w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/watres
Identifying the factors that influence the reactivity of effluent organic matter with hydroxyl radicals Olya S. Keen a,c,*, Garrett McKay b, Stephen P. Mezyk b, Karl G. Linden c, Fernando L. Rosario-Ortiz c a
Department of Civil and Environmental Engineering, University of North Carolina, 9201 University City Blvd., Charlotte, NC 28223, USA b Department of Chemistry and Biochemistry, California State University at Long Beach, 1250 Bellflower Blvd., Long Beach, CA 90840, USA c Department of Civil, Environmental and Architectural Engineering, 428 UCB, University of Colorado at Boulder, Boulder, CO 80309, USA
article info
abstract
Article history:
Advanced oxidation processes (AOPs) are an effective treatment technology for the
Received 8 May 2013
removal of a variety of organic pollutants in both water and wastewater treatment.
Received in revised form
However, many background constituents in water are highly reactive towards hydroxyl
15 October 2013
radicals (HO) and decrease the efficiency of the process towards contaminant oxidation.
Accepted 21 October 2013
Up to 95% of the HO scavenging can come from dissolved organic matter (OM). In this
Available online 1 November 2013
study, 28 wastewater effluent samples were analyzed to find correlations between the reactivity of HO with wastewater-derived OM (known as effluent organic matter, EfOM),
Keywords:
water quality parameters, treatment train characteristics, and fluorescence-derived data.
Effluent organic matter
Rate constants for the reaction between HO and EfOM (kEfOM-HO) were measured using a
Hydroxyl radical
bench scale UV-based AOP system with methylene blue as an HO probe and confirmed
Advanced oxidation
using an electron pulse radiolysis method for a subset of the samples. The EfOM was
Wastewater treatment
characterized using a series of physicochemical parameters, including polarity, average molecular size and fluorescence. The kinetic data were analyzed with principal component analysis and Akaike Information Criterion. Four predictors were identified as dominant: chemical oxygen demand, retention onto NH2 extraction medium, fluorescence index, and total organic carbon. These four variables accounted for approximately 62% of the variability in the value of kEfOM-HO The average kEfOM-HO value for EfOM in this study was 1 8 1 1 value determined for 2.5 108 M1 C s , which is about 31% lower than the 3.6 10 MC s
natural organic matter isolates and commonly used in AOP modeling. ª 2013 Elsevier Ltd. All rights reserved.
1.
Introduction
Advanced oxidation processes (AOPs) have been demonstrated as an efficient treatment technology for the oxidation
of a variety of organic contaminants (Huber et al., 2003; Gogate and Pandit, 2004; Rosenfeldt and Linden, 2004; Rosario-Ortiz et al., 2010). Its efficacy arises from the highly reactive and non-selective nature of the hydroxyl radicals (HO). HO reacts
* Corresponding author. Department of Civil and Environmental Engineering, University of North Carolina, 9201 University City Blvd., Charlotte, NC 28223, USA. Tel.: þ1 704 687 5048. E-mail address: [email protected] (O.S. Keen). 0043-1354/$ e see front matter ª 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.watres.2013.10.049
w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
List of abbreviations: AOP HO OM EfOM TOC DOC UV MB PCA AIC RC COD SRT SUVA254 MW d FI RI PS HIX
advanced oxidation process hydroxyl radical organic matter effluent organic matter total organic carbon dissolved organic carbon ultraviolet methylene blue principal component analysis Akaike Information Criterion retention coefficient chemical oxygen demand solids retention time specific UV absorbance at 254 nm molecular weight (weight average) dispersity fluorescence index redox index protein-like fluorescence signal humification index
with organic contaminants through different mechanisms, including hydrogen abstraction and addition to electron-rich sites, with reaction rate constants on the order of 10810 M1 s1 being reported for the reactions with many organic species (Buxton et al., 1988). Although the high reactivity of HO offers the benefit of oxidizing mixtures of organic compounds, it also has the drawback of reacting with the background water quality components. Carbonate species and dissolved organic matter (OM) are typically the primary HO scavengers in natural water samples. The application of AOPs for the oxidation of organic contaminants has been studied as a potential municipal wastewater treatment process for water reuse applications (Huber et al., 2003; Rosario-Ortiz et al., 2010; Keen et al., 2012a, 2012b). Typically, AOP efficiencies for organic compounds vary from no removal to complete oxidation, and are a function of the oxidation kinetics of the compounds and the scavenging capacity of the water matrix (Rosario-Ortiz et al., 2010; Keen et al., 2012b). In AOP systems, only a fraction of HO will react with the trace contaminants with the majority of the HO reacting with background scavengers. The consumption of HO via non-specific reactions is the scavenging capacity of the water matrix. Because of the high level of HO scavenging in wastewater matrices, AOPs are often viewed as not economical for wastewater treatment. However, some of the recent studies show that it may not be necessary to achieve full mineralization (Keen et al., 2012a). Combined with better ways to predict the process performance, AOPs may become a viable option for wastewater treatment. In wastewater treatment plant effluents, the main HO scavenger is effluent organic matter (EfOM), due to both its s1 moderate reactivity with HO on the order of 108 M1 C (Westerhoff et al., 1999; Rosario-Ortiz et al., 2008) and its relatively high concentrations (measured as total organic carbon-TOC). While natural OM isolates and standards were shown to have a relatively constant value for the reaction rate
SC TNF DN FE
409
secondary clarifier trickling nitrifying filter denitrifying filter final effluent
List of symbols: k reaction rate constant kEfOM-HO reaction rate constant between effluent organic 1 matter (EfOM) and HO, M1 C s , where MC is the molar concentration of organic carbon kHO,MB reaction rate constant between methylene blue (MB) and HO, M1 s1 E0 average fluence rate, mW/cm2 F fluence, mJ/cm2 ks HO reaction rate constant for a given scavenging compound, M1 s1 [S] concentration of a scavenging compound, M molar absorption coefficient at 254 nm, M1 cm1 ε254 F quantum yield of photolysis, e 254 nm wavelength energy, J/mol U254
constant with HO (kOM-HO) (Westerhoff et al., 1999), the reactivity of EfOM (kEfOM-HO) has been shown to have variable reactivity with the reported range of values from 1.39 108 to 1 across various studies (Rosario-Ortiz et al., 11.5 108 M1 C s 2008; Westerhoff et al., 2007; McKay et al., 2011; Katsoyiannis et al., 2011; Nagarnaik and Boulanger, 2011) with up to a factor of 4.5 difference in a single study (Rosario-Ortiz et al., 2008). The variable nature of kEfOM-HO has been attributed in part to differences in molecular weight composition across different samples (Westerhoff et al., 2007; Dong et al., 2010). Among other parameters that were previously suggested to influence reactivity of OM towards HO are specific UV absorbance (absorbance per mg/L of organic carbon) at 254 nm (Westerhoff et al., 1999), polarity measures and fluorescence index (Rosario-Ortiz et al., 2008). These parameters carry information about the structural composition and prevalence of certain functional groups within the bulk OM, such as aromatic rings or hydrophilic functional groups. For example, FI of the sample tends to be higher when more products of microbial activity (more aliphatic in structure) are present (McKnight et al., 2001). The main objective of this study was to conduct a statistical evaluation of the physicochemical properties of EfOM and of wastewater treatment process variables and their role in the reactivity of EfOM towards HO. In full-scale AOP systems, it is difficult to measure the concentration of HO in situ or in real time because of its low concentration. Therefore, engineers rely on calculations and safety factors when designing these treatment processes. Part of this study’s objective was to develop a model to predict the values of kEfOMHO to offer a full-scale reactor design tool that could increase confidence in the prediction of performance, and therefore minimize the safety factors and eventual costs associated with AOP treatment. A number of parameters were evaluated in this study for their ability to predict the value of kEfOM-HO. Those parameters
410
w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
included bulk properties of effluent organic matter (average molecular weight and its dispersity, and polarity), water quality parameters (chemical oxygen demand, specific UV absorbance at 254 nm, TOC), treatment train characteristics (solids retention time in the biological treatment process), and fluorescence derived data (fluorescence index, humification index, intensity of the protein-like peaks and the fraction of signal attributed to the oxidized vs. reduced quinones). Each of the parameters considered for the model was expected to influence the reactivity of EfOM towards HO, and the choice of each is discussed in detail in a corresponding section. The ability of these parameters to account for the variability of the reaction rate constants across 28 wastewater effluent samples was evaluated with principal component analysis (PCA). A model was fitted to the best predictors identified by the Akaike Information Criterion (Akaike, 1974) and is presented in this paper.
2.
Materials and methods
2.1.
Effluent samples
The samples were collected from 8 different wastewater treatment utilities, 6 of which were sampled on several separate occasions. The facilities were located mainly in Colorado with only 4 of the 28 samples from other states (Florida and California). This may be a limitation of the sample set because of the differences in source water. All utilities were sampled at the point in the treatment train right before disinfection. This sampling location is representative of the location of an AOP process in the treatment train. One of the utilities was sampled at several different points along the treatment train to determine how different biological processes affect the reactivity of EfOM. The sampling points included effluents from the secondary clarifier following the solids contact activated sludge, from the nitrifying trickling
Table 1 e Wastewater treatment trains of the participating utilities. Utility
1 2 3
4 5 6
7 8
Treatment process
Activated sludge with biological nutrient removal Activated sludge with biological nutrient removal Solids contact activated sludge þ nitrifying trickling filters þ denitrifying filters Activated sludge with biological nutrient removal Activated sludge with biological nutrient removal Activated sludge with biological nutrient removal þ denitrifying filters þ sand filters Extended aeration lagoon Membrane bioreactor
Average Solids plant flow, retention time, MGD days 12
12
80
5
23
0.7
55
10
6.3
10
22.5
15
0.04 0.014
100 e
filter, from the denitrifying trickling filter and the final effluent before disinfection consisting of a blend of water that passed through the denitrifying filters and the water that bypassed the denitrification. Table 1 summarizes the utilities and their corresponding treatment processes. All samples were filtered after collection through a 0.45 mm nylon filter (Nalgene, Rochester, NY) and stored at 4 C to minimize any further bacterial processes. Due to the large number of samples in this study, immediate analysis was not always possible. Most samples were analyzed within 10e15 weeks from collection. Water quality parameters (TOC, alkalinity, ammonia, nitrite and nitrate) were measured immediately after sample collection and immediately prior to measuring kEfOM-HO. Samples that showed signs of microbial activity (e.g. changes in inorganic nitrogen species or drop in TOC) were excluded from the study. Nitrate, nitrite and ammonia concentrations were measured using a Hach DR5000 spectrophotometer (Hach Corporation, Loveland, CO) and the corresponding Hach kits (TNT835, TNT839 and TNT830 respectively). Dissolved organic carbon (DOC) was measured using Shimadzu VCSH TOC analyzer (Shimadzu Scientific Instruments, Columbia, MD). Inorganic carbon was eliminated from the DOC samples by lowering the pH to about 2.0 with hydrochloric acid (Mallinckrodt, Hazelwood, MO), after which the sample was purged with ultra-zero air during the analysis. The sample pH was measured with a calibrated pH meter (F340, Beckman Coulter, Indianapolis, IN), alkalinity was measured with a Hach digital titrator (Hach Corporation, Loveland, CO), and absorbance scans were taken with Cary100Bio spectrophotometer (Agilent Technologies, Santa Clara, CA).
2.2.
Measurement of kEfOM-HO
2.2.1.
UV/H2O2 with methylene blue
Experiments measuring the overall HO scavenging of the water samples were performed with 5 mM of reagent-grade methylene blue (SigmaeAldrich, St.Louis, MO) as a probe. Use of methylene blue as an HO probe was described previously for Fenton reaction (Satoh et al., 2007) and for photocatalysis applications (Luo et al., 2009). It was determined to be an appropriate probe for this study as well based on two factors: (a) it does not degrade by direct photolysis at wavelengths >200 nm, and (b) it can be analyzed spectrophotometrically allowing fast and inexpensive processing of multiple samples. Hydroxyl radicals were generated by UV exposure of samples containing 10 mg/L of reagent-grade hydrogen peroxide (J.T.Baker, Phillipsburg, NJ). A low-pressure mercury vapor UV lamp emitting around 253.7 nm and housed in a collimated beam apparatus was used in the experiment. The experiments were performed in stirred crystallization dishes 5 cm in diameter with a Petri factor of 0.97. The average UV dose delivered to the sample was calculated using all appropriate correction factors (Bolton and Linden, 2003). Hydrogen peroxide was measured by the triiodide method (Klassen et al., 1994) prior to adding methylene blue to the sample to avoid interference. Methylene blue was measured spectrophotometrically at 664 nm. Although methylene blue is a photostable compound under UV > 200 nm, a blank run without hydrogen peroxide
411
w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
was performed to account for non-HO reaction pathways in the wastewater matrix. The difference between the sample and the blank was used to measure the extent of the probe degradation due to hydroxyl radicals alone. The concentration of steady-state HO ([HO]SS) was then calculated using the following relationship: [1]
where E0 is the average fluence rate (mW/cm2), F is the fluence (mJ/cm2) and kHO,MB is a time-based reaction rate constant between probe methylene blue (MB) and hydroxyl radicals (M1 s1). In this equation, the quantity kHO,MB [HO]ss/E0 is the slope of the plot of ln[MB]t/[MB]0 vs F, and [HO] can be calculated as ½HOss ¼
slope$E0 kHO;MB
[2]
The value for kHO,MB has been reported as 2.1$1010 M1 s1(Buxton et al., 1988). The UV/H2O2 model by Glaze et al. (1995) was then rearranged to calculate the total HO scavenging coming from the sample background: X
ks ½S ¼
E0 ε254 F½H2 O2 1 $ U254 ½HOss
ks ½S ¼
E0 ε254 F½H2 O2 kHO;MB ε254 F½H2 O2 kHO;MB $ ¼ $ U254 U254 slope$E0 slope
a b
8.5 3.9 2.7 1.0 3.6
106a 108a 107a 1010a 108b
Buxton et al., 1988. Westerhoff et al., 1999.
1 range of values for kEfOM-HO was (1.6e3.3) 108 M1 (>100% C s difference), which indicates that predicting the value well beyond the experimental noise is possible. EfOM accounted for 41e60% of scavenging in the experiments when H2O2 and the probe were taken into account, so the difference between the measured total scavenging and the calculated scavenging by the compounds other than EfOM was quantifiable. In the full-scale settings the fraction of scavenging attributable to EfOM would be even greater (45e95%) because the probe compound would not be affecting the calculations (Fig. 1). Therefore, EfOM is highly influential in the process performance.
[3]
2.2.2.
where ks is the HO reaction rate constant for a given scavenging compound (M1 s1), [S] is the concentration of the corresponding scavenging compound (M), ε254 is the molar absorption of hydrogen peroxide at 254 nm (M1 cm1), F is the quantum yield of hydroxyl radical formation by photolysis of hydrogen peroxide at 254 nm, [H2O2] is the concentration of hydrogen peroxide (M), and U254 is the wavelength energy (J mol1) Substituting Eq. (2) into Eq. (3) yields X
HCO 3 CO2 3 H2O2 NO 2 OM
kS-HO, M1 s1
[4]
Typical scavengers in wastewater matrices are EfOM, bicarbonate and occasionally nitrite, although the levels of nitrite at most properly operating wastewater treatment plants are very low. If hydrogen peroxide is added for HO generation, then hydrogen peroxide is a scavenger as well. Table 2 summarizes the most prominent HO scavengers in effluents along with the corresponding kHO values. Carbonate and ammonia are also scavengers of HO, however all of the effluents had pH values between 6.3 and 7.4, making the concentrations of either species negligible (aA2and aBH values 0.001e0.00005 and 0.01e0.001 respectively). The theoretical contribution by the scavenging compounds other than EfOM was calculated using the published values for the reaction rates between HO and bicarbonate, nitrite, hydrogen peroxide and methylene blue (Table 2). The difference between the measured scavenging (Eq. (4)) and the calculated scavenging by compounds other than EfOM was assumed to be the scavenging by EfOM. The EfOM scavenging value was then divided by the organic carbon concentration to find kEfOM-HO. The experiments were performed in duplicates with standard deviation within 25% difference between the values produced by the two methods.
416
w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
0.2 0.15 0.1
62 %
0.05 0 1
2
3
4
5
6 7 8 Variable
9
10 11 12
Fig. 3 e Principal component analysis results indicating the contribution to explaining the variability in the data by each additional variable.
surrogates for degree of bond conjugation. The statistical analysis performed in this study takes into account internal correlations among the variables and drops the redundant ones. The variables selected by the model account for many factors potentially influencing the reactivity. COD is a measure of the oxidation potential of EfOM. The RC on NH2 medium is a measure of the prevalence of charged functional groups (Rosario-Ortiz et al., 2007) and could be related to the spatial folding of the molecules which could affect how easily accessible some parts of the molecule are for the reaction with HO. FI is a measure of overall blue-shift associated with the degree of bond conjugation of the organic matter. TOC is indicative of the duration of biological treatment. It decreases with increased solids retention time, and, as a result, is indicative of average molecular weight of EfOM which also decreases with increased solids retention time. The following model coefficients were determined: 1 ¼ 0:00830 COD þ 0:0455 RCNH2 þ 0:208 FI kEfOMHO $108 0:00289 TOC 0:196
[5]
Fig. 4(a) shows the predicted vs. the observed values for all 28 samples, and Fig. 4(b) shows the results of cross-validation
Predicted kEfOM-HO×10-8, M-1s-1
4.00 3.50
(a)
3.00 2.50 2.00 1.50 1.00 1.00
2.00
3.00
Measured kEfOM-HO
×10-8,
4.00 M-1s-1
where each point is predicted by the model, the coefficients for which were calculated without that point being part of the data set. Cross-validation measures the ability of the model to accurately predict the rate constant for the samples that were not part of the data set, enabling a water reuse or wastewater treatment utility to evaluate the reactivity of their EfOM with HO for AOP design. In addition, it can determine if any data points have particular influence on the coefficients. Those points would show more deviation from the 1:1 line in crossvalidation than when predicted by the model that uses those points in coefficient determination. It appears that the model is reasonably good at predicting the rate constants. Only 4 of the 28 samples have >20% difference between the modeled and the observed values, and 18 samples had 0.99), suggesting that both are attributable to the same structural unit. The correlation of kEfOM-HO with TOC can be explained by the fact that lower TOC in effluent is generally associated with longer biological treatment which results in formation of smaller, more reactive EfOM molecules.
3.3.
Treatment train effects
When the reaction rate constants across the treatment train of a single facility were compared, the average reaction rate constant stayed largely unaffected by the nitrification or denitrification processes. Fig. 6 shows the reaction rate constant after the short SRT solids contact tank and secondary clarifiers (SC), then after the trickling nitrifying filters (TNF) followed by the denitrifying filters (DN) and in final effluent (FE). It is clear from the figure that different types of biological processes at this facility did not cause statistically significant changes in kEfOM-
TNF
DN
FE
Fig. 6 e kEfOM-HO after different biological treatment steps at a single wastewater treatment plant: after aerobic biodegradation and secondary clarification (SC), followed by trickling nitrifying filters (TNF), followed by denitrifying filters (DN), and in final effluent (FE). The values are the averages with error bars representing 95% confidence intervals.
HO. However, each additional process appears to stabilize the rate constant, as seen from the decrease of the error bars representing 95% confidence intervals with progressing treatment. This is not surprising, as utilities have variable influent quality, but the treatment processes are adjusted to meet the effluent goals, which are kept relatively constant (Fig. 7). It is also notable that the smaller of the participating utilities had less variability in the kEfOM-HO of the samples from one sampling event to another. The wide range of the rate constants at the largest facility used in the study could be the result of a wide variability in the influent with highly variable percent of industrial input compared to the smaller utilities treating mainly domestic waste. Some of the smallest utilities (community scale) could experience some variability due to limited operator oversight. However, more utilities would need to be surveyed for any conclusions to be drawn. Solids retention time did not show a clear correlation across all the utilities, but when separated into the short (
Available online at www.sciencedirect.com
ScienceDirect journal homepage: www.elsevier.com/locate/watres
Identifying the factors that influence the reactivity of effluent organic matter with hydroxyl radicals Olya S. Keen a,c,*, Garrett McKay b, Stephen P. Mezyk b, Karl G. Linden c, Fernando L. Rosario-Ortiz c a
Department of Civil and Environmental Engineering, University of North Carolina, 9201 University City Blvd., Charlotte, NC 28223, USA b Department of Chemistry and Biochemistry, California State University at Long Beach, 1250 Bellflower Blvd., Long Beach, CA 90840, USA c Department of Civil, Environmental and Architectural Engineering, 428 UCB, University of Colorado at Boulder, Boulder, CO 80309, USA
article info
abstract
Article history:
Advanced oxidation processes (AOPs) are an effective treatment technology for the
Received 8 May 2013
removal of a variety of organic pollutants in both water and wastewater treatment.
Received in revised form
However, many background constituents in water are highly reactive towards hydroxyl
15 October 2013
radicals (HO) and decrease the efficiency of the process towards contaminant oxidation.
Accepted 21 October 2013
Up to 95% of the HO scavenging can come from dissolved organic matter (OM). In this
Available online 1 November 2013
study, 28 wastewater effluent samples were analyzed to find correlations between the reactivity of HO with wastewater-derived OM (known as effluent organic matter, EfOM),
Keywords:
water quality parameters, treatment train characteristics, and fluorescence-derived data.
Effluent organic matter
Rate constants for the reaction between HO and EfOM (kEfOM-HO) were measured using a
Hydroxyl radical
bench scale UV-based AOP system with methylene blue as an HO probe and confirmed
Advanced oxidation
using an electron pulse radiolysis method for a subset of the samples. The EfOM was
Wastewater treatment
characterized using a series of physicochemical parameters, including polarity, average molecular size and fluorescence. The kinetic data were analyzed with principal component analysis and Akaike Information Criterion. Four predictors were identified as dominant: chemical oxygen demand, retention onto NH2 extraction medium, fluorescence index, and total organic carbon. These four variables accounted for approximately 62% of the variability in the value of kEfOM-HO The average kEfOM-HO value for EfOM in this study was 1 8 1 1 value determined for 2.5 108 M1 C s , which is about 31% lower than the 3.6 10 MC s
natural organic matter isolates and commonly used in AOP modeling. ª 2013 Elsevier Ltd. All rights reserved.
1.
Introduction
Advanced oxidation processes (AOPs) have been demonstrated as an efficient treatment technology for the oxidation
of a variety of organic contaminants (Huber et al., 2003; Gogate and Pandit, 2004; Rosenfeldt and Linden, 2004; Rosario-Ortiz et al., 2010). Its efficacy arises from the highly reactive and non-selective nature of the hydroxyl radicals (HO). HO reacts
* Corresponding author. Department of Civil and Environmental Engineering, University of North Carolina, 9201 University City Blvd., Charlotte, NC 28223, USA. Tel.: þ1 704 687 5048. E-mail address: [email protected] (O.S. Keen). 0043-1354/$ e see front matter ª 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.watres.2013.10.049
w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
List of abbreviations: AOP HO OM EfOM TOC DOC UV MB PCA AIC RC COD SRT SUVA254 MW d FI RI PS HIX
advanced oxidation process hydroxyl radical organic matter effluent organic matter total organic carbon dissolved organic carbon ultraviolet methylene blue principal component analysis Akaike Information Criterion retention coefficient chemical oxygen demand solids retention time specific UV absorbance at 254 nm molecular weight (weight average) dispersity fluorescence index redox index protein-like fluorescence signal humification index
with organic contaminants through different mechanisms, including hydrogen abstraction and addition to electron-rich sites, with reaction rate constants on the order of 10810 M1 s1 being reported for the reactions with many organic species (Buxton et al., 1988). Although the high reactivity of HO offers the benefit of oxidizing mixtures of organic compounds, it also has the drawback of reacting with the background water quality components. Carbonate species and dissolved organic matter (OM) are typically the primary HO scavengers in natural water samples. The application of AOPs for the oxidation of organic contaminants has been studied as a potential municipal wastewater treatment process for water reuse applications (Huber et al., 2003; Rosario-Ortiz et al., 2010; Keen et al., 2012a, 2012b). Typically, AOP efficiencies for organic compounds vary from no removal to complete oxidation, and are a function of the oxidation kinetics of the compounds and the scavenging capacity of the water matrix (Rosario-Ortiz et al., 2010; Keen et al., 2012b). In AOP systems, only a fraction of HO will react with the trace contaminants with the majority of the HO reacting with background scavengers. The consumption of HO via non-specific reactions is the scavenging capacity of the water matrix. Because of the high level of HO scavenging in wastewater matrices, AOPs are often viewed as not economical for wastewater treatment. However, some of the recent studies show that it may not be necessary to achieve full mineralization (Keen et al., 2012a). Combined with better ways to predict the process performance, AOPs may become a viable option for wastewater treatment. In wastewater treatment plant effluents, the main HO scavenger is effluent organic matter (EfOM), due to both its s1 moderate reactivity with HO on the order of 108 M1 C (Westerhoff et al., 1999; Rosario-Ortiz et al., 2008) and its relatively high concentrations (measured as total organic carbon-TOC). While natural OM isolates and standards were shown to have a relatively constant value for the reaction rate
SC TNF DN FE
409
secondary clarifier trickling nitrifying filter denitrifying filter final effluent
List of symbols: k reaction rate constant kEfOM-HO reaction rate constant between effluent organic 1 matter (EfOM) and HO, M1 C s , where MC is the molar concentration of organic carbon kHO,MB reaction rate constant between methylene blue (MB) and HO, M1 s1 E0 average fluence rate, mW/cm2 F fluence, mJ/cm2 ks HO reaction rate constant for a given scavenging compound, M1 s1 [S] concentration of a scavenging compound, M molar absorption coefficient at 254 nm, M1 cm1 ε254 F quantum yield of photolysis, e 254 nm wavelength energy, J/mol U254
constant with HO (kOM-HO) (Westerhoff et al., 1999), the reactivity of EfOM (kEfOM-HO) has been shown to have variable reactivity with the reported range of values from 1.39 108 to 1 across various studies (Rosario-Ortiz et al., 11.5 108 M1 C s 2008; Westerhoff et al., 2007; McKay et al., 2011; Katsoyiannis et al., 2011; Nagarnaik and Boulanger, 2011) with up to a factor of 4.5 difference in a single study (Rosario-Ortiz et al., 2008). The variable nature of kEfOM-HO has been attributed in part to differences in molecular weight composition across different samples (Westerhoff et al., 2007; Dong et al., 2010). Among other parameters that were previously suggested to influence reactivity of OM towards HO are specific UV absorbance (absorbance per mg/L of organic carbon) at 254 nm (Westerhoff et al., 1999), polarity measures and fluorescence index (Rosario-Ortiz et al., 2008). These parameters carry information about the structural composition and prevalence of certain functional groups within the bulk OM, such as aromatic rings or hydrophilic functional groups. For example, FI of the sample tends to be higher when more products of microbial activity (more aliphatic in structure) are present (McKnight et al., 2001). The main objective of this study was to conduct a statistical evaluation of the physicochemical properties of EfOM and of wastewater treatment process variables and their role in the reactivity of EfOM towards HO. In full-scale AOP systems, it is difficult to measure the concentration of HO in situ or in real time because of its low concentration. Therefore, engineers rely on calculations and safety factors when designing these treatment processes. Part of this study’s objective was to develop a model to predict the values of kEfOMHO to offer a full-scale reactor design tool that could increase confidence in the prediction of performance, and therefore minimize the safety factors and eventual costs associated with AOP treatment. A number of parameters were evaluated in this study for their ability to predict the value of kEfOM-HO. Those parameters
410
w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
included bulk properties of effluent organic matter (average molecular weight and its dispersity, and polarity), water quality parameters (chemical oxygen demand, specific UV absorbance at 254 nm, TOC), treatment train characteristics (solids retention time in the biological treatment process), and fluorescence derived data (fluorescence index, humification index, intensity of the protein-like peaks and the fraction of signal attributed to the oxidized vs. reduced quinones). Each of the parameters considered for the model was expected to influence the reactivity of EfOM towards HO, and the choice of each is discussed in detail in a corresponding section. The ability of these parameters to account for the variability of the reaction rate constants across 28 wastewater effluent samples was evaluated with principal component analysis (PCA). A model was fitted to the best predictors identified by the Akaike Information Criterion (Akaike, 1974) and is presented in this paper.
2.
Materials and methods
2.1.
Effluent samples
The samples were collected from 8 different wastewater treatment utilities, 6 of which were sampled on several separate occasions. The facilities were located mainly in Colorado with only 4 of the 28 samples from other states (Florida and California). This may be a limitation of the sample set because of the differences in source water. All utilities were sampled at the point in the treatment train right before disinfection. This sampling location is representative of the location of an AOP process in the treatment train. One of the utilities was sampled at several different points along the treatment train to determine how different biological processes affect the reactivity of EfOM. The sampling points included effluents from the secondary clarifier following the solids contact activated sludge, from the nitrifying trickling
Table 1 e Wastewater treatment trains of the participating utilities. Utility
1 2 3
4 5 6
7 8
Treatment process
Activated sludge with biological nutrient removal Activated sludge with biological nutrient removal Solids contact activated sludge þ nitrifying trickling filters þ denitrifying filters Activated sludge with biological nutrient removal Activated sludge with biological nutrient removal Activated sludge with biological nutrient removal þ denitrifying filters þ sand filters Extended aeration lagoon Membrane bioreactor
Average Solids plant flow, retention time, MGD days 12
12
80
5
23
0.7
55
10
6.3
10
22.5
15
0.04 0.014
100 e
filter, from the denitrifying trickling filter and the final effluent before disinfection consisting of a blend of water that passed through the denitrifying filters and the water that bypassed the denitrification. Table 1 summarizes the utilities and their corresponding treatment processes. All samples were filtered after collection through a 0.45 mm nylon filter (Nalgene, Rochester, NY) and stored at 4 C to minimize any further bacterial processes. Due to the large number of samples in this study, immediate analysis was not always possible. Most samples were analyzed within 10e15 weeks from collection. Water quality parameters (TOC, alkalinity, ammonia, nitrite and nitrate) were measured immediately after sample collection and immediately prior to measuring kEfOM-HO. Samples that showed signs of microbial activity (e.g. changes in inorganic nitrogen species or drop in TOC) were excluded from the study. Nitrate, nitrite and ammonia concentrations were measured using a Hach DR5000 spectrophotometer (Hach Corporation, Loveland, CO) and the corresponding Hach kits (TNT835, TNT839 and TNT830 respectively). Dissolved organic carbon (DOC) was measured using Shimadzu VCSH TOC analyzer (Shimadzu Scientific Instruments, Columbia, MD). Inorganic carbon was eliminated from the DOC samples by lowering the pH to about 2.0 with hydrochloric acid (Mallinckrodt, Hazelwood, MO), after which the sample was purged with ultra-zero air during the analysis. The sample pH was measured with a calibrated pH meter (F340, Beckman Coulter, Indianapolis, IN), alkalinity was measured with a Hach digital titrator (Hach Corporation, Loveland, CO), and absorbance scans were taken with Cary100Bio spectrophotometer (Agilent Technologies, Santa Clara, CA).
2.2.
Measurement of kEfOM-HO
2.2.1.
UV/H2O2 with methylene blue
Experiments measuring the overall HO scavenging of the water samples were performed with 5 mM of reagent-grade methylene blue (SigmaeAldrich, St.Louis, MO) as a probe. Use of methylene blue as an HO probe was described previously for Fenton reaction (Satoh et al., 2007) and for photocatalysis applications (Luo et al., 2009). It was determined to be an appropriate probe for this study as well based on two factors: (a) it does not degrade by direct photolysis at wavelengths >200 nm, and (b) it can be analyzed spectrophotometrically allowing fast and inexpensive processing of multiple samples. Hydroxyl radicals were generated by UV exposure of samples containing 10 mg/L of reagent-grade hydrogen peroxide (J.T.Baker, Phillipsburg, NJ). A low-pressure mercury vapor UV lamp emitting around 253.7 nm and housed in a collimated beam apparatus was used in the experiment. The experiments were performed in stirred crystallization dishes 5 cm in diameter with a Petri factor of 0.97. The average UV dose delivered to the sample was calculated using all appropriate correction factors (Bolton and Linden, 2003). Hydrogen peroxide was measured by the triiodide method (Klassen et al., 1994) prior to adding methylene blue to the sample to avoid interference. Methylene blue was measured spectrophotometrically at 664 nm. Although methylene blue is a photostable compound under UV > 200 nm, a blank run without hydrogen peroxide
411
w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
was performed to account for non-HO reaction pathways in the wastewater matrix. The difference between the sample and the blank was used to measure the extent of the probe degradation due to hydroxyl radicals alone. The concentration of steady-state HO ([HO]SS) was then calculated using the following relationship: [1]
where E0 is the average fluence rate (mW/cm2), F is the fluence (mJ/cm2) and kHO,MB is a time-based reaction rate constant between probe methylene blue (MB) and hydroxyl radicals (M1 s1). In this equation, the quantity kHO,MB [HO]ss/E0 is the slope of the plot of ln[MB]t/[MB]0 vs F, and [HO] can be calculated as ½HOss ¼
slope$E0 kHO;MB
[2]
The value for kHO,MB has been reported as 2.1$1010 M1 s1(Buxton et al., 1988). The UV/H2O2 model by Glaze et al. (1995) was then rearranged to calculate the total HO scavenging coming from the sample background: X
ks ½S ¼
E0 ε254 F½H2 O2 1 $ U254 ½HOss
ks ½S ¼
E0 ε254 F½H2 O2 kHO;MB ε254 F½H2 O2 kHO;MB $ ¼ $ U254 U254 slope$E0 slope
a b
8.5 3.9 2.7 1.0 3.6
106a 108a 107a 1010a 108b
Buxton et al., 1988. Westerhoff et al., 1999.
1 range of values for kEfOM-HO was (1.6e3.3) 108 M1 (>100% C s difference), which indicates that predicting the value well beyond the experimental noise is possible. EfOM accounted for 41e60% of scavenging in the experiments when H2O2 and the probe were taken into account, so the difference between the measured total scavenging and the calculated scavenging by the compounds other than EfOM was quantifiable. In the full-scale settings the fraction of scavenging attributable to EfOM would be even greater (45e95%) because the probe compound would not be affecting the calculations (Fig. 1). Therefore, EfOM is highly influential in the process performance.
[3]
2.2.2.
where ks is the HO reaction rate constant for a given scavenging compound (M1 s1), [S] is the concentration of the corresponding scavenging compound (M), ε254 is the molar absorption of hydrogen peroxide at 254 nm (M1 cm1), F is the quantum yield of hydroxyl radical formation by photolysis of hydrogen peroxide at 254 nm, [H2O2] is the concentration of hydrogen peroxide (M), and U254 is the wavelength energy (J mol1) Substituting Eq. (2) into Eq. (3) yields X
HCO 3 CO2 3 H2O2 NO 2 OM
kS-HO, M1 s1
[4]
Typical scavengers in wastewater matrices are EfOM, bicarbonate and occasionally nitrite, although the levels of nitrite at most properly operating wastewater treatment plants are very low. If hydrogen peroxide is added for HO generation, then hydrogen peroxide is a scavenger as well. Table 2 summarizes the most prominent HO scavengers in effluents along with the corresponding kHO values. Carbonate and ammonia are also scavengers of HO, however all of the effluents had pH values between 6.3 and 7.4, making the concentrations of either species negligible (aA2and aBH values 0.001e0.00005 and 0.01e0.001 respectively). The theoretical contribution by the scavenging compounds other than EfOM was calculated using the published values for the reaction rates between HO and bicarbonate, nitrite, hydrogen peroxide and methylene blue (Table 2). The difference between the measured scavenging (Eq. (4)) and the calculated scavenging by compounds other than EfOM was assumed to be the scavenging by EfOM. The EfOM scavenging value was then divided by the organic carbon concentration to find kEfOM-HO. The experiments were performed in duplicates with standard deviation within 25% difference between the values produced by the two methods.
416
w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 4 0 8 e4 1 9
0.2 0.15 0.1
62 %
0.05 0 1
2
3
4
5
6 7 8 Variable
9
10 11 12
Fig. 3 e Principal component analysis results indicating the contribution to explaining the variability in the data by each additional variable.
surrogates for degree of bond conjugation. The statistical analysis performed in this study takes into account internal correlations among the variables and drops the redundant ones. The variables selected by the model account for many factors potentially influencing the reactivity. COD is a measure of the oxidation potential of EfOM. The RC on NH2 medium is a measure of the prevalence of charged functional groups (Rosario-Ortiz et al., 2007) and could be related to the spatial folding of the molecules which could affect how easily accessible some parts of the molecule are for the reaction with HO. FI is a measure of overall blue-shift associated with the degree of bond conjugation of the organic matter. TOC is indicative of the duration of biological treatment. It decreases with increased solids retention time, and, as a result, is indicative of average molecular weight of EfOM which also decreases with increased solids retention time. The following model coefficients were determined: 1 ¼ 0:00830 COD þ 0:0455 RCNH2 þ 0:208 FI kEfOMHO $108 0:00289 TOC 0:196
[5]
Fig. 4(a) shows the predicted vs. the observed values for all 28 samples, and Fig. 4(b) shows the results of cross-validation
Predicted kEfOM-HO×10-8, M-1s-1
4.00 3.50
(a)
3.00 2.50 2.00 1.50 1.00 1.00
2.00
3.00
Measured kEfOM-HO
×10-8,
4.00 M-1s-1
where each point is predicted by the model, the coefficients for which were calculated without that point being part of the data set. Cross-validation measures the ability of the model to accurately predict the rate constant for the samples that were not part of the data set, enabling a water reuse or wastewater treatment utility to evaluate the reactivity of their EfOM with HO for AOP design. In addition, it can determine if any data points have particular influence on the coefficients. Those points would show more deviation from the 1:1 line in crossvalidation than when predicted by the model that uses those points in coefficient determination. It appears that the model is reasonably good at predicting the rate constants. Only 4 of the 28 samples have >20% difference between the modeled and the observed values, and 18 samples had 0.99), suggesting that both are attributable to the same structural unit. The correlation of kEfOM-HO with TOC can be explained by the fact that lower TOC in effluent is generally associated with longer biological treatment which results in formation of smaller, more reactive EfOM molecules.
3.3.
Treatment train effects
When the reaction rate constants across the treatment train of a single facility were compared, the average reaction rate constant stayed largely unaffected by the nitrification or denitrification processes. Fig. 6 shows the reaction rate constant after the short SRT solids contact tank and secondary clarifiers (SC), then after the trickling nitrifying filters (TNF) followed by the denitrifying filters (DN) and in final effluent (FE). It is clear from the figure that different types of biological processes at this facility did not cause statistically significant changes in kEfOM-
TNF
DN
FE
Fig. 6 e kEfOM-HO after different biological treatment steps at a single wastewater treatment plant: after aerobic biodegradation and secondary clarification (SC), followed by trickling nitrifying filters (TNF), followed by denitrifying filters (DN), and in final effluent (FE). The values are the averages with error bars representing 95% confidence intervals.
HO. However, each additional process appears to stabilize the rate constant, as seen from the decrease of the error bars representing 95% confidence intervals with progressing treatment. This is not surprising, as utilities have variable influent quality, but the treatment processes are adjusted to meet the effluent goals, which are kept relatively constant (Fig. 7). It is also notable that the smaller of the participating utilities had less variability in the kEfOM-HO of the samples from one sampling event to another. The wide range of the rate constants at the largest facility used in the study could be the result of a wide variability in the influent with highly variable percent of industrial input compared to the smaller utilities treating mainly domestic waste. Some of the smallest utilities (community scale) could experience some variability due to limited operator oversight. However, more utilities would need to be surveyed for any conclusions to be drawn. Solids retention time did not show a clear correlation across all the utilities, but when separated into the short (
