The prospect of synthesis of PES/PEG blend membranes using blend NMP/DMF for CO2/N2 separation

CO₂ is increasingly becoming a valuable commodity that has ignited interest surrounding carbon capture technologies [7]. Its exploitation may enhance industry value-chains and positively affect strategies for reducing CO₂ emissions. An alternative to curb CO₂ emissions is CCS. This method captures CO₂ from a large source then stores it for commercial use or injection [4]. For instance, the oil and gas processes often re-inject CO₂ into a reservoir for enhanced oil recovery [12]. This CCS application prevents CO₂ emission, re-routing the GHG into the reservoir instead of emitting it as flue gas [13]. A typical power plant uses a simple scrubber to remove impurities before flue gas is released through a furnace stack, resulting in the release of gases with high N₂ and CO₂ concentrations. The lack of CO₂ removal from residual gases in power plants resulted in an emission of 11.1 Gt of CO₂ in 2012, which amounted to around a third of global CO₂ emissions [14].

Furthermore, only a few power plants have started to reduce their CO₂ emissions using CCS technologies [15]. CO₂ is a by-product of combustion, and the choice of CO₂ removal scheme should depend on the combustion process within the system. Presently, CCS systems are available in the market but are costly in general. A complete CCS system can cost up to 70–80% of a power plant [16]. There are three different carbon capture concepts: pre-combustion capture, oxyfuel capture, and post-combustion capture system, as illustrated in Fig. 2 [17, 18].

In a pre-combustion carbon capture system, natural gas or coal transforms into syngas using oxygen (O₂) or steam reaction. This results in mostly carbon monoxide (CO) and hydrogen (H₂) gases that are free from pollutants [19]. Next, the syngas undergoes a water–gas shift reaction to CO₂ and creates more H₂ through CO conversion [20]. The water–gas shift results in a higher concentration of CO₂, which is then separated while the pure H₂ produced is used for combustion and mostly generates N₂ and water vapour [21, 22]. Integrated Gasification Combined Cycle (IGCC) power plants predominantly use this method to capture CO₂ [23]. However, this method leads to an efficiency loss of 8% for coal-fired IGCC power plants due to the need for a gasification unit [24].

The purpose of an oxyfuel carbon capture system is to achieve cleaner combustion using pure O₂ instead of air to create flue gas rich in CO₂ and water (H₂O) in its vapor phase [25]. A portion of the flue gas is redirected into the furnace to control the flame temperature [22, 26]. The residual CO₂ and H₂O in the remaining flue gases discharge via a purification process in cryogenic conditions [27]. The flexibility and customisability of the air separation unit (ASU) and supporting equipment, as well as its high purity (> 99.9%), make this method of CCS very efficient for coal-fired power plants [28]. However, the utility and energy costs remain a challenge for oxy-fuel combustion as it needs pure O₂ supply and consumes a large amount of energy for its boilers and ASU [11, 22, 25].

The post-combustion capture system captures CO₂ from the typical flue gas and prevents its release into the atmosphere. In this system, the carbon capture process occurs after the combustion process, which presents a retrofitting option without any change required on the pre-existing process [16, 29, 30]. Additionally, the post-combustion scheme's advantage lies in the maturity of the processes involved, such as amine scrubbing (absorption), leading to CO₂ purity higher than 99.99% [7]. However, the main limitation of this technology is the high energy load that often correlates with the cost accumulated from the solvent regeneration process and compression of CO₂ [26, 31]. The low amount of CO₂ levels in the combustion flue gas, typically less than 15%, means that more energy is required to separate CO₂ from flue gas [30, 32]. Furthermore, other technologies that are viable alternates to capture CO₂ selectively are adsorption, cryogenic separation, or membrane separation [18].

The CO₂ separation process consists of 75% of the overall CCS costs and 50% of the electricity production costs [33]. These figures vary for the various CCS systems; cost reduction for CO₂ separation remains the most significant dispute for CCS to be adopted in the energy sector [27]. Currently, there are wide varieties of processes for extracting CO₂ from gas streams. Physical or chemical properties are the driving force of these processes, such as absorption, adsorption, cryogenic, and membranes [34, 35].

One of the widely used technologies to separate CO₂ from power plant exhaust gas is absorption stripping [36]. In this approach, the CO₂ in the flue gas is cooled to about 320 °K. Then, it feeds into a column stripper where it comes into contact with a liquid solvent (absorbent), absorbing CO₂ from the flue gas mixture. The solvents used in this method are either physical or chemical solvents. The physical solvent method uses organic solvents to mechanically absorb the components without reacting while using pressure and temperature as a driving force [37]. Regarding chemical absorption, it mostly relies on acid–base reactions by applying alkaline solvents [29].

Amine-based solvents such as mono-ethanolamine (MEA) have been widely utilised in the industry for over 60 years and have become a highly developed and promising product to date [7]. Many companies have already installed the MEA absorption system in full-scale for CO₂ separation in fuel power plants. For example, the Shell gas-fired power plant in Norway and the RWE coal-fired power plant in the United Kingdom produces 860 MW and 500 MW of electricity, respectively [38]. However, these absorption processes are non-economical as they require high energy input and large-size equipment. A schematic of a typical adsorption process for CO₂ separation is presented in Fig. 3 [39].

Though absorption is currently used widely in the industry, a few disadvantages exist in the system. Firstly, the high corrosion rate of equipment used in the absorption of CO₂ [29]. Besides that, solvent degradation requires more fresh solvent feed, resulting in increased costs of the products and disposal of degraded solvents. The disposal of solvents would result in more pollution to the surroundings [38]. Additionally, this separation technique also involves high energy consumption due to the absorbent regeneration stage that requires high-temperature operations [40].

Another approach to separating CO₂ is adsorption. The adsorbents used in this process are in solid form rather than in liquid form. The use of a solid adsorbent is to accumulate all the CO₂ that passes through its surface [16]. Similarly, the adsorption method has two critical stages: adsorption and adsorbent regeneration, which is more commonly called desorption. The removal of the desired components from flue gas occurs in the adsorption stage using solid adsorbents such as zeolites, lithium zirconate, silica gel, activated carbon, and molecular sieves [37, 40]. The used adsorbents with high contents of the desired components are regenerated and recycled in the desorption stage [40].

There are two variations of adsorption used to remove and store adsorbed CO₂. These are temperature swing adsorption (TSA) and pressure swing adsorption (PSA) [41]. PSA is mostly used as CO₂ recovery technology for power plants with high efficiency (> 85%) [41]. This process involves the selective adsorption of CO₂ on the surface of a solid adsorbent at a higher pressure. Further, it swings to a lower pressure to regenerate the adsorbents and release CO₂ for subsequent transport [42]. As temperature does not change dramatically throughout the process, thermal and mechanical energy consumption is considerably lower than traditional TSA. Thus, PSA is more practical for industrial purposes [43]. In TSA, removing the absorbed CO₂ from the adsorbents involves increasing the device temperature by hot air or steam injection. The adsorbents typically take a longer time to regenerate compared to PSA, but higher CO₂ purity (> 96%) and higher recovery (> 80%) is achieved [44]. However, the TSA process requires more significant amounts of energy for the CO₂ adsorption stage, as it uses a high-temperature adsorbent [45]. The cost to operate a typical TSA process is estimated to be between US$ 80–150 per tonne of CO₂ captured [46]. Scaling up the adsorption process is also proven to be a challenge [47].

As for the cryogenic CO₂ removal process, the gas mixtures separation theory uses fractional condensation where the separation occurs at low temperatures [48]. This system is typically preferred when there is high CO₂ content in the stream, with more than 90% [27]. The process involves chilling of flue gases at very low temperatures to liquefy CO₂ and ease the subsequent processes [49]. The cryogenic separation allows the direct production of pure liquefied CO₂, making it more convenient for transport and storage [29]. Simplicity is another advantage for this technique as it does not require any additional solvent or other components. Furthermore, the simple principles of the process make it easy to scale-up [50].

While cryogenic separation has many advantages, it requires enormous power or energy for the refrigeration (chilling) stage and operational problems due to CO₂ solidification [51]. Incorporating gas hydrates (Fig. 4) for the sequestration of CO₂ can mitigate this problem [52]. This process includes a second stage where the residue of CO₂ from the cryogenic phase of condensation is solid hydrates that can easily be collected. This method decreases the energy required for feed gas cooling from the process streams [52].

Based on above, complex and costly operations are often associated with other CO₂ removal technologies. Hence, membrane technology is increasingly becoming an innovation that can compete against these technologies [53]. Membrane technology meets the general low-cost requirements for CO₂ removal [54]. Using a membrane with high selectivity, CO₂ separates from flue gas without needing a high concentration of CO₂ at the inlet [34].

A membrane is a semi-permeable barrier and has numerous transport mechanisms to aid in the recovery of specific components from inlet gas streams [55]. This technology shares a similar role as filters, where it only allows particular components to flow through. As a result, it produces a permeate or retentate stream with a high content of the desired component [56].

In a membrane, a pressure gradient induces the gas separation process. The permeate-side stream is normally open to the atmosphere, and the feed side carries the pressure from previous processes [57]. The component gases then flow from the side with higher pressure to the side with lower pressure. Compared to other approaches, membrane separation technology is a more straightforward and better energy-saving process, suitable for separating CO₂ [58]. This approach does not require a regeneration process as it does not use any separating agent. As a membrane separator is a piece of static equipment, the requirement for maintenance is very minimal [59]. Compared to other traditional separation methods, the capital cost is relatively lower as additional equipment, such as a heat exchanger, is not required to be installed. Therefore, this process saves more energy as phase transformation is not needed [49].

Moreover, mild conditions make this process relatively easy to control and operate, while its non-complex process makes it convenient to scale-up [29]. A notable advantage of membrane separation is its high separation efficiency, whereby it is common to achieve 80% purity of the desired product. Examples of its success can be seen in studies by Gielen [60] and Audus [61]. They recorded their success of achieving 82 to 88% efficiency for CO₂ separation when utilising the recently adapted technology of membrane separation [16, 60, 61].

Despite these advantages, membrane technology has its drawbacks. This technology is still in unknown territory compared to other conventional technologies as it is not mature yet. It is also difficult to gain more experience using this technology in a large-scale process as many companies prefer a more proven and conventional process [62]. Lifespan is also a significant issue with membrane technology due to fouling [63]. Based on Lu et al. [64], the Palladium membranes have a short lifespan; that often only requires months before a new replacement, which subsequently results in additional costs [64].

Further, selectivity and the required purity of the permeate highly influences the energy consumption of a membrane. On some occasions, one membrane is insufficient to achieve the required purity, which may require multiple-stage membranes or a membrane with larger areas. Therefore, poor-performing membranes can result in more costs through energy usage and replacements [34].

Nevertheless, in recent years developments in pilot-scale membrane plants have grown, particularly for post-combustion carbon capture [65]. Hägg et al. [66] reported that a pilot post-combustion membrane plant installed at a Northern Cement factory in Norway could capture 70% of CO₂ from low CO₂ content (17%) flue gas in a single-stage setup. The plant uses polyvinylamine (PVAm) based hollow fibre membrane modules (up to 18 m²) fabricated by Air Products [66]. Similar pilot-scale PVAm hollow fibre membrane modules were applied to separate real flue gas from a propane burner at the SINTEF Tiller plant in Trondheim (Norway), where CO₂ purity reached 60% in the permeate stream from a feed flue gas containing 9.5% of CO₂ [67]. However, from a process engineering perspective, the future development of membrane technology in post-combustion CO₂ capture should be focused on future CO₂/N₂ selective membranes with selectivity higher than 50 and CO₂ permeance higher than 4000 GPU. This precise membrane specification could offer a CO₂ capture cost lower than US$ 15 per ton of CO₂, which is below the US Department of Energy's (DOE) targeted goal of US$ 20$ per ton CO₂ [67, 68]. Merkel et al. [68] suggested that future developments in membrane separation should focus on enhancing permeability of the membrane as long as its CO₂/N₂ selectivity is above 30 [65, 68].

Furthermore, other operational properties should also be considered, such as chemical stability, heat resistance, lifespan, and durability against plasticisation [29]. These properties are usually affected by different membrane materials, which could be inorganic or organic [69]. Similarly, the membrane structure, whether non-porous or porous, can also alter its properties and gas separation performance [70].

The separation properties of a membrane are affected by the type and characteristics of the fabrication material [71]. There are primarily two types of materials that make up a membrane, either polymeric; or inorganic [29]. Both types of membranes can be categorised into dense, porous, and composite based on their morphology, as stated in Fig. 5 [72].

Dense homogeneous polymer membranes are only practical when the process requires highly permeable membranes, often requiring a minimum membrane thickness to be mechanically stable [72]. The permeate that flows through the membrane is usually very small, so a minimum thickness is necessary to ensure mechanical stability [73]. The dominant transport mechanism through the dense polymeric membranes is solution diffusion. In contrast to porous membranes, permeation of components in dense membranes is indirect. The most beneficial aspect of the solution-diffusion mechanism is the ability to adjust the permeation of the various components throughout the separation process [74]. This mechanism has three sequential stages. The first stage is the adsorption of the desired component on the feed side of the membrane. Next is the transfer of molecules through diffusion through the polymer matrix. Lastly, desorption occurs where the desired components evaporate on the surface of the membrane's permeate side [75]. The driving forces of this system are the variation in thermodynamic behaviours around the membrane and the magnitude of the gas-polymer interactions [76]. Concentration variances exist due to a gradient in thermodynamic activity, as presented in Fig. 6 [75, 77].

In this mechanism, permeability (P) defines the efficiency of a membrane, as shown in Eq. 1 [78]:where ‘D’ is the diffusivity coefficient, while ‘S’ is the solubility coefficient. The ratio of the permeabilities mostly describes the selectivity ( ∝). Equation 2 represents this ratio [78].

Additionally, a well-known equation to calculate the gas flow rates, Q is shown in Eq. 3.where ‘p₁′ and ‘p₂′ are the feed and permeate side pressures respectively, ‘A’ the membrane area, and ‘l’ the thickness of membrane [78].

Typically, porous membranes come in one of two forms, symmetric (isotropic) or asymmetric (anisotropic) membranes [70]. Porous membranes with a consistent structure throughout their whole area fall under the symmetric porous membranes category. On the other hand, membranes with a gradient in their structure are classified under the asymmetric membrane category [72]. Further, pore sizes are also useful for categorizing porous membranes, as summarised in Table 1 [79].

In addition to characterisation, pore sizes heavily affect the gas transport mechanisms through the membrane [65]. In a porous membrane, the transport mechanism is highly dependent on the pore size compared to the mean free path of molecules and the size of the transported molecule. The transport mechanisms in a porous membrane follow pore flow models, such as Knudsen diffusion, Poiseuille flow, molecular sieving, surface diffusion_, or capillary diffusion [81].

Knudsen diffusion occurs typically in a convective flow of a porous membrane. Generally, it takes place in a small diameter of the long pore where the frequent collision of molecules against the membrane walls occurs [82]. This mechanism is likely to happen in a membrane with pore sizes lower than the mean free path (λ) of the gas molecules, where λ is calculated by Eq. 4 [76].where ‘η’ is gas viscosity, ‘p’ is pressure, ‘T’ is temperature, ‘R’ is gas constant, and ‘M’ is the molecular weight. If the membrane pore size is smaller than its mean free path (r/λ < 0.05), molecules tend to move independently and frequently collide with the membrane walls than amongst themselves. The collisions result in a velocity gradient between molecules of different components, which would then be a medium for gas separation. The lighter components are more likely to go through the membrane, while the heavier components are less likely to pass. The molar flux is calculated by Eq. 5 [76]:where ‘L’ is the pore length, ‘r’ is the pore radius, ‘p₁′ and ‘p₂′ are the partial pressure of the feed and permeate gases, respectively. Additionally, prediction of selectivity for the Knudsen mechanism uses square-roots from the molecular weight ratios, as shown in Eq. 6 [76]:

Poiseuille flow occurs when the driving force for the diffusion is the pressure gradient between the feed and the permeate side. It usually happens with pore sizes of around 200 nm to 3,000 nm or when the membrane pore sizes are much greater than λ (r/λ > 3). It is also known as convective diffusion, which operates opposite to Knudsen diffusion. In contrast to Knudsen diffusion, this mechanism depicts gas diffusions through the molecular collisions instead of collisions with the pore walls. In Poiseuille flow, all the component molecules pass through the pores by an average drift velocity, independent of the shape, mass, and size of the molecules [74]. The molar flux is represented in Eq. 7, illustrating the Poiseuille flow mechanism [76, 82].

The molecular sieving mechanism takes place in membranes with small pore sizes (< 7 Å). For it to occur, the membranes must possess pore sizes between the molecular sizes of the gas pair that are about to be separated. Thus, separation occurs through the differences in the size of molecules where only gas molecules with an appropriate range of kinetic diameters can permeate, preventing larger molecules from passing through the pores [76].

The surface diffusion mechanism usually exists in porous membranes where the permeating molecules are strongly attracted to the membrane surface. The absorption of molecules occurs along the length of the pore walls [76]. The driving force for the separation using this mechanism is the difference in the affinity of the pore walls towards different molecules [82].

Capillary condensation transport occurs at specific critical relative pressures in which the condensed gas has filled the pores and excludes other components from entering the pores [76]. In this mechanism, the separation of gas molecules occurs by partial condensation of any one component from the gas mixtures. For this mechanism to happen, it usually requires the mesoporous pore (pore diameter > 3.0 nm) to facilitate condensation [76]. This mechanism can achieve high selectivity as the flow of non-condensable gas molecules is blocked [82]. Also, both surface diffusion and capillary diffusion can coincide due to similar underlying conditions [83].

Another critical group of membranes in the industry is composite membranes. This kind of membrane consists of a thin and dense selective surface above a supporting porous layer [72]. Membranes in this category are typically made up of either organic (polymeric) material, inorganic material, or both (mixed matrix), depending on their purpose [79]. The transport mechanism in composite membranes is determined using the Resistance model as it consists of several barrier layers with a distinct nature [84]. Unlike porous or dense membranes, the composite membrane has an apparent discontinuity at the boundary of two adjacent barrier layers. The discontinuity can arise from the chemical structure or in the morphology of the material [85]. In this case, the permeation rate, Q_i, can be calculated as a function of the driving force, which is the pressure difference, Δp, and the resistance to the flow, R′ as shown in Eq. 8 [84].

The connection between resistances in composite membranes uses the following configurations: series, parallel, and two resistance arms [85]. Two resistances are connected in series whenever two layers of membranes are combined in a series, as shown in Fig. 7(a). Two resistances are connected in parallel whenever distinct materials exist and secured on the same layer of the membrane’s surface, as shown in Fig. 7(b). The parallel combination of two resistance models is useful when a homogeneous film of relatively high permeability is laminated on top of the membrane, as shown in Fig. 7(c).

Metals, rigid polymers (ceramics), or pyrolysed carbons are common fabrication materials of an inorganic membrane. The purpose of using metals and ceramics in fabricating an inorganic membrane is to add mechanical strength with minimum resistance for mass transfer [86]. Besides, it promotes membrane surface interaction with the desired component, which increases permeation efficiency [40]. This type of membrane is typically stable and offers high thermal resistance for CO₂ separation from hot gases. However, selectivity and permeability are low compared to other membrane varieties, which require further development [29]. Regardless of the potential offered by inorganic membranes, the technology tends towards high manufacturing costs. Furthermore, the requirement of complex fabrication techniques often limits its application [87]. Inorganic membranes also face durability issues as they are fragile and brittle, requiring extra attention when handling [37].

The notion of MMM was discovered in the early 1970s when Paul and Kemp found out that incorporating 5A zeolite to a polydimethylsiloxane (PDMS) membrane leads to a large increase in CO₂ diffusivity along with an improved separation performance. Their findings motivated scientists to further their research and exploit the concept [88]. The MMM comprises of organic polymer matrix with inorganic filler particles. The coalition of these two different membrane materials results in a composite membrane [89]. The two materials create two phases: the bulk phase (polymeric) and the dispersed phase (inorganic). Generally, materials used as inorganic fillers in MMM are zeolite, silica, or carbon nanotubes (CNTs) [90].

In the MMM, gas separation efficiency improves due to the superior adsorptive properties of inorganic fillers added into a polymeric membrane [91]. Researchers focused on composite membrane structures by combining different fillers that improved the surpass of the separation performance compared to that of just one component [89]. However, this membrane may cost around 10–100 times more than neat membranes and is not preferable in situations where neat/polymeric membranes cannot satisfy separation requirements [92].

Glassy or rubbery organic polymers are the building blocks of a polymeric membrane [93]. Polymeric membranes tend to display a trade-off situation between permeance and ideal selectivity. An increase in ideal selectivity of gas pair results in a decrease in permeance and vice versa [94]. Rubbery polymeric membranes often possess a soft, elastic, flexible membrane structure, which can operate beyond its glass-transition temperature [95]. Generally, rubbery membranes exhibit properties such as high permeance and low ideal selectivity [96].

Consequently, its low selectivity will lead to lack of application in the industry; while its glassy membrane counterpart is widely used [95]. A glassy membrane tends to be more durable due to its rigidity in contrast to rubbery membrane, as it displays properties of high selectivity but low gas permeability [94]. Some commonly used glassy membranes that are suitable for CO₂ separations are polysulfone (PSF), polyethersulfone (PES), polyimides (PI), polyamide-bethylene oxide (Pebax®), and cellulose acetate (CA) [97].

Polymeric membranes often have lower capital costs when compared to inorganic membranes or MMMs. Additionally, polymeric membranes have higher mechanical stability and are easier to fabricate [34]. However, polymeric membranes often have low thermal stability, causing unstableness at high temperatures [29]. Nevertheless, polymeric membranes are highly preferable for CO₂ separations than inorganic membranes and MMMs due to low capital costs, leading to a broader industrial application [49]. The advantageous and disadvantageous aspects of polymeric, inorganic, and MMMs are summarised in Table 2 [71].

Besides all the advantages that polymeric membranes offer, there are still several issues to address to achieve a highly efficient membrane with a more stable performance. These issues include the upper bound trade-off between CO₂ permeance and CO₂/N₂ ideal selectivity and plasticisation [98].

One of the main issues associated with polymeric membranes is the trade-off between selectivity and permeation. The two parameters possess an inverse relationship where membrane selectivity for different gas-pairs can increase when the gas permeance is coherently decreasing. This notorious phenomenon is often addressed as the trade-off between these two parameters [99]. Therefore, a breakthrough occurred when Robeson [100] proposed an empirical relationship of upper-bound between both the selectivity and the permeability, as illustrated in Eq. 9, to formulate a point of reference for membrane performance [100].where ‘P_i’ and ‘ ∝ _ij’ respectively stand for permeability and selectivity, while ‘k’ and ‘n’ are the values calculated for the upper bound linear relationship for different gas pairs [100]. Freeman [101] further studied the upper-bound curves, such as the one shown in Fig. 8, where the upper-bound curves are obtained from previous studies from Robeson [100] with specific references and plotted the data as ∝ _ij versus log P_i. His study suggested that more attention should be put on increasing selectivity through chain rigidity and the inter-chain spacing in order to exceed the upper-bound successfully [101]. The CO₂/N₂ upper-bound curve in Fig. 7 was established in 2008 rather than 1991 due to insufficient data to illustrate the correlation [102].

Plasticisation is the absorption of small chemically benign molecules that migrate between molecular chains, causing a membrane to lose stiffness [103]. The occurrence of plasticisation depends on the volume of gas entering the polymer matrix and Henry’s Law. Thus, it is significantly affected by partial pressures. Irregular behaviour where permeance increases with increasing partial pressure is usually a sign of plasticisation [104].

Other than the trade-off issue, plasticisation also has a significant negative effect on CO₂/N₂ separation. It usually occurs when there is a high content of CO₂ in the feed gas, which affects the amount of dissolved gas within the polymeric matrix [105]. This particular plasticisation triggers a trade-off phenomenon where CO₂ permeance continuously increases and selectivity declines as a function of pressure [82]. This phenomenon also occurs when the partial pressure of the penetrant elevates and exceeds a specific point. This point is commonly known as plasticisation pressure (Fig. 9) [104]. Consequently, this tendency results in a lack of selectivity of the membrane, which impairs its efficiency. Currently, a few approaches exist to subdue plasticisation caused by CO₂, such as bonding, polymer blending, thermal treatment, and incorporation of inorganic fillers [99].

Polymer blends combine multiple polymers into a new product with enhanced properties [106]. A blend membrane with the desired properties is a combination of different polymers with distinctive physicochemical and separation properties [93]. However, polymer blends tend to be thermodynamically immiscible [107, 108]. Miscibility of polymers significantly affects the morphology and the specific volume fraction within the blend membrane. Consequently, this also affects its performance [109]. Soleimany, Hosseini [110] also stated that blending a polymer susceptible to plasticisation with one of a lower tendency towards plasticisation can reduce the overall plasticisation tendency of a membrane [110]. The blending of rubbery PEG and glassy PSF polymers results in a final membrane with properties beyond the upper bound for CO₂/N₂ in Robeson’s plot [111]. Jujie et al. [112] studied the performance of PEG/PSF blend membranes [112]. They found that a drop of 36% in the CO₂ permeance occurred due to encapsulation of the PSF chain by the PEG chain, which hindered gas diffusion [112]. However, the hindrance effect on the PEG chains had minimal influence on smaller CO₂ molecules compared with bigger N₂ molecules with CO₂/N₂ selectivity of 43.0. Thus, this led to separation performance beyond the Robeson upper bound compared with the respective PEG and PSF single polymeric membranes [112].

Polyethylene glycol (PEG) is a rubbery polymer used in large quantities in the pharmaceutical, cosmetics, and food industries due to its physiological acceptance [113]. PEG has a strong affinity towards CO₂ molecules due to its polar ether chain flexibility (Fig. 10) in the polymer [4]. This flexibility allows PEG to easily dissolve acidic gases, which results in higher CO₂ permeability [114]. For this reason, PEG is a popular choice as an additive or co-polymer in a polymer blend to improve the properties of the base polymers [106, 115]. Table 3 and Fig. 11 summarise the previous results found in the literature for the PEG blend membranes compared to its pure base membrane. According to Car et al. [116], blending PEG and Pebax® with a 50/50 weight ratio led to a 100% rise in CO₂ permeance and a slight increase in CO₂/N₂ selectivity from 75 to 85 at 1 bar and 283 ^oK, respectively [116]. Other literature also reported similar trends in CO₂ permeability and CO₂/N₂ selectivity, as shown in Table 3 and Fig. 11.

Moreover, the addition of PEG has two different influences on the morphology depending on its molecular weight [117]. PEG with low molecular weight is generally in the liquid phase, which tends to act as a pore-forming agent, increasing the number of pores formed and the pore sizes. Meanwhile, PEG with high molecular weight tends to be in the solid phase and has the opposite effect i.e., reduced pore formation and decreased pore size [118]. Therefore, PEG with low molecular weight is generally preferred for gas separation as an increase in pore size and pore density enhances gas permeance but sacrifices selectivity due to upper-bound trade-off [119]. Furthermore, permeation of gases through a PEG membrane may be obstructed due to its high crystalline nature, which can reach up to 71 vol%. Hence, crystallisation should be inhibited through blending with glassy polymers to increase permeability [120].

Polyethersulfone (PES) is a phenyl-based polymer with a glassy structure widely utilised for various membrane applications due to its high thermal resistance, stability, operability, low cost, and gas separation properties [127]. The most popular property of the PES membrane is its high-temperature properties. It can operate continuously at temperatures as high as 200 °C without causing structural changes or deterioration [82]. This property makes PES suitable for the separation of hot flue gases [128, 129]. Also, PES can attract CO₂ selectively as it possesses an ether-oxygen unit (Fig. 12) that provides a binding mode for CO₂ [130]. As PES has a regular and a polar backbone, its polymeric chain is more rigid than rubbery polymers such as PEG [127]. Table 4 and Fig. 13 summarise the performance of PES blend membranes up to date [123, 130‐135]. Generally, PES blend membranes have higher selectivity towards CO₂/N₂ pairs compared to other gas mixtures such as CO₂/methane (CH₄) or O₂/N₂ [136]. This is due to its ability to condense CO₂ and CH₄ without allowing O₂ and N₂ to permeate [130]. Moreover, PES has rapid ageing properties, which could impede gas permeation over time [70]. Chung and Khean Teoh [137] found that the O₂ gas permeance of their PES membrane exhibited a 79% decay over a one year period [137]. The decay in permeance flux of a PES membrane is caused by its tendency to release internal residual stress due to polymer chain relaxation. This accelerates the polymer chain packing which leads to a faster decrease in void fractions and gas permeance [138]. Furthermore, enhancement of permeance and ageing properties of PES membranes is possible through the application of modification approaches, such as polymer blending with rubbery polymers like PEG to decrease polymer chain relaxation [130, 139].

N-Methyl-2-Pyrrolidone (NMP) is an aprotic organic solvent consisting of a 5-membered lactam that is strongly attracted to water, as illustrated in Fig. 14. The solvent is highly polar with a high boiling point (202 °C), low melting point (-23 °C), low volatility, and low viscosity with a mild amine odour [140]. Additionally, NMP is non-toxic and able to withstand high temperatures. These characteristics enabled NMP to become a very useful solvent for a number of chemical reactions where a non-reactive medium is needed [141]. NMP is used in large number of engineering applications, such as material manufacturing, coatings, farm products, telecommunications, paint stripping and washing, among many others. Mubashir et al. [142] stated that the synthesis of CA membranes using the NMP solvent resulted in a higher excess free volume and gas permeation performance [142]. Their research also found that the CO₂ permeance increased with higher NMP concentrations in the casting solution. Askari and Chung [143] reported that polyimide (PI) membrane fabricated using the NMP solvent led to a 157% rise in CO₂/CH₄ ideal selectivity compared to the one fabricated using the dichloromethane (DCM) solvent [143, 144]. These results were mainly caused by the enhancement of the hydrogen bonding in the hydroxide (OH⁻) segments of the polymers by NMP, which reduces the formation of macrovoids [142, 145]. Further comparisons of membrane performances for the same gas pairs using NMP solvents and other solvents in previous literature are presented in Table 5 and Fig. 15. Additionally, the use of NMP solvent during the membrane synthesis results in a longer solvent evaporation process, leading to a thicker membrane which enhances CO₂ permeance. In fact, the longer evaporation time allows the polymer chains to be more rigid, improving the permeation of gases across the synthesized membrane [146].

Dimethylformamide (DMF) is a transparent liquid that has been widely applied as a solvent, an additive, or an intermediate in the industry due to its strong miscibility with water and most common organic compounds. The structure of DMF is presented in Fig. 16 [152].

Since DMF is strongly polar, its application as a solvent for polar polymers with strong intermolecular forces can facilitate for hydrogen bonding [153]. DMF has a lower density and viscosity compared to water, along with a high CO₂ solubility [154]. According to Karamouz et al. [155], the use of DMF for the synthesis of PEBAX-1074 membranes led to a faster phase-inversion process and higher CO₂ permeance (233%) than the one fabricated using NMP [155]. The phase inversion process was affected by the low thermal resistance of DMF compared to NMP, which was caused by its lower boiling point (Table 6), decreasing the time required for phase-inversion [146, 156]. Ahmad et al. [152] have also shown that the PES-DMF membrane obtained a much higher CO₂ permeance (45.7 barrer) when compared to PES-NMP (1.91 barrer) [152]. This trend is consistent in other literature, as shown in Table 7 and Fig. 17. Therefore, it can be established that the use of DMF could result in a permeable membrane with higher selectivity [157].

A top-quality membrane should be able to achieve close to 100% selectivity. High selectivity is usually associated with low gas permeance due to the trade-off [161]. The fabrication process of the membrane generally has an important role. In order to produce a membrane with decent separation performance, certain parameters need to be carefully considered. Phase inversion is generally used for the synthesis of blend membranes [162]. It is a process that transforms a homogeneous polymer solution from a liquid state to a solid state [163]. This process utilises the miscibility gap in the ternary phase diagram where the polymer/solvent/non-solvent system is unstable, causing de-mixing to occur through the formation of a polymer-rich phase and a polymer-lean phase [164]. Phase-inversion may be vapour-induced, non-solvent induced or thermally induced [165].

The vapour-induced phase separation (VIPS) method is done by exposing the casted solution with the volatile solvent to a vapour non-solvent (humid air) and letting evaporation take place [70]. The non-solvent vapour is absorbed by the polymer solution, triggering the de-mixing process, and eventually forms a membrane. Regarding the non-solvent induced phase separation (NIPS) method, the casted solution is immersed into a non-solvent bath, where solvent/non-solvent interaction occurs [165]. The thermally induced phase separation (TIPS) is widely used due to its simplicity, high reproducibility, low trend to create defeats, high porosity, and the ability to create microstructures with a narrow pore size distribution. Additionally, TIPS can handle polymer polymorphism [166]. In TIPS, a “latent solvent” is utilised for the phase-inversion [164]. A latent solvent is a compound that is only able to dissolve the polymers at high temperatures but not at low temperatures [167]. Therefore, TIPS involves preparing a cast solution through polymer blending with latent solvent at high temperatures and inducing phase transition through a temperature drop [167]. A diagram of the different phase-inversion processes is represented in Fig. 18 [168].

The ternary phase diagram in Fig. 19 can be used to describe the phase inversion process [169]. Each end of the triangle represents the three components of the casting solution, such as polymer, solvent, and nonsolvent, where any point on the triangle is comprised of all the three components.

Phase-inversion starts once the casting solution with a pre-determined polymer, solvent, and non-solvent concentration (Point A) is casted on the casting machine. Then, de-solvation occurs through solvent evaporation and solvent/non-solvent exchange. This process changes the composition of polymer on ABC to the two-phase area that consists of a solid porous phase and a liquid phase [170]. The first step of de-solvation occurs through solvent evaporation and instant formation of a thin skin layer of solid polymer at the top of the cast film due to the loss of solvent [171].

At point B, a transition takes places from one phase to the two-phase area in which the mixture breaks into a polymer-rich phase and a polymer-lean phase [169]. At this moment, solvent/non-solvent exchange process occurs where non-solvent diffuses into the polymer film through the thin solid layer while solvent diffuses out. This results in a low solvent composition in the solution film which can no longer hold polymer in one phase [172].

Point C shows the net membrane at which the two-phase region is in equilibrium, where point S is the solid (polymer rich) phase and point L is the liquid (polymer-lean) phase. On the line S-L, the position of C can be used to determine the membrane morphology [173, 174]. The polymer-rich phase becomes solid at point D. At this point, the thin layer that forms during the first evaporation step becomes the top skin layer, governing the selectivity and the flux of the membrane. Meanwhile, the porous structure that forms during the solvent/non-solvent exchange step becomes the porous sub-layer, providing mechanical strength [175]. Hence, an integrally skinned asymmetric membrane is obtained from the phase-inversion process.

It should also be noted that the morphology of the final membrane is developed in the phase-inversion process is highly dependent on the time taken for it to take place [176]. The size of the polymer-lean phase dispersed in the polymer-rich phase depends on the time taken to move from B to C [169]. Shorter time leads to a denser polymer gel formed at C instead of a porous structure due to the smaller size of the polymer-lean phase. Furthermore, the time taken from B to C depends on the distance from the cast polymer film/non-solvent interface where the initial thin dense upper layer is formed [170]. The speed of solvent/nonsolvent exchange is slowed down by the formation of the dense layer and a sub-layer with sponge-like pores is formed under the dense layer. However, a more permeable finger-like sub-layer can also be formed when the influx of non-solvent is much lower than the outflux of solvents, which enhances the growth of the polymer-lean phase [177].

The synthesis of membranes through blending of polymers has grown rapidly, regardless of the miscibility of the polymer blend [178]. Polymer blending is remarkable because it provides a practical method to combine the benefits of each polymer into a single new material. Further, a continuous range of performance is predicted by changing the composition of the blend. This method offers an easy and cost-effective combination of polymers with different separation and physicochemical properties to obtain the desired superior properties which could not be achieved by each polymers individually [106, 179]. Interestingly, blends that consist of glassy and rubbery polymers have proven to yield membranes with better CO₂ separation properties [180]. This is due to the distinct methods of gas separation provided by each polymer type. Rubbery polymers carry out gas separation based on condensability. Gas separation in glassy polymers is dependent on the molecular sizes of the gas particles [130]. Furthermore, the addition of rubbery PEG into glassy PES can increase excess free volume and add flexibility to the PES chain. This increases the permeance of PES by preventing polymer relaxation [139, 181]. Akbarian et al. [123] have reported that blending PEG and PES resulted in a 26% increase in CO₂ permeance and a 64% increase in perm-selectivity for CO₂/N₂ separation [123].

Additionally, the polymer concentration in a casting solution is defined as the most important parameter for the enhancement of membrane properties through blend membranes [182]. Generally, an increase of polymer content in a dope solution should form a denser skin layer which is more selective [183]. Moreover, an asymmetric membrane with a very thin surface layer yields a membrane with a high gas permeance. However, high gas permeance in a polymeric membrane leads to a trade-off in selectivity and vice versa. Furthermore, higher PEG content in the casting solution results in an increase of rubbery segments that leads to a higher dominance of gas separation through adsorption [184]. Further, the pore-forming and pore-reducing effects of PEG of different molecular weights must also be considered due to the relationship between CO₂/N₂ selectivity and CO₂ permeance [118, 185]. PEG with low molecular weights results in higher gas permeation with the cost of selectivity, while the PEG with high molecular weights results in the opposite effect [186]. In order to synthesise a membrane with optimum selectivity and permeability, it is necessary to decide the optimum concentration of the polymer in the casting solution.

Apart from polymers, the amounts of solvents chosen for the phase-inversion process is also an important parameter for achieving the desired separation performance of the final membrane. Ahmad et al. [152] and Mubashir et al. [142] have indicated that NMP and DMF are the most suitable solvents for separation of CO₂ from N₂ due to their high CO₂/N₂ selectivity and CO₂ solubility, respectively. However, NMP has better thermal resistance when compared to DMF, which results in a longer solvent evaporation time [142, 152]. Jami'an et al. [187] reported that solvent evaporation time increases with higher concentration of NMP, leading to a decreased surface porosity and a thicker final membrane that promotes CO₂/CH₄ selectivity [187]. On the contrary, increase of DMF solvent concentration leads to a faster phase-inversion process due to its lower boiling point [155]. Isanejad et al. [146] stated that CO₂ permeance and CO₂/N₂ selectivity increases with higher DMF concentration having minimal trade-off compared to NMP due to the higher formation of macrovoids [146]. In this regard, choosing one solvent above the other results in a trade-off between permeance and perm-selectivity. This trade-off decreases the overall separation efficiency. According to Fashandi and Karimi [188], mixing of NMP and DMF solvents may improve the fabrication of the PES membrane [188]. They reported that adding DMF to PES/NMP mixture, causes the VIPS process to proceed faster, promoting formation of long thin fibres, which are typically the site for adsorption. Based on the properties of NMP and DMF, the idea of mixing the two solvents to obtain the best properties is the logical option. Additionally, the solvent concentrations should be carefully considered as different solvent concentration affects the viscosity of the casting solution, influencing the polymer–solvent interactions during the phase-inversion process [189].

Casting thickness can be considered as one the most crucial factors that should be studied for the synthesis of an effective membrane [190]. This is mainly caused by the effect of casting thickness on the membrane structure, which alters its performance and properties [191]. Vogrin et al. [192] studied the influence of cast solution thickness on the structure of cellulose acetate (CA)//acetone/H₂O system [192]. Their work revealed that a structural transformation (finger-like to sponge-like structure) occurred with respect to the membrane thickness of CA/acetone/H₂O system i.e., between 300 and 500 μm. This indicated that membrane structures may vary depending on casting thickness. Meanwhile, Ahmad et al. [147] studied the performance of the PEI membranes of different thicknesses and found that a higher casting thickness of 300 µm led to an increase in CO₂/N₂ ideal selectivity from 1.74 to 2.56, as well as a major drop in CO₂ permeance from 10.37 to 2.59 barrer [147]. The CO₂ permeance decreased significantly due to the formation of a thicker selective skin layer and larger finger-like macrovoids leading to a denser membrane which hindered permeation of CO₂. Sugu and Jawad [193] studied the effects of casting thickness on CA membranes and found a similar trend in their research where CO₂ permeance decreases with increase in casting thickness [193, 194]. However, they also discovered that a drop in both permeance and selectivity of a CA membrane occurred when the casting thickness was above 300 µm. As a result, macrovoids started to show major defects due to high non-solvent influx into the polymer matrix in thicker membranes, forming a more porous structure [195]. Their research also stated that the optimum casting thickness for their CA membrane was 300 µm, achieving the optimum CO₂/N₂ ideal selectivity of 3.01 and CO₂ permeability of 401.17 GPU. Other than gas separation properties, thermal strength and mechanical properties are also affected by casting thickness. According to Rahman et al. [196], the thermal resistance of PolyActive membrane increases with casting thickness as the melting point of the thinner PolyActive membrane of 0.2 μm was 10 °C lower than that of its thicker 8 μm counterpart [196]. Therefore, casting thickness should be thoroughly studied as it affects gas separation performance, thermal properties, and mechanical characteristics of the membrane [147].

Akbarian et al. [123] examined the influence of different PEG molecular weights and concentrations on gas separation properties, morphologies, and mechanical strengths of PES/PEG blend membranes [123]. These blend membranes were fabricated using TIPS method at a temperature of 60 °C. Figure 20 presents the cross-sectional scanning electron microscope (SEM) micrographs of the PES/PEG membranes with variation in PEG molecular weights. Figure 20 also shows that no major changes were found in the morphologies of the membranes due to the rise in PEG or molecular weights. The presence of smooth and void-free surfaces also verified the suitability of the blend, as only one phase was visible in the blend membranes [109]. This resulted from the homogeneity of PES/PEG blend in the casting solution, which caused even distribution of the two polymers across the membrane [107]. The lack of pores on the surface of the membranes established that the membranes were dense and compact [123]. The formation of a dense structure is more desirable as gas-separation requires a dense layer to build up pressure and segregate small gas molecules [197].

Despite the ability to overcome the upper bound trade-off, PES/PEG blend membranes can still be developed further in order to lower CO₂ capture costs compared to the US DOE targets. Up to date, studies on PEG/PES blend membranes have only involved DMF as a solvent. In order to improve CO₂ permeation in PEG/PES blend membranes, one promising approach is to fabricate the blend membrane using NMP/DMF solvent mixture. Results from literature have indicated that the involvement of NMP results in a longer solvent evaporation time, which in turn increases the final thickness of the membrane and its CO₂/N₂ selectivity. With regards to that, future research on PEG/PES blend membranes should consider exploring the optimum PEG/PES/NMP:DMF blend concentration as well as its optimum casting thicknesses to yield a polymeric membrane with high permeability and CO₂/N₂ selectivity. This could lead to a membrane with properties that sits even closer or beyond the upper-bound in the Robeson’s plot and encourage its application in the gas separation industry.