Increasing membrane permeability of UV-modified poly(ether sulfone) ultrafiltration membranes

Increasing membrane permeability of UV-modified poly(ether sulfone) ultrafiltration membranes

Journal of Membrane Science 202 (2002) 1–16 Increasing membrane permeability of UV-modified poly(ether sulfone) ultrafiltration membranes John Pierac...

576KB Sizes 0 Downloads 0 Views

Journal of Membrane Science 202 (2002) 1–16

Increasing membrane permeability of UV-modified poly(ether sulfone) ultrafiltration membranes John Pieracci a , James V. Crivello b , Georges Belfort a,∗ a

Howard P. Isermann Department of Chemical Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA b New York State Center for Polymer Synthesis, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA Received 15 June 2001; received in revised form 27 July 2001; accepted 6 August 2001

Abstract The use of the chain transfer agent, 2-mercaptoethanol, and ethanol cleaning was successful in producing hydrophilic membranes with higher permeabilities than the unmodified poly(ether sulfone) (PES) membrane. Permeabilities of dip-modified membranes were three to five times higher than with ethanol cleaning alone. Membrane permeability increased by 50% for dip-modified membranes in the presence of a low 2-mercaptoethanol concentration and by 20–200% when high 2-mercaptoethanol concentrations were used. From the lower degree of grafting (DG) achieved, it was inferred that the graft chain density and chain length decreased with increasing concentration of chain transfer agent. However, the observed rejection was severely reduced indicating that dip-modification caused considerable pore enlargement. This was not previously observed because non-grafted homopolymer may have been formed which effectively blocked the pores. It is speculated that ethanol cleaning removed the homopolymer because it wets and possibly swells the membrane pore structure. This indicated that it was not possible to dip-modify 50 kDa PES UF membranes using 300 nm lamps and a benzene filter without causing significant change to the pore structure. In addition, the combination of irradiation and N-vinyl-2-pyrrolidinone (NVP) caused a more severe loss of observed rejection than irradiation alone. The molecular weight cut-off (MWCO) of membranes irradiated and/or dip-modified with or without 2-mercaptoethanol was estimated to be between 50 and 100 kDa using a calibration curve for observed rejection versus membrane MWCO. The high hydrophilicity, high permeability, and low fouling character of membranes modified with 2-mercaptoethanol and cleaned with ethanol make them desirable for the filtration of larger proteins (>500 kDa) or for applications with smaller proteins in which high protein transmission is required. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Membrane surface modification; UV-assisted graft polymerization; Protein fouling; Chain transfer agents; Wetting agents ultrafiltration; Porous membranes

1. Introduction Photochemical modification of ultrafiltration (UF) membranes by the UV-assisted graft polymerization of hydrophilic monomers has been shown to be successful in increasing surface hydrophilicity and decreasing membrane fouling during protein filtration [1–11]. ∗ Corresponding author. Tel.: +1-518-276-6948; fax: +1-518-276-4030. E-mail address: [email protected] (G. Belfort).

This has been accomplished with [1–4] or without the use of free radical photoinitiators [5–11]. A two-step UV process using photoinitiators has been used to modify poly(acrylonitrile) (PAN) and poly(sulfone) (PSf) UF membranes [1–4]. Yamagishi and coworkers [5–7] were able to achieve grafting on the surface of PSf and poly(ether sulfone) (PES) UF membranes in a single step without the use of a photoinitiator due to the intrinsic photoactivity of PSf and PES. Recently, a one-step technique was developed to render the

0376-7388/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved. PII: S 0 3 7 6 - 7 3 8 8 ( 0 1 ) 0 0 6 2 4 - X

2

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

modification process more easily adaptable to continuous membrane manufacturing processes [9,10]. Using this new dip-modification technique, very hydrophilic PES modified membranes were produced that exhibited as low protein fouling as regenerated cellulose membranes during the filtration of bovine serum albumin (BSA) without loss of observed BSA rejection [10]. However, membrane permeability was decreased after modification. This loss of membrane permeability has been widely observed [1–4,6–11] and has been linked to the blockage of membrane pores by the grafted polymer chains [1–4,6–11]. Specifically, it has been shown that a high grafted chain density and long chain length promotes permeability loss [12–14]. Ito et al. observed a loss of membrane permeability with increasing grafted chain density and chain length after modifying polycarbonate microporous membranes with acrylic [12] and methacrylic acids [13] using glow discharge treatment. They observed similar results when benzyl-l-glutamate was grafted onto the surface of microporous membranes composed of a tetrafluoroethylene/ethylene copolymer [14]. Grafting short polymer chains may be one method of maintaining more of the unmodified membrane permeability after modification. However, a high grafted chain density and long graft chain length may be essential to impart the necessary surface hydrophilicity to decrease membrane fouling. Therefore, the graft chain density and chain length should be optimized to impart the necessary surface hydrophilicity and maintain the permeability as high as possible. Chain transfer agents can be used to control the degree of polymerization during free radical polymerizations [15]. Chain transfer agents simultaneously terminate growing polymer chains and generate new radicals, resulting in a higher polymer chain density and a lower average polymer chain length. This leads to a lower molecular weight polymer with a broad molecular weight distribution [15]. By varying the chain transfer agent concentration used during dip-modification, the grafted chain density and chain length can be optimized. In this study, the chain transfer agent, 2-mercaptoethanol, was selected because thiol compounds have very high chain transfer constants [16] and, thus, can effectively control chain length at a low concentration. In addition, 2-mercaptoethanol is one of the few thiol compounds

with a high water solubility. Increasing membrane permeability after dip-modification may also be achieved through the use of solvents which better wet or even swell the pores of the membranes without causing an observable loss of protein rejection. Nystrom and Jarvinen [17] irradiated 6 kDa PSf UF membranes in the presence of ethanol which caused an increase in permeability without the loss of BSA rejection. Crivello et al. [7] observed higher permeability after modification of 50 kDa PES membranes immersed in methanolic solutions of 2-hydroethyl methacrylate (HEMA) than those in aqueous solutions. It was speculated that the increased permeability was caused by pore enlargement, but the unchanged BSA rejection suggests that better pore wetting or possibly slight pore swelling was responsible. The goal of this research was to produce highly retentive, low fouling 50 kDa PES UF membranes using dip-modification with high BSA retention while maintaining membrane permeability through the use of the chain transfer agents, ethanol cleaning, or by a combination of the two. The experimental materials and methods will be discussed first, followed by a discussion of the membrane characterization and protein filtration results. 2. Experimental 2.1. Materials The unmodified 10 kDa (Lot #7120D and 7190D), 30 kDa (Lot #8230B), 50 kDa (Lot #T0238A, 8139B, 9047G, and 9140G), 70 kDa (Lot #7309A), 100 kDa (Lot #7265G), and 300 kDa (Lot #T9336J) PES UF membranes and the 50 kDa regenerated cellulose (RC) membrane (Lot #8061D) were obtained from Pall-Filtron Corp. (East Hills, NY). All the PES membranes were modified by the manufacturer by an undisclosed process to increase hydrophilicity. N-vinyl-2-pyrrolidinone (NVP) was obtained from Aldrich (Milwaukee, WI) and purified before use by vacuum distillation to remove the inhibitor before use. 2-Mercaptoethanol (ME) and HPLC-grade benzene were obtained from Aldrich (Milwaukee, WI). Nitrogen gas and air, received from Matheson Co. (Secaucus, NJ), were of ultra high purity. Deionized water was produced from the tap water by an in-house

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

3

deionized water purification system consisting of reverse osmosis (FT-30, FilmTech, MN), UV irradiation treatment, and a train of polycarbonate, polypropylene, and Teflon® micron and submicron membranes (Nucleopore Corporation, Pleasanton, CA). Bovine serum albumin (BSA, essentially fatty acid free, >98% Lot #27H1256) and invertase (INV, Lot #83F81351) were obtained from Sigma (St. Louis, MO). Protein concentrations were determined spectroscopically at 280 nm using a Hitachi U 2000 Double-Beam UV/ VIS spectrophotometer (Hitachi Instruments, Inc., Danbury, CT).

The cosine of the contact angle is referred to as the wettability of a surface [20], and the higher the value of cos θ , the more wettable the surface is by a given solvent. Membranes were inverted in deionized water and air bubbles (1 mm) were placed in contact with the surface. The static contact angle, θ , was measured using a SIT camera (SIT66, Dage-MTI, Inc., Michigan City, IN) connected to a video screen. The contact angles were averages of 10 measurements with an error of ±3◦ .

2.2. Photochemical modification technique

A dead-end stirred cell filtration system was designed to characterize the filtration performance of unmodified and modified membranes and was described previously [10]. All filtration experiments were conducted at a constant transmembrane pressure of between 28 and 379 kPa (4–55 psig), a stirring rate of 300 rpm, and a temperature of 21 ◦ C. A 0.1 wt.% BSA solution or invertase (INV) in 10 mM phosphate buffered saline (PBS) at pH 7.4 was used as the test protein solution. BSA has a molecular weight of 67 kDa [21], with an isoelectric point of 4.7 [21], and at pH 7.4, it has a charge of −20.5 [21]. Invertase has a molecular weight of 270 kDa [22] with an isoelectric point of 4.2 [23].

A Rayonet Photochemical Chamber Reactor System (Model RPR-100, Southern New England Ultraviolet Company, CT) and a benzene light filter were used to photochemically dip-modify the PES membranes. The reactor system and the benzene filter was described previously [10]. The dip-modification protocol was also described previously [10]. When 2-mercaptoethanol was used, it was added to the aqueous NVP solution at a concentration of 10 or 50 mM. After modification was complete, the membranes were washed to remove any unreacted monomer or physically adsorbed polymer by shaking them in bottles of deionized water or pure ethanol for 2 h at room temperature. 2.3. Surface analysis Attenuated total reflection-Fourier transform infrared spectra (ATR-FTIR) of the membranes were obtained using a spectrometer (Nicolet Magna-IR 550 Series II, Nicolet Instrument Corp., Madison, WI). For each measurement, 256 scans were performed at a resolution of ±4 cm−1 using a germanium crystal at an incident angle of 45◦ . Membranes were prepared for FTIR by placing them in a vacuum oven overnight at room temperature. Membranes were measured in duplicate. The IR penetration depth for this incident angle was 0.1–1 ␮m [18]. 2.4. Contact angle measurements Static contact angles of the membrane surface were measured using the captive air bubble technique [19].

2.5. Filtration system

2.6. Constant volume diafiltration A schematic representation of the constant volume diafiltration protocol is shown in Fig. 1. All fluxes were converted to permeabilities by dividing the permeation flux by the operating transmembrane pressure. Membrane fouling performance was judged by comparing the flux loss during the filtration of BSA or INV. The fraction of the initial buffer flux lost during protein filtration or the total flux loss, 1−J p /J0 , is the sum of (i) the flux lost to reversible adsorptive fouling and osmotic effects from concentration polarization, (J1 − J p )/J0 , and (ii) fouling due to irreversible protein adsorption and aggregation, (J0 − J1 )/J0 . A high value of (1 − J p /J0 ) corresponds to large reduction in flux. The fraction of the initial buffer flux recovered after water cleaning, (J1 − J p )/J0 , is a measure of the degree to which the osmotic effects from concentration polarization and reversibly adsorbed protein have decreased the flux. The remaining unrecovered

4

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

Fig. 1. Constant volume diafiltration protocol: (a) to minimize compaction effects, 10 mM phosphate buffered saline (PBS) was passed through the membrane for 30 min at a transmembrane pressure (TMP) of 172 or 379 kPa (25 or 55 psig); (b) TMP was lowered to the operating pressure and flux (J0 ) noted when the difference between consecutive measurements was <2%; (c) 0.1 wt.% BSA solution was pumped into the cell using a secondary pump, the operating pressure was re-applied, and the filtration was continued until 10 ml of permeate was collected (Jp ); (d) the cell was rinsed with Di-H2 O three times for 1 min each and the flux measured (J1 ); (e) the flux was again measured (J2 ). All filtration steps were operated at 21 ◦ C. J0 − J p was the total flux loss, J1 − J p was the flux loss to reversible protein adsorption and osmotic effects from concentration polarization, and J0 − J1 was the flux loss to irreversible protein adsorption.

initial buffer flux, (J0 − J1 )/J0 , is a measure of the irreversibly adsorbed protein fouling and can, in principle, be regained by treatment with caustic [8]. The observed rejection, R, of BSA or INV indicates both the effect of photochemical grafting on the pore structure and the protein fouling of the membrane. It is desirable to modify the membrane surface without severely compromising the pore structure of the base membrane.

3. Results and discussion 3.1. Effect of irradiation alone The effect of irradiation on the pore structure of 50 kDa PES membrane is shown in Tables 1 and 2. Membranes were irradiated using 300 nm lamps and a benzene filter under a variety of experimental

conditions. Two different size proteins: BSA (67 kDa, Table 1) and INV (270 kDa, Table 2) were used to test membrane observed rejection and filtration performance. For 50 kDa PES membranes, irradiation in water followed by water cleaning caused the observed BSA rejection to decrease from 98 to 80% (approximately) and the permeability to increase by 45% (6.5–9.4 lm h/kPa) at an irradiation energy dose of 1776 mJ/cm2 (Table 1). This indicated that the pores were enlarged by irradiation. Treating unmodified membranes with ethanol caused a 57% permeability increase from 5.6 to 8.8 lm h/kPa without an observable change in BSA rejection (99.1–99.0%, Table 1). Irradiated membranes cleaned in ethanol exhibited a similar decrease in the observed BSA rejection as irradiated membranes cleaned in water, but the permeability increased by 5–28% depending on the energy dose used. Membranes irradiated in water

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

5

6

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

and cleaned in ethanol were also tested with invertase and the results were similar to those tested with BSA except for membranes irradiated with a high energy dose (1776 mJ/cm2 , Table 2). These membranes exhibited a lower observed INV (270 kDa) rejection (57%) than the observed BSA (67 kDa) rejection (80%) of the membranes irradiated at the same energy dose, despite the larger size of INV. The use of two different membrane lots might have caused this difference (Lots #9140G and 9047G). Finally, irradiation in nitrogen in the presence of the chain transfer agent 2-mercaptoethanol also caused a similar decrease in rejection and an increase in permeability as with irradiation in water (data not shown). These conditions were chosen as a control to mimic conditions used during dip-modification. The total and irreversible flux loss during BSA and INV filtration of membranes irradiated under different conditions were similar to those of the unmodified membranes. This was expected because irradiation alone did not significantly increase surface hydrophilicity [8].

7

The use of observed protein rejection provides some information on tracking the progress of pore enlargement, but by itself, is limited in estimating molecular weight cut-off (MWCO) of the membranes. However, by creating a calibration curve of observed rejection versus MWCO using both BSA and INV and a series of unmodified PES membranes with a MWCO between 10 and 300 kDa (Fig. 2), an estimate of MWCO is possible. The observed rejection data suggest that the MWCO of the irradiated membranes at almost all tested experimental conditions was between 50 and 70 kDa, commonly defined as the molecular weight at R = 90%. The most open of these irradiated membranes, produced by irradiation in water using a high energy dose (1776 mJ/cm2 ) then cleaned in ethanol and tested with INV, had a MWCO of >100 kDa. Care must be taken when using this calibration curve because it is well known that the observed rejection can be increased by protein aggregation and adsorption [24] and concentration polarization [25]. This might certainly be true for these membranes because of their high total flux loss during BSA (66–77%, Table 1) and

Fig. 2. Observed protein rejection of different molecular weight cut-off irradiated PES UF membranes: (䊊) bovine serum albumin (BSA); (䊉) invertase (INV). Protein solutions were made in 10 mM PBS at pH 7.4 and 22 ◦ C at a concentration of 1 g/l. The transmembrane pressure was 34–103 kPa (5–15 psig).

8

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

INV filtration (85–88%, Table 2), most of which was due to irreversible protein fouling. Despite this limitation, the data suggests that irradiation of membranes using the 300 nm lamps and the benzene filter did not cause significant pore enlargement. 3.2. UV-assisted graft polymerization of NVP 3.2.1. Modified membrane degree of graft polymerization The degree of grafting (DG) of dip-modified membranes using 5 wt.% aqueous solutions of NVP with and without the chain transfer agent 2-mercaptoethanol is shown in Fig. 3. Dip-modification followed by water cleaning produced membranes with a extremely high DG (>2), which increased approximately linearly with energy dose. When ethanol was used to clean the membranes, the DG was similar up to a an energy dose of approximately 700 mJ/cm2 .

At higher doses, the DG was significantly lower after ethanol cleaning and the difference increased with energy dose. This difference in DG might have been due to homopolymer that was removed more effectively from the surface and in the pores by the better wetting or swelling ability of ethanol. The DG of membranes cleaned with water (Fig. 3) was much higher than the DG obtained previously with 300 nm lamps and a benzene filter [10]. Those membranes were prepared for FTIR by soaking in ethanol for 10–30 min prior placement in the vacuum oven overnight. This ethanol treatment might have had the same effect on the DG as washing the membranes in ethanol. The use of the chain transfer agent, 2-mercaptoethanol (ME), caused the DG to decrease significantly. A concentration of 10 mM ME decreased the DG by 37–60% depending on the energy dose used. In addition, there was almost no change in the DG up to an energy dose of approximately 350 mJ/cm2 ,

Fig. 3. The effect of irradiation energy dose on the degree of NVP grafting during dip-modification in the presence of chain transfer agents. Base 50 kDa PES membrane (Lot #T0238A) was dip-modified in an aqueous solution of 5 wt.% NVP with or without 2-mercaptoethanol (ME) using 300 nm lamps and a benzene filter and then cleaned in water or ethanol: (䊊) 0 mM ME, cleaned in water; (䊉) 0 mM ME, cleaned in ethanol; (䉱) 10 mM ME, cleaned in ethanol; (䊏) 50 mM ME, cleaned in ethanol.

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

9

Fig. 4. The surface wettability of membranes as measured by cos θ vs. degree of grafting. All grafted membranes were dip-modified in an aqueous solution of 5 wt.% NVP using 300 nm lamps and a benzene filter: (䊐, 䊊) base 50 kDa PES membrane Lots #9047G and 9140G; (䉫) base 50 kDa PES membrane Lot #9140G dipped in 100% ethanol; (䉱) unmodified 10 kDa PES membrane; () 50 kDa regenerated cellulose membrane; (䊏) dip-modified 50 kDa PES membrane (Lot #9047G); base 50 kDa PES membrane (Lot #9140G) dip-modified using; (䊉) 10 mM ME; ( ) 50 mM ME. The measurement error was ±3◦ .

indicating that grafting was practically inhibited at low energy dose. When 50 mM ME was used, grafting was inhibited up to an energy dose of approximately 700 mJ/cm2 and the DG decreased by 60–74% at higher energy doses. Higher ME concentrations (>50 mM) would probably almost completely inhibit grafting in the range of irradiation energy doses used (150–1776 mJ/cm2 ). Since it is likely that approximately the same number radical sites were formed at the same irradiation energy dose, it is speculated that the lower DG achieved when using the chain transfer agent was caused by a lower average grafted chain length. However, it is possible that the lower DG was also the result of a lower number of grafted chains. 3.2.2. Wettability measurements The cos θ values of the static captive air bubble contact angles of the 50 kDa base membrane and

dip-modified membranes using 300 nm lamps and a benzene filter with and without 2-mercaptoethanol are shown in Fig. 4. The wettability of membranes dip-modified in the presence of 10 and 50 mM ME increased from cos θ = 0.67±0.04 to values similar to those of membranes dip-modified without ME present (cos θ = 0.70–0.87). The wettability increased with increasing DG and a high wettability was achieved at a modest DG (∼0.7). The most hydrophilic modified membranes (cos θ = 0.87 ± 0.02) had a slightly lower wettability than regenerated cellulose (0.91 ± 0.01), but were significantly more wettable than unmodified PSf membranes (0.62 ± 0.02). The two 50 kDa PES membrane lots used (Lot #9047G and 9140G) exhibited very dissimilar wettability (0.76 ± 0.04 and 0.67 ± 0.04). It is unclear why there was a large variation, but it is possible that this was the result of differences in surface roughness [26]. Treating the base

10

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

50 kDa PES membrane (Lot #9140G) with ethanol did not significantly change the wettability (0.69 ± 0.05 versus 0.67 ± 0.04). Overall, the high hydrophilicity of the membranes modified with 2-mercaptoethanol confirmed the hypothesis that chain transfer agents could be used to potentially shorten the length of grafted chains without affecting the imparted surface wettability. 3.2.3. BSA ultrafiltration experiments The results of the BSA filtration experiments conducted on dip-modified membranes are shown in Table 3 and Figs. 5 and 6. Membranes dip-modified and cleaned in water lost a large fraction of their permeability due to pore blockage by grafted chains and possibly by the presence of non-grafted homopolymer. Similar to previous results [10], the observed rejection remained unchanged even using an energy dose as high as of 324 mJ/cm2 (Fig. 5). The irreversible flux loss decreased significantly with DG

11

from 0.393 to −0.078 (Fig. 6), indicating that irreversible adsorptive fouling was reduced by the high hydrophilicity of these membranes (Fig. 4). The total flux loss also decreased significantly (0.694–0.144). While this was in part due to the lower irreversible fouling, it was mainly caused by lower reversible flux loss caused by the lower membrane permeability (0.3–0.6 lm h/kPa compared with 5.6 lm h/kPa). Modification followed by ethanol cleaning caused the permeability to increase substantially (between three and five times higher). However, the observed BSA rejection decreased severely from 99 to 64% at an energy dose of 324 mJ/cm2 . This corresponded to an increase in membrane MWCO from 50 to 70 kDa (Fig. 2). Energy doses higher than 528 mJ/cm2 caused the observed membrane rejection to decrease more sharply, to approximately 3% or a pore size of >100 kDa. This indicated that irradiation using the 300 nm lamps and a benzene filter enlarged the pores, despite the unchanged rejection after modification and cleaning

Fig. 5. The observed BSA rejection of modified membranes during the filtration of 0.1 wt.% BSA as a function of irradiation energy dose. Base 50 kDa PES membrane (Lot #T0238A) was dip-modified in an aqueous solution of 5 wt.% NVP with or without 2-mercaptoethanol (ME) using 300 nm lamps and a benzene filter and then cleaned in water or ethanol: (䉬) base 50 kDa PES membrane; (䉫) base membrane dipped in 100% ethanol; (䊊) 0 mM ME, cleaned in water; (䊉) 0 mM ME, cleaned in ethanol; (䉱) 10 mM ME, cleaned in ethanol; (䊏) 50 mM ME, cleaned in ethanol. A solution of 0.1 wt.% BSA in 10 mM PBS at pH 7.4 and 22 ◦ C was used.

12

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

Fig. 6. The irreversible flux loss of modified membranes during the filtration of 0.1 wt.% BSA as a function of irradiation energy dose. Base 50 kDa PES membrane (Lot #T0238A) was dip-modified in an aqueous solution of 5 wt.% NVP with or without 2-mercaptoethanol (ME) using 300 nm lamps and a benzene filter and then cleaned in water or ethanol: (䉬) base 50 kDa PES membrane; (䉫) base membrane dipped in 100% ethanol; (䊊) 0 mM ME, cleaned in water; (䊉) 0 mM ME, cleaned in ethanol; (䉱) 10 mM ME, cleaned in ethanol; (䊐) 50 mM ME, cleaned in ethanol. A solution of 0.1 wt.% BSA in 10 mM PBS at pH 7.4 and 22 ◦ C was used.

with water. Cleaning the unmodified membrane with ethanol increased permeability by 55%, but had no observed effect on the protein rejection (Table 3, Fig. 5). The increased membrane permeability indicated that the ethanol was better wetting agent than water and might swell the pore structure slightly, but not significantly enough to cause an observable decrease in BSA rejection. This suggested that homopolymer was formed during dip-modification and plugged the enlarged pores, but water cleaning was not sufficient to remove it. As a result, the permeability decreased and the observed rejection remained unchanged. Ethanol cleaning might have better wetted or even swelled the membrane pore structure and resulted in removal of the homopolymer which blocked the enlarged pores. The decrease in DG after ethanol washing (Fig. 3) supports this hypothesis. The decrease in observed

BSA rejection after dip-modification and ethanol cleaning (99–3% (approximately) at an energy dose of 1776 mJ/cm2 ) was more severe than after irradiation alone and ethanol cleaning (99–80% at an energy dose of 1776 mJ/cm2 , Table 1). This indicates greater pore enlargement after dip-modification. This could be caused by the enhanced wetting or possible swelling of the PES membrane porous structure by the monomer NVP. However, this is probably not due to the monomer alone because dipping the membrane in NVP without irradiation caused the permeability to increase by 49% without loss of observed rejection (Table 3). This was similar to the effect of ethanol treatment. It is speculated that irradiation of the membrane in a highly wetted or swollen state increased the pore size. However, the exact mechanism is not yet known. Irradiation of the membrane while immersed

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

in a 50% aqueous solution of ethanol was previously shown to cause greater pore enlargement than when irradiated in water [10]. Considering that both solvents caused a similar effect on the membrane porous structure (Table 3), the same mechanism may explain the effect of irradiating the membrane in the presence of these solvents. The total and irreversible flux loss (Table 3 and Fig. 6) decreased with increasing DG, but because the observed rejection also decreased it is difficult to compare the filtration performance to the highly retentive (>90%) control membranes (unmodified PES and regenerated cellulose membranes). In addition, it is not accurate to compare the filtration performance of membranes with similar low rejections (<90%) because the exact pore size or MWCO are not known. These membranes may however, be attractive for applications in which protein transmission is desired. Dip-modification with 2-mercaptoethanol followed by ethanol cleaning produced more permeable, but less retentive membranes. This was probably the result of the potentially lower average grafted chain length or grafted chain density as indicated by the lower DG achieved (Table 3 or Fig. 3). The initial buffer permeability (Lp0 ) of membranes dip-modified with 10 mM 2-mercaptoethanol increased with irradiation energy up to an energy of 528 mJ/cm2 (8.8–13.2 lm h/kPa), then decreased to a value (7.1– 7.8 lm h/kPa) slightly below that of the unmodified membrane (8.8 lm h/kPa). The permeability increase at lower energy doses was the result of a combination of pore enlargement and a low DG, while the decrease at higher energy dose was probably the result of the effect of the high DG achieved (>1). As previously speculated [9], a layer of chains might have been formed which blocked the pores as seen previously. The observed membrane rejection decreased more sharply (99–2% up to an energy dose of 528 mJ/cm2 ) than modified membranes without 2-mercaptoethanol present. It is speculated that membranes modified in the absence of 2-mercaptoethanol had a high surface chain density and long grafted chain length help to lessen the loss of rejection by blocking pores. If the use of a chain transfer agent decreased the surface chain density and shortened the average grafted chain length, the rejection would decrease more severely. The use of a higher 2-mercaptoethanol concentration (50 mM) significantly lowered the DG (Fig. 3),

13

increased membrane permeability, and had the same effect on rejection as with 10 mM ME. The permeability increased from 8.8 to 26.7 lm h/kPa at an energy dose of 1057 mJ/cm2 , then decreased to 18.8 lm h/kPa at high energy dose (1057 mJ/cm2 ) when the DG reached a high value (∼1). The total and irreversible flux loss of membranes modified with 10 mM 2-mercaptoethanol decreased significantly (0.54–0.01) with increased DG. However, since the membrane observed BSA rejection was considerably lower than the unmodified membrane, it would not be accurate to compare the flux loss of these membranes. Under these circumstances, it is not possible to uncouple the beneficial effect that surface hydrophilicity and low observed rejection each had on the flux loss. The MWCO of these membranes must first be accurately measured so that membrane filtration performance can be compared on an equal basis. Nonetheless, given their high surface hydrophilicity, high permeability, and low irreversible fouling, these membranes may be attractive for the filtration of larger proteins and for applications in which protein transmission is desired. 3.3. Invertase ultrafiltration experiments The results of the invertase filtration experiments using dip-modified membranes are shown in Table 4. Invertase was chosen because its higher molecular weight (270 kDa) could allow the MWCO of the modified membranes to be more precisely determined. The unmodified 50 kDa PES membrane exhibited much higher total flux loss (87%) and irreversible flux loss (51%) when filtering INV than BSA (69 and 39%, respectively, Table 3). The regenerated cellulose membrane exhibited similar total flux loss (85%), but lower irreversible flux loss (13%) during INV filtration than the unmodified PES membrane. This indicted that high membrane surface hydrophilicity could effectively combat the greater fouling tendency of this protein. The observed rejection of INV (95.6%) for the unmodified 50 kDa PES membrane was lower than that of BSA (99.1%) despite its higher molecular weight. This could be the consequence of using different membrane lots or that INV might have had some lower molecular weight contaminants, such as ␣-galactosidase (96 kDa), which passed through the membrane.

14

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

Increasing the DG (0.226–0.859) by dip-modification followed by ethanol cleaning decreased the total flux loss slightly (0.869–0.805), but significantly decreased the irreversible flux loss (0.508–0.309) while the observed rejection remained constant (95.6–95.0). The larger MW of INV probably explains why its rejection remained constant despite the increase in pore size after modification as indicated by the loss of BSA rejection (97.1–63.6%). At high DG, the rejection sharply decreased as observed during BSA filtration, indicating significant pore enlargement. The use of INV to test the membrane dip-modified using 2-mercaptoethanol corroborated the membrane MWCO estimated by BSA. The MWCO of all membranes dip-modified with or without chain transfer agent was between 50 and 70 kDa below an energy dose of 528 mJ/cm2 . At higher energy doses, the MWCO was above 100 kDa. Despite the use of a second, higher molecular weight protein in the calibration curve (Fig. 2), a more precise estimate of the MWCO was not possible because there was no membrane available with a MWCO between 100 and 300 kDa. Thus, it was not possible to determine the lowest membrane MWCO at which INV rejection was close to 0%. Due to the sharp loss of INV rejection with increasing irradiation energy dose and a lack of precise MWCO data, the total and irreversible flux loss of the modified and control membranes could not be compared on an equal basis. However, their increased surface hydrophilicity and low irreversible fouling by INV indicates that these membranes may be suitable for high protein transmission applications.

4. Conclusions The use of the chain transfer agent 2-mercaptoethanol and ethanol cleaning was successful in producing hydrophilic membranes with a permeability considerably higher than that of the unmodified PES, regenerated cellulose membranes, and membranes dip-modified without chain transfer agent and cleaned with water. From the lower DG achieved, it was inferred that the graft chain density and chain length decreased with increasing concentration of chain transfer agent. However, the low observed rejection of these membranes made it difficult to evaluate the filtration performance. A method to estimate the

15

changes in membrane pore size after modification was developed using a calibration curve of observed rejection versus membrane MWCO based on two different size proteins and a series of MWCO PES membranes. The MWCO of these membranes was estimated between 50 and 100 kDa. While the method proved useful for estimating the membrane MWCO between 50 and 100 kDa, it did not have the necessary sensitivity at higher MWCO. This emphasized the need to develop a more precise method to determine the MWCO of the modified membranes as this is crucial to both establishing the ultimate application of these larger pore size membranes and the proper evaluation of their filtration performance. Only by knowing the MWCO of the modified membranes can adequate controls be chosen to accurately evaluate their filtration performance. Nevertheless, it is very likely that the high hydrophilicity, high permeability, and low fouling character of these membranes can be utilized in the filtration of larger proteins (>500 kDa) or in applications in which high protein transmission is desired. Dip-modification was shown to cause considerable pore enlargement. This was not previously observed because it was thought that while the pores were enlarged by irradiation, they were effectively blocked by non-grafted homopolymer formed during the photochemical modification that was not removed by simple water cleaning. Ethanol was able to remove the homopolymer because it wetted and swelled the membrane more effectively. This indicates that it is not possible to dip-modify 50 kDa PES UF membranes using 300 nm lamps and a benzene filter without causing significant change to the pore structure. However, the observed BSA rejection of these membranes might be stable despite the pore enlargement since the observed BSA rejection of the dip-modified membranes cleaned in water was higher than that of the unmodified PES membrane after filtration using transmembrane pressures as high as 55 psig (379 kPa). This suggests that the homopolymer might be trapped in the pores unless the membranes are washed with a swelling agent. Therefore, dip-modified membranes might still potentially be used for high retention filtration if it can be proved that leaching of the homopolymer from the pores is negligible during filtration and after caustic cleaning. The ability of the monomer NVP to swell the PES membrane may be

16

J. Pieracci et al. / Journal of Membrane Science 202 (2002) 1–16

chiefly responsible for this severe observed protein rejection loss because irradiation alone caused only a moderate loss of observed rejection. NVP is a weak solvent for PES and might help to promote swelling and pore enlargement. This suggests that the selection of hydrophilic vinyl monomers that wet or swell PES less than NVP might help to maintain more of the pore structure after modification. Acknowledgements We acknowledge the support of the National Science Foundation (Grant No. CTS-94-00610) and the US Department of Energy (Grant No. DE-FG0290ER14114). The authors thank Dr. Barry Breslau and Dr. Michael Heath and Pall-Filtron Corp. for supplying the PES and regenerated cellulose membranes. References [1] M. Ulbricht, A. Oechel, J. Lehmann, G. Tomaschewski, H. Hicke, J. Appl. Polym. Sci. 55 (1995) 1707. [2] M. Ulbricht, H. Matuschewski, A. Oechel, H. Hicke, J. Membr. Sci. 115 (1996) 31. [3] M. Ulbricht, K. Richau, H. Kamusewitz, Coll. Surf. 138 (1998) 353. [4] V. Thom, M. Ulbricht, V. Kops, G. Jonsson, Macromol. Chem. Phys. 199 (1998) 2723. [5] H. Yamagishi, J. Crivello, G. Belfort, J. Membr. Sci. 105 (1995) 237. [6] H. Yamagishi, J. Crivello, G. Belfort, J. Membr. Sci. 105 (1995) 249.

[7] J. Crivello, Y. Yamagishi, G. Belfort, US Patent Number 5,468,390 (1995). [8] J. Pieracci, J. Crivello, G. Belfort, J. Membr. Sci. 156 (1999) 223. [9] J. Pieracci, D. Wood, J. Crivello, G. Belfort, Chem. Mater. 12 (2000) 2123–2133. [10] J. Pieracci, J. Crivello, G. Belfort, Chem. Mater., submitted for publication. [11] B. Kaeslev, J. Pieracci, G. Belfort, J. Membr. Sci., in press. [12] Y. Ito, S. Kotera, M. Inaba, K. Kono, Y. Imanishi, Polymer 31 (1990) 2157. [13] Y. Ito, M. Inaba, D. Chung, Y. Imanishi, Macromolecules 25 (1992) 7313. [14] Y. Ito, Y. Ochiai, Y. Park, Y. Imanishi, J. Am. Chem. Soc. 119 (1997) 1619. [15] M. Stevens, Polymer Chemistry, 2nd Edition, Oxford University Press, New York, 1990. [16] J. Brandrup, E. Immergut, Polymer Handbook, 2nd Edition, Vol. II., Intersciences, New York, 1966, pp. 126– 133. [17] M. Nystrom, P. Jarvinen, J. Membr. Sci. 60 (1991) 275. [18] Contact Sampler User’s Manual for Model #0012-490(T) Nicolet Magna-IR Spectrophotometer, SpectraTech, Inc., 1995. [19] W. Zhang, B. Hallstrom, Desalination 79 (1990) 1. [20] G. Sigal, M. Mrksich, G. Whitesides, J. Am. Chem. Soc. 120 (1998) 3464. [21] T. Creighton, Proteins: Structures and Molecular Properties, Freeman, New York, 1984. [22] D. Cavaille, D. Combes, J. Biotechnol. 43 (1995) 221. [23] P. Righetti, T. Caravaggio, J. Chrom. 127 (1976) 1. [24] M. Meireles, P. Aimar, V. Sanchez, J. Membr. Sci. 56 (1991) 13. [25] P. Aimar, C. Taddei, J. Lafaille, J. Sanchez, J. Membr. Sci. 38 (1988) 203. [26] M. Taniguchi, J. Pieracci, G. Belfort, Langmuir, 17 (2001) 4312.