fillers in transport control [221,222,226,230,231]. Among magnetic fillers, Fe@MWCNTs are uniquely multifunctional, combining CNT-mediated fast transport with magnetic tunability. Encapsulated iron cores enable ferromagnetic or superparamagnetic behavior, allowing alignment under an externally applied magnetic field during membrane fabrication. Magnetic-assisted alignment promotes anisotropic and continuous transport pathways, reducing diffusion tortuosity within the membrane [40,48,232]. Rybak et al. reported that MMMs containing magnetically aligned Fe@MWCNTs displayed enhanced permeability and selectivity for air separation and CO₂-containing mixtures relative to randomly dispersed systems [233,234,235]. These membranes achieved CO₂ permeability of 59.56 Barrer and CO₂/N₂ selectivity of 58.63, demonstrating that magnetic organization shifts the permeability–selectivity balance. Despite these advantages, incorporation of CNTs—both pristine and magnetic—remains challenging. Strong van der Waals and π–π interactions promote aggregation, leading to heterogeneous dispersion, non-selective interfacial voids, and degraded mechanical stability [225,227,232]. Random CNT aggregates force gas molecules to follow discontinuous and tortuous pathways, counteracting intrinsic transport benefits [227,232]. To mitigate these issues, combined strategies involving CNT functionalization, polymer matrix modification, and external-field-assisted processing have been proposed [224,225,232,233]. Magnetic-field-assisted fabrication is particularly effective for Fe@MWCNT-based MMMs, enabling in situ spatial organization without extensive chemical modification. Studies show that such alignment improves filler dispersion, reduces interfacial defects, and enhances gas separation performance [38,233,234,235]. In magnetically aligned MMMs, Fe@MWCNTs can form ordered suprastructures that act as preferential transport pathways. Paramagnetic or highly polarizable gases interact more strongly with magnetically active regions, whereas diamagnetic gases are mainly transported through the polymer matrix along more tortuous routes. This dual transport mechanism departs from the classical solution–diffusion model and highlights the potential of magnetic MMMs as a new generation of smart separation materials [222,233,234]. Recent studies further confirm the importance of magnetic-field-assisted organization. Zhu et al. incorporated Fe₃O₄-functionalized GO into Pebax, where magnetic alignment produced shorter, oriented transport pathways and improved interfacial compatibility, achieving 538 Barrer CO₂ permeability with CO₂/CH₄ and CO₂/N₂ selectivities of 47 and 75, respectively [236]. Yap et al. showed that alternating magnetic fields improved filler dispersion in γ-Fe₂O₃/TiO₂-based MMMs, doubling CO₂/CH₄ selectivity while reducing permeability due to suppressed non-selective defects [237].

In summary, magnetic nanofillers—particularly Fe@MWCNTs and iron-oxide-functionalized carbon nanostructures—enable magnetic-field-assisted alignment and controlled transport anisotropy in CO₂ separation membranes. With reported values of α_CO2/N2 ≈ 58–75, α_CO2/CH4 ≈ 3.4–47, and CO₂ permeability of ~59–538 Barrer, these systems demonstrate promising laboratory-scale performance. Nevertheless, optimization of filler functionalization, magnetic-field parameters, dispersion control, and scale-up strategies remains necessary for industrial deployment.

5.4.5. Metal–Organic Frameworks (MOFs)

Metal–organic frameworks (MOFs) are a rapidly expanding class of crystalline porous materials extensively investigated for gas separation due to their high surface area, tunable pore structures, and adjustable chemistry. In mixed matrix and hybrid membranes, MOFs combine selective adsorption and molecular sieving with polymer processability, often enhancing CO₂ separation performance. A prominent research direction involves integrating MOFs, particularly ZIF-8, with two-dimensional (2D) materials such as graphene oxide (GO), where GO interlayer nanochannels confine gas transport and MOFs introduce additional adsorption and sieving sites (

Table 6. Comparison of hybrid membranes based on various MOF fillers applied for CO₂ Separation.

Filler Type	Polymer Matrix	Filler Loading [wt.%]	Conditions	Permeability (Barrer)	Selectivity α*	Key Advantages	Main Limitations	Ref.
ZIF-8@GO	Pebax	20	p = 3 bar, T = 25 °C	PCO2 = 136.2	CO2/N2 = 77.9	Improved dispersion, reduced transport resistance, enhanced mechanical strength	High filler loading may reduce flexibility and scalability	[238]
ZIF-8 within GO interlayers	ZnEG on alumina HF support	mass ratio of ZnEG/GO = 90	p = 1 bar, T = 25 °C	PH2 = 365.4 PCO2 = 11.9	H2/CO2 = 30.8	Improved nanoparticle localization, minimized agglomeration	Multi-step synthesis; ceramic support limits scale-up	[239]
NG	PEO	3	p = 1 bar, T = 25 °C	PH2 = 32,240 PCO2 = 1289.6	H2/CO2 = 25	Strong CO2 affinity of PEO; ordered layered architecture	PEO crystallinity reduces permeability	[240]
ZIF-8 +[Emim][Ac] HKUST-1	Chitosan	10 5	T = 25 °C	PCO2 = 5413 PCO2 = 4754	CO2/N2 = 11.5 CO2/N2 = 19.3	Improved interfacial adhesion; reduced filler content	Phase stability; potential IL leaching	[171]
UiO-66/UiO-66–NH2 (60–80 nm)	-	-	-	PCO2 = 39.3	CO2/N2 = 31.3	High chemical & hydrothermal stability; humid resistance	Moderate permeability gains vs. ZIF systems	[241]
MOF-74(Ni)	-	-	p = 0.15 bar, T = 25 °C	-	CO2/N2 = 49	Strong adsorption via open metal sites	Limited permeability data	[172]
ZIF-8	Polyimide P84	17	p = 3 bar, T = 25 °C	PCO2 = 10.92	CO2/CH4 = 92.6	Synergistic adsorption–diffusion mechanism	Dependent on dispersion quality	[173]
UiO-66–NH2	Pebax 2533	10	p = 2 bar, T = 25 °C	PCO2 = 140.4	CO2/N2 = 37	Industrially relevant configuration; scalable geometry	Lower performance than flat-sheet membranes	[184]

, Figure 3). However, uniform dispersion and interfacial stability remain key challenges. For instance, Yang et al. developed a surface-growth strategy in which ZIF-8 crystals were grown directly on GO nanosheets, inducing electrostatic repulsion between GO layers, improving dispersion, and reducing transport resistance [238]. At 20 wt.% ZIF-8@GO loading, the membrane achieved a CO₂ permeability of 136.2 Barrer and CO₂/N₂ selectivity of 77.9, although such high filler contents may limit mechanical flexibility and scalability. Wu et al. proposed an alternative precursor-transformation method, forming ZIF-8 nanoparticles in situ within GO interlayers from zinc ethylene glycol fibers, which minimized agglomeration but required multistep synthesis and ceramic supports, potentially hindering large-scale application [239]. Beyond MOF–GO hybrids, layered membranes based on nanoporous graphene and polymers have been reported. Choi et al. fabricated NG/PEO membranes using GO liquid crystals as scaffolds, benefiting from the high CO₂ affinity of PEO, though polymer crystallinity limited permeability [240]. Zhu et al. replaced PEO with polyethylene glycol via in situ polymerization, improving permeability while maintaining stability, though durability under humid conditions remained a concern [104]. In conventional MMMs, Basu et al. compared multiple MOFs in PDMS and Matrimid matrices, finding that HKUST-1 and MIL-53(Al) yielded higher CO₂ selectivity due to open metal sites and framework breathing effects, albeit at the risk of defects and embrittlement at high loadings [170]. Interfacial compatibility improvements have been achieved using ternary and functionalized systems. Casado-Coterillo et al. introduced chitosan/ZIF-8/[Emim][Ac] MMMs, where the ionic liquid enhanced adhesion and enabled high CO₂/N₂ selectivity at low ZIF-8 loadings (~5 wt.%), although phase stability and leaching remain concerns [171]. UiO-66-based fillers have attracted attention due to their chemical and hydrothermal stability; incorporation of UiO-66 and UiO-66–NH₂ nanoparticles increased CO₂ permeability up to 2.5-fold, with UiO-66–NH₂ exhibiting up to 88% higher selectivity due to amine–CO₂ interactions and Zr–OH groups, while maintaining performance under humid conditions [241]. Lei et al. emphasized the role of open metal sites in MOF-74(Ni), achieving CO₂/N₂ selectivity of 49 [172], whereas Guo et al. demonstrated synergistic adsorption–diffusion mechanisms in ZIF-8-based MMMs for CO₂/CH₄ separation [173]. Although most studies focus on flat-sheet membranes, hollow fiber configurations offer significant industrial advantages. Li et al. fabricated Pebax 2533 hollow-fiber MMMs containing 10 wt.% amine-functionalized UiO-66, achieving CO₂/N₂ selectivity of 37 and CO₂ permeance of 26 GPU, corresponding to a CO₂ permeability of 140.4 Barrer [184]. While performance was lower than flat-sheet analogues, hollow fibers provide improved scalability potential [242,243,244]. Overall, MOF-based MMMs function as adsorption-dominated hybrid systems, where separation performance is governed primarily by CO₂-philic microporous fillers rather than the polymer matrix alone. Their three-dimensional crystalline architectures, tunable pore sizes, and open metal sites enable strong CO₂ adsorption and effective molecular discrimination, often allowing MMMs to surpass Robeson upper bounds under CO₂-rich conditions [245,246,247,248,249,250,251]. Nevertheless, poor polymer–filler compatibility frequently induces interfacial voids and non-selective transport pathways, making interfacial engineering critical [252,253]. Recent strategies, including surface functionalization and composite MOF@COF core–shell architectures, have improved dispersion and adhesion while preserving high adsorption capacity [252,253,254]. Consequently, future development is expected to emphasize rational hybrid design rather than high filler loadings, positioning MOFs as CO₂-selective components in application-specific multifunctional membrane systems [253,255].

5.4.6. MXene-Based Fillers

MXenes are an emerging family of two-dimensional (2D) transition-metal carbides and nitrides, first reported by Naguib et al. in 2011, with the general formula M_n+1X_nT_x, where surface terminations (–O, –OH, –F) are introduced during selective etching of MAX phases [256]. The abundance of surface functional groups imparts strong hydrophilicity, high polarity, and chemical reactivity, making MXenes attractive for membrane-based gas separation [257,258]. Compared with graphene-derived materials, MXenes exhibit higher mechanical robustness, lower diffusion resistance, and good processability, although their reactive surfaces raise stability and compatibility concerns [259]. To address interfacial challenges in mixed matrix membranes (MMMs), various surface modification strategies have been employed, including non-covalent interactions such as hydrogen bonding and electrostatic attraction [260], as well as covalent grafting and polymer coatings [261]. These approaches improve MXene–polymer adhesion, suppress non-selective voids, and enhance separation performance (

Table 7. Performance parameters of MXene-based hybrid membranes for CO₂ separation.

Membrane Type	Filler Type	Polymer Matrix	Filler Loading [wt.%]	Conditions	Permeability (Barrer)	Selectivity (α*)	Modification Strategy	Key Advantages	Ref.
Ti3C2Tx/Pebax1657 MMM	Ti3C2Tx MXene nanosheets	Pebax1657	0.1	p = 4 bar T = 25 °C	PCO2 = 126	CO2/N2 = 96 CO2/H2 = 12.8, CO2/CH4 = 21.9	Hydrogen bonding (MXene–amide groups)	Enhanced interfacial adhesion; reduced non-selective voids	[263]
NH2-MIL-125 hybrid membrane (MXene-derived Ti source)	-	-	0.025	T = 30 °C	PCO2 = 186.5 PH2 = 1492	H2/CO2 = 8.0	Solvothermal epitaxial growth	MXene enabled MOF growth and improved membrane integrity	[264]
Supported ionic liquid membrane (Ti3C2Tx + ChCl/EG)	-	-	-	p = 0.6 bar T = 25 °C	PCO2 = 52.7	CO2/N2 = 319.15, CO2/CH4 = 249.01, CO2/H2 = 12.38	Deep eutectic solvent confinement between layers	Stabilized lamellar channels via hydrogen bonding and electrostatic interactions	[265]
Self-crosslinked MXene Ti3CTx hollow fiber membrane	-	-	-	p = 1 bar T = 25 °C	PH2 = 15.53 PCO2 = 0.512	H2/CO2 = 30.3	Thermal self-crosslinking	Achieved uniform stacking and controlled interlayer spacing	[266]
Pd2+-intercalated MXene membrane				T = 25 °C	PH2 = 1108 PCO2 = 3.11	H2/CO2 = 356	Pd2+ ion intercalation	Reversible Pd–H interaction; exceptional selectivity under ambient conditions	[268]

). When MXenes function as primary membrane-forming units, gas transport occurs through slit-like interlayer nanochannels, making regulation of gallery spacing crucial. Shen et al. demonstrated that intercalation with polyethyleneimine and borate species enabled selective CO₂ transport, while suppressing swelling and enhancing CO₂/N₂ and CO₂/CH₄ selectivity [262]. Interfacial defects remain a major limitation in MXene–polymer composites. Shamsabadi et al. incorporated MXene nanosheets into Pebax-1657, where hydrogen bonding with polymer amide groups improved interfacial cohesion, promoted favorable phase separation, and increased both CO₂ permeability and selectivity [263]. MXenes have also been utilized as functional precursors in hybrid systems. Sun et al. used MXene as a titanium source for epitaxial growth of NH₂-MIL-125 membranes [264], while Lin et al. confined a deep eutectic solvent (ChCl/EG) between Ti₃C₂Tx layers, forming a supported ionic liquid membrane with enhanced CO₂ permeability stabilized by hydrogen bonding and electrostatic interactions [265]. Random stacking of MXene nanosheets can broaden interlayer spacing and reduce molecular sieving efficiency, particularly in hollow-fiber formats. Qu et al. addressed this issue by developing self-crosslinked MXene hollow fibers via thermal treatment, achieving uniform stacking and controlled spacing [266]. Zhang et al. suppressed interfacial defects through electrostatic assembly using polyethylenimine interlayers [267]. More recently, Wang et al. introduced Pd²⁺ ions into MXene galleries, enabling reversible Pd–H interactions and achieving exceptional H₂/CO₂ selectivity of 242 under ambient conditions, surpassing many benchmark membranes [268]. Despite these advances, MXene-based membranes face challenges related to oxidation, uncontrolled stacking, and interfacial incompatibility with polymers, which can compromise long-term stability and hinder industrial implementation. Nevertheless, MXene-based MMMs represent a distinct class of hybrid membranes, where separation performance is governed by 2D transport confinement rather than intrinsic porosity. Owing to their layered morphology and tunable surface chemistry, MXene-containing MMMs exhibit competitive performance (α_CO2/N2 ≈ 96–319; α_CO2/CH4 ≈ 22–249; PCO₂ ≈ 23–126 Barrer). While most applications focus on H₂-related separations, particularly H₂/CO₂, continued advances in surface stabilization, hybrid architectures, and nanosheet alignment are expected to further expand the role of MXenes as precision transport regulators in multifunctional membrane systems [269,270,271].

5.4.7. Oxide Nanoparticles

Oxide nanoparticles, including silica, alumina, TiO₂, ZnO, NiO, and related materials, are widely used as fillers in mixed matrix membranes (MMMs) to modify polymer structure and gas transport pathways (

Table 8. Performance parameters of oxide nanoparticles-based mixed matrix membranes for CO₂ separation.

Filler	Polymer Matrix	Loading (wt.%)	Conditions	Permeability (Barrer)	Selectivity (α*)	Advantages	Limitations	Ref.
MgO-Ag+	Matrimid® 5218	20	p = 3.5 bar T = 35 °C	PCO2 = 4.31	CO2/CH4 = 42.3	High CO2 affinity, reversible π-complexation	Low absolute permeability	[273]
SiO2	PIM	6.7	p = 0.28 bar T = 23 °C	PCO2 = 6200	CO2/N2 = 15	Increased permeability due to interfacial voids	Reduced CO2/N2 selectivity	[274]
TiO2	PVAc	10	p = 2 bar T = 30 °C	PCO2 = 5.26	CO2/N2 = 74.3	Improved thermal stability and gas permeability	Limited selectivity improvement	[275]
ZnO	PEBAX	10	p = 2 bar T = 30 °C	PCO2 = 149.81	CO2/CH4 = 23.9	Enhanced CO2 permeability and selectivity	Moderate loading required	[276]
Al2O3	PU	20	p = 1 bar T = 35 °C	PCO2 = 74.67	CO2/N2 = 67.89 CO2/CH4 = 23.48	Significant selectivity improvement	Decreased permeability	[277]
NiO	PU	5	p = 1 bar T = 30 °C	PCO2 = 321	CO2/N2 = 67.72 CO2/CH4 = 21.76	Improved selectivity at low loading	Slight permeability decrease at high loading	[278]
SiO2–APTMS	PEG	2.7	T = 35 °C	PCO2 = 134	CO2/N2 = 62 CO2/CH4 = 22	Improved permeability without selectivity loss	Low loading required	[279]
Dendritic Amino SiO2	6FDA-DABA	25	p = 1 bar T = 35 °C	PCO2 = 1920	CO2/CH4 = 23	High CO2 permeability with maintained selectivity	High filler content can be challenging	[280]
Dendritic Amino SiO2	PIM-1	50	p = 1 bar T = 35 °C	PCO2 = 15,200	CO2/N2 = 16.6 CO2/CH4 = 11.4	Surpassed Robeson 2008 upper bound	Very high loading may cause brittleness	[280]
POSS–NH2	GO	5	p = 2 bar T = 25 °C	–	CO2/CH4 = 74.5	Improved interlayer spacing, reduced swelling	Limited permeability data	[282]
PEG–POSS	PMHS	0.25	p = 2 bar T = 25 °C	PCO2 = 679	CO2/CH4 = 38.1	Tunable flexibility and transport	Membrane brittleness at high POSS	[283]
poly–POSS	PI	0.9	p = 10 bar T = 300 °C	PCO2 = 69 PH2 = 522	H2/CO2 = 7.6	Enhanced selectivity and permeance	Requires ionic liquid	[284]

) [272]. By tailoring polymer–filler interfaces or introducing additional diffusion channels, oxide fillers often enhance gas permeability, although their impact on selectivity strongly depends on filler chemistry and loading. Several studies have demonstrated improved CO₂ separation through specific chemical interactions. Hosseini et al. incorporated Ag⁺-functionalized MgO nanoparticles into Matrimid^® 5218, where reversible π-complexation between Ag⁺ and CO₂ increased CO₂ solubility, resulting at 20 wt.% loading in 50% and 35% improvements in CO₂/CH₄ and CO₂/N₂ selectivity, respectively [273]. By contrast, Ahn et al. reported that silica nanoparticles in PIM matrices increased permeability due to void formation around impermeable particles, but at the cost of reduced CO₂/N₂ selectivity [274]. More balanced performance enhancements have been observed for metal oxides such as TiO₂ and ZnO. Ahmad et al. showed that incorporation of up to 10 wt.% TiO₂ into PVAc improved thermal stability and increased CO₂, O₂, and H₂ permeability by 79%, 95%, and 62%, respectively, while simultaneously increasing selectivity by 14–38% [275]. Similarly, ZnO-filled PEBAX membranes achieved 13% and 21% increases in CO₂ permeability and CO₂/CH₄ selectivity at 10 wt.% loading [276]. Alumina and NiO fillers typically favored selectivity over permeability; alumina in polyurethane increased CO₂/CH₄ and CO₂/N₂ selectivity to 23.48 and 67.88 at 30 wt.%, albeit with reduced permeability [277], while NiO showed selectivity gains at low loadings and slight permeability losses at higher contents [278]. Surface functionalization has proven effective in optimizing oxide-based MMMs. Su et al. demonstrated that APTES-functionalized SiO₂ in crosslinked PEG membranes enhanced CO₂ permeability without sacrificing selectivity [279]. Kudo et al. achieved notable performance by anchoring dendritic amino groups on silica nanoparticles; in 6FDA-DABA polyimide at 25 wt.%, CO₂ permeability reached 1920 Barrer while maintaining CO₂/N₂ selectivity, and at 50 wt.% in PIM-1, pearl-necklace-like aggregates generated mesoporosity sufficient to surpass Robeson’s 2008 upper bound [280]. Polyhedral oligomeric silsesquioxane (POSS) represents a distinct class of silica-based nanostructures offering excellent polymer compatibility and tunable functionality. Amino-functionalized POSS incorporated into graphene oxide membranes increased CO₂/CH₄ selectivity from 10 to 74.5 by regulating interlayer spacing and suppressing swelling [281]. POSS–PEG composites enabled simultaneous control of flexibility and transport, with an 80:20 PEG:POSS ratio yielding 679 Barrer CO₂ permeability and CO₂/N₂ selectivity of 38.1 [282]. Ionic-liquid-functionalized POSS in polysulfone further improved CO₂/N₂ selectivity (24.1) and CO₂ permeance (34.98 GPU, corresponding to 679 Barrer) [282]. For CO₂/H₂ separation, functionalized polyPOSS-imide and Pebax/POSS membranes achieved H₂/CO₂ selectivity of 7.6 and CO₂/H₂ selectivity of 52.3, respectively, aided by CO₂-induced plasticization effects [283,284]. In summary, oxide nanoparticles—including TiO₂, ZnO, Al₂O₃, NiO, silica derivatives, and POSS—provide multiple routes to improve CO₂ separation in MMMs by enhancing permeability, tuning selectivity through chemical interactions, and increasing structural robustness. Typical performances (α_CO2/N2 ≈ 15–74, α_CO2/CH4 ≈ 22–74, CO₂ permeability ≈ 4–15,200 Barrer) indicate moderate selectivity enhancement compared with crystalline porous fillers. Nevertheless, aggregation, interfacial defects, and permeability loss at high filler loadings remain persistent challenges. Future research is expected to focus on hollow oxide structures, dual-functional fillers, and hybrid oxide–MOF/COF systems to further improve the permeability–selectivity balance.

5.4.8. g-C₃N₄-Based Membranes

Graphitic carbon nitride (g-C₃N₄) has recently attracted attention as a two-dimensional (2D) material for gas separation membranes due to its high chemical and thermal stability, intrinsic porosity, and atomic-scale thickness [285,286,287,288]. Structurally, g-C₃N₄ consists of stacked graphite-like layers built from tri-s-triazine units linked by amino groups, forming a π-conjugated sp² C–N network [286]. A key advantage of g-C₃N₄ is the presence of intrinsic sub-nanometer pores (3.1–3.4 Å), which enable molecular sieving between H₂ and larger gas molecules, making it attractive for selective gas transport [280]. Consequently, g-C₃N₄ nanosheets are increasingly employed as fillers in composite membranes (Table 9). Bulk g-C₃N₄ is typically synthesized via thermal polymerization and subsequently exfoliated using thermal oxidative etching, hydrothermal, ultrasonic, or chemical treatments. Among these, thermal oxidative etching in air is widely used due to its simplicity and controllable thickness reduction. However, producing high-quality few-layer nanosheets remains challenging, as exfoliation often induces fragmentation and defect formation. Surface functionalization and hybrid membrane design have been widely applied to improve compatibility and performance. Jomekian et al. fabricated a layer-by-layer membrane combining chitosan-modified g-C₃N₄ with ZIF-8 on a PES support, where amine-rich chitosan enhanced CO₂ affinity and interfacial cohesion, resulting in improved CO₂ separation and reduced brittleness [289]. Cheng et al. incorporated thermally etched g-C₃N₄ into Pebax, demonstrating that amine functionalities and intrinsic nanopores promote CO₂ transport through an affinity–sieving mechanism, with further performance improvements achieved by tuning the etching time [290]. To mitigate structural defects arising from exfoliation, Zhou et al. introduced graphene oxide as a reinforcing component, where strong interactions between g-C₃N₄ amine groups and GO carboxyl groups partially healed defects, improving mechanical integrity and long-term stability [291]. Guo et al. adopted an in situ growth strategy by nucleating ZIF-90 on g-C₃N₄ templates, which increased free volume, improved stacking order, and strengthened synergistic sieving–adsorption effects [292]. Enhancing CO₂ affinity through chemical modification has also proven effective. Voon et al. functionalized g-C₃N₄ with sulfonic acid and amine groups prior to incorporation into PIM-1, where sulfonated nanosheets exhibited superior performance due to strong dipole–quadrupole interactions with CO₂ [293]. Niu et al. confined the ionic liquid [EMIm][AcO] within g-C₃N₄-based nanochannels to form a supported ionic liquid membrane, combining high CO₂ affinity with improved structural stability [294]. Despite these advances, g-C₃N₄-based membranes face intrinsic limitations. Exfoliation-induced fragmentation and non-selective defects, incomplete accessibility of intrinsic nanopores within polymer matrices, interfacial void formation, random stacking, and poorly controlled interlayer spacing may reduce selectivity, particularly under humid or high-pressure conditions. Nevertheless, when defect healing, interlayer spacing control, surface functionalization, and hybridization with complementary materials are properly implemented, g-C₃N₄ remains a highly promising filler for advanced gas separation membranes. Overall, g-C₃N₄ offers a thermally stable, nitrogen-rich layered structure capable of enhancing CO₂ sorption and molecular sieving in mixed matrix membranes, achieving reported values of α_CO2/N2 ≈ 20–84 and CO₂ permeability of ~13–3740 Barrer. However, limited dispersibility and defect formation at high loadings constrain practical implementation. Future research should prioritize clear structure–property–performance correlations and scalable fabrication strategies to fully exploit the intrinsic advantages of g-C₃N₄ in membrane-based gas separation.

5.4.9. Layered Double Hydroxide Membranes

Layered double hydroxides (LDHs), also known as hydrotalcite-like compounds, consist of positively charged brucite-like two-dimensional layers separated by interlayer anions and solvent molecules [295]. Their tunable interlayer spacing and flexible metal–anion composition make LDHs attractive candidates for gas separation membranes (Table 9) [296]. Beyond membrane applications, LDHs have been explored in electrochemistry, polymer modification, catalysis, and CO₂ adsorption [297,298,299,300,301,302]. However, membrane fabrication remains challenging, with most approaches relying on top-down or bottom-up assembly strategies. Liu et al. prepared NiAl-CO₃ LDH membranes on porous alumina via in situ growth, achieving enhanced H₂ selectivity and demonstrating that dissolved CO₂ in the precursor solution critically influences membrane orientation and thickness [296]. Subsequently, Liu’s group developed ZIF-8–ZnAl-NO₃ LDH composite membranes through partial transformation, where ZIF-8 was generated in situ within LDH layers. Partial LDH dissolution reduced membrane thickness and increased permeability [297]. These studies highlight the sensitivity of LDH membrane performance to structural organization and phase evolution. Filler distribution and interfacial design strongly affect performance in LDH-based membranes. Zhang et al. fabricated ZIF-8@LDH hybrid fillers and embedded them into Pebax, where hydroxyl-rich LDH interfaces enhanced CO₂ solubility and directional ZIF-8 alignment improved permeability while reducing transport resistance [298]. Similarly, Fan et al. employed vertically aligned CoAl-LDH nanosheets as templates for in situ growth of COF-LZU1, yielding membranes with ultrathin interlayer spacing (0.3–0.4 nm) that exhibited ultrahigh H₂ permeability (7400 Barrer) and H₂/CO₂ selectivity of 31.6 [299]. Despite their hydroxyl-rich surfaces and strong CO₂ affinity, the intrinsically non-porous structure of LDHs can limit permeability. To overcome this, Zheng et al. fabricated three-dimensional hollow CoNi-LDH nanocages using ZIF-67 templates and incorporated them into Pebax. The hollow architecture shortened diffusion pathways, while abundant hydroxyl groups improved CO₂ solubility and polymer–filler compatibility, resulting in enhanced CO₂ permeability and CO₂/N₂ selectivity [300]. Overall, LDHs offer tunable interlayer spacing, compositional flexibility, and high anion-exchange capacity, enabling selective gas transport with reported values of α_CO2/N2 up to 71.7, α_CO2/CH4 up to 31.6, and CO₂ permeability of approximately 11–1300 Barrer. Nevertheless, relatively low mechanical strength, aggregation tendencies, and moisture sensitivity remain key limitations. Future research is therefore expected to focus on LDH functionalization, polymer hybridization to improve dispersion, and controlled layer orientation to enhance directional transport and achieve targeted gas separation performance.

5.4.10. Transition Metal Dichalcogenide Membranes

Transition metal dichalcogenides (TMDs) are two-dimensional (2D) nanomaterials with the general formula MX₂ (M = transition metal; X = S, Se, or Te), characterized by ultrathin layered structures, tunable crystal phases, and high chemical and thermal stability [301,302]. Owing to these properties, TMDs have been widely applied in catalysis, electronics, and energy-related fields [303,304], and more recently explored as fillers in gas separation membranes (Table 9). The first application of TMDs in composite membranes was reported by Shen et al., who incorporated MoS₂ nanosheets into Pebax-1657 supported on polysulfone with a PDMS gutter layer [305]. The PDMS layer preserved membrane integrity, while defect-free MoS₂ dispersion and the CO₂-philic Pebax matrix resulted in simultaneous improvements in permeability and selectivity compared with pristine Pebax membranes. Facilitated transport strategies have also been adopted using ionic liquids (ILs). Chen et al. confined [BMIM][BF₄] between WS₂ nanosheets to form supported ionic liquid membranes, where nanoscale confinement enhanced CO₂ interactions and compressive stability without sacrificing transport rates [306]. Polymer–TMD hybrid systems have further improved performance. Geng et al. blended WS₂ nanosheets with trifluoromethyl-functionalized polymers, increasing free volume, enhancing CO₂ solubilization selectivity, and generating continuous diffusion pathways [307]. More recently, Wang et al. developed Pebax/Cys-MoS₂ membranes using L-cysteine-functionalized MoS₂, where structural defects enhanced CO₂ adsorption, amine groups promoted reactive selectivity, and hydrogen bonding improved polymer–filler compatibility [308]. At only 1.5 wt.% loading, these membranes achieved a CO₂/N₂ selectivity of 120, exceeding previously reported upper bounds. Although research on TMD-based membranes is at an early stage compared with other 2D fillers, their high aspect ratio, tunable surface chemistry, and robust mechanical properties enable outstanding separation performance. Reported examples include amine-functionalized Cys-MoS₂ systems (α_CO2/N2 ≈ 120 at CO₂ permeability ≈ 297 Barrer) and fluoropolymer/WS₂ composites (α_CO2/N2 ≈ 30 at P_CO2 ≈ 472 Barrer) [307,308]. Nevertheless, synthesis complexity, nanosheet aggregation, and dispersion control remain significant challenges. Overall, TMD-based MMMs represent a promising class of emerging fillers for advanced gas separation membranes. Future research is expected to focus on optimized functionalization, controlled nanosheet orientation (e.g., via external fields), and hybrid TMD–COF or TMD–polymer architectures to improve interfacial compatibility and maximize directional transport efficiency.

5.4.11. Covalent Organic Framework Membranes

Covalent organic frameworks (COFs) are regarded as promising materials for gas separation due to their permanent porosity, high crystallinity, thermal stability, and large specific surface area [309]. As two-dimensional (2D) layered materials, COFs can be exfoliated into single- or few-layer nanosheets with high aspect ratios and short, ordered transport pathways [310]. However, the fabrication of continuous, defect-free COF membranes remains challenging, limiting their direct application (Table 9). Tang et al. synthesized TpPa COFs by mechanical grinding and employed graphene oxide (GO) as a structural guide to fabricate layered GO/COF composite membranes via vacuum filtration [311]. The GO layers improved membrane continuity, mechanical robustness, and defect suppression while preserving COF porosity. Interfacial compatibility between COF fillers and polymer matrices is another critical issue. Cao et al. functionalized COFs with polyvinylamine (PVAm) groups, achieving strong interfacial adhesion in PVAm matrices and reducing effective pore mobility, which promoted selective CO₂ sorption and transport [312]. This study emphasized the importance of surface functionalization and pore-environment regulation in COF-based mixed matrix membranes (MMMs). To mitigate aggregation and improve structural integrity, Zhang et al. developed a solvent-free, in situ bottom-up growth strategy for 2D COF/PEO hybrid membranes by molecular-level mixing of COF precursors and polymer chains, yielding continuous membranes with suppressed filler aggregation [313]. Nevertheless, the relatively large intrinsic pore sizes of many COFs (>1 nm) remain a major limitation, as they exceed the kinetic diameters of common gases and reduce molecular sieving efficiency. Several studies have addressed this limitation through multifunctional hybrid filler design. Liu et al. prepared hollow COF microspheres functionalized with polyethylene glycol monomethyl ether and incorporated them into Pebax matrices. The hollow structure shortened diffusion pathways, while PEG functionalization enhanced CO₂ affinity, decreased effective pore size, and improved polymer compatibility [314]. Similarly, Zhang et al. fabricated core–shell ZIF-8@COF hybrids by growing nitrogen-rich nanoporous polytriazine COFs on ZIF-8 cores followed by ethylenediamine functionalization. When embedded in PVAm, these fillers combined COF-facilitated CO₂ transport with ZIF-8 molecular sieving, resulting in improved filler–polymer adhesion and enhanced separation performance [315]. Overall, COF-based MMMs constitute hybrid systems distinguished by excellent polymer compatibility and fully organic, structurally ordered porous frameworks. The covalent nature of COFs minimizes interfacial void formation and contributes to improved mechanical integrity and membrane stability [316,317,318]. Reported performances typically include CO₂ permeabilities of approximately 234–1044 Barrer, CO₂/N₂ selectivities of ~61–91, and CO₂/CH₄ selectivities of ~19–24, particularly for functionalized COF-based MMMs. The principal limitation of COFs arises from their relatively large intrinsic pore sizes, which restrict molecular sieving when applied as standalone fillers [316,319,320]. Consequently, high separation performance generally relies on targeted pore size contraction, CO₂-philic functionalization, or synergistic integration with more selective fillers. In this context, COFs play a crucial interfacial role in hybrid architectures such as MOF@COF core–shell fillers, where they act as organic buffering layers that suppress filler agglomeration and eliminate polymer–filler incompatibility while preserving high permeability [318]. Future development of COF-based MMMs is therefore expected to focus on sub-nanometer pore engineering, advanced surface functionalization, and scalable fabrication approaches, positioning COFs as structurally versatile, interface-friendly components in rationally designed high-performance hybrid membranes [316,317,318,319,320].

Table 9. Performance parameters of g-C₃N₄-based, LDH-based, TMD-based and COF-based hybrid membranes for CO₂ separation.

Membrane Type	Filler	Polymer Matrix	Loading (wt.%)	Conditions	Permeability (Barrer)	Selectivity α*	Advantages	Ref.
Layer-by-layer membrane (Chitosan–g-C3N4/ZIF-8 on PES)	Chitosan– g-C3N4/ ZIF-8	PES	29	p = 2 bar T = 30 °C	PCO2 = 63.5	CO2/CH4 = 24.2	Amine-rich chitosan enhanced CO2 affinity; denser selective layer; reduced brittleness	[289]
g-C3N4/Pebax MMM	g-C3N4	Pebax	0.25	p = 3 bar T = 25 °C	PCO2 = 5900	CO2/N2 = 67.2	Affinity–sieving mechanism; etching time tuning optimized P and α	[290]
g-C3N4/GO composite membrane	GO	g-C3N4	30	T = 30 °C	PH2 = 451 PCO2 = 11.5	H2/CO2 = 39.2	Defect “healing” via –NH/–COOH interactions; enhanced mechanical integrity and stability	[291]
ZIF-90@g-C3N4 hybrid membrane	ZIF-90@g-C3N4	Pebax	8	p = 2 bar T = 25 °C	PCO2 = 110.5	CO2/N2 = 84.4	Increased free volume; enhanced sieving–adsorption synergy	[292]
Functionalized g-C3N4/PIM-1 MMM	g-C3N4	PIM-1	1	p = 3.5 bar T = 35 °C	PCO2 = 3740	CO2/N2 = 19.8 CO2/CH4 = 12.4	Strong dipole–quadrupole interactions with CO2	[293]
g-C3N4-based SILM		g-C3N4-		p = 1 bar T = 25 °C	PCO2 = 794 PCO2 = 928	CO2/N2 = 52.49 CO2/CH4 = 48.41	Combined high CO2 affinity with structural stability	[294]
ZnAl–NO3 LDH membrane on porous alumina		ZnAl–NO3 LDH		p = 1 bar T = 180 °C	PH2 = 267.5 PCO2 = 46.1	H2/CO2 = 5.8	CO2 in precursor solution influenced membrane orientation and thickness	[296]
ZIF-8@LDH/Pebax MMM	ZIF-8@LDH	Pebax	2	-	PCO2 = 1307	CO2/CH4 = 31.6	Hydroxyl-rich interfaces enhanced CO2 solubility; aligned ZIF-8 reduced resistance	[298]
COF-LZUI/CoAl-LDH membrane	-	-	-	p = 1 bar T = 25 °C	PH2 = 7200 PCO2 = 227.8	H2/CO2 = 31.6	0.3–0.4 nm interlayer spacing; ultrahigh H2 permeability	[299]
3D hollow CoNi-LDH/Pebax MMM	CoNi-LDH	Pebax	1	p = 8 bar T = 25 °C	PCO2 = 172.62	CO2/N2 = 71.66	Hollow structure shortened diffusion pathways; hydroxyl groups enhanced CO2 affinity	[300]
MoS2/Pebax-1657 MMM	MoS2	Pebax-1657	0.15	p = 2 bar T = 30 °C	PCO2 = 64	CO2/N2 = 93	Defect-free MoS2 dispersion; PDMS gutter prevents pore blockage; enhanced CO2-philic transport pathways	[305]
WS2–IL SILM	WS2–IL			p = 14 bar T = 25 °C	PCO2 = 18.9	CO2/N2 = 153.21 CO2/CH4 = 68.81 CO2/H2 = 13.56	Nanoscale IL confinement; facilitated CO2 transport; strong IL–CO2 interactions; improved compressive stability	[306]
WS2/fluoropolymer composite	WS2	FPPO	10	p = 0.7 bar T = 18 °C	PCO2 = 472	CO2/N2 = 29.6 CO2/CH4 = 39.4	Increased free volume via –CF3 groups; continuous CO2 diffusion channels; improved solubilization selectivity	[307]
Cys-MoS2/Pebax MMM	Cys-MoS2	Pebax	1.5	p = 1 bar T = 25 °C	PCO2 = 297	CO2/N2 = 120	Amine-enhanced CO2 affinity; defect-induced selective adsorption; strong hydrogen bonding improves interfacial compatibility	[308]
GO/COF layered composite membrane	GO/COF				PCO2 = - PH2 = 1.067 × 10−6 mol·m−2·s−1·Pa−1	H2/CO2 = 25.57	GO-guided layered architecture; continuous and defect-free membrane; improved mechanical robustness	[311]
PVAm- functionalized COF MMM	COF	PVAm/mPSf	10	p = 1.5 bar T = 25 °C	PCO2 = 234.6	CO2/H2 = 17.2	Enhanced interfacial adhesion; restricted pore mobility; selective CO2 sorption via amine groups	[312]
2D COF/PEO hybrid membrane					PCO2 = 803.9	CO2/N2 = 61.4 CO2/CH4 = 19.8 CO2/H2 = 15.0	Molecular-level mixing; reduced aggregation; continuous transport nanochannels	[313]
Hollow COF microspheres/Pebax MMM	Hollow COF microspheres	Pebax			PCO2 = 1044	CO2/CH4 = 24	Hollow structure lowers diffusion resistance; PEG enhances CO2 affinity; improved polymer compatibility	[314]
ZIF-8@COF (core–shell) MMM	ZIF-8@COF				PCO2 = 288	CO2/N2 = 91	Dual-function core–shell design: COF transport + ZIF-8 molecular sieving; improved filler–polymer adhesion	[315]

Table 9. Performance parameters of g-C₃N₄-based, LDH-based, TMD-based and COF-based hybrid membranes for CO₂ separation.

Membrane Type	Filler	Polymer Matrix	Loading (wt.%)	Conditions	Permeability (Barrer)	Selectivity α*	Advantages	Ref.
Layer-by-layer membrane (Chitosan–g-C3N4/ZIF-8 on PES)	Chitosan– g-C3N4/ ZIF-8	PES	29	p = 2 bar T = 30 °C	PCO2 = 63.5	CO2/CH4 = 24.2	Amine-rich chitosan enhanced CO2 affinity; denser selective layer; reduced brittleness	[289]
g-C3N4/Pebax MMM	g-C3N4	Pebax	0.25	p = 3 bar T = 25 °C	PCO2 = 5900	CO2/N2 = 67.2	Affinity–sieving mechanism; etching time tuning optimized P and α	[290]
g-C3N4/GO composite membrane	GO	g-C3N4	30	T = 30 °C	PH2 = 451 PCO2 = 11.5	H2/CO2 = 39.2	Defect “healing” via –NH/–COOH interactions; enhanced mechanical integrity and stability	[291]
ZIF-90@g-C3N4 hybrid membrane	ZIF-90@g-C3N4	Pebax	8	p = 2 bar T = 25 °C	PCO2 = 110.5	CO2/N2 = 84.4	Increased free volume; enhanced sieving–adsorption synergy	[292]
Functionalized g-C3N4/PIM-1 MMM	g-C3N4	PIM-1	1	p = 3.5 bar T = 35 °C	PCO2 = 3740	CO2/N2 = 19.8 CO2/CH4 = 12.4	Strong dipole–quadrupole interactions with CO2	[293]
g-C3N4-based SILM		g-C3N4-		p = 1 bar T = 25 °C	PCO2 = 794 PCO2 = 928	CO2/N2 = 52.49 CO2/CH4 = 48.41	Combined high CO2 affinity with structural stability	[294]
ZnAl–NO3 LDH membrane on porous alumina		ZnAl–NO3 LDH		p = 1 bar T = 180 °C	PH2 = 267.5 PCO2 = 46.1	H2/CO2 = 5.8	CO2 in precursor solution influenced membrane orientation and thickness	[296]
ZIF-8@LDH/Pebax MMM	ZIF-8@LDH	Pebax	2	-	PCO2 = 1307	CO2/CH4 = 31.6	Hydroxyl-rich interfaces enhanced CO2 solubility; aligned ZIF-8 reduced resistance	[298]
COF-LZUI/CoAl-LDH membrane	-	-	-	p = 1 bar T = 25 °C	PH2 = 7200 PCO2 = 227.8	H2/CO2 = 31.6	0.3–0.4 nm interlayer spacing; ultrahigh H2 permeability	[299]
3D hollow CoNi-LDH/Pebax MMM	CoNi-LDH	Pebax	1	p = 8 bar T = 25 °C	PCO2 = 172.62	CO2/N2 = 71.66	Hollow structure shortened diffusion pathways; hydroxyl groups enhanced CO2 affinity	[300]
MoS2/Pebax-1657 MMM	MoS2	Pebax-1657	0.15	p = 2 bar T = 30 °C	PCO2 = 64	CO2/N2 = 93	Defect-free MoS2 dispersion; PDMS gutter prevents pore blockage; enhanced CO2-philic transport pathways	[305]
WS2–IL SILM	WS2–IL			p = 14 bar T = 25 °C	PCO2 = 18.9	CO2/N2 = 153.21 CO2/CH4 = 68.81 CO2/H2 = 13.56	Nanoscale IL confinement; facilitated CO2 transport; strong IL–CO2 interactions; improved compressive stability	[306]
WS2/fluoropolymer composite	WS2	FPPO	10	p = 0.7 bar T = 18 °C	PCO2 = 472	CO2/N2 = 29.6 CO2/CH4 = 39.4	Increased free volume via –CF3 groups; continuous CO2 diffusion channels; improved solubilization selectivity	[307]
Cys-MoS2/Pebax MMM	Cys-MoS2	Pebax	1.5	p = 1 bar T = 25 °C	PCO2 = 297	CO2/N2 = 120	Amine-enhanced CO2 affinity; defect-induced selective adsorption; strong hydrogen bonding improves interfacial compatibility	[308]
GO/COF layered composite membrane	GO/COF				PCO2 = - PH2 = 1.067 × 10−6 mol·m−2·s−1·Pa−1	H2/CO2 = 25.57	GO-guided layered architecture; continuous and defect-free membrane; improved mechanical robustness	[311]
PVAm- functionalized COF MMM	COF	PVAm/mPSf	10	p = 1.5 bar T = 25 °C	PCO2 = 234.6	CO2/H2 = 17.2	Enhanced interfacial adhesion; restricted pore mobility; selective CO2 sorption via amine groups	[312]
2D COF/PEO hybrid membrane					PCO2 = 803.9	CO2/N2 = 61.4 CO2/CH4 = 19.8 CO2/H2 = 15.0	Molecular-level mixing; reduced aggregation; continuous transport nanochannels	[313]
Hollow COF microspheres/Pebax MMM	Hollow COF microspheres	Pebax			PCO2 = 1044	CO2/CH4 = 24	Hollow structure lowers diffusion resistance; PEG enhances CO2 affinity; improved polymer compatibility	[314]
ZIF-8@COF (core–shell) MMM	ZIF-8@COF				PCO2 = 288	CO2/N2 = 91	Dual-function core–shell design: COF transport + ZIF-8 molecular sieving; improved filler–polymer adhesion	[315]

Advanced two-dimensional (2D) fillers, including MXenes, transition metal dichalcogenides (TMDs), layered double hydroxides (LDHs), covalent organic frameworks (COFs) and graphitic carbon nitride (g-C₃N₄), have attracted significant attention for energy, environmental, and biomedical applications. Across these material families, long-term stability, oxidation or degradation processes, and scalable production routes remain recurring bottlenecks. MXenes, in particular, suffer from rapid oxidation, restacking, and structural deterioration, especially in aqueous colloids and at elevated temperatures, leading to pronounced losses in electrical conductivity and functional performance [321,322,323,324,325] HF-etched MXenes are especially prone to instability and oxidation under ambient conditions [321,322]. Their stability can be improved through non-HF or molten salt etching routes, which enable better control over surface terminations and interlayer spacing [321,322,326], as well as by low-temperature storage, inert atmospheres, non-polar solvents, and the use of larger flakes to suppress defect-driven oxidation [321]. Additional strategies such as ion intercalation or pillaring and heterostructure formation have been shown to mitigate restacking and preserve charge-transport pathways [321,322,325]. From a manufacturing perspective, HF or in situ HF wet etching remains the most mature and scalable MXene production route, although its industrial implementation is constrained by safety concerns and limited long-term material stability. Molten salt etching represents a scalable and versatile alternative with tunable surface terminations, but requires strict control of post-etching oxidation. In contrast, bottom-up approaches (e.g., CVD and PEPLD) enable the synthesis of high-quality, well-controlled MXene films, yet their applicability is currently limited to ultrathin layers and laboratory-scale production, restricting their industrial relevance. In comparison, g-C₃N₄ fillers are widely described as chemically and thermally stable, exhibiting high durability and long-term usability in membranes and photocatalytic systems, even under harsh operating conditions [323,327,328,329,330,331,332]. Their primary limitations are not oxidation but rather low surface area, charge-carrier recombination, and photochemical instability under strong illumination. These challenges are commonly addressed through doping, heterojunction engineering, and polymer blending, which also enhance mechanical robustness and membrane stability over repeated operating cycles [329,330,331,332]. Covalent organic frameworks (COFs) share with g-C₃N₄ a comparatively high chemical and thermal stability arising from their fully covalent backbones and crystalline architectures. In hybrid membranes, COFs exhibit strong interfacial affinity toward polymer matrices, which suppresses non-selective void formation and improves mechanical integrity [316,317,318]. Unlike MXenes, COFs are not inherently susceptible to oxidative degradation, and unlike LDHs they do not undergo pH-triggered structural collapse. However, their relatively large intrinsic pore sizes—often exceeding 1 nm—limit molecular sieving efficiency and intrinsic selectivity when COFs are employed as standalone fillers [316,319,320]. Consequently, high CO₂ separation performance in COF-based membranes generally relies on pore-size contraction, chemical functionalization with CO₂-philic groups, or synergistic integration with more selective fillers. In this context, COFs increasingly serve as interfacial and structural buffering layers in hybrid architectures such as MOF@COF core–shell systems, where they mitigate filler agglomeration, enhance dispersion, and preserve permeability while improving selectivity [318,319,320]. LDHs, by contrast, exhibit pH-responsive degradation due to their acidic sensitivity; excessively rapid degradation compromises functionality, whereas overly slow degradation raises long-term safety concerns. Fine-tuning particle size and thickness, as well as polymer or inorganic encapsulation, is therefore employed to balance degradability and structural stability [333,334]. Nevertheless, comprehensive long-term in vivo safety and chronic exposure data remain limited [333]. Meanwhile, TMDs have seen substantial advances in large-area epitaxial growth, including wafer-scale, single-crystal films, directly addressing industrial scalability requirements. Remaining challenges include precise defect control, thickness and compositional uniformity, and application-specific growth strategies [335]. Overall, across MXenes, TMDs, LDHs, and g-C₃N₄, several cross-cutting themes emerge: (i) oxidation and colloidal instability are particularly severe for MXenes; (ii) g-C₃N₄ and many TMDs exhibit comparatively robust chemical stability but require structural and optical engineering; (iii) LDHs demand careful control of pH-triggered degradation and long-term biosafety; and (iv) scalable and safer synthesis routes—such as molten-salt MXenes, epitaxial TMD growth, simple pyrolysis for g-C₃N₄, and tunable LDH syntheses—are advancing but have yet to achieve full industrial maturity.

6. Membrane Module Configurations and Process Design Considerations

For practical deployment, membranes must be integrated into leak-tight modules that ensure mechanical stability, protect fragile materials, and maximize productivity. A key parameter is packing density—the ratio of effective membrane area to module volume. Common configurations include plate-and-frame and spiral-wound modules for flat sheets, and hollow fiber or tubular modules for cylindrical membranes [336] (

Table 10. Comparative summary of membrane module configurations for CO₂ separation.

Module Type	Membrane Geometry	Typical Flow Pattern	Key Advantages	Major Limitations	Best-Suited Applications	Representative Notes/Studies
Plate-and-frame	Flat-sheet	Cross-flow	Simple design; easy membrane replacement; good control of operating conditions	Very low packing density; high module footprint; limited industrial scalability	Laboratory and pilot-scale studies	Mainly used for material screening rather than industrial CO2 separation
Spiral-wound	Flat-sheet	Cross-flow	Higher packing density than plate-and-frame; mature industrial technology; adjustable hydrodynamics via feed spacers; relatively low concentration polarization	Complex internal structure; difficult cleaning; moderate pressure drop	Flue gas CO2/N2 separation; post-combustion capture	Feed spacer optimization reduces boundary layer resistance
Hollow fiber	Cylindrical (fibers)	Counter-current or cross-flow	Highest packing density (3–4× spiral-wound); low cost per area; high surface-to-volume ratio; ideal for large gas flows	Prone to pressure drop and concentration polarization; sealing and mechanical stability challenges	Industrial-scale CO2 capture; high-throughput gas separation	Optimized designs outperform spiral-wound modules
Tubular	Tubes	Cross-flow	Excellent mechanical strength; high temperature and chemical resistance; easy cleaning	Very low packing density; high capital cost	High-temperature or aggressive gas environments

). Among these, spiral-wound, hollow fiber, and tubular designs are most relevant for CO₂ separation (Figure 4). Spiral-wound modules are widely used due to their relatively high packing density and flexible hydrodynamic design. By optimizing feed spacers, boundary layer thickness and concentration polarization can be reduced. However, their complex internal structure makes cleaning more difficult and increases fouling sensitivity (Figure 4a,b). Hollow fiber modules (Figure 4c) offer the highest packing density—typically three to four times greater than spiral-wound systems under comparable conditions [337]. Their large membrane area per unit volume makes them attractive for high gas flow rates and cost-efficient operation. On the downside, high packing density may induce significant pressure drops and concentration polarization if not properly engineered. Plate-and-frame modules have low packing densities and are therefore mainly used at laboratory or pilot scale. Tubular modules (Figure 4d), commonly applied for metallic or ceramic membranes, provide mechanical robustness and easy cleaning but suffer from a low surface-area-to-volume ratio, limiting large-scale economic viability. Module configuration strongly influences separation performance [338]. Flow patterns determine concentration gradients, boundary layer development, pressure drop, and mass transfer efficiency [339]. Spiral-wound modules typically operate in cross-flow, whereas hollow fiber systems often use counter-current flow. Although counter-current flow offers a higher theoretical average driving force [337], spiral-wound modules often display more favorable hydrodynamics and reduced polarization. In hollow fiber modules, concentration polarization can be mitigated through tangential feed flow, modified retentate withdrawal (e.g., central tubes), double-ended permeate collection, or slight reductions in packing density. When facilitated transport membranes (FTMs) are used, water management becomes critical, as water actively participates in CO₂ transport [340]. Likewise, conventional solution–diffusion membranes such as Nafion, Pebax [341], and Nexar [342] are highly sensitive to humidity, making uniform moisture control essential. To better capture coupled heat and mass transfer effects, Yang et al. [343] proposed a tanks-in-series model for amine-based FTMs, enabling prediction of spatial variations in water vapor, temperature, and pressure along modules. These models can be implemented in process simulators (e.g., Aspen HYSYS) to support scale-up and optimize module design. Adjusting local flow conditions—particularly in hollow fibers—can also prevent reverse permeation and maintain effective driving forces. With optimized design, hollow fiber modules can achieve mass transfer coefficients exceeding those of spiral-wound systems [344], making them promising for advanced gas separation [345]. However, pressure drop and gas leakage, especially at high pressure or temperature, remain significant challenges. Frictional losses and progressive permeate removal reduce transmembrane driving force along the module length. In counter-current hollow fiber systems, pressure drop can be mitigated by shifting to cross-flow operation using a central withdrawal pipe. While this reduces pressure losses, it may lower stage-cut efficiency, requiring careful optimization of fiber length and module diameter. Finally, thermal effects must be considered. Due to the high Joule–Thomson coefficient of CO₂ (~1.11 K·bar⁻¹), significant cooling may occur during permeation under large pressure gradients. Since gas permeability is often temperature-dependent [346], Joule–Thomson cooling should be incorporated into both module design and process modeling for reliable CO₂ separation performance.

A technical assessment is usually the first step before a new membrane is considered for industrial gas separation, with process design at the heart of the evaluation (

Table 11. Summary of membrane process design variables and their impact on CO₂ separation.

Design Variable	Primary Technical Role	Key Advantages	Main Drawbacks/ Trade-Offs
Membrane permeability	Determines gas flux and required membrane area	Enables compact modules	May increase concentration polarization; often trades off with selectivity
Membrane selectivity (CO2/other gases)	Controls product purity and stage requirements	Improves CO2 purity and recovery	Often reduces permeability; may increase membrane cost
Module configuration (spiral-wound, hollow fiber, etc.)	Defines packing density and flow hydrodynamics	Hollow fibers minimize module cost	Higher pressure drop; more complex sealing
Operating pressure	Provides driving force for permeation	Enhances productivity	Compression energy dominates
Operating temperature	Affects permeability and facilitated transport	Improves kinetics in FTMs	Thermal management cost; membrane stability limits
Number of stages	Determines achievable purity and recovery	Enables high-purity CO2 capture	Higher control and maintenance costs
Recycle ratio	Enhances recovery and purity	Allows flexible process tuning	Raises energy consumption and pressure drop
Membrane type selection per stage	Matches performance to stage function	Enables performance–cost optimization	Increases design complexity
Plant capacity	Scale of operation	Membranes cost-effective at small–medium scale	Absorption favored at very large scale
Relative humidity (RH)	Affects solution–diffusion and facilitated transport	Essential for FTMs and hydrophilic polymers	Humidity control adds system complexity

). The goal is to identify an optimized configuration suitable for real deployment, and among all variables, the intrinsic CO₂ separation performance of the membrane plays the decisive role in determining overall feasibility. Based on membrane properties, the process must be designed around operating conditions—pressure, temperature, feed flow rate, and inlet CO₂ concentration—as well as product requirements such as target purity and recovery. Under these constraints, engineers select the module type, define whether a single- or multi-stage system is needed, determine stage arrangements, and assess the role of recycle streams [182,347]. For CO₂ capture, single-stage systems are often inadequate, especially for dilute feeds or when high purity is required. Consequently, multi-stage configurations are widely applied. Although three- or four-stage systems can meet stringent specifications, two-stage processes are generally preferred due to lower capital costs and simpler operation. Cascade arrangements with partial recycle streams are commonly used to improve CO₂ recovery and product purity [348]. Various two-stage cascade designs and recycle strategies have been systematically compared in the literature [349]. A key advantage of multi-stage systems is the flexibility to use different membranes at different stages [350]. For example, highly permeable membranes may be placed in the first stage to maximize throughput, while more selective membranes can be used downstream to refine product purity. This staged optimization allows better control of system performance. Process simulations conducted by several research groups [350,351,352,353,354] highlights the particular strength of membranes in decentralized or modular carbon capture applications, while also indicating that further improvements in membrane performance and process integration are needed to enhance competitiveness at large industrial scales.

7. Industrial Applications of Membrane-Based CO₂ Separation

Membrane technology has been applied in industrial gas separation since the 1970s, including oxygen enrichment, hydrogen recovery, and CO₂ removal from natural gas. Despite long development, membranes historically occupy a small market share compared with absorption technologies, and large-scale post-combustion CO₂ capture remains unrealized. Recently, membrane-based CO₂ separation has gained attention due to industrial acceptance, CCS needs, and inherent advantages such as compactness, low energy demand, simplicity, and modularity, alongside advances in materials and fabrication. Although thousands of membranes surpass Robeson upper bounds (2008) for CO₂ separation, few have progressed beyond laboratory scale due to complex synthesis, limited scalability, and challenges in producing large-area asymmetric or thin-film composite membranes. Most studies still use small membranes under ideal conditions. This section focuses on membranes with commercial or pilot-scale evaluation for CO₂ separation in CO₂/CH₄, CO₂/N₂, and CO₂/H₂ applications, where requirements differ substantially (Table 12, Figure 5).

7.1. CO₂/CH₄ Separation

7.1.1. Natural Gas Sweetening

The largest industrial application, though membranes only cover approximately ~5% of the market, dominated by amine absorption [58]. Key commercial polymers include cellulose acetate (CA) with CO₂ permeability ~10 Barrer and CO₂/CH₄ selectivity ~40 (single-gas) or ~20 (mixed-gas) [168], Matrimid (selectivity 30–60) [355], and polysulfone (CO₂ permeability 20–40 Barrer, selectivity 15–35) [356]. Performance varies with selective layer thickness and module design, but these membranes offer acceptable stability, partial plasticization resistance, and low cost [357]. More than 20 large-scale facilities are operational worldwide [60,358]. Plasticization by CO₂ and CH₄ under high pressure reduces selectivity and increases CH₄ loss. Mitigation strategies include chemical/thermal crosslinking [359,360,361], thermal annealing [362,363], nanofiller incorporation [364,365], and polymer blending [366]. Treated membranes can operate stably at 30–50 bar [367,368].

7.1.2. Biogas Upgrading

Biogas contains higher CO₂ fractions at smaller plant scales, making membranes attractive due to lower capital and operating costs, energy efficiency, and compact design [369]. Polymeric membranes (CA, PSf, PEI, PI) dominate, while inorganic membranes (CMS, zeolite, silica) remain under investigation [370]. Pilot testing includes CMS hollow fiber membranes (~2 m²) at ~20 Nm³ h⁻¹]. PI membranes show moderate CO₂/CH₄ selectivity (~33), while CMS membranes offer high selectivity (~246) with lower CO₂ permeance (7.75 GPU, P_CO2 = 155 Barrer), improving energy efficiency [371]. Multi-stage systems, typically two-stage, are used to achieve high methane recovery and purity. Compression of raw, near-atmospheric pressure biogas is a major cost factor [369,370].

7.2. CO₂/N₂ Separation

Post-combustion CO₂ capture from flue gas is challenging due to low CO₂ partial pressure, large volumetric flow, and high humidity. Membranes with CO₂ permeance > 500 GPU, P_CO2 > 100 Barrer and CO₂/N₂ selectivity > 40 could compete with MEA absorption [372,373,374]. Despite many materials surpassing Robeson upper bounds, few have been made into thin-film composites with adequate CO₂ flux, and even fewer reached pilot scale [71,375]. Facilitated transport membranes (FTMs) represent the most advanced technology. NTNU has conducted multiple pilot and pre-pilot demonstrations since 2011 at coal-fired power plants, cement factories, and industrial facilities in Europe [376,377,378]. Initially using flat-sheet modules, NTNU transitioned to hollow fibers for higher packing density, demonstrating stable operation up to six months with resistance to SO_x and NO_x. Their second-generation FTM is now commercialized by Aqualung Carbon Capture AS [379]. Spiral-wound FTM modules from Ohio State University achieved CO₂ permeance of 800 GPU, P_CO2 = 128 Barrer and CO₂/N₂ selectivity of 150 at 57 °C [380,381]. Tianjin University developed a pilot-scale membrane facility (11,000 m² y⁻¹) and demonstrated a three-stage spiral-wound system reaching CO₂ purity of 96.2% and capture efficiency of 81.3% [382,383]. Challenges remain in water management, module packing density, and long-term stability. Pilot studies using polymeric membranes (PES, PolyActive, GO hybrids, Polaris by MTR) show promise but highlight sensitivity to flue gas impurities and humidity [373,374,384,385,386,387].

7.3. CO₂/H₂ Separation

CO₂/H₂ separation is associated with pre-combustion CO₂ capture from syngas and biohydrogen production. H₂ is small and highly diffusive, while CO₂ is highly soluble, allowing membranes to be either H₂- or CO₂-selective [266,388].

CO₂-Selective Membranes

Dense polymeric membranes with high CO₂ affinity, such as PEO-based systems, generally exhibit CO₂/H₂ selectivity < 30. Pilot evaluations showed insufficient performance under realistic conditions [389]. Facilitated transport membranes significantly improve separation due to water vapor in syngas [390]. For example, sterically hindered amine carriers achieved CO₂/H₂ selectivity up to 300 with CO₂ permeability > 3000 Barrer. Dual-phase ceramic–carbonate membranes also show promise at high temperatures, though scalability remains a challenge [391].

Table 12. Summary of industrially relevant membrane technologies for CO₂ separation.

Application	Membrane Type	Representative Materials/ Systems	Typical Performance Range	Key Advantages	Main Limitations	Ref.
CO2/CH4 (Natural gas sweetening)	Polymer	Cellulose acetate (CA), PSf, Matrimid	PCO2~10–40 Barrer; αCO2/CH4: 15–60	Low cost; mature technology; good long-term stability; commercial availability	Lower permeability than advanced materials; plasticization at high pressure	[58,168,355,356,357,358]
	Polymer (modified)	Crosslinked, blended, nanofiller-enhanced polymers	Stable operation at 30–50 bar; improved plasticization resistance	Improved durability; compatible with existing modules	Still below upper-bound lab-scale materials	[359,360,361,362,363,364,365,366,367,368]
CO2/CH4 (Biogas upgrading)	Polymer	CA, PSf, PI, PEI	CO2 permeance: ~10–20 GPU; PCO2~10–40 Barrer αCO2/CH4: ~30–40	Cost-effective for small/medium scale; compact systems	Multistage systems required; compression cost dominates	[369,370]
	Inorganic	CMS hollow fibers	αCO2/CH4: up to ~246; permeance ~7.75 GPU PCO2 = 155 Barrer	Very high selectivity; reduced CH4 loss	Lower permeance; scalability challenges	[371]
CO2/N2 (Post- combustion)	Polymer	PES, PolyActive, Polaris (MTR), GO hybrids	CO2 permeance: ~1000–1700 GPU; PCO2 = 100–170 Barrer αCO2/N2: 50–73	Scalable thin-film composites; good manufacturability	Sensitive to impurities and humidity	[373,374,384,385,386,387]
	Facilitated transport (FTM)	Fixed/mobile amine carriers; Pebax–amine systems	CO2 permeance: 500–1500 GPU PCO2 = 50–100 Barrer; αCO2/N2 100–200	Excellent low-pressure performance; water-compatible; high selectivity; pilot validated	Water management; carrier degradation; scale-up complexity	[378,379,380,381,382,383]
	Hybrid	GO-based, mixed-matrix membranes	CO2 permeance up to ~1000 GPU PCO2 up to 110 Barrer αCO2/N2 up to 80	Tunable nanochannels; potential high flux	Defect control at large scale remains challenging	[71,375,384]
CO2/H2 (Pre- combustion)	CO2-selective polymer	PEO-based membranes; FTMs	αCO2/H2 < 30 (PEO); up to ~300 (FTM); PCO2 > 3000 Barrer (FTM)	Effective under humid syngas conditions; strong facilitated transport effect	Limited thermal stability (polymeric systems)	[389,390]
	H2-selective polymer	PBI	H2 permeance up to ~100 GPU PH2 up to 35 Barrer αH2/CO2 = 18–40	High-temperature operation; good thermal stability	Moderate permeability	[266,388]
	Inorganic	ceramic–carbonate membranes	high flux at elevated T	Exceptional purity; high-temperature and pressure tolerance	High cost; poisoning sensitivity; scalability issues	[391]

Table 12. Summary of industrially relevant membrane technologies for CO₂ separation.

Application	Membrane Type	Representative Materials/ Systems	Typical Performance Range	Key Advantages	Main Limitations	Ref.
CO2/CH4 (Natural gas sweetening)	Polymer	Cellulose acetate (CA), PSf, Matrimid	PCO2~10–40 Barrer; αCO2/CH4: 15–60	Low cost; mature technology; good long-term stability; commercial availability	Lower permeability than advanced materials; plasticization at high pressure	[58,168,355,356,357,358]
	Polymer (modified)	Crosslinked, blended, nanofiller-enhanced polymers	Stable operation at 30–50 bar; improved plasticization resistance	Improved durability; compatible with existing modules	Still below upper-bound lab-scale materials	[359,360,361,362,363,364,365,366,367,368]
CO2/CH4 (Biogas upgrading)	Polymer	CA, PSf, PI, PEI	CO2 permeance: ~10–20 GPU; PCO2~10–40 Barrer αCO2/CH4: ~30–40	Cost-effective for small/medium scale; compact systems	Multistage systems required; compression cost dominates	[369,370]
	Inorganic	CMS hollow fibers	αCO2/CH4: up to ~246; permeance ~7.75 GPU PCO2 = 155 Barrer	Very high selectivity; reduced CH4 loss	Lower permeance; scalability challenges	[371]
CO2/N2 (Post- combustion)	Polymer	PES, PolyActive, Polaris (MTR), GO hybrids	CO2 permeance: ~1000–1700 GPU; PCO2 = 100–170 Barrer αCO2/N2: 50–73	Scalable thin-film composites; good manufacturability	Sensitive to impurities and humidity	[373,374,384,385,386,387]
	Facilitated transport (FTM)	Fixed/mobile amine carriers; Pebax–amine systems	CO2 permeance: 500–1500 GPU PCO2 = 50–100 Barrer; αCO2/N2 100–200	Excellent low-pressure performance; water-compatible; high selectivity; pilot validated	Water management; carrier degradation; scale-up complexity	[378,379,380,381,382,383]
	Hybrid	GO-based, mixed-matrix membranes	CO2 permeance up to ~1000 GPU PCO2 up to 110 Barrer αCO2/N2 up to 80	Tunable nanochannels; potential high flux	Defect control at large scale remains challenging	[71,375,384]
CO2/H2 (Pre- combustion)	CO2-selective polymer	PEO-based membranes; FTMs	αCO2/H2 < 30 (PEO); up to ~300 (FTM); PCO2 > 3000 Barrer (FTM)	Effective under humid syngas conditions; strong facilitated transport effect	Limited thermal stability (polymeric systems)	[389,390]
	H2-selective polymer	PBI	H2 permeance up to ~100 GPU PH2 up to 35 Barrer αH2/CO2 = 18–40	High-temperature operation; good thermal stability	Moderate permeability	[266,388]
	Inorganic	ceramic–carbonate membranes	high flux at elevated T	Exceptional purity; high-temperature and pressure tolerance	High cost; poisoning sensitivity; scalability issues	[391]

8. Big Data and Machine Learning for CO₂ Separation Membranes

The integration of big data analytics, machine learning (ML), and molecular simulation (MS) has profoundly reshaped membrane science by enabling accelerated prediction of transport properties and supporting the rational design of advanced CO₂ separation materials [392,393]. In particular, the coupling of ML with MS has emerged as a powerful strategy for exploring vast material spaces, identifying promising polymeric and porous membrane candidates, and guiding membrane and process design at unprecedented speed. Recent comprehensive reviews confirm that ML-based approaches are increasingly influential across gas separation, energy-related membranes, and water treatment technologies; however, they also highlight that substantial conceptual and practical barriers still limit their broader industrial adoption [394,395].

Although ML–MS integration enables rapid screening of large virtual material libraries, only a limited number of ML-predicted membranes have progressed beyond laboratory-scale validation. This persistent gap between computational promise and experimental realization underscores the necessity of critically examining the fundamental bottlenecks associated with ML-assisted membrane design. As observed in recent studies, most ML-driven discoveries remain confined to intrinsic material property prediction, while factors governing long-term stability, manufacturability, and scalability are rarely incorporated into data-driven workflows [395,396]. Polymeric membranes continue to dominate industrial gas separation technologies, particularly for CO₂ capture and upgrading. Consequently, ML has been widely applied to improve polymer CO₂ permeability and selectivity. Jason et al. [397] employed deep neural networks and random forest models trained on chemical descriptors to predict gas permeability across millions of hypothetical polymers, identifying candidates that exceeded established Robeson upper bounds for CO₂/CH₄ separation. Complementary molecular dynamics simulations clarified the structural features governing gas transport, including free volume distribution and chain packing. Similarly, at Caltech, Basdogan et al. [398] combined ML prediction with genetic algorithm–guided inverse design to screen more than 16,000 polymer candidates, such as imines and polyethers, for CO₂ separation applications. These studies illustrate how ML–MS coupling can dramatically accelerate polymer discovery while simultaneously providing insight into structure–property relationships. Beyond material discovery, ML has also been integrated with quantum chemical calculations and process-level modeling to optimize membrane-based CO₂ separation systems. Gasos et al. [399] developed a multi-objective artificial neural network (ANN) framework that combined density functional theory (DFT), ML, and process simulations to rapidly optimize CO₂ separation performance, balancing productivity and energy consumption within milliseconds. Such approaches demonstrate the potential of ML to bridge molecular design and process optimization, enabling holistic and time-efficient membrane development. In parallel, ML–MS approaches have expanded rapidly into the design of metal–organic frameworks (MOFs), which offer high porosity, tunable chemistry, and diverse topologies [150]. ML-assisted molecular simulations have been used extensively to evaluate gas adsorption and separation performance across large MOF databases [400,401]. Budhathoki et al. [402] carried out high-throughput screening of more than 100,000 MOFs and integrated polymer permeability data to identify millions of potential mixed-matrix membrane (MMM) combinations. Importantly, their work incorporated economic assessments, enabling selection of MMM candidates that balance membrane performance with practical feasibility. Cheng et al. [403] further advanced multi-scale modeling strategies by combining MS, ML, and process simulations for MOF membranes. Using IRMOF-1 as a model system, CO₂ adsorption and permeability data were incorporated into an ANN-based tank model for hollow-fiber membrane processes, allowing dynamic prediction of permeability and selectivity under varying operating conditions and supporting higher recovery with reduced membrane area requirements.

Despite these successes, significant limitations continue to constrain the broader impact of ML in membrane material design. One of the most critical challenges remains the availability, quality, and consistency of experimental data used to train ML models. The predictive performance of ML algorithms is strongly limited by the scarcity of high-quality, standardized datasets for gas permeability, diffusivity, and selectivity, particularly for technologically relevant gas pairs such as CO₂/CH₄ and CO₂/N₂ [396,404]. Experimental data reported in the literature are often obtained under disparate testing conditions, including variations in temperature, pressure, membrane thickness, and feed composition. Moreover, polymeric membranes are subject to aging, plasticization, and physical relaxation, introducing temporal variability and domain shifts between training datasets and real operating environments. As emphasized in recent ML-focused membrane reviews, such data heterogeneity undermines reproducibility and severely limits model transferability [395]. Another fundamental bottleneck lies in the selection and representation of molecular descriptors. Accurate ML predictions critically depend on how membrane materials are encoded numerically, yet no consensus exists regarding descriptor sets capable of capturing the multiscale nature of gas transport in membranes. Studies on polymeric and nanocomposite membranes demonstrate strong sensitivity of ML outcomes to descriptor choice, including physicochemical features, topological fingerprints, and graph-based representations [405]. However, many commonly used descriptors fail to explicitly represent key physical phenomena such as segmental mobility, free volume connectivity, and specific polymer–gas interactions, which govern membrane transport mechanisms. This disconnect between descriptor design and physical reality represents a core limitation of current ML-driven membrane strategies [396]. The limited interpretability of many ML models further restricts their practical utility. While deep neural networks and ensemble learning methods often provide superior predictive accuracy, they are frequently perceived as black boxes, offering limited mechanistic insight into structure–property relationships. This lack of transparency poses a significant barrier to industrial adoption, where understanding failure modes, ensuring reliability, and satisfying regulatory requirements are essential. Consequently, recent studies increasingly advocate for the integration of explainable artificial intelligence (XAI) approaches to identify dominant structural features, reveal transport-controlling factors, and improve trust in ML predictions [404]. Nevertheless, systematic application of XAI in membrane-focused ML workflows remains at an early stage.

Additional challenges arise when translating ML predictions to MMMs and industrial membrane configurations. Although ML screening has revealed promising MOF–polymer combinations and clarified the relative advantages of MOF-, polymer-, and COF-based membranes for CO₂ separation [395,406,407], critical issues such as filler dispersion, interfacial compatibility, mechanical robustness, and long-term membrane integrity are rarely encoded in training datasets. Similarly, process-level constraints—including module design, pressure drop, mixed-gas performance, impurity tolerance, and fabrication cost—are often neglected, limiting the industrial relevance of purely material-focused ML models [394,396,408]. Looking forward, converging evidence suggests that the most effective pathway toward industrially relevant ML-driven membrane design lies in holistic, multi-scale integration of ML, MS, and process modeling. Emerging strategies, including physics-informed ML, inverse design frameworks, and generative models such as variational autoencoders and generative adversarial networks, offer promising routes to bridge data-driven prediction with physical understanding [409]. When combined with process simulation and techno-economic analysis, ML–MS approaches can support the rational design of cost-effective, high-performance membranes and accelerate translation from computational discovery to industrial deployment [395,408]. Collectively, recent literature converges on the conclusion that ML is not merely an auxiliary tool, but a foundational component of next-generation membrane design frameworks—provided that data quality, interpretability, and process-level integration are properly addressed.

9. Summary and Future Development Directions

To sum up, it should be stated that MMM technology offers a path to high-performance CO₂ separation. By integrating advanced filler design, functionalization, and module optimization, it is feasible to achieve simultaneously high permeability and selectivity, bridging laboratory breakthroughs with industrial application. The experimental data collected from the previous chapters along with the advantages, disadvantages and future directions of the hybrid membranes under consideration are presented in Figure 6 and

Table 13. Characteristics and summary of considered hybrid membranes for CO₂ separation.

Filler	Thickness (µm)	α* CO2/N2	α* CO2/CH4	PCO2 (Barrer)	Advantages	Limitations	Future Directions
COFs	0.2–50	61–91	19–24	234–1044	Well-defined nanochannels; tunable chemistry; good CO2 affinity; fully organic	Large intrinsic pore size (>1 nm) can reduce selectivity; defect formation; complex fabrication	Pore size tuning < 1 nm; polymer functionalization; scalable defect-free membranes
MOFs	1–10	11–78	up to 92.6	11–5413	High surface area; molecular sieving; selective adsorption; tunable metal centers	Moisture sensitivity; particle aggregation; moderate stability	Amine-functionalized MOFs; core–shell MOF@COF hybrids; large-scale MMM fabrication
Graphene Oxide (GO)	40–70	24–104	15–41	27–5235	Layered structure; tunable interlayer spacing; good polymer adhesion	Aggregation; permeability–selectivity trade-off; water sensitivity	GO/COF or GO/polymer hybrids; crosslinking; aligned laminates
TMDs (e.g., MoS2)	0.4–50	29–153	39–69	19–472	High CO2 affinity via amine functionalization; 2D transport channels	Limited scalability; potential restacking; moderate mechanical strength	Cys-MoS2 nanosheets; functionalization to improve selectivity and adhesion
CNTs	40–55	22–81	16–85	5–742	High aspect ratio; fast transport channels; mechanical reinforcement	Aggregation; poor polymer compatibility; limited size control	Functionalized CNTs (–COOH, –NH2); aligned CNT membranes; mixed 2D/1D MMM
g-C3N4	0.8–179	20–84	12–48	6–3740	High thermal and chemical stability; tunable porosity; 2D structure	Lower permeability; aggregation; limited CO2 selectivity	Exfoliated nanosheets; surface functionalization; hybridization with MOFs/COFs
LDH (Layered Double Hydroxides)	2–2.5	Up to 71	Up to 32	11–1307	Tunable layer spacing; anion exchangeable; good thermal stability	Moderate CO2 permeability; poor long-term stability; aggregation	Interlayer modification; combination with polymers/COFs; exfoliated nanosheets
Zeolites	50–160	33–108	19–169	2–887	Molecular sieving; thermal stability; established industrial use	Low permeability; sensitivity to fouling; difficulty in thin-film fabrication	Nanozeolites; hierarchical porosity; zeolite@polymer hybrids
Metal Oxides (e.g., TiO2, Al2O3)	25–300	15–74	22–74	4–15,200	High chemical/thermal stability; robust; easy synthesis	Low intrinsic selectivity; aggregation; poor polymer compatibility	Surface functionalization; mixed filler approaches; thin-film integration
Magnetic Nanoparticles	30–170	58–75	3–47	59–538	Facilitated transport via magnetic alignment; potential for external control; reinforcement	Complex synthesis; limited long-term stability; aggregation	Functionalized Fe3O4; field-aligned MMM; hybrid 2D/NP systems
MXenes (e.g., Ti3C2Tx)	0.2–70	96–319	22–249	23–126	2D layered structure; high surface area; surface terminations for CO2 interaction; improved polymer adhesion	Aggregation at high loadings; oxidation; limited scalability	Surface functionalization; MXene/COF hybrids; aligned nanosheets for directional transport

.

Figure 6a,b illustrate the relationship between CO₂/N₂ and CO₂/CH₄ selectivity and CO₂ permeability for mixed-matrix membranes (MMMs) incorporating various filler particles. The data are plotted relative to the Robeson upper bound (2008), which defines the permeability–selectivity trade-off limit for polymer-based membranes. Most systems follow the classical inverse relationship between permeability and selectivity. As expected, α_CO2/CH4 values are generally lower than α_CO2/N2 at comparable permeability due to the closer kinetic diameters and competitive sorption behavior of CO₂ and CH₄. However, several hybrid membranes clearly exceed the Robeson upper bound line, particularly in the intermediate permeability range. In this range (P_CO2 ≈ 50–800 Barrer), multiple data points α_CO2/N2 are located above the Robeson line (Figure 6a), indicating simultaneous enhancement of permeability and selectivity. Quantitatively, these systems exhibit selectivity values approximately 20–80% higher than those predicted by the Robeson correlation at equivalent permeability levels.

The most pronounced exceedance is observed in hybrid systems incorporating two-dimensional nanofillers (e.g., graphene-derived materials (GO) or MXene-type structures), one-dimensional conductive fillers (e.g., CNT-based systems) and selected nanoparticles, magnetic particles and functionalized MOFs. For instance, Cys-MoS₂/Pebax (TMD-based MMM) exhibits α_CO2/N2 ≈ 120 at P_CO2 ≈ 297 Barrer, surpassing the upper bound due to amine-functionalized MoS₂ nanosheets that enhance CO₂ affinity and interfacial compatibility. GO/COF layered membranes and PVAm-functionalized COF MMMs demonstrate high selectivity driven by tailored nanochannels and strong CO₂–amine interactions. Some selected MXene-based MMMs (e.g., supported IL membrane) achieve α_CO2/N2 values above 100 at moderate permeability, benefiting from stabilized lamellar channels. These systems outperform conventional polymer membranes because of synergistic effects combining selective adsorption, facilitated transport mechanisms, and controlled nanochannel formation. By contrast, membranes with very high permeability (>1000 Barrer), such as some fluoropolymer- or free-volume-enhanced systems, tend to fall below the upper bound due to reduced size-sieving control. At higher permeability values (P_CO2 > 1000 Barrer), membranes containing MOFs, some zeolite-based systems, and high-loading oxide nanoparticles fall significantly below the Robeson line, with selectivity reductions exceeding 30–50%. This suggests that excessive free-volume generation or interfacial void formation dominates over selective transport mechanisms. For the CO₂/CH₄ gas pair, hybrid membranes can be grouped into three main performance regimes relative to the Robeson upper bound. In the range P_CO2 ≈ 80–800 Barrer could be observed membranes containing GO, CNTs, g-C₃N₄, LDH, COFs, and TMDs which typically operate close to the upper bound, with α_CO2/CH4 values within ~10–25% of the predicted limit. These systems provide balanced improvements in CO₂ sorption and diffusion without severe selectivity loss, effectively functioning at the edge of the classical solution–diffusion trade-off. Oxide nanoparticles, magnetic particles, functionalized zeolites and selected optimized GO/CNT systems exceed the upper bound at moderate permeability, often by ~20–60%. Their enhanced performance results from synergistic effects combining selective adsorption, strong CO₂–surface interactions, and well-defined nanochannels that suppress CH₄ transport. MOF-, zeolite-, and CNT-based membranes frequently fall below the upper bound (P_CO2 > 1000 Barrer), with selectivity reductions of 30–50%. Excessive free volume or interfacial voids dominate over selective transport, leading to permeability gains at the expense of discrimination.

Hybrid and mixed-matrix membranes (MMMs) have demonstrated clear potential to overcome the intrinsic permeability–selectivity trade-off that limits conventional polymeric membranes, particularly in the context of CO₂ separation for energy transition and CCUS applications. Experimental data confirm that modern MMMs incorporating advanced fillers such as COFs, MOFs, GO, TMDs, CNTs, g-C₃N₄, LDHs, zeolites, metal oxides, MXenes, and magnetic nanoparticles consistently outperform traditional polymers, which typically exhibit α_CO2/N2 ≈ 20–40 and P_CO2 ≈ 10–50 Barrer. Based on reported data, COF-based MMMs reach α_CO2/N2 ≈ 61–91 and α_CO2/CH4 = 19–24 at P_CO2 ≈ 234–1044 Barrer, while MOF-based systems show α_CO2/N2 ≈ 11–78 and α_CO2/CH4 up to 92.6 across a wide permeability span of 11–5413 Barrer. Two-dimensional fillers such as GO and MXenes further enhance directional transport, delivering α_CO2/N2 = 24–104 and α_CO2/CH4 = 15–41 at P_CO2 ≈ 27–5235 Barrer (GO) and α_CO2/N2 ≈ 96–319 and α_CO2/CH4 = 22–249 at moderate permeability P_CO2 ≈ 23–126 Barrer (MXenes). In selected multifunctional or hybrid systems, such as COF@MOF, MXene@COF, or PEG-functionalized hollow COFs, selectivity can exceed α_CO2/N2 > 150–200 at moderate permeability levels of 300–600 Barrer, clearly surpassing the Robeson upper bound. CNT-based membranes show α_CO2/N2 ≈ 22–81 and α_CO2/CH4 = 16–85 at P_CO2 ≈ 5–742 Barrer, whereas g-C₃N₄ systems combine moderate selectivity (α_CO2/N2 ≈ 20–84 and α_CO2/CH4 = 12–48) with a very broad permeability window (P_CO2 ≈ 6–3740 Barrer). LDHs-based membranes achieve α_CO2/N2 up to 71.7 and α_CO2/CH4 up to 31.6 at P_CO2 ≈ 11–1307 Barrer, while zeolite-based MMMs show α_CO2/N2 = 25–108 and α_CO2/CH4 = 19–169 at P_CO2 ≈ 9–887 Barrer. TMDs-based membranes reach α_CO2/N2 ≈ 29–153 and α_CO2/CH4 = 39–69 at P_CO2 ≈ 19–472 Barrer, metal oxide-based MMMs show α_CO2/N2 ≈ 15–74 and α_CO2/CH4 = 22–74 at P_CO2 ≈ 4–15,200 Barrer, and finally magnetic nanoparticle-based membranes achieve α_CO2/N2 ≈ 58–75 and α_CO2/CH4 = 3.4–47 at P_CO2 ≈ 59–538 Barrer. Despite these impressive laboratory-scale performances, several recurring limitations remain. Many crystalline fillers possess intrinsic pore sizes larger than 1 nm, which may reduce size-sieving precision unless fine-tuned below 1 nm. Interfacial defects, particle agglomeration, restacking of 2D nanosheets, and limited compatibility with polymer matrices frequently reduce the effective separation performance compared to theoretical predictions. Additionally, moisture sensitivity (particularly in MOFs and LDHs), oxidation susceptibility (notably in MXenes), and synthesis complexity (e.g., TMDs or core–shell hybrids) pose challenges for long-term stability and industrial scalability.

Experimental evidence indicates that the best overall MMM performance typically occurs at intermediate permeability values (P_CO2 ≈ 50–800 Barrer) combined with high selectivity (α_CO2/N2 > 100 and α_CO2/CH4 > 50). Excessively high filler loadings often lead to aggregation and defect formation, while insufficient interfacial engineering limits synergistic effects. Therefore, future development must prioritize precise pore engineering below 1 nm to optimize molecular sieving while maintaining fast diffusion pathways. Surface functionalization with amine or PEG groups has already demonstrated two- to three-fold improvements in selectivity compared to unmodified systems, confirming that interfacial chemistry is as critical as intrinsic filler porosity.

The alignment of two-dimensional nanosheets, including GO, MXene, g-C₃N₄, and TMDs, represents another decisive strategy. Controlled orientation—potentially assisted by magnetic or electric fields—can improve permeability–selectivity trade-offs by 20–40% compared to randomly dispersed fillers. Magnetic nanoparticles and magnetically functionalized nanosheets offer additional opportunities for field-assisted alignment, reduced aggregation, and even dynamic or switchable membrane behavior. For example, Cys-MoS₂ nanosheets in Pebax matrices demonstrate α_CO2/N2 ≈ 120 at P_CO2 ≈ 297 Barrer, exceeding the Robeson limit due to enhanced CO₂–amine interactions and improved interfacial compatibility.

Scalability and long-term stability remain critical for industrial implementation. Solvent-free in situ growth, vacuum-assisted assembly, controlled crosslinking, and layer-by-layer lamination techniques have enabled the fabrication of defect-free membranes thinner than 1 µm while preserving mechanical integrity. At the module level, hollow fiber configurations—with packing densities 3–4 times higher than spiral-wound modules—are particularly attractive for industrial deployment. Optimized hollow fiber systems have demonstrated CO₂ recovery rates above 80% with stable performance over six months in pilot tests, indicating technological readiness beyond laboratory scale.

The integration of multifunctional mechanisms within a single MMM—combining adsorption, molecular sieving, and facilitated transport—consistently outperforms single-mechanism membranes. Facilitated transport membranes (FTMs) already achieve P_CO2 ≈ 50–100 Barrer with α_CO2/N2 ≈ 100–200, while sterically hindered amine systems report α_CO2/H2 ≈ 300 at P_H2 > 3000 Barrer. Incorporating such facilitated transport principles into structurally controlled MMM architecture represents a promising pathway to next-generation membranes. A critical aspect in the evaluation of mixed-matrix membranes (MMMs) is the distinction between permeability and permeance, particularly in relation to membrane thickness. While permeability is an intrinsic material property, permeance determines the actual process performance and is inversely proportional to membrane thickness. The data summarized in

Table 14. Comparison of characteristic parameters of hybrid membranes used for permeation and separation of pure gases and gas mixtures.

Membrane Type	L (µm)	Mixture	Selectivity α*	Separation Factor α	Permeability (Barrer)	Permeability Gas Mixture (Barrer)	Ref.
GO/PIM-1	40–50	-	CO2/CH4 = 14.9	-	PCO2 = 5235 PCH4 = 359	-	[210]
PGO/Pebax	55	CO2:N2 = 10:90	CO2/N2 = 80.7	CO2/N2 = 87	PCO2 = 232.7	PCO2 = 1150	[213]
MWCNT/Pebax	-	-	CO2/N2 = 70 CO2/CH4 = 35	-	PCO2 = 567	-	[218]
ZIF-8@GO/Pebax			CO2/N2 = 77.9		PCO2 = 136.2		[238]
ZIF-8 within GO interlayers	0.9	H2:CO2 = 50:50	H2/CO2 = 30.8	H2/CO2 = 25	PH2 = 365.4 PCO2 = 11.9	PH2 = 1150 PCO2 = 46	[239]
NG/PEO	1.8		H2/CO2 = 25	H2/CO2 = 13	PH2 = 32,240 PCO2 = 1289.6	PH2 = 13,110 PCO2 = 1008	[240]
ZIF-8 +[Emim][Ac] HKUST-1/Chitosan			CO2/N2 = 11.5 CO2/N2 = 19.3		PCO2 = 5413 PCO2 = 4754		[171]
UiO-66/UiO-66–NH2 (60–80 nm)	3.5	CO2:N2 = 50:50	CO2/N2 = 31.3	CO2/N2 = 21.4	PCO2 = 386	PCO2 = 245	[241]
ZIF-8/Polyimide P84			CO2/CH4 = 92.6		PCO2 = 10.92		[173]
UiO-66–NH2/Pebax 2533	5–7		CO2/N2 = 37		PCO2 = 140.4		[184]
Ti3C2Tx/Pebax1657	60–70	CO2:N2 = 30:70	CO2/N2 = 42	CO2/N2 = 31	PCO2 = 139	PCO2 = 95	[263]
NH2-MIL-125 hybrid membrane (MXene-derived Ti source)	1		H2/CO2 = 8.0	-	PCO2 = 186.5 PH2 = 1492	-	[264]
Supported ionic liquid membrane (Ti3C2Tx + ChCl/EG)	2	-	CO2/N2 = 319.15, CO2/CH4 = 249.01, CO2/H2 = 12.38	-	PCO2 = 52.7	-	[265]
Self-crosslinked MXene Ti3CTx hollow fiber membrane	0.22	CO2:H2 = 10: 90	H2/CO2 = 30.3	H2/CO2 = 16	PH2 = 15.53 PCO2 = 0.51	PH2 = 3.74 PCO2 = 0.23	[266]
Pd2+-intercalated MXene membrane	0.78	CO2:H2 = 50: 50	H2/CO2 = 356	H2/CO2 = 242	PH2 = 1108 PCO2 = 3.11	PH2 = 620 PCO2 = 2.56	[268]
Layer-by-layer membrane (Chitosan–g-C3N4/ZIF-8 on PES)	0.8	CO2:CH4 = 50:50	CO2/CH4 = 24.2	CO2/CH4 = 17.8	PCO2 = 63.5	PCO2 = 43.7	[289]
g-C3N4/Pebax MMM	179	CO2:N2 = 50:50	CO2/N2 = 67.2	CO2/N2 = 50	PCO2 = 5900	PCO2 = 4600	[290]
g-C3N4/GO composite membrane	0.7	-	H2/CO2 = 39.2	H2/CO2 = 24	PH2 = 451 PCO2 = 11.5	PH2 = 543 PCO2 = 22.6	[291]
ZIF-90@g-C3N4 hybrid membrane	50–70	-	CO2/N2 = 84.4	-	PCO2 = 110.5	-	[292]
Functionalized g-C3N4/PIM-1 MMM	50	-	CO2/N2 = 19.8 CO2/CH4 = 12.4	-	PCO2 = 3740	-	[293]
g-C3N4-based SILM	0.8	CO2:N2 = 50:5 CO2:CH4 = 50:50		CO2/N2 = 52.49 CO2/CH4 = 48.41		PCO2 = 794 PCO2 = 928	[294]
MgO-Ag+/Matrimid	50		CO2/CH4 = 42.3		PCO2 = 4.31		[273]
TiO2/PVA	25–35		CO2/N2 = 74.3		PCO2 = 5.26		[275]
ZnO/Pebax	50–60		CO2/CH4 = 23.9		PCO2 = 149.81		[276]
Amino SiO2/6FDA-DABA	30–99		CO2/CH4 = 23		PCO2 = 1920		[280]
PEG–POSS/PMHS	200–300	CO2:N2 = 32:68	CO2/CH4 = 38.1	CO2/CH4 = 23.4	PCO2 = 679	PCO2 = 506	[283]
poly–POSS/PI	0.2		H2/CO2 = 7.6	H2/CO2 = 2.5	PCO2 = 69 PH2 = 522	PCO2 = 144 PH2 = 360	[284]
ZnAl–NO3 LDH membrane on porous alumina	2.5	H2:CO2 = 50:50		H2/CO2 = 5.8		PH2 = 267.5 PCO2 = 46.1	[296]
ZIF-8@LDH/Pebax MMM	2	CO2:CH4 = 50:50		CO2/CH4 = 31.6		PCO2 = 1307	[298]
COF-LZUI/CoAl-LDH membrane	2	H2:CO2 = 50:50		H2/CO2 = 31.6		PH2 = 7200 PCO2 = 227.8	[299]
3D hollow CoNi-LDH/Pebax MMM		-	CO2/N2 = 71.66	CO2/N2 = 38	PCO2 = 172.62	PCO2 = 125	[300]
MoS2/Pebax-1657 MMM	5	-	CO2/N2 = 93		PCO2 = 64		[305]
WS2–IL SILM	0.4	-	CO2/N2 = 153.21 CO2/CH4 = 68.81 CO2/H2 = 13.56		PH2 = 18.9		[306]
WS2/fluoropolymer composite			CO2/N2 = 29.6 CO2/CH4 = 39.4		PCO2 = 472		[307]
Cys-MoS2/Pebax MMM	30–50		CO2/N2 = 120	CO2/N2 = 116.5	PCO2 = 297	PCO2 = 285	[308]
GO/COF layered composite membrane			H2/CO2 = 25.57		PCO2 = - PH2 = 1.067 × 10−6 mol·m−2·s−1·Pa−1		[311]
PVAm- functionalized COF MMM	0.165	CO2:H2 = 40:60	CO2/H2 = 17.2	CO2/H2 = 17.2	PCO2 = 234.6	PCO2 = 211	[312]
2D COF/PEO hybrid membrane			CO2/N2 = 61.4 CO2/CH4 = 19.8 CO2/H2 = 15.0		PCO2 = 803.9		[313]
Hollow COF microspheres/Pebax MMM			CO2/CH4 = 24		PCO2 = 1044		[314]
ZIF-8@COF (core–shell) MMM	0.96	CO2:N2 = 15:85	CO2/N2 = 91	CO2/N2 = 101	PCO2 = 288	PCO2 = 358	[315]

clearly demonstrate that many high-performing MMMs are relatively thick. For example, POP-PGO/Pebax membranes exhibit a thickness of approximately 55 µm with a CO₂ permeability of 232.7 Barrer and an ideal CO₂/N₂ selectivity of 80.7 [213]. Similarly, g-C₃N₄/Pebax MMMs with a thickness of 179 µm reach very high CO₂ permeability values (~5900 Barrer), yet their large thickness significantly limits permeance under realistic conditions [290]. Even thicker systems, such as PEG–POSS/PMHS membranes (200–300 µm), require increased thickness to avoid mechanical failure, further emphasizing the trade-off between structural integrity and transport efficiency. In contrast, ultra-thin membranes such as PVAm/COF-based MMMs (≈0.165 µm) or layer-by-layer structures (~0.8 µm) exhibit significantly enhanced permeance due to reduced diffusion path length [289]. Likewise, NG/PEO membranes with a thickness of ~1.8 µm demonstrate high gas fluxes, highlighting the advantage of thin selective layers for practical applications. These results confirm that exceeding the Robeson upper bound based on permeability alone does not necessarily translate into superior industrial performance. Therefore, membrane thickness must be considered alongside permeability, and its systematic reporting (Table 14) is essential for meaningful comparison and scale-up assessment. It should also be noted that membrane thickness plays a crucial role not only in transport performance but also in physical aging behavior. Thin membranes, particularly those with high free volume, are prone to rapid structural relaxation, leading to a significant decline in permeability over time. For instance, ultra-thin membranes such as PVAm/COF (~0.165 µm) or g-C₃N₄/GO (~0.7 µm) rely on highly confined transport pathways, which may be susceptible to structural rearrangement and performance decay. In contrast, thicker MMMs such as POP-PGO/Pebax (~55 µm) [213] or Cys-MoS₂/Pebax membranes (30–50 µm) [308] exhibit more stable performance due to restricted polymer chain mobility and filler-induced stabilization. Additionally, the presence of fillers can mitigate aging by anchoring polymer chains and preserving free volume. However, it should be emphasized that the improved stability of MMMs is partly a consequence of their larger thickness. Therefore, direct comparison between thin-film and thick MMM systems must consider both geometric and material effects. In this work, a clear distinction is made between ideal selectivity (α*) and the separation factor (α), as these parameters are frequently misused in the literature. Ideal selectivity (α*) is calculated from single-gas permeation measurements, whereas the separation factor (α) is determined under mixed-gas conditions and reflects realistic separation performance. As shown in Table 13, the separation factor (α) is generally lower than the ideal selectivity (α*), reflecting non-ideal transport phenomena. For example, ZIF-8/GO membranes show a decrease from α* = 30.8 to α ≈ 25, while NG/PEO membranes exhibit a reduction from α* ≈ 25 to α ≈ 13 due to strong CO₂ interactions and polymer plasticization effects. More pronounced discrepancies are observed in certain systems. In polyPOSS-imide membranes, H₂/CO₂ selectivity decreases from 7.6 (ideal) to 2.5 under mixed-gas conditions, corresponding to a reduction of approximately 65%. Similarly, MXene-based hollow fiber membranes (~0.22 µm) exhibit a significant drop in selectivity (from 30.3 to 16) due to preferential CO₂ adsorption and transport pathway blocking. In contrast, specific systems may exhibit comparable or even enhanced mixed-gas performance. For example, POP-PGO/Pebax membranes show an increase in CO₂ permeability (from 232.7 to 1150 Barrer) and selectivity (from 80.7 to 87) under humid conditions, suggesting the contribution of facilitated transport mechanisms [213]. These results highlight the necessity of mixed-gas measurements for realistic membrane evaluation. Finally, machine learning and molecular simulations are expected to accelerate the discovery and optimization of MMM systems. High-throughput computational screening of more than 100,000 MOF and polymer combinations has already predicted permeability–selectivity pairs exceeding α_CO2/N2 = 38–61 at P_CO2 > 7000 Barrer, guiding experimental efforts toward previously unexplored material combinations. In conclusion, the future of MMM development lies in synergistic hybrid filler design, precise pore engineering, advanced surface functionalization, controlled 2D alignment, multifunctional approaches, scalable defect-free fabrication, energy-efficient deployment, machine learning- guided design and intelligent module engineering. When combined with computational design tools and process optimization, these strategies position hybrid membranes as industrially viable, energy-efficient solutions capable of supporting large-scale carbon capture and advancing the global transition toward low-carbon energy systems.

10. Conclusions

Hybrid and mixed-matrix membranes (MMMs) have clearly demonstrated their potential as an advanced platform for efficient CO₂ separation, offering realistic pathways to overcome the permeability–selectivity trade-off inherent to conventional polymeric membranes. The collective evidence reviewed here confirms that MMMs incorporating rationally designed fillers can approach or exceed the Robeson upper bound, particularly within the technologically relevant intermediate permeability range (PCO₂ ≈ 50–800 Barrer), where high selectivity can be combined with industrially meaningful flux. Across filler classes, the most successful MMMs exploit synergistic transport mechanisms, including selective adsorption, molecular sieving, facilitated transport, and controlled nanochannel formation. Low-dimensional fillers such as graphene oxide, MXenes, g-C₃N₄, TMDs, and carbon nanotubes are especially effective in enhancing directional transport and interfacial control when dispersion and alignment are optimized. Hybrid architectures (e.g., MOF@COF or MXene-based composites) further illustrate that interfacial chemistry and structural hierarchy are more decisive for performance than filler loading alone. In contrast, membranes relying primarily on excessive free-volume generation or very high filler contents frequently suffer from aggregation, interfacial voids, and diminished selectivity despite high permeability. Beyond low-dimensional materials, other fillers play complementary but application-specific roles. MOFs and zeolites offer well-defined microporosity and strong molecular sieving, enabling high selectivity, particularly for CO₂/CH₄ separation; however, their moisture sensitivity, particle aggregation, and challenges in fabricating thin, defect-free composites limit long-term stability and scalability. Metal oxide nanoparticles (e.g., TiO₂, Al₂O₃) provide excellent chemical and thermal robustness and ease of synthesis yet typically exhibit low intrinsic selectivity and require surface functionalization to avoid non-selective interfacial diffusion. Magnetic nanoparticles introduce unique opportunities for field-assisted alignment and tunable transport, but their complex synthesis, aggregation tendency, and limited long-term stability currently restrict industrial implementation. Overall, these fillers are most effective when used in hybrid or multifunctional combinations rather than as standalone additives. A key finding of this review is that optimal MMM performance is achieved not at maximum permeability, but at balanced permeability and selectivity, supported by robust interfacial engineering. Equally important is the distinction between permeability and permeance: membrane thickness, mixed-gas performance, aging behavior, and mechanical stability ultimately determine industrial viability. High permeability alone does not translate into superior process performance if thick selective layers, plasticization, or long-term degradation limits permeance and durability. Despite substantial progress, several recurring challenges remain, including filler aggregation, restacking of two-dimensional materials, oxidation or moisture sensitivity (notably for MXenes, MOFs, and LDHs), and difficulties in scalable fabrication of ultrathin, defect-free membranes. Future research should therefore prioritize: (i) precise pore engineering below 1 nm, (ii) advanced surface functionalization to enhance polymer–filler compatibility and CO₂ affinity, (iii) controlled alignment of low-dimensional fillers to reduce transport tortuosity, and (iv) scalable fabrication routes—particularly hollow-fiber configurations—with verified long-term stability. From an application perspective, MMMs constitute a versatile and energy-efficient separation technology for CCUS-related processes, including natural gas sweetening, biogas upgrading, post-combustion capture, and syngas purification. When combined with optimized module design, process integration, and emerging data-driven approaches such as machine learning and multiscale modeling, hybrid membranes are well positioned to transition from laboratory-scale demonstrations toward reliable industrial deployment, supporting low-carbon energy systems and the global energy transition.