Novel β-CD@ZIF-8 Nanoparticles-Doped Poly(m-phenylene isophthalamide) (PMIA) Thin-Film Nanocomposite (TFN) Membrane for Organic Solvent Nanofiltration (OSN).

Organic solvent nanofiltration (OSN) membranes are always troubled by the "trade-off" effect between solvent flux and solute rejection. Hence, a rapid, convenient, and effective way to synthesize novel β-cyclodextrin-enhanced zeolite imidazole framework-8 (β-CD@ZIF-8) nanoparticles was first proposed and the nanoparticles were doped into both selective layer and poly(m-phenylene isophthalamide) support for fabricating thin-film nanocomposite membranes. Transmission/scanning electron microscopy images and X-ray photoelectron spectroscopy results demonstrate the successful synthesis of β-CD@ZIF-8. Atomic force microscopy images illustrate the more rougher surface compared to the pristine membrane, while the pure acetone flux reached 62.3 ± 2.3 L m-2 h-1, and Rose Bengal rejection achieved 96.6 ± 1.8 and 94.5 ± 0.5% in methanol (MeOH) and tetrahydrofuran at 0.6 MPa, respectively, when the dosage was 0.05% (w/v). The molecular weight cutoff around 574 Da of PPA2505 containing β-CD@ZIF-8 in both support and selective layers shows the optimum properties and outstanding OSN performances in erythromycin concentration and purification in MeOH and butyl acetate. Additionally, polyimide nanofiber and the formed net structure may offer a potential way to fabricate "ultrathin" film in the OSN industry.

Introduction

Ultrafast molecular separation (n class="Disease">UMS) is an emerging notion in highly selective meclass="Chemical">pan class="Chemical">mbranes with pore size in the range of 0.5–5 nm, which consumes low energy while maintaining efficiency. One of the most important factors to develop UMS membranes is combining the state-of-the-art nanomaterials with advanced membrane material during preparation.[1]

There is one kind of green technology known as organic solvent nanofiltration (n class="Chemical">OSN) (Table S1 abbreviations) which finds applications in pharmaceuticclass="Chemical">pan class="Chemical">al ingredient recovery, dye rejection, catalyst recycling, food processing, and oil production.[2−7] Over the last decade, the OSN membranes had been widely prepared by adopting polyimide (PI) (P84 or Matrimid 5218) by nonsolvent induced phase separation (NIPS) method, followed by IP reaction to prepare thin-film composite (TFC)/N membranes.[8,9] During the preparation process, they were always cross-linked by different types of diamine. Shao et al. found that the diamine cross-linking reactivity was affected by the molecular lengths and nucleophilicity and the shorter molecular length may have the better effect. Moreover, different solvents would interact membranes differently, although the cross-linked OSN membrane has better stability, and the solvent permeance would be reduced due to phenomena like compact effect and cross-linker–membrane interaction.[10] Changelessly, the development of OSN membranes is aimed at reducing the “trade-off” effect between solvent flux and solute rejection, while maintain high resistance to organic solvents.[11]

A novel process to fabricate n class="Chemical">OSN meclass="Chemical">pan class="Chemical">mbrane, which breaks the traditional method, was presented using advanced PI electrospinning nanofiber as solvent-resistant support. The ultrafine nanofiber (∼33 μm) contains homogeneous pores with a nominal size of 1.273 μm and high porosity, which reduce the mass-transfer resistance during operation.[12,13]

n class="Chemical">Poly(m-phenylene isophthalamide) (class="Chemical">pan class="Chemical">PMIA) is a kind of aromatic polyamide (PA) material, which has several merits for membrane fabrication, such as its hydrophilic nature, high porosity, splendid mechanical properties, and operability (Figure S1).[14] Zhu et al. claimed that membranes prepared via PMIA could break through the trade-off effect between permeance and rejection, while Bruggen et al. developed an mixed matrix membrane (MMM) with it for the first time through combination with MIL-53(Al) and demonstrated its great potential in OSN performances.[15,16]

Compared to n class="Chemical">MIL-53(class="Chemical">pan class="Chemical">Al) and other metal-organic frameworks (MOFs), zeolite imidazole framework-8 (ZIF-8) is a group of prevailing porous materials with exceptional thermal and chemical stabilities, which play an important role in gas separation, catalyst, and building ultrathin ZIF-8 membranes.[17−21] With an sodalite (SOD) zeolite-type structure consisting of small apertures (3.4 Å), large pores (11.6 Å), and cubic space group (16.32 Å), ZIF-8 were always welcomed in OSN applications.[22] Besides, β-cyclodextrin (β-CD) found its capacity in TFC OSN membrane preparation and organic pollutants removal due to both its hollow truncated cone structure and well-arranged hydrophilicity or hydrophobicity distribution.[23−25]

Herein, novel β-CD-enhanced n class="Chemical">ZIF-8 (β-CD@class="Chemical">pan class="Chemical">ZIF-8) nanoparticles were first synthesized through a simple method and incorporated into the PA selective layer. The two PA layers were a PMIA support layer and a PA selective layer. The PMIA support layer was formed onto the PI nanofiber by spin coating of PMIA (without diamine cross-linking) with a spinner (Laurell model WS-650MZ-23NPPB) and IP between m-phenylenediamine (MPD) and trimesoyl chloride (TMC) with the improvement of β-CD@ZIF-8. The structure of the as-prepared thin-film nanocomposite (TFN) membrane is schematically shown in Table of Contents.

Results and Discussion

Chemical Compositions

Fourier transform infrared (FT-IR) spectra of β-CD, n class="Chemical">ZIF-8, and β-CD@class="Chemical">pan class="Chemical">ZIF-8 are displayed in Figure a. According to the references elsewhere, characteristic peaks at 689–756, 1171, 1588, and 2920 cm–1 attributed to the ZIF-8 particle could be clearly determined. Meanwhile, vibrations of primary, secondary, and tertiary alcohols in the β-CD structure could also be observed.[26,27] After the synthesis of β-CD@ZIF-8, the structure and the characteristic peaks were still retained, while those vibrations of alcohols were covered by the strong peaks of ZIF-8. These results indicated the successful synthesis of β-CD@ZIF-8 nanoparticles with both the merits of β-CD and ZIF-8 particles.

Figure 2

(a) FT-IR spectra of β-CD, ZIF-8, and β-CD@ZIF-8; (b) FT-IR spectra of as-developed TFN OSN membranes with their hydroxyl vibration strength and high-resolution image of C–O–C peaks; (c) X-ray diffraction (XRD) patterns of β-CD, ZIF-8, and β-CD@ZIF-8; (d) X-ray photoelectron spectroscopy (XPS) wide-scan images of PPA, PPA05, and PPA10 (inset: atomic percentages of the samples and XPS wide-scan images of ZIF-8 and β-CD@ZIF-8).

XRD patterns of β-CD, n class="Chemical">ZIF-8, and β-CD@class="Chemical">pan class="Chemical">ZIF-8 are displayed in Figure c. Likewise, the same conclusion could be acquired as the FT-IR analysis, in which the obtained patterns illuminate the characteristic peaks (in rectangle) of ZIF-8 and have never changed after being enhanced by β-CD. Some peaks attributed to β-CD overlapped with those of ZIF-8 and showed the strengthened peaks in the β-CD@ZIF-8 pattern.

With respect to the n class="Chemical">TFN OSN meclass="Chemical">pan class="Chemical">mbranes with or without β-CD@ZIF-8 nanoparticles, all of them exhibited a typical spectrum style of the MPD-TMC PA layer, while the influences of the nanoparticles could be defined conspicuously in terms of their hydroxyl vibration strength. In Figure b, we could clearly observe “C–O–C” at 1339 and 1418 cm–1, which are attributed to the ester group formed between the hydroxyl group of β-CD@ZIF-8 and TMC (inset). According to the reference, the peaks have some shifting mainly due to the low content and testing conditions such as testing temperature, humidity, and instrumental differences.[28] Furthermore, we judged the hydroxyl vibration strength of each membrane by a simple “dash dot line” tool. The PI nanofiber shows the highest hydrophobicity among all related membranes, while the membrane with PMIA support shows hydroxyl vibration strength of four dots indicating the excellent hydrophilicity of the PMIA nature for IP reaction. After the PA selective layer was established, the surface hydrophilicity decreased and the hydroxyl vibration strength reduced to only two dots. With the introduction of the β-CD@ZIF-8 nanoparticles, PPA05 owns four dots, which equaled to PMIA. As β-CD@ZIF-8 dosage increased in the PA selective layer, more hydroxyl groups in the β-CD@ZIF-8 structure dispersed onto/into the PA layer and reacted (Figure b), leading to the improvement of surface hydrophilicity, and the number of dots finally reached up to 7.

Figure 1

(a) Schematic diagram of β-CD@ZIF-8 preparation process. (b) Ideal reaction mechanism during PA selective layer formation stage.

(a) Schematic n class="Chemical">diagram of β-CD@class="Chemical">pan class="Chemical">ZIF-8 preparation process. (b) Ideal reaction mechanism during PA selective layer formation stage.

XPS ann class="Chemical">alysis was carried out to further determine the chemicclass="Chemical">pan class="Chemical">al composition of the as-developed β-CD@ZIF-8 nanoparticles and the TFN membranes. The wide-scan spectra and atomic percentages of the samples are shown in Figure d. Pure ZIF-8 shows rarely oxygen element content (0.01%) compared to β-CD@ZIF-8 (3.91%), which was successfully wrapped by β-CD. Meanwhile, TFC membrane PPA prepared without doping β-CD@ZIF-8 exhibited no zinc element, while TFN membrane PPA10 (4.87%) contains more zinc than PPA05 (4.33%), in accordance with the energy-dispersive X-ray (EDX) results (Table ). Detailed information about the TFC/N membranes’ high-resolution C 1s spectra is shown in Figure S2.

Table 2

Atomic Force Microscopy (AFM) Roughness

membrane code	Ra (nm)	Rms (nm)	Rz (nm)
PA	0.18 ± 0.01	0.23 ± 0.01	1.12 ± 0.11
PPA	0.53 ± 0.06	0.70 ± 0.17	3.83 ± 0.59
PPA05	10.60 ± 1.68	13.53 ± 2.74	57.60 ± 12.50
PPA10	16.23 ± 1.05	22.00 ± 0.92	127.20 ± 5.46

(a) FT-IR spectra of β-CD, n class="Chemical">ZIF-8, and β-CD@class="Chemical">pan class="Chemical">ZIF-8; (b) FT-IR spectra of as-developed TFN OSN membranes with their hydroxyl vibration strength and high-resolution image of C–O–C peaks; (c) X-ray diffraction (XRD) patterns of β-CD, ZIF-8, and β-CD@ZIF-8; (d) X-ray photoelectron spectroscopy (XPS) wide-scan images of PPA, PPA05, and PPA10 (inset: atomic percentages of the samples and XPS wide-scan images of ZIF-8 and β-CD@ZIF-8).

Morphology and Topography

From the inset of Figure , it could be observed that the morphology of the n class="Chemical">ZIF-8 class="Chemical">particle transformed from SOD structure to a bclass="Chemical">pan class="Chemical">all shape after wrapped by β-CD accompanied with the enlargement of the mean particle size from 65.5 ± 15.4 to 87.1 ± 10.7 nm. Particle size distribution curves also show differences between pure ZIF-8 and β-CD@ZIF-8, where the latter distributes in a larger region, while the former distributes in a smaller region. The changes of the particle sizes is the result of wrapping of β-CD onto the ZIF-8 nanoparticles during the preparing process. We could also observe these phenomena from the transmission electron microscopy (TEM) images evidently. Figure a,e displays a typically SOD structure and a ball shape of ZIF-8 and β-CD@ZIF-8 clearer than Figure a,b. Their sizes obtained from the images were around 51 and 82 nm, in the range of their particle size distribution curves. Comparing Figure b,c and 4f,b, enlarged particles and those “wrapping β-CD” organizations could be determined distinctly on the structure of β-CD@ZIF-8 in contrast to those of pure ZIF-8. Besides, β-CD@ZIF-8 displays more a homogeneous dispersion compared to pure ZIF-8 (Figure d,h). In addition, ζ ζ potential values of pure ZIF-8 and β-CD@ZIF-8 in deionized (DI) water 0.1% (w/v) were determined to be 19.9 ± 0.2 and 27.2 ± 0.2 mV. The differences of ζ potential values must be another evidence that β-CD@ZIF-8 nanoparticles were synthesized successfully. According to literature reports, the ball-shaped particles always have less surface energy and better dispersibility, and the properties of ZIF-8 has never changed, but may give this kind of nanomaterial the potential in the OSN membranes (Figure a,c).[29]

Figure 3

(a) Particle size distribution and cumulative distribution curves of ZIF-8 nanoparticles (inset: scanning electron microscopy (SEM) images (50 000× and 150 000×)). (b) Particle size distribution and cumulative distribution curves of β-CD@ZIF-8 nanoparticles (inset: SEM images (50 000× and 150 000×)).

Figure 4

TEM images of ZIF-8 nanoparticles: (a) 10 nm; (b) 20 nm; (c) 50 nm; and (d) 200 nm. TEM images of β-CD@ZIF-8 nanoparticles: (e) 10 nm; (f) 20 nm; (g) 50 nm; and (h) 200 nm.

(a) Particle size n class="Chemical">distribution and cumulative class="Chemical">pan class="Chemical">distribution curves of ZIF-8 nanoparticles (inset: scanning electron microscopy (SEM) images (50 000× and 150 000×)). (b) Particle size distribution and cumulative distribution curves of β-CD@ZIF-8 nanoparticles (inset: SEM images (50 000× and 150 000×)).

TEM images of pan class="Chemical">ZIF-8 nanoclass="Chemical">particles: (a) 10 nm; (b) 20 nm; (c) 50 nm; and (d) 200 nm. TEM images of β-CD@class="Chemical">pan class="Chemical">ZIF-8 nanoparticles: (e) 10 nm; (f) 20 nm; (g) 50 nm; and (h) 200 nm.

Researchers in the men class="Chemical">mbrane field know very well the importance of the class="Chemical">pan class="Chemical">dispersibility of a new material during the preparation process and that more hydrophilous material has better dispersibility. Figure a lists the powder water contact angle results of ZIF-8 and β-CD@ZIF-8, while the β-CD and β-CD, ZIF-8 mixture (molar ratio 1:1) was also prepared as control. Pure ZIF-8 powder possesses the highest water contact angle of 74.5° and hardly changed in 60 s test, while the compacted pure β-CD powder could not support the water droplet in a very short time due to its soluble nature. The β-CD and ZIF-8 mixture shows improved hydrophilicity than pure ZIF-8, but still could not surpass that of β-CD@ZIF-8, where the water droplet was vanished in 20 s. In addition, the variation trends could also prove the successful synthesis of β-CD@ZIF-8 with the exception of the instrumental analysis discussed above. Figure b illustrates the dispersion conditions of 0.1% (w/v) ZIF-8 and β-CD@ZIF-8 nanoparticles in DI, EtOH, and n-hexane. β-CD@ZIF-8 in DI water looks more transparent than ZIF-8 due to the presence of soluble β-CD wrapped around the particle, which endowed better hydrophilicity and even give the wrapped particles solubility. In EtOH, two particles display almost the same dispersibility, while two distinct results appeared in n-hexane. The better dispersion of β-CD@ZIF-8 in n-hexane guides us to introduce the particles into the organic phase to fabricate novel TFC OSN membranes.

Figure 5

(a) Powder water contact angle results of ZIF-8, β-CD, β-CD@ZIF-8, and ZIF-8, β-CD mixture; (b) comparison of dispersibilities of ZIF-8 (white cap) and β-CD@ZIF-8 (black cap) in DI water, EtOH, and n-hexane.

(a) Powder n class="Chemical">water contact angle results of class="Chemical">pan class="Chemical">ZIF-8, β-CD, β-CD@ZIF-8, and ZIF-8, β-CD mixture; (b) comparison of dispersibilities of ZIF-8 (white cap) and β-CD@ZIF-8 (black cap) in DI water, EtOH, and n-hexane.

The n class="Chemical">PMIA support class="Chemical">pan class="Chemical">always appeared to be a spongelike structure mixed with macropores, which endow the OSN membrane with excellent mechanical strength and high permeability (Figure a). With better surface hydrophilicity (Figure b), the PA selective layer could be easily formed onto the PMIA support (Figure b,c). After we incorporated specific contents of β-CD@ZIF-8 nanoparticles into the organic phase, the hydrophilicity of the as-formed PA layer changed as we expected. As shown in Figure b, the water contact angle shows the order of PPA10 < PPA05 < PPA. The decreased contact angle values were affected by these aspects. Adding fillers in the monomer solutions may vary the cross-linking degree of the formed PA selective layer and then vary the surface hydrophilicity of the membrane surface.[30] The presence of the hydroxyl wrapped outside the modified ZIF-8 nanoparticles and with dosage increase, the nanoparticles on the PA layer increases, further lowering the water contact angle. As we have dispersed the hydrophilic β-CD@ZIF-8 nanoparticles into hydrophobic TMC/n-hexane solution, which strongly affects the interfacial polymerization due to the reaction between hydroxyl and acyl chloride, the C–O–C and MPD-TMC-β-CD@ZIF-8 net structure was formed. Besides, hydrophilic nanoparticles in TMC/n-hexane solution would deposit them on the upper part of PA layer, which made them to exert their hydrophilicity nature more easily. These were also clearly reflected by the EDX and XPS results shown in Figure d and Table : the percentage of zinc atom increased remarkably as the increased dosage of β-CD@ZIF-8. Moreover, from the FT-IR spectra shown in the vibration between wavelengths 3100 and 3600 cm–1, PPA10 also becomes stronger than PPA05 (Figure b), in accordance with the results. Evidently, reacted β-CD@ZIF-8 nanoparticles could be easily found on the membrane surface (Figure f–h), which is confirmed by the C–O–C vibration and resulted in the increment of surface roughness (Figure b and Table ). Comparing Figure e and 6d, intensive peaks and a gully could be observed after doping the particles, and these structures would directly affect the solvent permeance. First, more effective membrane area could be used to receive the incoming feed solvent, and the solvent molecule could go through either β-CD@ZIF-8 cavities or the dense PA layer. Then, with better hydrophilicity of both the PA layer and the PMIA support, most of the solvent molecules would pass through the OSN membrane matrix rapidly and thus improve the permeance and separation efficiency of the membrane. Figure i reveals the overall cross-sectional perspective of PPA, which could be a typical example of the as-prepared OSN membranes. This series of OSN membrane consists of three layers, including PI nanofiber, PMIA support membrane, and PA selective layer. The electrospun PI shows a homogeneous thickness of 33 ± 1 μm, which supply the solvent-resistant backbone for the membrane and has the potential in preparing ultrathin OSN membranes. The bulk part is the PMIA support membrane formed by spin-coating technique and NIPS method, which shows a thickness of 85 ± 5 μm, and the thinnest layer is the PA selective layer, which was prepared via IP method. The PMIA support membrane and the PA selective layer with or without β-CD@ZIF-8 were developed to investigate their OSN performances. Those TFN OSN membranes containing 0.05% (w/v) β-CD@ZIF-8 show a thicker selective layer (102.1 ± 4.0 nm) than TFC membranes (97.1 ± 6.4 nm) due to the introduction of nanoparticles and the increased surface roughness. Moreover, the reacted β-CD@ZIF-8 nanoparticles, which formed the MPD-TMC-β-CD@ZIF-8 net structure, also integrated in the selective layer and give the TFN membrane additional OSN performances.

Figure 6

SEM images and AFM three-dimensional (3D) images of TFN OSN membranes: (a) cross-sectional structure of PPA (captured without PI nanofiber); (b) surface morphology of PPA (5000×); (c) cross-sectional structure of PPA (PA selective layer); (d) AFM 3D topography of PPA; (e) AFM 3D topography of PPA10; (f) surface morphology of PPA10 (1000×); (g) surface morphology of PPA10 (5000×); (h) surface morphology of PPA10 (20 000×); and (i) overall cross-sectional perspective of the as-prepared OSN membranes (take PPA as an example).

Figure 10

(a) Pure MeOH flux comparison of all kinds of membrane related in this work. (b) Dynamic water contact angle (DCA) curves.

Table 1

EDX Results of the TFN OSN Membrane Surface

membrane code	C (%)	N (%)	O (%)	Zn (%)
PPA	66.59	19.37	14.04
PPA05	70.49	19.02	10.38	0.11
PPA10	57.71	22.76	18.80	0.73

SEM images and AFM three-n class="Chemical">dimensionclass="Chemical">pan class="Chemical">al (3D) images of TFN OSN membranes: (a) cross-sectional structure of PPA (captured without PI nanofiber); (b) surface morphology of PPA (5000×); (c) cross-sectional structure of PPA (PA selective layer); (d) AFM 3D topography of PPA; (e) AFM 3D topography of PPA10; (f) surface morphology of PPA10 (1000×); (g) surface morphology of PPA10 (5000×); (h) surface morphology of PPA10 (20 000×); and (i) overall cross-sectional perspective of the as-prepared OSN membranes (take PPA as an example).

OSN Performances

The PI nanofiber has the pure n class="Chemical">MeOH flux just equclass="Chemical">pan class="Chemical">al to the feed speed at a relatively low cross-membrane pressure due to its homogeneous and a rather larger nominal pore size (1.273 μm). The PMIA support membrane was thus developed to control the solvent flux based on the concentration of PMIA in spinning solution. PMIA with concentration from 17.0 to 20.0 wt % were spin-coated and tested in pure MeOH and tetrahydrofuran (THF) under both 0.2 and 0.6 MPa. As shown in Figure a,b, the PMIA support membrane prepared by 18.0 wt % spinning solution is the most suitable support to further fabricate β-CD@ZIF-8-doped TFN OSN membrane because the quality of the IP reaction strongly depends on the properties of the support. The fluxes of MeOH and THF show the same variation trend with increase of the spinning solution concentration. However, the TFC membrane prepared by 17.0 wt % would have much higher solvent flux but lower solute rejection, while that prepared by 19.0 and 20 wt % would have extremely high solute rejection but almost no solvent flux. Therefore, we doped β-CD@ZIF-8 nanoparticles into the 18.0 wt % PMIA spinning solution and prepared PA25 support in which improved permeance performances were obtained and further developed PPA2505 TFN OSN membrane.

Figure 7

Pure solvent flux (PSF) of the PI nanofiber spin-coated with PMIA support membrane in different contents: (a) pure MeOH flux; (b) pure THF flux; control tests of TFN OSN membranes doped with pure ZIF-8, pure β-CD, β-CD, ZIF-8 mixture (molar ratio 1:1) and β-CD@ZIF-8 in the PA selective layer: (c) Rose Bengal (RB)/MeOH; (d) RB/THF.

Pure solvent flux (PSF) of the PI nanofiber spin-coated with n class="Chemical">PMIA support meclass="Chemical">pan class="Chemical">mbrane in different contents: (a) pure MeOH flux; (b) pure THF flux; control tests of TFN OSN membranes doped with pure ZIF-8, pure β-CD, β-CD, ZIF-8 mixture (molar ratio 1:1) and β-CD@ZIF-8 in the PA selective layer: (c) Rose Bengal (RB)/MeOH; (d) RB/THF.

Control tests of n class="Chemical">PPA class="Chemical">pan class="Chemical">OSN membranes with 0.05% (w/v) pure ZIF-8, pure β-CD, β-CD, ZIF-8 mixture (molar ratio 1:1), and β-CD@ZIF-8 in the PA selective layer were also carried out for illustrating the improvement effect of β-CD@ZIF-8 on the OSN performances. Comparing the results in Figure c,d, control tests show almost the same trend in RB/MeOH and RB/THF. TMC/n-hexane contains pure β-CD, which has the highest flux but the lowest RB rejection, with the reaction between hydroxyl and acyl chloride reducing the concentration of TMC and thus forming the less dense PA selective layer. The fluxes of PPA-containing mixture are acceptable, but their application in OSN membrane is still hindered by the low RB rejection induced by the large amounts of β-CD. For those membrane containing pure ZIF-8, a high RB rejection was achieved, but they still left behind the PPA membrane containing β-CD@ZIF-8, in which the formed MPD-TMC-β-CD@ZIF-8 net structure offers better hydrophilicity and extra dense layer for preventing those RB solutes from penetrating the membrane surface and thus reducing the trade-off effect in a novel way.

The solute selectivity was mainly dominated by the n class="Chemical">PA selective layer, and their performances were demonstrated by the dye/class="Chemical">pan class="Chemical">MeOH and dye/THF solvent system. Detailed results are plotted in Figure a,b and their insets. The as-prepared TFN OSN membranes exhibited excellent solvent-resistant properties for they could operate in a wide range of polar (a)protic solvents, including harsh solvents such as acetone and dimethylformamide (DMF). Same moderating trend could be observed from all prepared TFN OSN membranes. Besides the Hanson solubility effect, those mechanisms manipulating the pure polar protic solvent flux order of MeOH > EtOH > n-PrOH > isopropyl alcohol (IPA) are the outcomes of the influence of the molecular weight and the steric hindrance effect. Under the same transmembrane pressure, low-molecule-weight solvents have less mass transfer resistance to pass through membrane matrix (mainly the PA selective layer), while longer carbon chains may offer supererogatory structural resistance. IPA passes slower than n-PrOH, due to the secondary hydroxyl groups in its structure, which induces less hydrophilicity than n-PrOH (primary hydroxyl group). The PPA without incorporation of β-CD@ZIF-8 nanoparticles reveals the lowest pure flux of both protic and aprotic solvents. After introduction of the nanoparticles in the PA selective layer, pure fluxes of all solvents improved and higher loadings seem to show better performances. The characterization of β-CD (hydrophobic inner surface)-wrapped ZIF-8 nanoparticles was firm and offers better dispersibility (hydrophilous external surface) when preparing the reaction solution before IP treatment. β-CD@ZIF-8 in the PA selective layer faces the solvents when operating, the surface hydrophilicity improved due to the existence of β-CD, and additional ducts were established due to the function of ZIF-8 nanoparticles. These events could strengthen the solvent permeance of the OSN membranes. Besides, ZIF-8 itself has reported outstanding performances in OSN membrane due to its properties.[31] Dye/MeOH and dye/THF performances could also find influences of the β-CD@ZIF-8 nanoparticles. With the addition of the particles, except for RB/THF, methylene blue (MB), crystal violet (CV), and RB in MeOH and CV/THF, all showed increased dye solute rejection with the increment of the solvent, and the schematic of the ideal mechanism is shown in Figure b. Abnormal phenomena occurred in the dye/solvent performances, which were higher than those of the corresponding pure solvent, especially in CV/MeOH and MB/MeOH with β-CD@ZIF-8 nanoparticles in the PA selective layer. These could be attributed to the following reasons. First, solvent filtration systems are very different from aqueous filtration systems. In the latter, the flux of pure water is always larger than the solute and drives the concept of flux recovery (rate) performance because the solute is usually attached onto the membrane surface due to concentration polarization and further results in the growth of bacteria and microorganisms, which would reduce the separation efficiency. However, solvent filtration systems would seldom face the problem of microorganism contamination due to their toxicity and harsh environment. Then, those smaller dye molecules (CV, MB) cause more obvious phenomena. When operating at high pressure, smaller molecules can be easily pushed into the PA selective layer and leave the imprintings (most of them could not really penetrate the dense PA selective layer), which finally improve the dye/solution permeances. Foremost among these is the swelling effect of the solvent on the membrane, which may manipulate the OSN performance. Those membranes were tested with pure solvent flux first, followed by measuring the dye/solvent performance. Thus, more operating time was required for the sample membrane, and the small molecular part of the PMIA support was dissolved, just as the effect of “DMF activation”. Moreover, β-CD@ZIF-8 in the PA selective layer may enhance the permeance of the TFN membrane, which let more solvent pass through the membrane matrix and thus the effect of “solvent activation” enlarged and eventually give those TFN OSN membranes higher flux than the pure solvent. Furthermore, due to the presence of β-CD@ZIF-8 and the formed MPD-TMC-β-CD@ZIF-8 net structure by the reaction between hydroxyl and acyl chloride, this net structure was the strategic point for reducing the serious trade-off effect. Therefore, when the dye/solvent flux increases, the solute rejection does not reduce any more.

Figure 8

(a) Pure polar protic solvent flux and dye/MeOH permeance vs rejection performances (insets). (b) Pure polar aprotic solvent flux and dye/THF permeance vs rejection performances (insets). (Specifications of all solvents and solutes used in this work are displayed in Tables S2 and S3.)

(a) Pure polar protic solvent flux and dye/n class="Chemical">MeOH permeance vs rejection performances (insets). (b) Pure polar aprotic solvent flux and dye/class="Chemical">pan class="Chemical">THF permeance vs rejection performances (insets). (Specifications of all solvents and solutes used in this work are displayed in Tables S2 and S3.)

Figure a shows the n class="Chemical">OSN performances (class="Chemical">pan class="Chemical">RB/solvent separation) of PPA2505 membrane with β-CD@ZIF-8 nanoparticles in both layers. Similar RB rejection values were achieved (96.2 ± 1.6% of RB/MeOH, 95.0 ± 1.1% of RB/THF) compared to PPA05 (96.6 ± 1.8% of RB/MeOH, 94.5 ± 0.5% of RB/THF), while the polar aprotic solvent flux was approximately quadrupled (30.7 ± 1.1 L m–2 h–1 compared to 8.2 ± 0.3 L m–2 h–1). The enhancement could be confirmed by the influence of additional β-CD@ZIF-8 nanoparticles in the PMIA support compared to PPA05 or PPA10. From one aspect, the introduction of nanoparticles not only eliminated the macroporous structure but also strengthened the permeability of PMIA support membrane, which could be clearly observed in Figure a. The properties of the support may significantly affect the performances of TFC/N membrane.[32] In other aspect, as shown in Figure b, PA with additional 0.25 wt % β-CD@ZIF-8 (PA25) shows lower water contact angle values than PA and it is the lowest among all mentioned membranes in this work. All of these resulted in reducing the mass transfer resistances when operating with solvents. Moreover, when IP reaction took place, the β-CD@ZIF-8 in the PMIA support would also participate in the reaction and the formation of MPD-TMC-β-CD@ZIF-8 net structure, which offers extra OSN performance for separation of solute from the solvent. Furthermore, compared to the surface morphologies of PPA, PPA2505, and PPA05 (Figure b–d), no particle could be found on the surface of PPA, while particle agglomeration could be observed on PPA05. However, Figure e exhibits the uniformly dispersed particles on the surface of PPA2505, the better dispersibility of which in β-CD@ZIF-8 aids itself in the first PA layer dominated by these enhanced OSN performances.

Figure 9

(a) RB/solvent flux and rejection results of PPA2505 (contains β-CD@ZIF-8 nanoparticles in the PMIA support membrane); (b) surface morphology of PPA (20 000×); (c) surface morphology of PPA2505 (20 000×); (d) surface morphology of PPA05 (20 000×); (e) surface morphology of PPA2505 (50 000×); and (f) cross-sectional structure of PPA2505 (20 000×).

(a) n class="Chemical">RB/solvent flux and rejection results of class="Chemical">pan class="Chemical">PPA2505 (contains β-CD@ZIF-8 nanoparticles in the PMIA support membrane); (b) surface morphology of PPA (20 000×); (c) surface morphology of PPA2505 (20 000×); (d) surface morphology of PPA05 (20 000×); (e) surface morphology of PPA2505 (50 000×); and (f) cross-sectional structure of PPA2505 (20 000×).

(a) Pure n class="Chemical">MeOH flux comclass="Chemical">parison of class="Chemical">pan class="Chemical">all kinds of membrane related in this work. (b) Dynamic water contact angle (DCA) curves.

Upon n class="Chemical">all of these class="Chemical">pan class="Chemical">OSN performances discussed above, Figure a compared all membrane prepared in this work with pure MeOH flux, in which the differences could be figured out intuitively. The PI support has the highest MeOH flux in MF scale, PA was a little lower than PA25 with β-CD@ZIF-8 nanoparticles, while the order PPA < PPA05 < PPA10 was mainly dominated by its dosage. The performance of PPA2505 was induced by not only the nanoparticles but also the integrated MPD-TMC-β-CD@ZIF-8 net structure.

n class="Disease">Swelling behavior measurement results of the as-preclass="Chemical">pared β-CD@class="Chemical">pan class="Chemical">ZIF-8-doped TFN OSN membrane are listed in Figure . As the dosage of β-CD@ZIF-8 in the PA selective layer increases, the area swelling ratios decrease remarkably. These were benefited from the formed MPD-TMC-β-CD@ZIF-8 net structure discussed above. The integrated net structure strengthened the PA selective layer and PMIA support membrane from swelling like a reinforcement plate. PPA2505 with nanoparticles in the support formed a denser MPD-TMC-β-CD@ZIF-8 net than other TFN membranes, showing best antiswelling performance. Meanwhile, solvent uptake ratios display approximately the same variation tendency with area swelling ratios. However, as PMIA support may become weak in the DMF steam and small molecular part would dissolve gradually, the solvent uptake ratios were all very small due to the weight loss of PMIA except for PI nanofiber with excellent solvent-resistant properties. Nevertheless, higher dosage of nanoparticles could still offer more anti-DMF performance to the TFN membranes. In fact, PMIA support could sustain in almost all of the solvents and perform well, but become weak in the DMF steam. Luckily, this “drawback” of PMIA offers the as-prepared TFN membrane a potential way to become an “ultrathin” solvent-resistant film with only PI nanofiber and MPD-TMC-β-CD@ZIF-8 net integrated PA selective layer.

Figure 11

Results of swelling behavior measurements; inner: comparison of solvent uptake ratio of the as-prepared OSN membrane in different solvents; outer: comparison of area swelling ratio of the as-prepared OSN membrane in different solvents.

Results of n class="Disease">swelling behavior measurements; inner: comclass="Chemical">parison of solvent uptake ratio of the as-preclass="Chemical">pared class="Chemical">pan class="Chemical">OSN membrane in different solvents; outer: comparison of area swelling ratio of the as-prepared OSN membrane in different solvents.

The mechanicn class="Chemical">al properties of the β-CD@class="Chemical">pan class="Chemical">ZIF-8-doped TFN OSN membranes and their supports are shown in Table . As we excepted, after PMIA was spin-coated onto the PI nanofiber, their mechanical properties were improved obviously compared to the free-standing PMIA support membrane. The introduction of β-CD@ZIF-8 nanoparticles in the PMIA support improved the mechanical properties of pure PMIA due to their influences during the solvent–water exchange process (phase separation process) and enhanced the structure of the support.[33,34] Moreover, the excellent mechanical properties of PI nanofiber accompanied with nanoparticles would offer a potential way for fabricating PI–PA ultrathin solvent-resistant film for high-performance OSN application.

Table 3

Mechanical Properties of β-CD@ZIF-8-Doped TFN OSN Membranes and Their Supports

membrane no.	break strength (MPa)	elongation at break (%)	Young’s modulus (MPa)
PI	28.7 ± 1.3	42.1 ± 1.9	378.6 ± 7.4
PA	20.8 ± 1.6	58.2 ± 4.2	224.8 ± 12.3
PA (without PI nanofiber)	12.6 ± 1.4	47.6 ± 2.8	227.2 ± 10.7
PA25	21.6 ± 0.8	54.2 ± 3.7	204.2 ± 15.5
PA25 (without PI nanofiber)	13.0 ± 0.9	50.1 ± 3.5	213.6 ± 12.4
PPA	19.6 ± 2.1	55.0 ± 4.2	211.1 ± 9.9
PPA05	21.3 ± 1.1	53.3 ± 3.2	250.1 ± 8.8
PPA10	19.1 ± 1.7	47.4 ± 2.7	235.4 ± 7.6
PPA2505	22.1 ± 2.0	60.2 ± 3.4	253.5 ± 8.1

With the introduction of β-CD@n class="Chemical">ZIF-8 and the formation of the class="Chemical">pan class="Chemical">MPD-TMC-β-CD@ZIF-8 net structure, the molecular weight cutoff (MWCO) decreased sharply according to the poly(propylene glycol) (PPG) oligomers method (Figure a). The bulk structure usually exhibits higher rejection than the linear structure, while the chainlike structure of tripropylene glycol (192 Da) has lower rejection values than those of PPG oligomers, which were bulk structures, but showed a very high rejection due to the solute adsorption phenomenon.[35] Thus, the MWCO values below 192 and 250 Da are shown by dashed lines instead of abnormal value. For PPGs, the MWCO value seems larger than 800 Da, which were induced by the swelling effect of PPGs on PMIA, whose structure was similar to PEGs and results in the deformation of PMIA structure (Figure S3). Therefore, neutral organic solute in aqueous solution without swelling effect were adopted to determine the actual MWCO. The obtained results show the narrower distribution trend and smaller pores of PPA2505 compared to PPA05 as the MWCO value of PPA2505 (574 Da) is smaller than that of PPA05 (646 Da) due to the MPD-TMC-β-CD@ZIF-8 net structure, which were fit for the concentration and purification of erythromycin (EM) (733.93 Da).

Figure 12

(a) MWCO vs rejection curves determined by PPG method and pore size distribution accompanied with cumulative distribution curves of PPA05 and PPA2505 (insets). (b) EM/solvent concentration and purification performances.

(a) MWCO vs rejection curves determined by n class="Chemical">PPG method and pore size class="Chemical">pan class="Chemical">distribution accompanied with cumulative distribution curves of PPA05 and PPA2505 (insets). (b) EM/solvent concentration and purification performances.

Figure b concludes the performances of the as-prepared β-CD@n class="Chemical">ZIF-8-doped TFN OSN meclass="Chemical">pan class="Chemical">mbranes in EM concentration application. A high concentration and a high efficiency of purification of EM/solvent could be reached with the as-proposed β-CD@ZIF-8-doped TFN OSN membrane. The application of the membrane in the EM/polar protic solvent successfully avoided the trade-off effect and reduced the influence of concentration polarization phenomenon depending on the hydrophilic membrane surface. To the best of our knowledge, it was the first time to simulate the EM production environment by using butyl acetate (BA) and operate EM/BA solution directly with membrane process. With higher viscosity (0.685 mPa s, 25 °C), higher molecular weight (116.16 g mol–1), and more steric hindrance effect compared to MeOH (0.545 mPa s, 32.04 g mol–1), the EM/BA permeance could not catch up to the such high EM/MeOH solution, but obtained a high rejection value of EM. From the results, we can found that the PPA displays the highest EM/BA flux among all tested samples, which was mainly ascribed to the worse solvent-resistant properties of PPA than those containing β-CD@ZIF-8, its lowest EM rejection value could not meet the industry requirements. With the introduction of β-CD@ZIF-8 into the PA selective layer, the solvent-resistance properties improved greatly, and the as-prepared TFN membrane could be sustained in the BA environment and offer a high rejection value. As the dosage of nanoparticles increases, the TFN membrane shows increased EM/MeOH flux, while the flux of EM/BA decreased. This was induced by the formation of MPD-TMC-β-CD@ZIF-8 net structure in the PA selective layer. It is a kind of net structure formed based on the MPD-TMC structure with more hydrophilic character, but it is more denser than the MPD-TMC structure. Smaller and polar MeOH molecules could pass easier than bigger BA molecules and obstruct more EM molecules to penetrate through it. PPA2505 contains nanoparticles in both PMIA support and PA selective layers, which shows optimum performances in EM concentration due to its denser MPD-TMC-β-CD@ZIF-8 net structure, which offers a feasible way for high-purity EM production compared to the conventional high-cost, time-consuming extraction process.

DMF Long-Term Operation

Figure shows the n class="Chemical">DMF long-term operation results of class="Chemical">pan class="Chemical">PPA, PPA05, and PPA2505 in DMF steam. The PPA membrane was tested for only 6 h to figure out the PMIA support dissolving behavior. It is no doubt that PMIA membrane would be dissolved in the DMF steam gradually due to the breakage of hydrogen bonds and small chain fractured during the spinning solution preparation step by using strong solvent dimethylacetamide (DMAc) at high temperature.[36] However, this drawback could eventually offer the as-prepared TFC/N OSN membranes a potential way to become an PI–PA ultrathin selective film with only a solvent-resistant PI nanofiber and a dense PA layer with high selectivity and stability. The PI nanofiber itself could hardly form a uniform PA selective layer easily because of its superhydrophobic (Figure b) and huge mean pore size (1.273 μm).

Figure 13

Results of DMF long-term operating and the OSN performance of RB/DMF rejection.

Results of n class="Chemical">DMF long-term operating and the class="Chemical">pan class="Chemical">OSN performance of RB/DMF rejection.

Three n class="Chemical">OSN meclass="Chemical">pan class="Chemical">mbranes show different behaviors when facing DMF. The “starting fluxes” of DMF are at a very low level of all three OSN membranes, which were in accordance with the pure DMF flux reported in Figure b, and follow the order PPA < PPA05 < PPA2505. These are the function of β-CD@ZIF-8 in the PMIA support membrane and the selective layer, which enhanced the solvent permeance obviously. During this period, the PMIA support membranes of all three tested samples began dissolving (0.0–1.0 h), with the PMIA support of PPA without any β-CD@ZIF-8 dissolving the fastest, and its pure DMF flux reaches the DMF flux of PPA2505. PPA05 with 0.05% (w/v) β-CD@ZIF-8 in the PA selective layer shows a compact phenomenon instead of dissolving directly at first, which proves the enhanced solvent-resistant properties. Thus, we can also speculate PPA2505 with β-CD@ZIF-8 in both PMIA support and PA selective layer, the same phenomenon as PPA05, but the modification effects of the nanoparticles cover the compact phenomenon of PPA2505. Then, when the low-molecular-weight part of the PMIA support of PPA dissolved completely, the compact phenomenon appeared (1.5 h). At the same time, PPA05 and PPA2505 could not discover the compact phenomenon any longer (1.0–2.5 h) and reach a steady state (2.5–5.0 h). However, PPA was facing the “swell effect” of the high-molecular-weight part of the PMIA support (4.0–4.5 h), the DMF flux was lifted again and kept steady till the end (4.5–6.0 h), the 35 μM RB/DMF was pumped, and the RB rejection was calculated to verify the OSN performance. PPA05 and PPA2505 would follow the same procedure after 12 h operation. In the period of 5.5–7.0 h, PPA05 and PPA2505 also face the swell effect, but with the help of β-CD@ZIF-8, they sustained more time than PPA. Comparing their performances, at the phase of 7.0–12.0 h, PPA2505 exhibited much more steady and higher DMF flux than PPA05, which could benefit from the additional β-CD@ZIF-8 in the PMIA support that PPA05 guiltless. The insets illustrate the feature of tested samples after pure DMF operating and RB/DMF testing, the high-molecular-weight part of the PMIA support could still be observed, while PPA05 and PPA2505 exhibit corrugated topography due to the presence of the nanoparticles, which endowed them high roughness (Figure d,e). RB/DMF rejection results displayed in the right-hand side of Figure could demonstrate the fact that PMIA support dissolving in the DMF steam has almost no influence on the OSN performances of the as-prepared β-CD@ZIF-8 TFN OSN membranes.

Table n class="Chemical">displays the class="Chemical">pan class="Chemical">OSN performances of other recent research works. The as-prepared β-CD@ZIF-8-doped TFN OSN membranes are full of competence to meet the requirements of OSN industry.

Table 4

Comparison of OSN Performances with Other Research Works[4,37−42]

membrane material	membrane type	additive	OSN performancesa	sustained harsh solvent
			(L m–2 h–1 bar–1, %)
PI nanofiber	double-PA layer TFN membrane	β-CD@ZIF-8	pure acetone flux: 12.6	DMF
			RB/MeOH rejection: 97%
continuous ZIF-8 membranes	polymer-supported MOF membrane	none	RB/EtOH flux: 2.5	not mentioned
			RB/IPA rejection: 94%
PI P84	co-deposited nanocomposite membrane	POSS–NH3+Cl–	RB/MeOH flux: 2.2	DMF
			RB/MeOH rejection: 84%
polyacrylonitrile (PAN)	TFC membrane	tannic acid	congo red/N-methyl-2-pyrrolidone (NMP) flux: 0.09	NMP
			RB/NMP rejection: 93%
PI P84	MMM	MWCNTs-COOH	RB/EtOH flux: 9.6	not mentioned
			RB/EtOH rejection: 85%
polypropylene	surface-coated nanocomposite membrane	graphene oxide, HPEI, poly(diallyldimethylammonium chloride), and poly(styrene sulfonate)	pure EtOH flux: 8.5	not mentioned
			RB/EtOH rejection: 97%
PEEKWC	integrally skinned asymmetric membrane	none	pure MeOH flux: <2.2	not mentioned
			RB/MeOH rejection: 90%
PAN	TFC membrane	morin	solvent flux: not mentioned	NMP
			RB/NMP rejection: 97%

Due to the variety of solvent performances of OSN membranes, the present table displays two parameters of the OSN performances in relevant references, which included the most prominent solvent flux (reflected under 1 bar) and RB rejection effect (RB in low-molecular-weight solvent is the priority, such as in MeOH, EtOH and then in n-PrOH, IPA, etc.).

Due to the variety of solvent performances of n class="Chemical">OSN meclass="Chemical">pan class="Chemical">mbranes, the present table displays two parameters of the OSN performances in relevant references, which included the most prominent solvent flux (reflected under 1 bar) and RB rejection effect (RB in low-molecular-weight solvent is the priority, such as in MeOH, EtOH and then in n-PrOH, IPA, etc.).

Conclusions

In conclusion, we have successfully developed novel n class="Chemical">TFN OSN meclass="Chemical">pan class="Chemical">mbranes with doping β-CD@ZIF-8 nanoparticles also proposed by us first. The membrane exhibits a traditional support membrane, a PA selective layer, and an integrated MPD-TMC-β-CD@ZIF-8 net structure for outstanding OSN performances and offers an efficient way for EM concentration and purification. The membrane performance could be further improved through adjusting PMIA support membrane concentration, balancing the contents of aqueous and organic phases, optimizing the dosage of β-CD@ZIF-8 nanoparticles, or post-treating with a mixed polar aprotic solvent, which we would discuss in our future work.

Experimental Section

Agentia and Reagents

n class="Chemical">PMIA was bought from Dupont. PI nanofiber (average pore size, 1.273 μm) was fabricated in our laboratory. Briefly, class="Chemical">pan class="Chemical">polyamide acid nanofiber was electrospun by using PAA solution, followed by heat dehydration.[43] PAA solution was purchased from Changzhou Furun Special Plastic New Materials Co., Ltd. TMC was acquired from Qingdao Benzo Chemical Company (China). Zn(NO3)2·6H2O, LiCl, and solvents used in this work include MeOH, EtOH, n-PrOH, IPA, n-BuOH, THF, AcOEt, acetone, DMF, DMAc, BA, and n-hexane, which were procured from Shanghai Titanchem Co., Ltd., while MeCN (high-performance liquid chromatography (HPLC) grade) was obtained from Acros Organics. HmIM (98%), MPD (99%), β-CD (99%+), EM (98%+), tripropylene glycol (99%), glucose, sucrose, and raffinose were bought from Adamas-β. PPG (Mw = 425, 725, 1000 g mol–1) and RB (Mw = 1017.64 g mol–1) were labeled Sigma-Aldrich. CV (Mw = 407.98 g mol–1, AR) and MB (Mw = 799.80 g mol–1, BS) were fabricated by Shanghai Chemical Reagent Co., Ltd. All reagents were of reagent grade unless otherwise mentioned.

β-CD@ZIF-8 Preparation

The as-proposed β-CD@n class="Chemical">ZIF-8 nanoclass="Chemical">particles were first preclass="Chemical">pared in a simple and effective way in our laboratory. The schematic class="Chemical">pan class="Chemical">diagram of the β-CD@ZIF-8 preparation process is shown in Figure a. Zn(NO3)2·6H2O (4.0 mmol) and HmIM (280.0 mmol) were fully dissolved in 8.0 and 80.0 g of DI water under stirring, respectively; 4.0 mmol β-CD (Znmol2+/β-CDmol = 1:1) was added into the HmIM solution and stirred until it dissolves. Then, the Zn(NO3)2 solution was poured into the HmIM/β-CD solution and stirring was continued for 5 min. After the obtained white suspension stood for 24 h at room temperature, centrifugation was carried out with DI water three times to ensure purity. The white powder was collected after the precipitate was dried in a vacuum oven at 60 °C for 24 h.[44]

PMIA Spinning Solution

Prior to the preparation of the casting solution, n class="Chemical">PMIA was rinsed with plenty of class="Chemical">pan class="Chemical">DI water and dried in an oven at 105 °C. During this time, certain amounts of LiCl and DMAc were mixed and the solution was heated up to 80 °C.[15] Then, the preweighed dry PMIA was gradually added into the LiCl/DMAc solution under mechanical stirring and kept for 24 h. The homogeneous brown transparent spinning solution was stored in the oven at 80 °C overnight before spin-coating for degassing. Detailed information is presented in Table . We note for the membranes PPA2505, the nanoparticles were dispersed into the LiCl/DMAc solution by 30 min ultrasonic treatment before feeding.

Table 5

Compositions of the As-Prepared β-CD@ZIF-8-Doped TFN OSN Membranes

	PMIA content (wt %)	nanoparticles in the PMIA support (wt %)		solvent (wt %)	nanoparticles in the PA selective layer (w/v %)
membrane codea	PMIA	LiCl	β-CD@ZIF-8	DMAc	β-CD@ZIF-8
PA	18.00	4.50		77.50
PA25	18.00	4.50	0.25	77.25
PPA	18.00	4.50		77.50
PPA05	18.00	4.50		77.50	0.05
PPA10	18.00	4.50		77.50	0.10
PPA2505	18.00	4.50	0.25	77.25	0.05

To help understand those membrane codes, where “PA” represents the PMIA support without IP; “PPA” and the numbers represent the TFN OSN membranes and the dosage of β-CD@ZIF-8 nanoparticles in the organic phase solution, respectively.

To help understand those men class="Chemical">mbrane codes, where “class="Chemical">pan class="Chemical">PA” represents the PMIA support without IP; “PPA” and the numbers represent the TFN OSN membranes and the dosage of β-CD@ZIF-8 nanoparticles in the organic phase solution, respectively.

PMIA Support Membrane Formation

The n class="Chemical">PMIA support meclass="Chemical">pan class="Chemical">mbrane was prepared by the NIPS method with the assistance of spin-coating technique. Specifically, PMIA spin coating was achieved via a spinner produced by Laurell (model WS-650MZ-23NPPB) equipped with a Rocker 300 vacuum pump in the nitrogen environment (0.4 MPa). The PI nanofiber was rinsed with a sufficient amount of IPA to remove impurities and dried in air before stuck onto the PMMA disk. To obtain the uniform membrane structure and thickness, configurations of the spinner never changed during the experiment, during which the spin speed was set to 3500 rpm for 7.0 s, 4000 rpm for 3.0 s, and 5000 rpm for 5.0 s in sequence. After spinning, the disk was immersed in a 25 °C DI water coagulation bath immediately and solvent exchange occurred. The obtained membranes were washed with enough DI water and stored in the DI water for further use.

PA Selective Layer Formation

The formation of n class="Chemical">PA selective layer was achieved via IP reaction. The present work used 2.0% (w/v) class="Chemical">pan class="Chemical">MPD and 0.15% (w/v) TMC (with or without nanoparticles) as aqueous-phase and organic-phase solutions, respectively. Generally, the above prepared membrane was fixed onto a stainless steel hoop with membrane surface upward and the liquor outside the membrane was removed using a tissue paper. The as-formed PMIA support membrane was treated in an aqueous-phase solution first for 2 min, followed by 1 min polymerization in an organic-phase solution.[45] Subsequently, the overdosage was rapidly removed and volatilized in the ambient air and then ripened in an oven at 80 °C for 5 min. The as-prepared β-CD@ZIF-8-doped TFN OSN membranes were washed and stored in DI water for further study.[31]

Characterizations

Powder X-ray n class="Chemical">diffraction (XRD) device (MiniFlex600, Jaclass="Chemical">pan) was employed to characterize the class="Chemical">pan class="Chemical">crystal phase of ZIF-8 and β-CD@ZIF-8 nanoparticles, in which Cu Kα radiation and 2θ range of 20–80° were used under the 40 kV 100 mA.[46]

A field emission scanning electron min class="Chemical">croscope (Nova NanoSEM 450, Jaclass="Chemical">pan) was used to capture images of the as-preclass="Chemical">pared nanoclass="Chemical">particles and class="Chemical">pan class="Chemical">TFN OSN membranes. All samples were gold-sputtered for 50 s for electroconductibility.[47,48] Nanoparticles were also observed by a high-resolution transmission electron microscope (JEM 2100, Japan).

An effective area of 10 μm × 10 μm for each sample mepan class="Chemical">mbrane was scanned by an atomic force miclass="Chemical">pan class="Chemical">croscope (Veeco, NanoScope IIIa Multimode AFM) in tapping mode to illuminate the 3D surface topography and roughness.[49]

With respect to chemicn class="Chemical">al compositions, a Fourier transform infrared spectrometer (Nicolet-6700) was used and the spectra of nanoclass="Chemical">particles and class="Chemical">pan class="Chemical">TFN OSN membranes were collected from 4000 to 600 cm–1. XPS (Thermo ESCALAB 250Xi) and EDX (JEOL JSM-6306LV) analyses were also conducted for quantification.[50,51]

Dynamic n class="Chemical">water contact angle (class="Chemical">pan class="Chemical">DCA) measurement of the nanoparticles and TFN OSN membranes was performed to judge the wettability. A contact angle analyzer (JC2000D, Shanghai Zhongchen Digital Technology Apparatus Co., Ltd., China) was used to capture nanoparticles for 60 s and membranes for 180 s with 2 μL of water droplet on the compacted nanoparticles and membrane surface at 25 °C.[52,53]

The n class="Chemical">OSN performances of class="Chemical">pan class="Chemical">TFN OSN membranes were evaluated according to a variety of pure solvent fluxes (PSFs), dye/solvent separation performance, MWCO determination, and its application in pharmaceutical recovery. A self-designed cross-flow OSN filtration system with an effective filtration area of 4π cm2 was propelled by a gear pump (WT3000-1JB, Longer Pump, China) with sealed pipelines made of stainless steel. The operation pressure was maintained at 0.6 MPa. All reported values were measured at least thrice and averaged.

PSF Tests

PSF was tested through a series of polar protic solvents (e.g., n class="Chemical">MeOH, class="Chemical">pan class="Chemical">EtOH, n-PrOH, and IPA) and polar aprotic solvents (e.g., THF, AcOEt, acetone, and DMF) and their properties are shown in Table S2. The sample membrane was prewetted with the testing solvent and rinsed three times before loading onto the OSN cell. After 30 min preloading before collecting the permeance, a 5.0 mL graduated cylinder was placed under the permeate pipeline and sealed. After the graduated cylinder was filled with 5.0 mL (0.005 L) of the testing solvent, the collecting time was recorded and the PSF (L m–2 h–1) was obtained according to the equationwhere V is the permeate volume (L), A is the effective filtration area (m2), and t is the permeate time (h).

Dye/Solvent Separation

n class="Chemical">CV (Mw = 407.98 g mol–1), class="Chemical">pan class="Chemical">MB (Mw = 799.80 g mol–1), and RB (Mw = 973.67 g mol–1) were used as the target solutes in dye/solvent separation experiment. The solute properties are presented in Table S3. A small quantity (35 μM) of each dye in MeOH and THF were prepared separately.[54] The feed and the permeate solution were collected after preloading and determined by a UV spectrophotometer (UV-1800, Shimadzu, Japan). Then, the dye separation could be achieved through rejection according towhere Cf and Cp are the feed and permeate solution absorbances, respectively.[55]

MWCO vs Rejection Curve Determination

The MWCO determination of n class="Chemical">TFN OSN meclass="Chemical">pan class="Chemical">mbranes was achieved by a reverse-phase HPLC system with PPG oligomers as the sample detector, and detailed information was provided by Patterson et al.[35] In the present work, a Waters HPLC system consists of a 1525 binary pump, a 2420 ELSD, and a column oven within a column (InfinityLab Poroshell 120 EC-C18, 4.6 × 250 mm2, 2.7 μm) for analysis. DI water and MeCN were used as the mobile phase, and the system was operated under nitrogen atmosphere at 0.5 MPa. The purpose of HPLC system determination was to obtain the concentration of each unit of PPG in the feed and permeate. The feed solution was mixed with 4.0 g L–1 of PPG 425, PPG 725, and PPG 1000 in MeOH, and the permeate solution was collected after 30 min preloading. The rejection of each PPG unit could also be calculated from eq . In addition, pore size distribution curves of TFN OSN membrane were established according to the rejection value of n-BuOH, glucose, sucrose, and raffinose (74, 180, 342, and 504 g mol–1, Table S3) with further calculation through log-normal model using MATLAB software.[56]

Erythromycin Concentration Application

n class="Chemical">Erythromycin (EM, Mw = 733.93 g mol–1) is an antibiotic widely used in pharmacology, whose concentration and purification from fermentor still suffer from high cost and low efficiency. After the class="Chemical">pan class="Chemical">TFN OSN has been developed, it finds application in EM concentration and purification. We simulated an EM sample of low concentration (500 ppm) in both MeOH and BA. A total organic carbon (TOC) instrument was used to determine the concentration of EM in the feed and the permeate. In detail, 1.0 mL of feed and permeance were added into a sample bottle separately, dried in a vacuum oven, and redissolved by 15.0 mL of DI water, followed by TOC measurement. The data were tested thrice and their mean values were reported.

Swelling Behavior

n class="Disease">Swelling behavior experiment was conducted by immersing the class="Chemical">pan class="Chemical">TFN OSN membranes in eight solvents, including MeOH, EtOH, IPA, THF, EtOAc, acetone, DMF, and DI water, for 24 h, and the area swelling ratio and solvent uptake ratio were calculated according to eqs and 4.where Adry and Awet are the membrane areas before and after solvent immersion, respectively, and Wdry and Wwet are the membrane weights before and after solvent immersion, respectively.[57,58]

Mechanical Properties

The mechanicn class="Chemical">al properties were tested using an electronic mechanicclass="Chemical">pan class="Chemical">al property analyzer (QJ210A, Shanghai Qingji Instrument Technology Co., Ltd., China) to report the mean break strength, elongation at break, and Young’s modulus of the as-prepared OSN membranes. Three samples of each membrane were cut into 50 mm × 15 mm rectangles and tested at 50 mm min–1.[59,60]

DMF Long-Term Stability

n class="Chemical">DMF long-term stability was tested through operating the meclass="Chemical">pan class="Chemical">mbrane PPA, PPA05, and PPA2505, under pure DMF steam at 0.6 MPa. PPA05 and PPA2505 were tested consecutively for 12 h, and PPA for 6 h to figure out the DMF stability differences between the as-prepared TFC and TFN OSN membranes. The pure DMF flux was recorded every 0.5 h and then 35 μM RB/DMF was pumped through those OSN membranes instead of pure DMF. The permeance and feed were collected and analyzed by a UV-1800 spectrophotometer to determine the RB rejection.