A critical review of trihalomethane and haloacetic acid formation from natural organic matter surrogates

Transcription

1 Environmental Technology Reviews ISSN: (Print) (Online) Journal homepage: A critical review of trihalomethane and haloacetic acid formation from natural organic matter surrogates Tom Bond, Emma H. Goslan, Simon A. Parsons & Bruce Jefferson To cite this article: Tom Bond, Emma H. Goslan, Simon A. Parsons & Bruce Jefferson (2012) A critical review of trihalomethane and haloacetic acid formation from natural organic matter surrogates, Environmental Technology Reviews, 1:1, , DOI: / To link to this article: Accepted author version posted online: 03 Jul Published online: 22 Oct Submit your article to this journal Article views: 3485 Citing articles: 48 View citing articles Full Terms & Conditions of access and use can be found at

2 Environmental Technology Reviews Vol. 1, No. 1, November 2012, A critical review of trihalomethane and haloacetic acid formation from natural organic matter surrogates Tom Bond, Emma H. Goslan, Simon A. Parsons and Bruce Jefferson Cranfield Water Science Institute, Environmental Science and Technology Department, Cranfield University, Bedfordshire, MK43 0AL, UK (Received 15 February 2012; final version received 18 June 2012) Disinfection by-products (DBPs) in drinking water, including trihalomethanes (THMs) and haloacetic acids (HAAs), arise from reactions of natural organic matter (NOM) with chlorine and other disinfectants. The objective of this review was to investigate relationships between the molecular properties of NOM surrogates and DBP formation using data collated for 185 compounds. While formation of THMs correlated strongly with chlorine substitution, no meaningful relationships existed between compound physicochemical properties and DBP formation. Thus non-empirical predictors of DBP formation are unlikely in natural waters. Activated aromatic compounds are well known to be reactive precursors; in addition DBP formation from β-dicarbonyl, amino acid and carbohydrate precursors can be significant. Therefore effective DBP control strategies need to encompass both hydrophobic and hydrophilic NOM components, as well as consider data from NOM surrogates in the context of knowledge from representative treatment scenarios. In future experiments, employing surrogates of NOM is likely to remain a powerful tool in the search for unknown precursors and in understanding their response to various disinfection conditions. Keywords: disinfection by-products; model compounds; chlorination; drinking water; chloroform 1. Introduction The main reason for water disinfection is to prevent the spread of waterborne disease, through the inactivation of microbial pathogens. Partly due to its low cost, chlorine is the commonest chemical disinfectant used in the production of drinking water [1]. Another beneficial feature is its stability, which means a disinfectant residual is maintained in the distribution system, thus preventing bacterial re-growth. In addition to its activity as a disinfectant, chlorine also reacts with organic and inorganic molecules present in water. Reactions with organic molecules can give rise to disinfection by-products (DBPs), many of which are harmful or potentially harmful to human health [2]. The earliest published identification of DBPs in potable water came in 1974 [3]. Prior to this time, although it was appreciated that reactions of chlorine with organic material could produce chlorinated products, their identification was hindered by an absence of analytical methods. By the early 1970s, headspace gas chromatography (GC) was being utilized for the analysis of trihalomethanes (THMs), particularly chloroform (CHCl 3 ) [4]. These techniques were used to demonstrate the link between amount of organic material in water (as measured by colour) and levels of chloroform formed upon chlorination [3]. This breakthrough prompted the US Environmental Protection Agency (US EPA) to initiate a survey of four THMs and two other volatile organic chemicals in 80 waters across the country. THMs were detected in all 79 tap waters where chlorine or chloramine was the applied disinfectant [5]. Soon after, the USEPA began investigating the health impact of THMs and reported chloroform could act as a carcinogen in animal studies [6]. From this point onwards there has been much research dedicated to elucidating the formation, control and health risks of DBPs (Table 1). With growing interest and analytical sophistication has come the realization that many different products can arise from the reactions between organics and chemical disinfectants. Hence the focus has moved from solely THMs to incorporate other classes of DBPs. By 1980 it was appreciated that another group of DBPs, the haloacetic acids (HAAs), could occur in drinking water at levels similar to, or above, those of THMs [7]. In 1987, analysis of treated water from 10 utilities in the USA found 196 compounds thought to be produced from chlorination [8]. In addition to THMs and HAAs, those DBPs present in significant amounts included haloacetonitriles, haloaldehydes, haloketones and halonitromethanes (Table 2). In 2002 the non-halogenated compound N-nitrosodimethylamine (NDMA) was reported Corresponding author. b.jefferson@cranfield.ac.uk Current address: Department of Civil and Environmental Engineering, Skempton Building, Imperial College London, South Kensington Campus, London, SW7 2AZ, UK. ISSN print/issn online 2012 Taylor & Francis

3 94 T. Bond et al. Table 1. Milestones in DBP history. Year Milestone Reference 1974 Chlorination of organic matter in drinking water linked to chloroform formation [3] 1975 USEPA survey of THMs in drinking water across USA [5] 1976 National Cancer Institute classify chloroform as suspected human carcinogen 1976 USEPA investigation into health impact of THMs 1977 THMFP test developed [22] 1979 HAAs indentified in drinking water at levels similar to THMs [7] 1979 THMs regulated at 100 μgl 1 by USEPA 1989 UK regulations: THMs 100 μgl chlorination products identified in treated waters [8] 1993 WHO Guidelines: DCAA 50 μgl 1, TCAA 100 μgl First stage of USEPA D/DBP rule: THMs 80 μgl 1, HAA 5 60 μgl 1, based on annual average 2002 NDMA reported as DBP resulting from monochloramination [9] DBPs reported for chlorine, chloramines, ozone and chlorine dioxide [12] 2006 Second stage of USEPA D/DBP rule: THMs 80 μgl 1, HAA 5 60 μgl 1, based on locational running annual average (LRAA) as a DBP resulting from monochloramination [9] (Table 1), having initially been detected in Canadian drinking water in the 1980s [10]. NDMA and other nitrogenous DBPs, such as the haloacetonitriles and halonitromethanes, are the focus of much recent research because of their high cytotoxicity and genotoxicity compared with regulated DBPs. Their occurrence and control in drinking water has recently been reviewed [11]. Water providers have several available routes to minimize DBP formation. Altering disinfection practice or position, and removing DBP precursors before disinfection have received most attention. At the same time the risk from DBPs has to be balanced against that arising from microbial infection due to incomplete disinfection. However, even allowing for this caution it is likely that DBP regulations will in the future become more stringent and encompass additional DBPs as the health risk becomes less ambiguous. In total over 600 DBPs have been reported in drinking water or laboratory disinfection tests as arising from not only chlorine but also alternative disinfectants such as chloramine, ozone and chlorine dioxide [12]. Nevertheless, in surveyed water treatment works the sum of quantified halogenated DBPs typically only accounted for around 30% of total organic halides (TOX) [12]. In other words around 70% of halogenated DBPs are unidentified. Of all the known DBPs, the THMs and HAAs are still considered to be the dominant groups on a mass basis in potable water [12]. The total THMs and sum of five HAAs are regulated in the USA at 80 μgl 1 and 60 μgl 1 respectively, while the EU has a THM limit of 100 μgl 1. It follows that the identity of formed DBPs is affected by the disinfectant used, disinfection conditions and nature of precursors present in any water. Natural organic matter (NOM) is a complex mixture of many chemical groups that varies both temporally and spatially [13,14]. The major chemical groups in NOM are listed as humic species, carboxylic acids, amino acids, proteins and carbohydrates [15]. In reality this is something of an over-simplification as there are at least 12 important functional groups in NOM: carboxylic acid, enolic hydrogen, phenolic hydrogen, quinine, alcoholic hydroxyl, ether, ketone, aldehyde, ester, lactone, amide and amine [16]. Characterization of NOM is often achieved by fractionation into categories grouped by hydrophobicity [15]. These techniques employ adsorption columns (with ion-exchange or non-ionic resins (e.g. XAD)) to isolate NOM into operationally defined fractions. Hydrophobic NOM contains hydrophobic acids (HPOA, which can further be segregated into humic and fulvic acids) and hydrophobic neutrals (HPON). Carbohydrates, amino acids and carboxylic acids comprise much of the hydrophilic fraction (HPI), which is sometimes further split into hydrophilic acids (HPIA) and hydrophilic bases (HPIB) (Figure 1, Table 3). Some methodologies also separate additional fractions, such as the transphilic acids (TPHA) and hydrophobic/hydrophilic bases. It should be noted that segregations based on chemical group are ill-defined and, for example, amino acids can appear across a spectrum of fractions. Moreover, fractionation of model compounds has demonstrated that individual molecules can appear in multiple fractions [17]. Most NOM is of autochtonous (derived from biota in water) or allochthonous (from the terrestrial watershed) origin. Terrestrial NOM is commonly lignin derived and with high aromatic content, whereas microbially derived substances (for example, from algae, sewage and bacteria) tend to have low aromatic and high nitrogen contents [18]. Hence terrestrial NOM is often described as humic or non-polar and tends to be hydrophobic in character [19], whereas microbial NOM is often termed non-humic or polar and tends to be more hydrophilic. Thus catchment characteristics affect both fractional and chemical composition of NOM. Since NOM classification rarely extends to resolving specific chemical identity, there is uncertainty about the identity of reactive DBP precursors in drinking water. Model compounds have been used as surrogates of NOM since the early days of DBP research [20]. They allow

4 Environmental Technology Reviews 95 Table 2. Important DBPs. Class Structure Important DBPs Trihalomethanes (THMs) Chloroform (CHCl 3 ) Bromoform (CHBr 3 ) Haloacetic acids (HAAs) Dichloroacetic acid (DCAA) Trichloroacetic acid (TCAA) Nitrosamines N-Nitrosodimethylamine (NDMA) N-nitrosodiethylamine (NDEA) Haloacetonitriles (HANs) Dichloroacetonitrile (DCAN) Trichloroacetonitrile (TCAN) Haloacetamides Dichloroacetamide (DCAcAm) Haloaldehydes Dichloroacetaldehyde (DCA) Trichloroacetaldehyde (chloral hydrate) (TCA) Haloketones 1,1,1-trichloropropanone (TCP) Halonitromethanes (HNM) Trichloronitromethane (chloropicrin) (TCNM) for more specific investigation of formation mechanisms and kinetics than the use of natural waters. Further, and in contrast to NOM, model compounds have well-defined physicochemical properties. In general the most important DBP precursors identified have been aromatic compounds, and in this respect the recent report that several aliphatic β- dicarbonyl acid species generate high amounts of THMs and HAAs was notable [21]. Important DBP precursors identified from over thirty years of DBP research are shown in Table 4. The objectives of this review were to investigate relationships between the molecular properties of NOM surrogates and DBP formation and further to highlight the prevalence of different NOM classes as DBP precursors. The emphasis is thus on THMs and HAAs, rather than less-abundant DBPs, since more data is available for the former groups. Correlations between model compound physicochemical properties and DBP formation are then discussed with regard to drinking water studies.

5 96 T. Bond et al. Natural Organic Matter Hydrophobic Hydrophilic Polarity Acids Neutrals & Bases Acids Neutrals & Bases Acid-base character Humic acids Fulvic acids Hydrocarbons Tannins Aromatic amines Carboxylic acids Polyuronic acids Amino acids Peptides Carbohydrates Example compound classes Figure 1. Classification of NOM, based on ref [45]. Table 3. Chemical composition of NOM and significance for DBP formation (adapted and updated from [15]). Impact on DBP formation Additional Chemical group THMs HAAs N-DBPs references Humic species Primary source Primary source Possibly important for HNMs [33,60] Carbohydrates Important at ph8 Probably minor Insignificant [39] Amino acids Minor (except for tryptophan Important for: aspartic acid, Significant [17,31,70 72] and tyrosine) histidine, asparagine, tryptophan Proteins Important during algal blooms Not known, may be significant Uncertain [31,37] Carboxylic acids β-dicarbonyl acids important precursors β-dicarbonyl acids important precursors Probably minor [21,32,34] Note: N-DBP, nitrogenous disinfection by-product; HNM, halonitromethane. 2. Comparing model compound and natural water studies The general principle of formation potential (FP) tests, typically used to chlorinate NOM surrogates, is to provide uniform disinfection conditions and also maximize DBP formation [22]. Nonetheless, there are no standardized parameters for THMFP and HAAFP tests, with disinfectant dose (up to 472 M M 1 ), contact time ( h), temperature (15 25 C) and ph (5 10) all varying widely between studies (Table 5). During full-scale water treatment chlorine doses and contact times are typically much lower and therefore resultant DBP concentrations will also be relatively low. Another important difference is that, with a few exceptions, e.g. [17], the majority of model compound studies exclude bromide, whereas natural waters contain varying amounts of bromide for instance from 20 to 400 μgl 1 in raw water from 12 US water treatment works (WTWs) [12] and in turn generate brominated DBPs. Changes in the ratio of chlorine to bromide can affect the identity of formed DBPs. In general an increase in bromide concentration will shift DBP speciation to more species that are brominated [23]. However, it has been reported that with high chlorine doses, for example 35 M M 1 as used in laboratory DBPFP tests [17,21], excess chlorine can out-compete bromine [24], with bromine incorporation in THMs found to decrease with an increase in the Cl 2 /Br ratio [23]. Some DBPs are also unstable in the presence of free chlorine. In natural waters a decrease in formation of 1,1,1-trichloroacetaldehyde (TCA) and dichloroacetonitrile (DCAN) at higher chlorine doses [25] may reflect the subsequent formation of CHCl 3 and dichloroacetic acid (DCAA) respectively from these intermediate DBPs [26,27]. Thus, formation of such compounds may be suppressed by FP tests using excess chlorine. Finally, pre-oxidation or pre-disinfection steps in water treatment, normally with ozone, chloramines, chlorine dioxide or ultraviolet (UV) irradiation, are becoming more frequent, partly in response to THM regulations, and this will affect DBP speciation

6 Environmental Technology Reviews 97 Table 4. Important precursors of THMs and HAAs. Chemical group Model compound DBPFP Name Structure Name Structure (μg mgc 1 ) Reference Aromatic Substituted benzene Aniline THM: 207 ± 185 [28,36] Substituted phenol Resorcinol THM: 1500 ± 94 [17,20,21,28,30,33,35,36,38] Aliphatic β diketone 2,4-pentanedione THM: 1892 [34] β ketoacid 3-oxopentanedioic acid THM: 1424 ± 451 [17,21,32] HAA: 773 ± 1029 Amino acid L-aspartic acid DCAA: 693; [17] DCAN: 130 Table 5. Chlorination conditions used in selected model compound studies. Compound concentration Chlorine concentration Temperature Contact Study (M unless stated) (M Cl 2 unless stated) ph ( C) time (h) [20] M 7 and [32] [33] 0.55 to M Cl 2 M 1 C [36] Variable 7 and [34] , 7, [37] 5 mg L mg L [31] 2.5 mg L M M 1 or M M (free or combined amino acids) [28] 10 mg L mg L 1 7or [30] N/A 2M Cl 2 M 1 C 7.5 N/A 0.66 [29] 3 mg L 1 asc 28mgL [21] M Cl 2 M 1 compound 5.5 or [33] 2 mg L 1 asc 10mgL 1 asc 5,8, [17] or M Cl 2 M 1 compound [40] 0.1 and mg L Note: N/A, not available. in finished water by reacting with precursor sites. Thus DBPFP experiments using model compounds are designed to elucidate reactive precursors, kinetic behaviour, formation mechanisms and such information, rather than for direct comparison with natural waters. 3. Methods To elucidate the relationships between THM formation and compound physicochemical properties, Pearson productmoment correlations (r) were calculated for 184 compounds taken from 16 studies, spanning the years [17,20,21,25,28 40]. For 27 compounds THM data appears in multiple studies, albeit sometimes under different names (Table 6 and supplementary information). For these the mean and standard deviation of THMFP were quoted. Where HAA or chlorine demand data also appears in multiple studies the same approach was taken. All THM data was converted in to units of μgmgc 1 to facilitate data comparison. HAA data for a subset of 48 compounds were available [17,21,29,40]. Chlorine substitution efficiency for THMs (% mol Cl substituted in THMs per mol Cl 2 consumed)

7 98 T. Bond et al. was determined from THMFP and chlorine demand data where available. Compounds were assigned into predicted fractions based on the results of Bond et al. [17], who fractionated 21 NOM surrogates using a method based on that of Croué et al. [15]. Thus compounds of high hydrophobicity (log K OW 0.8 or over) were designated HPOA, compounds of low hydrophobicity (log K ow 1.1 or below) were HPI and compounds of intermediate hydrophobicity were of uncertain fraction, as they could be HPON, TPHA or the two fractions above. Molecular properties collated were: molecular weight (MW), octanol water partition coefficient (log K OW ),pk a, molar volume (Vm), surface tension (γ ), polarizability (α), density, soil water partition coefficient (log K OC ) and aqueous hydroxyl rate constant (k OH ) (supplementary information). Log K OC values were estimated using two different models: the Sabljic molecular connectivity method with improved correction factors, and the traditional method based on log K OW [41]. Remaining properties were taken from various chemical databases [41 44], with experimental values being used wherever available (this includes fractionation behaviour). Relationships were evaluated using the Pearson product-moment correlation coefficient (r). This coefficient is a dimensionless index used to measure the degree of linear relationship between two variables, and assumes a value between 1 and Factors affecting DBP formation Traditionally there has been a perception that humic substances are the major source of THM and HAA precursor sites [45]. Reflecting this, of the 185 compounds collated for the current study, 84 were assigned into the HPOA fraction, 53 into HPI and 44 were of uncertain fraction, though the latter also includes many hydrophobic compounds based on chemical functionality (Table 6). Of the 53 HPI compounds, 22 are amino acids, 12 carbohydrates and 12 carboxylic acids (Table 6). There are some conflicting reports about the identity of THM and HAA precursors in natural waters; although, owing to the limited HAA data, it is difficult to make inferences of this type from Table 6. To illustrate, one study concluded that HAA precursors have a higher aromatic content than THM precursors [46]. Conversely, other research proposes that the hydrophilic fraction produces a higher proportion of HAAs relative to THMs than the hydrophobic fraction [20,47]. It has been proposed that waters which produce high THM levels may also have a propensity to generate trichloroacetic acid (TCAA), and further that DCAA precursors are overall less hydrophobic than TCAA precursors [46], which correlates with high DCAA formation from aliphatic model compounds, particularly aliphatic amino acids and β-dicarbonyls [17,21]. DCAA and TCAA are thought to be produced as a result of differing mechanistic pathways [26], with TCAA not readily formed from direct chlorine substitution of DCAA. Meanwhile TCAA formation has been likened to THM formation and may proceed through common intermediates, whereas raised DCAA levels have been linked to the presence of diketones and then aldehydes after oxidation [25]. Formation of TCAA over CHCl 3 from a trichloroacetyl precursor structure is favoured by the presence of conjugation capable of stabilizing the formed carbonium ion [25]. In general the presence of bromide (Br ) increases levels of halogenated DBPs. This is significant as not only are brominated DBPs of higher mass than chlorinated analogues but they are often also more toxic. Bromine species (HOBr/OBr ) are known to be more effective substitution agents than the equivalent chlorine species [23]. In natural water studies, it has been observed that 5 10% of HOCl typically became incorporated into THMs, while bromine incorporation levels were higher at around 50% [48]. The higher reactivity of bromine than chlorine in HAA formation has also been reported [49]. The effect of ph on DBP formation is complicated and can favour formation of certain products over others (Table 7). Generally any effects occur because acidic or basic conditions increase the speed of a rate-determining reaction step. For instance, the higher THM formation from carbohydrates at ph 8 than at ph 5 has been explained by basic conditions promoting the rate-determining hydrolysis of the halogenated leaving group (Scheme 1), while for 3-oxopentanedioic acid an increased yield was associated with a fall in ph from 8 to ph 5.5 (Table 7). During full-scale water treatment DCAA was observed to be relatively insensitive to ph shifts, while higher ph levels have been reported to increase THM levels and decrease TCAA levels (Table 7). One explanation for this pattern would be THM and TCAA precursors being similar [26] and higher ph levels favouring base-catalysed hydrolysis of the halogenated leaving group. This route produces chloroform, while electron-pair donation gives rise to TCAA formation (Scheme 2). Both mechanisms are possible in postulated models [20,25]. The instability of DCAN and 1,1,1-trichloropropanone (TCP) at ph 7 and 8 (Table 7) is likely to translate to increased DCAA and CHCl 3 formation respectively from these intermediate DBPs [26,27]. 5. Chlorination of NOM Although chlorine is dosed as a gas or as sodium hypochlorite, it is hypochlorous acid (HOCl) which is the major reactive form during water treatment. Since hypochlorous acid is an electrophile, it tends to react with electron-rich moieties in NOM. Oxidation, addition and electrophilic substitution reactions are all possible pathways. Normally only electrophilic attack is significant in reactions with organics, based on kinetic analysis. Second-order rate constants for reactions of chlorine and organics vary widely, from 0.1 to 10 9 M 1 s 1 [50] and chlorine reacts selectively with certain chemical functionalities. Amines, reduced sulphur moieties and activated aromatic functionalities are all highly reactive towards chlorine and have rate constants

8 Table 6. Model compounds DBPFP, predicted fraction, chlorine demand and chlorine substitution efficiency. THMFP; Cl 2 Predicted HAAFP Other DBPs demand Cl substn % Name References fraction μg mgc 1 μg mgc 1 mol mol 1 mol mol 1 β-alanine [31] HPI ,2,3-trihydroxybenzene [36] HPOA ,2,4-trihydroxybenzene [36] Uncertain ,3,5-triazine-2,4,6-triamine [40] HPI 0 0 1,3,5-trihydroxybenzene [36] Uncertain ,3-dihydroxy-4-chlorobenzene [36] HPOA ,3-dimethoxybenzene [20] HPOA 0 1,4-dihydroxybenzene [36] Uncertain hydroxy-3-methoxybenzene [20] HPOA naphthol [36] HPOA ,3,6-trichlorophenol [36] HPOA ,3-dichlorophenol [36] HPOA ,4,6-trichlorophenol [36] HPOA ,4-dichlorophenol [36] HPOA ,4-dihydroxybenzoic acid [38] HPOA ,4-pentanedione [34] Uncertain ,6-dihydroxybenzoic acid [38] HPOA ,6-dihydroxytoluene [38] HPOA butanone [38] Uncertain 2 2-Ethyltoluene [28] HPOA 12 2-naphthol [36] HPOA oxobutyric acid [21] Uncertain oxopentanedioic acid [21] HPI pentanone [38] HPOA 2 3,4,5-trichlorophenol [36] HPOA ,5-dichlorophenol [36] HPOA ,5-dihydroxybenzoic acid [20,32,33,38] HPOA 697 ± ± ± ,5-dihydroxytoluene [38] HPOA ,5-dimethoxybenzoic acid [30,33] HPOA 8 ± ± ± ,5-heptanedione [34] HPOA 14 3-hydroxybenzoic acid [29] HPOA 44; hydroxybutyric acid [21] Uncertain 72; N,N-DAPSIS [40] Uncertain nitroaniline [36] HPOA nitrobenzoic acid [36] HPOA oxobutandioic acid [21] HPI 15; (Continued) Environmental Technology Reviews 99

9 Table 6. Continued THMFP; Cl 2 Predicted HAAFP Other DBPs demand Cl substn % Name References fraction μg mgc 1 μg mgc 1 mol mol 1 mol mol 1 3-oxohexanedioic acid [21] HPI 1378; oxopentanedioic acid [17,21,32] TPHA/HPI 1424 ± (1,1,1-TCP) 4.3 ± ± ± (4-hydroxyphenyl)phenol [36] HPOA ,6-dichloro-1,3-dihydroxybenzene [36] HPOA ,6-dioxoheptanoic acid [21] Uncertain 1223; chlorobenzoic acid [36] HPOA hydroxybenzoic acid [40] HPOA 23; oxoheptanedioic acid [21] HPI 7; ,7-dioxooctanoic acid [21] Uncertain 1133; Acetamide [17,36] HPI 1 ± 2; ± ± 0.35 Acetic acid [17,36] HPI 1 ± ± Acetone [28] Uncertain 564 Acetophenone [36] HPOA Acetylacetone [36] Uncertain Allantoin [40] HPI 0 0 Aniline [28,30,36] HPOA 207 ± ± ± 1.20 Anisole [36] HPOA Arabinose [17] HPI 12; Benzaldehyde [36] HPOA Benzene [28] HPOA 12 Benzoic acid [28,30,36] HPOA 4 ± ± ± 0.05 Benzenepropanoic acid, 4-hydroxy-3,5-dimethoxy- [33] HPOA Benzalkonium chloride (BKC) [40] HPOA 10; Bovine serum albumin (BSA) [37] Caffeine [40] Uncertain 0; 0 Catechol [20,36] HPOA 7 ± Chlorophyll [35] Uncertain Chloroxylenol [36] HPOA Citraconic acid [21] Uncertain 8; Citric acid [32] HPI 1293 Cytochrome [37] Uncertain 41 Diethylaniline [36] HPOA Dimethylamine [40] Uncertain 1; DL-Isoleucine [31] HPI T. Bond et al.

10 DL-Leucine [31] HPI DL-Threonine [31] HPI D-mannose [17] HPI 0; D-xylose [17] HPI 17; Erythrulose [39] HPI Ethanol [36] Uncertain Ethyl acetoacetate [25] Uncertain Ethylbenzene [28] HPOA 33 Ferulic acid [17,33] HPOA 29 ± 27; (TCA) Fructose [39] HPI Fumaric acid [32] Uncertain 573 Galactose [39] HPI Glucose [35,39] HPI 24 ± Glyceraldehyde [39] Uncertain Glycine [17,31,40] HPI 9 ± 14; 1 ± ± ± 0.18 Glyoxalic acid [36] HPI Hesperetin [20] HPOA 349 Hesperidin [20] Uncertain 114 Hexane [28] HPOA 18 Humic acid [37] HPOA 77 Hydroquinone [20] Uncertain 25 isocitric acid [32] HPI 17 Isopropanol [28] Uncertain 147 L-alanine [31] HPI L-arginine [31] HPI 1 ± 1; L-asparagine [17,31,36] HPI 2 ± 2; (TCA), (DCAN) 5.1 ± ± 0.26 L-aspartic acid [17,31] HPI 31 ± 33; (TCA), (DCAN) 5.6 ± ± 0.66 L-cysteine [31] HPI 0; L-glutamic acid [17,31] HPI 0 ± 1; ± ± 0.05 L-glutamine [31] HPI 0; L-histidine [31] HPI 13; L-leucine [17] HPI 0; L-lysine [31] HPI 3; L-methionine [31] HPI 0; L-ornithine chlorohydrate [31] HPI L-phenylalanine [31] HPI 0; L-proline [31] HPI L-serine [17,31] HPI 1 ± 2; ± ± 0.06 L-tryptophan [17,31] HPOA/ 219 ± 14; (DCA), ± ± 2.28 HPON 66 (TCA), (DCAN) L-tyrosine [17,31] HPI 115 ± 18; ± ± 0.45 L-valine [31] HPI (Continued) Environmental Technology Reviews 101

11 Table 6. Continued THMFP; Cl 2 Predicted HAAFP Other DBPs demand Cl substn % Name References fraction μg mgc 1 μgmgc 1 mol mol 1 mol mol 1 Maleic acid [30] Uncertain Malic acid [32,36] HPI 14 ± Malonic acid [25,36] Uncertain 2 ± ± Maltopentaose [39] HPI Maltose [39] HPI Maltotriose [39] HPI m-aminophenol [36] Uncertain m-chlorophenol [36] HPOA m-cresol [28] HPOA 157 Methoxyacetic acid [21] Uncertain 3; Methylfurfural [17] HPON 65; (1,1,1-TCP) m-hydroxyacetophenone [36] HPOA m-hydroxybenzaldehyde [36] HPOA m-hydroxybenzoic acid [36] HPOA m-methoxyphenol [36] HPOA m-methylphenol [36] HPOA m-nitrophenol [36] HPOA m-xylene [28] HPOA 60 N-acetylneuraminic acid [21] HPI 5; Naphthalene [28] HPOA 29 Naphthoresorcinol [35] HPOA n-butyraldehyde [36] HPOA Nitrobenzene [28,36] HPOA 12 ± 12 o-aminophenol [36] Uncertain o-chlorophenol [36] HPOA o-cresol [28] HPOA 143 o-hydroxyacetophenone [36] HPOA o-hydroxybenzaldehyde [36] HPOA o-methoxyphenol [36] HPOA o-methylphenol [36] HPOA o-nitrophenol [36] HPOA Orcinol [33,38] HPOA 969 ± ± ± Oxalic acid [17,36] HPI 42 ± 55; ± oxaloacetic acid [32] HPI 42 p-aminophenol [36] Uncertain p-chlorophenol [36] HPOA p-coumaric acid [17] HPOA 19; (TCA) p-cresol [28] HPOA T. Bond et al.

12 Pepsin [37] Uncertain 51 Phenol [28] HPOA 154 Phenylacetic acid [36] HPOA Phenylalanine [36] HPI Phenylthiourea [36] Uncertain Phlorizin [20] Uncertain 446 Phloroglucinol [29] Uncertain 332; p-hydroxyacetophenone [36] HPOA p-hydroxybenzaldehyde [36] HPOA p-hydroxybenzoic acid [36] HPOA p-methoxyphenol [36] HPOA p-methylphenol [36] HPOA p-nitrophenol [36] HPOA Propionaldehyde [36] Uncertain Pyrogallol [20] HPOA 0 Pyruvic acid [36] HPI Quercetin [35] HPOA Quinone [20] Uncertain 33 Rennin [37] 17 Resorcinol [17,20,21,28,30,33,35,36,38] HPOA 1500 ± 94; ± ± 3.04 Ribose [39] HPI Rutin [20] HPI 258 Salicylic acid [32,36] HPOA 42 ± Sinapic acid [17,33] HPOA 24 ± 5; (TCA) 7.7 ± ± 0.26 Styrene [28] HPOA 44 Succinic acid [36] Uncertain Syringic acid [30,33] Uncertain 5 ± ± ± 0.08 Tannic acid [17] HPOA Thioacetamide [36] Uncertain Thiourea [36] Uncertain Toluene [28] HPOA 23 Triethanolamine [40] Uncertain 0 0 Uracil [35] Uncertain Urea [36] HPI Vanillic acid [30,33] HPOA 84 ± ± Xylose [39] HPI Note: the two entries for hesperidin refer to two different compounds of the same name. HAAFP values are in bold. Environmental Technology Reviews 103

13 104 T. Bond et al. Table 7. Effect of ph on DBP formation. Precursor DBP/s ph change Effect References 3-oxopentanedioic DCAA [21] acid Citric acid THMs Higher levels at ph 7 [21,32,36] Amino acids DCAN/DCAA From acidic/neutral to alkaline Increase in hydrolysis of DCAN [27,70,71] to DCAA Amino acids TCA 7 8 [70] Carbohydrates THMs [39] Natural water THMs [61] Natural water TCAA Equal at 5 and 7, decrease at 9.4 [61] Natural water DCAA Insensitive [61] Natural water DCAN Stable at 5, decrease with time [61] at 7, insignificant formation at 9.4 Natural water TCA Equal at 5 and 7, decrease at 9.4 [61] Natural water TCP, DCAN TCP and DCAN unstable at 7 [66] and 8, stable at 6 Natural water TCAA Decrease at 8 [66] Natural water DCAA Insensitive [66] Scheme 1. Chlorination of carbohydrates, based on ref [39]. towards the upper end of the range listed; for example the apparent rate constant for the reaction between chlorine and the amino acid cysteine is at ph 7, because of the reactivity of a sulphur-containing side group [51]. Hypochlorous acid also reacts rapidly with amines to produce organic and inorganic chloramines. For the less reactive moieties, reactions with chlorine can be too slow to have an impact during the time span of water disinfection. For instance, reactions of HOCl with alkenes are typically too slow to be relevant during water treatment, as illustrated by the negligible apparent rate constant for reaction with the steroid progesterone [50,52]. The speed of chlorine addition to alkenes can increase if the double bond is activated by electron-donor groups. Similarly reactions with alcohols are very slow, e.g. the apparent rate constant of 0 at ph 7 for reaction with the monosaccharide ribose, but can lead to oxidation to ketones and aldehydes [53]. Likely reaction sites can be predicted based on the following order of reactivity, bearing in mind nearby electron-donor or electron-withdrawing groups will also have an effect: reduced sulphur groups > primary and secondary amines > phenols, tertiary amines >> double bonds, other aromatics, carbonyls, amides [50] Chlorination of humic substances Aquatic humic substances or hydrophobic NOM is operationally defined as coloured, polyelectrolytic acids isolated by sorption on to XAD resins or equivalent procedure. Major functional groups include carboxylic acids, phenolic and alcoholic hydroxyls, and keto groups [16]. The average pk a value of humic substances is 4.2 and their specific chemical structure is generally unknown [16]. They include both humic and fulvic acids, are generally derived from terrestrial vegetation and have high lignin, and consequently aromatic, content. Aquatic fulvic acids tend to have MW from 500 to 2000 Da; aquatic humic acids are larger and often colloidal, from 2000 to 5000 Da and sometimes up to 100,000 Da [54]. Humic substances typically comprise around 50% of the dissolved organic carbon (DOC) of an average river [16] and up to 76% for water draining a moorland catchment [55]. Chlorine reacts with aromatic compounds by electrophilic substitution. In the presence of an electrondonating and ortho-para directing group, for example phenol, stepwise chlorination occurs at the 2, 4, and 6 positions respectively, to give THM formation of 154 μg mgc 1 (Table 6). Work using model compounds has proposed the major reactive sites within fulvic acids as the carbon between two hydroxyl groups or one hydroxyl and one O-glucoside group [20], with resorcinol being the most important THM precursor at 1500 ± 94 μg mgc 1 (Table 6). In organic chemistry the classical route for THM synthesis is known as the haloform reaction [56]. It consists of a series of reactions of enolizable compounds, depicted for a methyl ketone (Scheme 3), whose rate is normally

14 Environmental Technology Reviews 105 Scheme 2. Chlorination of resorcinol. Cleavage at A will result in the production of CHCl 3 and cleavage at B will form TCAA (based on [20,38]). Scheme 3. Haloform reaction for a methyl ketone, redrawn from Larson and Weber [56]. determined by the rate of enolization of the initial carbonyl precursor molecule. Boyce and Hornig [38] proposed a related reaction mechanism for resorcinol whereby electrophilic substitution of chlorine and a complex series of hydrolysis and decarboxylation reactions lead to chloroform formation. A simplified version is shown in Scheme 2. Resorcinol can be redrawn as a masked β-dicarbonyl compound [56] (Scheme 4), in which case its participation in haloform reactions involving enolization followed by chlorine substitution becomes much clearer. Resorcinol-type structures have been classified as fast-reacting THM precursors, while more slowly reacting THM precursors include phenolic compounds [57]. This is also seen by comparison of rate constants for reaction with chlorine: 0.36 M 1 s 1 and M 1 s 1 for phenol and resorcinol respectively [57,58]. Resorcinol structures are thought to be commonly contained within macromolecular humic species found in natural waters [59]. Moreover, hydroxyl phenols, including resorcinol, as well as gallic acid, vanillin and catechol are degradation products of plant matter, including lignin or tannin. However, they are usually present at concentrations, under 1 μgl 1 [16], which are insufficient to account for THM formation in finished water. Aromatic compounds, including resorcinol, can also act as halonitromethane precursors when chlorinated in the presence of nitrite [60]. The reactivity of aromatic compounds can be explained in terms of the electron-donating or electron-withdrawing influence of substituents [50]. The high reactivity of resorcinol is thus ascribable to having two activating ortho-para hydroxyl groups in the one and three positions. Scheme 4. Resorcinol as a masked β-dicarbonyl, redrawn from Larson and Weber [56].

15 106 T. Bond et al Chlorination of carboxylic acids Carboxylic functional groups occur on aquatic humic substances at a frequency of 5 to 10 per molecule [16]. Under ambient ph conditions of most waters (ph 6 8) such groups are anionic (i.e. dissociated). Carboxylic acids of one type or another can represent approximately 90% of organic carbon in water. However, since they include aromatic, aliphatic and polymeric compounds, carboxylic acids are often contained within other classifications of NOM, notably colloids, humic substances and amino acids. Those which do not fall under such classification may account for 5 8% of DOC; in particular non-volatile fatty acids (over C5), such as palmitic and stearic acids, are the most abundant and may account for around 4% [16]. In general, simple carboxylic acid moieties, like fatty acids, are not reactive with chlorine, as shown by an apparent rate constant of 2.3 M 1 s 1 at ph 7.2 for reaction with sorbic acid [58], which also contains an alkene functionality. A corollary is the low DBP formation from simple carboxylic acids, as shown by a THMFP of 2 μg mgc 1 for acetic acid (Table 6). DBP formation from palmitic and stearic acids has been little studied but is presumably similarly low. A significant exception is the high chlorine reactivity and DBP formation found for certain β- dicarbonyl acids, illustrated by respective THM and HAA formation of 1424 ± 451 and 773 ± 1029μg mgc 1 for 3- oxopentanedioic acid (Tables 4 and 6). These high standard deviations are most likely explained by high 1,1,1-TCP formation, 987 μg mgc 1 in one study [17]. TCP is an intermediate DBP known to degrade to chloroform in the presence of free chlorine, especially at high ph [61]. 3- Oxopentanedioic acid is found in both TPHA and HPI fractions ([17], Table 6) As with the carboxylic acid functionality, simple carbonyl groups react slowly with chlorine, demonstrated by a negligible apparent rate constant for reaction of chlorine with the steroid progesterone [50]. Reaction with carbonyl groups normally proceeds through initial chlorine substitution at the α-carbon to the carbonyl group. With β-dicarbonyl species the electron-withdrawing effect of both carbonyls makes the hydrogen groups attached to the α-carbon more acidic. Both acid- and base-catalysed enolization can lead to DBP formation (Scheme 5). The higher TCA formation of fulvic acid isolates than humic acid isolates [26] was linked to higher methyl ketone content, which could include β-dicarbonyl species. Basecatalysed halogenation of β-dicarbonyls is dominant above ph 5 and kinetically controlled by keto-enolization [50]. Thus it may be expected that DBP formation from β- dicarbonyl species would increase with ph. However, the higher DBP formation reported for 3-oxopentanedioic acid (a β-dicarbonyl acid), at ph 5.5 compared with ph 8, with DCAA formation of 2062 and 1462 μg mgc 1 respectively [21], indicates this is not necessarily the case. THM formation from citric acid is also highly ph dependent, with high levels at ph 7 but not ph 8 explained by neutral ph being optimum for the rate-determining oxidative decarboxylation step (Table 7). As well as the complexities of ph dependence, the exact route by which HAAs and THMs are liberated from β-dicarbonyls has still to be elucidated. A route by which β-keto acids can give rise to DCAA is shown for 3-oxopropanoic acid (Scheme 5). Meanwhile, Scheme 6 shows a possible route by which 5,7-dioxooctanoic acid could give rise to both DCAA and CHCl 3. There is evidence DCAA precursors are different to TCAA and THM precursors. However, model compound work has identified a small number of precursors which produce high levels of both DCAA and THMs. The most striking example is the aforementioned 3-oxopentandioic acid, which can produce CHCl 3 at up to 1414 μg mgc 1 and DCAA at 1500 μg mgc 1 [21]). The most likely explanation for this is that various possible degradation pathways after chlorine substitution can liberate both DCAA and CHCl 3 (e.g. Scheme 6) Chlorination of amino acids and proteins Amino acids are typically present at mean levels of 0.3 mg L 1 in surface waters, representing some 2 5% of the total NOM, though in eutrophic lakes concentrations can rise to 13% (median concentration 600 μgl 1, range μgl 1 ) [16]. The amino group behaves as an organic base and, depending on the pk b (base dissociation Scheme 5. Chlorination of 3-oxopropanoic acid. Scheme 6. Chlorination of 5, 7-dioxooctanoic acid, based on [21].

16 Environmental Technology Reviews 107 constant) value, amino acids can behave as a cation, anion or zwitterion. Glutamic acid, glycine, serine and aspartic acid are considered the most abundant free amino acids [16]. These four species all have low THMFP (0 9 μg mgc 1, Table 6) and lie within the HPI fraction of NOM (log K OW = 3.07 to 3.89, Table 6). Nevertheless, in natural waters amino acids may commonly occur in the colloidal or hydrophobic fractions [16]. Combined amino acids are thought to be four to five times commoner than free amino acids [31], which is significant as amide groups involved in peptide links are unavailable for reaction with chlorine and do not act as a DBP precursor. The chlorine demand of polypeptides can be theoretically calculated from the demand of constituent parts, bearing in mind that the amide/peptide bond and also glycine and aspartic acid are non-reactive [31]. Free amino acids react strongly with chlorine, with chlorine demand as high as 12.6 ± 1.2 and 14.8 ± 1.7 mol mol 1 for tyrosine and tryptophan respectively (Table 6), which is related to their reactive side groups. Similarly, side groups including amine, sulphur or activated aromatic groups are presumed to be the main precursor sites of linked amino acids. Tryptophan generated 219 ± 14 μg mgc 1 of THMs, as well as 222 μg mgc 1 of TCA and 76 μg mgc 1 of DCAN (Table 6). For α-amino acids, initial formation of organic mono- or dichloramines, depending on the chlorine dose, followed by decarboxylation and deamination can produce carbonyl or nitrile compounds (Scheme 7). For L-aspartic acid it has been proposed that this can lead to the predominant formation of 3-oxopropanoic acid, a β-keto acid, as a reaction intermediate at ph 8 [31] (Scheme 8). This significance of this became apparent when the high DBP formation of several similar β-keto acids was reported [21], followed by the high DCAA formation of L-aspartic acid itself, 693 μg mgc 1 [17]. Like asparagine and tryptophan this precursor also forms significant amounts of TCA (77 μg mgc 1 ) and DCAN (130 μg mgc 1 ) (Table 6). Aspartic acid and asparagine are both thought to generate a β-keto acid from chlorine oxidation [21] and are unusual amongst amino acids in being represented by low chlorine demand but high DBP formation [31]. For the former to occur, the amino acid must have two terminal oxygenated groups and a four-carbon backbone, which can become a β-keto acid through loss of the alpha carboxylic group. Nitrile formation could also give rise to DCAA and TCAA based on the classical mechanism of amino acid chlorination. However, given the unfavourable kinetics of chlorination of single carboxylic groups, it is more likely that DCAA formation proceeds through the β-keto acid intermediate. It is known that DCAN hydrolyses to 2,2- dichloroacetamide (DCAcAm) and consequently DCAA in the presence of free chlorine [27]. Therefore it is likely that those amino acids which generate DCAA and DCAN will likewise produce DCAcAm. This was confirmed recently for aspartic acid and tyrosine, though DCAcAm yields were much lower than related DBPs [62]. Meanwhile, in finished drinking water DCAcAm has occurred at a similar level to DCAN [12], and HAN 4 (DCAN, bromochloroacetonitrile (BCAN), dibromoacetonitrile (DBAN) and trichloroacetonitrile (TCAN)) typically represents 7 8% of the sum of the four THMs and nine HAAs on a mass basis [11,63]. It has been reported that levels of total dissolved amino nitrogen, presumed to mainly comprise proteinaceous material, in a lake water rose from mg L 1 to 1.0 mg L 1 during an algal bloom [37]. From the algal bloom a THMFP of 115 μgl 1 was reported [37]. Meanwhile, the highest THMFP recorded from four model proteins was 51 μg mgc 1 for pepsin (Table 6) Chlorination of carbohydrates Total dissolved carbohydrates in rivers and lakes can respectively account for 5 10% and 8 12% of total DOC (mean concentration 500 μgl 1 for both, maximum Scheme 7. Chlorination of amino acids, based on [52,70,71]. Scheme 8. Chlorination of aspartic acid, based on [31].

17 108 T. Bond et al. 2mgL 1 in river water) [16]. A recent study found concentrations of 1 mg L 1, or 50% of the total DOC, in a Spanish river [39]. Glucose is considered the commonest carbohydrate in surface waters [16], while arabinose and mannose are also widespread [64]. All these carbohydrates presumably belong to hydrophilic NOM fractions (Table 6). As noted, functionalities contained within carbohydrates are slow to react with chlorine, illustrated by the negligible apparent rate constant for the monosaccharide ribose [58]. Navalon and co-workers [39] found ph to have a strong affect on the THM formation. At ph 5 only small amounts of THMs were observed, though production became significant at ph 8, for example from glucose 44 μg mgc 1 and from maltotriose 65 μg mgc 1 (Table 6). After 72 hours it was thought that reactions had still not reached completion, revealing the slow kinetics of carbohydrate chlorination. Though these THM values are still much lower than from more reactive precursors, they may become considerable given the ubiquity of carbohydrates in surface waters. The presence of bromide increases THMs still further. THM formation from glucose chlorination increased by 100% in the presence of 300 μgl 1 bromide, relative to no bromide, from 44 to 89 μg mgc 1 [39]. At bromide concentrations under 100 μgl 1, complete incorporation of bromide into THMs was observed during carbohydrate chlorination [39]. Most of the chlorine substituted at ph 8 can be accounted for by THMs, indicating formation of other DBPs is not significant. The proposed mechanism proceeds through chlorine substitution of the α-hydroxy aldehyde moiety (Scheme 1). The ph dependence has been ascribed to basic conditions promoting the rate-determining hydrolysis of the halogenated leaving group. 6. Correlations between model compound properties and THM formation Model compound THMFP was positively correlated (r = 0.849) with chlorine substitution efficiency (Table 8). The number of data points used for this correlation was 133, with a clear linear relationship observed between these two parameters, notwithstanding the presence of many data points close to zero (Figure 2). This underlines the importance of the chlorine substitution step to formation of DBPs. Conversely, there was no significant correlation (0.189) between THMFP and chlorine demand, which indicates most chlorine consumed is involved in oxidation rather than substitution reactions, as previously noted [50]. This trend is clearly illustrated by many of the aliphatic amino acids, such as L-glycine, which have significant chlorine demand but low THMFP, in this case 5.6 ± 0.0 mol mol 1 and 9 ± 14 μg mgc 1 respectively (Table 6). There were no meaningful relationships between any of the physicochemical properties and THM formation. This can be explained by compounds with similar physicochemical properties having disparate THM formation potential. In most cases this is due to the position of activating or deactivating groups. To Table 8. Correlations between compound properties and THMFP and HAAFP. THMFP HAAFP Cl2 demand Cl2 substn MW Vm γ α Density Log μg mgc 1 μg mgc 1 mol mol 1 mol mol 1 log KOW pka Da cm 3 dyne cm cm 3 gcm 3 KOC HAAFP Cl2 demand Cl substn Log Kow pka MW MV γ α Density Log KOC koh (10 8 M 1 s 1 )