Photocatalytic Performance and Degradation Mechanism of Aspirin by TiO 2 through Response Surface Methodology

: In the present work, the photocatalytic performance of P25TiO 2 was investigated by means of the degradation of aspirin, while the reaction system was systematically optimized by central composite design (CCD) based on the response surface methodology (RSM). In addition, three variables of initial pH value, initial aspirin concentration and P25 concentration were selected to assess the dependence of degradation efﬁciencies of aspirin. Meanwhile, a predicted model of degradation efﬁciency was estimated and checked using analysis of variance (ANOVA). The results indicated that the PC removal of aspirin by P25 was signiﬁcantly inﬂuenced by all these variables in descending order as follows: P25 concentration > initial aspirin concentration > initial pH value. Moreover, the parameters were optimized by the CCD method. Under the conditions of an initial pH value of 5, initial aspirin concentration of 10 mg/L and P25 concentration of 50 mg/L, the degradation efﬁciency of aspirin was 98.9%with 60 min of Xenon lamp irradiation. Besides, based on the liquid chromatography-mass spectrometry measurements, two main PC degradation pathways of aspirin by TiO 2 were deduced and the tentative degradation mechanism was also proposed.


Introduction
Among the large number of various organic pollutants that may enter into water resources, pharmaceuticals and personal care products (PPCPs) in surface and ground waters have been a major environmental concern [1,2]. Aspirin, known as acetylsalicylic acid (ASA), is commonly used as a classic non-steroidal anti-inflammatory drug (NSAID) to relieve minor aches and treat inflammation and cardiovascular diseases [3,4]. However, this type of pharmaceutical has been detected in the aquatic environment at trace concentrations (ng to µg/L), which can prolong exposure to environmentally relevant concentration and considerably increase the toxicity considerably, resulting in negative consequences for aquatic life and human beings [5][6][7]. Thus, it is indeed necessary to explore the way to remove these PPCPs, such as aspirin, from aqueous environments.
In recent decades, various wastewater treatments, such as anaerobic processes and activated sludge, have been employed for the degradation of organics, which are not so effective when treating the environment issue of PPCPs [8]. Therefore, researchers have found that advanced oxidation processes (AOPs) can eliminate the refractory contaminants (including emerging contaminants and PPCPs) in aqueous environments, which deserves major attention [9][10][11][12][13][14]. Moreover, one of these technologies is solar photocatalysis (PC), which can degrade organic pollutants by means of the photocatalytic reaction promoted by the action of light on the surface of a semiconductor acting as a photocatalyst [15]. Among various photocatalysts, TiO 2 has been widely studied since Honda pioneered the water splitting of TiO 2 photoanodes in 1972 [16], which displays strong chemical stability, non-toxicity, low cost and several unique characteristics. Thus, this has been proven to be a promising technology for the decontamination of these refractory pollutants [17,18]. Li et al. reported the C-N-S-TiO 2 electrode that was decorated with Pd exhibited enhanced PC degradation of aspirin [19]. However, during the conventional studies, the optimization of degradation efficiency of aspirin is usually performed using a single variable method, which is time consuming and cost expensive. Furthermore, it is difficult to predict the proper optimum and interactions between different variables [20]. As demonstrated, statistical programs have been invented for carrying out a large number of experiments in a shorter period of time and more efficiently to obtain optimum conditions, which can overcome these aforementioned defects rationally [21,22]. Response surface methodology (RSM), as a statistical method, is used to generate a model, analyze the effects of operating parameters and optimize the process conditions [23]. Furthermore, the central composited design (CCD) is a standard RSM design to analyze the relationship between the several variables and one or more response variables, which can also create a second-order response surface model in the optimization of processes [22,24].
Accordingly, in the present study, the photocatalytic degradation performance and TOC measurement of aspirin by TiO 2 (P25) under simulated solar irradiation was studied. Additionally, various influencing parameters, such as initial pH value, initial aspirin concentration and P25 concentration, were employed to assess the effect of individual variables and interaction of several variables on PC removal efficiency through the CCD method based on RSM. Besides, PC degradation intermediates and pathways of aspirin by TiO 2 were proposed based on the high-performance-liquid chromatography-mass spectrometry/mass spectrometry (HPLC-MS/MS) and ions chromatography (IC) measurements. In addition, the PC degradation mechanism by TiO 2 was also deduced.

PC Removal Degradation of Aspirin
Aspirin, as a common non-steroidal anti-inflammatory (NSAIDs) drug, has been widely used to reduce fever and relieve minor pains and aches. However, this type of pharmaceutical has already been detected in the effluents of sewage and might pose a potential risk to humans. Thus, this was employed to be the target pollutant to evaluate the PC performance of the P25 in this study ( Figure 1). Furthermore, the absorbance of the sample in the dark and the photolysis of the aspirin solution were also measured for references. As seen from Figure 1, less than 5% of aspirin could self-decompose when the solution was directly irradiated by the Xenon lamp without the addition of P25, indicating that aspirin was relatively stabilized under Xenon lamp illumination. Besides, when the catalysis existed with the lamp turned off, only 9% of aspirin was removed, which corresponded to the absorbance capacity of the P25 nano-catalyst. We noted that P25 resulted in higher degradation efficiency of aspirin, with68.5% of aspirin removed within 60 min of Xenon lamp illumination. This implied that TiO 2 had superior PC performance, which is consistent with other studies [25][26][27]. In addition, Figure 1 also showed TOC results of aspirin mineralization through PC process. As seen from Figure 1, the high PC mineralization rate of 60.9% was achieved under 60 min of irradiation, implying that P25 had high PC capacity for mineralization of aspirin, which leads to the formation of CO 2 and H 2 O molecules. that P25 had high PC capacity for mineralization of aspirin, which leads to the formation of CO2 and H2O molecules.

Multivariable Experimental Design
Commonly, the PC degradation efficiency of aspirin could be simultaneously influenced by several variables, such as initial aspirin concentration, initial pH and the content of photocatalyst [28,29]. However, it was difficult to find interacting effects among these independent variables from the single-variable optimization experiments, which has been reported by many studies. Hence, the CCD experiment based on the RSM method was aiming to overcome this problem. Following the method shown in Section 3.2, 17 experiments were conducted and the experimental data in CCD for PC degradation efficiency of aspirin are displayed in Table 1. In addition, the experimental conditions for response factor (Y) corresponded to the photocatalytic degradation efficiencies of aspirin (%) after 60 min of irradiation.
Generally, according to the experimental dates, a semi-empirical second-order polynomial equation consisting of 10 statistically significant coefficients was estimated by computer simulation programming, which applied the least square method using Design-Expert 8.0 as follows: where Y was the predicted response; and X1, X2 and X3 corresponded to initial pH value, initial aspirin concentration and P25 concentration, respectively. Besides, the coefficients in the polynomial equation, which included linear coefficients, cross product coefficients and quadratic coefficients, represented the weight of variables and the interaction between them [30]. As calculated from the model of Equation (1), the predicted values for degradation of aspirin by P25 are also listed in Table 1. Meanwhile, the correction between experimental values of degradation rate of aspirin and the predicted values are analyzed and shown in Figure 2. Noticeably, a high correlation coefficient (R 2 ) of 0.8506 was obtained, which indicates consistency between experimental and predicted values and suggests that the model had a good linear fitting and explained the studied experimental range very well.

Multivariable Experimental Design
Commonly, the PC degradation efficiency of aspirin could be simultaneously influenced by several variables, such as initial aspirin concentration, initial pH and the content of photocatalyst [28,29]. However, it was difficult to find interacting effects among these independent variables from the single-variable optimization experiments, which has been reported by many studies. Hence, the CCD experiment based on the RSM method was aiming to overcome this problem. Following the method shown in Section 3.2, 17 experiments were conducted and the experimental data in CCD for PC degradation efficiency of aspirin are displayed in Table 1. In addition, the experimental conditions for response factor (Y) corresponded to the photocatalytic degradation efficiencies of aspirin (%) after 60 min of irradiation. Generally, according to the experimental dates, a semi-empirical second-order polynomial equation consisting of 10 statistically significant coefficients was estimated by computer simulation programming, which applied the least square method using Design-Expert 8.0 as follows: where Y was the predicted response; and X 1 , X 2 and X 3 corresponded to initial pH value, initial aspirin concentration and P25 concentration, respectively. Besides, the coefficients in the polynomial equation, which included linear coefficients, cross product coefficients and quadratic coefficients, represented the weight of variables and the interaction between them [30]. As calculated from the model of Equation (1), the predicted values for degradation of aspirin by P25 are also listed in Table 1.
Meanwhile, the correction between experimental values of degradation rate of aspirin and the predicted values are analyzed and shown in Figure 2. Noticeably, a high correlation coefficient (R 2 ) of 0.8506 was obtained, which indicates consistency between experimental and predicted values and suggests that the model had a good linear fitting and explained the studied experimental range very well.  Furthermore, the model was checked using the analysis of variance (ANOVA), which could also analyze the statistical significance and effects of regression coefficients. In the CCD study, A p-value (significant probability value) lower than 0.05 indicated that the model terms were significant, while values greater than 0.1 indicates no significance [31]. As seen from Table 2, the variables of initial pH value (X1), initial aspirin concentration (X2) and P25 concentration (X3) were all highly significant as the p-values were all below 0.05. Thus, the response factor (degradation rate of aspirin) were indeed influenced by these three variables. In addition, the F-value was used to check the statistical significance with a higher value representing a larger influence [31]. This suggested that the significance of variables fitted the following order: X3 > X2 > X1, which is consistent with the weight of variables shown in the predicted model (Equation (1)). Furthermore, the ANOVA results also displayed that the predicted model (p-value < 0.05) was reliable for simulating and optimizing the PC decomposition of aspirin by P25.
Therefore, the graphical representations of the regression equation and the information on the primary impact of three factors were evaluated by the 3D response surface plots and the twodimension contours, which are shown in Figure 3. For aspirin removal, Figure 3a,b showed the Furthermore, the model was checked using the analysis of variance (ANOVA), which could also analyze the statistical significance and effects of regression coefficients. In the CCD study, A p-value (significant probability value) lower than 0.05 indicated that the model terms were significant, while values greater than 0.1 indicates no significance [31]. As seen from Table 2, the variables of initial pH value (X 1 ), initial aspirin concentration (X 2 ) and P25 concentration (X 3 ) were all highly significant as the p-values were all below 0.05. Thus, the response factor (degradation rate of aspirin) were indeed influenced by these three variables. In addition, the F-value was used to check the statistical significance with a higher value representing a larger influence [31]. This suggested that the significance of variables fitted the following order: X 3 > X 2 > X 1 , which is consistent with the weight of variables shown in the predicted model (Equation (1)). Furthermore, the ANOVA results also displayed that the predicted model (p-value < 0.05) was reliable for simulating and optimizing the PC decomposition of aspirin by P25. Therefore, the graphical representations of the regression equation and the information on the primary impact of three factors were evaluated by the 3D response surface plots and the two-dimension contours, which are shown in Figure 3. For aspirin removal, Figure 3a,b showed the combined effect of initial pH value and initial aspirin concentration with the constant concentration of P25 (50 mg/L). When P25 dosage was at a suitable concentration, the degradation rate increased with a decrease in initial pH as well as aspirin concentration. The most efficient removal rate was achieved at the initial aspirin concentration of 10 mg/L with a pH value of 5. As seen from Figure 3c,d, two graphs were developed for the aspirin removal efficiency with varying initial pH value and P25 concentration at a fixed initial aspirin concentration of 10 mg/L. It was clearly seen that the higher PC degradation rate of aspirin was obtained when the P25 concentration reached a maximum of 50 mg/L at a relatively lower pH value of 5, which was consistent with the former result. Subsequently, the relationship between the initial aspirin and P25 concentration affecting the degradation rate was carried out with the initial pH value fixed at 5 and displayed in Figure 3e,f. When the P25 concentration increased from 20 mg/L to 50 mg/L, the PC removal rate of aspirin rose sharply, indicating that P25 concentration had a huge influence on PC activity. Furthermore, the initial aspirin concentration also has an effect, with an optimized dosage of 10 mg/L for the maximum degradation efficiency.
Hence, the experiment in optimum process conditions through RSM based on CCD for the highest desirability was carried out, with the results shown in Figure 4. Obviously, the experimental degradation efficiency of aspirin was 98.9% at the optimized conditions of an initial pH value of 5, initial aspirin concentration of 10 mg/L and P25 concentration of 50 mg/L, respectively. In addition, the software optimized degradation efficiency was 94.1%, which was much closer to the experimental value. This suggests that the predicted model was able to optimize the parameters for PC degradation of aspirin by P25, which was consistent with ANOVA results.  Hence, the experiment in optimum process conditions through RSM based on CCD for the highest desirability was carried out, with the results shown in Figure 4. Obviously, the experimental degradation efficiency of aspirin was 98.9% at the optimized conditions of an initial pH value of 5, initial aspirin concentration of 10 mg/L and P25 concentration of 50 mg/L, respectively. In addition, the software optimized degradation efficiency was 94.1%, which was much closer to the experimental value. This suggests that the predicted model was able to optimize the parameters for PC degradation of aspirin by P25, which was consistent with ANOVA results. . 3D response surface plots (a,c,e) and two-dimension contour plots (b,d,f) of interaction between initial pH value (X 1 ), initial aspirin concentration (X 2 ) and P25 concentration (X 3 ) for PC degradation of aspirin by P25, where effect of initial pH value and initial aspirin concentration (a,b); effect of initial pH value and P25 concentration (c,d); effect of initial aspirin concentration and P25 concentration (e,f).

Analysis of Degradation Intermediates
As demonstrated already, the advanced oxidization process (AOP) was widely used in wastewater treatment of pharmaceutical and personal care products (PPCPs) due to being highly active and nonselective between OH radicals and target pollutants [32,33]. However, it might cause the aggregation of the toxicity and recalcitrance of parent compounds, which can cause harm to environment and human nature [34]. Thus, HPLC-MS/MS measurements were carried out to elucidate the intermediates and mechanism of PC decomposition of aspirin by P25 [34]. As displayed in Figure S1, LC-ESI/MS chromatography showed that aspirin was degraded by P25 nano-catalyst within 60 min of Xenon lamp illumination. Thus, we can conclude that several intermediates were formed. The LC-MS/MS spectra of corresponding LC-MS is displayed in Figure 5. Additionally, the structural assignments of all detected intermediates were studied, which are shown in Table 3. Noticeably, the degradation intermediates were detected by comparison with the target pollutant (aspirin, m/z 179) during the PC process, indicating the decomposition and mineralization of aspirin molecules. This is consistent with the TOC results. Therefore, according to the HPLC-MS/MS analyses, nine possible intermediates were identified and figured out, which was in accordance with the previous studies [35] and are listed in Table 3. Furthermore, the PC removal degradation diagram containing two main pathways for elimination of aspirin were proposed, which are plotted in Figure 6.
Compared with the aspirin solution (m/z, 179), there was an intermediate with m/z of 135 found, which might be ascribed to the decarboxylation (-CO2) that forms the compound of phenyl acetate ( Figure 6, pathway 1). Subsequently, the compound holding [M+H] + 93 was also detected, which was probably formed through the deacetylation (-COCH3) that reveals the formation of phenol. Furthermore, the addition of 16 mass units of phenol (m/z 93) could be assigned to the hydroxylation, resulting in two intermediates of o-dihydroxybenzene and p-dihydroxybenzene (m/z 109). Besides, we found that the intermediates with m/z values of 107 and 115 corresponded to the benzoquinone and butenedioic acid, respectively. In addition, the parent compound would be directly oxidized by the photogenerated holes (h + ), which consequently leads to the deacetylation (-COCH3) of aspirin. Therefore, this creates the formation of salicylic acid (m/z of 137), which was considered as the other degradation pathway (shown in Figure 6). Furthermore, the product (m/z 153) revealing the addition of 16 mass units compared with salicylic acid was detected, which was probably due to the addition D: Figure 4. Desirability ramp for numerical optimization of independent variables, initial pH value (X 1 ), initial aspirin concentration (X 2 ) and P25 concentration (X 3 ) for the PC degradation efficiency of aspirin by P25. A: initial pH value; B: initial aspirin concentration; C: P25 concentration; D: degradation efficiency

Analysis of Degradation Intermediates
As demonstrated already, the advanced oxidization process (AOP) was widely used in wastewater treatment of pharmaceutical and personal care products (PPCPs) due to being highly active and non-selective between OH radicals and target pollutants [32,33]. However, it might cause the aggregation of the toxicity and recalcitrance of parent compounds, which can cause harm to environment and human nature [34]. Thus, HPLC-MS/MS measurements were carried out to elucidate the intermediates and mechanism of PC decomposition of aspirin by P25 [34]. As displayed in Figure S1, LC-ESI/MS chromatography showed that aspirin was degraded by P25 nano-catalyst within 60 min of Xenon lamp illumination. Thus, we can conclude that several intermediates were formed. The LC-MS/MS spectra of corresponding LC-MS is displayed in Figure 5. Additionally, the structural assignments of all detected intermediates were studied, which are shown in Table 3. Noticeably, the degradation intermediates were detected by comparison with the target pollutant (aspirin, m/z 179) during the PC process, indicating the decomposition and mineralization of aspirin molecules. This is consistent with the TOC results. Therefore, according to the HPLC-MS/MS analyses, nine possible intermediates were identified and figured out, which was in accordance with the previous studies [35] and are listed in Table 3. Furthermore, the PC removal degradation diagram containing two main pathways for elimination of aspirin were proposed, which are plotted in Figure 6.
Compared with the aspirin solution (m/z, 179), there was an intermediate with m/z of 135 found, which might be ascribed to the decarboxylation (-CO 2 ) that forms the compound of phenyl acetate ( Figure 6, pathway 1). Subsequently, the compound holding [M+H] + 93 was also detected, which was probably formed through the deacetylation (-COCH 3 ) that reveals the formation of phenol. Furthermore, the addition of 16 mass units of phenol (m/z 93) could be assigned to the hydroxylation, resulting in two intermediates of o-dihydroxybenzene and p-dihydroxybenzene (m/z 109). Besides, we found that the intermediates with m/z values of 107 and 115 corresponded to the benzoquinone and butenedioic acid, respectively. In addition, the parent compound would be directly oxidized by the photogenerated holes (h + ), which consequently leads to the deacetylation (-COCH 3 ) of aspirin. Therefore, this creates the formation of salicylic acid (m/z of 137), which was considered as the other degradation pathway (shown in Figure 6). Furthermore, the product (m/z 153) revealing the addition of 16 mass units compared with salicylic acid was detected, which was probably due to the addition of hydroxyl radicals (-OH) to form monohydroxylated intermediates. Finally, all the intermediates could be decomposed into small molecules, such as CO 2 and H 2 O, during the PC process.
was added into the PC system [37], 49.7% of aspirin could be removed, revealing that H + could participate in the PC reaction. When tert-butyl alcohol (TBA) was added into the PC reaction system [38], the PC efficiency (5.4%) of aspirin sharply decreased. Additionally, 19.4% of PC efficiency for degradation of aspirin was obtained, suggesting H2O2 was involved in the aspirin degradation [39]. Similarly, BQ exhibited considerable inhibition of the PC degradation of aspirin (16.3%) [40]. Moreover, with the addition of K2Cr2O7 during the PC process [41], the PC efficiency of aspirin (13.1%) was partially suppressed, suggesting that the role of electron (e − ) was smaller than that of H2O2 in our PC system. These results also demonstrated that aspirin would occur several processes, which were finally converted to CO2 and H2O molecules.             Figure 7 displayed the proposed PC mechanism of degradation pollutant by P25. As seen, TiO2 with the valence band (VB) maximum potential of 2.91 eV and the conduction band (CB) maximum potential of −029 eV was easily excited to generate holes and electron pairs (h + /e − ) when turning on the light source [25,42]. Thus, under continuous light irradiation, the electrons were excited from the VB to CB, inducing the formation of the photogenerated H + in the VB. Subsequently, the residual H + reacted with H2O in the solution to yield OH radicals, which participated in the degradation of pollutants due to their highly oxidability. In addition, the separated e − were then trapped by O2 to produce O2 − radicals, which were then transferred to OH radicals to initiate the degradation process of organics. Thus, as seen from Figure 6, the side chain of aspirin could be altered via decarboxylation (-CO 2 ) of the compound, forming phenyl acetate (m/z of 135) in pathway 1. Therefore, the break in the C-O bond of phenyl acetate yielded phenol (m/z of 93), which subsequently could be attacked by hydroxyl radicals (·OH) on the benzene of ortho-position and para-position. This leads to the formation of p-dihydroxybenzene (m/z of 109). Besides, these products might be converted to 1,4-benzoquinone (m/z of 107) due to the self-electron transfer in the molecules. Moreover, we deduced that the benzene ring of benzoquinone was cleaved during the PC process and therefore, yielded butenedioic acid (m/z of 115). Ultimately, all these intermediates could be decomposed to acetic, oxalic acid and formic acid. With prolonged reaction time, these could be mineralized to CO 2 and H 2 O. Besides, the hydroxyl radicals (·OH) played a key role during the whole PC reaction. In pathway 2, the hydroxyl radicals (·OH) could attack the C=O double bonds of aspirin through demethylation (-CH 3 ) to form salicylic acid (m/z 137) and thus, would also attacked the aromatic hydrocarbons by the electrophilic addition on the C=C double bonds of aromatic ring to yield OH-adduct [35,36]. Finally, all the intermediates were completely converted to CO 2 and H 2 O under the Xenon lamp irradiation, which was highly consistent with the TOC results. Furthermore, we measured the contribution of each species to the PC degradation of aspirin, with the results illustrated in Figure S2. The PC efficiency was 68.5% after 60 min of Xenon lamp irradiation in the absence of any scavenger. When ammonium oxalate (AO) was added into the PC system [37], 49.7% of aspirin could be removed, revealing that H + could participate in the PC reaction. When tert-butyl alcohol (TBA) was added into the PC reaction system [38], the PC efficiency (5.4%) of aspirin sharply decreased. Additionally, 19.4% of PC efficiency for degradation of aspirin was obtained, suggesting H 2 O 2 was involved in the aspirin degradation [39]. Similarly, BQ exhibited considerable inhibition of the PC degradation of aspirin (16.3%) [40]. Moreover, with the addition of K 2 Cr 2 O 7 during the PC process [41], the PC efficiency of aspirin (13.1%) was partially suppressed, suggesting that the role of electron (e − ) was smaller than that of H 2 O 2 in our PC system. These results also demonstrated that aspirin would occur several processes, which were finally converted to CO 2 and H 2 O molecules. Figure 7 displayed the proposed PC mechanism of degradation pollutant by P25. As seen, TiO 2 with the valence band (VB) maximum potential of 2.91 eV and the conduction band (CB) maximum potential of −029 eV was easily excited to generate holes and electron pairs (h + /e − ) when turning on the light source [25,42]. Thus, under continuous light irradiation, the electrons were excited from the VB to CB, inducing the formation of the photogenerated H + in the VB. Subsequently, the residual H + reacted with H 2 O in the solution to yield OH radicals, which participated in the degradation of pollutants due to their highly oxidability. In addition, the separated e − were then trapped by O 2 to produce O 2 − radicals, which were then transferred to OH radicals to initiate the degradation process of organics.

Materials
Aspirin employed as the target pollutant in the spiked sample was kindly purchased from Sigma-Aldrich without any purification. Titanium dioxide (P25 TiO2) was purchased from Germany Degussa, which was used for photocatalysis before experiments. Besides, methanol and acetic acid were of HPLC grade, while other reagents were all of analytical grade. MilliQ-water was used throughout this experiment.

Design of Experiment
A CCD based on the RSM of the chemometric approach was used to distinguish the optimum condition for the degradation of aspirin. Analysis of this experimental data was supported by the Design-Expert software (trail version 8, Stat-Ease, Inc., Minneapolis, MN, USA).
In the CCD experimental design, three independent variables of initial pH value, initial aspirin concentration and P25 content on the PC removal efficiency of aspirin were analyzed for their responses. Besides, the rotatable experimental plan consisted of 17 experiments, which was determined by the expression as follows: N = 2 n + 2n + C0 (2) where N was the total number of experiments required; n was the number of variables with 2n being axial runs; and C0 was the center point runs (always chosen between 2 to 5), which was selected as 3 in this experiment. In addition, the three variables were converted to dimensionless ones of X1, X2 and X3 and the coded values at levels of −1 and +1 corresponded to the low and high levels, respectively. It should be noted that when there were 3 variables, α could be chosen as 1.682 [43]. Furthermore, after several pre-experiments were conducted, the determined values of the selected variables are presented in Table 4.

Materials
Aspirin employed as the target pollutant in the spiked sample was kindly purchased from Sigma-Aldrich without any purification. Titanium dioxide (P25 TiO 2 ) was purchased from Germany Degussa, which was used for photocatalysis before experiments. Besides, methanol and acetic acid were of HPLC grade, while other reagents were all of analytical grade. MilliQ-water was used throughout this experiment.

Design of Experiment
A CCD based on the RSM of the chemometric approach was used to distinguish the optimum condition for the degradation of aspirin. Analysis of this experimental data was supported by the Design-Expert software (trail version 8, Stat-Ease, Inc., Minneapolis, MN, USA).
In the CCD experimental design, three independent variables of initial pH value, initial aspirin concentration and P25 content on the PC removal efficiency of aspirin were analyzed for their responses. Besides, the rotatable experimental plan consisted of 17 experiments, which was determined by the expression as follows: N = 2 n + 2n + C 0 (2) where N was the total number of experiments required; n was the number of variables with 2n being axial runs; and C 0 was the center point runs (always chosen between 2 to 5), which was selected as 3 in this experiment. In addition, the three variables were converted to dimensionless ones of X 1 , X 2 and X 3 and the coded values at levels of −1 and +1 corresponded to the low and high levels, respectively. It should be noted that when there were 3 variables, α could be chosen as 1.682 [43]. Furthermore, after several pre-experiments were conducted, the determined values of the selected variables are presented in Table 4.

PC Degradation Experiments
Experiments on PC decomposition of aspirin were carried out in a self-made cylindrical quartz reactor to evaluate the PC performance of P25. Typically, the simulated solar light was generated by a 35-W Xenon lamp irradiation ( Figure S3). Besides, a radiation mater was used to adjust the luminous intensity of the Xenon lamp, which was controlled to 6.7 mW·cm −3 . In each run, 100 mL of an aspirin solution (10 mg/L) with a certain amount of P25 sample was added into the reactor, where a special glass atmolyzer was fixed at the bottom of it to maintain the constant air-equilibrated conditions. In order to obtain the adsorption−desorption equilibrium, the reactant mixture was kept in the dark for 30 min before irradiation. Subsequently, the PC activity was started when the Xenon lamp was turned on. At given time intervals, a certain amount of the solution was taken from the reactor and filtrated through a 0.22-µm membrance, which was measured at 35 • C using high performance liquid chromatography (HPLC, Shimadzu LC 10A, Shimadzu company, Kyoto, Japan) system. This system was equipped with a C18 column (150 mm × 20 mm × 4.6 mm i.d., Kromasil KR100-5, Shimadzu company, Tyoto, Japan). The mobile phase was a mixture of 35% methanol and 65% MilliQ-water, which contained 1% acetic acid, and the flow rate was maintained at 1 mL·min −1 . Furthermore, the sample with the injection volume of 20 µL was detected at the wavelength of 276 nm. Prior to the experiment, the mobile phase should be sonicated for 5 min to remove dissolved gas.
In addition, HPLC coupled with a thermos-LTQ Orbitrap Discovery mass spectrometer was employed in this experiment for detecting the PC degradation intermediates of aspirin. Besides, the measurement used ESI+/ESI− ionization and operated in full-scan mode between 50 and 400 uma. Separation was conducted using a C18 column (150 mm × 20 mm × 4.6 mm i.d., Kromasil KR100-5) with a flow rate of 0.3 mL·min −1 . It should be noted that the other conditions were in accordance with the HPLC measurements. Furthermore, total organic carbon (TOC) was measured to clarify the mineralization of aspirin using a TOC analyzer (Shimadzu, 5000, Kyoto City, Japan). Besides, several scavenging experiments were conducted to distinguish the contribution of each reactive species. It should be noted that prior to illumination, a scavenger (10 mM) was added into the system. After this, the light was turned on. At given time intervals, the collected sample was filtrated through a 0.22-µm membrane and measured by the HPLC method as mentioned above.

Conclusions
To conclude, the PC performance and degradation mechanism of aspirin by TiO 2 was evaluated and optimized by RSM based on CCD methods. In the present study, 68.5% of the aspirin could be removed within 60 min of Xenon lamp illumination by P25 and the process was further investigated by a decrease in TOC concentration. Besides, according to the RSM results, the variables of initial pH value (X 1 ), initial aspirin concentration (X 2 ) and P25 concentration (X 3 ) could significantly affect the PC degradation performance of aspirin and the significance of which followed the order of X 3 > X 2 > X 1 . In addition, the CCD method was employed as a powerful tool to optimize these parameters, suggesting that under the optimized conditions of initial pH value of 5, initial aspirin concentration of 10 mg/L and P25 concentration of 50 mg/L, an experimental degradation efficiency of 98.9% could be obtained. This is much closer to the predicted value calculated by the predicted model. Moreover, the tentative mechanism for PC decomposition of aspirin was deduced from HPLC-MS/MS measurements, where two main degradation pathways were proposed. Thus, this study provided the interacting effects among variables and the mechanism for PC degradation activity of aspirin by TiO 2 , which opened a gateway for further researches into wastewater treatment.