Introduction

Due to the fact that most of the dye pollutants in industrial wastewater are toxic and cause many environmental problems and are very harmful and dangerous for human health and aquatic organisms (Daraei et al. 2010); therefore, the removal of such dangerous pollutants is considered as one of the most important environmental measures in today's industrial societies (Li et al. 2014; Abbasi 2018; Gu, et al. 2024; Sun, et al. 2024). Although several methods such as ion exchange, reverse osmosis, and filtration have been investigated to remove dye pollutants, these methods have not been effective, which can be due to the complex structure of organic pollutants (Huang, et al. 2015; Abbasi 2020, 2021a; Daraei et al. 2017). In recent years, the advanced oxidation process as one of the most effective and widely used methods for the decomposition and removal of dye organic pollutants has attracted the attention of many researchers. The most important advantages of this method are the high efficiency of pollutant removal and the non-production of secondary pollutants (Mazinani et al. 2014; Abbasi 2021b). The advanced oxidation process is based on the excitation of valence layer electrons of the used semiconductor and the creation of active oxidizing radicals. Therefore, in order to increase the efficiency of removing organic pollutants, it is very important to choose a suitable semiconductor (Abbasi 2022, 2023a; Majlesi and Daraei 2016). Among the different semiconductors that are used in photocatalysis, TiO2 is the most suitable semiconductor in the advanced oxidation process due to its compatibility with the environment, non-toxicity, relatively low price, high abundance and wide band gap (Rétia et al. 2016; Abbasi 2019; Dastan et al. 2017; Manshouri et al. 2012). The efficiency of titania for the photocatalytic removal of organic pollutants is influenced by parameters such as particle size, specific surface area, morphology and crystal structure (Abbasi 2022, 2023b). Considering that titania has three different crystal structures such as anatase, rutile and brookite, each of these crystal structures has a unique band gap that affects its photocatalytic activity (Abbasi et al. 2021). Anatase crystalline phase has more photocatalytic activity than the other two phases. In case of combining different crystal phases, the photocatalytic activity may increase significantly compared to single phases (Navidpour et al. 2023). The photocatalytic activity of the semiconductor is initiated by the irradiation of radiation with energy equal to or greater than the band gap of the semiconductor. The contact of light rays with the semiconductor leads to the excitation of valence layer electrons and then transfers to the conduction layer (Abbasi et al. 2017a; Abbasi and Hasanpour 2017). For each electron that is transferred from the valence layer to the conduction layer, a vacancy (hole) is created in the valence layer and an electron is created in the conduction layer. The produced electron–hole pairs are prone to the formation of radical species such as \({\text{O}}_{2}^{ \cdot - }\) and \({\text{OH}}^{ \cdot }\), which have the ability to react with organic pollutants dissolved in the suspension and decompose them (Abbasi et al. 2017b; Ghaderi et al. 2018). Photocatalytic activity strongly depends on the amount of radical species and as a result the separation of electron–hole pairs. One of the most important strategies to prevent recombination and thus increase the stability of electron–hole pairs is the distribution of semiconductors on the surface of carbon materials with a high surface-to-volume ratio (Roozban et al. 2017a, 2017b). Graphene and its derivatives, such as graphene oxide (GO), which have a two-dimensional structure, have attracted the attention of many researchers for the preparation of composites due to their special and unique properties such as significant surface area and having functional groups containing oxygen (Gan et al. 2018; Abbasi et al. 2020; Banerjee et al. 2018). Despite the fact that the synthesis of photocatalytic nanoparticles on the surface of graphene oxide improves the dispersion of nanoparticles and prevents their accumulation and agglomeration, nevertheless, the separation of the photocatalyst from the pollutant solution is very complex and expensive. Therefore, this causes limitation of photocatalyst efficiency. One of the effective methods to solve this problem is to magnetize the photocatalyst. Fe3O4 is one of the best choices for photocatalyst magnetization due to its extraordinary magnetic properties, catalytic properties and biocompatibility (Ruíz-Baltazar et al. 2019; Fan et al. 2021; Abbasi 2024a). Separation of magnetic photocatalysts from suspension using an external magnetic force such as a magnet can increase the speed of separation and also reduce operating costs.

Although, so far, extensive studies have been reported on the removal of dye organic pollutants by single, hybrid and composite photocatalysts, also, the effect of effective factors on the efficiency of pollutant removal has been reported, but the mutual effect of effective parameters on the photocatalytic activity of magnetic composites has not been investigated much. Therefore, the purpose of this report is to study the changes in the photocatalytic activity of the magnetic composite containing Fe3O4, TiO2 and GO and to compare the photocatalytic efficiency of the synthesized composite with the hybrid without GO. The mutual influence of the effective parameters is investigated using graphical statistical methods as well as analysis of variance.

Materials and methods

Materials

The materials used for the synthesis of photocatalysts and photocatalytic applications include the following, which were used without any additional purification. All the applied chemicals have a purity greater than 99%, which is purchased from Merck: tetra chloride titanium (TiCl4), methyl orange (MO), ferric acetylacetonate (Fe(acac)3), ammonium acetate (NH4CH3CO2). Graphene oxide (GO) is the only material that was purchased from Novin Nano Negasht Company, Iran.

Photocatalytic activity for degradation of pollutants

The photocatalytic decomposition of methyl orange as a dye organic pollutant is investigated using two types of hybrid and composite photocatalysts, whose synthesis method is mentioned in the previously published articles of this team (Abbasi 2024b; Abbasi et al. 2023). Synthesized photocatalysts include magnetic titanium dioxide (Fe3O4-TiO2) as a hybrid and hybrid arranged on the surface of graphene oxide (MGO-TiO2) as a composite. A photoreactor equipped with a medium pressure mercury lamp is used to perform photocatalytic reactions and pollutant removal. Considering that the photocatalytic degradation of organic pollutants is influenced by the intensity of UV light emitted from the light source, therefore, the location of the UV lamp is one of the key parameters in the design of the photoreactor. For this purpose, for uniform contact of the pollutant with UV radiation in all directions, the UV lamp is fixed exactly in the center of the photoreactor. The light source embedded in the photoreactor can increase the temperature of the reaction medium. Therefore, in order to control the temperature and improve the performance of the photoreactor, it is important to equip the photoreactor with a cold water circulation system or a cold water bath. Irradiation duration, photocatalyst concentration and acidity of the suspension containing the pollutant are investigated as three effective factors in the removal efficiency of methyl orange. In order to investigate the effect of each of the mentioned factors, 70 ml of methyl orange aqueous solution with a concentration of 10 ppm is prepared and the synthesized photocatalyst is dispersed in different concentrations (0.05, 0.1 and 0.2%wt) in the pollutant solution. Then the acidity of the suspension is stabilized in the acidic (pH = 3), alkaline (pH = 11) or neutral (pH = 7) range. Before starting the irradiation of the suspension, the photoreactor is placed in a completely dark environment for one hour. Then, 3 ml of the suspension is drained and filtered several times to separate the photocatalyst completely. Finally, the pollutant concentration in the suspension, which is proportional to the amount of light adsorption at the wavelength of 464 nm, is measured using a spectrophotometer (Perkin Elmer Company) and is considered as A0. After measuring the initial concentration of the pollutant, the irradiation of the suspension is started and the absorption rate is measured and recorded as At in intervals of every 5 min. The removal efficiency of methyl orange can be calculated using Eq. 1.

$$ {\text{Removal efficiency }}\left( {\text{\% }} \right) = \frac{{A_{0} - A_{t} }}{{A_{0} }} \times 100 $$
(1)

Design of experiment using three variables including irradiation time, photocatalyst concentration and pH was used to conduct statistical studies. For this purpose design-expert version 7.0.0 software is used for statistical analysis and also to provide a model that has the ability to predict changes in pollutant removal efficiency using different photocatalysts. The model proposed by statistical analysis is a polynomial model that includes all single parameters and their binary and triple interactions. Also, the influence of single parameters (such as UV radiation time, photocatalyst concentration and pH), binary and triple interaction of effective factors are studied using analysis of variance with a confidence level of 5%. Due to the importance of the accuracy of the results, all tests are performed at least three times and the average results are reported as the final value.

Results and discussion

Investigation of the interaction of two effective factors on the removal efficiency

The mutual influence between the two effective factors on the photocatalytic activity of the Fe3O4-TiO2 and MGO-TiO2 for pollutant removal is shown in Figs. 1 and 2, respectively. In order to check the mutual effect between two factors on the removal efficiency, the third factor is fixed at the lowest level, so it has no effect on the response. According to the curves presented in these two figures, the non-parallelism of the curves is quite evident, which indicates the significant influence of the investigated factors on each other. Therefore, all the mutual effects of two factors including (A–B), (A–C) and (B–C) on the photocatalytic activity of both photocatalysts are significant. Therefore, in the statistical models that are presented to predict the changes in the removal efficiency of methyl orange, they have a nonzero coefficient. The interaction effect between the irradiation time and the concentration of the Fe3O4–TiO2 and MGO-TiO2 on the removal efficiency is observed in Figs. 1a and 2a, respectively. It is quite clear that at constant pH, with increasing irradiation time at all levels of concentration of both photocatalysts, the amount of photocatalytic activity for photodegradation of methyl orange increases. The increasing trend can be attributed to the effect of the irradiation time on the excitation of valence layer electrons and their transfer to the conduction layer. Because the longer the photocatalyst is exposed to UV radiation, the number of produced electron–hole pairs increases. Considering that there is a direct relationship between the number of electron–hole pairs and active radicals that play the role of pollutant oxidizers, therefore, increasing the time of irradiating the suspension with UV rays leads to an increase in pollutant removal efficiency (Ghaderi et al. 2015; Liu et al. 2012).

Fig. 1
figure 1

The interaction of two factors affecting the removal efficiency of methyl orange using the Fe3O4-TiO2

Full size image
Fig. 2
figure 2

The interaction of two factors affecting the removal efficiency of methyl orange using the MGO-TiO2

Full size image

Examining the effect of photocatalyst concentration also shows that with increasing the concentration of both photocatalysts from 0.05 to 0.2%wt, the removal efficiency increases significantly. The main reason for this increase can be caused by the increase in the active surface of the photocatalyst, which can be exposed to radiation. In fact, increasing the surface area increases the number of valence layer electrons that have the ability to be exposed to UV rays. As a result, it has a positive effect on the number of electron–hole pairs and oxidizing hydroxyl radicals (OH.) (Abbasi 2023b; Wang et al. 2022).

The interaction between the irradiation time and the pH of the suspension containing the Fe3O4-TiO2 and MGO-TiO2 is shown in Figs. 1b and 2b, respectively. As can be seen, with the increase of pH from 3 to 11, the changes in the removal efficiency of methyl orange using both photocatalysts do not show a uniform behavior. So that initially, with the increase in pH from acidic to neutral conditions, the removal efficiency decreases, and then the removal efficiency increases with the change of the environment of the suspension to alkaline conditions. Therefore, the lowest and highest removal efficiency of methyl orange using both photocatalysts occurs at pH = 3 and pH = 7, respectively. The greater photocatalytic decomposition of methyl orange in acidic conditions can be due to the large number of free hydrogen ions (H+) in the suspension. Therefore, the existing ions tend to absorb the created electrons and this causes the production of H. radicals. Considering that radical species are very active to participate in the reaction, they quickly react with dissolved oxygen in the suspension and produce hydroxyl radicals (OH.) (Abbasi et al. 2021; Abbasi and Hasanpour 2017). Another effective factor in the changes in the photocatalytic degradation rate of methyl orange with pH is the surface charge of the photocatalyst. Therefore, with the increase in the surface charge of the photocatalysts, their dispersion rate in the suspension containing the pollutant also increases, and this significantly improves the photocatalyst's exposure to UV rays and this has a significant positive effect on the level of electron excitation as well as pollutant oxidizing radicals (Abbasi 2021b). Therefore, in neutral conditions where the surface of the photocatalyst has the lowest amount of surface charge, the removal efficiency of methyl orange also decreases significantly (Yuan and Xu 2010). The comparison of the removal efficiency of methyl orange using the Fe3O4-TiO2 and MGO-TiO2, which can be seen on the vertical axis of the curves in Figs. 1 and 2, shows that under the same laboratory conditions, the photocatalytic activity of the MGO-TiO2 is significantly increased compared to the Fe3O4-TiO2. The difference in the removal efficiency of methyl orange using these two photocatalysts can be related to the change in their structure. As it is known, their structural difference is attributed to GO. The presence of GO in the structure of the MGO-TiO2 causes a more uniform distribution of titania magnetic nanoparticles and reduces their accumulation, and this causes an increase in the active surface that is exposed to UV rays. The increase in the active contact surface area of the photocatalyst implicitly leads to the growth of electron–hole pairs as well as methyl orange decomposing radicals.

Investigation of significance of the factors based on analysis of variance

The results of analysis of variance to determine the effective factors and their importance on the decomposition of methyl orange using the Fe3O4-TiO2 and MGO-TiO2 are presented in Tables 1 and 2, respectively. Considering that the level of confidence in this analysis is set at 5%, therefore, each of the main factors or their binary and triple interaction has a p-value of less than 5%, it is known as an effective factor on the removal efficiency. As can be seen, the p-value of all single parameters including irradiation time (A), photocatalyst concentration (B) and pH (C), binary interactions (A–B, A–C and B–C) as well as triple interaction (A–B–C) are less than 5%. Therefore, their significance is confirmed with a confidence of over 95%, which means that all the investigated factors and the interaction between them play a significant role in the statistical model that is presented to estimate the MO removal efficiency. Despite the effectiveness of all the factors, the importance and impact of the mentioned factors are different, which is determined according to the F-value presented in the analysis of variance tables, so that the parameters that are more effective on the response have a higher F-value (Namvar-Mahboub and Pakizeh 2014). Therefore, by comparing the F-value of Tables 1 and 2, it can be seen that individual parameters have the highest F-value and their triple interaction has the lowest F-value. It can also be concluded considering that the highest F-value belongs to the irradiation time, so the changes in pollutant removal efficiency using both photocatalysts are more influenced by time. The degree of freedom of the models presented using both photocatalysts is equal to 71, which is equal to the sum of the degrees of freedom of individual factors and their interactions, so this confirms the P-value results.

Table 1 Analysis of variance results for decomposition efficiency using the Fe3O4@TiO2 as a photocatalyst
Full size table
Table 2 Analysis of variance results for decomposition efficiency using the MGO@TiO2 as a photocatalyst
Full size table

The results of variance analysis showed that the statistical models provided by both photocatalysts have a P-value less than 5%, so they are suitable for estimating the changes in the removal efficiency of methyl orange. In addition to P-value, there are several statistical parameters to ensure the adequacy of the models. One of the most important parameters is the correlation coefficient (R2). According to the values presented in Table 3, it can be seen that the value of this parameter for the Fe3O4@TiO2 and MGO@TiO2 is equal to 0.9953 and 0.9923, respectively. A value close to one of this parameter indicates that the model has a very good correlation between the experimental values and the values predicted by the statistical models. According to the results of Table 3, it can be seen that there is no significant difference between R2 and adjusted R2 (R2adj) values, so the absence of non-significant parameters in both models can be confirmed. Another effective statistical parameter in determining the quality of the models is the R2pred. A difference of less than 0.2 between parameters R2adj and R2pred is also suitable to ensure the accuracy of the models for estimating the changes in the removal efficiency of methyl orange. It is clear that in both models, the difference between these two parameters is less than 0.2, so the results of other statistical parameters are confirmed. Another statistical parameter to confirm the adequacy of the model is adequate precision, which shows the difference between the values estimated by the model and the average error. The value of this parameter in the best statistical models is more than 4 (Abbasi and Hasanpour 2017; Kazemi-Beydokhti et al. 2015). According to the results of Table 3, it can be seen that the value of adequate precision in the Fe3O4@TiO2 and MGO@TiO2 models is equal to 80.667 and 57,936, respectively. Considering that the value of this parameter is much higher than 4 in both models, therefore the proposed models can be used with confidence.

Table 3 Statistical parameters belonging to the provided models by Fe3O4@TiO2 and MGO@TiO2
Full size table

Statistical investigation of the adequacy of the model

The accuracy of analysis of variance results depends on establishing effective conditions in this analysis. One of the main assumptions is the normal distribution of residuals. The plots of the internally studentized residuals compared to the predicted ones for Fe3O4@TiO2 and MGO@TiO2 are shown in Figs. 3 and 4, respectively. These plots indicate that there is no clear relationship between internally studentized residuals and the predicted. Therefore, the irregular distribution observed in Figs. 3 and 4 confirms the normal distribution of the residuals. Another main and significant assumption in the analysis of variance is the independent distribution of the error, as well as the constancy of the variance of individual independent factors. Therefore, examining the variations in the behavior of the internally studentized residuals with respect to all the individual factors affecting the removal efficiency, including irradiation time, photocatalyst concentration, and pH, is a practical solution for verifying the mentioned hypothesis. The internally studentized residuals dependence on individual factors affecting the removal efficiency of methyl orange using the Fe3O4@TiO2 and MGO@TiO2 is depicted in Figs. 5 and 6, respectively. According to Fig. 5, it is quite evident that all the design points in the model presented for the Fe3O4@TiO2 as a photocatalyst are in the range of − 3 to + 3, and there are no outliers or deviations, while in Fig. 6, a very small number of design points are on the border or outside the range. However, due to the small number of mentioned points, it can be ignored. Therefore, the verifications confirm the stability of the variance of independent single factors in both models. Due to the fact that the effective assumptions in the analysis of variance are fully confirmed, therefore, the removal efficiency values suggested by both statistical models have an acceptable compatibility with the obtained laboratory results.

Fig. 3
figure 3

The plot of the internally studentized residuals compared to the predicted ones for Fe3O4@TiO2

Full size image
Fig. 4
figure 4

The plot of the internally studentized residuals compared to the predicted ones for MGO@TiO2

Full size image
Fig. 5
figure 5

The internally studentized residuals dependence on individual factors for Fe3O4@TiO2

Full size image
Fig. 6
figure 6

The internally studentized residuals dependence on individual factors for MGO@TiO2

Full size image

One of the practical methods used to normalize non-normal dependent variables is Box–Cox transformation. The importance of normalizing such responses is due to the need to establish it in order to obtain statistical results with high accuracy. Therefore, if the dependent variables do not have a normal distribution, the Box–Cox transformation is able to create an almost normal distribution, and this improves the results of statistical analysis. Box–Cox transformation includes a power function that corrects asymmetry of variables, different variances, and also nonlinearity between variables. Therefore, this transformation is very useful for converting a non-normal variable into a variable that has a normal distribution. Therefore, the Box–Cox diagram is the most suitable solution for determining the acceptable power function that is applied to the response. The lower CL (pink) is the lower confidence limit for the 95% confidence interval for λ. The upper CL (red) is the upper confidence limit for the interval. Both confidence limits are displayed on the plot as dotted vertical lines. The lowest value that is determined on the vertical axis of the Box–Cox curve corresponds to the best \(\lambda\) value that has the minimum residual sum of squares in the propose model. Therefore, the degree of the proposed model can be controlled and confirmed using this curve. The Box–Cox diagram of methyl orange removal efficiency using the Fe3O4@TiO2 and MGO@TiO2 can be seen in Figs. 7 and 8, respectively. According to these curves, it is clear that in both curves, the optimal value and the current value of \(\lambda\) are almost equal, so there is no need to convert for any of the responses, and the degree of the proposed model can be confirmed.

Fig. 7
figure 7

Box–Cox plot for the removal efficiency of methyl orange using the Fe3O4@TiO2

Full size image
Fig. 8
figure 8

Box–Cox plot for the removal efficiency of methyl orange using the MGO@TiO2

Full size image

Conclusions

Magnetic photocatalysts based on two-dimensional nanostructures are used to remove methyl orange as an organic pollutant, and the dependence of their efficiency on irradiation time, photocatalyst concentration and pH of the suspension is investigated. Due to the importance of photocatalyst removal after pollutant decomposition process, magnetic photocatalysts were separated from the examined wastewater by applying a strong magnetic field. The findings show a significant change in removal efficiency with independent single factors, their binary and triple interaction at 5% level of probability. Also, the statistical models presented by both photocatalysts have a special ability to estimate the removal efficiency of methyl orange with an accuracy of over 99%. Considering that the importance of the factors was determined based on the analysis of variance, therefore, the statistical graphic curves fully confirmed the assumptions affecting the validity of the analysis, which include the normal distribution of the residuals.