Environ. Sci. Technol. 2004, 38, 3360-3369
Depolymerization of Chromophoric Natural Organic Matter JAMES THOMSON, ADELE PARKINSON, AND F E L I C I T Y A . R O D D I C K * ,† School of Civil and Chemical Engineering, RMIT University, GPO Box 2476V, Melbourne, Victoria 3001, Australia, and Cooperative Research Centre for Water Quality and Treatment, Private Mail Bag 3, Salisbury, SA 5108, Australia
Although the importance of natural organic matter (NOM) in the environment and in drinking water treatment is wellknown, its structure is still ill-defined. The fragmentation patterns of NOM treated by irradiation (various wavelengths - 185-400 nm), hydroxyl radicals, chlorine, ozone, and breakdown by a white rot fungus were studied to investigate the structure of chromophoric NOM molecules. Size exclusion chromatography was used to monitor the size distributions of NOM in two natural water waters and two NOM isolates. Three distinct fragmentation patterns were observed: ozone attack appeared to be nonsize specific, UV (g254 nm) irradiation preferentially removed higher molecular weight chromophores, while processes involving hydroxyl radical showed intermediate size specificity. For the samples studied, the UV (g254 nm) irradiationinduced fragmentation of NOM followed the patterns suggested by a simple trimer depolymerization model, supporting the viewpoint that NOM has repeating structural units joined by photolabile chemical bonds. The largest molecules reacted most rapidly, progressively fragmenting into slower reacting smaller molecules, which initially accumulated before breaking down to become nonchromophoric. This dependency of rate on molecular size appears to follow from the law of photochemistry which states the rate of reaction is proportional to the rate of light absorption: larger chromophores had higher molar absorptivities, absorbed more photons, and hence reacted faster than smaller chromophores.
Introduction Natural organic matter (NOM) plays an important role in aquatic systems both from environmental and drinking water treatment perspectives. NOM is environmentally significant because of its various roles in the global carbon cycle (1), its binding and transport of otherwise insoluble or less soluble metals (e.g., Cu, Fe, Al) and organics (e.g., DDT) (2), its biological functions (i.e., phosphorus and nitrogen nutrient cycling, trace metal availability, limiting potential metal toxicity) (2), and in the photochemistry of natural water bodies (3). In drinking water it is a problem due to the contribution of color and odor, it is a precursor to potentially harmful disinfection byproducts (4), and its contribution to the formation of biofilms in distribution systems. In drinking water treatment it is a nuisance because it consumes * Corresponding author phone: +61 3 9925 2080; fax: +61 3 9925 3746; e-mail:
[email protected]. † RMIT University. 3360
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 12, 2004
chemicals, competes with micropollutants for sites on activated carbon, consumes oxidants intended for micropollutant removal or microorganism inactivation, fouls membranes, and attenuates the intensity of radiation during UV disinfection. NOM is not a well-defined chemical entity but can be classified into a number of groups depending on source and season, for example, United States surface water comprises fulvic acids (45%), low molecular weight acids (25%), neutrals (15%), humic acids, bases, and contaminants (each 5%) (2). The distribution of NOM into these groups depends on whether the NOM was formed in the water body (autochthonous) or externally (allochthonous). In temperate zones the allochthonous sources are very important; they may be derived from lignin and are high in aromatic (particularly phenolic) carbon (5). Lignin, the second most abundant biopolymer in the plant kingdom, is a three-dimensional aromatic polymer of aryl propane structural units which cement intercellulose microfibrils together (6). There is a growing body of evidence suggesting that the aromatic parts of wood and nonwoody plants are the precursors of soil humic acids (6). The formation of aromatic hydroxy carboxylic acids and aldehydes during UV (254 nm) irradiation of lake NOM (7), and lignin phenols after sunlight irradiation of river NOM (8), suggests that lignin or lignin-derived compounds are important constituents of aquatic NOM. Size exclusion chromatograms of NOM treated using UV irradiation (912), ozonation, or chlorination (11-13) show that large NOM molecules fragment into smaller ones, suggesting a polymeric structure. The aim of this work was to compare the fragmentation patterns, as determined by using high performance size exclusion chromatography (HPSEC), of NOM chromophores caused by several chemical and physical oxidative techniques and to investigate the feasibility of mathematically modeling these processes as depolymerization reactions. UV irradiation (UV A, B, and C), the electrophiles O3, HOCl, and •OH (formed by irradiation at 185 nm in addition to 254 nm or UV/H2O2 processes), and enzymes produced in vitro by the white rot fungus Phanerochaete chrysosporium were used to fragment the NOM.
Methods NOM Samples. Four NOM samples were used in this study: two natural waters sourced from reservoirs located at Myponga, South Australia, and Upper Stony Creek (East Moorabool), Victoria, and the two NOM isolates, Aldrich Humic Acid and Hope Valley Magnetic Ion Exchange Resin (MIEX) Concentrate. The natural waters, properties of which are shown in Table 1, were used after membrane filtration (0.45 µm). Hope Valley Reservoir NOM was concentrated on MIEX resin (containing quaternary ammonium functional groups), eluted using 10% sodium chloride, the extract was concentrated by reverse osmosis to 30 g‚L-1, and the liquid was decanted after any salts had crystallized. Prior to use this was diluted to the concentrations shown in Table 1 using distilled or high purity water. Humic acid (Aldrich H1675-2, Lot 01816HH, extracted from brown coal) was dissolved in high purity water adjusted to pH 11 using NaOH, neutralized using HCl, and filtered (0.45 µm) prior to use at 6 mgC‚L-1. High purity water was produced by reverse osmosis and electrodeionization treatments (Millipore Elix10) followed by UV (254 and 185 nm) oxidation, adsorption, and ultrafiltration (Millipore Milli-Q Gr A10). Analytical Methods. Molecular size distribution was determined using high performance size exclusion chro10.1021/es049604j CCC: $27.50
2004 American Chemical Society Published on Web 05/11/2004
TABLE 1. Characteristics of the Aqueous NOM Samples Used in This Study sample Aldrich humic acid Hope Valley MIEX concentrate Myponga natural water East Moorabool natural water a
Calculated from supplier data.
b
DOC (mgC‚L-1)
A254 (cm-1)
pH
Fe (mg.L-1)
Mw (daltons)
Mn (daltons)
polydispersivity (Mw/Mn)
6 8
0.520 0.312
7.0 6.3
0.02a NDb
2060 1800
740 1140
2.8 1.57
14 12
0.552 0.457
7.8 7.7
0.3 1.1
1413 1180
730 790
2.0 1.5
ND - not determined.
matography (HPSEC). The operating system consisted of a Waters 2690 Alliance system with temperature controlled oven (30 °C) and Waters 996 Photodiode Array detector. A Shodex KW 802.5 glycol functionalized silica gel column, an isocratic flowrate of 1.0 mL‚min-1, and an injection volume of 100 µL of 0.22 µm filtered sample were used. The carrier solvent consisted of 0.02 M phosphate buffer at pH 6.8 adjusted to an ionic strength of 0.1 M with sodium chloride. Detection was by absorbance at 260 nm; this wavelength was chosen because substituted aromatics (e.g., phenol, aniline, anisole, benzoic acid) have peaks in their absorption spectra due to π f p* electron transitions at similar wavelengths (269-280 nm) (14), representative of similar structures in NOM. Also at this wavelength absorption by inorganics (such as nitrate) is negligible. In this paper chromophores refer to compounds absorbing at 260 nm. Polystyrene sulfonate standards were used to determine the relationship between apparent molecular weight and retention time, which, for example, was Log10Mi ) -0.377.ti + 7.12 when the N-lamp treated Myponga sample was analyzed. Dissolved organic carbon was determined using photocatalytic oxidation in an AnaTOC Series 2 Analyzer (SGE). Treatments. Table 2 details the matrix of samples and treatments used. Two UV reactors were utilized. For all Hope Valley MIEX experiments the reactor and lamps previously detailed by Parkinson (15) were used. This reactor was configured to allow three different UV wavelengths to be used: UV A (1.14 W‚L-1), B (1.34 W‚L-1), and C (1.62 W‚L-1) applied through an area of 454 cm2. For all other irradiation experiments an annular UV reactor with a working volume of 0.9 L, irradiated area of 322 cm2, and path length 1.94 cm was employed. In this reactor radiation from two lamps with identical physical dimensions was applied: one lamp with output primarily at 254 nm referred to hereafter as the “Nlamp” (Australian Ultra Violet Services G36T15NU) and the “H-lamp” with output at 254 and 185 nm (Australian Ultra Violet Services G36T15HU). The total absorbed light intensity, as determined by H2O2 actinometry (16), for the N-lamp was 1.7 × 10-5 einstein‚L-1‚s-1 or 7.9 W‚L-1 for an input electrical power of 39 W. The intensity of 185 nm radiation emitted by the H-lamp was measured using methanol actinometry (17) as 2.0 × 10-6 einstein‚L-1‚s-1 or 1.3 W‚L-1. The power draw of the H-lamp was quoted by the vendor as 46 W. During all experiments the air or nitrogen circulating past the lamp was maintained at 28 ( 1 °C and the water sample at 23 ( 1 °C by a peltier cooled water stream. The samples were mixed and aerated by humidified air during irradiation. For enhanced photooxidation processes hydrogen peroxide (H2O2) was dosed to give concentrations of 10 ((1) and 50 mg‚L-1, into the natural waters and Hope Valley MIEX solution, respectively. Hydrogen peroxide concentrations were determined by horseradish peroxidase catalyzed oxidation of DPD (18). Ozone was generated using 185 nm radiation from the H-lamp mounted in an empty reactor. Various doses were applied to a maximum of 6 mgO3‚mgC-1, estimated using iodometric titration (4). Various chlorine doses were applied to the water samples, incubated at 20 °C for 7 days and tested for residual chlorine.
To avoid contamination problems a quenching agent was not used, and only those samples without residual, corresponding to doses e 2 mgCl2‚mgC-1, were analyzed by HPSEC. MIEX concentrate diluted to 30 mgC‚L-1 inoculated with Phanerochaete chrysosporium (ATCC 34541) was incubated at 37 °C in shake flasks at 120 rpm for up to 16 days. Waksman’s special medium for counting soil fungi (19), without peptone, was used as a growth medium.
Results and Discussion Fragmentation Patterns. Three distinct fragmentation patterns were observed (Figure 1), and the behavior of East Moorabool NOM (20) was very similar to that of Myponga NOM. The removal of NOM chromophores by ozone was nonsize specific, whereas UV (g254 nm) irradiation and P. chrysosporium selectively removed larger molecular weight chromophores. The hydroxyl radical processes (combined 185/254 nm or UV/H2O2 processes) and chlorination also favored fragmentation of higher molecular weight material but not as strongly as UV (g254 nm) irradiation. For the ozone doses applied, the absorbance at 260 nm decreased without significant concomitant decrease in DOC (Figure 1 and Table 2), demonstrating removal of conjugated double bonds with minimal mineralization. Ozone was surprisingly nonsize specific (Figure 1). Frimmel et al. (13) also observed that ozone decreased the absorbance of Ruhr river water, but no chromophoric size fraction was preferentially removed; however, by using DOC detection they found decreases in the high molecular weight fractions with concomitant increases in the lower molecular weight fractions. Ozone reacts directly and indirectly (via •OH) with organic compounds in aqueous solution (4). For example, for each mole of phenol three moles of ozone are consumed, ring opening by the first ozone molecule being rate limiting to form an olefinic product (e.g., muconic acid) which reacts rapidly with the other two molecules of ozone to form less UV-absorbing products (e.g., a carboxylic acid and a carbonyl) (21). If this mechanism applies to NOM-ozone reactions, the higher reactivity of the ring-opened products would explain why chromophores did not accumulate in the smaller size fractions. Aqueous chlorine decreased the absorbance of Myponga NOM without significant reduction in the DOC concentration (Figure 1 and Table 2), demonstrating removal of conjugated double bonds with minimal mineralization. The removal of higher molecular weight chromophores appeared to be favored (Figure 1). Huber and Frimmel (11) and Frimmel et al. (13) reported little chromophoric size specificity, and DOC detection indicated formation of lower molecular weight material. This appears to be consistent with the mechanism of chlorine reaction: halogens react with activated (amine, hydroxy, or methoxy substituted) aromatics via electrophilic substitution and oxidation reactions which in some cases may lead to ring opening followed by further fragmentation (22). In this study the hydroxyl radical processes removed chromophores, mineralized NOM, and appeared to preferVOL. 38, NO. 12, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3361
TABLE 2. Treatment/Sample Combinations Used in This Study sample
treatment
Aldrich humic acid
N-lamp
Hope Valley MIEX
UV A
UV B
UV C
UV C with 50 mg/LH2O2
Myponga
P. chrysosporium N-lamp
N-lamp with 10 mgH2O2/L
H-lamp
H-lamp with 10 mgH2O2/L
chlorine ozone East Moorabool
N-lamp
N-lamp with 10 mgH2O2/L
H-lamp
H-lamp with 10 mgH2O2/L
time (h) fluence at 254 nm (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence (kJ‚m-2) time (days) time (h) fluence at 254 nm (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence at 254 nm (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence at 254 nm (kJ‚m-2) fluence at 185 nm (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence at 254 nm (kJ‚m-2) fluence at 185 nm (kJ‚m-2) DOC (mgC‚L-1) dose (mg‚L-1) DOC (mgC‚L-1) dose (mg‚L-1) DOC (mgC‚L-1) time (h) fluence at 254 nm (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence at 254 nm (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence at 254 nm (kJ‚m-2) fluence at 185 nm (kJ‚m-2) DOC (mgC‚L-1) time (h) fluence at 254 nm (kJ‚m-2) fluence at 185 nm (kJ‚m-2) DOC (mgC‚L-1)
entially oxidize the higher molecular weight material (Figure 1 and Table 2). Hydroxyl radicals react rapidly with aromatics, forming intermediate hydroxycyclohexadienyl-type radicals. These react rapidly with oxygen to form peroxyl radicals which either eliminate perhydroxyl radicals leaving a phenolic type compound or if an endoperoxide forms the aromatic ring can be opened (23). P. chrysosporium seemed to preferentially remove higher molecular weight chromophores with apparent overall increases in lower molecular weight material (Figures 1 and 2). This result suggested a depolymerization type reaction mechanism, where the extracellular enzymes preferentially broke certain bonds in larger molecules and hence formed smaller molecules. Ligninase (lignin peroxidase) catalyzes the one electron oxidation of activated (methoxyl substituted) aromatics forming an unstable cation radical which, if the substrate is lignin, undergoes further nonenzymic reactions resulting in CR-Cβ cleavage (24). This is significantly different from the chemical oxidation processes in which aromatic rings are opened, presumably leading to a loss of absorbance and size, whereas for ligninase-mediated degradation, the 3362
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 12, 2004
0 0 5.8 0 0 8 0 0 8 0 0 8 0 0 0 0 0 14 0 0 14.5 0 0 0 14.3 0 0 0 14 0 7 0 13.8 0 0 11 0 0 12 0 0 0 11 0 0 0 11.6
0.5 397 6.7 8 362 7.52 8 425 6.48 4 257 7.12 0.17 11 5 0.25 199 13 0.17 132 12.4 0.08 66 11 13.4 0.08 66 11 11 2 7.2 9.5 13.5 0.5 397 10.5 0.1 79 12.2 0.25 199 33 12 0.08 66 11 10.7
1 795 5.2 16 723 7.36 16 850 6.24 8 514 5.76 0.33 21 9 1 795 11.9 0.75 596 9 0.25 199 33 11.7 0.33 265 44 8 3 7.1 19 13.5 3 2385 9 0.32 252 10.7 1 795 131 7 0.25 199 33 9
3 2385 4 24 1085 7.12 24 1275 6 16 1028 4.8 0.5 32 16 4 3180 9 3 2385 4.8 1 795 131 7.1 1.5 1192 196 2 5 6.6 38 NA 5 3975 6 0.5 397 9 4 3180 523 1 0.5 397 65 7.2
5 3975 3 48 2170 NA 48 2550 NA 24 1541 4 0.75 48
64 2893 NA 64 3400 NA 48 3083 3.36 1 64
1.5 96
8 6359 6
4 3180 523 1
10 6.5 76 13.1
12 6.7
0.8 636 8.1
0.75 596 98 6.6
rings, and hence absorbance, are initially at least partially preserved while the size is reduced prior to cell uptake and metabolism. UV (g254 nm) irradiation led to chromatograms with 2-3 distinct peaks between 240 and 310 apparent molecular weight units and a sharply focused peak eluting at approximately 12 min (Figure 3). The height of this peak increased for small radiation doses for all NOM samples, and one or two of the other peaks also increased in aqueous samples prepared from Hope Valley MIEX concentrate and Aldrich humic acid. These increases may have been due to formation of products with higher molar absorptivity or by break up of larger chromophores with concomitant formation of smaller fragments, as discussed by Frimmel (12). It is unlikely that nonsize exclusion effects, such as decreased retention by electrostatic repulsion of negatively charged groups, or increased retention of hydrophobic compounds (25, 26), have affected the results because oxidation reactions are more likely to decrease hydrophobicity and increase negative charge, both effects being counter to the observed shift to longer retention times.
FIGURE 1. The size specificity of the various processes. The degree of treatment can be gauged by the area under the HPSEC curve and size specificity by the change in average retention time. Longer retention times correspond to smaller molecules that can penetrate the pores of the HPSEC column packing. Symbols show experimental data and lines indicate trends. For details of treatments see text and Table 2. of the well resolved peaks occurred at similar apparent molecular weight increments, suggesting that chromophoric NOM has repeating structural units separated by chemical bonds more susceptible to photoinduced cleavage. A mathematical model, based on trimer depolymerization kinetics, was developed to investigate this hypothesis. Description of a Depolymerization Model. The kinetics describing the sequential degradation of a trimer were adopted because this is the simplest reaction sequence that can be called depolymerization: k3X
3X 98 2X + 1X
(1)
k2X
2X 98 1X + 1X
(2)
k1X
1X 98 nonchromophoric
(3)
If first-order kinetics are assumed, performing a mass balance results in the following differential equations: FIGURE 2. Size distribution of Hope Valley MIEX NOM treated by lignolytic enzymes from P. chrysosporium.
d[3X] ) -k3X[3X] dt
(4)
Formation of products with higher molar absorptivity than their precursors is possible; the ligninase from P. chrysosporium converts dimethoxybenzene to more highly absorbing p-benzoquinone (27), and quinones have been identified as potential radical sources in white light-irradiated, alkaline, humic acid solutions (28). While this may be occurring at the molecular level, at a bulk property level the DOC has been shown to increase along with absorbance in these smaller size fractions (11, 12), suggesting that fragmentation or sequential degradation was taking place. Other studies using size exclusion chromatography (9) and product analysis (7, 8) of irradiated (UV g 254 nm) humics also indicated sequential degradation. Identification of increased concentrations of aromatic moieties in irradiated NOM samples (7, 8) may indicate that ring opening is not (initially) the dominant mechanism, fragmentation may occur elsewhere, and hence UV-absorbing groups are conserved. Furthermore, the similar size specificity of the UV-treated samples to that inoculated with P. chrysosporium (Figure 1) suggests a similar mechanism via radical cations, which are involved with NOM photochemistry (29), may apply. Several
d[2X] ) + k3X[3X] - k2X[2X] dt
(5)
d[1X] ) + k3X[3X] + 2k2X[2X] - k1X[1X] dt
(6)
Equations 5 and 6 are linear first-order differential equations that can be solved analytically using an integrating factor yielding
[3X] ) [3X]0‚exp(- k3Xt)
(7)
[2X] ) A‚exp(-k3Xt) + B‚exp(-k2Xt)
(8)
[1X] ) C‚exp(-k3Xt) + D‚exp(-k2Xt) + E‚exp(-k1Xt) (9) These equations were entered into a Microsoft Excel spreadsheet, and rate constants k3X, k2X, and k1X solved in turn by minimizing the sum of squares of the deviations between data and model using Solver. VOL. 38, NO. 12, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3363
FIGURE 3. The size distributions of UV (g254 nm) treated samples (left column) and depolymerization model kinetic plots (right column). In the kinetic plots the model fits are shown as lines and experimental data as symbols. The HPSEC chromatograms were resolved into peaks using PeakFit software (Figure 4). The extreme value 4 parameter (amplitude) tailed distribution, with width 0.12 min and tail factor of 1.2, was used based on the shapes of the elution curves of the polystyrene sulfonate standards. 3364
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 12, 2004
The sharply focused peak at 12 min required a width of 0.045 min. The featureless leading edge of the chromatogram was modeled as a series of peaks eluting at increments corresponding to the smallest resolved peak. The logarithmic relationship between apparent molecular weight and reten-
FIGURE 4. HPSEC chromatograms resolved into peaks assuming an underlying polymeric structure. Samples were irradiated using the N-lamp (254 nm). Dotted lines show sum of individual peaks. Stippled lines indicate peaks assigned to 2X category. tion time compresses the peaks associated with the higher molecular weight compounds and a good fit in this region of the chart was possible. Although somewhat conjectural, by this means the total area under the curve could be well approximated and areas corresponding to apparent molecular weights determined. The latter data were used to
calculate the “molar” concentration of the peak using
[iX] )
A260‚t Mi
(10)
the implicit assumption being that molar absorptivity is VOL. 38, NO. 12, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3365
TABLE 3. Results of Depolymerization Model Fitting - UV (g254 nm) Irradiation overall sample
R2
treatment
Aldrich humic N-lamp Hope Valley MIEX UV A (315-400 nm) UV B (280-315 nm) UV Ca Myponga N-lamp (254 nm) East Moorabool N-lamp (254 nm)
3X
Sx/y k3X (h-1)
0.978 4 0.366 10 0.915 7 0.989 7 0.958 9 0.980 9
0.80 0.041 0.10 0.53 0.84 1.45
2X
Sx/y k2X(h-1)
R2 0.933 0.917 0.972 0.999 0.950 0.999
0.8 2 2 0.3 2 0.1
0.36 0.013 0.036 0.33 0.49 0.68
R2 0.87 0.606 0.698 0.983 0.834 0.943
1X
Sx/y k1X (h-1 ) 1 2 6 2 6 4
0.26 0.006 0.03 0.21 0.37 0.54
R2
Sx/y k3X/k1X k2X/k1X
0.945 5 0.638 14 0.869 7 0.956 9 0.838 12 0.988 3
3.1 6.3 3.5 2.6 2.3 2.6
1.4 2.0 1.2 1.6 1.3 1.3
a UVC is defined as 200-280 nm but was supplied by low-pressure mercury vapor lamp approximately 90% at 254 nm with the remainder being longer wavelength or visible.
directly proportional to the apparent molecular weight. This is the same assumption made when the signal intensity from a UV absorbance detector is used to compute Mn and Mw from HPSEC chromatograms using the formulas given by Yau et al. (30). For model fitting purposes, the 2X fraction was defined so that the largest apparent molecular weight peak (in 2X fraction) when halved was approximately equal to the largest size fraction in 1X. For the Myponga NOM shown in Figure 4, apparent molecular weight increments of 5400-2040, 1700-1160, and 860-100 daltons were assigned to the 3X, 2X, and 1X categories. “Molar” concentrations of 3X, 2X, and 1X after various irradiation times were calculated by summing the “molar” concentrations of the peaks falling in these size ranges. Depolymerization Model - UV (g254 nm) Irradiation. With one exception the trimer depolymerization model gave good correlation to the experimental data (Figure 3 and Table 3). The UV A-treated MIEX concentrate model prediction did not correlate as well with the experimental data, possibly because much of the material detected at 260 nm was refractory to the longer (315-400 nm) wavelengths used. The model correlated well with the UV B data indicating that at these wavelengths (280-315 nm) there were enough chromophoric sites per large and medium sized molecule to observe substantial fragmentation. The photochemical rate equation from the laws of photochemistry (3) can be expressed as follows
d[C] ‚[C] ) - Φ‚I0λ‚ ‚F dt At(t) s
(11)
or, in words, that the rate of reaction is proportional to the product of the rate of light absorption by the chromophore and its quantum yield. The rate of light absorption is dependent on the molar absorptivity, concentration, incident intensity, and the fraction of light absorbed by the chromophore. It is expected that the molar absorptivity of NOM from natural sources increases with increasing molecular size. On initial irradiation the larger molecules will react fastest because they have higher molar absorptivities. If the chromophoric groups are conserved when the molecule fragments, then the two fragments could be expected to have a total absorbance similar to the original molecule. If the quantum yields of parent molecules and fragments were similar, then individually they would react slower than the original because of reduced molar absorptivity. Alternatively if chromophoric groups are only partly conserved, the quantum yields are similar, then the molar absorptivities of the fragments will be less than the original and react even slower. In either case the smaller, slower-reacting fragments should initially accumulate before reacting further, eventually to nonchromophoric material. Furthermore, the apparent first-order rate constants, approximately proportional to the molar absorptivity, should increase with increasing size. In 3366
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 12, 2004
FIGURE 5. Application of the “Multipool” model to photochemical bleaching of Aldrich humic acid, Myponga, and East Moorabool NOM samples. The model predictions are shown as lines and experimental data as symbols. The chart also shows model prediction for East Moorabool for each individual “pool” of chromophoric material. the e 315 nm irradiated systems the chromophores assigned to 3X reacted between 2 and 3.5 times faster than 1X and 2X between 1.2 and 2 times faster than 1X chromophores (Table 3), giving the model credibility. This dependence on molecular size, via molar absorptivity, explains the high size specificity of the UV photooxidation processes. While this model may not be the only explanation for the data, especially given the assumptions made, it supports the hypothesis that chromophoric NOM molecules have repeating structural units joined by photolabile chemical bonds. For solar irradiation, the active chromophores may be quinone groups (31); other functional groups could be involved at shorter wavelengths. Some interesting photochemical phenomena have been reported in the initial stages of irradiation; free radical concentrations are at their highest levels (28), and singlet oxygen production is at its maximum (32). Since the largest molecules react first it suggests that these molecules are responsible for generating free radicals and singlet oxygen. The decay of the absorbance of irradiated solutions of NOM can be mathematically modeled to test these ideas. Figure 5 shows that the absorbance versus time behavior of irradiated Myponga and East Moorabool water can be modeled using the “Multipool” model (32). Chromophores were divided into three groups, each independently reacting at a rate given by eq 12, which is eq 11 expressed in terms
TABLE 4. Results of Depolymerization Model Fitting - Hydroxyl Radical Processes overall sample
treatment
Hope Valley MIEX UV C/H2O2 Myponga N-lamp/H2O2 H-lamp/H2O2 H-lamp East Moorabool N-lamp/H2O2 H-lamp/H2O2 H-lamp
R2 0.978 0.895 0.968 0.976 0.942 0.939 0.962
3X
Sx/y k3X (h-1) 6 28 16 9 9 11 17
2.34 2.12 4.90 2.45 4.06 5.80 5.90
2X
1X
R2
Sx/y
k2X (h-1)
R2
0.993 0.978 0.996 0.989 0.999 0.994 0.999
0.7 1 0.6 0.8 0.2 0.7 0.02
2.27 1.66 4.55 2.12 2.49 3.84 4.14
0.973 0.939 0.987 0.955 0.944 0.956 0.999
Sx/y k1X(h-1) 2 6 3 4 3 4 0.8
3.82 2.56 8.72 2.78 7.38 12.6 5.94
R2 0.954 0.767 0.908 0.975 0.703 0.634 0.896
Sx/y k1X/k2X k3X/k2X 4 20 12 5 6 8 16
1.7 1.5 1.9 1.3 3.0 3.3 1.4
1.0 1.3 1.1 1.2 1.6 1.5 1.2
FIGURE 6. The size distributions of hydroxyl-radical treated NOM samples (left column) and depolymerization model kinetic plots (right column). In the kinetic plots the model predictions are shown as lines and experimental data as symbols. of absorbance:
-
dAi(t) Ai(t) ) iΦiIλ{1 - exp[-2.303LAt(t)]} (12) dt At(t)
The differential equations were solved using numerical methods. The majority of the chromophores fell into the slowest reacting group with a smaller quantity reacting about 10 times faster. To model the rapid initial absorbance drop of the natural waters a very small group of chromophores was predicted to react about 90 times faster than the bulk of the material. The size exclusion chromatograms showed that the largest molecular weight material reacted the quickest
and suggest the chromophoric reaction centers, with disproportionately large quantum yield × molar absorptivity products, were associated with the higher molecular weight fractions. For the humic acid sample this fast reacting group was absent (Figure 5). Depolymerization Model - Hydroxyl Radical Processes. The depolymerization model was also successfully applied to the data for the hydroxyl radical processes. The only sample treated with hydroxyl radicals which showed an apparent increase in a particular size fraction compared with the initial sample was Hope Valley MIEX irradiated in the presence of H2O2, which suggested that a depolymerization mechanism was responsible. For all other samples treated by processes VOL. 38, NO. 12, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3367
involving hydroxyl radicals the signal intensity was less than for the raw water at all retention times. For depolymerization without accumulation of smaller molecular weight material to occur, the smaller molecular weight material must have reacted as fast as or faster than it was formed. The alternative, that chromophoric material was bleached while the apparent molecular weight remained the same, is unlikely given the energy required to remove conjugated double bonds without fragmentation and would require a very specific oxidizing agent. Using the criterion set out above, the depolymerization model was applied to the data with reasonable correlation (Table 4 and Figure 6). As expected, 1X rate constants were the largest, followed by 3X, with 2X the smallest. Since UV (254 nm) irradiation was involved in all of the hydroxyl radical processes the high rates of 3X removal may have been partly due to direct photolysis. With the exception of the East Moorabool UV/H2O2 results, the differences between the rate constants were not as marked as for the UV (g254 nm) photooxidation systems, indicating that hydroxyl radical is less size specific than UV irradiation alone. Furthermore, the ratios of the rate constants for UV/H2O2 treatments were similar for Hope Valley MIEX and Myponga and also similar for Myponga and East Moorabool treated with the H-lamp. The 1X chromophores in the East Moorabool samples reacted about three times faster than 2X in the UV/H2O2 treatments; this water had a high iron content, and iron has been shown to accelerate the photooxidation of NOM (33), but how this might explain the result is unclear. The depolymerization model was used to interpret the fragmentation behavior of the chromophores caused by hydroxyl radical attack and UV photodegradation. More confidence can be placed on the modeling of photodegradation, because the result is in agreement with photochemical rate theory. While the model may not be the only explanation, the fact the model explains the photoinduced degradation data so well implies that at least some of the chromophoric groups are conserved and fragmentation of nearby weak links is occurring faster than ring opening. The mathematical model illuminates the previously unseen regularity in the structure of the chromophoric NOM samples studied.
Acknowledgments The authors wish to thank the CRC for Water Quality and Treatment for research funding and scholarship support, the Australian Water Quality Centre (AWQC) for HPSEC analyses, Farhad Younos (RMIT) for preparing the P. chrysosporium treated samples, Jim Morran (AWQC) for preparing and donating the MIEX concentrate, Associate Professor Malcolm Hobday (RMIT) for fruitful discussions, and the anonymous reviewers whose comments contributed to this paper.
Nomenclature A
A260‚t Ai(t)
B
C
3368
9
coefficient in eq 8 (M) k3X[3X]0 ) (-k3X + k2X) the area under the peak for component eluting at time ti absorbance divided by spectrophotometer path length at wavelength λ (cm-1), subscript i refers to chromophore pool i and subscript t to total coefficient in eq 8 (M) [2X]0(-k3X + k2X) - k3X[3X]0 ) -k3X + k2X coefficient in eq 9 (M) k3X[3X]0 + 2k3X‚A ) k1X - k3X ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 12, 2004
[C] D
E
Fs hi
I0λ [iX] K knX L Mi
Mn
concentration at time t (M), coefficient in eq 9 (M) 2k2X‚B ) k1X - k2X coefficient in eq 9 (M) k3X[3X]0 + 2k2XA 2k2X‚B ) [1X]0 k1X - k3X k1X - k2X fraction of incident light absorbed by the system (dimensionless). height of peak on chromatogram, with A260 detection and not mass detection as originally intended by Yau et al. (30) (see text) incident radiation intensity at wavelength λ (einsteins‚s-1) “molar” concentration of peak eluting at time ti number of variables used to describe the data rate constant for removal of size fraction nX (h-1) reactor path length (cm) the apparent molecular weight corresponding to time ti (i.e., equal to that of polystyrene sulfonate standard eluting at time ti) the number average molecular weight (30)
∑i)1 h N
) Mw
i
∑i)1 M
hi
N
i
the weight average molecular weight (30)
∑i)1 (h M ) ) ∑i)1 h N
i
i
N
N [nX]0 R2 Sr St Sy/x
i
number of data points initial concentration of nX (M) nonlinear correlation coefficient (34) ) 1 - Sr/St sum of square of residuals sum of squares about mean estimate of error of estimate
)
T ti η λ Φ
x
St
n-k time (h-1 or s-1) retention time of peak i (min-1) molar absorptivity (M-1‚cm-1) wavelength (nm) quantum yield (moles reacted per einstein)
Literature Cited (1) Hem, J. D. Study and Interpretation of the Chemical Characteristics of Natural Water; U.S. Govt Printing Office: Washington, 1992. (2) Malcolm, R. L. In Humic Substances in Soil, Sediment, and Water. Geochemistry, Isolation, and Characterization; Aiken, G. R., McKnight, D. M., Wershaw, R. L., MacCarthy, P., Eds.; John Wiley and Sons: New York, 1985; pp 181-209. (3) Leifer, A. The Kinetics of Environmental Aquatic Photochemistry; American Chemical Society: Washington, DC, 1988.
(4) Gottschalk, C.; Libra, J. A.; Saupe, A. Ozonation of Water and Wastewater; Wiley-VCH: Weinheim, 2000. (5) Aiken, G. R.; Cotsaris, E. J.-Am. Water Works Assoc. 1995, 87(1), 36-45. (6) Shevchenko, S. M.; Bailey, G. W. Crit. Rev. Environ. Sci. Technol. 1996, 26(2), 95-153. (7) Corin, N.; Backlund, P.; Kulovaara, M. Chemosphere 1996, 33(2), 245-255. (8) Opsahl, S.; Benner, R. Limnol. Oceanogr. 1998, 43(6), 12971304. (9) Parkinson, A. Evaluation of the Effects of Photooxidation and Enhanced Photooxidation on Aquatic NOM; Department of Chemical and Metallurgical Engineering, RMIT University: Melbourne, 2001. (10) Backlund, P. Chemosphere 1992, 25(12), 1869-1878. (11) Huber, S. A.; Frimmel, F. H. Intern. J. Environ. Anal. Chem. 1992, 49, 49-57. (12) Frimmel, F. H. Environ. Int. 1998, 24(5/6), 559-571. (13) Frimmel, F. H.; Hesse, S.; Kleiser, G. In Natural Organic Matter and Disinfection Byproducts - Characterisation and Control in Drinking Water; ACS Symposium Series 761; Barrett, S. E., Krasner, S. W., G. L., A., Eds.; American Chemical Society: Washington, DC, 2000; pp 84-95. (14) Lambert, J. B.; Shurvell, H. F.; Lightner, D. A.; Cooks, R. G. Organic Structural Spectroscopy; Prentice Hall: Upper Saddle River, NJ, 1998. (15) Parkinson, A.; Barry, M. J.; Roddick, F. A.; Hobday, M. D. Water Res. 2001, 35(15), 3656-3664. (16) Beltra´n, F. J.; Ovejero, G.; Garcia-Araya, J. F.; Rivas, J. Ind. Eng. Chem. Res. 1995, 34, 1607-1615. (17) Heit, G.; Neuner, A.; Saugy, P.-Y.; Braun, A. M. J. Phys. Chem. A 1998, 102, 5551-5561. (18) Bader, H.; Sturzenegger, V.; Hoigne´, J. Water Res. 1988, 22(9), 1109-1115. (19) Booth, C. Methods Microbiol. 1971, 4, 91. (20) Thomson, J.; Roddick, F. A.; Drikas, M. Water Sci. Technol. 2003, 2(5-6), 435-443.
(21) Bailey, P. S. Ozonation in Organic Chemistry. Volume II Nonolefinic Compounds; Academic Press: New York, 1982. (22) Rook, J. J. In Water Chlorination: Environmental Impact and Health Effects Vol. 3 Proceedings of the Third Conference on Water Chlorination: Environmental Impact and Health Effects, Colorado Springs, Colorado, October 28th-November 2nd; Jolley, R. L., Brungs, W. A., Cumming, R. B., Jacobs, V. A., Eds.; Ann Arbor Science: MI, 1979; pp 85-103. (23) von Sonntag, C. J. Water SRT- Aqua 1996, 45(2), 84-91. (24) Kirk, T. K.; Farrell, R. L. Annu. Rev. Microbiol. 1987, 41, 465505. (25) Perminova, I. V.; Frimmel, F. H.; Kovalevskii, D. V.; Abbt-Braun, G.; Kudryavtsev, A. V.; Hesse, S. Water Res. 1998, 32(3), 872881. (26) Pelekani, C.; Newcombe, G.; Snoeyink, V. L.; Hepplewhite, C.; Assemi, S.; Beckett, R. Environ. Sci. Technol. 1999, 33, 28072813. (27) Kersten, P. J.; Tien, M.; Kalyanaraman, B.; Kirk, T. K. J. Biol. Chem. 1985, 260(5), 2609-2612. (28) Senesi, N.; Schnitzer, M. Soil Sci. 1977, 123(4), 224-234. (29) Cooper, W. J.; Zika, R. G.; Petasne, R. G.; Fischer, A. M. In Aquatic Humic Substances; American Chemical Society: Washington, DC, 1989; pp 333-362. (30) Yau, W. W.; Kirkland, J. J.; Bly, D. D. In High Performance Liquid Chromatography; Brown, P. R., Hartwick, R. A., Eds.; John Wiley & Sons: New York, 1979; pp 277-316. (31) Slawinska, D.; Slawinski, J.; Sarna, T. J. Soil Sci. 1975, 26(2), 93-99. (32) Andrews, S. S.; Caron, S.; Zafiriou, O. C. Limnol. Oceanogr. 2000, 45(2), 267-277. (33) Gao, H.; Zepp, R. G. Environ. Sci. Technol. 1998, 32, 2940-2946. (34) Chapra, S. C.; Canale, R. P. Numerical Methods for Engineers; 2nd ed.; McGraw-Hill: New York, 1989.
Received for review March 14, 2004. Accepted March 22, 2004. ES049604J
VOL. 38, NO. 12, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
3369
