Fe3O4-Halloysite Nanotube Composites as Sustainable Adsorbents: Efficiency in Ofloxacin Removal from Polluted Waters and Ecotoxicity

The present work aimed at decorating halloysite nanotubes (HNT) with magnetic Fe3O4 nanoparticles through different synthetic routes (co-precipitation, hydrothermal, and sol-gel) to test the efficiency of three magnetic composites (HNT/Fe3O4) to remove the antibiotic ofloxacin (OFL) from waters. The chemical–physical features of the obtained materials were characterized through the application of diverse techniques (XRPD, FT-IR spectroscopy, SEM, EDS, and TEM microscopy, thermogravimetric analysis, and magnetization measurements), while ecotoxicity was assessed through a standard test on the freshwater organism Daphnia magna. Independently of the synthesis procedure, the magnetic composites were successfully obtained. The Fe3O4 is nanometric (about 10 nm) and the weight percentage is sample-dependent. It decorates the HNT’s surface and also forms aggregates linking the nanotubes in Fe3O4-rich samples. Thermodynamic and kinetic experiments showed different adsorption capacities of OFL, ranging from 23 to 45 mg g−1. The kinetic process occurred within a few minutes, independently of the composite. The capability of the three HNT/Fe3O4 in removing the OFL was confirmed under realistic conditions, when OFL was added to tap, river, and effluent waters at µg L−1 concentration. No acute toxicity of the composites was observed on freshwater organisms. Despite the good results obtained for all the composites, the sample by co-precipitation is the most performant as it: (i) is easily magnetically separated from the media after the use; (ii) does not undergo any degradation after three adsorption cycles; (iii) is synthetized through a low-cost procedure. These features make this material an excellent candidate for removal of OFL from water.


Introduction
In the current scenario of water shortage, there is an urgent need to favor water loops. For this purpose, preserving and guaranteeing water quality is mandatory, as reclaimed water can be directly reused and re-enter natural water bodies [1]. A critical aspect of water quality is represented by xenobiotics, such as heavy metals, dyes, pesticides, etc., detected in natural water bodies, also at trace levels, because of their recalcitrance in conventional wastewater treatment plants (WWTPs) [2]. In particular, pharmaceuticals and can provide useful tools to assess the risk related to nanomaterials and to select eco-friendly and sustainable ones for water remediation [32,33]. The application of standard and/or novel ecotoxicological tests completes the characterization of nanomaterials through the identification of possible toxicological targets and sheds light on the mechanism(s) of toxic action in aquatic species at different levels of the ecological hierarchy [34].
In the present study, we synthesized HNT/Fe 3 O 4 nanocomposites by using three different approaches: co-precipitation, sol-gel, and hydrothermal. Each material was characterized by FT-IR spectroscopy, X-ray powder diffraction (XRPD), scanning electron and transmission electron microscopy (SEM and TEM), energy dispersive spectroscopy (EDS), thermogravimetric analysis (TGA), and magnetization measurements. Moreover, the magnetite and halloysite amount in each sample was evaluated by EDS, TGA, and magnetization data. Lastly, potential ecotoxicity of these materials towards aquatic organisms was tested on the freshwater Cladoceran Daphnia magna according to the Daphnia sp. Acute Immobilization Test, OECD 202 guideline (OECD, 2004). Adsorption properties and mechanism of each nanocomposite were investigated, and compared with the commercial halloysite. The antibiotic ofloxacin (OFL) was chosen as the target molecule to assess the adsorption efficiency of HNT/Fe 3 O 4 nanocomposites for different reasons: (i) it is a very useful antibacterial agent belonging to the last class of antibiotics; (ii) it is largely detected in wastewaters and surface waters [3]; (iii) it is a recalcitrant to biological degradation [35]; (iv) it maintains a certain antibacterial activity after the first steps of its degradation [35]; (v) it has been used in our previous studies regarding both fluoroquinolones' environmental fate and their removal by adsorption processes [36][37][38][39][40][41]. The suitability of three materials for OFL removal under environmental conditions, i.e., tap and river waters, and wastewater treatment plant (WWTP) effluent, was also verified.

Synthesis
Halloysite nanotubes-magnetite composites (HNT/Fe 3 O 4 ) and magnetite alone (Fe 3 O 4 ) were synthesized by co-precipitation, sol-gel, and hydrothermal routes, as follows. Table 1 summarizes the synthesis approaches and the sample names.

Co-Precipitation Procedure
The HNT/Fe 3 O 4 -C sample was synthesized following the procedure of Xie et al. [27]. An amount of 0.5 g of HNT was added to an aqueous solution of 4.32 mmol of FeCl 3 ·6H 2 O and 2.16 mmol of FeSO 4 ·7H 2 O. The suspension was heated at 60 • C under N 2 flux, and an 8 M ammonia solution was added dropwise to reach pH 9-10. The suspension was further heated for 4 h at 70 • C, then the solid was magnetically recovered, washed three times, and dried for 3 h at 100 • C. The same procedure was applied to synthesize the Fe 3 O 4 alone (sample Fe 3 O 4 -C), by omitting the addition of HNTs.

Sol-Gel Procedure
The HNT/Fe 3 O 4 -SG sample was synthesized as reported by He et al. [29]. An amount of 1 g of HNT was dispersed in an ethanol solution containing 1.98 mmol of Fe(NO 3 ) 3 ·9H 2 O. The dispersion was sonicated, stirred 24 h at room temperature and dried for 24 h at 35 • C. An amount of 2 mL of ethylene glycol was added, and the sample was heated for 2 h at 400 • C under N 2 flux (N 2 99.999%; flow rate: 3 L h −1 ; heating and cooling rate: 5 • C min −1 ). The same procedure was applied to synthesize the Fe 3 O 4 alone (sample Fe 3 O 4 -SG), by omitting the addition of HNTs.

Hydrothermal Procedure
HNT/Fe 3 O 4 -H sample was synthesized following the procedure of Tian et al. [30], with some modifications. The procedure consists of two hydrothermal steps: the former to prepare HNT enriched with a carbonaceous component, and the latter to decorate it with magnetite. An amount of 0.5 g of HNT was added to a glucose solution (10 g L −1 ) and magnetically stirred. The dispersion was poured into a Teflon-lined stainless-steel autoclave and heated for 48 h at 160 • C. The obtained product was washed 5 times in ethanol, centrifuged, and dried for 18 h at 60 • C under vacuum. An amount of 0.5 g of the final product was added to a solution containing 3 mmol of FeCl 3 ·6H 2 O in ethylene glycol. After stirring for 24 h, 1.8 g of sodium acetate and 0.5 g of ethylene glycol were added, and the dispersion was poured into a Teflon-lined stainless-steel autoclave and heated for 8 h at 200 • C. The obtained magnetic composite was washed with distilled water and dried for 12 h at 80 • C. The procedure of the second step was also applied to synthesize the Fe 3 O 4 alone (sample Fe 3 O 4 -H), by omitting the addition of HNTs.

Characterization Techniques
X-ray powder diffraction measurements were performed using a Bruker D5005 diffractometer (Bruker, Karlsruhe, Germany) with the CuKα radiation, graphite monochromator, and scintillation detector. The patterns were collected in the 7-80 • two-theta angular range, step size of 0.03 • , and a counting time of 20 s/step. A silicon low-background sample holder was used.
FT-IR spectra were obtained with a Nicolet FT-IR iS10 Spectrometer (Nicolet, Madison, WI, USA) equipped with ATR (attenuated total reflectance) sampling accessory (Smart iTR with ZnSe plate) by co-adding 32 scans in the 4000-650 cm −1 range at 4 cm −1 resolution. Thermogravimetric measurements were performed by a TGA Q5000 IR apparatus interfaced with a TA 5000 data station (TA Instruments, Newcastle, DE, USA). The samples were scanned at 10 • C min −1 under nitrogen flow (45 mL min −1 ) in the 20-850 • C temperature range. Each measurement was repeated at least three times.
The specific surface area and porosity were investigated by N 2 adsorption using the BET method in a Sorptomatic 1990 Porosimeter (Thermo Electron, Waltham, MA, USA).
SEM measurements were performed using a Zeiss EVO MA10 (Carl Zeiss, Oberkochen, Germany) Microscope, equipped with an Energy Dispersive Detector for the EDS analysis. The SEM images were collected on gold-sputtered samples. HR-SEM images were taken from an FEG-SEM Tescan Mira3 XMU. Samples were mounted onto aluminum stubs using double sided carbon adhesive tape and were then made electrically conductive by coating Nanomaterials 2022, 12, 4330 5 of 20 in vacuum with a thin layer of Pt. Observations were made at 25 kV with an In-Beam SE detector at a working distance of 3 mm.
TEM micrographs were carried out on a JEOL JEM-1200 EX II (JEOL Ltd., Tokio, Japan) microscope operating at 100 kV high voltage (tungsten filament gun) and equipped with a TEM CCD camera Olympus Mega View III (Olympus soft imaging solutions (OSIS) GmbH, from 2015 EMSIS GmbH, Munster, Germany) with 1376 × 1032 pixel format. The samples were prepared by drop-casting the solution on nickel grids formvar/carbon coated.
Dynamic light scattering (DLS)-Nicomp 380 ZLS (Particle Sizing Systems, Lakeview Blvd. Fremont, CA, USA) was used. For analyses, samples were diluted 1:10 in MilliQ water. The main parameters set up were: channel 10, intensity 100 kHz, temperature 23 • C, viscosity 0.933 cPoise, and a liquid index of refraction 1.333. The values considered at the end of the analyses were: mean diameter (nm), standard deviation, and Zeta potential (mV).
To investigate the magnetic behavior of the materials, field dependence of magnetization was investigated using a vibrating sample magnetometer (VSM Model 10-Microsense) equipped with an electromagnetic producing magnetic field in the range ±2 T.

Adsorption and Kinetic Experiments
OFL adsorption on HNT/Fe 3 O 4 -C, HNT/Fe 3 O 4 -SG, HNT/Fe 3 O 4 -H, and commercial HNT was studied by a batch method. For adsorption equilibrium experiments, 20 mg of each material was suspended in 10 mL of tap water spiked with OFL in the range of 25-200 mg L −1 . Flasks were wrapped with aluminum foil to prevent light-induced drug decomposition and shaken for 24 h at room temperature with an orbital shaker. Subsequently, the suspensions were magnetically separated, and the supernatants were filtered (0.22 µm) and analyzed by UV-vis spectrophotometer at 287 nm to determine the antibiotic concentration in solution at equilibrium (C e ). The adsorbed OFL amount at equilibrium (q e , mg g −1 ) was calculated by Equation (1): where C 0 is the initial OFL concentration (mg L −1 ), C e is the drug concentration in solution at equilibrium (mg L −1 ), V is the volume of the solution (L), and m is the amount of the sorbent material (g). For the kinetic experiments, 20 mg of each material were suspended in 10 mL of 20 mg L −1 OFL tap water solution. Falcon tubes, wrapped with aluminum foil, were shaken by a roller shaker and, at selected times, the adsorbent was magnetically treated. Then, a few mL of the supernatant were collected, filtered (0.22 µm) in a quartz cuvette, and analyzed by a UV spectrophotometer at 287 nm. The analyzed solution was recovered to keep the suspension volume constant for all experiments. The adsorbed OFL amount at time t (q t , mg g −1 ) was calculated as (Equation (2)): where C 0 is the initial OFL concentration (mg L −1 ), C t is the drug concentration in solution at time t (mg L −1 ), V is the volume of the solution (L), and m is the amount of the sorbent material (g). All experiments were performed in duplicate. The thermodynamic and kinetic parameters were estimated by dedicated software (OriginPro, Version 2019b. OriginLab Corporation, Northampton, MA, USA).
The well-known Langmuir's and Freundlich's isotherm models were applied to fit the experimental data. The Langmuir model (Equation (3)) describes the adsorption process that takes place on specific homogeneous sites and in a monolayer on the material surface: 6 of 20 where K L is the Langmuir constant and q m is the monolayer saturation capacity. The Freundlich model defines non-ideal adsorption on the heterogeneous surface, and Equation (4) expresses it: where K F is the empirical constant indicative of adsorption capacity, and n is the empirical parameter representing the adsorption intensity. The time-dependent data were fitted by pseudo-first-order (Equation (5)) and pseudosecond-order kinetic (Equation (6)) models: where q t and q e are the drug adsorbed amount at time t and equilibrium, respectively, and k 1 and k 2 are the pseudo-first-order and the pseudo-second-order rate constants.

Analytical Measurements
For OFL analysis at mg L −1 , a UV-vis UVmini-1240 spectrophotometer (Shimadzu Corporation) was used. The instrument was set at 287 nm, corresponding to the maximum OFL absorption. Calibration in the range of 1-10 mg L −1 yielded optimal linearity (R 2 > 0.9988). The quantification limit was 0.8 mg L −1 .

Acute Toxicity Tests with Daphnia magna
The potential acute toxicity of the different materials, i.e., HNT, Fe 3 O 4 -C, and HNT/Fe 3 O 4 -C, was tested on the freshwater Cladoceran Daphnia magna according to the Daphnia sp. Acute Immobilization Test, OECD 202 guideline (OECD, 2004). Adult Daphnia magna individuals were cultured (30 individuals/L) in a commercial mineral water (San Benedetto ® ) under controlled laboratory conditions reported elsewhere [42]. Five replicates containing ten daphnids (i.e., <24 h old individuals) each were performed per each experimental condition, including control. In detail, daphnids were exposed for 48 h at 20 ± 0.5 • C and 16 h light: 8 h dark photoperiod under static, non-renewal conditions to 0.2 g L −1 of the materials. A single concentration mimicking the amount of residues in waters after depollution treatment was tested. This concentration reflected the amount of each material used in the experiments aimed at investigating their capability in the removal of OFL.
The viability of individuals was tested after 24 and 48 h of exposure. Individuals were considered dead when they did not swim for over 15 s after a slight stirring of the solutions. After checking for viability, all the individuals were observed under a Leica Microsystem EZ4 Stereoscopic microscope to check for the ingestion of materials by daphnids.

Results and Discussion
First, structure, morphology, composition, magnetic behavior, adsorption capacity, and adsorption kinetics of the magnetic HNT composites and the commercial HNT were investigated. Then the materials were tested under environmental conditions to remove the antibiotic OFL chosen as being representative of emerging contaminants. In addition, their potential ecotoxic effects, along with reusability, were evaluated. Figure 1a shows the XRPD pattern of the commercial halloysite. It compares to those reported in the literature [27,30,43,44] and deposited in JCPDS database (PDF# 028-1487).

Morphological, Structural, and Magnetic Characterization
The peak detected at about 12 • corresponds to the d 001 basal spacing of 7.35 Å, peculiar of the anhydrous form (halloysite-(7 Å)). The (002) reflection is observed at about 24 • . The peaks at 20 • and 62.8 • are typical of halloysites with nanotubular morphology [44,45]. No peaks are detected at about 8.8 • , assigned to the d 001 basal spacing of the di-hydrated halloysite (halloysite-(10 Å)). This is consistent with the easy loss of the interlayer water molecules near room temperature [46]. The very sharp reflections observed at 10.1, 26.6, and 27.3 • are attributed respectively to the small amount of kaolinite 1A (PDF# 074-1786), quartz (PDF# 046-1045), and rutile (PDF# 021-1276); these impurity phases are often detected in halloysite clay minerals.
solutions. After checking for viability, all the individuals were observed under a Leica Microsystem EZ4 Stereoscopic microscope to check for the ingestion of materials by daphnids.

Results and Discussion
First, structure, morphology, composition, magnetic behavior, adsorption capacity, and adsorption kinetics of the magnetic HNT composites and the commercial HNT were investigated. Then the materials were tested under environmental conditions to remove the antibiotic OFL chosen as being representative of emerging contaminants. In addition, their potential ecotoxic effects, along with reusability, were evaluated. Figure 1a shows the XRPD pattern of the commercial halloysite. It compares to those reported in the literature [27,30,43,44] and deposited in JCPDS database (PDF# 028-1487).

Morphological, Structural, and Magnetic Characterization
The peak detected at about 12° corresponds to the d001 basal spacing of 7.35 Å, peculiar of the anhydrous form (halloysite-(7 Å)). The (002) reflection is observed at about 24°. The peaks at 20° and 62.8° are typical of halloysites with nanotubular morphology [44,45]. No peaks are detected at about 8.8°, assigned to the d001 basal spacing of the di-hydrated halloysite (halloysite-(10 Å)). This is consistent with the easy loss of the interlayer water molecules near room temperature [46]. The very sharp reflections observed at 10.1, 26.6, and 27.3° are attributed respectively to the small amount of kaolinite 1A (PDF# 074-1786), quartz (PDF# 046-1045), and rutile (PDF# 021-1276); these impurity phases are often detected in halloysite clay minerals.     (Figure 2b), all the halloysite bands are detected. As for the Fe3O4 phase, only one broad band centered at about 3435 cm −1 attributed to OH-bending of hydroxyl groups was observed [43]. This broad band was not detected in the HNT/Fe3O4-SG sample, displaying a high amount of halloysite and a few magnetites (see XRPD results). The SEM images of the commercial HNT are shown in Figure S1a,b. The sample displayed 2-10 μm agglomerates of nanotubular particles, better highlighted in TEM micrographs (Figure 3a,b). The nanotubes exhibited an external diameter of 60-70 nm, a lumen of 20-30 nm, and variable length, from a few hundred nanometers to 1-2 µm. The DLS results showed a bimodal particle size distribution. The mean particle size is reported in Table S1. The SEM images of the commercial HNT are shown in Figure S1a,b. The sample displayed 2-10 µm agglomerates of nanotubular particles, better highlighted in TEM micrographs (Figure 3a,b). The nanotubes exhibited an external diameter of 60-70 nm, a lumen of 20-30 nm, and variable length, from a few hundred nanometers to 1-2 µm. The DLS results showed a bimodal particle size distribution. The mean particle size is reported in Table S1.
The SEM images of the commercial HNT are shown in Figure S1a,b. The sample displayed 2-10 μm agglomerates of nanotubular particles, better highlighted in TEM micrographs (Figure 3a,b). The nanotubes exhibited an external diameter of 60-70 nm, a lumen of 20-30 nm, and variable length, from a few hundred nanometers to 1-2 µm. The DLS results showed a bimodal particle size distribution. The mean particle size is reported in Table S1.   Figure S2 shows the SEM micrographs of the HNT/Fe 3 O 4 composites synthesized by co-precipitation ( Figure S2a,b), hydrothermal ( Figure S2c,d), and sol-gel ( Figure S2e,f) routes. All the composites displayed micrometric nanotubular particles, whose morphology well compares to the HNT sample one ( Figure S1a,b). In addition, nanometric rounded aggregates, possibly due to the magnetite phase, were observed on the nanotubes surface and between the nanotubes, interconnecting them; they were mainly detected in the HNT/Fe 3 O 4 -C sample ( Figure S2a,b) which was richer in magnetite, as suggested by XRPD and FT-IR results.  Table S1. The Fe 3 O 4 -C sample displayed wide particle size distribution. The HNT/Fe 3 O 4 -C and NHT/Fe 3 O 4 -SG samples displayed particle size >900 nm, slightly similar to the larger ones of the commercial halloysite. Instead, the HNT/Fe 3 O 4 -H composite displayed lower particle size. To better characterize the tendency of particles to aggregate and to investigate particles' surface charge changes, zeta-potential was evaluated. Commercial HNT exhibits a negative zetapotential of −31.77 mV; this value confirms that the outer nanotube surface is negatively charged and is in good agreement with the literature data [47].
In both the magnetite and composite samples, the Fe3O4 nanoparticles aggregate; particle size distribution was evaluated by DLS analysis and reported in Table S1. The Fe3O4-C sample displayed wide particle size distribution. The HNT/Fe3O4-C and NHT/Fe3O4-SG samples displayed particle size > 900 nm, slightly similar to the larger ones of the commercial halloysite. Instead, the HNT/Fe3O4-H composite displayed lower particle size. To better characterize the tendency of particles to aggregate and to investigate particles' surface charge changes, zeta-potential was evaluated. Commercial HNT exhibits a negative zeta-potential of −31.77 mV; this value confirms that the outer nanotube surface is negatively charged and is in good agreement with the literature data [47].  The Fe3O4-C sample (chosen as reference of the magnetite samples) exhibits a zetapotential of −7.16 mV, comparable to the literature values [48]; this value is not sufficient to achieve a stable suspension, and justifies particle aggregation (see TEM and DLS results).
Zeta-potential values of −36.36, −12.89 and −112.02 mV are obtained for HNT/Fe3O4-C, HNT/Fe3O4-SG and HNT/Fe3O4-H composites. The sample prepared by the hydrothermal process displays the most negative zeta-potential value; this may be due to the carbonaceous component (see TEM results and Section 3.2.) and explains the improved stability of the suspension and the lower mean particle size, as shown by DLS results.
The EDS analysis was applied to display the distribution map of halloysite and magnetite in each composite sample and to evaluate the weight percentage. Figures S3-S5 show the distribution maps of Al, Fe, and Si for the HNT/Fe3O4-C, HNT/Fe3O4-H, and HNT/Fe3O4-SG samples. Independently of the synthetic route, Al and Si were detected in the same areas. The Fe distribution was rather homogeneous in the sol-gel and hydrothermal samples (Figures S4 and S5, respectively), but also in some regions in which Fe prevails were detected. In the co-precipitation composite, Fe prevailed in areas poor in Al and Si, thus confirming the presence of magnetite aggregates connecting the halloysite particles.
From the EDS analysis, the Al, Si, and Fe atomic percentages were evaluated. Al:Si:Fe molar ratios of 5.  Table 2.  The Fe 3 O 4 -C sample (chosen as reference of the magnetite samples) exhibits a zetapotential of −7.16 mV, comparable to the literature values [48]; this value is not sufficient to achieve a stable suspension, and justifies particle aggregation (see TEM and DLS results).
Zeta-potential values of −36.36, −12.89 and −112.02 mV are obtained for HNT/Fe 3 O 4 -C, HNT/Fe 3 O 4 -SG and HNT/Fe 3 O 4 -H composites. The sample prepared by the hydrothermal process displays the most negative zeta-potential value; this may be due to the carbonaceous component (see TEM results and Section 3.2.) and explains the improved stability of the suspension and the lower mean particle size, as shown by DLS results.
The EDS analysis was applied to display the distribution map of halloysite and magnetite in each composite sample and to evaluate the weight percentage. Figures S3-S5 show the distribution maps of Al, Fe, and Si for the HNT/Fe 3 O 4 -C, HNT/Fe 3 O 4 -H, and HNT/Fe 3 O 4 -SG samples. Independently of the synthetic route, Al and Si were detected in the same areas. The Fe distribution was rather homogeneous in the sol-gel and hydrothermal samples (Figures S4 and S5, respectively), but also in some regions in which Fe prevails were detected. In the co-precipitation composite, Fe prevailed in areas poor in Al and Si, thus confirming the presence of magnetite aggregates connecting the halloysite particles.
From the EDS analysis, the Al, Si, and Fe atomic percentages were evaluated. Al:Si:Fe molar ratios of 5.  Table 2. The halloysite amount in the HNT/Fe 3 O 4 composites was also calculated by thermogravimetric analyses. The thermograms of commercial HNT and composites are shown in Figure 5. The halloysite TG curve (Figure 5a) well compared to the literature data [27]. The mass loss detected at low temperature (below 250 • C) was ascribed to the release of physisorbed water molecules. The steep mass loss observed at about 450 • C gave more insight, as it is due to the dehydroxylation process of the structural Al-OH groups of the aluminosilicate layers. A weight loss of 13.95% was calculated from halloysite stoichiometry. The mass loss detected in the commercial HNT was about 14.60%, in fair agreement with the calculated value.   Field dependence of magnetization was investigated for all the samples at 300 K (Figure 6a,b). Field dependence of magnetization was investigated for all the samples at 300 K (Figure 6a,b). For bare nanoparticles prepared with co-precipitation and sol-gel synthesis methods (Figure 6a), negligible value of reduced remanence magnetization (Mr/Ms) and small value of coercivity were obtained ( Table 3), suggesting that at 300 K most of the nanoparticles were in a superparamagnetic state and just a small fraction of nanoparticles showed a quasi-static behavior. While the zero coercivity in the nanoparticles synthesized with the hydrothermal procedure indicated that all nanoparticles were in a supermagnetic state. Fe3O4-C and Fe3O4-SG samples showed a weak non-saturating character at high field, with respect to the Fe3O4-H sample. Due to the small difference in size between the samples, a non-saturating character showed by samples prepared by sol-gel and co-precipitation techniques can be ascribed to an increase in surface anisotropy, probably due to the presence of magnetic disorder (i.e., canted spin) [49,50] at the particles' surface. This hypothesis was also confirmed by the decrease in MS in SG and C samples. All the HNT nanocomposites showed a decrease in MS with respect to bare nanoparticles in qualitative agreement with TGA and EDS measurements. This behavior confirmed that the amount of magnetic phase decreases along the order Fe3O4-C, Fe3O4-SG, and Fe3O4-H. From a quantitative point of view, if the agreement among magnetization measurements, TGA and EDS, was pretty good for Fe3O4-SG and Fe3O4-H, a difference was observed for Fe3O4-C nanocomposite. In particular, the particles prepared by co-precipitation looked to decrease their MS when prepared as nanocomposites. This can be ascribed to a decrease in nanoparticles' crystallinity that can be observed in the co-precipitation synthesis with respect to hydrothermal and sol-gel syntheses [51,52]. For bare nanoparticles prepared with co-precipitation and sol-gel synthesis methods (Figure 6a), negligible value of reduced remanence magnetization (Mr/Ms) and small value of coercivity were obtained ( Table 3), suggesting that at 300 K most of the nanoparticles were in a superparamagnetic state and just a small fraction of nanoparticles showed a quasi-static behavior. While the zero coercivity in the nanoparticles synthesized with the hydrothermal procedure indicated that all nanoparticles were in a supermagnetic state. Fe 3 O 4 -C and Fe 3 O 4 -SG samples showed a weak non-saturating character at high field, with respect to the Fe 3 O 4 -H sample. Due to the small difference in size between the samples, a non-saturating character showed by samples prepared by sol-gel and coprecipitation techniques can be ascribed to an increase in surface anisotropy, probably due to the presence of magnetic disorder (i.e., canted spin) [49,50] at the particles' surface. This hypothesis was also confirmed by the decrease in M S in SG and C samples. All the HNT nanocomposites showed a decrease in M S with respect to bare nanoparticles in qualitative agreement with TGA and EDS measurements. This behavior confirmed that the amount of magnetic phase decreases along the order Fe 3 O 4 -C, Fe 3 O 4 -SG, and Fe 3 O 4 -H. From a quantitative point of view, if the agreement among magnetization measurements, TGA and EDS, was pretty good for Fe 3 O 4 -SG and Fe 3 O 4 -H, a difference was observed for Fe 3 O 4 -C nanocomposite. In particular, the particles prepared by co-precipitation looked to decrease their M S when prepared as nanocomposites. This can be ascribed to a decrease in nanoparticles' crystallinity that can be observed in the co-precipitation synthesis with respect to hydrothermal and sol-gel syntheses [51,52]. It is well known that the adsorption capacity of the materials is strictly related to their specific surface area [53]. The BET method was applied to investigate the specific surface area of the commercial halloysite and the three HNT/

Preliminary Adsorption Experiments
Before starting the adsorption experiments, control samples (20 mg HNT/Fe 3 O 4 or HNT, 10 mL tap water), not containing OFL, were shaken for 24 h at room temperature. Then, the supernatants were magnetically separated for the pH measurement and analyzed by UV-vis spectrophotometer and HPLC-FD to check the instrumental baseline.
A pH value of 7.7-7.8, similar to that of natural waters, was measured in all samples, thus no additional pH adjustment was performed.
The background noise level was satisfactory for the commercial HNT, HNT/Fe 3 O 4 -C, and HNT/Fe 3 O 4 -H. On the contrary, HNT/Fe 3 O 4 -SG was rinsed with EtOH in an ultrasonic bath for 10 min, centrifuged for 5 min at 4000 rpm, separated, and dried at 50 • C for 1.5 h. The washing step was repeated twice to obtain a good signal-to-noise ratio.

Isotherm and Kinetic Studies
The behavior of the three magnetic HNT composites was evaluated through thermodynamic and kinetic experiments carried out under controlled conditions (see Section 2.3.1) and compared with the commercial HNT.
Adsorption isotherms are commonly used to describe the adsorption process in terms of maximum uptake and the relationship between the amount of adsorbed analyte (q e ) and its concentration in solution at equilibrium (C e ).
To fit the experimental data, the Langmuir and Freundlich models were considered. As shown in Figure 7, the Langmuir model gave the best fitting of the experimental data. Figure 7 shows that all materials were able to adsorb the antibiotic, although the maximum adsorption capacities were quite different. In detail, the highest value, 45 mg g −1 , was obtained for HNT/Fe 3 O 4 -H, while the lowest value was obtained for HNT/Fe 3 O 4 -C, which was equal to 23 mg g −1 . The HNT/Fe 3 O 4 -SG sample had an intermediate value of 31 mg g −1 , close to the commercial HNT (30 mg g −1 ). This trend can be due to both the different amount of HNT present in the samples, ranging from about 30% in HNT/Fe 3 O 4 -C to more than 80% in HNT/Fe 3 O 4 -SG (see Table 2), and to the possible presence of some carbonaceous component related to the glucose added during HNT/Fe 3 O 4 -H synthesis. In fact, as reported by Tian et al. [30], the carbon/organic groups formed on the HNTs not only favor the Fe 3 O 4 nanoparticle nucleation, but also may improve the analyte adsorption. On the contrary, no difference in the adsorption mechanism was observed among all materials. The Langmuir model, which describes a monolayer coverage, gives the best fitting of the experimental data, as confirmed by the good correlation coefficient R 2 and χ 2 values. rials 2022, 12, x FOR PEER REVIEW fitting of the experimental data, as confirmed by the good correlation coeff values. The experimental q max values of HNT/Fe3O4-C, HNT/Fe3O4-H, and were in agreement with the calculated ones, and fell within the OFL ad reported in the literature for other clays, i.e., 3.2 mg g −1 on kaolinite [54], calcined Verde-lodo bentonite clay [55]).
The isothermal parameters calculated by dedicated software are listed in Table 4. Concerning the kinetic aspect, quantitative adsorption occurred in less than five minutes in the presence of all the magnetic composites. As shown in Figure 8, a satisfactory fitting is obtained by applying the pseudo-second-order model, thus, considering a chemisorption process. For commercial HNT, the adsorption was instantaneous, thus, it was not possible to discriminate between the two models. The calculated kinetic parameters are shown in Table 5. Magnetic HNTs were also tested under environmental conditions, i.e., µg L −1 OFL concentration, tap and river waters, WWTP effluent (see Table S2 for the physicochemical parameters).
An amount of 20 mg of each material was suspended in 10 mL of each water sample, river water and WWTP effluent samples spiked with 10 µg L −1 OFL (C0) and shaken for 24 h. Then, the suspensions were magnetically separated and the supernatants were filtered on a 0.22 μm nylon syringe filter before HPLC-FD analysis to quantify the drug content (Ce).
The removal efficiency (R%) was calculated according to Equation (3): where C0 is the initial OFL concentration and Ce is the OFL concentration in solution at the equilibrium.
The obtained results were reported in Figure 9. The calculated kinetic parameters are shown in Table 5.

Ofloxacin Removal from Real Waters Samples
Magnetic HNTs were also tested under environmental conditions, i.e., µg L −1 OFL concentration, tap and river waters, WWTP effluent (see Table S2 for the physicochemical parameters).
An amount of 20 mg of each material was suspended in 10 mL of each water sample, river water and WWTP effluent samples spiked with 10 µg L −1 OFL (C 0 ) and shaken for 24 h. Then, the suspensions were magnetically separated and the supernatants were filtered on a 0.22 µm nylon syringe filter before HPLC-FD analysis to quantify the drug content (C e ).
The removal efficiency (R%) was calculated according to Equation (3): where C 0 is the initial OFL concentration and C e is the OFL concentration in solution at the equilibrium. The obtained results were reported in Figure 9. The investigated HNT/Fe3O4 composites gained an antibiotic removal ≥ 90% d different aqueous matrix constituents and other potential contaminants. The di amount of Fe3O4 in each composite did not affect the adsorption process; on the co the Fe3O4 percent in HNT/Fe3O4-C, higher than in HNT/Fe3O4-H and HNT/Fe3O4vored its complete magnetic recovery from the media after the use with no add centrifugation step.

Reusability and Post-Use Characterization of HNT/Fe3O4-C
Among the investigated magnetic HNTs, the HNT/Fe3O4-C sample ensured a titative OFL removal in different real water samples and excelled for its magnetic p ties. For these reasons, its reusability was explored.
The HNT/Fe3O4-C sample was suspended in 10 mL tap water containing OFL L −1 . After 1 h, HNT/Fe3O4-C was magnetically separated, and the supernatant w lyzed by HPLC-FD. Then the recovered sorbent material was suspended for a secon in 10 mL tap water samples containing OFL 10 µg L −1 . After 1 h contact, the susp material was magnetically separated, and the OFL concentration in the solution was ured. A third cycle was carried out following the same procedure. Figure 10 shows the adsorbed OFL percentage after each adsorption cycle. T sorbed antibiotic amount slightly decreased from 95% after the first use to 75% af third one.
This trend may be ascribed to a small loss of material during its magnetic sep from the sample solution and not to matrix interference, as XRPD analysis demons The recovered sorbent material after three adsorption cycles was analyzed by and compared to the synthesized HNT/Fe3O4-C sample. The two diffraction p

Reusability and Post-Use Characterization of HNT/Fe 3 O 4 -C
Among the investigated magnetic HNTs, the HNT/Fe 3 O 4 -C sample ensured a quantitative OFL removal in different real water samples and excelled for its magnetic properties. For these reasons, its reusability was explored.
The HNT/Fe 3 O 4 -C sample was suspended in 10 mL tap water containing OFL 10 µg L −1 . After 1 h, HNT/Fe 3 O 4 -C was magnetically separated, and the supernatant was analyzed by HPLC-FD. Then the recovered sorbent material was suspended for a second time in 10 mL tap water samples containing OFL 10 µg L −1 . After 1 h contact, the suspended material was magnetically separated, and the OFL concentration in the solution was measured. A third cycle was carried out following the same procedure. Figure 10 shows the adsorbed OFL percentage after each adsorption cycle. The adsorbed antibiotic amount slightly decreased from 95% after the first use to 75% after the third one.
This trend may be ascribed to a small loss of material during its magnetic separation from the sample solution and not to matrix interference, as XRPD analysis demonstrates.
The recovered sorbent material after three adsorption cycles was analyzed by XRPD and compared to the synthesized HNT/Fe 3 O 4 -C sample. The two diffraction patterns ( Figure S6) are really comparable, confirming the sorbent material does not undergo degradation processes with use.

Acute Toxicity Test with Daphnia magna
For the toxicity test, a single concentration, equal to 0.2 g L −1 of HNT, Fe3O4, and HNT/Fe3O4-C was tested. This concentration reflected a potential residual amount of each material in waters after depollution treatment.
All the individuals efficiently ingested the administered materials over 48 h of exposure (Figure 11), as shown by their presence in the digestive tract of exposed individuals. No mortality occurred in the control group. Despite the ingestion of all the materials, the 48 h exposure to 0.2 g L −1 of HNT and HNT/Fe3O4-C did not induce the mortality of any daphnid, while the viability of the individuals included in the Fe3O4 experimental group was slightly decreased compared to the corresponding control, accounting for the 96 ± 9%.

Conclusions
In the present work, magnetic halloysite nanotubes were successfully synthesized by three different approaches: co-precipitation, hydrothermal, and sol-gel method. The applied characterization techniques demonstrate that the nanometric-sized Fe3O4 (diameter of about 10 nm) were formed and connected to the HNT particles. Magnetic phase abundance depended on the synthetic route and was evaluated by EDS and TGA analyses, as well as by magnetization data. Thermodynamic and kinetic experiments suggested that

Acute Toxicity Test with Daphnia magna
For the toxicity test, a single concentration, equal to 0.2 g L −1 of HNT, Fe 3 O 4 , and HNT/Fe 3 O 4 -C was tested. This concentration reflected a potential residual amount of each material in waters after depollution treatment.
All the individuals efficiently ingested the administered materials over 48 h of exposure ( Figure 11), as shown by their presence in the digestive tract of exposed individuals.

Acute Toxicity Test with Daphnia magna
For the toxicity test, a single concentration, equal to 0.2 g L −1 of HNT, Fe3O4, and HNT/Fe3O4-C was tested. This concentration reflected a potential residual amount of each material in waters after depollution treatment.
All the individuals efficiently ingested the administered materials over 48 h of exposure (Figure 11), as shown by their presence in the digestive tract of exposed individuals. No mortality occurred in the control group. Despite the ingestion of all the materials, the 48 h exposure to 0.2 g L −1 of HNT and HNT/Fe3O4-C did not induce the mortality of any daphnid, while the viability of the individuals included in the Fe3O4 experimental group was slightly decreased compared to the corresponding control, accounting for the 96 ± 9%.

Conclusions
In the present work, magnetic halloysite nanotubes were successfully synthesized by three different approaches: co-precipitation, hydrothermal, and sol-gel method. The applied characterization techniques demonstrate that the nanometric-sized Fe3O4 (diameter of about 10 nm) were formed and connected to the HNT particles. Magnetic phase abundance depended on the synthetic route and was evaluated by EDS and TGA analyses, as well as by magnetization data. Thermodynamic and kinetic experiments suggested that No mortality occurred in the control group. Despite the ingestion of all the materials, the 48 h exposure to 0.2 g L −1 of HNT and HNT/Fe 3 O 4 -C did not induce the mortality of any daphnid, while the viability of the individuals included in the Fe 3 O 4 experimental group was slightly decreased compared to the corresponding control, accounting for the 96 ± 9%.

Conclusions
In the present work, magnetic halloysite nanotubes were successfully synthesized by three different approaches: co-precipitation, hydrothermal, and sol-gel method. The applied characterization techniques demonstrate that the nanometric-sized Fe 3 O 4 (diameter of about 10 nm) were formed and connected to the HNT particles. Magnetic phase abundance depended on the synthetic route and was evaluated by EDS and TGA analyses, as well as by magnetization data. Thermodynamic and kinetic experiments suggested that HNT/Fe 3 O 4 composites can be considered as performing materials for ofloxacin adsorption. All the investigated samples were able to quantitatively reduce the antibiotic concentration under realistic conditions and, more interestingly, the sample obtained by the co-precipitation synthetic approach-the most cost-effective-was also easily magnetically removed from the media after treatment and reused for three cycles with no degradation. The ecotoxicity test performed on the freshwater organism D. magna completed the characterization of this adsorbent material and confirmed that it might be safely applied in water depuration processes.  Figure S6: X-ray diffraction pattern of the HNT/Fe 3 O 4 -C sample as-prepared (black line) and after three cycles of OFL recover (red line); Table S1: Mean particle size and intensity determined by DLS analysis; Table S2: Physico-chemical characterization of tap and river water samples, and WWTP effluent.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.