Recent Advances in Molten-Carbonate Membranes for Carbon Dioxide Separation: Focus on Material Selection, Geometry, and Surface Modification

Membranes for carbon dioxide permeation have been recognized as potential candidates for CO2 separation technology, particularly in the energy sector. Supported molten-salt membranes provide ionic routes to facilitate carbon dioxide transport across the membrane, permit the use of membrane at higher temperature, and offer selectivity based on ionic affinity of targeted compound. In this review, molten-carbonate ceramic membranes have been evaluated for CO2 separation. Various research studies regarding mechanisms of permeation, properties of molten salt, significance of material selection, geometry of support materials, and surface modifications have been assessed with reference to membrane stabilities and operational flux rates. In addition, the outcomes of permeation experiments, stability tests, selection of the compatible materials, and the role of interfacial reactions for membrane degradation have also been discussed. At the end, major challenges and possible solutions are highlighted along with future recommendations for fabricating efficient carbon dioxide separation membranes.


Introduction
Carbon dioxide emission from industrial flue gases and fossil fuels consumption is increasing chronologically, which has become a major cause of global warming. Impact of global warming on climate change is one of the serious threats to human survival. In this regard, the first step is to reduce the CO 2 emissions by separating it from industrial flue gases and air [1]. CO 2 separation process should be energetically and economically viable and the separated CO 2 should be used as a potential carbon source in different chemical processes and fuel production [2]. Currently, in coal fired power plants, ethanolamines are being used for capturing CO 2 using chemical absorption.
is process is cost-effective but production of N-nitrosamines, alkanolamines, and ketones is one of the major disadvantages because of their toxicity, carcinogenic nature, and omnipresence in the environment [3][4][5]. Similarly, oceanic storage and mineralization are two other prominent methods for CO 2 capture. erefore, to limit the amount of CO 2 and avoid the harmful by-products, carbon dioxide separation is required by an alternative path which should be of low cost, efficient, and robust with low energy consumption [6].
Separation through membranes is one of the feasible approaches which could meet the above requirements. Membrane-based separations can be cost-effective in terms of operational time durations, temperature, pressure, and energy requirements as compared to separations based upon absorption/adsorption processes [7]. Possible membrane materials include polymers [8], inorganic materials [9], metal organic frameworks, and mixed-matrix membranes [10]. e major challenges for the membrane process are high permeability, selectivity, longer operational hours, and stability at high temperature. Organic polymeric membranes are not suitable for carbon dioxide removal at high temperatures in post-/precombustion industrial processes.
Microporous inorganic membranes are stable at high temperatures but their performance is decreased due to transport mechanisms such as sintering and molecular sieving at elevated temperatures [11]. Among different classes of membranes reported, one class of membranes suitable for above-mentioned challenges is the ceramic molten-carbonate (MC) dual-phase membrane [12]. ese membranes consist of molten-carbonate phase held within the pore structure of inorganic ceramic support. e ceramic support consists of pure oxygen ionic or mixed ionicelectronic conductor ceramic phase. Ceramic-MC dualphase membranes have been reported as one of the most potential candidates for CO 2 separation because their permeation flux reaches up to 2.05 ml (STP) cm −2 ·min −1 , greatly approaching the level of commercial use, and they can be coupled with many industrial reaction systems [13]. However, fabrication of MC membranes is still at its infancy stage and limited to lab scale because of some critical issues which hinder their complete utilization at large scale such as decreased stability at long operational hours, low efficiency, incompatibility to couple with industrial processes, membrane degradation, high cost, and complex fabrication techniques including optimization of ceramic supports.
is review elaborates MC membranes in view of general concept, fundamental mechanisms, components of MC membrane, fabrication methods, and the factors that influence membrane performance. Major challenges that are serving as huge obstacle for membrane performance, stability, and commercialization have also been highlighted along with prospective solutions and recommendations. e membrane fabrication techniques proposed will be the proof-of-concept work for enhanced CO 2 selectivity and high temperature stability for longer operational hours.

Membrane Gas-Separation Mechanism
e classic molten-carbonate dual-phase membrane consists of two phases. One is the molten-carbonate (MC) phase derived from mixture of carbonate salts. e second is the gaseous phase of carbon dioxide ( Figure 1). Difference in partial pressures of CO 2 at the feed (P f ) and permeate boundaries (P p ) creates a concentration gradient of carbonate ions in MC phase. When carbonate ions arrive at permeate side, CO 2 is regenerated by reverse reaction. Membranes separate selective components from a mixture with fast permeant transport. Pressure gradient across the membrane is the driving force for separating target component ( Figure 1) [14]. Normally pressure at the feed side (P f ) should be greater than pressure at the permeate side (P P ). Membrane performance is generally determined by permeation flux (Ji), which is defined as the volume of gas i passing through the membrane per unit area per unit time at given conditions of temperature and pressure. Permeation flux is measured in SI units, that is, m 3 ·m −2 ·s −1 , or other units, such as ml·cm −2 ·min −1 , and can be represented by the following equation [15]: where C i and Csg are the concentrations of component i and sweep gas, respectively, measured through gas chromatography (GC). S is the effective surface area of membrane and Q is the flow rate of sweep gas. However, Ji is a function of intrinsic properties of membrane (thickness, materials, and geometry) and operating conditions as well (feed gas concentration, temperature, and pressure). To ensure fair comparison, permeability and permeance are the two better parameters to evaluate membrane performance, as they depend upon intrinsic properties of membranes only. Permeability and permeance are defined as in equations (2) and (4) [16,17]: where ΔP � P f − P p and L � membrane thickness. Selectivity is highly influenced by permeability. In traditional polymeric membranes, driving force depends upon diffusivity and solubility coefficient of the permeant. Diffusion rates with lower resistance pathways can be increased by introducing free pore volume but the selectivity is compromised. e selectivity and permeability relationship can be characterized by Robeson upper bound ( Figure 2) [6]. One way to overcome the upper bound without the decrease in selectivity is introducing careers on porous membrane surface. Careers should have high affinity for permeant.
Permeabilities required for economically feasible CO 2 separation range from 10 −13 to 10 −12 mol·m −1 ·s −1 ·Pa −1 with carbon dioxide/nitrogen selectivity of 50-100 [18]. Polymeric membranes cannot meet the criteria because of the upper bound described in Figure 2. Zeolite membranes have high selectivity with moderate permeability rates, while metal organic framework (MOF) shows opposite behaviour to zeolite membranes (i.e., high permeability and low selectivity) [19]. However, ceramic supported molten-salt membranes have higher permeability of 10 −12 to 10 −10 mol·m −1 ·s −1 ·Pa −1 at 600°C and their intrinsic properties show higher selectivity for CO 2 [20]. Furthermore, these Feed side Dual phase membrance Permeate side Figure 1: Membrane separation mechanism based on partial pressure difference across the membrane. Pressure difference (driving force), membrane thickness, and surface area are key parameters for membrane performance [14].
membranes are operational at high temperature range (400-1000°C) which is an extra advantage for their commercial application, that is, CO 2 removal from hot flue gas streams [20,21]. Based upon the mode of CO 2 transport and reacting species, molten-carbonate (MC) membranes can be classified into three types: (a) mixed oxygen and carbonate ion conducting membranes (MOCC), (b) mixed electron and carbonate ion conducting membranes (MECC), and (c) mixed electron, oxygen, and carbonate ion conducting membranes (MEOCC) [16]. In MOCC membranes, gaseous CO 2 reacts with oxygen anions (O 2− ) to form CO 2− 3 ions (Figure 3(a)). ese CO 2− 3 ions travel through MC phase in the membrane and arrive at permeate side and then CO 2 is reformed. In MECC membranes, CO 2 reacts with oxygen atoms in the presence of electrons on an electron conducting porous support (Figure 3(b)). MEOCC membranes exhibit both mixed electron and carbonate ion conduction (Figure 3(c)).

Molten-Carbonate Membrane Components
Molten-carbonate membrane includes highly ceramic porous support with infiltration of molten salt in pore space.
e physical properties of the support such as wettability, thickness, size, volume of pores, and tortuosity have high impact on membrane performance and stability at high temperatures [22]. Supports also facilitate CO 2 diffusion through membranes by carbonates formation through reaction between CO 2 and ions in molten salt. erefore, properties of membrane support can be tuned for increasing CO 2 permeance. In addition, with the help of molten salts, physical properties of supports such as wettability and vapor pressure can also be modified to meet the requirement at high temperature range [23]. erefore, it is essential to review the fundamental role and properties of different components of membranes.
is section elaborates the various components of molten-carbonate membrane, in context to properties of molten salt, as well as materials and geometries of membranes.

Molten-Carbonate Salts.
e molten-carbonate salt commonly used in molten-carbonate membranes is a ternary eutectic carbonate mixture of Li 2 CO 3 -Na 2 CO 3 -K 2 CO 3 with molar ratio of 43.5 : 31.5 : 25 [12,24,25]. Binary eutectic carbonates have also been applied to membranes such as (Li-Na) 2 CO 3 , (Li-K) 2 CO 3 , and (Na-K) 2 CO 3 salts with molar ratios of 52 : 48, 62 : 38, and 41 : 59, respectively [26][27][28]. e purpose of using mixtures of carbonates is to decrease melting point. Melting points of some eutectic binary and ternary carbonates are enlisted in Table 1.  Figure 2: Comparison between selectivity and permeability for CO 2 for polymeric, MOF, and mixed matrix (both polymeric and MOF) membranes [6,12]. CO 2 flux strongly depends upon ionic conductivity of molten-carbonate salts infiltrated in ceramic supports. erefore, proper wetting of molten carbonates in ceramic supports is essential, which in turn is related to thermophysical properties of the molten salt such as viscosity, density, and surface tension. ermophysical properties depend upon the temperature and environmental conditions around molten salt. Usually, viscosity of the molten salt decreases with increasing temperature. At 650°C, viscosity of a ternary molten carbonate (8 m·Pa·S) becomes comparable to that of water at 20°C (1.02 m·Pa·S) [30,31]. High temperature is beneficial as low viscosity assists the proper infiltration of molten salt into the pores of ceramic support.
Surface tension of eutectic molten-carbonate mixture (220 mJ·m −2 ) is much lower than that of ceramic supports such as alumina (1.84 J·m −2 ) and yttria-stabilized zirconia, YSZ (1.53 J·m −2 ), at 650°C, leading to very low contact angles between infiltrated carbonates and support [32,33]. us, wetting of supports by carbonates is improved further due to reduced surface tension of carbonates at high temperatures.

Ceramic Supports.
Chung et al. pioneered to apply stainless steel as an electron conducting porous support to develop the initial molten-carbonate membrane (MECC) [12]. e conductivity of carbonates decreases at high temperatures because of interfacial reactions between steel and Li 2 CO 3 leading to formation of LiFeO 2 . Ag has been used as an electron carbonate conducting support. Ag offers high electron conductivity as compared to steel, along with reduction in interfacial reactions at high temperatures [27]. Metal oxides such as Al 2 O 3 , NiO, and ZrO 2 have also been applied to avoid sintering of Ag and other interfacial reactions [34][35][36].
Extensively investigated oxygen ion conducting supports (MOCC), such as samarium doped ceria (SDC) and yttriastabilized zirconia (YSZ), show good performance at high temperatures. SDC and YSZ exhibit fluorite type structure with AO 2 formula (Figures 4(a) and 4(b)). Type A cations (Ce, Zr) occupy the face-centered cubic positions, while O 2− anions fit into tetrahedral interstices [37]. To enhance oxygen ion conductivity (Figure 4(a)), oxygen vacancy concentration can be increased by doping bivalent and trivalent cations into the parent structure. Based on atomistic simulation results, scandium oxide-doped zirconia has higher oxygen ion conductivity than YSZ and calcium oxide-doped zirconia due to lower energy requirements for dopant solution [39]. Oxygen ion conductivity for rare Earth-doped ceria is reported to be 10 −2 S/cm. e inclusion of trivalent dopant decreases the activation energy for oxygen ion conduction due to large ionic radius, which results in improving oxygen ion transport [40]. Electronic conductivity is also increased when cerium ion (Ce +4 ) is converted into Ce +3 in reducing atmospheres or low oxygen partial pressure [37]. At high temperatures (800°C), cubic fluorite and bismuth oxide (Bi 2 O 3 ) show excellent performance for oxygen ion conductivity of 2.3 S/cm −1 [41]. If Bi 2 O 3 is doped with rare Earth cations like Bi 1.5 Y 1.3 Sm 0.2 O 3 (BYS), then fluorite structure is further stabilized, and oxygen ion conduction is improved at low temperature ranges [42]. However, at elevated temperatures, BYS wettability for carbonates decreases, which would then require some modification in pore size and membrane surface [43]. In this regard, alumina deposition on BYS improves the wettability for molten salts. Similarly, zirconia atomic layer deposition (ALD) also improves the wettability and enhances oxygen and electron ion conduction for longer operational hours [44].
For mixed electron, oxygen, and carbonate ion conduction (MEOCC), materials such as perovskites have been used as supports for molten-carbonate membranes. Figures 4(c) and 4(d) represent a typical perovskite structure ABO 3 , where

Ceramic-Support
Geometry. CO 2 flux in molten-carbonate membranes not only depends on type of materials used in ceramic supports but also is influenced by physical properties of supports such as three-dimensional geometry, thickness, porosity, and tortuosity of the membrane. erefore, it is essential to review various physical properties followed by corresponding membrane fabrication methods and impact of these properties on membrane performance.

Symmetry and ickness.
Based upon the shape and geometry of support, membranes can be divided into three types. (a) symmetric disk, (b) asymmetric disk, and (c) asymmetric tubular membranes. Generally, the symmetric disk shape membranes are made up of one layer with molten-carbonate salt infiltrated homogenously in the membrane material ( Figure 5(a)). ey are prepared in the form of pellets using isotactic compression and tape casting methods at lab scale [25,47]. ese approaches are used because of ease in the preparation and experimentation at lab scale. However, CO 2 permeation is limited in these symmetric disk-shaped membranes due to their significant thickness. Hence, advanced microstructural modifications have been applied for promoting thinner membranes such as asymmetric disks and hollow fibre/tubular membranes. Asymmetric disk type membrane consists of a thin dense membrane layer, infiltrated with molten salt, on a strong mechanical macroporous support ( Figure 5(b)). Meanwhile, in asymmetric tubular geometry, thin, porous membrane layer exists as the inner layer of the tube and porous support forms outer layer of the tube ( Figure 5(c)). For asymmetric geometry in oxygen ionic conducting membranes, reduction in membrane thickness enhances CO 2 and O 2 permeation flux [24,48]. Furthermore, tubular structure offers higher surface area, easy scale-up, and convenient sealing procedure at high temperature [49]. Asymmetric tubular membranes are usually prepared by phase inversion, tape casting, spin spraying, and centrifugal casting techniques [48][49][50][51]. Recent literature reveals extensively the use of ceramic hollow fibres for molten-salt membrane support because of high surface area and reduced thickness of membrane. For YSZ and LSCF hollow fibre membranes, acceptable carbon dioxide fluxes have been reported [52,53]. Fabrication of hollow fibre membrane is more manageable than pellet membranes, but single hollow fibre membrane possesses low mechanical strength and thermal stability. To solve this issue, multichannel hollow fibre membrane has been proposed, which consists of SrFe 0.8 Nb 0.2 O 3-δ and carbonates ( Figure 6) with improved mechanical strength and high CO 2 flux [45]. However, further research is required for determining new approaches to develop membrane supports for enhanced CO 2 permeation with reduced thickness.
Asymmetric membranes have been further modified by addition of a third layer to achieve stability under different operating environments such as H 2 S containing atmosphere. Chen et al. reported SDC-BYS-based three-layered asymmetric carbonate membrane, which showed excellent resistance against H 2 S gas [2]. In three-layered asymmetric membrane, two layers of SDC and BYS adsorb the H 2 S gas and prevent the third SDC dense carbonate layer from H 2 S attack ( Figure 7(a)). Operational hours of three-layered SDC-BYS-SDC asymmetric membranes increased 10-12 times compared to single-layer SDC-BYS. e adsorbed layer can be regenerated in reduced atmosphere and trapped sulfur in this layer can be stripped out by converting into H 2 SO 4 , leading to reducing the cost of overall separation process. However, twolayered asymmetric membrane (Figure 7(b)) showed higher CO 2 flux as compared to three-layered membrane which could be attributed to reduced thickness of dense layer.

Porosity.
Carbon dioxide permeation is influenced by pore volume, pore size, pore connectivity, and tortuosity.
ese properties have great impact on support and moltencarbonate conductivity [54]. Overall, a thin support with highly interconnected uniformly distributed pores and low tortuosity is desirable. Gas-liquid interfacial area, triplephase boundary length at feed and permeate side is also dependent upon surface porosity of the support. For pore formation, sacrificial materials such as metal oxides are used and porosity of support is controlled by varying quantity of sacrificial materials. After the formation of porous structure, sacrificial phase is removed by acid etching, firing (organic removal), or reduction [26,[55][56][57]. Zhang et al. used NiO as the sacrificial template in combination with coprecipitation of SDC precursors to synthesize a porous matrix support with compositional homogeneity and uniformly distributed pores for MOCC membrane [26]. Figure 8 portrays a sponge-like microstructure of SDC membrane, exhibiting high interconnectivity between pores and SDC matrix in 3D (a) and 2D (b) demonstration.
To create pores in metal supports such as Ag, for MECC type membranes, chemical-dealloying strategy has been applied. Fang et al. used alloy of Ag and Al, each with 50% composition with Al as the fugitive element [23]. After dealuminizing the samples in 3 M HCl solution for 48 hours, porous microstructures of Ag matrix were obtained, exhibiting pore size from 1 to 10 μm (Figure 8(c)). Electrochemical dealloying using Zn as the sacrificial element has been employed further for producing single and more homogenous nanoporous Ag matrix (Figure 8(d)) [28]. is method is useful for formation of well-connected micron porous structure with uniform size distribution. Ability of molten salt retention is enhanced by formation of these submicron pores leading to increase in carbon dioxide flux [28,57].

Performance Evaluation of MC Membranes
Different components of MC membranes are interrelated to one another in determining overall membrane performance in terms of CO 2 flux measurement and stability at high temperature. is section presents overview of membranes in operation. Since there are various intrinsic parameters that influence membrane performance significantly and have a direct impact on CO 2 permeation flux, membrane performance has been critically analysed based upon these parameters such as material selection, geometry, membrane thickness, support properties, and operating conditions (feed/sweep gas composition).  membranes was first introduced by Chung et al. A ternary mixture of Li/Na/K carbonate was infiltrated in stainless steel support [12]. Rates of single gas permeation for CO 2 and N 2 were found to be very low for stainless steel support; however, the permeance of CO 2 : O 2 (2 : 1) was 0.13 ml·min −1 ·cm −2 at 650°C (Figure 9(a)).

Effect of Material Selection and
Separation factor for CO 2 over N 2 with separately measured N 2 permeation was found to be 16. However, the fluxes were found to be lower than expected because of interfacial reactions due to the formation of LiFeO 2 . Huang et al. applied Ag porous matrix with molten carbonate which increased CO 2 flux 6 times, that is, 0.82 ml·min −1 ·cm −2 at 650°C as compared    [26]; (c) porous Ag matrix after chemical dealloying [23]; (d) porous Ag matrix after electrochemical dealloying [28].
to stainless steel support membrane [27]. Ag exhibits high electron conductivity and better wettability with molten carbonates. However, CO 2 flux decreases at higher temperatures due to sintering of Ag. Zhang and coworkers used NiO matrix as a support [35]. NiO showed enhanced CO 2 flux up to 1.0 ml·min −1 ·cm −2 at 850°C and long-term stability at high temperature. is membrane showed good results for initial 15 hours of operation for both CO 2 and O 2 in the ratio of 2 :1 ( Figure 9(b)). After 15hours, permeation was slow because of interfacial reactions between the support and molten carbonate due to the formation of lithiated NiO layer of 100 nm thickness. CO 2 fluxes for different types of MECC membranes are summarized in Table 2.

Mixed Oxygen and Carbonate Ion Conducting Support
Membranes (MOCC). In MOCC membranes, CO 2 is ionized only in the presence of O 2− ions and is converted to CO 2− 3 ions, consequently transported through the membrane. us, CO 2 flux depends upon conductivity of O 2− ions. Various types of O 2− ion ceramic conductors have been applied in fabrication of MOCC membranes such as yttria-stabilized zirconia (YSZ) [47,59], fluorite-structured [42], samarium-doped ceria (SDC) [26,60,61], and gadolinium-doped ceria (GDC) [47]. Wade and coworkers compared CO 2 flux measurements of molten carbonate membranes for YSZ and Al 2 O 3 . Al 2 O 3based membrane showed CO 2 flux of 0.019 mL·min −1 ·cm −2 , whereas YSZ-based membrane showed CO 2 flux up to 0.13 mL·min −1 ·cm −2 . e low flux in Al 2 O 3 -based membrane is because of nonoxide ion conducting nature of Al 2 O 3 [47]. However, YSZ material itself has low O 2− ion conductivity as it reacts with LiCO 3 , leading to formation of lithium zirconate irreversibly at low pressure of CO 2 . Rui et al. applied a fluorite-structured BYS material to fabricate MC membrane, which showed CO 2 flux of 0.083 mL·min −1 ·cm −2 [42].
SDC membranes doped with CeO 2 have been investigated extensively because of their good stability, high O 2− conductivity, and better wettability with molten carbonate. However, there are various parameters that can affect membrane performance such as method of fabrication, geometry of support, thickness of support, and composition of feed gas. Zhang et al. developed SDC membrane of 200 μm thickness using sacrificial template method [26].
is SDC membrane gave better CO 2 flux (1.84 mL·min −1 ·cm −2 ) as compared to SDC membrane giving CO 2 flux of 0.79 mL·min −1 ·cm −2 prepared by presssintering method [60]. Enhanced CO 2 flux could be attributed to the formation of intra-and interconnected ionic channels in membrane microstructure. CO 2 fluxes for different types of MOCC membranes are reviewed in Table 3. e thickness of ceramic supports has a significant impact on membrane performance. It has been observed that, in MOCC membranes, O 2− ion transport is the controlling factor; therefore, reduction in membraned thickness should enhance the ion transport, increasing CO 2 flux consequently.
In addition to thickness, different shapes of porous supports have also been fabricated and tested to enhance CO 2 permeation efficiency such as classical disc-shaped and tube-shaped MC membranes. In the past decade, most of the studies were focused on disc shape membranes. However, disc-shaped membranes offer limited area, whereas tubular/ hollow fibre configuration is expected to enhance the flux  Figure 9: (a) Gas permeation of CO 2 , N 2 , and CO 2 + O 2 from a porous stainless steel support at different temperatures [12]. (b) Flux densities and selectivity of CO 2 and O 2 at 850°C with NiO support [35].
because of high surface area and easy method of fabrication at high temperature [49]. Lin et al. fabricated thin asymmetric tubular membranes of 150 μm and 120 μm thickness, which showed significant increase in CO 2 flux, that is, 1.56 mL·min −1 ·cm −2 and 2.05 mL·min −1 ·cm −2 , respectively [49,62]. e latter membrane configuration was tested at 700°C in 10% H 2 in the feed gas in addition to N 2 , CO, and CO 2 (Figure 10(a)). Among MOCC membranes, the highest CO 2 flux of 5.46 mL·min −1 ·cm −2 has been reported by Chen et al. for a hollow fibre SDC− membrane; however, its stability was only up to 85 hours, which can be attributed to its low mechanical strength [63]. However, in few studies, disc-shaped membrane showed higher CO 2 flux as compared to tube-shaped membrane despite of exhibiting same thickness [61]. e authors attributed the better performance to high particle packing density in disc samples as compared to the tubular membrane samples (Figure 10(b)).    [25]. Anderson and Lin applied LSCF material and obtained a CO 2 flux of 0.3 mL·min −1 ·cm −2 (4.77 × 10 −8 mol·m −2 ·s −1 ·Pa −1 ) at 900°C for 375 μm thick support membrane [24]. Moreover, the effect of membrane thickness on CO 2 permeance was also investigated systematically for MEOCC membranes. It had been observed that reducing membranes thickness enhanced CO 2 permeance as surface reaction becomes significant on decreasing thickness (Figure 11(a)). However, LSCF was found to be unstable in an O 2 -free environment [64]. Figure 11(b) shows a fast decay in CO 2 flux, showing a steady-state value of 0.03 mL·min −1 ·cm −2 after 65 hours' exposure at 900°C. is could be attributed to the layer of SrCO 3 formed on the membrane surface by reaction of LSCF with CO 2 . is can be prevented by adding O 2 in the feed gas along with CO 2 . In the presence of O 2 , SrCO 3 is decomposed to SrO at 800°C, which would protect the LSCF material beneath [66]. CO 2 fluxes for different types of MEOCC membranes are summarized in Table 4.
Norton et al. achieved CO 2 flux up to 3.0 mL·min −1 ·cm −2 by introducing O 2 in the feed gas in addition to CO 2 and N 2 . e presence of O 2 improved the membrane stability, maintaining the ionic and electronic conductivity of LSCF as well.
In [64], Lin and coworkers applied La 0.5 Sr 0. 5 [65]. Perovskite material with A-site fee alkaline is known to be stable against reaction with CO 2 at high temperature. In this regard, La 0.85 Ce 0.1 Ga 0.3 Fe 0.65 Al 0.05 O 3-δ (LCGFA) was found to be compatible with molten-carbonate salts and chemically stable at high temperature, that is, 900°C [25]. However, the CO 2 flux (0.044 ml·min −1 ·cm −2 ) obtained for LCGFA was much lower than that of LSCF membranes. Support geometry other than disk shape has also been applied for MEOCC membranes such as multichannel hollow fibres. Asymmetric hollow fibre-based membranes (Figure 12(a)) offer high surface/volume ratio, less transport resistance, and easy method of sealing. CO 2 permeation flux of 0.64 mL·min −1 ·cm −2 at 850°C was reported by Jiang et al. for multichannel hollow fibre membrane of SrFe 0.8 Nb 0.2 O 3δ (SFN) support material with 220 μm (Figure 12(b)) [45]. Meanwhile stability test showed that CO 2 permeation flux was maintained at of 0.31 mL·min −1 ·cm −2 for 200 hr at 700°C (Figure 12(c)). SFN membrane showed good chemical stability as well.
Combination of fluorite and perovskite conducting material has also been used as ceramic support for MEOCC membranes such as Ce 0.85 Sm 0.15 O 2-Sm 0.6 Sr 0.4 Al 0.3 Fe 0.7 O 3 (SDC-SSAF) composite material [46]. ese composite dualphase membrane matrices are known to exhibit high CO 2 permeability and significant chemical resistance against CO 2 attack [67]. SDC-SSAF composite membrane of thickness of 1300 μm showed CO 2 flux of 0.24 mL·min −1 ·cm −2 at 900°C, which increased up to 0.28 mL·min −1 ·cm −2 , when O 2 was added in the feed gas [46]. [68]. Effect of membrane thickness on CO 2 flux is summarized in Table 5. From equation (3), it can be seen that permeation flux is inversely proportional to thickness. e  Figure 10: (a) High temperature stability test at 700°C for 0.12 mm asymmetric tubular dual-phase membrane in simulated syngas [62]; (b) comparison of CO 2 fluxes for disk-and tube-shaped membranes as a function of temperature [61]. general trend observed in MC membranes is that decreasing membrane thickness would enhance surface reactions, increasing CO 2 flux consequently. ere is a critical thickness, below which surface exchange reaction becomes slow. A critical thickness of 840 μm is reported for Ag-carbonate MECC membrane [58]. Above 840 μm, decrease in thickness increases CO 2 flux. However, decrease in membrane thickness does not cause the proportionate rise in CO 2 flux as expected, in case of MOCC membranes. Lin et al. reported CO 2 flux of 0.17 ml·min −1 ·cm −2 and 0.87 ml·min −1 ·cm −2 for thick symmetric SDC membrane (1500 μm) and thin asymmetric SDC membrane (150 μm), respectively [60,61]. e increase in CO 2 flux was less than expected, that is, only 5 times as compared to 10 times increase in thickness, suggesting the limitation from surface exchange reaction. Inverse relation between membrane thickness and flux is further evident from activation energy values. Activation energies of thin asymmetric membranes (150 μm: 60.3 kJ·mol −1 and 120 μm: 62.5 kJ·mole −1 ) are lower than that of thick symmetric membranes (1500 μm: 81.2 kJ·mol −1 and 1000 μm: 82.4 kJ·mol −1 ) [49,62]. Difference in activation energies could be attributed to the difference in porosity and tortuosity of the membrane materials, attained during fabrication procedures such as sintering conditions, which ultimately affects conduction of oxygen and carbonate ions [54].    In case of MEOCC membranes such as LSCF-MC membrane, the trend of activation energy change is opposite. Activation energies tend to increase with decreasing membrane thickness, while CO 2 flux increases with decreasing thickness (Table 4). Increase in activation energies can be attributed to the involvement of surface exchange reactions on decreasing thickness [24]. However, for LCGFA-MC membrane, activation energy values have been found to be constant for both 750 and 1500 μm thick membranes, whereas CO 2 flux is higher for thinner membrane, suggesting that CO 2 flux is controlled by bulk diffusion rather than surface exchange reaction [25].

Effect of Support Microstructure.
e microstructure properties of supports such as porosity, pore size distribution, tortuosity, and density of triple-phase boundary play a significant role in evaluating membrane performance because the porous support in molten-carbonate membranes not only supports MC phase but also provides medium for conduction of O 2− ions and electrons. Ortiz-Landeros et al. investigated that by decreasing the sintering temperature of LSCF-MC membrane from 1100 to 1000°C. CO 2 flux was increased three times [54]. is can be attributed to the changes in microstructure at low sintering temperature such as increase in porosity and decrease in tortuosity. Porosity can also be enhanced in matrix supports using pore forming agents such as carbon, cellulose, and metal oxides. Zhang et al. prepared a series of interconnected three-dimensional SDC supports of different porosity using NiO as a sacrificial pore former agent. With increase in porosity from 30 to 50%, tortuosity of the membrane decreased from 26.1 to 2.2 and CO 2 flux increased from 0.26 to 1.84 ml·min −1 ·cm −2 [26]. CO 2 flux and long-term stability of MC membranes is also influenced by pore size of the matrix. Capillary forces cause retention of carbonate phase in porous matrix, which depend upon pore size consequently. Large pores cannot generate capillary forces to withhold MC phase, leading to loss of MC phase and reduction in CO 2 flux. Large pores also reduce the density of triple-phase boundaries, causing decrease in CO 2 flux. Different types of pore formers and fabrication methods have been employed to reduce pore size for MC membranes (Table 6). By changing the pore former from cellulose (pore size: 15-20 μm) to carbon (pore size: 8-10 μm), CO 2 flux was increased from 0.39 to 0.61 ml·min −1 ·cm −2 in Ag-MECC membranes [58,69]. Fang and coworkers further decreased the pore size to 1 μm using electrochemical dealloying process for pore formation and achieved CO 2 flux up to 1.02 ml·min −1 ·cm −2 [23]. Same trend was observed by Zhang et al. for SDC membrane, where pore size was reduced to 0.55 μm along with subsequent increase in porosity leading to high CO 2 flux [26]. It can also be inferred from Table 6 that long term stability of MC membrane is enhanced by decreasing pore size of matrix support. is is because small pore size retains MC phase much longer at high temperatures, increasing CO 2 flux ultimately.

Effect of Feed/Sweep Gas.
Compositions of feed and sweep gas influence CO 2 transport across MC membranes. Various research efforts have been dedicated to studying the effects of different gases used as feed/sweep on CO 2 flux (Table 7). It can be inferred from equation (3) that partial pressure difference of CO 2 across the membrane at feed and permeate side has a significant impact on its transport in MC membranes. e higher the partial pressure gradient of CO 2 was, the higher would the CO 2 flux be.
Norton et al. studied the effect of increasing CO 2 partial pressure on SDC− membrane and found that, by increasing the CO 2 pressure at feed side from 0.1 to 0.9 atm, CO 2 flux across the membrane also increased from 0.39 to 0.79 ml·min −1 ·cm −2 [60]. MOCC membranes have been assessed in CH 4 gas environment. SDC-MC membrane of 1150 μm thickness has been tested to capture CO 2 using CH 4 gas in the feed. CO 2 flux was measured at different partial pressures of CO 2 at feed side. Highest CO 2 flux of 0.13 ml·min −1 ·cm −2 was obtained at CO 2 partial pressure of 0.375 atm (Figure 13(a)) [70]. us, MOCC membranes can also be employed in purification of shale gas/biogas. CO 2 flux of MC membranes can also be enhanced by decreasing partial pressure of O 2 at feed or sweep side such as adding H 2 on feed side. Since O 2− ions travel from sweep to feed side in membrane, addition of H 2 would decrease partial pressure at feed side causing rise in O 2− ion flux, thus leading to corresponding increase in CO 2 flux in opposite direction [16]. Since MOCC membranes involve transport of O 2− ions, they are more appropriate to capture CO 2 in feed environments containing H 2 such as from precombustion processes. Figure 13(b) shows linear relationship between CO 2 flux and partial pressure of H 2 at feed side for SDC membrane [26]. Chen and coworkers measured CO 2 flux of 4.78 ml·min −1 ·cm −2 and 5.46 ml·min −1 ·cm −2 in the absence and presence of 5% H 2 in feed gas, respectively, for an SDC hollow fibre membrane of 100 μm thickness (Figure 13(c)) [63].
For MECC membranes, CO 2 flux can be enhanced by adding H 2 to sweep side. Here the mechanism is different from that of MOCC membranes. H 2 reacts with the permeated O 2 , causing decrease in O 2 concentration. is would shift the equilibrium to the right side (CO 2− 3 ⇌CO 2 + 1/2O 2 + 2e − ), followed by increase in CO 2 permeation. Fang et al. investigated that, by increasing H 2 to 1.41% in sweep gas mixture, CO 2 flux could be enhanced to 2 times (1.02 ml·min −1 ·cm −2 ) as compared to CO 2 flux in pure Ar (0.49 ml·min −1 ·cm −2 ) for Ag-MC membrane [23].

Effect of Surface Modification.
Surface modification of membranes can be employed to facilitate surface reaction without changing the bulk properties. Various methods have been applied to modify surface such as colloidal deposition, chemical vapor deposition, and atomic layer deposition. is section elaborates in detail on each method.  (Figure 14(a)) at 750°C because of enhancement of surface reactions [65]. Surface of Ag-MC membrane was also coated with a thin layer of colloidal c-Al 2 O 3 solution before impregnation with molten-carbonate salt [69]. MC salt is expected to react with Al 2 O 3 to form a layer of LiAlO 2 on the surface, enabling better wettability between molten carbonate and Ag porous network. However, the Ag-MC membrane coated with 5% Al 2 O 3 gave the highest CO 2 flux (0.39 ml·min −1 ·cm −2 ) and stability as compared to that coated with 10% Al 2 O 3 and uncoated sample (Figure 14(b)). 5% Al 2 O 3 was the optimum concentration limit, above which bulk transport CO 2− 3 may be hindered, leading to reducing the CO 2 flux.

Chemical Vapor Deposition (CVD).
Chemical vapor deposition (CVD) involves coating of thin films on a heated substrate by means of gaseous phase precursors and coating can be obtained with tunable deposition rates [71]. In Al 2 O 3 modified Ag-MC membranes, with colloidal deposition of Al 2 O 3 , usually the thickness of coating cannot be controlled and subsequently stability improvements cannot be consistent from batch to batch. A uniform layer of Al 2 O 3 was deposited over the surface of Ag matrix by means of CVD, which enhanced the stability of MECC membrane [34]. No sign of degradation was observed for 100 hours at 650°C as compared to the pristine sample which lost 50% of its original flux in the first 20 hours (Figure 15(a)). Large pores in SEM images further indicate significant sintering of Ag particles and loss of metal carbonate in pristine sample, while the presence of dense microstructures in CVD coated sample shows decrease in sintering (Figure 15(b)).

Atomic Layer Deposition (ALD).
Atomic layer deposition (ALD) is a technique used for ultrathin films formation by gas phase deposition with thickness control at submicron and nanometer levels at wide temperature range [20]. ALD process involves use of precursors segregated from each other in a gas phase by purging and pulsing alternatively [72]. Due to consecutive pulsing, monolayer can be formed in each ALD cycle, and its thickness can be tuned by replicating ALD cycles. ALD is highly favourable for porous materials because of the following reasons: (a) Precursors used in ALD can be tuned into very small pores because of their vaporization phase. ese   Figure 13: CO 2 flux density of SDC membrane (a) as a function of logarithm of CO 2 partial pressure [70], (b) as a function of H 2 partial pressure [26], and (c) in the presence and absence of H 2 [63]. small pores can be adsorbed on pore walls and rejoin with previously formed precursors [73]. (b) ALD occurs on surface of substrate, so very-highquality and uniform thin film is deposited on highly dense ceramic porous support [74]. (c) By changing the ALD cycles, thickness of the layers can be accurately controlled [75].
ALD technique has been applied to fabricate ceramic membranes. Li et al. reported application of ALD for ceramic membranes which consisted of zirconia nanoparticles sintered on alumina supports. Precise pore tailoring was achieved by modification of uniform and conformal layer of metal oxides on BSA (bovine cerium and radium) ceramic support [76]. Figure 16 clearly indicates that a number of ALD cycles increase the thickness of deposited layer of alumina which increases the grain size leading to decrease in pore size subsequently. For maximum efficiency, porosity of membrane can be tuned by selecting specific ALD cycle.
Tran et al. fabricated alumina-titania composite membrane for H 2 separation by means of plasma-enhanced ALD (PE-ALD). A thin titania layer (approximately 10 nm thickness) was deposited on c-Al 2 O 3 support by adjusting the number of ALD cycles (280 cycles) [77]. SEM image in Figure 17(a) shows the top view of c-Al 2 O 3 which appears smooth with no pinholes. Although the TiO 2 layer deposited through PE-ALD is not clearly visible, the deposition of TiO 2 seems to be uniform and shows no significant effect on c-Al 2 O 3 layer (Figure 17(b)). ese membranes were tested under thermal and hydrothermal conditions for high CO 2 / H 2 selectivity in steam reforming and water-gas shift processes for H 2 gas separation [78].
ALD treatment has also been applied in Ag-MC membranes for CO 2 separation to further minimize the sintering problems of Ag. Zhang et al. deposited 25 nm thick ZrO 2 layer on Ag matrix using 200 cycles of ALD [36]. e ZrO 2 film was uniform and dense and adhered strongly to Ag matrix (Figure 18(a)). It not only increased CO 2 flux up to 0.8 ml·min −1 ·cm −2 but also led to much stabilized Ag matrix resistant to sintering resulting in prolonged operational hours (850 hours) at 700°C (Figure 18(b)).

Major Challenges and Prospective Solutions
Molten-carbonate membranes offer significant potential to separate CO 2 at high temperature, due to their inherent characteristics such as permeability, selectivity, reproducibility, and high temperature stability. However, their potential to be cost-effective and energy-efficient is still not investigated. In addition, their application at commercial scale is also limited till now, as most of the research work has been dedicated to their fabrication and testing at lab scale.
is section highlights the key areas of deficiencies of molten-carbon membranes including long-term stability, their methods of fabrication, material selection, and commercialization.

Long-Term Stability at High Temperature.
In current scenario of the research related to molten-carbonate membranes, the most crucial challenge is to achieve and maintain maximum efficiency in terms of CO 2 permeation flux for longer operational hours at high temperature. In this regard, long-term stability is more essential than achieving just high values of CO 2 fluxes. Up till now, the longest stability achieved at lab scale testing is 1000 hours approximately. e main causes of membrane degradation are loss of molten-carbonate phase at high temperatures and sintering of microporous supports, leading to decrease in CO 2 flux, which ultimately requires to direct research efforts at improving all components and mechanisms of MC membranes.

Material Selection.
Molten-carbonate membranes consist of a microporous solid support infiltrated with molten-carbonate salts. Among MECC membranes, Ag has shown promising results in terms of increasing CO 2 flux; however, sintering of Ag at high temperature and its high cost limits its application at industrial level. Although surface modifications have been applied to reduce degradation of Ag, high magnitudes of CO 2 fluxes have still not been achieved as compared to those of MOCC membranes. Other materials such as NiO showed better fluxes in MECC membranes; however, poor stability of NiO in reducing atmospheres limits its use at industrial level. erefore, finding new materials in MECC membranes is highly desirable.
For MOCC membranes, many materials have been investigated, which allow high conduction of O 2− ions, such as SDC and GDC. However, they degrade in flue gas environments containing SO 2 and H 2 S impurities. Among MEOCC membranes, majority of the research work has been conducted on LSCF type material. Overall, CO 2 fluxes achieved for MEOCC membranes are lower than those for MOCC membranes with no substantial improvement in long-term stabilities at high temperatures. us, new materials with good chemical stability and better shelf-life are needed to be explored for both MOCC and MEOCC membranes.

Microporous Support/Geometry/Surface Modifications.
Numerous research papers have been dedicated to optimizing the microporous structure and geometry of support. e general agreement is that molten-carbonate membrane exhibiting reduced thickness, high porosity, low tortuosity, and well-connected uniformly distributed pores of small size would promote CO 2 permeation. Overall, reduction in pore size and formation of well-connected three-dimensional porous structure has enhanced long-term stability of membranes, but reduction in membrane thickness is limited by surface exchange reactions. In addition, asymmetric geometries such as hollow fibres showed highest CO 2 fluxes because of high surface area. However, they exhibit low mechanical strength, which limits their utilization. Tailoring molten-carbonate membranes through surface modifications such as CVD and ALD has not only improved CO 2 flux but also increased stability in MECC membranes. Surface modification strategies should be applied to MOCC and MEOCC membranes as well to further tune support properties such as pore size and pore volume.

Fabrication Methods and Commercialization.
In view of optimizing support structure and geometry of molten-  carbonate membranes, many fabrication techniques have been used to improve CO 2 flux and stability such as tape casting, extrusion, phase inversion, centrifugal casting, and sacrificial-template synthesis. However, at present, most of these fabrication methods are limited to lab scale. eir complexity, high cost, and viability at industrial scale are still not evaluated in systematic manner. More efforts are needed not only to develop cost-effective methods of fabrication but also to make them compatible to capture CO 2 in industry according to the type of environment such as precombustion processes, water gas shift reactions (WGS), dry methane reforming (DMR), and biogas purification systems.

Conclusions
is review summarizes the recent progress in moltencarbonate membranes with focus on material selection, geometry, and surface modifications. Based upon mechanism, three types of membranes have been reviewed thoroughly, that is, MECC, MOCC, and MEOCC. In addition, the impact of physical properties of membranes (support microstructure, geometry, and membrane thickness) and operating conditions (feed/sweep gas composition and presence of impurities) on membrane performance has also been discussed in detail.
MC membranes have been found to be quite promising for CO 2 permeation because their intrinsic properties such as selectivity, permeability, and scalability can be easily tailored and fine-tuned by optimizing fabrication methods with operating conditions. However, the key challenges such as long-term stability at high temperatures, feasibility at commercial scale, and cost-effectiveness need to be addressed systematically. Moreover, selection of alternative materials and integration of fabrication methods with surface modifications (e.g., ALD) are highly recommended to improve CO 2 permeation along with chemical and long-term stability at lab scale.

Data Availability
e data used to support the findings of this study are provided within this article and are available from the corresponding author upon request.

Conflicts of Interest
ere are no conflicts of interest to declare.