Photocatalytic activity of ZrO2/TiO2/Fe3O4 ternary nanocomposite for the degradation of naproxen: characterization and optimization using response surface methodology

In this study, ZrO2, TiO2, and Fe3O4 components were synthesized by co-precipitation, sol–gel, and co-precipitation methods, respectively. In addition, solid-state dispersion method was used for synthesizing of ZrO2/TiO2/Fe3O4 ternary nanocomposite. The ZrO2/TiO2/Fe3O4 nanocomposite was characterized by different techniques including XRD, EDX, SEM, BET, FTIR, XPS, EELS, and Photoluminescence (PL). The FTIR analysis of ZrO2/TiO2/Fe3O4 photocatalyst showed strong peaks in the range of 450 to 700 cm−1, which represent stretching vibrations of Zr–O, Ti–O, and Fe–O. The results of FTIR and XRD, XPS analyses and PL spectra confirmed that the solid-state dispersion method produced ZrO2/TiO2/Fe3O4 nanocomposites. The EELS analysis confirmed the pure samples of Fe3O4, TiO2 and ZrO2. The EDAX analysis showed that the Zr:Ti:Fe atomic ratio was 0.42:2.08:1.00. The specific surface area, pores volume and average pores size of the photocatalyst were obtained 280 m2/g, 0.92 cm3/g, and 42 nm respectively. Furthermore, the performance of ZrO2/TiO2/Fe3O4 nanocomposite was evaluated for naproxen removal using the response surface method (RSM). The four parameters such as NPX concentration, time, pH and catalyst concentration was investigated. The point of zero charge of the photocatalyst was 6. The maximum and minimum degradation of naproxen using photocatalyst were 100% (under conditions: NPX concentration = 10 mg/L, time = 90 min, pH = 3 and catalyst concentration = 0.5 g/L) and 66.10% respectively. The stability experiment revealed that the ternary nanocatalyst demonstrates a relatively higher photocatalytic activity after 7 recycles.

. The mixing ratio of used components in the synthesized nanocomposite was selected based on the highest efficiency of naproxen degradation under the same operation conditions. The best ratio was achieved (4:8:1) of the ZrO 2 , TiO 2 , Fe 3 O 4 compounds.
Photocatalyst characterization. The mineralogical analysis of the prepared nanoparticles was characterized by X-ray diffraction (Inel France, Equinox 3000) with Cu Kα radiation. SEM(AIS2100). It was used to determine the surface morphology and microstructure of ZrO 2 /TiO 2 /Fe 3 O 4 nanocomposite photocatalyst. The Fourier-transform infrared spectroscopy (VERTEX 70, Bruker, USA) in the wavelength range from 400 to 4000 cm −1 was used to analyze the chemical bonding in a photocatalyst. Texture properties of nanoparticles was investigated using TriStar-II-Series, Micromeritics Instrument Corporation, USA. The specific surface area of photocatalysts was determined via the N 2 adsorption-desorption and Brunauer-Emmett-Teller (BET) model. Barrett-Joyner-Halenda (BJH) method was used to calculate the size of pores and pore size distribution using adsorption-desorption curves. Powder samples for XPS analysis are prepared in the form of pellets in the laboratory (the powder should be large enough to cover a surface of 1.5 × 1.5 cm), then the samples should be placed in a vacuum chamber after preparation. A PHI Perkin-Elmer Model 5400 instrument was used to record XPS spectra. The EELS analysis was obtained using a sub-nanometer probe size using the GIF2000 Filter. The PL Spectroscopy was recorded by Avaspec-2048-TEC model spectrometer.
Point of zero charge determination. An amphoteric oxide method was used to determine the point of zero charge of the photocatalyst 73 . In order to determine the point of zero charge (PZC), 0.25 g of the synthesized photocatalyst was added into the solution with the same ionic strength and different pH values. The synthesized nanocatalyst was added to 50 mL of 0.1 M NaNO 3 solution. The pH of solution was adjusted in the range of 3-9 using 0.1 M H 2 SO 4 and 0.1 M NaOH. The pH of solution was measured before and after 24 h mixing and those were called pH i and pH f respectively. The plot of pH changes (ΔpH) as a function of initial pH i was used for the determination of PZC 73 .
Optimization of the photocatalyst Activity. In order to optimize photocatalyst activity, the RSM has been used. Design Expert 8.0 software (Stat-Ease, Inc., USA) was used to design experiments and analyze mathematical modeling. Each experiment was performed in duplicate. Box-Behnken design (BBD) was applied in this www.nature.com/scientificreports/ study because of its rotatable or nearly rotatable second-order design. The "y" was the response (the percentage of naproxen degradation). We evaluated four independent variables, including catalyst loading dosage, initial concentration of naproxen, time and pH value in order to find an optimal condition: Photocatalyst activity. ZrO 2 /TiO 2 /Fe 3 O 4 Photocatalyst was used for the degradation of naproxen in a pharmaceutical synthetic solution. The experiments were conducted based on the conditions in Table 2. In this study, the effects of solution pH, time, initial naproxen concentration, and photocatalyst loading dose on naproxen degradation were evaluated. To determine the adsorption behaviour of the photocatalyst, the solution kept was in the dark for half an hour. The reactor was irradiated immediately after adding the photocatalyst. For each experiment, a 150 W UV light source has been used for irradiation (The intensity was kept as 15 W/m 2 ). At the end of each experiment, centrifugation with 12,000 rpm for 20 min have been performed for each withdrawn sample followed by absorbance measurement using UV-Vis (Shimadzu UV2401PC) at λ = 230 nm. The efficiency of designed photocatalysts for the degradation of naproxen was calculated as follows: where C i and C f are the naproxen concentration before and after degradation reaction respectively.
BET analysis. Figure 4a and b show the porosity distribution curve and pore size distribution of ZrO 2 /TiO 2 / Fe 3 O 4 nanocomposite. The specific surface area (SSA) was calculated using the N 2 isotherms. The BET-SSA, the volume of pores, the average size of pores of the photocatalyst obtained were 280 m 2 /g, 0.92 cm 3 /g, and 42 nm respectively. The sample is mesoporous as the diameter of pores is larger than 2 nm. Hysteresis loop of types IV-H1 was obtained for the sample which is related to cylindrical and spherical pores in the sample 82 . In terms of pore size distribution, wide pore size distribution was observed, especially, pores with size larger than 3 nm had significant volume in the sample. It was found that the pore size larger than 3 nm is appropriate for the reactants penetration into the porous media of photocatalyst 83,84 .  Figure 6d shows the peaks of 458.5 eV and 464.2 eV, which correspond to Ti 2p3/2 and Ti 2p1/2, respectively. The peak position between Ti 2p3/2 and Ti 2p1/2 at 464.2 eV indicates the presence of Ti 4+ oxidation state 89,90 . Figure 6e and f show the XPS spectrum of ZrO 2 nanoparticles, which indicate the presence of relevant elements. The Zr3d level spectra of Fig. 6f show the Zr3d5/2 and Zr3d3/2 peaks at binding energies of 185.66 eV and 188.1 eV, respectively. The energy difference of 2.44 eV between the two peaks indicates the presence of Zr +491 . The third suitable peak for the shoulder that appears at the base of Zr 3d3/2 can be attributed to the lack of oxygen, which can be due to under-coordinated Zr sites of very small ZrO 2 nanoparticles 92 . XPS     Fig. 7, the Ti structure in the EELS spectrum for Ti 4+ , indicates a lower oxidation state which has edges shift slightly towards the lower energy loss. Here, the blue curve is the fine structure of the Ti 4+ state. EELS was obtained using a sub-nanometer probe size using the GIF2000 Filter.
Photoluminescence spectra. The light absorption of photocatalysts significantly affects the activity of photocatalysts. Also, the spectrum of PL (Photoluminescence) and their intensity are closely related to its photocatalytic activity. Figure 8a- 98 . According to previous research [99][100][101][102] , the presence of these peaks in the visible range is probably due to the presence of oxygen vacancies, defects, surface states and other structural impurities. A high PL intensity of pure ZrO 2 appears to indicate that the ZrO 2 surface states are much lower, hence easy electron transfer can occur between VB and CB of ZrO 2 even with a low-energy laser excitation source (325 nm) 103 . Additionally, when Fe 3 O 4 and TiO 2 are loaded with ZrO 2 as ternary oxide catalysts, the PL intensity of ZrO 2 decreases abruptly (Fig. 8d). The formation of a chemical interaction between pure oxides when they accumulate together ( Photocatalyst performance. The experimental work was designed and the obtained actual results, as well as predicted values, were provided in Table 3. Box-Behnken design for RSM was used for the experimental design. Statistical investigation of experimental data was performed using linear, two-factor interaction, quadratic, and cubic models. The results were given in Table 4. According to Table 4 and R 2 values, the quadratic model was found the most accurate model for the prediction of experimental data in terms of naproxen removal. Its predicted R 2 value was found to be 0.9901 which means the developed model equation is able to predict the results accurately. Therefore, the following quadratic polynomial model in terms of coded factors was proposed for investigation of naproxen degradation from pharmaceutical synthetic solution as a function of operating parameters: where A, B, C, D are photocatalyst dose (g/L), time (min), solution initial pH, and initial concentration of naproxen (mg/L). The variables were defined at three levels including − 1, 0, and + 1. Figure 9 illustrates the individual effects of operational parameters on naproxen removal. The contour lines represent lines of equal response and can be visualized as response contours two factors at a time. In this study, contour lines mean the performance of photocatalyst for naproxen removal at different operating parameters. The contour map reflects the cross-interaction between two variables by keeping the other variable constant. Interaction of initial pH and naproxen concentration for photocatalyst dose of 0.3 g/L and time of 60 min was shown in Fig. 9a. At a pH value of 6, increasing the initial concentration of naproxen from 10 to 30 mg/L led to the reduction of naproxen removal from 91.81 to 82.9%. Naproxen removal is reduced due to solution turbidity and a decrease in light absorption by the photocatalyst. Also, the amount of hydroxyl radical in the solution decreases. Furthermore, the photocatalyst surface is covered by pollutants with increasing its concentration from 10 to 30 mg/L which decreases photon penetration into the photocatalyst. Therefore, the photocatalyst was unable to generate enough electron-hole pairs, resulting in a reduced removal of naproxen. In the interaction between time and initial pH of the solution (Fig. 9b), the photocatalyst dosage and initial concentration of naproxen were considered 0.3 g/L and 20 mg/L respectively. Firstly, the efficiency of naproxen degradation was increased from 90.10 to 91.34% with increasing pH of the solution from 3 to 6, then, it was decreased to 84.9% at a pH value of 9. These changes are related to the oxidation potential and surface charge of the photocatalyst. The pH of the media can have a significant impact on the adsorption and desorption of pollutants on the photocatalyst surface.
The amount of pK a is 4.15 for the naproxen. At pH values higher than 4.15, naproxen has a negative charge, otherwise, it has a positive charge. The neutral charge point of photocatalyst is 6. Therefore, it has a positive charge at pH below 6, whereas the surface charge of photocatalyst is negative at pH above 6. When the pH value is between 4.15 and 6, the naproxen and photocatalyst have opposite charges, so the adsorption of naproxen on the photocatalyst surface increases in this range. It is for this reason that maximum degradation occurs at pH 6 rather than pH 3 or 9. When the pH solution was 6, an increase in time from 30 to 60 min increased naproxen removal from 84.10 to 91.34%. The time and photocatalyst dosage interaction at constant initial pH and concentration  Fig. 9c. In Fig. 9c, and for a 60-min experiment run, there was an increase in naproxen degradation from 86.2 to 100% when the photocatalyst dosage was increased from 0.1 to 0.5 g/L. Photocatalyst dosage had a significant effect on naproxen degradation, increasing the number of available active sites for the generation of hydroxyl radicals. However, too much photocatalyst in the system can reduce naproxen degradation as a result of agglomeration of photocatalyst nanoparticles. The reduction in active site of the photocatalyst resulted from agglomeration. It was found that a decrease in pH can increase naproxen degradation (Fig. 9d), and the greatest degradation was obtained at pH 3. The solution pH had a complicated effect on the photocatalytic oxidation reaction. The optimum pH value is highly dependent on the type of pollutant and point of zero charges (PZC) of the photocatalyst. According to Fig. 10, it is 6 for the synthesized photocatalyst. At pH = pH PZC , the surface charge is neutral. The surface charge can be positive or negative at pH < pH PZC and  Table 5. The results suggested that a quadratic model is significant because of its high F-value (431.34) and very low p-value (< 0.001). The calculated higher F-value and lower p-value for the photocatalyst dosage (D factor) in comparison with other factors indicated that the photocatalyst dosage is the most important parameter in the system. The results in Table 5 showed that the proposed statistical model is accurately fitted to experimental data. The accuracy of the proposed model should be investigated in detail. Figure 11 shows a linear relationship between experimental data and predicted values in terms of naproxen removal and it confirms the accuracy of the proposed model as all the data accumulated around a 45-degree line. The normal probability plot of the residuals was shown in Fig. 11b and it can be clearly seen a linear scattering of modeling data. The linear scattering means normal distribution of errors in a defined matrix for experimental design. Therefore, the probability of random error intervention and effect of the sequence of experiments is considerably decreased in the proposed model. Optimization using RSM. An optimization was performed in order to find the optimal operating conditions for the removal of naproxen completely from pharmaceutical synthetic solutions. According to the proposed model, the following independent operating parameters are required for complete removal of naproxen (101.92%): initial naproxen concentration = 18.95 mg/L, initial pH value = 4.55, photocatalyst dosage = 0.49 g/L,  Figure 12 illustrates different operating conditions that can be used to achieve complete degradation of naproxen.

Reusability ZrO 2 /TiO 2 /Fe 3 O 4 photocatalyst.
The regeneration and re-use of the used or spent photocatalyst is very critical for determining the applicability of the process applied. The used photocatalyst was separated from the synthetic solution using 1.3 Tesla magnetic. Then, it was used for the treatment of pharmaceutical synthetic solution as given in Fig. 13. As can be seen in Fig. 13, the degradation efficiency only display partial reduction after seven cycles for degradation of naproxen, indicating excellent reusability of the synthesized photocatalyst as after 7 runs, its performance for naproxen degradation was decreased only 12%. Additionally, the photocatalyst was placed in distilled water under UV light for 12 h in order to regenerate it. After doing the regeneration process, the photocatalyst was able to degrade the naproxen by about 98.4%. However, after three regenerations, the activity of the photocatalyst decreased by 8.4%. The decrease in performance of photocatalyst may be related to the blocking of active pores due to precipitation of non-sensitive chemicals to light or destroying of effective nanoparticles during the recovery process. Each experiment was replicated three times and an error bar was shown in Fig. 13.  www.nature.com/scientificreports/ Analysis study of naproxen removal kinetics. The removal of naproxen using heterogeneous photocatalyst is divided into two main stages: physical adsorption reaction and chemical reaction. In order to describe the photocatalytic degradation rate of naproxen by plotting ln(C 0 /C) versus time (t), at different concentrations, the Langmuir-Hinshelwood (LH) kinetic model was used 106 : where C i (mg/L), the initial concentration of naproxen and C(mg/L), the concentration at time (t) after irradiation, and k is a pseudo-first-order rate constant. As shown in Fig. 14, the pseudo-first-order rate constant was determined based on a straight-line slope. According to the results of Fig. 14b, the maximum reaction rate at the initial concentration of 5 mg/L naproxen is 0.0514 min −1 , which is about 2 times higher than the obtained reaction rate at the initial concentration of 30 mg/L naproxen. The obtained results in Fig. 14b show that reaction rates decrease with increasing concentration. The naproxen absorbed amount q e (mg/g) by ZrO 2 /TiO 2 /Fe 3 O 4 nanocomposite was calculated using the following equation (Eq. (6)): where C i and C e are the initial and equilibrium concentrations of naproxen (mg/L), respectively. The volume of naproxen solution (ml) is V and the mass of ZrO 2 /TiO 2 /Fe 3 O 4 nanocomposite (mg) is m. The following equation was used to calculate the percentage of naproxen adsorption (%) (Eq. 7): As shown in Fig. 14 when the initial concentration of naproxen was increased from 5 to 30 mg/L, the initial photodegradation rate (r 0 ) also gradually increased from 0.26 to 0.77 mg/L. It was shown that naproxen photocatalytic degradation occurs on ZrO 2 /TiO 2 /Fe 3 O 4 nanocomposite surfaces and that the rate of photo degradation increases as a function of increasing of adsorption. As the initial concentration of naproxen increases, the coverage of naproxen molecules on the surface of ZrO 2 /TiO 2 /Fe 3 O 4 nanocomposite also increases accordingly. As a result, the electron transfer efficiency of naproxen molecules to the adsorbed surface and the charge produced by the light increases, which leads to an increase in the initial optical decomposition rate (r 0 ). As a result, the electron transfer efficiency of naproxen molecules, which leads to an increase in the initial photodegradation rate (r 0 ), increases with increasing absorbed surface and the produced charge by light. For different initial concentrations, the kinetic constant of naproxen photo degradation gradually decreased (from 0.0514 to 0.0255 min −1 ), while the R 2 correlation coefficient also decreased from 0.9961 to 0.9817. The relationship between the reaction rate (k) and the initial concentration of the substrate during the photocatalytic process can be generally explained by the following experimental formula (Eq. 8): Linear regression was used to analyze the relationship between the kinetic constant of NPX photo degradation and its initial concentration (5-30 mg/L). Figure 14 shows that the relationship between reaction rate k and NPX concentration is as follows: The results showed that the adsorption of naproxen on the surface of ZrO 2 /TiO 2 /Fe 3 O 4 nanoparticles is clearly time-dependent. As shown in Fig. 14a most naproxen removal occurred within the initial 50 min, and after 50 min, naproxen removals occurred almost in the flat part of the graph. The highest percentage of naproxen removal is at the initial concentration of 5 mg/L of solution. In the initial 50 min, two mechanisms were involved in the removal of naproxen: the free active sites at the adsorbent surface, and the generation of hydroxyl ions. Subsequently, a slow rate of naproxen removal was observed.
Different kinetic models were investigated to analyze the kinetic data and determine the kinetic mechanism of naproxen adsorption on the surface of ZrO 2 /TiO 2 /Fe 3 O 4 nanocomposite which is shown in Table 6.
Elovich model shows less linearity in the regression coefficient R 2 ; which was found to be 0.9202, On the other hand, the correlation coefficient value for a Langmuir-Hinshelwood (LH) kinetic model was obtained 0.9992 for NPX and greater than all other adsorption kinetic models. The calculated q e value was found to be 26.4 mg/g for NPX, which was very close to the obtained experimental values (shown in Table 6). According to the obtained results, it is expected that NPX and its degradation intermediates may be further degraded by reactive species, thus leading to ring openings and eventually oxidation to CO 2 and H 2 O. As can be seen in Fig. 15, the possible degradation pathways of NPX are proposed based on the identification of intermediates and mineralization results. The photocatalytic degradation of NPX by ZrO 2 /TiO 2 /Fe 3 O 4 mainly refers to three main pathways including decarboxylation and hydroxylation 39 . Three possible degradation pathways through oxidation processes are shown in Fig. 15. In the oxidation pathway I, NPX degradation was initiated by the electrophilic additive interaction between the naphthalene ring of NPX and ·OH [114][115][116] . In pathways II and III, NPX oxidation was performed by h + and ·O 2− and carbon-based radical species were formed by decarboxylation 117 . All produced intermediates during the process were decomposed by ring-opening reactions to malic acid, succinic acid, propionic acid and acetic acid, and finally mineralized to CO 2 and H 2 O 39,118,119 .
In order to confirm mineralization, we investigated total organic carbon (TOC) in the degradation of naproxen at ambient temperature and pH = 3. The amount of TOC was decreased with increasing irradiation time (Fig. 16). In this study, the mineralization of naproxen is confirmed using ZrO 2 /TiO 2 /Fe 3 O 4 photocatalysts. TOC removal was obtained 87% after 90 min irradiation. Naproxen may be oxidized to CO 2 , H 2 O, and some small molecules, according to TOC results.
As shown in Table 7, in comparison with previous studies, the synthesized photocatalyst in this study had the highest naproxen removal efficiency. The synthesis of ternary nanocomposites confirmed the excellent synergistic effect of these three types of nanoparticles on the photocatalytic process. Based on the Table 7, the synthesized photocatalyst has high specific surface area. In addition, it's proper capability in acidic environment increases efficiency and absorption over time for the removal of naproxen at low dosage of catalyst in comparison with other similar synthesized photo catalysts.

Conclusion
In summary, we synthesized a ternary ZrO 2 /TiO 2 /Fe 3 O 4 nanocomposite system using a solid-state dispersion method. Furthermore, its photocatalytic activity towards naproxen degradation was investigated. The effect of operating parameters including initial naproxen concentration, initial pH, photocatalyst dosage, and time on the degradation of naproxen was investigated. Characterization results confirmed the formation of the ternary nanocomposite. The existence of cylindrical and spherical pores in the sample was proved because its N 2 adsorption-desorption hysteresis followed IV-H1 type hysteresis. Based on optimization results using the RSM method, the optimal conditions for the complete removal of naproxen were determined as initial naproxen concentration of 18.95 mg/L, initial pH of 4.55, photocatalyst dosage of 0.49 g/L, and time of 74.31 min. Reusability results showed 12% reduction in naproxen degradation after 7 runs. It was found that Langmuir-Hinshelwood's (LH) K 1 : rate constant of pseudo 1st order adsorption process (min −1 ). K 2 : the rate constant of pseudo 2nd order adsorption process (g mg −1 min −1 ). α: the initial adsorption rate constant (mg/g min). β: constant related to surface coverage and the activation energy for chemisorptions (g/mg). K int : the intraparticle rate constant (mg/g min 1/2 ). C: value about the boundary thickness.