3.1. Material performance
The morphological structure of FTC was examined by SEM and TEM. Figure 1a shows spherical Fe3O4 nanoparticles, TiO2 nanosheets with characteristics corresponding to the 001 facet [33], and rod-shaped g-C3N4 aggregated by many irregularly shaped sheets [34]. Fe3O4, TiO2, and g-C3N4 coexisted in FTC, where Fe3O4 and g-C3N4 were accumulated on both sides of TiO2, successively, in a noncontact manner. The area corresponding to Fe3O4/TiO2 was observed in the front of the TEM image of FTC (Fig. 1b), which was dull and thick because of the sticking of Fe3O4 at the center of the sample. A rough area with TiO2/g-C3N4 caused by the wrinkled sheets of g-C3N4 was observed behind Fe3O4/TiO2. The Fe3O4/TiO2 and TiO2/g-C3N4 regions were extracted to obtain high-resolution (HR)-TEM images, where the lattice spacings of 0.24, 0.23, and 0.34 nm corresponded to TiO2(001), Fe3O4(311), and g-C3N4(002), respectively. This confirmed that Fe3O4/TiO2 and TiO2/g-C3N4 connected by TiO2 formed an entire nanometer FTC with a two-layer heterojunction. The models of FTC, TiO2/g-C3N4, and Fe3O4/TiO2 (Figs. 1b–c) intuitively described the structures of these materials.
Figure 1d shows the FT-IR spectra of the prepared materials. Compared with the known standard FT-IR spectrum, the absorption peaks of the prepared Fe3O4, TiO2, and g-C3N4 could be attributed to the stretching vibration of the corresponding bond in the standard [35, 36]. One of the outstanding peak bands observed at 617 cm− 1 in the pure Fe3O4 was the symmetric stretching of oxygen atoms along Fe–O–Fe bonds. This peak could not be found in FTC because of the overlap with the characteristic peak of the stretching vibrations of the Ti–O octahedron. The FT-IR results showed that the FTC heterojunction structure contained the undestroyed structures of Fe3O4, TiO2, and g-C3N4.
Compared with the standard XRD patterns of the pure anatase phase (JCPDS No. 21-1272), Fe3O4 (JCPDS No. 01-1111), and g-C3N4 (JCPDS No. 50-1550), Fig. 1e shows that the corresponding characteristic peaks were observed in the FT, TC, and FTC nanocomposites. The positions of these peaks did not shift, indicating that the composition of these composites met the requirements. In addition, compared with the patterns of g-C3N4 and FT, the 37.8° and 27.4° peaks of TC were not evident probably because a large heterojunction formed between g-C3N4 and {001} TiO2 restricted the stacking of g-C3N4 perpendicular to the (002) direction. The formation of the heterojunction also caused a series of peaks of Fe3O4 to be concealed in FT. Noteworthily, all the peaks in FT and TC were present in FTC, suggesting that FTC stably retained the ternary two-layer heterojunction with Fe3O4/TiO2 and TiO2/g-C3N4.
As shown in Fig. 1f, the structure of FTC was established by Fe3O4 [37], anatase TiO2 [38], and g-C3N4 because all the peaks of FTC correlated with those in the Raman spectra of Fe3O4, TiO2, and g-C3N4. The corresponding spectrum showed the following changes: the signals attributed to the Fe3O4 phase in FT were diminished, and the signals attributed to the g-C3N4 phase in TC became less intense and wide, suggesting a decrease in its crystallinity. These conversions attributed to the phonon confinement effect caused by the distortions led to the TiO2 interaction with the Fe3O4 or g-C3N4 lattice. The FTC results showed that the peaks further shifted based on retaining the characteristic peaks of FT and TC, implying that electronic interaction occurred among the TiO2 nanoparticles and Fe3O4 or g-C3N4 sheets. Thus, Fe3O4, g-C3N4, and TiO2 were successfully coupled into a two-layer heterojunction.
The band-gap energies and absorption thresholds were calculated using the Tauc [39] plot approach after UV–VIS diffuse reflectance spectroscopy (Figs. 2a–b). The band-gap energies of the as-prepared Fe3O4, g-C3N4, and TiO2 were 1.87, 2.80, and 3.20 eV, respectively, which correlated with reported results [40, 41]. According to the change trend of band-gap energies, the absorption intensity of FTC (675 nm) extended to UV–VIS light regions. The band-gap energies of FTC reduced to 1.96 eV after FT coupled with TC, which was mainly due to the addition of Fe3O4. Thus, the high photocatalytic activity of FTC was demonstrable by an almost full-wave band of visible sunlight.
FTC exhibited the highest intensity and fast photocurrent response, and the photocurrent slowly decreased (Fig. 2c). Fig. S6 shows the EIS results with a decreased electron-transfer resistance in FTC with the smallest diameter of the semicircular Nyquist plots. Thus, FTC exhibited more efficient charge separation. The transient photocurrent responses and EIS results of FTC and g-C3N4 were similar, showing that the heterojunction between g-C3N4 and TiO2 had a more evident influence on photocatalytic performance. The flat-band potentials of g-C3N4 and TiO2 shown in Mott–Schottky plots (Fig. 2d) correlated with reported results [40]. These plots, including the minimum slope belonging to FTC, exhibited positive slopes, indicating that they were all n-type semiconductors, and FTC exhibited the fastest charge-transfer rate. As shown in Fig. 2d, the flat-band potential (EFB) of FTC was − 0.36 eV (vs. Ag/AgCl [42]), signifying the CB potential (ECB) for FTC confirmed to be − 0.14 eV. Therefore, the corresponding VB potential (EVB) of FTC was calculated as 1.82 eV using the band-gap energy (1.96 eV) (Fig. 2a). The ECB, EVB values of g-C3N4, TiO2, and TC were calculated to be − 0.28, 2.52; −0.20, 3.00; and − 0.26 2.74 eV using the same method, respectively, indicating that the positions of g-C3N4 were lower than those of TiO2. Therefore, electrons formed a space-charge layer when transferring in the CB from g-C3N4 to TiO2. The amount of Fe3O4 (ECB = − 0.04 eV, EVB = 3.16 eV as reported) was relatively small, and the photoelectrochemical performance of Fe3O4/TiO2 was poor. The follow-up adsorption model showed that the Fe3O4/TiO2 heterojunction did not adsorb reactants. Therefore, the influence of Fe3O4 was not analyzed here.
3.1.2. Simultaneous removal experiment of U(Ⅵ) and Sb(Ⅲ)
Figures 3a–b show the changes in U(Ⅵ) and Sb(Ⅲ) with time or different materials. Figure 3b shows that there was no concentration change in U(Ⅵ) and Sb(Ⅲ) without FTC. Under simulated sunlight irradiation, the concentrations of U(Ⅵ) and Sb(Ⅲ) rapidly decreased in the solution over FTC after reaching an adsorption–desorption equilibrium. After 3 h of photocatalysis, approximately 93.0% of U(Ⅵ) and 83.0% of Sb(Ⅲ) were removed by FTC, which exhibited the best photocatalytic performance among TiO2, g-C3N4, Fe3O4, FT, and TC (Fig. 3b). The normalized U LIII-edge EXAFS and Sb K-edge EXAFS spectra (Fig. 3c) could verify the photocatalytic effect of FTC, i.e., the valence states of U(Ⅵ) and Sb(Ⅲ) were maintained during the adsorption process. After the 3-h photocatalysis, the valence state of U(Ⅵ) decreased close to + 4, while that of Sb(Ⅲ) increased to approximately + 5. Figure 3b also shows that the removal rate of FTC for the solution with only U(Ⅵ) was merely 70.8%, while it was just 63.9% for the pure Sb(Ⅲ) solution; these rates were considerably lower than that for the mixed solution. The removal rate for the mixed solution was higher because the major active species for photocatalysis of U(Ⅵ) and Sb(Ⅲ) were reductive e− and photooxidative groups (such as •OH), which competed for photogenerated electrons and holes, respectively [27], thereby fully harnessing the photocatalytic ability of FTC by preventing their recombination.
The magnetic property of these materials was characterized (Fig. 3d and Table S1). The saturation magnetic induction of FTC with a value of 16.02 emu/g was considerably higher than that of nonmagnetic TC. The coercive force of 116.75 Oe confirmed that FTC was a permanent magnet [43, 44] with stable magnetism and good dispersibility in the aqueous phase. Figure 3d shows that FTC could be easily separated using an external magnet and redispersed in the solution by slight shaking, which facilitated the recovery and reusability of the photocatalysts.
3.1.3. Recycling and Co-existing ions experiment
As shown in Fig. 4a, FTC could still maintain a relatively high removal rate for U(Ⅵ) and Sb(Ⅲ) after being recycled seven times. Although Fe3O4 was dissolved [22] after four cycles, the mass of photodissolution after seven cycles was still less than 4% of the mass of Fe3O4 in FTC. The barrier to the photodissolution of Fe3O4 was due to the effect of g-C3N4 with a low-lying CB, which hindered the transfer of photogenerated electrons in TiO2 to Fe3O4. With an emergence of Fe3+ and Fe2+, which could react with U(Ⅵ) [30, 45] and Sb(Ⅲ) [46, 47], a new redox path was observed, leading to an increase in the removal rate after four cycles. The TEM and VSM results (Fig. 4d) showed that the morphology was maintained with a two-layer heterojunction, and there was practically no change in the magnetism of FTC after a seven-run cycle test and cleaning. Compared with those of Fe3O4, TiO2, g-C3N4, and TC, the PL peak intensity of FTC was considerably less and just slightly changed after seven cycles (Fig. 4c), showing that FTC exhibited a low and stable recombination rate of photogenerated electrons and holes. These results confirmed that the properties of FTC were extremely stable. Furthermore, FTC could be used, as confirmed by the results shown in Fig. 4b, which indicated that the coexisting ions had only a minimal effect on the photocatalytic removal of U(Ⅵ) and Sb(Ⅲ) by FTC.
3.2. Reaction route
3.2.1. XPS analysis
For the XPS measurements (Fig. 5a), the peaks around 288, 399, 459, 531, 712, and 724 eV were attributed to the C 1s, N 1s, Ti 2p, O 1s, Fe 2p3/2, and Fe 2p1/2 photoelectrons. Notably, new peaks at 382 and 393 eV after adsorption and 380 and 391 eV after a 3-h photocatalysis were attributed to the U 4f7/2 and U 4f5/2 photoelectrons, respectively, indicating the successful adsorption and reaction of U(Ⅵ) [48] on FTC. The spectra of O 1s and Sb 3d partly overlapped; therefore, the analysis of Sb(Ⅲ) refers to the HR XPS spectra.
As shown in Fig. 5b, the HR O 1s spectrum exhibited two characteristic peaks at 530.0 and 531.5 eV, which could be assigned to the oxygen atoms in the lattice of TiO2 or Fe3O4 and the •OH group of FTC. It was further confirmed that the oxygen-containing functional groups contributed to the adsorption and photocatalysis of U(Ⅵ) and Sb(Ⅲ).
The HR XPS spectra of the N 1s (Fig. 5c) of FTC could be deconvoluted into three peaks at 398.7, 400.0, and 401.1 eV, which could be ascribed to C–N = C, N–(C3), and C–N–H, respectively. The peak of N–(C3) shifted to a low binding energy after adsorption till the end. The C–N = C bonds exhibited a similar change after the 3-h photocatalysis, which verified the stretching of bonds and a shift in the charge center. Therefore, preliminarily, it was assumed that N–(C3) was the location where the adsorption occurred; C–N = C and N–(C3) were regarded as the adsorption sites of the photocatalytic product.
It was concluded that the material exhibited excellent stability because the areas of all the peaks (including HR Ti 2p, C 1s, and Fe 2p XPS shown in Fig. S7) rarely changed after the reaction.
As shown in the HR U 4f XPS spectra (Fig. 5d), the peaks of U 4f5/2 and U 4f7/2 were deconvoluted into two components at 390.2 and 391.1 eV and 379.5 and 380.2 eV, respectively, where the peaks of 380.2 and 391.1 eV, instead of corresponding to U(Ⅵ), belonged to UO2 + x, and the rest belonged to UO2 [49, 50]. This confirmed that all the U(Ⅵ) was practically reduced. Similarly, as shown in the HR XPS spectrum of Sb 3d (Fig. 5b), the peaks corresponding to Sb(Ⅲ) at 529.9 and 539.8 eV after adsorption moved to 530.8 and 540.2 eV, respectively, corresponding to Sb(Ⅳ) or Sb(Ⅴ) [51] after a 3-h photocatalysis, confirming that all the Sb(Ⅲ) was nearly oxidized.
3.2.2. EXAFS
To investigate the binding environment and short-range structure of Sb(Ⅲ) and U(Ⅵ) interacting in FTC, EXAFS spectroscopy was employed to confirm the sorption mechanism between them. Figures 6a and 6c show the k3-weighted U LIII-edge Sb K-edge EXAFS spectra of the reference (UO2 + 2, UO2, Sb(OH)3, and Sb(OH)5); adsorption; and after-3-h-of-photocatalysis samples, which exhibited a distinct cyclic evolution. The corresponding Fourier-transforms results are shown in Figs. 6b and 6d, and the corresponding parameters are tabulated in Table 1. The R-factor showed that this fitting work had high reliability.
Table 1
Structural parameters around U and Sb uranium for the reference and in the solution over FTC after adsorption and 3h photocatalysis derived from EXAFS analyses.
Sample
|
Path
|
Na
|
R(Å)b
|
σ2(Å2)c
|
R factor
|
UO2 + 2
|
U-Oax
|
2
|
1.75
|
0.0026
|
0.0058
|
U-Oeq
|
5
|
2.51
|
0.0130
|
UO2
|
U-O
|
8
|
2.34
|
0.0086
|
0.0012
|
U-U
|
12
|
3.65
|
0.0026
|
U-O2
|
24
|
4.23
|
0.0013
|
Sb(OH)3
|
Sb-O
|
3
|
1.99
|
0.0039
|
0.0177
|
Sb(OH)5
|
Sb-O
|
6
|
1.97
|
0.0037
|
0.0155
|
After adsorption
|
U-Oax
|
2
|
1.82
|
0.0018
|
0.0184
|
U-Oeq/O
|
6
|
2.23
|
0.0064
|
U-N
|
1
|
2.74
|
0.0038
|
Sb-O
|
3
|
2.04
|
0.0040
|
0.0164
|
After photocatalysis
|
U-O
|
2
|
2.11
|
0.0035
|
0.0131
|
U-O1
|
6
|
2.36
|
0.0069
|
U-N
|
2
|
2.63
|
0.0038
|
U-U
|
3
|
3.65
|
0.0065
|
U-O2
|
14
|
4.23
|
0.0032
|
Sb-O
|
1
|
1.92
|
0.0030
|
0.0075
|
Sb-O1
|
5
|
2.02
|
0.0036
|
a Coordination number.
b Distance between absorber and backscatter atoms.
c Debye-Waller factor to account for both thermal and structural disorders.
d Indicates the goodness of the fit.
As shown in Figs. 6a–b, for the typical structure of UO2 + 2, the spectral fitting led to Oax (axial oxygen atom) at 1.75 Å and Oeq (equator oxygen atom) at 2.51 Å, which correlated with the distances reported for other uranyl aqueous species [52]. When U(Ⅵ) was adsorbed in the middle of the TiO2/g-C3N4 heterojunction by FTC, the aforementioned amplitudes were slightly stretched to 1.82 and 2.23 Å, respectively, and the other Fourier-transforms feature could be fitted by U–N with a bond distance of 2.74 Å. However, no new bond distance corresponding to that formed by the expected active-site formation of TiO2 and U(Ⅵ) after adsorption could be found in the results. Based on the change in coordination number, it was speculated that the active site of TiO2 should be oxygen, which formed a coordination bond in the heterojunction whose length was similar to that of U = O bonds in UO2 + 2 and was listed with them. The formation of the heterojunction and coordination could also explain why the oscillation feature near 2.34 Å in the typical structure of UO2 [53] changed to 2.11 and 2.36 Å after the 3-h photocatalysis. The coordination number of the U–N shell at 2.63 Å increased, suggesting that the nitrogen groups of g-C3N4 were more closely bound to U(Ⅳ).
As shown in Figs. 6c–d, the change in the bond length of Sb(Ⅲ) and its products after adsorption correlated with the aforementioned rule, i.e., squashing by the heterojunction altered the length Sb–O, compared with the standard bond length of the main species of Sb(Ⅲ) (Sb(OH)3) [54], and the bond formed by Sb(Ⅴ) and active oxygen on TiO2 was concealed in other Sb–O bonds. The fitting results of the samples tallied with those of the reference samples. Thus, it could be concluded that UO2 and Sb(OH)5 existed in the product.
Finally, the adsorption of U(Ⅵ) and U(Ⅳ) could be attributed to the inner-sphere surface complexation caused by N(g−C 3N 4)–U, O(TiO 2)–U(UO2+ 2)/U(UO 2), and O(TiO 2)–Sb(Sb(OH) 3)/Sb(Sb(OH) 5) within the TiO2/g-C3N4 heterojunction.
3.2.3. DFT of structures
The results of previous material characterizations showed that the reaction site could be related to the TiO2/g-C3N4 heterojunction. To confirm this, we optimized all the possible adsorption structures, including U(Ⅵ) or Sb(Ⅲ) adsorbed on one side of {001}TiO2 or g-C3N4 and in the middle of the TiO2/g-C3N4 and Fe3O4/TiO2 heterojunction (previous experimental results showed that Fe3O4 hardly participated in the reaction; therefore, no relevant adsorption structures were established). The corresponding adsorption energies (Table S2) showed that the adsorption energies of UO2 + 2 and Sb(OH)3 adsorbed within the TiO2/g-C3N4 heterostructure were the lowest, which correlated with the previous conclusion that the TiO2/g-C3N4 heterostructure was where the reaction occurred. The final adsorption models are shown in Figs. 7a and 7c. The models of the products in the heterojunctions after the reaction were optimized to determine the mechanism of photocatalysis (Figs. 7b and 7d). The removal rate in the solution over FTC was slightly higher than that over TC on the premise that no reaction occurred in the TiO2/Fe3O4 heterojunction. This could be because the addition of Fe3O4 increased the dispersion of active sites on TiO2 [38]. The redox potentials of UO22+/UO2, Sb(OH)3/Sb(OH)5, and Fe(Ⅲ)/Fe(Ⅱ) were 0.411, 0.59, and 0.77 V, respectively. Therefore, as mentioned in subsection 3.1.3, Fe3O4 could also participate in the reduction of U(Ⅵ) and the oxidation of Sb(Ⅲ) in an insoluble form, such that more U(Ⅵ) and Sb(Ⅲ) were removed. However, since these were not the main reactions, we do not discuss the TiO2/Fe3O4 heterojunction in detail herein.
The length of the bonds related to U or Sb shown in these models correlated with the EXAFS results, which confirmed the authenticity of these models. Furthermore, the models reflected the changes in material structures, and the bonding environment correlated with the XPS results, thereby confirming that the unsaturated O2c sites (Fig. S8) of {001}TiO2 were regarded as some of the reaction sites that participated in the adsorption and photocatalytic reaction of U(Ⅵ) and Sb(Ⅲ). Based on this, N–(C3) and C–N = C were the adsorption locations of U(Ⅵ) and U(Ⅳ), respectively, which were the products of the photoreduction reaction. The new bond formed after the reaction between U(Ⅳ) and TiO2/g-C3N4 strengthened the force between them, compared with that after adsorption. The corresponding adsorption energies (Table S2) showed that the adsorption energies of UO2 + TiO2/g-C3N4 and Sb(OH)5+TiO2/g-C3N4 were lower than those of UO2 + 2 + TiO2/g-C3N4 and Sb(OH)3+TiO2/g-C3N4, respectively, which were conducive for the removal of U and Sb from the solution after the reaction.
The analysis results of the total density of states (TDOS) (Figs. 8a–b), projected density of states (PDOS) (Figs. 8c–d), and models (Fig. 7) after the adsorption and reaction jointly illustrated the formation of the new force. After adsorption, the new chemical bond originated from the U(Ⅵ) 4f, Sb(Ⅲ) 3s, and Sb(Ⅲ) 3p orbitals hybridized with the O 2p, Ti 3d, and N 2p orbitals to form these heterojunction complexes, N–U(Ⅵ)–O–Ti and N…Sb(Ⅲ)–O–Ti. The photocatalytic product, UO2, existed above the Fermi level in the energy range of 0.71–2.11eV because the electrons occupied the antibonding orbitals between UO2 and the heterostructure. Through the analysis of the DOS redistribution caused by the exchange of electrons and the formation of new bonding displayed as s character states of Sb elevated to above the Fermi level of 2.9 eV, which hybridized with Ti, O, C, and N to form hybridization (Fig. 7d), we could infer that g-C3N4 used the accumulated holes to produce oxidized groups in the free space between it and Sb(OH)3 and oxidized Sb(Ⅲ) to create a strong interaction with g-C3N4 after the reaction. This phenomenon again clarified the role of g-C3N4 in Sb oxidation.
3.3. Catalytic mechanism
3.3.1. XPS analysis
As shown in Figs. 5b–c, after being optically excited, the peaks of O shifted to high binding energies, while the peaks of N slightly moved in the opposite direction, indicating that even if the photogenerated carriers were consumed by U(Ⅵ) and Sb(Ⅲ), the amount of e− around O increased, while that around N reduced. It reflected the strong interface interaction between them and U and Sb. It was speculated that photogenerated electrons and holes were detained on TiO2 to reduce U(Ⅵ) and g-C3N4 to oxidize Sb(Ⅲ), respectively, after they were directed to transfer in TiO2/g-C3N4.
3.3.2. DFT of Photocatalytic enhancement mechanism
The band structures, TDOS, and PDOS of the materials are shown in Figs. 9a and 9c. The calculated band gap of these materials was underestimated, compared with the experimental Eg because the actual materials had a certain thickness, while the models were single-layer structures. The DFT was limited in that the discontinuity in the exchange-correlation potential was not considered [55], but the trend of changes fitted with the experimental results and the calculated band gap of TiO2 were similar to reported results [54, 56], which verified the rationality of the DOS. Figures 9a and 8c show that the N 2p, O 2p, and Ti 3d orbitals overlapped the 2π* antibonding orbitals near the Fermi level, confirming that there was a strong interaction among them, endowing FTC with high reactivity and a strong binding energy. Figure 9a shows that the CB maximum (CBM) of TiO2/g-C3N4 was mainly occupied by the O and Ti atomic orbitals of TiO2, and the N atomic orbital of g-C3N4 mostly controlled the VB maximum (VBM), which showed that FTC had a heterostructure and again determined the direction of charge separation and transfer, similar to the results of the Mott–Schottky plots and UV–VIS diffuse reflectance spectroscopy.
The calculated results of the charge-density difference (Fig. 9a) showed that charge redistribution occurred in the heterostructure. Bader charge analysis showed that a 0.03-eV charge transferred from g-C3N4 to TiO2 on the other side of the heterojunction. The electrostatic potentials (Fig. S9) showed that g-C3N4 with a large work function (4.68 eV) was an oxidation-type photocatalyst, whereas TiO2 (4.00 eV) was a reduction-type photocatalyst. When the electrons flowed from TiO2 with a high Fermi level to g-C3N4 until the two Fermi energies reached the same level, the net charges distributed over both sides of the heterostructure formed a built-in electric field where negative and positive charges accumulated at g-C3N4 and TiO2, respectively. Under illumination, since the CB edge potential of g-C3N4 was more negative than that of TiO2, the photoinduced electrons on the g-C3N4 particle surfaces transferred easily to TiO2, and the photoinduced holes on TiO2 transferred to g-C3N4. The built-in electric field promoted the transfer of electrons from g-C3N4 to the CB of TiO2 and hindered the photoexcited electrons in the CB of TiO2 and holes generated in the VB of g-C3N4 from moving, which made them accumulate and difficult to compound. Oppositely, the recombination of photoexcited electrons and holes was more likely to occur in the VB of TiO2 and CB of g-C3N4, but it prevented the holes and electrons in TiO2 and g-C3N4, which had accumulated numerous electrons and holes, respectively, from being compounded by themselves.
From the charge-density difference after the adsorption and reaction (Figs. 9b–e), there was charge transfer among U(Ⅵ), U(Ⅳ), Sb(Ⅲ), Sb(Ⅴ), and TiO2/g-C3N4. Further Bader charge analysis of the TiO2/g-C3N4 + U model during the reaction showed that TiO2 interacted with U and delivered 0.15 eV, whereas the electron shift obtained by g-C3N4 was 0.61 eV. With the same method used to analyze Sb, the two aforementioned values became 0.25 and 0.04 eV, respectively, indicating that TiO2 provided electrons for U during the photoreaction process. Contrarily, g-C3N4 contributed more to the oxidation of Sb, This correlated with the expected FTC performance and showed that the electrons accumulated on TiO2 were fully utilized by U. In addition, the holes accumulated on g-C3N4 were fully utilized by Sb.
These results showed that the improvement in the photocatalytic ability of FTC was due to the enhancement of the separation of electrons and holes by the built-in electric field and the band structure through the regulation of the transmission path of photoexcited carriers. This process is intuitively shown in Fig. 10.
3.4. Reaction procedure: Photocatalytic enhancement mechanism
The sorption tendency of U(Ⅵ) and Sb(Ⅲ) on FTC accounted for the synergistic effect, including electrostatic interaction, surface complexation, and chemical precipitation. Sb(Ⅲ) existed as uncharged Sb(OH)3 over a wide pH range [54], suggesting that Sb(Ⅲ) was tightly bound on the surface sites via the formation of surface complexes rather than electrostatic interaction. Zeta potential values (Fig. 11a) revealed that the point of zero charge (pHpzc) of the surface was 5.03 mV. Electrostatic attraction between FTC and the positively charged UO22+ (the predominant U(Ⅵ) species at low pH [57]) could be formed at pH > 5.03, which increased the pHpzc to 5.42 mV after adsorption. The formation of electrically neutral UO2, low-mobility Sb(Ⅴ) species, or a negatively charged inner-sphere complex caused the pHpzc to move toward a lower zeta potential after the reaction. Furthermore, using a pH of 5.03 for the solution ensured an accurate analysis of the mechanism of the photocatalytic reaction.
To study the possible active species generated during the photocatalytic processes, ESR analysis was performed (Figs. 11b–c). After 5 min of light irradiation, paramagnetic signals showed a 1:2:2:1 typical cleavage pattern of OH• and six characteristic peaks of the superoxide radical. The intensity of these peaks decreased after 3 h of photocatalysis. This result confirmed that •OH and O2• ⁻ were the main active species generated in the photocatalytic reaction. Considering that •OH and O2• ⁻ were produced by photogenerated holes, the photogenerated electrons on FTC were fully utilized.
To elucidate the mechanism of the photocatalytic reaction, free-radical trapping tests were performed over FTC. Figures 11d–f show the changes in the redox rates and concentrations of U(Ⅵ) and Sb(Ⅲ), and the first-order kinetic models are shown in Fig. S10. The reduction rate of U(Ⅵ) was 87.9%, and the oxidation rate of Sb(Ⅲ) was 73.2% without a scavenger.
For antimony (Figs. 11d–e), the scavenging of O2• ⁻ and •OH led to a decrease in the oxidation rate of Sb(Ⅲ), indicating that they were both involved in the oxidation reaction. This conclusion correlated with that of the ESR experiment. EDTA-2Na, a scavenger of h+, blocked the formation of •OH [28], resulting in the poor oxidation of Sb(Ⅲ). This was mutually verified by the results of trapping •OH by TBA. Compared with bubbling N2 to discharge O2 and prevent the generation of O2• ⁻, the reaction speed and oxidation rate of Sb(Ⅲ) after BQ scavenged O2• ⁻ was higher, indicating that O2 could directly participate in the oxidation of Sb(Ⅲ) (Sb(Ⅲ) + O2→Sb(Ⅴ) + O2• ⁻). This could be verified by the fact that after K2Cr2O4 consumed e−, i.e., prevented O2 from forming O2• ⁻ with it [58], the oxidation rate of Sb(Ⅲ) only dropped slightly and was considerably higher than that in the experiment of bubbling N2. Moreover, the reduction rate and speed of Sb(Ⅲ) increased after oxygen was added to the system to supplement O2 and O2• ⁻. Thus, O2• ⁻, •OH, and O2 jointly played important roles in the oxidation of Sb(Ⅲ).
For uranium (Figs. 11d and f), the reduction rate of U(Ⅵ) significantly dropped after K2Cr2O4 consumed e−, confirming that e− was important for the reduction of U(Ⅵ). Upon the addition of EDTA-2Na as an h+ scavenger, the photogenerated electrons and holes were not easy to composite, and FTC provided more electrons for reduction. Therefore, it was again confirmed that e− was significant for the reduction of U(Ⅵ) through an increase in the uranium reduction rate and speed after h+ was scavenged. The oxygen-free environment caused by N2 and the absence of O2• ⁻ caused by BQ reduced the reduction rate of U(Ⅵ), which showed that O2• ⁻ had a positive effect on the reduction of U(Ⅵ) [59]. In addition, the reduction rate and speed of U(Ⅵ) after adding BQ were lower than those under N2. During the experiments filled with O2 or N2 (Figs. S10a and 11f), the reduction rate of the former slowed down, and the latter sped up. Based on these results and according to the premise that O2• ⁻ was the product of O2 and e−, it could be inferred that the interaction principle between O2• ⁻ and U(Ⅵ) was as follows: O2 competed with U(Ⅵ) for photogenerated electrons to generate O2• ⁻, which led to a reduction in speed; however, O2• ⁻ could react with U(Ⅵ) to generate low-valent uranium.
However, in theory, filling O2 would lead to excessive O2 competing for e−, lowering the reduction rate of U(Ⅵ) below that under N2. However, the opposite was the reality. A relationship was made with the oxidation process of antimony, i.e., excessive O2 oxidized antimony to generate O2• ⁻ without competing for electrons, which reduced uranium. This conclusion that the O2• ⁻ produced by the oxidation of Sb(Ⅲ) and consumed by the reduction of U(Ⅵ) was complementary could explain why ESR showed that the change in •OH content during the reaction was more evident than that in the O2• ⁻ content.
Summarily, the photocatalytic enhancement mechanism of FTC was to first use active substances, including •OH and e−, to compete for photogenerated electrons and holes, respectively, thereby fully harnessing the photocatalytic ability of FTC by preventing the recombination of the electrons and holes. The O2• ⁻ generated by Sb(Ⅲ) and O2 to reduce U(Ⅵ)) and the consumption of O2• ⁻ advanced the oxidation of Sb(Ⅲ).
The reaction process was summarized as follows (4)-(15) and is shown in Fig. 12.