Response surface design for removal of Cr(VI) by hydrogel-supported sulfidated nano zero-valent iron (S-nZVI@H)

As a metal pollutant, chromium (Cr) comes from industrial processes such as tanning, electroplating, metallurgy, and textile (Selvi et al. 2001). Large quantities of Cr-containing wastewater are drained into rivers, resulting in contamination of soil and groundwater. Cr usually exists in the environment in the form of trivalent chromium (Cr(III)) and hexavalent chromium (Cr(VI)) (Fendorf 1995). The toxicity of Cr(VI) is higher than that of Cr(III) mainly because of its higher solubility, fluidity, and bioavailability (Kotaś & Stasicka 2000). Cr(VI) can increase the risk of cancer by entering the body through direct contact with the skin or inhalation (Gheju 2011; Zhu et al. 2017). In order to effectively control the pollution of Cr, some methods of removing Cr(VI) have been developed, including activated carbon adsorption technology (Mohan & Pittman Jr 2006), polymer membrane filtration technology (Aroua et al. 2007), ion exchange technology (Lin & Kiang 2003), electrochemical precipitation technology (Kongsricharoern & Polprasert 1995). In order to further improve the removal efficiency of Cr(VI) and reduce the impact of the treatment processes on the environment, it is necessary to develop a green material with high removal efficiency, low cost, and little secondary pollution.

In recent years, nanomaterials have been widely applied in the removal of pollutants from water (Eskandari et al. 2018; Galdames et al. 2020; Li et al. 2021; Tsade et al. 2021). Among them, nano zero-valent iron (nZVI) has been widely used in the remediation of groundwater and soil pollution due to its high specific surface area and reactivity (Zhang 2003). Nevertheless, nZVI particles have a number of limitations that affect their reaction with contaminants. (1) nZVI easily aggregates based on the action of van der Waals force and magnetic force, and eventually reduces the migration in aquifer (Phenrat et al. 2007; Israelachvili et al. 2008; Li et al. 2017b). (2) nZVI particles with extremely high reactivity can react with other substances in the water in addition to the target contaminants, resulting in poor selectivity for pollutants (Cai et al. 2020; He et al. 2020). (3) When using nZVI in practice, a passivation layer can easily form on the surface of nZVI prohibiting the Fe⁰ core to react with pollutants, thus lowering the reactivity (Sarathy et al. 2008). The bare nZVI therefore needs to be modified to improve these limitations. A common approach is to load the nZVI onto inorganic carriers such as bentonite (Chen et al. 2011b) and biochar (Su et al. 2016). The author applied diatomite to support nZVI to remove Cr(VI) from water and achieved better dispersion and lower oxidation degree of nZVI with high removal rate (Zhang et al. 2019). In addition, the coating of polymer on the surface of nZVI is also an effective method. Chen et al. (2011a) conducted an experimental study on the removal of tetracycline from polluted water by modified nZVI with polyvinylpyrrolidone (PVP-K30). Another method for the modification of nZVI is adding transition metal to nZVI to form a bimetallic iron base. Zhu et al. (2016) reported the method of adding Cu metal to nZVI to repair Cr(VI) in groundwater.

The higher reactivity, selectivity, and lifetime of nZVI based on sulfidation are considered as promising characteristics (Xie & Cwiertny 2010; Fan et al. 2016a, 2017; Xu et al. 2020b). At present, there are two commonly used methods for the preparation of sulfide nanoparticles. Kim et al. (2011) used a facile one-pot synthesis process by adding a mixture of dithionite and sodium borohydride (NaBH₄) to a solution of ferric iron (Fe³⁺) to synthesize Fe/FeS nanoparticles, achieving high reduction rate for trichloroethylene (TCE). Another method of sulfidation is to reduce ferrous iron (Fe²⁺) with NaBH₄ and then add Na₂S, commonly known as the two-step synthesis method (Rajajayavel & Ghoshal 2015). Sulfurization breaks the application limit of nZVI. Fan et al. (2016a) reported that sulfidation of nZVI enhances the rate of dechlorination of TCE, indicating that the corrosion reaction with water of nZVI is suppressed. In view of improving lifetime of S-nZVI, Xie & Cwiertny (2010) proposed that the reduction ability of aged nZVI could be restored by using dithionite as a reducing agent to sulfurize the nZVI to prolong its lifetime. In addition, sulfidation significantly increases the crystal domain size and lattice spacing of iron. The larger lattice constant is beneficial for the electron transfer, thus improving the reactivity and selectivity for the reduction of contaminants (Xu et al. 2020a). These achievements are impressive and indicate that S-nZVI has great potential for the removal of pollutants from water (Su et al. 2015; Li et al. 2016, 2017a, 2020).

In order to improve the aggregation phenomenon and anti oxidation property of S-nZVI, using a porous material as the supported carrier of nanoparticles is an effective choice. Chen et al. (2020) removed TCE from groundwater by loading S-nZVI on biochar (BC). The promotion of removal and degradation of TCE benefited from the high adsorption capacity of BC and the high reduction capacity of S-nZVI. Other researchers have also applied inorganic porous material to support S-nZVI for the removal of heavy metal (Su et al. 2015; Qu et al. 2021) and toxic dye (Yang et al. 2018). Based on its strong response to environmental stimulation, biocompatibility, swelling property, and elasticity, hydrogels have been used in drug transportation, eyeglasses manufacturing, biosensor development, artificial muscles, and other aspects (Manetti et al. 2002; Sahiner et al. 2006; Satarkar & Zach Hilt 2008; Fuhrer et al. 2009; Jain et al. 2017). In this study, organic hydrogels with non-toxic, environmentally friendly, cheap and porous properties were selected as the carrier of S-nZVI. Hydrogel has super function of absorbing contaminants (Jin et al. 2009; Sahiner & Ilgin 2010) due to the three-dimensional (3D) network formed by cross-linking of polymer chains. In addition, hydrogels also have hydrophilic functional groups such as COOH, OH, CONH₂, and NH₂ in the structure, where negatively charged COOH can be combined with positively charged metal ions (Cu²⁺ and Fe²⁺) in aqueous solution by electrostatic forces (Chauhan et al. 2006; Sahiner et al. 2010a). Consequently, hydrogels can be used as reaction containers for in-situ reduction of metal nanoparticles (Sahiner 2006; Sahiner et al. 2010b). The metal nanoparticles are uniformly dispersed in the hydrogel network system, thus avoiding the serious aggregation of nanoparticles. Furthermore, hydrogels can be used as a protective structure to reduce the oxidation degree of nanoparticles to a certain extent. Obviously, it is worth expecting to combine S-nZVI with hydrogel (S-nZVI@H) to form a novel material to fully amplify the super reduction ability of S-nZVI particles, and then repair the pollution of Cr(VI) in groundwater.

Response surface methodology (RSM) is a design method that combines mathematics and statistics to establish polynomial models based on experimental data (Bezerra et al. 2008). After determining independent variables, designing experiments, data statistics, and model validation, the goal of determining the optimal independent variable conditions is achieved. Box–Behnken design (BBD) is one of the optimization analysis methods in the quadratic response surface design method, which can show the influence of interaction between multiple variables on the response value and avoid the combination of all variables in extreme circumstances (Ferreira et al. 2007; Bezerra et al. 2008).

In this study, nZVI@H and S-nZVI@H were synthesized and characterized for the removal of Cr(VI). Scanning electron microscopy (SEM) and Fourier transform infrared spectroscopy (FTIR) were applied to survey the surface morphology, dispersion phenomenon, and structure of original hydrogels, nZVI, nZVI@H, and S-nZVI@H. Comparison experiments were performed to research the removal efficiency of nZVI@H and S-nZVI@H. Considering the removal efficiency, S-nZVI@H was selected as the material for the response surface experiment design. The S-nZVI@H dose, initial Cr(VI) concentration, contact time, and initial pH were taken as independent variables to explore the influence of the interaction of four independent variables on the Cr(VI) removal efficiency to acquire the optimal conditions for removal Cr(VI).

Analytical pure levels of chemicals and reagents were used in this study. Ferrous sulfate (FeSO₄·7H₂O), sodium hydrosulfite (Na₂S₂O₄), and sodium borohydride (NaBH₄) were purchased from Tianjin Fuchen Chemical Reagent Factory (Tianjin, China). The solution required in the experiment was mixed with deionized water. Hydrochloric acid (HCl, 0.1 mol/L) and sodium hydroxide (NaOH, 0.1 mol/L) were used to adjust the pH of the solution.

The acquired S-nZVI@H composites were immersed in deionized water for 2 h in the presence of nitrogen to remove the residual NaBH₄ and Na₂S₂O_4. Finally, the fresh S-nZVI@H were dried in a freeze drier (FD-1 A-50) for 24 h.

The surface morphology and microscopic structure of the hydrogel, nZVI, nZVI@H, and S-nZVI@H were observed with SEM (JSM-7610F-Plus, JEOL Ltd, Japan). The functional groups of the hydrogel, nZVI, nZVI@H, and S-nZVI@H were analyzed using a FTIR (IRAffinity-1S, Shimadzu, Japan).

In this study, the experimental data are coded and analyzed with Design-Expert 8.0.6 Trial. Four independent variables and their levels are listed in Table 1. The experimental design procedure and results of Cr(VI) removal are shown in Table 2. Twenty-nine groups of experiments are designed, which includes five parallel experiments. Analysis of variance (ANOVA) is used to determine the fitting degree between the values obtained from the experiment and the predicted response values calculated by the regression mode (Silveira et al. 2017). The statistical significance is evaluated by the coefficient of determination R² (Khosravi & Arabi 2016).

FTIR was used to reveal the chemical functional groups of PANa-PAM hydrogel, nZVI, nZVI@H, and S-nZVI@H (Figure 1). The broad peak around 3,420 cm⁻¹ is assigned to hydroxyl groups (Dong et al. 2018). The peak near 2,370 cm⁻¹ is the saturated zone of CO₂ (Dong et al. 2018). The peak at 1,310 cm⁻¹ is attributed to C–C stretching (Tang et al. 2008). The absorption peaks at 1,670 cm⁻¹, 1,570 cm⁻¹, and 1,390 cm⁻¹ are assigned to C = O stretching, the N–H bending and –C–N– deformation of the –CONH₂ group in the hydrogel, respectively (Tang et al. 2008; Dong et al. 2019). Likewise, the absorption peak at around 1,420 cm⁻¹ is attributed to the symmetric stretching vibration of the –COO⁻ anions in the hydrogel (Tang et al. 2008). These absorption peaks of the hydrogel also present in nZVI@H and S-nZVI@H, indicating that nZVI and S-nZVI have been successfully supported by hydrogels. The adsorption peak at about 470 cm⁻¹ corresponds to Fe–O stretches due to the oxidation of nZVI in the air (Zhang et al. 2011). Occurrence of less obvious Fe–O stretches in nZVI@H and S-nZVI@H may be caused by the fact that the degree of oxidation of these two materials is modestly inhibited. The new band spectrum at 1,100 cm⁻¹ and 830 cm⁻¹ of S-nZVI@H compared with the infrared images of nZVI@H further indicates that nZVI has been successfully sulfidation (Chernyshova 2003; Reyes-Bozo et al. 2015).

SEM was applied to elucidate the surface morphology and structure of PANa-PAM hydrogel, nZVI, nZVI@H, and S-nZVI@H. The porous network structure of PANa-PAM hydrogel is clearly observed in Figure 2(a). The spherical structure of bare nZVI is shown in Figure 2(b). It can be observed that the diameter of nZVI particles is less than 100 nm with aggregation. The surface morphologies of nZVI@H and S-nZVI@H are shown in Figure 2(c) and 2(d), respectively. The porous structure of the supported hydrogel retains a rougher surface. Both nZVI and S-nZVI particles are successfully loaded in the hydrogel as small bulges in granular form and their dispersion performance is relatively better than that of bare nZVI. The unique 3D structure of PANa-PAM hydrogel can effectively inhibit the aggregation of nanoparticles. It is obvious from Figure 2(d) that the size of S-nZVI particles distributed in the porous hydrogel is larger and has a uniform distribution, which may be due to the minimization of aggregation by electrostatic repulsion and steric hindrance in the case of sulfidation.

According to previous studies on S-nZVI, the reactivity and selectivity of S-nZVI with pollutants are greatly enhanced compared with bare nZVI (Fan et al. 2016b, 2017). A comparative experiment was conducted to further explore Cr(VI) removal efficiency of S-nZVI@H and nZVI@H under certain experimental conditions (initial Cr(VI) concentration of 20 mg/L, pH 6, nZVI@H and S-nZVI@H dosage of 6 g/L, and temperature of 25 °C). Figure 3 shows that the removal efficiency of nZVI@H and S-nZVI@H increases rapidly in the first 30 min. After 60 min, the Cr(VI) removal efficiencies of both reach more than 90%. The removal efficiency of S-nZVI@H is 24.89% higher than that of nZVI@H at 5 min. During the reaction of Cr(VI) and S-nZVI@H, the sulfur species FeS enhances the transfer of electrons from the core to the surface of nZVI, thus increasing the probability of Cr(VI) reduction (Li et al. 2018; Brumovský et al. 2021). Brumovský et al. (2021) verified that S-nZVI was more effective in removing Cr(VI) than nZVI in laboratory conditions. Additionally, Su et al. (2015) reported that S-nZVI had a higher surface area rather than bare nZVI, generating more metal ion adsorption binding sites. Accordingly, the reason why S-nZVI@H can remove Cr(VI) more quickly and effectively compared with nZVI@H is that sulfidation can improve the dispersity of nZVI and selectivity of nZVI to Cr(VI). Specifically, sulfidation of nZVI reduces the magnetic interaction between nZVI nanoparticles by the formation of an iron sulfide compound, as a result of ameliorating the aggregation. Moreover, the outer layer created by sulfidation decreases the direct reaction of nZVI with water, thus improving the electron selectivity for the reduction of Cr(VI) and extending its reactive lifetime. PANa-PAM hydrogel, in particular, provides a protective layer for the passivation of S-nZVI, and makes S-nZVI more evenly distributed in the hydrogel network, thus improving the reduction effectivity of Cr(VI). Therefore, the removal of Cr(VI) by S-nZVI@H was further analyzed by RSM.

The residual graph is drawn to verify whether the residual of the model has the characteristics of normality, independence, and constancy (Un et al. 2014). Figure 5(a) is the residual normal possibility diagram, which indicates that the distribution of residuals conforms to the characteristics of Gaussian distribution, thus verifying the normal distribution characteristics of residuals. The chronological diagram of residuals is shown in Figure 5(b). Residual values are distributed between −3.93 and 3.93 with irregular distribution characteristics, which shows the independence of residuals. If the model is correct and reliable, the residual does not exhibit a special structure and does not have a relationship with other variables (Mansouriieh et al. 2019). Figure 5(c) is the residual graph based on predicted values. The distribution of residuals within a certain range is also irregular, which shows the constancy of residuals. The smaller residual is when the residual data points are close to the zero line. The more residual data points close to the zero line, the better the fitting degree between the predicted value and the experimental value. These three residual diagrams show that the quadratic polynomial model established by RSM has certain reliability and can effectively predict the experimental results.

Based on the quadratic polynomial model, the 3D response surface map can visualize the relationship between the interaction of the two independent variables and the Cr(VI) removal efficiency while keeping the other two variables fixed. Figure 6 shows the effects of the interaction between two of the four independent variables (S-nZVI@H dose (A), initial Cr(VI) concentration (B), contact time (C), and initial pH (D)) on the removal efficiency of Cr(VI) by S-nZVI@H. In each response surface diagram, under certain two variables, the highest point of the surface is the best reaction condition under which the maximum removal efficiency can be obtained.

The interactive impact of the combination of S-nZVI@H dose and initial Cr(VI) concentration on Cr(VI) removal efficiency is presented in Figure 6(a). It shows that the removal efficiency increases with decreasing initial Cr(VI) concentration and increasing S-nZVI@H dose. S-nZVI@H can provide more surface-active sites, improving the reduction ability of Cr(VI). The higher the initial concentration of Cr (VI), the more surface-active sites of S-nZVI can be occupied, resulting in the lower removal efficiency of Cr(VI) (Khosravi & Arabi 2016). At S-nZVI@H dose 7.72 g/L and initial Cr(VI) concentration 10.10 mg/L, a maximum Cr(VI) removal of 91.36% is determined.

The reciprocal effect of contact time and S-nZVI@H dose on Cr(VI) removal efficiency is shown in Figure 6(b). As S-nZVI@H dose and contact time increases, the removal efficiency of Cr(VI) improves. The removal efficiency of Cr(VI) increases rapidly at the initial contact time, which indicates that the reduction of Cr(VI) is predominant (Ponder et al. 2000). However, with the further extension of contact time, the Cr(VI) removal efficiency increases slowly, which is due to the accumulation of Fe(III)–Cr(III) hydroxides on the surface of S-nZVI@H hindering the reaction between Cr(VI) and S-nZVI (Wu et al. 2015; Li et al. 2017c). In addition, it can be clearly observed that the curvature of the surface is relatively large, which indicates that the interaction of contact time and S-nZVI@H dose has a great influence on the Cr(VI) removal efficiency. A maximum Cr(VI) removal (98.95%) is observed at S-nZVI@H dose of 8.76 g/L and contact time of 39.57 min.

Figure 6(c) shows the combined effect of initial Cr(VI) concentration and contact time on removal of Cr(VI). With the increase of contact time and the decrease of initial Cr(VI) concentration, the removal efficiency of Cr(VI) increases continuously. The extended contact time provides sufficient reaction time for the reduction and adsorption of Cr(VI) by S-nZVI@H. Under the condition of higher initial Cr(VI) concentration, the removal efficiency of Cr(VI) is reduced. This can be attributed to the limited number of active sites provided by a fixed amount of S-nZVI@H. At initial Cr(VI) concentration 10.02 mg/L and contact time 36.39 min, a maximum Cr(VI) removal of 93.63% is determined.

The plot for interactive effect of S-nZVI@H dose and initial pH (Figure 6(d)) shows that increasing initial pH results in a decreasing removal of Cr(VI); conversely, this increases with the increase of S-nZVI@H dose. This is because, under alkaline conditions, nZVI appears electronegative; under acidic conditions, nZVI appears electropositive (Gong et al. 2017). Cr(VI) exists mainly in the form of in acidic conditions, and only in the form of in alkaline solutions (Lv et al. 2011). Obviously, in alkaline solution, negatively charged nZVI and are more likely to generate electrostatic repulsion, which inhibits their interaction, leading to the decrease of Cr(VI) removal. A maximum Cr(VI) removal (90.33%) is determined at S-nZVI@H dose 8.30 g/L and initial pH 5.3.

The interactive influence of initial Cr(VI) concentration and initial pH on Cr(VI) removal is shown in Figure 6(e). Under the dual function of high pH value and high initial Cr(VI) concentration, the removal efficiency of Cr(VI) is reduced. Moreover, it can be observed that the curvature of the response surface is small, which indicates that the interaction between initial Cr(VI) concentration and initial pH has a slight influence on the removal efficiency of Cr(VI). At initial Cr(VI) concentration 10.07 mg/L and initial pH 3.72, a maximum Cr(VI) removal of 87.72% is determined.

The interaction of contact time and initial pH on Cr(VI) removal is shown in Figure 6(f). It may be noted that the trend of Cr(VI) removal efficiency is consistent with the change of contact time and is contrary to the initial pH. Similarly, the contact time and the initial pH play the same role as stated above. A maximum Cr(VI) removal (89.8%) is determined at contact time of 37.61 min and initial pH of 5.38.

As an optimization technology, RSM needs to be verified for its adaptability (Bezerra et al. 2008). To evaluate the validity and feasibility of the RSM model, three parallel experiments were carried out under the conditions of independent variables corresponding to the maximum Cr(VI) removal efficiency determined by the model (Table 4). The predicted optimum removal conditions obtained from RSM are S-nZVI@H dose (9.46 g/L), initial Cr(VI) concentration (30 mg/L), contact time (40.7 min), and initial pH (5.27). The relative error of prediction between the mean value of the experimental (97.75%) and predicted (99.76%) removal efficiency was calculated as 2.07%, which showed that the quadratic polynomial model can well predict the removal of Cr(VI) by S-nZVI@H (Ashrafi et al. 2018).

In this study, two materials of nZVI@H and S-nZVI@H are successfully synthesized for Cr(VI) removal. With better removal effects confirmed by SEM and FTIR characterizations and comparable experiments, S-nZVI@H is selected as the material designed by RSM to further explore its removal efficiency of Cr(VI). Furthermore, BBD experimental design is used to reveal the influence of the interaction of four independent variables on the Cr(VI) removal efficiency and the predicted optimum removal conditions are: S-nZVI@H dose 9.46 g/L, initial Cr(VI) concentration 30 mg/L, contact time 40.7 min, and initial pH 5.27. Additionally, the S-nZVI@H dose is the key factor affecting the removal of Cr(VI). The variance analysis and residual analysis of the experimental data are carried out by quadratic polynomial model. The analysis results show that the correlation between the experimental values and the predicted values is high (F = 118.73, P < 0.0001, R² = 0.9916), which corroborates the quality and reliability of the model. To verify the validity and feasibility of the model, three parallel experiments are carried out under the predicted optimum removal conditions. The removal efficiency of S-nZVI@H for Cr(VI) obtained from the experiment (97.75%) is very close to the predicted value of the model (99.76%). In conclusion, BBD based on RSM can be used to statistically optimize the experimental study of Cr(VI) removal by S-nZVI@H.

The authors thank the National Natural Science Foundation of China (No. 41907176) for supporting this project.

All relevant data are included in the paper or its Supplementary Information.