Environment protection is a continuing challenge for all engineering disciplines [1-3]. Wastewater discharged by industry frequently contains high concentrations of organic contaminants in large volumes. Such wastewater is usually poorly biodegradable and difficult to treat using chemical oxidants [4, 5]. Among the many reported water treatment technologies, photodegradation and adsorption are two promising technologies with high efficiencies and low energy requirements for removing pollutants like organic dyes [4-9]. TiO₂ has been extensively studied for this purpose, along with methods for improving its photocatalytic performance [10, 11]. These have included doping TiO₂ with other metal ions [12-15], and controlling the TiO₂ nanostructure by forming structures such as quantum dots (QDs) [16-19]. Loading QDs into porous materials with large surface areas is an effective method for avoiding QD agglomeration [20, 21]. Fe₂O₃ is also an efficient adsorbent for organic dyes [22-25] and has been widely investigated for water treatment because of its good chemical stability, low toxicity, and ease of preparation. However, the large dimensions of Fe₂O₃ limits its adsorption capacity and adsorption rate. Decreasing the Fe₂O₃ particle size from micrometers to nanometers can reportedly increase the adsorptive area by 100 to 1000 times [26].

The synergy of adsorption and photodegradation can enhance the removal of pollutants from water, in comparison to a single technology [27]. Herein, to realize this synergetic effect, we loaded both Fe-doped TiO₂ QDs and Fe₂O₃ QDs into MCM-41, using a direct hydrolysis method. The Fe-doped TiO₂ QDs exhibit high photoactivity due to their quantum-size effect and dopant-mediated band levels. The Fe₂O₃ QDs exhibit significant adsorption toward organic dye. Thus, the fabricated materials possess enhanced efficiency for degrading organic dye.

FeSO₄·7H₂O, rose bengal (RB), tetrabutyl titanate (TBT), commercial Fe₂O₃ (average diameter of 100 nm, abbreviated as C-Fe₂O₃) and ethanol were of reagent grade and purchased from Shanghai Aladdin Reagent Co. Ltd, P. R. China. MCM-41 was purchased from Tianjin Chemist Ltd, P. R. China. Milli-Q ultrapure water (>18 mΩ·cm) was used in all experiments. All reagents were used as received.

In a typical synthesis, 0.681 g of TBT was first adsorbed on the surface and pore channels of MCM-41 (0.800 g) in ethanol (50 mL), under stirring for 40 min. Then, 6.0 mL of FeSO₄·7H₂O aqueous solution was added dropwise to the mixture, which was stirred for another 24 h at room temperature. The resulting gel was collected by filtration, dried at 60 ℃ for 12 h, and then calcined at 550 ℃ for 2 h at a heating rate of 5 ℃/min. Ti/Fe molar ratios (x) of 30, 10, 3.0, and 1.0 were used in this study, and the resulting samples are referred to as FT/M-x. As shown in Scheme 1, pure TiO₂ QDs loaded on MCM-41 synthesized without FeSO₄·7H₂O is referred to as T/M. A series of samples loaded with Fe-TiO₂ QDs and Fe₂O₃ QDs were synthesized by adjusting the concentration of FeSO₄ solution. Pure Fe₂O₃/MCM-41 synthesized without TBT is referred to as F/M. As a reference sample, pure Fe₂O₃ was synthesized by the direct calcination of FeSO₄·7H₂O at 550 ℃ for 2 h at a heating rate of 5 ℃/min, and is referred to as S-Fe₂O₃.

X-ray diffraction (XRD) patterns were recorded using a D8-Focus X-ray diffractometer (BRUKER AXS GMBH) equipped with Cu K_α radiation, at 40 kV and 40 mA, at a scan rate of 5°/min. Fourier-transform infrared (FT-IR) spectra were obtained using a Bruker Vertex-70 FT-IR spectrometer. Transmission electron microscopy (TEM) and high-resolution TEM (HR-TEM) observations were carried out using a Tecnai G2 F-20 microscope with a field-emission gun, operated at 200 kV. Zeta potentials (ζ-potentials) were measured in suspensions (0.5 mg/min, 25 ℃) using a Zeta Nanosizer (Nano ZS, Malvern Ltd.). Ultraviolet-visible diffuse reflectance spectra (UV-Vis DRS) were recorded using a Shimadzu UV-2600 spectrometer equipped with a 60-mm-diameter integrating sphere, using BaSO₄ as the reference sample. X-ray photoelectron spectroscopy (XPS) was conducted with a PHI-1600 X-ray photoelectron spectrometer equipped with Al K_α radiation. Binding energies were calibrated to the C 1s peak (284.6 eV) of adventitious carbon.

The adsorption of the model organic dye RB was conducted in an opening jacketed glass beaker (150 mL) in the dark. Typically, 20 mg of catalyst and 0.2 mL of RB (20 mmol/L) were dispersed in 100 mL of pure water. The suspension was stirred until the adsorption of RB reached equilibrium. During this evaluation, 1.5 mL aliquots of suspension were withdrawn after given intervals, centrifuged, and the concentration of RB in the supernatant determined from the absorbance at 548 nm, using a Hitachi U-3010 UV-Vis spectrometer. 0.2 mL of RB (20 mmol/L) solution was added again, which was repeated several times until the adsorption reached saturation. In addition, appropriate volumes of pure water were added to compensate for the loss during sampling.

The photodegradation of RB was conducted in the same jacketed glass beaker (150 mL), which was irradiated from the vertical direction by a 300 W high-pressure xenon lamp (PLS-SXE300UV, Beijing Trusttech. Co., Ltd.). The irradiation area was ca. 20 cm². The reaction conditions were: temperature 25 ± 0.2 ℃, initial concentration of RB (C₀(RB)) 40 μmol/L, and synthetic photocatalyst concentration 0.2 g/L; no acidic or alkaline reagents were added. The reaction was magnetically stirred under atmospheric conditions, after stirring for 30 min in the dark to achieve adsorption equilibrium. Samples were then withdrawn, separated by centrifugation, and analyzed. Once UV-vis spectrometry indicated that the RB had been completely photodegraded, 0.200 mL of RB (20 mmol/L) was added again, and the overall process was repeated several times. To evaluate the efficiency of the decontamination process, we measured the chemical oxygen demand (dichromate method, COD_Cr) using a DR3900 apparatus (HACH), the final total organic carbon (TOC) contents of aqueous samples treated by synthesized catalysts in the first and tenth cycles using a Multi c/n2100s apparatus (Analytik Jena AG), and the concentrations of remanent Na⁺ and Cl^- using a PXS-270 apparatus (INESA Scientific Instrument Co., Ltd.).

As shown in Fig. 1(a), the XRD pattern of pure MCM-41 shows a wide diffraction peak at ca. 22°, while a new peak at 25.3° appears after loading with TiO₂ (T/M), which is characteristic of anatase TiO₂ (JCPDS No. 21-1272). The XRD pattern of T/M shows broadened diffraction peaks at about 25.3°, which indicates that the synthesized TiO₂ nanoparticles have a very small crystal size [28]. The XRD patterns of FT/M-x (with x ≥ 3.0) show characteristic peaks similar to the pattern of T/M, but a gradual main-peak shift from 25.3° to 23.7° is observed from x = 30 to x = 3.0. This is caused by distortion from crystal expansion resulting from the substitution of Ti⁴⁺ (0.075 nm) with Fe³⁺ (0.079 nm) [29, 30]. In the XRD pattern of FT/M-1.0, the diffraction peaks of α-Fe₂O₃ appear at 27.7°, 30.1° and 32.9° (JCPDS No. 52-1449). To understand the formation mechanism, we synthesized pure TiO₂ and Fe-doped TiO₂ without MCM-41 by the same method. The crystal sizes of undoped and Fe-doped (Ti/Fe = 10) TiO₂ are 65.2 and 79.5 nm, respectively. This demonstrates that loading TiO₂ into MCM-41 is an important factor accounting for the formation of quantum-sized TiO₂ [12]. As shown in Scheme 1, the surface hydroxyl groups of MCM-41 provide adsorption sites for TBT. The interaction between the host MCM-41 and TiO₂ is evidenced by peaks of Ti-O-Si vibrations at 960 cm^-1 in the FT-IR spectra of FT/M-10 (Fig. 1(b)).

The morphologies of samples were determined by TEM. TEM images of MCM-41 show porous channels with no obvious surface particles (Fig. 2(a)). For T/M (Fig. 2(b)), TiO₂ nanoparticles with an average diameter of 4.78 nm give rise of lattice fringes at intervals of 0.351 nm ((101) plane of anatase TiO₂), indicating that they are pure TiO₂ QDs. For FT/M (Fig. 2(b)-(f)), the anatase (101) lattice spacing increases from 0.351 nm to 0.380 nm with increasing Fe content, which is consistent with the XRD results. Meanwhile, the average diameter of the TiO₂ QDs increases from 4.78 to 6.65 nm. Fe₂O₃ QDs are formed when the Ti/Fe molar ratio is lower than 3.0. Fig. 2(e) and (f) show obvious Fe₂O₃ nanoparticles with lattice spacings of 0.271 and 0.321 nm, which are consistent with the (122) and (112) planes of α-Fe₂O₃, respectively. The average diameters of FT/M-3.0 and FT/M-1.0 are 3.94 and 4.51 nm, respectively, which are typical α-Fe₂O₃ QDs. The formation of Fe₂O₃ QDs is attributed to the solubility difference of FeSO₄·7H₂O in water and ethanol. FeSO₄·7H₂O is water-soluble but alcohol-insoluble, so precipitates and deposits on MCM-41 when adding high concentration FeSO₄·7H₂O aqueous solution to ethanol. Calcining at 550 ℃ causes the FeSO₄·7H₂O nanocrystals to decompose to Fe₂O₃ QDs. The energy dispersive X-ray (EDX) spectroscopy image in Fig. 2(i) shows uniform distributions of Fe₂O₃ and TiO₂ on MCM-41. In addition, F/M, i.e. Fe₂O₃ loaded on MCM-41 (Fig. 2(g)), has an average diameter of 40 nm, while Fe₂O₃ synthesized in the absence of MCM-41 has an average diameter of ca. 200 nm (Fig. 2(h)).

The Ti and Fe contents of the samples were measured by EDX. Table 1 shows that T/M contains 4.02 wt.% of TiO₂ loaded on MCM-41, and that the loading amount increases with the addition of FeSO₄. FT/M-3.0 and FT/M-1.0 contain about the same amount of TiO₂ (8.54-8.62) wt.%. The presence of the acidic FeSO₄ solution improves the hydrolysis of TBT to TiO₂.

XPS was used to characterize the chemical states of the sample surfaces, and XPS spectra are shown in Fig. 3. The Si 2p peaks in the spectra of T/M and FT/M show high-binding-energy-shifts of 0.12 eV, compared with the spectrum of pure MCM-41 (Fig. 3(a)). The Ti 2p peaks show low-binding-energy-shifts of 0.26 eV compared with the spectrum of pure anatase TiO₂ (Fig. 3(b)), which confirms the formation of Ti-O-Si bonds between TiO₂ and MCM-41 [12, 30, 31]. No Si 2p peak shift is observed in the spectrum of F/M, suggesting no chemical interaction between Fe₂O₃ and MCM-41.

Fig. 3(c) shows doublet Fe 2p_3/2 and Fe 2p_1/2 peaks of Fe(Ⅲ) at binding energies of 711 and 725 eV, respectively [32]. For FT/M-30 and FT/M-10, the satellite peak of α-Fe₂O₃ should be located at ca. 8 eV higher than the main peak of Fe 2p_3/2, but this is not observed. This suggests that Fe(Ⅲ) is doped into the inner lattice of TiO₂. The spectra of FT/M-3.0 and FT/M-1.0 show weak satellite peaks at 719 eV, because of the presence of α-Fe₂O₃. The valence band (VB) XPS spectra in Fig. 3(d) show obvious shifts, because of the newly formed Fe(Ⅲ) dopant-mediated band levels in TiO₂.

Fig. 4(a) shows UV-Vis DRS spectra of the samples. MCM-41 shows no absorption of the full-spectrum light. T/M strongly absorbs wavelengths below 380 nm, but not above 380 nm. FT/M-30 and FT/M-10 show red-shifted absorption, which is attributed to the newly formed Fe(Ⅲ) dopant-mediated band levels. For FT/M-3.0, FT/M-1.0, and M-Fe₂O₃, absorption peaks at about 540 nm and 690 nm are observed due to the presence of the Fe₂O₃ QDs. The band gaps of the samples were calculated from the corresponding plots of the transformed Kubelka-Munk functions (Fig. 4(b)). Pure TiO₂ QDs (3.40 eV) exhibit a typical quantum-size effect, and their bandgap is 0.20 eV wider than that of bulk TiO₂, because the VB shifts towards higher energy [33, 34]. For FT/M, the bandgap decreases from 3.40 to 2.65 eV with increasing Fe content. This is because Fe doping results in a new peak due to transition between Fe³⁺ and the VB. The energy of this transition changes with the Fe(Ⅲ) dopant content [33, 34], as confirmed by the VB density of states in Fig. 4(b).

The samples were applied to remove RB, by adsorption in the dark and then subsequent photodegradation under irradiation [35-37]. Fig. 5(a) shows the adsorption capacities of the samples in the dark. Pure MCM-41 and T/M show small adsorptions of RB, of less than 10 mg/g. FT/M-30 and FT/M-10 which contain Fe-doped TiO₂ QDs show slightly higher RB adsorptions. FT/M-3.0 and FT/M-1.0 which contain Fe₂O₃ QDs, show ca. 6.4 and 9.3 times higher RB adsorption than that of T/M, respectively. The RB adsorption capacity of F/M is the highest among the tested samples at ca. 75 mg/g. The adsorption rate constants (k₁) derived from these adsorption capacities were fitted by the Lagergren rate equation, as shown in Fig. 5c [38]. The adsorption capacities of the Fe-based materials should be related to their surface charge. The ζ-potentials of the samples follow the order F/M > FT/M-1.0 > FT/M-3.0 > FT/M-10 > 0 > FT/M-30 > T/M (Table 1), which correlates with their Fe contents and adsorption capacities. In addition, a sample prepared by thermally calcining FeSO₄ (S-Fe₂O₃), and another sample of commercial nano-Fe₂O₃ (C-Fe₂O₃), show saturation adsorption capacities of 46 mg/g and 5.68 mg/g, respectively. These are much lower than that of F/M, because F/M contains well-dispersed Fe₂O₃ QDs with a high surface area accessible for adsorption.

Fe₂O₃ is inactive in photodegradation due to its short hole diffusion length (ca. 2-4 nm) and significant recombination of photogenerated electron-hole pairs [39-43]. Fig. 5(a) shows that MCM-41, S-Fe₂O₃, and C-Fe₂O₃ exhibit no photoactivity, while weak photodegradation is observed for F/M (ca. 5.32 mg/g), which is attributed to the well dispersed Fe₂O₃ on MCM-41. T/M shows obvious photodegradation when using 30.2 mg/g of photocatalyst. The doping of Fe species in TiO₂ significantly increases the photoactivity. The highest photodegradation is observed for FT/M-3.0 (124 mg/g). FT/M-1.0 which contains a large amount of Fe₂O₃ shows lower photoactivity. In Fig. 5(d), the real reaction rate constant (k₂) of the photodegradation is fitted by the Langmuir-Hinshelwood equation [44] The photoactivity follows the order: FT/M-3.0 > FT/M-10 > FT/M-30 > FT/M-1.0 > T/M > F/M. The highest photoactivity of FT/M-3.0 is attributed to the effective doping of Fe(Ⅲ) into the TiO₂ lattice. This leads to a new impurity band above the VB, which benefits charge separation and optical absorption.

Fig. 5(c) and (d) show that the adsorption rate (k₁) is comparable to the photocatalytic rate (k₂) for T/M, FT/M-30, FT/M-10, and FT/M-3.0. This enables RB to be adsorbed on the Fe₂O₃ QDs and then photodegraded by the Fe-TiO₂ QDs. For FT/M-1.0, k₁ is 12.2 times higher than k₂, and the catalyst is covered by adsorbed RB after adsorption equilibrium, which will block light absorption and degradation. Therefore, during the ten cycle test (Fig. 5(b)) there are no obvious decreases in the adsorption capabilities of T/M, FT/M-30, FT/M-10, and FT/M-3.0, while the adsorption capability of FT/M-1.0 decreases significantly.

The initial COD_Cr, TOC, and C_Na of the RB solution are 32.93, 12.59, and 1.89 ppm, respectively, and there is nearly no Cl^-. The TOC and COD_Cr of samples treated by F/M and T/M decrease by about 23.9% and 21.5% in the first cycle, respectively, and decrease by about 79.6% when treated by FT/M-3.0 (Table 2). The increased Na⁺ concentration results from ionization of the RB dye. For the solution treated by F/M, the Cl^- concentration shows no significant increase (from 0.31 ppm to 0.43 ppm), while those treated by T/M and FT/M-3.0 show increases after one cycle of ca. 0.96 ppm and 5.11 ppm, respectively. This confirms that the Cl atoms of RB molecules are degraded to Cl^-, and dissolve into the water during the degradation process. F/M and T/M remove only about 2% and 12% of COD_Cr and TOC during the tenth cycle, respectively, while 96.6% of TOC and 96.9% of COD_Cr are removed by FT/M-3.0, respectively. The concentrations of Cl^- and Na⁺ increase from near-zero to 58.01 ppm and 18.51 ppm, respectively, which indicates the high efficiency and stability of FT/M-3.0 for degrading RB. The high efficiency is attributed to the synergic effects of adsorption and photodegradation. To confirm the active species during photodegradation, 0.1 mol/L methanol was introduced as a hole sacrificial reagent, and 0.1 mol/L isopropanol was introduced as a radical sacrificial reagent. The photodegradation rate does not significantly change in the presence of 0.1 mol/L methanol, but significantly decreases in the presence of 0.1 mol/L isopropanol. Thus, the photodegradation is considered to occur via a free radical pathway [45-47]. Scheme 2 shows that RB molecules are first concentrated on the surface by the strong adsorption capacity of the Fe₂O₃ QDs, then migrate to the Fe-TiO₂ QDs, and are finally degraded to CO₂, H₂O, Cl^-, etc. To achieve the synergetic advantage of adsorption and photoreaction, the structures of the Fe₂O₃ QDs and Fe-TiO₂ QDs should be suitably controlled. With its suitable structure, FT/M-3.0 provides a high RB elimination of ca. 160 mg/g in the first cycle, and retains this high efficiency during the ten cycle test. Even after ten cycles, FT/M-3.0 removes more than 96% of COD_Cr and TOC.

We fabricated Fe-doped TiO₂ and Fe₂O₃ QD co-loaded MCM-41 via simple hydrolysis, to combine the effects of adsorption and photodegradation. With increasing Ti/Fe molar ratio during synthesis, Fe-doped TiO₂ QDs form at low Fe precursor concentrations, while additional Fe₂O₃ QDs form at higher Fe concentrations. The material with a Ti/Fe molar ratio of 3.0 shows the best performance in the removal of RB (160 mg/g), which is two and four times higher than those of Fe₂O₃ QDs and TiO₂ QDs separately loaded on MCM-41. The best performing catalyst retains its high efficiency over ten consecutive cycles.