Structure of the Active Sites of Co− Mo Hydrodesulfurization Catalysts

288

J. Phys. Chem. B 2005, 109, 288-296

Structure of the Active Sites of Co-Mo Hydrodesulfurization Catalysts as Studied by Magnetic Susceptibility Measurement and NO Adsorption Yasuaki Okamoto,* Masatoshi Kawano, Takao Kawabata, Takeshi Kubota, and Ichiro Hiromitsu Department of Material Science, Shimane UniVersity, Matsue 690-8504, Japan ReceiVed: August 23, 2004; In Final Form: October 13, 2004

The elucidation of a molecular structure of the active sites (i.e., the Co-Mo-S phase) of Co-Mo hydrodesulfurization catalysts has received extensive attention. In the present study, we unambiguously determined, for the first time, the NO adsorption behavior and magnetic property of the Co-Mo-S phase by preparing unique Co-Mo/Al2O3 catalysts (CVD-Co/MoS2/Al2O3), in which all the Co atoms are present as the Co-Mo-S phase. The catalysts were characterized by NO adsorption (pulse technique and FTIR), Co K-edge XANES, and the magnetic susceptibility and effective magnetic moment of Co. Nitric oxide molecules were adsorbed on 33% of the Co atoms in CVD-Co/MoS2/Al2O3 after sulfidation and on only half of the Co atoms even after an H2-treatment of the sulfided catalyst at 573-673 K. The Co atoms in CVD-Co/MoS2/ Al2O3 exclusively exhibited an antiferromagnetic property, indicating that even-numbered Co atoms are interacting with each other in the Co-Mo-S phase. A Co-Mo/Al2O3 catalyst, prepared by a conventional impregnation technique, was composed of the antiferromagnetic Co sulfide species as observed in CVDCo/MoS2/Al2O3 in addition to Co9S8. On the basis of the NO adsorption behavior and magnetic property, it is empirically proposed that the structure of the Co-Mo-S phase is represented as a Co sulfide dinuclear cluster located on the edge of MoS2 particles. The magnetic property of Co/Al2O3 sulfide catalysts depended on the preparation method.

Introduction Hydrodesulfurization (HDS) catalysts have received increasing attention because of more severe legislation toward cleaner fuels.1-3 Supported Mo or W sulfides promoted by Co or Ni have been the main catalytically active components in industrial catalysts.4-7 Fundamental studies have been extensively conducted to elucidate the active phase in HDS catalysts.4,8,9 On the basis of a variety of physicochemical characterizations, such as 57Co Mo¨ssbauer emission spectroscopy (MES), FT-IR of NO adsorption, and XPS, Topsøe and co-workers4,10-12 proposed the so-called Co-Mo-S phase (structure), in which Co atoms are located on the edge of finely dispersed MoS2 particles, as active sites of Co-Mo HDS catalysts. Many catalytic and spectroscopic aspects of Co(Ni)-Mo(W) sulfide catalysts have been successfully interpreted in terms of the Co-Mo-S phase.4 The local structure of the Co-Mo-S phase has been investigated by means of XAFS techniques. Bouwens et al.13 and Niemann et al.14 suggested that the Co(Ni) atoms in the Co(Ni)-Mo-S phase are separated (no specific interactions between Co atoms) and are in a square pyramidal or octahedral configuration. Louwers and Prins15 proposed a similar structure for the Ni-Mo-S phase for carbon-supported Ni-Mo sulfide catalysts, although they found a considerable Ni-Ni contribution in their EXAFS due to adjacent Ni atoms in a square pyramidal configuration. Thus, no specific interactions between the Co or Ni atoms were assumed in the EXAFS analysis of the Co(Ni)Mo-S structure favorably accepted for more than a decade. From recent STM (scanning tunneling microscopy) studies of MoS2 and Co-MoS2 clusters fabricated on Au (111), * Corresponding author. Tel/Fax: [email protected].

+81-852-32-6466.

E-mail:

Besenbacher et al.16-18 proposed the location of Co atoms at the (1h010) surface (S-edge) of MoS2 in conjunction with the theoretical (DFT) calculations of the Co-Mo-S phase by Byskov et al.19 The STM images suggest that the Co atoms on the MoS2-edge form a one-dimensional array with bridging sulfur atoms between the adjacent Co atoms, in contrast to the Co-Mo-S structure proposed on the basis of the EXAFS analysis. More recent DFT calculations20,21 also suggested the interactions between adjacent Co atoms on the (101h0) surface (Mo-edge) of MoS2 particles. Nevertheless, the STM observations were made for Au-supported Co-MoS2 clusters specially prepared by a sputtering technique. It is, accordingly, intriguing to experimentally elucidate the molecular structure, in particular, the interactions between Co atoms situated at the MoS2-edges in conventional oxide-supported Co-Mo sulfide catalysts, such as Co-Mo/Al2O3. The magnetic property of Co sensitively reflects the chemical state of Co atoms and their mutual interactions. The magnetic study of Co-Mo catalysts has been conducted by several groups in oxidic22,23 and sulfided24-29 states. In their magnetic study, Topsøe et al.27,28 reported that the effective magnetic moment of Co decreased as the temperature decreased for an unsupported Co-MoS2 catalyst (Co/Mo ) 0.0625) in which all the Co atoms were present as the Co-Mo-S phase according to MES measurements. They observed a relatively low effective magnetic moment 0.78 µB at 275 K, leading them to conclude that extensive electron delocalization is caused by the interactions of Co with MoS2-edges. In addition, they noted some contribution of antiferromagnetic Co species. More recently, Richardson29 measured the effective magnetic moment of Co for an unsupported Co-MoS2 catalyst (Co/Mo ) 0.075) and obtained the value of 1.44 µB. They concluded that the discrepancies from

10.1021/jp0462052 CCC: $30.25 © 2005 American Chemical Society Published on Web 12/04/2004

Active Sites of Co-Mo Hydrodesulfurization Catalysts the value reported by Topsøe et al.27,28 are not unrealistic in view of the difficulty in resolving the Co contribution for such low Co loading. The magnetic property of the Co-Mo-S phase in conventional supported Co-Mo catalysts was studied by Topsøe et al.28 with Co-Mo/Al2O3 impregnation catalysts. In the impregnation catalysts, however, the Co-Mo-S phase, Co(II) ions dissolved in Al2O3, and Co9S8 were simultaneously present; hence, the accuracy of the magnetic moment of the Co-Mo-S phase was considerably low, despite the deconvolution of these components by MES. They merely suggested the presence of the interactions between Co atoms and MoS2 particles. The chemisorption of NO, O2, or CO molecules will also provide important experimental insights into the molecular structure of the active sites on HDS catalysts.4 In particular, NO chemisorption is very informative, when combined with IR techniques, to distinguish coordinatively unsaturated (cus) Mo and Co(Ni) sites in the case of supported Co(Ni)-Mo sulfide catalysts.11,30-33 It is well-established that NO molecules are adsorbed selectively on cus Mo and Co(Ni) sites as dinitrosyl forms.31 The location of Co atoms in the Co-Mo-S phase has been first elucidated on the basis of NO adsorption11 and later supported by other techniques including XAFS.13-15 However, despite wide use of NO as a probe molecule and its practical importance, crucial problems remain unresolved with the NO adsorption on HDS catalysts, for instance, as to the oxidation/ reconstruction of the adsorption sites34,35 and as to the fraction of the Co(Ni)-Mo-S phase accessible for NO. However, no unambiguous answers have been given yet. Preparation of CoMo sulfide catalysts is highly desirable, in which all the Co atoms are present as the Co-Mo-S phase and the edge of MoS2 particles is fully occupied by the Co-Mo-S phase, to study the magnetic property and NO adsorption behavior of the CoMo-S phase. In our previous study,36-38 we have reported the preparation of Co-Mo sulfide catalysts, in which the edge of MoS2 particles is preferentially and fully covered by Co atoms forming the Co-Mo-S phase, by using Co(CO)3NO (CVD-technique) as a precursor of Co (CVD-Co/MoS2 catalysts). A selective formation of the Co-Mo-S phase in the CVD-Co/MoS2 catalysts was evidenced by the Co2p XPS, Co K-edge XANES, a proportional correlation between Co/Mo and NO/Mo mole ratios, and a linear relationship between catalytic activity for thiophene HDS and the amount of Co accommodated.36-38 In addition, it has been shown that CVD-Co/MoS2/Al2O3 shows the maximum potential HDS activity for a given MoS2/Al2O3.39 CVD-Co/MoS2 catalysts would, therefore, provide an ideal basis to study the magnetic property and NO adsorption behavior of the Co-Mo-S phase. In the present study, we measured the magnetic susceptibility of Co atoms of highly active CVD-Co/MoS2 sulfide catalysts to elucidate the structure of the Co-Mo-S phase at a molecular level. In addition, we tried to determine the fraction of the Co atoms in the Co-Mo-S phase susceptible to NO adsorption. On the basis of the magnetic property and NO adsorption behavior, we propose a novel empirical structure of the CoMo-S phase, a Co sulfide dinuclear cluster. The catalytic results of CVD-Co/MoS2/Al2O3 have been reported elsewhere.36-39 Experimental Procedures Catalyst Preparation. An alumina-supported Mo oxide catalyst, MoO3/Al2O3 (Mo content, 8.7 wt %), was prepared by an impregnation technique using (NH4)6Mo7O24‚4H2O (Kanto Chemicals, analytical grade), followed by calcination

J. Phys. Chem. B, Vol. 109, No. 1, 2005 289 at 773 K for 5 h.39 The alumina was supplied by the Catalysis Society of Japan as a Reference Catalyst (ALO-7, 180 m2 g-1). MoO3/Al2O3 (0.1 g) was sulfided in a 10% H2S/H2 flow (100 cm3 min-1) at 673 K for 1.5 h. The sulfided catalyst is denoted MoS2/Al2O3 hereinafter. Co-Mo/Al2O3 (Co content, 2.53 wt %) was prepared by impregnating an aliquot of the MoO3/Al2O3 sample with an aqueous solution of Co(NO3)2, followed by being dried at 383 K for 16 h. The sample was not calcined after the addition of Co to avoid the formation of Co(II) species incorporated in the Al2O3 phase. Co-Mo/Al2O3 was sulfided in the same manner as MoS2/Al2O3 to prepare imp-Co-MoS2/Al2O3. Co/Al2O3 (Co content, 2.53 wt %) was also prepared by an impregnation of Al2O3. The sulfided sample is denoted imp-Co/Al2O3. Cobalt was also added to MoS2/Al2O3 by a chemical vapor deposition (CVD) technique using Co(CO)3NO as a precursor of Co. The detailed preparation method has been reported previously.36,37 Briefly, MoS2/Al2O3 was evacuated at 673 K for 1 h and then exposed to a vapor of Co(CO)3NO, kept at 273 K, for 5 min at room temperature, followed by evacuation for 10 min at room temperature to remove physisorbed Co(CO)3NO molecules and subsequent resulfidation at 673 K for 1.5 h in a 10% H2S/H2 flow. The Co-Mo/Al2O3 catalyst thus prepared by the CVD technique is designated as CVD-Co/ MoS2/Al2O3 hereinafter. H2S/H2-treated Al2O3, instead of MoS2/ Al2O3, was exposed to a vapor of Co(CO)3NO to prepare CVDCo/Al2O3. The Co content of the sulfided catalyst was determined by XRF.39 Catalyst Characterization. The amount of NO adsorption on CVD-Co/MoS2/Al2O3 was measured between 201 and 323 K using a pulse technique after the sulfidation at 673 K in a 10% H2S/H2 stream. The sample was cooled to room temperature in the H2S/H2 stream and then purged with a high purity He flow at room temperature. Subsequently, the temperature of the sample was adjusted to a desired one by using appropriate refrigerants, a cooling apparatus, or an electric furnace, before periodic introduction of NO pulses (10% NO/He, 5.0 cm3). The total amount of NO adsorption was determined from the cumulated amount of NO adsorbed in each pulse. The reproducibility of the amount of NO adsorption was usually better than (5%. The amount of NO adsorption was also measured for CVDCo/MoS2/Al2O3 after an H2-treatment between 523 and 673 K, following the sulfidation at 673 K. The temperature was decreased to a predetermined temperature and then the H2S/H2 flow was replaced by an H2 flow. The sulfided catalyst was H2-treated for 30 min at the temperature and then cooled to room temperature in the flowing H2. In the case of the H2treatment at 673 K, the catalyst was subjected to a prolonged treatment, 1 or 3 h, too. The FTIR spectra of NO adsorption on a catalyst were recorded on a FT/IR620 (JASCO) using an in situ IR cell with KBr windows. A self-supporting disk of the catalyst was presulfided at 673 K for 90 min in a 10% H2S/H2 stream and transferred to the cell in a vacuum type glovebox filled with N2 gas. The disk was sulfided again at 673 K for 30 min in the cell (26 kPa of 10% H2S/H2) and evacuated at the same temperature. A total of 6.7 kPa of NO was introduced to the cell at room temperature, followed by evacuation before the spectra were recorded. When a sulfided catalyst was H2-treated, ca. 26 kPa of H2 was admitted to the cell at 673 K. The H2 gas was replaced three times by fresh H2 gas every 30 min. After evacuation at 673 K, the FTIR spectra of adsorbed NO were recorded.

290 J. Phys. Chem. B, Vol. 109, No. 1, 2005

Figure 1. Amount of NO adsorption on CVD-Co/MoS2/Al2O3 as a function of the adsorption temperature.

Okamoto et al.

Figure 2. Amount of NO adsorption on CVD-Co/MoS2/support as a function of the loading of Co anchored by the CVD technique. CVDCo/MoS2/support (support: O, Al2O3; 4, SiO2; 0, TiO2; and ], ZrO2)36 and b, CVD-Co/MoS2/Al2O3 (present result).

The Co K-edge X-ray absorption near edge structure (XANES) spectra of the catalysts and reference compounds were measured at room temperature at BL-7C of KEK-IMSS-PF using an in situ XAFS cell in a fluorescence mode,40 with 2.5 GeV ring energy and 250-290 mA stored current by using a Litle-type detector. The static magnetic susceptibility measurement of the sulfided catalyst was done in situ with a Faraday method using a Cahn 2000 Electro-Balance system at between 4.2 and 300 K.41 The catalyst sample was evacuated at 673 K for 1 h before being fused into a glass ampule. The magnitude of magnetic field was fixed at 10 000 G. The effective magnetic moment and magnetic susceptibility were obtained by subtracting the magnetic contributions of the glass ampule and MoS2/Al2O3 or Al2O3 separately measured under the identical conditions. Results Figure 1 shows the amount of NO adsorption on CVD-Co/ MoS2/Al2O3 as a function of the adsorption temperature. Although the amount of NO adsorption was slightly smaller at 201 K, it kept constant between 253 and 323 K. This suggests that NO adsorption provides reliable information on the number of cus sites or the dispersion of Co or Mo sulfides. In our previous study,36 we reported a proportional correlation between the amount of NO adsorption (mmol/g) on MoS2/ support and the Co content (wt %) of CVD-Co/MoS2/support (support; Al2O3, TiO2, SiO2, and ZrO2). This fact provides strong evidence for the location of the anchored Co atoms on the edge of MoS2 particles in the CVD-Co/MoS2 catalysts.36 Figure 2 presents the amount of NO adsorption (mmol/g) and the Co loading (mmol/g) of the present CVD-Co/MoS2/Al2O3 together with the previous data36 for CVD-Co/MoS2/support. A single proportional line was obtained including the present CVDCo/MoS2/Al2O3, confirming the selective formation of the CoMo-S phase in the present catalyst. As reported previously,38 the FTIR spectra of NO adsorption on CVD-Co/MoS2/Al2O3 showed a set of intense doublet bands at 1845 and 1785 cm-1 characteristic of NO molecules adsorbed on Co sites in a dinitrosyl form (Figure 3b,d), accompanying a weak band around 1680 cm-1 due to NO molecules on Mo sites. The FTIR spectra indicate that the amount of NO adsorption on CVD-Co/MoS2/Al2O3 (Figure 2) represents essentially the amount of NO molecules adsorbed on Co atoms. Hence, the

Figure 3. FTIR spectra of NO adsorption on (a) MoS2/Al2O3,38 (b) CVD-Co/MoS2/Al2O3,38 (c) CVD-Co/MoS2/Al2O3 treated with H2, and (d) CVD-Co/Al2O3. The previous results are also shown for comparison.

slope of the line in Figure 2 shows the fraction of Co atoms available for NO adsorption. The slope in Figure 2 shows that the NO/Co mole ratio is 0.66 ( 0.03 for the CVD-Co/MoS2 catalysts. Taking into consideration the formation of dinitrosyl species,31 (NO)2/Co, it is concluded that NO molecules are adsorbed on only 33% of the Co atoms in the Co-Mo-S phase. It might be assumed that the rest of the Co atoms are possibly coordinatively saturated by adsorbed sulfur compounds such as H2S. To remove adsorbed sulfur species and to obtain the maximum amount of Co atoms accessible for NO molecules under typical HDS reaction conditions (reaction temperature: 573-673 K in an H2 stream), CVD-Co/MoS2/Al2O3 was treated in a stream of H2 between 523 and 673 K after the sulfidation at 673 K. As shown in Figure 4, the NO/Co ratio increased from 0.66 to 1.05 ( 0.05 by the H2-treatment at 573, 623, or 673 K for 30 min and remained constant for a prolonged treatment up to 3 h at 673 K. A slightly lower NO/Co ratio of 0.90 was obtained after the treatment at 523 K (Figure 4). As

Active Sites of Co-Mo Hydrodesulfurization Catalysts

J. Phys. Chem. B, Vol. 109, No. 1, 2005 291

Figure 4. Amount of NO adsorption as a function of the H2-treatment temperature and duration. 2, 473 K; b, 523 K; 4, 623 K; and O, 673 K.

shown in Figure 3c, the FTIR spectrum of NO adsorption was not essentially changed in the band positions and shape by an H2-treatment at 673 K (26 kPa of H2, 90 min in total), suggesting that the increase of NO adsorption capacity is apparently ascribed to the removal of mobile sulfur atoms as H2S to form cus Co sites during the H2 treatment. It should be noted in Figure 4 that (NO)2/Co ratio levels off at 0.53. This ratio strongly suggests that only half of the Co atoms in the Co-Mo-S phase are subjected to NO chemisorption at the highest coordinative unsaturation attainable under typical HDS reaction conditions. Figure 5 compares the Co K-edge XANES spectra for CVDCo/MoS2/Al2O3, CVD-Co/MoS2/Al2O3 after the H2 treatment at 673 K for 30 min, and reference compounds (Co9S8 and Co metal). The XANES spectrum (Figure 5a) of CVD-Co/MoS2/ Al2O3 shows a selective formation of the Co species characteristic of the Co-Mo-S phase13-15 and the absence of bulk Co9S8 and Co(II) oxide species (7725 eV), as pointed out previously.39 The intensity of a 1s-3d preedge peak around 7708 eV for CVD-Co/MoS2/Al2O3 is much weaker than that for Co9S8 as observed by other workers,13-15,42-44 indicating that the Co atoms in the Co-Mo-S phase are in a distorted octahedral symmetry or in a square pyramidal configuration. The H2-treatment of CVD-Co/MoS2/Al2O3 at 673 K caused no change in the XANES spectrum (Figure 5b), obviously indicating that the local structure and electronic state of the CoMo-S phase are virtually unchanged even by the H2 treatment and that no metallic Co species are formed. Figure 5 also presents the spectrum of CVD-Co/MoS2/Al2O3 after full adsorption of NO by the pulse technique. Small but significant changes were caused by NO adsorption (Figure 5c): an increase in the edge energy by 0.5 eV and fine structural changes, indicating the modification of the electronic state and geometry of the Co species by NO adsorption. It is worthy of note with the XANES spectrum (Figure 5c) that no white line peak appears around 7725 eV due to oxidized Co species, clearly demonstrating no notable oxidation of Co species by NO adsorption at room temperature in the present pulse technique. In contrast to the present observations, Nielsen et al.35 reported Ni K-edge XANES spectra for Ni-Mo/Al2O3 showing a gradual oxidation of Ni sites during the contact with gaseous NO.

Figure 5. Co K-edge XANES spectra of (a) CVD-Co/MoS2/Al2O3,39 (b) CVD-Co/MoS2/Al2O3 treated in an H2 stream for 30 min at 673 K, (c) CVD-Co/MoS2/Al2O3 after NO adsorption by a pulse technique, (d) Co9S8, and (e) Co metal foil. The previous results are also shown for comparison.

Figure 6 shows the effective magnetic moment and magnetic susceptibility of Co for CVD-Co/MoS2/Al2O3 (3.18 wt % Co) as a function of temperature. The magnetic property of Co was extracted by subtracting the contribution of MoS2/Al2O3 prepared under the same sulfidation and evacuation conditions. It is considered that the effects of Co addition on the chemical state and morphology of the preexisting MoS2 particles are minimal when Co is added by the CVD technique. With decreasing temperature, the magnetic susceptibility χ of Co increased and had a maximum at 16 K, followed by a sharp decrease at a lower temperature. This temperature dependency of χ is a typical antiferromagnetic behavior, clearly demonstrating, for the first time, that there are antiferromagnetic interactions between Co atoms in the Co-Mo-S phase. It seems difficult to explain the antiferromagnetism of the Co-Mo-S phase on the basis of the structural models previously reported,13-15 in which each Co atom is surrounded by four to six sulfur atoms and is isolated from each other. Instead, we consider a model in which Co sulfide clusters are located on the edge of MoS2 particles forming a one-dimensional array. The antiferromagnetism as observed in Figure 6 generates only when even numbered Co atoms are interacting.45 The smallest unit of such a structure is a dinuclear compound on the edge of MoS2 particles. When we assume the formation of such a dinuclear unit of two Co atoms (a spin pair model), the magnetic susceptibility χ of Co can be expressed by eq 1.45,46

χ ) RNAg2µB2/kBT[3 + exp(-2J/kBT)]

(1)

where NA is Avogadro’s constant, µB is Bohr magneton, g is gyromagnetic factor (assumed to be 2, here), kB is Boltzman’s constant, J is magnetic interaction strength defined by H )

292 J. Phys. Chem. B, Vol. 109, No. 1, 2005

Okamoto et al.

Figure 6. Magnetic susceptibility χ and effective magnetic moment (µB) of Co for CVD-Co/MoS2/Al2O3 as a function of temperature. The best fitting curve, assuming a dinuclear cluster, for the observed magnetic susceptibility is also shown.

Figure 7. Magnetic susceptibility χ and effective magnetic moment (µB) of Co for imp-Co-MoS2/Al2O3 as a function of temperature. The best fitting curve, assuming a dinuclear cluster, for the observed magnetic susceptibility is also shown.

TABLE 1: Magnetic Parameters Obtained by Fitting the Experimental Magnetic Susceptibility of Co Assuming a Dinuclear Co Clustera on the Edge of MoS2 Particles catalyst CVD-Co/MoS2/Al2O3 CVD-Co/MoS2/Al2O3b imp-Co-MoS2/Al2O3 CVD-Co/Al2O3 imp-Co/Al2O3c a

Mo (W) Co content/wt % content/wt % 8.7 8.7 8.7

3.18 3.26 2.53 2.74 2.53

J/K

R

-7.1 -6.9 -8.1 -7.5

0.84 0.96 0.67 0.31 0.30

Eq 1. b H2-treated for 30 min at 673 K. c A Curie-Weiss law (eq

2).

-2JS1‚S2, and R is the fraction of the paramagnetic spin per Co atom. It was found that the temperature dependency of the observed χ was rationally fitted by using the theoretical eq 1 as shown in Figure 6, suggesting that the structure of the CoMo-S phase is represented as a Co dinucler cluster. The fitting parameters, J and R, are summarized in Table 1. When CVD-Co/MoS2/Al2O3 was treated in a flow of H2 (1 atm) at 673 K for 30 min, the antiferromagnetic behavior was not changed significantly, and no formation of paramagnetic species was detected. The temperature dependency of χ was well-fitted by eq 1. The parameters are given in Table 1. The effective magnetic moments for CVD-Co/MoS2/Al2O3 before and after the H2 treatment are close to the theoretical value for Co(II) in a low spin state, 1.73 µB, at around 300 K. Other workers27-29 reported lower values of the effective magnetic moments for the Co sulfide species in Co-Mo sulfide catalysts. This may be due to coexistence of Co9S8 (Pauli paramagnetic compound,47 1.2 × 10-4 cm3 (mol Co)-1) in their catalysts and/ or possibly due to lower accuracy in some cases as noted by Richardson.29 It is worthy of note that the Co species in the Co-Mo-S phase are divalent in a low spin state, characteristic of sulfur environments, even after the H2 treatment at 673 K. The magnetic susceptibility and effective magnetic moment of Co are shown in Figure 7 for imp-Co-MoS2/Al2O3. Apparently, a similar magnetic behavior was observed to that for CVD-Co/MoS2/Al2O3. The antiferromagnetic component was also well-fitted assuming a dinucler Co sulfide cluster. The parameters are presented in Table 1. The value of J is in excellent agreement with that for CVD-Co/MoS2/Al2O3 within the experimental accuracy (-7.6 ( 0.5 K). The magnetic susceptibility of CVD-Co/Al2O3 showed some contribution of antiferromagnetic components, and the fitting

Figure 8. Magnetic susceptibility χ and effective magnetic moment (µB) of Co for imp-Co/Al2O3 as a function of temperature. The best fitting curve, assuming a Curie-Weiss law, for the observed magnetic susceptibility is also shown.

parameters assuming a Co sulfide dimer species are shown in Table 1. On the other hand, as presented in Figure 8, the magnetic susceptibility χ of imp-Co/Al2O3 obeyed a CurieWeiss law expressed by eq 2.

χ ) RNAg2µB2/4kB(T - TW)

(2)

where TW is Curie-Weiss temperature, and the other symbols are the same as those in eq 1. The best fitting curve is also shown in Figure 8 (TW ) -6.9 K and R ) 0.30). In both Co sulfide catalysts, the values of R are around 0.3, indicating that about 70% of Co is present as Co9S8, whereas about 30% as highly dispersed Co sulfide clusters (imp-Co/Al2O3) or as Co sulfide dimer species (CVD-Co/Al2O3). Discussion Fraction of Co Atoms Available for NO Adsorption. In the previous study,36,37 we have shown that the Co-Mo-S phase is selectively formed on the edge of supported MoS2 particles by the CVD technique. The edge of the MoS2 particles is fully occupied by the Co-Mo-S phase to show the maximum potential HDS activity of the supported MoS2 catalyst.37,39 It

Active Sites of Co-Mo Hydrodesulfurization Catalysts should be noted that the amount of Co in the CVD-Co/MoS2 catalyst corresponds to the amount of the Co-Mo-S phase. Figure 2 confirms the selective formation of the Co-Mo-S phase in the present CVD-Co/MoS2/Al2O3 catalyst. Nitric oxide has been widely employed for the characterization of HDS catalysts.4 However, important issues have remained unresolved for more than two decades, for example, as to the oxidation and/or reconstruction of the Co sites by NO adsorption and as to the fraction of the Co atoms detected by NO adsorption. The present study can answer these issues, for the first time, since all the Co atoms in the CVD-Co/MoS2 catalysts form the Co-Mo-S phase and the edge of MoS2 particles is fully covered by the Co atoms. Actually, the FTIR spectra of NO adsorption on CVD-Co/MoS2/Al2O3 show the adsorption of NO almost exclusively on the Co sites (Figure 3b,c). The selective formation of the Co-Mo-S phase in CVD-Co/MoS2/Al2O3 is consistent with the finding that the value of R (the fraction of the antiferromagnetic components) is close to unity, as summarized in Table 1 (vide infra). As shown in Figure 1, the amount of NO adsorption is constant in a wide range of adsorption temperatures. In addition, no oxidation of Co sites is detected by the XANES study (Figure 5c). These findings indicate that the NO adsorption technique provides a reliable method to count the number of the adsorption sites. The slope of the proportional line in Figure 2 suggests that only 33% of the Co atoms in the CVD-Co/MoS2 catalysts adsorb two NO molecules in a dinitrosyl form when the catalyst was cooled to room temperature in a stream of 10% H2S/H2. Even after CVD-Co/MoS2/Al2O3 was treated in an atmospheric pressure of H2 between 573 and 673 K, only half of the Co atoms can accommodate NO molecules. The finding in Figure 4 that the NO/Co ratio remains constant even after the prolonged treatment at 673 K indicates the formation of a stable structure of the Co-Mo-S phase during HDS reaction, at least, under atmospheric H2 conditions. In conformity with this finding, the Co K-edge XANES spectrum (Figure 5b) suggests no significant electronic and/or structural changes of the Co atoms after the H2 treatment of CVD-Co/MoS2/Al2O3. In consequence, the increase in the amount of NO adsorption by the H2 treatment is not caused by structural changes of the Co-Mo-S phase by the treatment but by the elimination of adsorbed sulfur compounds as H2S to produce cus Co sites. This is also supported by the FTIR spectrum of NO adsorption on the H2treated CVD-Co/MoS2/Al2O3 catalyst (Figure 3c). The finding that only half of the Co atoms in the Co-Mo-S phase adsorb NO molecules is, therefore, ascribed to the nature of the CoMo-S phase. Magnetic Property of the Co Atoms in the Co-Mo-S Phase. The present magnetic study of the Co-Mo-S phase unambiguously reveals, for the first time, that the Co atoms on the edge of MoS2 particles interact with each other to show an antiferromagnetic behavior. Topsøe et al.,28 briefly noted some contribution of antiferromagnetic components in their magnetic study of an unsupported Co-MoS2 sample (Co/Mo ) 0.0625). As stated above briefly, it seems difficult to interpret the antiferromagnetic interactions between the Co atoms on the basis of the structure of the Co-Mo-S phase proposed on the basis of EXAFS analyses13-15 since the Co(II) (3d7) ions in the structure are assumed to be coordinated by four to six sulfur atoms and separated from each other. Recently, Bollinger et al.18 suggested from ab initio calculations that the edge of MoS2 particles shows a one-dimensional metallic state. When Co atoms are situated on the edge of MoS2 particles, the spin of the Co(II) ion may be considerably delocalized over the edge.

J. Phys. Chem. B, Vol. 109, No. 1, 2005 293

Figure 9. Best fitting curve of the magnetic susceptibility χ (O) observed for CVD-Co/MoS2/Al2O3, assuming a Co dinuclear cluster (thick solid line, J ) -7.1 K), a Co4 cluster (dotted line, J ) -6.3 K), or an infinite Co chain (thin solid line, J ) -7.2 K).

However, on the basis of the metallic edge model, it seems still difficult to induce the antiferromagnetic property in Figure 6 without assuming a combination of even-numbered Co atoms. Instead, we propose a model of the Co-Mo-S phase in which Co atoms in a low spin state are located on the edge of MoS2 particles forming a one-dimensional array of Co sulfide clusters. In this model, the antiferromagnetic behavior is expected only when even-numbered Co atoms are interacting with each other.45 As shown in Figure 6, we can reasonably fit the observed magnetic susceptibility assuming a one-dimensional array of dinuclear clusters. It would also be possible to assume a larger size of one-dimensional clusters with evennumbered Co atoms.45 To examine the possibility, we tried to fit the observed magnetic susceptibility assuming a Co4 cluster and an infinite Co sulfide chain.45 The fitting results are compared in Figure 9. Obviously, the best fit is obtained when a dinuclear Co sulfide cluster is assumed. In addition, the size of MoS2 particles is very limited and usually 2-3 nm by TEM observations.37 Hence, it is considered reasonable to propose a dinuclear cluster as the structure of the Co-Mo-S phase. It is remarkable in Table 1 that the value of J for imp-Co-MoS2/ Al2O3 is in excellent agreement with that for CVD-Co/MoS2/ Al2O3 within the experimental accuracy (-7.6 ( 0.5 K), showing that the identical Co clusters are formed on the edge of MoS2 particles in the impregnation and CVD catalysts despite the completely different preparation methods. The fitting in Figure 6 is reasonable but not perfect even when we assume the formation of a dinuclear cluster. This may be due to limited experimental accuracy caused by the use of glass ampules instead of quartz ones, to neglecting possible interactions between the Co-Mo-S phase and residual paramagnetic Mo(V) species present in MoS2/Al2O3 in a small amount, and/ or to some contribution of larger Co sulfide clusters. Co-Mo/Al2O3 catalysts show HDS activity in the presence of H2 around 600-673 K. The antiferromagnetic behavior and their magnetic parameters of CVD-Co/MoS2/Al2O3 are not essentially varied by the H2 treatment at 673 K, as presented in Table 1. It is, therefore, concluded that the Co dinuclear clusters are stable during HDS reaction, at least at atmospheric pressure of H2 and 673 K. This is in excellent agreement with the results of the Co K-edge XANES (Figure 5b) and NO adsorption (Figure 3c and Figure 4). It is worthy to note that the value of

294 J. Phys. Chem. B, Vol. 109, No. 1, 2005

Figure 10. Schematic structure of the Co-Mo-S phase. Small gray balls, Co atoms; small black balls, Mo atoms; large gray balls, S atoms.

R for CVD-Co/MoS2/Al2O3 is close to unity irrespective of the H2 treatment, indicating that no significant amount of Co9S8 is formed by the CVD technique. The lower R value for impCo-MoS2/Al2O3 (Table 1) evidently suggests the formation of Co9S8 along with the Co dinuclear clusters. CVD-Co/Al2O3 and imp-Co/Al2O3 show different magnetic properties depending on the preparation method; antiferromagnetic Co sulfide dimers in CVD-Co/Al2O3 and highly dispersed Co sulfide clusters having antiferromagnetic exchange interactions (TW < 0) between a significant number of Co atoms in imp-Co/Al2O3, the fractions of which are more or less the same in the both catalysts (R ) 0.3). The rest of the Co sulfide species in both catalysts is ascribed to Co9S8. The formation of the antiferromagnetic dimers in CVD-Co/Al2O3 may result from a partial formation of dinuclear Co carbonyl species such as Co2(CO)8 during the decomposition of Co(CO)3NO, although Mauge´ et al.48 showed by FTIR that Co(CO)3NO adsorbed on Al2O3 loses CO to form an adsorbed Co(CO)2NO moiety in an initial stage of the decomposition. However, it should be stressed here that the formation of the Co dinuclear clusters in CVDCo/MoS2/Al2O3 is due to the chemical nature of the Co sulfide species on the edge of MoS2 particles since the identical Co dinuclear clusters are formed also in imp-Co-MoS2/Al2O3 despite the absence of such species in imp-Co/Al2O3. Structure of the Co-Mo-S Phase. As discussed above, on the basis of the structure of the Co-Mo-S phase proposed previously,13-15,42-44 it seems difficult to explain the amount of NO adsorption and magnetic property of the Co-Mo-S phase. Instead, we propose a dinuclear Co sulfide cluster on the edge of MoS2 particles as a structure of the Co-Mo-S phase. Figure 10 presents a possible structure of the Co-Mo-S phase constructed on the (101h0) surface (Mo-edge) of MoS2 particles. A similar Co dinuclear structure may also be possible on the (1h010) surface (S-edge). The DFT calculations by Schweiger et al.49 predicted that the Co atoms are located preferentially on the S-edge under usual sulfidation conditions as employed in the present study. The Co atoms on the S-edge are in a tetrahedral coordination. However, as noted previously, the XANES results in Figure 5 clearly show that the intensity of the preedge peak for CVD-Co/MoS2/Al2O3 is much weaker than that for Co9S8 in which 89% of the Co atoms are in a tetrahedral coordination and the rest in an octahedral one. After the H2 treatment, the intensity of the preedge peak remained weak. The weak preedge peak of the Co K-edge XANES spectra for Co-Mo sulfide catalysts have also been reported by other workers.13-15,42-44 These XANES results suggest that the Co atoms are in a distorted octahedral symmetry or in a square pyramidal configuration,13-15,42-44 in sharp contrast to the tetrahedral configuration predicted by Schweiger et al.49 by their DFT calculations. In addition, in agreement with the XANES results, the coordination number of Co-S bonds has been reported between 5 and 6.13,43,44 Therefore, it seems difficult to

Okamoto et al. assume that the Co atoms located on the S-edge and in a tetrahedral configuration are preferentially present as the CoMo-S structure. Accordingly, the nature of the Co-Mo-S structure will be discussed hereinafter on the basis of the model in Figure 10, although a similar counterpart may also be possible on the S-edge. The structure in Figure 10 obviously explains the antiferromagnetic interactions between two Co atoms as supported by the excellent fittings (Figure 6 and 7) using the theoretical eq 1 with reasonable sizes of the parameters (Table 1). Recently, Raybaud et al.20 made ab initio DFT calculations of energetically stable structures of the Co-Mo-S phase using a model in which the Mo atoms in MoS2-edges are substituted by Co atoms. They suggested a very similar structure to the one in Figure 10 as one of the stable structures, when all the Mo atoms on the Moedge are substituted by Co atoms (Co/Ms ) 1 and sulfur coverage ) 25% in their model20). Their structural parameters are consistent with those calculated from the EXAFS analysis.13-15,42-44 More recently, the DFT calculations by Travert et al.21 led them to similar conclusions for the structure of the Co-Mo-S phase. Byskov et al.19 concluded from DFT calculations that the Mo-substitution by Co on the S-edge provides energeticallty the most stable structure. In their calculations, a local structure (SMo)4Mo-S-Co(SMo)4 (SMo: S atom binding to Mo atoms) involving a bridging sulfur atom was assumed. In consistent with their calculations, the STM observations16,17 of Co-MoS2 model clusters fabricated on Au(111) suggested the formation of a one-dimensional array of Co atoms with bridging S atoms on the S-edge of MoS2 particles. The Co dinuclear structure in Figure 10, and possibly, its counterpart on the S-edge are consistent with the theoretical predictions,19-21 experimental observations,16,17 and the EXAFS parameters published so far.13-15,42-44 In particular, the present model is consistent with the EXAFS results by Louwers and Prins15 showing a considerable Ni-Ni contribution in the Fourier transforms for Ni-Mo sulfide catalysts, although they attributed the Ni-Ni bonds to the evidence of the presence of adjacent Ni atoms on the MoS2-edge. According to the calculations by Travert et al.21 of the electronic state of the structure analogous to the one in Figure 10, the substitution of Mo by Co leads to withdrawing of electrons from the metal-sulfur bonding states and an increase of the electron density in the metal-sulfur antibonding orbitals. Accordingly, it is expected that the spin pair interactions between two Co atoms are weak because of the increased antibonding character of the Co-bridging S bonds. This may be consistent with the relatively small J values (-7.6 ( 0.5 K). The present magnetic study suggests a preferential formation of a Co dinuclear cluster on the edge of MoS2 particles. The size of MoS2 particles in Co-Mo sulfide catalysts shows a distribution typically with a peak maximum around 2-4 nm. The number of the sites on the edge of MoS2 for the accommodation of Co atoms may be odd or even at the same probability. The present magnetic results may suggest that only even-numbered Co atoms are arrayed leaving one vacant site, when an odd number of the sites are available. This may be due to higher thermal stability of a Co sulfide dinuclear cluster as compared to that of isolated Co atoms. The amount of NO adsorption on the Co-Mo-S phase can be readily understood on the basis of the Co dinuclear structure in Figure 10 and their one-dimensional array on the MoS2-edge. Figure 11 illustrates a schematic model of the Mo-edge of MoS2 particles fully occupied by the Co-Mo-S phase. When CVDCo/MoS2/Al2O3 is cooled in H2S/H2 to room temperature after

Active Sites of Co-Mo Hydrodesulfurization Catalysts

Figure 11. Proposed mechanism of the adsorption of NO on the CoMo-S phase.

the sulfidation, some excess S atoms may adsorb on the Co atoms in the Co-Mo-S phase as bridging S atoms or SH groups to form an array of Co atoms connected each other. As shown in Figure 11a, the adsorption of one extra S atom between two dinuclear clusters produces a tetranuclear one-dimensional cluster that is also expected to show an antiferromagnetic property (even number of Co atoms) in agreement with the experimental results in Figure 6. When the catalyst is treated in an H2 stream at 573-673 K, it is considered that the excess S atoms adsorbed on the Co sites are removed as H2S to leave the dinuclear Co sulfide clusters (Figure 11b). It is considered that the bridging S atoms in the dinuclear clusters remain even after the H2-treatment since CVD-Co/MoS2/Al2O3 shows antiferromagnetic behavior even after the treatment. When NO molecules visit the dinuclear cluster (the CoMo-S phase), it is considered that one of the Co-S bonds is cleaved, possibly in a concerted manner, to allow two NO molecules adsorbed on the Co atom in a dinitrosyl form, concomitantly accompanying the formation of a Co-St (St: terminal sulfur atom) bond, as schematically presented in Figure 11c. This may be in accordance with a slight decrease in the Co-S peak intensity of the Fourie transform of k3-weighted Co K-edge EXAFS oscillations for CVD-Co/MoS2/Al2O3, when the catalyst was exposed to NO.50 The adsorption mechanism of NO clearly explains the fact that only half of the Co atoms of the Co-Mo-S phase at the most is available for the NO adsorption. When an extra S atom adsorbs between the dinuclear units, only one of the edge Co atoms may be accessible for NO since the Co atoms in the middle of the cluster are coordinatively saturated (octahedrally coordinated). When the same numbers of the dinuclear and tetranuclear clusters are simultaneously present as shown in Figure 11a, only 33% of the Co atoms can adsorb two NO molecules to form dinitrosyl species. The amount of NO adsorption on CVD-Co/MoS2/ Al2O3 was 0.33 as expressed by the ratio (NO)2/Co, when the catalyst was cooled in the stream of H2S/H2. The coincidence may suggest that similar amounts of dinuclear and teranuclear Co sulfide clusters are formed in the sulfided catalyst. Conclusions In the present study, we elucidated the NO adsorption behavior and magnetic property of the active sites (i.e., the CoMo-S phase) of Co-Mo sulfide catalysts using CVD-Co/ MoS2/Al2O3, in which all the Co atoms are present as the CoMo-S phase. On the basis of the findings, we propose a novel

J. Phys. Chem. B, Vol. 109, No. 1, 2005 295 empirical structure of the Co-Mo-S phase. The salient findings of the present study are as follows: (1) with CVD-Co/MoS2/ Al2O3, 33% of the Co atoms is available for NO adsorption when the catalyst is cooled in an H2S/H2 stream, whereas only half of the Co atoms is subjected to NO adsorption when the catalyst is treated in an atmospheric H2 stream at 573-673 K. (2) The Co atoms in CVD-Co/MoS2/Al2O3 exclusively exhibit an antiferromagnetic property, indicating that even-numbered Co atoms interact with each other in the Co-Mo-S structure. (3) imp-Co-MoS2/Al2O3, prepared by an impregnation technique, is composed of the antiferromagnetic Co sulfide species having the identical magnetic property with those in CVDCo/MoS2/Al2O3 despite the absence of such Co sulfide species in imp-Co/Al2O3. (4) On the basis of the NO adsorption behavior and magnetic property of CVD-Co/MoS2/Al2O3, it is proposed that the structure of the Co-Mo-S phase is represented as a dinuclear Co sulfide cluster on the edge of MoS2 particles. The structure of the cluster is stable under atmospheric pressure of H2 at 673 K. Acknowledgment. This study was supported by Grant-inAid for Scientific Research (16360404) from the Ministry of Education, Culture, Sport, Science, and Technology, Japan. References and Notes (1) Song, C. Catal. Today 2003, 86, 211. (2) Song, C.; Ma, X. Appl. Catal. B 2003, 41, 207. (3) Knudsen, K. G.; Cooper, B. H.; Topsøe, H. Appl. Catal. A 1999, 189, 205. (4) Topsøe, H.; Clausen, B. S.; Massoth, F. E. Catalysis Science and Technology; Anderson, J. R., Boudard, M., Eds.; Springer: Berlin, 1996; Vol. 11. (5) Prins, R. In Handbook of Heterogeneous Catalysis; Ertl, G., Kno¨zinger, H., Weitkamp, J., Eds.; VCH: Weinheim, 1997; p 1908. (6) Kabe, T.; Ishihara, A.; Qian, W. Hydrodesulfurization and Hydrodenitrogenation; Kodansha: Tokyo, 1999. (7) Whitehurst, D. D.; Isoda, T.; Mochida, I. AdV. Catal. 1998, 42, 345. (8) Prins, R.; de Beer, V. H. J.; Somorjai, G. A. Catal. ReV. Sci. Eng. 1989, 31, 1. (9) Eijsbouts, S. Appl. Catal. A 1997, 158, 53. (10) Clausen, B. S.; Mørup, S.; Topsøe, H.; Candia, R. J. Phys. Colloq. 1976, C6, 37. (11) Topsøe, N.-Y.; Topsøe, H. J. Catal. 1983, 84, 386. (12) Topsøe, H.; Clausen, B. S.; Topsøe, N.-Y.; Pederson, E. Ind. Eng. Chem. Fundam. 1986, 25, 25. (13) Bouwens, S. M. A. M.; van Veen, J. A. R.; Koningsberger, D. C.; de Beer, V. H. J.; Prins, R. J. Phys. Chem. 1991, 95, 123. (14) Niemann, W.; Clausen, B. S.; Topsøe, H. Catal. Lett. 1990, 4, 355. (15) Louwers, S. P. A. Prins, R. J. Catal. 1992, 133, 94. (16) Helveg, S.; Lauritsen, J. V.; Lagsgaad, E.; Stensgaad, I.; Nørskov, J. K.; Clausen, B. S.; Topsøe, H.; Besenbacher, F. Phys. ReV. Lett. 2000, 84, 951. (17) Lauritsen, J. V.; Helveg, S.; Lagsgaad, E.; Stensgaad, I.; Clausen, B. S.; Topsøe, H.; Besenbacher, F. J. Catal. 2001, 197, 1. (18) Bollinger, M. V.; Lauritsen, J. V.; Jacobsen, K. W.; Nørskov, J. J.; Helveg, S.; Besenbacher, F. Phys. ReV. Lett. 2001, 87, 196803. (19) Byskov, L. S.; Nørskov, J. K.; Clausen, B. S.; Topøe, H. J. Catal. 1999, 187, 109. (20) Raybaud, P.; Hafner, J.; Kresse, G.; Kasztelan, S.; Toulhoat, T. J. Catal. 2000, 190, 128. (21) Travert, A.; Nakamura, H.; van Santen, R. A.; Cristol, S.; Paul, J.-F.; Payen, E. J. Am. Chem. Soc. 2002, 124, 7084. (22) Lipsch, J. M. J. G.; Schuit, G. C. A. J. Catal. 1969, 15, 163, 174. (23) Ratmaswamy, A. V.; Sivasanker, S.; Ratnasamy, P. J. Catal. 1976, 42, 107. (24) Mitchell, P. C. H.; Trifiro, F. J. Catal. 1974, 33, 350. (25) Mehandjiev, D.; Zhecheva, E.; Aleksic, B. R.; Alkesic, B. D.; Markovic, B.; Bogdanov, S. React. Kinet. Catal. Lett. 1991, 43, 7. (26) Chiplunker, P.; Martinez, N. P.; Mitchell, P. C. H. Bull. Soc. Chim. Belg. 1981, 90, 1319. (27) Topsøe, H.; Topsøe, N.-Y.; Sørensen, O.; Candia, R.; Clausen, B. S.; ACS Symp. DiV. Petrol. Chem. 1983, 1252. (28) Topsøe, H.; Topsøe, N.-Y.; Sørensen, O.; Candia, R.; Clausen, B. S.; Kallesøe, S.; Pedersen, E.; Nevald, R. ACS Symp. Ser. 1985, 279, 235. (29) Richardson, J. T. J. Catal. 1988, 112, 313.

296 J. Phys. Chem. B, Vol. 109, No. 1, 2005 (30) Okamoto, Y.; Katoh, Y.; Mori, Y.; Imanaka, T.; Teranishi, S. J. Catal. 1981, 70, 445. (31) Topsøe, N.-Y.; Topsøe, H. Bull. Soc. Chim. Belg. 1981, 90, 1311. (32) Topsøe, N.-Y.; Topsøe, H. J. Catal. 1982, 75, 354. (33) Koizumi, N.; Yamazaki, M.; Hatanaka, S.; Yamada, M. Catal. Today 1997, 39, 33. (34) Shuxian, Z.; Hall, W. K.; Ertl, G.; Kno¨zinger, H. J. Catal. 1986, 100, 167. (35) Nielsen, L. P.; Ibsen, L.; Christensen, S. D.; Clausen, B. S. J. Mol. Catal. A 2000, 162, 375. (36) Okamoto, Y.; Ochiai, K.; Kawano, M.; Kobayashi, K.; Kubota, T. Appl. Catal. A 2002, 226, 115. (37) Okamoto, Y.; Ishihara, S.; Kawano, M.; Satoh, M.; Kubota, T. J. Catal. 2003, 217, 12. (38) Okamoto, Y.; Kubota, T. Catal. Today 2003, 86, 31. (39) Okamoto, Y.; Ochiai, K.; Kawano, M.; Kubota, T. J. Catal. 2004, 222, 143. (40) Kubota, T.; Hosomi, N.; Bando, K.; Matsui, T.; Okamoto, Y. Phys. Chem. Chem. Phys. 2003, 5, 4510.

Okamoto et al. (41) Hiromitsu, I.; Ikeda, N.; Handa, M.; Ito, T. Phys. ReV. B 1998, 57, 8501. (42) Bouwens, S. M. A. M.; Koningsberger, D. C.; de Beer, V. H. J.; Prins, R. Catal. Lett. 1988, 1, 55. (43) Bouwens, M. A. M.; van Zon, F. B. M.; van Dijik, M. P.; van der Kraan, A. M.; de Beer, V. H. J.; van Veen, J. A. R.; Koningsberger, D. C. J. Catal. 1994, 146, 375. (44) Craje, M. W. J.; Louwers, S. P. A.; de Beer, V. H. J.; Prins, R.; van der Kraan, A. M. J. Phys. Chem. 1992, 96, 5445. (45) Kahn, O. Molecular Magnetism; VCH: 1993; Ch. 6. (46) Bleaney, B.; Bowers, K. D. Proc. R. Soc. (London) Ser. A 1952, 214, 451. (47) Knop, O.; Huang, C.; Reid, K. I. G.; Carlow, J. S.; Woodhams, F. W. D. J. Solid State Chem. 1976, 16, 97. (48) Mauge`, F.; Vallet, A.; Bacherlier, J.; Duchet, J.; Lavalley, J. C. J. Catal. 1996, 162, 88. (49) Schweiger, H.; Raybaud, P.; Toulhoat, H. J. Catal. 2002, 212, 33. (50) Kubota, T.; Okamoto, Y. Unpublished results.