Ferric chloride hexahydrate (FeCl₃·6H₂O) and Ethanol were obtained from Merck, Germany. Sodium acetate trihydrate (CH₃COONa·3H₂O), ethylene glycol, HCl, NaOH, and NaHCO₃ were purchased from Merck, India. Humic acid and Cerium oxide nanoparticles dispersion (10 wt.%) were purchased from Sigma Aldrich (St. Louis, MO, USA). All chemicals were of analytical grade and were used without further purification.18.2 MMilli-Q water (Millipore) was used throughout the experiments.

The IONPs were synthesized through a solvothermal reaction with some modifications (^{Deng et al., 2008}; ^{Pang and Liu, 2013}; ^{Zhou et al., 2017}). Briefly, 2.7 g Ferric chloride hexahydrate (FeCl₃·6H₂O) was mixed with 7.2 g sodium acetate in 100 mL ethylene glycol solution at 65°C. Sodium acetate was used as a structure-directing and electrostatic stabilizing agent, while ethylene glycol was used as a reducing agent (^{He et al., 2018}). Besides, ethylene glycol has a characteristic of high boiling point and dielectric constant, which makes it efficient for the solvothermal process (^{He et al., 2018}). The mixture was stirred at 1300 rpm for 20 min and the obtained solution was sealed and heated at 200°C overnight. The product was collected by using a strong permanent magnet and was washed continuously with ethanol and Milli-Q water several times to remove the non-magnetic bi-products. The collected product was vacuum dried at 85°C temperature overnight. Finally, dark brown colored powder was obtained and stored for further characterization and sorption experiments.

To evaluate the phase and crystalline nature of the nanoparticles, powder X-ray diffraction (PXRD) measurements were carried out using a Rigaku (mini flex II, Japan) with Cu-K radioactive source (wavelength= 0.154 nm) at 40 kV/70 mA at a scanning rate of 2° per minute in the range of 20°‒70°. The morphology of the particles was investigated by field emission scanning electron microscopy (FESEM). Images were taken with a Carl Zeiss SUPRA 55VP FESEM. TEM images of particles were obtained using a JEOL, JEM 2100 HR model, along with an elemental line scan. Zeta potentials, Hydrodynamic diameter, Point of Zero charge (pH_PZC) in aqueous suspension solution were measured with a Malvern Zetasizer ZS-90 using dynamic light scattering (DLS) technique. Fourier-transform infrared (FT-IR) spectra were taken to check the presence of various surface functional groups on IONPs and CeO₂, using Thermoscie Nicolet iS5 coupled with iD1 transmission accessory by making KBr pellets. After the interaction of CeO₂ with IONPs, reaction precipitate was separated and characterized using TEM/STEM along with elemental distribution and FT-IR spectra to get detailed information about various interaction mechanisms.

For the good dispersion of particle, CeO₂ and IONPs stock solutions were sonicated for 15 min. CeO₂ working solutions (10 mL) in the range of 20‒120 mg/L initial concentrations were prepared by diluting a stock solution. Finally,0.5 mL of 2 mg/mL aqueous dispersions was added in CeO₂ aqueous solutions. These reaction mixtures were shaken at 300 RPM for 1 h in an amber glass vial at room temperature. Measured pH of the solutions was 5.8. The IONPs were separated through a permanent magnet, and the supernatant was collected. The concentration of CeO₂ in the supernatant was determined by using a UV-VIS spectrophotometer (Evolution 201, Thermo) at a wavelength of 300 nm. The adsorption capacity (Qe, mg/g) was calculated as follows.

where C_o (mg/L) and C_e (mg/L) are the initial concentration of CeO₂ and the remaining concentration of CeO_2, respectively. Q_e (mg/g) is the adsorption capacity of IONPs, V (mL) is the volume of solution, and M (mg) is the mass of IONPs. All experiments were done doubly to confirm Quality assessment/Original content (QA/OC) of data, error was within 10%, and therefore, error bars have been eliminated.

Both SEM and TEM images of IONPs (Fig. 1(a)) show the spherical nature of the particles with nearly 200 nm diameter. Fig. 1b shows the morphology of highly dispersed CeO₂ NPs, which are nearly of 25 nm in size and are cubic. To obtain the details of composition, Energy Dispersive X-ray (EDS) data was obtained with a TEM-EDS line scan (Fig. 1(c)) along an IONP sphere, and the result shows Fe/O= 3/4 which is for magnetite. To confirm the suspension stability of CeO₂ NPs, change in hydrodynamic diameter of these particles was observed for 180 min, and data in Fig. 1(d) clearly shows no significant change suggesting that these particles have no self-aggregation and are stable in suspension.

PXRD spectra of IONPs (Fig. 2(a)) shows six characteristic peaks at 30.1°, 35.4°, 43.1°, 53.4°, 56.9°, and 62.5°, corresponding to planes (220), (311), (400), (422), (511), and (440) of magnetite (JCPDS File No. 19-0629), respectively, were observed. Further, broad and smaller peaks signify short-range order. Zeta potential with varying pH was measured, and point of zero charges (PZC) for IONPs and CeO₂ were determined to be 5.8 and 8.1, respectively as shown in Fig. 2(b). IONPs tend to show negative zeta potential at circum-neutral pH because oxygen atoms on the surface are not fully coordinated with Fe³⁺ (^{Chen et al., 2017}). In CeO₂, there are oxygen vacancies acting as point defects due to which cerium oxide was represented as CeO_2—γ (Gunawan et al., 2019). Therefore, cerium oxide shows positive zeta potential at similar pH values. DLS characterization remains a critical factor for physisorption as the particles possess opposite charges between specific pH (5.8‒8.1),which is essential for electrostatic interaction.

Several solutions with a constant concentration of CeO₂ (70 mg/L) and IONPs (100 mg/L) were prepared and kept for different time intervals (0‒480 min) on a shaker to achieve the equilibrium time. It was observed that there are abundant sites available on IONPs, and the majority of the sites available were occupied rapidly by CeO₂NPs in the first 30 min of reaction. Then the active sites gradually decreased, and adsorption equilibrium was achieved, which shows the faster removal of CeO₂ in the system. Further data was interpreted with various kinetic models like the pseudo-first-order, pseudo-second-order, intra-particle model, and Elovich model(^{Largitte and Pasquier, 2016}).

The linear form of the pseudo-first equation is as follows

where Q_e and Q_t are the amount of CeO₂ (mg) adsorbed per unit gram of IONPs at equilibrium time and at time t, respectively. k₁ (min^‒1) is the pseudo-first-order rate constant. Fitting the data with pseudo-first-order gives the R² value of 0.75 (Fig. 3(a)) and calculated sorption capacity value Q_e (Table 1) was very different from the experimental value, which suggests that the sorption does not follow pseudo-first-order.The equation for second-pseudo order is (^{Ho and McKay, 1999})

where abbreviation for Q_e, Q_t, and t remains the same while k₂ is the rate constant for pseudo-second-order. Fitting the data with this model gives the R² value of 0.99 (Fig. 3(b)), which is high, and calculated sorption capacity value Q_e (Table 1) is matching the experimental value, suggesting that the sorption of CeO₂ into the IONPs agrees strongly to the pseudo-second-order, further supporting chemisorption as probable removal mechanism.

The Elovich equation is typically used to describe the kinetics of chemisorption on highly heterogeneous sorbents and linearly expressed as(^{Chien et al., 1980})

where α and k are the initial adsorption rate and extent of surface coverage and activation energy for chemisorption, respectively. Fitting the data with the Elovich equation gives the R² value of 0.93 (Table 1 and Fig. 3(c)) suggesting that the adsorption process was controlled by chemisorption.

The intraparticle diffusion model was used to interpret the movement of the particles from the surface of the adsorbent to the inner pores. It is based on the principle that the sorption on any material follows three steps process which is (1) film or surface diffusion, where the sorbate is transported from the bulk solution to the external surface of sorbent, (2) intraparticle or pore diffusion, where sorbate molecules move into the pores of sorbent particles, and (3) adsorption on the interior sites of the sorbent (^{Khandelwal et al., 2019}). The model is linearly expressed as(^{Srivastava et al., 1989})

where k_d is intraparticle diffusion rate constant or rate factor, and c is resistance in mass transfer due to the boundary layer. By comparing the data with the intraparticle diffusion model, it was observed in Fig. 3(d) that the sorption kinetics is very rapid, and diffusion does not affect the overall removal kinetics. Therefore, bulk transport was the dominant sorption process.

Sorption studies of CeO₂ on IONPs was carried out and compared with the two most prominent isotherm models Langmuir and Freundlich. Langmuir isotherm (^{Langmuir, 1918})

and the Freundlich model is expressed as

where Q_e, Q_m, C_e, and k_l are the adsorption capacity of IONPs, the maximum adsorption capacity of IONPs, remaining concentration, and Langmuir isotherm constant respectively while K_f and n are the Freundlich constant and adsorption intensity respectively. Langmuir model suggests about the homogeneity of the sorption process and accounts for the surface coverage by balancing the relative rates of adsorption and desorption (dynamic equilibrium). While Freundlich isotherm assumes heterogeneous adsorption of pollutants onto the adsorbent, and it gives an expression defining the surface heterogeneity and the exponential distribution of active sites and their energies. On fitting the data with Langmuir and Freundlich isotherm model as shown in Fig. 4, correlation factor R² comes out to be 0.99 and 0.95 for Langmuir and Freundlich respectively (Table 2) suggesting that the sorption of CeO₂ into IONPs follows better Langmuir isotherm model than Freundlich isotherm model supporting monolayer sorption with having uniform sorption energy of the surface of IONP. Adsorption intensity constant (n) value 0.33 (less than 1) suggests adsorption was favorable in the concentration range studied(^{Zhou et al., 2017}).

Bicarbonate is one of the commonly present ions in a natural environment and can affect the sorption; therefore, the effect of bicarbonate ions on the sorption capacity of IONPs for CeO₂ was studied in wide concentration range (0.1‒1 mM).It was observed in Fig. 5(a) that the sorption capacity decreases with the increasing concentration of bicarbonate, which can be attributed to the fact that bicarbonate ions can act as direct competitors for available binding sites on IONPs(^{Zachara et al., 1987}). A similar phenomenon was also observed in case of As(III) and As(V) sorption on Goethite(^{Stachowicz et al., 2007}).

Solution pH is the judgmental factor for the sorption experiment as it influences the form of functional groups on the surface of adsorbate or adsorbents. Predominantly, sorption capacity changes with pH due to the change in their electrostatic energy. In DLS measurement, change in zeta potential with pH for CeO₂ and IONPs has been observed resulting in change in electrostatic energy between the particles. Results in Fig. 5(b) show maximum sorption capacity near pH 6.5, which can be supported by zeta potentials suggesting maximum electrostatic attraction at the same pH due to drastically opposite charges, as shown in Fig. 2(b). Sorption decreased in acidic pH range due to electrostatic repulsion, but some amount of CeO_2, which was adsorbed, can be attributed to Fe-O-Ce-O direct interaction, as discussed in section 3.8.

To evaluate the practicality of sorption methodology of IONPs in the natural aqueous environment, for that waters with three different chemical composition, i.e. synthetic soft, hard and groundwater were prepared in the laboratory by mixing respective salts to carry out CeO₂ sorption experiments(^{Smith et al., 2002}; ^{Xu et al., 2016}). Besides, environmental samples of groundwater and river water were also utilized for the same to compare the change in the sorption capacity of IONPs. Similar CeO₂ concentrations, i.e., 70 mg/L was spiked in all solutions with 100 mg/L of IONPs and results in Fig. 6 show that removal was more than 90% suggesting that the matrix have negligible influence on the sorption capacity of IONPs. Therefore, IONPs could be utilized for the removal of CeO₂ NPs from contaminated environmental water samples.

For the removal of contaminants from the environment, cost factor remains one of the important criteria for the selection of the removal process and adsorbents.Therefore, the reusability of IONPs was evaluated by subjecting the reaction precipitate to ultrasonication, followed by magnetic separation. Sorption process shows the removal efficiency of nearly 88% of CeO₂ NPs. While the recovery efficiency was nearly 80% of CeO₂ NPs through ultrasonication. It suggests that the pores of IONPs are available again for adsorption and can be reused for the sorption of CeO₂ NPs. A schematic is given in Fig. 7 for adsorption-desorption experiments, and Fig. 8(f) provides a visual image of recovered CeO₂.

To confirm various removal mechanisms stated and interpreted through kinetic and isotherm modeling, CeO₂-IONPs reaction precipitate was collected and characterized using TEM and STEM. As shown in Fig. 8(a), CeO₂ NPs can be seen all along the circumference of IONPs, while STEM image in Fig. 8(b) confirms that these small particles were all along the surface of IONPs. TEM elemental mapping was performed to confirm the presence of CeO₂ on the IONP surface, and data in Fig. 8(c) confirms the STEM observations, as Ce was distributed all along with IONPs. Significant differences in hydrodynamic diameters of CeO₂ NPs and IONPs, i.e.,>100 nm for CeO₂ and>200 nm for IONPs, allowed us to monitor the change in hydrodynamic diameter with time in the system containing IONPs and CeO₂. Results show in Fig. 8(d) that separate intensity peak for CeO₂ at<100 nm, which was present till 100 s of interaction, completely disappears after 600 s and simultaneously single peak dominates with broad size fractions, suggesting attachment of CeO₂NPs with IONPs. To confirm whether the sorption was only due to electrostatic attraction or surface complexation was also involved in the removal process, reaction precipitate was analyzed using FT-IR to see the changes in functional groups stretching after sorption of CeO₂. Spectra provided in Fig. 8(e) shows that the intense band at 500 cm^‒1 which corresponds to the Ce-O stretching vibration and available in blank CeO₂ (Zamiri et al., 2015); ^{Pujar et al., 2018}) has also generated in CeO₂-IONP reaction precipitate confirming the electrostatic attachment of CeO₂. Further, sharp peak at 580 cm^‒1 which can be ascribed to the Fe-O vibration from the magnetite phase (^{Yang et al., 2015}) in case of IONP was disturbed and flattened after interaction with CeO₂ NPs suggesting that apart from electrostatic attraction, Fe-O-Ce-O type surface complexation was also helping in CeO₂ removal(^{Gan et al., 2017}). This observation also confirms why CeO₂ removal was also observed in acidic pH range where both the particles were positively charged.