Hierarchical Nanoflowers of MgFe2O4, Bentonite and B-,P- Co-Doped Graphene Oxide as Adsorbent and Photocatalyst: Optimization of Parameters by Box–Behnken Methodology

In the present study, nanocomposites having hierarchical nanoflowers (HNFs) -like morphology were synthesized by ultra-sonication approach. HNFs were ternary composite of MgFe2O4 and bentonite with boron-, phosphorous- co-doped graphene oxide (BPGO). The HNFs were fully characterized using different analytical tools viz. X-ray photoelectron spectroscopy, scanning electron microscopy, energy dispersion spectroscopy, transmission electron microscopy, X-ray diffraction, vibrating sample magnetometry and Mössbauer analysis. Transmission electron micrographs showed that chiffon-like BPGO nanosheets were wrapped on the MgFe2O4-bentonite surface, resulting in a porous flower-like morphology. The red-shift in XPS binding energies of HNFs as compared to MgFe2O4-bentoniteand BPGO revealed the presence of strong interactions between the two materials. Box–Behnken statistical methodology was employed to optimize adsorptive and photocatalytic parameters using Pb(II) and malathion as model pollutants, respectively. HNFs exhibited excellent adsorption ability for Pb(II) ions, with the Langmuir adsorption capacity of 654 mg g−1 at optimized pH 6.0 and 96% photocatalytic degradation of malathion at pH 9.0 as compared to MgFe2O4-bentonite and BPGO. Results obtained in this study clearly indicate that HNFs are promising nanocomposite for the removal of inorganic and organic contaminants from the aqueous solutions.


Introduction
There is a surging research interest in the fabrication of nanocomposites of ferrites with promising adsorptive and photocatalytic properties. Different morphologies of nanocomposites viz. flower-like [1,2], hollow-yolk shell [3], core-shell [4] and nanorods [5] are reported. These nanostructures have some unique features such as fast charge transfer, the reduced recombination rate of charge carriers and high surface area which allows their utilization in adsorption and photocatalysis [6]. For enhancing the activity of these nanocomposites, the right choice of the components is an important aspect.
In recent years, graphene-based materials have been extensively studied for the environmental applications because of their large surface area and tunable optical properties [7]. Graphene oxide (GO), a functionalized form of graphene, has adsorptive and catalytic inertness because of its low chemical potential and restricted surface functionalities [8]. GO is reported to have a wide bandgap (3.5 eV) and can act as a photocatalyst only under UV light illumination [8]. This feature limits its practical applications for visible light assisted Table 1 lists all the information about the synthesized samples, including sample codes and their description. Table 1. Information regarding the sample codes.    The HNFs' XPS full scan spectrum confirmed the existence of all elements of BPGO and MGF-B, indicating its fabrication. However, all the characteristic peak positions were slightly changed due to different coordination environments, which in turn caused a change in the observed binding energy. Strong interactions between MGF−B and BPGO were indicated by the red shift in binding energies across all elements when compared totwopristine materials. During the catalytic processes, these interactions favour fast migration and charge carrier separation caused by light. Table 2 lists the atomic concentration of each element. The HNFs' XPS full scan spectrum confirmed the existence of all elements of BPGO and MGF-B, indicating its fabrication. However, all the characteristic peak positions were slightly changed due to different coordination environments, which in turn caused a change in the observed binding energy. Strong interactions between MGF−B and BPGO were indicated by the red shift in binding energies across all elements when compared totwopristine materials. During the catalytic processes, these interactions favour fast migration and charge carrier separation caused by light. Table 2 lists the atomic concentration of each element. Mössbauer spectrum of MGF-B (Figure 2f) displayed two sextets and one quadruple doublet. One outer sextet is assigned to Fe 3+ ion in octahedral (O h ) sites and other in tetrahedral (T d ) sites [28]. The quadrupole splitting value of both the sextets is very small, indicating the symmetrical environment around Fe 3+ ions. Additionally, a quadruple doublet was foundwhich was ascribed to the super paramagnetic relaxation of small particles with a particle diameter less than 15 nm. These small particles are magnetically isolated and do not participate in the long-range magnetic ordering [29,30].

Sample Code Sample Description
There are more Fe 3+ ions in the Td sites, according to the relative area of the sextets (Table 3). On insertion of BPGO into MGF-B, the area of absorption attributing to the superparamagnetic part (doublet) decreased which may be due to the formation of nanoflowers, as they have large size (8.6 nm) as compared to MGF-B (4.0 nm). Size of NPs has a significant impact on the Mössbauer spectrum. The presence of broad sextets and doublet occurs when the relaxation rate and Larmor precession frequency are of the same order of magnitude in the case of larger NPs [31]. HNFs displayed a similar Mössbauer spectrum, confirming the presence of MgFe 2 O 4 .  FT-IR spectrum of HNFs ( Figure 3) confirmed the existence of different B-and Pcontaining functional groups, M-O bond and Si-O framework vibrations due to BPGO, MgFe2O4 and bentonite, respectively. The details are given in the Supplementary Text S1.
In GO, the XRD diffraction peaks at 2theta values of 10.9 • and 42.7 • can be indexed to (001) and (101) hkl planes of graphitic carbon, respectively ( Figure S1) [33]. After co-doping ( Figure S1), the fingerprint diffraction peak of GO at 2θ = 10.9 • disappeared. Low intensity and broad diffraction peaks at 2theta = 25.0 • and 43.6 • referring to (002) and (101) crystallographic planes, respectively, were observed. The diffraction peak at 2theta = 25.0 • was due to highly disordered graphitic carbon structure [34]. The peak observed at 2theta = 43.6 • was most probably due to the defects in the GO lattice.
In the HNFs (Figure 3b), the XRD peaks correspond to MGF-B and BPGO are retained, while their peak intensities are changed due to interactions between them. According to Bragg's equation, the d-spacing of most intense peak of MGF-B, BPGO and HNFs at 2theta ≈ 35.4 • , 25.0 • and 35.5 • was 0.25, 0.34 and 0.24 nm, respectively. The decrease in d-spacing of HNFs can be ascribed to the creation of more defective sites, which results in an increased number of smaller graphitic domains. Based on the Debye Scherrer's formula [35] and the peak at full width half-maximum of MGF-B (311) and HNFs (311) crystal plane, the average crystallite size was about 4.0 and 8.6 nm, respectively ( Table 2).

Morphological Studies
A TEM micrograph of MGF-B (Figure 4a) revealed the existence of spherical and agglomerated MgFe 2 O 4 NPs over the bentonite matrix. The agglomeration was due to the nanosized crystallites and magnetic character of the ferrite NPs. The average grain size was 30-35 nm. TEM image of BPGO showed folded/scrolled chiffon like nanosheets and twisted edges ( Figure 4b). After integrating MGF-B with BPGO nanosheets, the as-synthesized nanocomposite depicted nanoflower like microspheres with a 3D structure, produced from the interlayer Van der Waals forces and high surface energy of MGF-B (Figure 4c). In the presence of MGF-B, the nanostructures of BPGO gradually transform from the chiffon like nanosheet to nanoflower-like shapes. Chiffon-like BPGO nanosheets with wrinkled surface were wrapped on the MGF-B surface, producing a hierarchical nanohybrid with porous structure. Along with nanoflowers morphology, the nanocomposite having MGF-B particles on the BPGO surface was also observed (inset of Figure 4i). intensity and broad diffraction peaks at 2theta = 25.0° and 43.6° referring to (002) and (101) crystallographic planes, respectively, were observed. The diffraction peak at 2theta = 25.0° was due to highly disordered graphitic carbon structure [34]. The peak observed at 2theta = 43.6° was most probably due to the defects in the GO lattice.
In the HNFs (Figure 3b), the XRD peaks correspond to MGF-B and BPGO are retained, while their peak intensities are changed due to interactions between them. According to Bragg's equation, the d-spacing of most intense peak of MGF-B, BPGO and HNFs at 2theta ≈ 35.4°, 25.0° and 35.5°was 0.25, 0.34 and 0.24 nm, respectively. The decrease in d-spacing of HNFs can be ascribed to the creation of more defective sites, which results in an increased number of smaller graphitic domains. Based on the Debye Scherrer's formula [35] and the peak at full width half-maximum of MGF-B (311) and HNFs (311) crystal plane, the average crystallite size was about 4.0 and 8.6 nm, respectively ( Table 2).

Morphological Studies
A TEM micrograph of MGF-B (Figure 4a) revealed the existence of spherical and agglomerated MgFe2O4 NPs over the bentonite matrix. The agglomeration was due to the nanosized crystallites and magnetic character of the ferrite NPs. The average grain size was 30-35 nm. TEM image of BPGO showed folded/scrolled chiffon like nanosheets and twisted edges ( Figure 4b). After integrating MGF-B with BPGO nanosheets, the assynthesized nanocomposite depicted nanoflower like microspheres with a 3D structure, produced from the interlayer Van der Waals forces and high surface energy of MGF-B (Figure 4c). In the presence of MGF-B, the nanostructures of BPGO gradually transform from the chiffon like nanosheet to nanoflower-like shapes. Chiffon-like BPGO nanosheets with wrinkled surface were wrapped on the MGF-B surface, producing a hierarchical nanohybrid with porous structure. Along with nanoflowers morphology, the nanocomposite having MGF-B particles on the BPGO surface was also observed (inset of Figure 4i).  A SEM micrograph of MGF-B (Figure 4d) confirmed that MgFe 2 O 4 NPs were incorporated into the bentonite layers. Similar micrographs have been observed upon the insertion of polypropylene into the layers of bentonite clay [19]. The SEM micrographs of BPGO depict a translucent and rippled silk waves with numerous crinkles possibly due to dopants and oxygeneous groups. These wrinkled nanosheets adjust themselves physically to adapt to MGF-B (Figure 4e). SEM micrographs of HNFs displayed sponge-like morphology with structure like nanoflowers (Figure 4f).
The atomic and weight percentage of all the elements present in the synthesized materials was depicted via energy-dispersive spectroscopy (EDS) (Figure 4g-i). They confirmed the presence of different elements in MGF-B, BPGO and HNFs. The peak for B was not observed in the EDS spectra as lighter elements does not possess Bremsstrahlung scattering [36]. The elemental composition of each element found in the synthesized materials is listed in Table 2.
The Brunauer-Emmett-Teller (BET) surface area of HNFs calculated using the nitrogen adsorption-desorption technique was 187.56 m 2 g −1 which was higher than magnesium ferrite-bentonite (87.10 m 2 g −1 ) and BPGO (162.8 m 2 g −1 ). This may be ascribed to the inhibited agglomeration of pristine MGF-B in the presence of BPGO. The BPGO and HNFs followed type IV isotherm with H3 hysteresis loops indicating the existence of mesoporous particles (diameters 2-50 nm) according to IUPAC classification, whereas MGF-B indicates type IV isotherm with a H4 typical hysteresis curve, confirming the existence of mesoporous particles with internal slit-like pores [37] (Figure 5a). The pore size distribution curves of MGF-B, BPGO and HNFs (Figure 5a inset) exhibited sharp narrow peaks in the range of 3.11 nm to 3.89 nm ( Table 2). The pore volume of MGF-B, BPGO and HNFs was in the range of 0.07-0.29 cm 3 g −1 , respectively. Thus, the enlarged surface area and pore diameter of HNFs as compared to MGF-B can be ascribed to the presence of BPGO. It is a favourable feature for the adsorptive and photocatalytic process. A SEM micrograph of MGF-B ( Figure 4d) confirmed that MgFe2O4 NPs were inco porated into the bentonite layers. Similar micrographs have been observed upon the i sertion of polypropylene into the layers of bentonite clay [19]. The SEM micrographs BPGO depict a translucent and rippled silk waves with numerous crinkles possibly d to dopants and oxygeneous groups. These wrinkled nanosheets adjust themselves phy ically to adapt to MGF-B (Figure 4e). SEM micrographs of HNFs displayed sponge-li morphology with structure like nanoflowers ( Figure 4f).
The atomic and weight percentage of all the elements present in the synthesiz materials was depicted via energy-dispersive spectroscopy (EDS) (Figure 4g-i). Th confirmed the presence of different elements in MGF-B, BPGO and HNFs. The peak f B was not observed in the EDS spectra as lighter elements does not possess Bremsstra lung scattering [36]. The elemental composition of each element found in the synth sized materials is listed in Table 2.
The Brunauer-Emmett-Teller (BET) surface area of HNFs calculated using the n trogen adsorption-desorption technique was 187.56 m 2 g −1 which was higher than ma nesium ferrite-bentonite (87.10 m 2 g −1 ) and BPGO (162.8 m 2 g −1 ). This may be ascribed the inhibited agglomeration of pristine MGF-B in the presence of BPGO. The BPGO an HNFs followed type IV isotherm with H3 hysteresis loops indicating the existence mesoporous particles (diameters 2-50 nm) according to IUPAC classification, where MGF-B indicates type IV isotherm with a H4 typical hysteresis curve, confirming the e istence of mesoporous particles with internal slit-like pores [37] (Figure 5a). The po size distribution curves of MGF-B, BPGO and HNFs (Figure 5a inset) exhibited sha narrow peaks in the range of 3.11 nm to 3.89 nm ( Table 2). The pore volume of MGF-BPGO and HNFs was in the range of 0.07-0.29 cm 3 g −1 , respectively. Thus, the enlarg surface area and pore diameter of HNFs as compared to MGF-B can be ascribed to t presence of BPGO. It is a favourable feature for the adsorptive and photocatalytic pr cess.  Magnetic field dependent hysteresis curves of the MGF-B and HNFs are shown in Figure 6a. The values of saturation magnetization (M s ), coercivity (H c ) and remanence (M r ) are given in Table 2. This clearly indicated that a significant decline in M s value of HNFs (6.26 emu g −1 ) was observed after its fabrication with BPGO. Thus, fabrication of BPGO nanosheets on the surface of MGF-B altered the magnetic features as it decreased the M s value. Due to their ferrimagnetic nature, these materials showed a remarkable affinity for the external magnet. The decrease in H c value for the HNFs (70.65 Gauss) from MGF-B (138.00 Gauss) confirmed that less magnetic field was required to demagnetize it. This decrease showed enhancement in its soft magnetic feature. Magnetic field dependent hysteresis curves of the MGF-B and HNFs are shown in Figure 6a. The values of saturation magnetization (Ms), coercivity (Hc) and remanence (Mr) are given in Table 2. This clearly indicated that a significant decline in Ms value of HNFs (6.26 emu g −1 ) was observed after its fabrication with BPGO. Thus, fabrication of BPGO nanosheets on the surface of MGF-B altered the magnetic features as it decreased the Ms value. Due to their ferrimagnetic nature, these materials showed a remarkable affinity for the external magnet. The decrease in Hc value for the HNFs (70.65 Gauss) from MGF-B (138.00 Gauss) confirmed that less magnetic field was required to demagnetize it. This decrease showed enhancement in its soft magnetic feature.   UV-VIS diffuse reflectance (UV-Vis DRS) spectroscopy was used to determine the band gap of synthesized materials (Figure 6b). MGF-B, BPGO, and HNFs' optical band edges (Eg) were calculated to be 2.26, 2.40, and 2.10 eV, respectively. The chemical interactions between MGF-B and BPGO, which were supported by a decline in binding energy in XPS analysis, may be the reason of the HNFs' band edges narrowing. The absorbance of all the materials in the visible light region is of great significance in the area of visible light photocatalysis.
The photoluminescence emission spectra in the wavelength range of 200-900 nm was studied for MGF-B, BPGO and HNFs (Figure 6c). Emission peaks were observed at 430 nm for MGF-B and 720 nm for BPGO and 410, 720 nm for HNFs. Emission peaks at 410 and 430 nm were attributed to 3d 5 →3d 4 transitions in Fe 3+ and may be caused by radiative defects or surface imperfections. [38]. After its fabrication with BPGO, five-fold decrease in luminous intensity was observed, confirming higher quenching efficiency of HNFs over pristine materials. The findings also suggested that the transfer of electrons from the conduction band of the photo-excited MGF-B to BPGO sheets might slow down the photoinduced charge carrier recombination over pristine materials, resulting in the enhanced photo-catalytic degradation of organic pollutants.

Adsorption Experiments
HNFs were observed to be better adsorbent for Pb(II) ions than BPGO and MGF-B. Detailed explanation is given in Supplementary Text S3 and Figure S2a. HNFs combined the features of both MGF-B and BPGO, which enhanced their adsorption potential. The major factor for the enhancement in adsorption potential could be the increase in the surface area and porosity as depicted by BET analysis. It was further supported by DLS studies, which showed that the particle size of HNFs (130 nm) in solution was smaller than MGF-B (171 nm). The HNFs possessed appreciable Ms value, which led to their easy separation and reusability. Box-Behnken statistical analysis was applied to optimize the adsorption parameters for Pb(II) ions on HNFs. UV-VIS diffuse reflectance (UV-Vis DRS) spectroscopy was used to determine the band gap of synthesized materials (Figure 6b). MGF-B, BPGO, and HNFs' optical band edges (E g ) were calculated to be 2.26, 2.40, and 2.10 eV, respectively. The chemical interactions between MGF-B and BPGO, which were supported by a decline in binding energy in XPS analysis, may be the reason of the HNFs' band edges narrowing. The absorbance of all the materials in the visible light region is of great significance in the area of visible light photocatalysis.
The photoluminescence emission spectra in the wavelength range of 200-900 nm was studied for MGF-B, BPGO and HNFs (Figure 6c). Emission peaks were observed at 430 nm for MGF-B and 720 nm for BPGO and 410, 720 nm for HNFs. Emission peaks at 410 and 430 nm were attributed to 3d 5 →3d 4 transitions in Fe 3+ and may be caused by radiative defects or surface imperfections [38]. After its fabrication with BPGO, five-fold decrease in luminous intensity was observed, confirming higher quenching efficiency of HNFs over pristine materials. The findings also suggested that the transfer of electrons from the conduction band of the photo-excited MGF-B to BPGO sheets might slow down the photoinduced charge carrier recombination over pristine materials, resulting in the enhanced photo-catalytic degradation of organic pollutants.

Adsorption Experiments
HNFs were observed to be better adsorbent for Pb(II) ions than BPGO and MGF-B. Detailed explanation is given in Supplementary Text S3 and Figure S2a. HNFs combined the features of both MGF-B and BPGO, which enhanced their adsorption potential. The major factor for the enhancement in adsorption potential could be the increase in the surface area and porosity as depicted by BET analysis. It was further supported by DLS studies, which showed that the particle size of HNFs (130 nm) in solution was smaller than MGF-B (171 nm). The HNFs possessed appreciable M s value, which led to their easy separation and reusability. Box-Behnken statistical analysis was applied to optimize the adsorption parameters for Pb(II) ions on HNFs.

Statistical Analysis of Adsorption Studies
The randomised design matrix of the Box-Behnken statistical model with 4 factors, 3 central points, and 27 runs as well as the removal efficiency of the Pb(II) ions, are listed in Table 4 and their ANOVA results shown in Table 5. The present study's independent variable p-values were less than 0.05, proving that the experimental data may sufficiently characterize the proposed model derived using the Box-Behnken response surface method-ology. The predicted and experimental statistical results were found to agree, and the correlation coefficient (R 2 , 0.99) and adjusted R 2 (0.98) values were also quite high. The model's validity is further supported by the 0.86 p-value of lack-of-fit, which was found to be non-significant. This value was higher than the lowest limit of fit as recommended to be (0.05). The model's calculated value for adequate precision was 37.50, which supported a greater S/N ratio. Any process design that has a signal (response) to noise (deviation) ratio of >four is preferable, and the model can be moved around in the design space. According to Table 5    This expression demonstrates the empirical relationship between significant variables and Pb(II) removal (response). The plot ( Figure S2b) between normal probability and externally studentized residuals follows a normal distribution where the points lie on a straight line. The reliability of the assumptions and the independence of the residuals are indicated by this plot. The relationship between the actual and predicted values ( Figure S2c) shows that the experimental findings for this study arewell-accepted. All of these findings point to the proposed model's strong correlation and suitability for predicting the Pb(II) adsorption process utilizing the HNFs as an adsorbent.

Effect of pH, Adsorbent Dosage and Contact Time
The solution pH, adsorbent dosage and contact time played an vital role in adsorption of Pb(II) ions. In addition to affecting the charge on the adsorbent surface, the pH of the solution also affects the equilibrium between adsorption and desorption and the surface charge of Pb(II) ions. The results plotted in Figure 6d-f demonstrate that the percentage removal of Pb(II) ions increased progressively from 61% to 95% with rise in the solution pH from 2.0 to 6.0. At a lower solution pH, the surface adsorptive sites got protonated and became positively charged thus inducing an electrostatic repulsion for Pb(II) ions, whereas under alkaline conditions, the surface became more negative, providing the electrostatic attraction for Pb(II) ions. The pH ZPC studies ( Figure S2d) revealed surface charge on HNFs surface became zero at pH 4.3. Above pH 4.3, Pb(II) ions were strongly attracted towards the negatively charged surface resulting in maximum uptake of Pb(II) ions. Speciation of Pb(II) in aqueous solution using visual miniteq software showed that free ionic Pb(II) is the most predominated form (80-90%) up to pH 6.0, whereas Pb(OH) 2 (90-100%) dominates after this pH ( Figure S2e). Under alkaline conditions, Pb(II) started to form insoluble Pb(OH) 2 and the solubility of the metal hydroxides decreased at higher pH. These results were further supported by ξ-potential studies. The ξ-potential of HNFs at pH 6.0 was negative ( Figure S2f). The negative ξ-potential of HNFs was responsible for the electrostatic attraction between positively charged Pb(II) ions and negatively charge HNFs surface.
The adsorption of Pb(II) ions enhanced on increasing the nanoadsorbent dose up to 0.4 g L −1 due to more available surface area and binding sites for the removal of Pb(II) ions. After that, no increase was observed (Figures 6d and 7a) because of attainment of adsorption-desorption equilibrium. Adsorption was monitored as a function of contact time also. It followed three stages i.e., fast, slow and equilibrium stage (Figure 6e). The fast stage lasted for first 5 min, as the solute molecules freely attached to the available active sites. The slow stage persisted for up to 10 min. The equilibrium stage was attained after 30 min where insignificant difference in the percentage removal was observed. HNFs was responsible for the electrostatic attraction between positively charged Pb(II) ions and negatively charge HNFs surface. The adsorption of Pb(II) ions enhanced on increasing the nanoadsorbent dose up to 0.4 g L −1 due to more available surface area and binding sites for the removal of Pb(II) ions. After that, no increase was observed (Figures 6d and 7a) because of attainment of adsorption-desorption equilibrium. Adsorption was monitored as a function of contact time also. It followed three stages i.e., fast, slow and equilibrium stage (Figure 6e). The fast stage lasted for first 5 min, as the solute molecules freely attached to the available active sites. The slow stage persisted for up to 10 min. The equilibrium stage was attained after 30 min where insignificant difference in the percentage removal was observed. where C e stands for equilibrium concentration, q e and q t represents uptake capacity at equilibrium time "t" and at time "t", respectively.

Kinetic and Adsorption Isotherm Modelling
Comparative eValuation of linear and non-linear form of kinetic modelling was employed for determining the best fitting model. Linear pseudo second-order kinetic model (Figure 7b) provided higher 'R 2 ' value of 0.99 and lower statistical errors than its non-linear form (R 2 = 0.64) ( Figure S3d) suggested that the adsorption data followed the linear form of modelling. The better fitting of pseudo second-order kinetic model than pseudo-first order ( Figure S3a),intra-particle diffusion ( Figure S3b) and Elovich ( Figure S3c) specified the interaction of Pb(II) ions with two binding sites of the HNFs. Detailed explanation of kinetics results and parameters are given in Supplementary Text S4 and Table S1, respectively.
Adsorption isotherms, both linear and non-linear, were used to analyse the data from the adsorption equilibrium. The statistical functions shown in Table S2 were used to compare the two regressions. The linear Langmuir equation's goodness-of-fit over non-linear form was defined by its maximum "R 2 " (0.99) and lowest error values (Figure 7c). The separation factor 'R L ' was in the range of '0.0-1.0' depicting the adsorption of Pb(II) ions was favourable in nature. The q max value for Pb(II) ions was 745.4 mg g −1 and it was compared with other reported adsorbents (Table S3). Other adsorption isotherms are explained in detail in Supplementary Text S5, Figure S4 and Table S2. The decreasing order of goodness-offit of employed isotherms was as follows: Langmuir > Freundlich > Temkin > Dubinin-Radushkevich. Thus, Pb(II) ions interacted uniformly with the active sites of HNFs in the monolayered physical adsorption shown by the adsorption isotherm modelling.
Similar outcomes for the Pb(II) ions adsorption employing CaFe 2 O 4 -NGO nanocomposite were observed by Kaur et al. [22]. According to Buergisser et al. [39], natural adsorbents usually follownon-linear models. However, batch tests for the Cr (VI) adsorption over Rhizopus sp. conducted by Espinoza-Sánchez et al. [40] showed that linear regression provided a better fit. Yazdani et al. [41] showed the goodness-of-fit of the non-linear method for batch As (V) adsorption utilizing nano-TiO 2 /feldspar-embedded chitosan. However, in the current investigations, nanocomposite used linear regression and contained both natural and synthetic components. Thus, model fitting is dependent upon the nature of the adsorbent. Therefore, using both linear and non-linear modelling is essential for any adsorbent.

Thermodynamic and Regeneration Studies
The adsorption capacity of Pb(II) on HNFs increased with an increase in reaction temperature (Figures 6f and 7a) and could reach 94% at 50 • C. However, the adsorption capacity decreased above 50 • C. This may be due to the increase in average kinetic energy of ions with temperature leading to increase in adsorption as more number of ions interacted with the mesoporous surface of the nanostructure. After that, desorption dominates, leading to decline in their adsorption potential. The values of ∆G • was negative, indicating that the adsorption of Pb(II) on the HNFs was spontaneous ( Figure S5a). The absolute values of ∆G • increased as the temperature rose, indicating an increase in spontaneity. The endothermic nature of Pb(II) was illustrated by a positive ∆H • value, with an optimal temperature of 50 • C. The entropy change values (∆S • ) were positive, indicating that as the reaction temperature increased, the degree of freedom of the adsorption system also increased (Table S4).
The excellent recycling performance of adsorbent determines its application in practical environment. As seen in Figure S5b, after five adsorption-desorption cycles, the 89% removal efficiency was retained which confirmed reusability of HNFs. Moreover, the magnetic character facilitated their separation from the solution using an external magnetic field.

Effect of Co-Existing Cations and Anions
The adsorption capacity of the HNFs was tested in the presence of Pb(II), Cd(II), Zn(II) and Ni(II) ions ( Figure S5c). The observed trend for the removal was Pb(II) > Cd(II) > Zn(II) > Ni(II) and can be described by distribution and selectivity studies. The distribution coefficient and selectivity factor listed in Table S5 indicated that the adsorption of Pb(II) ions was preferred in the existence of other metal ions. The adsorption behaviour was also explained by ionic radii of the metal ions (Table S5). The larger the ionic size of ions, the lower is their hydrated radii hence have a greater affinity to get adsorbed. The adsorption capacity of HNFs for Pb(II), Cd(II), Zn(II) and Ni(II) ions decreased in the quaternary system than single system due to competition between metal ions for the adsorption sites available on HNFs surface.
The effect of presence of anions viz. Cl − , NO 3 − , SO 4 2− , CO 3 2− and PO 4 2− was also tested on the Pb(II) adsorption. The increase in anion concentration from 0.01 to 10.0 mM showed a negative effect on metal ions adsorption ( Figure S5d). However, this negative impact is maximum in the presence of PO 4 2− ions followed by CO 3 2− , SO 4 2− , NO 3 − and Cl − ions. This may be explained by the interaction of ligands with heavy metals to create ion pairs (Kaur et al., 2019). The availability of Pb(II) ions for adsorption decreases with increasing ion-pair stability. Here, PO 4 2− and Pb(II) ions form the most stable ion pair, which reduces the amount of Pb(II) that is adsorbed from the Pb 3 (PO 4 ) 2 solution.

Photocatalytic Degradation Studies
The photocatalytic potential of MGF-B, BPGO and HNFs using malathion as the model pollutant was eValuated. An insignificant increase in malathion degradation was observed from pH 1.0 to 3.0 ( Figure S6a), which indicated its excellent stability in an acidic medium as reported previously [42]. At basic pH, Organophosphate pesticides are more prone to oxidation and hydrolysis due to the presence of hydroxide ion (OH − ). Bavcon et al. [43], Zhao and Hwang [44] reported decomposition of malathion in the similar pattern. Thus, the optimum pH for degradation of malathion is 9.0.
The photocatalytic ability of HNFs was higher than BPGO and MGF-B. This improvement has been supposed to be due to (a) huge specific surface area, lowering in aggregation, easy charge transfer and layered structure of BPGO. The co-doped GO nanosheets act as a supporting substrate for the deposition of MGF-B, which enhanced the uptake capacity of the GO based photocatalyst.
The reduced aggregation resulted in higher specific surface area of HNFs (187.56 m 2 g −1 ) as compared to BPGO (162.81 m 2 g −1 ) and MGF-B (87.1 m 2 g −1 ) which provides more uptake sites for the pollutant molecules. (b) small size of HNFs (47 nm) as observed by DLS size distributions as compared to BPGO (60 nm) and MGF-B (77 nm) favoured higher adsorption, (c) Another factor was reduced recombination rate of electron-hole pair.
A control test was performed under visible light exposure in the pH range of 1.0 to 9.0. Figure S6a indicated poor self-degradation ability of malathion due to their stability in the absence of photocatalyst. The degradation was 38% under visible light illuminations which was significantly enhanced up to 99% in the presence of HNFs. Control experiment in the absence of light demonstrated less than 25% removal of malathion when treated with HNFs. The findings imply that photocatalytic degradation employing synthesized materials as a photocatalyst was mainly responsible for the malathion removal.
On increasing the photocatalyst dose from 0.04 to 0.2 g L −1 , the percentage of photodegradation was enhanced. Further increase in photocatalyst dosage displayed a decrease in degradation of pollutants. This may be ascribed to (1) the presence of huge specific surface area and active sites. The aggregation of HNFs, which results in the low penetration of visible light, may be the reason of the decline in photodegradation effectiveness at higher dosages.

Effect of Contact Time and Kinetic Studies
The photocatalytic decomposition of malathion with time was deduced by using UV-VIS absorption spectra (Figure 7f). The absorbance of base peak at 234 nm for malathion attenuated with the time, confirming its degradation. The photocatalytic degradation of malathion followed the Langmuir Hinshelwood first-order model. The value of apparent rate constant (k) calculated from ln(C t /C o ) vs. time was found to be 2.36 × 10 −2 min −1 (Figure 8a,b). Similar results were reported by Fakhri and Bagheri (2020) during the degradation of malathion and tetracycline using UiO@metal oxide/GO as a photocatalyst.
penetration of visible light, may be the reason of the decline in photodegradation effectiveness at higher dosages.

Effect of Contact Time and Kinetic Studies
The photocatalytic decomposition of malathion with time was deduced by using UV-VIS absorption spectra (Figure 7f). The absorbance of base peak at 234 nm for malathion attenuated with the time, confirming its degradation. The photocatalytic degradation of malathion followed the Langmuir Hinshelwood first-order model. The value of apparent rate constant (k) calculated from ln(Ct/Co) vs. time was found to be 2.36×10 −2 min −1 (Figure 8a,b). Similar results were reported by Fakhri and Bagheri (2020) during the degradation of malathion and tetracycline using UiO@metal oxide/GO as a photocatalyst.

. Statistical Analysis of Photodegradation Studies
A randomized design matrix and ANOVA for Box-Behnken regression model of malathion degradation is given in Tables 6 and 7, respectively. The terms F-value (79.73) and p-value (0.0001) are displayed in the ANOVA. Values less than 0.05 confirm that the given factors are significant at the 5% level of confidence. F test = 5.36 (p = 0.75) for "Lack of Fit" revealed that the model was adequately fitted and there was no lack of fit. The model's calculated value for adequate precision was 25.55, which supported a greater S/N ratio. The predicted quadratic regression model's R 2 (coefficient of determination) value was 99.7. As a result, this model may be used to more accurately predict the response at any level of the selected factors. The reduced expression of the quadratic regression model is presented below:    lines represents fitting of model, where C and C t represents concentration of malathion at initial time "t" and at time "t", respectively.

Statistical Analysis of Photodegradation Studies
A randomized design matrix and ANOVA for Box-Behnken regression model of malathion degradation is given in Tables 6 and 7, respectively. The terms F-value (79.73) and p-value (0.0001) are displayed in the ANOVA. Values less than 0.05 confirm that the given factors are significant at the 5% level of confidence. F test = 5.36 (p = 0.75) for "Lack of Fit" revealed that the model was adequately fitted and there was no lack of fit. The model's calculated value for adequate precision was 25.55, which supported a greater S/N ratio. The predicted quadratic regression model's R 2 (coefficient of determination) value was 99.7. As a result, this model may be used to more accurately predict the response at any level of the selected factors. The reduced expression of the quadratic regression model is presented below: Malathion degradation (%) = +77.0 + 33.5 A − 4.13 B + 4.63 C − 3.25 AB + 0.25 AC − 21.25 A 2 Table 6. A randomized design matrix for Box-Behnken methodology and observed responses for malathion degradation using HNFs.   Figure S6b displays the uniform distribution of data along a straight line and their strong correlation. The residuals of removal percentage are often plotted on a straight line in the normal probability plot ( Figure S6c). This demonstrates the validity of the normal distribution and proves to the model's applicability for accurate data fitting. Highest malathion degradation (%) was achieved at the maximum level (+1) of pH (9.0), minimum level (−1) of photocatalyst dose and median level (0) of contact time (Figure 7d,e).
Photocatalytic activity using malathion as model pollutant at pH ranging from 3.0 to 9.0, with variation in contact time = 30 to 120 min and variation in photocatalyst dose from 0.2 to 0.8 g L −1 . Figure 7d shows that as the photocatalytic activity increased with increasing pH and contact time, thestrong photocatalytic activity occurs at basic pH values and at a certain photocatalyst dose (Figure 7e). This could be a result of the catalyst particles aggregating at larger doses, which results in the low visible light penetration.

Quenching Studies
In order to eValuate the reactive species involved in the decomposition of organic pollutants, four quenchers viz. ascorbic acid, sodium azide, ethylene triamine tetraacetic acid and methanol were used as superoxide (O 2 − ), singlet oxygen (O 2 *), holes (h + ) and hydroxide (OH) quenchers, respectively (Figure 8c). The kinetic studies in the presence of quenchers indicated the photocatalytic degradation was significantly inhibited in presence of sodium azide and ascorbic acid. The addition of ascorbic acid produced the greatest inhibition, indicating that O 2 •− radicals are essential to the photodegradation process [45].
Here, it can be observedthat the reaction between holes and superoxide radicals produces singlet oxygen species. Contrary to these findings, the addition of EDTA and methanol slightly inhibited the degradation activity, confirming that h+ and • OH radicals also played a minor role in the photocatalytic degradation of malathion.

Analysis of Degraded Products
Degradation products formed at the end of the irradiation process were identified by GC-MS analysis ( Figure S6d). Oxidative desulfuration of malathion (m/z = 330) resulted in the formation of malaoxon (m/z = 314), which was further hydrolyzed into malaoxon monoacid (m/z = 287) and diethyl 2-mercaptosuccinate (m/z = 207) via. a carboxyl ester hydrolysis and a competing elimination reaction, respectively. When malathion was hydrolyzed, charged intermediates were created by the nucleophilic addition of − OH to the phosphorus atom. Malaoxon, which has a P=O moiety, is more likely to hydrolyze than malathion, which has a P=S moiety [20]. This enhances nucleophilic substitution by the − OH, which in turn encourages hydrolysis [46]. Diethyl [46]. However, HNFs have the advantage of higher photocatalytic potential.

Reusability Studies
One of the key factors determining a synthesizedphotocatalyst's effectiveness and practical usability is its stability. The reusability test of HNFs was run for seven consecutive cycles in order to assess the photocatalyst stability. After each cycle, the used photocatalyst was centrifuged from reaction vessel through centrifugation and washed with distilled water and used again forthe next cycle. A minimal loss in the percentage degradation was observed. The photocatalytic efficacy of HNFs decreases by~5%, after five consecutive cycles ( Figure S6e) indicating excellent stability of the HNFs photocatalyst.

Adsorption and Photodegradation Mechanism
HNFs as nanoadsorbent (Figure 9a) has the combined features of magnesium ferritebentonite (Path-A) and BPGO (Path-B) as: (1) The adsorption of Pb(II) ions on magnesium ferrite-bentonite part is mainly driven by Van der Waals forces, ion-exchange andelectrostatic interactions due to its negative charge.  In the FT-IR spectrum of Pb(II) adsorbed HNFs (Figure 8f), the bands affirmed to Bbonding, P-bonding and M-O vibrations were observed with reduced intensity and at lower wavenumbers which might be attributed to the electrostatic interactions of Pb(II) ions with these moieties present on the HNFs. All the bands except C=C and C-P shifted towards lower wavenumber due to complex formation of Pb(II) ions with the lone pair present on the oxygen atom of the adsorbed hydroxyl groups from the aqueous phase via electron sharing. XPS and FT-IR studies provided the strong eVidence for the adsorption of metal ions on the surface of HNFs.
A probable photocatalytic mechanism for the degradation of malathion has been proposed ( Figure 9b). Upon irradiation of light on the photocatalyst surface, the electrons present in the valance band are excited to the conduction band of magnesium ferritebentonite (Equation (1)). These photoinduced electrons can be readily accepted by BPGO nanosheets (as a matrix) to suppress the charge carrier recombination, which in turn improves the photodegradation ability (Equation (2)). Here, co-doped nanosheets played two vital roles for the improvement of the photocatalytic ability of the synthesized nanostructure. Briefly, the planar π-π framework of co-doped nanosheets makes it a good electron accepting material. The electrons present in the conduction band of magnesium ferritebentonite can readily move to co-doped nanosheets through percolation mechanism. This helps in decreasing recombination rate of charge carriers and thus more electrons are accessible for the generation of reactive species, boosting photocatalytic degradation. A good conducting nature of co-doped nanosheets allowed fast charge carrier transport, which helped in their effective carriers separation during the photocatalytic reaction. During this transfer of electrons, molecular oxygen reduces to form superoxide radical anion (O 2 •− ) (Equation (3)). Singlet oxygen species were generated by the reaction of holes with O 2 •− radicals (Equation (4)). The holes situated in the valence band oxidize hydroxyl ions (OH − ) or water molecules (H 2 O) to produce hydroxyl radicals (HO • ) (Equation (5)).The resulting reactive oxygen species degraded organic pollutants into the degradation products i.e., CO 2 and H 2 O (Equation (6)). The degradation mechanism can be expressed as: ROS + Organic pollutants → Degraded products (6) Hence, BPGO because of its better transporting and electron accepting nature combined with a united effect of magnesium ferrite-bentonite make the nanocomposite photocatalysts visibly active.
The magnesium ferrite-bentonite bound by the hexagonal BPGO hexagonal arrays, offered strong conductive network and electrical channels for magnesium ferrite-bentonite, thereby ensuring the structural integrity and enabling the effective transfer of electrons and holes during photocatalysis. The fabrication of BPGO with magnesium ferrite-bentonite offered more specific surface area (187.56 m 2 g −1 ) for the hierarchical nanohybrid ensuring the effective interaction space between organic contaminants and active sites.

Reagents
All the chemicals (analytical reagent grade), such as graphite powder, orthoboric acid, orthophosphoric acid, sulphuric acid, hydrogen chloride, malathion, sodium ni-trate, potassium permanganate, hydrogen peroxide, metal nitrates, bentonite, ethylene diammine tetraacetic acid, sodium azide, ascorbic acid, methanol and sodium hydroxide were procured from SD fine Pvt. Ltd. Deionized water was used to prepare all the solutions.

Synthesis and Characterization
3.2.1. Synthesis of Boron and Phosphorous Co-Doped Graphene Oxide (BPGO) BPGO was synthesized using orthoboric acid and orthophosphoric acid as the source for B and P, respectively. 50 mL of aqueous dispersion of GO (20 mg mL −1 ) was vigorously stirred with orthoboric acid (5.7 g) and orthophosphoric acid (1.60 mL) for 2 h at 25 • C. The solution was eVaporated to dryness. Remnant solid was annealed in the muffle furnace at 300 • C for 3 h to obtain BPGO. The aqueous suspension of BPGO was ultra-sonicated for 6 h, centrifuged and then washed with hydrogen chloride solution (5%) to eliminate unreacted reactants. Finally, it was washed with distilled water several times to get neutral pH followed by ethanol and dried at 80 • C overnight.

Synthesis of Hierarchical Nanoflowers (HNFs)
HNFs were synthesized in two steps. In the first step, binary composite of MgFe 2 O 4 and bentonite (MGF-B) was synthesized by sol-gel method, reported in our previous publication [20,21] which was further transformed into ternary hierarchical nanoflowers in the second step by adding BPGO to this binary composite. MGF-B and BPGO aqueous solutions were separately prepared by ultrasonication at room temperature. Finally, the ternary HNFs were collected by centrifugation and dried. The HNFs were synthesized in different w:w ratios (MGF-B:BPGO; 0.5:1, 1:1 and 2:1). The prepared samples were characterized by various physicochemical techniques discussed in Supplementary Text S6.

Adsorption Experiments
To perform batch mode adsorption experiments, stock solution of Pb(II) ions of 100 mg L −1 was prepared using Pb(NO 3 ) 2 ·4H 2 O. The first experiment was performed to determine the best adsorbent from MGF-B, BPGO and HNFs using 5 mg L −1 concentration of Pb(II) ions (100 mL) and 0.4 g L −1 of adsorbents at 25 • C. Then, using the best nanoadsorbent, next adsorption experiments were performed. To obtain the maximum adsorption efficiency, the batch mode adsorption experimental conditions, which included pH (2.0-10.0), contact time (2-120 min), adsorbent dosage (0.2-0.8 g L −1 ), and temperature (10-55 • C), were tuned. For pH adjustment, HCl or NaOH solutions were used. The adsorbate-adsorbent solutions were centrifuged and the concentration of Pb(II) in the centrifugates was eValuated using the ICAP-OES technique after shaking (240 rpm) on an orbital shaker for a fixedperiod of time (2 h). Limits of quantification (LOQ) and limit of detection (LOD) were also assessed for Pb(II) ions as explained in Supplementary Data S6. The adsorption efficiency (AE) and adsorption capacity (q t ) of HNFs was calculated as: where C i , C t , V and W are the initial metal concentration (mg L −1 ), concentration at time 't', volume of solution (L) and adsorbent dosage (g), respectively. The zero point charge (pH ZPC ) of samples was determined by plotting a graph between initial pH and ∆pH (pH final − pH initial ). The details of reusability studies and effect of coexisting cations and anions are discussed in Supplementary Text S7. All the experiments were run in three replicates, and average values are reported. Three error functions (Chi-square, sum of square of errors and residual root mean square error) were used to eValuate the validity of kinetic and isotherm models.

Photocatalytic Experiments
The first experiment performed to determine the photocatalyst showing the highest activity. For this, 0.2 g L −1 of pristine MGF-B, BPGO, and HNFs were combined with 50 mL of malathion (2 mg L −1 ) at pH 3.0-9.0. In order to reach equilibrium between adsorption and desorption, the mixture was agitated for 30 min. A 16 W visible light emitting diode was then turned on, and the photocatalytic reaction began. The suspension parts from the reaction vessel were removed after two hours and centrifuged. The residual malathion concentration was studied by a UV-VIS spectrophotometer at the wavelengths of 250 nm. Then, using best photocatalyst, next photocatalytic experiments were performed. An aliquot of 5 mL was taken for kinetic analysis at regular intervals ranging from 2 min to 12 h. Using the Langmuir Hinshelwood model, the apparent rate constant, k, was calculated [47]: ln(C o /C t ) = kt where C o and C t depicts organic pollutant concentration at initial time and after a definite period of light illumination, t stands for illumination time. The effect of the HNFs dose was investigated by changing the photocatalyst dose from 0.2 to 0.8 g L −1 . The details of quenching and reusability experiments are discussed in the Supplementary Text S8. The photodegradation studies were carried out in triplicate, and an average of three results is reported to test the validity and repeatability of the analytical data.

Statistical Analysis Using Box-Behnken Methodology
Using Stat-Ease Design-Expert (Version 13) software, a Box-Behnken experimental design was used to determine the effect of independent variables (pH, dose, contact time, and temperature) and their simultaneous interactions on the response function (removal efficiency of Pb(II) ions and degradation efficiency of malathion) with three central points. The various experimental levels of independent factors are displayed together with their codes in Table 8. The following quadratic polynomial equation was employed for analysing and predicting the response: Y = β o + Σ β i X i + Σ β ii X i 2 + ΣΣ β ij X i X j where Y and X m (m = i, j) stand for response and coded independent factors, respectively. β 0 is zero-order constant, whereas, β i , β ii , and β ij represents the linear, quadratic, and interaction coefficients of input independent variables, respectively. Selected design using independent variables "M" and central points "C 0 " proposed the N = 2M(M − 1) + C 0 experiments to performed trial runs, where N represents the frequency of samples. Based on this, a total of 27 runs for the adsorption experiment and 15 runs for the photocatalytic experiment were planned for three coded levels (1, 0, +1) of chosen components. Using the coefficient of determination (R 2 ), adjusted coefficient of determination (Adj. R 2 ), normal distribution of the residuals, and by plotting the actual values with predicted values, the suggested model's quality and goodness were assessed. The probability critical level (p-value) of 0.05 was used in the analysis of variance (ANOVA), which was used to determine whether the parameters of the proposed model were statistically significant.

Conclusions
Hierarchical nanoflowers (HNFs) was successfully fabricated via ultra-sonication approach. It combined features of MgFe 2 O 4 -bentonitenanocomposite with boronand phosphorus doped GO. The synthesized HNFs displayed excellent adsorptive and photodegradation performance as compared to pristine components towards Pb(II) and malathion removal. This enhancement in adsorption and photodegradation was mainly attributed to reduced aggregation, specific surface area and remarkable charge transfer ability. Parametric optimization by Box-Behnken design was successfully achieved for adsorption and photocatalytic performance of HNFs for the removal of Pb(II) ions and degradation of malathion. The 96% removal of Pb(II) and 98% degradation of malathion was achieved using HNFs as adsorbent and photocatalyst, respectively. Hence, synthesized HNFs can serve as a promising candidate for wastewater remediation.