- Email: [email protected]
Membrane and solution effects on solute rejection and productivity
PII: S0043-1354(01)00169-5
Wat. Res. Vol. 35, No. 18, pp. 4426–4434, 2001 r 2001 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0043-1354/01/$ - see front matter
MEMBRANE AND SOLUTION EFFECTS ON SOLUTE REJECTION AND PRODUCTIVITY AMY K. ZANDER1* and N. KEVIN CURRY2 1
Associate Professor, Department of Civil and Environmental Engineering, Clarkson University, Box 5710, Potsdam, NY 13699, USA and 2 Langan Engineering and Environmental Services, Doylestown, PA 18901-1699, USA (First received 1 August 2000; accepted in revised form 1 March 2001)
AbstractFLimited understanding of the physical and chemical processes involved in membrane processes affects their widespread application in drinking water treatment. Insight into these processes was attained through a systematic manipulation of solution chemistry in membrane filtration with three ‘loose’ nanofiltration membranes. Eighteen known solutions were created varying pH, ionic strength, and major cation valence in the presence of a commercial humic acid. The membrances varied as well, including a non-ionic hydrophobic, a non-ionic hydrophilic, and an anionic hydrophilic membrane surface. Specific membrane productivity and TOC and conductivity rejection were monitored. In all cases, the presence of divalent cations decreased the rejection of both conductivity and organic matter. Divalent cations also greatly increased the rate of productivity decline over equivalent tests in solutions with monovalent cations. The most hydrophobic membrane had the greatest productivity decline rate under all solution conditions. The lowest ionic strength solutions showed the greatest TOC and conductivity rejection and the greatest rate of productivity decline for each of the membranes. r 2001 Elsevier Science Ltd. All rights reserved Key wordsFnanofiltration, drinking water treatment, humic material, cations, membrane, productivity
NOMENCLATURE
CF CP R T
concentration of solute in membrane feed (mg/L) concentration of solute in membrane permeate (mg/L) percent rejection temperature (1C)
INTRODUCTION
Many factors are thought to contribute to nanofiltration membrane productivity and solute rejection, though an understanding of these factors is incomplete. The complex interactions between solution, membrane and natural organic matter (NOM) chemistry leading to solute rejection and membrane fouling require additional study. Some interactions are fairly well understood and are outlined here. One of the major mechanisms of solute rejection is physical sieving of solutes larger than the membrane molecular weight cutoff (MWCO). Theoretically, all molecules larger than the MWCO of a membrane should be rejected. However, MWCOs are seldom *Author to whom all correspondence should be addressed. Tel.: +1-315-268-6532; fax: +1-315-268-7985; e-mail: [email protected]
‘sharp’ cutoffs (Schweitzer, 1979; Seacord, 1998), and pH and ionic strength affect the size, shape and flexibility of a solute. For macromolecules with polyelectrolyte characteristics, such as NOM, rejection is controlled by several factors: molecular size, shape and flexibility, diffusional behavior in the bulk solution and membrane pores, and electrostatic and van der Waals forces between membrane and solutes (Weber, 1971). Ionic strength and pH are known to affect the molecular size distribution of NOM (Cornel et al., 1986). Ghosh and Schnitzer (1980) propose a theory for NOM configuration in solution as a continuum between two hypothetical endpoints. In low ionic strength and/or high pH solution, with large double layer thickness, the NOM will form a flexible, linear macromoleule. At high ionic strength and/or low pH the NOM will have a thin double layer and more readily form a spherical macromolecule. Adsorption can enhance rejection (Jonsson and Boeson, 1984; Nystrom et al., 1989), as even a small portion of NOM at high pH can impart a negative surface charge to the membrane and moreover, serve to reject the negatively charged NOM by charge repulsion (Aiken, 1984). Macromolecules can penetrate pores and produce a negative charge within the pore (Nystrom et al., 1989; Wiesner et al., 1989;
4426
Solution effect on rejection and productivity
Childress and Elimelech, 2000). A membrane with a negative surface charge can also employ anion repulsion as a rejection mechanism (LahoussineTurcaud et al., 1990; Marinˇas and Selleck, 1992; Peterson, 1993; Fu et al., 1994; Allgeier and Summers, 1995a). Thus, the rejection of both organic and inorganic solutes can be simultaneously controlled by sieving and by solution chemistry and membrane properties. Braghetta et al. (1997) examined the effect of solution pH and ionic strength on the permeability of negatively charged sulfonated polysulfone nanofiltration membranes, and compiled a discussion of the potential influence of pH and ionic strength on the membrane structure. In short, the double layer thickness surrounding a charged membrane can affect membrane characteristics. Conditions favoring a thick electrical double layer, high pH and/or low ionic strength solution, will result in an expansion of the membrane matrix, allowing greater water permeability. The same conditions will also result in a greater rejection of charged solutes due to the thick double layer overlapping or nearly overlapping within membrane pores. Under the conditions leading to a thinner electrical double layer, low pH and/ or high ionic strength solution, the membrane matrix can contract reducing water permeability. At the same time, charged solutes would be able to traverse the membrane better through the larger spaces between adjacent double layers within the pores. The same principles should apply, though to a lesser extent, even for less charged or ‘uncharged’ membrane surfaces. An uncharged membrane will experience adsorption of NOM under the conditions in which the NOM is also not strongly charged. Aqueous flux and van der Waals forces are then the predominant mechanisms bringing the solute to the membrane surface. Hydrophobic interactions can then lead to an adsorption of the solute to the surface. Other factors not studied here may also be important in interpreting membrane performance. These factors include, for example, transmembrane pressure (Schafer et al., 1998), polarity (Van der Bruggen et al., 1999) or aromaticity (Cho et al., 1999) of the organic
4427
material in solution, concentration of foulants (Manttari et al., 2000), and membrane surface roughness (Vrijenhoek et al., 2000).
RESEARCH OBJECTIVES
This work seeks to elucidate the effects of solution pH, ionic strength and cation valence on the rejection of total organic carbon (TOC) and conductivity in contact with three loose nanofiltration membranes of varying characters (charge, hydrophilicity and nominal MWCO). Potential effects on the membrane matrix and on the surface characteristics of the membranes are surveyed. In addition, the effects of the membrane character and solution chemistry on the decline in specific productivity of the membranes (an indicator of membrane fouling) are observed.
MATERIALS AND METHODS
Membrane system Membrane filtration was performed using a bench-scale membrane system that mimics the operating conditions of a spiral wound membrane element (Osmonics, 1995; Allgeier and Summers, 1995a, b). The membrane system uses a flat sheet membrane cell with an effective membrane surface area of 155 cm2 (24 in2), with feed and permeate spacers and tangential feed flow as found in spiral wound systems. The cell is held closed pneumatically and can provide pressures up to 700 kPa (100 psi). The membrane cell was incorporated into a system with a complete recycle of both permeate and concentrate to the feed tank to allow the approximation of steady-state conditions. Twenty liters of test solution were used for each test, with the exception of some RG03 tests in which 40 L of water was used to approximate steadystate concentration of humic acid in the test solution. The setup allowed independent control of the feed flow rate (crossflow velocity) and transmembrane pressure. Figure 1 presents a schematic diagram of the experimental setup. All tests were performed at a crossflow velocity of 0.10 m/s (0.33 ft/s), typical of a full-scale water production system. Control of the transmembrane pressure and the crossflow velocity at constant values required a permeate recovery rate (permeate flow rate 100%/feed flow rate), lower than the full scale average of 20% in a single pass. The recovery rate in this study was 10% or less and is given in Table 1.
Fig. 1. Experimental nanofiltration membrane system.
4428
Amy K. Zander and N. Kevin Curry Table 1. Membrane information
Membrane
SEPA BP02
YM10
SEPA RG03
Manufacturer Composition of surface layer Surface character MWCOd Recommended TMPd TMP employed pH ranged Recovery rate Initial specific productivity, SP0, L/m2 h kPa (gfd/psi)e
Osmonics, Minnetonka, MN ‘like’ sulfonated polysulfonea Anionic, hydrophilica 1000–5000 daltons 690 kPa (100 psi) 620 kPa (90 psi) 0.5–13 3% 0.06070.007 (0.24370.029)
Amicon Beverly, MA Regenerated cellulose acetateb Non-ionic, hydrophilicb,c 10,000 daltons 380 kPa (55 psi) 380 kPa (55 psi) 3–11 7% 0.27170.026 (1.1070.10)
Osmonics, Minnetonka, MN Acrylonitriled Non-ionic, hydrophobicd 4000–10,000 daltons 690 kPa (100 psi) 620 kPa (90 psi) 0.5–10 10% 0.25870.049 (1.0570.20)
a
Harrold (1996). Laı´ ne´ et al. (1989); Braghetta and DiGiano (1994); Braghetta et al. (1997). Clark and Jucker (1993). d As stated by the manufacturer. e This work. b c
Membranes studied The three membranes used in this study were chosen for their contrasting materials of composition, and the resulting variation in surface hydrophilicity and charge. Pertinent information regarding the membranes is tabulated in Table 1. The SEPA BP02 membrane is most similar in surface composition and MWCO to the membranes used by Braghetta et al. (1997) in a single membrane study of related pH and ionic strength effects in nanofiltration. Sulfonated polysulfone is rapidly becoming a popular choice for nanofiltration membranes, due to its perceived ability to resist fouling. The YM series of membranes have been shown by Clark and Jucker (1993) to be relatively unaffected by a change in solution pH and ionic strength. These membranes are hydrophilic like the BP02, but they do not possess a charged surface functional group. The RG03 membrane is non-ionic, similar to the YM10, but the membrane surface is hydrophobic, potentially increasing the adsorbance of hydrophobic materials from the solution and increasing the rate of decline in specific productivity. Membrane preparation The membranes were cleaned by soaking in 0.5 L of HCl solution at pH 2 for 30 min, followed by 0.5 L of NaOH solution at pH 10.5 for 30 min, using a procedure recommended by Allgeier and Summers (1995b). All membranes were then soaked in deionized water for 1 h, with at least three water changes, followed by filtration of approximately 1 gal (4 L) of deionized water through the membrane in the test cell. This allowed the removal of any impurities or preservatives from the membranes (Amicon, 1995). Each membrane was then compacted by operation under the desired operating pressure in the test cell, using pH 7 buffered water, until a stable flux was reached. This usually occurred within 20 h. The membrane was then characterized by a specific productivity measurement, which served as the initial specific productivity (SP0). These initial specific productivity measurements are given in Table 1. Test solution preparation The matrix of 18 test solutions used was chosen to represent a broad range of naturally occurring water conditions from a soft to hard water, river water to groundwater with saltwater intrusion. The solutions varied three pHs (4.0, 7.0 and 8.5), three ionic strengths (1.4 10@3, 6.0 10@3, and 1.4 10@2 M with approximate conductivities of 100, 700, and 1500 mmhos/cm), and either a monovalent (Na+) or divalent (Ca2+) predominant cation. Buffer materials accounted for 10% or less of the
solution ionic strength in all the cases. Phosphorous acid (pH 4.0 and 7.0; Aldrich, Milwaukee, WI) and boric acid (pH 8.5; J.T. Baker, Phillipsburg, NJ) served as buffer materials. The remaining ionic strength was made up of sodium chloride dihydrate or calcium chloride (J.T. Baker, Phillipsburg, NJ), as appropriate. Each solution also contained 10 mg/L of organic carbon added as 0.45 mm prefiltered commercially available humic acid (Lot #01828JZ, Aldrich, Milwaukee, WI). Despite some known differences between Aldrich humic acid and naturally occurring aquatic humic substances, the commercial humic acid has been found to be satisfactory as a surrogate for NOM for elucidating some of the mechanisms of membrane flux decline (Wiesner et al., 1989). Apparent molecular size fractionation of the humic material was performed following the protocol described by Owen et al. (1993). At a pH of 8.5 and an ionic strength of 0.0014 M the Aldrich humic acid had an apparent molecular size distribution of 67% greater than 30 kDa, 27% between 10 and 30 kDa, 3% between 5 and 10 kDa and 3% less than 5 kDa. At this fairly high pH and low ionic strength, the humic acid is likely to be in the form of a highly negatively charged, flexible, linear macromolecule (Braghetta et al., 1997). From this analysis, size exclusion alone should allow for 94% or greater rejection of TOC for all of the membranes studied. Analytical methods Organic concentrations were monitored by TOC analysis (Method 5310 B; APHA, AWWA, WEF, 1992) using a Dohrmann DC-190 TOC Analyzer (Rosemount Analytical Inc., Santa Clara, CA). Injection of large size samples (400 mL) allowed the detection of TOC to 0.2 mg/L and a relative standard deviation (RSD=standard deviation/ mean) of multiple TOC determinations of 0.044. Percent rejection of TOC (and conductivity) was determined using the following expression: R ¼ ðCF @CP Þ100%=CF where R is the percent of solute rejected (retained) by the membrane, CF the solute concentration in the feed water, and CP the concentration of solute in the permeate. Rejection was determined by averaging the calculated % rejection values following the observation of steady state rejection, which occurred soon after the system startup. Ionic strength was measured using electrical conductivity measurements (Method 2510; APHA, AWWA, WEF, 1992) as a surrogate parameter, and with a conductivity meter (Cole Parmer, Model #01481-61, Niles, IL). Measurement of solution pH was performed according to Method 4500-
Solution effect on rejection and productivity H+ (APHA, AWWA, WEF, 1992) using a pH meter (Cole Parmer, Model #05669-201, Niles, IL). Specific productivity measurements The specific productivity as used in this article refers to the flux of water through a membrane divided by the transmembrane pressure at which the flux was measured. Other works have called this term a water mass transfer coefficient (MTCW) (e.g. Morris and Taylor, 1991; Reiss and Taylor, 1991; Allgeier and Summers, 1995b; Taylor and Mulford, 1995; Curry and Zander, 1997), however the authors (and others, e.g. Fu et al., 1994) find specific productivity to be more precise description of the term. The time rate of decay of the specific productivity is an accurate procedure for measuring membrane productivity changes (Taylor and Mulford, 1995) and the rate of flux decline due to fouling (Morris and Taylor, 1991). Each specific productivity measurement was determined as the slope of a graph of eight individual measurements of flux vs. corresponding applied transmembrane pressure. All measurements for this were obtained at a feedwater cross-flow velocity of 0.10 m/s (0.33 ft/s) using deionized water at a pH of 5.6. Flux measurements were taken as soon as a stable flux developed for each pressure applied. Temperature was controlled within a few degrees Celsius and monitored throughout the tests. Flux measurements were corrected to flux at 201C using the correlation from Jacangelo et al. (1994) based on the variation of water viscosity with temperature:
Flux201C ¼ Meaured Flux eð@0:0239ðT@20ÞÞ where T is the temperature in 1C at which the flux was measured. The units of specific productivity are given as a flux per unit pressure: gal of permeate per square foot of membrane area per day per pound per square inch of applied pressure (gfd/psi) or liters of permeate per square meter of membrane area per h per kPa of applied pressure (L/m2 h kPa). These specific productivity determinations displayed excellent precision, with the coefficient of determination, R2 , greater than 0.99. The relative standard deviation of these measurements was less than 0.02.
4429
The rate of decline in specific productivity was determined by fitting a linear regression to data on specific productivity changes with time. As attempted in other studies (Solomon et al., 1993), the decline in specific productivity (flux decline) was fitted to several expressions, such as exponential decay, logarithm, power functions and linear regression. The best fit to all data was found to be a linear approximation of decline during the initial operation of the membrane system. The coefficients of determination, R2 , for these linear fits averaged 0.90 for the YM10 and RG03 membranes, but had a relatively poor mean fit ðR2 ¼ 0:66Þ for the BP02 membrane. This relatively poor fit may be a function of the small changes in specific productivity for this membrane over the course of the experiments. However, the relative magnitude of the specific productivity decline is still comparable between membranes.
EXPERIMENTAL RESULTS AND DISCUSSION
TOC rejection Figures 2–4 display the TOC rejection attained by each membrane (BP02, YM10, and RG03, respectively) under steady state conditions in each of the eighteen test solutions. The bars in each graph indicate the mean value of TOC rejection under each condition and the error bars indicate the relative standard deviation of the tests. Observation of Fig. 2 reveals no statistically significant differences in TOC rejection with pH or ionic strength for the BP02 membrane in solutions made with monovalent cation, though the overall TOC rejection was slightly less than the approximately 97% rejection expected from the apparent molecular size distribution of the humic acid. The expected trend of increasing rejection with increasing pH, though not experimentally substantiated here, is clearly not refuted. In the solutions made with a divalent cation, a trend toward a reduction in TOC
Fig. 2. The TOC rejection as a function of pH and ionic strength for BP02.
4430
Amy K. Zander and N. Kevin Curry
Fig. 3. The TOC rejection as a function of pH and ionic strength for YM10.
Fig. 4. The TOC rejection as a function of pH and ionic strength for RG03.
rejection as compared to monovalent cation solutions can be seen throughout the range of pH and ionic strength. This is likely due to the reduction in the thickness of the double layer on the membrane by the presence of Ca2+. This reduction is greatest at the highest pH, due to the effect of the Ca2+ on the configuration of the humic acid which is highly negatively charged at this pH. It is also possible that the charge repulsion experienced by the negatively charged humic acid in proximity to the negatively charged membrane is partially overcome by the presence of the divalent cation.
Similar results can be seen in Fig. 3 for the hydrophilic, nominally non-ionic YM10. The larger nominal MWCO for this membrane (10,000 DA as opposed to 5000 Da for the BP02) results in a slightly lower rejection of TOC in monovalent solution. Again, no statistically significant changes could be seen with pH or ionic strength for the monovalent solution. Exchange of the monovalent cation for a divalent cation results in a general reduction in TOC rejection, again potentially due to the effect of Ca2+ on the configuration of the humic acid, or the effect of the divalent cation on the electrostatic repulsion
Solution effect on rejection and productivity
between membrane and humic acid. Though the manufacturer touts this membrane as non-ionic, evidence exists that it actually bears a slightly negative potential in solution. A negative zeta potential was measured for this membrane under similar solution conditions using a streaming potential anlayzer (BI-EKA, Brookhaven Instruments, Holtsville, NY), (Zander and Desrosiers, 1999) and a negative zeta potential was observed by Elimelech et al. (1994) for other cellulose acetate membranes. The TOC rejection measured for the third membrane, RG03, is given in Fig. 4. This non-ionic, hydrophobic membrane behaves quite differently in solution than the other two membranes. In several of the solutions tested, the overall TOC rejection is greater for the solution containing divalent rather than monovalent cations. No significant difference is seen in the TOC removal with pH or ionic strength with divalent cations, though the overall rejection is slightly less than expected by steric hindrance using the manufacturers MWCO. The major mechanism for rejection in this case is likely to be the attraction of the humic acid and Ca2+ complex to the uncharged hydrophobic membrane surface. One can justify the case by the relatively high rate of decline in specific productivity for this membrane in divalent solution as seen in Fig. 7 and discussed in further detail later. Conductivity rejection Conductivity rejection varies greatly between the membranes and, for the BP02, between solution conditions. Figure 5 displays the conductivity rejection for the three membranes at the lowest ionic strength monovalent cation solution. This solution condition is also the only solution condition under which any measurable conductivity rejection was
4431
seen for the YM10 membrane (data not shown). Some slight conductivity rejection was seen for the RG03 in monovalent solution at the medium ionic strength (data not shown). In the presence of the divalent cation, and at high ionic strength in the monovalent solutions, essentially no conductivity rejection was seen for the YM 10 or RG03. This would be the expected result for these membranes. This illustrates that the rejection of salts under low ionic strength conditions is a function of interactions between the membrane and solution. At low ionic strength, especially with monovalent cations, the double layer surrounding the membrane is thicker than under the other conditions studied, allowing for greater rejection of charged solutes. As the double layer thickness decreases with increasing ionic strength, the charged solutes easily pass through the membrane. The anionic BP02 membrane has electrostatic repulsion as an additional mechanism of conductivity rejection, hence the significantly higher rejection for this membrane. A strong effect of ionic strength on conductivity rejection can also be seen for the BP02 in Fig. 6. For both the monovalent and divalent cation solutions, the rejection of conductivity decreases with increasing ionic strength. This is again a function of the effect of the solution on the double layer thickness surrounding the membrane. The double layer thickness will decrease with increasing ionic strength, and more so in the presence of divalent cations, leading to an increase in the area available for the passage of charged species. It is also apparent in Fig. 6 that at least for tests in the divalent solution, an increase in pH results in a decrease in rejection. This could be explained by a charge balancing of the negatively charged membrane by hydronium ions in solution and a reduction in electrostatic repulsion.
Fig. 5. Conductivity rejection for three membranes at 0.0014 M monovalent ionic strength.
4432
Amy K. Zander and N. Kevin Curry
Fig. 6. Conductivity rejection as a function of pH and ionic strength for BP02.
Fig. 7. Decline in specific productivity as a function of pH and divalent ionic strength for three membranes.
The observation that in this case the divalent ions are rejected less than the monovalent ions would seem to contradict the more common observation that NF membranes reject divalent cations very well (Nystrom et al., 1995; Hong and Elimelech, 1997). This is probably due to the fairly ‘loose’ pore size characteristics of the BP02 membrane. This membrane may be better described as an ultrafiltration type membrane. Productivity decline As membrane fouling is a major contributor to the operational cost of membrane systems, it is important to understand the factors involved in fouling.
Fouling here is measured as the decline in specific productivity of the membrane and the rate of that decline is assumed to be indicative of the operational cost involved in minimizing fouling or in membrane cleaning. A lower rate of specific productivity decline indicates a lower operational cost for the membrane under those conditions. For each of the membranes in the monovalent cation solutions the rate of decline is fairly low, generally less than 1%/h (data not shown). When the major cation in solution is changed from monovalent to divalent there is a significant increase in the rate of specific productivity decline from less than 1%/h to values up to 50%/h or more as seen in Fig. 7. This effect has been observed previously (e.g.
Solution effect on rejection and productivity
Hong and Elimelech, 1997) and is due to the interaction of the divalent cations with carboxyl functional groups on the humic material. This reduces electrostatic repulsion between the macromolecules and they form a dense surface layer on the membrane. Many of the solution conditions lead to productivity decline rates greater than 10%/h. The BP02 was the least affected by fouling, even in divalent solution. Adsorption of materials to the RG03 membrane surface at low ionic strength conditions led to a decline in productivity greater than 50%/h. At this very high decline rate, quantification is difficult as the flux rate is dropping rapidly. Decline rates greater than 50%/h are reported simply as greater than this value. In one test, carried out at a pH of 4.0, the RG03 membrane became completely fouled in less than 1 h. The fouling at the membrane surface of humic acid associated with Ca2+ is the likely reason for the measurable rejection of conductivity with this membrane under these conditions. The initial SP0 of each membrane is given in Table 1. The SP0 value for each membrane was relatively constant across test solution conditions. Changes in SP0 with time in a given test condition were generally much greater than the difference in the initial membrane productivity for a given membrane under differing solution conditions. Two of the membranes, RG03 and YM10, had similar values of SP0 at approximately 0.26 and 0.27 L/m2 h kPa, respectively. The BP02 membrane had a value of less than onefourth the value of the others, with an SP0 of 0.06 L/ m2 h/kPa. Comparing these initial productivity values, the BP02 would require more than 4 times the membrane surface area to achieve the same permeate flow rate for a similar applied pressure as the other membranes studied, greatly increasing the capital cost of a membrane plant.
CONCLUSIONS
Organic matter and conductivity rejection are not a direct function of manufacturers stated membrane MWCO. Instead, rejection is a complex interaction of steric hindrance, electrostatic repulsion, solution effects on the membrane and solute shape, and thickness of the double layer surrounding the membrane in solution. Presence of divalent cations in solution greatly affects the membrane performance. Solutions containing divalent rather than monovalent cations showed lower rejection of both aqueous conductivity and TOC. Specific permeability decline measurements are a useful tool for comparison of solution effects on membrane performance. At the lowest ionic strength tested, the more hydrophobic membrane had a greater rate of specific flux decline than hydrophilic membranes. Of the variables tested, membrane surface chemistry is the
4433
most important factor affecting the rate of flux decline. Ignoring the effect of membrane chemistry, specific flux decline is greatest for low ionic strength solutions. Divalent cations increased the rate of decline in specific flux significantly over monovalent cations, especially for the YM10 and the RG03 membranes. AcknowledgementsFThis work was sponsored by Grant #BES-9307851 from the National Science Foundation.
REFERENCES
Aiken G. R. (1984) Evaluation of UF for Determining Molecular Weight of Fulvic Acid. Environ. Sci. Technol. 18, 978–981. Allgeier S. C. and Summers R. S. (1995a) Evaluating NF for DBP control with the RBSMT. J. Am. Water Works Assoc. 87(3), 87–99. Allgeier S. C. and Summers R. S. (1995b) Development of a Rapid Bench Scale Membrane Test for the Evaluation of NF Performance. Proceedings of the 1995 AWWA Membrane Technology Conference, Reno, NV, pp. 207–227. Amicon Inc. (1995) Laboratory Products Catalog, Beverly, MA. APHA, AWWA, WEF. (1992) Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington, DC. Braghetta A. and DiGiano F. A. (1994) Organic solute association with nanofiltration membrane surface: Influence of pH and ionic strength on membrane performance. Proceedings of the 1994 AWWA Annual Conference, New York, NY. Braghetta A., DiGiano F. A. and Ball W. P. (1997) Nanofiltration of natural organic matter: pH and ionic strength effects. J. Environ. Eng. 123(7), 628–641. Childress A. E. and Elimelech M. (2000) Relating nanofiltration membrane performance to membrane surface charge (Electrokinetic) characteristics. Environ. Sci. Technol. 34(17), 3710–3716. Cho J. W., Amy G. and Pellegrino J. (1999) Membrane filtration of natural organic matter: initial comparison of rejection and flux decline characteristics with ultrafiltration and nanofiltration membranes. Water Res. 33(11), 2517–2526. Clark M. M. and Jucker C. (1993) Interactions between hydrophobic ultrafiltration membranes and humic substances. Proceedings of the 1993 AWWA Membrane Technology Conference, Baltimore, MD, pp. 259–272. Cornel P. K., Summers R. S. and Roberts P. V. (1986) Diffusion of humic acid in dilute aqueous solution. J. Colloid Interface Sci. 110(1), 149–164. Curry N. K. and Zander A. K. (1997) Effect of aqueous solution chemistry on UF and NF membrane performance. Proceedings of the 1997 AWWA Annual Conference, Atlanta, GA, pp. 603–609. Elimelech M., Chen W. H. and Waypa J. J. (1994) Measuring the Zeta (Electrokinetic) potential of reverse osmosis membranes by a streaming potential analyzer. Desalination 95, 269–286. Fu P., Ruiz H., Thompson K. and Spangenberg C. (1994) Selecting membranes for removing NOM and DBP precursors. J. Am. Water Works Assoc. 86(12), 55–72. Ghosh K. and Schnitzer M. (1980) Macromolecular structures of humic substances. Soil Sci. 129(5), 266–276. Harrold S. J. (1996) Personal Communication, Osmonics, Inc. Minnetonka, MN.
4434
Amy K. Zander and N. Kevin Curry
Hong S. and Elimelech M. (1997) Chemical and Physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. J. Membrane Sci. 132, 159–181. Jacangelo J. G., Laı´ ne´ J.-M., Cummings E. W., Deutschmann A., Mallevialle J. and Wiesner M. R. (1994) Evaluation of UF Membrane Pretreatment and NF of Surface Waters. AWWA Research Foundation and AWWA, Denver, Co. Jonsson G. and Boeson C. E. (1984) Polarization phenomena in membrane processes. Synthetic Membrane Process Fundamentals and Water Applications, ed. G. Belfort. Academic Press, Orlando, FL. Lahoussine-Turcaud V., Wiesner M. R., Bottero J.-Y. and Mallevialle J. (1990) Coagulation pretreatment for ultrafiltration of a surface water. J. Am. Water Works Assoc. 82(12), 76–81. Laine´ J.-M., Clark M. M. and Mallevialle J. (1989) Effects of ultrafiltration membrane composition. J. Am. Water Works Assoc. 82(11), 61–73. Manttari M., Puro L., Nuortila-Jokinen J. and Nystrom M. (2000) Fouling effects of polysaccharides and humic acid in nanofiltration. J. Membrane Sci. 165(1), 1–17. Marinˇas B. and Selleck R. E. (1992) Reverse osmosis treatment of multicomponent electrolyte solutions. J. Membrane Sci. 72, 211–229. Morris K. E. and Taylor J. S. (1991) The use of fouling indices to predict RO membrane fouling. Proceedings of the 1991 AWWA Membrane Processes Conference, Orlando, FL. Nystrom M., Kaipia L. and Luque S (1995) Fouling and retention of nanofiltration membranes. J. Membrane Sci. 98, 249–262. Nystrom M., Lindstrom M. and Matthiasson E. (1989) Streaming potential as a tool in the characterization of UF membranes. Colloids and Surfaces 36, 297–312. Osmonics, Inc. (1995) SEPAs CF Membrane Cell Instruction Manual. Minnetonka, Minnesota. Owen D. M., Amy G. L. and Chowdury Z. K. (1993) Characterization of Natural Organic Matter and its Relationship to Treatability. AWWA Research Foundation and AWWA, Denver, CO.
Peterson R. J. (1993) RO/NF membrane performance characteristics. Proceedings of the 1993 AWWA Membrane Technology Conference, Baltimore, MD, pp. 23–39. Reiss C. R. and Taylor J. S. (1991) Membrane pretreatment of a surface water. Proceedings of the 1991 AWWA Membrane Processes Conference, Orlando, FL. Schafer A. I., Fane A. G. and Waite T. D. (1998) Nanofiltration of natural organic matter: removal, fouling and the influence of multivalent Ions. Desalination 118(1–3), 109–122. Schweitzer P. A. (1979) Handbook of Separation Techniques for Chemical Engineers, 2nd ed. McGraw Hill, New York. Seacord T. F. (1998) Assessing nanofiltration membrane properties for water treatment applications, M.S. Thesis, Clarkson University, Potsdam, NY. Solomon T. D., Summers R. S. and Speth T. F. (1993) The reduction of membrane fouling by physical, chemical and biological pretreatment. Supplement to the Proceedings of the 1993 AWWA Membrane Technology Conference, Baltimore, MD. Taylor J. S. and Mulford L. A. (1995) Membrane protocol to meet the ICR. Proceedings of the 1995 AWWA Membrane Processes Conference, Reno, NV. Van der Bruggen B., Schaep J., Wilms D. and Vandecasteele C. (1999) Influence of molecular size, polarity and charge on the retention of organic molecules by nanofiltration. J. Membrane Sci. 156(1), 29–41. Vrijenhoek E. M., Elimelech M. and Hong S. K. (2000) Influence of membrane properties, solution chemistry, and hydrodynamics on colloidal fouling of reverse osmosis and nanofiltration membranes. Abst. Papers Am. Chem. Soc. Part 1, 220, 40-ENVR. Weber W. J. Jr. (1971) Physicochemical Processes for Water Quality Control. Wiley, New York, NY. Wiesner M. R., Clark M. M. and Mallevialle J. (1989) Membrane filtration of coagulated suspensions. J. Environ. Eng. ASCE 115(1), 20–40. Zander A. K. and Desrosiers K. (1999) Arsenite removal from water using UF/NF membranes. Proceedings of the 1999 AWWA Annual Conference, Chicago, IL.
Wat. Res. Vol. 35, No. 18, pp. 4426–4434, 2001 r 2001 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0043-1354/01/$ - see front matter
MEMBRANE AND SOLUTION EFFECTS ON SOLUTE REJECTION AND PRODUCTIVITY AMY K. ZANDER1* and N. KEVIN CURRY2 1
Associate Professor, Department of Civil and Environmental Engineering, Clarkson University, Box 5710, Potsdam, NY 13699, USA and 2 Langan Engineering and Environmental Services, Doylestown, PA 18901-1699, USA (First received 1 August 2000; accepted in revised form 1 March 2001)
AbstractFLimited understanding of the physical and chemical processes involved in membrane processes affects their widespread application in drinking water treatment. Insight into these processes was attained through a systematic manipulation of solution chemistry in membrane filtration with three ‘loose’ nanofiltration membranes. Eighteen known solutions were created varying pH, ionic strength, and major cation valence in the presence of a commercial humic acid. The membrances varied as well, including a non-ionic hydrophobic, a non-ionic hydrophilic, and an anionic hydrophilic membrane surface. Specific membrane productivity and TOC and conductivity rejection were monitored. In all cases, the presence of divalent cations decreased the rejection of both conductivity and organic matter. Divalent cations also greatly increased the rate of productivity decline over equivalent tests in solutions with monovalent cations. The most hydrophobic membrane had the greatest productivity decline rate under all solution conditions. The lowest ionic strength solutions showed the greatest TOC and conductivity rejection and the greatest rate of productivity decline for each of the membranes. r 2001 Elsevier Science Ltd. All rights reserved Key wordsFnanofiltration, drinking water treatment, humic material, cations, membrane, productivity
NOMENCLATURE
CF CP R T
concentration of solute in membrane feed (mg/L) concentration of solute in membrane permeate (mg/L) percent rejection temperature (1C)
INTRODUCTION
Many factors are thought to contribute to nanofiltration membrane productivity and solute rejection, though an understanding of these factors is incomplete. The complex interactions between solution, membrane and natural organic matter (NOM) chemistry leading to solute rejection and membrane fouling require additional study. Some interactions are fairly well understood and are outlined here. One of the major mechanisms of solute rejection is physical sieving of solutes larger than the membrane molecular weight cutoff (MWCO). Theoretically, all molecules larger than the MWCO of a membrane should be rejected. However, MWCOs are seldom *Author to whom all correspondence should be addressed. Tel.: +1-315-268-6532; fax: +1-315-268-7985; e-mail: [email protected]
‘sharp’ cutoffs (Schweitzer, 1979; Seacord, 1998), and pH and ionic strength affect the size, shape and flexibility of a solute. For macromolecules with polyelectrolyte characteristics, such as NOM, rejection is controlled by several factors: molecular size, shape and flexibility, diffusional behavior in the bulk solution and membrane pores, and electrostatic and van der Waals forces between membrane and solutes (Weber, 1971). Ionic strength and pH are known to affect the molecular size distribution of NOM (Cornel et al., 1986). Ghosh and Schnitzer (1980) propose a theory for NOM configuration in solution as a continuum between two hypothetical endpoints. In low ionic strength and/or high pH solution, with large double layer thickness, the NOM will form a flexible, linear macromoleule. At high ionic strength and/or low pH the NOM will have a thin double layer and more readily form a spherical macromolecule. Adsorption can enhance rejection (Jonsson and Boeson, 1984; Nystrom et al., 1989), as even a small portion of NOM at high pH can impart a negative surface charge to the membrane and moreover, serve to reject the negatively charged NOM by charge repulsion (Aiken, 1984). Macromolecules can penetrate pores and produce a negative charge within the pore (Nystrom et al., 1989; Wiesner et al., 1989;
4426
Solution effect on rejection and productivity
Childress and Elimelech, 2000). A membrane with a negative surface charge can also employ anion repulsion as a rejection mechanism (LahoussineTurcaud et al., 1990; Marinˇas and Selleck, 1992; Peterson, 1993; Fu et al., 1994; Allgeier and Summers, 1995a). Thus, the rejection of both organic and inorganic solutes can be simultaneously controlled by sieving and by solution chemistry and membrane properties. Braghetta et al. (1997) examined the effect of solution pH and ionic strength on the permeability of negatively charged sulfonated polysulfone nanofiltration membranes, and compiled a discussion of the potential influence of pH and ionic strength on the membrane structure. In short, the double layer thickness surrounding a charged membrane can affect membrane characteristics. Conditions favoring a thick electrical double layer, high pH and/or low ionic strength solution, will result in an expansion of the membrane matrix, allowing greater water permeability. The same conditions will also result in a greater rejection of charged solutes due to the thick double layer overlapping or nearly overlapping within membrane pores. Under the conditions leading to a thinner electrical double layer, low pH and/ or high ionic strength solution, the membrane matrix can contract reducing water permeability. At the same time, charged solutes would be able to traverse the membrane better through the larger spaces between adjacent double layers within the pores. The same principles should apply, though to a lesser extent, even for less charged or ‘uncharged’ membrane surfaces. An uncharged membrane will experience adsorption of NOM under the conditions in which the NOM is also not strongly charged. Aqueous flux and van der Waals forces are then the predominant mechanisms bringing the solute to the membrane surface. Hydrophobic interactions can then lead to an adsorption of the solute to the surface. Other factors not studied here may also be important in interpreting membrane performance. These factors include, for example, transmembrane pressure (Schafer et al., 1998), polarity (Van der Bruggen et al., 1999) or aromaticity (Cho et al., 1999) of the organic
4427
material in solution, concentration of foulants (Manttari et al., 2000), and membrane surface roughness (Vrijenhoek et al., 2000).
RESEARCH OBJECTIVES
This work seeks to elucidate the effects of solution pH, ionic strength and cation valence on the rejection of total organic carbon (TOC) and conductivity in contact with three loose nanofiltration membranes of varying characters (charge, hydrophilicity and nominal MWCO). Potential effects on the membrane matrix and on the surface characteristics of the membranes are surveyed. In addition, the effects of the membrane character and solution chemistry on the decline in specific productivity of the membranes (an indicator of membrane fouling) are observed.
MATERIALS AND METHODS
Membrane system Membrane filtration was performed using a bench-scale membrane system that mimics the operating conditions of a spiral wound membrane element (Osmonics, 1995; Allgeier and Summers, 1995a, b). The membrane system uses a flat sheet membrane cell with an effective membrane surface area of 155 cm2 (24 in2), with feed and permeate spacers and tangential feed flow as found in spiral wound systems. The cell is held closed pneumatically and can provide pressures up to 700 kPa (100 psi). The membrane cell was incorporated into a system with a complete recycle of both permeate and concentrate to the feed tank to allow the approximation of steady-state conditions. Twenty liters of test solution were used for each test, with the exception of some RG03 tests in which 40 L of water was used to approximate steadystate concentration of humic acid in the test solution. The setup allowed independent control of the feed flow rate (crossflow velocity) and transmembrane pressure. Figure 1 presents a schematic diagram of the experimental setup. All tests were performed at a crossflow velocity of 0.10 m/s (0.33 ft/s), typical of a full-scale water production system. Control of the transmembrane pressure and the crossflow velocity at constant values required a permeate recovery rate (permeate flow rate 100%/feed flow rate), lower than the full scale average of 20% in a single pass. The recovery rate in this study was 10% or less and is given in Table 1.
Fig. 1. Experimental nanofiltration membrane system.
4428
Amy K. Zander and N. Kevin Curry Table 1. Membrane information
Membrane
SEPA BP02
YM10
SEPA RG03
Manufacturer Composition of surface layer Surface character MWCOd Recommended TMPd TMP employed pH ranged Recovery rate Initial specific productivity, SP0, L/m2 h kPa (gfd/psi)e
Osmonics, Minnetonka, MN ‘like’ sulfonated polysulfonea Anionic, hydrophilica 1000–5000 daltons 690 kPa (100 psi) 620 kPa (90 psi) 0.5–13 3% 0.06070.007 (0.24370.029)
Amicon Beverly, MA Regenerated cellulose acetateb Non-ionic, hydrophilicb,c 10,000 daltons 380 kPa (55 psi) 380 kPa (55 psi) 3–11 7% 0.27170.026 (1.1070.10)
Osmonics, Minnetonka, MN Acrylonitriled Non-ionic, hydrophobicd 4000–10,000 daltons 690 kPa (100 psi) 620 kPa (90 psi) 0.5–10 10% 0.25870.049 (1.0570.20)
a
Harrold (1996). Laı´ ne´ et al. (1989); Braghetta and DiGiano (1994); Braghetta et al. (1997). Clark and Jucker (1993). d As stated by the manufacturer. e This work. b c
Membranes studied The three membranes used in this study were chosen for their contrasting materials of composition, and the resulting variation in surface hydrophilicity and charge. Pertinent information regarding the membranes is tabulated in Table 1. The SEPA BP02 membrane is most similar in surface composition and MWCO to the membranes used by Braghetta et al. (1997) in a single membrane study of related pH and ionic strength effects in nanofiltration. Sulfonated polysulfone is rapidly becoming a popular choice for nanofiltration membranes, due to its perceived ability to resist fouling. The YM series of membranes have been shown by Clark and Jucker (1993) to be relatively unaffected by a change in solution pH and ionic strength. These membranes are hydrophilic like the BP02, but they do not possess a charged surface functional group. The RG03 membrane is non-ionic, similar to the YM10, but the membrane surface is hydrophobic, potentially increasing the adsorbance of hydrophobic materials from the solution and increasing the rate of decline in specific productivity. Membrane preparation The membranes were cleaned by soaking in 0.5 L of HCl solution at pH 2 for 30 min, followed by 0.5 L of NaOH solution at pH 10.5 for 30 min, using a procedure recommended by Allgeier and Summers (1995b). All membranes were then soaked in deionized water for 1 h, with at least three water changes, followed by filtration of approximately 1 gal (4 L) of deionized water through the membrane in the test cell. This allowed the removal of any impurities or preservatives from the membranes (Amicon, 1995). Each membrane was then compacted by operation under the desired operating pressure in the test cell, using pH 7 buffered water, until a stable flux was reached. This usually occurred within 20 h. The membrane was then characterized by a specific productivity measurement, which served as the initial specific productivity (SP0). These initial specific productivity measurements are given in Table 1. Test solution preparation The matrix of 18 test solutions used was chosen to represent a broad range of naturally occurring water conditions from a soft to hard water, river water to groundwater with saltwater intrusion. The solutions varied three pHs (4.0, 7.0 and 8.5), three ionic strengths (1.4 10@3, 6.0 10@3, and 1.4 10@2 M with approximate conductivities of 100, 700, and 1500 mmhos/cm), and either a monovalent (Na+) or divalent (Ca2+) predominant cation. Buffer materials accounted for 10% or less of the
solution ionic strength in all the cases. Phosphorous acid (pH 4.0 and 7.0; Aldrich, Milwaukee, WI) and boric acid (pH 8.5; J.T. Baker, Phillipsburg, NJ) served as buffer materials. The remaining ionic strength was made up of sodium chloride dihydrate or calcium chloride (J.T. Baker, Phillipsburg, NJ), as appropriate. Each solution also contained 10 mg/L of organic carbon added as 0.45 mm prefiltered commercially available humic acid (Lot #01828JZ, Aldrich, Milwaukee, WI). Despite some known differences between Aldrich humic acid and naturally occurring aquatic humic substances, the commercial humic acid has been found to be satisfactory as a surrogate for NOM for elucidating some of the mechanisms of membrane flux decline (Wiesner et al., 1989). Apparent molecular size fractionation of the humic material was performed following the protocol described by Owen et al. (1993). At a pH of 8.5 and an ionic strength of 0.0014 M the Aldrich humic acid had an apparent molecular size distribution of 67% greater than 30 kDa, 27% between 10 and 30 kDa, 3% between 5 and 10 kDa and 3% less than 5 kDa. At this fairly high pH and low ionic strength, the humic acid is likely to be in the form of a highly negatively charged, flexible, linear macromolecule (Braghetta et al., 1997). From this analysis, size exclusion alone should allow for 94% or greater rejection of TOC for all of the membranes studied. Analytical methods Organic concentrations were monitored by TOC analysis (Method 5310 B; APHA, AWWA, WEF, 1992) using a Dohrmann DC-190 TOC Analyzer (Rosemount Analytical Inc., Santa Clara, CA). Injection of large size samples (400 mL) allowed the detection of TOC to 0.2 mg/L and a relative standard deviation (RSD=standard deviation/ mean) of multiple TOC determinations of 0.044. Percent rejection of TOC (and conductivity) was determined using the following expression: R ¼ ðCF @CP Þ100%=CF where R is the percent of solute rejected (retained) by the membrane, CF the solute concentration in the feed water, and CP the concentration of solute in the permeate. Rejection was determined by averaging the calculated % rejection values following the observation of steady state rejection, which occurred soon after the system startup. Ionic strength was measured using electrical conductivity measurements (Method 2510; APHA, AWWA, WEF, 1992) as a surrogate parameter, and with a conductivity meter (Cole Parmer, Model #01481-61, Niles, IL). Measurement of solution pH was performed according to Method 4500-
Solution effect on rejection and productivity H+ (APHA, AWWA, WEF, 1992) using a pH meter (Cole Parmer, Model #05669-201, Niles, IL). Specific productivity measurements The specific productivity as used in this article refers to the flux of water through a membrane divided by the transmembrane pressure at which the flux was measured. Other works have called this term a water mass transfer coefficient (MTCW) (e.g. Morris and Taylor, 1991; Reiss and Taylor, 1991; Allgeier and Summers, 1995b; Taylor and Mulford, 1995; Curry and Zander, 1997), however the authors (and others, e.g. Fu et al., 1994) find specific productivity to be more precise description of the term. The time rate of decay of the specific productivity is an accurate procedure for measuring membrane productivity changes (Taylor and Mulford, 1995) and the rate of flux decline due to fouling (Morris and Taylor, 1991). Each specific productivity measurement was determined as the slope of a graph of eight individual measurements of flux vs. corresponding applied transmembrane pressure. All measurements for this were obtained at a feedwater cross-flow velocity of 0.10 m/s (0.33 ft/s) using deionized water at a pH of 5.6. Flux measurements were taken as soon as a stable flux developed for each pressure applied. Temperature was controlled within a few degrees Celsius and monitored throughout the tests. Flux measurements were corrected to flux at 201C using the correlation from Jacangelo et al. (1994) based on the variation of water viscosity with temperature:
Flux201C ¼ Meaured Flux eð@0:0239ðT@20ÞÞ where T is the temperature in 1C at which the flux was measured. The units of specific productivity are given as a flux per unit pressure: gal of permeate per square foot of membrane area per day per pound per square inch of applied pressure (gfd/psi) or liters of permeate per square meter of membrane area per h per kPa of applied pressure (L/m2 h kPa). These specific productivity determinations displayed excellent precision, with the coefficient of determination, R2 , greater than 0.99. The relative standard deviation of these measurements was less than 0.02.
4429
The rate of decline in specific productivity was determined by fitting a linear regression to data on specific productivity changes with time. As attempted in other studies (Solomon et al., 1993), the decline in specific productivity (flux decline) was fitted to several expressions, such as exponential decay, logarithm, power functions and linear regression. The best fit to all data was found to be a linear approximation of decline during the initial operation of the membrane system. The coefficients of determination, R2 , for these linear fits averaged 0.90 for the YM10 and RG03 membranes, but had a relatively poor mean fit ðR2 ¼ 0:66Þ for the BP02 membrane. This relatively poor fit may be a function of the small changes in specific productivity for this membrane over the course of the experiments. However, the relative magnitude of the specific productivity decline is still comparable between membranes.
EXPERIMENTAL RESULTS AND DISCUSSION
TOC rejection Figures 2–4 display the TOC rejection attained by each membrane (BP02, YM10, and RG03, respectively) under steady state conditions in each of the eighteen test solutions. The bars in each graph indicate the mean value of TOC rejection under each condition and the error bars indicate the relative standard deviation of the tests. Observation of Fig. 2 reveals no statistically significant differences in TOC rejection with pH or ionic strength for the BP02 membrane in solutions made with monovalent cation, though the overall TOC rejection was slightly less than the approximately 97% rejection expected from the apparent molecular size distribution of the humic acid. The expected trend of increasing rejection with increasing pH, though not experimentally substantiated here, is clearly not refuted. In the solutions made with a divalent cation, a trend toward a reduction in TOC
Fig. 2. The TOC rejection as a function of pH and ionic strength for BP02.
4430
Amy K. Zander and N. Kevin Curry
Fig. 3. The TOC rejection as a function of pH and ionic strength for YM10.
Fig. 4. The TOC rejection as a function of pH and ionic strength for RG03.
rejection as compared to monovalent cation solutions can be seen throughout the range of pH and ionic strength. This is likely due to the reduction in the thickness of the double layer on the membrane by the presence of Ca2+. This reduction is greatest at the highest pH, due to the effect of the Ca2+ on the configuration of the humic acid which is highly negatively charged at this pH. It is also possible that the charge repulsion experienced by the negatively charged humic acid in proximity to the negatively charged membrane is partially overcome by the presence of the divalent cation.
Similar results can be seen in Fig. 3 for the hydrophilic, nominally non-ionic YM10. The larger nominal MWCO for this membrane (10,000 DA as opposed to 5000 Da for the BP02) results in a slightly lower rejection of TOC in monovalent solution. Again, no statistically significant changes could be seen with pH or ionic strength for the monovalent solution. Exchange of the monovalent cation for a divalent cation results in a general reduction in TOC rejection, again potentially due to the effect of Ca2+ on the configuration of the humic acid, or the effect of the divalent cation on the electrostatic repulsion
Solution effect on rejection and productivity
between membrane and humic acid. Though the manufacturer touts this membrane as non-ionic, evidence exists that it actually bears a slightly negative potential in solution. A negative zeta potential was measured for this membrane under similar solution conditions using a streaming potential anlayzer (BI-EKA, Brookhaven Instruments, Holtsville, NY), (Zander and Desrosiers, 1999) and a negative zeta potential was observed by Elimelech et al. (1994) for other cellulose acetate membranes. The TOC rejection measured for the third membrane, RG03, is given in Fig. 4. This non-ionic, hydrophobic membrane behaves quite differently in solution than the other two membranes. In several of the solutions tested, the overall TOC rejection is greater for the solution containing divalent rather than monovalent cations. No significant difference is seen in the TOC removal with pH or ionic strength with divalent cations, though the overall rejection is slightly less than expected by steric hindrance using the manufacturers MWCO. The major mechanism for rejection in this case is likely to be the attraction of the humic acid and Ca2+ complex to the uncharged hydrophobic membrane surface. One can justify the case by the relatively high rate of decline in specific productivity for this membrane in divalent solution as seen in Fig. 7 and discussed in further detail later. Conductivity rejection Conductivity rejection varies greatly between the membranes and, for the BP02, between solution conditions. Figure 5 displays the conductivity rejection for the three membranes at the lowest ionic strength monovalent cation solution. This solution condition is also the only solution condition under which any measurable conductivity rejection was
4431
seen for the YM10 membrane (data not shown). Some slight conductivity rejection was seen for the RG03 in monovalent solution at the medium ionic strength (data not shown). In the presence of the divalent cation, and at high ionic strength in the monovalent solutions, essentially no conductivity rejection was seen for the YM 10 or RG03. This would be the expected result for these membranes. This illustrates that the rejection of salts under low ionic strength conditions is a function of interactions between the membrane and solution. At low ionic strength, especially with monovalent cations, the double layer surrounding the membrane is thicker than under the other conditions studied, allowing for greater rejection of charged solutes. As the double layer thickness decreases with increasing ionic strength, the charged solutes easily pass through the membrane. The anionic BP02 membrane has electrostatic repulsion as an additional mechanism of conductivity rejection, hence the significantly higher rejection for this membrane. A strong effect of ionic strength on conductivity rejection can also be seen for the BP02 in Fig. 6. For both the monovalent and divalent cation solutions, the rejection of conductivity decreases with increasing ionic strength. This is again a function of the effect of the solution on the double layer thickness surrounding the membrane. The double layer thickness will decrease with increasing ionic strength, and more so in the presence of divalent cations, leading to an increase in the area available for the passage of charged species. It is also apparent in Fig. 6 that at least for tests in the divalent solution, an increase in pH results in a decrease in rejection. This could be explained by a charge balancing of the negatively charged membrane by hydronium ions in solution and a reduction in electrostatic repulsion.
Fig. 5. Conductivity rejection for three membranes at 0.0014 M monovalent ionic strength.
4432
Amy K. Zander and N. Kevin Curry
Fig. 6. Conductivity rejection as a function of pH and ionic strength for BP02.
Fig. 7. Decline in specific productivity as a function of pH and divalent ionic strength for three membranes.
The observation that in this case the divalent ions are rejected less than the monovalent ions would seem to contradict the more common observation that NF membranes reject divalent cations very well (Nystrom et al., 1995; Hong and Elimelech, 1997). This is probably due to the fairly ‘loose’ pore size characteristics of the BP02 membrane. This membrane may be better described as an ultrafiltration type membrane. Productivity decline As membrane fouling is a major contributor to the operational cost of membrane systems, it is important to understand the factors involved in fouling.
Fouling here is measured as the decline in specific productivity of the membrane and the rate of that decline is assumed to be indicative of the operational cost involved in minimizing fouling or in membrane cleaning. A lower rate of specific productivity decline indicates a lower operational cost for the membrane under those conditions. For each of the membranes in the monovalent cation solutions the rate of decline is fairly low, generally less than 1%/h (data not shown). When the major cation in solution is changed from monovalent to divalent there is a significant increase in the rate of specific productivity decline from less than 1%/h to values up to 50%/h or more as seen in Fig. 7. This effect has been observed previously (e.g.
Solution effect on rejection and productivity
Hong and Elimelech, 1997) and is due to the interaction of the divalent cations with carboxyl functional groups on the humic material. This reduces electrostatic repulsion between the macromolecules and they form a dense surface layer on the membrane. Many of the solution conditions lead to productivity decline rates greater than 10%/h. The BP02 was the least affected by fouling, even in divalent solution. Adsorption of materials to the RG03 membrane surface at low ionic strength conditions led to a decline in productivity greater than 50%/h. At this very high decline rate, quantification is difficult as the flux rate is dropping rapidly. Decline rates greater than 50%/h are reported simply as greater than this value. In one test, carried out at a pH of 4.0, the RG03 membrane became completely fouled in less than 1 h. The fouling at the membrane surface of humic acid associated with Ca2+ is the likely reason for the measurable rejection of conductivity with this membrane under these conditions. The initial SP0 of each membrane is given in Table 1. The SP0 value for each membrane was relatively constant across test solution conditions. Changes in SP0 with time in a given test condition were generally much greater than the difference in the initial membrane productivity for a given membrane under differing solution conditions. Two of the membranes, RG03 and YM10, had similar values of SP0 at approximately 0.26 and 0.27 L/m2 h kPa, respectively. The BP02 membrane had a value of less than onefourth the value of the others, with an SP0 of 0.06 L/ m2 h/kPa. Comparing these initial productivity values, the BP02 would require more than 4 times the membrane surface area to achieve the same permeate flow rate for a similar applied pressure as the other membranes studied, greatly increasing the capital cost of a membrane plant.
CONCLUSIONS
Organic matter and conductivity rejection are not a direct function of manufacturers stated membrane MWCO. Instead, rejection is a complex interaction of steric hindrance, electrostatic repulsion, solution effects on the membrane and solute shape, and thickness of the double layer surrounding the membrane in solution. Presence of divalent cations in solution greatly affects the membrane performance. Solutions containing divalent rather than monovalent cations showed lower rejection of both aqueous conductivity and TOC. Specific permeability decline measurements are a useful tool for comparison of solution effects on membrane performance. At the lowest ionic strength tested, the more hydrophobic membrane had a greater rate of specific flux decline than hydrophilic membranes. Of the variables tested, membrane surface chemistry is the
4433
most important factor affecting the rate of flux decline. Ignoring the effect of membrane chemistry, specific flux decline is greatest for low ionic strength solutions. Divalent cations increased the rate of decline in specific flux significantly over monovalent cations, especially for the YM10 and the RG03 membranes. AcknowledgementsFThis work was sponsored by Grant #BES-9307851 from the National Science Foundation.
REFERENCES
Aiken G. R. (1984) Evaluation of UF for Determining Molecular Weight of Fulvic Acid. Environ. Sci. Technol. 18, 978–981. Allgeier S. C. and Summers R. S. (1995a) Evaluating NF for DBP control with the RBSMT. J. Am. Water Works Assoc. 87(3), 87–99. Allgeier S. C. and Summers R. S. (1995b) Development of a Rapid Bench Scale Membrane Test for the Evaluation of NF Performance. Proceedings of the 1995 AWWA Membrane Technology Conference, Reno, NV, pp. 207–227. Amicon Inc. (1995) Laboratory Products Catalog, Beverly, MA. APHA, AWWA, WEF. (1992) Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington, DC. Braghetta A. and DiGiano F. A. (1994) Organic solute association with nanofiltration membrane surface: Influence of pH and ionic strength on membrane performance. Proceedings of the 1994 AWWA Annual Conference, New York, NY. Braghetta A., DiGiano F. A. and Ball W. P. (1997) Nanofiltration of natural organic matter: pH and ionic strength effects. J. Environ. Eng. 123(7), 628–641. Childress A. E. and Elimelech M. (2000) Relating nanofiltration membrane performance to membrane surface charge (Electrokinetic) characteristics. Environ. Sci. Technol. 34(17), 3710–3716. Cho J. W., Amy G. and Pellegrino J. (1999) Membrane filtration of natural organic matter: initial comparison of rejection and flux decline characteristics with ultrafiltration and nanofiltration membranes. Water Res. 33(11), 2517–2526. Clark M. M. and Jucker C. (1993) Interactions between hydrophobic ultrafiltration membranes and humic substances. Proceedings of the 1993 AWWA Membrane Technology Conference, Baltimore, MD, pp. 259–272. Cornel P. K., Summers R. S. and Roberts P. V. (1986) Diffusion of humic acid in dilute aqueous solution. J. Colloid Interface Sci. 110(1), 149–164. Curry N. K. and Zander A. K. (1997) Effect of aqueous solution chemistry on UF and NF membrane performance. Proceedings of the 1997 AWWA Annual Conference, Atlanta, GA, pp. 603–609. Elimelech M., Chen W. H. and Waypa J. J. (1994) Measuring the Zeta (Electrokinetic) potential of reverse osmosis membranes by a streaming potential analyzer. Desalination 95, 269–286. Fu P., Ruiz H., Thompson K. and Spangenberg C. (1994) Selecting membranes for removing NOM and DBP precursors. J. Am. Water Works Assoc. 86(12), 55–72. Ghosh K. and Schnitzer M. (1980) Macromolecular structures of humic substances. Soil Sci. 129(5), 266–276. Harrold S. J. (1996) Personal Communication, Osmonics, Inc. Minnetonka, MN.
4434
Amy K. Zander and N. Kevin Curry
Hong S. and Elimelech M. (1997) Chemical and Physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. J. Membrane Sci. 132, 159–181. Jacangelo J. G., Laı´ ne´ J.-M., Cummings E. W., Deutschmann A., Mallevialle J. and Wiesner M. R. (1994) Evaluation of UF Membrane Pretreatment and NF of Surface Waters. AWWA Research Foundation and AWWA, Denver, Co. Jonsson G. and Boeson C. E. (1984) Polarization phenomena in membrane processes. Synthetic Membrane Process Fundamentals and Water Applications, ed. G. Belfort. Academic Press, Orlando, FL. Lahoussine-Turcaud V., Wiesner M. R., Bottero J.-Y. and Mallevialle J. (1990) Coagulation pretreatment for ultrafiltration of a surface water. J. Am. Water Works Assoc. 82(12), 76–81. Laine´ J.-M., Clark M. M. and Mallevialle J. (1989) Effects of ultrafiltration membrane composition. J. Am. Water Works Assoc. 82(11), 61–73. Manttari M., Puro L., Nuortila-Jokinen J. and Nystrom M. (2000) Fouling effects of polysaccharides and humic acid in nanofiltration. J. Membrane Sci. 165(1), 1–17. Marinˇas B. and Selleck R. E. (1992) Reverse osmosis treatment of multicomponent electrolyte solutions. J. Membrane Sci. 72, 211–229. Morris K. E. and Taylor J. S. (1991) The use of fouling indices to predict RO membrane fouling. Proceedings of the 1991 AWWA Membrane Processes Conference, Orlando, FL. Nystrom M., Kaipia L. and Luque S (1995) Fouling and retention of nanofiltration membranes. J. Membrane Sci. 98, 249–262. Nystrom M., Lindstrom M. and Matthiasson E. (1989) Streaming potential as a tool in the characterization of UF membranes. Colloids and Surfaces 36, 297–312. Osmonics, Inc. (1995) SEPAs CF Membrane Cell Instruction Manual. Minnetonka, Minnesota. Owen D. M., Amy G. L. and Chowdury Z. K. (1993) Characterization of Natural Organic Matter and its Relationship to Treatability. AWWA Research Foundation and AWWA, Denver, CO.
Peterson R. J. (1993) RO/NF membrane performance characteristics. Proceedings of the 1993 AWWA Membrane Technology Conference, Baltimore, MD, pp. 23–39. Reiss C. R. and Taylor J. S. (1991) Membrane pretreatment of a surface water. Proceedings of the 1991 AWWA Membrane Processes Conference, Orlando, FL. Schafer A. I., Fane A. G. and Waite T. D. (1998) Nanofiltration of natural organic matter: removal, fouling and the influence of multivalent Ions. Desalination 118(1–3), 109–122. Schweitzer P. A. (1979) Handbook of Separation Techniques for Chemical Engineers, 2nd ed. McGraw Hill, New York. Seacord T. F. (1998) Assessing nanofiltration membrane properties for water treatment applications, M.S. Thesis, Clarkson University, Potsdam, NY. Solomon T. D., Summers R. S. and Speth T. F. (1993) The reduction of membrane fouling by physical, chemical and biological pretreatment. Supplement to the Proceedings of the 1993 AWWA Membrane Technology Conference, Baltimore, MD. Taylor J. S. and Mulford L. A. (1995) Membrane protocol to meet the ICR. Proceedings of the 1995 AWWA Membrane Processes Conference, Reno, NV. Van der Bruggen B., Schaep J., Wilms D. and Vandecasteele C. (1999) Influence of molecular size, polarity and charge on the retention of organic molecules by nanofiltration. J. Membrane Sci. 156(1), 29–41. Vrijenhoek E. M., Elimelech M. and Hong S. K. (2000) Influence of membrane properties, solution chemistry, and hydrodynamics on colloidal fouling of reverse osmosis and nanofiltration membranes. Abst. Papers Am. Chem. Soc. Part 1, 220, 40-ENVR. Weber W. J. Jr. (1971) Physicochemical Processes for Water Quality Control. Wiley, New York, NY. Wiesner M. R., Clark M. M. and Mallevialle J. (1989) Membrane filtration of coagulated suspensions. J. Environ. Eng. ASCE 115(1), 20–40. Zander A. K. and Desrosiers K. (1999) Arsenite removal from water using UF/NF membranes. Proceedings of the 1999 AWWA Annual Conference, Chicago, IL.