A large amount of SO_x compounds is released due to the combustion of fuel oil, causing serious environmental pollution and ecological damage. Recently, deep desulfurization of petroleum products to produce clean fuels is receiving extensive attention [1]. The conventional industrial hydrodesulfurization requires high temperature and pressure harsh reaction conditions, which is ineffective to heterocyclic sulfur compounds such as the dibenzothiophene (DBT) and its derivatives [2]. Up to now, several strategies have been reported for desulfurization of sulfur compounds, such as oxidative desulfurization [3,4], photocatalytic oxidative desulfurization [5], and physical adsorptive desulphurization [6]. Among these strategies, photocatalytic oxidative desulfurization (PODS) has been attracting increasing attention and been widely studied due to its high efficiency and low cost for the removal of heterocyclic sulfur compounds and their derivates under mild operating conditions [7]. However, the current desulfurization photocatalysts, such as Ag-TiO₂ [8], Ti₃C₂/g-C₃N₄ [9,10], CeO₂ [11] and heteropolyacid (HPA) [12], are still disadvantaged due to their poor light absorption, fast recombination of photogenerated carriers, and poor structural modification, which cannot achieve satisfactory and efficient deep desulfurization performance [13].

Metal-organic frameworks (MOFs), as novel porous structural materials, possess large surface area and abundant unsaturated sites and easy-tailored structure [14,15], which are favorable to adsorption and oxidation performance of sulfur compounds [16-18]. Especially, Ti-based MOFs exhibit certain oxidative desulfurization performance at 60 ℃ [19,20]. So far PODS using pristine MOFs is rarely reported because of the unsatisfactory visible-light responsiveness and limited active sites as well as ambiguous reaction mechanism. Semiconductor modification and metal doping to modify MOFs are effective strategies to improve the photo-response ability and photocatalytic activity. Bagheri et al. constructed MoO₃-MOF hybrid photocatalysts and exhibited excellent photocatalytic desulfurization performance due to the synergic effects of the active surface of MOFs together with active center of MoO₃ under UV and visible light irradiation [21]. The construction of hetero-structure not only improves photoexicited electron behavior, but also enhance catalytic activity [22,23]. Besides constructing heterojunction strategy, defect engineering via metal ion doping draws tremendous attention to expose more active sites and modulate electron interaction [24,25]. The incorporation of doping Ce with high abundance, low cost and inherent Ce⁴⁺/Ce³⁺ redox pairs into pristine MOFs has been widely investigated. For example, Ebrahim et al. prepared Ce(Ⅲ) doped Zr-based MOFs to exhibit excellent NO₂ adsorption at ambient conditions [26]. Yang et al. synthesized Ce ion-implanted MIL-96 via incipient wetness impregnation for high-efficiency removing fluoride from water [27]. The Ce⁴⁺/Ce³⁺ redox cycling might also endow MOFs with attractive adsorptive performance. Zhang et al. developed Ti-doped UiO-66(Ce) via cation-exchange method to improve the photocatalytic reduction performance, in which Ti mediator species induced abundant oxygen vacancies and active radicals (O₂^•− and ^•OH) as well as enhanced photoelectronic generation and transfer [28]. Patrycja et al. reported novel Ti-substituted (Ce)UiO-X MOFs@TiO₂ heterojunction to exhibit outstanding photocatalytic performance due to the synergistic effect of Ce-Ti species including reversible Ce⁴⁺/Ce³⁺ and Ti⁴⁺/Ti³⁺ redox mediators and the incorporation of TiO₂ [29]. However, utilizing metal-doped MOFs with longer-lived excited states and effective charge separation as photocatalyst for the PODS of the sulfur compound from light oil has been rarely exploited.

In this work, Ce-doped MIL-125-NH₂ samples (Ce/MIL-125-NH₂) with controlled Ce-doping content were systematically investigated for the photocatalytic oxidation of DBT. The experimental results show that appropriate Ce doping improves the separation efficiency of photo-generated carriers under visible light, thereby achieving the excellent conversion of DBT. In addition, the reaction temperature significantly enhanced the PODS performance. Finally, a thermo-enhanced PODS mechanism was explored through radical trapping experiment and control experiments.

The structures of as-synthesized Ce/MIL-125-NH₂ composites were revealed by XRD analysis. As presented in Fig. 1, the main diffraction peaks of MIL-125-NH₂ with good crystallinity are consistent with the XRD pattern of our previous report [30], which proves the successful configuration. Meanwhile, the XRD patterns of Ce-doped MIL-125-NH₂ samples are basically similar to that of MIL-125-NH₂. But the diffraction peak at 6.9 ° shifts to a lower angle due to the larger radius of Ce³⁺ (1.034 Å) than Ti⁴⁺ (0.68 Å) (Fig. 1b). The shift offset is greater with the increased Ce doping content, indicating that the Ce species are successfully doped into the MIL-125-NH₂ lattice without significantly affecting the crystal structure and crystallinity.

As shown in Fig. S1 (Supporting information), the morphology of the sample was characterized by SEM. MIL-125-NH₂ is disc-shaped with a diameter of 700-800 nm and a thickness of 200-300 nm. But the diameter and thickness of all the Ce-doped samples decrease sequentially. 2.0 mol% Ce/MIL-125-NH₂ has a diameter of 300-600 nm and a thickness of 50-60 nm. The doped Ce inhibits the growth of MIL-125-NH₂ or disturbs the interactions between the Ti-metallic centers and the organic ligands [26]. In addition, EDS spectra of 1.0 mol% Ce/MIL-125-NH₂ in Fig. S1e also indicate that Ce³⁺ is uniformly doped into the framework of MIL-125-NH₂.

The FT-IR spectra was used to confirm the bond structure of MIL-125-NH₂ (Fig. S2a in Supporting information). The peak at 3418 cm⁻¹ is related to the stretching vibration of the O-H bond. The strong absorption peak at 1648 cm⁻¹ is the characteristic peak of the carbonyl functional group in the O–C=O structure. The peak at 1391 cm⁻¹ is attributed to the C-N bond between the carbon atom connecting the benzene ring and the nitrogen atom on the amino group, while the peaks in the range of 540-800 cm⁻¹ belong to the stretching vibration of the O-Ti-O bond [31,32]. The incorporation of Ce does not induce new absorption peaks, which indicates that the Ce-doped MIL-125-NH₂ maintains the original framework structure. As shown in Fig. S2b (Supporting information), the MIL-125-NH₂ and Ce/MIL-125-NH₂ samples demonstrate a typical type Ⅰ curve. The specific surface areas of MIL-125-NH₂, 0.5 mol%, 1.0 mol% and 2.0 mol% Ce/MIL-125-NH₂ are 1274.9, 1158.3, 1213.8 and 1041.8 m²/g, respectively, indicating that slight Ce doping does not affect the BET to a large extent, but a large amount of Ce doping would induce the partial presence of Ce retains in the pores of the MOF, causing a decrease in specific surface area. The pore size distribution curves in Fig. S2c verify the similar microporous structure between MIL-125-NH₂ and Ce/MIL-125-NH₂.

In addition, the element valence state and chemical environment of the sample were further analyzed by XPS. In the spectra of Ti 2p (Fig. S3a in Supporting information), the binding energies located at 458.8 eV and 464.5 eV are assigned to Ti 2p_3/2 and Ti 2p_1/2. After Ce doping, it can be observed that both two peaks are shifted to lower binding energy, indicating Ce doping increases the electron density of Ti. Moreover, two new peaks corresponding to Ti³⁺ ions appear near 459 eV and 464 eV, which can be attributed to the redistribution of charges within the system due to Ce doping. In the N 1s spectra (Fig. S3b in Supporting information), the peaks at 399.0, 400.0 and 402.5 eV correspond to pyridine N, pyrrolic N and amino N, respectively. After Ce doping, the N 1s peaks shift towards lower binding energies which are caused by the increased electron density of N. In the spectra of O 1s (Fig. S3c in Supporting information), there is a Ce-O peak at 530.5 eV for Ce/MIL-125-NH₂ samples [33], which effectively indicates the interactions of Ce with Ti-O center or the organic linker via π-complexation or a chelation process [26]. The spectra of Ce 3d is shown in Fig. S3d (Supporting information) where eight characteristic peaks can be observed. The peaks at 885.1 eV and 903.5 eV are attributed to Ce³⁺, while the three pairs of peaks at 882.9 eV and 901.4 eV, 889.3 eV and 908.1 eV, and 898.9 eV and 917.2 eV are attributed to Ce⁴⁺ ions, indicating the coexistence of Ce⁴⁺ and Ce³⁺ in the Ce/MIL-125-NH₂ sample [34]. Incorporation of Ce into Ti-MOFs leads to the formation of the Ti-Ce clusters and generates abundant Ti³⁺ and Ce⁴⁺ on the surface. The generation of Ce⁴⁺/Ce³⁺ and Ti⁴⁺/Ti³⁺ redox mediators promote the formation of more Lewis acid site and enhances the photo-oxidation ability. Determined by the ICP-AES measurements of Ce/MIL-125-NH₂ in Table S1 (Supporting information), the actual amounts of Ce are calculated to be 0.73 mol%, 1.46 mol%, and 2.92 mol% for 0.5 mol%, 1.0 mol% and 2.0 mol% Ce/MIL-125-NH₂, respectively.

In order to explore the effect of Ce incorporation on the optical behavior, the photo-response properties and band structure of MIL-125-NH₂ and Ce/MIL-125-NH₂ were examined by UV-vis spectroscopy and Mott Schottky plots. It can be seen from Fig. S4a (Supporting information) that the absorption edge of MIL-125-NH₂ is about 500 nm, and the visible-light absorption range is slightly red-shifted after doping with Ce. The results indicate the doping of Ce can slightly broaden the visible light absorption range. The estimated band gaps of MIL-125-NH₂, 0.5 mol%, 1.0 mol%, and 2.0 mol% Ce/MIL-125-NH₂ are 2.32 eV, 2.31 eV, 2.30 eV and 2.29 eV, respectively. The Mott-Schottky plots in Fig. S5a (Supporting information) were measured to indicate the n-type semiconductor and the corresponding flat band potentials of MIL-125-NH₂ and 1.0 mol% Ce/MIL-125-NH₂ (−1.15 and −0.89 V vs. Ag/AgCl). Therefore, combined with band gap, the conduction band (CB) and valence band (VB) were calculated, as shown in Fig. S5b. In order to further verify that the Ce doping facilitates the separation and transfer efficiency of photogenerated carriers, photoluminescence (PL) spectroscopy was performed. As shown in Fig. 2a, the 1.0 mol% Ce/MIL-125-NH₂ sample has the lowest peak intensity at the excitation wavelength of 350 nm. In addition, as observed in Figs. 2b and c, 1.0 mol% Ce/MIL-125-NH₂ has the highest photocurrent density and the smallest charge transfer resistance, indicating that appropriate amount of Ce doping has stronger photo-excited carrier separation and transfer ability under visible light, which is conducive to improving the photocatalytic performance effectively [35].

To investigate the photocatalytic properties of Ce-doped MIL-125-NH₂, the PODS of DBT was firstly carried out under visible light irradiation at 30 ℃. As shown in Fig. 3a, the DBT extraction rate is 34% at the beginning of the reaction due to the addition of acetonitrile. Acetonitrile plays two important roles in the system. On the one hand, the catalyst was uniformly dispersed in this solution to avoid sticking to the wall during the reaction. On the other hand, the acetonitrile acts as an extractant of substrate DBT to contact well with the catalyst. In Fig. 3a, 1.0 mol% Ce/MIL-125-NH₂ has a 100% removal rate of DBT within 22 min at 30 ℃ while it is only 69% for MIL-125-NH₂. The enhanced performance is largely due to the existence of Ce⁴⁺/Ce³⁺ and Ti⁴⁺/Ti³⁺ redox mediators, which would promote the separation of photo-induced charge carriers, the generation of more Lewis acid sites and the decomposition of H₂O₂. In order to deeply investigate the effect of temperature and light, 1.0 mol% Ce/MIL-125-NH₂ was tested at different temperatures (e.g., 30 ℃, 40 ℃, 50 ℃ and 60 ℃) with or without visible light irradiation, respectively. As shown in Fig. 3b, the temperature significantly improves the oxidative removal rate of DBT, especially, the heat-driven effect is extremely obvious at 2-10 min. Furthermore, increasing the reaction temperature accelerates the reaction rate [36,37]. Particularly, the oxidative desulfurization rate reaches 100% only within 12 min at 60 ℃, indicating the positive synergistic effect of light and heat in the oxidative removal of DBT. However, only 75% and 80% removal rates are obtained with only light irradiation and without light irradiation at 30 ℃ after 25 min respectively. In addition, the effect of the initial DBT concentration was discussed for the 1.0 mol% Ce/MIL-125-NH₂ sample under the same O/S ratio, as shown in Fig. 3c. The results show that the 1000 ppm DBT is the optimal initial concentration, while the poor oxidation efficiency of DBT at 500 ppm concentration is attributed to the low concentration of active radicals generated by insufficient H₂O₂ content. The decrease in photocatalytic activity at higher DBT concentrations may be due to the high substrate content [38]. The PODS with different H₂O₂ content in 1000 ppm DBT concentration was performed to further investigate the effect of O/S molar ratio. As shown in Fig. 3d, the desulfurization rate gradually increases with the increase of H₂O₂ content within a certain range. The excessive H₂O₂ amount leads to negligible improvement because too excessive oxidant cannot be effectively activated by photocatalysts. When 50 uL H₂O₂ (O/S = 5) was added to the system, the best desulfurization performance was obtained under 15 min illumination at 30 ℃, demonstrating the important role of the O/S molar ratio in PODS. In addition, PODS over Ce-doped MIL-125-NH₂ at O₂ atmosphere was performed. The results in Fig. 4a indicate the reaction hardly proceeds under oxygen atmosphere, demonstrating the crucial role of H₂O₂ as active radical initiator in photocatalytic oxidative desulfurization.

To explore the general utilization of Ce-doped MIL-125-NH₂ catalysts, the thermo-assisted photocatalytic oxidation of other typical aromatic sulfides, such as benzothiophene (BT) and 4,6-dimethyldibenzothiophene (4,6-DMDBT), were carried out to prove the universality of the prepared catalysts under the same condition (Fig. 4b). The sulfur removal efficiency of 4,6-DMDBT reached 100%, and BT can reach 85% under visible light irradiation at 30 ℃. Due to the presence of steric hindrance effect and electron density of the sulfur atom, the order of reactivity is DBT > 4,6-DMDBT > BT [39]. The sequence of S electron densities is 4,6-DMDBT > DBT > BT, high S electron density is conducive to electrophilic addition reaction. However, due to the steric hindrance effect, 4,6-DMDBT with two methyl groups has lower desulfurization performance than DBT. Hence, the as-prepared catalyst has the potential application value for practical desulfurization. To evaluate the recycling stability of 1.0 mol% Ce/MIL-125-NH₂, the reacted catalysts were collected by centrifugation, washed and dried overnight at 80 ℃, and subjected to DBT desulfurization experiments under the same reaction conditions. As shown in Fig. S6a (Supporting information), the slightly decreased desulfurization rate may be due to the catalyst loss or the residual organic adsorbed within the pores of Ce/MIL-125-NH₂ even though the XRD patterns and SEM images in Figs. S6b and c (Supporting information) are consistent before and after cycling. As shown in Fig. S7 (Supporting information), the FT-IR and UV-vis spectra of 1.0 mol% Ce/MIL-125-NH₂ before and after cycling are different, indicating the possible adsorption of residual sulfur compounds. The similar BET area (1213.8 m²/g vs. 1251.7 m²/g) and the decreased pore volume (0.804 cc/g vs. 0.522 cc/g) indicate that pores of 1.0 mol% Ce/MIL-125-NH₂ may be blocked by the residual sulfur compounds. The presence of mesopores structure in Fig. S7d is attributed that H₂O₂ addition probably destroyed pore structure.

In order to explore the PODS mechanism of DBT, radical trapping experiments were performed to clarify the active radicals. p-Benzoquinone (P-BQ), tert-butanol (IPA), phenol and silver nitrate (AgNO₃) can act as scavengers for superoxide radicals (O₂^•−), hydroxyl radicals (^•OH), hole (h⁺) and electron (e⁻), respectively. As shown in Fig. 5a, the addition of IPA significantly inhibits the DBT oxidation efficiency, proving ^•OH is the main radicals. It is clear from Figs. 4a and 5a that O₂^•− is not dominant reactive radical. The addition of phenol and AgNO₃ scavengers could decrease the sulfur removal, indicating h⁺ and e⁻ may participate in the reaction. In addition, only about 37% extraction rate is achieved without catalyst or H₂O₂, demonstrating the important roles of catalyst and H₂O₂ during the oxidative desulfurization process. ESR tests were also performed on MIL-125-NH₂ and 1.0 mol% Ce/MIL-125-NH₂ to further verify the presence of ^•OH. As shown in Fig. S8 (Supporting information) and Fig. 5b, H₂O₂ partially decomposes into ^•OH in the dark, and ^•OH is generated at a faster rate with the involvement of visible light irradiation, leading to a faster reaction rate. It can be also seen that H₂O₂ can be decomposed faster in the 1.0 mol% Ce/MIL-125-NH₂ system than in MIL-125-NH₂, demonstrating the doped Ce facilitates the decomposition of H₂O₂.

In order to illustrate the conversion pathway of DBT in the extraction-PODS system, the collected acetonitrile liquid after 10 min and 20 min reaction was detected by GC-MS (Fig. S9 in Supporting information). After 10 min illumination, the DBT and DBTO₂ were detected in acetonitrile phase, while only DBTO₂ was detected after 20 min, indicating that there was obvious transfer behavior of DBT molecules between the oil phase and the acetonitrile phase [40]. During the reaction process, the acetonitrile phase extracts DBT in model oil to reach the extraction-dissolution equilibrium, and the product DBTO₂ is also soluble in the acetonitrile phase. Fig. 5c illustrates vividly the PODS process. The reaction process consists of two steps. Firstly, DBT molecules are rapidly extracted from the oil phase to the acetonitrile phase. Secondly, H₂O₂ partially decomposes into ^•OH via photogenerated electrons under visible light illumination. And the photogenerated holes react with water or OH⁻ to generate ^•OH. Finally, the adsorbed DBT is oxidized to DBTO₂ [29,41,42].

Summarily, Ce-doped MIL-125-NH₂ samples with different Ce-doping content were investigated as photocatalysts for the PODS performance of DBT under mild reaction conditions. Particularly, the desulfurization rate of DBT over 1.0 mol% Ce/MIL-125-NH₂ reaches 100% within 22 min at 30 ℃ under visible light irradiation. The electron interaction between Ce-Ti species leads to the formation of Ce⁴⁺/Ce³⁺ and Ti⁴⁺/Ti³⁺ redox mediators, which not only increase the Lewis acid sites, but also promote the separation and transfer of photogenerated electron-hole pairs, thereby realizing efficient desulfurization efficiency. The synergistic effect of external heat and visible light significantly increased the oxidative desulfurization rate. The feasible doping strategy and thermal enhancement effect will open new avenues for the design of other visible-light-responsive photocatalysts to solve environmental pollution and energy problems.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

This work was supported by the National Key Research and Development Program of China (No. 2021YFB3500700), the National Natural Science Foundation of China (No. 21976054), and Fundamental Research Funds for the Central Universities (No. FRF-TP-20-005A3).

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2022.107766.