Environ. Sci. Technol. 2008, 42, 5292–5297
Nanofiltration for Trace Organic Contaminant Removal: Structure, Solution, and Membrane Fouling Effects on the Rejection of Perfluorochemicals EVA STEINLE-DARLING AND MARTIN REINHARD* Civil and Environmental Engineering, Stanford University, Yang & Yamazaki E & E Building, 473 Via Ortega, Stanford, California 94305
Received December 20, 2007. Revised manuscript received March 31, 2008. Accepted April 2, 2008.
The use of nanofiltration (NF) membranes for water recycling requires an improved understanding of the factors that govern rejection of potentially harmful organic trace contaminants. Rejections of 15 perfluorochemicals (PFCs)s5 perfluorinated sulfonates, 9 perfluorinated carboxylates, and perfluorooctane sulfonamide (FOSA)sby four nanofiltration membranes (NF270, NF200, DK, and DL) were measured. Rejections for anionic species were >95% for MW > 300 g/mol. FOSA (MW ) 499 g/mol), which is uncharged at the pH of deionized water (5.6), was rejected as little as 42% (DL membrane). Decreasing the pH to less than 3 decreases rejection by up to 35%, effectively increasing the MWCO of NF270 by >200 g/mol, while a 2500 mg/L NaCl equivalent increase in ionic strength reduces rejections 99%) of perfluorooctane sulfonate (PFOS) by RO membranes and 90-99% removal by NF membranes at concentrations relevant to semiconductor wastewater (mg/L). The flux decline reported in these previous studies indicates some amount of PFOS sorption by the membrane. To ensure safe use of recycled water, it is imperative to know not only the specific rate of rejection for the particular contaminants studied here at concentrations observed in practice (19), but also to understand the factors that govern contaminant rejection in general. The objective of this paper is therefore to quantify the importance of membrane type, solute size, pH, ionic strength, sorption, and the existence of a fouling layer on the rejection of 15 PFCs by four NF membranes. Finally, rejection and sorption data are used to 10.1021/es703207s CCC: $40.75
2008 American Chemical Society
Published on Web 06/10/2008
TABLE 1. PFCs Used in This Study with Abbreviations, Chemical Formulas, Molecular Weights, and pKa Values full name
abbreviation
formula
MW
pKaa
perfluoropentanoate perfluorobutane sulfonate perfluorohexanoate perfluoroheptanoate perfluorohexane sulfonate perfluorooctanoate 1H,1H,2H,2H-perfluorooctane sulfonate perfluorononanoate perfluorooctane sulfonate perfluorooctane sulfonamide perfluorodecanoate perfluoroundecanoate perfluorodecane sulfonate perfluorododecanoate perfluorotetradecanoate
PFPnA PFBS PFHxA PFHpA PFHxS PFOA 6:2 FtS PFNA PFOS FOSA PFDA PFUnA PFDS PFDoA PFTA
CF3-(CF2)3-COOCF3-(CF2)3-SO3CF3-(CF2)4-COOCF3-(CF2)5-COOCF3-(CF2)5-SO3CF3-(CF2)6-COOCF3-(CF2)5-(CH2)2-SO3CF3-(CF2)7-COOCF3-(CF2)7-SO3CF3-(CF2)7-SO2-NH2 CF3-(CF2)8-COOCF3-(CF2)9-COOCF3-(CF2)9-SO3CF3-(CF2)10-COOCF-(CF2)12-COO-
263 299 313 363 399 413 426 463 499 499 513 563 599 613 713
-0.1 0.14 -0.16 -0.19 0.14 -0.2 0.36 -0.21 0.14 6.52b -0.21 -0.21 0.14 -0.21 -0.21
log Kocc
2.06 2.39 ( 0.09 2.57 ( 0.13 2.76 ( 0.11 3.30 ( 0.11 3.53 ( 0.12
a pKa values were obtained using the SPARC calculator (31), a model whose results have been shown to agree well with experimental data for other fluorinated compounds (32). b pKa given is for the transition between neutral and negative charge on the amide nitrogen. c logKoc values were obtained from Higgins and Luthy (30).
hypothesize the locations of sorption sites for the different types of PFCs.
2. Experimental Section 2.1. Materials. 2.1.1. Membranes. Four types of nanofiltration membranes were used in this study. Most experiments were performed with NF270 (Dow/FilmTec, Minneapolis, MN). For comparison, rejection experiments in deionized water were also performed with NF200 (Dow/FilmTec), DK, and DL (both GE Osmonics, Minnetonka, MN) NF membranes. All of these membranes are piperazine-based polyamide membranes. However, while the FilmTec membranes are composed only of polypiperazine, the GE Osmonics membranes have been modified from this chemistry (26). All the membranes have isoelectric points in the range of pH ) 4-5, as measured via zeta potential (7, 27). Selected rejection parameters for these membranes are listed in Table 2. 2.1.2. Chemicals. All reagents used in this study were of HPLC grade unless otherwise specified. The PFCs used in this study are shown in Table 1. Their sources and purities are given by Higgins et al. (20, 28) and Stevenson et al. (29). They are all linear molecules consisting of perfluorinated backbones ranging from 4 to 14 carbons, and one of three different functional groups (carboxylate, sulfonate, and one sulfonamide). Two additional descriptors, pKa and log Koc are given for those compounds for which the parameters were available. Due to their surface active nature, the traditional measure of hydrophobicity, Kow, has not been determined for these compounds. Instead, Koc, the distribution coefficient for water/sediment systems (normalized for the fraction of organic content in the sediment) is used as a surrogate (30). 2.2. Membrane Test Setup. The membrane apparatus used in this study has been described previously (3). Briefly, the setup consists of three flat sheet membrane cells configured in parallel and operated in a cross-flow configuration. Each cell has an active membrane area of 103 cm2 (8.1 cm wide × 12.7 cm long) and a channel height of 1 mm. No feed spacers were used. Retentate and permeate are recycled to the feed tank. Detailed descriptions of the experimental protocol as well as sample analysis can be found in the Supporting Information. The sample preparation and analysis method employing direct-injection liquid chromatography-tandem mass spectrometry (LC-MS/MS) are described in Plumlee et al. (23), which is based on a method by Higgins et al. (20). The extraction procedure for quantifying the sorbed PFCs is based on a method by Tang et al. (25).
2.3. Fouling Method. Some rejection experiments were conducted under simulated fouling conditions using a method similar to one described previously (3, 33). After the membranes were equilibrated in deionized water for at least 48 h, 60 mg/L alginate, 0.3 mM CaCl 2, and 10 mM NaCl were added to the feed as follows: Alginate (alginic acid sodium salt from brown algae; Sigma-Aldrich, St. Louis, MO) was predissolved in 2 L of 150 mM NaCl (Mallinckrodt-Baker, Phillipsburg, NJ) solution, which was then added to the feed tank. Calcium chloride (Mallinckrodt-Baker) was predissolved in ca. 200 mL of deionized water and added to the feed tank. After a flux decline of >30% was reached, the membrane coupons were temporarily removed from the membrane cells, the system was flushed with deionized water, and the fouled coupons were placed back into the cells. Then the feed tank was spiked with PFCs and the experiment proceeded as described in Section 2.2.
3. Results and Discussion 3.1. Steady-State PFC Rejections in Deionized Water. Figure 1 shows the steady-state rejections of PFCs in deionized water, defining steady-state as the drop in rejection over a 24 h period being less than 1%. The rejections are quite high in general. Nearly all the compounds are rejected to 95% or greater, with two notable exceptions: First, perfluoropentanoic acid (PFPnA), the smallest PFC studied here, is rejected 72% (by NF270; PFPnA was not chromatographically resolved for the experiments with the remaining membranes). This indicates that the size cutoff for compounds of this type is close to PFPnA’s molecular weight of 263 g/mol for this membrane. The second exception to the trend of high rejection is perfluorooctane sulfonamide (FOSA), a larger PFC at 499 g/mol, which is rejected as little as 42% (DL) and as much as 98.5% (NF200). FOSA was the only compound detected in the NF200 permeate. This compound differs from the rest of the set in that it is the only PFC expected to be at least partially uncharged at circumneutral pH levels. Thus, the lack of electrostatic repulsion of FOSA from the negatively charged NF membranes might explain the marked difference in steady-state rejections. While difficult to ascertain from Figure 1, the relative transmissions of most PFCs among the four membranes are consistent with relative descriptors of membrane “tightness”, as shown in Table 2. The charged PFCs are not detected in the permeate of NF200, which is the “tightest” according to salt rejection data. NF270 and DK are “looser” than NF200 and have similar MgSO4 transVOL. 42, NO. 14, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5293
FIGURE 2. Rejection versus molecular weight for the PFCs shown in Table 1 by NF270 at pH ) 2.8 and 5 < pH < 6. Note that with the decreased pH, the effective MWCO shifts upward from ca. 250 to ca. 500 g/mol. Error bars indicate one standard deviation from the mean of 6 samples. FIGURE 1. Steady-state rejection versus molecular weight of the PFCs listed in Table 1 for 4 membrane types in deionized water. Note that values given for NF200 are the detection limits (d.l.), since the permeate concentrations were not quantifiable, except in the case of FOSA. Error bars indicate one standard deviation from the mean of 6 samples (duplicates for each of three cells).
TABLE 2. Comparison of Transmissions for the Membranes (Ordered Tightest to Loosest): Salts versus PFCs
a
MgSO4 transmission NaCl transmission average PFC transmissiond FOSA transmission
NF200
NF270
DK
DL
