Relationship between chlorine consumption and chlorination by-products formation for model compounds

E.E. Chang

, P.C. Chiang

, S.H. Chao

, Y.L. Lin

aDepartment of Biochemistry, Taipei Medical University, 250 Wu-Shin Street, Taipei 110, Taiwan

bGraduate institute of Environmental Engineering, National Taiwan University, 71 Chou-Shan Road, Taipei 106, Taiwan Received 10 May 2005; received in revised form 10 November 2005; accepted 10 November 2005

Available online 10 January 2006

Abstract

The objective of this research is to investigate the relationship between chlorine decay and the formations of disinfection by-products (DBP), including trichloromethane (TCM) and chloroacetic acid (CAA) in the presence of four model compounds, i.e., resorcinol, phloro-glucinol, p-hydroxybenzoic acid, and m-hydroxybenzoic acid. The chlorine degradation in model compounds with OH and/or COOH functional groups were rapid after chlorination. The TCM yields of carboxylic group substituted compounds (3-hydroxybenzoic acid [3-HBA], 4-hydroxybenzoic acid [4-HBA]) were found to be lower than that of the m-dihydroxy substituted compounds. Phloroglucinol, with one more OH substitution group than resorcinol, tends to form significant amounts of CAA after chlorination. However, it was observed that with the COOH substitution of 3-HBA and 4-HBA tend to exhibit more CAA formation potential than resorcinol.

The developed parallel second and first-order reaction model for chlorine demand has been successfully utilized for TCM, CAA and DBP formation modeling. A high correlation between CAA and TCM was observed for the model compounds.

Ó 2005 Elsevier Ltd. All rights reserved.

Keywords: Chlorine consumption; Resorcinol (R); Phloroglucinol (P); Hydroxybenzoic acid (HBA); Tichloromethane (TCM); Chloroacetic acids (CAA);

Chlorine decay model

1. Introduction

After the humic fraction within nature organic matter (NOM) was identified as a major precursor for trihalome-thanes (THM) formation (Rook, 1976), most researches have focused their research on the humic portion of the NOM for disinfection by-products (DBPs) formation.

Marhaba and Van (2000) concluded that the hydrophilic acid fraction was the most reactive precursor to the THM formation; while the hydrophobic neutral fraction was related to the formation of HAA.Liang and Singer (2001) reported that hydrophilic carbon also plays an important role in DBP formation, especially for waters with low humic content. Recent studies indicate that all fractions of NOM

contribute to the formation of DBP (Sinha, 1999; Chang et al., 2001; Gang et al., 2003). It appears that the properties of humic substances have molecular weights of several hun-dred or larger, with weakly acidic functional groups (such as carboxylic group), and phenolic group which cause dif-ferent types and amounts of DBPs (Cook and Langford, 1998).

Several studies suggested that aliphatic carboxylic acids, hydroxybenzoic acids, phenols and pyrrole derivatives are reactive substrates of organic precursors for THMs forma-tion (Norwood et al., 1980; Korshin et al., 1997). Rook (1976)postulated that the m-dihydroxy structure of resor-cinol was the principal TCM precursor in aquatic humic materials and proposed a reaction mechanism. The reac-tion products, CHCl3 and CCl3COOH, were identified from the chlorination of resorcinol (Norwood et al., 1980). Chlorination could undergo electrophilic attack either at the chlorine atom (with displacement of hydroxyl,

0045-6535/$ - see front matter Ó 2005 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel./fax: +886 2 2736 9236.

E-mail address:[email protected](E.E. Chang).

www.elsevier.com/locate/chemosphere Chemosphere 64 (2006) 1196–1203

leading to chlorination) or at the hydroxyl group (with loss of chlorine). For example, p-hydroxybenzoic acid reacted rapidly to generate a mixture which is side-chain cleavage products of substitution and decarboxylation (Larson and Rockwell, 1979).Boyce and Hornig (1983) confirmed that the conversion of 1,3-dihydroxyaromatic precursors to THMs occurs in two stages. Extensive incorporation of halogen by electrophilic substitution and addition pro-cesses is followed by a complex series of hydrolysis and decarboxylation steps, which lead to TCM via carbon–car-bon carbon–car-bond cleavage about the C2site of the aromatic ring.

In many research reports, mathematical models were suggested to predict THM formation of specific source water (Engerholm and Amy, 1983; Amy et al., 1987; Chang et al., 1996; Gang et al., 2002). Gang et al. (2002) con-structed a mathematical model of chlorine decay to predict the THM formation. The authors (Gang et al., 2003) also indicated that the THM formation in fractionated NOM was a function of chlorine consumption. As the molecular weight of the fraction decreased, THM yield coefficients increased.Katz (1986)suggested that the total organic car-bon (TOC) had a strong correlation with chlorine demand, particularly when turbidity was less than 20 NTU in the fil-trate. The effect of chlorine demand on DBP formation is generally not well known because NOM is composed of many types of organics. Aromatics and humic substance strongly react with chlorine that could be responsible for the initial chlorine demand (Dotson and Helz, 1984).

Most organic matters contributing to major DBP pre-cursors in source water of Taiwan are small compounds, with a molecular weight of less than 1 KDa which was measured by the ultrafiltration membranes (Chang et al., 2001; Chiang et al., 2002) However, only limited research has been done on DBPs formation with different functional groups of small molecular aromatic compounds. The objectives of this research were to: (1) develop the appro-priate chlorine decay and DBP formation models for the selected four model compounds; (2) investigate the forma-tion potential of trichloromethane (TCM) and chloroacetic

acids (CAA) for the four model compounds; (3) evalu-ate the relationship between CAA/TCM and chlorine consumption with different functional groups of model compounds.

2. Material and methods 2.1. Sample preparation

Four model compounds with different functional groups of benzene i.e., carboxylic and phenolic groups were selected in this investigation to represent small molecular NOM. The four model compounds include phloroglucinol (1,3,5-trihroxybenzene), resorcinol (1,3-dihydroxybenzene), m-hydroxybenzoic acid (3-HBA), and p-hydroxybenzoic acid (4-HBA). The dissolved organic carbon (DOC) con-centrations for model compounds, using de-ionized water (Milli-Q SP), were prepared and adjusted to approximately 3.0 (±0.2) mg/l as C. The characteristics of the model com-pounds are listed inTable 1.

2.2. Evaluation of Chlorine consumption

A 7-day chlorine consumption study was performed using 28 mg/l chlorine dosage (as Cl2), about 9 times the DOC dosage, to determine the chlorine consumption, tri-chloromethane formation potential (TCMFP), and chloro-acetic acid formation potential (CAAFP). Throughout these chlorination experiments, all samples were chlori-nated by 13% free chlorine (sodium hypochlorite) stock solution and added phosphate buffer (pH 7.0). A blank sample was prepared using the same amount of deionized ultra filtered water, and chlorinated under the same condi-tions as the other samples. Samples were chlorinated in 6 liter glass bottles and then carefully transferred into 150 amber glass bottles with Teflon-lined caps. A separate bot-tle containing the four model compounds samples was used for each reaction kinetic test. There were 12 experimental data for 3-HBA, 4-HBA, resorcinol and phloroglucinol,

Table 1

Physical/chemical properties of model compounds

Model compound Phloroglucinol Resorcinol 3-HBA 4-HBA

Molecular formula C6H6O3 C6H6O2 C7H6O3 C7H6O3

Molecular weight 126.11 110.11 138.12 138.12

Dissociation constant (pKa) pK18.0 pK19.30 pK14.06 pK14.48

pK29.2 pK211.06 pK29.92 pK29.32

pK314

Solubility in water 10 g/l (20°C) 1000 g/l (20°C) slightly soluble (20°C) 5000 mg/l (25°C)

OH OH COOH COOH

E.E. Chang et al. / Chemosphere 64 (2006) 1196–1203 1197

respectively, obtained at the specific time intervals, i.e., 0.08, 0.17, 0.33, 0.50, 0.67, 0.83, 1, 3, 6, 24, 48 and 168 h during the reaction kinetic tests. The samples were kept headspace free in the dark at room temperature (25 ± 2°C) until they were analyzed. Chlorine residual, DOC and UV adsorption were measured at different times for each bottle.

2.3. Analytical methods

Chlorine concentration was measured by N,N-diethyl-p-phenylenediamine (DPD) titration methods. DOC (TOC), UV254, pH, and DBPFP analyses were conducted for each water sample. All analyses, unless otherwise noted, were performed according to the 19th edition of the Standards Methods (APHA, 1998). Water samples for DOC and UV analyses were conducted first by filtering through a prewashed 0.45 lm filter, and then the sample was ana-lyzed by a TOC instrument (O.I. Corporation model 700) and UV spectroscopy (Hitachi U-2000). TCM and CAA (including mono-, di-, and tri-chloroacetic acids) were analyzed by HP 6890GC/ECD according to Standard Methods 6230D and USEPA methods 552.2, respectively.

Duplicate analyses were performed on each sample, and the average was reported. If the difference between the two values was greater than 15%, a third analysis was per-formed, and the average of all three values was reported.

2.4. Models of chlorine decay and DBP formation

Owing to the unique characterization of the selected tar-get compounds, many models developed in the literature (Gang et al., 2002, 2003) can not fit the experimental data well. As a result, the parallel first-order reaction model, which was originally derived by Gang et al. (2002) could be modified as the following:

NOMRþ Cl2!^kRR X ðrapidÞ ð1Þ

in which CRis the chlorine concentration participating in a hypothetical separate rapid reaction; CSis the chlorine con-centration participating in a hypothetical separate slow reaction; R and X are chlorinated by-products; n and m are the order of the reaction with respect to the rapid and slow reactions, respectively.

The value of n and m are determined by the best fit as compared with the suggested reaction orders. Integrating these rate equations (Eqs.(2) and (4)) with CR0= fC0and

CðtÞ ¼ ½KR t  ðn þ 1Þ þ f  C^nþ1₀ ^nþ11

þ ½KS t  ðm þ 1Þ þ f  C^mþ1₀ ^mþ11 ðn; m 6¼ 1Þ ð5Þ CðtÞ ¼ C0 ff  e^KRtþ ð1  f Þ  e^KStg ðn; m ¼ 1Þ ð6Þ in which C(t) is the chlorine concentration at any time t (mg/l), C0is the initial chlorine concentration (dose), f is the fraction of the chlorine demand attributed to rapid reactions, kRis the rate constant for rapid reactions, and kS is the rate constant for slow reactions.

The coefficients (f, KR, and KS) obtained from the chlo-rine decay model (Eq.(5)or Eq. (6)) were used to predict the TCM, CAA, and DBP (TCM + CAA) formations.

Eqs.(7)–(9)assume that the TCM, CAA and DBP forma-tions are a function of chlorine consumpforma-tions with respect to the rapid and slow reactions:

TCM¼ AðCR0 CRÞⁿþ BðCS0 CSÞ^m ð7Þ CAA¼ CðCR0 CRÞⁿþ DðCS0 CSÞ^m ð8Þ DBP¼ EðCR0 CRÞⁿþ F ðCS0 CSÞ^m ð9Þ in which A and B are the TCM yield coefficient from the rapid and slow chlorine consumed, respectively; C and D are the CAA yield coefficients from the rapid and slow chlorine consumed, respectively; E and F are the DBP yield coefficients from the rapid and slow chlorine consumed, respectively.

The parameters n, m, f, kR, ks and yield coefficients (A, B, C, D, E, F) were determined by non-linear regression software (SYSTAT 5.01).

3. Results and discussion

3.1. Chlorine demand and decay modeling

Fig. 1shows the chlorine demand and residual chlorine associated with the hydroxybenzene and hydroxybenzoic acid during the chlorination process, respectively. It was observed that the chlorine consumption increased rapidly within the first 3 h and then gradually decayed after 3 h

0.0

Normalized chlorine demand (C/C0) Normalized residual chlorine (C/C0)

Normalized chlorine demand :

Normalized residual chlorine :

Time (h)

Fig. 1. Normalized chlorine demand and decay curves of model samples

1198 E.E. Chang et al. / Chemosphere 64 (2006) 1196–1203

of chlorination. Since the phenolates (dissociated form of phenols) from model compounds were responsible for the fast reaction with chlorine, all water samples consumed over 80% of the initial chlorine dose within the first 3 h, especially for the resorcinol, which had the highest chlorine consumption rate at 10 min.

The parallel second and first-order reaction model for chlorine demand derived in this study is the best fit as com-pared with the parallel first-order model (derived byGang et al., 2002), n-order chlorine decay model, parallel second order and parallel first order and second order.Table 2 pre-sents the chlorine decay constants and fitting parameters for the model compounds. The chlorine data of hydroxy-benzene and hydroxybenzoic acid in Fig. 2 fit the model well, yielding the correlation coefficients of 0.985–0.991.

The constants of rapid decay rates (KR= 0.32–

5.05 l mg^1h^1) in Table 2 were much higher than those of the slow decay rates (k_S= 0.006–0.028 h^1) for all model samples. The values of kR for the hydroxybenzoic acids were much smaller than those of the hydroxybenzene sam-ples. The proportion constants (f) shown inTable 2ranged from 76% to 91% of the chlorine consumption. Differences in the reaction kinetics observed between these four com-pounds may be separated into two groups. For resorcinol and phloroglucinol, the chlorine consumptions were higher at first and increased gradually afterwards; whereas for 3-HBA and 4-3-HBA, chlorine consumptions were lower at first and increased rapidly afterwards.

Larson and Rockwell (1979) and Gallard and Gunten (2002) revealed that resorcinol, with two activating –OH groups, could release electrons rapidly, leading to the

elec-trophilic addition and substitution reactions while chlori-nation was proceeding. Boyce and Hornig (1983) also pointed out that when both OH groups on an aromatic ring are located at an appropriated orientation to stabilize the transition state of the reaction through the donation of electron density, an electrophilic substitution mechanism could easily occur. These observations suggest that the aro-matic carbon site adjacent to the C1-hydroxyl group be inverted to electrophilic substitution by chlorine. However, phloroglucinol is highly symmetric and may form a reso-nance-stabilized intermediate because of three –OH groups. These three –OH groups may impede series of hydrolysis as well as decarboxylation with C–C bond cleav-age on the aromatic ring resulting in a lower kRvalue of phloroglucinol (1.225 l mg^1h^1) than that of resorcinol (5.051 l mg^1h^1).

As for hydroxybenzoic acids with moderately deactivat-ing substituents (–COOH), the electron density on the ben-zene ring would be lowered during the ionization process of carboxyl group. The chlorination of carboxyl groups pro-ceeds much slower than the chlorination of resorcinol and phloroglucinol, which is because that hydroxybenzoic acid reacts rapidly to give a decarboxylation product (Lar-son and Rockwell, 1979). As for the chlorine consumption rate between 3-HBA and 4-HBA, it was observed that there was a higher value for 4-HBA because the p-position of OH and COOH on the aromatic ring is more active than the m-position of OH and COOH which facilitates the chlorine reaction on hydroxybenzoic acid.

3.2. TCM, CAA and DBP formation kinetics and modeling Since the chlorine decay model was determined as the parallel second order (rapid reaction) and first order (slow reaction), the chlorine decay model could be simplified as:

CðtÞ ¼ C0 f

fC₀k_Rtþ 1þ ð1  f Þe^kst

 

ð10Þ With the above observations, the TCM, CAA and DBP formation models could also be simplified as

TCM¼ A fC₀ 1  1

Chlorine decay constants and fitting parameters of model compounds (chlorine dose = 28 mg/l)

Compounds f KR(l mg^1h^1) KS(h^1) R²

3-HBA 0.760 0.319 0.006 0.995

4-HBA 0.819 0.328 0.008 0.998

Resorcinol 0.782 5.051 0.008 0.995

Phloroglucinol 0.913 1.225 0.028 0.985

5

E.E. Chang et al. / Chemosphere 64 (2006) 1196–1203 1199

highest correlation coefficient (best fit) for the four model compounds with the exception of P for CAA model predic-tion.Figs. 3–5show that the data fit the TCM, CAA and DBP formation model quite well, with correlation coeffi-cients (0.854–0.996), which indicates that TCM, CAA and DBP formations were a function of chlorine

consump-other findings (Larson and Rockwell, 1979; Norwood et al., 1980; Boyce and Hornig, 1983).

The rate of CHCl₃production and chlorine consump-tion varied with each model compound, however, the over-all data showed two distinct patterns. The first pattern is exhibited by the m-dihydroxy substituted compounds and reflects a generally rapid and simultaneous exertion of both chlorine demand and TCM production (Figs. 1 and 3). The data suggests that this carbon between two hydroxyl groups should be responsible for TCM production (Larson and Rockwell, 1979; Rook, 1976).

The second pattern is demonstrated by the hydroxyben-zoic acids data and indicates that TCM is a minor reaction product. The hydroxybenzoic acids pattern produces approximately 5–10-fold less chloroform formation poten-tial than m-hydroxy substituted compounds (Table 3), although the chlorine demand remains relatively high. This phenomenon may be explained by the loss of a doubly acti-vated carbon between two free hydroxyls in 3-HBA and 4-HBA.

Because of the hydroxy configuration, the molecule will probably undergo oxidative decarboxylation with substitu-tion of chlorine in place of carboxyl (Larson and Rockwell, 1979), continuous chlorination and final cleavage could then occur at the chlorination site. The hydroxybenzoic acids pattern produces similar trichloromethane formation results. In addition to the chlorine demand for TCM pro-duction and oxidation, the evidence suggests that a portion of the chlorine demand is due to the incorporation of chlo-rine into non-CHCl3reaction products.

The chloroacetic acids (summation of mono-, di-, and tri-chloroacetic acid) were analyzed from the chlorination of model compounds as the disinfection by-products. In phloroglucinol, CAA formation rates also were initially rapid, corresponding with the rapid consumption of chlo-rine, followed by a slower, declining rate of production (Fig. 4). Above 90% CAA was generated within the first 3 h, as compared with the CAA formed at the end of reac-tion time—7 days.Christman et al. (1978)have noted that chlorination of resorcinol at high Cl2/substrate ratios enhance-the accumulation of chlorinated acids, including chlorobutenedioic acid, dichloroacetic acid, and trichloro-acetic acid etc. More electrophilic –OH groups of phloro-glucinol have lower pKa and higher SUVA254 (Table 1) which yields approximately 5-fold CAA and 2-fold DBP (TCM and CAA) formation potential than resorcinol (Table 3). Further, existing –COOH substitution sub-stances have lower pK_a values and generate more CAA as shown inFig. 4. Therefore, the distribution of various species of chlorinated products also depends on the acidity (pKa), SUVA254, and characteristics of the substrate in solution (Trussell and Umphres, 1978; Peters et al., 1980;

Gallard and Gunten, 2002).

The TCM (CAA) yield coefficient is defined as the ratio between the concentration (mg/l) of TCM (CAA) formed and the concentration of chlorine consumed (mg/l).Table

0

Fig. 3. The TCM formation and predictive data for model compounds during the chlorination process.

Fig. 4. The CAA formation and predictive data for model compounds during the chlorination process.

Fig. 5. The DBP formation and predictive data for model compounds during the chlorination process.

1200 E.E. Chang et al. / Chemosphere 64 (2006) 1196–1203

model compounds at different order of reaction. InTable 5, it was observed that there were two distinct patterns, i.e., hydroxybenzoic acid (3-HBA and 4-HBA) and hydroxyl benzene (R and P), exhibited their respective reaction order (n, m) and DBP yield coefficient.Reckhow et al. (1990)also found that the specific DBP formation was related to the activated aromatic matter, whereas activated aromatic con-tent was correlated with chlorine consumption.Gang et al.

(2003) reported that there was no strong correlation between molecular weight and chlorine decay kinetics.

With the above evidence, it suggests the amount of DBP generated be site-specific in practice, and the chlorine react-ing mechanism be dependent on the nature of target com-pounds in principle. In this study, although these four small model compounds have their respective functional group reacted with chlorine to form DBP, the DBP forma-tion is actually simulated by a chlorine demand model. The concept of DBP yield coefficient was useful for quantifying the difference in species production and evaluating the effect of organic precursor reduction.

3.3. Relationship between TCM and CAA

The specific chlorine demand (SCD) in Fig. 6 was defined as the ratio between the chlorine demand (mg/l) and the initial DOC concentration (mg/l) at the reaction times of 1, 3 and 168 h. In the first hour, the SCD and spe-cific DBPFP (DBP formation potential/DOC concentra-tion) of hydroxybenzenes are slightly higher than those of hydroxybenzoic acids, and the specific DBPFP of phloro-glucinol was the highest among the four model com-pounds. However, no relationships between specific DBPFP and SCD of model compounds were observed based on the limited data collected at various times.

Fig. 7 shows the relationship between TCM and CAA formation potential of model compounds under different chlorination times (1, 3 and 168 h). After linear regression of experimental data, a high correlation between CAA and TCM concentration was observed. However, there are two patterns of DBP correlation based on the slopes of linear curves in Fig. 7. The hydroxybenzoic acids pattern pro-duces a higher slope (>10) than that of the m-hydroxy substituted compounds (slope <1). These observations sug-gest that the aromatic carboxyl group has a strong correla-tion to the formacorrela-tion of CAA (Cook and Langford, 1998;

Pomes et al., 1999). However, oxidative decarboxylation of dihydroxybenzoic acid was not observed which was consis-tent with the findings suggested byNorwood et al. (1980).

Therefore, the formation of DBP is highly dependent on the nature of the organic matter.

4. Conclusions

Table 3

TCM, CAA and DBP formation for model compounds treated by chlorine

Model compound 3-HBA 4-HBA Resorcinol Phloroglucinol Initial concentration

(mg-C/l)

3.0 3.0 3.0 3.0

Specific chlorine demand (mg Cl2/mg-C)

1 h 6.0 6.5 7.1 8.3

Specific DBP^a(lg DBP/mg-C)

1 h 343 347 400 958

Correlation coefficients for TCM, CAA, and DBP formation models at different order of reaction

E.E. Chang et al. / Chemosphere 64 (2006) 1196–1203 1201