Degradation of odorous 2,4,6-trichloroanisole in chlorinated water by UV-LED/chlorination: kinetics and influence factors

2,4,6-Trichloroanisole (2,4,6-TCA) has aroused a special concern for their odor problem and potential threats. In this study, the degradation of 2,4,6-TCA by UV/chlorination with different UV sources was compared, including low-pressure mercury lamp (LPUV, 254 nm) and ultraviolet light-emitting diode (UV-LED, 275 and 285 nm). The maximum removal of 2,4,6-TCA can be achieved by 275-nm UV-LED/chlorination in neutral and alkaline conditions which was 80.0%. The reaction, kinetics, and water matrix parameters on 2,4,6-TCA degradation were also evaluated. During UV-LED (275 nm)/chlorination, 2,4,6-TCA degradation was mainly caused by direct UV photolysis and indirect hydroxyl radical (HO·) oxidation, while reactive chlorine radicals (RCSs) had a negligible contribution. The second-order rate constant between HO· and 2,4,6-TCA was determined as 3.1 × 109 M−1 s−1. Increasing initial chlorine dosage and decreasing 2,4,6-TCA concentration or pH value significantly promoted 2,4,6-TCA degradation during UV/chlorination process. The presence of natural organic matter (NOM) and bicarbonate (HCO3−) can inhibit 2,4,6-TCA degradation, while chloride ion (Cl−) had a negligible effect. The kinetic model for 2,4,6-TCA degradation was established and validated, and the degradation pathways were proposed based on the identified intermediates. Furthermore, UV-LED (275 nm)/chlorination also exhibited a promising effect on 2,4,6-TCA removal in real water, which can be used to control 2,4,6-TCA pollution and odor problems.


Introduction
Off-flavors have aroused great concern in the past decades, as the esthetic properties are often used to judge water safety and quality (Antonopoulou et al. 2014;Feng et al. 2020).
Trichloroanisoles (TCAs) have attracted much attention in recent years, because they can cause an earthy-musty odor and unpleasant taste (Vlachos et al. 2008;Zhang et al. 2013). Except for causing odor problems, as aromatic compounds, TCAs have certain toxicological effects and accumulate in the food chain, potentially endangering the aqueous environment and human health. TCAs are produced through the microbiological O-methylation of chlorophenols or the chlorination of anisoles as disinfection by-products (DBPs) (Ge et al. 2006;Zhang et al. 2016a). TCAs are first found on wine corks and are often detected in natural water bodies (Buser et al. 1982). Additionally, they can also be formed during water treatment processes. Diaz et al. studied the Abrera drinking water plant in Spain and found that when there were no trihaloanisoles in the raw water, a total concentration of TCAs was detected as 8.13 ng L −1 in the finished water, implying that TCAs could be formed in the water treatment process (Diaz et al. 2005). In the meanwhile, odor problems also arise in drinking water distribution networks, Responsible Editor: Sami Rtimi * Tian-Yang Zhang tianyang1815@126.com 1 because many microorganisms deposited on the pipe surface can react with 2,4,6-trichlorophenol (2,4,6-TCP), and produce TCAs by microbial O-methylation process. Zhang et al. reported that the pipe materials had important effect on the formation of TCAs, especially stainless steel, and ductile iron pipes producing much more TCAs than polyethylene (PE) pipes do (Zhang et al. 2018a). TCAs have three kinds of structures in general: 2,4,6-trichloroanisole (2,4,6-TCA); 2,3,6-trichloroanisole (2,3,6-TCA); and 2,3,4-trichloroanisole (2,3,4-TCA). 2,4,6-TCA is often detected in water systems with an odor threshold concentration of 0.05 to 4 ng L -1 (Benanou et al. 2003). Urase and Sasaki analyzed the concentration of 2,4,6-TCA in wastewater of Tokyo as 4.3-37.7 ng L −1 (Urase and Sasaki 2013). Zhang et al. used solid-phase microextraction (SPME) and gas chromatography-mass spectroscopy (GC-MS) analysis method to investigate the presence of haloanisoles in the tap water from 22 cities in China and discovered that 2,4,6-TCA was detected in every sample with concentrations ranging from 0.03 to 15.9 ng L −1 (Zhang et al. 2016b).
Among these AOPs, UV/chlorination has received growing interest in recent years. Due to the comparable UV absorption rate (59 M −1 cm −1 for HOCl and 66 M −1 cm −1 for OCl − ) and quantum yield (about 1.0 mol Es −1 ) (Fang et al. 2014;Jin et al. 2011;Xu et al. 2022), UV/chlorination is considered an alternative technology to destruct emerging contaminants, such as trichloroethylene, caffeine, and carbamazepine (Guo et al. 2017;Wang et al. 2012;Xu et al. 2021). Moreover, UV/chlorination does not require extra quenching chemicals in the water treatment plants and has beneficial effects on the inactivation of microorganisms . In this regard, the research on the degradation of 2,4,6-TCA by using UV/chlorination is promising and valuable. The direct UV photolysis of free chlorine primarily generates the radicals of Cl• and HO•; then, the secondary radicals such as O − • and Cl 2 − • are further conversed (Cao et al. 2023). It is worth mentioning that process efficiency has a significant relationship with solution pH and UV wavelength Zou et al. 2019a). The main UV source for UVbased AOPs is medium-and low-pressure mercy lamps, but they have many disadvantages in terms of installation and operation (Chevremont et al. 2013). Recently, ultraviolet light-emitting diode (UV-LED) industry is emerging Nyangaresi et al. 2019). The superiorities of UV-LEDs to conventional mercury lamps include better durability, environmental friendliness, no warm-up time, flexible installation, and diversity in wavelengths (Kheyrandish et al. 2017, Xiao et al. 2018, Yin and Shang 2020. It is reported that at the same UV intensity, degradation of micropollutants including propranolol, sulfamethoxazole, oxytetracycline hydrochloride, and gatifloxacin was better using UV-LED/ chlorination than using UV (254 nm)/chlorination (Gao et al. 2020a;Kim et al. 2020). Peroxymonosulfate (PMS) activated by UV-LEDs can also effectively degrade emerging pollutants (Hassani et al. 2023;Madihi-Bidgoli et al. 2021). To sum up, the control of TCAs, which are odorous substances generated during water treatment and distribution, is particularly important at the end of the water supply system. There is a paucity of relevant studies. Moreover, UV/chlorination is proven to be an effective process for removing odorous substances, especially for applications at the end of water supply systems (Sun and Chang 2013). But traditional UV sources have many technical bottlenecks in the application at the end of water supply system, while UV-LEDs are more suitable due to their compact shape and lower energy consumption Lu et al. 2022;Yeom et al. 2021). Finally, the multiple wavelengths of UV-LEDs provide a wavelength-to-pollutant match. And there is a lack of reports on which wavelength is most suitable for the removal of TCAs in the UV-LED/chlorination process.
In this study, the elimination of 2,4,6-TCA by UV/chlorination process was investigated, and the performances of different UV sources including LPUV (254 nm) and UV-LEDs (275 and 285 nm) were compared. For the best performing UV source, the degradation mechanism and kinetic model of 2,4,6-TCA degradation were evaluated. In the meantime, effects of the chlorine dosage, TCA concentration, and water matrix (HCO 3 − , Cl − and natural organic matter (NOM)) on the removal of 2,4,6-TCA were also explored.
Then, the accuracy of the kinetic model was verified, which was consistent with the actual degradation effect. The degradation pathway of 2,4,6-TCA in the UV/chlorination process was proposed. Finally, the removal of 2,4,6-TCA in real water was also proved to be feasible for practical application. The developed model can well simulate the degradation of 2,4,6-TCA in a variety of water conditions, and the experimental results were crucial for TCAs and odorant substance control in drinking water systems.

Chemicals and reagents
All the chemicals utilized were of at least reagent grade. The details are shown in Text S1.

UV sources
UV sources including LPUV (254 nm) and UV-LEDs (275 and 285 nm) were supplied by Qingdao Jason Electric Co., Ltd. (Qingdao, China). The irradiation system's schematic diagram is described in Fig. S1 of Supplementary Material. The parallel light beam generated by LPUV or UV-LED was irradiated in a quartz container (diameter of 10.2 cm, water depth of 1.5 cm, and solution volume 100 mL). The average irradiation (0.28, 0.42, and 0.31 mW cm −2 for 254, 275, and 285 nm UV sources, respectively) was measured by the KI/KIO 3 method (Zou et al. 2019b). Text S2 of the Supplementary Material contains detailed information. Each UV light has an air fan for cooling. The temperature of the reaction solutions was maintained at 25 ± 1 °C during the experiment by using a stirring hot plate.

Experimental procedures
In each group of experiments, LPUV was preheated for 30 min until the UV intensity was steady, while UV-LEDs were not. The reaction solution was prepared with 0.03-100 μg L −1 2,4,6-TCA and 10 mM phosphate buffer, and the pH was adjusted to the target value (pH = 5.0-9.0) with a small volume of sodium hydroxide (3.0 M) or phosphate acid (1.0 M). Then, 100 mL of 2,4,6-TCA solution was transferred to a quartz container covered with a quartz glass sheet to prevent volatilization of the odoriferous substances during the reaction. The reaction was started by adding different concentrations of NaClO stock solution with slow stirring in the whole time. After reaching a certain UV irradiation dosage, the UV light was closed to stop the reaction. Ten 10 ml of the sample was withdrawn, and the remaining oxidant was quenched using sodium thiosulfate (0.1 M). The molar ratio of the quenching agent to chlorine was 1.2. The residual 2,4,6-TCA concentration was detected by using GC-MS. Each set of experiments was repeated at least twice.

Analytical methods
Free chlorine concentration was conducted with buffer solution and N, N-diethyl-p-phenylenediamine (DPD) colorimetric method (Association and Association 1995). The concentration of 2,4,6-TCA was liquid-liquid extracted (LLE) with n-hexane and then analyzed by using GC-MS (QP2010 Ultra, Shimadzu, Japan). Details of the method were listed in Text S3.

Toxicity estimation
The ecological structure-activity relationship (ECOSAR) computer program (version 2.0) was used to evaluate the acute and chronic toxicity of 2,4,6-TCA and its TPs.

Wavelength-dependency of 2,4,6-TCA degradation during UV/chlorination
The degradation of 2,4,6-TCA by chlorination, UV irradiation, and UV/chlorination processes with different UV wavelengths (254, 275, and 285 nm) is displayed in Fig. 1. Chlorine alone was not effective for the removal of 2,4,6-TCA (< 3%). It has been reported that chlorination of anisole in acidic conditions produced a large amount of chlorinated anisole (Zhang et al. 2016a). Chlorine is the main cause of 2,4,6-TCA production. UV alone can degrade 2,4,6-TCA, because of π-systems (e.g., aromatic rings and C = C), carbonyl groups and heteroatoms (sulfur, nitrogen, and halogens) that promote the absorption of UV light (Carlson et al. 2015). The structure of 2,4,6-TCA is displayed in Table S1. However, the degradation effect of 2,4,6-TCA using different UV wavelengths was different. When the UV fluence reached 443 mJ cm −2 , the removal of 2,4,6-TCA declined in the order of 254 nm > 285 nm > 275 nm. The difference in the photolysis of 2,4,6-TCA by direct UV photolysis was mainly due to the different molar absorption coefficients at different UV wavelengths, similar to other micropollutants that can be photolyzed (Carlson et al. 2015;Gao et al. 2019a;Schnabel et al. 2021). Compared with direct photolysis, the removal of 2,4,6-TCA was significantly improved in UV/ chlorination, with the removal efficiency in the descending order of 275 nm > 254 nm > 285 nm (Fig. 1). Among them, the degradation efficiency of 275 nm for 2,4,6-TCA was improved by 50.4% compared with UV alone, which showed great superiority. This could be attributed to the process's production of reactive radicals such as HO• and RCSs (Fang et al. 2014;Gao et al. 2019b;), which will be covered in greater detail in Section of Kinetic model of 2,4,6-TCA degradation during UV/chlorination.

Effect of pH on UV/chlorination systems with different wavelengths
It is worth mentioning that the UV alone performed generally in the removal of 2,4,6-TCA, but the combination of UV and chlorine showed superiority at the wavelength of 275 nm. To further investigate the effect of UV wavelength on 2,4,6-TCA degradation, trials were conducted at different pH values of 5.0-9.0, and the UV fluence-based rate constant (k obs-UV ) was calculated with Eq. (1) (Yin et al. 2018).
where I represents the UV fluence (cm −2 mJ), k obs is the UV fluence-based rate constant (cm 2 mJ −1 ), and C t and C 0 are the concentrations of 2,4,6-TCA after and before UV irradiation (μM).
The k obs for the 2,4,6-TCA degradation at different pH values and wavelengths is summarized in Fig. 2. For 254 nm, the k obs decreased sharply as pH increased from 5.0 to 7.0 (from 1.2 × 10 −2 to 2.9 × 10 −3 cm 2 mJ −1 ). On the contrary, the k obs of 275 nm and 285 nm decreased relatively slow (from 9.2 × 10 −3 to 3.6 × 10 −3 cm 2 mJ −1 and from 6.7 × 10 −3 to 2.5 × 10 −3 cm 2 mJ −1 , respectively). At pH < 6.5, the k obs at 254 nm was the highest among the three wavelengths, while at pH ≥ 6.5, the k obs at 275 nm became the highest. The pH value of the actual water system is usually at neutral or slightly alkaline conditions; therefore, the wavelength of 275 nm is more suitable for 2,4,6-TCA degradation in the water treatment system. A similar tendency has been observed in the removal of HA (Gao et al. 2019b).

Chlorine photodecay rates and quantum yields
The effects of different UV wavelengths and solution pH values on chlorine degradation under UV irradiation (k obs, chlorine ) were studied, and the results are displayed in Fig. S2 and Fig. S3. Under UV irradiation at 254 nm, k obs, chlorine decreased by 95.1% as pH increased from 5.0 to 9.0, while at 275 nm and 285 nm, k obs, chlorine decreased slowly by 54.8% and 72.0%, respectively. At pH 7.0, the k obs, chlorine of 275 nm was higher than that of 254 nm, which was consistent with the higher degradation efficiency of 2,4,6-TCA by UV-LED at 275 nm in Fig. 1. The results showed that the degradation of chlorine by UV is a crucial factor for 2,4,6-TCA degradation.
The chlorine degradation characteristics can be explained in part by the molar absorption coefficient of chlorine, which reflects the absorption degree of the chlorine under UV irradiation (Fig. S4). As pH increased from 5.0 to 9.0, the molar absorption coefficient of chlorine at 254 nm did not change much, while in alkaline conditions, the molar absorbance coefficient of chlorine increased rapidly at 275 nm and 285 nm (Fig. S4). Consequently, the UV-LED/chlorination groups performed well. On the other hand, the influence of pH and UV wavelength on the molar absorbance coefficient of chlorine can be discussed in terms of the distribution of chlorine species. In acidic condition, HOCl is the dominant chlorine species, while in basic condition, OCl − is the dominant one. The molar absorption and photolysis quantum yield of HOCl and OCl − are significantly different at different wavelengths (Yin et al. 2018). In acidic conditions, HOCl is the dominant chlorine species, and the absorbance coefficient of HOCl at 254 nm is greater than that at 285 and 275 nm (Fig. S5), while in alkaline conditions, OCl − is the dominant chlorine species. Therefore, the distribution of chlorine species resulted in a higher absorbance coefficient of 254 nm under acidic conditions and of 275 and 285 nm under alkaline conditions, which was consistent with previous research (Yin et al. 2018).
However, the degradation trend of 2,4,6-TCA was not in perfect agreement with the molar absorption coefficient of chlorine, so other factors such as the quantum yield of chlorine should also be considered (Fig. S5). Quantum yield is the ratio of photochemical reaction photons to the total number of absorbed photons, and the detailed calculation method of the quantum yield of chlorine was the same as that described by Zou et al. (2019b) in previous research. The changing trend of chlorine quantum yield was consistent with the degradation efficiency of 2,4,6-TCA, namely, in acidic conditions, 254 > 275 > 285 nm, while in neutral and alkaline conditions, 275 > 254 > 285 nm ( Fig. 2 and Fig. S5). Although 285 nm had a high light absorption coefficient (Fig. S4), its quantum yield was relatively poor (Fig. S5), leading to an unsatisfactory degradation efficiency of 2,4,6-TCA. A similar tendency has been observed for the decomposition of iopamidol during UV/chlorination (Gao et al. 2019a).

EEO analysis
The photolysis reactions consume a large amount of electrical energy and the running costs are high. The economic efficiencies of different wavelength light sources were evaluated. The electrical energy of UV sources was calculated, and the chlorine dosages were also considered (Text S5) (Sun et al. 2016). The E EO-total value of different UV sources is shown in Fig. S6, where the hotter the color, the higher the E EO-total values. The E EO-total value of UV-LED was significantly lower than that of LPUV (254 nm), which represented the energy consumption of UV-LED/chlorination which was lower than LPUV/chlorination. Moreover, when pH values increased from 5.0 to 9.0 and UV fluence was up to 443.9 mJ cm −2 , the E EO-total values of UV-LED (275 nm) and UV-LED (285 nm) increased from 3.2308 × 10 −4 to 5.6800 × 10 −3 kWh L −1 and 4.2692 × 10 −4 to 3.0100 × 10 −3 kWh L −1 . Although the E EO-total values of UV-LED (275 nm) had a greater range of variation, the wavelength of 275 nm had lower E EO-total value under alkaline condition, which represented the possibility for application in water treatment plants.
In summary, the wavelength of UV irradiation and pH values have a great impact on the removal of 2,4,6-TCA during UV/chlorination by affecting the distribution of chlorine species and the corresponding molar absorption coefficient and quantum yield. Under neutral and alkaline conditions, the wavelength of 275 nm has the best removal efficiency and the lowest energy consumption (Fig. 2, Fig. S6); therefore, all subsequent studies were conducted at 275-nm UV wavelength.

Kinetic model of 2,4,6-TCA degradation during UV/ chlorination
The role of radicals on the degradation of 2,4,6-TCA As discussed in Fig. 1, UV-LED/chlorination at the wavelength of 275 nm exhibited the most efficiency in 2,4,6-TCA degradation of 77.9% in 15 min (k obs = 0.034 min −1 ). UV irradiation can excite free chlorine to produce reactive HO• and reactive chlorine species (RCSs, such as Cl•, Cl 2 − •), RCSs are selective oxidants, while HO• is a broad-spectrum  (Gao et al. 2020b). The coexistence of multiple reactive species in UV/chlorination system can promote the degradation of pollutants (Xu et al. 2021). During UV-LED (275 nm)/chlorination, the degradation of 2,4,6-TCA followed pseudo-first-order kinetics, and the pseudo-first-order reaction rate constant (k obs ) can be calculated using Eq. (2).
where k dobs represents the pseudo-first-order reaction rate constant of direct UV photolysis (s −1 ), and k idobs represents the pseudo-first-order reaction rate constant of indirect oxidation (s −1 ).
The contribution of different reactive species in the degradation of 2,4,6-TCA was determined by dosing an excess of nitrobenzene (NB) and tert-butyl-alcohol (TBA) as free radical-quenching agents . TBA eliminates HO• and RCSs (k HO•-TBA = 6.0 × 10 8 M −1 s −1 , k Cl•-TBA = 3.0 × 10 8 M −1 s −1 ), while NB only reacts with HO• (k HO•-NB = 3.9 × 10 9 M −1 s −1 ) (Kong et al. 2018), and the reactions of NB with chlorine and RCSs are negligible (Watts and Linden 2007). As depicted in Fig. 3, when excessive NB or TBA was added to the solution, the degradation efficiency of 2,4,6-TCA was reduced to almost the level of UV irradiation alone. The results of the experiments indicated that other radicals had less effect on 2,4,6-TCA degradation (3.7%) . And as displayed in Fig. S7, the removal of 2,4,6-TCA during UV/ chlorination was mainly through direct UV photolysis and indirect oxidation (HO•), contributing 23.7% and 72.6%, respectively.

Establishment of the kinetic model
In this study, a simplified kinetic model was established to simulate 2,4,6-TCA degradation with the following assumptions: (1) the UV/chlorination process of 2,4,6-TCA only involved the reactions summarized in Table 1; (2) the model did not take into account reactions and conversions between radicals and intermediate products; (3) HO• was the main radical to attack 2,4,6-TCA in the system; and (4) other factors of the solution were ignored as they can hardly affect the reaction system.
Based on the reactions considered in Table 1 during 2,4,6-TCA degradation, Eqs. (3)-(5) represent the differential equations for each species in the system. Some rate constants in Table 1 were acquired from other researches, and the missing ones (k 1 , k 2 , k 9 , and k 10 ) were calculated in this study. The photodecay rate of 2,4,6-TCA (k 9 ) was obtained by its degradation in UV alone process, and the other reaction rates were determined by fitting the differential equations and the experimental data of 2,4,6-TCA degradation during UV/chlorination in MATLAB 2018a using the ode15s function. According to the simulation results, the photodecay rates of HOCl and OCl − (k 1 and k 2 ) were solved as 1.5 × 10 −3 and 9.3 × 10 −3 s −1 , respectively, and the second-order rate constant of HO• reacting with 2,4,6-TCA (k 10 ) was calculated as 3.1 (± 0.01) × 10 9 M −1 s −1 , which was of the same magnitude with the experimental data (Peter and Gunten 2007). The 95% confidence interval was obtained by the function NLPARCI (Zhu et al. 2019b). The verification of model accuracy will be discussed in Section of Validation of the model by the effect of NOM.

Influence factors on 2,4,6-TCA degradation
Effects of solution pH, initial 2,4,6-TCA concentration, and chlorine dosage Figure 4a shows the effect of pH on 2,4,6-TCA degradation during UV-LED (275 nm)/chlorination. When the pH increased from 5.0 to 9.0, the k obs of 2,4,6-TCA   (Feng et al. 2007) 6 2,4,6-TCA + hv → products k 6 = 4.0 × 10 −4 s −1 This study 7 2,4,6-TCA + HO• → products k 7 = 3.1 (± 0.01) × 10 9 M −1 s −1 This study 8 HOCl ⇔ OCl − + H + k ac = 2.88 × 10 −8 (Feng et al. 2007) degradation decreased rapidly from 9.3 × 10 −3 cm 2 mJ −1 to 2.4 × 10 −3 cm 2 mJ −1 . This can be explained by the form and molar absorption of different chlorine species at different pH values, which has been discussed in the previous Section of Chlorine photodecay rates and quantum yields. At pH ≤ 7.5, however, free chlorine mainly exists in the form of HOCl (Yin et al. 2018), which can increase the production of HO• (Fang et al. 2014), thus improving the degradation efficacy and removal rate of the target pollutants. Figure 4b displays the effect of initial 2,4,6-TCA concentration on its degradation during UV/chlorination. As the 2,4,6-TCA concentration increased from 0.03 to 100 μg L −1 , its removal efficiency gradually decreased. A similar trend had been reported during UV/H 2 O 2 oxidation of 2,4,6-TCA (Luo et al. 2016). This can be interpreted by the fact that a certain oxidant concentration produced limited free radicals, the higher the substrate concentration, the lower the ratio of free radicals to react with target pollutants. Furthermore, 2,4,6-TCA can be direct photolysis and compete with chlorine for UV irradiation (Fig. 1), resulting in reduced HO• production in UV/chlorination. It is worth mentioning that 0.03 μg L −1 is very close to the level of 2,4,6-TCA in the actual water, and the removal efficiency is 82.9% when the UV fluence reached 443 cm −2 mJ, which proves the high potential of using UV-LED (275 nm)/chlorination in actual water treatment. Figure 4c depicts the effect of chlorine concentration on 2,4,6-TCA removal. With the chlorine levels ranging from 50 to 400 μM, the degradation of 2,4,6-TCA was positively correlated with chlorine concentration. The value of k obs increased from 1.9 × 10 −3 cm 2 mJ −1 to 4.1 × 10 −3 cm 2 mJ −1 as the chlorine concentration increased from 50 to 400 μM, indicating that the increasement of oxidant dosage can increase the steady-state concentration of HO• in the solution, so as to improve the degradation efficiency of 2,4,6-TCA (Watts and Linden 2007). However, it is worth noting that the increase in removal efficiency gradually slowed down when the chlorine concentration reached 200 μM, which could be explained in two aspects. First, when the oxidizer is in excess, the reaction between the generated free radical and 2,4,6-TCA is saturated (Zhang et al. 2018b). Secondly, excessive chlorine could quench HO•, forming less active species such as ClO• (Fang et al. 2014).

Validation of the model by the effect of NOM
NOM is ubiquitous in aquatic environments and is an important HO• trapping agent. Therefore, the presence of NOM is a critical factor affecting the efficiency of AOPs (Gao et al. 2020a). In this experiment, humic acid (HA) was selected as a representative of NOM to investigate its effect on UV-LED (275 nm)/chlorination of 2,4,6-TCA.
To validate the developed kinetic model's accuracy, a new set of differential equations were supplied in Text S6, and the reactions involved in the degradation process are listed in Table S2. The predicted results of the model and experimental data are displayed in Fig. 6. The correlation coefficients between the model predictions and experimental data were calculated (> 0.9818) and are listed in Table S3, which confirmed the validity of the second-order rate constant of HO• reacting with 2,4,6-TCA (k 10 in Section of Establishment of the kinetic model and Table 1) and the proposed model. The function nlparci was used to calculate the 95% confidence interval for each rate constant.
It can be seen from Fig. 6 that NOM significantly inhibited the degradation of 2,4,6-TCA in UV-LED (275 nm)/ chlorination. The k obs values decreased rapidly from 3.4 × 10 −3 to 1.1 × 10 −4 cm 2 mJ −1 when the HA concentration increased from 0 to 10 mg L −1 , which indicates that HA is a strong radical scavenger in UV/chlorination AOP. The inhibitory effect of HA on the degradation of 2,4,6-TCA can be explained in three ways: (1) HA was oxidized with the consumption of chlorine (Gao et al. 2019b;Li et al. 2016); (2) HA would absorb UV light and act as an inner light filter, thus slowing down the photolysis of chlorine (Goslan et al. 2006); (3) NOM can compete with 2,4,6-TCA for HO• and RCSs during the experiment (Goslan et al. 2006;Lutze et al. 2015). NOM has a significant inhibitory effect on UV/ chlorination. Because of its widespread presence, AOPs for drinking water treatment require special attention.

Degradation pathways and degradation of 2,4,6-TCA in real water during UV/chlorination
The degradation products after UV-LED (275 nm)/chlorination of 2,4,6-TCA were analyzed by using GC-MS and UPLC-TOF-Q-MS, and the degradation pathways are proposed accordingly in Fig. S8. A total of six oxidation intermediates were mainly detected and are listed in Table S4. The degradation of 2,4,6-TCA during UV/chlorination system is mainly by hydroxylation and dichlorination, as well as dichlorination followed by hydroxylation (Luo et al. 2016;Zhao et al. 2012;Zhu et al. 2019a). The toxicity of individual intermediates is often difficult to be detected directly. Table S5 shows the acute and chronic toxicity of 2,4,6-TCA and its TPs predicted by ECOSAR, and it can be seen that there is no significant increase in toxicity.
To further investigate the application of the UV-LED (275 nm)/chlorination in real water matrix, the treatment efficacy in ultrapure water (UPW), raw water (RW), finished water (FW), and tap water (TW) was discussed. As can be seen from Fig. 7, compared with the UPW, the RW, FW, and TW all affected the degradation efficiency of 2,4,6-TCA during UV/chlorination. The inhibition effect was the most significant in RW, with the k obs decreasing by 50% from 3.4 × 10 −3 to 1.7 × 10 −3 cm 2 mJ −1 . The effect of the FW and TW on degradation was relatively small with the k obs slightly decreased to 2.6 × 10 −3 and 2.7 × 10 −3 cm 2 mJ −1 , respectively.
Considering the different water matrices for different water sources (Table S6), the strong inhibitory effect of RW can be attributed to the relatively high turbidity and DOC. High turbidity can reduce the transmission of ultraviolet light and reduce the UV intensity , while organic compounds in water can compete with 2,4,6-TCA for HO• . The slightly alkaline pH value of RW may also be a factor affecting the efficiency of 2,4,6-TCA degradation because its degradation rate was slightly reduced under alkaline conditions (in Fig. 4a). Compared with TW, FW and TW had lower turbidity, pH value, and DOC concentration, so the inhibitory effect on UV/chlorination degradation of 2,4,6-TCA was much less. Notably, UV-LED/chlorination of 2,4,6-TCA performed well in TW, confirming the advantage of applying this technique to end water treatment of home-taps.

Conclusions
UV-LED/chlorination can effectively degrade 2,4,6-TCA at three wavelengths (254, 275, and 285 nm). In the indirect oxidation process, HO• was the dominant radical and the k HO•-2,4,6-TCA calculated as 3.1 × 10 9 M −1 s −1 using the developed kinetic model in this study. Decreasing pH and increasing chlorine dosage could enhance the degradation efficiency of 2,4,6-TCA. Both HCO 3 − and HA inhibited 2,4,6-TCA degradation, while Cl − had a negligible effect. Another kinetics model for the presence of NOM was proposed and with excellent prediction of the experimental results. Five degradation intermediates of 2,4,6-TCA during UV/chlorination were detected, and appropriate degradation pathways were proposed. Moreover, the experiments were also carried out in real water sources. Results showed that different water matrices had different degrees of inhibition on the degradation of 2,4,6-TCA. However, UV-LED/chlorination performed satisfactorily for the degradation of 2,4,6-TCA.