Journal of Membrane Science 198 (2002) 75–85 Performance of ultrafiltration membranes in ethanol–water solutions: effect of membrane conditioning Rishi Shukla1 , Munir Cheryan∗ Agricultural Bioprocess Laboratory, Department of Food Science, University of Illinois, 1302 West Pennsylvania Avenue, Urbana, IL 61801, USA Received 29 May 2001; received in revised form 15 August 2001; accepted 20 August 2001 Abstract Several commercial polymeric ultrafiltration membranes were screened for their performance with aqueous ethanol solutions. The method of conditioning the membrane has a major effect on solvent flux, membrane integrity and their pressure ratings. Gradual solvent exchange with successively higher concentrations increased in small doses appears to work best with completely miscible solvents such as those studied here (ethanol–water mixtures). Rapid solvent exchange between water and high concentrations of alcohol disrupts the polymer matrix in many cases. The Darcy model was used to correlate the data and it indicated that viscosity differences of the ethanol solutions could account for part of the variations in solvent flux with some membranes. Exposure to organic solvents significantly reduces the pressure rating of the membranes. Several membranes that provided acceptable rejection of ethanol-soluble proteins at low pressures (138 kPa, 20 psi) lose its properties at higher pressures (413 kPa, 60 psi) if conditioned incorrectly, and vice versa. © 2002 Elsevier Science B.V. All rights reserved. Keywords: Ultrafiltration; Organic separations; Ethanol; Protein 1. Introduction Organic solvents are being increasingly used in extraction, purification and processing of pharmaceuticals, food, nutraceuticals and flavor compounds. Membrane technology would be the method of choice for separating the desired compounds and for recycling the solvent. However, almost all current applications of membrane technology are with aqueous systems. The few nonaqueous applications discussed in the literature usually deal with streams containing organic compounds with concentrations only up to a ∗ Corresponding author. Fax: +1-217-244-2455. E-mail address: [email protected] (M. Cheryan). 1 Present address: James R. Randall Research Center, Archer Daniels Midland Co., Decatur, IL, USA. few thousand of parts per million such as oily waste streams and cleaning solvents [1]. There are very few examples of membrane applications with feeds containing 50–100% organic solvents. In addition, much of the membrane work to date with organic solvents has been with nanofiltration membranes, e.g. in the vegetable oil industry for degumming, deacidification and solvent recovery [2–4]. There are fewer reports on the use of ultrafiltration (UF) membranes with relatively high concentrations of organic solvents. In one of the earliest reports on this subject, Nguyen et al. [5] observed that, in the absence of solutes, membrane permeability of several commercial UF membranes increased with some solvents (ethanol, methanol) and decreased with others (chloroform, decane, benzene). Lencki and Williams [6] studied 10,000 and 30,000 molecular weight cut-off 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 3 8 - X 76 R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 (MWCO) membranes with methanol, ethanol and acetonitrile. Solvents with solubility parameters similar to the membrane reportedly led to the greatest change in flow resistance, but solvents with a similar solubility parameter but low hydrogen bonding capabilities could disrupt the structure of anisotropic polysulfone membranes to such an extent that a dramatic drop in flow resistance is observed. Jaffrin and Charrier [7] observed an 80% drop in flux in 40% ethanol compared to water in the manufacture of plasma proteins from human serum. The reduction in flux was explained by an increase in viscosity when ethanol was added to water and the formation of a thicker polarization gel layer due to lower back diffusion of the solutes in 40% ethanol. Gupta et al. [8] reported a 25–35% drop in UF flux when ethanol concentration was increased from 20 to 30% with albumin in mineral (Carbosep) membranes. Most polymers used in manufacturing membranes and/or their supports are first dissolved in organic solvents as part of their casting process. Consequently, they could swell or dissolve in the solvent, leading to unacceptable changes in solvent flux and solute separation. For example, of the 15 membranes screened by Koseoglu et al. [4] for vegetable oil processing, only 5 were found to be stable to hexane. Raman et al. [9,10] tested 13 membranes for deacidification of soybean oil of which only 6 were stable in methanol, and only one was compatible with hexane. Kuk et al. [11] reported significant degradation in membrane performance on exposure to aqueous and anhydrous ethanol solutions with cellulose acetate and composite reverse osmosis (RO) membranes with vegetable oil–solvent mixtures. Koike et al. [12] screened 18 membranes (including some developed for gas separations) for separation of fatty acids and glycerides from lipase hydrolysates of high oleic sunflower oil. Cellulosic membranes gave good results but suffered from poor long-term stability. Zwijnenberg et al. [13] used prototype polyamide and cellulose-based membranes for deacidification of vegetable oils in acetone. This paper reports on our studies on polymeric UF membranes with ethanol–water solutions. The ultimate goal was to develop a process for the manufacture of zein, which is a hydrophobic, ethanol-soluble protein in corn with a molecular weight of ∼22,000. Zein has a variety of industrial uses, from fibers and adhesives to chewing gums and biodegradable plastics [14]. The commercial application of this natural polymer is limited by its high manufacturing cost, due primarily to the high cost of separating and purifying the zein from the ethanol extract and recovering the ethanol solvent. Ultrafiltration could be used readily to recover and purify the zein while simultaneously recycling the ethanol solvent [15]. The best solvent for extracting zein from corn is 70% (v/v) ethanol [16] and thus our studies were limited to a maximum ethanol concentration of 70% (v/v). This paper specifically focuses on the effect of conditioning on solvent flux and rejection of zein. 2. Materials and methods 2.1. Membrane screening Ultrafiltration membranes were obtained from several manufacturers and are listed in Table 1. Flat sheets were evaluated in an Amicon dead-end stirred cell (model 502) which used 62 mm membrane discs of area 28.7 cm2 . Pressure was provided by a nitrogen gas cylinder and turbulence was created by a magnetic stirrer which was operated at 300 rpm. The hollow fibers were tested as-is in their housings in a recycle system. A pump provided pressure and the required cross-flow for turbulence. All experiments were conducted at room temperature (24 ± 2◦ C). 2.2. Ethanol solutions Ethanol (anhydrous, 200 proof) was obtained from McCormick Distillation Co., Weston, MO. Aqueous solutions of ethanol (EtOH) were prepared on a volume/volume basis. Deionized water was used for all experiments. Ethanol and deionized water were microfiltered through a 0.2 m filter before use. Viscosity of ethanol solutions (with and without protein) were measured using Ubbelohde viscometers (model OC with an instrument coefficient of 0.002752 cSt/s and model 1 with a coefficient of 0.00963 cSt/s) obtained from Cannon Instrument Co., State College, PA. Densities of the solutions were measured using precalibrated pycnometers. Viscosity measurements were performed in triplicate while density data are the mean of five observations. R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 77 Table 1 Ultrafiltration membranes selected for screening studies Materiala Membrane MWCOb Manufacturerc Configurationd Cellulose ester Cellulose acetate Regen. cellulose Regen. cellulose Compositee Compositee Compositee PAN-m PAN-based PES-m PES-m PES PES PS PS PS PS PVDF Type C Cell PLGC YM10 U20S G80 H051 MX25 U20T Alpha Omega UFC PES4H UFP10 PS10 PM10 PM30 AN09 10000 10000 10000 10000 20000 5000 – 25000 20000 10000 10000 10000 10000 10000 10000 10000 30000 25000 Spectrum-Microgon Pall Filtron Millipore Millipore Koch Osmonics Osmonics Osmonics Koch Pall Filtron Pall Filtron Hoechst Hoechst A/G Technology Sartorius Koch Millipore Osmonics FS FS FS FS FS FS FS FS FS FS FS FS FS HF FS HF FS FS a PAN: polyacrylonitrile, PES: polyethersulfone, PS: polysulfone, PVDF: polyvinylidine fluoride, Regen.: regenerated, m: modified. Molecular weight cut-off from manufacturers’ specifications. c A/G Technology, Needham, MA; Hoechst, Wiesbaden, Germany (through US Tech., Cincinnati, OH); Koch Membrane Systems, Wilmington, MA; Millipore, Bedford, MA; Osmonics, Minnetonka, MN; Spectrum-Microgon, Laguna Hills, CA; Pall Filtron, Northborough, MA; Sartorius, Edgewood, NY. d FS: flat sheet, HF: hollow fiber. e Composition is proprietary. b 2.3. Membrane conditioning Most membranes (except Koch U20S) were received from the manufacturer preserved with glycerol or a similar humectant. Prior to use, the membranes were conditioned by gradual exposure to solvents using the following protocol: 1. In each case, the membrane was thoroughly rinsed with large amounts of the solvent that the membrane first came in contact with (depending on the method of conditioning as discussed below). 2. The membranes were then soaked in the solvent for a period of at least 6 h (overnight in most cases) and then fitted in to the appropriate membrane test cell. The Amicon cell was then filled with 200 ml of the conditioning solvent (for the hollow fibers, the solvent was placed in a separate feed tank connected to the module via a pump). 3. The system was pressurized to 69 kPa (10 psi) in all cases except with the U20S and G80 membranes which were pressurized to 138 kPa (20 psi), and 4. 5. 6. 7. 8. the hollow fibers which were pressurized to 35 kPa (5 psi). At least 25% of the initial solvent volume was allowed to permeate through the membrane. Flux was then measured. The pressure was then raised to the next higher level (which varied for each membrane as reported later) and steps 4 and 5 repeated. The pressures tested in each case depended on the pressure rating of the membrane. In most cases, flat sheet polymeric membranes were tested up to 413 kPa (60 psi) and hollow fibers up to 138 kPa (20 psi). After the tests, the membrane was removed from the cell and stored in a petri dish soaked in the appropriate solvent for at least 6 h at 24◦ C and the flux measurements (steps 4–6) repeated. The membrane was then conditioned in the next consecutive solvent, which depended on the method of conditioning, as discussed below. Four methods of membrane conditioning were evaluated. The upper ethanol concentration was limited to 78 R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 70% ethanol in most of our studies since this was the optimum concentration for the extraction of zein from corn [16]. Method 1: Gradual change from 0 to 70% ethanol. The membrane was initially exposed to water and then conditioned to 70% ethanol in increments of 10% ethanol concentration (i.e. experiments were performed with water, 10% ethanol, 20, 30, 40, 50, 60 and 70% ethanol). At each stage, steps 2–7 described above were repeated. Method 2: Direct change from 0 to 70% ethanol. The membrane was first exposed to water through steps 2–7 and then solvent-exchanged directly with 70% ethanol without any intermediate ethanol concentrations. Method 3: 70% ethanol. The membrane was exposed directly to 70% ethanol using steps 2–7. Method 4: 100 ethanol to 70% ethanol. The membrane was initially exposed to 100% ethanol and then conditioned to 70% ethanol in steps of 10% ethanol concentration (i.e. experiments were performed for 100, 90, 80 and 70% ethanol). At the end of each conditioning experiment, the membranes were tested for flux and zein rejection with a model protein solution using the method described below. This would provide an indication of the effect of membrane conditioning on membrane integrity and performance. Conditioning data were expressed in terms of a convective transport model: J = LP PT µ (1) where J is the flux, PT the transmembrane pressure, µ the viscosity of the permeate and LP is the permeability coefficient of the membrane. Membranes not affected by the solvent should give a linear plot of flux versus 1/µ. In cases where the solvent does have an effect on the membrane (e.g. swelling of polymer, dilating of pores), the plot will not be linear. All polymeric membranes listed in Table 1 were first tested using method 1 for solvent stability. Membranes that gave linear or nearly linear plots were then evaluated with methods 2–4. In order to include pressure effects, the data are also plotted in terms of relative resistance (R/Rw ), where R is the membrane resistance in the presence of ethanol solutions, and Rw is the membrane resistance mea- sured with pure water. From Eq. (1), the membrane resistance (R or Rw ) can be calculated as follows: R= 1 PT = LP µJ (2) 2.4. Membrane screening with zein solutions Flux and rejection of an ethanol-soluble protein was measured using a solution of 5 g/l zein (F4000, Freeman Industries, Tuckahoe, NY) made up in 70% (v/v) aqueous ethanol. For each experiment, the preconditioned membrane was placed in the Amicon stirred cell with 250 ml of the zein solution. The cell was pressurized and flux measurement started. Flux measurement continued at each pressure until at least three consecutive values were constant. This is reported as the steady-state flux. Permeate samples (approximately 2–3 g each) were collected and analyzed for protein content by the Kjeldahl method. Rejection is defined as: CP R = 1− × 100 (3) CR where CP and CR are the concentrations of zein in permeate and retentate, respectively. Zein concentrations for the model solutions were measured spectrophotometrically by the procedure of Craine et al. [17]. All experiments were repeated within 24 h with a fresh membrane disc. The membranes were cleaned after the zein experiments by soaking in a solution of 5 g/l NaOH in 70% ethanol for periods up to 6 h. The membrane was then thoroughly rinsed with several volumes of fresh 70% ethanol. 2.5. Static swelling Membrane swelling was determined as an increase in weight due to solvent absorption (Sw ) (%): Ww − Wd Sw = × 100 (4) Wd where Wd and Ww are the weights of dry and wet membrane samples, respectively. The wet weight, Ww , was obtained by incubating membranes at 24◦ C for up to 24 h in the appropriate solvent. All swelling experiments were replicated with at least five samples of each membrane and the mean value is reported. R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 79 3. Results and discussion 3.1. Membrane conditioning Figs. 1–4 show typical flux behavior as a function of ethanol concentration in the solvent and transmembrane pressure when membranes were conditioned using method 1. Many membranes showed a minimum in flux at 25–50% (v/v) ethanol (0.12–0.26 M ethanol) and the flux was higher at higher pressures. Due to space limitations, data for all combinations of membranes and pressures are not shown here. The complete data set for all membranes studied is available in Shukla [18]. Each data point in Figs. 1–4 was repeated within 24 h with a fresh membrane disc. Reproducibility of flux data was high (±5%) for membranes that were minimally or not affected by ethanol solutions, while reproducibility was poor for those membranes affected by the solvent. As a group, the composite Fig. 2. Effect of ethanol concentration on flux of polysulfone and polyethersulfone membranes. Membranes were conditioned using method 1. Pressure was 275 kPa except for PM10 membrane, which was 69 kPa. membranes (G80, H051 and U20S) shown in Fig. 3 displayed lower solvent fluxes of <50 l/m2 h (LMH) than most other membranes. Among the polysulfone and polyethersulfone family of membranes, the PES4H had much lower flux (5–15 LMH) than others Fig. 1. Effect of ethanol concentration and transmembrane pressure on flux of a modified PAN and cellulose ester membrane. Membranes were conditioned using method 1. Fig. 3. Effect of ethanol concentration and transmembrane pressure on flux of composites, polysulfone and polyethersulfone membranes. Membranes were conditioned using method 1. Pressure was 275 kPa except for UFP10 membrane, which was 124 kPa ((䊉) UFC; (䊐) UFP10; (䉬) PES4H; () G80; (䊏) H051; (䊊) U20S). 80 R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 Fig. 4. Effect of ethanol concentration on flux of cellulose-based and PAN-based membranes. Membranes were conditioned using method 1. Pressure was 275 kPa. (Figs. 2 and 3). This could be due to differences in hydrophobicity of the membrane materials, or due to differences in density of the membrane top layer which results in higher membrane resistance and lower permeability. Surfaced-modified PES membranes that are usually cross-linked after membrane formation displayed lower solvent fluxes. Since the viscosity of ethanol solutions is maximum at 50% ethanol concentration (Fig. 5), the flux should show a corresponding minimum value at this ethanol concentration, assuming the membrane is not affected by the solvent in any other manner. This phenomenon Fig. 5. Effect of ethanol concentration on viscosity and density of ethanol solutions at 24◦ C. Fig. 6. Darcy plot for a modified PAN and a PES membrane. Effect of pressure and ethanol concentration (expressed in terms of viscosity) on flux. Membranes were conditioned using method 1. (MX25: (䊏) 138 kPa, (䊉) 275 kPa, (䉱) 413 kPa; PES4H: (䊐) 138 kPa, (䊊) 275 kPa, () 413 kPa). is governed by the Darcy or Hagen–Pouseuille laws shown in Eq. (1). Several membranes do show a flux minima in the middle range of ethanol concentrations (Figs. 1–4). Thus, a linear correlation should be obtained between flux and the reciprocal of viscosity of the permeating solvent, according to Eq. (1). A typical Darcy plot for two membranes at different pressures is shown in Fig. 6. Of the 18 membranes tested, 15 generally followed Darcy’s law in that flux generally decreased with increasing viscosity of the ethanol solutions [18]. The three exceptions were the composite membranes (G80, U20S and H051). This phenomenon could be explained by viscosity considerations alone, in a manner similar to that observed by others [6,19,20]. Our results are similar to those of Machado et al. [21] but in contrast to those of Iwama and Kazuse [20]. The latter group used binary solvent mixtures that followed the viscosity additive rule, whereas the solvents used by us (aqueous ethanol solutions) and by Machado et al. [21] (aqueous acetone solutions) do not follow the viscosity additivity rule. Viscosity of acetone–water mixtures also go through a maximum at ∼0.15 M acetone. However, acetone–water flux did not go through a minimum at this acetone–water concentration, indicating that solvent transport is also affected by other parameters such as surface tension and solubility parameter of the permeating solvent. R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 Fig. 7. Resistance ratio plots for polysulfone and polyethersulfone membranes. Membranes were conditioned using method 1. ((䊏) PM10, (䊊) Omega, () UFC, (䊐) PM30, (䉫) PES4H, (䊉) Alpha, (䉱) PS10). 81 Fig. 9. Resistance ratio plots for hydrophilic polyacrylontrile-based and cellulose-based membranes. Membranes were conditioned using method 1. 3.2. Membrane resistance The influence of different water fluxes of the membranes was factored out by using the resistance ratio (R/Rw ) to compare the membranes, as shown in Figs. 7–9. As a group, the more hydrophobic membranes, such as PS and PES (Fig. 7) and the composite membranes (Fig. 8) show a decrease in R/Rw Fig. 8. Resistance ratio plots for composite membranes. Membranes were conditioned using method 1. with increase in ethanol concentration. The exceptions in this group are the PM10 and PES4H membranes: their resistance ratios were about 1 but increased at higher ethanol concentrations (Fig. 7). However, the more hydrophilic membranes shown in Fig. 9 did not behave as a group. Except for the U20T and PLGC, the others in this group also showed a decrease in membrane resistance (or the resistance ratio remained close to 1) with an increase in ethanol concentration. This suggests that factors other than, or in addition to, hydrophilicity and viscosity (e.g. surface tension, solubility parameter or dielectric constant in the case of charged membranes) may be governing solvent flow through the membrane. With the composite membranes (G80, H051 and U20T), resistance was high with water and decreased when exposed to ethanol (Fig. 8). The pores on the surface of G80 and U20S membranes, which have very dense barrier layers, apparently dilated leading to lower membrane resistance and higher fluxes. This behavior does not appear to be reported in the literature. With cross-linked PAN-chitosan membranes, Musale and Kumar [22] observed the highest flux with methanol followed by ethanol and isopropanol. These differences in flux were explained on the basis of the combined effects of increase in 82 R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 molecular weight, viscosity, hydrophobicity and dielectric constant of the alcohol. Reddy et al. [23] observed lower flux of primary and secondary alcohols with increase in molecular weight and hydrophobicity of the solvent with a polyamide-polyphenylene sulfone membrane. There was no correlation with viscosity. Only Machado et al. [21] reported an increase in flux with increase in solvent (acetone) mole fraction. Their nonlinear exponential increase was thought to be due to decreasing surface tension at higher acetone concentrations. 3.3. Membrane swelling Swelling of several polymeric membranes is shown in Table 2. Except for two membranes (Alpha and Omega), water resulted in higher degrees of swelling by weight while 100% ethanol resulted in the least weight-swelling. The polysulfone hollow fiber UFP10 showed the greatest degree of swelling in water (211%) and in neat ethanol (94%). In general, the dielectric constant of the solvent and difference in solubility parameters of the polymer and solvent govern swelling behavior. Membrane swelling is higher in polar solvents like water due to its high dielectric constant (ε = 80), and less with ethanol (ε = 25) and nonpolar solvents like hexane (ε = 1.9). Musale and Kumar [22] observed over 50% swelling by weight with cross-linked chitosan-PAN membranes Table 2 Weight swelling of polymeric UF membranesa Material Membrane Water 70% EtOH 100% EtOH Cellulose Cellulose Composite Composite Composite PAN-m PAN PES-m PES-m PES PS PVDF PVDF Type C Cell U20S G80 H051 MX25 U20T Alpha Omega PES4H UFP10 AN09 AF5 112.5 25.9 83.7 65.8 31.0 71.2 85.8 44.7 51.0 26.0 211.4 54.3 21.0 60.2 29.9 48.4 61.5 26.4 56.3 39.1 59.9 46.5 18.0 161.4 46.2 9.5 45.9 22.5 34.6 51.3 24.5 46.8 32.5 52.8 56.3 17.1 94.4 34.6 6.1 a Percentage increase in weight after exposure to the solvents for 24 h. with ethanol compared to 20% swelling in hexane. It is interesting that the three membranes manufactured by Pall Filtron (Alpha, Cell and Omega) showed essentially no change or no trend in swelling, even though these membranes are made of two different materials. Length-swelling measurements have been reported in the literature and were also done in this study [18]. However, they were largely inconclusive because the dimensional changes were so small that measurements were difficult and imprecise. A typical sample of 5 cm length experienced a length-swelling of 0.1–0.2 cm. Consequently this experimental method had high errors and is not useful. 3.4. Effect of methods of conditioning on permeability The effect of different methods of conditioning is shown in Table 3 in terms of permeability coefficients of 70% ethanol. Some membranes (Cell, MX25, Alpha, Omega, PS10) show high LP values (>100 × 10−15 m) at all pressures. On the other hand, the composite membranes (U20S, G80, H051) and modified PES membrane (PES4H) had significantly lower LP values. This is expected since composites and surface-modified membranes have denser layers and consequently higher membrane resistance. AN09, which is a PVDF-based high MWCO membrane, displayed much higher fluxes when conditioned by methods 2–4. The reason for this is not clear since PVDF is resistant to ethanol. Protein rejection values for this membrane were poor when conditioned by methods 2–4 (Table 4), indicating that high solvent fluxes might be due to pore dilation, which also results in low protein rejection. If solvent flux increased in proportion to the pressure (i.e. if the value of LP was unaffected by pressure), we can assume the membrane integrity remains intact. Exposure to water followed by exposure to 70% ethanol (method 2) appears to cause some damage to some membranes, e.g. type C, Cell, G80 and PES4H, which show an increase in LP at higher pressure. The order of magnitude of the fluxes with method 2 are comparable to those observed with method 1, except at the highest pressure in the above cases. However, some membranes (PS10 and AN09) show a decrease in LP at higher pressure. R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 83 Table 3 Effect of methods of conditioning and pressure (kPa) on membrane permeability (LP ) with 70% ethanol solventa Membrane Method 1 Type C Cell PLGC U20Sb G80b H051 MX25 Alpha Omega PES4H PS10 AN09 a b Method 2 Method 3 Method 4 138 275 413 138 275 413 138 275 413 138 275 413 68 142 51 23 38 62 247 424 642 21 205 50 58 132 42 31 54 46 235 414 666 16 150 35 58 240 49 28 62 36 220 443 655 14 132 39 75 127 45 50 25 71 254 411 554 20 432 282 150 130 53 50 50 70 261 397 576 20 303 214 256 618 45 51 81 84 271 445 585 37 265 182 62 136 38 22 15 66 241 362 739 18 853 323 140 232 40 38 104 95 249 371 760 49 726 242 489 412 75 38 665 145 337 463 855 136 585 285 71 79 26 21 19 49 137 356 101 16 150 219 141 119 101 22 22 92 164 371 142 18 334 209 370 1112 278 42 21 428 179 655 483 126 445 214 LP values calculated from Eq. (1) (×10−15 m). Pressures were 275, 310 and 345 kPa. Direct exposure to 70% ethanol (method 3) seems to have the most destructive effect on membranes such as type C, Cell, PLGC, G80, H051, MX25, Omega, PS10 and PES4H, as measured by their higher LP values compared to method 1 and the increase in LP at higher pressure. Only AN09 and Alpha appear to remain intact. Method 4 (direct exposure to 100% ethanol) also appears to damage the polymer matrix in all cases except U20S, G80 and AN09. 3.5. Effect of conditioning methods on protein flux and rejection A better indicator of membrane damage is changes in rejection properties. This was studied using model zein solutions. As expected, LP values are generally lower with zein present (Table 4) than in its absence (Table 3), in some cases by a substantial margin, e.g. compare Alpha, Omega and MX25 in Tables 3 and 4. Table 4 Effect of methods of conditioning and pressure (kPa) on membrane permeability (×10−15 m) and rejection (%) of proteina Membrane Method 1 LP Type C Cell PLGC U20Sb G80b H051 MX25 Alpha Omega PES4H PS10 AN09 Method 2 Rejection LP Method 3 Rejection LP Method 4 Rejection LP 138 275 138 275 138 275 138 275 138 275 138 275 138 275 138 275 56 63 42 N.D.c 8 34 41 70 45 22 114 63 46 55 29 9 15 34 39 68 N.D. 23 123 52 92 99 94 N.D. 87 96 98 99 78 88 90 79 97 94 95 93 96 92 99 98 N.D. 95 86 51 21 24 12 91 59 6 18 20 20 7 23 20 81 127 71 726 136 21 82 82 73 40 300 112 91 92 97 18 31 91 95 93 92 79 48 80 94 72 96 11 20 96 58 96 94 95 4 58 56 39 23 14 15 9 26 35 32 10 153 28 199 32 25 17 14 25 24 83 54 13 138 397 96 94 94 78 100 72 96 98 98 91 25 87 55 93 99 80 93 36 97 53 56 90 23 0 68 76 59 18 5 15 51 97 63 20 616 74 89 428 40 16 9 166 87 109 41 19 539 70 93 69 96 98 90 94 84 88 87 84 15 81 99 9 98 87 79 10 52 26 88 96 19 62 The protein was 5 g/l zein in 70% ethanol solvent. LP values were calculated from Eq. (1) (×10−15 m). Pressures were 275 and 345 kPa. c N.D.: not determined. a b Rejection 84 R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 Among the methods, membranes conditioned by method 1 appear to retain their integrity and provide reasonably high LP values and high zein rejections. Membranes conditioned using the other methods result in lower rejections in almost all cases. The U20S membrane gave >75% rejection in all cases except with method 2. This is probably because the manufacturer ships the U20S membrane in 50% ethanol, and exposing it first to water, followed by 70% ethanol, could cause drastic structural changes. This also explains why protein rejection with this membrane remains high (∼80%) after direct exposure to 70% ethanol even though this conditioning method 3, damages most of the other membranes tested. Some membranes that have high zein rejections at low pressures (138 kPa) lose their rejection and become permeable to the protein at higher pressures (>275 kPa). Organic solvents tend to lower the glass transition temperature of polymers by acting as polymer plasticizers, which in turn reduces their ability to resist high pressures [6]. For instance, soaking polysulfone in an organic solvent with a similar solubility parameter will lead to cracking of the matrix [24]. There is evidence that hydrogen bonding solvents are usually much less disruptive to the polymer matrix than strongly nonpolar solvents like hexane [25]. Consequently, on exposure to organic solvents, the pressure ratings of the membranes are significantly reduced in many cases, except if they are conditioned using method 1. 4. Summary and conclusions The method of conditioning has a strong effect on the solvent flux, membrane integrity and pressure rating of polymeric membranes. Gradual solvent exchange with successively higher concentrations increased in small doses appears to work best with completely miscible solvents such as those studied here (ethanol–water mixtures). Rapid solvent exchange between water and high concentrations of alcohol disrupts the polymer matrix in many and leads to pore degradation. Exposure to organic solvents significantly reduces the pressure rating of the membranes. A membrane that provides acceptable rejection of protein up to 413 kPa (60 psi) under normal conditioning may lose its rejection even at 138 kPa (20 psi) if conditioned incorrectly. Acknowledgements This research was supported by the Illinois Corn Marketing Board, Illinois Department of Commerce and Community Affairs Bureau of Energy and Recycling, US Department of Agriculture through the NRICGP program (Award No. 97-35504-4296) and the Illinois Agricultural Experiment Station. Contributions of membranes by Koch Membrane Systems, Osmonics and Pall Filtron are gratefully acknowledged. Analytical assistance was provided by Amanda D. Popp and Melissa M. Stein. Useful discussions were held with D.A. Musale, M. Balakrishnan, H. Yacubowicz and J. Yacubowicz. References [1] M. Cheryan, N. Rajagopalan, Membrane processing of oily streams: wastewater treatment and waste reduction, J. Membr. Sci. 151 (1998) 13–28. [2] M. Cheryan, Ultrafiltration and Microfiltration Handbook, Technomic, Lancaster, PA, 1998. [3] V. Kale, S.P.R. Katikaneni, M. Cheryan, Deacidifying rice bran oil by solvent extraction and membrane technology, J. Am. Oil Chem. Soc. 76 (1999) 723–727. [4] S.S. Koseoglu, J.T. Lawhon, E.W. Lusas, Membrane processing of crude vegetable oils: pilot plant scale removal of solvent from oil miscellas, J. Am. Oil Chem. Soc. 67 (1990) 315–322. [5] Q.T. Nguyen, P. Aptel, J. Neel, Characterization of ultrafiltration membrane. Part I. Water and organic-solvent permeabilities, J. Membr. Sci. 5 (1979) 235–251. [6] R.W. Lencki, S. Williams, Effect of nonaqueous solvents on the flux behavior of ultrafiltration membranes, J. Membr. Sci. 101 (1995) 43–51. [7] M.Y. Jaffrin, J.P. Charrier, Optimization of ultrafiltration and diafiltration processes for albumin production, J. Membr. Sci. 97 (1994) 71–81. [8] B.B. Gupta, A. Chaibi, M.Y. Jaffrin, Ultrafiltration of albumin–ethanol solutions on mineral membranes, Sep. Sci. Technol. 30 (1995) 53–71. [9] L.P. Raman, M. Cheryan, N. Rajagopalan, Deacidification of soybean oil by membrane technology, J. Am. Oil Chem. Soc. 73 (1996) 219–224. [10] L.P. Raman, M. Cheryan, N. Rajagopalan, Solvent recovery and partial deacidification of vegetable oils by membrane technology, Fette/Lipid 98 (1996) 10–14. R. Shukla, M. Cheryan / Journal of Membrane Science 198 (2002) 75–85 [11] M.S. Kuk, R.J. Hron, G. Abraham, Reverse osmosis membrane characteristics for partitioning trigylceride–solvent mixtures, J. Am. Oil Chem. Soc. 66 (1989) 1374– 1380. [12] S. Koike, M. Yokoo, H. Nabetani, M. Nakajima, Membrane separation of fats and oils in organic solvents, Prog. Biosep. Eng. (Jpn. Soc. Chem. Eng.) 33 (1992) 84– 87. [13] H.J. Zwijnenberg, A.M. Krosse, K. Ebert, K.V. Peinemann, F.P. Cuperus, Acetone-stable nanofiltration membranes in deacidifying vegetable oil, J. Am. Oil Chem. Soc. 76 (1999) 83–87. [14] R. Shukla, M. Cheryan, Zein: the industrial protein from corn, Ind. Crops Products 13 (2001) 171–192. [15] M. Cheryan, Corn Oil and Protein Extraction Method, US Patent Application Serial Number 09/313,690 (1999). [16] R. Shukla, M. Cheryan, R. Devor, Solvent extraction of zein from dry-milled corn, Cereal Chem. 77 (2000) 724–730. [17] E.M. Craine, C.A. Jones, J.A. Boundy, A rapid turbidimetric method for determination of zein, Cereal Chem. 34 (1957) 456–462. [18] R. Shukla, Production of Zein from Dry Milled Corn by Solvent Extraction and Ultrafiltration, Ph.D. Thesis, University of Illinois at Urbana-Champaign, 2000. 85 [19] Y. Isooka, Y. Imamura, Y. Sakamoto, Recovery and reuse of organic solvent solutions, Met. Finish. 82 (1984) 113–118. [20] A. Iwama, Y. Kazuse, New polyimide ultrafiltration membranes for organic use, J. Membr. Sci. 11 (1982) 297– 309. [21] D.R. Machado, D. Hasson, R. Semiat, Effect of solvent properties on permeate flow through nanofiltration membranes. Part I. Investigation of parameters affecting solvent flux, J. Membr. Sci. 163 (1999) 93–102. [22] D.A. Musale, A. Kumar, Solvent and pH resistance of surface cross-linked chitosan-poly(acrylonitrile) composite nanofiltration membranes, J. Appl. Polym. Sci. 77 (2000) 1782–1793. [23] K.K. Reddy, T. Kawakastsu, J.B. Snape, M. Nakajima, Membrane concentration and separation of l-aspartic acid and l-phenylalanine derivatives in organic solvents, Sep. Sci. Technol. 31 (1996) 1161–1178. [24] E. Klein, K. Smith, The use of solubility parameters for solvent selection in asymmetric membrane formation, in: H.K. Lonsdale, H.E. Podall (Eds.), Reverse Osmosis Membrane Research, Plenum Press, New York, 1972. [25] R.C. Lukaszewicz, T.H. Meltzer, On the structural compatibilities of membrane filters, J. Parenteral Drug. Assoc. 34 (1980) 463–472.
© Copyright 2026 Paperzz