Generation of hydroxyl radical during chlorination of hydroxyphenols and natural organic matter extracts

The generation of hydroxyl radicals ( OH) during the chlorination of air saturated solutions of different hydroxyphenols (hydroquinone, resorcinol, catechol, gallic and tannic acids) at pH 7 has been determined by the formation of phenol (in presence of benzene in excess) or 2-hydroxyterephthalic acid (in presence of terephthalic acid). Formation of OH was only detected during the chlorination of oor phydroxyphenols, compounds that react with chlorine by electron transfer forming the corresponding semiquinones/quinones. In aerated solutions, oxygen is reduced by the semiquinone to the superoxide radical, O2 , which reacts with HOCl to OH. Compared to the studied o-hydroxyphenols, the lower reactivity of hydroquinone towards chlorine favours the reaction between chlorine and O2 , and its OH formation potential is ~50 times higher. The extent of OH generated increased with the concentration of the hydroxyphenol and chlorine, but the OH yield (moles formed per mole of hydroxyphenol eliminated), decreased due to the formation of the quinone, that acts as O2 scavenger. The yield was almost not affected by the pH (6 pH 7.5), whereas a strong impact of dissolved O2 was observed. The OH production was null in absence of O2 and 2.5e3 times higher at oxygen saturated conditions compared to air-saturated. Contrary to chlorination, during bromination of hydroquinone OH was not formed, which can be attributable to a much faster consumption of the oxidant, with no chance for O2 to react with bromine. Formation of OH during the chlorination of different NOM extracts (SRHA, SRFA, PLFA and Nordic Lake NOM) and water from Lake Greifensee (Switzerland) was also studied using terephthalic acid as OH scavenger. For SRHA, SRFA and Nordic Lake NOM (all of allochthonous origin and presenting high electron-donating capacity, EDC), OH yields expressed as moles formed per mole of DOC0 (%), were between 1.1 and 2.0, similar to that of hydroquinone (~1.5). For PLFA and Lake Greifensee water (autochthonous, lower EDC) much lower OH yields were observed (0.1e0.3). Both chlorination rate and EDC, the later favouring the formation/stabilization of O2 , seem to be key factors involved in OH generation during the chlorination of NOM. A mechanism for these findings is proposed based on kinetic simulations of hydroquinone chlorination at pH 7. © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Dissolved organic matter (DOM) is typically the main sink for oxidants commonly applied in drinking water treatment (Wenk et al., 2013;. The oxidants react with electronrich functional groups such as phenolic moieties, amines, olefines, etc., partially leading to the formation of low molecular weight organic disinfection byproducts (DBPs) (Lee et al., 2007;Bond et al., 2012;von Sonntag and von Gunten, 2012;Le Roux et al., 2016). Among the different structures present in DOM, phenolic moieties are highly reactive towards many different oxidants. The content of aromatic moieties in DOM has been usually measured in terms of the specific ultraviolet absorbance at 254 nm, SUVA 254 (Weishaar et al., 2003), however, recently more specific measurements such as the electron-donating-capacity (EDC) are being used as a proxy of the phenolic content (Aeschbacher et al., 2012;Walpen et al., 2016;€ Onnby et al., 2018). Although, in general, phenolic structures react with chlorine by electrophilic aromatic substitution (EAS) forming chlorophenols, recent studies have demonstrated that chlorination of ortho and para hydroxyphenols occurs via an electron transfer (ET) with the formation of the corresponding quinones. These reactions are likely responsible for the high initial hypohalous acid consumption by DOM through ET reactions (Criquet et al., 2015).
The chemistry of o-and p-hydroxyphenols in water is complex. According to Eyer (1991), hydroquinone (HQ) can be oxidized by dissolved oxygen in aqueous solution through the following mechanism, based on the hydroquinone/semiquinone/benzoquinone (HQ/SQ À /BQ) equilibria: At pH 7 equilibrium 1 is almost completely displaced to the left (K 1~1 0 À14 at pH 7, room temperature and air saturated, Eyer, 1991), and the oxidation of HQ to SQ À by O 2 , with generation of superoxide radical (O 2 À ) is minimal. If SQ À were formed, it would be quickly oxidized by O 2 to BQ (reaction 2, k 2 ¼ 5 Â 10 4 M À1 s À1 at pH 7 and room temperature, Eyer, 1991), under formation of more O 2 À . BQ could then react with HQ (k 3 ¼ 58 M À1 s À1 , k -3 ¼ 8 Â 10 7 M À1 s À1 at pH 7, Yamazaki and Ohnishi, 1966), and also with O 2 À (k -2 ¼ 10 9 M À1 s À1 at pH 7, Eyer, 1991), both reactions leading to SQ À . With a pKa ¼ 4.8 for equilibrium 4 (Bielski et al., 1985), at pH 7 the formation of hydroperoxyl radical (HO 2 ) from superoxide protonation is expected to be very low, whereas O 2 À disproportionation to H 2 O 2 (reaction 5) is slow leading to a certain stability of O 2 À in solution under these conditions (Sheng et al., 2014). Although the reaction between HO 2 and O 2 À to HO 2 À has a high second-order rate constant (k 6 ¼ 9.7 Â 10 7 M À1 s À1 , Sheng et al., 2014), at pH 7 this reaction is not favored due to the low concentration of HO 2 (~0.6% of [O 2 À ]). In the above mechanism the possible reaction between HQ and O 2 À was not considered. For this reaction, a second-order rate constant of 1.7 Â 10 7 M À1 s À1 was reported by Rao and Hayon (1975), in disagreement with the findings of Nadezhdin and Dunford (1979), among others, who determined a second-order rate constant of 1.7 Â 10 4 M À1 s À1 for the HQ/HO 2 reaction. Since HO 2 is a stronger oxidant than O 2 À , the rate constant of HQ-O 2 À reaction is expected to be < 1.7 Â 10 4 M À1 s À1 . According to the mechanism provided in reactions 1e6, superoxide radical will not be formed in absence of dissolved O 2 and its steady state concentration will be very low in an aerated medium at pH 7.
The presence of an agent that can oxidize HQ to BQ could promote the formation of SQ À through reaction 3 and hence the generation of O 2 À (reaction 2). As stated before, Criquet et al. (2015) determined that chlorine and bromine react with o-and phydroxyphenols, HQ among them, via ET, with formation of the corresponding quinones. In the case of HQ, the apparent secondorder rate constant of its reaction with free available chlorine (FAC, sum of [HOCl] and [ClO À ]) at pH 7 (reaction 7; k 7 ¼ 21.6 M À1 s À1 measured in terms of BQ formation) is comparable to that of reaction 3. Hence, depending on the experimental conditions as BQ is formed, some SQ À can be generated (reaction 3), thereby promoting the formation of O 2 À . Also, the oxidation of According to Candeias et al. (1993), superoxide reacts with HOCl to form OH (reaction 9; k 9 ¼ 7.5 Â 10 6 M À1 s À1 , determined by monitoring the decay of O 2 À ; Long and Bielski, 1980).
A previous study revealed that an initial bromination of natural organic matter (NOM) mainly occurs through ET instead of EAS (Criquet et al., 2015). Hence, with the presence of phenols being structural moieties of NOM and in analogy to bromination, formation of SQ À during chlorination of NOM, with a concomitant formation of HOCl À and/or O 2 À and subsequently OH, cannot be ruled out.
Because of the short lifetime of hydroxyl radicals, it is common to use scavengers to determine their formation (Flyunt et al., 2003;. Moreover, if the product from the reaction between OH and a scavenger is known and can be measured, the extent of the formation of OH can be estimated. In addition, it is also important to exclude reactions of the scavenger and/or the reaction products with other substances present in the reaction medium (reagents and/or intermediates and/or products). Therefore, the following criteria need to be considered for the selection of a scavenger to quantify OH formation: (i) The reactivity of the scavenger, the organic surrogate (hydroxyphenols or NOM in this case) and the oxidant (chlorine in this study) towards the radical species; (ii) the nature of the products formed from the reaction between the scavenger and the radical (stability, reactivity towards oxidants, etc.); (iii) easy detection and quantification of the products; and (iv) the reactivity of the scavenger, the transformation products and the organic surrogate towards the oxidant. Furthermore, the scavenger has to be present in a sufficiently high concentration, so that a fraction of >95% of the OH is scavenged. Even though the product from the reaction between the scavenger and OH is an indicator of the generation of this species, there are some uncertainties on its quantification due to possible effects of the reaction conditions on the product yield (moles of product formed per mole of OH generated).
In this study, to investigate the formation of OH, dimethyl sulfoxide (DMSO), tert-butyl alcohol (TBA), benzene (BZ) and terephthalic acid (TPA) were initially selected as scavengers. However, DMSO, that reacts with OH to form methane sulfinic acid (MSI) (k OH-DMSO ¼ 4.5 Â 10 9 M À1 s À1 ; Bardouki et al., 2002), also has a relatively high reactivity with FAC (k FAC-DMSO ¼ 315 M À1 s À1 at pH 7, 20 C; Amels et al., 1997), wherefore its use is impossible for this system. The second order rate constants for the reactions of the different OH scavengers towards OH and FAC (pH 7 and 25 C) and their reaction products are provided in Table 1.
In aqueous solution and in presence of dissolved oxygen the reaction between OH (generated radiolytically) and BZ (k OH-BZ ¼ 7.8 Â 10 9 M À1 s À1 ; Buxton et al., 1988) proceeds through the formation of a hydroxycyclohexadienyl radical that further reacts with O 2 , to mainly phenol (Pan et al., 1993). The molar phenol yields were found in the range of 53e93% depending on the pH, and O 2 À / HO 2 , small amounts of various aldehydes and formic acid were also formed. By photolysis of aerated NO 3 À to produce OH at pH~6, the molar phenol yield was 95% (Deister et al., 1990). An average value of 85% has been adopted by different authors (Dong and Rosario-Ortiz, 2012). Contrary to BZ, which does not react with FAC, phenol (PHEN) reacts with an apparent second order rate constant k FAC-PHEN ¼ 18 M À1 s À1 , at pH 7 and 25 C (Gallard and von Gunten, 2002). The reaction of OH with TPA in aqueous solution (k OH-TPA ¼ (3.3e4.4) x 10 9 , Fang et al., 1996;Page et al., 2010;Charbouillot et al., 2011) (Fang et al., 1996).
The molar yield of hTPA based on moles of OH produced is~30% at pH 7 and 25 C and slightly increasing with increasing pH and temperature (Charbouillot et al., 2011). The reactivity of the cyclohexadienyl-type radical formed from the TPA-OH reaction with O 2 is~30 times lower than that of the radical formed from BZ-OH reaction (Fang et al., 1995). Because of that, in presence of a stronger oxidant such as IrCl 6

2À
, the hTPA yield can increase to up to 85% without formation of O 2 À /HO 2 (Fang et al., 1996). Since there are no data in literature about the reactivity of TPA nor hTPA towards FAC, the corresponding apparent second-order rate constants (pH 7 and 25 C) have been experimentally determined in this study.
The above mechanism (reactions 1e10) suggests that OH might be formed during the chlorination of hydroxy-phenolic compounds. However, so far, there is no experimental evidence for this process. The main objective of this study was to determine whether chlorination of NOM leads to OH formation and identify the main NOM structures involved. To this end, different o-, m-and phydroxyphenols (hydroquinone, catechol, resorcinol, gallic and tannic acid) and NOM extracts (humic and fulvic acids) were chlorinated, and the formation of OH and the role of O 2 À were investigated with the aid of multiple radical scavengers. The influences of hydroxyphenol, NOM, FAC and dissolved oxygen concentrations as well as pH on OH production were also studied. From the results obtained a kinetic model for OH formation during hydroquinone chlorination at pH 7 was developed to support the proposed mechanism.

Chemicals
With the exception of the sodium hypochlorite solution (10e15% active chlorine, Sigma-Aldrich, reagent grade), all organic and inorganic compounds (purchased from Sigma-Aldrich, Fluka Analytical, Merck or Carlo Erba), were of analytical grade or higher and used without further purification. Stock solutions of the selected hydroxyphenols (15 mM for hydroquinone (HQ), catechol (CAT), resorcinol (RES) and gallic acid (GAL); 2.5 mM for tannic acid (TAN)), were freshly prepared in Milli-Q ultrapurified water. Their structures are presented in Table 2 together with the corresponding second order rate constants for the reactions with FAC, OH and O 2 À .
Stock solutions of the different scavengers were prepared in ultrapurified water: tert-butyl alcohol (TBA) 3.5 M; saturated aqueous benzene (BZ)~23 mM, stirred overnight and kept in a fume hood; terephthalic acid (TPA) 50 mM, adjusted to pH 6.5 by adding some drops of concentrated NaOH; and MnCl 2 $4H 2 O 20 mM, the latter used as a superoxide scavenger. When tetranitromethane (TNM) was used as superoxide scavenger, 4 mL of TNM 8.2 M (pure compound) was added to 20 mL of the reaction medium (final [TNM] ¼ 1.65 mM) in the fume hood.

Analytical methods
Stock solutions of chlorine and bromine were photometrically standardized at pH 11 before use (e ClO-at 292 nm ¼ 350 M À1 cm À1 , Hand and Margerum, 1983;e BrO- Heeb et al., 2017). Residual concentrations of FAC were determined using a colorimetric method based on diethyl-p-phenylene diamine (DPD), measuring the absorbance at 515 nm (Rodier et al., 2009).
Formaldehyde (FAL) was firstly derivatized with 2,4dinitrophenylhydrazine (DNPH), to form the corresponding hydrazone, and then analysed by HPLC-DAD at 360 nm (U.S. EPA, 1996). Acetone and BQ, if present, were also quantified by this method.
A description of the preparation of reagents, conditions of the HPLC analyses, measuring ranges and limits of detection (LOD) and quantification (LOQ) are provided in Text S1 (SI).
The nitroform anion (NF À ) formed from the tetranitromethane (TNM) reduction by O 2 À was determined spectrophotometrically at 350 nm (e 350nm ¼ 14600 M À1 cm À1 ; Rabani et al., 1965). The dissolved organic carbon (DOC) and inorganic carbon (IC) content in LG water were measured using a Shimadzu total organic carbon (TOC) analyzer. Initial DOC concentration (DOC 0 ) of the hydroxyphenols and NOM extract solutions was calculated from their molecular structures and chemical composition, respectively.

Chlorination of phenolic surrogates, NOM extracts and LG water in the presence of OH scavengers
To determine the formation of OH during chlorination, 0.4 mL phosphate buffer (PBS) 0.5 M (pH 6e7.5), and fixed volumes of hydroxyphenol solutions (or NOM solutions), radical scavenger stock solutions and ultrapurified water were added, to obtain the desired concentrations in a final volume of 20 mL (22 mL amber glass vials). In the case of LG water, only the scavenger solution was added, decreasing the pH from 8.37 to 7.85.
After mixing, different volumes of a FAC stock solution were added to each vial while stirring and the vials were closed. A control sample was always prepared under the same conditions but without adding FAC. The mixtures were allowed to react at room temperature for 1e72 h (depending on the nature of the surrogate and the concentration of reactants) and then analysed by HPLC. In some cases bromine instead of chlorine was added.
To establish the initial concentrations of all reagents, the reactivity of the OH scavenger, the organic surrogates and FAC towards OH was taken into account to ensure that more than 95% of the generated OH (f ð OHÞ!0.95), is captured by the OH scavenger, according to the following equation: where k OH-scavenger represents the second order rate constant for the reaction between OH and the selected scavenger (Table 1); k OH-hydroxyphenol the second order rate constant for the reaction between OH and the phenolic surrogate (Table 2); k OH-HOCl (1.21 Â 10 9 M À1 s À1 , Bulman et al., 2019) and k OH-ClO-(6.37 Â 10 9 M À1 s À1 ; Bulman et al., 2019), are the second order rate constants for the reactions of OH with HOCl and ClO À , respectively.  When needed, the dissolved O 2 concentration in the reaction medium was increased or reduced by purging with pure O 2 or N 2 , respectively. When BZ was used, bubbling was performed before adding the scavenger to avoid losses by stripping. molar ratios a predetermined volume of a FAC stock solution was added to each vial while stirring and the vials were closed. The mixtures were allowed to react at room temperature in the dark, and after 72 h the residual FAC was analysed. For LG water, the procedure was the same but without adding PBS or ultrapurified water. Once the chlorine demand (mM) of each NOM solution was determined, the specific chlorine demand was calculated by dividing this value by the DOC 0 (mM) content.
2.3.3. Determination of the apparent second order rate constants for the reactions between chlorine and TPA, hTPA or BQ at pH 7 (10 mM PBS) and 25 C Apparent second order rate constants for the reactions of TPA, hTPA or BQ with chlorine at pH 7 (10 mM PBS, 25 C), were determined in excess of FAC (pseudo-first order conditions; see experimental conditions in Fig. S1, SI), by measuring the evolution of the parent compounds (TPA and hTPA by HPLC-DAD/ fluorescence; BQ by the decrease of the absorbance at 247 nm). In all cases but BQ, residual chlorine was quenched before the analysis by adding a few mL of a concentrated solution of Na 2 S 2 O 3 (0.5 M). For TPA, experiments in excess of TPA (4 mM) were also performed measuring the evolution of residual FAC . For BQ, a competition kinetics method was also applied using phenol as competitor (k FAC-PHEN ¼ 18 M À1 s À1 at pH 7, 25 C; Gallard and von Gunten, 2002). In this case, vials containing BQ (50 mM) and PHEN (5e10 mM) (10 mM PBS, pH 7 at 25 C), were dosed with different volumes of a FAC stock solution to obtain [FAC] 0 in the range of 0e50 mM. The samples were mixed, and after a reaction time of 3 h, the final concentrations of BQ and PHEN were determined by HPLC. In all cases, control samples were prepared in absence of chlorine. All the experiments were carried out at least in duplicate.

Modelling of OH generation during chlorination of HQ
The kinetics of the formation of OH during HQ chlorination at pH 7, 25 C and air saturated conditions (concentration of dissolved oxygen assumed to be 3 Â 10 À4 M), was modeled by the Tenua kinetic simulator (http://bililite.com/tenua/).

Results and discussion
3.1. Determination of OH formation from hydroquinone chlorination 3.1.1. Influence of the nature of the selected OH scavengers on HQ chlorination In a first series of experiments, chlorination of an air-saturated HQ solution 62.5 mM (DOC 4.5 mg L À1 ) at pH 7 (PBS 10 mM) and room temperature (22 ± 2 C) was carried out, using different molar [FAC] 0 /[HQ] 0 ratios, in absence/presence of the selected OH scavengers: TBA 50 mM, BZ 10 mM or TPA 4 mM (Fig. 1). Under these experimental conditions, and as expected due to their low reactivity towards FAC (see Table 1 and Fig. S1, SI) the scavengers had practically no effect on the extent of HQ oxidation. Thus, for a given [FAC] 0 /[HQ] 0 ratio, the extent of HQ abatement after 1.5 h contact time (Fig. 1a) was very similar regardless of the presence of a scavenger. Furthermore, the HQ abatement (solid symbols) corresponded well with the BQ formed (open symbols), i. e., the BQ yields were close to 1 mole per mole of HQ oxidized. For [FAC] 0 /[HQ] 0 ratios > 1.5, a decrease in the final BQ concentration is observed, most probably due to its reaction with FAC (k FAC-BQ ¼ 3 M À1 s À1 at pH 7, 25 C obtained in this study; Fig. S1, SI). Also, as the dose of FAC increased, the BQ concentration was higher in presence of a scavenger, which could indicate that under these conditions some BQ abatement is prevented. The apparent global stoichiometry of the chlorine-HQ reaction was close to 2 mol of chlorine per mole of HQ consumed (Fig. 1b), which is consistent with previous observations (Criquet et al., 2015). [HQ] 0 molar ratios are shown. In absence of FAC (control), FAL, PHEN or hTPA were not detected. The FAL concentration was much lower than PHEN or hTPA, results that cannot be attributable to the low FAL yield of the TBA-OH reaction (~25%; Flyunt et al., 2003). Moreover, the FAL concentration initially increased with increasing molar [FAC] 0 /[HQ] 0 ratios and decreased for ratios > 0.5. These results indicate that, for the tested conditions, FAL and acetone (not shown) are not stable in solution due to unknown secondary reactions. Therefore, in the further experiments the formation of FAL from TBA was not considered to determine the formation of OH from HQ chlorination.
For molar [FAC] 0 /[HQ] 0 ratios 0.5, the PHEN and hTPA concentrations were similar despite the different yields of PHEN (from BZ þ OH) and hTPA (from TPA þ OH) reported in air saturated and neutral pH conditions (~85% and 30%, respectively). As it will be discussed in the kinetics section, this could indicate that in presence of chlorine, similarly to what was observed in presence of IrCl 6 2À (Fang et al., 1996), the yield of hTPA is higher than when using  Table 2 and Fig. S1, SI). The yield of PHEN or hTPA as moles formed per mole of HQ eliminated (%), is shown in Fig. 2b. For the lowest molar [FAC] 0 / [HQ] 0 ratio, yields of PHEN and hTPA were~20% and~15%, respectively. As the [FAC] 0 /[HQ] 0 ratios increased, both yields decreased, in case of PHEN to < 5% for molar ratios ! 1.5, potentially due to the reaction of PHEN with chlorine. For hTPA, as will be discussed later, the decrease in the yield is probably related to a superoxide scavenging effect of the formed BQ, and remained almost constant at about 9% for molar [FAC] 0 /[HQ] 0 ratios ! 0.5.
For the conditions tested, the compliance with equation (11) (f ð OHÞ!0.95) was also experimentally corroborated as shown in Fig. S2 (SI) as an example, where the influence of the TPA concentration on hTPA formation from HQ chlorination ([HQ] 0 62.5 mM; [FAC] 0 93.7 mM), at pH 7 and room temperature, is presented. As observed, a constant hTPA concentration was determined when TPA >3 mM was added.
The fact that both PHEN and hTPA are detected implies that under the applied experimental conditions, chlorination of HQ leads to the formation of OH. Moreover, to suppress the OH -TPA reaction, the formation of hTPA was also tested in presence of high TBA concentrations. The results (Fig. S3, SI) indicate that TBA drastically inhibited the formation of hTPA. Therefore, it is very likely that OH is the main species involved in hTPA formation and the same can be assumed for PHEN when BZ was used as a scavenger.

Influence of the initial concentration of HQ
The influence of the initial HQ concentration on OH formation at pH 7 for TPA as OH scavenger is shown in Fig. 3 (Fig. S4, SI).
These observations could be at least partly explained by the nature of the intermediates and/or products formed during HQ chlorination. BQ, which reacts very fast with O 2 À (k -2 ¼ 10 9 M À1 s À1 ), could compete with HOCl (k 9 ¼ 7.5 Â 10 6 M À1 s À1 ) for O 2 À , thereby lowering the OH yield. To test this, BQ (0e60 mM) was added to samples containing 62.5 mM HQ and 10 mM BZ. Under these conditions, BQ did not impact the f ( OH) of the scavenger ( In absence of FAC (control samples) the concentration of HQ varied very slightly during the reaction time considered (between 1.5 and 4 h), indicating that a contribution of reaction 1 (HQ autooxidation) to O 2 À and SQ À formation was likely negligible. This means that BQ is mainly formed from HQ chlorination (reaction 7), allowing the formation of SQ À through reaction 3. Depending on the experimental conditions, the SQ À formed could be further oxidized to BQ by O 2 (reaction 2), with concomitant generation of O 2 À , and/or HOCl (reaction 8) to HOCl À , which could also  To test the potential role of superoxide radicals, experiments were performed with TNM or Mn(II) as O 2 À scavengers. Unfortunately, due to the reactivity of TNM and Mn(II) towards HQ, that at the conditions tested was oxidized to BQ by these reagents in absence of chlorine, their use as O 2 À scavenger seems to be impossible for this system (for further information see Text S2, SI).

Influence of pH
To test the influence of speciation of chlorine on OH production from HQ chlorination, experiments were performed in the pH range 6e7.5 (PBS 10 mM, [HQ] 0 62.5 mM, air saturated), using TPA as OH scavenger. At pH > 7.5, HQ is not stable in solution due to its fast oxidation by O 2 (James et al., 1938;WHO, 1994;Criquet et al., 2015). At pH 7.5, the formation of a BQ~1.5 mM was determined in the control sample (absence of chlorine), attributable to partial HQ autooxidation by O 2 through reactions 1e2.  (Fig. 5a). However, since the extent of HQ oxidized per mole of FAC also slightly increased with increasing pH, the differences in terms of hTPA yield are negligible (Fig. 5b). In presence of oxygen as the only oxidant, an increase of pH causes a slight increase in the yield of hTPA formed from TPA-OH (from 28% at pH 6 to~33% at pH 7.5, at 25 C) (Charbouillot et al., 2011). Although k 3 also increases with increasing pH (58 M À1 s À1 at pH 7, Yamazaki and Ohnishi, 1966; 570 M À1 s À1 at pH 7.4, Eyer, 1991), which could favor the formation of O 2 À , an increase in k 7 is also expected counteracting this effect.

Influence of dissolved oxygen concentration
In contrast to HOCl À (reaction 8), the formation of O 2 À requires the presence of oxygen (reaction 2). Hence, if this species is mainly responsible for OH generation during HQ chlorination, the formation of hTPA or PHEN must strongly depend on the O 2 concentration. In addition, it is important to note that, in absence of other oxidants, the mechanisms of hTPA and PHEN formation (from the TPA-OH or BZ-OH reactions) also require the presence of O 2 (Mathews, 1980;Fang et al., 1996;Pan et al., 1993). The reactions of the radical intermediates with O 2 leading to PHEN or hTPA could also produce molar equivalents of O 2 À , which could contribute to the OH formation by reaction 9. In absence of O 2 , products different than hTPA or PHEN could be formed. All these aspects will be discussed in more detail in the section on reaction kinetics.
To investigate the effect of O 2 , air saturated ([O 2 ]~300 mM), oxygen saturated ([O 2 ]~1200 mM) and N 2 saturated (O 2 purged by stripping with N 2 ) solutions containing 62.5 mM HQ were chlorinated (pH 7 and room temperature), using TPA as OH scavenger. The results in Fig. 6, show that hTPA is not formed in absence of O 2 , both hTPA concentrations (Fig. 6a) and yields (Fig. 6b) increased with increasing O 2 levels. For oxygen saturated conditions (the O 2 concentration increased four fold compared to air saturation), hTPA concentrations were 2e3 times higher and the yield increased~3e4 times, the latter due to the lower efficiency of HQ abatement for increasing O 2 concentrations (Fig. 6c). Overall, an increase in O 2 favors the formation of O 2 À through reaction 2 and, at  the same time, the chlorine consumption by reaction 9. From these results, it seems quite evident that O 2 À participates in the generation of OH. Moreover, the strong influence of O 2 levels on hTPA and PHEN generation would indicate that a contribution of HOCl À to OH production through reaction 8 is less important.
Similar results were obtained for the influence of O 2 on PHEN formation and yields when BZ was used as OH scavenger. However, in this case, to avoid BZ volatilization, the level of O 2 in the samples could only be partially reduced or increased (Fig. S6, SI). The possible contribution to OH formation from superoxide that can be generated from BZ-OH or TPA-OH reactions will also be discussed in the kinetics section.
The experiments were performed at pH 7 (PBS, 10 mM, air saturated) with 62.5 mM [HQ] 0 , using 4 mM TPA as a scavenger and molar [HOBr] 0 /[HQ] 0 ratios between 0 and 1.5. The apparent stoichiometry of HOBr:HQ reaction was~1:1 (Fig. S7, SI), in agreement with Criquet et al. (2015). In these experiments hTPA was not detected. A plausible explanation for this observation could be the  much higher value of k HOBr-HQ compared to that of k FAC-HQ , wherefore, HOBr would be quickly and totally consumed by HQ with no chance to react with O 2 À so neither HOBr À nor OH would be generated in this reaction system.

OH formation from chlorination of different hydroxyphenols at pH 7
The behavior of other hydroxyphenols (CAT, RES, GAL and TAN) was investigated to explore the applicability of our findings to a wider range of compounds. Their structures and reactivities towards different species (FAC, OH and O 2 À ) are summarised in Table 2. At pH 7, all of these compounds except RES lead to the formation of quinones upon chlorination and have higher FAC reactivities than HQ (Criquet et al., 2015).
The experimental procedure was the same as for HQ, i.e., chlorination in 10 mM PBS air saturated at pH 7, using 4 mM TPA as scavenger with [CAT] 0 , [RES] 0 , or [GAL] 0 , being 62.5 mM and [TAN] 0 ¼ 6.25 mM. Hence, DOC 0 as hydroxyphenol was 375 mM for CAT and RES, and 475 mM for GAL and TAN. Relative hTPA yields normalized to DOC 0 (%) are shown in Fig. 7a. For comparative purposes, hTPA yields corresponding to HQ 0 62.5 mM (DOC 0 ¼ 375 mM) are also included, and were much higher than for the rest of the tested compounds (squares, right Y-axis in Fig. 7a). In absence of FAC (control runs), formation of hTPA was negligible.
As expected, hTPA was not formed when RES was chlorinated (stars in Fig. 7a). Unlike the other selected hydroxyphenols, due to the m-position of the two OH substituents, RES chlorination proceeds through EAS, with chlorine addition, instead of ET (Criquet et al., 2015). Since there is no quinone/semiquinone formation, neither O 2 À nor HOCl À can be generated. In contrast, for CAT, GAL and TAN, formation of hTPA was detected, although the yields were significantly lower than for HQ. Again, in line to what was proposed in the case of HQ bromination, a possible explanation could be the higher reactivity of these o-hydroxyphenols with FAC compared to HQ (Table 2). Since FAC is consumed more quickly, its reaction with O 2 À can take place to a much lower extent.
Previously, it was shown that the presence of borate buffer decreased the reactivity of o-hydroxyphenols towards chlorine through the formation of a borate complex (Criquet et al., 2015). To test if the reactivity of these compounds with chlorine is a key parameter for O 2 À formation, solutions containing 60 mM CAT were chlorinated at pH 7 in 10 mM PBS, pH 8 in 25 mM PBS and pH 8 in 50 mM borate buffer with 4 mM TPA as OH scavenger. Fig. 7b shows that the presence of borate leads to a much higher normalized hTPA yield. This supports the hypothesis, that the rate of reaction 7 is important for generation of OH through reaction 9.
Taking into account the complexity of these reaction systems, the reactivity between FAC and the hydroxyphenol (reaction 7) is probably not the only factor affecting OH formation. For example, both the relative position of the OH groups and the presence of electron donating/withdrawing groups on the aromatic ring can affect the standard reduction potentials (E 0 ) of the quinone/semiquinone pairs and, therefore, reaction 2 and the stability of O 2 À (Song and Buettner, 2010). To test this, methylhydroquinone (MeHQ), with higher electron donating properties than HQ due to the presence of the methyl group, was chlorinated at pH 7 alone ( and hTPA were determined. Although the reactivity of MeHQ towards FAC was higher than for HQ as expected (see Fig. S8, SI), the hTPA yields were very similar and close to [HQ] 0 60 mM (Fig. S8, SI).
These results can be explained by the influence of the methyl group on E 0 (quinone/semiquinone) and hence the ratio of k 2 /k -2 . For MeHQ, k 2 ¼ 1.1 Â 10 6 M À1 s À1 and k 2 /k -2 ¼ 1.4 Â 10 À3 , whereas for HQ k 2 ¼ 5 Â 10 4 M À1 s À1 and k 2 /k -2 ¼ 5 Â 10 À5 (Song and Buettner, 2010). This means that the presence of the methyl group favors the formation/stabilization of O 2 À counteracting the negative effect of its higher reactivity towards FAC.

OH formation from chlorination of different NOM extracts at pH 7
The formation of OH during the chlorination of NOM extracts (SRHA, SRFA, PLFA and NL-NOM; Table S1, SI) was investigated in presence of TPA, and the formation of hTPA was measured after 72 h. The [DOC] 0 of the NOM extracts was 375 mM (SRHA), 392 mM (SRFA), 442 mM (PLFA) and 248 mM (NL-NOM). The specific chlorine demand of each extract (pH 7, 72 h contact time) was determined, and resulted to be 0.49, 0.34, 0.38 and 0.36 moles of FAC per mole of DOC 0 for SRHA, SRFA, PLFA and NL-NOM, respectively (Fig. S9, SI).
The DOC 0 -normalized yields of hTPA (mol/mol, %) are shown in Fig. 8 as a function of the specific chlorine dose. The maximum hTPA yields followed the order PLFA (0.3) < SRFA (1.1) < SRHA (1.6) < NL-NOM (2.0). They were achieved for specific chlorine doses slightly higher than the corresponding specific chlorine demands, which is due to some FAC consumption by TPA under these conditions. After reaching these maxima, higher chlorine doses led to lower hTPA yields probably due to its reaction with FAC (k FAC-hTPA ¼ 0.15 M À1 s À1 at pH 7 and 25 C; this study). It has to be mentioned that hTPA was also formed in very low concentrations in absence of FAC (control runs). This means that the NOM extracts can generate OH in aqueous solution in the dark. This effect was also observed previously during aeration of different anoxic lake waters and attributed to the development of Fenton-like reactions that involve HQ/SQ À and/or Fe(II) (Minella et al., 2015). After 72 h, the hTPA yields (mol hTPA/mol DOC o , %) of the control samples were about 2 orders of magnitude smaller, but following the same sequence as for chlorination (0.002 (PLFA) < 0.006 (SRFA) < 0.02 (SRHA) < 0.03 (NL-NOM)).
Among the selected extracts, PLFA (autochthonous NOM) presents the lowest SUVA 254 , EDC, aromatic C and phenolic content (Table S1 in SI, and references therein). PLFA has the lowest OH yield, which seems to indicate that phenolic structures in NOM are involved in OH formation during chlorination of NOM. With the exception of PLFA, the DOC-normalized hTPA yields for the selected NOMs were close or even higher than for HQ (Fig. 7a). Thus, in case hydroxyphenols in NOM are the main source of OH during chlorination, HQ moieties with higher OH formation potential than HQ must be present. This points towards the presence of HQ-like structures in NOM with lower reactivity with FAC than HQ, thus allowing FAC to react with O 2 À . Alternatively, this could be explained by the presence or formation of quinone structures with higher k 2 and k 2 /k -2 values than HQ (that is, structures with low E 0 quinone/semiquinone) favoring the formation/stabilization of O 2 À .
Also, NOM moieties different from phenols could be responsible for OH formation upon chlorination. Based on the information in Table S1 (SI), N-containing structures are probably not involved, since PLFA has the lowest OH formation potential, but contains 4e10 times more N than the other extracts. And finally, the higher the iron content, the higher the amount of OH produced. To this end, a study by Santana-Casiano et al. (2010) carried out in seawater demonstrated that Fe(III)-CAT complexes are formed, followed by Fe(III) reduction to Fe(II) and CAT oxidation to SQ À , with generation of O 2 À . Hence, the enhancement of OH formation due to iron complexes in NOM cannot be disregarded.
Water from Lake Greifensee (LG, Switzerland) was also chlorinated (DOC 0 250 mM; pH 7.85 after the addition of TPA; specific chlorine demand 0.22 moles of FAC per mole of DOC), and the formation of hTPA was measured. In absence of FAC (control runs), after 72 h contact time, the molar DOC-normalized hTPA yield (%) was <0.003. The results obtained (see Fig. S10, SI), indicate that DOM in LG water was capable of producing OH when chlorinated. However, the maximum DOC-normalized hTPA formation was low, 0.1% mol (mol DOC 0 ) À1 , similar to PLFA (also of autochthonous origin; Wenk et al., 2011), and more than ten times lower than for HQ and the other investigated NOM extracts. In the last part of the study, the formation of OH during HQ chlorination at pH 7 has been modeled based on the experimental results and the kinetic data available. All the reactions considered and their rate constants are compiled in Table S3 in SI. The kinetic model is based on reactions (1)e(3) and (7) assuming values of 10 À7 and 10 7 M À1 s À1 for k 1 and k -1 , respectively (K 1 ¼ 10 À14 ; Eyer, 1991). At pH 7, k 5 is very low (Sheng et al., 2014), and from equilibrium 4 the contribution of reaction 6 can be disregarded (see below). Also, due to the strong effect of O 2 concentration on hTPA and PHEN formation (see section 3.1.5), keeping in mind that reactions 8 and 10 are oxygen-independent, the contributions of these reactions to OH formation is expected to be minimal and they have not been considered. For reaction 7, according to the experimental results a molar stoichiometry 2:1 FAC:HQ has been assumed (see Fig. 1b; and also Fig. S11, SI).
According to the pKa of FAC (7.49, equilibrium 12), at pH 7, the molar fractions of HOCl and ClO À are 0.756 and 0.244, respectively. Since ClO À does not react with O 2 À (Long and Bielski, 1980), reaction 9 can be rewritten in terms of FAC with an apparent second order rate constant k 9b ¼ 5.6 Â 10 6 M À1 s À1 at pH 7 (Long and Bielski, 1980): In presence of BZ as OH scavenger, the following reactions have been considered (Pan et al., 1993;Fang et al., 1995): where CHDt is a cyclohexadienyl-type radical intermediate (k 16 ¼ 7.8 Â 10 9 M À1 s À1 ; Buxton et al., 1988), CHDHPt the cyclohexadienylhydroperoxide-type radical (k 17 ¼ 3.1 Â 10 8 M À1 s À1 , k -17 ¼ 1.2 Â 10 4 s À1 ; Fang et al., 1995), and P1BZ another final product other than PHEN. Since the rate constant of CHDHPt decrease is 800 s À1 (Fang et al., 1995), by assuming a yield of 85% for PHEN and 15% for P1BZ (Dong and Rosario-Ortiz, 2012), k 18 and k 19 values have been assumed to be 680 s À1 and 120 s À1 , respectively. In presence of TPA as OH scavenger, the following reactions have been considered (Fang et al., 1996): where P1TPA and P2TPA represent products different from hTPA and CHDt the cyclohexadienyl-type radical formed when OH was added to the ortho position of TPA. From the overall rate constant of the TPA -OH reaction (4.4 Â 10 9 M À1 s À1 , Page et al., 2010) and the corresponding yields (85% for CHDt and 15% for P1TPA; Fang et al., 1996), values of 0.66 Â 10 9 M À1 s À1 and 3.74 Â 10 9 M À1 s À1 have been assigned for k 20 and k 21 , respectively. According to Fang et al. (1995), k 22a ¼ 1.6 Â 10 7 M À1 s À1 and k -22a ¼ 3.4 Â 10 3 s À1 , whereas the rate constant of CHDHPt decomposition (reactions 23 and 24) is 390 s À1 . By considering a global yield of 30% for hTPA and 55% for P2TPA, k 23 and k 24 values were assumed to be 138 s À1 and 252 s À1 , respectively. Similarly to what was observed by Fang et al. (1995) during the TPA -OH reaction in presence of IrCl 6 2À , another possible scenario in which FAC can also oxidize CHDt directly and irreversibly to hTPA (without formation of CHDHPt or O 2 À ) has been considered (reaction 22b), initially assuming the same rate constant for this reaction as for reaction 21, that is, k 22b ¼ 3.74 Â 10 9 M À1 s À1 . Finally, the reactions of FAC and OH with other compounds have also been introduced in the mechanism (reactions 25e28 and 29e34, respectively. See Table S3, SI). However, since at the experimental conditions applied in this study f( OH) ! 0.95 for TPA and BZ (Table S1, Fig. 9a (symbols: experimental results, lines: model calculations). When TPA was used as OH scavenger the two possible scenarios (that is, without/ with participation of 22b) were considered, in the latter case initially assuming k 22b ¼ k 21 as indicated before. In Fig. 9a, a good agreement between experimental and modeled data is observed for PHEN, whereas in the case of hTPA its formation through the O 2 À dependent mechanism (reactions 22a and 23, without participation of 22b), clearly underestimate the experimental values (Fig. 9a, dotted line). In contrast, when reaction 22b is also considered, the experimental/calculated hTPA agree within the experimental errors (Fig. 9a, solid line).
To further investigate the role of reaction 22b, the two scenarios were tested for lower HQ 0 and FAC 0 conditions, again initially assuming k 22b ¼ k 21 . From the evolution of experimental and modeled hTPA as a function of the FAC concentration (see Fig. S12, SI) it seems that for [FAC] 0 > 1 mg L À1 reaction 22b (solid lines) is the main source of hTPA, and hence~0.85 moles of hTPA are formed per mole of OH generated. At lower FAC doses (<1 mg/L) the participation of the O 2 -based mechanism (reactions 22a and 23) increases, with hTPA yields in the range of 30%e85%. The fact that in presence of FAC the yield of hTPA formation from the TPA -OH reaction is higher than 30% would explain why the experimental concentrations and yields of PHEN and hTPA were similar for [FAC] 0 /[HQ] 0 molar ratios 1 (see Fig. 2). The value for k 22b was optimized by various model calculations to minimize the differences between experimental and calculated hTPA for different initial HQ and FAC concentrations. The optimized value is k 22b ¼ 10 8 M À1 s À1 (see Fig. S13, SI) which was then applied for all model calculations.
In any case, the yields of both hTPA (from the TPA -OH reaction) and PHEN (from the BZ -OH reaction) are lower than 100%. This implies that the total formation potential of OH ( OHFP), is higher than the measured PHEN or hTPA concentrations. The OHFPs of HQ 62.5 mM as a function of [FAC] 0 /[HQ] 0 at pH 7 in presence of BZ 10 mM or TPA 4 mM were modeled. As shown in Fig. 9b, the total production of OH in presence of BZ (dashed line) is higher than for TPA (solid line), attributable to the O 2 À generated through reaction 18. The calculated relative contribution (%) of O 2 formed from reactions 18 or 23 to OHFP, is also indicated in Fig. 9b (reaction 18: dashed line; reaction 23: solid line). For TPA this contribution was minimal (lower than 1%), whereas for BZ it decreased from~22% to 10% as the FAC dose increased. At the conditions tested and as expected, the introduction of equilibrium 4 and reaction 6 in the mechanism did not cause any change. The presence of FAC and/or BQ in the reaction medium together with the low concentration of O 2 À and HO 2 À in solution (~10 À11 and 10 À13 M, respectively) prevent the formation of HO 2 À through 6. The effect of the initial HQ concentration on the experimental hTPA yield (that is, for a given [FAC] 0 /[HQ] 0 the hTPA yield decreases with increasing [HQ] 0 , see Fig. 3) is also well predicted by the proposed mechanisms (Fig. S14, SI).
The negative effect of BQ on OHFP experimentally observed (see Fig. 4), is also corroborated by this mechanism. As an example, Fig. S15 (SI) shows the modeled reduction of the OHFP (%) during the chlorination of HQ (62.5 mM, pH 7, air saturated) in presence of different concentrations of BQ, with 4 mM TPA or 10 mM BZ as scavengers. As observed, the higher the initial BQ concentration, the higher the reduction of the OHFP. For a given [BQ] 0 , this effect is attenuated by the BQ formed from HQ chlorination as the [FAC] 0 increases. Thus, the formation of BQ from the HQ -chlorine reaction seems to be the main reason why, for a given [HQ] 0 , OHFP (and hence hTPA or PHEN yields) decreases as the dose of FAC increases, BQ acting as O 2 À scavenger.
Finally, the role of dissolved O 2 on OH generation and hence hTPA formation is also well predicted by the proposed mechanisms, as shown in Fig. S16 (SI) for the experiments performed using TPA as OH scavenger. Fig. S16a (SI) shows that for O 2 -saturated conditions ([O 2 ] ¼ 1200 mM) reaction 22b seems to be the main source of hTPA, although in this case the model for the optimized k 22b value (dashed line) slightly underestimates the actual hTPA concentrations. However, the model fit is still very good considering the significant number of reactions. Fig. S16b  [O 2 ] air sat ¼ 4), it results to be 2.5e3. In agreement with the experimental results, the relationship between O 2 concentration and the OHFP is not linear. Based on the findings of this study, a mechanism for OH formation during HQ chlorination is proposed in Fig. 10.
The mechanism depicted in Fig. 10 shows that in a first step BQ is formed from the reaction of HQ with chlorine. HQ and BQ react to two semiquinone radicals, which then react with molecular oxygen to form superoxide radical. Finally, HOCl reacts with superoxide to form hydroxyl radical. BQ can also react with superoxide, leading back to molecular oxygen. Therefore, the BQ concentration can affect the hydroxyl radical yield significantly.

Conclusions
With the aid of different scavengers the generation of hydroxyl radical ( OH) during the chlorination of o-and p-hydroxyphenols at neutral pH conditions has been demonstrated. The mechanism of OH generation requires the formation of semiquinones capable of reducing dissolved oxygen to superoxide radical, which further reacts with HOCl to OH. For p-hydroxyphenols the OH yield (in terms of moles formed per mole of DOC 0 ) was clearly higher and comparable to that of SRHA, SRFA and NL NOM, all these extracts presenting high electron donating capacities. According to the results obtained, a low reactivity towards chlorine and a high presence/formation of quinone structures with low E 0 favor the formation/stabilization of O 2 À , thereby increasing the potential of OH formation during the chlorination of NOM. From the combination of experimental results with kinetic modelling, a mechanism for hydroquinone chlorination at pH 7 has been proposed, which can explain the findings well.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.