Facilitated Prediction of Micropollutant Degradation via UV-AOPs in Various Waters by Combining Model Simulation and Portable Measurement

Yanyan Huang; Mengkai Li; Zhe Sun; Wentao Li; James R. Bolton; Zhimin Qiang

doi:10.1016/j.eng.2023.10.009

Engineering ›› 2024, Vol. 37 ›› Issue (6) :97 -105. DOI: 10.1016/j.eng.2023.10.009

Research

Article

Facilitated Prediction of Micropollutant Degradation via UV-AOPs in Various Waters by Combining Model Simulation and Portable Measurement

Author information +

History +

PDF (1277KB)

Abstract

The degradation of micropollutants in water via ultraviolet (UV)-based advanced oxidation processes (AOPs) is strongly dependent on the water matrix. Various reactive radicals (RRs) formed in UV-AOPs have different reaction selectivities toward water matrices and degradation efficiencies for target micropollutants. Hence, process selection and optimization are crucial. This study developed a facilitated prediction method for the photon fluence-based rate constant for micropollutant degradation (k′_p_,MP) in various UV-AOPs by combining model simulation with portable measurement. Portable methods for measuring the scavenging capacities of the principal RRs (RRSCs) involved in UV-AOPs (i.e., $ \mathrm{HO}^{·}$, $ \mathrm{SO}_{4}^{·-}$, and $ \mathrm{Cl}^{·}$) using a mini-fluidic photoreaction system were proposed. The simulation models consisted of photochemical, quantitative structure–activity relationship, and radical concentration steady-state approximation models. The RRSCs were determined in eight test waters, and a higher RRSC was found to be associated with a more complex water matrix. Then, by taking sulfamethazine, caffeine, and carbamazepine as model micropollutants, the k′_p_,MP values in various UV-AOPs were predicted and further verified experimentally. A lower k′_p_,MP was found to be associated with a higher RRSC for a stronger RR competition; for example, k′_p,MP values of 130.9 and 332.5 m²·einstein^–1, respectively, were obtained for carbamazepine degradation by UV/H₂O₂ in the raw water (RRSC = 9.47 × 10⁴ s⁻¹) and sand-filtered effluent (RRSC = 2.87 × 10⁴ s⁻¹) of a drinking water treatment plant. The developed method facilitates process selection and optimization for UV-AOPs, which is essential for increasing the efficiency and cost-effectiveness of water treatment.

Graphical abstract

Keywords

UV-AOPs / Micropollutant degradation / Reactive radicals / Water matrix / Model simulation

Highlight

• Water matrix scavenging capacities for $ \mathrm{HO}^{·}$, $ \mathrm{SO}_{4}^{·-}$ and $ \mathrm{Cl}^{·}$ were measured portably.

• Model simulation consisted of photochemical, QSAR, and SSA models.

• k′_p,MP values in UV-AOPs were predicted in real waters and verified experimentally.

• The developed method facilitates the selection and optimization of UV-AOPs.

Cite this article

Download citation ▾

Yanyan Huang, Mengkai Li, Zhe Sun, Wentao Li, James R. Bolton, Zhimin Qiang. Facilitated Prediction of Micropollutant Degradation via UV-AOPs in Various Waters by Combining Model Simulation and Portable Measurement. Engineering, 2024, 37(6): 97-105 DOI:10.1016/j.eng.2023.10.009

登录浏览全文

4963

注册一个新账户忘记密码

1. Introduction

Micropollutants, such as endocrine-disrupting chemicals (EDCs), pharmaceuticals and personal care products (PPCPs), have attracted increasing attention because of their potential to threaten ecosystem and public health [1], [2]. In 2022, the issued Action Plan for the Control of Emerging Contaminants emphasized the need to efficiently remove micropollutants from water bodies in China. Ultraviolet (UV)-based advanced oxidation processes (UV-AOPs) are regarded as effective approaches for micropollutant degradation. The UV/hydrogen peroxide (H₂O₂) process, which can generate reactive radicals (RRs) such as the hydroxyl radical ($ \mathrm{HO}^{·}$), has been widely applied in treating drinking water, wastewater, and reclaimed water [3], [4], [5]. Emerging UV-AOPs such as UV/chlorine and UV/peroxydisulfate (PDS) have also demonstrated effectiveness in micropollutant degradation, attracting significant recent interest [6], [7].

The principal RRs involved in UV-AOPs are $ \mathrm{HO}^{·}$, $ \mathrm{SO}_{4}^{·-}$ and $ \mathrm{Cl}^{·}$; each has a different reactivity toward micropollutants, which influences the degradation efficiency for a target micropollutant, as quantified by the photon fluence-based pseudo-first-order rate constant (k′_p,MP) [8]. Moreover, real water matrices are complex, and major constituents such as natural organic matter (NOM) and inorganic ions (e.g., HCO₃⁻/CO₃²⁻ and NO₃⁻) are present at substantially higher concentrations than the target micropollutants. These constituents create intense competition for RRs against the target micropollutants, comprising a primary factor that weakens the efficiency of UV-AOPs (i.e., k′_p_,MP). In addition, the different reaction selectivities of RRs toward water matrices may lead to varied competitions in UV-AOPs. Therefore, the process selection and optimization of UV-AOPs based on real water matrices and target micropollutants are critical for achieving the efficient degradation of micropollutants, mitigating RR competition, and thereby increasing efficiency and cost-effectiveness.

To select and optimize UV-AOPs, evaluation experiments are typically conducted with real waters to determine the k′_p,MP [9]. Aside from the target micropollutants and water matrices (drinking water, wastewater, and reclaimed water) that strongly scavenge RRs, k′_p,MP is influenced by other factors such as the type of UV-AOP (e.g., UV/H₂O₂, UV/PDS, and UV/chlorine) and the operating conditions (e.g., reagent dose). Consequently, a significant number of experiments are required to identify the optimal process and operating conditions in order to enhance the efficacy and reduce the costs for target micropollutants removal in real waters. Such experiments involve frequent field water collections/transportations, as well as a high level of professional skill in the operation of the photoreactor (e.g., a quasi-collimated beam apparatus) and advanced analytical instruments (e.g., an ultra-high-performance liquid chromatograph-tandem mass spectrometer (UPLC-MS/MS) for micropollutant analysis), making them time-consuming, laborious, and expensive. Therefore, it is crucial to develop a facilitated method for predicting the k′_p,MP values of various UV-AOPs, considering the different influencing factors.

Photochemical and radical concentration steady-state approximation (SSA) models can be used to rapidly predict the k′_p,MP of UV-AOPs under various operating conditions [10], [11], [12]. In these models, the second-order reaction rate constants of target micropollutants with RRs (k_RR,MP), which are closely correlated with the molecular structures of the micropollutants, can be obtained using a variety of methods, including kinetics experiments, literature reports, and the quantitative structure-activity relationship (QSAR) model [13], [14]. Moreover, the water matrix strongly influences the efficiencies of UV-AOPs by competing with target micropollutants for photons and RRs, which must be considered in the model simulation. UV photon competition can be characterized by measuring the UV absorption coefficient of water at 254 nm (UV₂₅₄). Assessing the RR competition in real water is challenging yet critical for evaluating UV-AOP efficiency.

It is difficult to assess the competition for RRs between water matrices and target micropollutants because it is nearly impossible to analyze all the water matrix constituents and their concentrations, as required for the model simulation. Previous studies have proposed the RR scavenging capacity (RRSC) for $ \mathrm{HO}^{·}$ (i.e., HRSC) as a means of representing the competition of the water matrix for $ \mathrm{HO}^{·}$ against the target micropollutant [15]. Furthermore, a corresponding portable measurement method that does not require advanced analytical instruments has been developed. In this way, the k′_p_,MP by UV/H₂O₂ can easily be predicted using a model simulation (photochemical, SSA, and QSAR) in combination with portable measurement (the MS-PM method). However, this method does not consider other principal RRs in UV-AOPs such as $\mathrm{SO}_{4}^{·-}$ and $ \mathrm{Cl}^{·}$. Therefore, it is necessary to develop complete RRSC determination methods for all principal RRs involved in UV-AOPs, which are needed for k′_p_,MP prediction in real waters during the selection and optimization of UV-AOPs.

This study aimed to develop a facilitated prediction method for k′_p_,MP in real waters by various UV-AOPs, including the conventional UV/H₂O₂ process and the emerging UV/PDS and UV/chlorine processes. Portable measurement methods to determine the RRSCs for all principal RRs in UV-AOPs (i.e., $ \mathrm{HO}^{·}$, $\mathrm{SO}_{4}^{·-}$, and $ \mathrm{Cl}^{·}$) were proposed. Sulfamethazine (SMN), caffeine (CAF), and carbamazepine (CBZ) were selected as the model micropollutants in this study because of their high detection frequencies and concentrations in various waters [16], [17]. The k_RR_,MP values between the RRs and model micropollutants were predicted by means of the QSAR model. By combining the simulation models (i.e., photochemical, SSA, and QSAR models) and portable measurements of RRSCs with a calibration of UV₂₅₄, the k′_p,MP values in various UV-AOPs in eight different real waters were predicted and experimentally verified. This study provides an important tool for the process selection and optimization of UV-AOPs, which could greatly benefit efficiency improvement and cost savings in water treatment.

2. Materials and methods

2.1. Chemicals and test waters

All chemicals used were of reagent grade or higher. SMN, CAF, CBZ, PDS, and methylene blue (MB) were purchased from Sigma-Aldrich (USA). H₂O₂, chlorine, sodium bisulfite (NaHSO₃), isopropyl alcohol (IPA), benzoic acid (BA), and nitrobenzene (NB) were purchased from Sinopharm Chemical Reagent Co. (China).

Eight real waters with different matrices were studied. Ultrapure water (UP) was produced using a Milli-Q Advantage A10 system (Millipore, USA) with a resistivity of 18.2 MΩ·cm. Raw water (RW1) and sand-filtered effluent (SF) were collected from a drinking water treatment plant in the city of Yancheng, Jiangsu Province. Raw water (RW2), ultrafiltration effluent (UF), and polymeric ferric sulfate coagulation (PFS)/UF effluent were collected from a rural drinking water treatment facility in the city of Changzhou, Jiangsu Province. Primary and secondary sedimentation effluents (PrS and SeS) were collected from a municipal sewage treatment plant in Beijing. The major water quality parameters are listed in Table 1. The test waters were filtered through 0.45 μm membrane filters and then stored at 4 °C before the experiments.

2.2. Analytical methods

SMN, CAF, and CBZ were detected via an UPLC-MS/MS (Agilent, USA). NB and BA were detected by means of a high-performance liquid chromatograph (Agilent). The detailed analysis methods are described in Text S1 in Appendix A, and the mass spectrometer (MS) conditions for the three micropollutants are listed in Table S1 in Appendix A. MB was analyzed using an online DR1900 spectrophotometer (Hach, USA) at 664 nm. Dissolved organic carbon (DOC) and ferric ions (Fe³⁺) were determined by means of a total organic carbon (TOC)-VCPH/CPN analyzer (Shimadzu, Japan) and an Optima 2000 inductively coupled plasma optical emission spectrometer (PerkinElmer, USA), respectively. The initial concentrations of SMN, CAF, and CBZ were all set at 0.1 mg·L⁻¹. All experiments were conducted at least twice.

2.3. Simulation models

2.3.1. Photochemical model

The micropollutant degradation rate constants by UV-AOPs (k′_p,MP) are equal to the sum of the direct photolysis rate constants (k′_d,MP) and the indirect oxidation rate constants by RRs (k′_i,MP), as shown in Eq. (1):

(1)

k_{p, MP}^{'} = k_{d, MP}^{'} + k_{i, MP}^{'}

The k′_d,MP can be further expressed by Eq. (2) [18]:

(2)

k_{d, MP}^{'} = \frac{Φ_{MP} q_{0} f_{MP} (1 - 1 0^{- a l^{'}})}{V E_{p, UV}^{0} [MP]}

where Φ_MP is the quantum yield (mol·einstein^-1) of the micropollutant by UV photolysis; f_MP is the UV photon absorption fraction of the micropollutant; q₀ and E_p,UV⁰ are the incident UV photon flux (einstein·s⁻¹) and fluence rate (einstein·m⁻²·s⁻¹), respectively; V and a are the total volume (L) and absorption coefficient (cm⁻¹) of the reaction solution, respectively; and l′ is the effective optical path-length of the photoreactor (cm). In this study, a previously developed mini-fluidic photoreaction system (MFPS) equipped with an 8 W cold-cathode low-pressure mercury lamp was used, whose configuration is described in Text S2 in Appendix A. The values for E⁰_p,UV, q₀, l′, and V were 3.98 × 10^-4 einstein·m⁻²·s⁻¹, 4.33 × 10^-6 einstein·s⁻¹, 0.46 cm, and 50 mL, respectively.

RRs ($ \mathrm{HO}^{·}$, $\mathrm{SO}_{4}^{·-}$, and $ \mathrm{Cl}^{·}$) were formed through the UV photolysis of the oxidants (H₂O₂, PDS, and chlorine), whose formation rates ($r_{\mathrm{HO}^{·}}$, $r_{\mathrm{SO}_{4}^{·-}}$, and $r_{\mathrm{Cl}^{·}}$) were calculated as follows [5], [11], [12]:

(3)

r_{H O^{\cdot}, UV / H_{2} O_{2}} = \frac{Φ_{H_{2} O_{2}} q_{0} f_{H_{2} O_{2}} (1 - 1 0^{- a l^{'}})}{V}

(4)

r_{S {O_{4}}^{\cdot -}, UV/PDS} = \frac{Φ_{PDS} q_{0} f_{PDS} (1 - 1 0^{- a l^{'}})}{V}

(5)

r_{H O^{\cdot}, UV/chlorine} = \frac{Φ_{HOCl} q_{0} f_{HOCl} (1 - 1 0^{- a l^{'}})}{V}

(6)

r_{C l^{\cdot}, UV/chlorine} = \frac{(Φ_{HOCl} f_{HOCl} + Φ_{OC l^{-}} f_{OC l^{-}}) q_{0} (1 - 1 0^{- a l^{'}})}{V}

where $\Phi_{\mathrm{H}_{2} \mathrm{O}_{2}}$, Φ_PDS, Φ_HOCl, and Φ_OCl^– are the quantum yields (mol·einstein^–1) of H₂O₂, PDS, HOCl, and OCl^– by UV photolysis, respectively; and $f_{\mathrm{H}_{2} \mathrm{O}_{2}}$, f_PDS, f_HOCl, and f_OCl⁻ are the UV photon absorption fractions of H₂O₂, PDS, HOCl, and OCl^–, respectively (the calculation is provided in Text S3 in Appendix A).

2.3.2. SSA model

The SSA model has been widely used in kinetic simulations of micropollutant degradation by UV-AOPs [19], [20]. For the UV/PDS process, the steady-state concentration of its dominant RRs (i.e., [$\mathrm{SO}_{4}^{·-}$]_ss and [$ \mathrm{HO}^{·}$]_ss) can be calculated as follows [11]:

(7)

{[S {O_{4}}^{\cdot -}]}_{ss} = \frac{r_{S {O_{4}}^{\cdot -}}}{\sum k_{S {O_{4}}^{\cdot -}, S_{i}} [S_{i}] + k_{S {O_{4}}^{\cdot -}, MP} [MP] + k_{S {O_{4}}^{\cdot -}, PDS} [PDS] + k_{S {O_{4}}^{\cdot -}, H_{2} O} [H_{2} O] + k_{S {O_{4}}^{\cdot -}, O H^{-}} [O H^{-}]}

(8)

{[H O^{\cdot}]}_{ss} = \frac{(k_{S {O_{4}}^{\cdot -}, O H^{-}} [O H^{-}] + k_{S {O_{4}}^{\cdot -}, H_{2} O} [H_{2} O]) \times {[S {O_{4}}^{\cdot -}]}_{ss}}{\sum k_{H O^{\cdot}, S_{i}} [S_{i}] + k_{H O^{\cdot}, MP} [MP] + k_{H O^{\cdot}, PDS} [PDS] + k_{H O^{\cdot}, O H^{-}} [O H^{-}]}

where $\Sigma k_{\mathrm{SO}_{4}^{·-}, \mathrm{S}_{i}}\left[\mathrm{~S}_{i}\right]$ and $\Sigma k_{\mathrm{HO}^{·}, S_{i}}\left[\mathrm{~S}_{i}\right]$ are the RRSCs for $\mathrm{SO}_{4}^{·-}$ and $ \mathrm{HO}^{·}$ of a water matrix (s⁻¹), respectively; $k_{\mathrm{SO}_{4}^{·-}, \mathrm{MP}}$, $k_{\mathrm{SO}_{4}^{·-}, \mathrm{PDS}}$, $k_{\mathrm{SO}_{4}^{·-}, \mathrm{H}_{2} \mathrm{O}}$, and $k_{\mathrm{SO}_{4}^{·-}, \mathrm{OH}^{-}}$ are the second-order reaction rate constants of the micropollutant, PDS, H₂O, and OH⁻ with $\mathrm{SO}_{4}^{·-}$, respectively; and $k_{\mathrm{HO}^{·}, \mathrm{MP}}$, $k_{\mathrm{HO}^{·}, \mathrm{PDS}}$, and $k_{\mathrm{HO}^{·},\mathrm{OH}^{-}} $ are the second-order reaction rate constants of the micropollutant, PDS, and OH⁻ with $ \mathrm{HO}^{·}$, respectively. The k′_i_,MP by UV/PDS can be calculated using Eq. (9) [11]:

(9)

k'_{i,MP} = \frac{k_{S {O_{4}}^{\cdot -}, MP} {[S {O_{4}}^{\cdot -}]}_{ss} + k_{H O^{\cdot}, MP} {[H O^{\cdot}]}_{ss}}{E_{p, UV}^{0}}

By analogy, the steady-state concentration of $ \mathrm{HO}^{·}$ in the UV/H₂O₂ process and $ \mathrm{Cl}^{·}$ and $ \mathrm{HO}^{·}$ in the UV/chlorine process, as well as the k′_i_,MP, were calculated using the SSA model. The calculation is shown in Text S4 in Appendix A.

2.3.3. QSAR model

The k_RR_,MP was predicted using the QSAR model. A standardized procedure was adopted, involving dataset establishment, descriptor calculation, and model development and assessment [21]. It should be noted that the $k_{\mathrm{HO}^{·}}$,MP was predicted using the QSAR model established previously [15]. For $k_{\mathrm{SO}_{4}^{·-}}$,MP and $k_{\mathrm{Cl}^{·}}$,MP, first, organic compounds (OCs) with diverse functional groups were selected to establish the dataset, and their second-order reaction rate constants with RRs (i.e., k_RR,OC) were collected from the literature. In this study, 53 OCs (40 in the training set and 13 in the validation set) were selected for the $k_{\mathrm{SO}_{4}^{·-}}$,OC dataset, and 20 (16 in the training set and four in the validation set) were selected for the $k_{\mathrm{Cl}^{·}}$,OC dataset. Second, 19 descriptors reflecting the constitutional, electrostatic, and quantum chemical properties that were reported as being pertinent to the radical reactivity were calculated for each OC in the dataset [14], [22]. The local minimum-energy conformation of each molecule was achieved using density functional theory with the B3LYP function, 6-31+G(d) basis set, and polarizable continuum model of the Gaussian 09 program. Finally, Pearson correlation analysis, principal component analysis (PCA), and stepwise multiple linear regression (MLR) were performed in the training set using SPSS 19.0 to establish the QSAR model between lnk_RR,OC and the most significant descriptors. The robustness and predictability of the established QSAR model were further assessed using the correlation coefficient (R²), leave-one-out cross-validation (Q_LOO²), and external explained variance (Q_Ext²), as detailed in Text S5 in Appendix A. The k_RR_,MP of a target micropollutant could be obtained by substituting its specific descriptors into the established QSAR model.

2.4. Portable measurement of RRSCs

Portable measurements of various RRSCs, including the scavenging capacities for $ \mathrm{HO}^{·}$ (HRSC), $\mathrm{SO}_{4}^{·-}$ (SRSC), and $ \mathrm{Cl}^{·}$ (CRSC), were performed using the MFPS. The detailed portable measurement procedure for the HRSC can be found in Ref. [15]. In brief, MB and IPA were selected as the probe and surrogate matrix component (SMC), respectively. The relationship between the pseudo-first-order rate constants of MB degradation (k′_p_,MB) by UV/H₂O₂ and the HRSC of the water matrix were derived based on the SSA model, as described in Eq. (10). The detailed derivation is provided in Text S6 in Appendix A. To establish the standard curve of HRSC (i.e., the linear relationship between 1/k′_p_,MB and the HRSC, as shown in Fig. S1 in Appendix A), the k′_p_,MB values were measured in a series of IPA solutions with known HRSCs. Finally, the k′_p_,MB by UV/H₂O₂ in real water was measured, and the HRSC of the real water was obtained by referring the k′_p,MB to the standard curve of the HRSC.

(10)

\frac{1}{k_{p, MB}^{'}} = \frac{{Ef}_{P,UV}^{0}}{k_{H O^{\cdot}, MB} r_{H O^{\cdot}}} HRSC + \frac{E_{P,UV}^{0} (k_{H O^{\cdot}, MB} [MB] + k_{H O^{\cdot}, H_{2} O_{2}} [H_{2} O_{2}])}{k_{H O^{\cdot}, MB} r_{H O^{\cdot}}}

The portable measurement procedures for the SRSC and CRSC were generally similar to those for the HRSC. BA was selected as an SMC instead of IPA (for the HRSC) because of its high second-order reaction rate constants with $\mathrm{SO}_{4}^{·-}$ and $ \mathrm{Cl}^{·}$ ($k_{\mathrm{SO}_{4}^{·-}, \mathrm{BA}}$=1.2 × 10⁹ (mol·L^–1)^–1·s⁻¹, $k_{\mathrm{Cl}^{·}}$,BA = 1.8 × 10¹⁰ (mol·L^–1)^–1·s⁻¹) [11], [12]. In addition, for the UV/PDS and UV/chlorine processes, 0.5 mmol·L^–1 NB was spiked for $ \mathrm{HO}^{·}$ quenching to ensure that only $ \mathrm{HO}^{·}$ and $ \mathrm{Cl}^{·}$ were present [23], [24].

When measuring RRSCs, discrepancies in the UV₂₅₄ between the SMC solutions and the real waters can result in different formation rates of RRs, thereby introducing errors during RRSC measurement. To account for this, a UV₂₅₄ calibration factor (F_a) was proposed, defined as the formation rate of the RR in the SMC solution (r_RR,SMC) divided by that in a test water (r_RR,TW) as shown in Eq. (11):

(11)

F_{a} = \frac{r_{RR,SMC}}{r_{RR,TW}}

Hence, to account for the errors resulting from discrepancies in UV₂₅₄, the measured 1/k′_p,MB was divided by F_a before being applied to calculate the RRSCs according to the standard curves.

2.5. Prediction procedures

The prediction procedures for the k′_p,MP in the UV/H₂O₂, UV/PDS, and UV/chlorine processes using the MS-PM method are illustrated in Fig. 1 and listed below:

(1) Portably measure the RRSCs of the test waters on the MFPS;

(2) Predict the k′_d,MP (i.e., Eq. (2)) and the r_RR of the applied UV-AOPs (i.e., Eqs. (3), (4), (5), (6)) using the photochemical model;

(3) Predict the k_RR,MP of the target micropollutants using the QSAR model;

(4) Predict the k′_i,MP by substituting the RRSCs, r_RR, and k_RR,MP into the SSA model;

(5) Calculate the k′_p,MP by adding up the k′_d,MP and k′_i,MP (i.e., Eq. (1)).

3. Results and discussion

3.1. Determination of RRSCs of test waters

The established standard curves for the HRSC, SRSC, and CRSC, which can be used for the SC determination of real water matrices, are shown in Fig. 2(a). The slopes of the HRSC, SRSC, and CRSC standard curves were determined to be 3.10 × 10⁻⁸, 2.78 × 10⁻⁷, and 1.58 × 10⁻⁸, respectively. The F_a values for the UV/H₂O₂, UV/PDS, and UV/chlorine processes were almost identical, because the F_a was primarily influenced by the UV₂₅₄ value of the test water and the l′ of the photoreactor rather than the added oxidant, according to its definition in Eq. (11). The F_a values were calculated as 1.000, 1.029, 1.047, 1.041, 1.049, 1.054, 1.079, and 1.225 for UP, SF, RW1, PFS/UF, UF, RW2, SeS, and PrS, respectively.

The RRSCs were obtained by referring the k′_p,MB values, which were measured in various test waters, to the standard curves with an F_a calibration. Fig. 2(b) shows the measured RRSCs of eight test waters. A higher RRSC was found to be associated with a more complex water matrix. The highest HRSC (1.51 × 10⁶ s⁻¹), SRSC (4.22 × 10⁵ s⁻¹), and CRSC (3.97 × 10⁶ s⁻¹) were observed in the PrS due to its complex water matrix, with DOC of 95.3 mg·L^-1 (Table 1). In contrast, UP exhibited the lowest HRSC (9.33 × 10³ s⁻¹), SRSC (3.71 × 10³ s⁻¹), and CRSC (5.59 × 10⁴ s⁻¹). In addition, the RRSCs of the SF, PFS/UF, and SeS were lower than those of RW1, RW2, and PrS, respectively, indicating that the treatment processes can effectively remove the RR-scavenging constituents in the water matrices.

In each test water, the RRSCs followed the order of SRSC < HRSC < CRSC, which can be ascribed to the different reaction reactivities of the RRs toward the water matrix. For example, $\mathrm{SO}_{4}^{·-}$ has the lowest second-order reaction rate constant with NOM (2.0 × 10³ (mg·L^–1)^–1·s⁻¹) among the three RRs, whereas $ \mathrm{Cl}^{·}$ has the highest second-order reaction rate constant with HCO₃⁻ (2.2 × 10⁸ (mol·L^–1)^–1·s⁻¹) [12], [25]. However, it is difficult or even impossible to analyze the concentration of each constituent in a water matrix during practical application. Hence, proposing the RRSC and its portable measurement method could be helpful in facilitating the evaluation of the RR competition of a water matrix against the target micropollutants.

**3.2. Prediction of k′_d,MP and r_RR via the photochemical model**

The k′_d,MP values of the three model micropollutants by UV-AOPs were predicted using Eq. (2). For each model micropollutant, no obvious difference was observed among the UV/H₂O₂, UV/PDS, and UV/chlorine processes, as shown in Fig. S2 in Appendix A. SMN exhibited the highest k′_d,MP because of the highest quantum yield and molar absorption coefficient (ε) of SMN at 254 nm (Φ_SMN = 0.005 mol·einstein^-1, ε_SMN = 17 000 (mol·L^-1)^-1·cm⁻¹), while CBZ exhibited the lowest k′_d,MP (Φ_CBZ = 0.0006 mol·einstein^-1, ε_CBZ = 8983 (mol·L^-1)^-1·cm⁻¹) [18], [26].

The r_RR values in eight test waters by UV-AOPs were calculated using the photochemical model and were found to range from 5.2 × 10⁻⁸ to 8.4 × 10⁻⁷ mol·L^-1·s⁻¹ at [H₂O₂]₀ = 15.0 mg·L^-1, [PDS]₀ = 50.0 mg·L^-1, and [chlorine]₀ = 1.0 mg·L^-1 (Table S2 in Appendix A). The difference in the r_RR values among various test waters arose from their diverse UV₂₅₄ values. For a given test water, the r_RR was nearly identical for various model micropollutants because the UV₂₅₄ of the micropollutants (0.004-0.006 cm⁻¹) was negligible compared with that of the reaction solutions (0.015-0.423 cm⁻¹).

3.3. Prediction of k_RR,MP via the QSAR model

The $k_{\mathrm{HO}^{·}}$,MP values for various model micropollutants were predicted by means of the previously developed QSAR model [15]. To predict the $k_{\mathrm{SO}_{4}^{·-}, \mathrm{MP}}$ and $k_{\mathrm{Cl}^{·}, \mathrm{MP}}$, a Pearson correlation analysis was first performed between ln$k_{\mathrm{SO}_{4}^{·-}, \mathrm{OC}}$ (or ln$k_{\mathrm{Cl}^{·}, \mathrm{OC}}$) and each descriptor. Out of 19 descriptors, ten (#C, #H, #H:C, #O:C, #ringatoms, #nonHatoms, DBE, HLG, EA, and IP) were selected with an absolute Pearson correlation coefficient higher than 0.4 at the 0.01 significance level for ln$k_{\mathrm{SO}_{4}^{·-}, \mathrm{OC}}$, and another ten (#C, #H, #O, #N:C, #acid, #ringatoms, #nonHatoms, DBE, DM, and EA) for ln$k_{\mathrm{Cl}^{·}, \mathrm{OC}}$ (for the full description of the descriptors, please see Table S3 in Appendix A). PCA was then performed for further reduction of the variables. As illustrated in Figs. 3(a) and (b), Components 1 and 2 accounted for 73.5% and 85.6% of the total variance for ln$k_{\mathrm{SO}_{4}^{·-}, \mathrm{OC}}$ and ln$k_{\mathrm{Cl}^{·}, \mathrm{OC}}$, respectively. For ln$k_{\mathrm{SO}_{4}^{·-}, \mathrm{OC}}$, HLG, #H:C, and DBE had loading scores with high absolute values in Component 1, as did #H and #O:C in Component 2 (Fig. 3(c)). For ln$k_{\mathrm{Cl}^{·}}$,OC, #C, DM, and #acid had high values in Components 1 and 2; EA was also selected because it did not cluster with other descriptors (Fig. 3(d)). Finally, stepwise MLR was used to establish the relationship between ln$k_{\mathrm{SO}_{4}^{·-}, \mathrm{OC}}$ (or ln$k_{\mathrm{Cl}^{·}, \mathrm{OC}}$) and the selected descriptors (Table S4 in Appendix A). Considering both the accuracy (R², Q²_LOO, and Q²_Ext) and the conciseness of the equation, the QSAR models of $k_{\mathrm{SO}_{4}^{·-}}$,OC and $k_{\mathrm{Cl}^{·}}$,OC were established as follows:

(12)

ln k_{S {O_{4}}^{\cdot -}, OC} = 26.335 - 4.103 \times (# O : C) - 0.69 \times HLG

(13)

ln k_{C l^{\cdot}, OC} = 20.731 + 0.588 \times DM + 0.239 \times EA

As illustrated in Figs. 3(e) and (f), the predicted ln$k_{\mathrm{SO}_{4}^{·-}, \mathrm{OC}}$ and ln$k_{\mathrm{Cl}^{·}}$,OC values agreed well with the corresponding literature values [13], [22], [27], [28], [29], indicating the accuracy of the established QSAR models. The values of the involved descriptors in the QSAR models and the literature and the predicted $k_{\mathrm{SO}_{4}^{·-}, \mathrm{OC}}$ and $k_{\mathrm{Cl}^{·}}$,OC values of each OC in the entire dataset are respectively listed in Tables S5 and S6 in Appendix A. By substituting the descriptors of the target micropollutants (i.e., SMN, CAF, and CBZ) into the QSAR models, $k_{\mathrm{SO}_{4}^{·-}, \mathrm{MP}}$ and $k_{\mathrm{Cl}^{·}, \mathrm{MP}}$ were obtained. As shown in Table 2 [13], [15], [30], [31], [32], [33], [34], the predicted k_RR_,MP values had the same order of magnitude as those in the literature or experimentally measured values, indicating an accuracy that was considered to be acceptable [35].

3.4. Prediction of k′_p,MP and experimental verification

By substituting the determined RRSCs, r_RR, and k_RR_,MP into the SSA model, the [RR]_ss in various UV-AOPs was calculated. For the UV/PDS process, the [SO₄^·−]_ss and [HO^·]_ss values (Table S7 in Appendix A), calculated using Eqs. (7), (8), exhibited a negative correlation with the SRSCs and HRSCs of the test waters (Fig. 2(b)). For example, the lowest [SO₄^·−]_ss value for SMN degradation (1.00 × 10⁻¹² mol·L^–1) was found in the PrS with the highest SRSC value (4.22 × 10⁵ s⁻¹), indicating the strong competition of the water matrix for SO₄^·−. For a certain test water, [SO₄^·−]_ss (or [HO^·]_ss) varied with the type of micropollutant (SMN, CAF, and CBZ) because of their different $k_{\mathrm{SO}_{4}^{·-}}$,MP (or $k_{\mathrm{HO}^{·}, \mathrm{MP}}$) values (Table 2). Similarly, the results for the [HO^·]_ss in the UV/H₂O₂ process and the [Cl^·]_ss and [HO^·]_ss in the UV/chlorine process (Tables S8 and S9 in Appendix A) were similar to those in the UV/PDS process in eight test waters.

With the calculated [RR]_ss, the k′_i,MP values for the various model micropollutants by UV-AOPs were predicted in the eight test waters, allowing the k′_p,MP values to be readily obtained (Eq. (1)). The predicted k′_p,MP values and the corresponding experimental verification values determined using the MFPS are illustrated in Fig. 4 and shown in Table S10 in Appendix A. Almost all predicted k′_p,MP values fall within the 95% confidence intervals of the measured values, which demonstrates the feasibility of the developed MS-PM method. Both the predicted and measured k′_p_,SMN values were higher than the k′_p,CAF and k′_p,CBZ values (Table S10) because of the higher k′_d,MP and k_RR,MP values of SMN (Table 2). Moreover, most of the predicted k′_p,CAF and k′_p,CBZ values by UV/PDS were higher than the measured ones (Table S10), which could be attributed to the overestimated $k_{\mathrm{SO}_{4}^{·-}, \mathrm{ CAF }}$ and $k_{\mathrm{SO}_{4}^{·-}, \mathrm{ CBZ }}$ values from the QSAR model (Table 2).

3.5. Discussion

This study proposed an MS-PM method for the facilitated prediction of k′_p_,MP by various UV-AOPs in real waters, with consideration of all principal RRs. According to the challenge of assessing the RR competition in real water, portable RRSC measurement methods for $ \mathrm{HO}^{·}$, $\mathrm{SO}_{4}^{·-}$, and $ \mathrm{Cl}^{·}$ were proposed. The MS-PM method does not require an advanced analytical instrument; rather, it can be carried out in the field via a small photoreactor (e.g., MFPS) and a portable spectrophotometer (e.g., HACH DR1900) using a mobile power source, thereby saving the trouble of frequent field water collection and transportation. The verification results demonstrated the acceptable accuracy of the developed method, indicating its feasibility for the selection and optimization of UV-AOPs in practice.

The potential errors in the k′_p,MP prediction may originate from the degradation contribution of neglected RRs, as well as overestimated RRSCs during the portable measurement, aside from the errors in k_RR,MP predicted by the QSAR model, as discussed above. On the one hand, $ \mathrm{HO}^{·}$ (UV/H₂O₂), $\mathrm{SO}_{4}^{·-}$ and $ \mathrm{HO}^{·}$ (UV/PDS), and $ \mathrm{Cl}^{·}$ and $ \mathrm{HO}^{·}$ (UV/chlorine) were considered during the model simulation, but other RRs such as $\mathrm{O}_{2}^{·-}$ in the UV/H₂O₂ process, $\mathrm{S}_{2} \mathrm{O}_{8}^{·-}$ in the UV/PDS process, and $\mathrm{Cl}_{2}^{·-}$ and $\mathrm{ClO}^{·}$ in the UV/chlorine process were not considered, which may have led to an underestimation of k′_p,MP [13], [36], [37]. Hence, the developed method should consider the RRSCs and k_RR_,MP for these RRs in order to improve the prediction accuracy in the future. On the other hand, different RRSCs (i.e., HRSC, CRSC, and SRSC) of a given water matrix were measured individually, yet some active sites of the water matrix constituents can react with multiple RRs formed in UV-AOPs. This could lead to an overestimation of the RRSCs and a subsequent underestimation of [RR]_ss and the k′_p,MP values.

The k′_p,MP in various UV-AOPs and under various operating conditions was easily predicted by the developed MS-PM method in practice. For example, taking CBZ as the target micropollutant, by using the measured HRSCs, SRSCs, and CRSCs in the eight test waters, as well as the k_RR,CBZ predicted by the QSAR models, the k′_p,CBZ by various UV-AOPs using oxidants with identical molar concentrations (0.2 or 0.5 mmol·L^-1) were predicted (Fig. 5). In general, the k′_p,CBZ can be influenced by multiple factors, such as micropollutants (k′_d,CBZ and k_RR,CBZ), water matrices (RRSCs), UV-AOPs, and operating conditions, which are considered in the MS-PM method. The prediction results indicated that UV/PDS had the highest degradation efficiencies at an oxidant concentration of both 0.2 and 0.5 mmol·L^-1 for most test waters, which was primarily attributed to its lowest SRSC values among the three RRSCs. Therefore, based on the facilitated prediction obtained using the MS-PM method, appropriate process selection and optimization could be conducted according to various real water matrices and target micropollutants.

In practice, a change in the inflow water matrix can impact the stability of a UV-AOP. The developed portable measurement method for various RRSCs also presents a feasible approach for monitoring changes in the water matrix. Furthermore, based on the online monitored RRSCs and UV₂₅₄, the real-time efficiency of the UV-AOP could be obtained in practice by using simulation models with preset model parameters. This could guide the online adjustment of operating parameters, such as UV fluence (e.g., lamp power) and oxidant dose, to ensure adequate micropollutant degradation at a high RRSC value of the water matrix, or to avoid energy waste at a low RRSC value.

4. Conclusions

In this study, the k′_p,MP values of various UV-AOPs in eight test waters were evaluated using a novel MS-PM method. Based on the experimental results, the following conclusions can be drawn:

· Portable measurement methods for the RRSCs for the principal RRs involved in UV-AOPs (i.e., $ \mathrm{HO}^{·}$, $\mathrm{SO}_{4}^{·-}$, and $ \mathrm{Cl}^{·}$) were proposed, and the error introduced from the UV₂₅₄ discrepancies between the SMC and the test waters was calibrated. The HRSC, SRSC, and CRSC values for eight test waters were measured. A higher RRSC was found to be associated with a more complex water matrix. For a given test water, SRSC < HRSC < CRSC because of the different reaction selectivities of the RRs.

· The model simulation consisted of photochemical, QSAR, and SSA models. The k′_d,MP and r_RR for various UV-AOPs were calculated via the photochemical model. The k_RR,MP values for three model micropollutants were predicted via the developed QSAR models for various RRs, which generally agreed well with the literature values. By substituting the measured RRSCs and simulated r_RR and k_RR,MP into the SSA models, the k′_p,MP could easily be predicted.

· The predicted k′_p,MP values by UV-AOPs in the eight test waters exhibited a descending order, as follows: UP > SF > RW1 > PFS/UF > UF> RW2 > SeS > PrS. This order was verified by the experimental results. A lower k′_p,MP was found to be associated with a higher RRSC, indicating that a more complex water matrix competes for more RRs against the target micropollutants.

· The developed prediction method can be used to guide the process selection and optimization of UV-AOPs. It can also provide a feasible approach for monitoring water matrix change and enable the online adjustment of operating parameters, which is important for increasing the efficiency and cost-effectiveness of UV-AOPs.

Acknowledgments

This work was financially supported by the National Natural Science Foundation of China (52222002), Bureau of International Cooperation of Chinese Academy of Sciences (032GJHZ2022035MI), and State Key Laboratory of Environmental Aquatic Chemistry (23Z01ESPCR).

Compliance with ethics guidelines

Yanyan Huang, Mengkai Li, Zhe Sun, Wentao Li, James R. Bolton, and Zhimin Qiang declare that they have no conflicts of interest or financial conflicts to disclose.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.eng.2023.10.009.

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	R.P. Schwarzenbach, B.I. Escher, K. Fenner, T.B. Hofstetter, C.A. Johnson, U. von Gunten, et al. The challenge of micropollutants in aquatic systems. Science, 313 (5790) (2006), pp. 1072-1077.

[2]	N.Z. Arman, S. Salmiati, A. Aris, M.R. Salim, T.H. Nazifa, M.S. Muhamad, et al. A review on emerging pollutants in the water environment: existences, health effects and treatment processes. Water, 13 (22) (2021), p. 3258.

[3]	Y. Huang, M.H. Kong, S. Coffin, K.H. Cochran, D.C. Westerman, D. Schlenk, et al. Degradation of contaminants of emerging concern by UV/H₂O₂ for water reuse: kinetics, mechanisms, and cytotoxicity analysis. Water Res, 174 (2020), Article 115587.

[4]	K.H. Guo, Z.H. Wu, S.W. Yan, B. Yao, W.H. Song, Z.C. Hua, et al. Comparison of the UV/chlorine and UV/H₂O₂ processes in the degradation of PPCPs in simulated drinking water and wastewater: kinetics, radical mechanism and energy requirements. Water Res, 147 (2018), pp. 184-194.

[5]	J.E. Grebel, J.J. Pignatello, W.A. Mitch. Effect of halide ions and carbonates on organic contaminant degradation by hydroxyl radical-based advanced oxidation processes in saline waters. Environ Sci Technol, 44 (17) (2010), pp. 6822-6828.

[6]	X.J. Kong, J. Jiang, J. Ma, Y. Yang, W.L. Liu, Y.L. Liu. Degradation of atrazine by UV/chlorine: efficiency, influencing factors, and products. Water Res, 90 (2016), pp. 15-23.

[7]	J. Lee, U. von Gunten, J.H. Kim. Persulfate-based advanced oxidation: critical assessment of opportunities and roadblocks. Environ Sci Technol, 54 (6) (2020), pp. 3064-3081.

[8]	W. Li, T. Jain, K. Ishida, H.Z. Liu. A mechanistic understanding of the degradation of trace organic contaminants by UV/hydrogen peroxide, UV/persulfate and UV/free chlorine for water reuse. Enviro Sci-Wat Res, 3 (1) (2017), pp. 128-138.

[9]	E.J. Rosenfeldt, K.G. Linden. The ROH, UV concept to characterize and the model UV/H₂O₂ process in natural waters. Environ Sci Technol, 41 (7) (2007), pp. 2548-2553.

[10]	B.A. Wols, D.J.H. Harmsen, J. Wanders-Dijk, E.F. Beerendonk, C.H.M. Hofman-Caris. Degradation of pharmaceuticals in UV (LP)/H₂O₂ reactors simulated by means of kinetic modeling and computational fluid dynamics (CFD). Water Res, 75 (2015), pp. 11-24.

[11]	P.C. Xie, J. Ma, W. Liu, J. Zou, S.Y. Yue, X.C. Li, et al. Removal of 2-MIB and geosmin using UV/persulfate: contributions of hydroxyl and sulfate radicals. Water Res, 69 (2015), pp. 223-233.

[12]	J.Y. Fang, Y. Fu, C.I. Shang. The roles of reactive species in micropollutant degradation in the UV/free chlorine system. Environ Sci Technol, 48 (3) (2014), pp. 1859-1868.

[13]	Y. Lei, S.S. Cheng, N. Luo, X. Yang, T.C. An. Rate constants and mechanisms of the reactions of Cl^· dot and Cl₂^·- with trace organic contaminants. Environ Sci Technol, 53 (19) (2019), pp. 11170-11182.

[14]	S. Sudhakaran, G.L. Amy. QSAR models for oxidation of organic micropollutants in water based on ozone and hydroxyl radical rate constants and their chemical classification. Water Res, 47 (3) (2013), pp. 1111-1122.

[15]	Y.Y. Huang, Z.M. Qiang, Z. Sun, M.K. Li. Micropollutant degradation by UV/H₂O₂ in drinking water: facilitated prediction through combination of model simulation and portable measurement. Water Res, 221 (2022), Article 118794.

[16]	S. Sundararaman, J. Aravind Kumar, P. Deivasigamani, Y. Devarajan. Emerging pharma residue contaminants: occurrence, monitoring, risk and fate assessment—a challenge to water resource management. Sci Total Environ, 825 (2022), Article 153897.

[17]	C.C. Wang, Y.L. Lu, C. Wang, C. Xiu, X.H. Cao, M. Zhang, et al. Distribution and ecological risks of pharmaceuticals and personal care products with different anthropogenic stresses in a coastal watershed of China. Chemosphere, 303 (2022), Article 135176.

[18]	M.K. Li, W.T. Li, D. Wen, J.R. Bolton, E.R. Blatchley III, Z.M. Qiang. Micropollutant degradation by the UV/H₂O₂ process: kinetic comparison among various radiation sources. Environ Sci Technol, 53 (9) (2019), pp. 5241-5248.

[19]	W.H. Glaze, Y. Lay, J.W. Kang. Advanced oxidation processes—a kinetic model for the oxidation of 1,2-dibromo-3-chloropropane in water by the combination of hydrogen peroxide and UV radiation. Ind Eng Chem Res, 34 (7) (1995), pp. 2314-2323.

[20]	J.C. Crittenden, S.M. Hu, D.W. Hand, S.A. Green. A kinetic model for H₂O₂/UV process in a completely mixed batch reactor. Water Res, 33 (10) (1999), pp. 2315-2328.

[21]	X. Luo, X.H. Yang, X.L. Qiao, Y. Wang, J.W. Chen, X.X. Wei, et al. Development of a QSAR model for predicting aqueous reaction rate constants of organic chemicals with hydroxyl radicals. Environ Sci Process Impacts, 19 (3) (2017), pp. 350-356.

[22]	R.Y. Xiao, T.T. Ye, Z.S. Wei, S. Luo, Z.H. Yang, R. Spinney. Quantitative structure-activity relationship (QSAR) for the oxidation of trace organic contaminants by sulfate radical. Environ Sci Technol, 49 (22) (2015), pp. 13394-13402.

[23]	M.J. Watts, K.G. Linden. Chlorine photolysis and subsequent ·OH radical production during UV treatment of chlorinated water. Water Res, 41 (13) (2007), pp. 2871-2878.

[24]	Y. Yang, J. Jiang, X.L. Lu, J. Ma, Y.Z. Liu. Production of sulfate radical and hydroxyl radical by reaction of ozone with peroxymonosulfate: a novel advanced oxidation process. Environ Sci Technol, 49 (12) (2015), pp. 7330-7339.

[25]	Z.C. Hua, X.J. Kong, S.D. Hou, S.Q. Zou, X.B. Xu, H. Huang, et al. DBP alteration from NOM and model compounds after UV/persulfate treatment with post chlorination. Water Res, 158 (2019), pp. 237-245.

[26]	X. Yang, J.L. Sun, W.J. Fu, C.I. Shang, Y. Li, Y.W. Chen, et al. PPCP degradation by UV/chlorine treatment and its impact on DBP formation potential in real waters. Water Res, 98 (2016), pp. 309-318.

[27]	P. Neta, V. Madhavan, H. Zemel, R.W. Fessenden. Rate constants and mechanism of reaction of SO₄^·- with aromatic-compounds. J Am Chem Soc, 99 (1) (1977), pp. 163-164.

[28]	G.V. Buxton, M. Bydder, G.A. Salmon, J.E. Williams. The reactivity of chlorine atoms in aqueous solution. Part III. The reactions of Cl^·- with solutes. Phys Chem Chem Phys, 2 (2) (2000), pp. 237-245.

[29]	National Institute of Standards and Technology (NIST). NDRL/NIST Solution Kinetcis Database on the [Internet]. c 2023 [cited 2023 Mar 30]. Available online: https://kinetics.nist.gov/solution/.

[30]	A.L. Boreen, W.A. Arnold, K. McNeill. Triplet-sensitized photodegradation of sulfa drugs containing six-membered heterocyclic groups: identification of an SO₂ extrusion photoproduct. Environ Sci Technol, 39 (10) (2005), pp. 3630-3638.

[31]	S.P. Mezyk, T.J. Neubauer, W.J. Cooper, J.R. Peller. Free- radical-induced oxidative and reductive degradation of sulfa drugs in water: absolute kinetics and efficiencies of hydroxyl radical and hydrated electron reactions. J Phys Chem A, 111 (37) (2007), pp. 9019-9024.

[32]	G.V. Buxton, C.L. Greenstock, W.P. Helman, A.B. Ross. Critical review of rate constants for reactions of hydrated electrons, hydrogen atoms and hydroxyl radicals (^·OH/^·O^-) in aqueous solution. J Phys Chem Ref Data, 17 (2) (1988), pp. 513-886.

[33]	M. Nihemaiti, D.B. Miklos, U. Huebner, K.G. Linden, J.E. Drewes, J.P. Croué. Removal of trace organic chemicals in wastewater effluent by UV/H₂O₂ and UV/PDS. Water Res, 145 (2018), pp. 487-497.

[34]	L.S. Lian, B. Yao, S.D. Hou, J.Y. Fang, S.W. Yan, W.H. Song. Kinetic study of hydroxyl and sulfate radical-mediated oxidation of pharmaceuticals in wastewater effluents. Environ Sci Technol, 51 (5) (2017), pp. 2954-2962.

[35]	S.Q. Zhou, W.Q. Zhang, J.L. Sun, S.M. Zhu, K. Li, X.Y. Meng, et al. Oxidation mechanisms of the UV/free chlorine process: kinetic modeling and quantitative structure activity relationships. Environ Sci Technol, 53 (8) (2019), pp. 4335-4345.

[36]	W. Li, H.G. Guo, C.J. Wang, Y.L. Zhang, X. Cheng, J.Q. Wang, et al. ROS reevaluation for degradation of 4-chloro-3,5-dimethylphenol (PCMX) by UV and UV/persulfate processes in the water: kinetics, mechanism, DFT studies and toxicity evolution. Chem Eng J, 390 (2020), Article 124610.

[37]	K.H. Guo, Z.H. Wu, C.I. Shang, B. Yao, S.D. Hou, X. Yang, et al. Radical chemistry and structural relationships of PPCP degradation by UV/chlorine treatment in simulated drinking water. Environ Sci Technol, 51 (18) (2017), pp. 10431-10439.

RIGHTS & PERMISSIONS

THE AUTHOR