The continuing decline in the quality of petroleum feedstocks and the recently enacted regulations limiting the emissions of pollutants have made sulfur and nitrogen removal one of the paramount problems in the refining industry^［1-2］. Researchers throughout academy and industry have developed a variety of new hydrotreating catalysts for the ultra⁃deep hydrodesulfurization （HDS），among which the transition metal phosphides showed high activities and were expected to be a new generation of hydrotreating catalysts replacing the traditional metal sulfide catalysts.

A number of metal phosphides （such as MoP^［3-5］，WP^［4，6］，Fe₂P^［7-8］，FeP^［8-9］，CoP^{［7，10-13］}，Co₂P^{［4，11-13］} and Ni₂P^{［7，13-17］}） have been successfully synthesized and used for HDS reactions. Oyama et al^［7］ prepared SiO₂ supported Ni₂P，CoP and Fe₂P by the temperature⁃programmed reduction （TPR） method，and found that the activity of the phosphides for the HDS reactions followed the order of Ni₂P/SiO₂>CoP/SiO₂>Fe₂P/SiO₂，while that for the hydrodenitrogenation （HDN） reactions followed the order of CoP/SiO₂>Ni₂P/SiO₂>Fe₂P/SiO₂. In particular，the Ni₂P/SiO₂ was shown to be more active （90% conversion of dibenzothiophene （DBT）） than the commercial Ni⁃Mo⁃S/Al₂O₃ （76% conversion of DBT）. In comparison，the CoP/SiO₂ exhibited 32% conversion of DBT at the same conditions. Bussell et al^［11］ found that the HDS activity of thiophene on the Co_xP_y/SiO₂ catalysts prepared by the TPR method was affected by the composition of Co_xP_y，and the pure phase of CoP/SiO₂ showed the highest activity. In addition，Co₂P/SiO₂ and CoP/SiO₂ catalysts were also prepared by TPR method for the hydrodechlorination of chlorobenzene^［12］. It was found that the cobalt phosphides were more active than metallic Co，and the activity increased when more P were incorporated for the transformation from Co₂P to CoP. Thus，the catalytic activity for the reaction followed the order of CoP/SiO₂>Co₂P/SiO₂>Co/SiO₂.

The metal phosphide catalysts reported in literature were mostly obtained by the TPR method. Although this process was simple，the required temperature was relatively high，leading to catalysts with large particles and poor dispersions. Thus，developing metal phosphide catalysts with higher dispersion and more active sites is still highly desirable. Recently，a series of highly dispersed Ni₂P catalysts supported on SiO₂^［18］，Al₂O₃^［19-20］，MgAlO^{［19，21］} and ZrO₂⁃Al₂O₃^［22］ were prepared in our lab by a liquid phase phosphidation process using triphenylphosphine （PPh₃）. The above catalysts possessed small Ni₂P particles with diameters of about 6~8 nm，high CO uptakes of about 269~324 μmol·g^-1 and thus high activities for the HDS of DBT，HDN of quinoline and hydrogenation of tetralin to decalin. One of the key factors to the above success could be attributed to the usage of highly loaded （60wt%~80wt%），reduced and dispersed Ni catalysts as precursors. In this way，Fe₂P/C and FeP/C were also prepared^［23］. The FeP/C was found to be active and stable for the HDS of DBT while the Fe₂P/C was unstable and converted into Fe_0.96S during the hydrotreating reactions. When Al₂O₃ was used as the support，the dispersion of supported Fe was enhanced and the FeP/Al₂O₃ with smaller FeP particles （~16 nm） was prepared，which exhibited significantly higher activity than the FeP/C for the HDS reactions^［24］.

Compared with Ni₂P catalysts，the studies on the cobalt phosphides for HDS reactions attracted less attention over the past decades owing to their relatively lower activity for the HDS reactions^［1］. In addition，the cobalt phosphide catalysts reported in literature usually possessed large CoP or Co₂P particles with small amounts of surface active sites^{［4，7，10-11］}. Previously，we prepared highly loaded Co/SiO₂ catalysts （containing 60wt%~80wt% Co） for Fischer⁃Tropsch synthesis^［25-26］ which exhibited the high reducibility and dispersion of cobalt with small Co particles （11~13 nm） and high H₂ uptakes （71~118 μmol·g^-1）^［26］. In this work，a 60%Co/Al₂O₃ catalyst was prepared and then phosphided into Co₂P/Al₂O₃ and CoP/Al₂O₃ catalysts using PPh₃ in the liquid phase. Due to the strong interactions of cobalt species with Al₂O₃，the Co/Al₂O₃ catalyst possessed highly dispersed cobalt species. A small amount of carbon was added into the Co/Al₂O₃ in order to promote the reduction of supported cobalt species^［26］. A Co₉S₈/Al₂O₃ catalyst was also prepared by the sulfidation of the above 60%Co/Al₂O₃ using CS₂ in liquid phase. The activities of these catalysts were compared for the HDS reactions in a model diesel.

The 60%Co/Al₂O₃ catalyst was prepared by the co⁃precipitation method^{［24，26］}. In order to promote the reduction of cobalt，a resorcinol formaldehyde resin gel （RFG） was added during the precipitation process，which would be converted into carbon （with a theoretical content of 1wt%） during the reduction process. The RFG was prepared as reported previously^［26］. Specifically，desired amounts of resorcinol and 37% aqueous formaldehyde （with the molar ratio of 1∶2） were dissolved in 400 mL distilled water and heated at 353 K for 4 h to form a gel in the presence of Na₂CO₃ as a catalyst. Desired amounts of Co（NO₃）₂ and Al（NO₃）₃ were dissolved in 100 mL distilled water to form a solution （Ⅰ）. Another solution （Ⅱ） was obtained by dissolving a desired amount of Na₂CO₃ in 100 mL distilled water. Then，the two solutions （Ⅰ and Ⅱ） were dropwise added into the RFG solution at 353 K under vigorous stirring. The precipitate formed was filtered and washed thoroughly with distilled water. Then，the filter cake was dispersed into 200 mL n⁃butanol and heated at 353 K for 12 h to remove water and then n⁃butanol. The sample was further dried at 393 K overnight.

Cobalt phosphides were prepared in the same way as reported previously^［18-22］. Typically，the Co/Al₂O₃ was loaded into a micro⁃reactor and sandwiched with quartz sands，followed by the reduction in flowing H₂ （101.3 kPa，40 mL·min^-1） at 773 K for 2 h. Then，the sample was cooled down to 443~593 K，at which a heptane solution containing 2% PPh₃ was pumped into the reactor with liquid hourly space velocity （LHSV） of 2 h^-1 and H₂/oil of 300 （v∶v） for 18~36 h. Finally，the sample was treated in flowing H₂ at 673 K for 3 h.

The sulfidation of the Co/Al₂O₃ was carried out using CS₂ in liquid phase^［24］. The pre⁃reduced Co/Al₂O₃ （in H₂ at 773 K for 2 h） was sulfided by 2% CS₂ in heptane first at 503 K for 2 h and then at 593 K for 4 h，with LHSV of 2 h^-1 and H₂/oil of 300 （v∶v）.

The catalysts for characterizations were phosphided or sulfided same way as described above. Then，they were unloaded from the reactor when they were soaked in heptane，which was slowly evaporated at room temperature for 24 h，during which the samples were slowly oxidized by air （passivated）. The passivated samples were characterized with different methods.

X⁃ray diffraction （XRD） patterns were collected on a Shimadzu XRD⁃6000 powder diffractometer （Japan） using a Cu Kα radiation. The 2θ scans covered the range of 10°~80° with a scan speed of 6°·min^-1.

N₂ adsorption⁃desorption measurements were carried out at 77.3 K using a Micromeritics Tristar Ⅱ 3020 autosorption analyzer. The specific surface areas were calculated according to the Brunauer⁃Emmett⁃Teller （BET） equation. Samples were degassed in flowing N₂ at 573 K for 3 h before the measurements.

The chemical compositions of catalysts were analyzed by an ARL⁃9800 X⁃ray fluorescence spectrometer （XRF）.

The adsorption of O₂ was carried out in a home⁃made volumetric apparatus. The catalysts were reduced in H₂ at 773 K for 2 h and evacuated for 1 h before the measurements. The adsorption of O₂ was performed at 673 K. The uptake of O₂ was obtained by extrapolating the coverage of corresponding isotherms to P=0. The degree of reduction （reducibility） was calculated based on the amount of O₂ adsorbed and the loading of cobalt （assuming the molar ratio of O∶Co=4∶3）^［26-28］.

The microcalorimetric adsorption of CO was performed at room temperature （308 K） on a Tian⁃Calvet heat⁃flux microcalorimeter （C⁃80 from Setaram） connected to a gas⁃handling system. A Baratron capacitance manometer was equipped for precise pressure measurements. The passivated samples were re⁃reduced in flowing H₂ at 673 K for 2 h and then evacuated at 673 K for 1 h prior to the adsorption measurements.

The hydrotreating reactions of model diesel containing 1.72% DBT （3000 ppm S），0.185% quinoline （200 ppm N），5% tetralin and 0.5% n⁃octane （as an internal standard） in balanced n⁃tetradecane （solvent） were carried out in a fix⁃bed reactor with an inner diameter of 10 mm at 3.1 MPa with weight hourly space velocity （WHSV） of 2 h^-1 and H₂/oil ratio of 1500 （v∶v）. The model diesel was fed into the reactor using a 2ZB⁃1L10 high pressure syringe pump and flowed down through the catalyst bed. For each test，about 1 g catalyst （20~40 mesh） diluted with quartz sands was loaded and sandwiched by two layers of quartz sands （20~40 mesh），and the bed length was about 40 mm. The catalytic activity was measured at 533~633 K. The products were collected after 24 h when the activity was stabilized and analyzed by using GC. Two SE⁃30 capillary columns were connected to a flame ionization detector （FID） and a flame photometric detector （FPD） for the analysis of hydrocarbons and DBT，respectively，while a SE⁃54 capillary column was connected to a nitrogen⁃phosphorus detector （NPD） for the analysis of quinoline.

Table 1 summarizes the textural and structural properties of the catalysts Co/Al₂O₃，Co₂P/Al₂O₃，CoP/Al₂O₃ and Co₉S₈/Al₂O₃. The phases and particle sizes in the catalysts were derived by XRD （shown shortly below）. The reducibility of cobalt was estimated according to the uptake of O₂ for the reduced catalyst. The molar ratio of Co∶O was assumed to be 3∶4 since only Co₃O₄ was detected by XRD for the sample after the O₂ titration. The uptake of O₂ for the reduced Co/Al₂O₃ catalyst was about 4030 μmol·g^-1，and thus the reduction degree of cobalt in the Co/Al₂O₃ could be estimated to be 65.7%.

According to the broadening of Co （111） peak at 44.2° and Scherrer equation，the average crystallite size （D_XRD） of Co formed in the reduced Co/Al₂O₃ was estimated to be about 3 nm. Then the dispersion （D%） of metallic Co could be calculated to be about 32% using the equation： D%=6*14.6*M_Co/（ρ*D_XRD*NA），in which M_Co，ρ and 14.6 are the molar weight （g·mol^-1），metal density （g·cm^-3） and surface atomic density （atoms·nm^-2） of metallic cobalt，respectively，while the NA is the Avogadro constant^［4，29］. When the metallic Co was sulfided or phosphided into Co₉S₈，Co₂P and CoP，the crystallite sizes increased to 5.1，8.1 and 8.7 nm，respectively.

Table 1 also gives the information about the specific surface areas and pore parameters. The specific surface area and pore size changed remarkably after the phosphidation and sulfidation of Co/Al₂O₃，while the pore volumes remained almost unchanged. The specific surface area decreased from 161 m²·g^-1 for the reduced Co/Al₂O₃ to 136，113，and 120 m²·g^-1 for the phosphided Co₂P/Al₂O₃ （at 468 K for 18 h），CoP/Al₂O₃ （at 593 K for 36 h） and sulfided Co₉S₈/Al₂O₃，respectively. In contrast，the pore sizes of phosphided and sulfided catalysts were increased （to 15~17 nm） as compared with the reduced catalyst （9 nm）.

The adsorption of CO is usually used to titrate the surface metal atoms and to probe the number of the active sites for the transition metal phosphide catalysts^［1，7］. In particular，the coverage and heat for the adsorption of CO could be obtained simultaneously by the microcalorimetric adsorption measurement^［18-22］. The results were shown in Fig.1. The coverages and initial heats for the CO adsorptions on the catalysts Co₂P/Al₂O₃，CoP/Al₂O₃ and Co₉S₈/Al₂O₃ were listed in Table 1. It is seen that the initial heats were measured to be 103，92 and 107 kJ·mol^-1 with the saturation coverage of 69，76 and 169 μmol·g^-1 on the Co₂P/Al₂O₃，CoP/Al₂O₃ and Co₉S₈/Al₂O₃，respectively. The uptakes of CO on Co₂P/Al₂O₃ and CoP/Al₂O₃ were significantly higher than those on Co₂P/SiO₂ （42 μmol·g^-1）^［30］ and CoP/SiO₂ （16 μmol·g^-1）^［7，14］ reported in literature，indicating that the Co₂P/Al₂O₃ and CoP/Al₂O₃ catalysts derived from the liquid phase phosphidation of highly dispersed Co/Al₂O₃ precursor possessed more surface active sites. The comparable uptakes of CO on the Co₂P/Al₂O₃ and CoP/Al₂O₃ indicated their similar amounts of surface active sites that might be due to the similar average particle sizes and Co contents^［19］. On the other hand，the initial heat for the adsorption of CO on the CoP/Al₂O₃ was found slightly lower than that on the Co₂P/Al₂O₃，indicating that the higher P content decreased the heat of CO adsorption on Co^{［14，18］}. Fig.1 also showed that the CO uptake on the Co₉S₈/Al₂O₃ （169 μmol·g^-1） was much higher than those on the Co₂P/Al₂O₃ and CoP/Al₂O₃，indicating that the Co₉S₈/Al₂O₃ possessed more surface active sites，probably due to the significantly higher dispersion of Co₉S₈ in Co₉S₈/Al₂O₃. As shown in Table 1，the particle size of Co₉S₈ （5.1 nm） were much smaller than those of CoP （8.7 nm） and Co₂P （8.1 nm） in the CoP/Al₂O₃ and Co₂P/Al₂O₃，although they were prepared from the same Co/Al₂O₃ precursor，but with different treatments.

Fig.2 shows the XRD patterns for the samples after the phosphidation of reduced Co/Al₂O₃ by PPh₃ at different temperatures （443~593 K） for 36 h. The weak and broad diffraction peaks around 2θ=44.2°，51.5° and 75.9° characteristic of metallic Co （PDF 15⁃0806） were clearly observed on the reduced Co/Al₂O₃，indicating that metallic Co was formed and highly dispersed. The weak diffraction peaks at 2θ=36.8°，44.8°，65.2° and 77.3° for Co₃O₄ （PDF 43⁃1003） detected for the reduced Co/Al₂O₃ and the sample phosphided at 443 K might be due to the oxidation of surface metallic Co during the passivation.

The results in Fig.2 showed that the phosphidation temperature affected significantly the phases of Co_xP/Al₂O₃ formed. When the Co/Al₂O₃ was phosphided at 443 K，most metallic Co remained，while the diffraction peaks at 2θ=40.7°，43.3° attributed to Co₂P （PDF 32⁃0306） were also observed，indicating that part of metallic Co was phosphided to Co₂P at this temperature. When the Co/Al₂O₃ was phosphided at 493 K，the diffraction peaks of metallic Co disappeared，and those of Co₂P and CoP （PDF 29⁃0497） were clearly seen. When the phosphidation temperature increased to 593 K，only the phase of CoP was observed in the sample.

The phosphidation time also affected the phases of Co_xP/Al₂O₃. Fig.3 shows the XRD patterns for the samples after the phosphidation of Co/Al₂O₃ at 468 and 493 K for different time. When the sample was phosphided at 468 K for 18 h，diffraction peaks of Co₂P were observed clearly. The presence of weak peaks for Co₃O₄ might be due to the oxidation of Co during the passivation. When the phosphidation time prolonged to 24 h at 468 K，CoP appeared in addition to Co₂P. Thus，it was difficult to obtain a pure Co₂P phase at 468 K. However，it was possible to obtain a catalyst with Co₂P as the dominant phase when phosphided at 468 K for 18 h.

When the sample was phosphided at 493 K for 24 h，CoP and Co₂P were formed. Even if a longer phosphidation time （36 h） was used at this temperature，both CoP and Co₂P phases remained.

The Co₂P/Al₂O₃ and CoP/Al₂O₃ catalysts obtained from the phosphidation of reduced Co/Al₂O₃ by PPh₃ at 468 K for 18 h and 593 K for 36 h，respectively，were tested for the hydrotreating reactions. The reduced Co/Al₂O₃ itself and its sulfided sample （Co₉S₈/Al₂O₃） were also tested in the same way for comparison.

Fig.4 shows the XRD patterns for these catalysts before and after the use for the hydrotreating reactions. Strong diffraction peaks around 2θ=29.8°，31.1°，36.2°，44.8°，47.6° and 51.9° for Co₉S₈ （PDF 02⁃1459） were clearly seen （without any other phases） in the used Co/Al₂O₃ after the hydrotreating reactions，indicating that metallic Co could be sulfided into Co₉S₈，during which the crystallite size was increased from 3 nm for metallic Co to 9 nm for Co₉S₈.

When the reduced Co/Al₂O₃ was sulfided directly by CS₂，metallic Co was completely transformed into Co₉S₈ （5.1 nm）. After the use for the HDS reactions，the Co₉S₈/Al₂O₃ still exhibited the small size of Co₉S₈ particles （5.0 nm），indicating the high stability of Co₉S₈ for the HDS reactions.

The phases CoP and Co₉S₈ were formed after the Co₂P/Al₂O₃ was used for the HDS reactions. The formation of CoP indicated the disproportionation of Co₂P into Co and CoP，and the formed Co might be then converted into Co₉S₈. According to the broadening of Co₂P peak at 40.7° and Scherrer equation，the average sizes of Co₂P particles in the fresh and used Co₂P/Al₂O₃ catalysts were estimated to be about 8.1 and 10.5 nm （Table 2），respectively，indicating the growth of Co₂P particles during the HDS reactions.

In contrast，the CoP phase did not change after the HDS reactions，indicating the high stability of this phase. However，according to the broadening of CoP peak at 48.1° and Scherrer equation，the average sizes of CoP particles in the fresh and used CoP/Al₂O₃ were estimated to be about 8.7 and 11.2 nm （Table 2），respectively，indicating the growth of CoP particles during the HDS reactions.

Table 2 summarizes the chemical compositions of the fresh and used catalysts. The used Co/Al₂O₃ contained about 13.1% S due to the formation of Co₉S₈ during the HDS of DBT. The fresh and used Co₉S₈/Al₂O₃ had the S/Co atomic ratios of 0.86 and 0.89，respectively，close to the stoichiometric ratio of 0.89 in Co₉S₈，indicating again the high stability of Co₉S₈ for the HDS reactions.

The fresh CoP/Al₂O₃ exhibited the P/Co atomic ratio of 1.11，close to the stoichiometric ratio of 1 for CoP. However，the fresh Co₂P/Al₂O₃ possessed the P/Co atomic ratio of 0.37，lower than the stoichiometric ratio of 0.5 for Co₂P，probably owing to the presence of some un⁃phosphided cobalt species （Fig.3）. After the hydrotreating reactions，the P/Co atomic ratios in the used CoP/Al₂O₃ and Co₂P/Al₂O₃ changed little. However，S was detected by XRF in the used CoP/Al₂O₃ and Co₂P/Al₂O₃. Specifically，the used CoP/Al₂O₃ and Co₂P/Al₂O₃ contained about 4.1% and 6.8% S，indicating that CoP was significantly more resistant to S poisoning than Co₂P，which was also observed by Bussell et al^［11］. In fact，Co₉S₈ was detected by XRD in the used Co₂P/Al₂O₃，while no any other phases were observed in the used CoP/Al₂O₃.

Fig.5 shows the TEM images of fresh and used catalysts. It can be seen that metallic Co particles were well dispersed in the reduced Co/Al₂O₃. The statistically averaged particle size of Co in the fresh （Fig.5a） and that of Co₉S₈ in the used Co/Al₂O₃ （Fig.5a'） were about 2.5 and 10 nm，respectively，consistent with those estimated by the Scherrer equation （3 and 9 nm，respectively）.

The TEM images of fresh and used Co₉S₈/Al₂O₃ catalysts were shown in Fig.5b and 5b'. The averaged particle sizes of Co₉S₈ in the fresh and used catalysts were estimated to be about 5.5 and 5.0 nm，respectively，also in agreement with the results derived by XRD.

The TEM images of fresh and used Co₂P/Al₂O₃ are shown in Fig.5c and Fig.5c'. The average particle sizes in the fresh and used catalysts were estimated to be about 8.8 and 11 nm，respectively，close to those （8.1 and 10.5 nm） estimated by XRD，though some fresh Co₂P were converted into other phases after the use. The TEM images （Fig.5d and Fig.5d'） also showed that the CoP particles in the CoP/Al₂O₃ grew during the HDS reactions，as already evidenced by XRD.

Co usually acts as a promoter in Co⁃Mo⁃S catalysts for the hydrotreating reactions in industry^［2，31］. However，Co itself was also active for the HDS reactions of thiophene and DBT^［32-36］. In an earlier work^［37］，we prepared a Co⁃O⁃Si complex material with a high specific surface area （560 m²·g^-1） at a high Co loading （~34wt%）. After sulfidation，the material exhibited a high activity in HDS of thiophene （the conversion of thiophene reached 99.4% at 573 K and 1.5 MPa）. In this work，the highly dispersed Co/Al₂O₃ was sulfided by CS₂ to obtain a cobalt sulfide catalyst and compared with metallic Co and cobalt phosphide catalysts for the HDS of DBT.

Fig.6a shows the catalytic activities of the catalysts for the HDS of DBT. The conversion of DBT increased with temperature （533~633 K）. At 573 K，the conversion of DBT reached 40% and 65%，respectively，over the Co₂P/Al₂O₃ and CoP/Al₂O₃. Apparently，the CoP/Al₂O₃ was significantly more active than the Co₂P/Al₂O₃ for the HDS of DBT. At 633 K，the conversion of DBT reached 99% and 100%，respectively，over the Co₂P/Al₂O₃ and CoP/Al₂O₃.

The Co/Al₂O₃ and Co₉S₈/Al₂O₃ exhibited the similar activities to the Co₂P/Al₂O₃ at 533~573 K for the HDS of DBT. However，at 593 K，the conversion of DBT increased to 48%，60% and 66%，respectively，over the Co/Al₂O₃，Co₉S₈/Al₂O₃ and Co₂P/Al₂O₃. Thus，the activities of these catalysts for the HDS of DBT followed the order of CoP/Al₂O₃>Co₂P/Al₂O₃>Co₉S₈/Al₂O₃>Co/Al₂O₃.

The HDS of DBT usually undergoes through two pathways. One is the direct desulfurization pathway （DDS） with the formation of biphenyl （BP） as the desulfurization product，while another is the indirect hydrodesulfurization pathway with the formation of cyclohexylbenzene （CHB） as the desulfurization product （the desulfurization occurs only after an aromatic ring in DBT is hydrogenated）. Fig.6b shows the selectivity to CHB at different temperatures. Apparently，the selectivities to CHB on the CoP/Al₂O₃ and Co₂P/Al₂O₃ were always significantly higher than those on the Co/Al₂O₃ and Co₉S₈/Al₂O₃. For example，at 613 K，the selectivities to CHB on CoP/Al₂O₃ and Co₂P/Al₂O₃ were 79% and 78%，while those on Co/Al₂O₃ and Co₉S₈/Al₂O₃ were 50% and 46%，respectively. Since the selectivity to CHB （>70%） was always higher than that to BP on the CoP/Al₂O₃ and Co₂P/Al₂O₃，the CoP/Al₂O₃ and Co₂P/Al₂O₃ seemed more active for the hydrogenation of aromatic rings in DBT.

Fig.7a shows the results for the conversion of quinoline at different temperatures. The conversion of quinoline reached 100% on the four catalysts at 533~633 K. Propylbenzene （PB） and propylcyclohexane （PCH） were the products of HDN of quinoline. The former is the product of hydrodenitrogenation without the hydrogenation of another aromatic ring. The latter involves not only the denitrogenation，but also the hydrogenation of another aromatic ring in quinoline. Fig.7b shows the selectivity to PCH at different temperatures. With the increase of temperature，the selectivity to PCH increased on all the four catalysts. At 533 K，the selectivity to PCH on different catalysts followed the order of CoP/Al₂O₃ （84%） > Co₂P/Al₂O₃ （71%） > Co₉S₈/Al₂O₃ （62%） > Co/Al₂O₃ （52%）. Thus，the Co₂P and CoP seemed more active than Co and Co₉S₈ for the hydrogenation of aromatic rings in quinoline at the low reaction temperatures，consistent with the results for the HDS of DBT.

Aromatic hydrocarbons with multi⁃rings have low cetane numbers，and their contents must be reduced to enhance the quality of diesel fuels. In this work，the hydrogenation of tetralin was used to probe the activities of catalysts for the hydrogenation of hydrocarbons with multiple aromatic rings. It can be seen from Fig.8a that the conversions of tetralin were low on all the catalysts. However，the conversions of tetralin were higher on the Co₂P/Al₂O₃ and CoP/Al₂O₃ than on the Co/Al₂O₃ and Co₉S₈/Al₂O₃. For example，the conversions of tetralin on the CoP/Al₂O₃ and Co₂P/Al₂O₃ were 4.4% and 4.6%，while those on the Co/Al₂O₃ and Co₉S₈/Al₂O₃ were only 1.1% and 1.7% at 633 K，respectively.

Tetralin might be hydrogenated to decalin or dehydrogenated to naphthalene. Fig.8b shows the selectivity to naphthalene over the catalysts at different temperatures. With the increase of temperature，the selectivity to naphthalene increased over all the catalysts. However，the selectivities to naphthalene （dehydrogenation） on the Co/Al₂O₃ and Co₉S₈/Al₂O₃ were significantly higher than those on the Co₂P/Al₂O₃ and CoP/Al₂O₃，indicating that Co₂P and CoP were more active than Co and Co₉S₈ for the hydrogenation reactions.

A 60%Co/Al₂O₃ catalyst was prepared and reduced. Metallic Co formed was highly dispersed with small Co particles （~3 nm），which could be phosphided to Co₂P and CoP by using PPh₃ in liquid phase.

The reduced Co/Al₂O₃ could be phosphided into Co₂P and CoP by monitoring the phosphidation temperature and time. Specifically，metallic Co was phosphided mainly into Co₂P at 468 K for 18 h，and purely into CoP at 593 K for 36 h. For comparison，metallic Co was also sulfided into Co₉S₈ by CS₂ at 583 K for 4 h.

The CoP/Al₂O₃ exhibited the significantly higher activity than all the other catalysts （freshly Co₂P/Al₂O₃，Co₉S₈/Al₂O₃ and Co/Al₂O₃） studied in this work for the HDS of DBT. In addition，the phase of CoP was stable and it did not change during the HDS reactions.

The phase of Co₂P was not stable since a large amount of Co₂P changed into CoP and Co₉S₈ during the HDS reactions，leading to the decreased activity of this catalyst for the HDS reactions.

Metallic Co in the reduced Co/Al₂O₃ （~3 nm） was completely converted into Co₉S₈ with significantly increased particle size （~9 nm） during the HDS reaction. In contrast，the phase of Co₉S₈ and its particles （~5 nm） in the pre⁃sulfided Co₉S₈/Al₂O₃ were highly stable and quite active for the HDS reactions.