However, they can break both C–N bonds and C–C bonds. This so-called hydrogenolysis property of metals leads in many instances to substantial cracking of the product molecules to smaller (gaseous) molecules. In the petroleum industry, this would result in a lower yield of liquid product, and consequently metal catalysts are not used in HDN. A second reason for not using metal catalysts is that HDN in isolation, without HDS, is not possible because oil fractions always contain sulfur as well as nitrogen. The relatively high concentrations of H2S that are produced during normal hydrotreating would directly lead to sulfidation of any metal catalyst.
Recently, however, schemes for a hydrotreating process in which metals may play a role have come under consideration. New legislation will require very low levels of sulfur in gasoline and diesel fuel (<50 ppm) by the Year 2005. This goal might be attained by means of a two-stage hydrotreating process (2). In this process configuration, the majority of the original sulfur is removed with a classic metal sulfide catalyst in a first-stage reactor. The product is sent to a second-stage hydrotreating reactor in which the final sul- fur specification of 50 ppm is reached. The relatively small amount of sulfur present in the product of the first reactor (e.g., 200 ppm) is very difficult to remove because this sulfur is present in highly unreactive molecules such as 4,6-dimethyl dibenzothiophene. This calls for highly active HDS catalysts, which are difficult to find among the metal sulfides. Because the concentra- tion of H2S in the second reactor is low, metal catalysts that are not overly sensitive to sulfur can be used, such as Pt–Pd catalysts on acidic supports (172). Other metals, or metal carbides or metal nitrides, may also play a role in future two-stage hydrotreating processes.
Catalysis on metal carbides and nitrides is a new branch of catalysis that started in the 1980s. Volpe and Boudart (173) discovered that it is easy to prepare Mo2N and W2N samples with high surface areas by nitriding MoO3 and WO3, respectively, with ammonia. Lee et al. (174) synthesized high- surface-area Mo2C by carbiding MoO3with methane. Subsequently, it was
discovered that the same methods also lead to supported carbides and ni- trides. The resulting materials were tested for several catalytic reactions, such as hydrocarbon transformations (isomerization and hydrogenolysis), HDS, and HDN. Since the early 1990s, the number of investigations carried out in this field has increased dramatically, and many papers deal not only with their preparation and characterization but also with their catalytic proper- ties. This section is focused on investigations that describe these carbides and nitrides as potential HDN catalysts. Results of related studies ( for HDS and hydrogenolysis) are mentioned only when necessary to aid understanding.
A. STRUCTURE OF THECATALYST
Metal carbides and metal nitrides are basically metals with carbon and nitrogen atoms, respectively, that are incorporated at interstitial positions.
Several structures are possible (175). Complete filling of all octahedral sites in an fcc metal lattice gives an NaCl-type structure common to transition metal monocarbides such as VC and NbC. Random filling of half the octa- hedral sites results in a subcarbide structure such as Mo2C and W2C. Orderly filling of half of the trigonal-prismatic sites in the hexagonal metal lattice gives the WC structure. Bimetallic carbides have more complex structures.
Oxycarbides have also been described (176). They are of interest for catalysis because oxophilic metals may not form carbides easily; in prepa- rations from metal oxides by treatment with methane, oxycarbides rather than carbides may form. Furthermore, carbides and nitrides are usually pas- sivated after preparation; they are slowly exposed to oxygen or air to form a thin protective oxide layer on the metal carbide surface. In this passivated form, these materials are not pyrophoric and can be transported through the air. After insertion into a catalytic reactor, they must be reactivated either by methane or ammonia treatment, as used in their preparation, or by re- duction with hydrogen. In these cases, pure carbide or nitride is not always reformed at the surface; an oxycarbide or oxynitride is often formed instead.
B. MECHANISMS
Pure metal carbides and nitrides have metallic properties. The interstitial C or N atoms increase the distance between the metal atoms somewhat, leading to a decrease in the width of the metaldband and a weaker metal character.
Thus, metal carbides formed from metals in the middle of the periodic table, such as molybdenum and tungsten, behave more like metals on the right side of the table, such as platinum. Consequently, the hydrogenation and hydrogenolysis properties of these metal carbides, although still present, are not as strong. Like many pure metals, carbides and nitrides of molybdenum and tungsten are still active for hydrogenolysis (177), which will prevent
CATALYTIC HYDRODENITROGENATION 453 applications in classic HDN because of the loss of the liquid product to gaseous molecules.
Product loss through hydrogenolysis is a well-known problem. In catalytic reforming, hydrocarbons must be transformed into other hydrocarbons with higher octane numbers. Metals can do this, but even a metal with an almost filleddband, such as platinum, still catalyzes unwanted hydrogenolysis. The solution to this problem in catalytic reforming was to temper the hydrogenol- ysis properties of platinum by adsorbing sulfur atoms on the metal or by alloying platinum with an inactive metal such as tin (178). The remaining uncovered platinum atoms in the surface are isolated, and no clusters of neighboring platinum atoms are available to break the C–C bonds. At the same time, the hydrocarbon isomerization capability of platinum is sup- pressed, however, and another catalytic function has to be added to bring back isomerization capability. This was realized by using an acidic support that can isomerize hydrocarbons but is not acidic enough to catalyze crack- ing. In industry, Al2O3acidified with chlorine or fluorine is often used (179).
The problem related to the hydrogenolysis of metal carbides and nitrides in hydrotreating can be solved in a similar way by isolating the metal atoms by sulfur adsorption on the surface or by alloying with inactive metals.
Oil fractions contain substantial amounts of sulfur, and the problem is to prevent bulk sulfidation. Therefore, a feed that contains only a relatively small amount of sulfur, such as the product from a first-stage hydrotreating reactor, will be more appropriate for the application of metal carbides or nitrides in hydrotreating. Sulfur adsorption or alloying will suppress not only C–C bond breaking but also C–N bond breaking. Therefore, as in catalytic reforming, a new catalytic function must be built in to enable breaking of the C–N bond and removal of nitrogen. Acidity of the support or acidity on the metal carbide or nitride may be the solution. Oxycarbides and oxy- nitrides contain oxygen atoms in the lattice which may form hydroxyl groups at the surface. These hydroxyl groups may have Brứnsted acid properties, or a metal atom at the surface next to an oxygen atom may have Lewis acid properties. Consequently, oxycarbides and oxynitrides have metallic as well as acidic properties, and they are bifunctional (180, 181). In this respect, they are like metal oxides with metals in low oxidation states and metal sulfides and can catalyze the removal of nitrogen from aromatic heterocycles.
C. HDNOFSPECIFICCOMPOUNDS
Interest in the use of metal carbides and nitrides in hydrotreating was aroused by a publication by Schlatter et al.(182) in 1988. They reported that Mo2C and Mo2N (unsupported as well as supported on Al2O3) were highly active for HDN of quinoline in autoclaves under industrially realistic
conditions (e.g., 6.9 MPa and 673 K) (182). The activities that were attained in this exploratory study were similar to the activity of a commercial sul- fided Ni–Mo/Al2O3catalyst. When tested in the absence of sulfur, Mo2C in particular showed a much higher selectivity for aromatic hydrocarbons than sulfided Ni–Mo/Al2O3. This result shows that Mo2C and Mo2N can remove nitrogen from propylaniline directly by hydrogenolysis of the C–N bond, without having to hydrogenate the phenyl ring. When sulfur was added to the feed (as CS2), however, the PB to PCH ratio was much lower and more representative of a sulfided Mo/Al2O3catalyst. It was suggested that a sur- face layer of MoS2forms on the Mo2C under such conditions since the metal sulfide is thermodynamically preferred. In that case, the HDN mechanism is similar to that discussed in Section II.B for metal sulfide catalysts. This inference means that hydrogenolysis of the aniline C–N bond is relatively insignificant and that HDN occurs mainly via hydrogenation of the phenyl ring of the aniline. As a result, the aromatic content of the product is lower than when the molybdenums carbide or nitride are used in the absence of sulfur.
Subsequent work by Leeet al.(183) and Abeet al.(184) with Mo2N con- firmed the results of Schlatteret al.They obtained much higher selectivities to aromatic products in the absence of sulfur because their lower pressure (0.1 MPa) and higher temperature (723 K) favored the formation of aro- matics. Under these conditions, they observed substantial hydrogenolysis.
Thus, benzene and toluene rather than PB were the main products of the HDN of quinoline. Stanczyket al.(185) observed a high selectivity for aro- matic products and substantial hydrogenolysis as well. They studied the HDN of 1,2,3,4-THQ on molybdenum and niobium oxynitrides at 673 K and 4.5 MPa. The ratio PB to PCH was close to 1, indicating that the reaction of propylaniline to PB was much more important on MoNxOythan on sulfided Ni–Mo/Al2O3(Section II.B.2.c). The relatively large amounts of ethylaniline methylaniline, and aniline (as well as ethylbenzene and toluene) demon- strated that hydrogenolysis of the propylaniline, or rather of the propylben- zene product, was substantial.
Several unsupported metal carbides and nitrides (of titanium, vanadium, niobium, molybdenum and tungsten) were investigated as catalysts for HDS, HDN, and hydrodeoxygenation by Ramanathan and Oyama (186). A model feed containing 3000 ppm sulfur (as dibenzothiophene), 2000 ppm nitrogen (as quinoline), 500 ppm oxygen (as benzofuran), and 20% aromatic hydro- carbon (as tetralin) was tested at 643 K and 3.1 MPa in a microflow reactor.
Mo2C, Mo2N, and WC showed HDN activities to hydrocarbons comparable to that of a commercial sulfided Ni–Mo/Al2O3catalyst. Only Mo2C showed a good HDS activity, and VN showed a high hydrodeoxygenation activity
CATALYTIC HYDRODENITROGENATION 455 under these conditions with a high nitrogen content in the feed. The VN catalyst produced mainly ethylbenzene in the conversion of benzofuran, whereas the Ni–Mo/Al2O3 catalyst yielded mostly ethylcyclohexane. Hy- drogenolysis of the C–O bond in phenol (which is weaker than the C–N bond in aniline) is probably substantial, even on surface-sulfided VN. The PCH to PB ratio resulting from the HDN was greater than 1 for all these catalysts, as it was for the sulfided Ni–Mo/Al2O3. This result suggests that the large amount of sulfur in the feed transformed the metal carbides and nitrides into sulfides, at least at the surface. No traces of sulfide phases were revealed by X-ray diffraction, however, and only small amounts of sulfur were found by X-ray photoelectron spectroscopy. Therefore, the authors concluded that the materials are sulfur tolerant.
Similar experiments, again with a mixed feed of dibenzothiophene, quino- line, benzofuran, and tetralin, were conducted with a series of bimetallic oxynitride compounds, such as V–Mo–O–N, Mo–W–O–N, Co–Mo–O–N, and Co–W–O–N (187). Several of these materials had higher HDN activi- ties than their monometallic equivalents and were comparable to a Ni–Mo/
Al2O3catalyst, but their HDS and hydrodeoxygenation activities were lower.
Whereas molybdenum and tungsten, alloyed with early transition metals such as vanadium, chromium, and niobium, were found to be sulfur resis- tant, according to X-ray diffraction data, manganese and cobalt oxynitride alloys showed bulk sulfidation to give MnS and Co4S3, respectively. Oyama et al.(188) also studied bimetallic oxycarbides such as Nb–Mo–O–C, Ni–
Mo–O–C, Mo–W–O–C, and Ni–W–O–C, which were found to have moder- ate activities similar to those of the monometallic carbides. Only Nb–Mo–
O–C had a higher activity than NbC or Mo2C. X-ray diffraction patterns measured before and after the reaction showed that whereas Mo–W–O–C and Nb–Mo–O–C were sulfur resistant during the reaction, Ni–Mo–O–C and Ni–W–O–C were transformed into sulfides.
Sajkowski and Oyama (189) reported high activities of unsupported Mo2N and Mo2C catalysts supported on Al2O3in a hydrotreating study of a coal- derived gas oil and a coal-derived residuum feed at 633 K and 13.7 MPa.
These feeds contained about 4000 ppm nitrogen, 120 or 800 ppm sulfur, and 3500 or 18,600 ppm oxygen, respectively. The authors compared the activ- ities based on sites titrated by CO on the nitride and carbide and titrated by O2on sulfided Mo/Al2O3and Ni–Mo/Al2O3catalysts. They reached the conclusion that the carbide and nitride were intrinsically five times more ac- tive than the classic catalysts. Although this result points to a high intrinsic activity of these new materials, it does not demonstrate that they are indeed intrinsically more active because currently there is no reliable method for counting the number of active sites at the surfaces of metal sulfides. Under
certain conditions, oxygen chemisorption does indeed give results that cor- relate with activity, but it has been shown that O2chemisorption on sulfides is corrosive and therefore dependent on the measurement conditions. O2
chemisorption probably overestimates the number of sites. Thus, the quoted fivefold higher intrinsic activity of Mo2N and Mo2C has to be taken as an upper limit. In any case, industry is interested in actual activities rather than intrinsic catalytic activities, based on volume rather than on weight, because existing reactors have to be filled with these catalysts. In both cases, the Mo2N and Mo2C catalysts were less active per unit of reactor volume than the classic Ni–Mo/Al2O3catalyst.
In the same study (189), Sajkowski and Oyama reported that X-ray diffrac- tion did not show any crystalline MoS2, but that X-ray photoelectron spec- troscopy showed a sulfur-to-molybdenum ratio of 0.26 at the surface. They concluded that the formation of MoS2at the surface of the nitride and car- bide crystallites could not be ruled out. The question of surface sulfidation was addressed by van Veenet al.(190) and Aegerteret al.(191). The TEM results of van Veenet al.showed that the surface of an unsupported Mo2N sample was transformed into molybdenum sulfide (Fig. 13), whereas the X- ray data showed that the bulk structure of molybdenum carbide (or nitride) was retained. The same X-ray result was obtained by Aegerteret al.(191).
Their infrared spectra of adsorbed CO, however, indicated that the surfaces of the Mo2C and Mo2N catalysts had become sulfided. Furthermore, the HDS product selectivities did not differ from those of a sulfided Mo/Al2O3 catalyst. The Mo2C and Mo2N particles apparently act as rigid supports on which a thin layer of sulfided molybdenum forms under HDS conditions.
Indole conversion was investigated at atmospheric pressure by Liet al.
(192) and at 9 MPa in the presence of 70 ppm of H2S by Migaet al.(181).
In the latter study, more of the fully hydrogenated ethylcyclohexane than ethylbenzene was produced in the presence of bulk molybdenum oxynitride, suggesting that MoS2was the active component. The HDN of pyridine was investigated at atmospheric pressure with a Mo2C catalyst (193) and with several metal nitrides (194): The results show substantial hydrogenolysis of the pentane product. At 3 MPa and 573 K, pyridine was almost totally denitrogenated on nitrided Ni–Mo/Al2O3catalysts in the absence of sulfur (195). As expected for a metal or metal nitride catalyst, some hydrogenol- ysis was observed. The Ni–MoN/Al2O3catalysts not only gave higher con- versions of pyridine than a sulfided Ni–Mo/Al2O3 catalyst but also were much more active for denitrogenation. Metallic nickel, but not Ni3N, was observed by X-ray diffraction, in line with thermodynamics. A Ni3Mo3N phase was probably responsible for the high activity and the promotion effect of nickel.
CATALYTIC HYDRODENITROGENATION 457
FIG. 13. Micrograph of a spent Mo2N catalyst with a surface layer of MoS2(190).