Bioadhesion-inspired fabrication of robust thin-film composite membranes with tunable solvent permeation properties

Abstract

Dopamine chemistry arising from marine mussel bioadhesion principles has attracted growing interest in designing and fabricating robust thin films/membranes for various chemical separation processes. In this study, inspired by the whole adhesion system of marine mussels, polydopamine nanoparticles (PDNPs) are employed to fabricate thin-film composite (TFC) membranes via conventional interfacial polymerization method. To be specific, PDNPs are dispersed into the aqueous polyethyleneimine (PEI) solution so as to construct “cross-linked particles in matrix” architectures. The incorporated PDNPs are found to play two major functions: (1) PDNPs reduce the hydrophilicity of membrane surface and enhance the permeate flux of non-polar organic solvents; (2) PDNPs react with PEI and result in covalent bonding between fillers and polymer matrix, which not only effectively inhibits the chain mobility and enhances solvent resistance, but also enables size-dependent selectivity. Compared with the three-layer composite membrane in which a polydopamine interlayer is present between the PEI layer and the porous support layer, the TFC membrane with mixed-matrix active layer shows enhanced swelling resistance and rejection ability, hinting that PDNPs can reinforce the entire active layer of the membrane, rather than merely the interfacial region. In addition, the gradual increment of rejection ratio and slight decrement of flux during a constant 720 min test show excellent operational stability of membrane.

---

1. Introduction

Thin-film composite (TFC) membranes have been widely applied in various membrane processes due to the combination of high permeation flux and excellent mechanical strength.^1–7 Presently, interfacial polymerization (IP) is regarded as the predominant method for the preparation of TFC membranes due to the following two advantages: (i) facile and mild fabrication process, reducing environmental pollution and simplifying the scale-up production; (ii) tunable membrane thickness, allowing the formation of active layer with ultrathin and detect-free structure.^8–11 Nevertheless, the active layer prepared via IP is mostly made of polyamide because of the limited kinds of monomers. Such polymeric materials are often subjected to severe swelling in organic solvents due to the similar solubility parameters between polymer and solvent molecules. Although increasing the crosslinking density can restrict swelling, while a simultaneous decline of permeation flux often occurs.¹² Therefore, it remains still a huge challenge to fabricate robust membrane materials with tunable solvent transport properties.

In recent years, dopamine chemistry arising from marine mussel's bioadhesion principles has been widely employed to design and fabricate robust thin films/membranes.^5,13–19 To be specific, inspired by the chemical structure of mussel's protein adhesives, Messersmith et al. found that dopamine, a small-molecule mimic, could spontaneously polymerize into an ultrathin film which could firmly adhere to virtually all type of surfaces due to multiple interactions.²⁰ Furthermore, polydopamine could proceed to react with thiol- or amino- containing species via Michael addition or Schiff based reactions.^20,21 In this sense, polydopamine allows multi-functional surface modification, which could be employed to enhance the interfacial adhesion between two incompatible phases. For example, fabrication of TFC membranes using polydopamine-modified support membrane could significantly enhance the adhesion between the adjacent active layer and support layer, and thus prevent the asynchronous swelling of them.^5,22,23 On the other hand, Waite's group and Wilker's group found that mussel's bioadhesion phenomenon not only relies on its firm attachment to rocky surface, but also results from its robust byssus, where numerous rigid submicron granules (mainly consist of Fe³⁺-oxidized DOPA nanoaggregates) are crosslinked by flexible byssus proteins.^24–26 Collectively, it is thus envisaged that crosslinking polydopamine nanoparticles (PDNPs) with polyamine chains during IP process could result in robust “cross-linked particles in matrix” architectures and improve the anti-swelling properties. Simultaneously, according to our pervious study, PDNPs are expected to efficiently interrupt the polymer chain packing and render favorable free volume for solvent permeation.²⁷

In this work, a series of TFC membranes comprising crosslinked polyethyleneimine (PEI) and PDNPs were prepared by interfacial polymerization method for the first time. The PDNPs were introduced into the thin-film active layer by subsequently immersing PAN support membrane into PDNP-dispersed PEI aqueous solution and trimesoyl chloride (TMC) solution, respectively. The microstructures and physicochemical properties of the membranes were manipulated via altering PDNPs content. The solvent uptake, area swelling, permeation flux, and rejection of the membranes for both polar and non-polar solvents were investigated in detail. Additionally, the long-term operation stability was also explored to predict the potentials in practical applications.

2. Experimental

2.1. Materials and chemicals

Dopamine hydrochloride was supplied from Yuancheng Technology Development Co., Ltd. PEI (M_w 20 000 g cm⁻¹), and TMC were purchased from Alfa Aesar and used without further purification. Polyacrylonitrile (PAN) ultrafiltration membrane with molecular weight cut-off (MWCO) of 50 kDa, which served as a support, was provided by shanghai MegaVision membrane engineering & Technology Co., Ltd. Isopropanol, ethyl acetate, hexane, n-heptane, toluene, and sodium hydroxide were obtained from Kewei Chemistry Co., Ltd. Polyethylene glycol (PEG, M_w 200, 400, 600, 800, 1000, and 2000 g cm⁻¹), and sodium carbonate were supplied from Guangfu Fine Chemical Research Institute. Deionized water was used throughout the experiment.

2.2. Synthesis of PDNPs

The PDNPs were synthesized via covalent oxidative polymerization and non-covalent self-assembly in the atmosphere, and the detailed mechanism of dopamine polymerization had been clearly demonstrated in this literature.²⁸ Dopamine hydrochloride (2.0 g) was dissolved in deionized water (1.0 L) under mild stirring at 50 °C. Afterwards, 8.44 mL of 1 N NaOH solution was rapidly injected into the dopamine hydrochloride solution under vigorous stirring. The color of this mixed solution immediately turned into pale yellow as soon as NaOH solution was added and gradually changed to dark brown. After reacting for 5 h, the PDNPs were retrieved by a PAN ultrafiltration membrane with a molecular weight cut-off of 50 kDa and washed repeatedly and sequentially with deionized water and ethanol to remove residual dopamine oligomers. In the end, the PDNPs were dried in a vacuum oven overnight at 60 °C. During this procedure, dopamine was firstly oxidized and polymerized into oligomers consisting of dihydroxyindole, indoledione, and open-chain dopamine units, and then the oligomers self-assembled into nanoaggregates driven by multiple interactions.^29,30

2.3. Preparation of TFC membranes

The TFC membranes were prepared via interfacial polymerization technique and the preparation process was shown in Scheme 1 (note that the PDNPs were not in scale with the rest of the materials in this scheme). To begin with, PAN support was immersed into water for 30 min to retard the invasion of PEI molecules into PAN pores. PEI aqueous solutions (4 wt%, 50 g) (containing a certain amount of the PDNPs and 0.05 g Na₂CO₃ as acid-acceptor) and trimesoyl (TMC) organic solution (2 wt%, 50 g) were prepared, respectively. The resultant aqueous solution was gently dip-coated upon the top of PAN support at 25 °C. After 10 min immersion, the membrane was taken out of the solution, and the excess solution was removed by holding the membrane perpendicularly until no liquids remained. Afterwards, the TMC organic solution was cast onto the coated PEI membrane for another 10 min at 25 °C, which resulted in the formation an ultra-thin film over the surface of PAN support. The immersion time in both aqueous solution and organic solution was determined by seeking for a compromise between permeation flux and rejection ratio. Finally, the resulting TFC membrane was air dried for 1 h, followed by being cured in a 60 °C oven for another 2 h to ensure interfacial reaction has been completed and the residual solvent has been removed. For comparison, PAN/PEI membrane was prepared by the same procedure as above without adding PDNPs. For simplicity, due to the “mixed-matrix” structure of the active layer, the PDNPs-embedded membranes were named as MMM-PDNPs-X, where MMM meant “mixed matrix membrane” and X (X = 0.1, 0.2, 0.3, or 0.5) represented the weight percentage (wt%) of the PDNPs to PEI.

Scheme 1 Schematic fabrication process and network structures of PDNPs-filled TFC membrane.

2.4. Characterization

The morphology of the PDNPs was observed by transmission electron microscopy (TEM, JEOL JEM-100CXII). The chemical structure of the PDNPs and the membranes were characterized by attenuated total reflection-Fourier transform infrared spectra (ATR-FTIR) using a Nicolet MAGNA-IR 560 instrument in the range of 4000–700 cm⁻¹ with the resolution of 4 cm⁻¹ under ambient atmosphere without special pretreatment. The size distribution of the PDNPs was determined by a Particle Size Analyzer 90 Plus (Brookhaver Instruments Corporation). The surface and cross-section images of the membranes were observed by scanning electron microscope (SEM, JSM7500F) after being freeze-fractured in liquid nitrogen and then sputtered with Au. Surface morphologies and roughness of the membrane samples were obtained by atomic force microscopy (AFM, Veeco MultiMode, United States). The static contact angles of the TFC membranes for water were obtained at room temperature with a contact angle goniometer (JC2000C Contact Angle Meter, Powereach Co.).

2.5. Measurement of solvent uptake and area swelling

The stability and chemical resistance behavior of the TFC membranes were evaluated by solvent uptake and area swelling measurement in n-heptane, toluene, isopropanol, and ethyl acetate. The membranes were first placed in a 60 °C oven for 48 h and then tailored into squared-shaped samples (about 1.5 cm × 1.5 cm). The weight (W_dry, g) and area (A_dry, cm²) were measured accurately. Subsequently, the samples were immersed in a certain above pure solvent for 48 h at room temperature. Finally, both the weight (W_wet, g) and size (A_wet, cm²) were retested rapidly after removing the residual solvent on the membrane surface. All measurements were performed three times and an average of the three results with the error less than 5% was regarded as the final value. The solvent uptake and area swelling were calculated by eqn (1) and (2), respectively:

2.6. Solvent permeation test

All solvent permeation experiments were performed in a home-made dead-end stirred cell with a volume capacity of about 200 mL and an effective membrane area of 18.2 cm² as presented in Scheme 2. During the test, the feed solution was driven to get through the membrane by the pressurized N₂. A membrane was first immersed in a pure selected solvent (including n-heptane, toluene, isopropanol, and ethyl acetate) for at least 48 h for complete equilibration. Prior to the test, the membranes were compacted by pure solvent for at least 30 min. Upon different pressures (4 and 10 bar), the amount of permeation that transported through the membrane was measured by weighing the liquid collected in a flask. Solvent flux (J, L m⁻² h⁻¹) was defined as the volume permeated per unit area per unit time according to following eqn (3):where V, A, and t were the permeate volume (L), membrane surface area (m²), and permeation time (h), respectively.

Scheme 2 Schematic representation of filtration unit.

2.7. Rejection measurement

For the rejection measurement, feed solutions (500 mg L⁻¹) were prepared by dissolving a series of PEG oligomers with the molecular weights ranging from 200 to 2000 Da in water. The rejection of all membranes was determined under the 10 bars at room temperature and constant speed of 300 rpm. The rejection ratio (R, %) was evaluated applying eqn (4) in which C_p and C_f correspond to PEG concentration in permeate and in feed solution, respectively, measured by a UV-vis spectrophotometer (PerkinElmer Lambda25).

All the results presented were the average data from three measurements, with the error less than 5%.

3. Results and discussion

3.1. Physicochemical properties of PDNPs

The physicochemical properties of PDNPs were investigated by TEM, FTIR as presented in Fig. 1. The TEM image in Fig. 1a suggests that PDNPs display regular spherical shapes with sizes about 100 nm.²⁸ Particle size analysis suggests a narrow size distribution around 100 nm (Fig. 1b), which is consistent with TEM observation for PDNPs. The FTIR spectrum of PDNPs is depicted in Fig. 1c. The absorption band at 1290 cm⁻¹ is ascribed to the phenolic C–O stretching vibration. Moreover, the characteristic bands at 1512 and 1602 cm⁻¹ are related to the N–H shearing and bending vibration of the amine group, respectively.²⁷ Additionally, we also attempted to tune the size of PDNP from about 100 nm to 300 nm by varying the synthesis conditions. The particles with the size below 100 nm were not selected because the severe agglomeration was found to occur. However, by using larger particles with the size about 200 or 300 nm, we found that it is difficult to obtain a thin-film active layer (thickness of the active layer often exceeds 1 μm to prevent defects), and hence the permeation flux became rather low.

Fig. 1 (a) TEM image, (b) lognormal size distribution, (c) FTIR spectrum of PDNPs.

3.2. Physico-chemical characterization of membranes

FTIR spectrum of PAN support, PAN/PEI, and MMM-PDNPs-0.5 are investigated and presented in Fig. 2 to analyze the chemical structural and corresponding reactions within the membranes. The strong adsorption peak of PAN support at 2238 cm⁻¹ is assigned to the –C≡N stretching vibration. By comparison, two new bands around 1353 and 1554 cm⁻¹ are observed and that at 2238 cm⁻¹ disappears in the spectrum of PAN/PEI. These results demonstrate that the surface of PAN support is completely covered by a PEI layer.

Fig. 2 FTIR spectra of PAN support, PAN/PEI, and MMM-PDNPs-0.5.

Besides, the formation of amide bonds, verified by the C=O stretching vibration peak at 1616 cm⁻¹, is likely due to the cross-linking between –NH–/–NH₂ in PEI and –COCl in TMC through nucleophilic substitution reaction. Owing to the addition of PDNPs, a new absorption band at 1654 cm⁻¹ appears in the spectrum of MMM-PDNPs-0.5, which is ascribed to the formation of C=N bonds between PEI and PDNPs.^31,32

The surface and cross-section micrographs of MMM-PDNPs-X are presented by SEM images in Fig. 3. It can be clearly seen from the surface SEM images in Fig. 3a–c that the interfacial polymerization between PEI-PDNPs and TMC generated a relatively dense and rough selective layer on the PAN support without visible defects.⁹ Spherical granular protuberances are clearly observed and uniformly dispersed within the PEI matrix according to the respective insets of Fig. 3b and c. Compared with PAN/PEI in Fig. 3a, the granular protuberance increased the surface area of MMMs due to the presence of PDNPs, which is beneficial to solvent adsorption and permeation. Fig. 3d shows the cross-section morphology of MMM-PDNPs-0.1, which reveals that the thickness of the active layer is around 200 nm. Moreover, the interface between the adjacent layers is not very clear, indicative of the good compatibility between the mixed matrix active layer and the PAN support layer.

Fig. 3 SEM images of the surface of: (a) PAN/PEI, (b) MMM-PDNPs-0.1, (c) MMM-PDNPs-0.5, and (d) the cross-section of MMM-PDNPs-0.1.

AFM provides further analysis of the surface characteristics of MMM-PDNPs-X. The typical three-dimensional AFM surface topography pictures for the prepared membranes are presented in Fig. 4. It could be observed that the PAN/PEI has a relatively rough surface, with the root mean square roughness (RMS) value of about 51.1 nm, consistent with the description of roughness in the other literature.³³ Compared with PAN/PEI, incorporating 0.1% PDNPs leads to a higher RMS value of about 80.2 nm.

Fig. 4 AFM analysis of (a) PAN/PEI, (b) MMM-PDNPs-0.1, (c) MMM-PDNPs-0.5.

Simultaneously, the maximum roughness (R_max) increases from 309 to 681 nm, and the observed R_max difference (372 nm) is quite larger than the diameter of PDNPs. These phenomena may be ascribed to the following two reasons: (i) some PDNPs (including their sub-micron aggregates) appearing on the membrane surface contribute to the enhancement of RMS; (ii) PDNPs increase the degree of cross-linking by covalence bond and hydrogen bond as depicted in Scheme 1, which is expected to increase chain stiffness and generate more protrude wrinkles.³⁴ By further increasing PDNPs loading to 0.5%, the R_max value does not show obvious change, indicative no further agglomeration of PDNPs with the increase of filling content. Simultaneously, the RMS value further increases up to 88.4 nm, suggesting that higher loading of PDNPs is expected to result in higher degree of crosslinking and more protrude wrinkles.

The adsorption and permeation of solvent molecules are strongly influenced by the hydrophilic/hydrophobic nature of membrane surface. Contact angle values of PAN, PAN/PEI, and MMM-PDNPs-X are shown in Fig. 5. The contact angle of PAN support is measured to be 55.5°. After interfacial polymerization, the contact angle of PAN/PEI is reduced to 39.9°, which is probably attributed to the fact that PEI chains on the surface of membrane provide abundant unreacted hydrophilic groups, including –NH– and –NH₂. Compared with the PAN/PEI, the presence of PDNPs notably increases the contact angle of MMM-PDNPs-0.1 to 56.1°. As is well known, the surface hydrophilicity of membrane is mainly determined by both surface chemistry and surface roughness. Firstly, the incorporation of PDNPs greatly change the surface chemistry of MMM-PDNPs-0.1, which is reasonably attributed to the following two reasons: (1) the hydrophilic groups (–NH– and –NH₂) from PEI chains and PDNPs react with acyl chloride group to form amide groups;^35,36 (2) PDNPs inherit catechol and N–H groups from their starting material, which can react with PEI chains.^20,21 On other hand, the introduced PDNPs gives rise to a significant increase in RMS (revealed by Fig. 4). Although the increased surface roughness is beneficial for enhancement of hydrophilicity, the sharply reduced amount of hydrophilic groups on membrane surface determines the overall decrease of surface hydrophilicity. With the increase of the PDNPs content, the contact angle of MMM-PDNPs-X (0.2, 0.3, 0.5) are measured to be 50.9°, 52.3° and 49.1°, respectively. The partial recovery of surface hydrophilicity at high PDNPs loading may be attributed to the appearance of unreacted hydrophilic groups (–OH, –NH– and –NH₂) from PDNPs as well as the higher surface roughness.

Fig. 5 Contact angles of PAN, PAN/PEI, and MMM-PDNPs-X.

Herein, four typical solvents, namely n-heptane, isopropanol, toluene, and ethyl acetate, were employed to evaluate the solvent resistance of the as-prepared TFC membranes. As shown in Fig. 6, PAN support displays low area swelling (below 1.4%) in four solvents, but relatively high solvent uptake (around 20%). The high solvent uptake and low area swelling imply that the adsorbed solvent molecules are mainly stored in the pores of PAN support.^37,38 Compared with PAN support, PAN/PEI shows about one-fold increment in area swelling and a bit increment in solvent uptake for each solvent. These results indicate that pristine PEI as active layer can be easily swollen by the adsorbed solvent. Fortunately, the incorporation of PDNPs reduces the solvent uptake and area swelling of membrane for both polar solvents and non-polar solvents (except MMM-PDNPs-0.1 which may result from the relatively high hydrophobicity). For example, embedding 0.3 wt% PDNPs confers low area swelling down to 0.9%, 1.8%, 0.8%, and 1.5% on membrane in n-heptane, isopropanol, toluene, and ethyl acetate, respectively. The reduction of uptake and swelling is reasonably ascribed to the presence of PDNPs, which restricts the mobility of PEI chains through covalent and hydrogen bonding interactions. Under identical conditions, the solvent uptake and area swelling of MMM-PDNPs-X in polar solvent (isopropanol, ethyl acetate) are distinctly higher than that in nonpolar solvent (n-heptane, toluene). This finding is reasonably ascribed to the hydrophilicity of PEI chains, which enhances the adsorption capability of the MMMs for polar molecules but decreases the adsorption capability for nonpolar molecules. For the two nonpolar solvents, the solvent uptake and area swelling of MMM-PDNPs-X in n-heptane are slightly higher than that in toluene, probably due to the relatively larger molecule size of toluene over n-heptane. For the polar two, the solvent uptake and area swelling of MMM-PDNPs-X follow the order that ethyl acetate < isopropanol, in contrast to the order of their dipole moments. This finding suggests that the hydrophilic groups (–NH– and –NH₂) in PEI chains and the hydrophilic groups (–OH, –NH– and –NH₂) in PDNPs provide abundant hydrogen-bonding sites, which promote the adsorption and diffusion of isopropanol via hydrogen-bonding interactions. In addition, further testified by Hansen solubility parameters, the hydrogen bonding solubility parameter of isopropanol (16.4 MPa^1/2) is larger than that of ethyl acetate (7.2 MPa^1/2).³⁹

Fig. 6 The (a) solvent uptake and (b) area swelling of PAN, PAN/PEI, and MMM-PDNPs-X using n-heptane, isopropanol, toluene and ethyl acetate at room temperature.

3.3. Nanofiltration performances of MMM-PDNPs-X

Fig. 7 exhibits the permeation fluxes of PAN/PEI and MMM-PDNPs-X for pure n-heptane, toluene, isopropanol, and ethyl acetate under the trans-membrane pressures of 4 and 10 bar. Generally speaking, due to the hydrophilicity of membrane surface, polar solvents (isopropanol, ethyl acetate) exhibit higher permeation ability than nonpolar solvents (n-heptane, toluene) for all the membranes. Compared with toluene, n-heptane exhibits apparently higher permeation ability in all the membranes, in accordance with the larger dynamic diameter of toluene (5.9 Å) than n-heptane (4.3 Å). For two polar solvents, due to the mutual hydrogen-bonding interactions between isopropanol and active layer, isopropanol possesses the higher permeability than ethyl acetate. On the other hand, incorporating PDNPs decrease the permeability of four solvents for most cases. The reduction of permeance should be ascribed to the following factors: (i) the restrained chain mobility in the active layer, which would weaken the mobility of polymer chains, and (ii) the reduced area swelling of the membrane, which would raise the mass transfer resistance for solvent molecules. These influences become stronger with the increase of PDNPs content, and consequently the permeance continuously reduced.

Fig. 7 Influence of the content of PDNPs on permeance of (a) n-heptane, (b) toluene, (c) isopropanol, and (d) ethyl acetate at room temperature and under the transmembrane pressures of 4 bar and 10 bar.

It should be noted that the permeate flux of either n-heptane or toluene for MMM-PDNPs-0.1 is higher than that of other membranes (including PAN/PEI), which is reasonably ascribed to the increase of hydrophobicity (revealed by contact angle) and area swelling. According to the solution-diffusion mechanism, an enhanced hydrophobic surface of membrane could enhance the adsorption capability for nonpolar molecules and facilitate the dissolution of non-polar solvents to lower transporting resistance. Similarly, relatively high permeate fluxes are observed for isopropanol and ethyl acetate when PDNPs concentration reaches 0.2 wt%. Such interesting phenomena can be interpreted by the following aspects. (1) MMM-PDNPs-0.1 possesses the most hydrophobic surface (contact angle 56.1°), which favors the adsorption and permeation of nonpolar solvents (n-heptane, toluene); (2) MMM-PDNPs-0.2 shows relatively more hydrophilic surface (contact angle 50.9°) than MMM-PDNPs-0.1, which allows higher flux of polar solvents; (3) embedding PDNPs increases the crosslinking density and restricts the PEI chain mobility, rendering the MMMs lower free volume and permeate flux for all the four solvents. Furthermore, it can be found that MMM-PDNPs-0.2 shows higher ethyl acetate flux but lower isopropanol flux than PAN/PEI. This tiny difference implies that the permeate flux at low PDNPs loading is dominated by surface adsorption, which is determined by whether the polarity of solvent matches the hydrophilicity of membrane surface. Nevertheless, the appearance of a maximum permeate flux for MMMs at certain PDNPs loading demonstrates that the MMMs prepared in this study allow simultaneously enhancement in permeability and anti-swelling ability, especially when the solvents possess weak polarity.

Another interesting finding lies in the pressure-dependent flux of different solvents. As the pressure increases from 4 to 10 bar, a significant increment of n-heptane flux (from 2.8 to 9.1 L m⁻² h⁻¹) is observed for MMM-PDNPs-0.1, while the toluene flux only increases by 60%. Furthermore, it is calculated that MMM-PDNPs-0.1 displays a rather high ideal n-heptane/toluene selectivity of 5.69 under 10 bar. These results appear surprising because n-heptane and toluene possess similar polarity, while they can be interpreted by the following hypothesis. Since polymeric membranes tend to be compacted and densified under high pressure, the diffusivity of solvent molecules may play more important roles than solubility. Heptane is a linear and flexible molecule, and it is expected to diffuse into membrane more easily than toluene, which contains rigid benzene rings. This hypothesis is supported by the case of polar solvents, because the flux enhancement of isopropanol (1.75 fold) is larger than that of ethyl acetate (1.13 fold), and the latter exhibits lower flexibility due to the rigid planar ester bond. Therefore, it is reasonable to assume that both the shape of a solvent molecule can significantly affect its diffusivity in polymers, especially when they are highly compacted.

The rejection ability of membrane is reflected by MWCO, which is calculated based on the rejection of solute above 90%.⁴⁰ Fig. 8 depicts the rejection curves of PAN/PEI and MMM-PDNPs-X using polyethylene glycol (PEG) 200-2000 as solutes. Compared with PAN/PEI (MWCO 695), the addition of PDNPs remarkably enhances the rejection ability of membranes. For instance, the MMMs exhibit the MWCOs of 380 and 253 when the concentration of PDNPs are 0.1 and 0.2 wt%, respectively. Besides, when the concentration of PDNPs exceeds 0.3 wt%, the rejections of MMMs for PEG 200 are higher than 90%, which means that the MWCOs are less than 200 Da. The rejection enhancement is most likely due to the presence of PDNPs, which effectively reduces the chain mobility and meanwhile lengthens the transfer pathway for PEG.⁴¹

Fig. 8 Rejection curves of PAN/PEI and MMM-PDNPs-X under the transmembrane pressure of 10 bar (the membranes were immersed in isopropanol for 48 h prior to testing).

3.4. The comparison of embedding PDNPs with inserting polydopamine interlayer

In this study, TFC membrane with PEI-PDNPs active layer is facilely prepared inspired by bioadhesion for the first time, and the obtained MMMs display fairly good performances in terms of solvent resistance, nanofiltration, and operational stability. On the other hand, it is known that employing polydopamine coating as interlayer could also enhance the resistance against swelling. To compare the two ways which both involve bio-adhesion inspired polydopamine, another two kinds of membranes, namely PAN/PD and PAN/PD/PEI, were prepared. The PAN/PD was prepared by immersing PAN into dopamine solution (100 mL, 2 g L⁻¹) for 1 h at 25 °C under continuous shaking. The as-prepared PAN/PD was selected as the support membrane to prepare PAN/PD/PEI, of which the active layer was fabricated via interfacial polymerization under the same conditions as PAN/PEI.

The influence of the introduced way of polydopamine upon the solvent resistant ability is investigated by testing the area swelling and solvent uptake. The results in Fig. 9 reveal that all the membranes display relatively high solvent uptake but low area swelling, inferring that the adsorbed solvent molecules are mainly in the pores of PAN support. Among these membranes, PAN/PD displays the lowest area swelling and solvent uptake in four solvent, indicating that the thin polydopamine layer has little influence on the uptake and swelling of composite membrane. Compared with PAN/PEI, PAN/PD/PEI and MMM-PDNPs-0.1 show slightly enhanced solvent resistance in the four solvents.

Fig. 9 The (a) area swelling and (b) solvent uptake of PAN/PD, PAN/PD/PEI, PAN/PEI and MMM-PDNPs-0.1 using n-heptane, isopropanol, toluene, and ethyl acetate at room temperature.

However, there is an important difference between PAN/PD/PEI and MMM-PDNPs-0.1. That is, the former shows relatively high swelling in polar solvents but apparently low swelling in nonpolar solvents, while the latter shows moderate swelling for each solvent. This tiny difference suggests that the PEI layer of PAN/PD/PEI possesses relatively high hydrophilicity, because the polydopamine interlayer cannot affect the bulk and surface properties of PEI layer. In this way, this membrane may suffer from excessive swelling in polar solvents and insufficient swelling in nonpolar solvents. On the contrary, PDNPs can efficiently affect the bulk and surface properties of PEI layer, resulting in highly crosslinked membrane bulk and well controlled surface hydrophilicity. Consequently, MMM-PDNPs-0.1 allows moderate swelling in different organic solvents, which is important to reach a compromise between permeate flux and stability.

The nanofiltration performances of the four types of membranes were evaluated in terms of permeation and rejection ability. Table 1 shows remarkably high flux for PAN/PD, implying PAN/PD is still an ultrafiltration membrane. Compared with PAN/PEI, PAN/PD/PEI shows lower permeate fluxes for n-heptane, toluene, isopropanol, and ethyl acetate. The flux reduction is probably due to the reduced membrane swelling, inhibited chain mobility, and increased membrane thickness, all of which will raise the mass transfer resistance for solvent molecules. It is notable that MMM-PDNPs-0.1 displays higher flux of these four solvent than PAN/PD/PEI, especially for two non-polar solvents. This result is likely due to the following two reasons: (i) MMM-PDNPs-0.1 is expected to possess lower interfacial adhesion than PAN/PD/PEI, which reduces the mass transfer resistance near the interfacial district; (ii) MMM-PDNPs-0.1 may displays higher hydrophobicity, which enhances its adsorption ability for nonpolar molecules. Also, under high pressure (10 bar), the n-heptane/toluene selectivity of PAN/PD/PEI (4.00) is found much lower than that of MMM-PDNPs-0.1 (5.68), indicating that the dense packing of PEI chains in the active layer of MMM-PDNPs-0.1 produces size-dependent selectivity.

Table 1 Permeate flux (L m⁻² h⁻¹) of PAN/PD, PAN/PD/PEI, PAN/PEI, and MMM-PDNPs-0.1 for n-heptane, toluene, isopropanol, and ethylacetate

Membrane	Pressure	n-Heptane	Toluene	Isopropanol	Ethyl acetate
PAN/PD	4 bar	17.4	20.1	35.9	31.2
	10 bar	61.1	60.8	83.3	76.7
PAN/PD/PEI	4 bar	1.1	0.3	10.5	5.7
	10 bar	2.4	0.6	21.5	11.6
PAN/PEI	4 bar	1.6	0.5	13.4	7.5
	10 bar	3.8	1.0	28.2	16.1
MMM-PDNPs-0.1	4 bar	2.8	1.0	8.0	6.2
	10 bar	9.1	1.6	22.0	13.2

---

Fig. 10 presents the MWCO curves of PAN/PD, PAN/PD/PEI, PAN/PEI, and MMM-PDNPs-0.1 using PEG 200-2000 as solutes. PAN/PD displays poor rejection ability with the rejection of 36.5% for PEG 2000, confirming that it remains a ultrafiltration membrane. Compared with PAN/PEI, the MWCO of PAN/PD/PEI is slightly reduced from 695 Da to 641 Da, while MMM-PDNPs-0.1 shows much lower MWCO (380 Da). This comparison demonstrates that embedding PDNPs can reinforce the entire active layer of membrane, while inserting polydopamine interlayer can merely enhance the adhesion of the adjacent layers. This result appears to be surprising, because the interfacial adhesion in MMM-PDNP membranes is truly lower than that inserting polydopamine interlayer. However, the interfacial adhesion is not the only factor that determines solvent resistance, unless the active layer thickness is comparable to the polymer radius of gyration.⁴² It has also been verified that the best performance of adhesion occurs when the cohesive force (within the bulk material) and adhesion force (at the interface) can be well balanced.²⁴ In this study, the average thickness of polyamide layer is found up to 200 nm, and hence the bulk of polyamide layer may also play important roles. That is, if the bulk of polyamide layer is not robust enough, the active layer and the underneath support layer may not swell in a coordinated manner. In this way, it is reasonable to conclude that the reinforcement of the bulk of active layer effectively enhances the swelling resistance of MMM-PDNPs-0.1, although the interfacial adhesive force is not as strong as the PAN/PD/PEI.

Fig. 10 The rejection curves of PAN/PD, PAN/PD/PEI, PAN/PEI, and MMM-PDNPs-0.1 under the transmembrane pressure of 10 bar, and the membranes were equilibrated in isopropanol for 48 h prior to testing.

3.5. Operational stability of the MMM-PDNPs

Long-term operational stability is a key index for assessing the practical application of as-prepared membranes. Allowing for the optimum comprehensive performance of flux and rejection, MMM-PDNPs-0.1 is chosen as representative for the stability evaluation within the operation time of 720 min and the results are depicted in Fig. 11.⁴³ The membrane has been immersed in isopropanol for 48 h for complete equilibration prior to the test. It could be noted that the permeate flux of isopropanol reduces from 22.7 to 17.3 L m⁻² h⁻¹ with a reduction of 23.8% during the initial 480 min. Such phenomenon is likely ascribed to the compaction of membrane, the pore blockage by solute molecules, and the increase in PEG concentration at the feed side, which raises the mass transfer resistance. Certainly, these factors endow the membrane with a distinct increase of rejection from 94.5% to 98.9%. Thereafter, both the permeance and rejection tend to reach the constant values of about 17.1 L m⁻² h⁻¹ and 99.5%, respectively. Similar results have been previously reported in other literatures.^44,45 In spite of the decrease of flux, the membrane exhibits promising operation stability for solvent-resistant nanofiltration or pervaporation, benefiting from its excellent solvent resistance ability.

Fig. 11 Operational stability of MMM-PDNPs-0.1 in isopropanol solution at room temperature under the transmembrane pressure of 10 bar.

4. Conclusions

In summary, a series of ultrathin MMMs with excellent solvent resistance are prepared by incorporating PDNPs into PEI matrix during the interfacial polymerization process. These membranes combine the benefits of PDNPs (nanosize and surface functionalities) and interfacial polymerization method (easy processability). Through the systematical characterizations and measurements, it is found that the presence of PDNPs plays the following two functions: (1) PDNPs reduce the hydrophilicity of membrane surface and enhance the permeate flux of non-polar organic solvents, especially for n-heptane; (2) PDNPs react with PEI and result in covalent bonding between fillers and polymer matrix, which not only effectively inhibits the chain mobility and enhances solvent resistance, but also produces size-dependent selectivity. We also find that embedding PDNPs into PEI layer is a much better method than introducing a polydopamine interlayer between PEI and PAN support, because PDNPs can reinforce the entire active layer of membrane, rather than merely the interfacial region. In addition, the stable structure of the crosslinked PEI-PDNPs layer afforded acceptable long-term operational stability. The method delineated herein may be helpful to understand the functions of hydrophilic–hydrophobic hybrid materials on solvent resistance and transport behaviors and to design composite membrane with high performances for practical applications in both polar and nonpolar solvents.

Acknowledgements

We gratefully acknowledge the financial supports from National Natural Science Foundation of China (21506196, 21476215 and 21576244), China Postdoctoral Science Foundation (2015M570633), and Outstanding Young Talent Research Fund of Zhengzhou University (1521324002).

References

1. W. J. Lau, A. F. Ismail, N. Misdan and M. A. Kassim, Desalination, 2012, 287, 190–199.
2. H. Mao, H. Zhang, Y. Li, Y. Xue, F. Pei, J. Wang and J. Liu, ACS Sustainable Chem. Eng., 2015, 3, 1925–1933.
3. Y. Li, S. Wang, H. Wu, R. Guo, Y. Liu, Z. Jiang, Z. Tian, P. Zhang, X. Cao and B. Wang, ACS Appl. Mater. Interfaces, 2014, 6, 6654–6663.
4. W. C. Chao, S. H. Huang, Q. An, D. J. Liaw, Y. C. Huang, K. R. Lee and J. Y. Lai, Polymer, 2011, 52, 2414–2421.
5. J. Zhao, C. Fang, Y. Zhu, G. He, F. Pan, Z. Jiang, P. Zhang, X. Cao and B. Wang, J. Mater. Chem. A, 2015, 3, 19980–19988.
6. P. Li, Z. Wang, W. Li, Y. Liu, J. Wang and S. Wang, ACS Appl. Mater. Interfaces, 2015, 7, 15481–15493.
7. M. Namvar-Mahboub, M. Pakizeh and S. Davari, J. Membr. Sci., 2014, 459, 22–32.
8. B. B. Tang, T. W. Xu and P. Y. Wu, Prog. Chem., 2007, 19, 1428–1435.
9. Y. Zhang, H. Zhang, Y. Li, H. Mao, G. Yang and J. Wang, Ind. Eng. Chem. Res., 2015, 54, 6175–6186.
10. X. Yu, Z. Wang, J. Zhao, F. Yuan, S. Li, J. Wang and S. Wang, Chin. J. Chem. Eng., 2011, 19, 821–832.
11. M. Namvar-Mahboub and M. Pakizeh, Korean J. Chem. Eng., 2014, 31, 327–337.
12. P. Vandezande, L. E. Gevers and I. F. Vankelecom, Chem. Soc. Rev., 2008, 37, 365–405.
13. J. Zhao, Y. Su, X. He, X. Zhao, Y. Li, R. Zhang and Z. Jiang, J. Membr. Sci., 2014, 465, 41–48.
14. S. Kim, T. Gim and S. M. Kang, Prog. Org. Coat., 2014, 77, 1336–1339.
15. W. Zhang, F. K. Yang, Y. Han, R. Gaikwad, Z. Leonenko and B. Zhao, Biomacromolecules, 2013, 14, 394–405.
16. J. Jiang, L. Zhu, H. Zhang, B. Zhu and Y. Xu, ACS Appl. Mater. Interfaces, 2013, 5, 12895–12904.
17. J. Wu, L. Zhang, Y. Wang, Y. Long, H. Gao, X. Zhang, N. Zhao, Y. Cai and J. Xu, Langmuir, 2011, 27, 13684–13691.
18. F. Pan, H. Jia, S. Qiao, Z. Jiang, J. Wang, B. Wang and Y. Zhong, J. Membr. Sci., 2009, 341, 279–285.
19. B. Li, W. P. Liu, Z. Y. Jiang, X. Dong, B. Y. Wang and Y. R. Zhong, Langmuir, 2009, 25, 7368–7374.
20. H. Lee, S. M. Dellatore, W. M. Miller and P. B. Messersmith, Science, 2007, 318, 426–430.
21. J. H. Waite, Nat. Mater., 2008, 7, 8–9.
22. J. Chen, X. Chen, X. Yin, J. Ma and Z. Y. Jiang, J. Membr. Sci., 2009, 344, 136–143.
23. Y. Li, Y. Su, J. Li, X. Zhao, R. Zhang, X. Fan, J. Zhu, Y. Ma, Y. Liu and Z. Jiang, J. Membr. Sci., 2015, 476, 10–19.
24. J. J. Wilker, Angew. Chem., Int. Ed., 2010, 49, 8076–8078.
25. J. J. Wilker, Curr. Opin. Chem. Biol., 2010, 14, 276–283.
26. M. J. Harrington, A. Masic, N. Holten-Andersen, J. H. Waite and P. Fratzl, Science, 2010, 328, 216–220.
27. W. Liu, Y. Li, X. Meng, G. Liu, S. Hu, F. Pan, H. Wu, Z. Jiang, B. Wang, Z.-X. Li and X.-Z. Cao, J. Mater. Chem. A, 2013, 1, 3713–3723.
28. K. Y. Ju, Y. Lee, S. Lee, S. B. Park and J. K. Lee, Biomacromolecules, 2011, 12, 625–632.
29. S. Hong, Y. S. Na, S. Choi, I. T. Song, W. Y. Kim and H. Lee, Adv. Funct. Mater., 2012, 22, 4711–4717.
30. J. Liebscher, R. Mrówczyński, H. A. Scheidt, C. Filip, N. D. Hădade, R. Turcu, A. Bende and S. Beck, Langmuir, 2013, 29, 10539–10548.
31. Y. Tian, Y. Cao, Y. Wang, W. Yang and J. Feng, Adv. Mater., 2013, 25, 2980–2983.
32. H.-C. Yang, K.-J. Liao, H. Huang, Q.-Y. Wu, L.-S. Wan and Z.-K. Xu, J. Mater. Chem. A, 2014, 2, 10225–10230.
33. D. Wu, Y. Huang, S. Yu, D. Lawless and X. Feng, J. Membr. Sci., 2014, 472, 141–153.
34. M. Peyravi, M. Jahanshahi, A. Rahimpour, A. Javadi and S. Hajavi, Chem. Eng. J., 2014, 241, 155–166.
35. J. Lee, A. Hill and S. Kentish, Sep. Purif. Technol., 2013, 104, 276–283.
36. M. F. Jimenez Solomon, Y. Bhole and A. G. Livingston, J. Membr. Sci., 2013, 434, 193–203.
37. H. Zhang, Y. Zhang, L. Li, S. Zhao, H. Ni, S. Cao and J. Wang, Chem. Eng. Sci., 2014, 106, 157–166.
38. H. Zhang, H. Mao, J. Wang, R. Ding, Z. Du, J. Liu and S. Cao, J. Membr. Sci., 2014, 470, 70–79.
39. C. M. Hansen, Hansen solubility parameters: a users handbook, CRC Press, 2nd edn, 2007.
40. H. J. Zwijnenberg, S. M. Dutczak, M. E. Boerrigter, M. A. Hempenius, M. W. J. Luiten-Olieman, N. E. Benes, M. Wessling and D. Stamatialis, J. Membr. Sci., 2012, 390, 211–217.
41. M. Fang, C. Wu, Z. Yang, T. Wang, Y. Xia and J. Li, J. Membr. Sci., 2015, 474, 103–113.
42. W. Yave, H. Huth, A. Car and C. Schick, Energy Environ. Sci., 2011, 4, 4656–4661.
43. J. P. Sheth, Y. J. Qin, K. K. Sirkar and B. C. Baltzis, J. Membr. Sci., 2003, 211, 251–261.
44. S. S. Luthra, X. J. Yang, L. M. F. dos Santos, L. S. White and A. G. Livingston, J. Membr. Sci., 2002, 201, 65–75.
45. P. Ahmadiannamini, X. Li, W. Goyens, N. Joseph, B. Meesschaert and I. F. J. Vankelecom, J. Membr. Sci., 2012, 394, 98–106.

---