Carbon molecular sieve hollow fiber composite membrane derived from PMDA-ODA polyimide for gas separation

Abstract

Carbon molecular sieve (CMS) membranes have excellent gas separation property over conventional polymeric membranes and superior anti-swelling property. PMDA-ODA polyimide has high thermal stability and good mechanical property. It has been extensively adopted as the precursor of CMS membrane. However, due to the insoluble nature, PMDA-ODA CMS membranes are limited to configurations like dense symmetric films or composite membranes using porous inorganic or metal substrates. In this work, CMS hollow fiber composite membranes based on an asymmetric PMDA-ODA hollow fiber were successfully prepared for the first time. The neat PMDA-ODA hollow fiber membrane was crosslinked by polyethyleneimine to alleviate pore collapsing during carbonization and then dip-coated by a PMDA-ODA PAA solution to seal the surface defects. The PDMA-ODA CMS composite hollow fiber membranes showed gas permeances of 93.4 GPU, 19.6 GPU, 6.5 GPU, and 4.7 GPU for CO₂, O₂, N₂, and CH₄, respectively, with an ideal selectivity of 14.4, 3.0, and 19.8 for CO₂/N₂, O₂/N₂, and CO₂/CH₄ gas pairs, respectively. The attractive gas separation property shows a great potential for industrial application.

Keywords

Carbon molecular sieve hollow fiber membrane PMDA-ODA dip-coating polyethylenimine crosslinking

Introduction

Over the last two decades, membrane-based separation technologies have attracted a great deal of attention because they have merits such as high energy efficiency, low capital investment,¹ environmentally friendly, simple operation, and small footprint.^2–4 However, in harsh operational conditions such as natural gas purification⁵ or recovery of landfill gas, conventional polymeric membranes may undergo physical aging or plasticization that deteriorates the gas separation property.^1,6 Therefore, it is necessary to develop membranes having a better stability in extreme conditions. Carbon molecular sieve (CMS) membrane is one of the candidates, which exhibits excellent gas separation performance surpassing the Robeson Upper Bound for polymeric membranes,² and superior chemical/thermal stability as well.⁷ CMS membranes are typically obtained by pyrolyzing polymer precursors under precisely controlled conditions. It has micropores with a size between 7 and 20 Å resulting from the formation of turbostratic carbon comprising disordered sp²-hybridized sheets and ultra-micropores below 7 Å representing the inter-layer distance among carbon sheets.⁸ The formation of microstructure is mainly controlled by factors including: (a) polymeric precursor, (b) membrane configuration, and (c) pyrolysis condition.

Polyacrylonitrile, polyvinylidene chloride, poly (furfuryl alcohol), phenolic resin, and polyimide are the most studied polymer precursors for CMS membranes in forms of flat sheet, tubular, capillary, or hollow fiber configurations.^9–13 Hollow fiber CMS membrane has advantages including high surface area to volume ratio, high packing density, and good mechanical property.¹³ Soluble polyimides such as P84,¹⁴ Matrimid^® 5218,¹⁵ and 6FDA/BPDA-DAM¹⁶ have been fabricated into hollow fibers and then converted to CMS membranes. PMDA-ODA polyimide with a trade name of Kapton is another well studied polyimide for CMS membrane due to its high mechanical property, excellent thermal stability, easy to synthesize, abundant carbon content, and relatively low cost.¹⁷ However, due to the insoluble nature of PMDA-ODA, its soluble precursor, polyamic acid (PAA), is used to prepare asymmetric membranes by the nonsolvent induced phase inversion process.^18,19 Since PAA is very hydrophilic, PAA hollow fiber membrane is difficult to spin because of the slow phase inversion rate. To our best knowledge, there is only one report of the fabrication of Kapton based CMS hollow fiber membranes.²⁰ However, the membrane was made by coating a thick PMDA-ODA PAA gel layer on the inner surface of a Teflon tube. This method was difficult to be scaled up. In our previous work,²¹ a series of integrally-skinned PMDA-ODA hollow fiber membranes were prepared. The key was to speed up the phase inversion rate by adjusting the dope composition and spinning condition.

In this work, we aimed to prepare a CMS membrane based on the PMDA-ODA hollow fiber membrane. Special attentions were paid to avoid the pore collapsing of the PMDA-ODA hollow fiber during carbonization by chemical crosslinking using polyethylenimine (PEI). This was to prevent the formation of low permeable CMS hollow fiber membranes. As shown in Figure 1, a PMDA-ODA PAA solution was coated onto a crosslinked PI hollow fiber membrane to seal the surface defects. Finally, the hollow fiber composite membrane was carbonized to form the defect-free CMS membrane.

Experimental

Materials

Pyromellitic dianhydride (PMDA) and 4,4′-oxydiphenylene diamine (ODA) were purchased from Aladdin (China) and dried in vacuum at 80°C for 12 h before use. A branched polyethylenimine (PEI) with an M_w of 70000 g·mol⁻¹ was purchased from Macklin Biochemical Technology Co. Ltd (Shanghai, China). 1-methyl-2-pyrrolidine (NMP), tetrahydrofuran (THF), glycerinum (GLY), isopropanol (IPA), n-hexane (HEX), and methanol(MeOH)were obtained from Beijing Chemical Works (China) and used as received. CO₂ (99.99%), O₂ (99.99%), CH₄ (99.96%), and N₂ (99.999%) were obtained from Air liquide (China).

Dope preparation

PMDA-ODA PAA was synthesized following a reported method.²² PMDA, ODA, and a three-neck glass flask were fully dried. NMP was dehydrated by molecular sieve. ODA and NMP were added to the flask. After ODA was fully dissolved, a stoichiometric amount of PMDA to ODA was added. The solution was stirred in an ice-bath for 12 h till it became very viscous. After that, certain amount of GLY and THF were added as nonsolvents to the PMDA-ODA PAA dope solution. The solution was continuously stirred till it became a homogenous mixture.

Preparation of the PMDA-ODA PAA hollow fiber membrane

The PAA hollow fiber membrane was prepared using a similar procedure to references 21,23. Details of the spinning parameters are listed in Table 1. Specifically, a PAA dope solution was poured into a feed tank 1 day prior spinning to remove air bubbles. During the hollow fiber spinning process, the dope and bore solutions were co-extruded through a spinneret with an outer and inner diameter of 1.0 and 0.4 mm, respectively. The nascent gel-like hollow fiber passed through an air gap of 14 cm and entered a water bath to solidify via the non-solvent induced phase inversion process.²⁴ Finally, the solidified PAA hollow fiber was collected by a winding wheel at a take-up speed of 4 m/min. Since there was no observable stretching to the hollow fiber, this take-up rate was considered free-fall. The water-wet PAA hollow fiber membranes were dried using an IPA-HEX solvent exchanged method. Water in the hollow fibers were exchanged by IPA for three times and then by HEX for three times. After that, the HEX-wet hollow fibers were dried in air overnight. The dry PAA hollow fibers were placed in a tube furnace and gradually heated to 300°C for imidization. The heating program is shown in Figure 2.

Table 1.

The spinning condition of the PMDA-ODA PAA hollow fiber membranes.

Spinning conditions	Value
Dope composition (wt%)	22%PAA+15%THF+15%GLY+48%NMP
Dope flow rate (mL/min) Bore fluid composition	3 IPA:H₂O=1:9
Bore fluid flow rate (mL/min)	1
Air gap (cm) Take-up speed (m/min)	14 4
Water coagulant temperature (°C)	8
Relative humidity (%)	22
Room temperature (°C)	13

The crosslinking of PMDA-ODA PI hollow fiber

PEI was dissolved in a water/IPA (1:1) co-solvent at a concentration of 1 wt%. Then, the PEI solution was heated to 70°C and the PMDA-ODA PI hollow fibers were soaked in the solution for 1 h for crosslinking. After that, the hollow fibers were taken out and rinsed with deionized water for three times to remove any residual PEI. Finally, the crosslinked fibers were vacuum dried at 100°C for 1 h. The crosslinking mechanism is shown in Figure 3.

Preparation of the CMS hollow fiber composite membranes

To seal surface defects of the PMDA-ODA hollow fiber membranes, we dissolved the PMDA-ODA PAA polymer in a THF/MeOH (8:2 w/w) co-solvent and dip-coated the PAA solution onto the hollow fiber outer surface. Figure 4 shows the procedure of dip-coating. First, a single filament hollow fiber module was prepared. Then, the hollow fiber was immersed in the coating solution for 90 s. Later, the hollow fiber was vacuum dried at 70°C for 2 h to remove any residual solvent. Before carbonization, the fibers were cut off from the module. At last, the composite hollow fibers were pyrolyzed in a tube furnace and purged by N₂ in a flow rate of 200 mL/min using a heating protocol shown in Figure 2(b).

Characterization methods

A Nicolet iS50 Fourier transform infrared spectrometer (Thermo Scientific Inc., USA) was used to determine the chemical structures of the PAA, PI, and PEI crosslinked PI hollow fiber membranes. The scan numbers were 32 times and the scanning range was from 400 to 4000 cm⁻¹.

The thermal stability of the hollow fiber membranes before and after crosslinking was evaluated using a Metter-Toledo thermogravimetric analysis instrument (TGA). Hollow fiber samples were cut into small pieces and heated from room temperature to 800°C in an N₂ atmosphere at a heating rate of 10 K/min.

The glassy to rubbery transition (T_g) of polyimide hollow fiber membrane with or without crosslinking was measured using a Q-800 dynamic mechanical analysis instrument (TA instrument, USA). A frequency of 1 Hz and a temperature ramping rate of 3 °C/min was used.

Morphologies of the hollow fiber membranes were characterized using a Hitachi JSM-7800F scanning electron microscope (Japan). To obtain a smooth cross-section, the hollow fibers were fractured in liquid nitrogen and sprayed with gold before test.

The shrinkage% of hollow fiber membranes after carbonization was calculated using equation (1)

S = \frac{R_{1} - R_{2}}{R_{1}} \times 100 %

(1)

where S was the shrinkage%; R₁ (μm) was the outer diameter of the hollow fiber membrane before carbonization; and R₂ (μm) was the outer diameter of the hollow fiber membrane after carbonization.

The mean pore size (r) of the outer surface of the hollow fiber membranes was determined using a method as introduced by Liu et al.²⁵ Since the gas diffusion followed the Knudsen diffusion mechanism, gas permeance J_i was linear dependence of the pressure P_i

J_{i} = \frac{2}{3} {(\frac{8 R T}{Π M})}^{0.5} \frac{1}{R T} \frac{r ψ}{L_{P}} + \frac{P_{i}}{8 μ R T} \frac{r^{2} ψ}{L_{P}}

(2)

where J_i was the gas permeance (GPU, 1 GPU=1×10⁻⁶cm³ (STP)·cm⁻²·s⁻¹·cmHg⁻¹)); R was the universal gas constant (8.314 J·mol⁻¹ K⁻¹); T was the absolute temperature (K); M was the molecular weight of test gas (g·mol⁻¹); r was the mean pore size (m);

ψ

was the surface porosity; L_P was the effective length; P_i was the transmembrane pressure difference; and

μ

was the viscosity of test gas (Pa·s). Equation (2) can be simplified to equation (3)

J_{i} = K_{0} + P_{0} \times P_{i}

(3)

where

K_{0} = \frac{2}{3} {(\frac{8 R T}{Π M})}^{0.5} \frac{1}{R T} \frac{r ψ}{L_{P}}; P_{0} = \frac{1}{8 μ R T} \frac{r^{2} ψ}{L_{P}}

. We measured the permeance of N₂ at elevated pressure gradients and then drew a plot of J_i versus P_i. K₀ (intercept) and P₀ (slope) could be gotten after linear fitting of the plot. The mean pore size (r) was calculated using equation (4)

r = \frac{16}{3} (\frac{P_{0}}{K_{0}}) {(\frac{8 R T}{Π M})}^{0.5} μ

(4)

Determining the gas separation property of the hollow fiber membranes

The gas permeances and ideal selectivity of the hollow fiber membranes were determined using pure gases. The details of hollow fiber module preparation and testing process have been described in our previous study.²⁵ The membranes were tested at 25°C with a transmembrane pressure of 7 bar. The permeance was calculated using equation (5)

\frac{P}{L} = \frac{Q}{A \times T \times Δ P}

(5)

where

P / L

was the gas permeance in a unit of GPU; Q (cm³) was the permeate volume of test gas; A (cm²) was the outer surface area of the hollow fiber membrane; T(s) was the time for recording the volume of gas permeate; and ΔP (cmHg) was the transmembrane pressure difference. The ideal selectivity was calculated using equation (6)

α_{A / B} = \frac{{(\frac{P}{L})}_{A}}{{(\frac{P}{L})}_{B}}

(6)

where α_A/B was the ideal selectivity of gas A to gas B and P_A, P_B was the permeances of gas A and gas B, respectively.

Results and discussion

Hollow fiber membrane morphology

Hollow fiber membranes with regular cross-sectional morphology (round inner and outer surfaces) is important for achieving good mechanical strength, stable separation property and a uniform thickness of the coating layer. GLY and THF were selected as nonsolvent in the dope solution to facilitate phase inversion. This was to shift the composition of polymer dope closer to the binodal curve in the triangular diagram. Meanwhile, GLY as a more viscous nonsolvent would be kept in the polymer-rich phase during phase inversion. After the hollow fiber membrane solidified, the residual GLY would leach out. Thus, an interconnected structure that was beneficial to enhance permeance was formed. THF as a low boiling point nonsolvent would evaporate rapidly, causing a quick increase in the polymer concentration at dope surface. Thus, a denser skin layer having less surface defects would form.²²

Figure 5 shows the cross-sectional and surface morphologies of the hollow fiber membranes. A smooth outer surface with no observable defects was obtained. It was reported that a substrate with a smooth outer surface and small pores was preferred to form a uniform and thin coating layer.^26,27 Moreover, the hollow fiber inner surface had micropores of 1–3 μm that indicated the hollow fiber membrane was outer selective. Although the outer surface showed no observable pores, the mean surface pore size was calculated to be 1.5±0.1 nm (by equation (4)). Since the kinetic diameters of the testing gases are 3.3 Å (CO₂), 3.46 Å (O₂), 3.64 Å (N₂), and 3.76 Å (CH₄), respectively, these gases shall follow the low selective Knudsen diffusion mechanism in the neat PMDA-ODA hollow fiber membrane.²⁸ The gas separation performance shown in Figure 6 further proved the Knudsen diffusion mechanism. Therefore, a dip-coating treatment was adopted to seal these defects.

The chemical structures of the PMDA-ODA based membranes

Figure 7 shows the FTIR spectra of the PMDA-ODA PAA, PI, and the PEI crosslinked PMDA-ODA PI. The characteristic peaks of PAA could be found at 1542 cm⁻¹ (C-N stretching) and 1655 cm⁻¹ (C=O stretching). After the thermally induced imidization reaction, the two peaks disappeared. The observation of the new peaks at 1375 cm⁻¹ (C-N< stretching), 1720 cm⁻¹ (>C=O stretching), and 1779 cm⁻¹ (>C=O stretching) indicated the formation of the imide group. The result proved the successful formation of polyimide. After crosslink, the peaks of amide structures were observed at 1542 cm⁻¹ (C-N stretching) and 1637 cm⁻¹ (C=O stretching).^29–31 The two peaks proved the occurrence of the PEI crosslinking reaction (Figure 3). Note that, the peaks at 1382 cm⁻¹ (C-N<stretching), 1720 cm⁻¹ (>C=O stretching), and 1779 cm⁻¹ (>C=O stretching) still existed, which indicated that some imide groups remained in the crosslinked PI.

Effects of crosslinking on the thermal stability of the PMDA-ODA PI

As shown in Figure 8, both the crosslinked and uncrosslinked PIs had insignificant weight losses at temperature below 250°C. This could be attributed to the evaporation of absorbed water or high boiling point solvent such as NMP. At higher temperatures, the crosslinked PI showed a better thermal stability. The uncrosslinked PI started to decompose at 500°C.This was consistent with the reported T_d for PMDA-ODA PI.^32,33 However, the crosslinked PI begun to decompose at 550°C. Clearly, the PEI crosslinking improved the thermal stability of the PMDA-ODA PI. And the improved thermal stability of the crosslinked PI would alleviate the pore collapsing of the hollow fiber membrane during carbonization.

As shown in Figure 9, the uncrosslinked PMDA-ODA polyimide hollow fiber had a T_g of 392°C. After crosslinking, the T_g disappeared that indicated a thermal-set nature of the crosslinked PMDA-ODA hollow fiber membrane. Therefore, when the crosslinked polyimide hollow fiber was carbonized, the glassy to rubbery state transition would not happen. So, the polymer chains would be rigid during carbonization and the pore-collapsing phenomena should be mitigated.

Effects of crosslinking on suppressing the pore collapsing during the carbonization process

It is well known that polymer tends to shrink during carbonization.³² For porous membranes, pores would collapse during heating. This will aggravate the shrink ratio of the polymer and cause delamination or defect formation of the composite carbon molecular sieve membrane.²⁷ In addition, pore collapsing will lead to densification of the hollow fiber substrate and decrease the gas permeance. According to Figure 10(a), the PMDA-ODA hollow fiber membrane had an outer diameter of 665 μm. After carbonization, it reduced to 518 μm (Figure 10(b)). However, the PEI crosslinked hollow fiber had an outer diameter of 604 μm (Figure 10(c)) after carbonization. As listed in Table 2, the shrinkage% of the outer diameter, membrane thickness, and inner diameter of the uncrosslinked and crosslinked hollow fiber membranes were 22.21%, 16.03%, and 27.4% and 9.17%, 6.41%, and 11.58%, respectively. The results proved the effectiveness of crosslinking in inhibiting the membrane shrinkage during the carbonization process. The CO₂ permeances of the uncrosslinked and crosslinked CMS hollow fiber membranes had significant difference. The crosslinked CMS membrane showed 4-fold higher gas permeances (4231.4±334.2 GPU) to the uncrosslinked CMS membrane (1102.9±158.9 GPU). Again, it proved that the crosslinking treatment inhibited the pore collapsing during carbonization and increased the gas permeance of the CMS membrane.

Table 2.

Dimensions of the hollow fiber membranes.

Code	Outer diameter	Fiber wall	Inner diameter
PI	665±20 μm	156±18 μm	354±23 μm
CMS membranes without crosslinking	518±30 μm	131±27 μm	257±21 μm
CMS membranes with crosslinking	604±24 μm	146±15 μm	313±26 μm
Shrinkage%	22.21%→9.17%	16.03%→6.41%	27.4%→11.58%

In the carbonization process, the skin layer would be thickened due to the pore collapsing. Figures 10(d–f) show the thicknesses of the skin layers of different hollow fiber membranes. The skin layer of the CMS hollow fiber without crosslinking was thickest. After the PEI crosslinking, the CMS hollow fiber membrane showed a thinner skin layer thickness than the uncrosslinked sample. This justified our estimation that the PEI crosslinking could alleviate pore collapsing of the Kapton hollow fiber membrane and maintain its cross-section morphology during carbonization. Note that, both the CMS membranes showed a typical Knudsen diffusion behavior that indicated the existence of defects. Later, we would show that these defects could be repaired by dip-coating.

Gas separation performance

To seal the defects of the PMDA-ODA PI hollow fiber membrane, we dip-coated the membranes using the PMDA-ODA PAA coating solutions with different PAA concentrations. As listed in Table 3, as the PAA concentration increased, all gas permeances decreased with an increase in gas selectivity. This indicated the surface defects were gradually sealed. When the PAA concentration was 2wt%, the gas separation mechanism still followed the Knudsen diffusion. This was due to that the dilute coating solution could not seal all the surface defects of the hollow fiber membrane.²⁶ Thus, the gas selectivity was low. As the PAA concentration of the coating solution increased, the more viscose coating solution could seal all the surface defects so that the gas selectivity increased greatly and reached to the intrinsic gas selectivity of the PMDA-ODA polyimide. When the PAA concentration was 5 wt%, the CMS composite hollow fiber membrane showed gas permeances of 93.4 GPU, 19.6 GPU, 6.5 GPU, and 4.7 GPU for CO₂, O₂, N₂, and CH₄, respectively. And the ideal selectivities to CO₂/CH₄, CO₂/N₂, and O₂/N₂ gas pairs were 19.8, 14.4, and 3.0, respectively. Since the kinetic diameters of the tested gases decreased in an order of CH₄>N₂>O₂>CO₂, the result indicated that the gas transport in the CMS membranes followed a typical molecular sieving mechanism.

Table 3.

The gas separation performance of the CMS composite membranes.

PAA Content (wt%)	Gas permeance (GPU)				Selectivity(α)
PAA Content (wt%)	CO₂	O₂	N₂	CH₄	CO₂/CH₄	CO₂/N₂	O₂/N₂
2	605±201	928±228	999±277	1134±333	0.5	0.6	0.9
3	417±155	213±54	150±31	127±24	3.2±0.6	2.7±0.5	1.4±0.1
4	180±41	61±22	28±4	24±5	7.4±0.2	6.5±0.5	2.2±0.5
5	93.4±11.3	19.6±3.8	6.5±1.6	4.7±1.7	19.8±2.4	14.4±1.7	3±0.4

To optimize the separation performance, the PMDA-ODA hollow fibers were coated using the 5 wt% PAA solution by multiple times and then carbonized. When increasing the coating times, there was a higher chance to seal the surface defects but at the cost of the increase in the coating layer thickness. As a result, the hollow fiber composite membrane showed higher gas selectivity and lower permeance. As listed in Table 4, after the second coating, gas permeabilities dropped approximately 50% and the CO₂/CH₄ selectivity increased by 2.3. After the third coating, gas selectivities were barely changed but gas permeances dropped significantly. Figure 11 shows that the coating layer thickness was about 764 nm for the CMS membrane after one-time coating. Consider that commercial gas separation membranes typically have a skin layer thickness below 1 μm, multiple coating is not suggested in this study.

Table 4.

Gas separation performance of CMS membranes based on different coating times (PAA content 5 wt%).

Coating times	Gas permeance(GPU)				Slectivity(α)
Coating times	CO₂	O₂	N₂	CH₄	CO₂/CH₄	CO₂/N₂	O₂/N₂
1	93.4±11.3	19.6±1.8	6.5±0.6	4.7±0.7	19.8±2.4	14.4±1.7	3.0±0.4
2	43.7±5.6	11.3±1.1	2.5±0.4	2±0.2	22.1±1.8	17.5±1.1	4.5±0.3
3	9.2±1.3	2.7±0.3	0.5±0.1	0.4±0.1	23.1±0.6	18.1±0.6	5.3±0.1

Evaluating the gas separation performance of the Kapton CMS hollow fiber membrane

As listed in Table 5, gas transport performances of the PMDA-ODA CMS composite membranes are comparable to literature reported data.^26,32,34 Our PMDA-ODA CMS membranes show better performance than the flat sheet Kapton CMS membranes. This is due to the thinner dense layer of our membrane. Note that, this is the first time that CMS Kapton hollow fiber membranes are successfully prepared for gas separation. And we believe the fabrication method can be applied to other CMS hollow fiber membranes spun from polyamic acids.

Table 5.

Comparison of the gas separation performances of typical CMS membranes.

Precursor	configuration	Gas pair	Selectivity	Permeance (GPU)	Reference
Kapton	Dense film	CO₂/N₂ O₂/N₂ CO₂/CH₄	8.6 1.7 9	25.7 4.95	[34]
Kapton	Flat composite	CO₂/N₂ O₂/N₂	12.3 10.4	25.9 21.9	[26]
Kapton	Dense film	CO₂/N₂ O₂/N₂	22.2 4.7	82.7 17.4	[32]
6FDA:BPDA-DAM	Hollow fiber	CO₂/CH₄	58.8	310	[13]
Matrimid 5218	Hollow fiber	CO₂/CH₄	69-83	11-13	[35]
6FDA/BPDA-TMPD	Hollow fiber	CO₂/N₂ O₂/N₂	55-56 8.5-11.5	20-50	[36]
PAN	Hollow fiber	O₂/N₂	1.3–3.7	6.8–71.7	[9]
Phenolic resin	Tubular membrane	O₂/N₂	1.8–2.8	0.75-1	[37]
Phenolic resin	Tubular membrane	O₂/N₂ CO₂/CH₄	10.8 170	6.8 31	[38]
Kapton-this work	Hollow fiber	CO₂/N₂ O₂/N₂ CO₂/CH₄	14.4 3 19.8	93.4 19.6	1^st coating
Kapton-this work	Hollow fiber	CO₂/N₂ O₂/N₂ CO₂/CH₄	17.5 4.5 22.1	43.7 11.3	2^nd coating

Conclusion

A series of PMDA-ODA CMS hollow fiber composite membranes have been successfully prepared. The PEI chemical crosslinking agent is very effective to suppress the membrane shrinkage during carbonization. The surface defects of the CMS membrane can be sealed by dip-coating. By adjusting the coating condition, the CMS membranes exhibit gas separation properties comparable to literature reported data for CMS membranes. This work provides a guidance to prepare CMS hollow fiber membranes derived from polyamic acid polymers.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This research is funded by National Natural Science Foundation of China (51773011).

ORCID iD

Pei Li

References

Zhang

. Preparation and gas separation properties of spirobichroman-based polyimides. Macromol Chem Phys 2018; 219: 1800157.

Han

Zhang

Zhao

, et al. Post-modification of PIM-1 and simultaneously in situ synthesis of porous polymer networks into PIM-1 matrix to enhance CO₂ separation performance. J Membr Sci 2021; 636: 119544.

Xue

Huang

Lau

, et al. Tailoring the molecular structure of crosslinked polymers for pervaporation desalination. Nat Comm 2020; 11: 1461.

Meng

Lau

Xue

, et al. Compatibilizing hydrophilic and hydrophobic polymers via spray coating for desalination. J Mater Chem A 2020; 8: 8462–8468.

Zhang

. Synthesis and characterization of bis (phenyl) fluorene-based cardo polyimide membranes for H₂/CH₄ separation. J Mater Sci 2019; 54: 10560–10569.

Zhang

Deng

. Decarboxylation cross-linking of triptycene-based tröger’s base polymers for gas separation. Ind Eng Chem Res 2020; 59: 18640–18648.

Bird

Trimm

. Carbon molecular sieves used in gas separation membranes. Carbon 1983; 21: 177–180.

Steel

Koros

. Investigation of porosity of carbon materials and related effects on gas separation properties. Carbon 2003; 41: 253–266.

David

LIB

Ismail

. Influence of the thermostabilization process and soak time during pyrolysis process on the polyacrylonitrile carbon membranes for O₂/N₂ separation. J Membr Sci 2003; 213: 285–291.

10.

Rao

Sircar

. Nanoporous carbon membranes for separation of gas mixtures by selective surface flow. J Membr Sci 1993; 85: 253–264.

11.

Acharya

Foley

. Spray-coating of nanoporous carbon membranes for air separation. J Membr Sci 1999; 161: 1–5.

12.

Centeno

Fuertes

. Supported carbon molecular sieve membranes based on a phenolic resin. J Membr Sci 1999; 160: 201–211.

13.

Cao

Zhang

Sanyal

, et al. Carbon molecular sieve membrane preparation by economical coating and pyrolysis of porous polymer hollow fibers. Angew Chem Int Ed 2019; 58: 12149–12153.

14.

Barsema

Van der Vegt

NFA

Koops

, et al. Carbon molecular sieve membranes prepared from porous fiber precursor. J Membr Sci 2002; 205: 239–246.

15.

Favvas

Kapantaidakis

Nolan

, et al. Preparation, characterization and gas permeation properties of carbon hollow fiber membranes based on Matrimid ® 5218 precursor. J Mater Process Technol 2007; 186: 102–110.

16.

Zhang

Wenz

Williams

, et al. Purification of aggressive supercritical natural gas using carbon molecular sieve hollow fiber membranes. Ind Eng Chem Res 2017; 56: 10482–10490.

17.

Zhang

Weng

. Preparation and gas separation properties of triptycene-based microporous polyimide. Macromol Chem Phys 2019; 220: 1900047.

18.

Ragosta

Abbate

Musto

, et al. Effect of the chemical structure of aromatic polyimides on their thermal aging, relaxation behavior and mechanical properties. J Mater Sci 2012; 47: 2637–2647.

19.

Zhao

Yong

, et al. Highly solvent-durable thin-film molecular sieve membranes with insoluble polyimide nanofibrous substrate. Chem Eng J 2021; 409: 128206.

20.

Ogawa

Nakano

. Gas permeation through carbonized hollow fiber membranes prepared by gel modification of polyamic acid. J Membr Sci 1999; 162: 189–198.

21.

Cao

. Fabrication of PMDA-ODA hollow fibers with regular cross-section morphologies and study on the formation mechanism. J Membr Sci 2017; 544: 1–11.

22.

Yang

Zhang

, et al. Preparation of thermally imidized polyimide nanofiltration membranes with macrovoid-free structures. Ind Eng Chem Res 2020; 59: 14096–14105.

23.

Woo

Lee

Dong

, et al. Fabrication of thermally rearranged (TR) polybenzoxazole hollow fibermembranes with superior CO₂/N₂ separation performance. J Membr Sci 2015; 490: 129–138.

24.

Kim

Han

Lee

. Thermally rearranged (TR) polybenzoxazole hollow fiber membranes for CO₂ capture. J Membr Sci 2012; 403-404: 169–178.

25.

Liu

Cao

. Effects of spinning temperature on the morphology and performance of Poly(ether sulfone) gas separation hollow fiber membranes. Ind Eng Chem Res 2018; 57: 329–338.

26.

Song

Jiang

, et al. Preparation and gas separation performance of supported carbon membranes with ordered mesoporous carbon interlayer. J Membr Sci 2014; 450: 469–477.

27.

Wei

Xia

Liu

, et al. Interfacial adhesion between polymer separation layer and ceramic support for composite membrane. AIChE J 2010; 56: 1584–1592.

28.

Gilron

Soffer

. Knudsen diffusion in microporous carbon membranes with molecular sieving character. J Membr Sci 2002; 209: 339–352.

29.

Gao

Sun

Zhu

, et al. Polyethyleneimine (PEI) cross-linked P84 nanofiltration (NF) hollow fiber membranes for Pb²⁺ removal. J Membr Sci 2014; 452: 300–310.

30.

Shao

Chung

Pramoda

, et al. Polyimide modification by a linear aliphatic diamine to enhance transport performance and plasticization resistance. J Membr Sci 2005; 256: 46–56.

31.

Wang

Sun

. Designing scalable dual-layer composite hollow fiber nanofiltration membranes with fully cross-linked ultrathin functional layer[J]. J Membr Sci 2021; 628: 119243.

32.

Suda

Haraya

. Gas permeation through micropores of carbon molecular sieve membranes derived from Kapton polyimide. J Phys Chem B 1997; 101: 3988–3994.

33.

Boroglu

Gurkaynak

. The preparation of novel silica modified polyimide membranes: synthesis,characterization, and gas separation properties. Polym Adv Technol 2011; 22: 545–553.

34.

Lua

. Effects of carbonisation on pore evolution and gas permeation properties of carbon membranes from Kapton polyimide. Carbon 2006; 44: 2964–2972.

35.

Koros

Miller

. High pressure CO2/CH4 separation using carbon molecular sieve hollow fiber membranes. Ind Eng Chem Res 2002; 41: 367–380.

36.

Jones

Koros

. Carbon molecular sieve gas separation membranes I––preparation and characterization based on polyimide precursors. Carbon 1994; 32: 1419–1425.

37.

Wei

Qin

, et al. Preparation of supported carbon molecular sieve membrane from novolac phenol–formaldehyde resin. J Membr Sci 2007; 303: 80–85.

38.

Zhou

Yoshino

Kita

, et al. Preparation and gas permeation properties of carbon molecular sieve membranes based on sulfonated phenolic resin. J Membr Sci 2003; 217: 55–67.