Membrane separation technology has evolved significantly because of its relatively easy operation and low-energy needs. 1,2 High-performance membranes should be constructed with materials that simultaneously provide high permeability, good selectivity, and robust stability. [2][3][4] Two-dimensional (2D) covalent organic frameworks (COFs) are porous crystalline polymers with periodic structures linked by strong covalent bonds and have received tremendous attention recently. 5 Their unique properties, including highly ordered in-plane pores, tunable functionalities, and robust thermal and chemical stabilities, [6][7][8] make them great candidate materials for developing next-generation membrane-based separation technologies. Unlike other crystalline porous materials (e.g., zeolites), the wide diversity of available COF monomers confers unique functionalities and applications derived from their tailorable pore sizes, geometries, and crystallinity. [9][10][11] However, their relatively large pore sizes compared to the common gases' kinetic diameters (0.25-0.5 nm) 8,12,13 and rather complex membrane fabrication methods pose challenges to their use in gas separations.
There have been a limited number of studies 10,[13][14][15][16][17][18][19][20] on microporous 2D COF membranes for gas separations (see Table S1 in the Supporting Information for a detailed literature review). In most of these previous studies, the COF membrane syntheses involved the use of mesoporous inorganic supports including anodic alumina oxide (AAO) and modified α-Al 2 O 3 . One of the earliest studies was performed by Li et al., 10 who exfoliated COF-1 (1.5 nm diameter pore) into 2D nanosheets first and then assembled these nanosheets into an ∼100 nm thick membrane on a modified α-Al 2 O 3 support, which reported an ideal H 2 /N 2 selectivity of ∼3.3. In another study, Fan et al. 21 synthesized an ∼8 μm thick azine-based ACOF-1 (0.94 nm diameter pore) membrane by in situ solvothermal growth on an α-Al 2 O 3 support. The resulting membrane showed effective pore size narrowing due to crystal intergrowth and exhibited molecular sieving with a high CO 2 / CH 4 separation factor of ∼86.
Other efforts 14,17,19,22 have also been explored to reduce 2D COF membranes' effective pore sizes for improving their gas separation performance. One approach is to stack different pore-sized 2D COFs in a layer-by-layer fashion (Table S1). This strategy has been employed to prepare multilayer COF membranes with reduced effective pore sizes on inorganic supports such as AAO. The resulting multilayer COF membranes reported remarkable H 2 /CO 2 separation factors of 23-26, 14,19,22 thus holding promise for high-performance H 2 separation. Another commonly explored approach is the staggered stacking of single-type 2D COF nanosheets. For example, Wang et al. 17 prepared ultrathin 2D COF membranes using bulk COF powders (pore size range of 1.3-1.8 nm) as precursors via a hot drop coating method. The staggered stacking reduced the effective membrane pore size and endowed the membranes with selective CO 2 separation (CO 2 /H 2 separation factor ∼ 22). Recently, Guo et al. 18 prepared ionic Dha-TG Cl (1.5 nm diameter pore) 2D COF membranes by vacuum filtration of ionic COF nanosheet colloidal suspensions. The resulting ionic membranes reported a separation factor of ∼21 for CO 2 /CH 4 under humidified conditions, which was further improved to ∼49 by subsequent COF pore-wall functionalization, enabling facilitated and selective CO 2 transport. 18 Research on conventional polyamide (PA)-based gas separation membranes prepared by scalable liquid-liquid (L-L) interfacial polymerization (IP) has made significant progress over the past decade (Table S2). 24 These PA-based thin-film composite membranes (with selective PA layer being 100-600 nm thick) have been tested for CO 2 /N 2 , 25 CO 2 / CH 4 , 26 H 2 /CO 2 , 27 and O 2 /N 2 28 separations, reporting high selectivities of 70, 94, 50, and 10 for these respective gas mixtures. Notably, Ali et al. reported a PA membrane with a mixed-gas H 2 /CO 2 selectivity of 50 and a H 2 permeability of ∼70 Barrer at 140 °C. 27 In general, the gas permeabilities of these PA-based membranes are relatively low (2-17 Barrer) when compared to those of the microporous 2D COF membranes prepared by the same IP method.
Previous 2D COF membrane gas separation studies 10,13-23 have focused more on the synthesis of microporous COF membranes but less on their fundamental gas transport and separation mechanisms. For example, previous studies have studied only a few gas mixtures (e.g., H 2 /CO 2 , CH 4 /CO 2 , or N 2 /CO 2 ) separately 14,18,21,22 (Table S1), and the possible concurrent gas transport mechanisms and their relative contributions through a particular type of 2D COF membrane still require systematic studies. In addition, most of these previous studies have relied on relatively complex and lengthy membrane preparation strategies (Table S1) and used expensive, brittle, inorganic porous supports. Although microporous 2D COF membranes have been synthesized on flexible polymeric supports using simpler methods such as substrateassisted L-L IP, the assessment of their separation performance has focused on the separations of larger molecules such as dye sieving, 24 desalination, 25 and alcohol dehydration. 26 Therefore, the underlying gas transport/separation mechanisms of microporous 2D COF membranes, especially those prepared by easier and more scalable fabrication methods such as L-L IP, remain unclear and need to be studied.
In this study, we prepared TpHz (a commonly reported microporous β-ketoenamine-linked 2D COF 25,27 ) membranes on a flexible polymeric support (nylon 6,6) by L-L IP growth and characterized their structural properties (Figure 1). We then systematically studied their gas permeation and separation characteristics under both single and binary gas feed conditions to elucidate the underlying gas transport mechanisms and the effects of gas species/COF's pore-wall interactions. Greenhouse gases such as CO 2 and CH 4 were examined in the present study due to their significance in global warming and their roles in the heating values of natural gas important to the fuel-energy industries. 28,29 The gas permeation/separation performance of these TpHz membranes was correlated to the membrane pore structures (e.g., pore size), membrane gas adsorption properties, and gas permeation conditions (e.g., temperature). The gas transport mechanisms through these supported TpHz membranes were rationalized by Knudsen diffusion and preferential surface flow mechanisms.
2.1. Materials and Membrane Synthesis. 2.1.1. Surface Treatment of Polymeric Support. Nylon 6,6 disks (47 mm diameter coupons, 0.1 μm nominal pore size, ∼114 μm thick, Sterlitech) 30,31 were used as the polymeric support for TpHz membranes. The nylon support was immersed in a 0.25 wt % polyethylenimine (M n ∼ 60 kg mol -1 , 50 wt % in H 2 O, Sigma-Aldrich) aqueous solution for 10 min and then dried at 60 °C in air for 10 min. This cycle was repeated 5 times to yield a uniform, colorless polyethyleneimine coating after drying overnight under ambient conditions. The modified nylon support was then fixed vertically in a glass diffusion cell for subsequent membrane fabrication.
2.1.2. Fabrication of TpHz Membrane. TpHz membranes were fabricated by substrate-assisted L-L IP growth of TpHz atop of nylon supports, following a synthesis protocol reported by Wang et al. 25 with minor modifications. Briefly, 0.1 mmol (21.6 mg) of Tp (2,4,6-triformylfloroglucinol, 99.12%, ChemScene) monomers were dissolved in 80 mL of n-hexane (Acros Organics, 99%) to form the organic phase. To form the aqueous phase, 15.4 mmol (748 μL) of Hz•H 2 O (hydrazine monohydrate, 98%, Sigma-Aldrich) monomers and 15.4 mmol (2936 mg) of PTSA•H 2 O (catalyst, p-toluene sulfonic acid monohydrate, 99%, Sigma-Aldrich) were dissolved in 80 mL of deionized (DI) water. Hz monomers in large excess were used to ensure a unidirectional diffusion of Hz during L-L IP growth of TpHz. 25,32,33 Both aqueous and organic solutions were sonicated (Fisherbrand, CPX2800) for 30 min and then transferred into the assembled nylon support/glass diffusion cell above (Figure 1a; see a zoom-in view of the interface in Figure S1 and more experimental details in the Supporting Information).
The first TpHz growth was carried out for 6 h without any external perturbation at room temperature. Subsequently, the membrane was dried at 60 °C for ∼12 h. The resulting dry TpHz-seeded nylon sample was then subjected to the second growth for 48 h under the same conditions as the first TpHz growth. Secondary growth is essential to the formation of a uniform and continuous thin TpHz selective layer. 25 After synthesis, the TpHz membranes were soaked thoroughly in DI water and anhydrous ethanol (95%, Fisher Chemicals) to remove residual precursors. Finally, the supported TpHz membranes were vacuum-dried at 40 °C for 24 h.
2.1.3. Syntheses of TpHz Support-Free Films and Bulk Powders. To gather sufficient TpHz film samples for subsequent characterizations, multiple large batches of support-free TpHz thin films were synthesized without the nylon support in 1 L beakers using the same L-L IP method. Note that the synthesis conditions closely mirrored those used for synthesizing the supported TpHz membranes. This approach, as seen in previous studies, 13,34 was utilized to address the challenge of characterizing the TpHz thin films obtained in small quantities via IP growth. TpHz bulk powders were also prepared by solvothermal synthesis (see the Supporting Information). 17 2.2. Material Characterization. Attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectra were collected by using a Perkin-Elmer Frontier spectrometer. All samples were background-corrected for H 2 O/CO 2 . The nylon-containing samples' spectra were normalized to the characteristic N-H stretching band (∼3300 cm -1 ) of nylon 6,6.
Powder X-ray diffraction (PXRD) patterns were collected with a Malvern PAN analytical Aeris unit equipped with a Cu Kα X-ray source operated at 40 kV and 15 mA and a PIXcel1D detector. The 2θ acquisition range was from 2 to 40°at a scan rate of ∼1.9°min -1 .
Grazing incidence wide-angle X-ray scattering (GIWAXS) measurements were performed on a Xenocs Xeuss 3.0 SAXS/ WAXS. A GeniX3D Cu high flux very long (HFVL) focus source at 50 kV and 0.6 mA generated X-ray radiation with a wavelength of 1.5419 Å. An EIGER2 R 1 M DECTRIS hybrid pixel photon counting detector, positioned 80 mm away from the sample, was employed to collect scattering signals at a grazing angle of 0.2°. The GIWAXS sample was prepared by transferring an IP-grown support-free TpHz film onto a smooth Si wafer required for sample alignment, rinsed with various organic solvents (i.e., tetrahydrofuran, acetone, and ethanol), and finally dried overnight under vacuum at room temperature.
Scanning electron microscope (SEM) images were obtained on a Zeiss Auriga SEM microscope operating at 5 kV. Prior to SEM characterizations, the samples were gold-coated using a 108 auto Cressington sputter coater. The coating time was 120 s, yielding an ∼20 nm thick gold coating. To prepare the samples for cross-sectional SEM imaging, they were frozen by immersion in liquid N 2 for 20 min and fractured by using a razor blade.
Nitrogen adsorption isotherms at 77 K were measured on a Micromeritics ASAP2020 instrument. The pore size distributions were calculated using the H-K model (Saito-Foley) for cylindrical pore geometry (carbon-graphite as a sorbent solid) using the MicroActive V6 software. The COF samples were activated at 120 °C under N 2 flow for 16 h before measurement.
Room-temperature gas (i.e., CO 2 , CH 4 , and N 2 ) adsorption experiments were performed on a Micromeritics Tristar II 3020 instrument. The samples were activated at 120 °C under N 2 flow for 16 h in a Micromeritics Vacprep 061 unit. The sample tube was immersed in DI water to avoid temperature fluctuations during analysis.
2.3. Single/Mixed-Gas Permeation Studies. Gas permeation experiments were conducted on a multicomponent gas permeation system (i.e., the Wicke-Kallenbach configuration; Figure 1b). A TpHz membrane was attached to a porous stainless steel disc (Mott Co., 0.2 μm pore size), 35 and the two stacked objects were sealed in between silicone Orings. The effective membrane area was 2.5 × 10 -4 m 2 . The temperature was controlled by a Vulcan 3-550 furnace, where the heating/cooling rate was 1 °C min -1 . The volumetric flow rates were controlled by mass flow controllers (MFC, Brooks Instruments, SLA5850A). On-stream membrane activation was performed at 100 °C for 5 h using helium (10 mL min -1 ) to remove any adsorbed gases prior to gas permeation experiments. 36 During permeation/separation measurement, a single gas or equimolar binary gas mixture was sent to the feed side at 50 mL min -1 . Ultra-high-purity argon (99.999%) was used as the sweep gas at 50 mL min -1 . The pressure of the feed side was kept at ∼1.2 atm by adjusting the needle valve. The permeate side pressure was kept at 1 atm. Because the COF membranes are microporous and the sweep gas is inert gas, the presence of argon should have negligible effects on single gas permeation. 19 Gas permeance was reported as the mean of at Industrial & Engineering Chemistry Research least three measurements with a standard deviation of < 3%. The gas compositions were determined by injecting gas aliquots into a gas chromatography unit (Agilent Technologies, 6890N) 37 coupled with thermal conductivity detectors (TCDs, 200 °C) and the column (HayeSep DB packing, 30 ft length, 1/8 in OD) held at 100 °C. The same ultra-high-purity argon was used as the carrier gas.
The permeance (F i ) of each gas species i was measured using eq 1
where Q is the total molar flow rate in the permeate side, y i is the mole fraction of gas i in the permeate side, x i is the mole fraction of gas i in the feed side, A is the effective membrane area, and P f and P p are the pressures on the feed and permeate sides, respectively. For the measurement of single gas permeation, x i = 1. The single gas selectivity (α SG ij ) was calculated using eq 2 38
The mixed-gas separation factor (α i/j ) was determined using eq 3
where y i and y j are the mole fractions of gases i and j in the permeate stream, and x i and x j are the mole fractions of gases i and j in the retentate stream, respectively.
The Knudsen selectivity (α i/j K ) was calculated using eq 4 39,40
where M i and M j are the molecular weights of gases i and j, respectively.
3.1. TpHz Membrane Characterization. Thin TpHz films were grown on an amino-modified nylon support placed in a glass diffusion cell (Figure 1a) that separates the organic phase containing Tp monomers from the aqueous phase containing Hz monomers and PTSA catalysts. A secondary COF growth method is employed to avoid gap-like defects generally observed with a single-step growth. 25 After TpHz membrane synthesis, the organic-phase side of the nylon support becomes uniformly red in color (Figure 2a, optical images), consistent with the characteristic color of TpHz. 25,27 The aqueous-phase side remains almost intact and is white in color, indicative of negligible TpHz growth on this side. These results demonstrate that the formation/growth of TpHz is solely confined on the organic-phase side caused by the unidirectional diffusion of the Hz monomers in large excess from the aqueous to the organic phase. 25,32,33 Cross-sectional SEM imaging (Figure 2a) shows that the assynthesized TpHz COF selective layer is compact and dense in nature, reporting a thickness of ∼290 nm on average. No obvious pinholes, cracks, or uncovered areas are present throughout the entire TpHz layer. There are tiny TpHz flakes/ particles randomly deposited on the walls of nylon support's large cavities. However, they are not expected to fully obstruct the macroporous structure of the nylon support, 41 considering the relatively low TpHz yield (∼10%) after IP synthesis and the relatively large void fraction of the nylon support (75-80% by volume). In addition, deep intrusion of the organic aldehyde/n-hexane solution into the porous nylon support fully wetted by the aqueous solution is minimized by the unidirectional diffusion of the Hz monomers in large excess from the aqueous to the organic phase, thus confining the TpHz growth at the L-L interface. Figure 2b displays the FTIR spectra of the support-free TpHz films, the nylon-supported TpHz membrane, and the nylon support. The synthesis conditions for the support-free TpHz films were almost identical to those for synthesizing the nylon-supported TpHz membranes, except for the absence of a nylon support. Therefore, it is reasonable to anticipate that the support-free TpHz film's characteristics closely resemble, if not exactly match, those of the TpHz layer in the membrane. The FTIR spectrum of the support-free films shows the characteristic bands of TpHz COF in the ketoenamine form, i.e., C�C stretching band at ∼1588 cm -1 , aromatic C-C ring stretching band at ∼1450 cm -1 , and C-N stretching band at 1278 cm -1 . 42 Such peaks are also found in the previously reported FTIR spectra of TpHz membranes and solvothermally synthesized TpHz bulk powders. 9,25,42 The intensified C-N stretching peak at 1278 cm -1 in the nylon-supported TpHz membrane in Figure 2b could be attributed to the newly formed C-N linkage after TpHz synthesis, although this signal is also intrinsic to the nylon support.
Figure 2c shows the PXRD patterns of support-free TpHz films, a nylon-supported TpHz membrane, and a nylon support. The PXRD patterns for the nylon-supported TpHz membrane and nylon support are almost identical, indicating that nylon support's PXRD signals 30,31 pose a masking effect on the PXRD signals of the TpHz layer. The PXRD pattern of the nylon-supported TpHz membrane obtained in the Bragg-Brentano geometry (i.e., reflection mode) does not show discernible peaks from the thin TpHz COF layer, possibly due to its much smaller sample mass relative to the semicrystalline nylon support and a relatively large X-ray penetration depth during measurement. This phenomenon has been reported in many other studies 14,19,25,27,43,44 on COF thin films prepared by similar IP growth. 12,45,46 Hence, the PXRD results of the support-free TpHz films are examined instead, with the assumption that their microstructure should resemble that of the TpHz layer in the membrane synthesized under the same conditions. Specifically, the PXRD pattern of the support-free TpHz film in Figure 2c shows a relatively broad and weak peak at 2θ ∼ 6.9°(100), indicative of a crystallinity lower than that of the TpHz bulk powders prepared by solvothermal synthesis (Figure S2). The (110) and (001) reflection planes are, respectively, identified at 2θ ∼ 12°(110) and 2θ ∼ 27°(001), consistent with the previous literature on TpHz lamellar crystallites in an eclipsed (AA) stacking geometry. 17,25,42,47 To gain further qualitative insight into the microstructure of the thin TpHz film, GIWAXS was measured. Figure 2d shows the wedge-corrected GIWAXS profile of an ∼400 nm thick support-free TpHz film prepared by the IP method and subsequently transferred onto a Si wafer. The (001) peak at q = 1.84 Å -1 shows a stronger intensity toward q z than q xy , indicating that the c-axis of the COF crystallites (i.e., the direction parallel to the COF's pore channels) tends to align with the film's normal direction. This anisotropy is also evident from the in-plane and out-of-plane one-dimensional (1D) profiles (Figure S3). In addition to the stronger (001) peak along q z , the in-plane scattering shows a strong (100) peak at q xy = 0.48 Å -1 (Figure S3), which is coming from the q y direction (perpendicular to the detector plane, not readily visible in the 2D profile). The presence of the (100) peak along q xy further supports that COF's pore channels prefer to align perpendicular to the Si substrate (also TpHz film) surface, [48][49][50] although the alignment was not perfect (see the Supporting Information for a more detailed discussion).
Figure 3a shows the N 2 adsorption isotherm at 77 K for the support-free TpHz film. It exhibits combined type I and type IV isotherm characteristics, indicating the formation of both micropores and interstitial voids between 2D and threedimensional (3D) COF lamellar crystallites. The BET surface area of this TpHz film is approximately 45 m 2 g -1 , whereas the BET surface area of the TpHz bulk powders is around 267 m 2 g -1 (Figure S4). The lower surface area of the TpHz film can be attributed to its lower crystallinity as compared to that of the TpHz bulk powder (Figures 2c vs S2). Moreover, these structural differences can be explained by the significantly distinct polymerization conditions, such as the concentration, temperature, and solvents. A comparison between the unidirectional L-L IP used to produce support-free TpHz films and the solvothermal method employed for producing the TpHz bulk powders reveals the underlying reasons. The synthesis at higher temperatures favors the reversibility of the ketoenamine condensation reaction, facilitating the availability of reactive monomer and its diffusion during crystallization. This, in turn, enables a better error-correction process within the cross-linked networks formed during the solvothermal process than the one obtained in the IP method. Note that such structural differences 51 induced by interfacial growth vs solvothermal synthesis have been reported in many moderately crystalline 2D COF systems, including TpHz, 12,26,27,52,53 TpPa-1, 54,55 and ACOF-1 34 (Table S3).
The pore size distribution of the TpHz film (Figure 3b) is broader and has a higher mesopore fraction than that of the TpHz bulk powder (Figure S4), yet it exhibits a peak centered at ∼1.1 nm, which is close to the TpHz's intrinsic channel size
predicted by AA-stacking structural models. 9,47,56 Notably, such broadening effects of the pore size distribution of moderately crystalline 2D COF films, as compared to their bulk powder counterparts, have been observed in the previous literature. 34,54 Since the crystallinity of the TpHz films is lower than that of the solvothermally synthesized TpHz bulk powders (thermodynamically favored), the TpHz COF films could contain a larger fraction of nanosized defects from the dangling bonds in 2D COF networks that fail to undergo sufficient error correction due to the lower synthesis reversibility during interfacial growth. 44,[57][58][59] 3.2. Gas Permeation and Separation. Figure 4a shows the permeance (F) of five single gases (i.e., H 2 , He, CH 4 , N 2 , and CO 2 ) through the synthesized TpHz membrane at room temperature. Notably, F(He) through the assembled membrane (∼9.5 × 10 -8 mol s -1 m -2 Pa -1 ) is more than 1 order of magnitude lower than that through the combined porous metal disk and nylon support (∼1.4 × 10 -6 mol s -1 m -2 Pa -1 ). This indicates that the TpHz COF selective layer dictates the overall gas transport resistance of the assembled membrane, which is reasonable considering the microporous nature of the TpHz COF layer vs the macroporous nature of the metal disk and nylon support.
As shown in Figure 4a, the single gas permeance decreases as the gas molecular weight (M i ) increases in the following order:
, where F i scaled reasonably well with M i -0.5
. This suggests that the transport of these single gases through the TpHz membrane is dominated by a Knudsen diffusion regime 39,60,61 (eq 5)
where μ k is the reciprocal of tortuosity, r is the mean pore radius of the membrane, ε is the membrane porosity, L is the membrane thickness, R is the gas constant, and T is the permeation temperature. This observation is reasonable considering the mean free path length of the studied gases ranging from 55 to 137 nm at ambient conditions 40 and considering that the pore size distribution of the TpHz film (Figure 3b) exhibits a small fraction of mesopores up to ∼2.5 nm. Notably, the lowest possible Knudsen number (defined as the ratio of the mean free path length and pore diameter), i.e., 55/2.5 = 22, is larger than the Knudsen diffusion regime threshold number, i.e., 10. Moreover, it is worth pointing out that the H 2 permeability of the TpHz membrane synthesized in this study is ∼113 Barrer, comparable to those of the previously reported multilayer 2D COF membranes (105-143 Barrer). 14,19 Such a good agreement in permeability further supports the idea that the macroporous structure of the nylon support is almost not altered after IP growth of TpHz, as discussed previously in Section 3.1.
Figure 4b shows the single gas permeance data of the five different gases through the TpHz membrane at various temperatures. For all five gases, the single gas permeance decreases slightly as the temperature increases. Notably, the order of F(H 2 ) > F(He) > F(CH 4 ) > F(N 2 ) > F(CO 2 ) was held at any given temperature tested in this study. These results further provide evidence that these single gases permeate through the TpHz membrane following the Knudsen diffusion characteristics. It is worth noting that gas permeation is best described by the Knudsen mechanism, which is not an activated transport process with an activation energy. Thus, the calculation of the activation energy for permeation by Knudsen diffusion does not have a physical meaning.
Table 1 compiles the single gas selectivity calculated from the single gas permeance data (eq 2), together with the theoretical Knudsen selectivity (eq 4). Note that selectivity in most of the gas pairs in this study is reported with respect to the lighter gas. The most significant variations from the Knudsen selectivity are observed for the gas pairs involving a single-CO 2 permeance. This could be indicative of additional gas/COF interactions that may affect their mass transport. Note: reported gas permeance is the mean of at least three measurements with a standard deviation of < 3%.
To gain further insight into the concurrent gas interactions with the TpHz membrane, equimolar mixtures of CH 4 /He, CO 2 /He, CH 4 /N 2 , and CO 2 /N 2 were used for binary gas permeation experiments. Table 2 summarizes the experimentally measured permeance for the individual gas species at room temperature and the corresponding separation factors (α i/j ) calculated by eq 3. Notably, the individual permeance of most gases is reduced in the binary feed state compared to that in the single feed state. The permeance reduction is more pronounced for the noninteracting helium gas (24-29%) in the CO 2 /He and CH 4 /He mixtures. On the other hand, the individual CO 2 permeance within the CO 2 /He mixture remains nearly unchanged (only an ∼3% decrease) when compared to its permeance in the single gas transport. As a result, the CO 2 /He separation factor (α i/j ∼ 0.35) was higher than the selectivity calculated from the single gas permeance measurements (0.25) and its Knudsen selectivity (0.30).
Similarly, in the equimolar binary CO 2 /N 2 gas mixture, a relatively large decline in N 2 permeance (∼24%) is observed. Meanwhile, the CO 2 flux is slightly higher than that observed for the single gas permeance. Note that the separation factor (0.90) measured for the binary CO 2 /N 2 mixture is higher than the selectivity calculated from the single gas measurements (0.63) and the gas pair's Knudsen selectivity (0.79). All of these observations highlight intriguing gas transport outcomes arising from the existence of diverse gas species in the binary gas permeation. Section 3.3 provides a detailed discussion of our gas permeation/separation results.
For the equimolar CH 4 /N 2 mixture, a discrete decrease of individual gas permeance is observed (Table 2) and the measured separation factor (1.05) is smaller than the Knudsen selectivity (1.32) and fairly similar to the calculated selectivity from single gas permeance (1.11), possibly due to the limited difference in the TpHz film's CH 4 and N 2 gas uptakes at low pressure, which will be discussed in detail in Section 3.3.
Binary CH 4 /N 2 and CO 2 /N 2 separation studies were also performed at various temperatures ranging from 298 to 363 K to gain deeper insight into the role of permeating gas/TpHz interactions in dictating the separation performance. Figure 5a shows the T-dependent permeation data of the CH 4 /N 2 binary mixture and the corresponding separation factors. The lighter (and thus more mobile) CH 4 exhibits a larger permeance than N 2 in the CH 4 /N 2 binary mixture. In addition, the individual gas permeance of both CH 4 and N 2 in the CH 4 /N 2 mixture decreases as the permeation T increases, resulting in a separation factor almost invariant with T and smaller than the Knudsen selectivity. These observations suggest that these two gases concurrently permeate through the TpHz membrane by Knudsen diffusion (eq 5), where an inverse relationship between the permeance and T is expected.
Figure 5b shows the T-dependent permeation data of the CO 2 /N 2 binary mixture and the corresponding separation factors. At room temperature, the separation factor (0.90) is higher than the calculated Knudsen selectivity (0.79), indicative of additional transport pathways for CO 2 within the TpHz membrane. As T increases, the individual CO 2 permeance in the CO 2 /N 2 mixture decreases slightly, whereas the individual N 2 permeance increases slightly, resulting in a decrease in the separation factor. Notably, the measured CO 2 / N 2 separation factor approaches the Knudsen selectivity (0.79; Table 2) at T > 340 K and appears to plateau at higher T.
Our results of the gas permeance through the synthesized TpHz membrane could be described qualitatively by eq 6, which correlates the Knudsen diffusion and surface flow contributions, assuming negligible viscous flow 39,40,61 where v is the average gas molecule velocity (which is equal to (8RT/πM i ) 1/2 ), ρ is the membrane density, D s is the surface diffusion coefficient, k s is the tortuosity of the surface, and dq/ dP is the average slope of the gas adsorption isotherm.
From Table 1 and Figure 4, it is observed that the single gas CO 2 permeance aligns better with the Knudsen straight line. Such an effect in the single gas state is the opposite of some previous reports, where similar 2D COF membranes 17,21 prepared by different methods and other microporous inorganic membranes 38 (e.g., zeolites and MOFs) could exhibit additional CO 2 flux by surface flow due to preferential CO 2 adsorption. However, this difference could be explained by a negligible surface flow term derived from several factors, such as a diminished solubility term (reduced dq/dP) at a higher partial CO 2 pressure, a relatively large pore size distribution (r) enhancing the Knudsen diffusion term, and possibly a less mobile adsorbed phase (smaller D s ) 62 due to the strong interactions between CO 2 and TpHz. Such a negligible surface flow contribution in the single-component gas transport of strongly adsorbing gases (e.g., CO 2 ) has been previously observed in graphene oxide membranes 38 and other microporous 2D COF membranes. 63 On the other hand, the individual gas permeance of He and N 2 in the binary CO 2 /He and CO 2 /N 2 mixtures (Table 2) at room temperature experiences a significant decrease when compared to that in the single gas feed, possibly due to the reduced driving force (i.e., partial pressure) and the presence of other more favorably interacting gas species (i.e., CO 2 ) that modify the TpHz pore-wall surfaces by adsorption. More dramatic effects have been observed in other previous binary gas permeation studies where CO 2 displays confinement effects that largely suppress the diffusion of other gases in the permeating gas mixture. 17,18,21 In our case, the lower partial pressure of CO 2 in the binary mixture could enhance the surface flow contribution (eq 6) due to higher solubility (i.e., larger average slope of gas adsorption isotherm or bigger dq/ dP) as it is proportional to the permeance. Such behavior is reasonable as CO 2 exhibits a nonlinear adsorption isotherm in TpHz bulk powders. 42,47 Moreover, these observations agree well with our gas adsorption results, which will be discussed in detail at the end of this section.
Regarding the temperature effect on the permeance of the CH 4 /N 2 mixture (Figure 5a), the surface flow component (i.e., the second term in eq 6) could be minimal and hence neglected. By assuming negligible viscous flow, the Knudsen flow component (i.e., the first term in eq 6) would play a dominant role in dictating the individual gas permeances of CH 4 and N 2 through the TpHz membrane. As a result, the individual permeance of both CH 4 and N 2 would decrease with increasing T (Figure 5a), with the lighter gas (i.e., CH 4 ) exhibiting a higher permeance at any given T. Meanwhile, the separation performance of the CH 4 /N 2 mixture would be dictated by the Knudsen transport mechanism, resulting in an almost T-independent separation factor that is close to theoretical Knudsen selectivity.
Concerning the CO 2 /N 2 mixture vs T (Figure 5b), the individual CO 2 permeance at lower T could be composed of contributions from both Knudsen transport and surface flow (by forming a surface CO 2 layer). However, as T increases, the surface flow component (i.e., the second term in eq 6) may decrease and become less important because of the reduced dq/dP term at higher T, thus leading to a decrease in the overall individual CO 2 permeance. Simultaneously, the reduced dq/dP could decrease the thickness of the CO 2 surface layer at higher T, which would in turn enlarge the effective transport pore size (Figure 6) to favor the Knudsen flow component (i.e., the first term in eq 6). This could help explain the increase in the individual permeance of N 2 with increasing T since its permeation is dictated by Knudsen transport. As will be discussed in more detail later, N 2 does not adsorb significantly onto the TpHz COF and thus would probably not induce any surface flow. As a result, the separation factor of the CO 2 /N 2 mixture would decrease with increasing T because of the reduced CO 2 surface flow contribution, finally approaching a value closely similar to the Knudsen selectivity at higher T.
To ascertain that the synthesized TpHz membrane shows preferential adsorption for CO 2 over other gases, roomtemperature gas adsorption experiments were performed on the IP-grown support-free TpHz films using three different gases (i.e., N 2 , CH 4 , and CO 2 ), and Figure 7 compiles the obtained room-temperature gas adsorption isotherms. The preferential gas adsorption order is CO 2 ≫ CH 4 > N 2 , e.g., at 100 kPa, the CO 2 uptake is ∼3.6 cm 3 STP g -1 , much higher than those of CH 4 (∼0.3 cm 3 STP g -1 ) and N 2 (∼0.1 cm 3 STP g -1 ). Similarly, the gas adsorption isotherms on the TpHz bulk powder obtained by solvothermal synthesis (Figure S5) show the same CO 2 ≫ CH 4 > N 2 gas uptake order while reporting much higher gas uptake values, probably because of its higher crystallinity and surface area as compared to those of the TpHz film. It is also noteworthy that the absolute gas uptakes of these TpHz bulk powders are comparable to those reported by other research groups under various solvothermal synthesis conditions. 17,42,47
Interestingly, the room-temperature CO 2 adsorption isotherm on the TpHz film showed a nonlinear dependence on pressure (Langmuir-type), displaying a much larger slope (bigger dq/dP, indicative of an increased solubility) 39 due to preferential gas adsorption at lower pressure (Figure 7). In contrast, both CH 4 and N 2 exhibited almost linear uptakes onto the TpHz film, and their uptake values were nearly identical at lower pressure (e.g., < 30 kPa).
As permeability is defined as the product of diffusivity and solubility, the relatively small room-temperature CO 2 permeance difference observed in the binary gas feed (i.e., 2.56 × 10 -8 mol s -1 m -2 Pa -1 in CO 2 /N 2 and 2.30 × 10 -8 mol s -1 m -2 Pa -1 in CO 2 /He; Table 2) vs single gas feed (i.e., 2.38 × 10 -8 mol s -1 m -2 Pa -1 ; Table 1) could be attributed to a larger CO 2 solubility at reduced partial pressure in the binary gas feed. Therefore, the enhanced CO 2 surface flow due to its preferential adsorption on the TpHz's pore walls over weakly interacting N 2 and He gases can help explain the roomtemperature separation factors of CO 2 /N 2 (0.90) and CO 2 /He (0.35) being higher than the Knudsen selectivities of CO 2 /N 2 (0.79) and CO 2 /He (0.30) and the corresponding selectivities derived from single gas permeation tests (Table 1).
Overall, compared to the previously reported 2D COF membranes (Table S1) comprising some mesopore fractions and/or defects, 10,20,23 the TpHz membrane synthesized in this study exhibits almost equal or slightly better single gas selectivity. These observations further prove the relatively low fraction of nanosized defects of the TpHz COF film interfacially grown on top of the nylon support. However, the selectivity of this TpHz membrane for gas mixture separations is lower than those of the state-of-the-art microporous 2D COF membranes and relevant thin-film PA membranes reported elsewhere with molecular sieving characteristics (Tables S1 and S2). It must be pointed out that most of these previously reported microporous COF-based membranes were based either on multilayers of different types of 2D COFs 14,19 or on a single type of 2D COF prepared by less scalable techniques. 21
Well-intergrown, thin (∼290 nm) TpHz 2D COF membranes could be efficiently prepared on a macroporous nylon support by confining the substrate-assisted interfacial polymerization in the organic-phase side via a unidirectional diffusion/secondary growth approach under mild conditions. Qualitative insight into the microstructure of the TpHz selective layer in the supported membrane is inferred from the support-free TpHz film prepared by the L-L IP under nearly identical conditions. GIWAXS characterizations qualitatively suggest that the IPgrown TpHz film consists of 2D COF crystallites of moderate crystallinity with a preferential c-axis orientation normal to the film's surface. The support-free TpHz 2D COF films are stacked layers adopting an eclipsed (AA) geometry, centered on the effective pore size distribution at ∼1.1 nm. The temperature and molecular size dependence of single gas permeation data show that gas transport through the TpHz membranes is governed by the Knudsen diffusion mechanism. In the binary gas permeation of the equimolar CO 2 /N 2 mixture, the presence of the CO 2 surface flow slightly hinders the N 2 flow at room temperature due to the decreased membrane channel size by the adsorbed CO 2 gas layer without exhibiting gate closing/confinement effects. In contrast, permeation of the equimolar CH 4 /N 2 mixture does not exhibit discernible surface flow of both gases due to their much lower gas uptake, and their transport mechanism follows Knudsenlike behavior. We observe that the selectivity of the TpHz membrane prepared by the IP method is modest. Our findings show that the gas permeation through these microporous 2D COF membranes is dominated by the Knudsen diffusion mechanism, where selectivity is determined by the square root of the molecular weight. Thus, such a moderate selectivity is consistent with the permeation mechanism we identified. Though membranes with Knudsen selectivity may find limited applications where high selectivity is not required (e.g., the gas diffusion process for uranium isotope separation is based on the Knudsen separation mechanism), our results suggest the need to identify COF membranes with smaller pores for enhanced selectivity.
Overall, this study provides fundamental insight into how the underlying gas permeation mechanisms (i.e., Knudsen flow and surface flow) dictate the transport of industrially relevant gases through 2D COF membranes. Notably, the polymeric substrate-assisted L-L IP synthesis in this study is easy and scalable, holding promise for the industrial-scale production of gas separation COF membranes.
1 2 Literature survey on gas permeation studies of 2D COF membranes Table S1. Literature survey on previous gas permeation studies of state-of-the-art microporous 2D COF membranes, including details on their fabrication strategies. (L-L: Liquid-Liquid, V-L: Vapor-Liquid) COF type (Intrinsic pore size) COF layer thickness H2 permeance (mol s -1 m -2 Pa -1 ) H2 permeability (Barrer) Porous substrate used Ideal selectivities (𝑰𝑺) and/or Separation factors (𝜶 𝒊/𝒋 ) Fabrication protocol Gas transport mechanism(s) Ref. COF-1 (1.5 nm) 100 nm 1.7×10 -6 507.5 α-Al2O3-SiO2-ZrO2 (𝑰𝑺) H2/N2: 3.3-4.1 H2/He: 1.4 H2/SF6: 5.6 Solvothermal / Exfoliation / Dip coating Knudsen 1 TpAzo (2 nm) 800 nm (He) 3.76x10 -8 (He) 89.86 Al2O3 hollow fiber (𝑰𝑺) N2/He:0.5 O2/He: 0.43 CO2/He: 0.5 In situ L-L IP Knudsen/ viscous flow 2 Tp-TAPA (0.6 nm) 10-20 μm NA NA Selfstanding (𝜶 𝒊/𝒋 ) Xe/Kr: 1.11 Oil-water-oil IP Knudsen/ viscous flow 3 ACOF1 (1.27 nm) 8 μm 1.6×10 -8 382.1 α-Al2O3-APTES (𝜶 𝒊/𝒋 ) CO2/CH4 :86 (𝑰𝑺) H2/CO2 ~ 1.88 H2/N2 ~ 12.31 H2/CH4 ~ 50.9 Solvothermal (in situ growth) Molecular sieving / Adsorption diffusion (CO2/CH4) 4 LZU1-ACOF1 (1.8 nm / 1.27 nm) 1 μm 2.4×10 -7 716.4 α-Al2O3-APTES (𝜶 𝒊/𝒋 ) H2/CO2: 24.2 H2/N2: 83.9 H2/CH4: 100.2 Multilayer, solvothermal (in situ growth) Molecular sieving 5 TpEBr-TpPaSO3 (1.66 nm / 1.4 nm) 41 nm 8.54×10 -7 105 α-Al2O3 (𝜶 𝒊/𝒋 ) H2/CO2: 22.6 (𝑰𝑺) H2/CO2: 26.2 H2/N2: 40.5 H2/CH4: 74.2 V-L IP / Layerby-Layer (Langmuir-Schaeffer) Molecular sieving 6 TpTGCl/ TpPaSO3H (1.3 nm / 1.45 nm) 155 nm 7.2×10 -7 333.3 AAO (anodic alumina oxide) (𝜶 𝒊/𝒋 ) H2/CO2: 26 (𝑰𝑺) H2/CO2: 28.2 H2/N2: 35.9 H2/CH4: 47.9 Multilayer, Self-standing L-L (gutter layer), In situ L-L IP Molecular sieving 7 LZU1vertically aligned (0.37 nm interlayer), -CoAl-LDH 2 μm 1.22×10 -6 7266.4 α-Al2O3-APTES (𝜶 𝒊/𝒋 ) H2/CO2 ~ 31.6 H2/CH4 ~ 29.5 Solvothermal (in situ growth), hybridized Molecular sieving / CO2 adsorption with blocking effect 4 TpHz (1.3 nm) 10 nm 3.33×10 -8 0.99 α-Al2O3 (𝜶 𝒊/𝒋 ) CO2/H2 ~ 10 Solvothermal (precursors) / Exfoliationball milling / Hot drop coating Surface diffusion + Activated diffusion 9 Tp-Pa2 (1.5 nm) 10 nm 2×10 -8 0.60 α-Al2O3 (𝜶 𝒊/𝒋 ) CO2/H2 ~ 22 Solvothermal (powder precursor)/ Exfoliationball milling / Hot drop coating Surface diffusion + Activated diffusion 9 TpPa-1 (1.8 nm) ~ 1 μm 1.17×10 -6 3479.1 α-Al2O3-APTES (𝜶 𝒊/𝒋 ) H2/CH4 ~ 15 H2/CO2 ~ 15 H2/C3H6 ~ 45 H2/C3H8 ~ 70 Substrate assisted L-L IP at high T (120 °C) / special reactor Molecular sieving 10 TAPB-BTCA (1.2 nm) 55 μm 1.22×10 -8 2000.0 Self-Standing (𝜶 𝒊/𝒋 ) CO2/N2 ~ 22 CO2/CH4 ~ 16.8 (𝑰𝑺) H2/CO2 ~ 4.6 H2/N2 ~ 28.6 H2/CH4 ~ 33.3 Aerogel synthesis + Compression method Adsorption diffusion / Molecular sieving 11 DhaTG-Asp (1.5 nm) 65 nm NA NA PAN (𝜶 𝒊/𝒋 ) CO2/CH4 ~ 49 Nanosheets synthesis by L-L IP (oilwater-oil) / Vacuum filtration Facilitated transport 12 TAPB-NTCA (2.9 nm) ~50 nm NA NA PAN (𝜶 𝒊/𝒋 ) CO2/N2 ~ 4.7 Self-standing, CVD Solutiondiffusion 13 3 Literature survey on gas permeation studies of PA membranes Table S2. Literature survey on previous gas permeation studies of thin-film PA membranes prepared by L-L IP fabrication method without gutter layers. Organic phase monomers Aqueous Phase Monomers Selective Layer Thickness (nm) Target gas permeance (mol s -1 m -2 Pa -1 ) Target gas permeability (Barrer) Ideal selectivity (𝑰𝑺) and/or Separation factor (𝜶 𝒊/𝒋 ) Ref. TMC PIP, PPD, MPD 170-503 8.64×10 -9 ~13 (𝑰𝑺) H2/CO2 TMC-PIP: 10.1 TMC-PPD: 10.8 TMC-MPD: 14.3 14 TMC TETA 200 4.46×10 -9 2.66 (𝜶 𝒊/𝒋 ) CO2/CH4: 94.1 15 TMC DNMDAM 100 5.8×10 -9 17.3 (𝜶 𝒊/𝒋 ) CO2/N2: 70 CO2/CH4: 37 16 TMC MPD 200 (avg.) 1.17×10 -7 70 (𝜶 𝒊/𝒋 ) H2/CO2: 50 17 TMC PIP 270 2.59×10 -9 ~2.1 (𝜶 𝒊/𝒋 ) O2/N2: 10.43 18 TMC MPD 500 2.0×10 -9 ~3 (𝑰𝑺) He/CH4: 36.9 (𝑰𝑺) He/CO2: 38.05 19 4 Experimental details More details on the fabrication of TpHz membrane. A thin layer of inert sealing grease (Krytox™ grease GPL106) was applied on the edges of both glass flanges to ensure reliable sealing and prevent solvent leak during TpHz growth (Figure 1a). A pair of disposable plastic syringes (24 mL) were used to simultaneously dispense 10 mL of the required organic-and aqueous-phase solutions in each cell halve, respectively.
Solvents and COF reagents, including dichloromethane (99%), tetrahydrofuran (99%), 1,4dioxane (99%), polyethyleneimine (PEI, Mn ~60 kg/mol, 50 wt% in H2O), hydrazine monohydrate (Hz•H2O, 98%), glacial acetic acid (99%), and p-toluene sulfonic acid monohydrate (PTSA•H2O, 99%), were acquired from Sigma-Aldrich. N-hexane (99%), methanol (99.8%), and acetone (99%)
were obtained from Oakwood Chemicals. Mesitylene (99%) was purchased from Across Organics.
Anhydrous ethanol (95%) was obtained from Fisher Chemicals.
2,4,6-triformylphloroglucinol (Tp, 99.12%) was obtained from ChemScene. All chemicals were used as received. Commercial hydrophilic macroporous nylon 6,6 supports (nominal pore size = 0.1 μm; 47-mm-diameter coupons with a thickness of ~114 μm) were obtained from Sterlitech-USA (NY0147100) and used as received. All compressed gases were research-grade purity (99.999%) and sourced from local suppliers.
To collect adequate TpHz film samples for subsequent characterizations (e.g., gas adsorption studies), several larger batches of TpHz thin film syntheses were repeated by interfacial polymerization without the nylon support in 1 L glass beakers. This strategy is used to overcome the characterization challenge posed by the small sample mass of the TpHz thin film obtained via interfacial polymerization. Notably, the reaction conditions for synthesizing support-free TpHz films, including solvents for each phase, temperature, and molar ratios of monomers and catalyst, were almost identical to those for synthesizing TpHz/nylon composite membranes (Figure S1; see description in the main text). In a typical synthesis, 36 mmol (1.76 mL) of Hz•H2O and 36 mmol (6.84 g) of PTSA•H2O were dissolved in 320 mL of DI water to form the aqueous phase. The organic phase was formed by adding 0.24 mmol (50.4 mg) of Tp in 320 mL of n-hexane. Both aqueous and organic solutions were sonicated for 60 min. The aqueous solution was first transferred into a 1 L beaker and degassed for 5 min. Subsequently, ~40 mL pure n-hexane was added dropwise using a glass pipette to form a buffer layer atop of the aqueous solution. Finally, used to perform four freeze-pump-thaw cycles to ensure an air-free reaction. The reactive mixture was then stirred at 300 rpm and heated at 120 °C in an oil bath for 72 h. After the reaction, the resulting mixture was centrifuged at 5000 rpm for 10 min to isolate and purify the as-synthesized TpHz solids. The purification step involved immersing/shaking (~10 min in total) the precipitated TpHz solids in dichloromethane, methanol, and acetone in a 50 mL centrifuge tube (2 times per The 1D profiles revealed three broad Bragg peaks at q = 0.48, 0.84, and 1.84 Å -1 , corresponding to the (100), (110), and (001) peaks, respectively. The observed q values from GIWAXS are comparable to those of TpHz support-free films and TpHz bulk powders obtained from PXRD (q = 0.49, 0.86, and 1.9 Å -1 ) that correspond to the 2θ Bragg's angle values of 6.9°, 11.9°, and 27°. To quantify the orientational preference, Hermans orientation function f 20,21 was calculated from the azimuthal profile of (001) peak: 𝒇 = (3〈cos 2 𝜃〉 -1)/2, where θ is the 0.0 0.2 0.4 0.6 0.8 1.0 0 50 100 150 200 250 300 350 400 5 10 15 20 25 30 35 0.00 0.01 0.02 0.03 0.04 0.05 0.06 dV/dW (cm³/g•Å) Pore Width (Å) Quantity Adsorbed (cm 3 g -1 STP) Relative Pressure (P/Po) 0 10 20 30 40 50 60 70 80 90 100 0 5 10 15 20 25 30 35 40 45 50 Gas Uptake (cm 3 g -1 STP) Pressure (kPa) CO 2 CH 4 N 2
Industrial & Engineering Chemistry Research
https://doi.org/10.1021/acs.iecr.3c04303 Ind. Eng. Chem. Res. 2024, 63, 3684-3694