Recent Advances in Mixed-Matrix Membranes for Light Hydrocarbon (C1–C3) Separation

Light hydrocarbons, obtained through the petroleum refining process, are used in numerous applications. The separation of the various light hydrocarbons is challenging and expensive due to their similar melting and boiling points. Alternative methods have been investigated to supplement cryogenic distillation, which is energy intensive. Membrane technology, on the other hand, can be an attractive alternative in light hydrocarbon separation as a phase change that is known to be energy-intensive is not required during the separation. In this regard, this study focuses on recent advances in mixed-matrix membranes (MMMs) for light hydrocarbon (C1–C3) separation based on gas permeability and selectivity. Moreover, the future research and development direction of MMMs in light hydrocarbon separation is discussed, considering the low intrinsic gas permeability of polymeric membranes.


Introduction
Light hydrocarbons (specifically methane, acetylene, ethylene, ethane, propylene, and propane) are major raw materials in petrochemical industries. In particular, the increasing demand for polyethylene, poly(vinyl chloride), ethane, and polypropylene, which are used in the production of everyday materials [1], is driving the substantial production of ethylene and propylene (~191 and 120 million tons, respectively, worldwide in 2019) [2][3][4]. These two hydrocarbons are typically produced through the steam cracking of hydrocarbon sources, such as naphtha, natural gas, coal, and shale gas [5]. Irrespective of the source, the purities of the extracted ethylene and propylene streams should ideally be at least 99.5% [6,7]. The presence of acetylene (>40 ppm) in the ethylene stream, for instance, hampers ethylene polymerization because acetylene could poison the Ziegler-Natta catalyst [8]. This polymerization process is essential because it nullifies contaminants that may trigger side reactions, which could substantially reduce the molecular weight of the resulting polymers [9].
High-pressure cryogenic distillation, a mature technology, has been adopted in industrial operations for light alkene/alkane separation. For example, effective ethylene/ethane separation requires an operating temperature and pressure of −160 • C and 23 bar, respectively, given their similar melting and boiling points (Table 1) [10]; specifically, the boiling and melting points of ethylene and ethane differ by 14 and 15 • C, respectively. Such energy-intensive operating conditions are required for propylene/propane separation as well, because the melting and boiling points of this pair are even more similar (differing by only 6 and 3 • C, respectively). In particular, 75-85% of the input energy is utilized for  Mixed-matrix membranes (MMMs), which combine the advantages of polymeric and molecular sieves (adsorbents), have been a focus of research attention. Numerous papers have reviewed the research on membranes for hydrocarbon separation [2,5,[33][34][35]; therefore, the present study focuses mainly on the performance of MMMs and composite membranes in separating C 1 to C 3 hydrocarbons. In general, the utilization of MOFs as potential fillers for light hydrocarbon separation has been the main focus due to their higher flexibility in both pore size control and post-synthetic functionalization compared to conventional porous materials, such as zeolites [23]. In addition, their performance with respect to the upper bound limit (C 2 H 4 /C 2 H 6 and C 3 H 6 /C 3 H 8 ) [36,37] is discussed. Finally, insights on the future research and development direction of MMMs in light hydrocarbon separation are shared.

C 2 Hydrocarbon Separation
Membranes in C 2 hydrocarbon separation have been mainly investigated with respect to their potential in C 2 H 4 /C 2 H 6 separation (Table 2). However, although polymeric membranes, especially those commercially available, are robust and processable, they do not exhibit high permeabilities in ethylene/ethane separation. For example, Naghsh et al. [38], Davoodi et al. [39], and Jeroen et al. [40] demonstrated that the C 2 H 4 permeabilities of Matrimid ® , cellulose acetate (CA), and polyimide (P84) under ambient conditions were merely 0.26, 0.051, and 0.049 barrer, respectively, which is attributable to their low fractional free volume (FFV). C 2 H 4 /C 2 H 6 selectivities, on the other hand, were reported to be moderate, with values of 3.25, 2.27, and 4.00, respectively. It is expected that the reported selectivity is comparatively higher than those of polymers with large FFV, such as 6FDAbased polymers and PIM-1, as observed in CO 2 separation [24,28,41]. Therefore, these low intrinsic gas permeabilities make the lab-scale (gas chromatography) investigation of such membranes challenging, as it requires specific devices (high-accuracy pressure transducers or thermal conductivity detectors) or conditions (high feed pressure). The advantages of incorporating filler materials into polymeric membranes have been investigated. For instance, Naghsh et al. [38] and Davoodi et al. [39] incorporated amorphous silica nanoparticles into Matrimid ® and CA membranes, respectively, and investigated their performance in C 2 H 4 /C 2 H 6 separation; the incorporation of 20 wt% silica in these two polymeric membranes was found to increase C 2 H 4 permeability by 58% and 76%, with no substantial change in C 2 H 6 permeability, in comparison with pure polymer matrix. Although silica nanoparticles are not porous, the residual hydroxyl (-OH) functionalities make interactions with polar C 2 H 4 molecules feasible. Furthermore, the incorporation of silica nanoparticles disrupts polymer packing, which increases the FFV and facilitates rapid gas transport.
These advantages notwithstanding, the use of non-porous silica nanoparticles constrains improvements in C 2 H 4 /C 2 H 6 separation performance due to their lack of intrinsic porosity, highlighting the need for the incorporation of porous materials into separation membranes. In this regard, MOFs with coordinative unsaturated open metal sites are of particular interest. These metal sites interact favorably with the π-bond in C 2 H 4 , thus facilitating C 2 H 4 adsorption. Bachman et al. [42] identified M 2 dobdc (M = Co, Fe, Mg, Mn, Ni, Zn), which possess reasonable C 2 H 4 /C 2 H 6 selectivity (c.a. 8-12 at 50/50 vol/vol C 2 H 4 /C 2 H 6 and 35 • C, based on ideal adsorbed solution theory [IAST] calculations), as a feasible adsorbent. Studies on MMMs incorporated with M 2 dobdc nanocrystals have revealed that the addition of Mg 2 dobdc and Mn 2 dobdc, unlike Ni 2 dobdc and Co 2 dobdc, creates undesirable interfacial gaps in the polymer matrix, leading to non-selective gas transport [42,43]. Nevertheless, given their high C 2 H 4 adsorption, these M 2 dobdc-incorporated MMMs have exhibited large increases in C 2 H 4 permeability, making it increasingly likely to surpass the upper bound limit for C 2 H 4 /C 2 H 6 separation. The advantages of incorporating MOFs with open metal sites were also demonstrated by Chuah et al. [44], who used HKUST-1 as the adsorbent; specifically, the incorporation of 20 wt% HKUST-1 in 6FDA-TMPDA membranes substantially improved their C 2 H 4 permeability, pushing their performance close to the C 2 H 4 /C 2 H 6 upper bound limit ( Figure 2).
MMMs have also been applied in acetylene separation. However, adsorbents with high acetylene/ethylene selectivity are rare [6]. MOFs are generally more favorable for acetylene/ethylene separation than conventional adsorbents, such as zeolites and activated carbon, because MOFs offer more systematic control over pore size (through the selection of ligands with varying strut lengths) and are compatible with pre-and post-synthetic functionalization. Qian et al. [45] recently demonstrated the C 2 H 2 /C 2 H 4 selectivity of two types of MOF, namely SIFSIX-2-Cu-i and ZRFSIX-2-Ni-i ( Figure 3a); this favorable performance is attributable to the presence of MF 6 2− moieties (M = Cu, Ni), which enables strong interactions with C 2 H 2 . Furthermore, poly(ionic liquids) have also been used in this manner given their reasonable C 2 H 2 permeability and filler-polymer compatibility [46][47][48]. Specifically, single-gas permeation experiments have shown that 30 wt% SIFSIX-2-Cu-i and ZRFSIX-2-Ni-i in poly(ionic liquids) improve C 2 H 2 permeability by 129% and 100%, respectively (Figure 3b), without a substantial change in C 2 H 2 /C 2 H 4 selectivity.

Figure 2.
Performance of mixed-matrix membranes (MMMs) with 10 and 20 wt% HKUST-1 loading in ODPA-TMPDA and 6FDA-TMPDA membranes; 20 wt% HKUST-1, whose performance is comparable with those of other membranes reported in previous research, has the potential to surpass the upper bound limit of C 2 H 4 /C 2 H 6 separation. Reprinted with permission from [44], Creative Commons Attribution 4.0.
The feasibility of using MMMs for C 2 H 6 /CH 4 separation has also been explored in research. In contrast to MOFs, which are effective at C 2 H 4 /C 2 H 6 separation, the applicability of zeolite adsorbents in this separation is relatively less common than in CO 2 separation (e.g., post-combustion or pre-combustion carbon capture) [49,50] due to the comparable polarizability and quadrupole moment of C 2 H 4 and C 2 H 6 molecules (Table 1). Hence, C 2 H 6 /CH 4 separation was investigated instead due to a more substantial difference in their polarizabilities. Tirouni et al. [51] investigated C 2 H 6 /CH 4 separation using polyurethane MMMs incorporated with zeolite 4A and ZSM-5. With this zeolite-based approach, the polarizability of C 2 H 6 is higher than that of CH 4 ; accordingly, the permeability of C 2 H 6 is also higher than that of CH 4 . However, this approach is not promising due to its low C 2 H 6 /CH 4 selectivity, which in turn is attributable to the poor compatibility between the polymer matrix and the zeolite particles; specifically, the zeolite-polymer configuration has been reported to exhibit the sieve-in-a-cage morphology [52][53][54][55]. These shortcomings can be potentially overcome via post-synthetic treatments, such as the incorporation of silane coupling agents and post-synthetic amine grafting. However, these treatments may jeopardize the effective transport of the desired gas through the membranes [23]. Reprinted with permission from [45], copyright 2020, Elsevier.  [1] Permeance in GPU; [2] incorporated with reactive ionic liquid (RIL); [3]

C 3 Hydrocarbon Separation
C 3 hydrocarbon separation has primarily been examined for the propylene/propane (C 3 H 6 /C 3 H 8 ) gas pair (Table 3), with few studies examining the C 3 H 8 /CH 4 gas pair. As discussed in Section 2, commercial polymer membranes do not exhibit intrinsically high C 3 H 6 permeability and C 3 H 6 /C 3 H 8 selectivity due to their considerably larger kinetic diameter compared with other common gaseous molecules, such as CO 2 , N 2 , C 2 H 4 , and C 2 H 6 . Naghsh et al. [38] and Davoodi et al. [39] demonstrated that the C 2 H 6 permeabilities of Matrimid ® and CA under ambient conditions were as low as 0.09 and 0.046 barrer, respectively, due to their low FFV. Thus, it is highly desirable to use polyimide membranes with high intrinsic C 3 H 6 permeabilities, such as 6FDA-based polymers and rubbery polymers (e.g., polydimethylsiloxane (PDMS]) and cross-linked poly(ethylene oxide) (XLPEO)) to achieve sufficient flux for gas separation.
Regarding other porous materials, the effectiveness of ZIF-8 (ZIF = zeolitic imidazolate framework), which is generally stable toward water and organic solvents (e.g., methanol), in C 3 H 6 /C 3 H 8 separation has been investigated. Based on single-crystal X-ray diffraction data, the aperture of ZIF-8 has been reported to be 3.4 Å [63], which is fairly flexible compared with rigid frameworks (e.g., zeolites and zeotype materials). Thus, considering the flexibility of the ligands present in ZIF-8, Zhang et al. [64] estimated (revised) the effective aperture size. According to the adsorption kinetics of C 3 H 6 and C 3 H 8 , C 3 H 6 has a shorter equilibration time than C 3 H 8 (Figure 4a), indicating a potential means of size discrimination between C 3 H 6 and C 3 H 8 . However, a precise cut-off for the ZIF-8 aperture size cannot be determined. As an alternative, based on the diffusivities of gaseous molecules against pure ZIF-8 membrane (as calculated from the Maxwell model [23,65]), the diffusion selectivity of C 3 H 6 /C 3 H 8 has been reported to exceed 100 at 35 • C ( Figure 4b); this result indicates the feasibility of using the ZIF-8 aperture size to achieve effective molecular sieving between C 3 H 6 and C 3 H 8 . As evident from Table 1, kinetic diameter alone is not satisfactory for characterizing the molecular diffusion for selective C 3 H 6 transport. Thus, as suggested by Zhang et al. [64], a hybrid molecular dimension scale (i.e., a combination of kinetic and van der Waals diameters) is required to effectively explain the feasibility of ZIF-8 to perform effective C 3 H 6 /C 3 H 8 separation. Numerous researchers have investigated the use of ZIF-8 for effective C 3 H 6 /C 3 H 8 separation. The first widely known investigation of ZIF-8 as a filler in MMMs for C 3 H 6 /C 3 H 8 separation was conducted by Zhang et al. [66]. Based on the gas adsorption isotherms and adsorption kinetics of both C 3 H 6 and C 3 H 8 , their sorption capacities and fractional uptake are comparable. The incorporation of 48 wt% ZIF-8 in 6FDA-TMPDA membranes was found to increase the C 3 H 6 permeability and C 3 H 6 /C 3 H 8 selectivity by 258% and 150%, respectively. These conflicting results are attributable to the difficulty of accurately characterizing the adsorption kinetics (i.e., determining the mass and heat transfer resistance and particle size) during the equilibrium gas adsorption measurements [12,67].
Efforts have also been made to improve the separation performance of ZIF-8-containing MMMs. Specifically, polymers other than 6FDA-TMPDA [68] have been investigated with the aim of surpassing the upper bound limit for C 3 H 6 /C 3 H 8 separation. For example, Ma et al. [69] examined the use of PIM-6FDA-OH, which possesses higher C 3 H 6 /C 3 H 8 selectivity (30 vs 12 for 6FDA-TMPDA) due to the hydroxyl (-OH) group. As evidenced by X-ray photon spectroscopy (XPS) data (Figure 5a), hydrogen bonding (N . . . H-O) between the nitrogen moiety in the ZIF-8 ligand and the -OH group in the polymer improves the compatibility between the filler and the polymer (Figure 5b), in turn improving the C 3 H 6 permeability and C 3 H 6 /C 3 H 8 selectivity by 1.086% and 43%, respectively (Figure 5c), at 65 wt% ZIF-8. The incorporation of ZIF-8 in rubbery polymers, such as PDMS and XLPEO, has also been explored given their high C 3 H 6 permeability [70,71], but these compounds have yet to be used in industrial operations due to their low inherent brittleness [24]. Other porous materials have also been explored for C 3 H 6 /C 3 H 8 separation. For instance, cobalt-substituted ZIF-8 (ZIF-67) has been proposed to be more efficient in this regard than ZIF-8 because the Co-N bond in ZIF-67 is stiffer than the Zn-N bond in ZIF-8. In particular, due to the higher electronegativity of Co compared with Zn, the generation of a stiffer ionic bond has the potential to restrict the flipping motion of the organic ligands present in the respective MOFs (Figure 6a) [72][73][74]. An et al. [75] demonstrated that the incorporation of 20 wt% ZIF-67 in 6FDA-TMPDA improved the C 3 H 6 /C 3 H 8 selectivity by 165%, whereas a comparable ZIF-8 incorporation resulted in an 83% improvement. Furthermore, Oh et al. [76] developed Co 2+ and Zn 2+ mixed-metal hybrids (ZIF-8-67) to take advantage of the benefits of both fillers. A comparison study of ZIF-8, ZIF-67, and ZIF-8-67 revealed that ZIF-8-67 could improve C 3 H 6 permeability to a much larger extent than could the ZIF-8 and ZIF-67 frameworks (Figure 6b). The use of composite fillers, such as ZIF-67 with porous graphene oxide (GO) [77], has also been explored. The creation of three-dimensional architectures with MOFs and GO has been reported to increase the accessible surface area due to the favorable interactions between the functional groups in GO and the ligands in MOFs [78][79][80]. Frameworks other than those involving ZIFs have also been investigated in previous research. For example, Liu et al. [81] examined the potential of Zr-fum-fcu-MOF crystals in C 3 H 6 /C 3 H 8 separation and found that the resulting C 3 H 6 /C 3 H 8 sorption selectivity is comparable with that of ZIF-8, and a pore aperture size of 3.3-4.6 Å was found to further improve the C 3 H 6 /C 3 H 8 diffusion selectivity. Furthermore, it has been postulated that C 3 H 8 has a comparatively higher energy barrier to pass through Zr-fum-fcu-MOF crystals than does C 3 H 6 , due to the C-C bond rotation of C 3 H 8 . Similarly, Lee et al. [82] developed defect-engineered MOFs derived from hydrolytically-stable [83] UiO-66 to create additional sorption sites for effective C 3 H 6 /C 3 H 8 separation; their defect-engineered UiO-66-based MMM significantly improved C 3 H 6 permeance (1.005%) relative to the control (UiO-66based MMM, 308%), without substantially jeopardizing the C 3 H 6 /C 3 H 8 selectivity.

Comparison of Membrane Performance against the Upper Bound
In this study, the hydrocarbon separation performance of membranes is benchmarked against the upper bound curve for C 2 H 4 /C 2 H 6 and C 3 H 6 /C 3 H 8 separation, based on the data in Tables 2 and 3, respectively. The parameters used to obtain the upper bound plot (Figure 7) are listed in Table 4. The results reveal that achieving excellent light hydrocarbon separation performance is highly challenging. For example, the trend in Figure 7a indicates that the upper bound limit could easily be surpassed by using polymeric membranes with M 2 dobdc series fillers, considering their high C 2 H 4 adsorption and C 2 H 4 /C 2 H 6 selectivity (based on IAST calculations). However, MOFs with open metal sites generally exhibit poor hydrolytic stability, which may affect their long-term separation performance. The incorporation of deep eutectic solvents (DES) into polymeric membrane [62] has been proven to be feasible in substantially improving C 2 H 4 /C 2 H 6 selectivity, the challenges of developing cost-effective and highly stable DES notwithstanding [89,90]. As discussed in Section 3, the use of ZIF-8 as porous material in membranes for C 3 H 6 /C 3 H 8 separation has been proven effective, as ZIF-8 substantially improves C 3 H 6 /C 3 H 8 selectivity. Thus, the incorporation of ZIF-8 in polymeric membranes has the potential to surpass the C 3 H 6 /C 3 H 8 upper bound limit, as indicated in Figure 7b. λ [2] β [2] C 2 H 4 /C 2 H 6 0.14 7.3 C 3 H 6 /C 3 H 8 0.17 20.4 [1] The upper bounds for C 2 H 4 /C 2 H 6 and C 3 H 6 /C 3 H 8 are based on the data obtained in 2013 and 2003, respectively; [2] λ and β are described as the slope and front factor of the upper bound line, for which the relation between permeability and selectivity can be correlated as follows: α = β/P λ .  Tables 2 and 3, respectively. The parameters used to generate the upper bound line [36,37] are summarized in Table 4.

Conclusions and Future Perspectives
This perspective paper presents an overview of MMMs used for C 2 H 4 /C 2 H 6 and C 3 H 6 /C 3 H 8 separation. Such light hydrocarbon separations are generally challenging due to the similar polarizabilities, kinetic diameters, and van der Waals diameters of the adsorbates, which make it difficult to realize high selectivity. Moreover, because light hydrocarbon molecules are bulkier than carbon dioxide, nitrogen, and methane (which have been extensively investigated for CO 2 separation [23,28,91]), membranes with high FFV are highly desirable for separation. Thus, most relevant studies have investigated high-performance polyimides, namely 6FDA-based polymers and PIMs [44,92,93]. However, these polymers require extensive monomer purification, making their large-scale fabrication difficult [53]. Composite membranes are thus a feasible alternative for light hydrocarbon separation. This can be potentially achieved with the utilization of porous fillers possessing large accessible surface areas that are not limited to zeolites and MOFs, as they can effectively increase the diffusion rates of gas-molecule MMMs.
Future efforts should focus on the investigation of MMMs for acetylene separation. Acetylene is typically produced via the thermal cracking of CH 4 and typically contains CO 2 as the byproduct. Acetylene is also present as an impurity in C 2 H 4 production, which tends to disrupt the effective polymerization of C 2 H 4 . Acetylene separation is largely reliant on the use of adsorbents, but effective membrane-based separation is difficult because of the comparable properties of the adsorbates (Table 1). Nevertheless, the membrane performance of MMMs can be expected to improve through the use of porous materials proven to be effective in previous studies.

Conflicts of Interest:
The authors declare no conflict of interest.