First, γ-Al₂O₃ was prepared by the strip extrusion method with the following process conditions: pseudo-boehmite (Shandong Aluminum Corporation) and deionized water were mixed thoroughly at a mass ratio of 1:1, and then, 2 wt.% sesbania powder (Shandong Xunda Chemical Group Co., Ltd.) and 5 wt.% nitric acid (HNO₃, Aladdin, 65%) were added. The mixture was extruded by using an extruding machine at 30 MPa pressure, and the extruded strips were shaped into a clover type with a diameter of 1.5 mm. The supports was naturally dried in a place of protection from light and ventilation for 24 h and then dried in a drying oven at 120°C for 12 h, and finally, the required γ-Al₂O₃ supports were obtained in a muffle furnace at a heating rate of 2°C min⁻¹, kept at 500°C for 4.0 h, and naturally cooled to room temperature. The obtained supports were sieved into particles with the size of 20–40 meshes.

The incipient wetness co-impregnation method was used to synthesize Ni–Mo/γ-Al₂O₃ catalysts, and the active metal loading was MoO₃:16 wt.% and NiO:4 wt.%, respectively. The specific preparation process is as follows: a certain mass of ammonium heptamolybdate tetrahydrate [(NH₄)₆Mo₇O₂₄⋅4H₂O, Aladdin, ≥99.8%] was added to an appropriate amount of deionized water and ammonia to dissolve it completely, and then, nickel nitrate hexahydrate [Ni(NO₃)₂⋅6H₂O, Aladdin, ≥99.8%] solution was added and stirred uniformly to obtain the nickel–molybdenum co-impregnated solution. The prepared Ni–Mo co-impregnation solution was evenly added dropwise to the alumina surface, and the samples were naturally dried at room temperature for 24 h, then dried in an oven for 6 h, and finally, heated to 500°C at a rate of 2°C min⁻¹ in a muffle furnace kept at a constant temperature for 4.0 h.

A series of Ga-modified Ni–Mo/γ-Al₂O₃ catalysts were prepared using the stepwise impregnation method, in which the Ga₂O₃ loading amount was 2 wt.%, 4 wt.%, and 6 wt.%, respectively, and the specific steps were as follows: gallium nitrate (Ga(NO₃)₃⋅H₂O, Aladdin, ≥99.5%) was dissolved in the deionized water, and then, the solution was loaded using incipient wetness impregnation on the Ni–Mo/Al₂O₃. The sample was dried in air for 24 h and then dried in an oven at 120°C for 4 h. Finally, it was heated to 500°C at 2°C min⁻¹ in a muffle furnace, with a constant temperature for 4.0 h. The obtained catalysts were denoted as SCal (2 wt.% Ga₂O₃), Scam (4 wt.% Ga₂O₃), and SCah (6 wt.% Ga₂O₃) depending on Ga₂O₃ content from low to high.

All catalyst samples synthesized were characterized on a PANalytical advanced powder diffractometer using Cu Kα radiation with an accelerating voltage of 40 kV and current of 40 mA in the 2θ interval of 5–80 °via a scanning rate of 0.2 °s⁻¹, by which the patterns of X-ray diffraction (XRD) were recorded. The surface areas of the tested samples were carefully calculated by using the Brunauer–Emmett–Teller (BET) equation in the relative pressure (P/P0) range of 0.05–0.30. The pore diameter distributions and pore volumes of all the investigated samples were calculated using the Barrett–Joyner–Halenda (BJH) method from the N₂ adsorption isotherms. The catalysts were sulfided in a JQ-III fixed-bed microreactor with 2 wt.% CS₂ cyclohexane solution at 320°C for 4 h. Afterward, the investigated catalysts were taken on a Philips Tecnai G2 F20 instrument with an acceleration voltage of 200 keV to obtain the images of active phases via a high-resolution transition electron microscope (HRTEM). To analyze the results of the average slab length and average stacking number of MoS₂, no less than 300 MoS₂ slabs were counted using the equations reported elsewhere (Ortega–Domínguez et al., 2015; W. Zhou et al., 2017). Average   slab   length   L ¯ = ∑ i = 1 n n i l i / ∑ i = 1 n n i , (1)   Average   stacking   number   N ¯ = ∑ i = 1 n n i N i / ∑ i = 1 n n i , (2) where l _i is the length of the slab, n _i is the number of slabs with length l _i , and N _i is the number of layers in slab i.

The acidity of the catalysts was assessed by NH₃ temperature programmed desorption (NH₃-TPD), which was performed on an Auto Chem II 2920 automatic chemical adsorption instrument. All catalysts were pretreated in an Ar and air mixture (v:v = 3:1) at 773k for 60 min and then cooled to 373 k in Ar flow and adsorbed NH₃ for 40 min. Then, the physically adsorbed NH₃ was removed in a continuous Ar flow for 90 min. The reactor temperature was then programmed to increase with a heating rate of 10°Cmin⁻¹ for 65 min, and the amount of desorbed NH₃ was detected by using a continuous effluent gas monitor with a thermal conductivity detector (TCD). After the IR spectra were recorded on a Magna 560 FT-IR analyzer, pyridine-adsorbed Fourier transform infrared measurements (Py-FTIR) was used to assess the acidity properties of all the samples by using pyridine as a probe molecule. Firstly, the sample was dehydrated for 3 h at 623 K under 10⁻² Pa vacuum. Second, to obtain the saturated adsorption of the Py-FTIR spectrum, pure pyridine vapor was added to the measured samples at room temperature for 30 min. Finally, the adsorbed pyridine was evacuated, respectively, at 523k and 623k for an hour so as to obtain the desorbed Py-FTIR spectra.

4,6-DMDBT was used as a probe to assess the HDS performances of SCal, Scam, and SCah catalysts. On the fixed-bed reactor whose length was 200 mm and inner diameter was 8 mm, 2.0 g of catalyst was loaded with a particle size of 20–40 meshes. In the first place, the pre-sulfidation of tested catalysts was carried out with 2.2 wt.% CS₂ cyclohexane solution at 330°C and 4 MPa with H₂/oil (v/v) of 60 for 4 h. Then, the reactor temperature was allowed to drop to the temperature required for the reaction at the previous condition. Stabilization of the 4,6-DMDBT cyclohexane solution at a mass concentration of 1.0% was carried out for 4 h at 4.0 MPa and 270°C, 280, and 290°C, respectively. The H₂/oil (v/v) was maintained at 120 when all catalysts were evaluated. The collected products of 4,6-DMDBT was determined by the GC–MS technique on an Agilent 4890D gas chromatograph equipped with a 60 m capillary Rtx-1 column (0.25 mm, RESREK). The column temperature was increased from 50 to 300°C at a rate of 15°C min⁻¹, while the N₂ pressure was maintained at 0.3 MPa and the flow rate was 30 ml min⁻¹. The 4,6-DMDBT conversions and DDS route selectivity (3,3′-DMBP selectivity) were counted using the following equations, respectively: 4,6 − DMDBT   conversions   C o n = ( 1 − m 4,6 − D M D B T ) × 100 % , (3) DDS   Route   selectivity   S D D S = m 3,3 ′ − D M B P / ( 1 − m 4,6 − D M D B T ) × 100 % , (4) where m _4,6-DMDBT is the molar fraction of remaining 4,6-DMDBT and m _3,3′-DMBP is the molar fraction of 3,3′-DMBP tested in the liquid products.

The spectra of Ni–Mo/Al₂O₃ modified by the stepwise impregnation method with different Ga loadings determined by powder X-ray diffraction are shown in Figure 1. It can be seen from the diagrams the characteristic diffraction peaks (4.4.0), (4.0.0), and (3.1.1) of γ-Al₂O₃ are almost unchanged, which shows that the crystal structure of γ-Al₂O₃ is not affected. No characteristic peaks of Ga₂O₃ and other Ga compounds are found in Figure 1, which indicates that Ga₂O₃ was highly dispersed on Ni–Mo/Al₂O₃ and loaded well. However, diffraction peaks (0.2.2), (-3.1.2), and (2.2.0) of NiMoO₄ were observed, illustrating that the introduction of Ga₂O₃ brought about the agglomeration of Ni and Mo elements. In addition, the characteristic peak intensity of NiMoO₄ increased with the increase in the content of Ga₂O₃, which manifests that the introduction of excessive Ga₂O₃ is not conducive to the dispersion of Ni and Mo elements.

The method of N₂ physical adsorption–desorption was carried out to confirm the size of the specific surface area, average pore diameter, and pore volume of Ni–Mo/Al₂O₃, SGal, SGam, and SGah. The abovementioned textural properties are calculated and displayed about the four synthesized samples in Table 1, and the results of pore size distribution for two samples are recorded in Figure 2. From the data in Table 1, it is clearly found, especially for SGam, that the specific surface area increased from 214 m² g⁻¹ to 227 m² g⁻¹with the addition of Ga₂O₃ in comparison with Ni–Mo/Al₂O₃, indicating that the introduced Ga₂O₃ had a wide distribution in the inner and outer surface of the catalyst; however, the pore volume and pore size decreased to some extent because parts of Ga₂O₃ blocked the pore channel of alumina. It can be seen from Figure 2 that the pore size distribution curve moved to the direction of small pores, and the peak area decreases slightly, which also proved that a large number of pores were blocked or even disappeared with the introduction of gallium.

In order to observe the morphologies, characteristics, and dispersion about the sulfide active phase of Ni–Mo/Al₂O₃ and SGax (x stands for l, m, and h) catalysts, the images obtained by HRTEM characterization are shown in Figure 3 . At least, 300 MoS₂ slabs of each catalyst were counted to statistically analyze the results of the average slab length and average stacking number of the active-phase slabs in Table 2. It can be clearly seen from Table 2 that the introduction of Ga contributed to an increase in the average stacking number of MoS₂ and its change rule kept pace with the increase in Ga loading. Especially, the largest increase of average stacking number for SGal is from 1.1 to 3.6 layers for Ni–Mo/Al₂O₃, which displays that the introduction of Ga weakened the Mo–O–Al bond energy and caused a weaker interaction between MoS₂ and the support, resulting in the formation of more type II Ni–Mo–S reactive phases with high stacking layers. In accordance with the abovementioned regulation, the variation of the average length also has a positive correlation with the Ga₂O₃ content. Also, the degrees of average length increase in the following order: SGal (2.4 nm) < Ni–Mo/Al₂O₃ (2.7 nm) <SGam (3.1 nm) <SGal (3.3 nm). The average length slightly increased from 2.7 nm for sample Ni–Mo/Al₂O₃ to 3.3 nm for sample SGah. This can also be explained by the abovementioned data.

Acidity plays a crucial role in the formation of the active phase and the HDS reaction, so NH₃-TPD characterization of the series samples was carried out and is displayed in Figure 4. The peak temperature and peak area of the NH₃ desorption were calculated for each of the investigated samples and are summarized in Table 3 . The NH₃-TPD profiles in Figure 4 reveal the peak temperature of the Ga–Ni–Mo/Al₂O₃ catalyst existed mainly between 150 and 300°C, indicating that the catalyst was mainly the prince of weak acid sites and medium–strong acid sites, which was weaker than that of the Ni–Mo/Al₂O₃ catalyst. It can be seen from Table 3 that the introduction of Ga made the NH₃ desorption peak shift to a lower temperature, manifesting that the weak acid sites of the catalyst mainly increased. According to the comparison of peak area, the number of acid sites of the Ni–Mo/Al₂O₃ catalyst was increased by introducing Ga₂O₃, but it was negatively correlated with introduced Ga₂O₃ content. Thus, the acid sites varied significantly for SGal. That phenomenon explained that the acid sites were covered by Ga₂O₃, leading to reduction in the degree of increase on the number of acid sites.

It is well known that acid types and strength are the key points to decide the performance of the corresponding catalysts. As a consequence, pyridine desorption FTIR analyses were performed at different temperatures, and the results are listed in Table 4, where the specific data of weak and strong Brönsted acid sites (BAS) or Lewis acid sites (LAS) were derived from the information of pyridine desorption at 200 and 350°C, respectively. It can be clearly observed from Table 4 that the introduction of 2 wt.% Ga₂O₃ enhanced the weak B-acid, and the B-acid enhancement facilitates the desulfurization of 4,6-DMDBT through the isomeric desulfurization pathway (ISOM). Moreover, after the isomerization of the methyl group on 4,6-DMDBT, the steric hindrance to sulfur atoms decreased significantly, which provided access to direct desulphurization just by hydrogenolysis of sulfur atoms, and the path to remove sulfur was considered to be very ideal. In addition, the introduction of different contents of Ga increased the Lewis acid, among which the weak Lewis acid increased more and the strong Brönsted acid decreased slightly. This is consistent with the characterization of NH₃-TPD. Except for the enhancement of weak Brönsted acid by SGam, the amount of Ga₂O₃ introduced had little effect on the acid type and amount of Brönsted acid and Lewis acid of Ni–Mo/Al₂O₃, but the amount of Ga₂O₃ introduced increased the total acid amount.

As is concretely confirmed, though the appropriate content of Ga₂O₃ modification can effectively improve the physicochemical properties of the catalysts from the characterization results mentioned above, the favorable conditions for conversion and selectivity at different temperatures need to be explored. Thus, 4,6-DMDBT served as the probe molecules in the assessment of the HDS performance of Ni–Mo/Al₂O₃ and SGax series catalysts on a fixed-bed reactor under a total pressure of 4 MPa and the liquid hourly space velocities (LHSVs) of 2.5 h⁻¹ at different reaction temperatures (in the range of 270–290°C). The variation of conversions on 4,6-DMDBT at different temperatures is clearly displayed in Figure 5A, suggesting that the activities of all the investigated catalysts increased with the increase in the reaction temperature. These results also pointed out that there was no significant disparity in the conversion of SGax series catalysts with different Ga₂O₃ loadings, and the highest conversion rate was obtained when the Ga₂O₃ loading was 2 wt.%. In addition, with the increase in reaction temperature, the difference of conversion rate between SGax series catalysts and the Ni–Mo/Al₂O₃ catalyst became smaller, which resulted from the increase in the conversion rate of 4,6-DMDBT at high temperature close to the reaction end point.

Scheme 1 shows the HDY and DDS reaction route over Ga–Ni–Mo/Al₂O₃ catalysts separately. HYD pathway desulfurization is a tedious step to form the products 3,3′-dimethylcyclohexanebenzene (3,3′-DMCHB) or 3,3′-dimethylbicyclohexyl (3,3′-DMBCH), while DDS pathway desulfurization is achieved with fewer steps and without excess aromatic rings and produces 3,3′-dimethylbiphenyl (3,3′-DMBP). The DDS route selectivity over different catalysts at the same conditions mentioned above was obtained, and the results are shown in Figure 5B. It can be observed from Figure 5B that, with the increase in reaction temperature, the DDS selectivity of the tested catalysts on the DMDBT decreased, owing to that Ni gradually began to be reduced as the reaction temperature exceeded 290°C. Especially, the metal Ni has a strong hydrogenation performance. The reason mentioned above led to the enhancement of the HYD path. In addition, it was also found that if the reaction temperature is higher than 290°C, the high hydrogenation performance of Ni can even convert a small amount of DDS products into HYD products, but it was generally considered that this reaction will not occur, which was the unique characteristic of Ni-containing hydrotreating catalysts. Further increase in temperature would lead to a profound increase in the proportion of hydrogenation conversion of DDS products to HYD products. In order to properly deal with this situation, the metal promoter can be replaced with the load of Co with low hydrogenation performance, and the amount of the Ni loading or the Ni/Mo atomic ratio ought to be reduced.

The influence of different LHSVs in the investigated series catalysts on the 4,6-DMDBT HDS reaction is shown in Figure 5C. As can be clearly noticed from the figure, with the decrease in LHSVs which meant increasing the residence time of raw materials, the conversion rate of 4,6-DMDBT increased gradually and over the tested catalysts was in the order of Ni–Mo/Al₂O₃<SGah < SGam < SGah. When the LHSV decreased gradually, the increasing rate of conversion rates slowed down gradually as the reaction is close to complete conversion. Thus, the conversion rate gap between different Ga₂O₃ loading catalysts was bridged.

Figure 5D reveals the transformation about selectivity of DDS route by stepwise impregnation of the Ga–Ni–Mo/Al₂O₃ catalyst with different LHSVs on 4,6-DMDBT at a reaction temperature of 280°C. As it is vividly shown in Figure 5D, the gallium-modified catalysts exhibited higher DDS route selectivity than the Ni–Mo/Al₂O₃ catalyst at high LHSV. However, the DDS route selectivity of SGax series catalysts for 4,6-DMDBT exhibited a negative correlation with the variation of LHSV. With the decrease in LHSV, the DDS selectivity of SGax series catalysts for 4,6-DMDBT even decreased to about the corresponding selectivity of the Ni–Mo/Al₂O₃ catalyst. On the one hand, the incorporation of Ga improved the average stacking number of the active-phase slabs and generated more type II NiMoS active phases which weekly interact with the support or do not interact with the support at all, and it is considered to be very active (Nikulshin et al., 2014; van Haandel et al., 2015), so that the HYD route got strengthened. On the other hand, the residence time of raw materials became longer while the LHSV decreased, which is beneficial to the HYD route with longer reaction path and slower reaction rate. In addition, because of the strong hydrogenation ability of the reduced metal Ni, a small part of DDS products was also transformed into HYD products. All mentioned above were identified as the reason for which the variation of LHSV had impact on the DDS route selectivity.