Oxytetracycline Degradation by Heat-Activated Peroxydisulfate and Peroxymonosulfate 1 Oxidation: Optimization by Box-Behnken Design

This study investigates the removal of oxytetracycline performed by persulfate (PS) and 21 peroxymonosulfate (PMS) processes. For this purpose, response surface methodology was used to 22 examine the effectiveness of PS and PMS processes under both alkaline and thermal conditions. 23 The effect of four independent variables, which were selected as pH, PS/PMS concentration, 24 temperature, and time, were analyzed in a wide range. The working pH was between pH 3–11 to 25 compare acidic and alkaline conditions, and the temperature was selected between 30 and 90 °C 26 to evaluate the effect of thermal activation of sulfate radicals for both processes. According to the 27 results of prepared Response Surface Methodology (RSM) models, all four independent variables 28 were determined to be highly significant for both PS and PMS. Especially in the PMS process, the 29 highest PMS concentration was observed to complete degradation of OTC. The conditions for the 30 highest removal were pH 9 and PS/PMS concentration was approximately 4mM for both 31 processes, while the temperature and required time were 72.9°C and 75°C, 26.5 min and 20 min, 32 for PS and PMS processes, respectively. PS process has higher kinetic constants at all pH values 33 than the PMS process.

(lakes, rivers, etc.), remains in nature for a long time due to its persistence and poses a threat to 48 the deterioration of the ecosystem by triggering the formation of antibiotic-resistant bacteria. 49 Therefore, OTC was chosen as the target pollutants to be degraded by persulfate and 50 peroxymonosulfate processes. 51 Due to the ineffectiveness of traditional biological treatment methods, especially advanced 52 oxidation techniques are among the issues that need to be investigated (Cuerda-Correa et al., 2020). 53 Today, many advanced oxidation processes are successfully applied in many treatment facilities 54 As can be understood from Eq1 and Eq.2, only sulfate radicals are formed in PS activation, while 75 by using PMS, sulfate radical (SO4•) and hydroxyl radicals (•OH) can be formed. However, the 76 reaction of water and SO4• radicals to form •OH radicals is too slow to be significant in most 77 processes, it is stated in the literature that SO4• radicals transform more quickly into •OH radicals 78 with the effect of temperature and pH (Eq.3). 79 4 • + 2 → 4 2− + • + + (Eq.3) 80 It has been encountered in some studies that SO4• radicals are effective when pH is acidic, and 81 hydroxyl radicals are seen in PS activation when pH is alkaline. Besides, PS can be activated only 82 by adjusting the pH to alkaline conditions. In this mechanism, the perhydroxyl radicals formed to 83 play an active role in the formation of sulfate radicals (Eq.4 and Eq.5), hydroxyl radicals (Eq. 6), 84 and superoxides (Eq. 5) (Wang and Wang, 2018 Response surface methodology (RSM) is a set of statistical methods used to find the optimum 98 point based on independent variables. RSM is an experimental design that is widely used, and it 99 both minimizes the amount of waste that can be generated and provides an optimized result faster. 100 In light of this information, the effects of PS and PMS on oxytetracycline degradation, activated 101 by the use of both heat and initial pH, were compared in this study. PS and PMS processes were 102 evaluated with a 4-factor 5-level central composite design, in which PS/PMS concentration, pH, 103 temperature, and time were considered as independent variables. According to the optimized 104 results, the intermediate products were evaluated via UV-vis spectrum. This study is the first study 105 that compares the PS and PMS processes for OTC degradation. 106

Chemicals and Reagents 108
The chemicals used in this study are in lab quality. Oxytetracycline (C22H24N2O9) was purchased 109 from Sigma Aldrich (Turkey). Sulfuric acid (H2SO4, purity 98%) and sodium hydroxide (NaOH, 110 purity ≥97.0%) were obtained from Merck, Turkey, and both were applied in 0.1N and 0.1N 111 concentration for pH adjustment. 112

Experimental studies and reactor design 113
The experiments were conducted in 50 mL Erlenmeyer flasks placed on a magnetic stirrer with a 114 heater (Weightech Instruments, WH-H320, Turkey). The flasks were closed and the temperature 115 of the samples was measured regularly by a temperature probe during experiments. The pH of each 116 sample was measured before and after the experiments by WTW Multi 9620 IDS, Turkey. The 117 initial concentration of OTC was 10 mg/L and measured at 355 nm wavelength by UV-visible 118 spectrophotometer (WTW-Photolab 6600 UV-vis, Turkey). 119

Response Surface Methodology 120
Response surface methodology was used to optimize the study both with PS and PMS experiments. 121 Four factor-five level design was applied by performing the independent variables, initial pH (A), 122 PS/PMS concentration (B), initial temperature (C), and time (D) on OTC degradation (y). The 123 levels were coded according to the α value of ±2 which can be calculated via Eq.7. Actual and 124 coded values of independent variables can be seen in Table 2. 125 In Eq.7, α is code for independent variables, xi is the actual value of xi variable, xo is the actual 126 value in the med-point, and x shows the changes in xi variable. Experimental data were analyzed 127 with the quadratic equation as given in Eq. 8. 128 show the coefficients of the linear, quadratic, and interaction, xi and xj shows the independent 130 variables. In this study, the response is selected as OTC degradation efficiency. According to 131 selected levels and the number of independent variables, the number of runs in the central 132 composite design (CCD) were given in Table 3. 133 The OTC degradation efficiencies were calculated according to Eq. (9) 135 where C0 and Ct are the initial and effluent OTC concentrations after a treatment time of t, 136 respectively. 137 Before determining the levels given in the matrix in Table 2, pre-studies were done, according to 141 the studies conducted in the literature (Table 1), which are mostly for PS activation. Thermal and 142 alkali activation of PS and PMS were studied for the points given in Table 4, and it was observed 143 that the effect of pH and temperature were significant. Therefore, the level of pH was selected, 144 starting from pH 3 to pH11 and the levels for temperature was ranged from 30 -90°C. Since high 145 degradation efficiencies were achieved under high PMS concentration and high required time, 146 lower values of both parameters were chosen as levels ( Table 2). 147 According to the central composite design, the runs were designed in a batch mode as given in 149 Table 3. The response of CCD was selected as OTC degradation. The PS and PMS processes were 150 evaluated separately, and compared based on degradation, active species, and intermediate 151 products. 152

Model Accuracy 153
All runs given in Table 3 were studied with the PS process by considering the concentrations of B 154 parameter as PS concentrations (Table 2). Regarding the obtained responses (Table 3), the 155 quadratic model was used as the most suitable model to explain the relationship between the 156 experienced and the values of OTC degradation estimated using ANOVA results in the PS process. 157 The results of the statistical analysis of ANOVA are shown in detail in Table 5. The results 158 obtained in the studies with PS show that the oxidation performance gives satisfactory results. By 159 considering the regression coefficients, the reliability between the actual data and the predicted 160 in Table 5 to better understand the optimization performance of the study. The regression 162 coefficient (R 2 ) was determined as above 90%, which shows the model performance is good. The 163 fit of the model appears to be in good accordance with the R 2 with adjusted R 2 values. Besides, 164 there is a slight difference between the experimental and predicted values shown in Table 3, 165 resulting in a low coefficient of variation value (C.V. value). The accuracy and repeatability of the 166 experiments were confirmed by this low C.V. value (C.V. <10%) (Can-Guven et al., 2020; Varank, 167 2020). The adequate precision value is required to be greater than 4 to measure the predicted 168 response and the corresponding error value (the signal-to-noise ratio) (Rahdar et al., 2020). In this 169 study, it was found to be over 4 ( Table 5). The fitness of the model can be considered as desired. 170 The non-significant lack of fit values can be submitted as the best fit (Varank, 2020). 171 confidence interval. Parameters with a value of P less than 0.05 were considered as significant and 175 those with a value lower than 0.0001 were considered as highly significant. It can be seen from 176 Table 5 that the linear terms of the selected independent variables were significant, but quadratic 177 forms were not so effective. The equation obtained based on these data is obtained as given in 178 Eq.10. The regression coefficient of the model was obtained as 0.89 and C.V was less than 10%. Besides, 193 as mentioned above, the adequate precision value was used to measure the signal-to-noise ratio. A 194 higher value than 4 shows the accuracy of the model. P-values of the parameters can also be seen 195 in The Pareto chart also illustrates the influence of independent variables on OTC degradation using 201 Eq. 12, where βi represents the coefficient of the independent variables (Majumdar and Pal, 2020). 202 = ( 2 ∑ 2 ) . 100( ≠ 0) (Eq.12) Pareto chart was prepared for PS and PSM processes and showed in Figure 1a and Figure 1b, 203 respectively. The independent parameters which are effective more than 1% are considered for concentration (B) and followed by temperature (C), pH (A), and time (D). The P values of the 206 interaction and quadratic terms of the models were higher than 0.05 which shoes the low/non-207 impact on the OTC degradation. The percentage of the effects of these terms is less than 4%. 208 a.  be evaluated within the ranges given in Table 1. The effect of independent variables can be 219 observed via the 3D plot generated via CCD. These plots were prepared for PS and PMS processes 220 separately and the comparison of the processes was established accordingly. As can be seen from 221 the surface graphs given in Figure 3, the OTC degradation efficiencies were increasing with the 222 increment on all independent variables. This was also gathered from ANOVA results. The higher 223 degradation efficiencies were expected with the increase of PS concentrations since it is the most 224 important variable of the process (Figure 3.a)(Can-Guven et al., 2020). According to Figure 3.a, 225 when the PS concentration was higher than 3mM, and pH is higher than pH7, the degradation 226 efficiencies can be obtained at a higher value than 80%. As per Figure 3.b and Figure 3.c, with 227 higher values than 60°C for temperature, 30 min, and pH7, the degradation efficiencies are higher 228 than 80%. This can also be understood from the intercept of Eq.9, which is 80. All 3D figures were 229 illustrated for OTC degradation efficiencies versus 2 independent variables, while the other two 230 parameters were selected at the center point. The highest degradation efficiency can be seen in 231  variables on OTC degradation. Considering that the increase in temperature has a positive effect 235 on both increasing the efficiency of sulfate radicals and degradation of OTC, a value at high 236 temperature will be optimum as expected. Likewise, the effect of the pH factor is an effective 237 parameter on PS processes. PS can be activated at low pH values but also at alkaline conditions. 238 Therefore, the effect of pH was screened in a very broad spectrum (pH3 -pH 11). According to 239 Figure 3.a, 3.b, and 3.c, OTC degradation obtained at alkaline conditions. 240 d.
e. f. which is less than other parameters, can be seen from Figure 4.c, 4.e, and 4.f and also from P-251 Value from Table 5. The graph with the highest removal efficiency (approximately 98%) was 252 attained in Figure 4.d, and it was obtained in the case where the PMS concentration was the highest. 253 Also, from Figure 4.d, higher OTC degradation with the increment of temperature can be caused 254 by the formation of hydroxyl radicals at higher temperatures. To have a higher degradation than 255 80%, all parameters should be in a higher alpha value than 0.9, which means that the conditions 256 higher than pH9, 4 mM PMS concentration, 75°C, and 40 min of reaction time. However, this 257 situation is different for the case where the PMS concentration is 5mM at the highest point, the 258 effectiveness of all other parameters is low (Figure 4.a, 4.d, and 4.e). Especially, the time parameter 259 is not affecting, when PMS concentration is 5mM; the OTC degradation efficiency is 260 approximately 90% regardless of time (Figure 4.e). This shows that the effectiveness of PMS can 261 be seen at the highest concentration. 262

Optimization of PS and PMS Processes 263
Eq. 10 and Eq. 11 were optimized to understand the conditions under which the highest removal 264 efficiency is at the levels of the experimental matrix given in Table 2. Both processes were 265 optimized taken into account all parameters, and it was obtained that the degradation efficiencies 266 that could be obtained for PS and PMS processes were 89.7% and 84.0%, respectively (Table 7). 267 Accordingly, the optimum pH is approximately 9, and the optimum temperature was obtained as 268 72.9 and 75.0°C for PS and PMS processes, respectively. It has been gathered that while 26.5 min 269 of reaction time is needed for the PS process, 20 min is sufficient for PMS. Figure 2.e shows that 270 the degradation efficiency is the same regardless of the time when the PS/PMS is 4mM. 271 Nevertheless, a low reaction time is required, the efficiency of the PMS process was lower than 272 the PS process. Ji et al. (2016) emphasized in their study that 240 min was needed to achieve 100% 273 removal in the persulfate process they activated with heat (Table 1) Where C0 and C were the initial and effluent concentration (mg/L) at "t" time (min) and kt is the 291 pseudo-first-order constant (min -1 ). Exponential function versus time was prepared at different pH 292 values to calculate kt, and the results can be seen in Figure 5. Accordingly, the optimum pH can 293 also be obtained from kt as pH 9. After pH 9, while the kt values for PS were reduced, it stayed the 294 same for the PMS process. The form of OTC 2which exits on pH11 does not react well with PS 295 activation (Liu et al., 2016b). Suggested that the active species on the PS process is known that 296 mostly SO4• radicals existed under alkaline conditions; the reaction of SO4• radicals and OHcan 297 decrease the effectiveness of SO4• radicals which causes low OTC degradation efficiencies (Eq.6). 298 When kt values at acidic and alkali conditions were compared, the PS process seems to be more 299 effective. Since it is known that only SO4• radicals exit at the PS process, it can be concluded that 300 both SO4• radicals are more effective than OH• radicals in both PS and PMS processes. Where CI is the confidence interval, ̅ is the sample mean, z is the value of confidence level value, 308 s is the sample standard deviation and n is the number of samples. The confidence level was 309 selected as 95% for this study, thus, z was used as 1.96. CI was determined as ± 4.2 and ± 4.6 for 310 respectively, which are acceptable for the prepared model.   Table 7. The spectrums of the samples before 317 and after treatment by PS and PMS processes were given in Figure 6.a and Figure 6.b, respectively. 318 Two wide peaks were observed at the initial OTC samples. While the wide peak between 320 nm 319 and 400 nm disappeared in the effluent samples of both processes, the peak at 265 nm seen in the 320 initial sample was shifted to 255 nm in the effluent sample of the PMS process ( Fig.6.b) but was 321 not observed in the PS process ( Fig.6.a). Since OTC is a naphthol ring, naphthalene is an expected 322 by-product as a result of degradation. The presence of naphthalene has been stated in the literature 323 that the OTC compound will be formed by breaking the stable naphthol ring (Kumar Subramani 324 et al., 2019). The wide peak seen at 255nm can be seen as the transformation of naphthene as a 325 result of degradation. Looking at the absorbance values around 220 nm, an increase in height is 326 seen as a result of both processes. This case shows simple acidic compounds with -C=O-bonds, 327 such as acetic acid. This can be stated as the last step (by-product) of degradation. The issue of 328 determining the degradation by-products and complete mineralization will be investigated in 329 further studies. 330

Conclusion 331
This study was conducted to compare PS and PMS processes on OTC degradation by optimization 332 with central composite design. The independent variables were selected as initial pH, PS/PMS 333 concentration, temperature, and time for both processes in which PS/PMS concentration was 334 determined as the most important variables within them. For both processes, the importance of the 335 independent variables was determined as PS/PMS concentration > temperature > initial pH > time. 336 According to CCD models, only the linear terms were concluded as the effective independent 337 variables for both processes. Kinetic constants for pseudo-first-order for different pH values 338 calculated at highest PS/PMS concentrations showed that PMS has slightly higher k constants, 339 meaning better in OTC degradation. Kinetic constants were attained as 0.068 and 0.071 min -1 for 340 PS and PMS processes, respectively. According to optimum conditions, optimum pH is pH9, 341 PS/PMS concentration is approximately 4 mM, the temperature is 72.9°C and 75°C and the 342 required time was 26.5 min and 20 min for PS and PMS processes, respectively. Besides, according 343 to the UV-vis spectrum, while the peak for OTC decreased for both processes, peak height at 225 344 nm showed that compounds with -C=O-bonds were increased at both PS and PMS processes All processes can be used in PS/PMS activation to lower the required PS/PMS concentration. 347 Funding -Not applicable 353

Availability of data and materials 354
The datasets obtained during this study are available from the corresponding author on reasonable 355 request. 356

Author contributions 357
All authors are contributed to designing the study. SYG, ECG, and FI are contributed to the whole 358 experiments, KUA and GV helped partially on experiments. KUA, SYG, and GV were contributed 359 to analyzing response surface methodology. KUA was the major contributor in writing the 360 manuscript, while all authors were contributed to writing, revising, and finalizing the manuscript. 361 All authors approved the final manuscript.