Fe3O4-TiO2 Thin Films in Solar Photocatalytic Processes

The optical properties of 5wt% Fe3O4-TiO2 thin films were evaluated in detail with the aim of proposing a mechanism for solar photocatalytic processes and highlighting the advantages over the use of bare TiO2. The results showed that the incorporation of 5wt% Fe3O4 enhanced the optical properties by a redshift to a wavelength in the visible range, reducing the anatase/rutile band gap energy from 3.2 eV to 2.8 eV. Photoluminescence studies reveal a superior separation efficiency of photoexcited electron-hole pairs when Fe3O4 nanoparticles (NPs) are present in the photocatalyst. X-ray photoelectron spectroscopy spectra confirm the presence of Fe3O4 and existence of a chemical bonding between TiO2 and Fe3O4 NPs. Moreover, in this study, a mechanism of solar photocatalytic processes involving Fe3O4-TiO2 thin films is proposed and it is supported by experimental results. Finally, solar photocatalytic experiments were carried out, indicating that the effectiveness for the removal of the selected pharmaceutical is considerably improved when the composite material is used as catalyst. Furthermore, it was demonstrated that the photocatalytic activity of the prepared Fe3O4-TiO2 thin films depends on their thickness, achieving the highest pharmaceutical removal yields using the 2 µm thick sample. The stability and reusability of the catalyst was confirmed studying the photocatalytic activity over three cycles.


Introduction
In recent years, semiconductor photocatalytic processes have found wide applications in water treatment [1]. Titanium dioxide (TiO 2 ) has been considered as the most widely used semiconductor oxide photocatalyst due to its attractive attributes, such as low cost, corrosion resistance, excellent chemical stability, non-toxicity and high photocatalytic activity [2]. However, the use of TiO 2 nanoparticles (NPs) in solar photocatalysis for environmental remediation has some technical limitations, such as the high band gap of TiO 2 (~3.2 eV) [3,4], close to UV spectral range. For this reason, this photocatalyst is triggered by near UV radiation. Thus, it takes advantage of only a small range of the solar spectrum (about 3-4%), which limits the potential of solar light as an energy source. Additionally, the photogenerated electron-hole pairs are liable to recombination, resulting in low quantum efficiency of TiO 2 [5].
Moreover, the use of TiO 2 as a photocatalytic material would increase the overall cost if it is utilized in powder form, which constitutes a drawback. This is so due to the necessity of performing an additional filtration stage in order to separate this powder from the water in solutions [6]. Thus, the separation of the semiconductor TiO 2 catalyst after water treatment remains as the major obstacle towards the practicality of this technology [7]. In addition, nanoparticles, dispersed in the medium, can be dangerous due to their possible cytotoxic and inflammatory effects [8,9].
In order to overcome these issues, there are several studies in the literature in which Fe 3 O 4 -TiO 2 nanoparticles systems have already been investigated. These studies revealed that the presence of Fe 3 O 4 enhance the photocatalytic activity of TiO 2 and that modified TiO 2 NPs can be separated from water by means of an external magnetic field [10][11][12][13][14]. It is important to note that composites/heterojunction catalysts enhance the electron-hole separation and prevent charge back transfer [15,16].
However, we consider that modifying TiO 2 with Fe 3 O 4 would not lead to a feasible solution, since: (1) the suspended NPs tend to aggregate, especially when high concentrations are used [6] and furthermore, (2) the use of magnetic separation is complicated in real water treatment applications.
Therefore, a more realistic proposal would consist of depositing thin film coatings onto a substrate material [17]. In this sense, it is important to note that there are only few studies about the immobilization of Fe 3 O 4 -TiO 2 NPs on substrates as thin films, being very few of them those related to its use in photocatalytic processes. For instance, Wei et al. [18] prepared Fe 3 O 4 -TiO 2 composite films to enhanced the ability to store energy under low operating electric field. Zhang et al. [19] used Fe 3 O 4 nanoparticles and molecularly imprinted TiO 2 in the fabrication of a novel photoelectrochemical biosensor. Yeh et al. [20] evaluated the antimicrobial efficiency of immobilized Fe 3 O 4 /TiO 2 powder on glass substrate. Nevertheless, to the best of our knowledge, no studies have been reported on the use of Fe 3 O 4 -TiO 2 thin film for solar photocatalytic processes. Therefore, the current work represents a highly innovative study, which could open a new field of application for this particular material.
Thus, the purpose of this paper is to develop a Fe 3 O 4 -TiO 2 composite material with high photocatalytic activity and good stability and their immobilization on substrates. In this study, thin films are prepared using a "sandwich" type structure [21], which prevents the release of NPs to the medium. Furthermore, the photocatalytic efficiency is evaluated by studying the degradation of an antibiotic, sulfamethoxazole (SFX), in the presence of these films. SFX is a sulphonamide antibiotic, which is one of the common pollutants in surface water [4,[22][23][24][25].
The main motivation for using an antibiotic as a contaminant model in solar photocatalytic experiments is that conventional methods implemented in wastewater treatment plants are not efficient enough to remove pharmaceutical contaminants [25][26][27]. Therefore, these pollutants can arrive in rivers, lakes, seas, aquifers and occasionally, they can be found in tap water, although in very small quantities [28]. The presence of pharmaceutical compounds in surface and wastewater, all over the world, involves a significant risk for aquatic ecosystems and for all living beings and the environment [29][30][31]. Considering the effects and consequences to the environment, animals and plants, and regarding the frequency with which pharmaceuticals have been detected in water, it is expected that these types of pollutants will be candidates for future regulations.
Therefore, the research on alternative, resultative and cost-efficient technologies is a field of great interest for all the technological and scientific community.

Chemicals
Iron (II) chloride (FeCl 2 ), iron (III) chloride (FeCl 3 ) and ammonium hydroxide (NH 4 4 ) nanoparticles have been prepared following a classical bottom-up chemical route based on a co-precipitation reaction. With the aim to obtain the mixed (II and III) iron oxide, we have started from Fe 2+ and Fe 3+ precursor salts, FeCl 2 and FeCl 3 , respectively, and we have promoted the controlled co-precipitation of both salts. Briefly, 0.99 g of FeCl 2 and 2.76 g of FeCl 3 are dissolved in 100 mL of MilliQ water. The solution is placed in a three-necked flask and heated at 80 • C under reflux and with magnetic stirring, in a nitrogen atmosphere, in order to avoid the formation of non-magnetic iron oxide phases. After heating the solution for 10 min, 10 mL of 25% ammonium hydroxide solution is swiftly added to the flask, promoting the fast and homogeneous nucleation step to form the magnetite cores. The solution is still heated, keeping the temperature at 80 • C, for 30 additional minutes, in which the previously formed nuclei continue growing. Once the solution was cooled down to room temperature, it was filtered three times (using a 0.1 µm Millipore membrane), adding ethanol, in order to eliminate the excess of precursor species and/or residues from the reaction. As result, we obtained a magnetic black powder composed by magnetite NPs.

Preparation of Fe 3 O 4 -TiO 2 Nanoparticles
For preparation, 0.15 g of synthetized Fe 3 O 4 NPs were initially dispersed in 10 mL deionized water and kept under ultrasonic irradiation for 30 min. Throughout this procedure, a stable ferrofluid was obtained. Next, the dispersion was added, drop by drop, to 3 g of TiO 2 NPs, forming a slurry that was finally dried at 110 • C for 24 h. Finally, samples were annealed at 400 • C for 30 min. This processing would lead to a 5wt% load of Fe 3 O 4 NPs in TiO 2 . This concentration value (around 5wt%) has shown to give the best optical and structural properties as well as superior photocatalytic activity in FeTiO 2 samples [32][33][34].

Preparation of Photocatalyst Films
Photocatalyst coatings on flat glass supports were prepared according to a multilayer approach proposed in previous works by scientists of our research group [4,21]. According to the methodology, the structure of the photocatalyzer responds to a "sandwich" approach consisting of two very thin anatase films (~50 nm), obtained by sol gel [35], delimiting a thicker layer obtained from a suspension of TiO 2 /Fe 3 O 4 nanoparticles.
Anatase films were deposited from a precursor sol prepared by hydrolysis of titanium n-butoxide (TBOT) in a solution containing acetylacetone, ethanol and acidified water (HNO 3 , pH = 1) in a molar ratio TBOT:acetylacetone:H 2 O:EtOH of 1:0.5:2:35 as described in Blanco et al. (2015) [35]. Dip coating was performed from aged sol at precisely controlled withdrawal speed (100 mm/min) in a homemade apparatus on a microscope slide and the resulting film was subsequently dried at 150 • C for 30 min and annealed at 1 • /min until 500 • C, in order to ensure that TiO 2 films is formed by small anatase crystallites while preserving a certain porosity level in the films which provides high surface areas [36].
For the central part of the sandwich coating, a suspension of the Fe 3 O 4 -TiO 2 nanoparticles in an acetylacetone-ethanol solution (4:1 vol ratio, 40 g/L) was prepared by using high power ultrasounds (20 KHz). Films were dip coated on the top of the previous anatase film at 100 mm/min, layer by layer. After every five depositions, the layers are dried at 150 • C for 30 min and the process is repeated until reaching the total number of nanoparticulated layers are deposited.
Finally, the coating was sealed with a new single anatase thin film prepared as described before. The final photocatalyst is obtained after annealing in air for 1 h at 400 • C. For comparison, a similar non-doped "sandwich" coating was also prepared using a suspension of TiO 2 NPs a precursor of the central layer. The resulting film was used as a reference.

Photocatalyst Characterization
Magnetite NPs were studied by Transmission Electron Microscopy (TEM) and Scanning-Transmission Electron Microscopy (STEM), using a JEOL JEM-2100 EX TEM and a FEI TALOS F200X STEM microscope, with thermionic effect and field effect electron emission guns, respectively, both working at 200 kV. Both microscopes are located at the facilities of Servicios Centrales de Investigación, Ciencia y Tecnología (SC-ICYT) of the University of Cádiz (UCA). NPs were deposited in TEM holey carbon grids according to a previously stablished method [37]. Dynamic Light Scattering (DLS) was utilized for measuring the hydrodynamic NP size. Measurements were carried out at 25 • C, using a Malvern Zetasizer Nano-ZS ("Malvern Instruments", Worcestershire, U.K.), with a 1 cm path cell. This equipment also allows for the Z-potential to be measured which evaluate colloidal solutions stability, as well as NP surface charge.
In order to study the effect of incorporating Fe 3 O 4 NPs on the optical properties of TiO 2 NPs and estimate their energy band gap, UV-Vis absorbance and diffuse reflectance measurements were carried out using an Agilent Cary 5000 UV-Vis-NIR double-beam spectrophotometer. The spectra in the 200 to 800 nm range were registered in an integrating sphere. The resulting diffuse reflectance spectra were transformed into apparent absorption spectra using the Kubelka-Munk function (F(R)). The direct and indirect optical band gap of the materials was determined through the construction of Tauc plots by plotting (F(R)hν) n against (hν), with n = 2 or n = 1 2 , for direct and indirect transitions, respectively. The optical band gap was obtained by extrapolating the linear part of this plot to the energy axis. Furthermore, the photoluminescence spectra were obtained using a Fluorolog-QM spectrometer (Horiba) equipped with a continuous 75 W Xenon arc lamp, double monochromators at the excitation and the emission and a photomultiplier tube detector (PMT-920IS). Temperature controlled experiments were performed with a cryostat (CS204-FMX-1SS, ARSCRYO) that can be operated between 4 and 350 K.
X-ray photoelectron spectroscopy (XPS) was recorded using a Kratos Axis Ultra DLD spectrometer. XPS measurements were performed using a monochromatic Al Kα radiation at 1486.6 eV and X-ray power of 150 W. The analyser was operated in constant analyser energy transmission (CAE) mode, with pass energy of 80 eV and 20 eV for the survey scan spectrum and for high-resolution core levels, respectively. Surface charging effects were compensated by making use of the Kratos coaxial neutralization system. The binding energy (BE) scale was calibrated with respect to the C 1 s signal at 284.8 eV. XPS data analysis was performed by using CasaXPS Software, version 2.3.19rev1.1 m (Neal Fairley, Casa Software Ltd., Teignmouth, Devon, UK).
Morphology and thickness of prepared thin films was followed by Scanning Electron Microscopy (SEM), recording cross-section images obtained using the secondary electron detector of a FEI Nova NanoSEM 450 scanning microscope (also located at the UCA-SC-ICYT facilities), with field effect electron emission gun, working at a voltage and a distance of 20 kV 4.6 mm, respectively.
Atomic Force Microscopy (AFM) images were obtained with a DI Veeco Nanoscope Multimode IIIa equipment using Bruker TESP-V2 probes and working in tapping mode. AFM images were processed with the open-source software Gwyddion [38].

Photocatalytic Experiments
Experiments were carried out using a custom-made photocatalytic system, consisting of a magnetic-stirred reactor exposed to sunlight, equipped with a recirculating peristaltic pump (Selecta Percom N-M) and coupled to an Avantes fiber optic UV-Vis-NIR spectrometer (Ocean Optics DT-Mini-2GS light source) that allows continuous monitoring [4].
For each experiment, 20 mL of an aqueous solution of SFX was charged into the photoreactor (5.6 mm diameter). SFX initial concentration was 5 ppm. A Fe 3 O 4 -TiO 2 thin film (2.5 × 2.8 cm 2 ) was also place into the reactor. Experiments were conducted at Campus de Puerto Real, University of Cadiz (Spain, local latitude 36 • 31 N) on sunny days. A radiometer UV34 (PCE Ibérica, Spain) was used to measure the intensity of solar radiation.
Absorbance measurements were determined using a UV-Vis-NIR fibre optic spectrometer and its evolution over radiation dose was monitored due to a peristaltic pump that recirculated (6 mL/min) the solution treated by solar photocatalysis.
Cumulative UV dose was calculated as shown in Equation (1).
where Q UVn , Q UVn−1 is the cumulative irradiated UV energy received at times n and n − 1, respectively; ∆t n the time interval between two sampling times; UV n the average incident radiation for each time interval (W/m 2 ); Vi the illuminated volume of the reactor and V T the total volume of the reactor. Experimental results can also be expressed using t 30W (see Equation (2)). It is a normalized value that indicates the time spent on the experiment in the hypothetically situation of a constant UV radiation intensity of 30 W/m 2 , during its development [39]. A UV intensity of 30 W/m 2 resembles conditions around a sunny noon in the vicinity of Cádiz.
where t 30Wn t 30Wn−1 are the normalised illumination time of samples at times n and n − 1, respectively.

Fe 3 O 4 and Fe 3 O 4 -TiO 2 NPs Morphology, Size, Crystallinity and Composition
Figure 1a corresponds to a SEM micrograph, showing a general view of the synthetized magnetite NPs. This image and others alike were utilized as a first indication of the quality of the synthetized NPs and to calculate the average particle size. Micrograph reveals the formation of relatively small nanoparticles with homogeneous size distribution. Although the strong magnetic character of the material leads to the NP aggregation, the higher atomic number of iron gives a good contrast that allows distinguishing the formation of approximately spherical particles whose concentration is high enough for performing a statistical study on the size distribution. In this sense, the histogram of this magnitude is presented in Figure 1b. As it can be observed, the experimental particle size distribution fits well with a Gaussian distribution. This fit indicates a NP average size of 8.7 nm with a small standard deviation (1.1 nm). NP crystalline structures can be observed in High Resolution TEM (HRTEM) images, as the ones in Figure 1c and, at higher magnification, Figure 1d. The averaged experimental distances between crystalline planes are 3.00 and 3.43 Å, respectively. These distances best agree with those of the theoretical ones for (111) and (200) planes of magnetite [40]. These measured planes have been labelled and marked in the figure.
On the other hand, Table 1 shows the average hydrodynamic size, PDI (polydispersity index) and Z-Potential values for magnetite NPs. As expected, the hydrodynamic size determined by the DLS technique is significantly higher in comparison with the one obtained by TEM. This effect is assigned to the presence of extra hydrate layers in aqueous medium. Moreover, some agglomeration of the nanoparticles in aqueous solutions is expected due to the magnetism of magnetite NPs. However, this average hydrodynamic size, together with the relatively high value of the Z-potential, indicates the NPs colloidal stability. In addition, the positive value of the Z-potential suggests the presence of positive charge at the NP surface, which could promote subsequent crosslinking with the matrix. On the other hand, Table 1 shows the average hydrodynamic size, PDI (polydispersity index) and Z-Potential values for magnetite NPs. As expected, the hydrodynamic size determined by the DLS technique is significantly higher in comparison with the one obtained by TEM. This effect is assigned to the presence of extra hydrate layers in aqueous medium. Moreover, some agglomeration of the nanoparticles in aqueous solutions is expected due to the magnetism of magnetite NPs. However, this average hydrodynamic size, together with the relatively high value of the Z-potential, indicates the NPs colloidal stability. In addition, the positive value of the Z-potential suggests the presence of positive charge at the NP surface, which could promote subsequent crosslinking with the matrix. In order to find out whether or not the utilized TiO 2 doping process produce a material in which the magnetite NPs are actually imbibed in the titanium oxide matrix, small volumes of Fe3O4-TiO2 were deposited on holey-carbon TEM grids, for their compositional study. One of these volumes is presented in Figure 2a, which corresponds to a TEM micrograph obtained in Diffraction Contrast-Bright Field (BF) mode. The darker areas correspond to orientations of the material (TiO2) along which the electron beam is parallel to high atomic density crystallographic planes. This doped TiO2 volume was later observed using STEM techniques, such as High Angle Annular Dark Field (HAADF) and Energy Dispersive X-ray spectroscopy (EDX), as presented in Figure 2b,c. HAADF images intensity highly depends on the atomic mass and thickness of the material. Therefore, and taking into consideration the small size of the NPs and roughness and irregularity of the TiO2 volume under study, it is difficult to distinguish these NPs in HAADF images. EDX spectra maps, on the other hand, easily revealed the positions of such particles in the titanium oxide matrix, as it can be observed in Figure 2c. In this map, red spots correspond to Fe characteristic X-rays emitted from the material, and green dots correspond to Ti-signal. Weaker Fe signals from the volume could be explained by different particle sizes or, more probably, because of the high possibility of the NP being totally contained inside the TiO2 volume. This was corroborated by EDX point spectra taken from the NPs. Figure 2d shows  In order to find out whether or not the utilized TiO 2 doping process produce a material in which the magnetite NPs are actually imbibed in the titanium oxide matrix, small volumes of Fe 3 O 4 -TiO 2 were deposited on holey-carbon TEM grids, for their compositional study. One of these volumes is presented in Figure 2a, which corresponds to a TEM micrograph obtained in Diffraction Contrast-Bright Field (BF) mode. The darker areas correspond to orientations of the material (TiO 2 ) along which the electron beam is parallel to high atomic density crystallographic planes. This doped TiO 2 volume was later observed using STEM techniques, such as High Angle Annular Dark Field (HAADF) and Energy Dispersive X-ray spectroscopy (EDX), as presented in Figure 2b,c. HAADF images intensity highly depends on the atomic mass and thickness of the material. Therefore, and taking into consideration the small size of the NPs and roughness and irregularity of the TiO 2 volume under study, it is difficult to distinguish these NPs in HAADF images. EDX spectra maps, on the other hand, easily revealed the positions of such particles in the titanium oxide matrix, as it can be observed in Figure 2c. In this map, red spots correspond to Fe characteristic X-rays emitted from the material, and green dots correspond to Ti-signal. Weaker Fe signals from the volume could be explained by different particle sizes or, more probably, because of the high possibility of the NP being totally contained inside the TiO 2 volume. This was corroborated by EDX point spectra taken from the NPs. Figure 2d shows one of such spectra, obtained from the region "1" indicated in Figure 2a with a red square. This spectrum presents a Fe-peak, corresponding to the NP, and a Ti-peak (since the STEM probe interacts with a certain volume surrounding the NP). EDX point spectrum in Figure 2e, taken from the region labelled as "2" in Figure 2a, proves that the Fe-peak in the previous spectrum is not due to sample contamination. Cu-peak in both spectra comes from the Cu mesh of the utilized TEM grid. It can be concluded, thus, from this EDX map, that the utilized process indeed produces a Fe 3 O 4 -TiO 2 slurry, with NPs randomly distributed in the TiO 2 . HRTEM images were obtained in order to investigate possible changes in the TiO 2 structure. In this sense, images such as the one in Figure 2f (obtained from the region labelled as "3" in Figure 2a) and, at higher magnification, 2g allow measuring averaged lattice distances of 2.36 and 3.52 Å, which are in good agreement well with the theoretical distances for (110) and (101) for the atomic planes of the anatase phase of TiO 2 [41]. The dashed lines in the figure are representative of the indicated lattice distances. Neither of the HRTEM images recorded shows, in any case, a change in the TiO 2 crystalline structure, indicating the absence of a possible phase change or measurable stress due to the introduction of NPs.
This spectrum presents a Fe-peak, corresponding to the NP, and a Ti-peak (since the STEM probe interacts with a certain volume surrounding the NP). EDX point spectrum in Figure  2e, taken from the region labelled as "2" in Figure 2a, proves that the Fe-peak in the previous spectrum is not due to sample contamination. Cu-peak in both spectra comes from the Cu mesh of the utilized TEM grid. It can be concluded, thus, from this EDX map, that the utilized process indeed produces a Fe3O4-TiO2 slurry, with NPs randomly distributed in the TiO2. HRTEM images were obtained in order to investigate possible changes in the TiO2 structure. In this sense, images such as the one in Figure 2f (obtained from the region labelled as "3" in Figure 2a) and, at higher magnification, 2g allow measuring averaged lattice distances of 2.36 and 3.52 Å, which are in good agreement well with the theoretical distances for (110) and (101) for the atomic planes of the anatase phase of TiO2 [41]. The dashed lines in the figure are representative of the indicated lattice distances. Neither of the HRTEM images recorded shows, in any case, a change in the TiO2 crystalline structure, indicating the absence of a possible phase change or measurable stress due to the introduction of NPs.

Optical Properties: UV-Vis Spectra, Bandgap Energy and Photoluminescence Studies
The catalysts were deposited in the form of thin films. Figure 3 shows the UV-Vis absorption spectra of Fe3O4-TiO2 and TiO2 photocatalysts, where it is possible to assign some common features in the spectral range beyond 300 nm, despite the strong absorption of the substrate at this wavelength. TiO2 spectrum shows a strong absorption corresponding to the absorption edge of semiconductor TiO2 below 400 nm but, in the spectral range above 420 nm, shows very little absorption. Conversely, spectrum corresponding to Fe3O4-TiO2 sample exhibits a broad absorption across the whole visible spectrum, as was previously described in the literature [42]. The catalysts were deposited in the form of thin films. Figure 3 shows the UV-Vis absorption spectra of Fe 3 O 4 -TiO 2 and TiO 2 photocatalysts, where it is possible to assign some common features in the spectral range beyond 300 nm, despite the strong absorption of the substrate at this wavelength. TiO 2 spectrum shows a strong absorption corresponding to the absorption edge of semiconductor TiO 2 below 400 nm but, in the spectral range above 420 nm, shows very little absorption. Conversely, spectrum corresponding to Fe 3 O 4 -TiO 2 sample exhibits a broad absorption across the whole visible spectrum, as was previously described in the literature [42].
Optical band gap energy determination of the doped catalyst and its components were evaluated from UV-Vis diffuse reflectance spectra by applying the Kubelka-Munk (K-M) function [21,43]. The K-M function, F(R), establishes a relationship between absorbed, scattered and diffused reflected light by the sample, and is defined as: where R is the wavelength dependent sample diffuse reflectance, measured with the spectrophotometer integrating sphere. The use of this model for this kind of coating has been justified by considering the granular structure for these optically rough films [36].  Optical band gap energy determination of the doped catalyst and its components were evaluated from UV-Vis diffuse reflectance spectra by applying the Kubelka-Munk (K-M) function [21,43]. The K-M function, F(R), establishes a relationship between absorbed, scattered and diffused reflected light by the sample, and is defined as: where R is the wavelength dependent sample diffuse reflectance, measured with the spectrophotometer integrating sphere. The use of this model for this kind of coating has been justified by considering the granular structure for these optically rough films [36]. Thus, the Tauc method can be applied to the function (F(R)*E) m versus the photon energy, E, with m = 1/2 or m = 2 for indirect or direct interband transitions, respectively. The (F(R)*E) m function will present a linear region for E values close to Eg where the relationship (F(R)*E) m = k(E − Eg) applies. In this way, the gap energy of Fe3O4-TiO2 thin film, Eg, can be obtained from the extrapolation to zero of the linear regression in this region of the plot. Both direct and indirect bandgaps ( Table 2) calculated in this way from plots using m = 1/2 and m = 2, respectively, are collected in Table 2. It can be noticed that materials shown in Table 2 exhibit an indirect bandgap smaller than the direct band gap, suggesting that the fundamental bandgap for these materials could be considered as an indirect bandgap [44,45]. The direct or indirect nature of the bandgap transition is an essential parameter of semiconductors for photocatalytic applications. Thus, when a semiconductor has an indirect bandgap and the electric dipole transition from valence band maximum to conduction band minimum is allowed, then the electron-hole pairs recombine with a lower probability [46].
As it can be seen in Table 2, the calculated indirect bandgap of TiO2 is nearly 3.2 eV but it is reduced when Fe3O4 NPs are included in the material. The small indirect energy Thus, the Tauc method can be applied to the function (F(R)*E) m versus the photon energy, E, with m = 1/2 or m = 2 for indirect or direct interband transitions, respectively. The (F(R)*E) m function will present a linear region for E values close to E g where the relationship (F(R)*E) m = k(E − E g ) applies. In this way, the gap energy of Fe 3 O 4 -TiO 2 thin film, E g , can be obtained from the extrapolation to zero of the linear regression in this region of the plot. Both direct and indirect bandgaps ( Table 2) calculated in this way from plots using m = 1/2 and m = 2, respectively, are collected in Table 2. It can be noticed that materials shown in Table 2 exhibit an indirect bandgap smaller than the direct band gap, suggesting that the fundamental bandgap for these materials could be considered as an indirect bandgap [44,45]. The direct or indirect nature of the bandgap transition is an essential parameter of semiconductors for photocatalytic applications. Thus, when a semiconductor has an indirect bandgap and the electric dipole transition from valence band maximum to conduction band minimum is allowed, then the electron-hole pairs recombine with a lower probability [46].
As it can be seen in Table 2, the calculated indirect bandgap of TiO 2 is nearly 3.2 eV but it is reduced when Fe 3 O 4 NPs are included in the material. The small indirect energy gap obtained for magnetite NPs is consistent with the fact that Fe 3 O 4 -TiO 2 materials present absorption in the visible spectrum region.
This lower bandgap energy of the doped material is due to the fact that the presence of Fe 3 O 4 NPs introduce a series of localized occupied states into the bandgap of TiO 2 lattice [47].
Narrowing the bandgap of TiO 2 is an important strategy to enhance its photocatalytic activity. It has been shown that TiO 2 surface loading with Fe 3 O 4 nanoparticles can be used to enhance the electron-hole pairs generation efficiency to stimulate photocatalytic reaction [48].
PL studies were conducted to elucidate the effect of the electron-hole annihilation on the photocatalytic properties. The steady state photoluminescence spectra of synthesized photocatalysts were registered using an excitation wavelength of 325 nm and different temperatures (77,206 and 273 K). Our results showed (Figure 4) that pure TiO 2 exhibits the highest PL emission intensity, what can be directly related to a high electron-hole recombination rate [49]. PL intensity of Fe 3 O 4 -TiO 2 was significantly smaller, indicating a noteworthy lower recombination rate, i.e., a superior separation efficiency of the photoexcited electron-hole pairs [49]. Thus, the addition of Fe 3 O 4 nanoparticles on the surface of TiO 2 can significantly reduce the electron-hole pairs recombination by accepting the electrons from TiO 2 [48]. activity. It has been shown that TiO2 surface loading with Fe3O4 nanoparticles can be used to enhance the electron-hole pairs generation efficiency to stimulate photocatalytic reaction [48].
PL studies were conducted to elucidate the effect of the electron-hole annihilation on the photocatalytic properties. The steady state photoluminescence spectra of synthesized photocatalysts were registered using an excitation wavelength of 325 nm and different temperatures (77,206 and 273 K). Our results showed (Figure 4) that pure TiO2 exhibits the highest PL emission intensity, what can be directly related to a high electron-hole recombination rate [49]. PL intensity of Fe3O4-TiO2 was significantly smaller, indicating a noteworthy lower recombination rate, i.e., a superior separation efficiency of the photoexcited electron-hole pairs [49]. Thus, the addition of Fe3O4 nanoparticles on the surface of TiO2 can significantly reduce the electron-hole pairs recombination by accepting the electrons from TiO2 [48].

Composition, Structure and Morphology
X-ray photoelectron spectroscopy was carried out in order to investigate the electronic features and elemental composition of the prepared materials. Figure 5 shows the survey scan XPS spectrum of Fe3O4-TiO2 thin films, which contains O 1 s, Ti 2p and Fe 2p peaks.

Composition, Structure and Morphology
X-ray photoelectron spectroscopy was carried out in order to investigate the electronic features and elemental composition of the prepared materials. Figure 5 shows the survey scan XPS spectrum of Fe 3 O 4 -TiO 2 thin films, which contains O 1 s, Ti 2p and Fe 2p peaks.  From those computationally obtained peaks, the ones centred at energies of 529.8 and 530.4 eV can be assigned to the oxygen in the lattice Ti-O and to the oxygen in Fe-O, respectively. The peak located at 531.6 eV is related to the oxygen atoms in 'bridging' hydroxyl groups that are bound to titanium or iron atoms (Me-OH) [50]. Two types of hydroxyl groups are defined in the literature [51,52]: one of them, referred to as a 'bridging'   O 1s: a high-resolution O 1s XPS spectrum is deconvoluted into four peaks. From those computationally obtained peaks, the ones centred at energies of 529.8 and 530.4 eV can be assigned to the oxygen in the lattice Ti-O and to the oxygen in Fe-O, respectively. The peak located at 531.6 eV is related to the oxygen atoms in 'bridging' hydroxyl groups that are bound to titanium or iron atoms (Me-OH) [50]. Two types of hydroxyl groups are defined in the literature [51,52]: one of them, referred to as a 'bridging' hydroxyl, is attributed to O 2 − anions in the bulk structure of the material. The other one, which called 'terminal' and resides at an acidic site, is attributed to chemisorbed OH groups on the sample surface. As can be deduced from spectrum shown in Figure 6a, no terminal hydroxyl groups were detected in the fabricated material.
The O atom in the bridge OH should be energetically favourable to attract positive holes, because it has negative charge. Consequently, a certain fraction of trapped holes probably release • OH into bulk solution [53].
Fe 2p: in order to confirm the presence of Fe 3 O 4 , the elemental and valence state analysis were performed using XPS. Figure 6b shows the core-level XPS spectra of Fe 2p corresponding to Fe 3 O 4 . The peaks located at 711 and 724.6 eV are associated to Fe(III) 2p 3/2 and Fe(III) 2p 1/2 , respectively. Peaks at 709.3 eV and 722.5 eV can be assigned to Fe (II) 2p 3/2 and Fe(II) 2p 1/2 , respectively. These results are in agreement with Fe 2p XPS spectrum of Fe 3 O 4 reported in the literature [54]. Moreover, the area ratio of Fe 3+ /Fe 2+ is about 0.50, which is consistent with the stoichiometry of Fe 3 O 4 [55].
Ti 2p: the core-level XPS spectra of Ti 2p is fitted with three peaks. Peaks located at 458.6 eV and 464.4 eV can be assigned to Ti 4+ , confirming the main valence state of + 4 for Ti. Moreover, the energy difference between Ti2p 1/2 and Ti2p 3/2 is 5.8 eV, which is associated to the TiO 2 standard binding energy value [56]. Additionally, the calculated peak that corresponds to the Ti in the Ti-O-Fe bond is centred at an energy of 459.7 eV. It is worth noting that the existence of this peak, which is a necessary result of the deconvolution of the experimental XPS signal, is proof of the existence of a chemical bonding between TiO 2 and Fe 3 O 4 NPs [50,56].
Once the composition of the Fe 3 O 4 -TiO 2 thin film was analysed by XPS, and taking into account the expected influence of the thickness in the photocatalysis experiments, the photocatalysts with 5, 10 and 15 layers deposited were prepared, and SEM in the cross section was performed to get thickness measurements. Samples were called "1 µm:Fe 3 O 4 -TiO 2 ", "1. Fe3O4 reported in the literature [54]. Moreover, the area ratio of Fe 3+ /Fe 2+ is about 0.50, which is consistent with the stoichiometry of Fe3O4 [55]. Ti 2p: the core-level XPS spectra of Ti 2p is fitted with three peaks. Peaks located at 458.6 eV and 464.4 eV can be assigned to Ti 4+ , confirming the main valence state of + 4 for Ti. Moreover, the energy difference between Ti2p1/2 and Ti2p3/2 is 5.8 eV, which is associated to the TiO2 standard binding energy value [56]. Additionally, the calculated peak that corresponds to the Ti in the Ti-O-Fe bond is centred at an energy of 459.7 eV. It is worth noting that the existence of this peak, which is a necessary result of the deconvolution of the experimental XPS signal, is proof of the existence of a chemical bonding between TiO2 and Fe3O4 NPs [50,56].
Once the composition of the Fe3O4-TiO2 thin film was analysed by XPS, and taking into account the expected influence of the thickness in the photocatalysis experiments, the photocatalysts with 5, 10 and 15 layers deposited were prepared, and SEM in the cross section was performed to get thickness measurements. Samples were called "1 μm:Fe3O4-TiO2", "1.5 μm:Fe3O4-TiO2" and "2 μm:Fe3O4-TiO2", according to their nominal thicknesses. SEM micrographs of the prepared photocatalysts are shown in Figure 7. These images allowed to obtain the experimental thicknesses of the Fe3O4-TiO2 layer for each sample: (0.97 ± 0.03), (1.51 ± 0.04) and (1.99 ± 0.02) μm, respectively. These experimental values are in very good agreement with the nominal ones.
SEM images in Figure 7 indicates, beside the spatial scale, the expected 1, 1.5 and 2 μm thicknesses for each case. Note that, due to sample orientation and the clivationmethod followed in order to observe the materials in the SEM microscope, images in Figure 7 shows not only the layers profiles, but also, in the cases of Figure 7b and c, the surface of the preparations. Such surface can be distinguished from the profile by the intensity contrast in the image. Surface roughness of the titanium oxide layer for these three photocatalysts samples (1 μm:Fe3O4-TiO2, 1.5 μm:Fe3O4-TiO2 and 2 μm:Fe3O4-TiO2) was also determined, using atomic force microscopy ( Figure 8). Measurements were repeated in three different locations on their surfaces. These measurements and their averaged values are collected in Table 3. The values in this table reveal that, the appearance of surface, in all cases, is similar and quite homogeneous, having a roughness which seems to be independent of the layer thickness. It means that this method of fabrication would allow using even thicker layers without a modification in the surface roughness, which is an important factor that influences the physical behaviour of TiO2 -based thin films [57]. SEM images in Figure 7 indicates, beside the spatial scale, the expected 1, 1.5 and 2 µm thicknesses for each case. Note that, due to sample orientation and the clivationmethod followed in order to observe the materials in the SEM microscope, images in Figure 7 shows not only the layers profiles, but also, in the cases of Figure 7b and c, the surface of the preparations. Such surface can be distinguished from the profile by the intensity contrast in the image.
Surface roughness of the titanium oxide layer for these three photocatalysts samples (1 µm:Fe 3 O 4 -TiO 2 , 1.5 µm:Fe 3 O 4 -TiO 2 and 2 µm:Fe 3 O 4 -TiO 2 ) was also determined, using atomic force microscopy ( Figure 8). Measurements were repeated in three different locations on their surfaces. These measurements and their averaged values are collected in Table 3. The values in this table reveal that, the appearance of surface, in all cases, is similar and quite homogeneous, having a roughness which seems to be independent of the layer thickness. It means that this method of fabrication would allow using even thicker layers without a modification in the surface roughness, which is an important factor that influences the physical behaviour of TiO 2 -based thin films [57].

Proposed Mechanism
Based on Equations (4) and (5) and using the experimental band gap energies (see Table 2), the valence band energy, EVB, of TiO2 and Fe3O4 can be calculated to be 2.91 eV and 1.74 eV, respectively.
being the absolute electronegativities ( ) of TiO2 and Fe3O4 5.81 eV and 5.77 eV, respectively, and E C = 4.5 eV is the scaling factor relating the normal hydrogen electrode scale (NHE) to absolute vacuum scale [58]. As can be seen in Figure 9, when Fe3O4-TiO2 is solar irradiated (see Solar spectrum shown in the embedded Figure [59]), the different steps that take place during the photocatalytic process are: 1. Electron-hole pairs are produced in TiO2 and Fe3O4 nanoparticles.

Proposed Mechanism
Based on Equations (4) and (5) and using the experimental band gap energies (see Table 2), the valence band energy, E VB , of TiO 2 and Fe 3 O 4 can be calculated to be 2.91 eV and 1.74 eV, respectively.
being the absolute electronegativities (χ) of TiO 2 and Fe 3 O 4 5.81 eV and 5.77 eV, respectively, and E C = 4.5 eV is the scaling factor relating the normal hydrogen electrode scale (NHE) to absolute vacuum scale [58].
As can be seen in Figure 9, when Fe 3 O 4 -TiO 2 is solar irradiated (see Solar spectrum shown in the embedded Figure [59]), the different steps that take place during the photocatalytic process are: On the other hand, holes located in the valence band of TiO 2 are transferred to the surface of TiO 2 generating hydroxyl radicals because of the oxidation of water or surface hydroxyl species [60]. 7.
Radiation corresponding to wavelength range above 388 nm (E < 3.2 eV) is not energetic enough to excite TiO 2 and generate electron-hole pairs, whereas Fe 3 O 4 could be easily activated in this spectral range and produces charge carriers. Subsequently, the photogenerated electrons should be involved in the processes previously described in the step 3 of this proposed mechanism. 8.
The holes that are accumulated in the valence band of Fe 3 O 4 will react with OH − species existing on the surface of the catalyst, producing reactive hydroxyl radicals ( · OH).
Most of the considerations made in the proposed mechanism could be supported by experimental results. Thus, PL studies revealed a superior separation efficiency of the photoexcited electron-hole pairs in Fe 3 O 4 -TiO 2 NPs. Results obtained from the deconvolution of the high-resolution O 1s XPS show that positive holes generate hydroxyl radicals. In addition, as can be seen in Figure 10, absorption bands can be related to the energy levels proposed.
On the other side, the determined optical bandgap (Table 2) is originated by electronic coupling between the wide band gap of TiO 2 and the narrow band gap of Fe 3 O 4 [63].
Thus, via the proposed mechanism, it can be deduced that in Fe 3 O 4 -TiO 2 material, not only the amount of photogenerated electron-hole pairs are increased, but also the separation efficiency of the photoexcited charges is enhanced. It is essential due to photocatalytic activity of photocatalysts which mainly depend on the extent of electron-hole pairs production and separation efficiency of these charge carriers [64]. getic enough to excite TiO2 and generate electron-hole pairs, whereas Fe3O4 could be easily activated in this spectral range and produces charge carriers. Subsequently, the photogenerated electrons should be involved in the processes previously described in the step 3 of this proposed mechanism. 8. The holes that are accumulated in the valence band of Fe3O4 will react with OH − species existing on the surface of the catalyst, producing reactive hydroxyl radicals (˙OH). Most of the considerations made in the proposed mechanism could be supported by experimental results. Thus, PL studies revealed a superior separation efficiency of the photoexcited electron-hole pairs in Fe3O4-TiO2 NPs. Results obtained from the deconvolution of the high-resolution O 1s XPS show that positive holes generate hydroxyl radicals. In addition, as can be seen in Figure 10, absorption bands can be related to the energy levels proposed. On the other side, the determined optical bandgap (Table 2) is originated by electronic coupling between the wide band gap of TiO2 and the narrow band gap of Fe3O4 [63].
Thus, via the proposed mechanism, it can be deduced that in Fe3O4-TiO2 material, not only the amount of photogenerated electron-hole pairs are increased, but also the separation efficiency of the photoexcited charges is enhanced. It is essential due to photocatalytic activity of photocatalysts which mainly depend on the extent of electron-hole pairs production and separation efficiency of these charge carriers [64].

Photocatalytic Experiments
Photocatalytic degradation of SFX is selected as a probe reaction to evaluate the pho-

Photocatalytic Experiments
Photocatalytic degradation of SFX is selected as a probe reaction to evaluate the photocatalytic activity of the so-prepared Fe 3 O 4 -TiO 2 thin films. Solar photocatalysis experiments were performed treating 5 mg·L −1 aqueous solutions of SFX. The goal of photocatalytic experiments was not only to evaluate whether the incorporation of Fe 3 O 4 NPs enhance the photocatalytic activity or not, but also to study the influence of thickness of the thin films on the SFX removal rates. Thus, in order to compare the effectiveness of Fe 3 O 4 -TiO 2 films with that of pure TiO 2 , solar photocatalysis experiments were carried out using thin films with a similar thickness (~2 µm). Experimental results show that SFX follows first order kinetics according to Equation (6), in which t 30W can be calculated from experimental data by using Equation (2).
From this fitting process, pseudo-first order constants for TiO 2 and Fe 3 O 4 -TiO 2 thin film were calculated to be 6.95·10 −3 min −1 and 1.35·10 −2 min −1 , respectively. As expected from the previously exposed optical characterization, when the composite material is used, the removal performance is considerably improved when solar photocatalytic processes are applied. Moreover, adsorption experiments indicated that SFX does not adsorb on the TiO 2 and Fe 3 O 4 -TiO 2 photocatalyst surface. These results are in agreement with the literature [22]. In addition, direct photolysis contribution is negligible. Thus, no elimination of SFX was observed when it was exposed to solar radiation, without the presence of the photocatalyst. In conclusion, it can be accepted that free radical oxidation is the main way of oxidation and, as was described in Section 3.3, hydroxyl radicals can be generated by different pathways.
In order to prove that the elimination is due to the presence of hydroxyl radicals, an experiment was carried out in the presence of T-butanol. T-Butanol is a very commonly used hydroxyl radical scavenger in aqueous matrices [24]. Results show that the presence of t-butanol completely inhibits the SFX removal.
On the order hand, in addition to the Fe 3 O 4 -TiO 2 film with a thickness of 2 µm, thin Fe 3 O 4 -TiO 2 films with thickness of around 1 and 1.5 µm were also prepared, as was mentioned above. Photocatalysis efficiency of these three samples is evaluated in Figure 11, in which the normalized concentration of SFX versus UV dose is represented. The presented results were calculated as an average value of the three repeated experiments per photocatalyst. Under solar irradiation, the concentration of the selected pharmaceutical evolves according to Equation (6) for all the cases and the calculated pseudo-first order constants are shown in Table 4. In view of the Figure 11 and Table 4, it is evident that a higher removal degree is achieved as the thickness of the Fe 3 O 4 -TiO 2 film increases in the studied range. literature [22]. In addition, direct photolysis contribution is negligible. Thus, no elimination of SFX was observed when it was exposed to solar radiation, without the presence of the photocatalyst. In conclusion, it can be accepted that free radical oxidation is the main way of oxidation and, as was described in Section 3.3, hydroxyl radicals can be generated by different pathways.
In order to prove that the elimination is due to the presence of hydroxyl radicals, an experiment was carried out in the presence of T-butanol. T-Butanol is a very commonly used hydroxyl radical scavenger in aqueous matrices [24]. Results show that the presence of t-butanol completely inhibits the SFX removal.
On the order hand, in addition to the Fe3O4-TiO2 film with a thickness of 2 μm, thin Fe3O4-TiO2 films with thickness of around 1 and 1.5 μm were also prepared, as was mentioned above. Photocatalysis efficiency of these three samples is evaluated in Figure 11, in which the normalized concentration of SFX versus UV dose is represented. The presented results were calculated as an average value of the three repeated experiments per photocatalyst. Under solar irradiation, the concentration of the selected pharmaceutical evolves according to Equation (6) for all the cases and the calculated pseudo-first order constants are shown in Table 4. In view of the Figure 11 and Table 4, it is evident that a higher removal degree is achieved as the thickness of the Fe3O4-TiO2 film increases in the studied range.
As an example, for an accumulated radiation dose of 150 Wh/m 2 (t30w = 300 min), elimination yields of 84.2, 95.4 and 98.3% were reached, when 1 μm:Fe3O4-TiO2; 1.5 μm:Fe3O4-TiO2 and 2 μm:Fe3O4-TiO2 were used, respectively. Furthermore, differences observed when 1.5 μm:Fe3O4-TiO2 and 2 μm:Fe3O4-TiO2 samples were used are less significant than those observed when comparing the results obtained with the 1 and 1.5 μm thick samples, which may be due to the existence of a critical film thickness. Figure 11. Degradation of SFX using thin films with different thickness. Table 4. Values of the apparent pseudo-first order rate constant using thin films with different thicknesses in the oxidation of SFX in water.
Taking into account that there is no elimination of SFX by adsorption or direct photolysis, the presence of light and photocatalyst are necessary.
Experimental results obtained ( Figure 11) demonstrate that activity is governed by the catalyst film thickness. This result allows us to conclude that all the deposited material participates in the solar photocatalysis process.
Finally, in order to check the stability of the fabricated thin films, solar photocatalytic experiments were carried out reusing one of these films. Specifically, the 1.5 µm:Fe 3 O 4 -TiO 2 photocatalyst was used in three consecutive reactions and no decrease in SFX elimination efficiency was observed.

Conclusions
The main conclusions of this work are: Magnetite NPs and TiO 2 , as examined by electron microscopy, were of crystalline nature, and the utilized method to dope TiO 2 resulted in imbibed magnetite NPs, randomly distributed in the titanium oxide. NPs size was (8.7 ± 1.1) nm. XPS confirmed the presence of Fe 3 O 4 and the chemical bonding between TiO 2 and Fe 3 O 4 NPs.
The process to obtain the modified titanium oxide layer for the "sandwich" structure was well controlled, and tested for thicknesses between 1 and 2 µm.
The calculated indirect bandgap is reduced when Fe 3 O 4 NPs are included in the material. The small indirect energy gap obtained for magnetite NPs is consistent with the fact that Fe 3 O 4 -TiO 2 materials present absorption in the visible spectrum region.
The proposed mechanism of Fe 3 O 4 -TiO 2 thin films in solar photocatalytic processes could be supported by experimental results. It was determined that, in Fe 3 O 4 -TiO 2 material, not only are the number of photogenerated electron-hole pairs increased, but also the separation efficiency of the photoexcited charges is enhanced. Solar photocatalytic experiments evidenced the effectiveness improvement when modified thin films are used. Thus, pseudo-first order constants for TiO 2 and Fe 3 O 4 -TiO 2 2 µm thin films were calculated to be 6.95·10 −3 min −1 and 1.35·10 −2 min −1 , respectively.
Regarding the thickness of the fabricated thin films, AFM measurements revealed that surface roughness seems to be independent of the number of deposited layers. However, a higher removal degree is achieved as the thickness of the Fe 3 O 4 -TiO 2 film increases in the studied range.
Finally, experiments reusing 1. Data Availability Statement: The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.