Environ. Sci. Technol. 2006, 40, 7343-7349
Use of Reverse Osmosis Membranes to Remove Perfluorooctane Sulfonate (PFOS) from Semiconductor Wastewater† CHUYANG Y. TANG, Q. SHIANG FU,* A. P. ROBERTSON, CRAIG S. CRIDDLE, AND JAMES O. LECKIE Environmental Engineering and Science, Department of Civil and Environmental Engineering, Stanford University, Stanford, California 94305-4020
Perfluorooctane sulfonate (PFOS) and related substances are persistent, bioaccumulative, and toxic, and thus of substantial environmental concern. PFOS is an essential photolithographic chemical in the semiconductor industry with no substitutes yet identified. The industry seeks effective treatment technologies. The feasibility of using reverse osmosis (RO) membranes for treating semiconductor wastewater containing PFOS has been investigated. Commercial RO membranes were characterized in terms of permeability, salt rejection, scanning electron microscopy (SEM), transmission electron microscopy (TEM), and membrane surface zeta potential (streaming potential measurements). Filtration tests were performed to determine the membrane flux and PFOS rejection. Over a wide range of feed concentrations (0.5 - 1500 ppm), the RO membranes generally rejected 99% or more of the PFOS. Rejection was better for tighter membranes, but was not affected by membrane zeta potential. Flux decreased with increasing PFOS concentration. While the flux reduction was severe for a loose RO membrane probably due to its higher initial flux, very stable flux was maintained for tighter membranes. At a very high feed concentration (about 500 ppm), all the membranes exhibited an identical stable flux. Isopropyl alcohol, present in some semiconductor wastewaters, had a detrimental effect on membrane flux. Where present it needs to be removed from the wastewater prior to using RO membranes.
1. Introduction Perfluorooctane sulfonate (PFOS), an emerging contaminant, is an eight-carbon perfluorinated alkane with a sulfonate group at one end (CF3(CF2)7SO3H). PFOS and its derivatives have been widely used as a coating material on paper, photographs, packaging, and textile products as they repel both water and oil (1). The surface activity of these compounds and their stability even at high temperature also allow them to be used in fire-fighting foam, hydraulic fluids, and metal plating solutions (2). In addition, PFOS and its derivatives play an important role in the semiconductor industry because of their unique optical characteristics and †
This article is part of the Emerging Contaminants Special Issue. * Corresponding author phone: 650-724-5310; fax 650-723-3162; e-mail:
[email protected]. 10.1021/es060831q CCC: $33.50 Published on Web 10/05/2006
2006 American Chemical Society
acid-generating efficiency (2). The semiconductor industry in the European Union (EU) alone uses an estimated 470 kg annually (2). PFOS is principally used in wafer photolithography, where it serves as (a) a photoacid generator (PAG) in positive photoresist, (b) an antireflective coating (ARC), and (c) a surfactant in the developer solution. Typical concentrations are in ranges of 0.02-0.1%, 0.1%, and 0.01-1%, respectively (2, 3). In the photoresist coating process, photoresist and ARC are sequentially applied to the center of the wafer and spread by spinning the wafer. About 93-99% of the solution is spun off, drained to solvent waste, and subsequently incinerated (2). However, some photolithographic facilities drain about 40% of ARC to wastewater (2). Photoresist that remains on the wafer is then patterned by exposing selective areas to UV radiation. For positive photoresist, the exposed portions become soluble in a developing solution. Photoresist that does not wash off during the developing step is removed either by dry stripping using oxygen plasma in which case PFOS is completed destroyed, or by wet stripping in which case it is dissolved in a stripping chemical such as sulfuric acid. Currently, wastewater from the developing and wetstripping washes passes into a main wastewater collection system, is neutralized, and sent to municipal sewage treatment plants. Considerable dilution occurs as this stream combines with streams from other rinse activities (3). The long-term fate of PFOS and related compounds is poorly understood, though some generalizations are possible. PFOS has been found in the blood sera, livers, and bladders of human beings and a wide range of animals all over the world (4, 5), including the Arctic. Animals high in the food chain such as minks, bald eagles, and polar bear, had the highest concentrations, suggesting it is bioaccumulative (4, 6). It is toxic to rats, birds, and monkeys (6). The perfluorinated chain is resistant to enzyme attack, and, while it does not defluorinate, attached functional groups may be modified (7). Wastewater is likely the prime conduit for entry of PFOS into the environment (5, 8). PFOS and PFOS precursor compounds have been found in sludge samples from wastewater treatment plants (8, 9). There is no reliable evidence of PFOS biodegradation (5, 7, 10). Due to its environmental persistence, bioaccumulation, and potential toxicity to humans, the European Union is proposing legislation to ban the use of PFOS (11). Despite the above concerns, it appears that PFOS will remain a key ingredient in photoresist and ARC formulations for many years to come (11). Development of a PFOS substitute could take 10-15 years before high-volume manufacturing is possible (3). The semiconductor industry thus seeks to obtain an exemption from the ban for the critical uses of PFOS in photolithography. PFOS generated from the semiconductor industry is presumably discharged primarily in the manufacturing wastewater. Therefore, the industry is seeking a cost-effective process for removal of PFOS from the wastewater (3). Reverse osmosis (RO) is one possible option. RO membranes are effective in removing most organic (12-14) and inorganic compounds from water solutions. In recent years, new polymer chemistry and manufacturing processes have brought about dramatic improvements in efficiency, lowering operating pressures and reducing costs. As a result, RO membranes are increasingly used by industry to concentrate or remove chemicals. This paper describes an investigation to determine the feasibility of removing PFOS from semiconductor wastewater by RO membranes. PFOS concentrations in a semiconductor VOL. 40, NO. 23, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7343
wastewater were measured by liquid chromatograph and tandem mass spectrometry (LC/MS/MS). Flux performance and PFOS rejection efficiency of different membranes were determined for feed solutions containing a wide range of PFOS concentrations.
2. Materials and Methods 2.1 Chemicals and Materials. 2.1.1 General Chemicals. Unless otherwise specified, all reagents and chemicals were analytical grade with purity over 99%. Purified water was supplied in-house from a MilliQ water system (Millipore, Billerica, MA) with a resistivity of 18.2 Mohm-cm. Sodium chloride, sodium hydroxide (1 N), and hydrochloric acid (1 N) were purchased from Fisher Scientific (Santa Clara, CA). Isopropyl alcohol was purchased from EMD Chemicals (Gibbstown, NJ). The acid form of PFOS (100%) was supplied by AZ Electronic Materials USA Corp. (Somerville, NJ) through SEMATECH (Austin, TX). LR White resin and ethanol used for transmission electron microscopy (TEM) sample preparation were obtained from Polyscience (Warrington, PA) and Gold Shield (Hayward, CA), respectively. 2.1.2 Chemicals Used in LC/MS/MS. Perfluorooctanesulfonate (PFOS, 100%) was provided by AZ Electronic Materials USA Corp. (Somerville, NJ). The internal standard perfluorononanoic acid (PFNA, 97%) was purchased from Fluka through Sigma-Aldrich (St. Louis, MO). Optima grade methanol was purchased from Fisher Scientific (Pittsburgh, PA), and ammonium acetate and glacial acetic acid were purchased from Mallinckrodt (Phillipsburg, NJ). Stock solutions of PFOS and PFNA were prepared in methanol and used for calibration and quantitation. 2.1.3 RO Membranes. Four commercial thin-film composite (TFC) polyamide (PA) RO membranes were used: ESPA3 and LFC3 were kindly supplied by Hydranautics (Oceanside, CA), BW30 was provided by Dow FilmTec (Minneapolis, MN), and SG was purchased from GE Osmonics (Minnetonka, MN). All membranes were supplied and stored as dry coupons. Before being loaded into test cells, membranes were thoroughly rinsed with milliQ water and soaked in a milliQ water bath for 24 h. 2.2 LC/MS/MS. LC/MS/MS has been demonstrated as the most sensitive and specific technique for analysis and measurement of PFOS. A method similar to one described by Higgins et al. (8) was used to determine the concentration of PFOS for the PFOS-containing wastewater. 2.2.1 Sample Preparation. Four bottles of semiconductor manufacturing wastewater were collected at a facility that used PFOS, isopropyl alcohol, and ultrapure water. The bottles were shipped to our laboratory, where they were immediately stored at 4 °C. The pH of the four samples ranged from 2.85 to 3.0. A 100 µL sample from each bottle was diluted by methanol 100 000 times. Samples from RO membrane feed and permeate solutions were diluted by methanol to concentrations within the calibration range (0.1-25 ng/mL). One-mL aliquots of diluted sample were transferred to 2-mL glass autosampler vials where 100 µL of 11 ng/mL internal standard solution (in methanol) was added prior to analysis. 2.2.2 Analysis via LC/MS/MS. Samples were analyzed under conditions similar to those previously reported (8). Chromatography was performed using an aqueous ammonium acetate (2 mM) and methanol gradient delivered at a flow rate of 200 µL/min by a Shimadzu LC system (LC10ADvp pumps controlled by a SCL10Avp controller, Columbia, MD). Samples and standards were injected (20 µL) by a Shimadzu SIL10ADvp autosampler onto a 50 mm by 2.1 mm Targa C18 column (5 µm pore size, Higgins Analytical, Mountain View, CA) equipped with a C18 Guard Column (Higgins Analytical). Initial eluent conditions were 40% methanol, and the methanol content was increased to 95% at 5.5 min, ramped 7344
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 23, 2006
to 100% over 8.5 min and held for 1 min, and lowered to 40% at 11.5 min. The total run time was 12 min. A Sciex API 3000 triple quadropole mass spectrometer (MDS Sciex, Ontario, Canada) operating in negative ESI multiple reaction monitoring (MRM) mode was employed for sample analysis. Two MRM transitions (primary transition PFOS- f FSO3-; and secondary transition PFOS- f SO3-) were used for quantitation and confirmation and a dwell time of 100 ms was used for each transition. Optimal instrumental source parameters were set at the following levels: ionspray voltage -2800 V, curtain gas flow 8 arbitrary units (au), nebulizer gas flow 14 au, turbo gas flow 8 au, collision gas flow 12 au, and source temperature 425 °C. Zero air provided by a Parker-Balston 76-818 Zero Air Generator (Haverhill, MA) was used for the nebulizer and drier gas, and nitrogen (provided by a Parker-Balston N2-4000 generator) was used as the curtain and collision gas. 2.2.3 Quantitation. A six-point calibration spanning the 0.1-25 ng/mL range was run at the beginning and end of every sample batch. The coefficient of determination (R 2) was greater than 0.99 for curves generated from both the primary and secondary transitions. The fitted values for all points were within 15% of their actual values. Compound confirmation was based on comparison of retention times of known standards as well as peak detection of both primary and secondary transitions. 2.3 Membrane Testing Setup. Permeate flux and PFOS rejection were determined using a custom-made setup (Figure A, Supporting Information). Membrane coupons (14.6 cm × 9.5 cm) were housed in stainless steel membrane cells (CEPA CFII, GE Osmonics, Minnetonka, MN) assembled in parallel. Feed solution was pumped with a variable-speed HydraCell diaphragm pump (model D-03, Minneapolis, MN) from a polypropylene tank. The temperature inside the feed tank was maintained constant (25 ( 1 °C) with a recirculation water bath (VWR, West Chester, PA). The pressure inside the membrane cells was set with a backpressure regulator (HydraCell, Minneapolis, MN). The cross-flow was adjusted with a needle valve downstream of each membrane cell. Cross-flow was monitored with in-line manual flowmeters (Cole Parmer 32477-06, Vernon Hills, IL). Flux was determined gravimetrically by weighing the mass of feedwater collected at predetermined time intervals. Electronic data monitoring was available for temperature (Type J Transmitter, Cole Parmer 94770-02), pressure (Gauge Transmitter, Cole Parmer 68073-14) and pH (pH Transmitter, Cole Parmer 56717-20). A FieldPoint network module (National Instrument, Austin, TX), programmed with customary LabView codes, was used to log and process the electronic data. 2.4 Membrane Rejection Tests. PFOS rejection tests were performed over a range of feed solution concentrations (0.51600 ppm). The 1600 ppm feed solution contained 5% isopropyl alcohol (representative of the waste we received). The alcohol increases the solubility of PFOS and prevents phase separation. No isopropyl alcohol was added to other PFOS solutions. To eliminate the effect of membrane compaction, membranes were precompacted by filtering MilliQ water through them for 24 to 48 h, until a steady-state flux was observed. PFOS and, if necessary, isopropyl alcohol were then introduced. Tests were continued for at least 24 h before the feed PFOS concentration was ramped up to a higher level. The solution pH was adjusted to 4.0 ( 0.1 with NaOH or HCl right after each PFOS concentration adjustment. Intermittently during the test period, membrane permeate flux was measured gravimetrically, and both the feed tank and permeate samples were collected for analysis. For all rejection tests, the pressure inside the membrane cells was maintained at 1380 kPa (200 ( 2 psi) and the cross-flow was 1.37 L/min ((10%, corresponding to a superficial velocity of 20 cm/s)
during precompaction. They were fine-tuned to 200 ( 1 psi and 1.37 L/min ( 5% once PFOS was introduced to the feed tank. To assess the effect of isopropyl alcohol on membrane permeate flux, isopropyl alcohol concentration was ramped up to a final concentration of 5% in 4 steps (1.25%, 2.50%, 3.75%, and 5.00%), with each intermediate step lasting about 30 min. 2.5 Membrane Characterization. Virgin membranes were characterized in terms of scanning electron microscopy, transmission electron micrroscopy, streaming potential, membrane permeability, and salt rejection. 2.5.1 Scanning Electron Microscopy (SEM). SEM micrographs were obtained by an FEI XL30 Sirion Microscope at an accelerating voltage of 5 kV. Prior to imaging by SEM, samples were sputter coated with a uniform layer of platinum, approximately 10 nm thick, to avoid charge effects of the nonconductive samples. 2.5.2 Transmission Electron Microscopy (TEM). TEM micrographs were obtained by a JEM 1230 Electron Microscope at an accelerating voltage of 80 kV. The aromatic acrylic LR White resin was used to embed membrane samples. No staining was necessary to obtain high contrast images with this resin. Samples were dehydrated and embedded by immersing membrane coupons in the following solutions (in order): 100% ethanol, three changes for 15 min each; 1:1 ethanol/LR White resin mixture for 1 h; 1:2 ethanol/LR White resin mixture for 2 h; 100% LR White resin, 3 h; and another change of 100% LR White resin for 2 h. Samples were then transferred to dry gelatin capsules filled with fresh LR White resin. The capsules were tightly capped to minimize the presence of oxygen which inhibits the polymerization process. They were then left in a 48 °C oven for 3 days to allow complete polymerization. Thin TEM sections less than 100 nm thick were cut with a diamond knife by a Leica Ultracut S ultramicrotome (Leica, Wetzlar, Germany) and transferred onto copper TEM grids for imaging. 2.5.3 Streaming Potential. Streaming potential measurements were performed using an Electro Kinetic Analyzer (EKA) equipped with an asymmetric clamping cell (Anton Paar, Graz, Austria). Zeta potentials for membrane samples were computed from the Helmoltz-Smoluchowski equation using the Fairbrother and Mastin substitution (15). The EKA included an automatic pH titrator, external pH and conductivity electrodes, an asymmetric flow cell with a platinum electrode on each end of the flow cell, and a control system connected to a computer. Unlike conventional measuring cells where two identical flat samples are mounted face to face and separated by a channel spacer such as a thin piece of Teflon, streaming potential values were measured in this work using a single piece of sample mounted to a reference poly(methyl methacrylate) (PMMA) channel plate for the asymmetric cell. A full description on the theory and applications of the asymmetric clamping cell can be found elsewhere (16, 17). The measured zeta potential (ζM) is the average of the sample zeta potential (ζS) and the zeta potential of reference channel plate (ζR):
1 ζM ) (ζS + ζR) 2
(1)
The reference zeta potential can be determined by replacing the sample with a reference flat plate also made of PMMA from the same batch as that for the channel plate. Membrane samples were thoroughly rinsed and soaked in MilliQ water for 24 h prior to analysis. The EKA system was cleaned by rinsing with 2 L each of the following solutions, in order, without recirculation: MilliQ water, 1 mM NaOH, MilliQ water, 1 mM HCl, and MilliQ water. After mounting
the membrane samples the system was flushed with 2 L of 10 mM NaCl at pH 3. Streaming potential measurements were performed in a background electrolyte of 10 mM sodium chloride over a pH range from 3 to 10 at room temperature (22-24 °C). The background electrolyte was degassed at pH 3 before analysis, and a nitrogen headspace was maintained to eliminate potential artifacts from the presence of carbon dioxide. Solution pH was increased in small steps by adding aliquots of 1 N sodium hydroxide with the automatic pH titrator. The equilibration time at each pH was at least 15 min. 2.5.4 Membrane Permeability and Salt Rejection. Membrane permeability and salt rejection were evaluated in the setup described in Section 2.3. The procedures and conditions were similar to those in Section 2.4: temperature at 25 °C, pressure at 1379 kPa (200 psi), and superficial velocity at 20 cm/s. Membranes were thoroughly rinsed with MilliQ water, soaked for 24 h, and pre-compacted for 48 h. “Pure water flux” was defined as the measured flux at the end of the compaction period. Sodium chloride solution was then added to the feed tank to achieve a concentration of 10 mM, and the test continued for another 24 h to determine salt rejection at steady state. The conductivity of both the feed and the permeate was measured by a Ultrameter II (Myron L Company, Carlsbad, CA) to determine membrane salt rejection. In all cases, the sodium chloride concentration in the feed tank was stable (within ( 1%).
3. Results and Discussion 3.1 PFOS Concentration in Semiconductor Wastewater. Four diluted samples (100 000× dilution) from a wastewater generated from photolithographic processes were analyzed by LC/MS/MS and the PFOS concentration was calculated based on the calibration curves generated from both the primary (PFOS- f FSO3-) and the secondary transitions (PFOS- f SO3-). The measured PFOS concentration in the original semiconductor wastewater was ∼1650 mg/L. This value is much higher than its solubility in pure water (570 mg/L as potassium salt of PFOS at 25 °C (2)) and can be attributed to the fact that this water contained about 5% isopropyl alcohol, added to enhance PFOS solubility (3). 3.2 Virgin Membrane Properties. SEM and TEM images of the RO membranes are shown in Figure 1. Typical structures for thin film composite polyamide RO membranes are illustrated by the SEM images. These membranes are composed of three layers: a top dense polyamide layer responsible for selectivity, a microporous polysulfone layer, and a nonwoven fabric layer as support. The polysulfone layer is typically several tens to hundreds of micrometers thick, and the polyamide layer is usually less than one micrometer thick. Polyamide top layer details are shown in the TEM images. The bright region represents the embedding LR White resin, while the polyamide layer is the darker layer of a few hundred nanometers thick with a rough surface. The observed surface roughness is common in membranes formed by an interfacial polymerization process. In general, tight RO membranes such as BW30 and LFC3 tend to have higher salt rejection than loose RO membranes such as ESPA3 (Table 1). SG seems to be an exception that it has similar sodium chloride rejection with ESPA3 although it is the tightest membrane based on flux. Our salt rejections were generally lower than those specified by manufacturers, possibly due to the different testing conditions. Membrane modules tested by manufacturers are typically stored wet in pressure vessels, while our membrane samples were all supplied as dry coupons. Figure 2a shows the zeta potential of the PMMA reference plate in 10 mM sodium chloride background electrolyte over the pH range 3-10 under nitrogen headspace (8 replicates). VOL. 40, NO. 23, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7345
FIGURE 1. SEM and TEM images of virgin membranes. (a) ESPA3 and (b) BW30 SEM images with 20 µm scale bars. Both SEM samples were inclined at 45° to the mounting stage. (c) ESPA3, (d) BW30, and (e) LFC3 TEM images with 2 µm scale bars. The results are consistent and agree reasonably well with literature values (16, 17), except at higher pH values. These other studies used 10 mM potassium chloride as the background electrolyte and did not exclude carbon dioxide. Our data fit well to a third-order polynomial (R 2 ) 0.99). The zeta potential of the four commercial membranes is shown in Figure 2b. Duplicates were run for each membrane. Values were calculated using eq 1 with ζR determined from the regression equation shown in Figure 2a. From Figure 2b, it is clear that membranes were more negatively charged at higher pH. The isoelectric points for ESPA3, BW30, and LFC3 were all between pH 4 and 5, which is typical for polyamide membranes. LFC3 had surface charge properties similar to those of BW30, but was slightly more negatively charged between pH 5-7. Indeed, both membranes could be considered to be nearly neutral over the entire pH range tested. 7346
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 23, 2006
ESPA3 was positively charged at pH values less than 5, likely due to protonation of carboxyl and amine groups. At pH values greater than 5,ESPA3 was negatively charged likely due to deprotonation of these groups. Finally, SG was negatively charged over the entire pH range. At pH ∼4, where most of the rejection runs were performed, the zeta potentials of the four membranes were as follows: ESPA3 (15 mV) > BW30 ≈ LFC3 (3mV) > SG (-12 mV (Figure 2b). 3.3 PFOS Rejection by RO Membranes. Figure 3a illustrates the percentage of PFOS passing through 4 membrane types versus PFOS concentration in the feed solution. Both permeate and tank samples were taken at 24 h after the feed concentration was adjusted. Additional measurement on feed tank concentration showed no significant changes (1 ppm, rejection exceeded 99% for all 4 membrane types. Rejection increased at higher PFOS concentrations, except at a feed concentration of 1600 ppm when 5% isopropyl alcohol was also present. A three-log reduction in concentration was achieved for BW30 and LFC3 membranes at concentrations 100 ppm and above. A similar effect of solute concentration has been noted for organic compounds such as diclofenac, phenacetine, and primidone (13). In general, surfactants form micelles at high concentration. For PFOS, the critical micelle concentration (CMC) in water is greater than 1 mM (1.13 mM based on electrical conductance measurements, and 1.05 mM based on surface tension measurements (18)). Thus, only the highest aqueous concentration tested in the current study (500 ppm, equivalent to 1 mM) approaches the CMC. Previous studies have emphasized two rejection mechanisms for organic molecules by RO membranes: electrostatic repulsion and size exclusion. Electrostatic repulsion of some negatively charged organic molecules has correlated with enhanced rejection by negatively charged membranes
FIGURE 3. PFOS passage for RO membranes. (a) PFOS passage versus feed tank concentrations. Permeate samples were taken at 24 h after the feed concentration was adjusted. The feed with 1600 ppm PFOS included 5% isopropyl alcohol by volume. (b) PFOS passage versus sodium chloride passage. For BW30 membrane where three replicates were performed, the mean values and range (error bar) are indicated. (13). This trend was not observed for PFOS rejectionsthe order of solute rejection by the membranes was BW30 ≈ LFC3 > SG > ESPA3 (Figure 3a). This sequence differs from the sequence based on zeta potential (ESPA3 > BW30 ≈ LFC3 > SG at pH 4, as shown in Figure 2b). We conclude that zeta potential of the virgin membranes is not a useful predictor of PFOS rejection. Zeta potential values of fouled membranes might be a better predictor of solute rejection. In this particular case, zeta potentials of PFOS fouled membranes were essentially identical (unpublished data). Rejection by size exclusion was previously reported for rejection of the natural hormone estrone by tight RO membranes (14). This mechanism is likely important for PFOS given that PFOS molecules are reasonably large (molecular weight of 500 as compared to 270 for estrone) and rigid due to the strong C-F bonds and lack of rotation at the carbon-carbon bonds. Indeed, it is somewhat surprising to see that PFOS, a sizable molecule, was able to pass through the RO membranes. One possible mechanism might be that PFOS molecules first dissolve into the membrane and subsequently diffuse through the polyamide layer, in a way similar to the passage of hormones (19). Alternatively, passage of PFOS might be explained by its molecular shape. PFOS is a highly linear molecule with a smaller cross section compared to organic molecules with aromatic or aliphatic rings. This is consistent with the previous finding that passage of organic molecules through membranes might be determined by their minimum molecular dimension (12). Finally, Figure 3b shows a plot of PFOS rejection against sodium chloride rejection for the four membranes at various PFOS concentrations. PFOS rejection agrees reasonably well with sodium chloride rejection and better PFOS rejection was observed for tighter RO membranes. 3.4 Permeate Flux. No significant flux reduction was observed with feed PFOS concentrations of 2.5 ppm and less. Figure 4a illustrates flux changes over time for the three membranes tested (ESPA3, BW30, and LFC3) at feed conVOL. 40, NO. 23, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7347
FIGURE 4. Membrane flux performance: (a) flux performance at various feed tank concentrations; (b) stable flux at various PFOS concentrations plotted against initial pure water flux. The initial water flux was measured after 48 h pre-compaction with MilliQ water, and the stable flux was measured at 24 h after PFOS concentration was adjusted to a particular level (except 48 h used for 500 ppm PFOS). Three replicates were tested for BW30, error bars (based on the range) were about the size of the symbols. centrations ranging from 10 to 500 ppm. Flux decreased as PFOS concentration increased, although reasonable flux was still achievable at 500 ppm PFOS when the feed solution was nearly saturated (solubility of PFOS is about 570 mg/L (2)). The flux decline is probably associated with the entrapment of PFOS molecules in the polyamide layer and their accumulation on the membrane surfaces. The high flux membrane ESPA3 appears to be more sensitive to feed concentration than the other membranes. While the flux reduction for ESPA3 was 10% at a concentration of 10 ppm, it increased to 60% at 500 ppm PFOS. For BW30 and LFC3, significant flux reduction (>10%) occurred only at concentrations greater than 100 ppm. The flux reductions at 500 ppm were 40% and 16% for BW30 and LFC3, respectively. Although greater flux reduction (as a percentage) was observed for ESPA3, its absolute flux was still greater than that for BW30 and LFC3 at PFOS concentrations up to 250 ppm. At 500 ppm, however, all three membranes reached almost identical fluxes. The flux performance at 500 ppm PFOS appeared to be independent of the virgin membrane properties. ESPA3’s greater flux reduction as PFOS concentration increased is likely due to its higher initial flux compared to the other membranes. Similar dependence has been reported by many researchers (20, 21). The effect of initial flux is also shown in Figure B of the Supporting Information for the membrane BW30. At a low initial flux of 0.6 m/day, no flux reduction was noticed. On the other hand, a 9% reduction occurred when the initial flux was doubled. The stable flux for each membrane at various PFOS concentrations is plotted against the initial pure water flux in Figure 4b). The lines for 0%, 10%, and 50% flux reductions are also shown. The extent of flux reduction (fouling) was highly dependent upon the initial fluxshigher initial flux values led to greater flux reduction at any given PFOS 7348
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 40, NO. 23, 2006
concentration. This fouling pattern has been well documented for other contaminants (20, 21). In addition, the amount of flux reduction also depends on the PFOS concentration in the feed. At low concentrations, the reduction in flux was minimal, such that the stable flux was essentially determined by intrinsic membrane propertiess the pure water flux. Such dependence on membrane properties became progressively weaker as the concentration of PFOS in the feed increased. At concentrations of 250 and 500 ppm, the initial pure water flux seemed to have little effect on the stable flux. From an operational point of view, this indicates that high-flux RO membranes should be avoided when treating high concentrations of PFOS as any initial high flux exceeding the stable flux would not be sustainable, and the rejection of high-flux membranes is less than that achieved using tighter membranes. On the other hand, highflux membranes would perform reasonably well when treating low strength PFOS solutions, providing around 99% rejection efficiency while maintaining higher stable fluxes than tighter membranes. The flux performance at 1600 ppm PFOS with 5% isopropyl alcohol is shown in Figure C(a) of the Supporting Information. The addition of PFOS and isopropyl alcohol led to a drastic immediate flux reduction of about 90% for all the membranes, and flux became nonmeasurable 12 h later. Similar results were observed for a feed solution containing 5% isopropyl alcohol and no PFOS, where a 90% immediate flux drop was observed. Isopropyl alcohol, not PFOS, seems to cause a severe reduction in flux probably due to the increased osmotic pressure. Flux reduction for a range of isopropyl alcohol concentrations (0-5%) is presented in Figure C(b) of the Supporting Information. The flux reduction was more severe at higher isopropyl alcohol concentrations. It appears that application of membrane technology to remove PFOS may require an alcohol removal pretreatment step. Our results indicate that PFOS can be efficiently rejected by commercially available RO membranes over a wide range of feed concentrations from 0.5 to 1600 ppm. With a suitable membrane, a three-log reduction in concentration is achievable while maintaining stable flux. Multi-stage membrane arrays could be designed to further increase removal efficiency. Consequently, membrane separation appears to be a promising and effective technology for removal of PFOS from PFOS-containing semiconductor wastewaters.
Acknowledgments This research was partially supported by the Clean Water Programme, Nanyang Technological University (NTU), Singapore. C.Y.T. was funded by an NTU Overseas Scholarship. We are thankful for SEMATECH for providing wastewater samples and PFOS standard. We also thank Hydranautics and Dow FilmTec for providing free membrane samples used in this study.
Supporting Information Available Figure A, Membrane test setup; Figure B, effect of initial flux on flux decline; Figure C, effect of isopropyl alcohol on flux. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited (1) Kissa, E. Fluorinated Surfactants: Synthesis, Properties, and Applications; Marcel Dekker: New York, 1994. (2) Brooke, D.; Footitt, A.; Nwaogu, T. A. Environmental Risk Evaluation Report: Perfluorooctanesulphonate (PFOS); Building Research Establishment Ltd., Risk and Policy Analysts Ltd., and UK Environment Agency’s Science Group, 2004. (3) Personal communication; Walter Worth, SEMATECH, 2005. (4) Giesy, J. P.; Kannan, K. Global distribution of perfluorooctane sulfonate in wildlife. Environ. Sci. Technol. 2001, 35 (7), 1339.
(5) Schultz, M. M.; Barofsky, D. F.; Field, J. A. Fluorinated alkyl surfactants. Environ. Eng. Sci. 2003, 20 (5), 487. (6) Renner, R. Growing concern over perfluorinated chemicals. Environ. Sci. Technol. 2001, 35 (7), 154A. (7) Fu, Q. S.; Rhoads, K. R.; Higgins, C. P.; Luthy, R. G.; Criddle, C. S. Microbial transformations of perfluorinated organics. Abstr. Papers Am. Chem. Soc. 2004, 228 (Part 1), U609. (8) Higgins, C. P.; Field, J. A.; Criddle, C. S.; Luthy, R. G. Quantitative determination of perfluorochemicals in sediments and domestic sludge. Environ. Sci. Technol. 2005, 39 (11), 3946. (9) Schultz, M. M.; Barofsky, D. F.; Field, J. A. Quantitative determination of fluorinated alkyl substances by large-volumeinjection liquid chromatography tandem mass spectrometry characterization of municipal wastewaters. Environ. Sci. Technol. 2006, 40 (1), 289. (10) Key, B. D.; Howell, R. D.; Criddle, C. S. Defluorination of organofluorine sulfur compounds by Pseudomonas sp. strain D2. Environ. Sci. Technol. 1998, 32 (15), 2283. (11) DEPRA. Perfluorooctane Sulfonate. Risk Reduction Strategy and Analysis of Advantages and Drawbacks; Prepared by RPA in Association with BRE for Department for Environment, Food and Rural Affairs and the Environment Agency for England and Wales: London, 2004. (12) Kiso, Y.; Sugiura, Y.; Kitao, T.; Nishimura, K. Effects of hydrophobicity and molecular size on rejection of aromatic pesticides with nanofiltration membranes. J. Membr. Sci. 2001, 192 (1-2), 1. (13) Kimura, K.; Amy, G.; Drewes, J. E.; Heberer, T.; Kim, T. U.; Watanabe, Y. Rejection of organic micropollutants (disinfection by-products, endocrine disrupting compounds, and pharmaceutically active compounds) by NF/RO membranes. J. Membr.
Sci. 2003, 227 (1-2), 113. (14) 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 (1), 182. (15) 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 (2), 253. (16) Walker, S. L.; Bhattacharjee, S.; Hoek, E. M. V.; Elimelech, M. A novel asymmetric clamping cell for measuring streaming potential of flat surfaces. Langmuir 2002, 18 (6), 2193. (17) Fa, K. Q. The significance of electrokinetic characterization for interpreting interfacial phenomena at planar, macroscopic interfaces. Phys. Chem. Chem. Phys. 2005, 7 (4), 678. (18) Gente, G.; La Mesa, C.; Muzzalupo, R.; Ranieri, G. A. Micelle formation and phase equilibria in a water-trifluoroethanolfluorocarbon surfactant system. Langmuir 2000, 16 (21), 7914. (19) Nghiem, L. D.; Schafer, A. I.; Elimelech, M. Removal of natural hormones by nanofiltration membranes: Measurement, modeling, and mechanisms. Environ. Sci. Technol. 2004, 38 (6), 1888. (20) Kim, K. J.; Fane, A. G.; Fell, C. J. D.; Joy, D. C. Fouling mechanisms of membranes during protein ultrafiltration. J. Membr. Sci. 1992, 68 (1-2), 79. (21) Hong, S. K.; Elimelech, M. Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes. J. Membr. Sci. 1997, 132 (2), 159.
Received for review April 6, 2006. Revised manuscript received August 1, 2006. Accepted August 21, 2006. ES060831Q
VOL. 40, NO. 23, 2006 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
7349
