Chemosphere 77 (2009) 228–235
Contents lists available at ScienceDirect
Chemosphere journal homepage: www.elsevier.com/locate/chemosphere
Removal of toxic ions (chromate, arsenate, and perchlorate) using reverse osmosis, nanoﬁltration, and ultraﬁltration membranes Jaekyung Yoon a, Gary Amy b, Jinwook Chung c, Jinsik Sohn d, Yeomin Yoon e,* a
Korea Institute of Energy Research, New and Renewable Energy Research Division, 71-2 Jang-Dong, Yuseong-Gu, Daejeon 305-343, South Korea Water Desalination and Reuse Center, King Abdullah University of Science and Technology, Box 55455, Jeddah 21534, Saudi Arabia c Samsung Engineering Co., Ltd., R&D Center, 39-3, Sungbok-Dong, Yongin, Gyeonggi-Do 449-844, South Korea d Kookmin University, Civil and Environmental Department, Seoul 136-702, South Korea e Department of Civil and Environmental Engineering, University of South Carolina, Columbia, SC 29208, USA b
a r t i c l e
i n f o
Article history: Received 20 April 2009 Received in revised form 10 July 2009 Accepted 16 July 2009 Available online 12 August 2009 Keywords: Chromate Arsenate Perchlorate Water treatment Membrane rejection
a b s t r a c t Rejection characteristics of chromate, arsenate, and perchlorate were examined for one reverse osmosis (RO, LFC-1), two nanoﬁltration (NF, ESNA, and MX07), and one ultraﬁltration (UF and GM) membranes that are commercially available. A bench-scale cross-ﬂow ﬂat-sheet ﬁltration system was employed to determine the toxic ion rejection and the membrane ﬂux. Both model and natural waters were used to prepare chromate, arsenate, and perchlorate solutions (approximately 100 lg L1 for each anion) in mixtures in the presence of other salts (KCl, K2SO4, and CaCl2); and at varying pH conditions (4, 6, 8, and 10) and solution conductivities (30, 60, and 115 mS m1). The rejection of target ions by the membranes increases with increasing solution pH due to the increasingly negative membrane charge with synthetic model waters. Cr(VI), As(V), and ClO4 rejection follows the order LFC-1 (>90%) > MX07 (25–95%) ﬃ ESNA (30–90%) > GM (3–47%) at all pH conditions. In contrast, the rejection of target ions by the membranes decreases with increasing solution conductivity due to the decreasingly negative membrane charge. Cr(VI), As(V), and ClO4 rejection follows the order CaCl2 < KCl ﬃ K2SO4 at constant pH and conductivity conditions for the NF and UF membranes tested. For natural waters the LFC-1 RO membrane with a small pore size (0.34 nm) had a signiﬁcantly greater rejection for those target anions (>90%) excluding NO 3 (71–74%) than the ESNA NF membrane (11–56%) with a relatively large pore size (0.44 nm), indicating that size exclusion is at least partially responsible for the rejection. The ratio of solute radius (ri,s) to effective membrane pore radius (rp) was employed to compare ion rejection. For all of the ions, the rejection is higher than 70% when the ri,s/rp ratio is greater than 0.4 for the LFC-1 membrane, while for di-valent ions 2 2 (CrO2 4 , SO4 , and HAsSO4 ) the rejection (38–56%) is fairly proportional to the ri,s/rp ratio (0.32–0.62) for the ESNA membrane. Ó 2009 Elsevier Ltd. All rights reserved.
Toxic ions such as Cr(VI), As(V), and perchlorate (ClO4 ) have been an important drinking-water quality issue. River systems in the United States have been found to have chromium concentrations that range from <1 to 30 lg L1 (ATSDR, 1993). A recent epidemiological study has indicated that the cancer risk posed by arsenic in drinking water may be greater than previously considered (Smith et al., 2002). A survey of perchlorate occurrence survey has shown detectable perchlorate ions found in 69 of 232 well samples, wherein 24 had levels above 18 lg L1; samples that were obtained from highly contaminated sites contained up to 8000–9000 lg L1 (EPA, 2002). The United States Environmental Protection Agency (USEPA) has implemented the maximum allow* Corresponding author. Tel.: +1 803 777 8952; fax: +1 803 777 0670. E-mail address: [email protected]
(Y. Yoon). 0045-6535/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.chemosphere.2009.07.028
able contaminant levels for arsenic (10 lg L1) and chromium (total100 lg L1) in public drinking water supplies. Although perchlorate is not listed as a USEPA drinking water contaminant, the California Department of Health Service (CDHS) has recently revised a notiﬁcation level of 6 lg L1 (March 2004) – that is, the level of perchlorate at which, if it is exceeded, the CDHS advises water utilities to remove drinking water supplies from service. The removal of chromate, arsenate, perchlorate from drinking source water is critical for the protection of human health. Several different technologies have been investigated to remove these toxic ions from drinking source water and/or wastewater, including ion exchange, coagulation, and sorption-based metal oxide (Kim and Benjamin, 2004; Golder et al., 2007; Habuda-Stanic et al., 2008; Van Ginkel et al., 2008). Nevertheless, membrane ﬁltration using reverse osmosis (RO) and nanoﬁltration (NF) plays an important role in the removal of chromate, arsenate, and/or perchlorate and has also been a promising technology for water
J. Yoon et al. / Chemosphere 77 (2009) 228–235
treatment (Kosutic et al., 2005; Yoon et al., 2005a; Moore et al., 2008; Muthukrishnan and Guha, 2008). Several recent studies have investigated the transport mechanisms of ions including chromate, arsenate, and perchlorate through RO, NF, and ultraﬁltration (UF) membranes (Childress and Elimelech, 2000; Brandhuber and Amy, 2001; Sato et al., 2002; Yoon et al., 2004, 2005a; Ergican et al., 2005; Uddin et al., 2007; Muthukrishnan and Guha, 2008). These studies have shown that for inorganic compounds the degree of rejection is governed by both size exclusion and electrostatic exclusion. These studies, however, were still limited to a few membranes (e.g., RO, NF, or UF), focused on synthetic solutions, or examined only a single ion (i.e., chromate, arsenate, or perchlorate) at limited solution pH and conductivity ranges. Therefore, a systematic rejection assessment for those target toxic ions is necessary to investigate the rejection mechanisms for RO, NF, and UF membranes in the presence of co- and counter-ions in natural source waters. Although our study focuses primarily on the rejection of chromate, arsenate, and perchlorate using RO, NF, and UF membranes, it is still necessary to evaluate the rejection of other ionic compounds, since they always coexist in drinking water. For ions, the solution pH and conductivity are major factors that inﬂuence the solute rejection due to electrostatic repulsion between the ions and negatively charged membranes. This is since the solution pH and conductivity inﬂuence signiﬁcantly the membrane charge. Previous studies have shown that the rejection of various ions such as sodium, calcium, chloride, and sulfate ions increases with increasing solution pH and decreasing solution conductivity (Childress and Elimelech, 1996, 2000; Yoon et al., 2004), since the membrane charge becomes more negative under such conditions. The objective of this study is to verify the existing mechanisms (i.e., size exclusion and electrostatic exclusion) for the rejection of chromate, arsenate, and perchlorate by RO, NF, and UF membranes. To accomplish this, the rejections of chromate, arsenate, and perchlorate were measured at various pH and conductivity conditions + in the presence of co- and counter-ions such as Cl, SO2 4 , K , and Ca2+ for synthetic water and at ambient pH and conductivity conditions for drinking source waters. In addition, the ratio of ri,s to rp was employed to compare ion rejection.
2. Materials and methods
from the manufacturer’s nominal molecular weight cut-offs (MWCO)). In Table 1, ESNA and GM show relatively high pure water permeability (PWP) values, while LFC-1 and MX07 have relatively low PWP values, presumably because permeability is inﬂuenced by pore size, pore density (porosity), and thickness. The surface charge of the membranes was estimated by measuring streaming potential using a commercial electrokinetic analyzer measurement apparatus. These membranes are relatively hydrophobic based on their contact angle values. A commercial test cell (SEPA, Osmonics, Minnetonka, MN, USA) was employed for all membrane experiments. The ﬁltration module has an active ﬁltration area of 139 cm2 and without a spacer, assuredly provides for a laminar ﬂow condition. The experiments were performed at room temperature, 295 ± 1 K, and the initial transmembrane pressure was varied to obtain a constant ratio (i.e., 0.5) of the initial pure water ﬂux (Jo) to the estimated back-diffusional mass transfer coefﬁcient (k) through the boundary layer. A hydrodynamic operating parameter (Jo/k ratio) was used to facilitate comparison of results from different membranes in benchscale cross-ﬂow ﬂat-sheet tests (Cho et al., 1999; Yoon et al., 2005a,b). The Jo/k ratio is a function of recovery (i.e., a ratio of permeate volume to feed volume), which inﬂuences concentration polarization. Each new membrane tested was preﬁltered with a non-recirculated pure water feed volume of 8 L (at a feed ﬂow rate of approximately 200 mL min1 and a permeate ﬂow rate of approximately 5 mL min1) to clean any humectant materials which may be coated onto/in the membrane surface/pores. Before testing, the membrane coupons were ﬂushed with pure water for at least 12 h until the membrane was stable with regard to pure water permeability. A new membrane coupon was used for each experiment. The feed solution was drawn from a 10 L reservoir, and supplied to the ﬂat-sheet tester. Because of a limited volume of natural feed water, the unit was operated in a recycle mode, and the retentate and permeate streams were returned to the reservoir after passing through the test cell. The feed waters were kept at room temperature for 24 h prior to the ﬁltration tests to assure thermal equilibration. The retentate ﬂow, permeate ﬂow, and temperature were monitored over time and the feed ﬂow was calculated by summing the measured retentate and permeate ﬂows. Duplicate analytical samples were taken during rejection tests. Rejection, R, was calculated based on the feed concentration using:
2.1. Membranes and testing unit
Ri ð%Þ ¼
2 The chromate (Cr(VI), CrO 4 ), arsenate (As(V), HAsO4 ), and per chlorate(ClO4 ) rejection measurements were conducted using one RO (LFC-1), two NF (ESNA and MX07), and one UF (GM) membranes. In addition, a model uncharged species, arsenite (As(III)), rejection was conducted using those membranes. The properties of the membranes used in this study are described in Table 1. These membranes have ionizable functional groups such as unreacted carboxylic acids and have different membrane pore sizes (assumed
Cp 100ð%Þ Cf
where Cp and Cf are the solute concentrations in the permeate and feed, respectively. 2.2. Source waters Model water solutions were prepared with pure water from a commercial laboratory puriﬁcation system (Milli-Q, Millpore
Table 1 RO, NF, and UF membranes and their characteristics along with measured water ﬂux, zeta potential, and contact angle. Membrane type
Product name/ manufacturer
Pure water permeability (L m2 d1 kPa1)
Zeta potential (mV)c
Contact angle (°)d
RO NF NF UF
LFC-1/Hydranautics ESNA/Hydranautics MXO7/Desal-Osmonics GM/Desal-Osmonics
Polyamide TFC Polyamide TFC Polyamide TFC Proprietary
NA 200 400 8000
0.75 1.05 0.47 2.81
3–10 2–10 4–11 3–10
4.5 11.1 36.8 32.2
70 57 45 46
NA = not available. Data obtained from the manufacturer. c Measured at pH 8 and conductivity 30 mS m1 with KCl. d Average value (three measurements).
J. Yoon et al. / Chemosphere 77 (2009) 228–235
Water Puriﬁcation System, Bedford, MA, USA). Chromate, arsenate, and perchlorate anions (present at a concentration of 100 lg L1 for each anion) in mixtures were fed to the membrane test apparatus either as pure components or in binary mixtures with other salts (KCl, K2SO4, and CaCl2); and at varying pH conditions (4, 6, 8, and 10) buffered by 1 mM potassium bicarbonate and solution conductivities (30, 60, and 115 mS m1). Potassium hydroxide or hydrochloric acid was added to the pure water to adjust solution pH. Two ground waters that provide raw water to water treatment plants were used to perform bench-scale cross-ﬂow ﬂat-sheet membrane tests. These source waters were obtained from the City of Glendale Water and Power (Glendale water) and the Los Angeles Department of Water and Power (LADWP water) in California. The LADWP and Glendale waters were contaminated with 70 and as chromate, respectively, but arsenate and 120 lg L1 of CrO2 4 perchlorate were not detected. Unless otherwise indicated, the LADWP and Glendale waters were preﬁltered with a 0.45 lm nylon ﬁlter (Millipore) for all membrane tests. Table 2 lists the characteristics of the feed waters used in this study. Several analyses were performed including dissolved organic carbon (DOC), ultraviolet adsorption at 254 nm (UVA254) – a measure of natural organic matter (NOM) aromaticity, conductivity, and pH. Both the Glendale and LADWP waters contained relatively low NOM contents (<0.7 mg L1). The anions and cations in the Glendale and LADWP waters were measured to assess how 2 + + 2+ 2+ the co- and counter-ions, Cl, NO 3 , SO4 , K , Na , Ba , Mg , and 2 2+ Ca , would inﬂuence CrO4 , HAsO4 , and ClO4 removal. As shown in Table 2, the Glendale and LADWP waters have a relatively high conductivity values from a variety of mono- and di-valent ions, therefore we would expect signiﬁcant electrostatic screening to occur.
(ICP) mass spectroscopy (Perkin–Elmer Sciex, Model Elan 5000, cone spray pneumatic nebulizer, Toronto Canada) with a detection limit of 0.1 lg L1 As. ClO4 concentration was measured using a Dionex DX300 (Dionex Corp., Sunnyvale, CA, USA) including a CDM-2 conductivity detector, a GPMII gradient pump, and an auto-sampler. A ClO4 calibration curve was constructed with a linear coefﬁcient of determination (r2) of P0.95 having a MDL of 1.4 lg L1. Uncertainty was determined through replicate analyses with the resultant data used to calculate the means and a standard deviation. Other anion concentrations were measured on a Dionex DX300, and cation concentrations were measured by ICP emission spectroscopy (Liberty-Series II, Varian, Australia).
2.3. Analytical measurement of chromate, arsenate, and perchlorate
3. Results and discussion
Samples were analyzed for Cr(VI) using a Dionex DX-320 ion chromatograph with an AD25 post-column UV–Vis detector (Dionex Corp., Sunnyvale, CA, USA) according to USEPA Method 1636 (Diphenylcarbohydrazide Colorimetric Method). The method detection limit (MDL) was 0.2 lg L1. Samples were preserved with 12 mM soda ash prior to analysis. Arsenic samples were preserved for analysis with trace metal hydrochloric acid to pH < 2 at the time of sampling and chilled to 4 °C until analyzed. Dissolved arsenic analysis was performed using an inductively coupled plasma
3.1. Zeta potential measurements
Table 2 Water quality compositions of synthetic and natural waters.
3.1.1. Effect of solution pH and conductivity ZP results were obtained at varying pH and conductivity levels. These measurements establish the trends for the electrostatic con tribution to Cr(VI), As(V), and ClO4 rejection from charge repulsion between these target ions and the negatively charged membranes (under varying pH conditions and salts). ZP versus pH and conductivity data for experiments performed in the presence of KCl is shown in Fig. 1. The ZPs for the LFC-1 and ESNA membranes did not differ signiﬁcantly from each other and were consistently more negatively charged with increasing pH. The ZP of the MX07 and GM membranes is higher compared with the LFC-1 and ESNA membranes. The error bars for the GM membrane data are calculated based on the standard deviation from triplicate measurements of the ZP and reﬂect the relative uncertainty of the measurements: Note that the error bars are smaller than the symbols in most cases. For all the membranes, the surface charge is negative throughout the entire pH range (at a constant conductivity of 30 mS m1 KCl) excluding pH 4 for LFC-1 (ZP = 3.2 mV) and becomes more negative with increasing pH (Fig. 1a). This is presumably since anions such as Cl and OH are less hydrated than cations in aqueous solutions, they can more closely approach the membrane surface. The shape of the ZP curves is indicative of
pH Conductivity (mS m1) DOC (mg L1) UVA254 (1 cm1) SiO2 (mg L1) Ba2+ (mg L1) Ca2+ (mg L1) Mg2+ (mg L1) Na+ (mg L1) K+ (mg L1)
4, 6, 8, 10 30, 60, 115 <0.1 NA NA NA NA NA NA NA NA
7.3 78 0.47 0.011 17.1 0.05 89.9 29.2 40.6 2.4 105.7
8.0 76 0.62 0.015 10.5 0.12 91.7 20.1 26.1 4.3 68.7
NA NA 0.1a
37.4 64.3 0.07b/0.1a
58.2 34.2 0.12b/0.12a
1 SO2 4 (mg L ) 1 NO (mg L ) 3 Cl (mg L1) 1 CrO2 4 (mg L ) 1 HAsO2 4 (mg L ) ClO4 (mg L1)
NA = not available. a Total concentrations of chromate (CrO2 4 ), arsenate, and perchlorate after spiking. b Ambient concentrations of chromate (CrO2 4 ), arsenate, and perchlorate in feed waters.
2.4. Streaming potential measurements A commercial streaming potential analyzer (Electrokinetic Apparatus-EKA, Brookhaven Instruments Corp., Holtsville, NY, USA) was used to measure the membrane surface charge. This instrument includes an analyzer, a measuring cell, electrodes, and a data control system. We followed the same procedures and used the same EKA to measure streaming potential as described by a previous study (Wilbert et al., 1999). For the streaming potential measurements, the LFC-1, ESNA, MX07, and GM ﬂat-sheet membrane samples were cut to ﬁt the measurement cell and then wetted in KCl, K2SO4, or CaCl2 solution at the desired pH (4, 6, 8, and 10) and stored in a refrigerator for the pre-soaking time. Zeta potential (ZP) was calculated from the measured streaming potential based on the Helmholtz–Smoluchowski relationship with the assumption that the electrolyte solution, with conductivity k (X1 m1), carries most of the current.
Natural water sources such as surface and ground waters for use as drinking water have complex ionic compositions with various pH levels. Thus, we must consider the effect of solution pH, conductivity, and electrolyte on membrane charge (i.e. ZP). Zeta potential of the RO, NF, and UF membranes was analyzed at various pH and conductivity conditions in the presence of different electrolytes.
J. Yoon et al. / Chemosphere 77 (2009) 228–235
At pH 8, all membranes become slightly less negative with increasing conductivity (Fig. 1b), presumably due to a decrease in the electrical double layer with increasing solution conductivity, since the anion bonding/adsorption occurs primarily in the inner plan, which is affected by the size of the electrical double layer. During formation of the surface complex, the ion may form an innersphere complex (coordinating bond), and an outer-sphere ion pair, or be in the diffuse layer of the electrical double layer (Stumm, 1992).
0 -10 -20 -30 -40
0 -10 -20 -30 -40 -50 20
(b) conductivity 40
Conductivity, mS m-1 Fig. 1. Dependence of zeta potential on pH and conductivity for RO (LFC-1), NF (ESNA, MX07), and UF (GM) membranes. (a) At conductivity = 30 mS m1 KCl and (b) at pH 8 (d, LFC-1; s, ESNA; ., MX07; 4, GM). Error bars are calculated based upon the standard deviation from triplicate measurements of the zeta potential.
amphoteric surface (Childress and Elimelech, 1996). Ionizable carboxyl and amine functional groups may be expected on the membrane surface, since the thin-ﬁlm composite membranes are made by the interfacial polymerization reaction of a monomeric polyamine with a polyfunctional acyl halide (Elimelech et al., 1994).
3.1.2. Effect of mono- and di-valent co- and counter-ions Three different types of salt were used to evaluate the effect of mono- and di-valent ions on the membrane surface charge. Fig. 2 illustrates ZP versus pH curves for KCl, K2SO4, and CaCl2 at a constant conductivity value of 30 mS m1. All experiments were performed at varying conductivity values, 30, 60, and 115 mS m1. At constant pH and solution conductivity, the absolute value of the MX07 and GM membranes’ ZP (it is typically negative) follows the order CaCl2 < K2SO4 < KCl. However, the mono- and di-valent ions have insigniﬁcant effect on the ZP of the LFC-1 and ESNA membranes. The pH exerts a major inﬂuence on the membrane surface charge, presumably, through ionizable species and many previous studies reported that both pH and ion composition affect the surface charge of RO and NF membranes (Childress and Elimelech, 1996, 2000; Hong and Elimelech, 1997). All four tested membranes acquired more negative charge when pH increased at a constant conductivity value of 30, 60, and 115 mS m1, regardless of the salt. In addition, these membranes become less negatively charged with increasing solution conductivity in the presence of the salts. These results are all consistent with trends expected from the simple theories of electrokinetic phenomena, that is, both higher salt concentration (conductivity) and higher ionic valency decreases the Debye length thus lowering the ZP (Yoon et al., 2004). The manner in which other factors, such as, the relative dielectric strength of the different ion solutions, co-ion binding; and counter-ion mobility that, would affect the ZP are also consistent with these measurements. For example, the ZP of membranes
-10 -20 -30 -40
(a) LFC-1 (RO)
(b) ESNA (NF)
(c) MX07 (NF)
(d) GM (UF))
-50 10 0
-10 -20 -30 -40 -50 4
Fig. 2. Inﬂuence of electrolyte salt type on the zeta potential variation with pH for RO (LFC-1), NF (ESNA and MX07), and UF (GM) membranes (conductivity = 30 mS m1 KCl). (d, KCl; s, K2SO4; ., CaCl2.) Error bars are calculated based upon the standard deviation from triplicate measurements of the zeta potential.
J. Yoon et al. / Chemosphere 77 (2009) 228–235
is decreased by CaCl2 due to it being a di-valent co-ion with stronger ion complexation (binding) with surface groups. We also interpret that adsorption of Cl co-ion (in hydrophobic regions) is also reduced with increasing conductivity (ionic strength) due to the decrease of the electrical double layer due to ﬁxed ionizable groups (Yoon et al., 2004). Overall, independently of the type of mono- and di-valent co- and counter-ions, the ZP of the membranes becomes less negatively charged with increasing solution conductivity at pH 8. 3.2. Measurement of chromate, arsenate, and perchlorate rejection in model water The surface charge of the RO, NF, and UF membranes was measured to predict chromate, arsenate, and perchlorate transport through the membranes. Several qualitative predictions for the transport of these target toxic ions can be made with characterizations with the streaming potential measurements. The rejection of the target ions (Cr(VI), As(V), and ClO4 ) may (i) increase with increasing pH due to the increasingly negative membrane charge, (ii) decrease with increasing solution conductivity due to the decreasingly negative membrane charge, and (iii) decrease in the presence of di-valent counter-ions (Ca2+) due to a less negative membrane charge. All the measurements were made at a constant Jo/k ratio of 0.5. 3.2.1. Effect of solution pH and conductivity Measurements of Cr(VI), As(V), As (III), and ClO4 rejection by the RO, NF, and UF membranes were made for 16 h. The pure water was spiked with 100 lg L1 for each ion. The ﬁltration experiments were performed at pressures of 513 kPa (LFC-1), 366 kPa (ESNA), 818 kPa (MX07), and 137 kPa (GM). Flux decline trends were monitored at the same time. In addition, the ﬂux measurements were made at the constant Jo/k ratio (i.e., a function of recovery) where the same concentration polarization occurs. Flux declines ranging from 4% to 16% of the pure water ﬂux were observed depending on the membrane and the synthetic water composition. The relatively higher conductivity and pH water showed a greater ﬂux decline for the RO, NF, and UF membranes. The ﬂux declines follow the order, LFC-1 > MX07 ﬃ ESNA > GM. This pattern is presumably attributed to the LFC-1 membrane having the highest ion rejection leading to the highest osmotic pressure at the membrane surface, which lowers ﬂux by reducing the effective transmembrane pressure. The rejections of Cr(VI), As(V), As (III), and ClO4 are plotted as a function of pH at a constant conductivity value of 30 mS m1 (Fig. 3a). In general, the rejection of ions by the membranes is dependent on the solution pH. Cr(VI), As(V), and ClO4 rejection follows the order LFC-1 > MX07 ﬃ ESNA > GM at the same pH conditions. These results are consistent with the inﬂuence of pore size on both steric and electrostatic effects based on their nominal MWCOs, LFC-1 < ESNA ﬃ MX07 < GM. In addition, a general trend was observed that the rejection of these ions increases as the solution pH is increased from 4 to 10. This effect is possibly attributable to OH adsorption occurring in the inner plane at the membrane surface (Elimelech and Omelia, 1990) or a greater degree of dissociation of ﬁxed ionizable functional groups in the membrane. A further increase in the pH, however, results in slightly decreased rejection by the MX0 and GM membrane for the ions excluding As(III). These results can be explained by electrostatic exclusion, since the membrane charge becomes more negative with increasing pH, as shown in Fig. 1a, resulting in increased electrostatic repulsion between the target ions and the membranes thus increasing the ion rejection. However, for As(III) the rejection by the NF membranes only slightly varies over the range of pH when the pH is lower than 10 (Fig. 3a), since the As(III) exists primarily as
an uncharged species below pH 9.13 (i.e., its pKa). In contrast, As(III) rejection signiﬁcantly increases at pH 10, when it becomes anionic. This indicates that steric/size exclusion is the determining mechanism for the uncharged As(III) species until it becomes anionic at pH > 9.13 in which the electrostatic exclusion mechanism begins to play an important role. For LFC-1 and GM, As(III) rejection was fairly constant over the entire pH range, 92–96% for LFC-1 and 7–11% for GM. This is presumably since steric/size exclusion is dominant for both LFC-1 and GM membranes. While the rejection of uncharged As(III) was lowest among the ions, perchlorate rejection was signiﬁcantly lower than chromate and arsenate for the NF membranes, presumably since the hydrated di2 valent ions have a larger size (0.27 nm for HAsO4 ) and/or greater charge than the hydrated monovalent perchlorate ion (ClO4 , 0.14 nm). The solute radii are calculated using the Stokes–Einstein equation (Bowen and Mohammad, 1998). For target toxic ions, the RO membrane with small pore size (the measure of which was discussed in a previous study (Yoon and Lueptow, 2005)) had the highest rejection (>90%), indicating that size exclusion is at least partially responsible for the rejection. However, the rejection of the ions was 4–80% lower for the NF membranes depending on solution pH and target ion, while the UF membrane with relatively large pore size had the lowest rejection ranging from 7% to 43%. The rejections of Cr(VI), As(V), As (III), and ClO4 are plotted as a 1 KCl) at pH 8 function of conductivity (30, 60, and 115 lg L (Fig. 3b). The results show that the rejection of all the target ions excluding As(III) decreases with increasing solution conductivity for the membranes excluding LFC-1. The LFC-1 membrane shows constantly the highest ion rejection (>90%) over the conductivity range, while the rejection of the target ions excluding As(III) decreases signiﬁcantly with increasing conductivity, 34–86% for ESNA, 27–94% for MX07, and 3–43% for GM. It is inferred that these results are due to a reduction in electrostatic repulsion, and the corresponding increase in both the hindered diffusion and parti tioning (between the Cr(VI), As(V), and ClO4 ions and the membrane) with increasing conductivity. In addition, the rejection of Cr(VI) and As(V) was signiﬁcantly greater than that for ClO4 for the NF and UF membranes. This is presumably since di-valent 2 chromate (CrO2 4 ) and arsenate (HAsO4 ) ions have a larger size and/or greater charge than the monovalent ClO4 . For As(III) the rejection by the NF (<26%) and UF (<11%) membranes only slightly varies over the conductivity range, since at pH 8, the dominant form of As(III) exists primarily as an uncharged species, which is difﬁcult to reject by the negatively charged membranes. 3.2.2. Effect of mono- and di-valent co- and counter-ions The effects on Cr(VI), As(V), As(III), and ClO4 rejection by the LFC-1, ESNA, MX07, and GM membranes were determined with the ideal, model water (spiked with 100 lg L1 for each target ion) having three different salts (KCl, K2SO4, and CaCl2) as the primary background electrolyte at constant pH 8 and conductivity (30 mS m1) conditions. These results are shown in Fig. 4. Cr(VI), As(V), and ClO4 rejection follows the order CaCl2 < KCl ﬃ K2SO4 at the constant pH and conductivity conditions for the NF and UF membranes tested. However, the RO membrane shows relatively high rejections (>92%) for all the target ions including As(III). The rejection of Cr(VI) and As(V) was greater than that for ClO4 in + 2+ the presence of each co- and counter-ion species, K , Ca , Cl, for the NF and UF membranes. However, Cr(VI), As(V), and SO2 4 and ClO4 rejections are much lower in solutions having CaCl2 (di-valent Ca2+) than in those with KCl and K2SO4 (monovalent K+) for the NF and UF membranes. This is presumably since in terms of electrostatic interactions, the Ca2+ binding causes the membrane’s surface charge to decrease signiﬁcantly in absolute value. The concentration of each co- and counter-ion species, K+, Ca2+, Cl, and SO2 4 , as well as Cr(VI), As(V), As(III), and ClO4 in
J. Yoon et al. / Chemosphere 77 (2009) 228–235
100 80 60 40 20 (a.1) LFC-1 (RO)
(a.2) ESNA (NF)
(a.3) MX07 (NF)
(a.4) GM (UF)
80 60 40 20 0 4
100 80 60 40 20 (b.1) LFC-1 (RO)
(b.2) ESNA (NF)
100 80 60 40 20 (b.3) MX07 (NF) 0 20 40 60
(b.4) GM (UF) 80
Conductivity (mS m-1)
Fig. 3. Effect of (a) pH and (b) conductivity (KCl) on arsenate, chromate, perchlorate, and arsenite rejection for RO (LFC-1), NF (ESNA and MX07), and UF (GM) membranes. (a) At conductivity = 30 mS m1 (KCl) and (b) at pH = 8. (d, arsenate; s, chromate; ., perchlorate; and 4, arsenite).
the feed and permeate during ﬁltration was measured to determine how the target solute’s composition at the interface of the membranes may change. When dilute solutions containing Cl or SO2 4 anions are brought in contact with a membrane possessing a ﬁxed negative charge, Donnan exclusion (Mulder, 1996) shows that rejection of chloride and sulfate may be greater than if the membrane were completely uncharged (Yoon et al., 2004). In addition, it was mentioned earlier how Ca2+ binding screens the co-ion adsorption. When the charge screening by Ca2+ occurs for the negatively charged membranes, anion rejection is signiﬁcantly reduced. The background electrolyte’s cation and anion rejections vary from 5% to 98% at the same conditions (pH 8 and conductivity 30 mS m1) for the RO, NF, and UF membranes. The background ion concentration at the membrane interface inﬂuences the further
Donnan exclusion (e.g., greater exclusion at low ion concentration, high pH, and in the presence of more di-valent anions). As discussed previously, the Cr(VI), As(V), and ClO4 rejection is reduced due to the effect on the reduced electric double layer from increased concentration of the other ions, K+, Cl, and SO2 4 when some concentration accumulation occurs. For an uncharged species (As(III)) the rejection by the NF (11–30%) and UF (5–11%) membranes only slightly varies in the presence of KCl, K2SO4, and CaCl2. 3.3. Determination of chromate, arsenate, perchlorate, and co-ion rejection with natural water Two ground waters (Glendale and LADWP) were selected as a natural feed water, since they already contained some amounts
J. Yoon et al. / Chemosphere 77 (2009) 228–235
80 60 40 20 (a) LFC-1 (RO)
(b) ESNA (NF)
80 60 40 20 (d) GM (UF)
(c) MX07 (NF) 0
Fig. 4. Effect of variable salts on arsenate, chromate, perchlorate, and arsenite rejection for RO (LFC-1), NF (ESNA, MX07), and UF (GM) membranes (pH = 8 and conductivity = 30 mS m1). (d, arsenate; s, chromate; ., perchlorate; and 4, arsenite.)
(a) 100 80 60 40 20 0
(b) 100 Rejection (%)
of CrO2 (70 lg L1 in Glendale water and 120 lg L1 in LADWP 4 water) and have been widely used as a drinking water source. In addition, they have relatively high levels of other typical anions and cations. Both LADWP and Glendale waters were spiked with 100 lg L1 As(V) and ClO4 , while additional 30 lg L1 CrO2 4 , was spiked into Glendale water. The ﬁltration experiments were performed at the same operating conditions (Jo/k ratio = 0.5; recovery = 15%) where the rejection was performed with a model water for the RO, NF, and UF membranes. Measurements of DOC, UVA254, conductivity, and co-/counter-ion rejection as well as Cr(VI), As(V), and ClO4 rejection by the RO, NF, and UF membranes were made for 16 h. Flux decline trends showed that the ﬂux declines of these membranes follow the order, LFC-1 (25– 30%) > MX07 (19–25%) ﬃ ESNA (18–23%) > GM (12–13%) for the waters. The LADWP water showed a slightly greater ﬂux decline than the Glendale water for the RO, NF, and UF membranes. This is presumably since the LADWP water contains the slightly higher DOC concentration compared with the Glendale water. Several previous studies have shown that the membrane ﬂux declines decrease with increasing solution NOM concentration due to NOM fouling (Cho et al., 1999, 2000; Yoon et al., 2005b). The RO membrane had a slightly greater ﬂux decline than the NF and UF membranes, indicating that the increases in osmotic pressure for RO are greater than for NF and UF. Since the osmotic pressure increases with increasing concentration polarization occurring at the membrane surface, the concentration polarization follows the order for the solutions and the membranes, LADWP > Glendale and RO > NF > UF, respectively. Glendale and LADWP waters are relatively less complex water to treat based on their low DOC (0.47– 0.62 mg L1), UVA254 (0.011–[email protected]
cm1), and conductivity (76–78 mS m1) values (see Table 2). The rejections of DOC, UVA254, and ions (based on conductivity) for Glendale and LADWP waters follow the order for the RO, NF, and UF membranes, LFC1 > MX07 > ESNA > GM (see Supporting Information). In order to compare the detailed properties of the RO, NF, and UF membranes, the rejections of individual ions are compared for the Glendale and LADWP waters in Fig. 5. Below this ﬁgure, ri,s and ri,s/rp are noted. The membrane pore radii were estimated from
80 60 40 20 0
Cl- NO3 ClO4 CrO4 SO4 HAsO4 ri,s (nm) : 0.12 0.13 0.14
0.14 0.23 0.27
ri,s/rp LFC-1 ( ) : 0.36 0.38 0.40 ESNA (∇) : 0.27 0.29 0.31
0.41 0.68 0.80 0.32 0.52 0.62
Fig. 5. Rejection of different ions by RO (LFC-1), NF (ESNA, MX07), and UF (GM) membranes for Glendale and LADWP water. (a) Glendale water and (b) LADWP water (d, LFC-1; 5, ESNA; j, MX07; and e, GM).
experiments using uncharged organic compounds (urea and creatine) and an analysis, similar to the previous studies (Lee and Lueptow, 2001; Yoon and Lueptow, 2005). Two clear trends are observed for rejection data. First, the LFC-1 RO membrane with a small pore size (0.34 nm) had a greater rejection for all the anions (>90%) excluding NO 3 (71–74%) than the ESNA NF membrane (11– 56%) with a relatively large pore size (0.44 nm), indicating that size exclusion is at least partially responsible for the rejection. However, the small radius ions such as chloride and nitrate were slightly rejected by the NF and UF membranes (approximately
J. Yoon et al. / Chemosphere 77 (2009) 228–235
<20%), since the ri,s/rp ratios are very low (0.27–0.29) for ESNA. This indicates that electrostatic repulsion is insigniﬁcant to enhance the rejection, since for small monovalent ions the rejection is mainly governed by size exclusion for the relatively large pore size and negatively charged NF and UF membranes. Second, the rejection 2 2 of di-valent ions (CrO2 4 , SO4 , and HAsO4 ) is much larger than that for monovalent ions. This is because the di-valent ions have a larger size and greater charge than the monovalent ions, contributing to both size exclusion and electrostatic exclusion. Monova lent perchlorate ion (ClO4 ) rejection is also signiﬁcantly higher (approximately 20–35%) than the other monovalent ions (Cl and NO 3 ) by the ESNA NF membrane, even though the monovalent ions have very similar ionic radii (0.12–0.14). It is unclear why this phenomenon occurs. For di-valent ions the rejection is fairly proportional to the ri,s/rp ratio for the ESNA membrane. 4. Conclusions In this study, RO, NF, and UF membranes were tested to determine rejection of toxic anions (chromate, arsenate, and perchlorate) using laboratory scale experiments. The chromate, arsenate, and perchlorate rejections by the negatively charged RO and NF membranes are signiﬁcantly greater than expected based exclusively on steric/size exclusion due to electrostatic repulsion. The results also show that rejection of ions by negatively charged NF membranes is signiﬁcantly inﬂuenced by solution pH, since membrane surface charge becoming more negative with increasing solution pH enhances the chromate, arsenate, and perchlorate rejection. The rejection of the target toxic ions decreases with increasing solution conductivity for the membranes due to a reduction of electrostatic repulsion with increasing conductivity. The ratio of solute radius to effective membrane pore radius was employed to interpret the ion rejections. The rejection of all the ions by the negatively charged RO membrane is very signiﬁcant (71–99%), although the ri,s/rp ratio range is relatively low (0.36– 0.80), verifying that the rejection of a charged compound is governed by electrostatic exclusion in addition to steric exclusion. Acknowledgments The authors wish to acknowledge the American Water Works Association Research Foundation (now Water Research Foundation) for support of this study. References ATSDR, 1993. Toxicological Proﬁle for Chromium. Department of Health Services, US Public Health Service, La Porte County, Indiana. Bowen, W.R., Mohammad, A.W., 1998. Characterization and prediction of nanoﬁltration membrane performance – A general assessment. Chem. Eng. Res. Des. 76, 885–893. Brandhuber, P., Amy, G., 2001. Arsenic removal by a charged ultraﬁltration membrane – inﬂuences of membrane operating conditions and water quality on arsenic rejection. Desalination 140, 1–14. Childress, A.E., Elimelech, M., 1996. Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanoﬁltration membranes. J. Membr. Sci. 119, 253–268.
Childress, A.E., Elimelech, M., 2000. Relating nanoﬁltration membrane performance to membrane charge (electrokinetic) characteristics. Environ. Sci. Technol. 34, 3710–3716. Cho, J., Amy, G., 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, 2517– 2526. Cho, J., Amy, G., Pellegrino, J., 2000. Membrane ﬁltration of natural organic matter: factors and mechanisms affecting rejection and ﬂux decline with charged ultraﬁltration (UF) membrane. J. Membr. Sci. 164, 89–110. Elimelech, M., Chen, W.H., Waypa, J.J., 1994. Measuring the zeta (electrokinetic) potential of reverse-osmosis membranes by a streaming potential analyzer. Desalination 95, 269–286. Elimelech, M., Omelia, C.R., 1990. Effect of electrolyte type on the electrophoretic mobility of polystyrene latex colloids. Colloids Surf. 44, 165–178. EPA, 2002. Perchlorate Environmental Contamination: Toxicological Review and Risk Characterization Based on Emerging Information EPA, Report. NCEA-10535. Ergican, E., Gecol, H., Fuchs, A., 2005. The effect of co-occurring inorganic solutes on the removal of arsenic (V) from water using cationic surfactant micelles and an ultraﬁltration membrane. Desalination 181, 9–26. Golder, A.K., Chanda, A.K., Samanta, A.N., Ray, S., 2007. Removal of Cr(VI) from aqueous solution: Electrocoagulation vs chemical coagulation. Sep. Sci. Technol. 42, 2177–2193. Habuda-Stanic, M., Kalajdzic, B., Kules, M., Velic, N., 2008. Arsenite and arsenate sorption by hydrous ferric oxide/polymeric material. Desalination 229, 1–9. Hong, S., Elimelech, M., 1997. Chemical and physical aspects of natural organic matter (NOM) fouling of nanoﬁltration membranes. J. Membr. Sci. 132, 159– 181. Kim, J., Benjamin, M.M., 2004. Modeling a novel ion exchange process for arsenic and nitrate removal. Water Res. 38, 2053–2062. Kosutic, K., Furac, L., Sipos, L., Kunst, B., 2005. Removal of arsenic and pesticides from drinking water by nanoﬁltration membranes. Sep. Purif. Technol. 42, 137– 144. Lee, S., Lueptow, R.M., 2001. Reverse osmosis ﬁltration for space mission wastewater: membrane properties and operating conditions. J. Membr. Sci. 182, 77–90. Moore, K.W., Huck, P.M., Siverns, K., 2008. Arsenic removal using oxidative media and nanoﬁltration.. J. Am. Water Works Assoc. 100 (12), 74–83. Mulder, M., 1996. Basic Principles of Membrane Technology, second ed. Kluwer Academic Publishers, Dordrecht, The Netherlands. Muthukrishnan, M., Guha, B.K., 2008. Effect of pH on rejection of hexavalent chromium by nanoﬁltration. Desalination 219, 171–178. Sato, Y., Kang, M., Kamei, T., Magara, Y., 2002. Performance of nanoﬁltration for arsenic removal. Water Res. 36, 3371–3377. Smith, A.H., Lopipero, P.A., Bates, M.N., Steinmaus, C.M., 2002. Public health – arsenic epidemiology and drinking water standards. Science 296, 2145– 2146. Stumm, W., 1992. Chemistry of the Solid–Water Interface. Wiley, New Work. Uddin, M.T., Mozumder, M.S.I., Islam, M.A., Deowan, S.A., Hoinkis, J., 2007. Nanoﬁltration membrane process for the removal of arsenic from drinking water. Chem. Eng. Technol. 30, 1248–1254. Van Ginkel, S.W., Ahn, C.H., Badruzzaman, M., Roberts, D.J., Lehman, S.G., Adham, S.S., Rittmann, B.E., 2008. Kinetics of nitrate and perchlorate reduction in ionexchange brine using the membrane bioﬁlm reactor (MBfR). Water Res. 42, 4197–4205. Wilbert, M.C., Delagah, S., Pellegrino, J., 1999. Variance of streaming potential measurements. J. Membr. Sci. 161, 247–261. Yoon, J., Yoon, Y., Amy, G., Her, N., 2005a. Determination of perchlorate rejection and associated inorganic fouling (scaling) for reverse osmosis and nanoﬁltration membranes under various operating conditions. J. Environ. Eng.-ASCE 131, 726– 733. Yoon, Y., Amy, G., Cho, J., Her, N., 2005b. Effects of retained natural organic matter (NOM) on NOM rejection and membrane ﬂux decline with nanoﬁltration and ultraﬁltration. Desalination 173, 209–221. Yoon, Y., Amy, G., Cho, J., Pellegrino, J., 2004. Systematic bench-scale assessment of perchlorate rejection mechanisms by nanoﬁltration and ultraﬁltration membranes. Sep. Sci. Technol. 39, 2105–2135. Yoon, Y., Lueptow, R.M., 2005. Removal of organic contaminants by RO and NF membranes. J. Membr. Sci. 261, 76–86.