2.1 Catalyst preparation and characterization

All catalysts were prepared by incipient impregnation method using γ-alumina (213 m²/g) as support. The catalysts were prepared using a water solution of ammonium metatungstate hydrate (Sigma-Aldrich, 89 %). After the impregnation step, the samples were dry at 120 °C for 12 h and then calcined by 5 h at 500 °C in a quartz reactor with a continuous feed of dry air (15 mL/min). The catalysts were labelled according to their tungsten loading as Wx, where x indicates the metal loading (% wt). Some calcined catalysts were reduced up to 750 °C in TPR-H₂ equipment and the letter “R” was added to the label, i.e. W17R corresponding to the catalyst with 17 wt.% of W and reduced at 750 °C.

In order to characterize the surface oxide species, TPR-H₂ studies were carried out, the TPR-H₂ experiments were performed in a conventional temperature programmed reduction apparatus. The test was carried out using a flow of an H₂/Ar mixture (70% H₂, 25 cm³/min) with a heating rate of 10 °C/min up to1000 °C, if necessary, the sample was kept at this temperature until the reduction was completed. In the case of the partially reduced catalysts, the reduction was carried out up to 750 °C using the same flow and heating rate.

Elemental composition was evaluated by SEM-EDS using a Jeol JSM-5900 LV microscope equipped with an energy dispersive X-ray elemental analysis system (EDS) Oxford-ISIS. Textural properties were measured by N₂ adsorption-desorption isotherms of the samples at 77 K using a Tristar Micrometrics apparatus. The isotherms were analyzed by the BJH method. Before each textural analysis, the samples were degassed for 8 h in vacuum at 350°C. XRD studies were carried out with a Siemens D500 powder diffractometer with Cu Kα radiation from 20 to 70 degree in two theta scale with a step size of 0.02 s. Raman spectroscopy was used to obtain further information of the tungsten oxide surface species, the Raman spectra were recorded from a stationary sample at room temperature and pressure. The samples were analyzed with a DxR Raman microscope Thermo Scientific using a 532 nm line as the excitation source and the spectral resolution was 1 cm^-1. The acquisition time for each spectrum was 20 s and 50 scans were averaged.

2.2 Catalytic experiments

Activity tests were carried out in a glass batch reactor jacketed in a thermally controlled water bath at 60 °C and atmospheric pressure fitted with a condenser and mechanical stirrer. A model mixture with a total S content of 609 ppm was prepared using acetonitrile (Empura, 99%) as solvent and S-compounds (DBTs): dibenzothiophene (DBT, Sigma-Aldrich, 98%), 4-methyl dibenzothiophene (4-MDBT, Sigma-Aldrich, 96%) and 4,6-dimethyl dibenzothiophene (4,6-DMDBT, Sigma-Aldrich, 97%). In a typical run, 50 cm³ of the model mixture, catalyst (100 mg) and 0.5 mL of H₂O₂ (PQF, 30%, V/V) were added to the reactor with vigorous stirring (750 rpm) for avoiding mass transfer issues (^{Cedeño-Caero et al., 2008}; ^{González-García & Cedeño-Caero, 2009}; ²⁰¹⁰). The initial molar ratio of the oxidizing agent/sulfur compound (O/S) was 6. If the O/S ratio decreased below 2, small amounts of H₂O₂ were added for maintaining the O/S ratio around 6. Reaction products were analyzed by gas chromatography, the analysis was performed in an HP5890 Series II Gas Chromatograph with a PONA capillary column (Methyl silicone Gum, 50 m x 0.2 mm x 0.5 μm film thickness). Reactant and product identifications were achieved with GC-PFPD (Varian CP-3800) and GC-MS (HP5890 Series II with MS detector). H₂O₂ concentration was measured by standard iodometric titration.

3.1 Tungsten oxide based catalysts

In order to measure the intrinsic catalytic activity of the catalysts, ODS tests were carried out in a simplified 2 phases ODS system (extraction solvent and solid catalyst), considering extraction step of sulfur compounds from hydrocarbon phase is faster than oxidation step (^{Bernal & Caero 2005}; ^{Ismagilov et al., 2011}). The Fig. 1 shows the DBT-sulfone yield in function of time, for simplicity, only DBT-sulfone yield is shown since all the other DBTs show similar tendencies.

Fig. 1 Catalytic performance for DBT oxidation of tungsten based catalysts. 

The catalytic performance of the tungsten oxide catalysts is highly dependent on the metal loading in each catalyst. Low loading catalysts such as W1, W4 and W9 (1, 4, and 9 % wt tungsten loading) exhibit low yield throughout the catalytic tests. W13 and W17 exhibit a better performance than the catalysts with lower loading, reaching around the 100% of the DBT-sulfone yield throughout the reaction time. However, the catalysts with a higher tungsten loading (W21 and W25) show a decrease in the catalytic performance in comparison with W17.

In general, the results showed in Fig. 1. fit very well with a first-order reaction comportment, in consequence, the kinetic constants associated with these catalysts were calculated using a first-order fit and are shown in Table 1, moreover, the molar percentage of H₂O₂ that reacted to form sulfones (H₂O₂ to sulfones yield) is shown. Additionally, the kinetic constants of DBTs oxidation are plotted in Fig.2 in function of the surface density of tungsten, considering a homogeneous distribution of the metal loading on the support.

Fig. 2 Pseudo-first order kinetic constants of DBTs oxidation in function of surface density of tungsten. 

The value of the kinetic constants increases with the tungsten metal loading until 2.68 atoms/nm², then, there is a decrease of the kinetic constants with the increase in metal loading. Furthermore, the activity increase is more remarkable at around 2 and 2.7 atoms/nm², while the increase in activity between 0 and 2 atoms/nm² is less representative. This behaviour is observed for all the DBTs tested. It is important to point out that the increase in the kinetic constant value, between 0 and 2.68 atoms/nm², is nonlinear with the metal loading increase, suggesting that the increase in activity is not only related to an increase in active phase in the catalysts.

The XRD patterns of the tungsten oxide based catalysts are shown in Fig. 3. In general, there are not crystalline phase with the exception of W25 and W21 which present small diffraction peaks at 23, 24, 33 and 34 degrees. According to the references, these diffraction peaks are assigned to WO₃ (monoclinic) (^{Kuzmin, Purans, Cazzanelli, Vinegoni, & Mariotto, 1998}).

Fig. 3 XRD patters of (a) γ-alumina, (b) W1, (c) W4, (d) W9, (e) W13, (f) W17, (g) W21 and (h) W25. 

The SEM-EDS and N₂-physisorption analysis of the tungsten oxide based catalysts are shown in Table 2. In general, the theoretical loading matches with the experimental composition detected. Moreover, the SEM images (not shown) do not show agglomerates of oxide metal except for W25 and W21, this result supports the results obtained from XRD. Regarding the textural properties, the surface area and pore volume decrease with the increase in metal loading, this trend has been observed and discussed in previous works (^{Loefanti et al., 1997}; ^{Massoth, 1979}; ^{Mitra, Wachs, & Deo, 2006}; ^{Schneider, Hudec, & Solcova, 2008}) and it is related to porosity obstruction by the increase in metal oxide loading.

TPR-H₂ profiles obtained from the tungsten oxide based catalysts are shown in Fig. 4, moreover, the reduction profile of bulk WO₃ is shown as a reference. Bulk WO₃ reduction profile exhibits two main peaks with a temperature of reduction rate maximum (T_max) at 605 and 730 °C, furthermore, there are two shoulders, one before the first peak and the second before the second peak. WO₃ reduction takes place in three steps (^{Ogata, Kamiya, & Ohta, 1973}), the first step corresponds to reduction WO₃ to W₂₀O₅₈ and it is associated with the shoulder between 490 and 550 °C. The second step is the W₂₀O₅₈ reduction to WO₂ and corresponds with the reduction peak with T_max at 605 °C. Finally, the third step is the WO₂ reduction to W and is associated with the reduction peak with T_max at 730 °C. The shoulder after the peak at 730 °C is associated with diffusional issues in the third reduction step (^{Magnus, Bos, & Moulijn, 1994}; ^{Vermaire & Van Berge, 1989}).

Fig. 4 TPR-H₂ profiles of tungsten based catalysts and bulk WO₃. 

W1 reduction profile does not show any reduction peak, this could be due to the low tungsten loading (0.15 atoms/nm²) that leads to form very strong interactions with alumina, which cannot be reduced at temperatures of 1000 °C. W4 shows a reduction profile with a dominant peak between 870 and more than 1000 °C with the T_max above the 1000 °C, W9 exhibits a peak like the peak of W4, but the T_max is at 976 °C. W13 exhibits a peak with a T_max of around 940 °C and W17 exhibits a broader peak with a T_max at 877 °C. In W4, W9, W13 and W17, as the metal loading increases, the reduction begins at lower temperatures and the peak is broader.

The peak observed in W4, W9, W13 and W17 profiles is associated with the reduction of WO₄ tetrahedral surface species and WO₆ octahedral surface species (Fierro, 2006; ^{Salvati, Makovsky, Stencel, Brown, & Hercules, 1981}). WO₆ species are polymerized in continuous two-dimensional chains, their formation is enhanced at high tungsten oxide loadings and began to reduce between 620 °C and more than 900 °C (^{Horsley, Wachs, Brown, Via, & Hardcastle, 1987}; ^{Thomas, De Beer, & Moulijn, 1981}). The WO₆ species are reduced in a large temperature range due to different degrees of polymerization which imply different resistances to be reduced. WO₄ species are isolated or as a dimer (^{Vuurman & Wachs, 1992}; ^{Vuurman, Stufkens, Oskam, Deo, & Wachs, 1996}; ^{Weckhuysen, Jehng, & Wachs, 2000}) and they are reduced at temperatures up 900°C, in fact, the catalyst W4 points out a T_max above the 1000 °C.

W21 and W25 show reduction profiles with the same main peak than W4, W9, W13 and W17. Moreover, there are several shoulders between 550 and 750 °C, like WO₃ reduction profile, suggesting that exist tungsten oxide crystals in these catalysts in addition to the single surface species discussed for the catalysts W4, W9, W13 and W17. Finally, the small peak around 520 °C observed in W13, W17, W21 and W25 corresponds with a small formation of microcrystals or clusters of WO₃ (^{Reddy & Varma, 2004}; ^{Thomas et a.,1981}; ^{Vermaire & Van Berge, 1989}), and there are formed due to small cationic impurities on the alumina (^{Vurrman et al.,1996}).

Raman spectra of W25, W21, W17 and W13 are shown in Fig. 5. W13 exhibits a spectrum with a strong Raman band at 963 cm^-1, moreover, there is a band centred at 880 cm^-1 and a pair of weak bands at 352 and 219 cm^-1. The Raman band at 963 cm^-1 is assigned to the symmetric stretch of W=O bond. For hydrated conditions, the Raman band associated with the symmetric stretch of W=O changes according to the predominance of WO₄ (951 cm^-1) or WO₆ species (975 cm^-1) (^{Kim, Burrows, Kiely, & Wachs, 2007}; ^{Ostromecki, Burcham & Wachs 1998}; ^{Wachs & Roberts, 2010}). Therefore, the band at 963 cm^-1 for W13 is generated by a combination of WO₄ and WO₆ species. The Raman band at 880 cm^-1 is assigned to the symmetric stretch of O-W-O bonds from WO₆ polymerized species (^{Ostromecki, Burcham, Wachs, Ramani, & Ekerdt, 1998}; ^{Vuurman & Wachs, 1992}). The band at 352 cm^-1 is related to bending mode of WO₆ and WO₄ species (^{Ostromecki et al., 1998}), it shifts its position according to the WO₆/WO₄ ratio change, taking lower values when to WO₆/WO₄ ratio increases. The band at 219 cm^-1 is associated with the bending mode of polymeric W-O-W linkages (^{Horsley et al., 1987}; ^{Vuurman & Wachs, 1992}).

Fig. 5 Raman spectra of high loading tungsten oxide based catalysts. 

W17 possesses the same bands that W13, however, the main band in W17 is centred at a higher value (968 cm^-1) and the band at 352 cm^-1 in W13 is shifted to 338 cm^-1 in W17. According to the references, the shift of these bands is related to a change in the WO₆/WO₄ ratio, the shift in these bands is associated with an increase in WO₆ species. Moreover, the peak associated with the W=O symmetric stretch in W9, W4 and W1 (not shown) tend to 951 cm^-1 indicating a higher presence of WO₄ species while the tungsten loading decrease.

Regarding W21 and W25, the Raman bands at 806, 714 and 270 cm^-1 are associated with WO₃ crystalline phase (^{Daniel, Desbat, Lassegues, Gerand, & Figlarz, 1987}). Moreover, W21 shows a weak Raman band at 977 cm^-1 indicating WO₆ species, this result points out that W21 (3.29 atoms/nm²) possesses tungsten surface species and tungsten crystalline phase at the same time which could be an indicator of this catalyst is just above the monolayer coverage.

3.2 Partially reduced tungsten based catalysts

Some tungsten oxide based catalysts were partially reduced at 750 °C in a continuous H₂ flow. The results of the catalytic tests of reduced catalysts are shown in Table 3, the letter “R” was added to the label to indicate that the catalysts were partially reduced up to 750 °C.

All reduced catalysts show a high kinetic constant value and possess a high H₂O₂ to sulfones yield. Moreover, the W loading in the reduced catalyst does not affect significantly the value of the kinetic constants. Results in SEM micrographs and N₂-physisorption (not shown) displayed the formation of metal particles over the surface and specific surface areas around 125 m²/g for all reduced catalysts. These results suggest a loss of area due to a collapse of micropores in the support during the high-temperature reduction and a decrease of the dispersion of the active phase owing to the formation of clusters and metal particles.

4.1 Tungsten oxide surface species and ods activity

Below the monolayer coverage, the tungsten oxide catalysts showed a nonlinear increase in their activity whit the increase in the tungsten loading. There is a considerable increase in activity around the catalysts W13 (2.06 atoms/nm²), which is 3-fold more active than W9 and only possesses around 50% more active phase. This same phenomenon can be observed between the catalysts W17 and W13 showing that W17 is 2-fold more active than W13 but W17 only possesses around a 30% more active phase than W13. Supporting the previous, the catalyst W9 is around the double of active than W4, which corresponds with a linear relationship between the tungsten loading and the increase in activity. These differences point out that the activity does not increase just for the increase in the tungsten loading. Furthermore, the activity decreases for W21 and W25, according to the characterizations, this is due to the loading is above the monolayer coverage and WO₃ crystalline phase has been formed on the surface blocking the active sites.

According to the results in TPR-H₂, there is a change in the type and predominance of tungsten surface species over the alumina surface, which coincide with the change in activity observed in the catalytic tests. The TPR-H₂ results pointed out that at low tungsten loadings, species WO₄ is formed over the surface, and when the tungsten loading increases, there is a change in the peak of reduction associated with the formation of WO₆ species, especially after the 2.06 atoms /nm² (W13). Moreover, the whole reduction peak is shifted to lower temperatures pointing out the presence of several WO₆ polymerized chains. At the same time, the peak associated with WO₄ becomes less representative with the loading increase.

Regarding Raman spectroscopy, the same change in the predominance of surface species (WO₄ and WO₆) is observed. High tungsten loading enhances the formation of WO₆, while the WO₄ species becomes fewer commons. This could be due to the WO₄ species are forced to polymerize with other WO₄ species to form WO₆ due to a decrease in the number of hydroxyl groups available over the alumina surface. Small tungsten loadings allow the formation of WO₄ species which have a stronger interaction with the alumina than WO₆. However, with the increase in tungsten loading and due to the strong interaction between the alumina and the tungsten oxide, the WO₄ species could be forced to change their coordination and polymerized between them to allow more tungsten oxide surface species over the alumina.

In order to explain how the surface tungsten oxide species are related to the activity of the catalysts, it is necessary to consider previous studies related to the ODS mechanism (^{Bregeault, 2003}; ^{Campos-Martin, Capel‐Sanchez, Perez‐Presas, & Fierro, 2010}; ^{García-Gutiérrez et al., 2008}; ^{Ishihara et al., 2005}; ^{Rafiee, Sahraei, & Moradi, 2016}; ^{Rezvani, Oveisi, & Asli, 2015}). The ODS reaction mechanism in aprotic solvents, as acetonitrile, using H₂O₂ as oxidizing agent, occurs by a nucleophilic attack of the sulfur compound to peroxo complex formed by the metal oxide surface species and the H₂O₂ molecule. The Fig. 6 shows the possible mechanism for the oxidation of DBTs by WO₆ surface species.

Fig. 6 A possible mechanism for the oxidation of DBT by H₂O₂-WO₆ surface species. 

The mechanism consists in three steps: (1) the formation of a hydroperoxytungstate by a nucleophilic attack of the H₂O₂ to the surface species, (2) a peroxo complex is formed by the loss a water molecule and (3) the sulfur molecule makes a nucleophilic attack on the peroxo complex to form a sulfoxide and then, by a similar way, to form the sulfone.

The high degree of coordination of WO₆ in comparison with WO₄ species suggests that WO₆ polymerized species are more susceptible to nucleophilic attacks. In WO₆, the electronegativity of the 6 oxygen atoms generates a high coordinated tungsten atom susceptible to nucleophilic attacks. The formation of the hydroperoxytungstate and the oxidation of the sulfur compound by the peroxo complex occurs by nucleophilic attacks, therefore, the surface species with metal cores with lack of electrons favour the steps 1 and 3 of the mechanism, showing a higher activity in the oxidation of DBTs.

Regarding the catalysts W21 and W25, the decrease in activity can be explained by the results of XRD, TPR-H₂ and Raman spectroscopy. All of these characterizations showed evidence of crystalline phase over W21 and W25. Possibly, the crystalline phase blocks the surface tungsten oxide species, and at the same time, it points out that the crystalline phase is less active than the surface tungsten oxide species. Moreover, the results of H₂O₂ consumption (Table 1) of W21 and W25 indicate that there is an increase in H₂O₂ consumption but a decrease in the sulfones production, this result suggests that the WO₃ crystalline phase promotes the decomposition of H₂O₂ instead of to use it to produce sulfones.

4.2 Partially reduced catalysts and ods activity

According to the activity tests of partially reduced catalysts, there is a remarkable change in their behaviour in comparison with the calcined tungsten based catalysts. All of the reduced catalysts showed high activity, and a better H₂O₂ to sulfones yield than the calcined catalysts (Table 1 and Table 3) without considering the tungsten loading. Moreover, during the catalytic test, all partially reduced catalysts showed poor activity the first 15 min of reaction (not shown), but then, the activity increased considerably. This phenomenon could be associated with a “kinetic delay period”, which could be related to a reoxidation of the tungsten species over the catalysts which take the role of the active phase.

These results point out a new way to catalyze the ODS reaction, the partially reduced catalysts possess different active phases and exhibit activities and selectivities (H₂O₂ to sulfones yield) comparable to the best tungsten oxide based catalyst (W17). These results are similar to a previous work where a system based on Mo-V was studied (^{Alvarez-Amparan & Cedeño-Caero, 2017}; ^{González-García & Cedeño-Caero, 2010}; ²⁰⁰⁹), in such study, an increase in ODS activity was reported when the catalysts were partially reduced, regardless the loss of surface area and the metal loading. These results suggest that the partially reduced catalysts possess a high activity and overall DBTs yield. Nevertheless, an exhaustive study is needed for understanding completely the performance of partially reduced tungsten based catalysts.