Beyond polyimide_ Crosslinked polybenzimidazole ...

Journal of Membrane Science 457 (2014) 62–72

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Beyond polyimide: Crosslinked polybenzimidazole membranes for organic solvent nanofiltration (OSN) in harsh environments Irina B. Valtcheva, Santosh C. Kumbharkar, Jeong F. Kim, Yogesh Bhole, Andrew G. Livingston n Department of Chemical Engineering, Imperial College, Exhibition Road, South Kensington Campus, London SW7 2AZ, UK

art ic l e i nf o

a b s t r a c t

Article history: Received 3 November 2013 Accepted 27 December 2013 Available online 18 January 2014

In this work, we report a new class of organic solvent nanofiltration (OSN) membranes fabricated from polybenzimidazole (PBI) which exhibit superior chemical stability compared to other well-known polymeric membranes such as polyimide. Integrally skinned asymmetric PBI membranes were prepared and crosslinked using either an aliphatic or an aromatic bifunctional crosslinker. Three batches of membranes with the same composition were prepared and crosslinked with each crosslinker. The membrane performance showed excellent reproducibility in organic solvents and water in terms of flux and retention profile. Also, both membranes showed good tolerance towards extreme pH conditions. To critically assess their chemical stability the membranes were tested in realistic chemical process conditions that employ different types of acids and bases, e.g. concentrated dichloroacetic acid in acetonitrile, piperidine in N,N-dimethylformamide (DMF) and monoethanolamine in water. The membranes modified with aliphatic crosslinker could not retain their properties when DMF was used as the organic solvent. This was found to be due to dissolution of PBI in DMF rather than degradation due to pH exposure. On the other hand, it was shown that the membrane modified with the aromatic crosslinker exhibits superior stability and higher permeances in comparison to the membrane crosslinked with the aliphatic crosslinker. The results obtained show that membranes fabricated from crosslinked PBI were stable and have the potential for applications in chemically harsh conditions found in processes ranging from pharmaceutical to petrochemical industries. & 2014 Elsevier B.V. All rights reserved.

Keywords: Organic solvent nanofiltration (OSN) Polybenzimidazole (PBI) Crosslinking Extreme pH Acid/base filtration

1. Introduction Separations of solutes in organic solvents are widely employed in pharmaceutical and chemical industries. Typical techniques used for fractional separation are distillation, adsorption and flash chromatography. These processes usually involve the utilisation of fresh solvents and energy. In particular, flash chromatography is relatively cumbersome to scale-up due to its aspect ratio limitation [1]. These operations result in increased cost of goods, and raise environmental concerns. As a result, cost effective alternatives are desired in order to minimise the plant footprint, waste production and energy consumption. Organic solvent nanofiltration (OSN) is a membrane-based separation process which can provide the necessary molecular discrimination [2], and has already shown great potential in terms of savings [1,3] and applicability [4–6] to industrial processes. Nanofiltration (NF) is already a state-of-the-art process for water purification and treatment [7,8]. However, the available NF and reverse osmosis (RO) membranes are in most cases not stable in harsh and corrosive environments typically required for OSN [2]. The solvent and pH resistance of the membrane is one of the main challenges for implementing OSN in relevant processes. Inorganic materials are ideal for such conditions since they

n

Corresponding author. Tel.: þ 44 20 7594 5582; fax: þ 44 20 7594 5639. E-mail address: [email protected] (A.G. Livingston).

0376-7388/$ - see front matter & 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.memsci.2013.12.069

do not easily dissolve or deform in organic solvents. On the other hand, they are difficult to fabricate, handle and are more expensive as compared to organic membranes [2,9]. For this reason the majority of reported OSN membranes are made out of polymeric materials. The most exhaustively studied polymer for application in OSN is polyimide (PI) [2,10,11]. The usual solvents in which PI dissolves are polar aprotic ones, such as N,N-dimethylformamide (DMF) and dimethyl sulfoxide (DMSO) and these are typically used in membrane fabrication. For a PI membrane to be stable in applications using these solvents, it needs to be chemically crosslinked with diamines [10,11]. PI is stable to some extent in weak acids and low concentrations of bases but is not recommended for use in inorganic acids and degrades when exposed to high concentrations of organic and inorganic bases [12]. Apart from solvent stability, tailored membrane performance is a desirable feature, e.g. varying the molecular weight cut-off (MWCO)1 of PI membranes [13]. A few polymeric membranes have been reported in literature which can withstand either acids or bases in aqueous feed streams [14–16]. However, these membranes are based on sulfonated polyether ether ketone (SPEEK) or polysulfone (PSf) which will not withstand many organic solvents. Hence, there is a need for

1 Molecular weight cut-off (MWCO) is defined as the lowest molecular weight of a solute rejected at 90% by the membrane.

I.B. Valtcheva et al. / Journal of Membrane Science 457 (2014) 62–72

alternative membrane materials which can withstand both organic solvents and acidic/basic conditions. Surface crosslinked chitosan/ polyacrylonitrile composite NF membranes have been shown to maintain their stability under basic pH and in several important organic solvents [17]. However, the acid/base resistance was only demonstrated in aqueous media. Polybenzimidazole (PBI, Fig. 1) has been studied extensively for reverse osmosis [18–21]. More recently PBI has gained much attention for applications in gas separation [22], aqueous NF [23], forward osmosis [24] and fuel cells [25,26] due to its outstanding properties (thermal, mechanical and chemical stability in corrosive environments) [27]. In addition, PBI has the advantage of possessing excellent stability towards acids and bases [12]. PBI dissolves in polar aprotic solvents, such as N,N-dimethylacetamide (DMAc), N-methylpyrrolidone (NMP) and DMSO, from which dope solutions can be prepared and cast. Uncrosslinked PBI hollow fibre membranes were shown to be effective in the removal of chromate from alkaline wastewater [28]. PBI flat sheet membranes crosslinked with α,α0 -dichloro-p-xylene were reported for the concentration and separation of cephalexin from aqueous solutions over a pH range from 2 to 10 [23]. Similar to polyimide membranes, unmodified PBI membranes cannot be used for polar aprotic solvents. Several chemical modification procedures have been reported in literature including treatment of PBI membranes with α,α0 -dichloro-p-xylene [23], divinyl sulphone [29] and 1,4-dibromobutane [30]. In this study, we report on the fabrication and performance of crosslinked PBI NF membranes for applications in organic solvents containing acids or bases. The conditions of two pharmaceutical and one chemical process have been chosen to demonstrate the solvent and acid/base resistance of the prepared membranes. Also, the batchto-batch reproducibility of crosslinked PBI membranes has been evaluated. The membranes were characterised with FTIR, contact angle measurement and SEM.

2. Experimental 2.1. Materials Celazoles S26 polybenzimidazole (PBI, MW ¼27,000 g mol 1) solution was purchased from PBI Performance Products Inc. (USA). The solution contains 26 wt% polymer solids and 1.5 wt% lithium chloride (stabiliser) dissolved in DMAc. Non-woven polypropylene

63

fabric Novatexx 2471 was from Freudenberg Filtration Technologies (Germany). All solvents such as DMAc, propan-2-ol (IPA), acetonitrile (MeCN), and DMF were HPLC grade and were used as received from VWR (UK). The chemicals for crosslinking were α,α0 dibromo-p-xylene (DBX) and 1,4-dibromobutane (DBB) from VWR (UK) and Sigma Aldrich (UK), respectively. Polyethylene glycol (PEG) of three different molecular weights was purchased from VWR (UK) and Sigma Aldrich (UK). Dichloroacetic acid (DCA) and piperidine (PIP) were both from Sigma Aldrich (UK); monoethanolamine (MEA) was from VWR (UK). De-ionised water was produced by passing tap water through an RO filtration unit.

2.2. Fabrication of integrally skinned (IS) asymmetric PBI membranes and chemical crosslinking Celazoles S26 was diluted with DMAc to 17 wt% polymer concentration and stirred continuously at 2170.5 1C until a homogeneous dope solution was obtained. The polymer solution was then left overnight to remove air bubbles. Membranes were cast on polypropylene non-woven using a bench top laboratory casting machine with an adjustable knife set at 250 mm (Elcometer, UK). Following this, the membranes were immersed in a water precipitation bath at 2271 1C for 24 h, and subsequently washed with IPA to remove residual solvent and water. The viscosity of the dope solutions was measured using a rotary viscometer Model 2020 from Cannon Instrument Company (USA) with a spindle size 16 suitable for high viscosities. All viscosities were recorded at 10 rpm spindle speed and at 21 1C. Membranes were crosslinked by immersion in a solution containing 3 wt% DBX or 10 wt% DBB in acetonitrile (Fig. 1). The reaction was carried out at 80 1C for 24 h under constant stirring and reflux. After crosslinking, the membranes were first immersed in IPA to remove residual reagents and later on impregnated with PEG400 by immersion in a PEG400/IPA (1:1) solution for 4 h to preserve the pore structure and allow dry storage. After obtaining the final membrane, the thickness was recorded using a digital micrometre purchased from Mitutoyo (UK). The stated membrane thicknesses include the impregnated polymer film and the polypropylene non-woven and are the mean between six measured points in different parts of the membrane sheet.

2.3. Selection of model solutions

Fig. 1. (a) Chemical crosslinking mechanism of 2,2-(m-phenylene)-5,5-bibenzimidazole (PBI) with a) α,α0 -dibromo-p-xylene (DBX) and (b) 1,4-dibromobutane (DBB).

Typically, a range of polystyrene (PS) [31] or polyethylene glycol (PEG) oligomers [32] is used to determine the rejection performance of OSN membranes. The main advantage of both methods is that the rejection of a range of homologous solutes can be analysed simultaneously, and an MWCO curve can be obtained for each membrane. Also, both types of oligomers are soluble in a wide variety of organic solvents. However, PS solutes are more expensive and not soluble in water due to their hydrophobic character. Since the application of crosslinked PBI membranes is intended for but not limited to iterative synthesis of therapeutic drugs [5,33,34], which closely resemble PEG compounds, PEGs in different MW were chosen as the marker solutes in the current study. Three solvent systems, tested in this work, were chosen based on three commercial processes which employ acidic or basic conditions in the production. The first one mirrors one of the reaction conditions used in oligonucleotide synthesis, which is 3 wt% DCA in MeCN [35]. The second system replicates one of the reaction conditions used in peptide synthesis – 20 wt% piperidine in DMF [36]. The third filtration solution – 20 wt% MEA in DI water – is representative of liquids used in carbon capture and storage (CCS) [37].

64


2.4. Membrane characterisation 2.4.1. Fourier transform-infrared spectroscopy (FT-IR) Infrared spectra were recorded on a Perkin Elmer-Spectrum 100. The samples were fixed on a zinc/selenium diamond plate with the separating layer facing the beam. Prior to FT-IR measurements the membrane samples were immersed in water followed by IPA to remove traces of residual chemicals and PEG400. 2.4.2. Soak test for stability The stability of crosslinked PBI membranes was compared with the stability of commercially available polyimide membranes (Table 2). Small pieces of DuraMem 200 [38] (a crosslinked polyimide membrane from Evonik-MET) and lab prepared DBX and DBB crosslinked PBI membranes were cut and washed with DI water and IPA to remove any traces of conditioning chemicals. The pieces were then dried in vacuum and their weight was measured and recorded. Five soak solutions were prepared: 20 wt% DCA/ acetonitrile, 20 wt% PIP/DMF, 20 wt% MEA/water, 20 wt% NaOH/ water and 20 wt% HCl/water. Two pieces from two batches of membranes with the same composition were soaked in each solution to evaluate reproducibility. Along with these tests, one of each five solutions was left blank. This test was carried out for two months at 20 1C, after which the membrane pieces were withdrawn from the solutions, washed with water and vacuum dried. Then the weight of the sample was recorded and compared to the initial value. 2.4.3. Contact angle The contact angle of the membranes was obtained with an Easy-Drop Instrument (Kruess, Germany). Water drops of constant size were deposited on the membrane surface using a Table 1 Summary of crosslinked PBI membranes prepared from eight different dope solutions of the same composition. Four membranes were crosslinked with DBX and the other four with DBB. Membranes with entry nos. 1–6 were used for reproducibility evaluation. Entries 7 and 8 were used to demonstrate the chemical resistance of crosslinked PBI membranes. Entry no.

Membrane code

Composition Viscosity (cP) at 21 1C

Crosslinking Thickness (μm)

1 2 3

M1.1 – DBX 17 wt% PB Iin 7 720 M1.2 – DBX DMAc 8 090 M1.3 – DBX 6 850

3 wt% DBX 3 wt% DBX 3 wt% DBX

262.5 7 5 242.77 4 251.7 7 4

4 5 6

M2.1 – DBB 17 wt% PBI in 8 100 M2.2 – DBB DMAc 7 500 M2.3 – DBB 7 100

10 wt% DBB 10 wt% DBB 10 wt% DBB

228.0 7 4 230.7 7 4 215.2 7 8

7 8

M1 – DBX M2 – DBB

17 wt% PBI in 7 500 DMAc 7 690

3 wt% DBX 10 wt% DBB

250.8 7 5 241.3 7 7

micropipette. The digital image from the camera was then analysed using a built-in drop shape analysis tool. The reported contact angle for each membrane is an average of 10 drop measurements.

2.5. Membrane performance and analysis The filtration experiments were all carried out in crossflow cells connected in series (Fig. 2), each holding membrane discs with an effective area of 14 cm2. Pressure and temperature were kept constant at 10 bar and 30 1C, respectively, throughout the experiment. The pump flow rate was set at 40 L h 1 for all experiments. The permeance B is defined as the volumetric flow rate of solution per unit membrane area per unit pressure drop and can be calculated using the given equation B¼

V 1 1 ðL m 2 h bar Þ At Δp

ð1Þ

where V is the collected permeate volume, A is the effective membrane area, t is the time and Δp is the applied transmembrane pressure. Feed and permeate samples were taken at different time intervals for rejection Ri determination. Rejection was calculated using Eq. (2), where Cp,i and Cf,i represent the concentration of solute i in the permeate and the feed, respectively. C p;i 100 ð%Þ Ri ¼ 1 C f ;i

ð2Þ

A solute mixture containing polyethylene glycols (PEG) of three different molecular weights (400, 2000 and 8000 g mol 1) was used to determine the rejection properties of the membranes. The PEGs were dissolved in MeCN, DMF and DI water at a concentration of 1 g L 1 for each MW (referred to as standard solution from here on). Collected samples containing PEGs, DCA and MEA were analysed using an Agilent HPLC coupled to an evaporative light scattering detector (ELSD, Varian-385). The ELSD evaporator was set at 40 1C and the nebuliser at 55 1C. Nitrogen gas was supplied to the detector at a flow rate of 1.5 SLM (standard L m 1). A reverse phase column (C18-300, 250 mm 4.6 mm, ACE Hichrom) was used and the mobile phases were methanol and DI water buffered with 0.1 M ammonium acetate. The HPLC pump flowrate was set at 1 ml min 1 and the column temperature was set at 30 1C. The concentration of PIP was analysed using an Agilent GC with an HP-5 column (5% phenyl methyl siloxane; capillary: 30 m 0.530 mm 1.50 mm) coupled to a flame ionisation detector (FID). The temperature ramp was from 40 1C (hold for 1 min) until 200 1C with an increase of 10 1C min 1.

Table 2 Percentage weight loss of membrane samples which were inserted in five different acidic/basic solutions. Membrane Code

Weight loss (%)a 20 wt% DCA/MeCN

20 wt% PIP/DMF

20 wt% MEA/water (pH ¼ 12) 20 wt% HCl/water (pH ¼ 0) 20 wt% NaOH/water (pH¼ 14)

DuraMem 200

1 70

397 2

227 2

817 3

157 0

M1 – DBX M2 – DBB

0 70 0 70

17 0 37 2

27 1 27 0

17 0 47 0

07 0 17 0

a

The error indicates the standard deviation in % from the average weight loss value of two membrane samples.


65

Fig. 2. (a) Schematic representation of the crossflow filtration system with pressure gauge P, flow meter F and temperature control T. The solid lines represent the feed/ retentate flow and the dashed lines – the permeate flow; (b) top and cross-section view of a crossflow cell with (1) feed inlet, (2) retentate outlet, (3) permeate outlet, (4) o-rings, (5) membrane disc, and (6) sintered disc.

Fig. 3. Filtration sequences followed for evaluating the chemical stability of crosslinked PBI membranes: Sequence 1 was used for one set of M1 – DBX and M2 – DBB membranes and sequence 2 was used for a new set of the membranes.

2.6. Viscosity and molar volume of solvent solutions

2.8. Experimental sequence

The kinematic viscosity of the solvents and solvent mixtures used was determined using a BS/U/M2 miniature viscometer at 30 1C. Three consecutive measurements were taken (with coefficient of variation less than 1%) and the mean was used to calculate the kinematic viscosity. By multiplying with the corresponding density, the dynamic viscosity for each liquid was obtained. The molar volume Vm for pure solvents was taken from literature [39] and the molar volume Vm,mix of acidic and basic solutions was calculated according to the equation

Eight membranes were cast from eight different dope solutions by diluting the commercial PBI solution to 17 wt% of polymer solids. Four membranes were crosslinked with DBX and the other four with DBB. The procedure is described in detail in Section 2.2, and Table 1 summarises the composition and physical properties of the eight membranes. Two membrane discs were cut out of each of the eight batches to eliminate errors resulting from defective membrane coupons. First, the reproducibility of crosslinked PBI membranes was determined in different solvents – acetonitrile, DMF and DI water using entries 1–6 listed in Table 1. The pure solvent permeance was established for each of the membranes after washing out PEG400 preservative. The steady state value was taken as such after three consecutive hourly measurements were within 5% of each other. Then the standard PEG/solvent mixture was filtered for 24 h and rejection was taken at that point. Secondly, the filtrations of acidic and basic solvent mixtures were performed as solvent swap experiments (Fig. 3). The evaluation of the membrane stability was done by first establishing the membrane performance with solutions containing only the PEG marker solutes for 24 h. Then, without removing the membranes from the test cells they were exposed for 24 h to the acidic or basic conditions. Next, by filtering pure respective solvents the rig was cleaned from the previous mixtures. The last step involved filtration of the initial solvent containing the PEG markers. The chemical stability of the membranes was evaluated by comparing membrane performance before and after acid/base exposure. The composition and physical properties of the two membranes M1 – DBX and M2 – DBB can be found in Table 1 entry nos. 7 and 8. Due

V m;mix ¼

∑i xi M i

ρmix

ðmol cm 3 Þ

ð3Þ

where xi and Mi are the mole fraction and molecular weight of component i, respectively, and ρmix is the measured density of the mixture at 30 1C.

2.7. Scanning electron microscopy (SEM) Scanning electron micrographs of membrane surface and crosssection were recorded using a JEOL 5610 LV. The samples were first washed with IPA in order to remove PEG400. The surface samples were prepared by cutting small squares and pasting them onto SEM sample holders covered with carbon tape. For the preparation of cross-section samples, small pieces of membrane were snapped under liquid nitrogen and pasted vertically onto SEM stubs. Finally, the samples were sputtered with gold under argon atmosphere (Emitech K550 coater). The SEM was operated at an acceleration voltage of 10 kV.

66


to failure of M2 – DBB after completing the first filtration sequence (Fig. 3 – Sequence 1), fresh membrane discs were taken to evaluate their performance upon exposure to MEA (Fig. 3 – Sequence 2). All results are reported as the mean between the tested samples and the error values represent the one standard deviation from the mean.

3. Results and discussion 3.1. Membrane characterisation 3.1.1. FT-IR Prior to filtration tests, small membrane samples were used to confirm that the crosslinking was successful. Fig. 4 shows the IR spectra of uncrosslinked, DBX and DBB crosslinked membranes. The characteristic peaks of PBI were confirmed with spectra of PBI published in literature [40]. The peak at 3415 cm 1 can be attributed to non-hydrogen bonded N–H stretching and the peak at 3145 cm 1 to hydrogen bonded N–H stretching. Unfortunately, these signals are not useful because O–H stretching is registered as a broad peak in the same area, overlapping the N–H stretches. The peaks in the fingerprint area (at 1612, 1590, 1443, 1286, 801 and 705 cm 1) are all attributed to benzene and imidazole rings and their conjugation. Two distinct peaks at 2920 cm 1 and 2850 cm 1 were observed in the spectra of the crosslinked membranes which are attributed to C– H (the terminal C of the crosslinker) and C–N (the link between the crosslinker and the polymer backbone) stretches, confirming the crosslinking between the imidazole rings. 3.1.2. Soak test for stability To assess the stability of crosslinked PBI membranes, the membranes were soaked in different chemical solutions and the observed stability was compared to that of commercially available PI membranes (DuraMem 200). Table 2 summarises the percentage weight loss of the three membranes after withdrawing them from the corresponding solutions. Polyimide membrane degraded in MEA, PIP, HCl and NaOH. The DCA soak test showed localised swelling of polyimide membrane surface but did not result in membrane disintegration. The two types of crosslinked PBI membranes were found to be stable in all five test solutions based on visible observation and weight loss measurement, except for M2 – DBB exposed to HCl/water and PIP/DMF. The PIP/DMF solution turned a slight yellow colour due to dissolution of PBI in DMF. As the five blank solutions remained clear for the tested period, it can be concluded that any visible observation was due to the degradation

of the membrane pieces. From Table 2 it can be concluded that DBX crosslinked PBI membranes, compared to DuraMem 200 and DBB crosslinked membranes, showed superior chemical stability in the pH range from 0 to 14. 3.1.3. Contact angle measurements The contact angles of crosslinked PBI membranes were measured and compared with uncrosslinked PBI samples. It can be seen in Table 3 that the contact angles of crosslinked PBI membranes decreased compared to uncrosslinked PBI, suggesting that the membranes became more hydrophilic after the crosslinking treatment. The lowest contact angle was measured for DBX treated PBI membranes, decreasing by more than 40% from the value of untreated PBI. This is in agreement with the observed higher permeances of M1 – DBX in all tested solvents compared to M2 – DBB (described in Section 3.2). Such behaviour has previously been reported by Soroko et al. [41] for TiO2 doped PI membranes. The higher TiO2 content resulted in more hydrophilic surfaces and an increase in ethanol flux. Bhanushali et al. [42] tested various polar and non-polar solvents with commercial hydrophilic and hydrophobic membranes. They concluded that hydrophilic membranes provide higher fluxes with polar as opposed to non-polar solvents. 3.2. Membrane performance 3.2.1. Reproducibility of crosslinked PBI membranes Crosslinked PBI membranes (entry nos. 1–6 in Table 1) were tested for their reproducibility in the three solvents of interest – acetonitrile, DMF and DI water. Six membrane discs from each batch were tested in each solvent and a new set of discs was used for each solvent. First, the pure solvent permeance of the membranes was measured after 24 h of operation and the data is summarised in Table 4. Comparing the pure solvent permeances shown in Table 4, a correlation is evident between solvent viscosity, molar volume, surface tension (Table 5) and the respective permeance. Consistent with the trend reported in literature [43,44], solvent transport through crosslinked PBI membranes increased with decreasing viscosity, molar volume and surface tension (Table 5). Hence, the highest permeance, observed for acetonitrile, can be attributed to the lowest viscosity. On the other hand the lowest permeance was obtained for DMF due to its higher viscosity and higher molar volume. Water has the highest Table 3 Contact angles for unmodified and modified PBI membranes. Membrane

Contact angle (deg)a

Uncrosslinked PBI M1 – DBX M2 – DBB

60.5 7 2.3 35.37 5.4 55.0 7 2.3

a The error indicates the standard deviation between 10 drop measurements on the same membrane.

Table 4 Pure solvent permeances for DBX and DBB crosslinked membranes measured after establishing steady state. Membrane

DBX crosslinked PBI DBB crosslinked PBI Fig. 4. FT-IR spectra for uncrosslinked, DBB and DBX crosslinked PBI membrane samples.

Pure solvent permeance (L m 2 h 1 bar 1)a Acetonitrile

DMF

DI water

377 7 177 7

77 1 17 0

127 7 47 2

a The error given in the table is the standard deviation from the average calculated value of the total of three membrane batches (entry nos. 1–6 in Table 1).


viscosity and surface tension, which should result in the lowest permeance. However, due to its small molar volume, water had an intermediate flux, slightly higher than DMF (Table 4). After obtaining the pure solvent permeance, the pure solvent was replaced by a solution containing the PEG markers. The flux was monitored over time and PEG rejection was measured after 24 h. The results are shown in Fig. 5. In all filtrations DBB membranes showed a lower permeance than DBX membranes. This was likely due to the more hydrophilic surface of M1 – DBX (Table 3) which enhanced the permeance of polar solvents. In Fig. 5(a), it is interesting to note that filtration of PEG/acetonitrile through DBX crosslinked membranes led to around 50% decrease in permeance (from 21 to 11 L m 2 h 1 bar 1) whereas filtration of PEG/DMF and PEG/DI water resulted in no significant permeance loss (Fig. 5(a)). On the other hand, for the membranes crosslinked with DBB no impact of the solvent/solute

Table 5 Solvents and solvent mixtures used in the study and their characteristics. Solvent mixture

Viscositya (cP)

Molar volumeb (cm3 mol 1)

Surface tensionb (mN m 1)

Pure Acetonitrile 3 wt% DCA/acetonitrile Pure DMF 20 wt% PIP/DMF DI water 20 wt% MEA/DI water

0.34 0.36 0.78 0.82 0.83 1.27

52.7 52.8 77.4 81.4 18.1 21.2

28.03 – 33.90 – 71.67 –

a

Measured viscosity at 30 1C. Values for pure solvents taken from [39] at 30 1C and for mixtures calculated using Eq. (3) at 30 1C. b

67

system was observed (Fig. 5(b)). It is speculated that such trend comes from the initial swelling of DBX membranes by MeCN, slowly compensated by pressure compaction. Furthermore, Fig. 5(c) and (d) show the rejection of PEG markers in the three different solvents. It can be noted that DBX crosslinked membranes had the same rejection profile in all tested solvents, while DBB crosslinked ones appeared to have a higher rejection of PEGs when water is the solvent and lower in the case of MeCN and DMF. Overall, the error bars in Fig. 5 indicate that crosslinked PBI membranes have a consistent performance from batch to batch and it can be concluded that all demonstrated effects are reproducible.

3.2.2. Filtration under acidic/basic conditions To assess the chemical stability membranes M1 – DBX and M2 – DBB (entry nos. 7 and 8 in Table 1) were used. The experimental solvent sequences, shown in Fig. 3, were used to investigate the resistance in acidic and basic environments. The permeance results collected from the filtrations are shown in Fig. 6. Similar to the solvent permeance data in Fig. 5(a), M1 – DBX membranes had higher permeances than M2 – DBB membranes (Fig. 5(b)) in all tested solvents. The solvent/solute permeances followed the same trend as seen in Fig. 5 and increased in the order: PEG/DMFoPEG/DI wateroPEG/acetonitrile. A steady state flux was reached after 24 h filtration of PEG/solvent solutions in each case. After washing the rig with fresh solvents to ensure removal of PEGs, the corresponding acidic or basic solution was loaded and circulated for 24 h. The permeances for acidic and basic solutions are shown in Fig. 6 in the middle panel of each graph. In all three cases, an instant drop

Fig. 5. Average solvent permeance profiles over time of (a) PBI membranes crosslinked with DBX and (b) PBI membranes crosslinked with DBB; average rejection of PEGs after 24 h for (c) PBI membranes crosslinked with DBX and (d) PBI membranes crosslinked with DBB. The error bars represent one standard deviation from the average value of the measurement, where each point summarises the average value from tests on M1.1 – DBX, M1.2 – DBX, M1.3 – DBX panels (a) and (c) and M2.1 – DBB, M2.2 – DBB, M2.3 – DBB panels (b) and (d).

68


Fig. 6. Permeance profiles through M1 – DBX and M2 – DBB PBI membranes using the following sequences: (a) PEG/Acetonitrile, 3 wt% DCA in PEG/Acetonitrile and PEG/ Acetonitrile (b) PEG/DMF, 20 wt% PIP in PEG/DMF and PEG/DMF (c) PEG/DI water, 20 wt% MEA in PEG/DI water and PEG/DI water.

in flux was observed. The reason for such drop in flux may be attributed to a combination of four effects: (1) change in solution viscosity and molar volume, (2) polymer degradation, (3) build-up of osmotic pressure across the membrane, and (4) interaction of acids and bases with the crosslinked polymer backbone. To understand the nature of this effect, the viscosities and molar volumes of the three tested solutions were obtained, summarised in Table 5. It can be concluded that the addition of DCA or PIP did not result in significant increase of solution viscosity (only 5–6%) and molar volume (less than 5%). Hence, in these cases the contribution of solution viscosity and molar volume to the flux drop (between 27% and 34%) was insignificant. However, the addition of 20 wt% MEA to DI water led to more than 50% increase in solution viscosity and 17% increase in solution molar volume. The drop in flux in the case of MEA/DI water was found to be more significant (51% for M1 – DBX and 61% for M2 – DBB decrease) than with the other solutions, possibly due to a more pronounced effect of solution viscosity and molar volume. Another possible cause could be polymer degradation, which is only likely in the case of PIP/DMF filtration through M2 – DBB according to the stability soak test (Table 2). Addition of PIP to DMF resulted in 70% decrease in permeance for membrane M2 – DBB. Of the four effects, it is more likely that the osmotic pressure increased and/or the acids and bases interacted with the

Table 6 Summary of osmotic pressure differences across M1 – DBX and M2 – DBB due to addition of DCA, PIP and MEA with the respective rejection data.

3 wt% DCA 20 wt% PIP 20 wt% MEA

Rejection (%)

Osmotic pressure difference (bar)

M1 – DBX

M2 – DBB

M1 – DBX

M2 – DBB

5 5 9

3 3 11

0.3 2.6 7.0

0.1 N/A 9.0

crosslinked polymer. Osmotic pressure difference Δπ can be estimated using Van0 t Hoff relations [45], shown in the given equation

Δπ ¼ ðC f ;i C p;i ÞRT ðbarÞ

ð4Þ

where Cf,i and Cp,i are the molar concentrations of solute i on the feed and permeate side, respectively, R represents the universal gas constant, and T is the temperature. The contribution of the calculated difference in osmotic pressure across membranes M1 – DBX and M2 – DBB to the decrease in driving force is summarised in Table 6. The addition of 3 wt% DCA to acetonitrile did not result in a decrease of driving force for both


types of membranes. In this case it is likely that other interactions between the crosslinked polymer and the acid occurred. In the filtration of 20 wt% PIP in DMF the osmotic pressure difference accounted for 30% drop in driving force for membrane M1 – DBX, which is consistent with the observed 32% decrease in flux (Fig. 6 (b)). M2 – DBB gave negative rejection values of PIP, thus validating polymer degradation by DMF. The decrease of driving force was most significant during filtration of 20 wt% MEA/DI water. The effective driving force was only 3 bar for M1 – DBX and 1 bar for M2 – DBB compared to the applied 10 bar, predicting a higher drop in permeance than that observed in the experiment (Fig. 6c)). We ascribe this discrepancy to limitations of the Van0 t Hoff relation when applied to non-ideal concentrated solutions [45]. The equation was used here to give a rough estimate of the magnitude of osmotic pressure difference. Nevertheless, we can conclude that increased solution viscosity and molar volume, coupled with high rejections of MEA, contributed to a flux drop of more than 50% for the filtration of MEA/DI water.

69

The contributions of the four aforementioned effects to the observed permeance drop are verified by the permeance behaviour upon removal of the acidic or basic media (Fig. 6 last panel of each graph). The permeance for M1 – DBX membranes regained the original value almost completely in acetonitrile and DI water and in the case of DMF, an increase in flux of about 9% of the original value was observed. M2 – DBB membranes regained their permeance in the case of acetonitrile and DI water filtrations but failed in DMF, which is in agreement with the stability soak test shown in Table 2. These observations were also confirmed by the PEG rejection profiles shown in Figs. 7 and 8. The rejections before and after DCA, PIP and MEA filtration are overlapping for both types of membranes. A slight increase observed in PEG400 rejection was due to further compaction. PEG rejection for M2 – DBB after DMF filtration showed no NF properties. On the contrary, the comparison of graphs (a) and (e) in Fig. 7 confirmed the conclusion that filtering DMF through DBX crosslinked PBI membranes had no adverse effect on membrane performance. However, DBB membranes lost their

Fig. 7. Rejection profiles of PEG standards dissolved in acetonitrile and DMF: (a) before and (b) after filtration of DCA/acetonitrile for M1 – DBX and M2 – DBB crosslinked PBI membranes; (c) before and (d) after filtration of PIP/DMF for M1 – DBX and M2 – DBB crosslinked PBI membranes and (e) final PEG rejection in acetonitrile to compare with the initial rejection shown in (a).

70


Fig. 8. Rejection profiles of PEG standards dissolved in water (a) before and (b) after filtration of MEA for M1 – DBX and M2 – DBB crosslinked PBI membranes.

Fig. 9. SEM pictures of surface and cross-sections of M1 – DBX (a) before filtration, (b) after DCA/MeCN and PIP/DMF filtration and (c) after MEA/water filtration.

Fig. 10. SEM pictures of surface and cross-sections of M2 – DBB (a) before filtration, (b) after DCA/MeCN and PIP/DMF filtration and (c) after MEA/water filtration.


MWCO, likely due to dissolution in DMF than to the basic conditions. Considering that PBI is soluble in polar aprotic solvents, insufficiently crosslinked PBI membranes would dissolve in DMF over time. 3.3. Membrane morphology under SEM After completing the filtration experiments small pieces from the tested membranes were taken and inspected using SEM. The pictures of the membranes are shown in Fig. 9 for M1 – DBX and in Fig. 10 for M2 – DBB. Fig. 9(a) represents the structure of M1 – DBX before being used and the pictures in Fig. 9(b) and (c) show M1 – DBX after exposure to DCA/acetonitrile and PIP/DMF and MEA/DI water, respectively. As can be seen, no visible change can be observed for DBX membranes before and after filtrations. The membrane retained its features on the surface and across the membrane cross-section. On the other hand, visible changes were detected for DBB membranes. The initial membrane morphology of M2 – DBB is shown in Fig. 10(a) and the pictures shown in (b)) Fig. 10(b) reveal the membrane structure after DCA/acetonitrile and PIP/DMF filtration and Fig. 10(c) after MEA/DI water. It is clear that the morphology of M2 – DBB changed after the first set of acidic and basic filtrations. The surface of the membrane Fig. 10(b)) became rougher and the cross-section of the membrane looked distorted. Comparing the flux and rejection data shown in Figs. 6 and 7, respectively, for DBB membranes, it can be deduced that the deformation of membranes occurred during the DMF filtration stage. In line with the flux and rejection data, it can be concluded that DBB crosslinked PBI membranes are not stable in the presence of DMF, possibly due to insufficient crosslinking of the polymer backbone. Apart from the DMF filtration, it can be seen that MEA/DI water had no significant effect on M2 – DBB.

4. Conclusions Integrally skinned asymmetric polybenzimidazole (PBI) nanofiltration membranes were successfully prepared and crosslinked using halogenated compounds – dibromoxylene (DBX) and dibromobutane (DBB). Both membranes showed consistent and reliable batch-to-batch separation performance in acetonitrile, DMF and DI water using a range of PEGs as nanofiltration markers. In addition, DBX crosslinked PBI membranes exhibit permeabilities superior to those of DBB crosslinked ones in all tested solvents. Further, crosslinked PBI membranes have been tested under acidic and basic conditions, as well as in organic solvent environments. They have shown good chemical tolerance towards these aggressive conditions and recovered their initial performance upon neutralisation. The only exception was filtration of DMF through PBI membranes crosslinked with DBB. The membranes failed to regain their initial performance and SEM pictures revealed a distorted morphology. This indicates insufficient crosslinking and dissolution of polymer in DMF. Crosslinking with DBX resulted in much more stable membranes which could withstand harsh chemical environments including DMF. The filtration experiments carried out showed that crosslinked PBI membranes could lead to a wider implementation of membrane technology in separation processes in harsh and corrosive environments. In a more general comparison, PBI membranes show similar performance to polyimide membranes, but can withstand much wider chemical environments. Hence, they have great potential to be used as OSN membranes in various pharmaceutical purification processes, which also employ acids and bases. In addition, PBI seems to be a promising membrane material for the needs of CCS.

71

Acknowledgement Funding for this work was provided by the 7th Framework Program of the European Commission0 s Marie Curie Initiative (PITN-GA-2008-238291-MEMTIDE). The authors wish to thank Dr. Ludmila G. Peeva for her help in setting up the analytical methods. References [1] G. Szekely, M. Gil, B. Sellergren, W. Heggie, F.C. Ferreira, Environmental and economic analysis for selection and engineering sustainable API degenotoxification processes, Green Chem. 15 (2013) 210–225. [2] P. Vandezande, L.E.M. Gevers, I.F.J. Vankelecom, Solvent resistant nanofiltration: separating on a molecular level, Chem. Soc. Rev. 37 (2008) 365–405. [3] P. Marchetti, A. Butte, A.G. Livingston, Chapter 8: reactive peptide nanofiltration, in: N. Shamim, V.K. Sharma (Eds.), Sustainable Nanotechnology and the Environment: Advances and Achievements, ACS, Washington, DC, 2013, pp. 121–150. [4] R.M. Gould, L.S. White, C.R. Wildemuth, Membrane separation in solvent lube dewaxing, Environ. Prog. 20 (2001) 12–16. [5] S. So, L.G. Peeva, E.W. Tate, R.J. Leatherbarrow, A.G. Livingston, Organic solvent nanofiltration: a new paradigm in peptide synthesis, Org. Process Res. Dev. 14 (2010) 1313–1325. [6] G. Szekely, J. Bandarra, W. Heggie, B. Sellergren, F.C. Ferreira, Organic solvent nanofiltration: a platform for removal of genotoxins from active pharmaceutical ingredients, J. Membr. Sci. 381 (2011) 21–33. [7] G.M. Geise, H.-S. Lee, D.J. Miller, B.D. Freeman, J.E. McGrath, D.R. Paul, Water purification by membranes: the role of polymer science, J. Polym. Sci. B Polym. Phys. 48 (2010) 1685–1718. [8] J. Schaer, B. Van der Bruggen, S. Uytterhoeven, R. Croux, C. Vandecasteele, D. Wilms, E. Van Houtte, F. Vanlerberghe, Removal of hardness from groundwater by nanofiltration, Desalination 119 (1998) 295–302. [9] B. Van der Bruggen, Fundamentals of membrane solvent separation and pervaporation, in: E. Drioli, L. Giorno (Eds.), Membrane Operations: Innovative Separations and Transformations, Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, 2009, pp. 45–61. [10] Y.H. See Toh, F.W. Lim, A.G. Livingston, Polymeric membranes for nanofiltration in polar aprotic solvents, J. Membr. Sci. 301 (2007) 3–10. [11] K. Vanherck, P. Vandezande, S.O. Aldea, I.F.J. Vankelecom, Cross-linked polyimide membranes for solvent resistant nanofiltration in aprotic solvents, J. Membr. Sci. 320 (2008) 468–476. [12] W. Woishnis, S. Ebnesajjad, (Eds.), Chemical Resistance of Specialty Thermoplastics, Elsevier Inc., 2012, pp. 863–870, pp. 1399–1417. [13] Y.H. See-Toh, M. Silva, A. Livingston, Controlling the molecular weight cut-off curves for highly solvent stable organic solvent nanofiltration (OSN) membranes, J. Membr. Sci. 324 (2008) 220–232. [14] M. Dalwani, G. Bargeman, S.S. Hosseiny, M. Boerrigter, M. Wessling, N.E. Benes, Sulfonated poly(ether ether ketone) based composite membranes for nanofiltration of acidic and alkaline media, J. Membr. Sci. 381 (2011) 81–89. [15] T.J.K. Visser, S.J. Modise, H.M. Krieg, K. Keizer, The removal of acid sulphate pollution by nanofiltration, Desalination 140 (2001) 79–86. [16] M. Liu, G. Yao, Q. Cheng, M. Ma, S. Yu, C. Gao, Acid stable thin-film composite membrane for nanofiltration prepared from naphthalene-1,3,6-trisulfonylchloride (NTSC) and piperazine (PIP), J. Membr. Sci. 415–416 (2012) 122–131. [17] D.A. Musale, A. Kumar, Solvent and pH resistance of surface crosslinked chitosan/poly(acrylonitrile) composite nanofiltration membranes, J. Appl. Polym. Sci. 77 (2000) 1782–1793. [18] R.L. Goldsmith, B.A. Wechsler, H. Shigeyoshi, K. Mori, Y. Taketani, Development of PBIL low pressure brackish-water reverse osmosis membranes, Desalination 22 (1977) 311–333. [19] U.S. Pat. No. 4693825, 1987. [20] U.S. Pat. No. 3720607, 1973. [21] L.C. Sawyer, R.S. Jones, Observations on the structure of first generation polybenzimidazole reverse osmosis membranes, J. Membr. Sci. 20 (1984) 147–166. [22] S.C. Kumbharkar, P.B. Karadkar, U.K. Kharul, Enhancement of gas permeation properties of polybenzimidazoles by systematic structure architecture, J. Membr. Sci. 286 (2006) 161–169. [23] K.Y. Wang, Y. Xiao, T.S. Chung, Chemically modified polybenzimidazole nanofiltration membrane for the separation of electrolytes and cephalexin, Chem. Eng. Sci. 61 (2006) 5807–5817. [24] M.F. Flanagan, I.C. Escobar, Novel charged and hydrophilized polybenzimidazole (PBI) membranes for forward osmosis, J. Membr. Sci. 434 (2013) 85–92. [25] Q. Li, R. He, J.O. Jensen, N.J. Bjerrum, PBI-based polymer membranes for high temperature fuel cells—preparation, characterization and fuel cell demonstration, Fuel Cells 4 (2004) 147–159. [26] S. Yu, B.C. Benicewicz, Synthesis and properties of functionalized polybenzimidazoles for high-temperature PEMFCs, Macromolecules 42 (2009) 8640–8648. [27] T.S. Chung, A critical review of polybenzimidazoles, J. Macromol. Sci. C Polym. Rev. 37 (2006) 277–301. [28] K.Y. Wang, T.S. Chung, Fabrication of polybenzimidazole (PBI) nanofiltration hollow fibre membranes for removal of chromate, J. Membr. Sci. 281 (2006) 307–315.

72


[29] U.S. Pat. No. 4666996, 1987. [30] US Pat. No. 2013/0118983 (A1),2013. [31] Y.H. See Toh, X.X. Loh, K. Li, A. Bismarck, A.G. Livingston, In search of a standard method for the characterisation of organic solvent nanofiltration membranes, J. Membr. Sci. 291 (2007) 120–125. [32] X. Li, F. Monsuur, B. Denoulet, A. Dobrak, P. Vandezande, I.F.J. Vankelecom, Evaporative light scattering detector: toward a general molecular weight cutoff characterization of nanofiltration membranes, Anal. Chem. 81 (2009) 1801–1809. [33] WQ. Chen, M. Cristau, M. Giraud, A.G. Livingston, Novel liquid phase peptide synthesis (LPPS) technology: elongation using organic solvent nanofiltration (OSN), J. Pept. Sci. 18 (2012) S133. [34] G. Szekely, M. Schaepertoens, P.R.J. Gaffney, A.G. Livingston, Iterative synthesis of monodisperse PEG homostars and linear heterobifunctional PEG, Polym. Chem. 5 (2014) 694–697. [35] G.M. Bonora, G. Biancotto, M. Maffini, C.L. Scremin, Large scale, liquid phase synthesis of oligonucleotides by the phosphoramidite approach, Nucl. Acids Res. 21 (1993) 1213–1217. [36] A.A. Zompra, A.S. Galanis, O. Werbitzky, F. Albericio, Manufacturing peptides as active pharmaceutical ingredients, Future Med. Chem. 1 (2) (2009) 361–377. [37] H. Lepaumier, D. Picq, P.-L. Carrette, New amines for CO2 capture. I. Mechanisms of amine degradation in the presence of CO2, Ind. Eng. Chem. Res. 48 (2009) 9061–9067. [38] Evonik MET, Ltd. 〈http://duramem.evonik.com/product/duramem-puramem/ en/about/membrane-properties/pages/default.aspx〉 (accessed September 2013).

[39] C.L. Yaws, Yaws0 Thermophysical Properties of Chemicals and Hydrocarbons, (electronic edition), Knovel, 2010, available at: http://app.knovel.com/hotlink/ toc/id:kpYTPCHE02/yaws-thermophysical-properties. [40] P. Musto, F.E. Karasz, W.J. MacKnight, Fourier transform infra-red spectroscopy on the thermo-oxidative degradation of polybenzimidazole and of a polybenzimidazole/polyetherimide blend, Polymer 34 (1993) 2934–2945. [41] I. Soroko, A. Livingston, Impact of TiO2 nanoparticles on morphology and performance of crosslinked polyimide organic solvent nanofiltration (OSN) membranes, J. Membr. Sci. 343 (2009) 189–198. [42] D. Bhanushali, S. Kloos, C. Kurth, D. Bhattacharyya, Performance of solventresistant membranes for non-aqueous systems: solvent permeation results and modeling, J. Membr. Sci. 189 (2001) 1–21. [43] D.R. Machado, D. Hasson, R. Semiat, Effect of solvent properties on the permeate flow through nanofiltration membranes. Part I: Investigation of parameters affecting solvent flux, J. Membr. Sci 163 (1999) 93–102. [44] J. Geens, B. Van der Bruggen, C. Vandecasteele, Transport model for solvent permeation through nanofiltration membranes, Sep. Purif. Technol. 48 (2006) 255–263. [45] L.G. Peeva, E. Gibbins, S.S. Luthra, L.S. White, R.P. Stateva, A.G. Livingston, Effect of concentration polarisation and osmotic pressure on flux in organic solvent nanofiltration, J. Membr. Sci. 236 (2004) 121–136.