Environ. Sci. Technol. 2002, 36, 3497-3503
Fractionation of Natural Organic Matter in Drinking Water and Characterization by 13C Cross-Polarization Magic-Angle Spinning NMR Spectroscopy and Size Exclusion Chromatography S. WONG Institut fu ¨ r Polymerforschung, Hohe Strasse 6, Dresden D01069, Germany J. V. HANNA ANSTO NMR Facility, c/o Materials Division, Private Mail Bag 1, Menai, New South Wales 2234, Australia S. KING, T. J. CARROLL, R. J. ELDRIDGE,* D. R. DIXON, AND B. A. BOLTO CSIRO Molecular Science, Bag 10, Clayton South, Victoria 3169, Australia S. HESSE, G. ABBT-BRAUN, AND F. H. FRIMMEL Engler-Bunte-Institut, Wasserchemie, Universita¨t Karlsruhe, Engler-Bunte-Ring 1, D-76131 Karlsruhe, Germany
Natural organic matter from drinking water sources was fractionated, and the fractions were characterized by NMR and SEC with the aim of relating NOM structure to treatability. Organic matter was isolated from two Australian surface waters (Horsham, Moorabool) by reverse osmosis and from a groundwater (Wanneroo) by anion exchange. The isolates were fractionated according to polarity and charge by resin adsorption. 13C NMR spectra of the freezedried fractions showed the most hydrophobic fraction to be high in aliphatic and aromatic carbon while slightly hydrophobic organics have more carbonyl and alkoxyl carbon. The Horsham and Wanneroo hydrophilic fractions show strong alkoxyl signals attributed to carbohydrate. Moorabool hydrophilics contain aromatic (phenolic) carbon; the apparent absence of carbohydrate appears to be an artifact. Size-exclusion chromatograms were recorded on the original and fractionated organics with both UV and dissolved organic carbon detection. The Horsham and Moorabool organics have similar molecular size distributions while Wanneroo is dominated by strongly absorbing species having large hydrodynamic radii. The hydrophobic and charged hydrophilic fractions also have high apparent MW, while the neutral fraction is higher in low-MW compounds of relatively low specific absorbance, suggestive of carbohydrates.
Introduction Natural organic matter (NOM) represents a broad range of structurally complex compounds derived from the degrada* Corresponding author phone: +61 3 9545 2222; fax: +61 3 9545 2415; e-mail:
[email protected]. 10.1021/es010975z CCC: $22.00 Published on Web 07/13/2002
2002 American Chemical Society
tion of plants and microorganisms. It influences the availability and migration of nutrients and pollutants in the environment and largely controls drinking water purification processes. Characterization of NOM is made difficult by the heterogeneous size, structure, and functional chemistry of its constituent compounds, and by the variation in NOM with origin, climate, and season (1-3). One approach to characterization is to fractionate the NOM into broad chemical classes by resin adsorption (4, 5) before gathering information on chemical composition (by techniques such as pyrolysis GC-MS or 13C NMR) and molecular size [by sizeexclusion chromatography (SEC) (6) or field-flow fractionation]. The characterization of NOM has historically focused upon hydrophobic components such as humic and fulvic acids. These compounds contribute color, transport hydrophobic pollutants such as pesticides, and generate byproducts on disinfection of drinking water (7). Anionically charged hydrophilic compounds contribute to the transport of heavy metal cations. However, hydrophobic and charged compounds are relatively effectively removed from drinking water by conventional treatment with metal salt coagulants (810). Therefore, NOM entering the drinking water distribution system after treatment will be relatively enriched in uncharged hydrophilic compounds. These compounds also contribute to disinfection byproducts (8, 11) and may have a role in bacterial regrowth in distribution systems (12). We have fractionated NOM from several Australian drinking water sources into two hydrophobic and two hydrophilic groups by adsorption on nonpolar and ion exchange (IX) resins. We previously reported the removal of these fractions by a broad range of coagulants and their potential to form disinfection byproducts (13). The results were consistent with the findings cited above, with the hydrophobic and charged fractions being most readily removed and all fractions forming trihalomethanes on chlorination. The aim of the work reported here was to gather information on the composition and size of the treatable and recalcitrant fractions as a first step toward correlating structure and treatability and ultimately developing appropriate removal strategies. We used solid-state 13C NMR analysis and SEC to obtain structural information on hydrophilic and hydrophobic NOM fractions from three sources. Solid-state 13C NMR is extensively used by soil scientists to identify NOM compounds by determining the carbon functional groups in whole soils and humic extracts (1418). The heterogeneity and complexity of NOM typically limit the level of detail available from 13C NMR to broad functional group classes such as aromatics or aliphatics. There is potential to improve this level of detail toward definitive identification of specific signature compounds by adsorptive fractionation. In particular, selective removal of hydrophobic NOM isolates the narrower class of hydrophilic compounds that dominate residual NOM after drinking water treatment. SEC separates solutes primarily according to their size and shape, although separation is also influenced by interactions such as ionic exclusion and hydrophobic attraction (19). SEC of NOM has usually been performed with ultraviolet detection, which biases the results toward strongly absorbing constituents such as humic substances. We used an instrument equipped with both ultraviolet and organic carbon detectors, allowing all organic constituents to be detected and the specific ultraviolet absorbance (SUVA) of each size fraction to be determined. The molecular size of SEC fractions can be expressed in terms of the molecular weight (MW) of simple polymers having similar retention VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3497
times and hence hydrodynamic radii. Approximate MW averages can be estimated by reference to model compounds of known MW, size, and hydrophobicity (20), but it cannot be assumed that the analyte and the model compound are similar in shape and structure. Because NOM is a complex mixture of macromolecules and/or aggregates of ill-defined MW (21, 22), SEC was used here simply to resolve each sample into characteristic fractions according to retention time.
Materials and Methods Raw Water Fractionation. NOM was extracted from two surface waters and one groundwater. The surface water NOM was concentrated by reverse osmosis from the Moorabool River near Anakie, Victoria, and from the Mount Zero reservoir near Horsham, Victoria. The groundwater NOM was obtained as a concentrate from a MIEX magnetic ion-exchange pilot plant at Wanneroo, Western Australia. The plant was operated so as to recover about 55% of the dissolved organic carbon (DOC) in the raw water (23). The sample is therefore not fully representative of the source NOM but rather represents the material removed by IX. This material was used in our earlier treatability studies (13). The anion exchanger was contacted repeatedly with raw water, and then regenerated with brine. The regenerant was contacted repeatedly with loaded resin, giving finally a concentrate containing 8-10 g/L DOC. The RO and IX concentrates were filtered (0.45 µm), adjusted to pH 2, and passed successively through Supelite DAX-8, Amberlite XAD-4, and (after neutralization to pH 8) Amberlite IRA-958 (24, 25). DAX-8 and XAD-4 were eluted with 0.1 M NaOH and the eluates acidified on Amberlite IR-120. These fractions are denoted strongly hydrophobic acids (SHA) and weakly hydrophobic acids (WHA), respectively. Negatively charged hydrophilic NOM adsorbing on IRA-958 was eluted with 1 M NaOH/1 M NaCl. IRA-958, a macroporous anion exchanger having a moderately polar acrylic skeleton, is known to take up aquatic organic matter across a wide range of molecular size and is readily regenerated with alkaline brine. Irreversible adsorption is minimized by regenerating promptly (26). The three eluates and the remaining material not adsorbed on any resin (denoted neutral hydrophilic, although possibly including cationic components) were freeze-dried. Before use, DAX-8 and XAD-4 were extracted with three changes of methanol and acetonitrile (5). These resins and IRA-958 were rinsed alternately with 0.1 M NaOH and 0.1 M HCl until the effluent contained WHA > neutral. High values are indicative of double bond character (>CdC< or >CdO) and are seen for humic and fulvic acids (29, 36). The observed trend is therefore consistent with a progressive decrease in humic character. The absorbance of the charged fractions could be due to aromatic compounds, as indicated by NMR in the case of Moorabool, or to proteins and amino acids, which have been detected in the humic fractions of water from various sources (35, 37). The Horsham and Moorabool chromatograms show a peak at tR ) 35 min with a relatively low SUVA. In previous studies, similar peaks were attributed to anthropogenic organic matter (29, 37). The SEC Fractions. For further interpretation, the chromatograms were divided into three characteristic SEC fractions as illustrated for Horsham in Figure 3. Table 2 shows the weight-average molecular weight, Mw, of the corresponding PEG fractions. Substances eluting at 22 min < tR < 34 min (SEC fraction I) were assigned to large molecules (or aggregates) such as polysaccharides and humic substances (20, 29). Brown water and aqueous soil samples are generally dominated by such material (35, 36). Substances eluting at 34 min < tR < 39 min (SEC fraction II) were assigned to building blocks of refractory organic substances (29, 37), while in SEC fraction III low-MW molecules (e.g., carbohydrates, aldehydes, ketones, or alcohols) elute at tR > 39 min (29, 34). Small peaks at times well beyond the fully retained limit (∼42 min) indicate nonsteric interactions on the column. The DOC chromatograms in Figure 3 show distinct differences between the raw water and the different NOM components. There is no clear difference between the SHA and WHA fractions of the Horsham sample, implying virtually identical size distributions. The charged fraction is dominated by relatively large species eluting near tR ) 31 min, while the neutral isolate covers a broad spectrum of molecular size, with a high proportion of smaller molecules (tR > 34 min). The shift of the sharp peak from tR ) 42 min to shorter retention times can be attributed to the higher salt concentration in the sample as a result of the isolation procedure. Table 2 shows for each NOM sample the amount of DOC in each SEC fraction as a percentage of the total DOC signal. SUVA values are also shown for each SEC fraction. The unfractionated NOM and the hydrophobic fractions are dominated by large species eluting in SEC fraction I. The proportion of DOC eluting in this SEC fraction decreases in the order Wanneroo > Horsham > Moorabool. SUVA values
for the hydrophobic fractions decrease in the same order, suggesting that the large hydrophobic species in the Moorabool NOM have a lower content of highly unsaturated and aromatic compounds than the corresponding material in the other waters. In all three waters, the charged hydrophilic fractions are also dominated by SEC fraction I. Wanneroo charged has the lowest proportion of DOC in SEC fraction I but the highest SUVA for this SEC fraction, implying more unsaturation than the corresponding component from the other sources. In contrast, SUVA for charged substances in SEC fraction III is in the order Moorabool > Horsham > Wanneroo. The high proportion of DOC eluting in SEC fractions II and III for the neutral fraction of all three waters shows that this fraction contains the greatest proportion of smaller molecules, as already noted for the Horsham case. This fraction generally has the lowest SUVA, suggesting a high carbohydrate content.
Discussion The NMR spectra of the hydrophobic fractions exhibit the general features of aliphatic, alkoxyl, unsaturated, and carbonyl carbons. The SHA fractions have a higher aliphatic and unsaturated carbon content coupled with less alkoxyl and carbonyl carbon than the WHA fractions. These features are consistent with the fractionation expected from XAD-4 and DAX-8 resins and with the SEC results, where the high SUVA values observed for the SHA fractions again imply high levels of unsaturation. The aliphatic and alkoxyl carbon regions are broader and more featureless than other spectral regions due to larger residual carbon-proton dipolar couplings (higher proton concentration) and possibly larger dispersion resulting from high MW and broad molecular weight distribution. The cross-polarization technique precludes further, more detailed quantitative analysis based on spectral intensities. However, the 13C CPMAS NMR spectra of both hydrophobic fractions of the three waters studied here have the same features as numerous other 13C CPMAS NMR spectra of humic and fulvic acids isolated from water and soil sources (5, 38). Even in cases where NMR spectra were taken with quantitative techniques such as direct single-pulse excitation or using liquid samples at low pH, it is difficult to progress beyond broad characterizations of humic and fulvic acid functional groups. The size, structure, and functional chemistry of these compounds are too complex to permit extraction of individual features. Fractionation of hydrophobic NOM does not produce simplified13C NMR spectra when compared to the spectra of unfractionated NOM. Indeed, the spectrum of unfractionated Wanneroo NOM (not shown) is very similar to that of the SHA fraction. Again this is consistent with SEC findings: the chromatograms of the SHA fractions are most like those of the unfractionated samples, whereas those of the hydrophilic fractions are significantly different. The unsaturated carbon region of the hydrophobic acid spectra has higher resolution than other regions. This is due to a lower degree of carbon-proton dipolar coupling in unsaturated functional groups. The lower proton density also reduces spectral intensity in this region, but it is still possible to assign specific functionality and more detailed characterization. The chemical shifts in this region of the Moorabool hydrophobic fraction spectra can be assigned to molecular fragments containing aromatic rings, possibly lignin or tannin building block materials. These features are more distinct in the hydrophilic fraction spectra, as discussed below. A comparison of the NMR spectra of the hydrophobic fractions of the three waters gives some insight into the effectiveness of fractionation of hydrophobic NOM. The spectra of the Moorabool SHA and WHA fractions are relatively similar, indicating that fractionation was least effective in this case. The Moorabool hydrophobic fractions VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3501
are also distinct from the same Horsham and Wanneroo fractions in lack of features in the alkoxyl region of the spectra, suggesting amorphous polysaccharides or possibly protein. (The Moorabool hydrophobics contained 2-3 times as much nitrogen as the corresponding Wanneroo fractions.) The isolation of this NOM in the hydrophobic fractions could be a result of lower fractionation effectiveness. Molecular association or entanglement may influence fractionation. The NMR spectra of the Horsham strongly and weakly hydrophobic acid fractions are relatively distinct, indicating more effective fractionation in this case. The enhanced resolution in the alkoxyl region suggests less amorphous material than in Moorabool. The Wanneroo hydrophobic fractions support similar arguments. Wanneroo is dominated by SHA, which NMR and SUVA measurements both imply is the most aromatic and least cellulosic NOM component of the three waters studied. It seems plausible to postulate that NOM composition, MW, and morphology strongly influence the fractionation performance. Further support for these inferences is available from analysis of the hydrophilic NOM. The NMR spectra of the hydrophilic fractions have substantially higher resolution than those of the hydrophobic fractions. The NOM retained in these fractions is therefore a sufficiently smaller subset of the original NOM to permit extraction of individual features. In principle, this could result from reduced heterogeneity in chemical functionality and/ or average MW as a result of the retention of the more chemically diverse and/or higher MW NOM in the hydrophobic fractions. The former effect would reduce the number of chemically distinct but overlapping NMR carbon signals; the latter effect may also reduce the line widths of the remaining signals. The SEC results rule out the role of MW because the charged hydrophilic fractions are dominated by high-MW NOM while the neutral fractions are rich in lowMW material. Unfortunately, the hydrophilic NOM samples have a low organic carbon content and a high salt content, which may reduce cross-polarization efficiency. Hygroscopic salts are particularly likely to cause loss of signal intensity. On the other hand, desalting to increase the organic carbon content can compromise the representativeness of the sample. Despite these limitations, 13C CPMAS NMR provides some insight into the character of hydrophilic NOM and the effectiveness of the fractionation scheme. The NMR spectra of the Horsham and Wanneroo hydrophilic NOM are dominated by alkoxyl carbon, attributed to carbohydrate. The SEC results show that this carbohydrate mainly consists of or is associated with large molecules. The NMR spectra also show minor amounts of aliphatic, possibly proteinaceous, carbon in the charged fractions. Wanneroo has the greatest proportion of low-MW charged material (SEC fraction III), consistent with the narrow line widths and terminal methyl groups observed in the NMR. The Horsham neutral fraction has a higher carbohydrate content than the charged fraction. The SEC shows this to be mainly low-MW. The dominant carboxyl peak of the charged fraction is virtually absent in the neutral fraction. As would be expected, SUVA is low. These results are consistent with the predicted passage of simple sugars, alcohols, and ketones into the unadsorbed neutral fraction (4) and the expected retention of protein in the charged fraction. However, the presence of charged groups (sulfate or phosphate) on polysaccharide NOM might facilitate retention on IRA-958. In any case, the use of IRA-958 to separate the putative protein and polysaccharide is less than 100% effective. The NMR spectra of the Moorabool hydrophilic NOM are most inconsistent with the expected fractionation outcome and the SUVA results. The spectra are markedly distinct from those of the Horsham and Wanneroo fractions and suggest that the Moorabool hydrophilic fractions are predominantly 3502
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 16, 2002
aromatic and phenolic. NMR signals at 160 and 165 ppm are consistent with highly oxygenated or amidic substances, suggesting proteins and carbohydrates, but there is insufficient NMR signal in the alkoxyl and aliphatic regions of the spectrum to substantiate this assignment. The unsaturated carbon signals in the Moorabool hydrophilic fractions could be explained by guaiacyl- or syringyl-type carbons, both of which are exclusively aromatic. However, it seems more likely that the expected aliphatic and alkoxyl carbon is in fact present, but invisible because of the salt effect proposed above. In this way, the low SUVA of the Moorabool hydrophilic fractions can be reconciled with the NMR spectra. We have reported elsewhere comparisons of NOM removal by several established and innovative water treatment methods (13, 39-41). The ready removal of hydrophobic and charged hydrophilic fractions is understandable in terms of their interactions with coagulants or ion exchangers. Further work on the neutral hydrophilic fraction is warranted because this poorly removed material contributes significantly to disinfection byproducts (8, 11, 13) and disproportionately to membrane fouling (42). Better molecular characterization is needed to guide the development of moreeffective removal processes.
Acknowledgments Funding was provided by the Cooperative Research Centre for Water Quality and Treatment and the Deutscher Verein des Gas- und Wasserfaches. We thank Domenico Petruzzelli, IRSA, Bari, Italy, for the thermal desorption GC-MS analyses.
Literature Cited (1) Aiken, G. R.; McKnight, D. M.; Wershaw, R. L.; MacCarthy, P. Humic substances in soil, sediment and water; Wiley: New York, 1985. (2) Frimmel, F. H.; Christman, R. F. Humic substances and their role in the environment; Wiley: Chichester, 1988. (3) Hayes, M. H. B.; MacCarthy, P.; Malcolm, R. L.; Swift, R. S. Humic Substances II: In Search of Structure; Wiley: New York, 1989. (4) Leenheer, J. A. Environ. Sci. Technol. 1981, 15, 578-587. (5) Aiken, G. R.; McKnight, D. M.; Thorn, K. A.; Thurman, E. M. Org. Geochem. 1992, 18, 567-573. (6) Mu ¨ ller, M.-B.; Schmitt, D.; Frimmel, F. H. Environ. Sci. Technol. 2000, 34, 4867-4872. (7) Rook, J. J. Water Treat. Exam. 1974, 23, 234-243. (8) Bose, P.; Reckhow, D. A.; Bezbarua, B. K. Proc. Annu. Conf.; AWWA: Denver, CO, 1993; Water Research pp 417-429. (9) Singer, P. C.; Harrington, G. W. Proc. WQTC Conf.; AWWA: Denver, CO, 1993; pp 1-19. (10) Dryfuse, M. J.; Miltner, R. J.; Summers, R. S. Proc. Annu. Conf.; AWWA: Denver, CO, 1995; Water Quality pp 217-241. (11) Croue´, J. P.; Lefebvre, E.; Martin, B.; Legube, B. Water Sci. Technol. 1993, 27 (11), 143-152. (12) van der Kooij, D. J. AWWA 1992, 84 (2), 57-65. (13) Bolto, B.; Abbt-Braun, G.; Dixon, D.; Eldridge, R.; Frimmel, F.; Hesse, S.; King, S.; Toifl, M. Water Sci. Technol. 1999, 40 (9), 71-79. (14) Schnitzer, M.; Preston, C. M. Soil Sci. Soc. Am. J. 1986, 50, 326331. (15) Wilson, M. A. NMR techniques and applications in geochemistry and soil chemistry, 1st ed.; Pergamon: Oxford, U.K., 1987; pp 70-74. (16) Hanna, J. V.; Johnson, W. D.; Quezada, R. A.; Wilson, M. A.; Lu, X.-Q. Environ. Sci. Technol. 1991, 25, 1160-1164. (17) Preston, C. M. Soil Sci. 1996, 161, 144-166. (18) Lu, X.-Q.; Hanna, J. V.; Johnson, W. D. Appl. Geochem. 2000, 15, 1019-1033. (19) De Nobili, M.; Gjessing, E.; Sequi, P. In Humic Substances II: In Search of Structure; Hayes, M. H. B., MacCarthy, P., Malcolm, R. L., Swift, R. S., Eds.; Wiley: New York, 1989; pp 561-591. (20) Perminova, I. V.; Frimmel, F. H.; Kovalevskii, D. V.; Abbt-Braun, G.; Kudryavtsev, A. V.; Hesse, S. Water Res. 1998, 32, 872-881. (21) Wershaw, R. L. Soil Sci. 1999, 164, 803-813. (22) Conte, P.; Piccolo, A. Environ. Sci. Technol. 1999, 33, 16821690. (23) Bourke, M.; Slunjski, M.; O’Leary, B.; Smith, P. Proc. Fed. Conv., 18th; Aust. Water Wastewater Assn.: Sydney, 1999; CD-ROM.
(24) Croue´, J. P.; Martin, B.; Deguin, A.; Legube, B. Proceedings of Natural Organic Matter in Drinking Water Workshop; AWWA: Denver, CO, 1994; pp 73-81. (25) Bolto, B.; Dixon, D.; Eldridge, R.; King, S.; Toifl, M. In Chemical Water and Wastewater Treatment V; Hahn, H. H., Hoffmann, E., Ødegaard, H., Eds.; Springer: Berlin, 1998; pp 173-185. (26) Symons, J. M.; Fu, P. L.-K.; Kim, P. H.-S. In Ion Exchange Technology; Sengupta, A. K., Ed.; Technomic: Lancaster, PA, 1995; pp 149-192. (27) Petruzzelli, D. React. Funct. Polym. 1997, 35, 89-98. (28) Huber, S. A.; Frimmel, F. H. Vom Wasser 1996, 86, 277-290. (29) Huber, S. A.; Frimmel, F. H. Environ. Sci. Technol. 1994, 28, 1194-1197. (30) AIST. http://www.aist.go.jp/RIODB/SDBS/menu-e.html. (31) Smernik, R. http://www.waite.adelaide.edu.au/NMR/nmrlist. html. (32) Owen, D. M.; Amy, G. L.; Chowdhury, Z. K.; Paode, R.; McCoy, G.; Viscosil, K. J. AWWA 1995, 87 (1), 46-63. (33) Chow, C.; van Leeuwen, J.; Fabris, R.; King, S.; Withers, N.; Spark, K.; Drikas, M. Proc. Watertech; Aust. Water Wastewater Assn.: Sydney, 2000; CD-ROM. (34) Abbt-Braun, G.; Frimmel, F. H. Environ. Int. 1999, 25, 161-180.
(35) Hesse, S.; Frimmel, F. H. Acta Hydrochim. Hydrobiol. 1999, 27, 94-97. (36) Frimmel, F. H.; Jahnel, J.; Hesse, S. Water Sci. Technol. 1998, 37 (2), 97-103. (37) Hesse, S.; Balz, A.; Frimmel, F. H. Vom Wasser 1997, 88, 103117. (38) Mahieu, N.; Powlson, D. S.; Randall, E. W. Soil Sci. Soc. Am. J. 1999, 63 (2), 307-319. (39) Bolto, B.; Dixon, D.; Eldridge, R.; King, S. Water Res. 2001, 35, 2669-2676. (40) Bolto, B.; Dixon, D.; Eldridge, R.; King, S. Water Res., submitted for publication. (41) Bolto, B.; Dixon, D.; Eldridge, R.; King, S.; Linge, K. Water Res., submitted for publication. (42) Carroll, T.; King, S.; Gray, S. R.; Bolto, B. A.; Booker, N. A. Water Res. 2000, 34, 2861-2868.
Received for review May 15, 2001. Revised manuscript received May 13, 2002. Accepted June 7, 2002. ES010975Z
VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3503
