Recent Advances in Poly(Ionic Liquid)-Based Membranes for CO2 Separation

Poly(ionic liquid)-based membranes have been the subject of intensive research in the last 15 years due to their potential for the separation of CO2 from other gases. In this short review, different types of PIL-based membranes for CO2 separation are described (neat PIL membranes; PIL-IL composite membranes; PIL-polymer blend membranes; PIL-based block copolymer membranes, and PIL-based mixed matrix membranes), and their state-of-the-art separation results for different gas pairs (CO2/N2, CO2/H2, and CO2/CH4) are presented and discussed. This review article is focused on the most relevant research works performed over the last 5 years, that is, since the year 2017 onwards, in the field of poly(ionic liquid)-based membranes for CO2 separation. The micro- and nano-morphological characterization of the membranes is highlighted as a research topic that requires deeper study and understanding. Nowadays there is an array of advanced structural characterization techniques, such as neutron scattering techniques with contrast variation (using selective deuteration), that can be used to probe the micro- and nanostructure of membranes, in length scales ranging from ~1 nm to ~15 μm. Although some of these techniques have been used to study the morphology of PIL-based membranes for electrochemical applications, their use in the study of PIL-based membranes for CO2 separation is still unknown.


Introduction
The anthropogenic emission of greenhouse gases, such as CO 2 , to the atmosphere has been considered responsible for the accelerated global warming of our planet and, consequently, a serious threat to the future of mankind [1]. Most of the anthropogenic carbon dioxide emissions result from the burning of fossil fuels like coal, natural gas, and oil. The Paris United Nations agreement for Climate in 2015 [2] highlighted the importance of reducing CO 2 emissions from fossil fuel combustion. There are currently several carbon capture and storage (CCS) technologies being developed to control CO 2 emissions in an efficient and economical way. These major CCS approaches can be applied to pre-combustion [3,4], post-combustion [3,5], and oxyfuel combustion [3,6].
Pre-combustion capture [4] involves CO 2 removal before combustion, and takes place in three stages: (i) the hydrocarbon fuel (methane, or gasified coal) is converted primarily into H 2 and CO (synthesis gas); (ii) synthesis gas is converted into CO 2 and H 2 rich streams by the water gas shift reaction; and (iii) CO 2 is separated from H 2 and then it can be compressed into liquid and transported to a storage site. Post-combustion capture [5] involves separating CO 2 from the exhaust gases (flue gas) created by burning fossil fuels. The flue gas consists mostly of N 2 and CO 2 . The post-combustion CO 2 separation techniques may involve cryogenic separation, amines solvent-based absorption, and membrane separation. Currently, the chemical absorption of CO 2 by aqueous amine solutions occupies 90% of the market for CO 2 separation [7]. A major drawback of a typical amine-based CO 2 absorption system is that it requires being heated up to~120 • C to release the captured CO 2 and regenerate the amine solvent. This high temperature regeneration process makes the technology very energy intensive [8]. The oxyfuel combustion [6] consists in burning the fuel with nearly pure oxygen instead of air. However, as the amount of oxygen required in oxyfuel combustion is significantly larger than in pre-combustion, the CO 2 capture costs are higher.
There is a need to develop technological solutions for carbon capture and storage that are simpler to operate, more environmentally friendly, and more energy efficient, and in this context membrane separation technology is the most promising option. There are several different types of membranes currently being studied for the separation of CO 2 from different gas streams (CO 2 /N 2 , CO 2 /CH 4 , CO 2 /H 2 ), and these include polymeric membranes [9], carbon molecular sieve membranes [10][11][12][13][14], and poly(ionic liquids) (PIL) membranes [15][16][17][18].
Membranes for CO 2 separation based on poly(ionic liquids) (PIL), also named polymerized ionic liquids, were first reported in 2007 by Bara et al. [19]. PIL membranes are obtained by polymerizing ionic liquid (IL) monomers, or crosslinking polymer chains containing IL-based functional groups, forming a charged macromolecular architecture. PIL membranes combine the advantages of polymers, namely improvement of the mechanical stability and durability of the membrane, and the advantages of ILs, namely the ability to tailor the chemical and physical properties of the membrane. PILs often exhibit a significantly higher CO 2 uptake capacity than their corresponding IL monomers [20,21]. Different types of PIL-based membranes have been investigated: (i) neat PIL membranes; (ii) PIL-IL blend membranes; (iii) PIL copolymer membranes; and (iv) PIL-IL-inorganic particle mixed matrix membranes.
Neat PIL membranes, due to their compact, solid nature, typically present low gas diffusivities and permeabilities, and selectivities well below the Robeson Upper Bound. Additionally, they are usually very brittle and difficult to process into free standing films, requiring the use of expensive rigid and porous supports. Despite many studies being performed that addressed different PIL polycation structures and functionalization, it became evident that neat PIL membranes cannot achieve the gas separation performances needed to be technologically competitive. This intrinsic limitation of neat PIL membranes has stimulated the research on other types of PIL-based membranes [15][16][17].
PIL-IL composite membranes are produced by blending PIL with ILs. The addition of free ILs into PIL membranes can increase drastically the CO 2 permeability of PIL-IL membranes [22], and the membrane properties can be tuned with respect to the IL content to guarantee high CO 2 permeabilities without compromising the mechanical stability that is provided by the PIL. These membranes are very stable and can withstand large pressure gradients without leaching, because the Coulombic attraction between PIL and free ILs largely outweighs the external pressure [17]. The gas permeation properties of the free ILs largely affect the CO 2 separation performance of the composite membranes, and therefore the appropriate choice of an IL is crucial in tuning the PIL-IL membrane performance for CO 2 separation.
The gas transport in poly(ionic liquid) membranes obeys the sorption-diffusion mechanism, and the gas permeation through a membrane involves three consecutive steps: (1) gas molecules dissolve into the membrane at the high-pressure side; (2) gas molecules diffuse through the membrane under the concentration gradient across the membrane; (3) gas molecules desorb at the low-pressure side of the membrane. Therefore, the permeability (P) can be explained as being equal to the product of gas diffusivity (D) and sorption (S) in the membrane, i.e., P = D·S. The membrane ideal permeability selectivity (α i/j ), permselectivity, is the ratio of permeabilities (P i /P j ) of two permeating species (i and j) and can also be represented as the product of diffusivity selectivity (D i /D j ) and sorption selectivity (S i /S j ) [23]. PIL membranes for CO 2 separation should have both a high permeability to CO 2 and a low permeability to the other gases X of the component mixture, i.e., they should have a high selectivity to CO 2 /X mixtures. A higher permeability to CO 2 decreases the area of membrane needed to separate a certain amount of gas mixture, and a higher selectivity produces CO 2 with higher purity, both contributing to a reduction in the capital cost of the purification process. In 1991, Robeson showed that in polymeric membranes for gas separation there is a trade-off relationship between permeability and selectivity, as selectivity tends to decrease when the permeability increases [24]. In the case of polymeric membranes, this relationship is described by the so-called Robeson upper-bound, which is a straight line with a negative slope in the log-log plot of selectivity versus permeability of the more permeable gas. Later, in 2008, Robeson revised the upper bound for several gas mixtures based on newly available experimental data [25]. More recently, Araújo et al. [26] proposed a new figure of merit, the Robeson Index (θ), to characterize the separation performance of membranes. The Robeson Index is the ratio between the actual selectivity value α i,j and the one corresponding to the Robeson upper bound (α i,j (RUB)), both for a given permeability: The permeability to CO 2 in poly(ionic liquid) membranes is usually much higher than the permeability to other gases such as H 2 , N 2 , and CH 4 . These are CO 2 -selective membranes, that can permeate CO 2 and retain other gases and impurities. Furthermore, the sorption-diffusion gas transport mechanism in PIL-based membranes can be facilitated by a CO 2 -selective transport mechanism, as shown in Figure 1. In this mechanism, complexation reactions between CO 2 and CO 2 carriers increase the transport of CO 2 through the membrane. By contrast, other non-reacting gases (such as H 2 , N 2 , CO, and CH 4 ) do not experience such a transport enhancement, being transported mostly by the simple sorption-diffusion mechanism. This results in a significantly improved permeability of the membrane to CO 2 , with high selectivities towards H 2 , N 2 , CO, and CH 4 [27]. j) and can also be represented as the product of diffusivity selectivity (Di/Dj) and sorption selectivity (Si/Sj) [23].
PIL membranes for CO2 separation should have both a high permeability to CO2 and a low permeability to the other gases X of the component mixture, i.e., they should have a high selectivity to CO2/X mixtures. A higher permeability to CO2 decreases the area of membrane needed to separate a certain amount of gas mixture, and a higher selectivity produces CO2 with higher purity, both contributing to a reduction in the capital cost of the purification process. In 1991, Robeson showed that in polymeric membranes for gas separation there is a trade-off relationship between permeability and selectivity, as selectivity tends to decrease when the permeability increases [24]. In the case of polymeric membranes, this relationship is described by the so-called Robeson upper-bound, which is a straight line with a negative slope in the log-log plot of selectivity versus permeability of the more permeable gas. Later, in 2008, Robeson revised the upper bound for several gas mixtures based on newly available experimental data [25]. More recently, Araújo et al. [26] proposed a new figure of merit, the Robeson Index (θ), to characterize the separation performance of membranes. The Robeson Index is the ratio between the actual selectivity value αi,j and the one corresponding to the Robeson upper bound (αi,j (RUB)), both for a given permeability: The permeability to CO2 in poly(ionic liquid) membranes is usually much higher than the permeability to other gases such as H2, N2, and CH4. These are CO2-selective membranes, that can permeate CO2 and retain other gases and impurities. Furthermore, the sorption-diffusion gas transport mechanism in PIL-based membranes can be facilitated by a CO2-selective transport mechanism, as shown in Figure 1. In this mechanism, complexation reactions between CO2 and CO2 carriers increase the transport of CO2 through the membrane. By contrast, other non-reacting gases (such as H2, N2, CO, and CH4) do not experience such a transport enhancement, being transported mostly by the simple sorption-diffusion mechanism. This results in a significantly improved permeability of the membrane to CO2, with high selectivities towards H2, N2, CO, and CH4 [27]. In this article, we review the most relevant studies of PIL-based membranes, targeting their application in the purification of CO2 and performed within the last 5 years (from 2017 onwards). As a benchmark for performance improvements, in Figure 2 we compare In this article, we review the most relevant studies of PIL-based membranes, targeting their application in the purification of CO 2 and performed within the last 5 years (from 2017 onwards). As a benchmark for performance improvements, in Figure 2 we compare the performances reported in these last 5 years with those reported in a very comprehensive review by Marrucho et al. [15] published in 2016. The gas separation performances of PIL-based membranes for different gas pairs (CO 2 /N 2 , CO 2 /H 2 , and CO 2 /CH 4 ) are also shown in Table 1. the performances reported in these last 5 years with those reported in a very compre sive review by Marrucho et al. [15] published in 2016. The gas separation performanc PIL-based membranes for different gas pairs (CO2/N2, CO2/H2, and CO2/CH4) are shown in Table 1.    19 (a) Mixed-gas selectivity.

Neat PIL Membranes
In 2017, Nikolaeva et al. [28]  ] exhibited a pure gas permeability to CO 2 (P CO 2 ) of 8.9 barrer and a perm-selectivity α(CO 2 /N 2 ) of 26.8; this separation performance, however, was worse than that of the original CA membrane (P CO 2 = 13.8 and α(CO 2 /N 2 ) = 39.5). However, under mixed gas conditions, membranes based on P[CA][Tf 2 N] exhibited a higher permeability to CO 2 than membranes based on the original CA.
More recently, Yin et al. [30] synthesized neat, crosslinked poly(ionic liquids) (PILs) and studied the effects of the PIL Mw, crosslinker type, and the mass ratio of PIL:crosslinker on the gas separation performances of the corresponding membranes. Both the use of an ether-containing crosslinker and an increase in the crosslinker content, improved the CO 2 solubility and diffusivity, with the best neat PILs membrane, named LP(1:2), exhibiting a P CO 2 of 170 barrer and a CO 2 /N 2 permselectivity of 36. This membrane was fully amorphous, as determined by XRD. Semi-crystalline neat-PIL membranes, named LT, exhibited a much poorer permeability and permselectivity than other membranes.

PIL-IL Composite Membranes
The group of Tomé and Marrucho at the University of Lisbon, in Portugal, has been particularly active and has contributed significantly to this field of PIL-IL composite membranes for CO 2 separation [31][32][33][34][35][36]. In 2017, Tomé et al. [31]  The permeability results obtained demonstrated that the incorporation of siloxane-based ILs into PILs increases the membrane permeability to CO 2 (P CO 2 ) as well as the CO 2 /N 2 permselectivity. However, the CO 2 /CH 4 permselectivity is not significantly increased. For the same amount of IL, the membranes containing the [NTf 2 ] − anion exhibit higher permeabilities to CO 2 than membranes containing the [C(CN) 3 ] − anion. However, membranes containing the [C(CN) 3 ] − anion display higher CO 2 /N 2 permselectivities. No structural or morphological studies of these composite membranes were reported.
Later, 42 PIL-IL composite membrane combinations, prepared by the simple solvent casting technique, were tested by Teodoro et al. [32]. While the CO 2 and N 2 permeability in these PIL-IL membranes was found to be mainly controlled by gas diffusivity, the CO 2 /N 2 permselectivity was found to be controlled by the gas solubility. The study of the micro-and nano-morphologies of the membranes was not reported.
Tomé et al. [33] studied the effect of the PIL molecular weight (namely high M w (average 400-500 kDa), medium M w (average 200-350 kDa), and low M w (average < 100 kDa)), on the physical and gas permeation properties of PIL-IL composite membranes based on pyrrolidinium-based PILs, having [C(CN) 3 ] − as the counter-anion and different amounts (20,40, and 60 wt%) of free [C 2 mim][C(CN) 3 ] IL. Free standing PIL-IL membranes could only be obtained with high and medium M w PIL. PIL-IL membranes based on the medium M w PIL exhibited higher CO 2 permeabilities (14.6-542 barrer) than those based on the high M w PIL (8.0-439 barrer). The membrane permeability to CO 2 , and the CO 2 /N 2 permselectivity, both increased with the addition of IL. The performance value reported for the CO 2 /N 2 separation with the medium M w membrane with 60 wt% of IL (P CO 2 = 542 barrer and α(CO 2 /N 2 ) = 54.0) is the best reported in these last 5 years for this membrane type. No morphological studies of the membranes have been reported.
Gouveia et al. [34] prepared, by solvent casting, free-standing PIL-IL membranes using two pyrrolidinium-  3 ] IL, and achieved a P CO 2 = 438 barrer and α(CO 2 /H 2 ) = 15.1 at 20 • C, and a P CO 2 = 505 barrer and α(CO 2 /H 2 ) = 12.5 at 35 • C. The study of the membrane's morphology was not reported. More recently, the same authors tested [35] the permeability of free-standing membranes similar to these, at five different temperatures (20,35,50,65, and 80 • C) and using a multicomponent gas mixture with composition: 57.1 vol% of H 2 , 40 vol% of CO 2 , and 2.9 vol% of N 2 , at a total feed pressure of 1 bar. The same trend of gas permeabilities P CO 2 > P H 2 > P N 2 was obtained for all the membranes, like in the previous single gas experiments. Also, similarly to the results previously obtained in single gas testing [34], the membrane with the best separation performance was again based on poly( Four novel anionic poly(IL)-IL composite membranes were synthesized by Kammakakam et al. [37] using photopolymerization. Two types of photopolymerizable methacryloxy-based IL monomers (MIL-CF 3 and MIL-C 7 H 7 ) were synthesized and photopolymerized with two distinct amounts of free IL (0.5 equiv and 1 equiv) containing the same cation ([C 2 mim][Tf 2 N]) and 20 wt% PEGDA cross-linker. The best separation performance for the three different gas separations: CO 2 /H 2 , CO 2 /N 2 , and CO 2 /CH 4 , was attained with the membrane (MIL-C 7 H 7 /PEGDA(20%)/IL(1 equiv.)), namely, P CO 2 = 20.4 barrer and α(CO 2 /H 2 )~4.1, α(CO 2 /N 2 )~87, and α(CO 2 /CH 4 )~119. Wide-Angle X-ray Diffraction (WAXD) of the membranes revealed that they were mostly amorphous as no sharp peaks could be observed. The main halo observed in the four membranes was used to determine the interchain spacing using the Bragg equation, and the largest d-spacing of 6.63 Å was obtained for the best performing membrane.
In a continuation of previously reported work, Yin et al. [30] prepared crosslinked PIL-IL composite membranes and studied the effect of the PIL M w , crosslinker type, and mass ratio of PIL:crosslinker on the gas separation performance of the resulting membranes. The optimized membrane achieved a P CO 2 of 2070 barrer and a CO 2 /N 2 permselectivity of 24.6.
In 2021, Vijayakumar et al. [38] synthesized the poly(ionic liquid) 1-bromohexyl-1 methylpiperidinium bromide (Br-6-MPRD) grafted poly(2,6 dimethyl 1,4 phenylene oxide) (ILPPO). Then free-standing PIL/IL composite membranes were made by mixing the PIL with different amounts of free Br-6-MPRD (0, 2, 5, and 10 wt%), and these membranes are named (ILPPO/Br-6-MPRD-X, where X = 2, 5, 10). Permeability measurements showed that the permeability to CO 2 (P CO 2 ) increased significantly from 69.58 barrer in the neat ILPPO membrane, to 907.20 barrer in the PIL/IL composite membrane with 2 wt% of IL. However, further increases in the IL content led to a drastic reduction in the permeability to CO 2 -this drastic reduction was attributed to a space-filling effect. SEM was used to study the cross-section of the membranes and all of them looked very homogeneous with no visible phase domains.

PIL-Polymer Blend Membranes
In 2017, Zhang et al. [39] synthesized a series of semi-interpenetrating polymer network (semi-IPN) membranes by incorporating linear polyvinyl acetate (PVAc) into a crosslinked poly (ionic liquid) (c-PIL) network. For the preparation of PVAc/c-PIL semi-IPN membranes, films were first cast from a solution containing appropriate amounts of PVAc and PIL, and a cross-linker and photo-initiator. Then the cast films were placed under a UV-lamp to promote the cross-linking, and finally the cross-linked PVAc/c-PIL membranes were dried in a vacuum oven. For reference, similar uncross-linked membranes with linear PIL (l-PIL) were prepared by the same procedure, just lacking the cross-linker and photo-initiator (membranes PVAc/l-PIL). A structural analysis of the cross-section of the membranes was made using SEM (magnification between ×500 and ×10,000). While the pure c-PIL membrane exhibited no visible morphological structure (Figure 3e), a microscale morphology, with an average domain size of 2 µm, of the minor phase was observed in the PVAc/c-PIL membranes with 30 and 60 wt% c-PIL (Figure 3b,d), respectively. Furthermore, an interconnected co-continuous microstructure of two phases was observed in the PVAc/c-PIL membranes with 50 wt% c-PIL (Figure 3c). By contrast, the uncross-linked membranes (PVAc/l-PIL) exhibited a significant macroscopic phase separation for l-PIL contents above 10 wt%. The permeability of the PVAc/c-PIL semi-IPN membranes to CO 2 (P CO 2 ) and to N 2 (P N 2 ) was shown to increase with the amount of c-PIL in the membranes. The membrane reached its best permeability (P CO 2 ) of 36.1 barrer and permselectivity (CO 2 /N 2 ) of 59.6, when the c-PIL content was 50 wt%. For c-PIL contents above 60 wt% the PVAc/c-PIL semi-IPN membrane was very brittle and could not be tested. The performance of the PVAc/c-PIL-50 semi-IPN membrane was also studied at different temperatures, and the permeability to CO 2 increased with increasing temperature, while the CO 2 /N 2 permselectivity decreased. the uncross-linked membranes (PVAc/l-PIL) exhibited a significant macroscopic phase separation for l-PIL contents above 10 wt%. The permeability of the PVAc/c-PIL semi-IPN membranes to CO2 (PCO2) and to N2 (PN2) was shown to increase with the amount of c-PIL in the membranes. The membrane reached its best permeability (PCO2) of 36.1 barrer and permselectivity (CO2/N2) of 59.6, when the c-PIL content was 50 wt%. For c-PIL contents above 60 wt% the PVAc/c-PIL semi-IPN membrane was very brittle and could not be tested. The performance of the PVAc/c-PIL-50 semi-IPN membrane was also studied at different temperatures, and the permeability to CO2 increased with increasing temperature, while the CO2/N2 permselectivity decreased.

PIL-Based Block Copolymer Membranes
Nellepalli et al. [40] synthesized a series of imidazolium-based homo-PILs and copolymer-PILs, having different side chain groups (ethyl, pentyl, benzyl, and naphthyl) at the imidazolium ring. The membrane forming ability of these homo-PILs and copolymer-PILs was tested both in their neat state and when mixed with different amounts of free [C2mim][NTf2] IL. Among all the tested combinations, only three originated stable and homogeneous free standing solid membranes: (i) poly(ViPenIm)(Sty) NTf2 with 10 wt% of IL; (ii) poly(ViBenIm)(Sty) NTf2 with 25 wt% of IL; and (iii) poly(ViNapIm)(Sty) NTf2 with 30 wt% of IL. These three membranes, with IL contents of 10 wt%, 25 wt%, and 30 wt%, exhibited permeabilities to CO2 (PCO2) of 21.6, 16.5 and 24.5 barrer, respectively, and CO2/N2 selectivities of 31.7, 32.9 and 34.4, respectively. The anomalous variation of PCO2 with the amount of the IL content was hypothesized as being due to specific structural features of the copolymers-however no structural characterization was performed to elucidate this. The selectivities observed were attributed to the much higher solubility of CO2 in the membranes relative to N2. The difference in the chemical structures of the co-PILs did not significantly affect the performance of the membranes.
In 2021, Wang et al. [41] compared membranes based on the synthesized block copolymer-grafted SiO2 particle brush SiO2-g-PMMA-b-PIL, with membranes based on the homopolymer SiO2-g-PIL. The mechanical and permeability tests demonstrated that by introducing a PMMA segment on a PIL-based block copolymer membrane, the mechanical properties of the membrane can be improved without compromising its gas separation performance.

PIL-Based Block Copolymer Membranes
Nellepalli et al. [40] synthesized a series of imidazolium-based homo-PILs and copolymer-PILs, having different side chain groups (ethyl, pentyl, benzyl, and naphthyl) at the imidazolium ring. The membrane forming ability of these homo-PILs and copolymer-PILs was tested both in their neat state and when mixed with different amounts of free [C 2 mim][NTf 2 ] IL. Among all the tested combinations, only three originated stable and homogeneous free standing solid membranes: (i) poly(ViPenIm)(Sty) NTf 2 with 10 wt% of IL; (ii) poly(ViBenIm)(Sty) NTf 2 with 25 wt% of IL; and (iii) poly(ViNapIm)(Sty) NTf 2 with 30 wt% of IL. These three membranes, with IL contents of 10 wt%, 25 wt%, and 30 wt%, exhibited permeabilities to CO 2 (P CO 2 ) of 21.6, 16.5 and 24.5 barrer, respectively, and CO 2 /N 2 selectivities of 31.7, 32.9 and 34.4, respectively. The anomalous variation of P CO 2 with the amount of the IL content was hypothesized as being due to specific structural features of the copolymers-however no structural characterization was performed to elucidate this. The selectivities observed were attributed to the much higher solubility of CO 2 in the membranes relative to N 2 . The difference in the chemical structures of the co-PILs did not significantly affect the performance of the membranes.
In 2021, Wang et al. [41] compared membranes based on the synthesized block copolymer-grafted SiO 2 particle brush SiO 2 -g-PMMA-b-PIL, with membranes based on the homopolymer SiO 2 -g-PIL. The mechanical and permeability tests demonstrated that by introducing a PMMA segment on a PIL-based block copolymer membrane, the mechanical properties of the membrane can be improved without compromising its gas separation performance.
Very recently, the block copolymers [NBM-mIM] [Tf2N] and [NBM-ImCnmIm] [Tf 2 N] 2 (n = 4 and 6) were synthesized by Ravula et al. [42], and then they were cast into thin membranes and tested for their permeability to several pure gases (CO 2 , N 2 , CH 4 , and H 2 ). However, all the prepared membranes displayed very modest permeabilities to CO 2 . The cross-sections of the membranes were studied by SEM. The membranes were also subjected to WAXD analysis, that revealed their essentially amorphous nature.

PIL-Based Mixed Matrix Membranes (MMM)
PIL-based mixed matrix membranes (MMMs) combine the benefits of both polymeric PILs and inorganic materials, and they have gained a special interest in the research community over the last few years as a strategy to circumvent the performance trade-off shown by polymeric membranes. [BETI] IL were prepared with different loadings of MOF-5 (between 10 and 30 wt% of the total mass) and were tested for their CO 2 /CH 4 separation performance by Sampaio et al. [45]. The CO 2 single gas permeabilities of the prepared membranes are shown in Table 1 as a function of the MOF-5 loading. Although P CO 2 increased with the loading of MOF-5, all the performances were well below the Robeson upper bound limit. Furthermore, the addition of MOF-5 originated brittle membranes. SEM images of the cross-sections of the membranes revealed dense morphologies and a uniform dispersion of the MOF into the PIL/IL matrix on a micrometer length scale.

Advanced Structural Characterization of PIL-Based Membranes
Structural and morphological studies of PIL-based membranes for CO 2 separation have been so far very much limited to the use of some relatively common techniques such as SEM [28,29,38,39,42,44,46] and TEM [41], as well as some WAXS [37,42], with their associated limitations. Furthermore, the reports on the use of small angle X-ray scattering (SAXS) are also still very scarce [38].
Electrons and X-rays interact with atomic electrons, meaning that the greatest contrast is obtained between elements with significantly different atomic numbers. Therefore, in PIL-IL composites the contrast between different phases-as observed using electron microscopy or X-ray scattering techniques-may be weak due to the small difference in scattering length density between the different phases containing atoms with similar atomic numbers. In this sense, neutron scattering can be used as a technique complementary to Xray scattering, because neutrons interact with the atomic nucleus and the neutron−nucleus interaction can be very different between nuclei of similar atomic numbers. Therefore, the X-ray and neutron scattering profiles can be very different and provide complementary information. Furthermore, whereas X-ray scattering of a system typically produces only one scattering profile of the total structure, neutron scattering using selective isotopic substitutions (isotopes just differ in their nucleus) can produce many different neutron scattering profiles that highlight different atoms, leading to a more complete picture of the total structure, this latter approach, where H and D atoms are replaced, being termed selective deuteration.
Contrary to the structure of PILs for CO 2 separation membranes, the micro and nanostructures of PILs has been previously studied in greater detail for some other applications, such as for electrochemical applications. An array of techniques has been used in those studies, including a combination of molecular dynamics (MD) simulations, X-ray scattering (WAXS and SAXS), and neutron scattering (SANS), besides some other more common techniques such as electron microscopy (SEM and TEM). These studies are briefly reviewed below.
The Paddison group has developed important work on elucidating the nanomorphologies of PILs for electrochemical applications [47][48][49]. In 2016, using atomistic molecular dynamics (MD) simulations this group investigated the structural properties of a homologous series of poly(nalkyl-vinylimidzolium bistrifluoromethylsulfonylimide) poly-(nVim Tf 2 N) [47]. The backbone-to-backbone distance, and the size of the nonpolar nanodomains, were shown to increase with the alkyl chain length at a rate of 1 Å/CH 2 . Excellent agreement was obtained between the MD simulations and the results from the X-ray scattering experiments. The alkyl chain length dependence of backbone-to-backbone distance on the complete homologous series of PIL poly(C n Vim Tf 2 N) (n = 2-8), was later studied by the same group [48], using extensive atomistic MD simulations. Excellent agreement was once again observed between the atomistic simulations and the experimental X-ray scattering results, and the backbone-to-backbone correlation length was once again shown to increase with the alkyl chain length. In another similar work performed by the same group [49], atomistic simulations revealed a progressive change in the nanoscale morphology of imidazolium-based PILs with increasing alkyl chain length: discrete apolar islands form initially inside the continuous polar network and then grow beyond the percolation threshold, finally forming a bi-continuous nanostructure of polar and apolar domains. These complex 3D networks were considered very important for the ionic conductivity.
Doughty et al. [50], using WAXS in conjugation with DFT calculations, have shown that the size of the mobile anions has a large impact on chain packing in PILs. A wellpacked structure is formed in the presence of larger mobile ions, while smaller ions frustrate the packing of PILs chains.
Very recently, Corvo et al. [51] studied the nanostructure of poly(1-vinyl-3-alkylimidazolium)s in the bulk, with varying alkyl side-chain lengths (n from n = 1 to n = 10) and counter-anions, using WAXS and SANS (small-angle neutron scattering). The WAXS patterns of bulk PIL features three peaks in the Q range 0.1 Å −1 < Q < 2.5 Å −1 , as shown in Figure 4a. The low-Q peak (<0.5 Å −1 ) is assigned to the distance between two neighboring macromolecular chains, and when the alkyl chain length increases it sharpens, becomes more intense, and shifts to lower-Q values (higher distance between macromolecular chains). The intermediate peak shows a very small dependence on the alkyl side-chain length and is attributed to the correlation length of the counter-anion network. The higher-Q peak is ascribed to close contact between alkyl side chains inside the alkyl domain. Modelling of the SANS data, shown in Figure 4c,d, allowed the determination of the influence of the alkyl chain length on the radius of gyration R g , on the chain cross section and on the backbone-to-backbone distance. The backbone-to-backbone distance exhibited a non-monotonic variation with n, as side-chains tend to interpenetrate as their length increases. more intense, and shifts to lower-Q values (higher distance between macromolecular chains). The intermediate peak shows a very small dependence on the alkyl side-chain length and is attributed to the correlation length of the counter-anion network. The higher-Q peak is ascribed to close contact between alkyl side chains inside the alkyl domain. Modelling of the SANS data, shown in Figure 4c,d, allowed the determination of the influence of the alkyl chain length on the radius of gyration Rg, on the chain cross section and on the backbone-to-backbone distance. The backbone-to-backbone distance exhibited a non-monotonic variation with n, as side-chains tend to interpenetrate as their length increases. SANS, supported by MD simulations and SAXS, has also been used in recent years to study the bulk-phase structure of ionic liquid mixtures [52][53][54]. Using the isotopic contrast variation technique, it was determined that the structure of these ionic liquid mixtures changes substantially as a function of composition. Evidence has been emerging in recent years that points to the existence of an enhanced level of structural complexity in IL-based systems [54], that results in hierarchical complex morphologies that play a role in the bulk properties of the systems (be it membranes, or others).
Despite all these advanced characterization studies performed on elucidating the nanostructure of PIL-based membranes for electrochemical applications, as well as eluci- SANS, supported by MD simulations and SAXS, has also been used in recent years to study the bulk-phase structure of ionic liquid mixtures [52][53][54]. Using the isotopic contrast variation technique, it was determined that the structure of these ionic liquid mixtures changes substantially as a function of composition. Evidence has been emerging in recent years that points to the existence of an enhanced level of structural complexity in IL-based systems [54], that results in hierarchical complex morphologies that play a role in the bulk properties of the systems (be it membranes, or others).
Despite all these advanced characterization studies performed on elucidating the nanostructure of PIL-based membranes for electrochemical applications, as well as elucidating the structure of ionic liquid mixtures, similar detailed morphological studies performed on PIL-based membranes for CO 2 separation are still missing in the literature. As far as we know, neutron scattering techniques with contrast variation, such as small angle neutron scattering (SANS) and Spin-Echo-SANS (SESANS), have still not been used to study the micro and nanostructures of PIL-based membranes for CO 2 separation. SANS probes the~1-300 nm length scales, albeit in reciprocal space, whereas SESANS probes the 50 nm-15 µm length scale range [55,56]. Several studies have explored polymer systems such as colloids [57] and fibrous calcium caseinate gels [58]. However, it is only recently that new instrumentation at large facilities [59], and straightforward analysis techniques, have emerged for structural determination [60]. There also exist possibilities to use this technique to look at the kinetics [61], and a similar approach could be undertaken to look at the kinetics of film formation.
It is our belief that the design of PIL-based membranes for CO 2 separation would greatly benefit from a better micro, nano, and molecular level understanding of microand nano-structure-performance relationships, which is still clearly missing. Relating the micro-and nano-structural features of the PIL-based membranes, to the understanding of their CO 2 separation performances, remains therefore an open question up to now. Probing in more detail the structure of PIL-based membranes for CO 2 separation will require the synergic use of advanced experimental (X-ray and neutron scattering) and computational tools.

Conclusions
PIL-based membranes have been studied for the separation of CO 2 from gas mixtures. Among the different CO 2 separations considered, namely CO 2 /CH 4 , CO 2 /H 2 , and CO 2 /N 2 , PIL-based membranes demonstrate more promising results in the separation of CO 2 from N 2 . The best performing PIL-based membranes for CO 2 /N 2 separation developed at laboratory scale in the last 5 years, exhibited a permeability of 542 barrer and a selectivity of 54, which is slightly above the 2008 Robeson upper bound.
The very high number of possible combinations of cations and anions creates the expectation of the large potential of this technology for CO 2 separation. However, there seems to exist still a very large margin for further development. One particular aspect that requires much research improvement is related to the relationship between the microstructure and nanostructure of the PIL-based composite membranes, and their corresponding separation performances-this relationship is still poorly understood. In fact, studies regarding the relationship between PIL-based membranes' nanostructure morphologies and their gas separation properties are still very scarce in the literature. The nanostructured morphologies are likely to impact on the transport properties, as well as on the mechanical properties of the membranes. To achieve this understanding, the micro-and nanostructures of the composite membranes should be studied with greater detail, using a battery of X-ray and neutron scattering techniques combined with theoretical ab-initio and molecular dynamics simulations. Although these advanced characterization techniques have been used in studies of PIL-based membranes for electrochemical applications, their use in the study of PIL-based membranes for CO 2 separation has still not been reported.