The surface residue of waste oyster shells was cleaned with detergent and cleaning ball, and the surface was polished to white with cutting machine and cut into small pieces, dried at 65˚C in blast drying oven and crushed in sealed sample making machine, sieved at 120 mesh, then soaked in 0.1% HCl, vacuum dried and calcined at 900˚C for 2 h in muffle furnace to obtain modified oyster shell powder, sealed and kept for backup.

Measure 36 mL of anhydrous ethanol and quantitative modified oyster shell powder mixed with ultrasonic shaking for 10 mins, then under the action of magnetic stirring, tetrabutyl titanate was added to the above mixture drop by drop at a rate of 1 - 2 drops/s and stirred evenly for 20 min to form solution A; Measure 36 mL of anhydrous ethanol, 3.0 mL of distilled water, 0.2 mL of hydrochloric acid and mix thoroughly, add a certain rate of urea and cerium nitrate hexahydrate, and form solution B by ultrasonic shaking for 10 min. The solution B was slowly added dropwise to the above solution A at a rate of 30 - 35 drops/min and stirred until a uniform transparent sol was formed. After aging in a vacuum drying oven at 25˚C for 24 hours, it was put into a blast dryer at 85˚C for overnight drying, crushed, and finally calcined in a muffle furnace at a certain temperature for 2 h to form the modified oyster shell powder/Ce-N-TiO₂ composite photocatalyst. Alternatively, the method of preparing the Ce-N-TiO₂ photocatalyst is the same as that described above except that the modified oyster shell powder is not added.

It can be seen from Figure 1 that the degradation rate of PMG increases with the increase of Ce doping rate. When the Ce doping rate is 0.1 mol%, the maximum degradation rate is 79.5%, and when the Ce doping rate increases to 1.0 mol%, the degradation rate of glyphosate decreases to 52%. The results show that, there is an optimal doping value for cerium doped with different photocatalysts. Ce

doping makes Ce³⁺ and Ce⁴⁺ coexist on the crystal surface or gap. Ce³⁺/Ce⁴⁺ is called “oxygen tank” because of its special redox performance [20]. Ce⁴⁺ has strong electron capture ability to improve the separation efficiency of photogenerated e⁻/h⁺ pairs, while Ce³⁺ can absorb oxygen to form superoxide radicals to participate in the photocatalytic reaction, as shown in Equation (2) and Equation (3). However, with the increase of cerium doping rate, the TiO₂ anatase lattice will cause structural changes, which will instead produce the compound phenomenon of photogenerated e⁻/h⁺ pairs [21], thus reducing its photodegradation activity to PMG.

Ce 4 + + e − → Ce 3 + (2)

Ce 3 + + O 2 → O • 2 − + Ce 4 + (3)

As shown in Figure 2, the N doping reaches a maximum at 10 mol% and the

degradation rate of PMG is even inferior to that of the undoped nitrogen photocatalyst after the N doping exceeds 50 mol%. It has been shown [22] [23] [24] that N doping forms a new N 2p energy level band between the TiO₂ conduction and forbidden bands. The valence band electrons can be excited and leap to the impurity energy level N 2p, and then move further to the conduction band, shortening its forbidden band width to make its absorption edge appear red-shifted and improve the photocatalytic efficiency. When the rate of N doping is excessive, it will form deposits on the surface of TiO₂, which will not be uniformly dispersed and the active sites will be covered, thus reducing the photocatalytic activity.

As can be seen from Figure 3, the calcination temperature has a greater effect on the degradation rate of PMG than the Ce and N doping, with the degradation rate of PMG reaching over 80% at 500˚C, however dropping to 40% at 700˚C. When the calcination temperature was low, TiO₂ did not reach the crystallization critical temperature of amorphous phase-anatase, and the photocatalytic degradation performance of the formed amorphous phase TiO₂ was low. When the calcination temperature was too high, firstly, the agglomeration and sintering phenomena on the catalyst surface during the high temperature calcination led to a decrease in the specific surface area of the modified oyster shell powder/Ce-N-TiO₂ composite photocatalyst [25] Secondly, when the calcination temperature is excessive, the rate of nitrogen or cerium doping may be lost and thus affect the effective formation of impurity energy levels [26], reducing the use of visible light portion of the photocatalyst.

Based on the Box-Behnken Design (BBD) design response surface test set shown

in Table 2, a ternary linear regression equation with N doping (X₁), Ce doping (X₂) and calcination temperature (X₃) as the response variables and PMG degradation rate (Y) as the response value was established and is shown in Equation (4).

Y = 78.80 − 2.52 X 1 − 0.9375 X 2 + 3.58 X 3 + 1.37 X 1 X 2 + 0.3333 X 1 X 3     + 0.8333 X 2 X 3 − 2.30 X 1 2 − 2.46 X 2 2 − 14.17 X 3 2 (4)

The results of the preliminary model statistical significance analysis based on analysis of variance (ANOVA) of the model data based on Design-Expert 11.0 software are shown in Table 3. the ANOVA results show a p-value of <0.0001 and an F-value of 101.69, indicating that the model is highly significant; on the other hand, the lack of fit term has an F-value of 2.39 and the response has a p-value of 0.2092 > 0.05, which indicates that the model is not significant in terms of its shortcomings in predicting the data. The X₁, X₂ and X₃ terms in the model were all significant levels, with the interaction term X₁X₂ being more significant, indicating that the three factors selected through the one-way test had a more significant effect on the photocatalytic performance of modified oyster shell powder/Ce-N-TiO₂.

Based on Pareto analysis, the effect of each independent variable, the interaction effect variable, on the photocatalytic degradation of PMG was assessed [27]. The relative magnitude of the effect of each factor on the response was calculated according to Equation (5) [28] as follows.

P i = β i 2 Σ β i 2 (5)

where β_i denotes the regression coefficient for the linear, quadratic and interaction effects of the response of the second order polynomial. As shown in the Pareto graphical analysis in Figure 4, where calcination temperature had the greatest effect on the response, followed by N doping, the interaction effect of X₁X₂ was more pronounced than X₂X₃ and X₁X₃. In summary, the results of the Pareto analysis remained consistent with the results of the ANOVA analysis. The sign of the single factor coefficients in Equation (4) indicates positive or negative effects, and within the range of this study, catalyst calcination temperature has a positive effect, while N and Ce doping are both negative effects; for the crossed product coefficients, positive and negative values represent synergistic and antagonistic effects, respectively, and there are mutual synergistic effects between the three factors within the range of this study.

In response surface experimental analysis, the model validity depends on a number of assumptions expressed by diagnostic statistics [29] [30], such as scatter plots of actual versus predicted values, scatter plots of studentized residuals versus predicted values, fold plots of studentized residuals versus number of runs and normal probability plots. From Figure 5(a), it is found that the scatter points corresponding to the actual and predicted values are evenly scattered around a certain straight line, indicating that the predicted response is in good agreement with the actual values and that the model shows a good fitting trend. Figure 5(b) compares the studentized residual values with the predicted response values and then tests the assumption of constant variance. The scatter plot has some random dispersion, with residuals randomly distributed between the range of [−3, 3]. In addition, the uniform random distribution of each point does not indicate that the variance has a peculiar tendency of linear increase or decrease, so the assumption of constant variance is tenable. Figure 5(c) shows the studentized residuals from a model run test, usually used to test hypothetical outliers. The studentized residuals should obey N (0, 1) interval distribution and be approximately independent of each other, and by the nature of the standard normal distribution there exists approximately 95% falling in the range of [−2, 2] horizontal band and further approximately 97% falling in the range [−3, 3] horizontal band [31], whereas Figure 5(c) shows that the model studentized residual values fit perfectly and do not show any trend. Figure 5(d) is the normal graph of simulation residuals, indicating that the model residuals follow the normal distribution. Obviously, it is proved that it is sufficient to describe the relationship between the research variables and the response values.

From the statistical analysis of the errors, the correlation coefficient R² of 0.9924 indicates that the model can explain more than 99.24% of the deviation of the data; the corrected coefficient of determination R a d j 2 of 0.9827 indicates that there is a high correlation between the experimental and predicted values; the signal-to-noise ratio (Adeq Precision) of 27.2764 is much greater than 4, indicating that the model has sufficient signal and likewise The CV of 1.57% < 10%

indicates that the test data is accurate and the experimental operation is reliable. In summary, the response surface regression model designed by this central combination can be used to predict the optimal preparation process parameters of modified oyster shell powder/Ce-N-TiO₂ for the degradation of PMG aqueous solution.

By fixing one preparation factor parameter, the relationship between the other two preparation parameters and the response can be analyzed by 3D response surface plots and contour plots [32]. Figure 6 shows the contour and 3D response surface of the degradation rate of PMG solution (150 mL, 50 mg/L, reaction time of 300 min) by simulated Ce doping rate, N doping rate and calcination temperature under fluorescent light.

The contours formed by Ce doping and N doping at a calcination temperature of 500˚C are elliptical, indicating a more significant interaction between the two factors. From the (Figure 6(a), Figure 6(b)), the resulting 3D surface has less curvature and is a gentler convex surface, indicating that both have less influence

on the degradation rate within the range of values taken for this test. However, the slope of the N doping curve is steeper and denser compared to the Ce doping curve, indicating a greater contribution of N doping to the effect on response values. This is consistent with the results of comparing the F-values of Ce doping and N doping in Table 3, where the contribution of N doping to the response value is much greater than that of Ce doping.

From the Figures 6(c)-(f), the interaction between N doping and calcination temperature is not significant; the contour plot of Ce doping and calcination temperature is approximately elliptical, indicating that there is some interaction between the two, and this result is more consistent with the results of the p-test in Table 3. The 3D response surface shows that the calcination temperature plays an important role in the degradation of PMG by the photocatalyst, with the degradation rate of PMG solution increasing and then decreasing with the increase in calcination temperature while the Ce or N doping is kept constant, with the maximum response value around 500˚C.

The results of the contour and 3D response surface analysis are consistent with the ANOVA as derived from the graph. Both the steepness of the 3D surface and the F-value showed that the degradation rate of PMG degradation by modified oyster shell powder/Ce-N-TiO₂ was affected in the order of calcination temperature > N-doping rate > Ce-doping rate. The model was solved using Design-Expert 11, and the optimum conditions for the preparation of modified oyster shell powder/Ce-N-TiO₂ to achieve the highest activity in degrading PMG solution were as follows: N doping rate of 17.057 mol%, Ce doping rate of 0.165 mol% and calcination temperature of 505.440˚C. The predicted degradation rate of PMG solution (100 mL, 50 mg/L, 300 min reaction time) was 82.15%. Considering the actual conditions, the optimal preparation conditions were finally determined: N doping rate of 17 mol%, Ce doping rate of 0.165 mol% and calcination temperature of 505˚C.