Solvent resistant nanofiltration membranes based on crosslinked polybenzimidazole

Abstract

Solvent resistant nanofiltration membranes based on crosslinked polybenzimidazole (PBI) were designed and fabricated. The modification was carried out by crosslinking PBI asymmetric nanofiltration (NF) membranes with p-xylylene dichloride. The morphology of the membranes was tuned by changing crosslinking time and was investigated in detail by atomic force microscopy (AFM) and scanning electronic microscopy (SEM). The membranes have a typical asymmetric structure with a dense skin layer and a porous sublayer. The roughness of the membranes increases after crosslinking. The results showed that the solvent stability of PBI membranes improved obviously after crosslinking. The modified PBI membranes were applied in solvent resistant nanofiltration (SRNF) processes. Dyes with different molecular weights (ranging from 400 to 1100) and different charges were selected as solutes for filtration experiments. The modified PBI membranes showed very good rejection up to 99.9% for the tested dyes in the selected solvents. Compared to pristine PBI membranes, the modified membranes showed much higher retention of dyes with different molecular weights in different solvents. Therefore, the modified PBI membranes show very promising prospects in solvent resistant nanofiltration processes.

---

Introduction

With the rapid development of various industries, a huge amount of organic solvent is widely used in productive processes. In order to search for a greener and more sustainable way to recycle or purify these solvents, a membrane process becomes an efficient way for separation and purification. Solvent resistant nanofiltration (SRNF) is an emerging technology for separation of compounds with molecular weight cut-offs (MWCOs) between 200 and 1000 Da by simply applying a pressure gradient across a membrane in organic solvents.^1,2 Compared to traditional separation methods, a SRNF membrane process is highly energy saving, and therefore is widely applied in different fields like petrochemistry,³ pharmaceuticals,^4–6 catalysis⁷ and food industries,⁸ which normally feature harsh and corrosive environments. Thus the chemical stability of the membranes is vital to realize their application in SRNF. Currently, two major types of membrane materials have been applied in SRNF. One is based on inorganic membranes e.g. ceramic membranes.^9,10 Ceramic membranes typically show excellent solvent stability, however, suffer from the disadvantages of brittleness and difficulty in further scale up. Another type of SRNF is polymer membranes, which normally exhibit very good mechanical stability and very promising prospect in SRNF application.^11,12 However, the low chemical stability of polymer membranes in a wide range of organic solvents inspires great efforts on the development of new SRNF membranes.

Many efforts have been made to explore SRNF membranes, among which the crosslinked polymeric membranes are the mostly studied systems, such as crosslinked polyimide (PI).^13,14 Generally, SRNF membranes are divided into two types according their structures: the integrally skinned membranes¹⁵ and the thin film composite membranes.¹⁶ The PI SRNF membranes normally showed a typical integrally skinned asymmetric structure, which is normally fabricated by a phase inversion process. The stability of PI membranes in organic solvents can be further improved via cross-linking modification.¹⁷ Even though, the crosslinked PI membranes show very good stability in different solvents, the stability of the formed functional groups like amide groups in diamine crosslinked polyimides after crosslinking is still limited.^17–19 In addition, PI membranes are not stable in acids and base and will be degraded when exposed to high concentrations of organic and inorganic acid/base. Therefore, to explore new SRNF membranes, which can resist to both organic solvents and acidic/basic conditions, is still one of the most important topics for SRNF.^20–22

As a class of thermal plastics, polybenzimidazole (PBI) was commercially developed by the Celanese Corporation in 1983. PBI has received extensive attention due to its outstanding thermal stability and excellent chemical resistance to harsh environments.^23,24 Especially, the excellent stability of PBI membranes under acidic and basic condition was well confirmed by their applications in fuel cells and battery. For example, PBI membranes have successfully applied in strong acidic flow battery and strong basic direct borohydride fuel cells, confirming their excellent stability.^25,26 In addition, PBI membranes have been widely used in osmosis,²⁷ reverse osmosis, forward osmosis, nanofiltration^23,24,28 and fuel cells²⁹ due to their exceptional thermal and chemical stabilities. Recently, PBI membranes have been investigated and successfully applied in SRNF as well by group of Livingston.²⁴ The membranes show very high stability in different kind of solvents especially in apolar solvents after crosslinking with α,α′-dibromo-p-xylene. However, the traditional commercial PBI polymers, especially with high molecular weight, normally show very poor solubility in solvents. Additives (LiCl) and higher pressure are normally needed to dissolve these polymers, which bring high difficulty in making membranes.

In our recent work, we have designed and prepared a new kind of PBI polymer with ether functional groups. The prepared PBI membranes showed very good solubility in apolar solvents like dimethylformamide (DMF), dimethylacetamide (DMAc), N-methyl-2-pyrrolidone (NMP) and their application in SRNF was investigated in detail, showing very promising selectivity. However, the solvent stability of the membranes need to be further improved.³⁰

In this study, we aim to fabricate solvent stable SRNF PBI membranes by chemical modification with p-xylylene dichloride. In this modification, the –NH– groups from heterocyclic imidazole rings can be directly reacted with p-xylylene dichloride. The chemical structure of modified PBI was confirmed by FTIR. The morphology of the membrane before and after modification was investigated by AFM and SEM. The performance of the membranes in actual SRNF-applications together with the relation between membrane morphology and performance were investigated in detail.

Experimental

Materials and methods

3,3′-Diaminobenzidine (DABz) was purchased from Acros Organics and used as received. 4,4′-Dicarboxydiphenyl ether (DCDPE) was obtained from Peakchem (Shanghai) and dried prior to be used. Methanesulfonic acid (MSA) and phosphorus pentoxide were used commercially. p-Xylyene dichloride was purchased from Aladdin Reagent Company (Shanghai, China). Rose bengal, crystal violet and bromothymol blue were used as solutes. All solvents such as N,N-dimethylacetamide, tetrahydrofuran, 2-propanol (IPA), ethyl acetate (EA), dimethyl sulphoxide (DMSO) and ethanol with analytical grade were purchased commercially and used as received. All the above reagents were supplied with analytical grade.

The PBI polymer with ether groups (Scheme 1) was synthesized as reported previously.²⁵ PBI was synthesized via condensation polymerization of DCDPE and DABz by using MSA as solvent.

Scheme 1 The chemical structure of synthesized PBI.

Fabrication of the pristine PBI nanofiltration membranes

The membranes were prepared via the phase inversion method from casting solutions with predetermined amounts of PBI, DMAc and THF. The weight concentration of PBI is 15%, while the weight ratio of DMAc/THF was kept at 75/15. The solution was directly cast on a glass plate by using an automatic film applicator (Elcometer), after which the nascent polymer film with a wet thickness of 350 μm was immersed in de-ionized water. Then the membranes were stored in water for use. Before immersing in the water bath, the cast film was evaporated for 0 s, 30 s, 60 s and the resulted membranes were labelled as M1, M2, M3 respectively.

Modification of PBI nanofiltration membranes

First, a 2.0 wt% homogeneous solution of crosslinking reagents was prepared by dissolving p-xylylene dichloride in ethanol.²⁸ Before crosslinking, pristine PBI membranes (M1, M2 and M3) were washed with fresh ethanol to replace the contained water in membranes. Then, the membranes were immersed in a solution of p-xylyene dichloride for 12 h at ambient temperature. Scheme 2 exhibits the chemical modification reaction. Finally, the modified membranes were taken out from p-xylylene dichloride solution and then washed with fresh ethanol to remove the residual p-xylylene dichloride, and stored in water for test. The modified PBI membranes from M1, M2, M3 were labelled as M1′, M2′, M3′ respectively.

Scheme 2 Chemical modification of PBI by p-xylylene dichloride.

Characterization

SEM (JEOL JCM-6000) was applied to observe the cross-section and surface morphology of the membranes. The cross-sections of membranes were obtained by breaking membranes in liquid nitrogen. Samples were coated with gold before SEM analysis.

The surface morphology and the average surface roughness (R_a) of the PBI membranes were investigated by AFM operating in a tapping mode with a Nanoscope V controller (Veeco/Digital Instruments, Santa Barbara, USA). The average roughness (R_a) is expressed as eqn (1):

where Z_i is the current surface height, Z_avg is the average surface height within the given area, and N is the number of data points in the given area.

The chemical structure of prepared polymer was characterized by ¹H-NMR (BRUKER DRX400 spectrometer) with DMSO-d₆ as a solvent and tetramethylsilane (TMS) as internal standard and Fourier Transform Infrared Spectroscopy (FTIR) (BRUKE TENSOR 27) in attenuated total reflectance mode (FTIR-ATR).

Solvent stability

The solvent stability of the membranes was evaluated by immersing the membranes in various solvents at room temperature for at least 72 h. The solvent stability of the pristine and modified PBI membranes was determined by visual inspection of the dimension change of the membranes.

Filtration experiments

Nanofiltration experiments were carried out in a stainless steel dead-end pressure cell with the membrane area of 19.0 cm². The feed solutions of 35 μM solutes in solvents were poured in the cell and the pressure was kept at 25 bar. The solvent flux test was carried out on four different solvents e.g. ethanol, 2-propanol, ethyl acetate and tetrahydrofuran. The properties of the selected solvents are listed in Table 1. Standard feed solutions were prepared by dissolving the three different solutes (rose bengal, crystal violet, bromothymol blue) in the selected solvents. The concentration of solutes in the permeate (C_p) and feed samples (C_f) was analysed by using a double-beam UV-vis spectrophotometer (TU-1901). The properties of the solutes used in the nanofiltration experiments are listed in Table 2. All the permeate samples were taken after two hours' filtration, when the steady state was reached. All nanofiltration experiments were based on at least three membrane samples, which were individually measured and the average value was taken. The average deviation of solvent fluxes is less than 5%.

Table 1 Physical properties of the solvents used

Solvent	Molecular weight [g mol^−1]	Molar volume [m^3 mol^−1]	Viscosity [mPa s]	Molar volume/viscosity	Surface tension [mN m^−1]	Solubility parameter [(MPa)^1/2]
Ethyl acetate	88.1	98.2	0.426	230.56	23.24	18.2
THF	72.0	82.1	0.480	171.0	31.10	19.4
Ethanol	46.0	58.2	1.10	52.90	21.99	26.5
2-Propanol	60.0	76.9	2.00	38.50	21.01	23.7

---

Table 2 Properties of solutes used for the filtration experiments

Component (abbreviation)	Structure	Charge (pKa)	Molar volume (cm^3 mol^−1)	Molar weight (g mol^−1)
Bromothymol blue (BTB)		0/7.2	281	624.39
Crystal violet (CV)		+/1.8	231	407.99
Rose bengal (RB)		−/4.5	272.8	1017

---

The flux J was calculated by using the following equation (eqn (2)), where V is the total volume of permeated solvent, A is the effective filtration area, Δt is the flow time across the membrane.

The effective solute rejection was calculated by the following equation (eqn (3)):

where C_p is the concentration of permeate solution and C_f is the concentration of the original feed solution.

Results and discussion

Chemical structure of prepared membranes

The chemical structure of synthesized polymers was detected by ¹H-NMR and the corresponding spectrum is shown in Fig. 1s in ESI,† which identified the expected structure. FTIR was carried out on the membranes before and after crosslinking to confirm the structure of modified membrane. The ATR-FTIR spectra of the pristine PBI membrane (M1) and p-xylylene dichloride modified PBI membrane (M1′) are shown in Fig. 1. The strong absorption bands ranging from 2500 to 3600 cm⁻¹ are attributed to the N–H group of imidazole ring in the pristine PBI and the modified PBI membranes. The strong peak at 3410 cm⁻¹ corresponds to the stretching vibration of the isolated non-hydrogen bonded N–H group, while the peak at 3194 cm⁻¹ can be attributed to the stretching vibration of the involved hydrogen bonded N–H group in the pristine PBI membrane. However, the absorption at 3194 cm⁻¹, which corresponds to non-hydrogen bonded N–H group, disappeared after modification and a new peak located at around 1464 cm⁻¹ appeared. The bands at 2920, 2850 and 1464 cm⁻¹ in the spectrum of modified PBI membrane are attributed to –CH₂– stretching. The results confirmed the successful crosslinking reaction between PBI and p-xylylene dichloride. The bond at 1625 cm⁻¹ is caused by the C=C and C=N stretching vibrations of the five-member heterocyclic ring in the pristine PBI and modified PBI membranes.

Fig. 1 FTIR spectra of the pristine PBI membrane and the modified PBI membrane.

Morphology of the PBI membranes

The cross section morphologies of pristine and modified PBI membranes were presented in Fig. 2. All the membranes exhibit a typical asymmetric structure, which includes a selective skin layer and a porous support layer. The morphology of PBI membrane was tuned via adjusting evaporation time of volatile co-solvent. Traditionally, the evaporation of volatile THF before immersing the cast film in the coagulation bath will lead to a localized increase in polymer concentration. Hence, a thin resistance barrier between the membrane and the non-solvent will be formed, which makes the diffusion between solvent and non-solvent much slower and results in a delayed demixing and a more compact skin layer.^31,32 After cross-linking with p-xylylene dichloride, no visible change in the membrane overall cross section was found by SEM.

Fig. 2 Cross section morphology of pristine and modified PBI membranes.

The magnified skin layer of pristine and modified PBI membranes was shown in Fig. 3. The pore size of skin layers decreases from M1 to M3, which further confirms above conclusion. After modification, the skin layer becomes more compact as well, due to the formed cross linking networks between PBI chains. It is expected that higher selectivity and lower permeability will be obtained after cross linking.

Fig. 3 SEM image of the magnified skin layer of pristine and modified PBI membranes.

Surface morphology of membranes was recognized by AFM. Fig. 4 shows the height images of pristine and modified PBI membranes. The pristine PBI membranes show relatively smooth surface with spherical nanoparticles. With increasing evaporation time of THF, the particle size decreases, however, their surface roughness is at the similar level. With THF evaporating, a more concentrated membrane surface with lower viscosity and less mobility will be formed. Upon immersing in the coagulation bath, less perturbation thus occurs, so the particle size decreases with increasing THF content. However, after modification, the surface becomes much rougher compared to pristine membranes and the surface becomes very irregular, which is possibly due to introduction of crosslinked layer, inducing more uneven surface. For example, the average roughness of M1 (R_a) increased from 2.866 nm to 6.145 nm after modification (M1′), while the roughness of M3 increased from 2.846 nm to 5.013 nm after crosslinking. The results indicated that surface of PBI was covered with crosslinked networks after modification and induced rougher surface.

Fig. 4 The AFM height images of prepared membranes.

Solvent stability

To investigate the stability of the PBI membranes in different solvents, the pristine (M3) and modified PBI membranes (M3′) were immersed in various solvents for 72 hours. Table 3 gives an observation of the solvent stability of the pristine and modified PBI membranes. All membranes were stable in ethanol, 2-propanol, ethyl acetate and acetone. The pristine PBI membrane (M3) dissolved in most of the other solvents like DMAc, NMP and DMSO and swelled in THF. However, the modified PBI membranes (M3′) were stable in most solvents, despite that they are swollen with certain degree in DMAc, NMP and DMSO. Compared with pristine PBI membrane, the modified PBI membranes were more stable in most organic solvents. The results can further prove the success of crosslinking reaction between PBI and p-xylylene dichloride.

Table 3 Solvent stability of the pristine PBI membrane (M3) and modified PBI membrane (M3′)

Membrane	Ethanol	IPA	EA	Acetone	THF	DMAc	NMP
a (+): stable.b (Sw): swelling.c (−): dissolving.d Residual weight after immersing.
M3	^a+	+	+	+	^bSw	^c−	−
	^d98.8%	98.9%	98.2%	98.5%	97.9%	−	−
M3′	+	+	+	+	+	Sw	Sw
	99.2%	98.7%	98.5%	99.5%	98.3%	85.2%	82.3%

---

Pure solvent flux

Fig. 5 shows the flux of EA, THF, ethanol and IPA through the pristine and modified PBI membranes. Among all the membranes, M3 and M3′, which prepared with the longest evaporation time of THF (60 s), show the lowest flux on all selected solvents. With longer evaporation time of THF, the membrane fluxes decrease. As discussed early, the longer evaporation time of THF leads to an increased polymer concentration on the surface of the cast polymer film prior to immersion in the coagulation bath and will induce a more compact skin layer and further decreased fluxes.

Fig. 5 Organic solvent fluxes of all the membranes.

In addition, all the modified membranes exhibited obviously lower solvent fluxes than those of pristine PBI membranes for all the selected solvents, due to the fact that the pore size of the skin layer in modified PBI membranes is smaller than the pristine ones. In addition, the formed crosslinking networks limited the mobility of polymer chain, increased the solvent stability and further resulted in the lower fluxes as well.

Moreover, the solvent fluxes of all the tested membranes follow the order of EA > THF > ethanol > IPA. The transport mechanism of SRNF process is quite complicate.³³ The solvent fluxes of SRNF membranes were found to be influenced by many factors, such as the solvent–membrane interactions, the membrane morphology and the physical properties of the solvent.³⁴

The solvent viscosity and the ratio of molar volume to solvent viscosity have quite high impact on solvent transportation as well.²³ Fig. 6a showed the correlation between flux and the solvent viscosity, where the flux of the solvent decreases with increasing viscosity. For example, the fluxes of tested solvents are in the order of EA > THF > ethanol > IPA, while, their viscosities are reversed. Inspection of the relation between solvents fluxes and the ratio of the molar volume to solvent viscosity is shown in Fig. 6b. Compared to pristine PBI membranes, the modified membranes show higher correlation with solvent viscosity and MV.

Fig. 6 Solvent flux though pristine membrane (M3) and modified PBI membrane (M3′) as function of viscosity (a) and MV/viscosity (b).

Obviously, the flux for EA, which has the highest ratio of the molar volume to viscosity, is the highest. On the contrary, the flux for IPAaaa, which possesses the lowest ratio of the molar volume to viscosity, is the lowest. The results indicate that the solvent viscosity and the molar volume are the two of the most important factors that affect the solvent flux through the pristine and modified PBI membranes.

Filtration performance

To investigate the filtration performance of prepared membranes, dyes with different size and different charges were selected as solutes for further experiment. To optimize the crosslinking parameters, the PBI membranes were firstly modified with different time (0, 2, 4, 12 and 24 hours) and their filtration performance on neutral BTB in IPA and ethanol was shown in Fig. 7. With increasing of reaction time, an increase of rejection and a decrease of solvent flux can be obtained during the first 12 hours, afterward, the rejection and flux were kept constant, which well confirmed the successful chemical crosslinking reactions between PBI and p-xylylene dichloride over time. And the reaction equilibrium is reached after 12 hours. Therefore, in the following part, all the performance of crosslinked PBI membranes is based on 12 hours' cross linking. With increasing reaction time from 0 to 24 h, the rejection of the M1 increased from 56.8% to 97.6%, while the solvent flux decreased from 19.6 to 8.45 L m⁻² h⁻¹ (ethanol). After 12 hours' crosslinking, the rejection for BTB of all the membranes exceeded 94%. When keeping the crosslinking time at 12 h, the solvent fluxes of M3 series are obviously lower than those of M1, due to the longer evaporation time of THF during membrane fabrication. These results are in good agreement with the morphology changes. As mentioned above, longer evaporation time can induce more compact skin layer, smoother surface and further lead to lower fluxes and higher rejections. After modification, the formed crosslinking networks could make membrane pores smaller, which will induce higher rejections as well.³⁵ Compared to ethanol, the membranes show lower solvent fluxes and retention on the solutes, when using IPA as solvent. This is possibly due to the different physical properties of solvents. The higher viscosity of IPA possibly increases the force of friction between solvent and membrane and further leads to the harder transport of solvent through the membrane. Apart from the viscosity, the weaker interaction between the IPA and membranes than that of ethanol induced by their different polarities would cause the different fluxes as well. The interaction is based on functional groups of –OH in alcohols and –NH– in PBI. Due to relatively higher acidity of ethanol than IPA, where the interaction between –OH of ethanol with PBI is stronger than does of IPA, which possibly induce the flux difference as well.

Fig. 7 The filtration data of BTB with different crosslinking time in ethanol (a, a′) and IPA (b, b′).

Tables 4, 5 and 6 indicate the filtration performance of different solutes in the selected solvents for the pristine and the modified PBI membranes. The chemical structure, charge, molar volume and molecular weight of the selected solutes are summarized in Table 2. Among all the solutes, RB has the highest MW, while carries negative charge, CV has lowest MW with positive charge. And BTB is neutral. As expected, with increasing evaporation time of THF, the fluxes of all the solvents decreased, while the rejection of all the solutes increased (Table 6). This result can well agree with the membrane morphology changes. Compared to the pristine PBI membranes, the modified PBI membranes exhibited entirely much higher rejection and a bit lower fluxes. For example, the rejection for RB of pristine PBI membranes (M1 to M3) increased from less than 70% to more than 96%, when applying THF, ethanol, IPA and EA as solvents. This can be explained by the morphology changes after modification. The denser structure formed by crosslinking reaction between PBI and p-xylylene dichloride limited the mobility of polymer chain and further resulted in the higher rejection. The modified PBI membranes showed very high rejection on all the tested dyes in different solvents (in the range from 93.9% to 99.9%), except on CV in ethanol solutions, showing very promising performance in SRNF. The low rejection on CV in ethanol is possibly due to the smaller molar volume and the strong interaction between CV and ethanol. As we mentioned, the transport mechanism of solutes through SRNF is a very complicate process, different factors can lead to the different transport behaviour, including the interaction between solvent and solutes, between membranes and solvent, membrane and solutes.^33,36 In this system, all the membranes showed the lowest rejection on CV, when applying ethanol as solvent. The reason may be derived from the different solvent–solute interactions, the strong interaction between CV and ethanol possibly induced the low rejection and high fluxes, however, the pristine and modified PBI membranes show very high retention on CV in other related solvents like THF, IPA and EA, even their molecular weight and molar volume are the lowest among all the dyes. This may be attributed to the charge properties of the solutes and the different physical properties of solvents. Compare with ethanol, THF and EA have lower viscosity and weaker interaction with CV, therefore induces higher fluxes and higher retention. For IPA, as previously reported, the imidazole group within PBI unit has amphoteric property, which illustrates negative charge at high pH and positive charge at low pH.³⁷ IPA can possibly offer very minor protons to offer the positively charged surface for prepared pristine and modified PBI membranes. Combining the effect of Donnan repulsion between the positively charged CV and the interaction between IPA and CV, the membranes showed more than 99% retention on CV in IPA solution. On the other hand, the pristine PBI membranes showed much lower rejection on the negatively charged RB and neutral BTB than that of CV when applying IPA as solvents. However, after crosslinking, the role of charge becomes vague, all the membranes showed similar and very high retention (95%) on all the solutes in different solvents. Compare with reported crosslinked PBI membranes, the membranes showed comparable even higher selectivity.²⁴

Table 4 Filtration data of the pristine and modified PBI membranes on BTB in the selected solvents

Membrane	THF		Ethanol		IPA		EA
	^aR	^bJ	R	J	R	J	R	J
a Rejection (%).b J: flux (L m^−2 h^−1).
M1	68.4	8.56	56.8	19.6	56.3	4.97	97.1	8.31
M2	83.3	11.2	60.9	18.2	63.5	3.41	99.3	11.1
M3	87.2	11.2	69.0	18.4	64.9	2.65	99.7	10.8
M1′	93.9	6.10	97.6	8.45	94.4	1.96	98.6	5.64
M2′	94.5	8.02	95.5	7.83	97.2	1.91	99.5	10.5
M3′	96.5	5.69	98.9	5.30	99.0	1.24	99.6	6.25

---

Table 5 Filtration data of the pristine and modified PBI membranes on CV in the selected solvents

Membrane	THF		Ethanol		IPA		EA
	^aR	^bJ	R	J	R	J	R	J
a Rejection (%).b J: flux (L m^−2 h^−1).
M1	97.5	17.0	22.3	29.6	99.9	2.65	98.0	21.2
M2	98.7	9.14	23.4	14.6	99.1	2.45	99.0	10.9
M3	98.2	14.5	25.4	11.3	99.6	1.61	98.4	17.3
M1′	97.2	8.30	62.8	16.8	99.9	2.28	99.3	10.5
M2′	95.3	14.2	60.6	7.89	99.9	1.00	98.5	17.0
M3′	98.2	9.19	75.7	7.42	99.9	0.835	98.4	10.2

---

Table 6 Filtration data of the pristine and modified PBI membranes on RB in selected solvents

Membrane	THF		Ethanol		IPA		EA
	^aR	^bJ	R	J	R	J	R	J
a Rejection (%).b J: flux (L m^−2 h^−1).
M1	49.7	74.6	60.9	22.1	43.3	5.53	98.8	23.2
M2	62.9	56.2	63.0	14.6	51.6	3.53	99.2	10.0
M3	68.8	57.2	70.2	12.5	59.9	2.53	99.0	11.5
M1′	97.5	37.2	96.1	10.6	98.2	1.45	98.8	9.43
M2′	97.9	35.2	97.0	6.03	98.6	0.873	99.2	9.60
M3′	97.5	37.6	97.2	5.37	99.1	0.776	99.2	8.41

---

Conclusions

In this work, the modified PBI polymer membranes were designed and fabricated for solvent resistant nanofiltration application. The integrally skinned asymmetric PBI nanofiltration membranes were prepared and chemically modified by p-xylylene dichloride and the morphology of PBI membranes was tuned via introducing volatile solvent. SEM results showed that a more compact skin layer will be formed with longer evaporation time of THF. The modified PBI membranes were investigated in SRNF by changing solutes and solvents. The results showed that the modified PBI membranes showed much higher retention on solutes in different solvents than the pristine PBI membranes due to the formed crosslinking networks after modification. The solvent stability of PBI membranes was dramatically increased after modification. The modified PBI membranes showed retention of more than 95% on different solutes (MW ranging from 400 to 1017) in different solvents, showing very promising prospect in SRNF.

Acknowledgements

The authors greatly acknowledge the financial support from China Natural Science Foundation (No. 21206158, 21476224) and the Outstanding Young Scientist Foundation, CAS, Dalian Municipal Outstanding Young Talent Foundation (2014J11JH131).

Notes and references

1. P. Vandezande, L. E. Gevers and I. F. Vankelecom, Chem. Soc. Rev., 2008, 37, 365–405.
2. P. Marchetti, M. F. Jimenez Solomon, G. Szekely and A. G. Livingston, Chem. Rev., 2014, 114, 10735–10806.
3. L. S. White, J. Membr. Sci., 2006, 286, 26–35.
4. L. Peeva, J. D. S. Burgal, I. Valtcheva and A. G. Livingston, Chem. Eng. Sci., 2014, 116, 183–194.
5. R. Abejón, A. Garea and A. Irabien, AIChE J., 2014, 60, 931–948.
6. M. G. Buonomenna and J. Bae, Sep. Purif. Rev., 2014, 44, 157–182.
7. M. Janssen, C. Müller and D. Vogt, Green Chem., 2011, 13, 2247.
8. K. Hendrix, S. Vandoorne, G. Koeckelberghs and I. F. J. Vankelecom, Polymer, 2014, 55, 1307–1316.
9. S. Zeidler, P. Puhlfürß, U. Kätzel and I. Voigt, J. Membr. Sci., 2014, 470, 421–430.
10. S. Rezaei Hosseinabadi, K. Wyns, V. Meynen, R. Carleer, P. Adriaensens, A. Buekenhoudt and B. van der Bruggen, J. Membr. Sci., 2014, 454, 496–504.
11. I. Strużyńska-Piron, M. R. Bilad, J. Loccufier, L. Vanmaele and I. F. J. Vankelecom, J. Membr. Sci., 2014, 462, 17–27.
12. L. Liu, X. Wang, Y. Wang, L. Li, K. Pan, J. Yang and B. Cao, Mater. Lett., 2014, 132, 11–14.
13. H. Siddique, Y. Bhole, L. G. Peeva and A. G. Livingston, J. Membr. Sci., 2014, 465, 138–150.
14. K. Hendrix, K. Vanherck and I. F. J. Vankelecom, J. Membr. Sci., 2012, 421–422, 15–24.
15. A. K. Hołda, M. de Roeck, K. Hendrix and I. F. J. Vankelecom, J. Membr. Sci., 2013, 446, 113–120.
16. M. Namvar-Mahboub and M. Pakizeh, Korean J. Chem. Eng., 2013, 31, 327–337.
17. K. Vanherck, P. Vandezande, S. O. Aldea and I. F. J. Vankelecom, J. Membr. Sci., 2008, 320, 468–476.
18. K. Vanherck, G. Koeckelberghs and I. F. J. Vankelecom, Prog. Polym. Sci., 2013, 38, 874–896.
19. K. Vanherck, A. Cano-Odena, G. Koeckelberghs, T. Dedroog and I. Vankelecom, J. Membr. Sci., 2010, 353, 135–143.
20. L. Shao, X. Cheng, Z. Wang, J. Ma and Z. Guo, J. Membr. Sci., 2014, 452, 82–89.
21. I. B. Valtcheva, P. Marchetti and A. G. Livingston, J. Membr. Sci., 2015, 493, 568–579.
22. Y. C. Xu, X. Q. Cheng, J. Long and L. Shao, J. Membr. Sci., 2016, 497, 77–89.
23. D. Y. Xing, S. Y. Chan and T.-S. Chung, Green Chem., 2014, 16, 1383.
24. I. B. Valtcheva, S. C. Kumbharkar, J. F. Kim, Y. Bhole and A. G. Livingston, J. Membr. Sci., 2014, 457, 62–72.
25. H. Dai, H. Zhang, H. Zhong, H. Jin, X. Li, S. Xiao and Z. Mai, Fuel Cells, 2010, 10, 754–761.
26. D. Chen, S. Yu, X. Liu and X. Li, J. Power Sources, 2015, 282, 323–327.
27. M. F. Flanagan and I. C. Escobar, J. Membr. Sci., 2013, 434, 85–92.
28. K. Y. Wang, Q. Yang, T.-S. Chung and R. Rajagopalan, Chem. Eng. Sci., 2009, 64, 1577–1584.
29. F. Mack, S. Heissler, R. Laukenmann and R. Zeis, J. Power Sources, 2014, 270, 627–633.
30. D. Chen, S. Yu, H. Zhang and X. Li, Sep. Purif. Technol., 2015, 142, 299–306.
31. K. Hendrix, G. Koeckelberghs and I. F. J. Vankelecom, J. Membr. Sci., 2014, 452, 241–252.
32. A. K. Hołda, B. Aernouts, W. Saeys and I. F. J. Vankelecom, J. Membr. Sci., 2013, 442, 196–205.
33. S. Postel, G. Spalding, M. Chirnside and M. Wessling, J. Membr. Sci., 2013, 447, 57–65.
34. A. Dobrak-van Berlo, I. F. J. Vankelecom and B. van der Bruggen, J. Membr. Sci., 2011, 374, 138–149.
35. K. Y. Wang, Y. Xiao and T.-S. Chung, Chem. Eng. Sci., 2006, 61, 5807–5817.
36. S. Zeidler, U. Kätzel and P. Kreis, J. Membr. Sci., 2013, 429, 295–303.
37. J. Lv, K. Y. Wang and T.-S. Chung, J. Membr. Sci., 2008, 310, 557–566.

---

Footnote

† Electronic supplementary information (ESI) available: H-NMR spectrum of the synthesized PBI. See DOI: 10.1039/c5ra27044h

---