Environ. Sci. Technol. 2002, 36, 3393-3399
Variations of Molecular Weight Estimation by HP-Size Exclusion Chromatography with UVA versus Online DOC Detection N A M G U K H E R , * ,† G A R Y A M Y , † DAVID FOSS,§ AND JAEWEON CHO‡ Civil & Environmental Engineering, University of Colorado, ECOT 441, Boulder, Colorado 80309, Wright Water Engineers, Inc., Denver, Colorado 80211, and Kwangju Institute of Science and Technology, Kwangju, Korea, 500-712
High performance size exclusion chromatography (HPSEC) with ultraviolet absorbance (UVA) detection has been widely utilized to estimate the molecular weight (MW) and MW distribution of natural organic matter (NOM). However, the estimation of MW with UVA detection is inherently inaccurate because UVA at 254 nm only detects limited components (mostly π bonded molecules) of NOM, and the molar absorptivity of these different NOM constituents is not equal. In comparison, a SEC chromatogram obtained with a DOC detector showed significant differences compared to a corresponding UVA chromatogram, resulting in different MW values as well as different estimates of polydispersivity. The MWs of Suwannee River humic acid (SRHA), Suwannee River fulvic acid (SRFA), and various mixtures thereof were estimated with HPSEC coupled with UVA and DOC detectors. The results show that UVA is not an adequate detector for quantitative analysis of MW estimation but rather can be used only for limited qualitative analysis. The NOM in several natural waters (Irvine Ranch, California groundwater, and Barr Lake, Colorado surface water) were also characterized to demonstrate the different MWs obtained with the two detectors. The results of the SEC-DOC chromatograms revealed NOM constituent peaks that went undetected by UVA. Utilizing online DOC detection, a better representation of NOM MWs was suggested, with NOM displaying higher weightaveraged MW (Mw) and lower number-averaged MW (Mn) as well as higher polydispersivity. A method for estimation of the MWs of NOM fractional components and polydispersivities is presented.
Introduction Natural organic matter (NOM), ubiquitous in all ground and surface waters, is comprised of a heterogeneous mixture of humic and fulvic acids, lignins, carbohydrates, and proteins. These compounds are chemical and biological products of plant and animal residues (1, 2). NOM is known to play an important role in pollutant chemistry and geochemistry because it can form metal complexes, bind solubilized * Corresponding author phone: (303)735-2433; fax: (303)492-7317; e-mail:
[email protected]. † University of Colorado. § Wright Water Engineers, Inc.. ‡ Kwangju Institute of Science and Technology. 10.1021/es015649y CCC: $22.00 Published on Web 07/04/2002
2002 American Chemical Society
nonpolar organic compounds, influence colloid stability, and affect redox behavior in soil (3, 4). Molecular weight (MW) and MW distributions are important factors in NOM characterization because they relate to disinfection byproduct formation potential (5), proton and metal binding, organic pollutant partitioning, adsorption onto minerals and activated carbon, and NOM persistence and removal during water treatment (6). MWs have been determined by ultrafiltration (7), field flow fractionation (8), vapor pressure osmometry (9), analytical centrifugation (10), low angle X-ray scattering (11), and high performance size exclusion chromatography (HPSEC) methods (12). NOM MW estimation by HPSEC can be influenced by numerous factors such as calibration standards, undesirable column packing/ resin interactions, suitable data handling of chromatograms, and detection methods (13, 17). In recent years, HPSEC, with an ultraviolet absorbance (UVA) detector, has been widely employed due to various advantages (small sample volume, minimal pretreatment, availability of equipment, and ease and speed of analysis) (13). To improve accuracy of MW estimation by HPSEC, numerous investigators have reported their efforts (13-17). O’Loughlin and Chin (13) examined the effect of UVA detector wavelength on the determination of MW distribution of humic and fulvic acids. They found that both the number-averaged (Mn) and weight-averaged (Mw) MWs increased with increasing wavelength for humic substances. Zhou and co-workers (17) showed the effects on the Mn, Mw, and polydispersivity (F, a measure of the sample heterogeneity) by the definition of low MW (LMW) cutoff. They recommended either 2% of the maximum chromatogram height or MW ) 50 as the LMW cutoff, whichever is the higher value, and 1% of the maximum chromatogram height as the high MW (HMW) cutoff (17). Despite these recent and intensive efforts to improve MW estimation by HPSEC with UVA detection, significant inaccuracy remains due to the unequal molar absorptivities () of the organic components contained in NOM. The energies of the various types of molecular orbitals differ significantly. The electronic transitions among certain of the energy levels can be brought about by the absorption of radiation through σfσ*, nfσ*, nfπ*, and πfπ* (18). However, most UVA by organic compounds is based on transitions of n or π electrons to the π* excited state, because the energies required for these processes bring the absorption peaks into an experimentally convenient spectral region (200-700 nm) (18). Both transitions require the presence of unsaturated groups (double or triple bonds contained in NOM) to provide π orbitals. Therefore most of the compounds that have distinct chemical bonds show different molar absorptivities. The HPSEC-UVA-DOC technique has shown effective separation of different NOM fractions based on MW by both UVA and DOC online detectors. This system, in addition to detecting the aromatic NOM moieties with UVA, can also detect aromatic as well as aliphatic carbon fractions with the use of the DOC detector (19). The DOC detector has shown less interference than a UVA detector (254 nm) by inorganics (e.g., NO3-). The research objectives of this study were to compare and contrast the MW variation between UVA detection with the results obtained by the DOC detector and to suggest an alternative MW representation for NOM comprised of multiple fractional components.
Materials and Methods Suwannee River humic acid (SRHA) reference material (1R101H) and fulvic acid (SRFA) standard material (1S101F) obtained from the International Humic Substances Society VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3393
were analyzed to compare MWs obtained with the UVA and DOC detectors. The isolates (15.1 mg of SRHA and 15.3 mg of SRFA) were first dissolved in 1 L of Millipore Milli-Q (MQ) water to produce 8 mg C/L. After adjusting the pH to 6.8 with NaOH solution and filtering with a 0.45 µm cartridge (nylon) filter, molar absorptivity was measured with a UV spectrophotometer from 500 to 200 nm (UV-160A UV/Visible Spectrophotometer, Shimadzu). The ionic strength of these samples was then adjusted to 0.1 M with a concentrated eluent solution to match the ionic strength of the HPSEC mobile phase for the MW determination. Additional samples that were evaluated included an Irvine Ranch Water District (IRWD), California groundwater, and a Barr Lake, Colorado surface water (BL-SW). These samples were also analyzed by HPSEC system to illustrate the MW differences by detectors and to demonstrate an improved MW representation. Sample pretreatments also consisted of 0.45 µm cartridge (nylon filter) filtration and ionic strength adjustment. The HPLC (LC600 Shimadzu) was coupled with a UVA detector (SPD-6A Shimadzu) operated at 254 nm and an online DOC analyzer (Modified Sievers Turbo Total Organic Carbon Analyzer). The DOC detector was connected to the UVA detector outlet. The system, adapted based on the pioneering work of Huber and Frimmel (20, 21), employed a TSK-50S column (35 µm Toyopearl HW resin) that had a length of 25 cm and an inner diameter of 2 cm. A detailed description of this HPSEC system was provided by Her et al. (19). The HPSEC mobile phase was prepared with a phosphate buffer (0.0024 M NaH2PO4 + 0.0016 M Na2HPO4, pH 6.8) and 0.025 M Na2SO4, producing an ionic strength of 0.1 M. Helium gas was sparged into the mobile phase reservoir to eliminate inorganic carbon and oxygen that can cause interferences or react with the mobile or stationary phases. The flow rates were 1 mL/min, and sample injection volume was 2 mL. UVA and DOC data over time were collected every 6 s by digital signal processing using a modified Labview software. When the same operating conditions were applied, the reproducibility of both UVA and DOC chromatograms showed indistinguishable traces with multiple injections. The difference of peak area for three injections is less than 3% for this system, based on an injected DOC of 5 mg/L. Polystyrene sulfonate (PSS) standards have been shown to be more representative of NOM than PEG in SEC characteristics (hydrodynamic radii, viscosity etc.) (22). However, this system was calibrated with poly(ethylene glycol) (PEG) standards ranging from 200 to 10 000 g/mol because PSS displayed more undesirable interactions than did the PEG with the Toyopearl HW resin within the column. A semilog calibration curve (r2 > 0.99) was used to calculate the MWs. The baseline of the chromatograms was changed due to tailing and was set as 0 at 2% of the maximum chromatogram height, based on the approach of Zhou et al. (17). Mn, Mw, and F (number-averaged MW, weight-averaged MW, and polydispersivity, respectively) were determined using the following equations n
Mw )
n
∑(h ‚M )/∑h i
i
i)1 n
Mn )
i
n
∑h /∑(h /M ) i
i)1
(1)
i)1
i
i
(2)
i)1
F ) Mw/Mn
(3)
where hi and Mi are the height of the HPSEC chromatogram and the molecular weight at eluted volume i (16). Mp,ME (measured peak maximum MW) is defined as the MW corresponding to the peak maximum location measured 3394
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 15, 2002
FIGURE 1. Absorptivities based on 1 mol C/L for SRHA, SRFA, albumin, and sucrose in pure water (pH 7) as a function of wavelength. by HPSEC when the peak shape shows a Gaussian distribution. However, because the peaks of the samples did not follow a Gaussian distribution, Mp,ME values were determined as the MW at the median value of the cumulative weight chromatograms (corresponding to the centroid of the chromatogram area), instead of the using equations composed of Mn and Mw (Mp,ME ) xMnMw) (6). To compare the deviation between measured and calculated MWs for the mixture samples, the concept of calculated Mp (Mp,CA, calculated peak maximum MW) was introduced based on Mp,ME values of individual components. Using an arithmetic scale, Mp,CA can be determined for the known concentrations of SRHA and SRFA using the following equation
Mp,CA )
a(Mp,SRHA) + b(Mp,SRFA) a+b
(4)
where a is the DOC concentration of SRHA, b is the DOC concentration of SRFA, and Mp,SRHA and Mp,SRFA are MWs at the median value of the cumulative weight chromatograms. However, MW distributions were depicted on a logarithmic scale. Therefore, Mp,CA was calculated using eq 8, which was derived from equations from 5-7.
log Mp,CA ) log Mp,CA )
a log(Mp,SRHA) + b log(Mp,SRFA) a+b
(a +1 b)(log(M
p,SRHA)
a
+ log(Mp,SRFA)b)
log Mp,CA ) log((Mp,SRHA)a(Mp,SRFA)b)1/(a+b) Mp,CA )
a+b
x(Mp,SRHA)a(Mp,SRFA)b
(5) (6) (7) (8)
Results and Discussion Chin et al. (12) pointed out that MWs measured with HPSEC generated higher values than those measured with other methods due to the UVA detector normally used in the HPSEC system. These researchers found that the higher MW fractions have a greater molar absorptivity () and that high weight fractions appeared to be more abundant than they actually were, while the lower MW fractions (with lower ) appeared to be lower in concentration. The net result is an overestimation of the humic substance components of NOM (that have high ) and an under-representation of the nonhumic constituents. Carbohydrates and proteins, for example, are additional NOM components. Figure 1 shows the absorptivities based on 1 mol/L of organic carbon of four different NOM components: SRHA, SRFA, albumin (a protein), and sucrose (a small simple carbohydrate). As expected, the is
FIGURE 2. Mw and Mn values estimated by UVA and DOC chromatograms for (a) 8 mg C/L of SRHA and (b) 8 mg C/L of SRFA. not the same for all samples across all wavelengths. Albumin shows highest at lower wavelengths, decreasing rapidly with increasing wavelength. At 254 nm, SRHA and SRFA have high ; however, albumin and sucrose show low or negligible at this wavelength. A UVA detector connected to the HPSEC quantifies the response intensity based on the . As a result, the MW determined with a UVA detector (at 254 nm) is primarily the MW of only the high components such as humic and fulvic acids, leading to inherent inaccuracy, and over- or underestimation of MWs. Therefore, MW estimation with a UVA detector is problematical for a hetero-mixture of NOM, and a UVA detector cannot accurately be used for quantitative MW measurements of NOM. The estimated MWs of SRHA and SRFA are displayed in Figures 2. SRHA and SRFA were selected because they share similar chemical properties. The MWs (Mn ) 1385 daltons (Da) and Mw ) 2114 Da) of SRFA by the UVA detector were similar or slightly higher than reported values (Mn ) 1112 Da and Mw ) 1950 Da (9); Mn ) 1360 Da and Mw ) 2310 Da (16)), even with the use of PEG standards, which have been suggested to underestimate the MW of humic substances (22). The integrated areas under the DOC chromatograms for both samples were equal. In contrast, the areas of the UVA chromatograms for equivalent sample concentrations displayed significant differences due to the different molar absorptivities of humic and fulvic acid. This discrepancy between SRHA and SRFA is clearly seen in the UVA cumulative responses in Figure 3, portraying cumulative weight derived from integration of the chromatograms in Figure 2. Cumulative response is assumed to be equivalent to cumulative weight based on the UVA and DOC signal being proportional to the concentration (12). The cumulative weight of SRHA obtained with the UVA detector is much higher than that of SRFA. Because the chromatograms in Figure 2 were not Gaussian distributions, Mp,ME values were determined as the
FIGURE 3. UVA and DOC cumulative responses and Mp,ME values estimated for (a) 8 mg C/L of SRHA and (b) 8 mg C/L of SRFA.
TABLE 1. Comparison of MWs of Mixed Samples (SRHA and SRFA) by HPSEC-UVA-DOC Mp (Da)
detector
sample (mg C/L)
Mn (Da)
Mw (Da)
G
DOC
SRHA 12 + SRFA 4 SRHA 8 + SRFA 8 SRHA 4 + SRFA 12 SRHA 12 + SRFA 4 SRHA 8 + SRFA 8 SRHA 4+SRFA 12
1544 1393 1272 1809 1680 1534
3084 2759 2436 3102 2851 2531
2.00 1.98 1.91 1.71 1.70 1.65
UVA
Mp,ME Mp,CA
diff
2501 2220 1982 2697 2419 2126
-9 -14 -6 +79 +81 +38
2510 2234 1988 2618 2338 2088
MW of the median value of the cumulative response chromatograms (Figure 3). Mp,ME values for SRHA and SRFA with the UVA detector were 2932 and 1865 Da, respectively, and these values were higher than those measured by the DOC detector (2821 Da for SRHA and 1770 Da for SRFA). The MWs of three different mixtures of the isolates (12 mg/L SRHA and 4 mg/L SRFA, 8 mg/L SRHA and 8 mg/L SRFA, and 4 mg/L SRHA and 12 mg/L SRFA (as DOC)) were also determined with the UVA and DOC detectors (Table 1). Mn and Mw values were higher and F values were lower with UVA detection compared to the values obtained by DOC detection. Figure 4 shows the cumulative response distributions of the three different mixtures. The DOC cumulative chromatograms (Figure 4b) have equal total heights for these three samples, each with the same total DOC concentration of 16 mg C/L. However, the UVA cumulative chromatograms (Figure 4a) show different heights at these same total DOC concentrations. With an increasing portion of SRHA, the total height of the UVA cumulative chromatogram increased due to the higher of SRHA compared to SRFA. The Mp,CA for mixture samples (in Figure 5) was calculated with eq 8 using the individual Mp,ME values of the SRHA and SRFA obtained with the UVA and DOC detectors from Figure VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3395
FIGURE 4. Cumulative responses and Mp,ME values for three different mixtures of SRHA and SRFA: (a) with UVA detector and (b) with DOC detector. 3. As shown in Table 1 and Figure 5, the Mp,ME values obtained by DOC detection were very similar to the Mp,CA values, within the error of the DOC detector. However, in contrast, the Mp,ME values obtained by UVA detection showed slightly more deviation from the Mp,CA values. Higher Mp,ME than Mp,CA values were obtained due to the greater contribution of the higher MW SRHA. As discussed earlier, this discrepancy results from the differences in . The MW deviation between SRHA and SRFA with the UVA detector displayed in Figure 5 showed minor differences likely attributable to the relatively similar and MW for these NOM components. Figures 6 and 7 further elucidate the different MW distributions depending on the detector. Figure 6 shows the HPSEC-UVA-DOC chromatograms for Barr Lake, a 32 000 acre-foot (3.95 × 107 m3), hypereutrophic reservoir. The Mn and Mw values measured with the DOC detector were 443 Da and 5609 Da. However, the values measured with the UVA detector were significantly different. The component at 13 500 Da showed a very low response in the UVA chromatogram compared to DOC chromatogram, illustrating the low of this component. Therefore, the UVA-determined Mw value was much lower, 2010 Da, and the F value was also significantly different (12.7 with the DOC detector versus 5.3 with the UVA detector). Figure 7 presents another example of the MW deviation between the two detectors using a groundwater sample collected at 1090 ft depth from the Irvine Ranch Water District (IRWD). An aliphatic carbon fraction displaying a strong DOC response and no discernible UVA response was observed at 3396
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 15, 2002
a high MW range (19 600 Da, could represent polysaccharide like substances (23)) and a low MW range (210 Da). With these two additional large responses, the DOC detector resulted in a higher Mw (6300 Da) and a lower Mn (750 Da) compared with those values measured by UVA (Mw: 3400 Da and Mn: 1770 Da) and also a significantly higher F (8.4 with the DOC detector and 1.9 with the UVA detector). The higher F indicates greater NOM heterogeneity. Thus, singular MW representations utilizing Mw and Mn are not appropriate for NOM that consists of a high F (e.g., 8.40). Therefore, in the case of multiple MW fractions, each fraction should be represented independently, possibly including each fractional concentration. Figure 8 presents the normalized cumulative response using the DOC data from Figure 7. To represent the MWs of multiple components of NOM, each fractional component should be expressed. Chromatogram peaks were divided into four different fractions based on the “inflection points” in Figure 8. Then, Mn, Mw, F, and Mp,ME values were obtained for each fraction. Each fractional concentration was also determined by multiplying total concentration by fractional height. Table 2 shows a comparison of currently accepted and proposed representations of MWs. The F values, which ranged from 1.02 to 1.27 for each fraction, are reasonable values for a single MW constituent. This proposed representation can express MW more specifically and accurately and provides much greater utility and insight, especially when evaluating NOM persistence and removal across water treatment processes.
FIGURE 5. Mp,ME and Mp,CA values obtained by UVA and DOC chromatograms of mixture samples: (a, d) 12 mg C/L SRHA and 4 mg C/L SRFA, (b, e) 8 mg C/L SRHA and 8 mg C/L SRFA, and (c, f) 4 mg C/L SRHA and 12 mg C/L SRFA.
FIGURE 6. HPSEC-UVA-DOC chromatograms for Barr Lake. The estimation of MW is an important factor for understanding the physical and chemical properties of NOM (24, 25) and determination of appropriate water treatment process selection, design, and operation. Even though the molar absorptivities of NOM components are significantly different,
UVA detection has been widely used for determining MWs. Instead, for estimating MW based on equitable consideration of all NOM components, a DOC detector that registers equal intensity for both aliphatic and aromatic compounds should be employed. The importance of employing a DOC detector VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3397
FIGURE 7. HPSEC-UVA-DOC chromatograms for IRWD groundwater.
FIGURE 8. Cumulative response distributions (normalized) by DOC detector of IRWD groundwater.
TABLE 2. Comparison of Currently Accepted and Proposed Representations of MWs (by DOC Detector) proposed representation currently accepted fraction fraction fraction fraction representation 1 2 3 4 concn (mg C/L) Mn (Da) Mw (Da) F Mp (Da)
2.69
0.77
1.06
0.37
0.49
750 6300 8.40 2460
17780 19200 1.08 18870
2234 2838 1.27 2607
509 528 1.04 517
210 215 1.02 213
in MW determination relates to the measurement of nonaromatic compounds (e.g., polysaccharides) and partially aromatic compounds (e.g., proteins), which are NOM constituents, and are important in water treatment processes. Understanding these aliphatic and partially aromatic compounds is especially significant, as proteins can form chlorination byproducts and polysaccharides have been implicated as major foulants in membrane treatment (26). UVA detection should only be used for qualitative analysis, 3398
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 15, 2002
such as comparison of the aromaticity of NOM components as a function of MW.
Acknowledgments We gratefully acknowledge Ionics-Sievers Instruments for supplying the modified TOC 800 analyzer and Mr. Paul Kosenka for his continuous technical support.
Literature Cited (1) Nystrom, M.; Ruohomaki, K.; Kaipia, L. Desalination 1996, 106, 79. (2) Yuan, W.; Zydney, A. L. J. Membr. Sci. 1999, 157, 1. (3) Aiken, G. R.; McKnight, D. M.; Wershaw, R. L.; MacCarthy, P. Humic Substances in Soil, Sediment, and Water; Wiley-Interscience: New York, 1985; 692 pp. (4) Scott, D. T.; McKnight, D. M.; Blunt-Harris, E. L.; Kolesar, S. F.; Lovely, D. R. Environ. Sci. Technol. 1998, 32, 2984. (5) Amy, G. L.; Collins, M. R.; Kuo, C. J.; King, P. H. J. Am. Water Works Assoc. 1987, 79, 43. (6) Cabaniss, S. E.; Zhou, Q.; Maurice, P. A.; Chin, Y.; Aiken, G. Environ. Sci. Technol. 2000, 34, 1103. (7) Buffle, J.; Deladsey, J. P.; Haerdi, W. Anal. Chim. Acta 1978, 101, 339. (8) Giddings, J. C.; Williams, P. S.; Beckett, R. Anal. Chem. 1987, 59, 28.
(9) Figini, R. V.; Marx-Figini, M. Macromol. Chem. Phys. 1981, 182, 437. (10) Ende, H. A. J. Polym. Sci. 1965, 3, 139. (11) Wershaw, R. L.; Pinckney, D. J. J. Res. U.S. Geol. Surv. 1973, 1, 701. (12) Chin, Y.; Aiken, G.; O’Loughlin, E. Environ. Sci. Technol. 1994, 28, 1853. (13) O’Loughlin, E.; Chin, Y. Water Res. 2000, 35, 333. (14) Mori, S.; Kato, H.; Nishimura, Y. J. Liq. Chromatogr. Relat. Technol. 1996, 19, 2077. (15) Mori, S.; Marechal, H.; Suzuki, H. Int. J. Polym. Anal. Charact. 1997, 4, 87. (16) Yau, W. W.; Kirkland, J. J.; Bly, D. D. Wiley: New York, 1979; 476 pp. (17) Zhou, Q.; Cabaniss, S. E.; Maurice, P. A. Water Res. 2000, 34, 3505. (18) Skoog, D. A.; Leary, J. J. Principles of instrumental analysis, 4th ed.; Saunders College Publishing: 1992; p 150. (19) Her, N.; Amy, G. L.; Foss, D.; Cho, J.; Yoon, Y.; Kosenka, P. Environ. Sci. Technol. 2002, 36, 1069.
(20) Huber, S.; Frimmel, F. H. Fresenius J. Anal. Chem. 1992, 342, 198. (21) Huber, S. A.; Frimmel, F. H. Vom Wasser. 1996, 86, 277. (22) Perminova, I. V.; Frimmel, F. H.; Kovalevskii, D. V.; Abbt-Braun, G.; Kudryavtsev, A. V.; Hesse, S. Water Res. 1998, 32, 872. (23) Hesse, S.; Kleiser, G.; Frimmel, F. H. Water Sci. Technol. 1999, 40, 1. (24) DeWit, J. C. M.; van Riemsdijk, W. H.; Koopal, L. K. Environ. Sci. Technol. 1993, 27, 2005. (25) Vermeer, A. W. P.; Koopal, L. K. Langmuir 1998, 14, 4210. (26) Cho, J.; Amy, G. L.; Pelligrino, J.; Yoon, Y. Desalination 1998, 118, 101.
Received for review August 16, 2001. Revised manuscript received May 5, 2002. Accepted June 3, 2002. ES015649Y
VOL. 36, NO. 15, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3399
