Environ. Sci. Technol. 2004, 38, 1124-1132
Ultrafiltration of Nonionic Surfactants and Dissolved Organic Matter MARGIT B. MU ¨ LLER, WOLFGANG FRITZ, ULRICH LANKES, AND FRITZ H. FRIMMEL* Engler-Bunte-Institut, Bereich Wasserchemie, Universita¨t Karlsruhe (TH), Engler-Bunte-Ring 1, 76131 Karlsruhe, Germany
A brownwater sample with a high content of humic substances (HS) was fractionated by multistage ultrafiltration (mst-UF) into five fractions with nominal molecular weights ranging from >30 to 30 kDa fraction and decreased with decreasing MW. To evaluate whether separation mechanisms other than size exclusion were of importance during the fractionation, the behavior of low molecular weight model compounds (MC) with a range of polarities was studied. Recoveries decreased with increasing hydrophobicity of the MC. For selected nonylphenol ethoxylates and 4-nonylphenol the recovery correlated well with the hydrophile-lipophile balance value. The presence of dissolved organic matter (DOM) caused an additional loss of hydrophobic MC, possibly because of sorption of the compounds onto DOM fouling layers. The hydrophilic MC caffeine was recovered almost completely (85-86%) regardless of the DOM content of the model solution. It was concluded that size exclusion was the dominant fractionation mechanism for caffeine, whereas hydrophobic interactions played a major role during the mstUF fractionation of nonpolar contaminants. For a better understanding of the behavior of polyfunctional molecules such as HS, the effect of other physicochemical properties needs to be investigated in further studies.
Introduction Dissolved organic matter (DOM) in natural waters consists to a large extent (40->80%) of aquatic humic substances. Humic substances (HS) are polyfunctional natural compounds that exhibit a wide range of structures and properties (1, 2). Molecular size and molecular weight (MW) are fundamental properties of HS, which are known to affect * Corresponding author phone: +49 721 608 2580; fax: +49 721 699154; e-mail:
[email protected]. 1124
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 4, 2004
their behavior with respect to interactions with metals, organic contaminants, and inorganic colloids (3-5) as well as HS behavior during water treatment operations such as adsorption, flocculation, or ozonation (6). Molecular size fractions of humic substances have been shown to differ in their chemical composition (structure) as well as their physicochemical properties (7-12). Data obtained from chemical and physical characterization of HS size fractions provide useful information for the interpretation of their interactions with other water constituents. Clear differences in MW values between HS of various origins as well as between HS fractions [humic and fulvic acids (HA, FA)] have recently been reported by Perminova et al. (13), who statistically evaluated MW data determined by size exclusion (SEC) measurements for a large set of humic substances. The fact that differences in MW can be observed for various HS implies that a separation of HS based on molecular size should be possible. Ultrafiltration (UF) is a technique that has been widely used for the fractionation of HS on the basis of molecular size differences (8, 10, 14-16). Early on, it became, however, obvious that separation according to molecular size is not the only fractionation mechanism during UF of the polyfunctional, negatively charged HS solutions (14). Effects that have to be considered during the UF of DOM are hydrophobic as well as electrostatic interactions between the solutes and the membrane (adsorption, repulsion) and the formation of fouling layers, which may change the molecular weight cutoff (MWCO) of the membrane and cause loss of DOM material (14, 17, 18). Membrane fouling caused by HS is of particular importance during technical applications of UF as well as nanofiltration in water treatment (19). To relate the behavior of humic substances during ultrafiltration to their composition, HS fractions of different polarities and acidities have been produced and subjected to UF (20). An alternative and more straightforward approach is to study the behavior of model compounds with defined physicochemical properties. The behavior of organic molecules in aqueous solutions depends on, in addition to other parameters such as molecular weight, conformation, and charge, largely their polarity. We therefore investigated the influence of solute polarity on the separation of organic compounds by ultrafiltration. To avoid size exclusion effects, we selected model compounds (MC) with molecular weights far below the MWCO of the UF membranes. Nonylphenol ethoxylates (NPEO), 4-nonylphenol (4-NP), and caffeine were used as MC (Figure 1). NPEO are nonionic surfactants that are mostly used for industrial purposes [cleaning agents, emulsifiers, etc. (21)]. Metabolites of NPEO such as monoand diethoxylates (NP1EO, NP2EO), mono- and dicarboxylates (NP1EC, NP2EC), and 4-NP have been shown to cause estrogenic effects in different test systems (22, 23). The technical surfactants are mixtures of NPEO with various degrees of alkyl chain branching and ethoxylate (EO) chain length. NPEO with more than six ethoxy units have hydrophile-lipophile balance (HLB) values >10 and are thus classified as hydrophilic molecules. The HLB value has been introduced by Griffin (24) and describes the ratio of hydrophilic to hydrophobic structures in nonionic surfactants. The HLB value can be determined experimentally or calculated according to
HLB ) mrel/5
(1)
where mrel is the mass proportion of the hydrophilic group relative to the molar mass M of the molecule. The polarity 10.1021/es0300416 CCC: $27.50
2004 American Chemical Society Published on Web 01/16/2004
TABLE 1. Characterization of the Brownwater DOM Sample, Concentrate, and mst-UF Fractions sample original concentrate F1 F2 F3 F4 F5
DOCa (mg/L)
% Kb SUVA254c SUVA436c DOC (µS/cm) pH L/(m*mg) L/(m*mg)
22.1 ( 0.1 100 104 ( 9 91 208 ( 14 63 48.3 ( 0.2 10 51.3 ( 0.4 11 8.28 ( 0.16 2 2.39 ( 0.06 9
32 76 158 65 52 24 8
4.5 3.9 3.5 3.8 4.2 5.6 5.9
4.77 4.72 6.64 4.60 4.18 3.06 2.60
0.42 0.39 0.63 0.28 0.21 0.14 0.09
a Mean (standard deviation (n ) 3)). b κ ) electrical conductivity (at 25 °C). c Specific UV absorbance at wavelength λ.
FIGURE 1. Structures of the model compounds 4-nonylphenol (4NP), nonylphenol ethoxylates (NPEO), and caffeine. n ) number of ethoxy units per molecule. of NPEO decreases with decreasing length of the EO chain. This is reflected in both HLB values and octanol-water partitioning coefficents (log P values). Log P values for longchain NPEO to our knowledge have not been reported so far. 4-NP, NP1EO, and NP2EO have log P values of 4.48, 4.17, and 4.21, respectively (25). Caffeine was selected as MC because it has a molecular weight similar to that of 4-NP (194.20 vs 220.24 g/mol), but with a log P of -0.07 it is much more polar than the latter (26). The aims of this study were as follows: (i) evaluation of the separation efficiency and recovery of the multistage ultrafiltration (mst-UF) procedure; (ii) fractionation of a brownwater DOM sample that had been investigated in previous studies (11, 16, 27) and structural characterization of mst-UF fractions by 13C magic angle spinning nuclear magnetic resonance (MAS NMR) spectroscopy; and (iii) fractionation of model compounds in the absence and presence of DOM.
Experimental Section Chemicals. Two technical nonylphenol ethoxylate surfactants were obtained from Sasol Germany GmbH. The surfactant Marlophen NP10 (average of 10 EO units) consists of homologues with 1-20 EO units, and Marlophen NP3 (average of 3 EO units) contains homologues with 0-6 EO units. 4-NP (technical grade; 85% para-isomers) and caffeine (> 99%) were from Fluka. For spiking of the model solutions, two different stock solutions were prepared. The NPEO stock solution consisted of 1 g/L Marlophen NP10 (total ethoxylate concentration), 1 g/L Marlophen NP3, and 0.5 g/L 4-NP in methanol (HPLC grade). This mixture is further referred to as “NPEO mixture”. The caffeine stock solution was prepared at a concentration of 100 mg/L in methanol. New stock solutions were prepared every 3 months and stored at 4 °C in the dark until use. All solvents used were from Merck and of LiChroSolv grade. Ammonia solution (25% in water; Suprapur) and paraffin (highly liquid) were also from Merck. DOM Samples. The brownwater (BW) sample was obtained from a bog lake in the northern Black Forest of Germany. Due to its high content of dissolved organic matter and low content of inorganic salts, it was suitable for mst-UF fractionation with subsequent NMR analysis. Humic substances make up 55-66% of the BW DOC. Fulvic acids account for 32-45%, and HA for ∼20% of the DOC (28). The bog lake has no in-flowing rivers and receives its water primarily from precipitation. The BW was therefore expected to be free from NPEO surfactants as well as from caffeine, which are known to be released to the aquatic environment mainly via sewage treatment plant effluents. Three hundred
liters of brownwater was collected in July 2001. The BW sample was filtered through a 5 µm glass fiber filter and a 0.45 µm cellulose nitrate filter (both Sartorius, Germany) before mst-UF fractionation and stored at 10 °C in the dark until further use. The filtered sample had a DOC concentration of 22.1 ( 0.1 mg/L (mean ( SD; n ) 3) and a pH value of 4.5 (Table 1). For experiments with model compounds, BW (diluted to a DOC concentration of 10 mg/L) as well as demineralized water and tap water were used as DOM matrixes. The demineralized water had a DOC concentration of 0.0769 ( 0.0004 mg/L and was adjusted with dilute hydrochloric acid to the pH of the brownwater sample (pH 4.5) prior to use. The tap water had a DOC of 0.647 ( 0.009 mg/L. Fractionation of BW DOM by mst-UF. All membrane filtrations were performed using stirred cells (Millipore) and regenerated cellulose UF membranes (type YM, Millipore). Filtrations were carried out in the concentration mode. To obtain amounts of organic carbon sufficient for NMR analysis, the BW sample was preconcentrated before mst-UF fractionation using a 1 kDa (1000 g/mol) UF membrane. The permeate of the concentration step was collected as fraction 5 (F5; DOM < 1 kDa). For the mst-UF fractionation of the concentrate, UF membranes with nominal MWCO of 30, 10, and 3 kDa were used. Fractionation was carried out serially from high to low MWCO as described in ref 16. The following fractions were obtained: fraction 1 (F1), 30 kDa < DOM < 0.45 µm; fraction 2 (F2) 10 kDa < DOM < 30 kDa; fraction 3 (F3) 3 kDa < DOM < 10 kDa; and fraction 4 (F4) 1 kDa < DOM < 3 kDa. To evaluate the reproducibility of the procedure, three replicates were fractionated in parallel. Details of the fractionation procedure are given in the Supporting Information. DOM Characterization. Values for pH and electrical conductivity (at 25 °C) were determined using WTW electrodes pH 325 and LF 318, respectively. Absorbances at λ ) 254 and 436 nm were measured with a Varian Cary 50 spectrophotometer (Varian) using a 10 mm quartz cell. DOC concentrations were determined using a total carbon analyzer TOC 5000A (Shimadzu). A DOC mass balance was carried out to determine the recovery of the fractionation procedure. To evaluate the separation efficiency of the UF fractionation, the molecular size distribution of all samples was investigated using the SEC system LC-OCD, which was equipped with a TSK HW 50-(S) column and on-line UV absorbance (254 nm) and DOC detection. The system has been described in detail ref 29. A phosphate buffer solution (0.028 mol/L) with a pH of 6.8 was used as the eluent. Samples were diluted to a DOC of 10 mg/L and brought to the same buffer concentration prior to analysis to avoid ionic strength gradients during SEC separation. The SEC column was calibrated using a series of polystyrene sulfonate (PSS) standards with Mp (molecular weight at peak maximum) values of 35.7, 15.8, 6.43, and 4.48 kDa. PSS standards were VOL. 38, NO. 4, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1125
dissolved in eluent at DOC concentrations of ∼5 mg/L and detected using the on-line DOC detector. The exclusion (V0 ) 22.2 mL) and permeation volumes (Vp ) 47 mL) of the column were determined using blue dextran and water, respectively (11). The equation of the calibration curve was log Mp ) -0.059 Ve + 5.8259 (R2 ) 0.9553). Number- and weight-average MW (Mn and Mw) and polydispersity (P) were calculated as described in ref 30. Fractions from replicate mst-UF fractionations were pooled, analyzed using the previously described methods, and then freeze-dried and characterized by 13C MAS NMR spectroscopy and elemental analysis. Because of the low DOC content of fractions F4 and F5 (Table 1), this was possible for the original sample as well as fractions F1-F3 only. Solid state 13C MAS NMR spectra of freeze-dried samples were recorded on a Bruker Avance 400 MHz spectrometer at a 13C resonance frequency of 100.6 MHz (for experimental details see Supporting Information). The number of accumulated scans per sample was (30-45) × 103, resulting in signal-tonoise ratios (S/N) > 30 for carboxyl (∼173 ppm), carbohydrate (∼71 ppm), and branched aliphatic (∼40 ppm) signals. With S/N > 30, the error of the measurement is 15% and were thus significant. Structural assignments of chemical shift regions according to ref 31 were as follows: carbonyl carbon (220 to 190 ppm), carboxyl carbon (190 to 160 ppm), O-/ N-substituted aromatic carbon (160 to 140 ppm), non-/Csubstituted aromatic carbon (140 to 100 ppm), O-alkyl carbon (100 to 50 ppm), and aliphatic carbon (50 to 0 ppm). The total integral (220 to 0 ppm) of each spectrum was set to 100% (normalization). Relative carbon contents of the different functional groups were calculated by determining the signal areas of individual chemical shift regions. Elemental analysis (C, H, and N content) was conducted in duplicate after drying of the freeze-dried fractions at 42 °C and 0.107 kPa until a constant weight was achieved. Fractionation of Model Compounds by mst-UF. Fractionation of model compounds was performed in a similar manner as described above. The concentration step in this case was not required because no NMR analysis was made and was thus omitted. This was considered to be permissible due to the fact that the concentrate contained 91% of the DOC of the BW (Table 1). Each model solution was fractionated in triplicate. Model solutions of caffeine at a concentration of 100 µg/L in demineralized water, tap water, and BW were prepared by spiking the appropriate volume of the stock solution into the DOM samples. Similarly, model solutions of the NPEO mixture at concentrations of 1 mg/L of Marlophen NP10 (total ethoxylates), 1 mg/L of Marlophen NP3, and 0.5 mg/L of 4-NP in demineralized water, tap water, and BW were prepared. Six different model solutions were thus obtained and fractionated as described above. Replicate fractions in this case were not pooled but analyzed separately for contents of caffeine and NPEO by HPTLC after solid phase extraction (SPE). After the mst-UF of the model solutions, all UF membranes were collected, extracted with methanol, and analyzed for the presence of model compounds. As low concentrations of the MC were expected, membranes of identical MWCO that had been used for the same model solution were extracted together. The UF membranes were cut into small pieces, transferred to 40 mL glass vials with Teflon caps, and extracted for 4 h with 6 mL of methanol (head-over-head shaker). The methanol was transferred to an 8 mL glass vial and evaporated in a stream of nitrogen at 40 °C. Samples were then dissolved in 250 µL of methanol and analyzed by HPTLC. SPE and HPTLC Analysis of Model Compounds. SPE using Oasis HLB extraction cartridges was performed prior to HPTLC analysis. A concentration factor of 1000 was applied 1126
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 4, 2004
for all samples. Analysis of caffeine, nonylphenol, and the selected ethoxylates was performed using HPTLC analysis with absorbance or fluorescence detection (for details, see Supporting Information). All samples were analyzed in duplicate, and blanks (methanol) and standards were included in every analysis. Plates were developed either manually in twin-through glass chambers or automatically using an automated multiple development (AMD) device (model AMD 2, Camag). Absorbance or fluorescence was measured with a TLC scanner 3 (Camag) in the reflectance mode. Data recording and evaluation were performed using the software CATS (version 4.06, Camag). From the different ethoxylates in the NPEO mixture, three were selected for further analysis. The ethoxylates with 11 (NP11EO), 6 (NP6EO), and 2 (NP2EO) ethoxy units were considered to be representative of the range of polarities of the different homologues present in the NPEO mixture, and their behavior during fractionation was compared to that of the other model compounds 4-NP and caffeine. Comparison of MC was performed on the basis of recoveries and relative concentrations. Recoveries were calculated as mass balances and were used for quantification of losses of the MC during fractionation of a given model solution. Relative concentrations (Frel) were calculated as
Frel in % ) (Ffraction/Foriginal) × 100%
(2)
where Ffraction is the mass concentration of the MC in a particular fraction and Foriginal is the mass concentration of the MC in the original model solution. For comparison of the different model compounds relative concentrations were preferred to actual concentrations because they directly illustrate any increase (Frel > 100%) or decrease (Frel < 100%) in MC concentrations during fractionation, irrespective of the actual concentration in solution (which was different for different model compounds). For caffeine and 4-NP absolute concentrations were determined on the basis of linear regression of peak heights of the five standards. Limits of detection (LOD) and quantification (LOQ) were determined according to the calibration curve procedure described in ref 32. In contrast, due to batch-to-batch variations in the composition of the technical surfactants, the exact ethoxylate distribution in the NPEO mixture was unknown. Concentrations of the selected ethoxylates were therefore expressed in terms of total EO concentration instead of giving mass concentrations of individual ethoxylates. This was considered to be appropriate on the basis of the fact that separation of the ethoxylates of interest was possible with the HPTLC method and that for each ethoxylate a constant response factor should be obtained under constant experimental conditions (pH, solvent, etc.). Relative concentrations of the selected ethoxylates were then calculated on the basis of total ethoxylate concentrations.
Results and Discussion Fractionation of BW DOM by mst-UF and DOM Characterization. The precision of the mst-UF procedure is illustrated by the coefficients of variation (CV) of the DOC and UVA254 values of the three replicates, which were better than 7 and 9%, respectively, for fractions F1, F3, and F4. Fraction F2 showed higher CV with 18% for DOC and 13% for UVA254 values (data not shown). For a fractionation procedure involving several manual steps, the observed recoveries were considered to be acceptable. Good precision was also inferred from LC-OCD analysis of the mst-UF fractions, which showed that the molecular size distributions of the DOM in the three replicates of each of the five fractions were practically identical (data not shown). Ninety-one percent of the total DOC was recovered in the concentrate and further fractionated into F1-F4 (Table 1). Fractions F1-F4 contained 86% of the total
DOC. More than 60% of the DOC were recovered in the >30 kDa fraction (F1), whereas fractions F2-F4 contained e10% of the DOC. The overall loss of the mst-UF procedure was 5%. The DOC input (e.g., contamination from UF membranes) of the procedure was very low (∼2 mg of C) when a blank sample (demineralized water) was fractionated. With the YM30 membranes in particular, a rapid flux decline during the first hour of filtration as well as a discoloration of the membranes was observed. Those observations were taken as indicators of the formation of fouling layers, which, however, obviously did not result in any considerable loss in DOC. The SEC chromatograms of the original sample, the concentrate, and the five mst-UF fractions (pooled replicates) are shown in Figure 2a). The chromatograms of the original sample and the concentrate are very similar because the BW DOM was recovered almost quantitatively (91% of DOC) and, furthermore, qualitatively unchanged (similar SUVA values of both samples, Table 1), in the concentrate. The sum curve of the chromatograms of fractions F1-F5 (Figure 2b) was constructed after normalization of peak areas to the % DOC data (Table 1). The similarity between the original and the constructed chromatograms indicates that neither destruction of molecules nor irreversible aggregate formation have taken place to a considerable extent during the fractionation. The relatively broad shape of the SEC chromatograms shows that the mst-UF fractions still contained molecules with a wide range of molecular sizes. This is a consequence of the polydisperse nature of the HS as well as of the pore size distribution of UF membranes and of fractionation in concentration mode. Several differences between the samples are obvious and are both reflected in the SEC chromatograms as well as in the MW data (Table 2). It is stated here that the determination of MW values was not the purpose of this work but that in particular the Mp and P values serve only to describe the results of the SEC analysis in a more quantitative way. Whereas the Mn and Mw values determined in this study are higher than values reported in the literature, the polydispersity data agree well with results by other groups (9, 13, 33, 34). Three distinct peaks were detected in the samples (Figure 2a). Peak 1 was present only in the original and concentrated samples as well as in mstUF fractions with MWCO >10 kDa (F1 and F2). It contained ∼8% of the DOC detected during SEC analysis in the former samples and 30 kDa) to a greater extent than small ones, resulting in relative enrichment of those compounds and a shift in the maximum of the molecular size distribution toward greater molecular weight. With increasing fraction number, that is, decreasing MWCO of the UF membrane, peak 2 shifted toward higher elution volumes and smaller molecular weight. SEC analysis thus demonstrated that the mst-UF procedure is suitable for the production of DOM fractions, which differ in their molecular size distribution. The polydispersity progressively decreased with an increasing number of fractionation steps. Whereas the original sample had a polydispersity value of 1.88, considerably lower values were determined for F3 and F4 (P ) 1.39 and 1.34, respectively). On the basis of earlier investigations, Swift (35) stated that the polydispersity of HS
FIGURE 2. (a) SEC chromatograms of the brownwater sample (original), concentrate, and mst-UF fractions F1-F5. DOC detection; Ve ) elution volume. (b) SEC chromatograms of original sample, sum of fractions F1-F5 (calculated), and difference (original sum; calculated). Signal areas normalized to percent DOC data (Table 1). samples can be reduced substantially by combining several increasingly fine fractionation steps in one procedure. Our results confirm this statement. Several observations which mostly apply to the F1 fraction, however, indicate that the mst-UF fractionation was not solely based on molecular size differences. The DOC recovery in fraction F1 (30 kDa membrane) with 63% is rather high and in a range that has been reported for UF membranes with a MWCO of 1 kDa (36). The Mp value for peak 2 in F1 is well below 30 kDa, which suggests that the corresponding DOM molecules should not have been retained by the YM30 membrane. It has already been noted that the apparent retention of molecules of small MW is also a consequence of filtration in the concentration mode, but other factors VOL. 38, NO. 4, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1127
TABLE 2. Elution Volumes and Molecular Weight Data Obtained from Analytical SEC Measurements peak 1
peak 2
peak 3
sample
Vea
Mpb
Ve
Mp
Ve
original concentrate F1 F2 F3 F4 F5
21.5 21.4 21.8 21.5 nd nd nd
36100 36600 34700 36100
31.8 31.8 30.4 33.4 34.2 36.5 36.5
8900 8800 10800 7200 6400 4700 4700
nd nd nd nd nd 40.0 40.0
total sample
Mp
Mnc
Mwd
Pe
2900 2900
5600 5900 6700 4900 4100 3200 2900
10600 11100 11900 7400 5700 4300 5000
1.89 1.88 1.78 1.51 1.39 1.34 1.72
a V ) elution volume at peak maximum (mL). b M ) molecular weight at V . c M ) number-average molecular weight. e p e n molecular weight (PSS calibration) in Da. e P ) polydispersity. f Not detected. For peak assignment see Figure 1.
d
Mw ) weight-average
TABLE 3. 13C MAS NMR and Elemental Analysis Data for the Brownwater DOM Sample and mst-UF Fractions F1, F2, and F3. Values for the Loss (DOM in F4 and F5, and DOM lost during fractionation) Are Calculated Values relative carbon content of chemical shift regions (TMS scale) aromatic C
elemental analysis
sample
carbonyl C 220-190 ppm
carboxyl C 190-160 ppm
O/N-subst 160 -140 ppm
non/C-subst 140-100 ppm
O-alkyl C 100-50 ppm
aliphatic C 50-0 ppm
H/C ratio
N/C ratio
original F1 F2 F3 loss
4.1 4.2 3.6 5.4 3.4
17.5 17.7 18.2 20.0 14.8
8.4 9.3 8.1 7.3 5.6
20.3 21.7 20.5 19.7 14.6
29.7 29.0 28.8 27.9 34.4
19.9 18.0 20.7 19.7 27.2
0.974 0.926 0.918 0.874 1.246
0.028 0.021 0.019 0.021 0.061
have to be considered as well. A simple explanation would be to conclude that polystyrene sulfonates do not give reliable MW data for humic substances. They have, however, been reported to yield values that agree well with data from, for example, vapor pressure osmometry (33). The presence of fouling layers (pore fouling) could have resulted in a change of the MWCO distribution of the YM30 membrane toward smaller pore sizes, thereby leading to an overestimation of the large (>30 kDa) molecular size proportion of the DOM. Negatively charged DOM fouling layers could have resulted in electrostatic repulsion of DOM. Aggregation of DOM molecules at high DOC concentrations (Table 1) could also have caused an increased DOC rejection. DOM aggregates have not been not observed during SEC analysis of the fractions after dilution to a DOC of 10 mg/L and can thus only have been of reversible nature. Reversible aggregate formation has been reported to occur for a peat HA (37) and can therefore not be excluded in this work. At present it is, however, not possible to determine which of the discussed mechanisms are of actual relevance for fraction F1. Results from 13C MAS NMR analysis are shown in Figure 3 and Table 3, respectively. Structural differences between the analyzed samples were most pronounced in the O-alkyl (100 to 50 ppm), carboxyl carbon (190 to 160 ppm), and aliphatic (50 to 0 ppm) chemical shift regions. Within the O-alkyl structures, the relative contents of carbohydrates (∼71 ppm) progressively decreased from the original sample toward mst-UF fractions of smaller MW. Carbohydrates have been reported to be relatively enriched in both large MW fractions as well as in fouling layers after membrane filtration of DOM (19, 38). The presence of large MW carbohydrates in fouling layers is a consequence of their small diffusion coefficients, which makes them more susceptible to concentration polarization effects during UF. It is assumed that carbohydrates have also partially been deposited with fouling layers in this work. Relative contents of carboxyl groups increased with decreasing nominal molecular size and were highest in fraction F3 (3-10 kDa). The integration data for aliphatic carbon (50 to 0 ppm) showed no clear trend regarding their distribution over UF fractions. As in the O-alkyl region, individual signals have to be interpreted. The signal 1128
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 4, 2004
FIGURE 3. 13C MAS NMR spectra of the brownwater sample (original) and mst-UF fractions F1, F2, and F3. intensity of methyl groups (23 ppm) decreased with decreasing MW. Those functional groups can be found, for example, in N-acetylated carbohydrates, and their content is thus related to the content of carbohydrates. Methylene groups (30 ppm) were significantly more prominent in the original sample as well as in F1, whereas branched alkyls (40 ppm) were most abundant in F3. For the BW DOM branched alkyls, which have also been detected in the permeate of a 10 kDa UF membrane in previous studies, are ascribed to terpenoid structures (38). The results agree well with literature data, where similar trends regarding the distribution of carbohydrates and methylene groups (decreasing relative contents with decreasing MW), as well as branched aliphatic and carboxyl structures (increasing relative contents with decreasing MW) over size fractions of HS have been reported (7, 9, 10, 38). This suggests that fulvic acids are enriched in the DOM of F3 on the basis of the facts that they are generally smaller than HA, have lower SUVA254 values, and contain more carboxylic functions per molecule (28). Elemental analysis (EA) data are reported in Table 3. Due to small amounts of sample, only C, H, and N contents could be determined. More detailed EA data of the BW sample are
FIGURE 4. HPTLC chromatograms of (a) 100 ng of caffeine standard and (b) caffeine in brownwater fraction F1. Absorbance detection was at 275 nm. Peaks: a, brownwater DOM; b, caffeine; c, nonretarded compounds (dashed line indicates baseline position. HPTLC chromatograms of (c) 12 µg of NPEO standard (total EO concentration) and (d) NPEO in brownwater fraction F1. Fluorescence detection was at 227 nm (cutoff filter 280 nm). Peak numbers indicate number of ethoxylate units per molecule.
TABLE 4. Recoveries (Percent; Mean ( Standard Deviation, n) 3) of Model Compounds after Solid Phase Extraction compound
tap water
brownwater
4-NP NP2EO NP6EO NP11EO caffeine
80 ( 9 100 ( 7 100 ( 6 100 ( 3 100 ( 10
90 ( 4 90 ( 3 100 ( 4 100 ( 2 100 ( 3
given in ref 28. The BW has a very low ash content, which is in the range of 4-5% for the original sample and even lower for a UF concentrate (3.5%; concentrate of a 4 kDa membrane). Similar values can be assumed for the sample that was investigated in this study and, consequently, H/C and N/C data are considered to be reliable. The oxygen content of the BW DOM is ∼40% (28). Elemental analysis data agree well with NMR data. The H/C ratio decreased from the original sample toward fractions of decreasing MW (∼10% decrease from original sample to F3). This is a consequence of a lower relative content of both carbohydrates and methylene and methyl structures in fractions F2 and F3. The N/C ratio decreased by 25% from the original sample to F3. By 15N NMR analysis of the BW DOM (except from the fulvic acids) it has been shown that nitrogen is present mainly in amide structures such as N-acetylic carbohydrates or peptides (39). The term “loss” in Table 3 refers to the DOM that was not analyzed by NMR and EA, that is, DOM in fractions F4 and F5 (DOM < 3 kDa; 11% of DOC) as well as lost material (5% of DOC). NMR and EA data for this part of the DOM were calculated on the basis of mass balance assumptions and
using the % DOC data from Table 1. Both NMR and EA data indicate that this part of the DOM is more aliphatic than the original sample as well as fractions F1-F3. O-alkyl and nitrogen-containing structures are more abundant in this collective DOM fraction, whereas relative carboxyl as well as aromatic carbon contents must be lower than in the original sample. With this estimation it is not possible to assign the various structures to particular MW fractions (mst-UF fractions) or defined Ve regions (SEC chromatograms). This can be achieved only by analyzing the respective fractions and is part of another study. Specific absorbances of the mst-UF fractions ranged from 2.60 to 6.64 L/(m*mg) for SUVA254 and from 0.09 to 0.63 L/(m*mg) for SUVA436 (Table 1). Both values were highest in fraction F1 and decreased with decreasing molecular size of the mst-UF fractions. A correlation between SUVA254 and the MW of HS has also been reported by other authors (8, 33, 40). The high SUVA254 value of 6.64 L/(m*mg) of fraction F1 is in the upper range of data reported for aquatic HS. An increasing SUVA254 correlates with an increasing content of UV-absorbing aromatic and other unsaturated structures (41). The differences in the SUVA254 values are more pronounced than those of the NMR data for the aromatic carbon region because of the greater sensitivity of this parameter. The brownwater sample contains various dissociable groups such as carboxyl, phenolic, and alcoholic hydroxy as well as amino groups (38, 41). Due to the low concentration of inorganic electrolytes (28), the pH and electrical conductivity values correlate with the DOC content of the mst-UF fractions (Table 1). At pH values 85% in all three model solutions. Our experiments demonstrate that the polarity of a molecule is an important property which greatly affects its behavior during mst-UF fractionation. This is of relevance when UF is used for the fractionation of DOM. Although it has been shown in the present work that the BW DOM could be separated into fractions with different Mp and polydispersities using the mst-UF procedure, non-size separation mechanisms could not be completely excluded. HS in general, similarly to the BW sample investigated in this study, are known to contain both hydrophilic and hydrophobic structures. In contrast to the model compounds used in this work, humic substances generally carry negative charges in aqueous solutions. Therefore, a similar study using MC of various structures and charge densities should be conducted to assess the role of other physicochemical properties during UF fractionation. Our results are also of importance for the use of membrane filtration in water treatment, because uncontrolled adsorption of hydrophobic compounds onto membranes and other solid surfaces will result in uncontrolled and, therefore, unreliable removal of, for example, hydrophobic pollutants. By carrying out further experiments with a greater number of model compounds, the observed correlation between MC recovery and polarity may be used to predict the behavior of different compounds during membrane filtration. Finally, for NPEO and 4-NP in particular, treatment operations based on adsorption should be more successful for the removal of those compounds than membrane filtration.
Acknowledgments This work was financially supported by a research grant obtained from the German Ministry for Education and Research (BMBF, project number 02WU9856/4). We thank Sasol Germany GmbH for providing us with Marlophen samples and HLB data. We are grateful to Dr. Gudrun AbbtBraun for assistance with data interpretation, and to Gabi Kolliopoulos for LC-OCD analysis of DOM samples. Furthermore, we thank the reviewers for their valuable comments to this manuscript.
Supporting Information Available
FIGURE 6. Relative concentrations (Grel) of model compounds in UF fractions of different DOM matrixes: (a) demineralized water; (b) tap water; (c) brownwater. Mean ( SD (n ) 3).
Additional information regarding the experimental procedures for mst-UF fractionation of brownwater DOM, 13C MAS NMR analysis, and SPE and HPTLC analysis of model compounds. This material is available free of charge via the Internet at http://pubs.acs.org.
Literature Cited caused by adsorption onto DOM molecules which were deposited in solid form onto the UF membranes (fouling layers). This explanation is supported by the fact that the greatest differences between the Frel of 4-NP and NP2EO in the three model solutions (Figure 6) were observed for fraction F1, that is, the fraction which had also caused the greatest flux decline and membrane discoloration.
(1) Frimmel, F. H. In Biopolymers; Steinbu ¨ chel, A., Ed.; Wiley: Weinheim, Germany, 2001; pp 301-324. (2) Leenheer, J. A.; Croue´, J.-P. Environ. Sci. Technol. 2003, 37, 18A26A. (3) Chin, Y.-P.; Aiken, G. R.; Danielsen, K. M. Environ. Sci. Technol. 1997, 31, 1630-1635. (4) Specht, C. H.; Kumke, M. U.; Frimmel, F. H. Water Res. 2000, 34, 4063-4069. VOL. 38, NO. 4, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1131
(5) Schmitt, D.; Mu¨ller, M. B.; Frimmel, F. H. Acta Hydrochim. Hydrobiol. 2000, 28, 400-410. (6) Frimmel, F. H.; Abbt-Braun, G.; Hesse, S.; Kleiser, G. In Proceedings of the 9th International Conference of the International Humic Substances Society (IHSS), IHSS Publications: Adelaide, Australia, 2001; pp 453-465. (7) Swift, R. S.; Leonard, R. L.; Newman, R. H.; Theng, B. K. G. Sci. Total Environ. 1992, 117/118, 53-61. (8) Everett, C. R.; Chin, Y.-P.; Aiken, G. R. Limnol. Oceanogr. 1999, 44, 1316-1322. (9) Pelekani, C.; Newcombe, G.; Snoeyink, V. L.; Hepplewhite, C.; Assemi, S.; Beckett, R. Environ. Sci. Technol. 1999, 33, 28072813. (10) Christl, I.; Knicker, H.; Ko¨gel-Knabner, I.; Kretzschmar, R. Eur. J. Soil Sci. 2000, 51, 617-625. (11) Mu ¨ ller, M. B.; Schmitt, D.; Frimmel, F. H. Environ. Sci. Technol. 2000, 34, 4867-4872. (12) Khalaf, M.; Kohl, S. D.; Klumpp, E.; Rice, J. A.; Tomba´cz, E. Environ. Sci. Technol. 2003, 37, 2855-2860. (13) Perminova, I. V.; Frimmel, F. H.; Kudryavtsev, A. V.; Kulikova, N. A.; Abbt-Braun, G.; Hesse, S.; Petrosyan, V. S. Environ Sci. Technol. 2003, 37, 2477-2485. (14) Aiken, G. R. Environ Sci. Technol. 1984, 18, 978-981. (15) Becher, G.; Carlberg, G. E.; Gjessing, E. T.; Honslo, J. K. Environ. Sci. Technol. 1985, 19, 422-426. (16) Mu ¨ ller, M. B.; Frimmel, F. H. Water Res. 2002, 36, 2643-2655. (17) Cho, J.; Amy, G.; Pellegrino, J. J. Membrane Sci. 2000, 164, 89110. (18) Scha¨fer, A. I.; Mauch, R.; Waite, T. D.; Fane, A. G. Environ. Sci. Technol. 2002, 36, 2572-2580. (19) Gorenflo, A. Ru ¨ ckhalt und Fouling von natu ¨ rlichen organischen Substanzen bei der Nano- und Ultrafiltration; Ph.D. thesis, Universita¨t Karlsruhe (TH), 2003. (20) Lin, C.-F.; Huang, Y.-J.; Hao, O. J. Water Res. 1999, 33, 12521264. (21) Thiele, B.; Gu ¨ nther, K.; Schwuger, M. J. Chem. Rev. 1997, 97, 3247-3272. (22) Soto, A. M.; Justicia, H. W.; Wray, J. W.; Sonnenschein, C. Environ. Health Perspect. 1991, 92, 167-173. (23) Routledge, E. J.; Sumpter, J. P. Environ. Toxicol. Chem. 1996, 15, 241-248. (24) Griffin, W. C. J. Soc. Cosmet. Chem. 1949, 1, 311-326. (25) Ahel, M.; Giger, W. Chemosphere 1993, 26, 1461-1470. (26) SRC PhysProp Database: http://esc.syrres.com/interkow/ webpro.exe?CAS)58-08-2. (27) Refractory Organic Substances in the Environment; Frimmel, F. H., Abbt-Braun, G., Heumann, K. G., Hock, B., Lu¨demann, H.D., Spiteller, M., Eds.; Wiley: Weinheim, Germany, 2002.
1132
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 4, 2004
(28) Abbt-Braun, G.; Frimmel, F. H. In Refractory Organic Substances in the Environment; Frimmel, F. H., Abbt-Braun, G., Heumann, K. G., Hock, B., Lu ¨ demann, H.-D., Spiteller, M., Eds.; Wiley: Weinheim, Germany, 2002; pp 1-38. (29) Huber, S.; Frimmel, F. H. Fresenius’ J. Anal. Chem. 1992, 342, 198-200. (30) Wershaw, R. L.; Aiken, G. R. In Humic Substances in Soil, Sediment, and Water; Aiken, G. R., McKnight, D. M., Wershaw, R. L., MacCarthy, P., Eds.; Wiley: New York, 1985; pp 477-492. (31) Lambert, J.; Lankes, U. In Refractory Organic Substances in the Environment; Frimmel, F. H., Abbt-Braun, G., Heumann, K. G., Hock, B., Lu ¨demann, H.-D., Spiteller, M., Eds.; Wiley: Weinheim, Germany, 2002; pp 89-95. (32) Chemical Analysis. Decision Limit, Detection Limit and Determination Limit. Estimation in Case of Repeatability, Terms, Methods, Evaluation; Norm DIN 32645, 1994. (33) Chin, Y.-P.; Aiken, G.; O’Loughlin, E. Environ. Sci. Technol. 1994, 28, 1853-1858. (34) Zhou, Q.; Cabaniss, S. E.; Maurice, P. A. Water Res. 2000, 34, 3505-3514. (35) Swift, R. W. Soil Sci. 1999, 164, 790-802. (36) Egeberg, P. K.; Christy, A. A.; Eikenes, M. Water Res. 2002, 36, 925-932. (37) Avena, J. M.; Wilkinson, K. J. Environ Sci. Technol. 2002, 36, 5100-5105. (38) Lankes, U. Festko¨rper-NMR-Spektroskopie zur Charakterisierung von refrakta¨ren organischen Sa¨uren in Gewa¨ssern; Ph.D. thesis, Universita¨t Regensburg, 2001. (39) Lankes, U.; Lu ¨ demann, H.-D. In Refractory Organic Substances in the Environment; Frimmel, F. H., Abbt-Braun, G., Heumann, K. G., Hock, B., Lu ¨ demann, H.-D., Spiteller, M., Eds.; Wiley: Weinheim, Germany, 2002; pp 96-114. (40) Abbt-Braun, G.; Weis, M.; Frimmel, F. H. Finnish Humus News 1991, 3, 151-156. (41) Abbt-Braun, G.; Lankes, U.; Jahnel, J. B.; Lambert, J.; Lu ¨demann, H.-D.; Frimmel, F. H. In Refractory Organic Substances in the Environment; Frimmel, F. H., Abbt-Braun, G., Heumann, K. G., Hock, B., Lu ¨demann, H.-D., Spiteller, M., Eds.; Wiley: Weinheim, Germany, 2002; pp 302-309. (42) Yamamoto, H.; Liljestrand, H. M.; Shimizu, Y.; Morita, M. Environ Sci. Technol. 2003, 37, 2646-2657.
Received for review March 20, 2003. Revised manuscript received November 3, 2003. Accepted November 10, 2003. ES0300416
