Environ. Sci. Technol. 2005, 39, 6487-6493
Ion Implantation: Effect on Flux and Rejection Properties of NF Membranes† JOSHUA OLUFEMI ABITOYE,‡ PARNA MUKHERJEE,§ AND K I M B E R L Y J O N E S * ,‡ Department of Civil Engineering, Howard University, Washington D.C. 20059, and Civil and Environmental Engineering Department, California Polytechnic State University, San Luis Obispo, California 93407
Nanofiltration (NF) membranes typically carry a net electric charge, enabling electrostatic interactions to play a pivotal role in the rejection of species such as metals, nitrates, and other charged contaminants. In this study, two types of polymeric NF membranes, polyamide and cellulose acetate, were modified by ion implantation to increase the effective surface charge of the membranes. The modified membranes contain implanted ions in the membrane matrix, inducing a discrete, permanent charge in the active membrane layer. The presence of a permanent charge in the membrane matrix allows for increased electrostatic repulsive forces throughout the entire pH range. Streaming potential measurements were conducted as a function of pH for the modified and unmodified membranes to determine the effect of ion implantation on the zeta potential of the membranes. Rejection experiments were performed in order to quantify the effect of increased electrostatic repulsion on ion rejection, and flux measurements quantified the effect of the modification on permeability. Results indicate that electrostatic interactions near the membrane surface can affect rejection; however, the extent of the effect of increased membrane charge depends on physical-chemical characteristics of the membrane. Increased negative zeta potential of the modified membranes resulted in slightly higher rejection of salts with divalent co-ions from the membrane, with less increase observed with salts of monovalent co-ions. Modified membranes were less permeable than the unmodified membranes. Results of this research hold implications in membrane synthesis and modification studies as well as choice of membranes for water treatment applications.
Introduction Nanofiltration (NF) membranes have properties between ultrafiltration (UF) and reverse osmosis (RO) membranes in terms of rejection and flux. Many commercially available NF membranes are negatively charged throughout the upper pH range (1-3), allowing electrostatic interactions between the membrane and ions in solution to affect transport of * Corresponding author phone: (202)806-4807; fax: (202)8065271; e-mail:
[email protected]. † This paper is part of the Charles O’Melia tribute issue. ‡ Howard University. § California Polytechnic State University. 10.1021/es050102v CCC: $30.25 Published on Web 07/09/2005
2005 American Chemical Society
charged organic and inorganic solutes through the membrane (4-7). The charge of the membranes is pH dependent, as the charge evolves from dissociation or deprotonation of functional groups such as sulfonic or carboxyl groups within the polymeric structure of the membrane. Electrical properties of NF membranes significantly affect separation efficiency, as NF membranes are used to remove hardness (8), heavy metals such as arsenic (9, 10), endocrine disrupting compounds (11, 12), dissolved organics (8, 13), and pharmaceutical and biomedical compounds (14, 15) from aqueous solutions. Models such as the extended Nernst-Planck (16-18) and solution-diffusion models (19-21) have been widely applied to NF and RO membranes to describe ion transport across membranes. The extended Nernst-Planck model describes mechanisms for ion rejection in NF membranes as steric hindrance due to size of permeating solute, Donnan exclusion due to electrostatic interaction, and dielectric exclusion due to polarization of the membrane/water interface. Thus, the membrane pore size, membrane charge, and solute properties are important parameters in determining flux (and thus rejection) of solute across the membrane surface. The solution diffusion model assumes that the electrolytes (dissociated cations and anions) first get dissolved in the membrane water interface and then permeate through the membrane in the product water. Transport is then dependent on three mass transport coefficientssbulk solution (transport from bulk to boundary layer), membrane-solute (transport from boundary layer into membrane), and diffusion (transport through membrane). Increased electrostatic interactions will alter the membrane-solute mass transfer coefficient to reduce transport into the membrane, changing the overall solute permeability coefficient of the ion-implanted membranes as compared to the unmodified or virgin membranes (21, 22). Some researchers have attempted to exploit the inherent charged nature of NF membranes by operating within a defined pH value or utilizing electric fields or using a higher charged membrane to increase rejection (23-25). In most studies of the effect of electrostatic interactions on membrane fouling, changes in both the membrane and solute charge were due to ionized functional groups. It would be beneficial to add a fixed charge to the membrane structure to increase the membrane surface charge. Conceptually, the magnitude and sign of the charge could be specifically designed to optimize different environmental applications (water treatment, wastewater treatment, removal of specific contaminants). In this study, F- was implanted into the membrane matrix to induce a discrete, permanent negative charge on the surface. Although charged ion exchange membranes have been used to exploit the effect of charge on solute rejection, it is important to note that the strategy described here involves injecting quasi-permanent charge into the active layer of the membrane; the exchange of ions, which is the dominant mechanism in ion exchange membranes, is not expected in this case. The presence of a permanent charge on the membrane surface avoids the case where the measured membrane charge may change with different salt mixtures (26). Most importantly, a negatively charged surface should increase electrostatic repulsion and Donnan exclusion mechanisms at the membrane-bulk solution interface.
Experimental Section Membranes. Two types of NF membranes were used in this study (Table 1): a cellulose acetate membrane, denoted as SP28 (Osmonics Inc., Minnetonka, MN), and a thin film VOL. 39, NO. 17, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6487
TABLE 1. Membrane Specifications membrane membrane material membrane thickness, µm active layer thickness, µm pH range at 25 °C temperature range, °C
SP-28 cellulose acetate 177.8 0.2 2-8 0-135
NF-90 polyamide 177.8 0.2 2-11 0-122
polyamide membrane, denoted as NF 90 (FilmTec Inc., Minneapolis, MN). Membranes were supplied dry in rolled sheets and were stored as received at room temperature. Once used, membranes were stored in deionized water at 3 °C. Ion Implantation. Membrane coupons were cut from dry membrane sheets and sent for ion implantation (Implant Sciences Corporation, Massachusetts). Fluoride ion (F-) was chosen for implantation because of its high electronegativity. The implantation was carried out at two different dosing intensitiessa low and a high dose with F- ion concentrations of 1 × 10 atoms/cm2 and 5 × 10 atoms/cm2, respectively. Ions were implanted into the active layer of the membrane (Figure 1) using an applied vacuum with a linear accelerator to create a beam of charged ions. The ion beam was then shaped and directed toward the membrane surface, embedding ions into the active layer of the membrane. The depth of implantation is a function of membrane material and the ion energy used for implantation. It was desirable to have the implanted ions deep enough in the membrane surface to avoid leaching but not too deep as to prevent electric fields to extend to the boundary layer near the membrane surface. Appropriate ion energies and corresponding projected depth for fluoride ion in polyamide and cellulose acetate were determined using Stopping and Range of Ions in Matter (SRIM 2000) software. A low ion energy was used for the implantation process in order to minimize potential damage to the membrane surface. Specifications for implantation are as summarized in Table 2. Rejection and Flux Experiments. Rejection and flux experiments were carried out in batch mode using a high pressure stirred cell (Sterlitech Corporation, Washington). The cell is made from 316 stainless steel and has a maximum pressure rating of 1000 psi and a low hold up volume (1 mL). Concentration polarization in the cell is minimized by a Teflon-coated magnetic stirring bar, which is driven by a magnetic stirring plate. The membranes were cut to a diameter of 47 mm to fit the cell, and nitrogen was used to deliver pressure to the stirred cell. The membrane area in the unit is 11.95 cm2. A schematic of the filtration set up is shown in Figure 2. Flux experiments were conducted using purified water at different transmembrane pressures. The permeability was calculated as the slope of the linear flux versus pressure line. A limited number of experiments were also conducted with water with different dielectric constant values. These experiments were conducted at constant pressure of 100 psi with water at three dielectric constant values: ) 80 (DI water), ) 63.29 (70% water and 30% ethanol solution), and ) 52.15 (50% water and 50% ethanol solution). Note that the dielectric constant of water at 80 °F ) 80 and the dielectric constant of ethanol at 77 °F ) 24.3. No salt was added to any of the feed solutions for these experiments. Rejection experiments were carried out at a constant flux of 3.5 × 10-6 m/s. The duration for each filtration run was 1-2 h with samples collected at equal time intervals. The stirring speed was constant throughout the experiment at approximately 600 rpm. The permeate volume was measured using a graduated cylinder and was analyzed to determine flux (J) and salt rejection (R). At each time interval, feed and permeate samples were analyzed for salt concentrations and 6488
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 17, 2005
fluoride concentrations. No fluoride was detected in any of the samples. A clean water flux test was performed before and after each rejection experiment. If initial and final steadystate flux values were constant, membrane coupons were reused. The water flux (J) was calculated as
J)
V A‚t
(1)
where V is the volume collected at a particular time (t) and A is the active surface area of the membrane in contact with the solution. The observed rejection (R) of the solute was calculated using eq 2
R ) 100
() Cp Cf
(2)
where Cp is the solute concentration in the permeate stream and Cf is the solute concentration in the feed solution. Salt Solutions. Reagent grade single salt solutions of potassium chloride (KCl), sodium nitrate (NaNO3), and sodium sulfate (Na2SO4) were used in rejection experiments. All salt solutions were obtained from Sigma Chemical Company, St. Louis, MO. The feed salt concentration was 0.005 M for all experiments. This relatively low salt concentration was chosen to minimize the Donnan potential at the interface between the charged membrane and salt solution. Salt solutions were prepared using water which was processed by a reverse osmosis membrane, GAC adsorber, and a mixedbed deionizer. A dual column Ion Chromatograph DX-120 (Dionex Corporation) was used for ion analysis in the feed and permeate. Membrane Characterization. Membranes were characterized by streaming potential measurements to estimate charge and by uncharged solute rejection experiments to estimate pore size. These characterization experiments were conducted on the virgin membranes (unmodified) and the membranes after implantation (modified). Membrane zeta potential was measured using an Electro Kinetic Analyzer (EKA) (Brookhaven Instruments Corp., NY), which utilizes streaming potential to calculate zeta potential by the Helmholtz-Smoluchowski equation. Membrane and solution preparation for these experiments was as described in ref 27. Zeta potential measurements are sensitive to background electrolyte utilized (3); therefore, these experiments were performed with 0.001 M KCl. This dilute salt concentration is lower than the solution conditions for the rejection experiments. The ZP conditions were chosen in order to minimize the effect of the salt solution on the zeta potential of the membrane, as zeta potential decreases with increasing salt concentration (or ionic strength). Higher salt concentration will decrease the thickness of the diffuse double layer, which will consequently decrease the zeta potential. Rejection experiments using uncharged solutes were utilized to yield the pore radius of the modified and unmodified membranes (28, 29). Rejection and flux data using 20 mg/L glucose and vitamin B-12 (Sigma-Aldrich Company, MI) were fit to transport models as in ref 16. Organic concentrations in the permeate and feed samples were analyzed using a UV-visible spectrophotometer (Shimadzu Scientific, Columbia, MD) and Phoenix 8000 TOC analyzer (Tekmar Company, OH). The experiments were carried out at varied incremental pressures until an asymptotic rejection value was achieved. It should be noted that the calculated pore radius represents an effective pore size; this estimation was considered appropriate, as the pore size distribution of NF membranes is typically narrow (30).
FIGURE 1. Schematic of implanted membrane adapted from ref 22.
TABLE 2. Specifications for Ion Implantation ion species implantation depth, SP28/NF 90, µm low ion dose/high ion dose, atoms/cm2 ion beam energy, keV sample temperature gas carrier beam current, µA source vacuum, Torr (N/m2) implant time, s
fluoride ion 0.0368/0.0281 1 × 10 10/5 × 1010 10 room BF3 0.01 6 × 10-7 (8.0 × 10-5) 33
Results and Discussion Results are presented for each type of membrane (cellulose acetate SP28 and polyamide NF90) separately, with results of unmodified and modified membranes (high and low dose) presented on the same graph for comparative purposes. Zeta Potential. Zeta potential measurements for the modified and unmodified membranes are shown in Figure 3 as a function of solution pH. The decrease in zeta potential as pH increases is typical of NF membranes (1, 27, 31). Polyamide and cellulose acetate membranes have carboxyl groups which are protonated at low pH to bring about a positive charge (positive zeta potential); the carboxyl groups will deprotonate as pH increases, which will make the membrane charge more negative (negative zeta potential). Also, at the isoelectric point (iep), the membrane has a net charge (zeta potential) of zero. Above the iep, the zeta potential becomes negative due to deprotonation of the functional groups of the membrane. All zeta potential curves (modified and unmodified) show characteristics of amphoteric surfaces, i.e. surfaces with both acidic and basic functional groups. The basic mechanism for charging by deprotonation of functional groups is still apparent in the modified membranes, although the absolute value of the zeta potential is more negative due to Fimplantation. The unmodified polyamide membrane (NF90) has an isoelectric point (iep) of 4.2; the polyamide membrane modified with a low F- dose has an iep of 3.9, and the polyamide membrane modified with a high F- dose has an iep of 3.4. The cellulose acetate (SP28) membranes are negatively charged throughout the entire pH range; however, the zeta potential is more negative for the low F- dose at each pH value. The high F- dose resulted in a less negative zeta potential for the cellulose acetate membrane, and these results are included for completeness; however, the higher ion dose likely damaged the cellulose acetate membranes. Cellulose acetate is a weak ion exchange material and has
a relatively narrow pH range due to susceptibility to hydrolysis. It is possible that the higher ion dose may have resulted in a breaking of the C-H bonds leaving dangling C-bonds, which may link together and form a more rigid ladderlike structure (32). This new structure may be less electronegative than the original structure of the SP28 membrane resulting in a less negative zeta potential. For both membranes, and at each pH, the zeta potential decreased in a consistent manner after implantation. Implanted F- ions represent a fixed charge that exists along with the pH and ionic strength-dependent membrane charge typically found on NF and RO membranes, resulting in increased charge density of the membrane. The zeta potential difference between the unmodified and modified NF90 membranes at pH 6.5, where zeta potential values were -20 mV (unmodified), -24.0 mV (low dose), and -27.5 mV (high dose). Zeta potential values for the SP28 membranes were -22 mV (unmodified) and -25 mV (low dose). Subsequent permeation and rejection experiments were conducted at pH 6.5. Permeability Measurements. Permeability experiments were performed in order to determine if the membrane pore structure was changed as a result of implantation process. Clean water flux tests were performed on the unmodified and modified membranes before and after each experiment. Steady-state clean water flux values at a pressure of 100 psi for the unmodified, low dose modified, and the high dose modified NF90 membrane were 5.67 × 10-6 m/s, 3.26 × 10-6 m/s, and 2.00 × 10-6 m/s, respectively. These values for the SP28 membrane were 5.09 × 10-6 m/s, 2.44 × 10-6 m/s, and 1.92 × 10-6 m/s for the unmodified, low dose and high dose SP28, respectively. Results from permeability experiments are shown in Table 3. In each case, membrane permeability decreased after implantation, with higher increase in permeability at the higher ion dose. An obvious concern is that the membranes may have been damaged during the ion implantation process. AFM images of the membranes indicated that the basic surface roughness, porosity, and pore size values were unchanged for both membranes before and after implantation. Roughness RMS values decreased from 5.18 nm for the unmodified membrane to 4.98 nm for the high dose membrane. The RMS values for the polyamide membrane increased slightly from 16.4 to 18.0 nm. Although the AFM micrographs are consistent before and after implantation, clean water flux values indicate that the inherent membrane permeability, Lp, decreased after ion implantation. A reduction in membrane permeability for charged membranes is not uncommon, and increased osmotic pressure for higher charged membranes can decrease net driving force and reduce flux; VOL. 39, NO. 17, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6489
FIGURE 2. Schematic of NF system experimental setup, adapted from ref 22.
TABLE 3. Permeability of Membranes membrane type
permeability (m3/N-h)
% reduction in permeability
unmodified low dose modified high dose modified
NF90 0.0296 0.0149 0.0093
50 68
unmodified low dose modified high dose modified
SP28 0.0258 0.0126 0.0098
51 62
feed solution and the value in the polymeric matrix. This phenomenon has been documented for dielectric solvents in narrow pores (33, 34). For a charged membrane, the dielectric constant will decrease from bulk (less charge interactions) to concentration polarization layer to membrane surface due to increased electric field. In this case, flux experiments were conducted with solutions with no ions present but with feed solutions with differing dielectric constants (Figure 4) to test this hypothesis in the absence of electrolytes in solution. These experiments are identical to the clean water flux experiments; only ethanol has been added to the feed to decrease the dielectric constant of the solution. The flux decreases as the dielectric constant decreases, simulating the behavior of a polar solvent entering pores of a membrane with increased charge.
FIGURE 3. Zeta potential of unmodified and modified (a) polyamide (NF90) and (b) cellulose acetate (SP28) membranes as a function of pH. Background electrolyte 0.001 M KCl. however, this phenomenon is typically observed in the presence of ionic solutes. One possible explanation for the decreased permeability is a difference between the dielectric constant of the aqueous 6490
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 17, 2005
Decreased permeability may be related to “electroviscous” effects thought to affect permeability through charged pores (4, 34, 35) The electroviscous effect occurs when a solution is pressed through a narrow capillary or pore with charged surfaces. This effect refers to the back-flow of counterions and water in the double layer adjacent to the capillary pore surface due to a streaming potential that develops between the capillary ends. The increase in electric surface charge in the pores of these membranes as a result of fluoride implantation may cause the membrane to be “tighter” due
FIGURE 4. Pure water flux for NF90 membrane using feedwater with different dielectric constants (E). Constant pressure of 100 psi (689 kPa). to increased attraction between the functional groups of the membrane and the implanted charge. When water is forced through the pores of these membranes, ions are displaced from their preferred positions in the double layer. The polar water molecules near the charged wall are aligned in energetically favorable positions, and the water near the walls will, therefore show a higher viscosity (34). The higher the charge on the membrane the more the extra energy needed to achieve the same displacement of ions. This is evident from the higher transmembrane pressure necessary for the modified membranes to achieve the same flux as the unmodified membranes. Although the resistance in Darcy’s law (J ) ∆p/µR) does not explicitly account for electroviscous effects, other researchers have used derivations based on this law to explain change in permeability of membranes due to electroviscous effects (36, 37), leading to the general observation that electroosmotic permeability will generally decrease with increasing fixed membrane charge, which is consistent with the permeability results of this study. Pore Size Measurements. Change in permeability can also be affected by change in pore size. Uncharged solute experiments were conducted on the unmodified and modified membranes. Figure 5a,b contains representative graphs for uncharged solute rejection reaching asymptotic values at high permeate flux. Although the increase in permeate flux is a physical phenomenon that occurs with increase in transmembrane pressure, an asymptotic rejection of solute will always be reached because of the steric hindrance from the membrane. The asymptotic rejection values were used to calculate the pore sizes. A summary of the calculated pore sizes is listed in Table 4. It should be noted that the identification of an effective pore size does not necessarily imply the existence of geometrically well-defined pores in nanofiltration membranes. A more general interpretation will be that the hindrance to transport for the passage of solutes through the polymer structure of such a membrane is equivalent to that for passage through pores of that effective size. The pore size of the NF90 membrane did not change with modification of the membrane. There was a significant difference in the rejection value of the high dose SP28 membrane hence a difference in the pore size. As shown in Table 4, the pore size of the high dose SP28 membrane increased from 1.15 to 1.32 nm after implantation. This increase in pore size for the high dose SP28 was likely caused by the ion implantation process; recall that the high dose showed less negative zeta potential results when compared to the unmodified and low dose SP28 membrane.
FIGURE 5. Uncharged solute (vitamin B12) rejection as a function of permeate flux for modified and unmodified (a) NF90 and (b) SP28 membranes.
TABLE 4. Summary of Uncharged Solute Rejection Experiments membrane
average real rejection (%)
pore radius (nm)
unmodified low dose modified high dose modified
NF90 98.3 ( 0.1 98.7 ( 0.2 98.5 ( 0.1
0.84 0.82 0.83
unmodified low dose modified high dose modified
SP28 81.7 ( 0.2 72.3 ( 0.2 81.4 ( 0.1
1.15 1.32 1.16
Rejection. Rejection values for each salt solution have been reported as an average of the rejection values of their respective cations and anions, which had similar rejection values due to electroneutrality in the permeate. It should be noted that, in some cases (e.g. excess proton in KCl), there can be a significant passage of uncoupled H+ due to its higher ionic mobility and smaller size than the K+, resulting in selective passage of H+ and less passage of K+ (3, 38, 39). More Cl- will transport through the membrane to maintain electroneutrality in the permeate and hence result in different rejection of the K+ and Cl- ions, increasing the pH in the permeate solution. In this study, experiments were conducted at a near neutral pH of 6.4 ( 0.2 where there is little excess H+ or OH-. Furthermore, identical anionic and cationic rejection values were measured and corroborated by a stable pH in feed and permeate solution. Zeta potential values for the low dose NF90 membrane were slightly more negative than that of the unmodified membrane, which should have resulted in increased ion rejection due to increased Donnan exclusion. However rejection of monovalent and divalent salts from the low dose VOL. 39, NO. 17, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6491
when compared with KCl and NaNO3 which is typical of divalent salts as a result of increased co-ion repulsion. The small increase in rejection after implantation brings to question the use of ion implantation as a viable option for increasing rejection efficiency of NF membranes. Modification by ion implantation also decreased the membrane permeability, which influences rejection. These interdependent processes have to be considered in tandem when considering options for enhancing performance. Salt Flux. In ref 22, some of the data presented in this paper were presented as salt fluxes of individual ions, which were compared for the unmodified and the modified membranes (both polyamide (NF90) and cellulose acetate (SP28)) from multicomponent single salt experiments. The results showed that for NF90 membranes the salt fluxes decreased significantly with the higher membrane modification. The lower membrane modification did not have any significant effect on the salt flux of ions through NF 90 membranes. For the SP28 membrane, there was significant decrease in salt flux for both the modifications. Thus, interpretation of the ion transport values as a bulk rejection value may not be the optimal method for comparing salt flux (ion transport) through membranes of different permeabilities. Ion implantation decreased ion transport through the membrane but resulted in lower water permeability in the modified membranes.
Acknowledgments FIGURE 6. Salt rejection for unmodified and modified (a) polyamide (NF-90) and (b) cellulose acetate (SP28) membranes. Flux ) 3.5 × 10-6 m/s. ∆P ) 70 psi (483 kPa), 140 psi (966 kPa), and 190 psi (1310 kPa), respectively, for unmodified, low modified, and high modified. Salt concentration ) 0.005 M KCl, 0.005 M NaNO3, and 0.005 M Na2SO4 (ionic strength ) 0.0125, 0.0125, and 0.005 M, respectively). Feed pH ) 6.4 ( 0.2 polyamide membrane was not increased significantly by ion implantation (Figure 6a). Recall the extended Nernst-Planck equation, where transport is modeled as a function of the diffusive, electric, and convective forces. In the case of the low dose NF90 membrane, implantation of the low fluoride dose may not have caused enough electrostatic repulsion to increase the electric field gradient enough to overcome convective forces. At the experimental pH of 6.4, zeta potential of the unmodified NF90 membrane (Figure 3) was -20.5 mV, while that of the modified high dose was -27.5 mV, a difference of 34%. This change in zeta potential resulted in a slight increase in rejection of 4-6% for the salt with a monovalent co-ion and a slightly larger 8% with the salt with the divalent co-ion. This increase in rejection is admittedly modest. It should be noted that the effect of increased charge may have been influenced by the dead end operation of the stirred cell. Care was taken to reduce concentration polarization by using a high stirring speed and a low salt concentration; however, there was some concentration polarization during the course of these rejection experiments. The rejection behavior of the SP28 membrane did not follow Donnan nor steric exclusion. Rejection results indicate that the ion implantation procedure caused structural modifications to the cellulose acetate membranes. Rejection of the monovalent salts was lower for the high and low dose SP28 membranes (Figure 6b). For the Na2SO4 with a divalent co-ion, the rejection trend was consistent with zeta potential values of the membrane, but the increases were still slight. The slight increase in the rejection of the Na2SO4 for the high dose SP28 membrane when compared to the unmodified membrane is a result of increased elecronegativity of the high dose SP28. Na2SO4, owing to its divalent nature, showed the greatest rejection 6492
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 17, 2005
The authors would like to acknowledge the Keck Center for the Design of Nanoscale Materials for Molecular Recognition and the National Science Foundation (BES 9734429) for funding part of this study. The authors also acknowledge Jermey Matthews, Lily Wan, and Lekan Ogedengbe for their help with the characterization studies.
Literature Cited (1) Schaep, J, C.; Vandecasteele, A.; Mohammad, W.; Bowen, W. R. Modelling the retention of ionic components for different nanofiltration membranes. Sep. Purif. Technol. 2001, 22-23, 169-179. (2) Afonso, M. D.; Hagmeyer, G.; Gimbel, R. Streaming potential measurements to assess the variation of nanofiltration membranes surface charge with the concentration of salt solutions. Sep. Purif. Technol. 2001, 22-23, 529-541. (3) Peeters, J. M. M.; Mulder, M. H. V.; Strathmann, H. Streaming potential measurements as a characterization method for nanofltration membranes. Colloids Surf., A 1999, 150, 247259. (4) Childress, A. E.; Elimelech, M. Relating Nanofiltration Membrane Performance to Membrane Charge (Electrokinetic) Characteristics. Environ. Sci. Technol. 2000, 34, 17 3710-3716. (5) Vezzani, D.; Bandini, S. Donnan equilibrium and dielectric exclusion for characterization of nanofiltration membranes. Desalination 2002, 149, 477-483. (6) Anne, C.; Trebouet, D.; Jaouen, P.; Quemeneur, F. Nanofiltration of seawater: fractionation of mono- and multivalent cations. Desalination 2001, 140, 67-77. (7) Peng, W.; Escobar, I. Rejection Efficiency of Water Quality Parameters by Reverse Osmosis and Nanofiltration Membranes. Environ. Sci. Technol. 2003, 37, 4435-4441. (8) Watson, D.; Hornburg, C. Low-energy membrane nanofiltration for removal of color, organics and hardness from drinking water supplies. Desalination 1989, 72, 11-22. (9) Seidel, A.; Waypa, J.; Elimelech, M. Role of Charge (Donnan) Exclusion in Removal of Arsenic from Water by Negatively Charged Porous Nanofiltration Membrane. Environ. Eng. Sci. 2001, 105-113. (10) Sato, Y.; Kang, M.; Kamei, T.; Magara, Y. Performance of nanofiltration for arsenic removal. Water Res. 2002, 36, 33713377. (11) Schafer, A. I.; Nghiem, L. D.; Waite, T. D. Removal of the Natural Hormone Estrone from Aqueous Solutions Using Nanofiltration and Reverse Osmosis. Environ. Sci. Technol. 2003, 37, 182-188. (12) Nghiem D., Schafer, A.; Elimelech, M. Removal of natural hormones by nanofiltration membranes: measurement, mod-
(13) (14) (15) (16) (17)
(18) (19) (20) (21)
(22) (23) (24) (25) (26)
(27)
eling, and mechanisms. Environ. Sci. Technol. 2004, 38, 6, 18881896. Whu, J.; Baltzis, B.; Sirkar, K. Modeling of nanofiltration assisted organic synthesis. J. Membr. Sci. 1999, 163, 319-331. Capelle, N.; Moulin, P.; Charbit, F.; Gallo, R. Purification of heterocyclic drug derivatives from concentrated saline solution by nanofiltration. J. Membr. Sci. 2002, 196, 125-141. Christy, C.; Vermant S. The state-of-the-art of filtration in recovery processes for biopharmaceutical production. Desalination 2002, 147, 1-4. Bowen W. R.; Mohammad, A. W. Characterization and prediction of nanofiltration membrane performance - A general assessment. Trans. IChemE 1998a, 76, Part A. Tsuru T.; Nakao, S.; Kimura, S. Calculation of ion rejection by extended Nernst-Planck equation with charged reverse osmosis membranes for single and mixed electrolyte solutions. J. Chem. Eng. Jpn. 1991, 24, 4, 511-51. Mohammad, A.; Ali, N. Understanding the steric and charge contributions in NF membranes using increasing MWCO polyamide membranes. Desalination 2002, 147, 205-212. Wijmans J. G.; Baker, R. W. The solution-diffusion model: a review. J. Membr. Sci. 1995, 107, 1-21. Bhattacharya, D.; Williams, M. E. Reverse Osmosis. In Membrane Handbook; Ho, W. S. W., Sirkar, K. K., Eds.; Van Nostrand Reinhold: New York, 1992. Mukherjee P.; Sengupta A. S. Ion Exchange Selectivity as a Surrogate Indicator of Relative Permeability of Homovalent Ions in Reverse Osmosis Processes. Environ. Sci. Technol. 2003, 37, 1432-1440. Mukherjee, P.; Abitoye J.; Jones, K. Surface Modification of Nanofiltration Membranes. J. Membr. Sci. 2005, 254/1-2, 303-310. Jagannadh, S. N.; Muralidhara, H. S. Electrokinetic methods to control membrane fouling. Ind. Eng. Chem. Res. 1996, 35, 1133. Jitsuhara, I.; Kimura, S. Rejection of inorganic salts by charged ultrafiltration membranes make of sulfonated polysulfone. J. Chem. Eng. Jpn. 1983, 16, 394. Good K.; Escobar, I.; Xinglong, X.; Coleman, M.; Ponting, M. Modification of commercial water treatment membranes by ion beam irradiation. Desalination 2002, 146, 259-264. Schaep, J.; Vandecasteele, C.; Peeters, B.; Luyten, J.; Dotremont, C.; Roels, D. Characteristics and retention properties of a mesoporous γ-Al2O3 membrane for nanofiltration. J. Membr. Sci. 1999, 163, 229-237. Childress, A. E.; Elimelech, M. Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanofiltration membranes. J. Membr. Sci. 1996, 119, 253-268.
(28) Bowen, W.; Welfoot, J. Predictive modeling of nanofiltration: membrane specification and process optimization. Desalination 2002, 147, 197-203. (29) Bowen, W. R.; Mukthar, H. Characterisation and prediction of separation performance of nanofiltration membranes. J. Membr. Sci. 1996, 112, 263-274. (30) Bowen, W. R.; Mohammad, A. W.; Hilal, N. Characterisation of nanofiltration membranes for predictive purposes - use of salts, uncharged solutes and atomic force microscopy. J. Membr. Sci. 1997, 126, 91-10. (31) Szymczyk, A.; Pierre, A.; Reggiani, J.; Pagetti, J. Characterisation of the electrokinetic properties of plane inorganic membranes using streaming potential measurements. J. Membr. Sci. 1997, 134, 59-66. (32) Lee, E. H.; Lewis, M. B.; Blau, P. J.; Mansur, L. K. Improved surfaces of polymer materials by multiple ion beam treatment. J. Mater. Res. 1991, 6, (3). (33) Hall, M.; Starov, V.; Lloyd, D. Reverse osmosis of multicomponent electrolyte solutions Part I. Theoretical development. J. Membr. Sci. 1997, 128, 23-37. (34) Hunter, R. J. In Zeta Potential in Colloid Science; Academic Press: London, 1981. (35) Tu S.; Ravindran, V.; Den, W.; Pirbazari, M. Predictive membrane transport model for nanofiltration processes in water treatment. AIChE J. 2001, 47, (6). (36) Straatsma J.; Bargeman, G.; van der Horst, H. C.; Wesselingh, J. A. Can nanofiltration be fully predicted by a model? J. Membr. Sci. 2002, 198, 273-284. (37) Huisman I. H.; Dutre, B.; Persson, K. M.; Tragardh, G. Water permeability in ultrafiltration and microfiltration: Viscous and electroviscous effects. Desalination 1997, 113, 95-103. (38) Chaufer B.; Rabiller-Baudry, M.; Guihard, L.; Daufin, G. Retention of ions in nanofiltration at various ionic strength. Desalination 1996, 104, 37-36. (39) Canas A.; Ariza, M.; Benavente, J. A comparison of electrochemical and electrokinetic parameters determined for cellophane membranes in contact with NaCl and NaNO3 solutions. J. Colloid Interface Sci. 2002, 246, 150-156.
Received for review January 15, 2005. Revised manuscript received May 12, 2005. Accepted June 1, 2005. ES050102V
VOL. 39, NO. 17, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
6493
