Optimization of a MOF Blended with Modified Polyimide Membrane for High-Performance Gas Separation

The preparation, characterization and gas separation properties of mixed matrix membranes (MMMs) were obtained from polyimide capped with ionic liquid and blended with metal-organic frameworks (MOFs). The synthesized MOF was amine functionalized to produce UiO-66-NH2, and its amino group has a higher affinity for CO2. Mixed matrix membranes exhibited good membrane forming ability, heat resistance and mechanical properties. The polyimide membrane exclusively capped by ionic liquid exhibited good permselectivity of 74.1 for CO2/CH4, which was 6.2 times that of the pure polyimide membrane. It is worth noting that MMM blended with UiO-66-NH2 demonstrated the highest ideal selectivity for CO2/CH4 (95.1) with a CO2 permeability of 7.61 Barrer, which is close to the 2008 Robeson upper bound. The addition of UiO-66-NH2 and ionic liquid enhanced the permselectivity of MMMs, which may be one of the promising technologies for high performance CO2/CH4 gas separation.


Introduction
Concerns about global warming have brought unprecedented public attention on the issue of carbon emissions [1][2][3][4]. As a result, an effective technique for separating the carbon dioxide from a mixture is required. Gas separation membrane technology is an effective method. The objective that researchers have been trying to achieve is a membrane material with high permeability and high selectivity. While gas separation membranes have advanced significantly, numerous gas separation membranes with excellent characteristics have emerged, such as polymer membranes [5]; metal-organic framework membrane (MOF) [6][7][8]; carbon membranes and zeolite membranes [9,10]; and various mixed matrix membranes (MMM) [11][12][13]. Polymer membranes are the most important commercial membranes for gas separation due to their advantages, which include the ease of membrane formation and low cost. However, polymer membranes are usually limited in their equilibrium relationship between selectivity and permeability, which makes it difficult to achieve both simultaneously [14,15]. Robeson applied extensive experimental data to demonstrate the inverse relationship between selectivity and permeability of polymer membranes and defined the Robeson upper bound [16,17]. For the past two decades, the focus of separation membrane research has been on how to exceed the upper bound. Common mem-Membranes 2022, 12, 34 3 of 12 chemical reactor with 800 W for 30 min. The solution was centrifuged and washed with DMF and methanol to exchange solvents. The obtained solids were evaporated in a vacuum oven at 110 • C overnight [32].

Synthesis of PI-IL/UiO-66-NH 2 Membranes
As shown in Scheme 1, PI-IL membranes were synthesized by polycondensation reaction. BTDA (1.1 mol, 1.4 g) and ODAs (1.0 mol, 0.60 g) were dissolved in DMF (3 mL) to form a polyamic acid (PAA) solution for 4 h at 20 • C. Ionic liquid 1-carboxyethyl-3methylimidazolium hexafluorophosphate (0.10 mmol, 0.020 g) was dissolved in DMF solution, then the solution was added into polyamic acid solution. The solution was stirred for 6 h to react sufficiently. After that, UiO-66-NH 2 (1 wt%, 2 wt%, 3 wt%, 4 wt% and 5 wt%) was dispersed in DMF and sonicated for 15 min, then the dissolved UiO-66-NH 2 solution was mixed with the PI-IL solution and stirred for 24 h. After standing for 2 h to remove bubbles (no sedimentation of MOF due to the uniform dispersion in the casting solution [33]), the mixed solution was casted in a clean glass dish and dried at 80-280 • C for 10 h in order to obtain the mixed matrix membrane. The compositions of the synthesized membranes are listed in Table 1.

Preparation of UiO-66-NH2
UiO-66-NH2 was synthesized according to a published procedure [31], with some improvements. Briefly, ZrCl4 (0.87 mol, 2.01 g) and 2-aminoterephthalic acid (0.87 mol, 1.56 g) were mixed with 90 mL DMF in a 250 mL flask. The mixture was heated by Microwave chemical reactor with 800 W for 30 min. The solution was centrifuged and washed with DMF and methanol to exchange solvents. The obtained solids were evaporated in a vacuum oven at 110 °C overnight [32].

Synthesis of PI-IL/UiO-66-NH2 Membranes
As shown in Scheme 1, PI-IL membranes were synthesized by polycondensation reaction. BTDA (1.1 mol, 1.4 g) and ODAs (1.0 mol, 0.60 g) were dissolved in DMF (3 mL) to form a polyamic acid (PAA) solution for 4 h at 20 °C. Ionic liquid 1-carboxyethyl-3-methylimidazolium hexafluorophosphate (0.10 mmol, 0.020 g) was dissolved in DMF solution, then the solution was added into polyamic acid solution. The solution was stirred for 6 h to react sufficiently. After that, UiO-66-NH2 (1 wt%, 2 wt%, 3 wt%, 4 wt% and 5 wt%) was dispersed in DMF and sonicated for 15 min, then the dissolved UiO-66-NH2 solution was mixed with the PI-IL solution and stirred for 24 h. After standing for 2 h to remove bubbles (no sedimentation of MOF due to the uniform dispersion in the casting solution [33]), the mixed solution was casted in a clean glass dish and dried at 80-280 °C for 10 h in order to obtain the mixed matrix membrane. The compositions of the synthesized membranes are listed in Table 1.

Characterization of the Materials
The crystal structure of the sample was determined by using X-ray diffraction (XRD) (D8 advanced diffractometer, Bruker AXS, Karlsruhe, German). The morphology of the MOF and cross sections of the membranes were observed by scanning electron microscope (SEM) (S-3400, Hitachi, Tokyo, Japan). Fourier Transform Infrared Spectroscopy Scheme 1. Structure of UiO-66-NH 2 and PI-IL.

Characterization of the Materials
The crystal structure of the sample was determined by using X-ray diffraction (XRD) (D8 advanced diffractometer, Bruker AXS, Karlsruhe, Germany). The morphology of the MOF and cross sections of the membranes were observed by scanning electron microscope (SEM) (S-3400, Hitachi, Tokyo, Japan). Fourier Transform Infrared Spectroscopy (FT-IR) (Spectrum Two, PE company, Los Angeles, CA, USA) was used to determine the chemical structure of the samples. The nitrogen adsorption-desorption isotherm of the sample was measured with an adsorption instrument (ASAP 2020 Plus HD88, Micromeritics, Norcross, GA, USA), and the test was carried out after degassing at 120 • C for 12 h (80 mg of the sample). The particles were analyzed according to their specific surface area and pore size Membranes 2022, 12, 34 4 of 12 distribution at 77 K. A thermogravimetry analyzer (TGA 8000, Perkin Elmer, Waltham, MA, USA) was used for a thermal performance test, the flow rate of N 2 was 40 mL/min and the heating rate was 5 • C/min.

Gas Permeation Measurements of Membranes
The mixed gas permeability and selectivity of the membranes were measured at 34 • C using a GTR-11MH gas permeability analyzer (GTR Tec Corporation, Uji, Kyoto). The test area was 0.785 cm 2 , the test pressure is maintained at 49 KPa, the test gas is a 1:1 mixture of CO 2 and CH 4 , the carrier gas is H 2 and the pressure is 0.1 MPa for testing. The gas permeability coefficient was calculated using Formula (1).
The diffusion coefficients (D) and the solubility coefficients (S) were calculated by using Equations (2) and (3) [34]: where q is the gas permeation measured by the instrument, mL; K is the auxiliary positive coefficient of the instrument, with a fixed value of 1.25; L (cm) is the thickness of the gas separation membrane; a is the permeation area in the instrument (set at 0.785 cm 2 ); p (cmHg) is the pressure while the instrument is testing; and t (s) is the time it takes for the instrument to measure the gas separation membrane, which was determined by the specific operation of the operator.

Fabrication and Characterization of UiO-66-NH 2
Microwave heating is an effective tool in organic chemistry synthesis, but it has also recently been used in the synthesis of inorganic and inorganic/organic materials. In addition, the conversion of microwave radiation to heat is often efficient and homogeneous throughout the sample, which reduces energy consumption and the necessity for heat transfer in the mixture [35]. For the reasons stated above, we employed microsynthesis technology to prepare UiO-66-NH 2 , and the reaction time decreased to 30 min, which presented a 47-fold reduction in reaction time compared to the conventional method. The yield of MOF using microwave synthesis was 50%, while the yield of MOF using traditional solvothermal process was 35%. The crystal structure of UiO-66-NH 2 was characterized by XRD. In Figure 1a, the powder X-ray diffraction pattern of the synthesized UiO-66-NH 2 shows excellent agreement with the simulated diffraction pattern. All of the diffraction peaks of the micro-assisted synthesized MOFs correlate well with the simulated spectrum [31,36]. Figure 1b depicts the morphology of UiO-66-NH 2 , which has a particle size of 100-200 nm and an ortho-octahedral structure, and exhibited good crystal shape regularity, particle size uniformity and crystal perfection. The nitrogen adsorption-desorption isotherm and pore size distribution curve of UiO-66-NH 2 are presented in Figure 1c. UiO-66-NH 2 follows Type 1 isotherm, which is indicative of microporosity. The surface area of the Brunauer-Emmett-Teller (BET) surface area was 813.25 m 2 /g, and the pore volume was 0.44 cm 3 /g, which can contribute to gas permeability. The BET area of micro-synthesized MOF is lower than that of conventional solvothermal process, while the pore is bigger. This phenomenon may be due to fewer defects; thus, the specific surface area is smaller [35]. The results of XRD, SEM and N 2 adsorption-desorption isotherm confirmed the successful preparation of UiO-66-NH 2 nanoparticles using microwave-assisted synthesis.
contribute to gas permeability. The BET area of micro-synthesized MOF is lower than that of conventional solvothermal process, while the pore is bigger. This phenomenon may be due to fewer defects; thus, the specific surface area is smaller [35]. The results of XRD, SEM and N2 adsorption-desorption isotherm confirmed the successful preparation of UiO-66-NH2 nanoparticles using microwave-assisted synthesis.

Characterization of MMMs
In order to verify whether PI is fully imidized, the FT-IR spectra of PAA and PI are shown in Figure 2. The peak at 1665 cm −1 is the absorption vibration peak of -NH on the polyimide amide group, which indicated the formation of PAA. For the PI spectrum, this peak becomes weaker, indicating that PAA has been completely imidized to form PI. After imidization, the stretching vibration peak of C-N in polyimide is at 1238 cm −1 , and the absorption peak of C=O in 1729 cm −1 was weakened, which proves that polyimide was formed [37]. Before SEM characterization, the membrane samples were brittled with liquid nitrogen, and the broken side was marked. As shown in Figure 3, the aggregation phase of IL is not visible in cross-section electron microscopy, which proves that the ionic liquids are not comingled in the matrix membrane but capped into polyimide. The bulky parts are the polyimide-ionic liquid matrices. The SEM images of the PI-IL/3% MOF membrane revealed that UiO-66-NH2 particles are well dispersed in the matrix and there is no significant agglomeration of the filler. It can be observed from Figure 3 that the number of UiO-66-NH2 particles in the electron micrographs increases as loading content in-

Characterization of MMMs
In order to verify whether PI is fully imidized, the FT-IR spectra of PAA and PI are shown in Figure 2. The peak at 1665 cm −1 is the absorption vibration peak of -NH on the polyimide amide group, which indicated the formation of PAA. For the PI spectrum, this peak becomes weaker, indicating that PAA has been completely imidized to form PI. After imidization, the stretching vibration peak of C-N in polyimide is at 1238 cm −1 , and the absorption peak of C=O in 1729 cm −1 was weakened, which proves that polyimide was formed [37]. Before SEM characterization, the membrane samples were brittled with liquid nitrogen, and the broken side was marked. As shown in Figure 3, the aggregation phase of IL is not visible in cross-section electron microscopy, which proves that the ionic liquids are not comingled in the matrix membrane but capped into polyimide. The bulky parts are the polyimide-ionic liquid matrices. The SEM images of the PI-IL/3% MOF membrane revealed that UiO-66-NH 2 particles are well dispersed in the matrix and there is no significant agglomeration of the filler. It can be observed from Figure 3 that the number of UiO-66-NH 2 particles in the electron micrographs increases as loading content increases. The size of the nanoparticles in the membranes is 200-300 nm, which is comparable to and slightly larger than the size of pure UiO-66-NH 2 . The particles are wrapped in a PI matrix, and the interface is blurred, demonstrating that membrane compatibility has improved. This is in accordance with the experimental theory, as the viscosity of the polyimide casting solution is very high, and the MOF particles added to it are enveloped by the casting solution. Consequently, the Membranes 2022, 12, 34 6 of 12 particle pattern in the cross-sectional electron micrographs of the film is inclined to be round, which proves that the MOF is successfully mixed into the matrix membrane. As shown in Figure S2 (in Supplementary Materials), by applying surface scanning electron microscope in mixed matrix membranes, all membranes are basically uniformly dispersed, and there was only a very small amount of agglomeration of MOF.
creases. The size of the nanoparticles in the membranes is 200-300 nm, which is comparable to and slightly larger than the size of pure UiO-66-NH2. The particles are wrapped in a PI matrix, and the interface is blurred, demonstrating that membrane compatibility has improved. This is in accordance with the experimental theory, as the viscosity of the polyimide casting solution is very high, and the MOF particles added to it are enveloped by the casting solution. Consequently, the particle pattern in the cross-sectional electron micrographs of the film is inclined to be round, which proves that the MOF is successfully mixed into the matrix membrane. As shown in Figure S2 (in Supplementary Materials) by applying surface scanning electron microscope in mixed matrix membranes, all membranes are basically uniformly dispersed, and there was only a very small amount of agglomeration of MOF.  creases. The size of the nanoparticles in the membranes is 200-300 nm, which is comparable to and slightly larger than the size of pure UiO-66-NH2. The particles are wrapped in a PI matrix, and the interface is blurred, demonstrating that membrane compatibility has improved. This is in accordance with the experimental theory, as the viscosity of the polyimide casting solution is very high, and the MOF particles added to it are enveloped by the casting solution. Consequently, the particle pattern in the cross-sectional electron micrographs of the film is inclined to be round, which proves that the MOF is successfully mixed into the matrix membrane. As shown in Figure S2 (in Supplementary Materials), by applying surface scanning electron microscope in mixed matrix membranes, all membranes are basically uniformly dispersed, and there was only a very small amount of agglomeration of MOF.

Mechanical Properties of Membranes
As shown in Table 2, the yield strength (σs); elongation at break (εb); tensile strength (TS); and modulus of elasticity (E) of the ionic liquid-terminated polyimide membranes were measured from the tensile strength test. Figure 4 is a visualization of the trends in the mechanical properties and corresponds to the folded line point diagram of tensile strength and elongation at the break of the membranes.
The ionic liquid-capped polyimide membrane has better mechanical properties than a pure polyimide membrane, with the elongation of the break increasing from 8.96% to 37.3% and tensile strength increasing from 53.6 MPa to 6203 MPa. When IL was added to cover the end groups of the polyimide matrix, the unique electronic structure of the ionic liquid influenced the orientation of the molecular chains and increased the orientation. Simultaneously, it restricted the mobility of polyimide chain segments, increasing the degree of build up and enhancing intermolecular interactions, which resulted in its increased mechanical strength.
The tensile strength and elongation at the break of PI-IL/1% MOF reached a maximum of 37.4% and 6434 MPa, respectively. It may be due to the large interfacial interaction between the doped UiO-66-NH2 and the substrate polyimide. With the increase in MOF content, the tensile strength and elongation at the break of the membranes tended to decrease; this tendency may be due to the rigid structure of MOF, which makes the original excellent mechanical properties slightly inferior but still maintains good mechanical properties. This shows that all the produced mixed matrix membranes have good mechanical properties and completely fulfill the requirements under which the membranes may be used.

Mechanical Properties of Membranes
As shown in Table 2, the yield strength (σs); elongation at break (εb); tensile strength (TS); and modulus of elasticity (E) of the ionic liquid-terminated polyimide membranes were measured from the tensile strength test. Figure 4 is a visualization of the trends in the mechanical properties and corresponds to the folded line point diagram of tensile strength and elongation at the break of the membranes.

Thermal Properties of the Membranes
A TGA measurement in a N2 atmosphere was used to determine the thermal stability of PI, PI-IL and PI-IL/x% MOF membranes. The TGA curve of MOF can be divided into three stages ( Figure 5). As temperature Figure homogeneous co-blending. The thermal decomposition mass of MOF uniformly distributed in the matrix is represented by the 5-   The ionic liquid-capped polyimide membrane has better mechanical properties than a pure polyimide membrane, with the elongation of the break increasing from 8.96% to 37.3% and tensile strength increasing from 53.6 MPa to 6203 MPa. When IL was added to cover the end groups of the polyimide matrix, the unique electronic structure of the ionic liquid influenced the orientation of the molecular chains and increased the orientation. Simultaneously, it restricted the mobility of polyimide chain segments, increasing the degree of build up and enhancing intermolecular interactions, which resulted in its increased mechanical strength.
The tensile strength and elongation at the break of PI-IL/1% MOF reached a maximum of 37.4% and 6434 MPa, respectively. It may be due to the large interfacial interaction between the doped UiO-66-NH 2 and the substrate polyimide. With the increase in MOF content, the tensile strength and elongation at the break of the membranes tended to decrease; this tendency may be due to the rigid structure of MOF, which makes the original excellent mechanical properties slightly inferior but still maintains good mechanical properties. This shows that all the produced mixed matrix membranes have good mechanical properties and completely fulfill the requirements under which the membranes may be used.

Thermal Properties of the Membranes
A TGA measurement in a N 2 atmosphere was used to determine the thermal stability of PI, PI-IL and PI-IL/x% MOF membranes. The TGA curve of MOF can be divided into three stages ( Figure 5). As temperature figure homogeneous co-blending. The thermal decomposition mass of MOF uniformly distributed in the matrix is represented by the 5-10% mass difference. All prepared membranes can be used in normal environments and can withstand a temperature range of 0-480 • C.

Gas Permeation Performance of the MMMs
The prepared PI-IL/x%MOF membranes were tested for the separation performance of CO2 and CH4 mixed gas at a 1:1 ratio. Table 3 shows the gas separation performance of membranes. When PI was capped with IL, the membrane's permeability increased to 5.19 Barrer and 74.15, which was 4.36 and 6.21-times that of pure PI membranes, respectively.

Gas Permeation Performance of the MMMs
The prepared PI-IL/x%MOF membranes were tested for the separation performance of CO 2 and CH 4 mixed gas at a 1:1 ratio. Table 3 shows the gas separation performance of membranes. When PI was capped with IL, the membrane's permeability increased to 5.19 Barrer and 74.15, which was 4.36 and 6.21-times that of pure PI membranes, respectively. The addition of ionic liquid increased CO 2 permeability and CO 2 /CH 4 selectivity of the PI membrane. This phenomenon was most likely due to the affinity of ionic liquid for CO 2 . The ionic liquid's imidazole groups can interact with CO 2 to enhance CO 2 adsorption; consequently, the PI-IL membrane achieved higher CO 2 permeability. It is also hypothesized that the effect of ionic liquid on gas permeability is partly due to the plasticizing effect of ionic liquids on polymers. The addition of ionic liquids to polymer membranes reduces glass transition temperature, while increasing chain mobility resulted in higher permeability [38]. Meanwhile, inorganic anions (BF 4 , NO 3 and Cl) had a distinctive effect on its CO 2 sorption properties. Overall, it was concluded that increasing the basicity of anions is known to enhance CO 2 sorption [39]. The effect of IL in membranes is to increase their CO 2 solubility while decreasing CH 4 solubility [40]. As shown in Table 4, the solubility coefficient of the membrane increased after adding ionic liquid and MOF. In summary, the addition of IL increases the separation effect of CO 2 /CH 4 .  As shown in Table 3, introducing UiO-66-NH 2 particles to PI-IL/x% MOF mixed matrix membranes clearly enhanced their gas permeability properties. The permeability of PI-IL/3% MOF was the highest with a value 7.61 Barrer. This behavior may be due to the addition of porous UiO-66-NH 2 particles, which improves the transport passage of gas through the membrane and provides a higher affinity of UiO-66-NH 2 for CO 2 than CH 4 . The selectivity of CO 2 /CH 4 increased to 95.10 when UiO-66-NH 2 loading content was 3%; this could be ascribed to the molecular sieving effect of UiO-66-NH 2 nanoparticles [41]. As the loading content increased to 4% and 5%, the selectivities dropped compared to the PI-IL/3% MOF membrane but remained higher than the pure PI membrane. These values might be attributed to a smaller degree of agglomeration of MOF particles in the matrix, blocking a small part of gas transfer. Meanwhile, the presence of MOF seems to rigidify polymeric chains; therefore, the rigidified polymeric chains around the MOF and polyimide surfaces may be one of the causes for permeability reduction. Other possibilities for the decline in selectivities include partial pore blockage of MOF by polymer chains [42]. In the case of PI-IL/1-3% MOF MMMs, the results show that the presence of IL increases the adhesion of UiO-66-NH 2 in MMMs, since IL can also act as a wetting agent for UiO-66-NH 2 [43]. On the other hand, IL may enhance interfacial interaction via plasticization; the presence of IL in PI-IL/1-3% MMMs improves gas separation performance.
The Robeson upper bound of PI, PI-IL and PI-IL/x% MOF membranes for the gas pair of CO 2 /CH 4 is plotted in Figure 6. The PI-IL membrane exhibits higher CO 2 permeability and permselectivity compared to pure PI membrane, for which its plot moves up on Robeson's 1991 upper bound. CO 2 and CH 4 permselectivities of PI-IL/x% MOF mixed matrix membrane were higher than those of the PI-IL membrane, and gas separation data approaches Robeson's 2008 upper bound when UiO-66-NH 2 filler loading content reaches 1-2%. PI-IL filled with UiO-66-NH 2 at 4 and 5% exhibit lower permselectivities than the PI-IL membrane, but it is still higher than the 1991 upper bound. In summary, this increase in permselectivity confirmed that the introduction of IL and UiO-66-NH 2 can enhance gas separation performance of the fabricated membranes. The Robeson upper bound of PI, PI-IL and PI-IL/x% MOF membranes for the gas pair of CO2/CH4 is plotted in Figure 6. The PI-IL membrane exhibits higher CO2 permeability and permselectivity compared to pure PI membrane, for which its plot moves up on Robeson's 1991 upper bound. CO2 and CH4 permselectivities of PI-IL/x% MOF mixed matrix membrane were higher than those of the PI-IL membrane, and gas separation data approaches Robeson's 2008 upper bound when UiO-66-NH2 filler loading content reaches 1-2%. PI-IL filled with UiO-66-NH2 at 4 and 5% exhibit lower permselectivities than the PI-IL membrane, but it is still higher than the 1991 upper bound. In summary, this increase in permselectivity confirmed that the introduction of IL and UiO-66-NH2 can enhance gas separation performance of the fabricated membranes.

Conclusions
In conclusion, UiO-66-NH 2 particles were used to improve CO 2 gas permselectivity in an ionic liquid capped polyimide matrix. IL was selected to enhance the permeability of a polyimide. All the obtained mixed matrix membranes exhibited excellent mechanical and thermal properties, allowing them to meet conventional conditions of membranes. More importantly, CO 2 permeability and selectivity of MMM exhibited an obvious improvement over PI-IL. The mixed matrix membranes capped by IL and blended with UiO-66-NH 2 provide an efficient method to improve CO 2 selectivity while retaining gas permeability.