RBS Characterization of Arsenic(III) Partitioning from Aqueous Phase

RBS spectra for samples of the ESNA membrane with PA active layer and the ..... NRA, RBS and ERDA, Proceedingsof the 15th International Conference on ...
0 downloads 0 Views 360KB Size
Download PDF
Environ. Sci. Technol. 2007, 41, 3290-3295

RBS Characterization of Arsenic(III) Partitioning from Aqueous Phase into the Active Layers of Thin-Film Composite NF/RO Membranes B A O X I A M I , †,§ BENITO J. MARIN ˜ A S , * ,†,§ A N D D A V I D G . C A H I L L ‡,§ Department of Civil and Environmental Engineering, Department of Materials Science and Engineering, and National Science Foundation Science and Technology Center of Advanced Materials for the Purification of Water with Systems, University of Illinois at UrbanasChampaign, Urbana, Illinois 61801

The main objective of this study was to apply Rutherford backscattering spectrometry (RBS) for characterizing the partitioning of arsenic(III) from aqueous phase into the active layer of NF/RO membranes. NF/RO membranes with active layer materials including polyamide (PA), PApolyvinyl alcohol derivative (PVA), and sulfonatedpolyethersulfone (SPES) were investigated. The partition coefficient was found to be constant in the investigated As(III) concentration range of 0.005-0.02 M at each pH investigated. The partitioning of As(III) when predominantly present as H3AsO3 (pH 3.5-8.0) was not affected by pH. In contrast, the partition coefficient of As(III) at pH 10.5, when it was predominantly present as H2AsO3-, was found to be approximately 33-49% lower than that of H3AsO3. The partition coefficients of H3AsO3 and H2AsO3- for membranes containing PA in their active layers were within the respective ranges of 6.2-8.1 and 3.6-5.4, while the corresponding values (4.8 and 3.0, respectively) for the membrane with SPES active layer were approximately 30% lower than the average values for the PA membranes.

Introduction Nanofiltration/reverse osmosis (NF/RO) membrane processes are emerging as increasingly attractive technologies for drinking water treatment because they are capable of providing effective control for a broad range of contaminants. However, the rejection of small neutral molecules tends to be limited. For example, the intrinsic rejection of arsenic(III), predominantly arsenious acid (H3AsO3) in most natural waters, by commercial NF/RO membranes has been reported to be in the range of 20-90% (1, 2). The permeation of neutral molecules through NF/RO membranes is typically a combination of partition/diffusion through the polymer matrix and advection across membrane nanopores and larger imperfections. Although the contribution of each of these two transport components to overall * Corresponding author phone: (217)333-6961; fax: (217)333-6968; e-mail: [email protected]. † Department of Civil and Environmental Engineering. ‡ Department of Materials Science and Engineering. § National Science Foundation Science and Technology Center of Advanced Materials for the Purification of Water with Systems. 3290

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 9, 2007

solute permeation could be affected by many physical/ chemical properties of both solute and membrane (3-5), small neutral molecules such as H3AsO3 have been shown to permeate through NF/RO membranes predominantly by partition/diffusion (6). The solute flux for H3AsO3 at steady state is characterized by a permeation coefficient (B), which is the product of the partition coefficient (K) at the membrane/water interface and the diffusion coefficient (D) within the membrane phase divided by the membrane active layer thickness (δm), or B ) KD/δm. The development of new membrane materials with enhanced rejection of small neutral molecules will be facilitated if the mechanism for the relatively high permeation, e.g., high partitioning (K) or high diffusivity (D) of the solute, can be isolated. However, no approach has been developed to date for the independent determination of the partition and diffusion coefficients of solutes into the active layer of thin-film composite NF/RO membranes. Therefore, by introducing a new approach to characterize partition coefficient, this work will help improve our current understanding regarding the relatively high permeation (i.e., low rejection) of neutral molecules through NF/RO membranes. In previous work, Rutherford Backscattering Spectrometry (RBS) was used to characterize the elemental composition and thickness of the active layer of thin-film composite membranes (7). The primary objective of this study is to apply RBS for characterizing the partitioning of arsenic(III) at the aqueous/active layer interface for six commercial NF/ RO membranes as a function of As(III) concentration and pH, with potassium iodide used as a surrogate background electrolyte.

Materials and Methods Target Membranes. Six commercial NF/RO membranes were used: ESNA NF (Hydranautics, Oceanside, CA), TFCS NF (Koch Membrane Systems, Wilmington, MA), and FT30 RO (Dow Liquid Separation, Midland, MI) membranes with PA active layers, NTR729 NF and LF10 RO membranes (Nitto Denko, Shimohozumi, Japan) with PA-PVA active layers, and NTR7450 NF membrane (Nitto Denko, Shimohozumi, Japan) with SPES active layer. All of these membranes had a composite structure consisting of an active layer (50-250nm thick) on top of an asymmetric polysulfone (PSf) support (∼50-µm thick), in turn backed by a layer of unwoven polyester fibers (∼200-µm thick). Target Solutes. Arsenic(III), a drinking water contaminant of current interest, was selected as a small neutral molecule. Arsenious acid (H3AsO3) with an acid-base dissociation constant (pKa) of 9.2 is the predominant form of As(III) at typical natural water pH of 6-9, while arsenite ion (H2AsO3-) becomes predominant at pH above 9.2. Arsenic concentrations were determined as described previously (6). Potassium iodide (KI) was used as a background electrolyte surrogate for sodium chloride (NaCl), a more common electrolyte in natural waters, because compared to the lighter elements sodium and chlorine, potassium and iodine could be detected by RBS at lower concentrations and without interference from membrane polymer constituents. KI concentrations was determined as iodide ion by Standard Method 4110B (8) using an Ion Chromatograph Model DX-300 (Dionex, Sunnyvale, CA). Test solutions were prepared by dissolving arsenic trioxide in a 0.5 mM KOH solution at 100 °C and then adding KI. Target concentrations were in the range of 5-20 mM for As(III) and 0-32 mM for KI. The pH was adjusted with KOH and HI as needed. 10.1021/es062292v CCC: $37.00

 2007 American Chemical Society Published on Web 03/27/2007

Partitioning Experiments. Experiments were designed to characterize the partitioning of As(III) from aqueous solution into the membrane active layers. Membrane samples were first allowed to equilibrate with aqueous solutions at various target concentrations of As(III) and KI for 24 h. Once equilibrated they were taken out of the solution and immediately placed between two Grade No. 2 Whatman filter papers (Whatman Inc., Florham Park, NJ) for external drying. Then, each membrane sample was placed inside a 50 mL centrifugation vial with its polyester backing layer resting on a sponge placed at the bottom of the vial. Then the membrane surface was covered with a piece of clean filter paper and a sponge on top to keep the membrane in position while centrifuging. The sample tube was then spun at 12 000 rpm for 20 min in a RC-5B PLUS Superspeed Refrigerated Centrifuge (Thermo Electron Corporation, Asheville, NC), which provided a centrifugal force of approximately 20 000×g. The goal of this step was to remove liquid residue from the membrane support layer. Finally, the active and support layers of the membrane samples were separated from the polyester backing and fixed on a silicon wafer with thermally conductive double-sided tape. The samples were analyzed by RBS to obtain the concentrations of target solutes in the membrane active layer. Rutherford Backscattering Spectrometry Analyses. The concentrations of target chemical species in the membrane active layer were determined by performing RBS analyses at room temperature with a 2-MeV He+ beam generated with a Van de Graaff accelerator (High Voltage Engineering Corp., Burlington, MA). The incident angle, exit angle, and scattering angle of the incident He+ beam were 22.5°, 52.5°, and 150°, respectively. The beam current was controlled at approximately 25 nA. The shape of the incident He+ beam was circular with a diameter of 3 mm. The fluence of helium ions was controlled below the threshold value of 3 × 1014 He+/ cm2 by scanning the beam over an area of around 2 cm2. This threshold was selected to avoid losses in the physicochemical integrity of the polymers (9). A commercial computer simulation program, SIMNRA (10), was used to analyze the RBS data to determine the concentration of target solutes within the active and support layers. Additional details about the RBS analyses and background on the approach for model fitting were described previously (7). Attenuated Total Reflectance Fourier Transform Infrared Spectrometry Analyses. Attenuated total reflectance Fourier transform infrared (ATR-FTIR) spectrometry was used to characterize functional groups of membrane polymers. Analyses were performed with a NEXUS 670 FT-IR spectrometer (Thermo Nicolet Corporation, Madison, WI) equipped with a smart golden gate accessory, DTGS-KBr detector, KBr beam-splitter, and diamond crystal. An IR source at 45o incident angle was employed. Dry specimens of membrane samples were mounted with active layer facing the crystal surface.

Results Membrane Elemental Composition and Active Layer Thickness. RBS has been used in a previous study (7) to determine the elemental composition of the active and support layers as well as the thickness and roughness of the active layer of the NF/RO membranes selected for this study. The same approach was used to characterize the concentration of target chemicals in the membrane active layer in the present study. Because the penetration depth of the helium ion beam was on the order of 2 µm, RBS was able to determine the presence of target solutes in the active layer (∼50-250 nm thick) and part of the support layer. RBS spectra for samples of the ESNA membrane with PA active layer and the NTR7450 membrane with SPES active layer that were unexposed to As(III) or KI are shown in Figures 1 and 2 (spectra labeled

FIGURE 1. RBS spectra of ESNA membrane samples. Red represents membrane before partitioning experiments, and green represents samples soaked in 0.01 M H3AsO3 and 0.01 M KI solution at pH 6.5. The percentage values labeled on each peak are the concentration, E, of each element in the membrane active layer.

FIGURE 2. RBS spectra of NTR7450 membrane samples. Red represents membrane before partitioning experiments, and green represents sample soaked in 0.01 M H3AsO3 and 0.01 M KI solution at pH 6.5. The percentage values labeled on each peak are the concentration, E, of each element in the membrane active layer. as untreated samples). The spectra for the untreated samples have several major features expected for the polymers used to make these membranes (7). Plateau signals can be seen for the three major components of the PSf support layer, sulfur, oxygen and carbon, in both figures with respective high energy onsets of 1.27, 0.78, and 0.60 MeV. Oxygen and nitrogen peak signals associated with the PA active layer of the ESNA membrane appear on top of the PSf oxygen plateau in Figure 1. Oxygen and sulfur peak signals associated with the SPES active layer of the NTR7450 membrane appear on top of oxygen and sulfur plateau signals of the PSf support layer (Figure 2). The width and height of the peak signals defined the layer thickness and elemental concentration in the PA active layer (Table 1), which were obtained by fitting the experimental spectra with SIMNRA (7). The corresponding fit curves for the untreated samples, shown in Figures 1 and 2 as continuous lines, support that the model provided a good representation of the major characteristics of all RBS spectra. Consistent with previous results (7), RBS model fitting of these signals resulted in elemental composition in good agreement with those of PSf (C27H22O4S), fully aromatic PA (C18H12O3N3), and SPES (C12H8O3S). Also consistent with previous observations (7), the RBS spectra obtained for the TFCS, LF10, FT30 and NTR729 membranes (shown in Figures S1-S4 of the Supporting Information), all containing PA in their active layer, were generally similar to that shown in Figure 1 for the ESNA membrane. The average thickness and elemental composition of their active layers obtained by model fitting of the corresponding RBS spectra are summarized in Table 1. VOL. 41, NO. 9, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3291

TABLE 1. Elemental Composition and Active Layer Thickness Obtained from RBS Analyses active layer material projected density of active layer (excluding H), θ (atoms/cm2) mean membrane active layer thickness, δm (nm) elemental concentrations in active layer of original membranes (excluding H),  (%)

C N O S Cl Ca2+

ESNA

TFCS

FT30

LF10

NTR729

NTR7450

PA 4.8 × 1017

PA 4.0 × 1017

PA 9.4 × 1017

PA-PVA 1.1 × 1018

PA- PVA 2.3 × 1017

SPES 8.4 × 1017

100

85

201

231

48

158

73.5 11.8 13.2

73.1 10.7 15.0

72.3 8.0 19.7

70.3 7.3 21.7

77.0 8.0 15.0

72.7

1.4 0.13

0.9 0.19

0.07

0.7 0.14

0.12

0.82

Table 1 revealed that besides the major elements (carbon, oxygen, and nitrogen/sulfur) expected for PA/SPES, chlorine and calcium were also present in the active layer of some of the membranes. Ion exchange tests performed in a previous study (7) revealed that chlorine was covalently bonded to the polymer matrix, and calcium was ionically associated with charged functional groups. Target Solute Concentrations in Active Layers. The RBS spectra of the ESNA and NTR7450 membrane samples equilibrated with a solution containing 0.01 M H3AsO3 and 0.01 M KI with pH adjusted to 6.5 are compared to those of untreated membrane samples in respective Figures 1 and 2. In addition to the plateau and peak signals characteristic of the untreated membranes, peak signals for the three elements potassium, arsenic, and iodine that partitioned into the membrane active layers were also obtained with respective high-energy onsets of approximately 1.4, 1.65, and 1.8 MeV. Fitting of these spectra with SIMNRA resulted in the elemental concentrations of the active layer indicated under the corresponding elements in Figures 1 and 2. Experimental and fit spectra for the TFCS, FT30, LF10, and NTR729 membrane samples equilibrated with the same As(III)/KI solution are shown in Figures S1-S4 of the Supporting Information. The molar concentration of partitioned elements in the membrane active layer was calculated with the expression

cm )

θ LAδm

(1)

in which cm is the concentration of the element in the active layer (mM),  is the elemental concentration in the active layer expressed as elemental fraction, θ is the projected atomic density of the active layer (atoms/cm2) (Table 1), δm is the average thickness of the active layer (cm) (Table 1), and LA is Avogadro’s number (6.022 × 1023 atoms/mol). Partition Coefficients. The partition coefficient K was defined as the ratio of solute concentration in the membrane active layer cm to the concentration in aqueous solution cl at equilibrium:

K)

cm cl

(2)

The partition coefficient of As(III) from aqueous phase into the active layers of the membranes investigated was determined at aqueous phase concentrations in the range of 0.005-0.02 M. This concentration range, higher than what is normally found in natural waters (0.01-1 mg/L), was selected based on the detection limit of RBS. Also, it should be pointed that because the aqueous solution volume was in the order of 106 times that of membrane active layer, the aqueous concentration of As(III) should not change significantly during the equilibration experiment. A representative illustration of the RBS spectra obtained at various solution 3292

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 9, 2007

20.4 6.0

FIGURE 3. RBS spectra of TFCS membrane samples with As(III) partitioning at various liquid-phase concentration (cl). The pH of aqueous solution is at 6.5. concentrations is shown in Figure 3 for the TFCS membrane and solution pH of 6.5. As depicted in the figure, the As(III) peak increased with increasing As(III) concentration in the aqueous phase. The concentration of As(III) in the membrane active layer, cm, was obtained from the percent concentration from RBS model fitting of spectra such as those shown in Figure 3 using eq 1. The resulting values are plotted against aqueous phase concentration, cl, in Figure 4. As depicted by the figure, cm and cl had linear relationships that passed through the origin, which indicated that the arsenic partitioning was not affected by aqueous phase concentration within the range investigated. The slope of each line, therefore, represented the partition coefficient, K, of As(III) at the solution/membrane active layer interface.

Discussion pH Effect on As(III) Partitioning. The partitioning of As(III) at the aqueous/membrane active layer interface was studied at pH 3.5-10.5. Neutral arsenious acid (H3AsO3) was predominant at pH 3.5-8.0, while the monovalent arsenite anion (H2AsO3-) was predominant at pH 10.5. A representative illustration of the RBS spectra obtained is shown in Figure 5 for the TFCS membrane at an aqueous As(III) concentration of 0.02 M. As depicted in the figure, the As(III) peak increased with decreasing pH, consistent with greater absorption of the neutral As(III) species. Once again the concentrations of As(III) in the membrane active layer, cm, were obtained from the percent concentration obtained from RBS model fitting of spectra such as those shown in Figure 5 using eq 1, and the resulting values are plotted against aqueous phase concentration, cl, in Figure 4. As shown in Figure 4 for each membrane the partition coefficient of arsenious acid was constant within experimental variability for the pH range of 3.5-8.0. In contrast, the partition coefficient was 33-49% lower when most of the As(III) was present as H2AsO3- (pH 10.5). The lower partitioning of the ionized form of As(III) might have resulted, at least in part, from Donnan exclusion by charged carboxylate groups.

FIGURE 6. Effect of KI concentration on H3AsO3 partitioning into the ESNA membrane at pH 6.5. The partition coefficient of H3AsO3 remained constant when varying the KI concentration in the aqueous phase.

FIGURE 4. Arsenic partitioning into RO/NF membrane active layers at various pH and aqueous phase concentrations. Arsenic partitioning has a linear relationship between the aqueous and membrane active concentration for the range of experimental conditions investigated in this study. Therefore, the slope of each fitting line represents the As(III) partition coefficient, K. As(III) is present primarily as a neutral molecule (H3AsO3) at pH 3-8 and as negatively charged ions (H2AsO3-) at pH 10.5.

FIGURE 5. RBS spectra of TFCS membrane samples with As(III) partitioning at various pH. The As(III) concentration in aqueous solution (cl) is 0.02 M. Role of Membrane Material in As(III) Partitioning. The equilibrium partitioning of neutral molecules at the interface of aqueous solution and membrane active layer could be affected by two processes: steric effects and solutemembrane affinity interactions. The molecules and ions of interest in this study are all relatively small, and so steric effects would play a relatively lesser role. With respect to solute-membrane interactions with As(III) species they could involve the formation of hydrogen bonds between polar groups (5). Therefore, membrane materials with different types and concentrations of functional groups would result in different levels of As(III) partitioning. The partition coefficients of neutral H3AsO3 obtained for all five membranes with PA or PA-PVA active layer (ESNA, TFCS, FT30, LF10, and NTR729 membranes) are in the range of 6.2-8.1. In contrast, as shown in Figure 4, the arsenic partition coefficients obtained for the NTR7450 membrane with SPES active

FIGURE 7. Effect of iodide ion concentration (up to 0.11 M) on the partitioning of H2AsO3- (0.06 M) into the ESNA membrane at pH 10.5. The partition coefficient of H2AsO3- remained constant when varying the KI concentration in the aqueous phase. layer were approximately 30% lower than the average values of the five membranes with PA in their active layers. This would be consistent with groups in PA providing greater overall level of hydrogen bonding for As(III) compared to groups in SPES for the entire range of experimental conditions investigated. Effect of KI Concentration on As(III) Partitioning. The effect of KI concentration on As(III) partitioning was studied with 0.02 M (pH 6.5) or 0.014 M (pH 10.5) solutions of As(III) to which KI was added at various concentrations up to 0.02 M. Figures 6 and 7 illustrated the results obtained from analyzing RBS spectra for the partitioning experiments performed with the ESNA membrane. The data revealed that the partition coefficient of As(III) not only was the same for both pH levels tested but also was not affected by the presence of ionic species within the KI concentration range investigated. Validity of Partitioning Measurement. The most significant result of this study is the experimental quantification of As(III) partitioning within the active layer of RO/NF membranes. It is important, therefore, to discuss the uncertainties in our data that are introduced by the method used to prepare the samples. When we derive the partition coefficient, we assume that the areal density of atoms measured by RBS accurately reflects the concentrations of ions in the membrane when the membrane was in contact with the aqueous solution; i.e., we assume that the process of drying the membrane has not caused the concentration of ions to change significantly. Such a change in concentration could conceivably arise from (i) incomplete removal of solution from the external surface of the membrane or (ii) by the fact that any solution left in the support layer might VOL. 41, NO. 9, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3293

become more concentrated as the membrane dries and this increase in the concentration of ions in the support layer could drive additional ions into the active layer by diffusion. Although we have no direct evidence of these possible errors in the measurement of the partition coefficients, the validity of measurements can be supported by looking at two aspects of the data. First, the validity of partitioning data is supported by the relative values of the areal densities observed for As(III), K+, and I- (as shown in Figures 1, 2, and S1-S4). Drying of residual aqueous solution on the external surface would increase the concentrations of As(III), K+, and I- proportionally, and thus the partition coefficients for all three species should be the same. This was not the case. For example, the measured concentrations () of As(III), K+, and I- in the NTR7450 membrane active layer at pH 6.5 were 0.04%, 1.5%, and 0.02%, respectively. Second, the validity of the data can also be confirmed by the different As(III) partitioning behavior at low pH (3.5, 6.5, and 8.0) and at high pH (10.5). The As(III) partition coefficients at pH 3.5-8.0 was 1.5-2 times higher than that at pH 10.5. Since drying would result in the same value for all pH, the observed pH effects further verified the reliability of the partitioning data. The presence of nanometer scale porosity within the active layer could also impact the interpretation of the RBS data. Consider a ∼10 nm diameter pore in the active layer that is completely surrounded by polymer membrane. In equilibrium, the concentrations of ions in the nanopore would be nearly the same as the concentration of solutes in the solution external to the membrane. Ions in this nanopore would contribute to the RBS signal even though these ions are not dissolved within the matrix of the polymer membrane. Because the partition coefficient of As(III) is larger than 1, we do not believe that nanoporosity significantly affects our interpretation of the data, but nanoporosity would be important to consider for solute species that are strongly excluded from the polymer matrix. The partition coefficients of As(III) measured in this study are within the range of 3-8. These values are comparable to the partition coefficient of ∼2 estimated by atomic simulations of salt transport through the FT30 membrane (11) and partition coefficients of 1-7 reported for As(III) permeation through polyamide NF membranes (12). Hypothesis of Partitioning Mechanism of Neutral H3AsO3. The partition coefficients measured in this study are not in agreement with an empirical expression solely based on steric hindrance (13, 14), see eq 3, that is often used to estimate the partition coefficient of neutral molecules into charged NF membranes

K ) (1 - λ)2

(3)

where λ is the ratio of uncharged solute radius rs to pore radius rp. Equation 3 was originally derived to describe the steric hindrance at the pore entrance in a pore flow model (15) and used to evaluate the transport of viruses and large molecules through porous membranes (16, 17). Since λ < 1, eq 3 predicts K < 1. The measured partition coefficients of As(III) from the present study, however, are larger than unity, which indicates that partitioning controlled by steric hindrance is not a valid description for small neutral molecules. It is hypothesized, instead, that the partitioning of H3AsO3 into polymer membranes involves the formation of hydrogen bonds between the molecule and the membrane consistent with previous work (5) that found a reduction in the rejection of solutes with increasing polarity. Among the possible hydrogen bonds formed between different functional groups, OH...N with binding energy of 29 kJ/mol is stronger than OH...O (21 kJ/mol) and NH...O (8 kJ/mol) (18). Although the strength of a hydrogen bond is also affected by the specific 3294

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 9, 2007

FIGURE 8. ATR-FTIR spectra of ESNA, TFCS, FT30, LF10, and NTR729 membranes and PSf support without active layer. NTR729 does not show amide bands at 1660 and 1540 as the other four membranes do, although it is also classified as a PA-PVA membrane, thus indicating that the PVA component might be predominant. structure of the molecule containing an OH group, it is plausible that the hydrogen bond formed between OH from H3AsO3 (as hydrogen donor) and N from polyamide (as hydrogen acceptor) would play an important role in the partitioning process. This hypothesis was partially supported by the correlation found between partition coefficient and N concentration in the PA and PA-PVA membranes: the membrane with higher nitrogen contents has higher H3AsO3 partition coefficients. The NTR729 membrane, however, did not follow the trend. The discrepancy is most possibly caused by the different polymer chemistry of NTR729, as evidenced by the ATR-FTIR spectra shown in Figure 8 that NTR729 does not show amide bands at 1660 and 1540 as the other four membranes do. However, we need to point out that the correlation between partition coefficient and nitrogen concentration of polymer is not conclusive due to two reasons. One is that the partition coefficient could also be affected by membrane physical structure through steric effects, and the other is that the difference in partition coefficients for the various PA membranes investigated is relatively small, which introduces uncertainty by data variability.

Acknowledgments RBS analyses were carried out in the Center for Microanalysis of Materials, University of Illinois, partially supported by the U.S. Department of Energy under grant DEFG02-91-ER45439. The authors acknowledge Michael J. Williams and Doug Jeffers for assistance in RBS analyses, Jim Lozier from CH2M Hill, Tasuma Suzuki from the University of Illinois at UrbanaChampaign for facilitating information about membrane materials, and the companies Hydranautics, Oceanside, CA, Koch Membrane Systems, Wilmington, MA, and Nitto Denko, Shimohozumi, Japan for providing membrane materials and related information. This work was supported by the National Science Foundation Environmental Engineering and Technology program under agreement number BES-0332217 and the WaterCAMPWS, a Science and Technology Center of Advanced Materials for the Purification of Water with Systems under agreement number CTS-0120978. The opinions in this paper do not necessarily reflect those of the sponsor.

Supporting Information Available Additional spectra (Figures S1-S4). This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Sato, Y.; Kang, M.; Kamei, T.; Magara, Y. Performance of nanofiltration for arsenic removal. Water Res. 2002, 36 (13), 3371-3377.

(2) Oh, J.-I.; Lee, S.-H.; Yamamoto, K. Relationship between molar volume and rejection of arsenic species in groundwater by lowpressure nanofiltration process. J. Membr. Sci. 2004, 234 (1-2), 167-175. (3) Kimura, K.; Amy, G. L.; Drewes, J.; Watanabe, Y. Adsorption of hydrophobic compounds onto NF/RO membranes-an artifact leading to overestimation of rejection. J. Membr. Sci. 2003, 221, 89-101. (4) Freger, V.; Arnot, A. C.; Howell, J. A. Separation of concentrated organic/inorganic salt mixtures by nanofiltration. J. Membr. Sci. 2000, 178, 185-193. (5) Van der Bruggen, B.; Schaep, J.; Vilms, D.; Vandecasteele, C. Influence of molecular size, polarity and charge on the retention of organic molecules by nanofiltration. J. Membr. Sci. 1999, 156, 29-41. (6) Mi, B.; Marin ˜ as, B. J. Role of pore size distribution in the transport of water contaminants through nanofiltration membranes, American Water Works Association Water Quality Technology Conference, San Antonio, Texas, November 14-18, 2004. (7) Mi, B.; Coronell, O.; Marin ˜ as, B. J.; Watanabe, F.; Cahill, D. G.; Petrov, I. Physico-chemical characterization of NF/RO membrane active layers by Rutherford backscattering spectrometry. J. Membr. Sci. 2006, 282, 71-81. (8) American Public Health Association, American Water Works Association, and Water Environment Federation. Standard Methods for the Examination of Water and Wastewater, 18th ed.; American Public Health Association: Washington, DC, 1992. (9) Mi, B.; Cahill, D. G.; Marin ˜ as, B. J. Physico-chemical integrity of nanofiltration/reverse osmosis membranes during characterization by Rutherford backscattering spectrometry. J. Membr. Sci. 2007, 291, 77-85. (10) Mayer, M.; Duggan, J. L.; Morgan, I. L. SIMNRA, a simulation program for the analysis of NRA, RBS and ERDA, Proceedings

(11) (12)

(13) (14) (15) (16) (17) (18)

of the 15th International Conference on the Application of Accelerators in Research and Industry, Denton, TX, 1998; p 541. Kotelyanskii, M. J.; Wagner, N. J.; Paulaitis, M. E. Atomic simulation of water and salt transport in a reverse osmosis membrane FT30. J. Membr. Sci. 1998, 139, 1-16. Oh, J.-I.; Lee, S. H.; Yamamoto, K. Relationship between molar volume and rejection of arsenic species in groundwater by lowpressure nanofiltration process. J. Membr. Sci. 2004, 234, 167175. Bowen, W. R.; Welfoot, J. S. Modelling the performance of membrane nanofiltration-critical assessment and model development. Chem. Eng. Sci. 2002, 57, 1121-1137. Sarrade, S.; Rios, G. M.; Carles, M. Dynamic characterization and transport mechanisms of two inorganic membranes for nanofiltration. J. Membr. Sci. 1994, 97, 155-166. Pappenheimer, J. R.; Renkin, E. M.; Borrero, L. M. Filtration, diffusion and molecular sieving through peripheral capillary membranes. Am. J. Physiol. 1951, 167, 13-46. Nakao, S.; Kimura, S. Models of membrane transport phenomena and their application for ultrafiltration data. J. Chem. Eng. Jpn. 1982, 15 (3), 200-205. Herath, G.; Yamamoto, K.; Urase, T. Mechanism of bacterial and viral transport through microfiltration membranes. Water. Sci. Technol. 1998, 38 (4-5), 489-496. Chen, J. Y.; Naidoo, K. J. Evaluation of intramolecular hydrogen bond strength in (1-4) linked disacharides from electron density relationships. J. Phys. Chem. B 2003, 107, 9558-9566.

Received for review September 25, 2006. Revised manuscript received February 23, 2007. Accepted February 24, 2007. ES062292V

VOL. 41, NO. 9, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3295