Improving the Selectivity of ZIF-8/Polysulfone-Mixed Matrix Membranes by Polydopamine Modification for H2/CO2 Separation

Gas separation membranes are essential for the capture, storage, and utilization (CSU) of CO2, especially for H2/CO2separation. However, both glassy and rubbery polymer membranes lead a relatively poor selectivity for H2/CO2 separation because the differences in kinetic diameters of these gases are small. The present study establishing the mixed matrix membranes (MMMs) consist of a nano-sized zeolitic imidazolate frameworks (ZIF-8) blended with the polysulfone (PSf) asymmetric membranes. The gas transport properties (H2, CO2, N2, and CH4) of MMMs with a ZIF-8 loading up to 10 wt% were tested and showing significant improvement on permeance of the light gases (e.g., H2 and CO2). Moreover, the depositional polydopamine (PDA) layer further enhanced the ideal H2/CO2 selectivity, and the PDA-modified MMMs approach the Robeson upper bound of H2/CO2 separation membranes. Hence, the PDA post-modification strategy can effectively repair the defects of MMMs and improved the H2/CO2selectivity.


INTRODUCTION
The continuous rise of atmosphere carbon dioxide (CO 2 ) concentration caused by excessive anthropological combustion of fossil fuels leads to global warming and extreme climate events (Gao et al., 2017). In this situation, capture, storage, and utilization (CSU) of CO 2 from other sources have been a worldwide attention Zheng et al., 2016). At present, there are three technically plausible strategies for CO 2 CSU: post-combustion CO 2 capture (mainly for CO 2 /N 2 separation), oxy-combustion (mainly for O 2 /N 2 separation), and pre-combustion CO 2 capture (mainly for H 2 /CO 2 separation; Ramasubramanian et al., 2013;Yan et al., 2015). Among them, pre-combustion CO 2 capture is a promising technology, which can reduce the CO 2 emission and mitigate energy crisis (Wang et al., 2012;Liao et al., 2015). Since the H 2 /CO 2 syngas from water-gas shift reaction can provide H 2 as a preferred fuel or chemical feedstock, it is very important for the separation technologies for H 2 /CO 2 .
Membrane separation technology has obtained a great deal of attention for H 2 /CO 2 separation due to the fewer environmental impacts and lower energy costs, compared to conventional industrial methods (e.g., adsorption or cryogenic distillation; Rabiee et al., 2014;Liu et al., 2016;Wang et al., 2017;Ibrahim and Lin, 2018). Besides, membrane separation processes can be employed for the capture of CO 2 , while H 2 is subjected to combustion, which is due to a very high permeation rate of H2 relative to most other gases (Fu et al., 2016). However, polymer membrane performance has been limited by a trade-off between gas permeation and selectivity, known as "Robeson upper bound" (Chua et al., 2011;Zhang et al., 2012;Sánchez Laínez et al., 2018a). Both glassy and rubbery polymer membranes show a relatively poor selectivity for separation of H 2 /CO 2 because the differences in kinetic diameters of these gases are small (2.89 Å for H 2 and 3.3 Å for CO 2 ; Hosseini et al., 2010;Kim et al., 2018). The inorganic membranes generally have uniform pore size and excellent resistance to high temperature and pressure, e.g., silica membranes (Song et al., 2016a,b), which can achieved both high permeability and selectivity (Xiang et al., 2017). However, the expensive price limited the large-scale gas application of inorganic membranes. Recent gas separation membranes have been focused on the mixed matrix membranes (MMMs), which compensated the limitations of polymeric and inorganic membranes, while offering an ease in processability and moderate processing cost (Nordin et al., 2015).
Zeolitic imidazolate frameworks (ZIFs) are promising materials for gas separation membrane fabrication, for example, zeolitic imidazolate framework (ZIF-8) as one of the most investigated MOFs with the sod topology and the smaller window of 3.4 Å, which is close to the kinetic diameter of H 2 (2.89 Å; Sánchez Laínez et al., 2018b). A highly oriented ZIF-8 membrane on a porous α-alumina support was reported by Bux et al. (2011). The results showed that the H 2 permeance of ZIF-8 membrane was ∼4,032 Barrer, while the H 2 /CO 2 selectivity was only six. Besides, the pure ZIF-8 membranes are difficult to be reproduced on a large-scale and are too brittle to withstand high operating pressures (Gascon et al., 2012). Recently, the number of studies focused on the utilization of ZIF-8 for MMM preparation, which could potentially overcome the H 2 /CO 2 Robeson upper bound of gas separation membranes .  Song et al. (2012) incorporated the ZIF-8 as a nanofiller into a model polymer matrix (Matrimid R 5,218) via a mixing solution, showing enhanced permeability of the MMMs with negligible loss in selectivity. Wijenayake et al. (2013) fabricated a polyimide MMM with 33.3 wt% ZIF-8, and H 2 permeability of prepared MMM showed an approximate 400% improvement. Nevertheless, excessive ZIF-8 loadings would increase the chances to agglomerate and increase the defective risks of MMMs.
Developing defect-free ZIF-8/polymer MMMs is a major challenge because the defect leads to the deterioration of the membrane performance (Dechnik et al., 2017). Recently, surface post-treatment could effectively repair defects of MMMs (Nordin et al., 2014). Polydopamine (PDA), which is prepared through dopamine self-polymerization in weak alkaline solutions with the participation of oxygen, forming a PDA coating adhere onto nearly all kinds of substrates, has drawn much attention. (Lu et al., 2017;Yang et al., 2018). This work aims to develop a new type of asymmetric MMMs via phase inversion and PDA modification (Figure 1), further improving the H 2 /CO 2 selectivity. By using our designed strategy, the gas permeance of ZIF-8/PSf MMMs was significantly improved in the presence of a certain amount of ZIF-8 nanoparticles. Furthermore, the effect of polydopamine (PDA)-modified MMMs on the gas transport was studied.

Synthesis of ZIF-8 Nanoparticles
ZIF-8 nanoparticles were obtained by the solvent method as described in a previous study (Sánchez-Laínez et al., 2016). First, 1.47 g of Zn(NO 3 ) 2 •6H 2 O and 3.25 g of mIm were added in 100 ml of MeOH as solutions A and B, respectively. Then, solution A was rapidly poured into solution B under stirring. The mixture solution was reacted for 30 min with continuous stirring at room temperature. Finally, the ZIF-8 nanocrystals were separated from the milky dispersion by centrifugation and then were washed with fresh MeOH three times. The obtained ZIF-8 nanoparticles were directly used for the preparation of MMMs without drying.

Characterization
An XRD diffractometer (Bruker D8, Germany) was used to detect the crystal structure of the ZIF-8 at 2θ = 5 • -40 • with 0.02 step size. The infrared spectral analysis of the ZIF-8 nanoparticles and MMMs were tested by attenuated total reflectance-Fourier transform infrared spectroscopy (ATR-FTIR; Vertex 70; Bruker). The morphology and structure of the samples were observed by scanning electron microscopy (Nova NanoSEM450, USA). The N 2 adsorption and desorption isotherms of the ZIF-8 nanoparticles were observed by 3Flex physical adsorption instrument (Micromeritics, USA) at 77 K, and the adsorption isotherms of H 2 and CO 2 were tested at 273 K. A thermogravimetric analyzer (TGAQ50; TA Instruments-Waters LLC) was used for evaluating the thermal stability of the ZIF-8 nanoparticles and MMMs. All samples were heated from 30 to 800 • C with 10 • C min −1 heating rate under N 2 with a flow of 50-ml min −1 .

Gas Permeation Tests
Single-gas permeability of membranes was measured using a constant-volume variable-pressure method (Zhao et al., 2018). The entire permeation cell was placed in an oven to keep the temperature at 30 • C. The permeation cell was kept under vacuum for 12 h to remove other gases. The effective area of membranes is about 0.3 cm 2 , and the gas permeability of each samples has been tested at least three times. The gas permeation (in terms of Barrer, 1 Barrer = 1 × 10 −10 cm 3 (STP) cm cm −2 s −1 cmHg −1 ) was calculated by the following Equation (1): where, P is the gas permeability in Barrer, V is the constant volume container (cm 3 ), l is the thickness of dense layer of membrane (cm), which are obtained from SEM images. A is the membrane surface area (cm 2 ), T is the temperature (K), P 0 is the upstream (feed) pressure (psia), and dp/dt is the change in pressure against time (mmHg/s). The gas ideal selectivity (α i/j ) for components i and j was defined as the ratio of gas permeability of the two components by the following Equation (2):

Characterization of ZIF-8 Nanoparticles
The ZIF-8 nanoparticles were synthesized by the solvent method, and the XRD characteristic peaks of ZIF-8 are shown in Figure 2A, comfirming the typical sodalite (SOD) type structure of ZIF-8 nanoparticles (Park et al., 2006;Cravillon et al., 2009). Figure 2B shows the resembling spherical morphology of ZIF-8 nanoparticles by SEM characterization, and the average size Frontiers in Chemistry | www.frontiersin.org of ZIF-8 nanoparticles was about 150 nm. The microporous structure of ZIF-8 was confirmed by N 2 adsorption and desorption isotherms, and the ZIF-8 nanoparticles exhibited a typical Type-I isotherm, as shown in Figure 2C. The BET surface area and pore volume of ZIF-8 nanoparticles were 1,371 m 2 g −1 and 0.72 cm 3 g −1 , respectively. The pore size distribution (PSD) provided further insight into the pore structure of ZIF-8 nanoparticles ( Figure 2D). The 0.64 and 0.75 nm PSD centers of ZIF-8 represented the flexible six-membered ring, which arises from the vibrations of imidazole ligands (Guo et al., 2018). In addition, the largest PSD center of ZIF-8 was about 1.0 nm, which corresponds to the diameter of the ZIF-8 SOD cage.

Effect of Evaporation for Asymmetric PSf Membranes
To optimize the preparation process of asymmetric PSf membranes, the effect of evaporation for gas separation performance of membranes was investigated as shown in Table 1.
It can be found that the H 2 and CO 2 permeabilities significantly were reduced with the increasing evaporation time. This is due to the evaporation process that induced the skin layer formed of PSf membranes, and the skin layer thickness was improved with increasing evaporation times as shown in Figure 3. The formed skin layer is a resistance barrier between the PSf membrane and the coagulation bath. Presence of this resistance barrier induced the densification of skin layer as the evaporation time (Hołda et al., 2013). Moreover, all asymmetric PSf membranes exhibited lower N 2 and CH 4 gas permeabilities than the other gases, owning to their larger kinetic diameters. The crosssection morphology of asymmetric PSf membranes with different evaporation times is shown in Figure 3, which consist of the extremely well-defined dense skin layers supported on a highly open-celled structure. Based on the gas separation test results, the PSf-30 membrane was used for the ZIF-8/PSf MMM preparation.
Effect of ZIF-8 Loading for ZIF-8/PSf-30 MMMs Table 2 summarizes the effects of the ZIF-8 gas loading on the gas separation performance of MMMs. Compared with the asymmetric PSf-30 membrane, it was confirmed that permeability of MMMs for the four gases was improved with increasing ZIF-8 loading below 10 wt.%. However, an excessive amount of ZIF-8 can cause the separation performance to decline in the case of the 15 wt% ZIF-8/PSf-30 MMM. Figure 4 shows the cross-section and top surface of the five different MMMs. The surface defect was significantly increased due to the incorporation of ZIF-8 with different loadings. The defects on the surface of MMMs could be attributed to the ZIF-8 agglomeration (Nafisi and Hägg, 2014;Boroglu and Yumru, 2017). Owing to these defects of surfaces, the gas selectivity of MMMs did not change significantly; only the improvement of H 2 /N 2 selectivity was found. In addition, the thickness of sublayers under the dense skin layers was improved (Figure 4) with the increasing ZIF-8 loading. This morphological change in the MMMs was attributed to the delayed demixing during the phase separation, leading to the dense skin layer transformation to the porous layer (Lu et al., 2016).

PDA Coating the 10 Wt% ZIF-8/PSf-30 MMMs
The PDA modified method can overcome the limitations caused by the traditional self-assembly, entrapment, and chemical binding methods (Lu et al., 2017;Wang et al., 2019). After the PDA modification, the stability of MMMs could be improved, and the defects of surfaces will be repaired (Liu et al., 2013;Huang et al., 2014;Yuan et al., 2014). In order to further study the gas separation of MMMs after PDA coating with different time, the H 2 and CO 2 gases were selected as the representatives. Figure 5A showed the H 2 and CO 2 permeabilities and H 2 /CO 2 selectivity of 10 wt% ZIF-8/PSf-30 MMM with the different PDA coating time. Either H 2 or CO 2 permeability follows a decreasing trend when increasing PDA coating times. This is because the improved denser layers of MMMs and the enhanced gas transport resistance. However, for smaller kinetic diameter gas such as H 2 , the permeability inhibition was not obvious than that of CO 2 . Hence, the H 2 /CO 2 selectivity of MMMs was improved, such as the selectivity of PDA-2/10 wt% ZIF-8/PSf-30 MMM achieved 9.3 at 4 bar and 30 • C. Figure 5B showed the H 2 and CO 2 permeabilities and H 2 /CO 2 selectivity of PDA-2/10 wt%    Wang et al., 2015). In addition, the surface property of PDA-2/10 wt% ZIF-8/PSf-30 MMM was investigated by ATR-FTIR as shown in Figure 5C. Compared with the 10-wt% ZIF-8/PSf-30 MMM, a different peak at 1,540 cm −1 was attributed to the N-H bending vibration of PDA (Habibi et al., 2015;Zhou et al., 2015), which proved the successful PDA modified layers. However, the TG curves showed good thermal stability of MMMs as shown in Figure 5D. The surface and cross-section SEM images of the 10-wt% ZIF-8/PSf-30 with different coating times are shown in Figure 6. The SEM images were taken to investigate the changes in surface morphology of MMMs with different coating times. With increasing PDA coating time, it can be observed that the defects in the surface almost disappeared, and the little rough surface becomes quite smooth.
The correlation between selectivity and permeability for H 2 /CO 2 is shown in Figure 7. Embedding ZIF-8 nanoparticles into the PSf-30 asymmetric membrane had a positive effect on the H 2 permeability. However, the selectivity of H 2 /CO 2 was not improved. After the PDA coating, the H 2 /CO 2 selectivity was significantly improved, while the H 2 permeability was reduced, and the result of PDA-2/10 wt% ZIF-8/PSf-30 MMMs was very close to the 2008 Roberson upper bound. Compared to the reported MMMs, the PDA-modified MMMs showed higher selectivity of H 2 /CO 2 as shown in Table 3.

CONCLUSIONS
The novel MMMs have been developed for gas separation application by the PDA post-modified strategy. The ZIF-8 nanoparticles have been embedded in the PSf asymmetric membranes as the MMMs, and the gas permeability of MMMs was significantly improved. The optimal ZIF-8 concentration of 10 wt% produced an H 2 permeability of 87 Barrer, but the H 2 /CO 2 selectivity was only 2.38. The PDA modification has been considered as an effective method for improving the properties of membranes. This binding method can overcome the limitations caused by the traditional self-assembly, entrapment, and chemical binding methods (Lu et al., 2017(Lu et al., , 2019. Coating the MMMs with PDA repaired most defects of the surfaces, which reduced the H 2 permeability of MMMs and improved the H 2 /CO 2 selectivity. For PDA-2/10 wt% ZIF-8/PSf-30 MMM, the H 2 permeability was 23.3 Barrer and the H 2 /CO 2 selectivity achieved 9.3 at 30 • C under 4 bar. Prepared PDA-modified MMMs were highly promising for H 2 /CO 2 separation, owing to the simple manufacturing process and effective improvement. These results demonstrated the availability of the PDA postmodified MMMs for gas separation application.

DATA AVAILABILITY STATEMENT
All datasets presented in this study are included in the article/ supplementary material.

AUTHOR CONTRIBUTIONS
XM designed, analyzed data, and wrote the main manuscript. SY performed the experiments and performance evaluation. PL conceptualized, designed, and edited the main manuscript. YZ and JZ helped design complement experiments and reviewed and edited the manuscript. All authors contributed to the review and approval of the manuscript.