J O U R N A L O F C H E M I S T R Y Materials In situ X-ray absorption spectroscopic studies at the cobalt K-edge on an Al2O3-supported rhenium-promoted cobalt Fischer–Tropsch catalyst. Comparing reductions in high and low concentration hydrogen Arild Moen,a David G. Nicholson,a Magnus Rønningb and Hermann Emerichc aDepartment of Chemistry, Norwegian University of Science and Technology, N-7055 Trondheim, Norway bDepartment of Industrial Chemistry, Norwegian University of Science and Technology, N-7055 Trondheim, Norway cSwiss-Norwegian Beamline, European Synchrotron Radiation Facility BP 220, F-38043 Grenoble, France Received 5th June 1998, Accepted 10th August 1998 In situ XAFS spectroscopic studies have been carried out at 450 °C on the hydrogen reduction of a rheniumpromoted Co3O4/Al2O3 catalyst.Reductions carried out using 100% hydrogen and 5% hydrogen in helium gave diVerent results. Whereas the reduction using dilute hydrogen yielded bulk-like metallic cobalt particles (hcp or fcc), the reaction with pure hydrogen led to a more dispersed system with smaller cobalt metal particles (<40 A° ) the crystal form of which could not be established so that the recently reported metastable nonclose-packed bodycentred cubic form cannot be excluded.Reoxidation of a similar catalyst in water-containing gas mixtures has been reported in the literature; it is suggested that the diVerent outcome in the case of the 100% hydrogen protocol may be due to a similar mechanism. This would involve the in situ water produced by the reduction with reoxidation–reduction of cobalt metal particles in the water vapour–hydrogen mixture.However, this mechanism cannot be established by the present study. Additionally, in both reduction protocols a small fraction (3–4 wt.%) of the cobalt content is randomly dispersed over the tetrahedral vacancies of the alumina support with Co–O bond lengths of 1.96±0.01 A° .This dispersion occurs during reduction and not calcination.The cobalt in these sites cannot be reduced at 450 °C, a temperature that is too low to permit formation of the spinel CoAl2O4. X-Ray absorption spectroscopy (XAS) is a useful method for part involves the diVusion of cobalt ions into the structure of characterising local structural features of selected elements in c-Al2O3 where they occupy vacant octahedral and/or tetraheterogeneous catalysts.Although the technique is actually a hedral positions. This apparently leads to two main types of bulk technique it is also valuable for studying chemical reac- cobalt-containing phase: (a) cobalt bonded to the alumina tions that take place at the surfaces of heterogeneous catalysts support in the form of a surface phase that is resistant to because the overall contribution to the XAS signal arises from reduction and (b) the easily reduced spinel Co3O4 which the significant proportion of active sites that are highly dis- dominates above a certain cobalt concentration (3–4%).A persed over a large number of surfaces. Indeed, XAS is major factor determining catalytic activity is the influence of particularly suited for characterising such systems because promotors; e.g.there is a two-fold increase in the rates of their inherent lack of long range order precludes study by X- hydrogenation of carbon monoxide in the presence of platinum ray diVraction. In addition, XAS is capable of giving infor- and rhenium promoters.28,29 mation on interactions between the support and the metal We report here the results of in situ XAS experiments on catalyst, an aspect which is relevant to this work since such the reduction with dilute hydrogen (5% in helium) at 450 °C interactions are known to influence activity and selectivity.1–3 of the rhenium-promoted Co/Al2O3 (cobalt content 26 wt.%).This paper is concerned with an alumina-supported cobalt This is an extension of previous work35 in which the catalyst catalyst used in the Fischer–Tropsch process.In this process was reduced by 100% hydrogen at the same temperature. We linear high-molecular weight aliphatic hydrocarbons are syn- also report here the results of experiments carried out at lower thesised by catalytically hydrogenating carbon monoxide.4–30 temperatures (200 °C) on the same catalyst and the results of A comprehensive review of the literature dealing with this the reduction at 450 °C of a similar system but one with a system and its variants is given by Holmen et al.31 much lower cobalt content (4.6%).XAS has been used previously to study several modifications of Co/Al2O3 catalysts. Probably the earliest study was that Experimental reported by Greegor et al.32 Fay et al.24 used the method to characterise a boron-modified system, and HuVman et al.33 Synthesis used XAS in an in situ study of potassium-promoted cobalt Essentially following the procedure described in the catalysts supported on alumina and silica during reduction in literature29(but in our case with a triple cobalt loading) the hydrogen at 200 °C with subsequent reaction in a synthesis alumina-supported cobalt catalyst with 26% cobalt loading gas mixture.The latter support was also used by Takeuchi (analysis by atomic absorption spectroscopy) was prepared by et al.34 Recently, Moen et al.35 in a preliminary report described the incipient wetness impregnation of c-alumina (sieved to an in situ XAS study at the Co K-edge on the reduction of 40–50 mesh) with aqueous solutions of Co(NO3)2·6H2O and the rhenium-promoted Co/Al2O3 catalyst (cobalt content HReO4(calculated to give 1.0 wt.% rhenium). The resulting 26 wt.%) at 450 °C using 100% hydrogen.From these studies it appears that catalyst preparation in material was dried overnight at 120 °C and then calcined at J. Mater. Chem., 1998, 8(11), 2533–2539 2533500 °C for 8 h.A sample of the same system but with a much EXAFS data analysis. The data were corrected for dark currents, converted to k-space, summed and background sub- lower cobalt loading (4.6%) was prepared and treated similarly.29 tracted to yield the EXAFS function xobsi(k) using the EXCALIB and EXBACK programs.38 Model fitting was Hydrogen reduction of the catalysts was carried out in situ (see below).carried out with EXCURV90 using curved-wave theory and ab initio phase shifts.38,39 The edge positions were determined from the first inflection points (after any pre-edge features) of X-Ray absorption data the derivative spectra. The k3 weighting scheme compensates for the diminishing XAS data were collected using the facilities of the bending photoelectron wave at higher k.The curve-fitting was per- magnet Swiss-Norwegian Beamline (SNBL) at the European formed on data that had been Fourier filtered over a wide Synchrotron Radiation Facility (ESRF), Grenoble, France. range (1.0–25.0 A° ). This filter removes low-frequency contri- Spectra were obtained on station EH1 (SNBL-ESRF) at the butions to the EXAFS below 1 A° , but does not smooth the cobalt K-edge (l=1.6086 A° ; energy=7 709 eV).spectrum (i.e. the noise is not removed). The ranges for the A channel-cut silicon(111) monochromator with an Fourier transformations were 3–14 A° -1. unfocussed beam was used to scan the X-ray spectra. The The model compounds, cobalt Tutton salt,40 cobalt alumin- beam currents ranged from 80–130 mA at 6.0 GeV.Higherate41 and cobalt(II) oxide42 were used to check the validity of order harmonics (ca. two orders of magnitude) were rejected the ab initio phase shifts and establish the general parameter by means of a gold-coated mirror angled at 7.3 mrad from a AFAC (proportion of absorption causing EXAFS) and VPI beam of size 0.6×4.7 mm which was defined by the slits in (allows for inelastic scattering of the photoelectron).38 The the station.The maximum resolution (DE/E) of the Si(111) cobalt spinel (Co3O4), which contains both cobalt(II ) and bandpass is 1.4×10-4. cobalt(III),43 was used as a reference. Gas ion chamber detectors with their gases at ambient temperature and pressure were used for measuring the intensities of the incident (I0) and transmitted (It) X-rays.The detector Results and discussion gases were as follows: I0, detector length 17 cm, 100% N2; It, We have previously shown35 that the XAS of the unreduced length 31 cm, 35% Ar, 65% N2. material (sample A) is indistinguishable from that of the reference compound Co3O4. Clearly, the method of preparation, together with the comparatively high cobalt loading, Data collected at the Co K-edge.The spectral energy calibration was checked by measuring the spectrum of a Co- yields a material in which Co3O4 is the dominant phase and that this phase is distributed throughout sample A in the form foil (thickness 5 mm; the energy of the first inflection point being defined as the edge 7 709 eV; accurate calibrations are of relatively large (>100 A° ) crystallites.This conclusion is consistent with the sizes (140–190 A° ) derived from the line particularly important for the pre-edge and XANES regions where the need for comparisons between the diVerent spectra broadening observed in the X-ray diVractogram.29 The reduction at 450 °C with 5% hydrogen was monitored make it necessary to define the absolute energies of the spectral features; for the EXAFS the energy is relative to the individual by a series of quick short scans taken just before and after the edge.The intensity of the white line (characteristic of Co3O4) edges which are therefore defined as zero). The energy scale calibration of each spectrum was carried out using an rapidly decreased with time and on the basis of this indicator the reduction was essentially completed within approximately in-house program.The XAS of the catalysts, before and after reduction, and 15 min, although the process was continued for a further 6 h. Fig. 1 shows the XAS of sample A (26 wt.% loading of of the model compounds CoAl2O4, CoO, the cobalt Tutton salt, [NH4]2[Co(H2O)6][SO4]2, and Co3O4 were also meas- cobalt) reduced by 100% and 5% hydrogen (to give samples B and C, respectively) and sample D (4.6 wt.% loading) ured.The amounts of material in the samples were calculated from element mass fractions and the absorption coeYcients of reduced in 5% hydrogen; the spectrum of bulk cobalt metal is also shown. The X-ray absorption near edge structure the constituent elements36 above the absorption edge to give an absorber optical thickness of 1.5 absorption lengths.The (XANES) regions for the model/reference compounds and samples A, B, C, D are depicted in Fig. 2. The model/reference well-powdered samples of the model and reference compounds were mixed with boron nitride so as to give a sample thickness compounds, chosen for their tetrahedral and octahedral cobalt environments, are represented by the following compounds: of ca. 1.0 mm and placed in aluminium sample holders and held in place by Kapton tape. Co3O4/Al2O3 destined for the CoAl2O4 (tetrahedral CoII(O)4); [NH4]2[Co(H2O)6][SO4]2 (distorted octahedral CoII(O)6); CoO (octahedral CoII(O)6); in situ reduction was ground and sieved (7–125 mm) and mixed with the requisite amount of boron nitride or graphite to and Co3O4 (tetrahedral CoII(O)4 plus octahedral CoII(O)6).Fig. 3 contains the first derivatives of the edge region because achieve the desired absorber thickness. (The experiment using graphite was to establish whether that material aVects the they are useful for highlighting characteristic features about the edges (particularly the pre-edge features) and establishing product; the spectra of the samples diluted with boron nitride and graphite were similar.) their energies.35,44 A comparison of the XAS of cobalt metal (as a foil with The catalyst was then loaded into a Lytle in situ reactor-cell37 and reduced in a mixture of H2 (5%) in He (purity; 99.995%: the face centred cubic or hexagonal close-packed structure, their spectra are similar35) and sample B in Fig. 1 reveals that flow rate 60 ml min-1) by heating from room temperature to 450 °C and maintaining at that temperature for 6 h.there are significant diVerences between them. Yet, the spectrum of sample C closely resembles that of the cobalt metal. The same procedure was repeated for other samples but with the diVerence that the hydrogen flow was started at The edge regions of bulk cobalt metal and sample B are also diVerent, as emphasised by the first derivatives (Fig. 3). 450 °C. Similar measurements were also carried out at 200 °C and for 13 h. The samples are designated as follows. Sample Evidently, sample B, unlike sample C, is not composed of bulk metal particles. Hence, reduction by 100% and 5% A: the catalyst containing 26% cobalt (shown below to be mainly Co3O4/Al2O3).Sample B: produced by reducing sample hydrogen yields products (B and C) with diVerent physical characteristics. A in 100% hydrogen at 450 °C.35 Sample C: prepared by reducing sample A in 5% hydrogen at 450 °C. Sample D was The most prominent features in the XAS of sample B are the pre-edge peak at 7709 eV and the absence (or considerable prepared by reducing the catalyst containing 4.6% cobalt in 5% hydrogen at 450 °C.reduction) of the white line so characteristic of Co3O4, 2534 J. Mater. Chem., 1998, 8(11), 2533–2539Fig. 3 The first derivatives of the XANES spectra of samples A, B, C, D and the reference/model compounds. The positions of the preedge peaks are marked ($). CoAl2O4, CoO, cobalt Tutton salt and its precursor, sample A. The white line is also absent in the XAS of bulk cobalt metal which suggests that sample B contains a high degree of reduced material consistent with the estimate (80%) for reducibility of rhenium-promoted Co3O4/Al2O3 under the same Fig. 1 Normalised XAS of samples B, C, D and the reference cobalt conditions.29 metal. Pre-edge peaks The XAS at the K-edge of certain valence states of some transition compounds often contain an electronically interesting pre-edge feature a few eV below the edge.This feature is useful because it yields structural and electronic information, especially when combined with the extended X-ray absorption fine structure (EXAFS) region of the same spectrum.44,45 The peak results from the absorption process 1sA3d and for solid macromolecular materials (such as CoO, CoAl2O4 and Co3O4 ) the final state(s) are unoccupied bands.The transition probability (intensity) is related to the symmetry (A1AT2 for T d symmetry) and to the occupancy of the 3d shell. For a given occupancy, the transition is most intense when the first coordination shell lacks inversion symmetry. In the case of the cubic point groups this applies to tetrahedral (T d) environments but not to octahedral (Oh) symmetry; although for the latter point group a considerably weaker preedge feature does actually occur (despite being forbidden by the centre of symmetry).This is because the crystallographic point group represents a static model derived by time-averaging the asymmetric vibrations within the molecule whereas the XAS reacts to the individually and constantly changing structures of the local environment as vibrations momentarily Fig. 2 Normalised Co K-edge XANES of the samples and reference materials. eliminate the centre of symmetry. An example is the spectrum J. Mater. Chem., 1998, 8(11), 2533–2539 2535(Fig. 2 and 3) of CoO, with its very weak pre-edge feature the bond length of a tetrahedral CoII(O)4 environment the latter appears to be somewhat shorter than that yielded by the (better revealed in the derivative of the spectrum), in which the octahedrally coordinated cobalt is in an Oh environment.44 EXAFS of the bulk cobalt metal reference (2.48±0.02 A° , Debye–Waller type factor (2s2)=0.013 A° 2).Apai et al.47 For symmetries lower than Oh, as in the distorted octahedrally coordinated cobalt environment (Co(H2O)62+) in the reported that metal–metal distances in very small metal particles are contracted relative to those in the bulk, an obser- Tutton salt, [NH4]2[Co(H2O)6 ][SO4]2,40 the intensity is somewhat enhanced, although still comparatively weak.vation that has subsequently been reported by others.48,49 Although a shortened Co–Co distance would accompany an By contrast, the pre-edge peaks associated with tetrahedral cobalt environments are more intense in accordance with the edge shift a more detailed study that focuses on this particular aspect is required before the final conclusion can be drawn.noncentrosymmeric Td point group. This is exemplified in the spectrum of CoAl2O4 (Fig. 2 and 3). These figures also show XAS of sample C the pre-edge region for the spinel Co3O4 which has one third cobalt(II) in tetrahedral sites and two thirds Co(III) in octa- The XAS spectrum (Fig. 5 shows the magnitude of the Fourier hedral sites. The pre-edge feature is composed of the tetratransforms of the EXAFS region) of sample C shows that hedral peak and the very weak (forbidden) octahedral peak, reduction at 450 °C in 5% hydrogen produces bulk-like cobalt the overall peak intensity being reduced because the proportion metal (hcp or fcc) particles with only a minor fraction of of the cobalt content in tetrahedral sites is now only one third cobalt being incorporated into the alumina support (the the total cobalt content and not unity as in CoAl2O4 (but nonmetallic phase); the latter is discussed below.see below). XAS of sample D XAS of sample B It is evident from the full XAS spectrum (Fig. 1) and the Fig. 4 shows the XANES regions of samples A and B and EXAFS region (Fig. 6) that the major cobalt-containing com- bulk cobalt metal. The edge position of sample B is at a higher ponent consists of cobalt incorporated into the alumina sup- energy (7714 eV) than that in bulk cobalt metal.The pre-edge port. This cobalt fraction is not reduced by hydrogen at peak at 7710 eV in sample B serves both as a convenient 450 °C. The pre-edge peak shows that the cobalt environment control of the edge energy and as an important diagnostic is tetrahedral CoII(O)4 the EXAFS yielding a Co–O distance feature for tetrahedral CoII(O)4 environments. Another sigof 1.96±0.01 A° that is consistent with this.43 nificant feature in the sample B spectrum is the absence of the white line. As shown in the same figure, this is also character- The eVects of using concentrated versus dilute hydrogen istic of bulk metallic cobalt.Since the edge energy of bulk metallic cobalt is 7 709 eV, any significant contribution from This study shows that the main reduction product of sample that metal in a composite spectrum of two-or-more phases A at 450 °C using 100% hydrogen (sample B) diVers from that would obscure this pre-edge peak.This does not occur in sample B because the edge is actually moved to a higher energy. The significance of the edge shift, the pre-edge peak and a considerably reduced EXAFS amplitude have been discussed in terms of small metal particles (<40 A° ) and the consequent deviation from bulk properties.35 The eVect of particle size on the EXAFS amplitude is significant for dimensions below ca. 30 A° .26 Recent studies on the size determination by EXAFS confirm this.46 It is therefore apparent that a minor fraction of cobalt in sample B is sited in the tetrahedral environment that we know is present from the pre-edge peak.The distances extracted35 from the EXAFS are 1.92±0.01 A° (Debye–Waller type factor (2s2)=0.011 A° 2) for the Co(O)4 tetrahedral site (which is close to the distance extracted from the cobalt-containing support in sample D, see below) and 2.49±0.01 A° (Debye–Waller type factor (2s2)=0.023 A° 2) due to Co–Co backscattering. Whereas the former distance corresponds to Fig. 5 Fourier transforms of samples B and C (top and bottom, Fig. 4 Normalised Co K-edge XANES regions of samples A and B respectively). The amplitudes of the peaks for sample B are much reduced compared with bulk cobalt metal (centre). and the reference cobalt metal (bulk). 2536 J. Mater. Chem., 1998, 8(11), 2533–2539ing gas mixtures. They found that unpromoted and rheniumpromoted catalysts behave diVerently, with cobalt in the latter being more easily reoxidised and the resulting oxidised phase being more easily reduced.Other results in the literature also show that water has an eVect; the introduction of water to Fischer–Tropsch reactions that are catalysed by variants of the Co/Al2O3 system aVect the rate of reaction and distribution of the reaction products.51,52 Turning to the present experiments, since in situ water (ca. 2 mg) is produced during the reduction some degree of reoxidation of the freshly reduced cobalt metal would fit in with the results of Hilmen et al.50 If the action of in situ water is responsible for the diVerent outcomes of 100% and 5% hydrogen reduction at 450 °C then it seems likely that the partial pressures of water must be highest for the former and hence more eVective in reoxidising metal particles which are again reduced by hydrogen. Further experiments are necessary to ascertain whether this suggested mechanism is viable and also to establish whether a reoxidation–reduction process increases the metal dispersion by breaking up the initially large cobalt particles into smaller particles.The nonmetallic phase Information concerning the amount of cobalt incorporated into the alumina support is forthcoming from the XAS of samples C and D with 26% and 4.6% cobalt (as noted above). The data for the latter show that the amount of metallic cobalt produced is small. In order to estimate the fraction of cobalt that is incorporated into the alumina support the spectra of the high- and low-cobalt samples were first normalised and then a series of summed spectra were constructed by adding the low-cobalt spectrum (the major component in the support being cobalt incorporated into the support) to progressively increasing numbers of cobalt-metal spectra until the actual spectrum of the high-cobalt sample was closely reproduced.The best simulation was obtained with the fractions cobalt metal5low-cobalt catalyst being in the ratio 851.For the 26% loading this translates into ca. 3% cobalt being incorporated into the alumina support. (If the small fraction of metallic cobalt present in sample D (0.77%, see below) is taken into account then this figure would be slightly higher.) In order to check this result, the following alternative procedure was used.On comparing the pre-edge regions of samples C and D (Fig. 2) it is evident that the pre-edge peak assigned to tetrahedral Co(O)4 environments is obscured by the dominating contribution from metallic cobalt in the former (the position of the edge for cobalt metal being the same as that for the pre-edge peak.) Consistent with the low-cobalt metal content in sample D, the pre-edge peak in the spectrum of that material can be discerned because it is only partially merged into the metal edge.The spectrum of the nonmetallic phase was separated from the observed spectrum by subtracting the minor cobalt metal spectral contribution from the spectrum of sample D. The appropriately weighted cobaltmetal spectrum was obtained by multiplying the normalised cobalt-metal spectrum by a series of factors and subtracting each resulting spectrum from the actual spectrum of D until Fig. 6 Fourier transforms of sample A and sample D together with those of the spinel reference compounds. a spectrum (and its derivative) was obtained which exhibited a pre-edge peak that closely matched the same feature in CoAl2O4. A good match was generated by subtracting the cobalt-metal obtained using 5% hydrogen in helium (sample C). An explanation for this diVerence is sought in a mechanism that we spectrum weighted by the fraction 0.167 from the observed composite spectrum.This corresponds to a metal content of suggest may be connected with the reduction of this catalyst. Although the present results cannot establish this mechanism 0.77% cobalt for the 4.6% cobalt loading in sample D, the low-cobalt catalyst.(The same procedure has been described the concentration of hydrogen must be the key factor here. The background for the mechanism is provided by Hilmen recently in ref. 44.) Hence, the nonmetallic phase contributes 3.8% to the total cobalt loading of the catalysts. This can be et al.50 who used temperature programmed reduction and gravimetry to show that cobalt is reoxidised in water-contain- rationalised in terms of the structure of c-alumina.The sup- J. Mater. Chem., 1998, 8(11), 2533–2539 2537port, c-Al2O3, has a defect spinel structure in which not all of alumina lattice28 but the present study clearly excludes the presence of CoAl2O4 in all of the reduced samples (B, C and the cation sites are occupied, i.e.Al21Hh22IO32, where h designates vacant tetrahedral and octahedral sites.53 If cobalt D) (see Fig. 3 and 5) because the Co,Co distance in that compound is considerably longer (2.83 A° ).41 Instead, it is is distributed as Co(II) over the vacant tetrahedral sites in alumina (one third of the total vacant sites) then this would evident that only a small fraction of the total cobalt contents of samples B and C are randomly dispersed over vacant yield AlIII21HCoIIU~V O32U~V , with the cobalt content being 4.5%, a value which is close to that found (3.8%) in the nonmetallic tetrahedral sites of the alumina support.This conclusion is supported by an X-ray diVraction study57 which also phase.Thus, the data for the high and low-cobalt loadings are shows that CoAl2O4 is not formed and additionally by the observation28 that much higher calcination temperatures consistent with 3–4% cobalt being randomly dispersed over vacant tetrahedral sites of the alumina support. Fig. 7 shows (>1200 °C) are needed to form CoAl2O4 from mixtures of Co3O4 and Al2O3. the EXAFS and the magnitude of the Fourier transform together with the parameters obtained from the least-squares refinement.The Co–O bond distance (1.96±0.01 A° ) is typical Reduction at 200 °C for tetrahedral coordinated cobalt(II),44 a similar value is also HuVman et al.33 reported that reduction of potassium- obtained for sample B (see above) and the pre-edge peak in promoted Co3O4/Al2O3 at 200 °C yields cobalt-metal particles that material’s spectrum is consistent with some of the cobalt in which the coordination numbers are less than those for being incorporated in the support in a similar manner.bulk cobalt metal. On this basis they estimated the average The XANES spectrum and its derivative of the particle size to be as small as 10–20 A° although they doubted cobalt-incorporated support (designated sample D-Co and this estimate noting that the actual size must be larger.Unlike obtained by removing the cobalt metal component from the the present work, no other cobalt-containing phase was spectrum of sample D) are shown in Fig. 2 and 3. There are detected. The present study also diVers in another respect since marked similarities with the XANES of CoAl2O4. Like the it shows that the rhenium-promoted Co3O4/Al2O3 catalyst is latter the derivative spectrum is also consistent with a minor not reduced at 200 °C even when reacted with hydrogen for amount of cobalt entering some of the vacant octahedral sites as long as 13 h.of the bearer.44 It has been assumed that CoAl2O4 is the phase generated when cobalt occupies the vacant tetrahedral positions of the The role of rhenium In a temperature-programmed reduction (TPR) study of the mechanism of rhenium promotion of alumina-supported cobalt Fischer–Tropsch catalysts, Holmen et al.31 reported that direct contact between rhenium and cobalt particles does not appear to be necessary for promotion.It is suggested that the mechanism by which rhenium promotes the reduction of Co3O4 is by hydrogen spillover.(The influence that rhenium exerts with regard to reduction and oxidation is mentioned above.) They also reported that cobalt diVuses into the support during the reduction and not during calcination which agrees with our findings. The increased dispersion that they found can account for the fact that the reduced rhenium-promoted catalyst relative to the unpromoted catalyst is consistent with the small particle sizes found in this XAS study.Conclusion XAS shows that high temperature reduction (450 °C) by 100% hydrogen of rhenium-promoted Co3O4/Al2O3 yields sample B which contains highly dispersed cobalt in the form of small (<40 A° ) metal particles (Co–Co distance 2.49±0.01 A° ) together with a minor fraction of cobalt atoms that are randomly spread over the tetrahedral vacancies of the alumina support (Co(II )–O distance 1.92±0.01 A° ).The contribution to the XAS from this phase was isolated by reducing a catalyst in which essentially all of the cobalt entered the alumina support (sample D). These results are consistent with those of Holmen et al.29–31 who found that the metal particles constitute Fig. 7 (Top) k3-Weighted experimental and least-squares fitted 78% of the cobalt content. Also in agreement is the finding EXAFS of sample D adjusted for 0.77% cobalt metal (see main text). that cobalt diVuses into the support during reduction and not (Bottom) The magnitude of the Fourier transform (FT). The solid during calcination and that this phase is resistant to reduction line shows the experimental data and the broken line represents the at 450 °C and is not the spinel CoAl2O4. calculated EXAFS with its corresponding FT using a single Co–O A particularly interesting feature of the XANES of sample shell.The final round of refinement yielded the following: distance B is the 3–4 eV shift of the K-edge to higher energy relative RCoMO=1.963(3) A° , coordination number N=3.2(2), and Debye– Waller-like factor A=2s2=0.0010(9) A° 2, and the refined correction to bulk cobalt metal.This, together with the reduced EXAFS to the threshold energy E0=21.98 eV for an R-factor=54.3%. The amplitudes, is attributed to an enhanced dispersion of cobalt standard deviation in the last significant digit as calculated by in the reduced material relative to the large bulk-like crystallites EXCURV90 is given in parentheses.However, note that such estimates of Co3O4 the calcined product. This has been attributed to of precision (which reflect statistical errors in the fitting) overestimate diminished shielding of the 1s electrons by the valence electrons the accuracy. The estimated errors for distances are 0.01 A° at R<2.5 A° which attends this heightened dispersion.35 The higher disper- with 20% accuracy for N and A, although the accuracy for these is increased by refinements using k1 vs.k3 weighting.44 sion and smaller particle sizes in sample B expresses the 2538 J. Mater. Chem., 1998, 8(11), 2533–253923 C. Bai, S. Soled, K. Dwight and A. Wold, J. Solid State Chem., positive eVect that rhenium-promotion has on the reducibility 1991, 91, 148. of sample A. 24 M. J. Fay, A. Procter, D. P. HoVmann, M. Houalla and No diVerence was observed when graphite was mixed with D. M. Hercules, Appl. Spectrosc., 1992, 46, 345. the catalyst instead of boron nitride. It is therefore clear that 25 E. A. Blekkan, H. Holmen and S. Vada, Acta Chem. Scand., 1993, carbon does not play a role here in transforming the metal 47, 275. 26 M. Shirai, T. Inoue, H. Onishi, K. Asakura and Y. Iwasawa, from the hexagonal or face-centred cubic structures to the J. Catal., 1994, 145, 159. metastable body-centred form (ref. 35 and refs. therein). 27 A. Kogelbauer, J. C.Weber and J. G.Goodwin, Catal. Lett., 1996, 34, 259. 28 P. G. Dimitrova and D. R. Mehandjiev, J. Catal., 1994, 145, 356.Support from the Nansen Foundation, the Norwegian 29 S. Vada, A. HoV, E.A° dnanes, D. Schanke and A. Holmen, Top. Research Council (including a NATO Postdoctoral Fellowship Catal., 1995, 2, 155. to A. Moen) and VISTA-Statoil is much appreciated. The 30 D. Schanke, A. M. Hilmen, E. Bergene, K. Kinnari, E. Rytter, preliminary work (contribution No. 98-9) was carried out at E.A° dnanes and A. Holmen, Catal. Lett., 1995, 34, 269. the Swiss-Norwegian Beamline (SNBL) for which we thank 31 A. M. Hilmen, D. Schanke and A. Holmen, Catal. Lett., 1996, 34, 143. the Norwegian University of Science and Technology and the 32 R. B. Greegor, F. W. Lytle, R. Chin and D. M. Hercules, J. Phys. Norwegian Research Council for grants towards its construc- Chem., 1981, 85, 1232.tion. We also thank the European Synchrotron Radiation 33 G. H. HuVman, N. Shah, J. Zhao, F. E. Huggins, T. E. Hoost, Facility for beamtime (CH-440) on the SNBL. The assistance S. Halvorsen and J. G. Goodwin, J. Catal., 1995, 151, 17. of the rest of the SNBL Project Team (K. Knudsen, P. Pattison 34 K. Takeuchi, T. Hanaoka, T. Matsuzaki, M. Reinikainen and and H.P.Weber) is also much appreciated and we thank Y. Sugi, Catal. Lett., 1991, 8, 253. 35 A. Moen, D. G. Nicholson, B. S. Clausen, P. L. Hansen, Professor Anders Holmen for providing us with some of A. Molenbroek and G. SteVensen, Chem.Mater., 1997, 9, 1241. the samples. 36 International Tables for X-Ray Crystallography, Kynoch Press, Birmingham, 1962, vol. 3, 175. 37 W. Lytle, R.B. Greegor, E. C. Marques, D. R. Sandstrom, References G. H. Via and J. H. Sinfelt, J. Catal., 1985, 95, 546. 38 EXCALIB, EXBACK and EXCURV90 programs, N. Binsted, 1 F. Solymosi, Catal. Rev., 1968, 1, 233. J. W. Campbell, S. J. Gurman and P. C. Stephenson, SERC 2 M. LoJacono and M. Schiavello, Preparation of Catalysts, ed. B. Daresbury Laboratory, 1990. Delmon, P.A. Jacobs and G.Poncelot, Elsevier, Amsterdam, 39 S. J. Gurman, N. Binsted and I. Ross, J. Phys. C: Solid State 1976, p. 473. Phys., 1984, 17, 143. 3 R. C. Reuel and C. H. Bartholomew, J. Catal., 1984, 85, 78. 40 F. A. Cotton, L. M. Daniels, C. A. Murillo and J. F. Quesada, 4 J. H. Ashley and P. C. H. Mitchell, J. Chem. Soc., A, 1968, 2821. Inorg. Chem., 1993, 32, 4861. 5 G. N. Asmolov and O. V.Krylov, Kinet. Katal., 1971,12, 463. 41 K.Toriuma, M. Ozima, M. Akaogi and Y. Sato, Acta Crystallogr., 6 H. Ueda and N. Todo, J. Catal., 1972, 27, 281. Sect. B, 1978, 34, 1093. 7 M. LoJacono, J. L. Verbeek and G. C. A. Schuit, J. Catal., 1973, 42 S. Sasaki, K. Fujino, and Y. Takeuchi, Proc. Jpn. Acad., Ser. B, 29, 463. 1979, 55, 43. 8 J. Grimblot, J. P. Bonnelle and J. P. Beaufils, J.Electron Spectrosc. 43 N. N. Greenwood and A. Earnshaw, Chemistry of the Elements, Relat. Phenom., 1976, 8, 473. Pergamon, Oxford, 1984, p. 1294. 9 R. I. Declerck-Grime�e, R. M. Caresson, R. M. Friedman and 44 Arild Moen, David G. Nicholson, M. Rønning, Geraldine J. J. Fripiat, J. Phys. Chem., 1978, 83, 885. M. Lamble, Jyh-Fu Lee and Hermann Emerich, J. Chem. Soc., 10 P. Dufresne, J. Grimblot and J. P. Bonnelle, Bull. Soc. Chim. Fr., Faraday Trans., 1997, 93, 4071. 1980, 1, 89. 45 A. Moen, D. G. Nicholson, M. Rønning, J. Chem. Soc., Faraday 11 K. S. Chung and F. E. Massoth, J. Catal., 1980, 64, 320. Trans., 1995, 91, 3189 and refs. therein. 12 K. S. Chung and F. E. Massoth, J. Catal., 1980, 64, 332. 46 M. Borowski, J. Phys. IV, 1997, 7, C2, 259. 13 R. B. Greegor, F. W. Lytle, R. L. Chin and D. M. Hercules, 47 G. Apai, J. F. Hamilton, J. Stohr and A. Thompson, Phys. Rev. Lett., 1979, 43, 165. J. Phys. Chem., 1981, 85, 1232. 48 H. Toepsoe, L. B. Hansen, P. Stoltze and J. K. Norskov, Catal. 14 H. Topsøe, B. S. Clausen, R. Candia and C. Wivel, J. Catal., 1981, Today, 1994, 21, 49. 68, 433. 49 S. Diaz-Moreno, D. C. Koningsberger and A. Munoz-Paez, Nucl. 15 N. Topsøe and H. Topsøe, J. Catal., 1982, 75, 354. Instrum. Methods Phys. Res., Sect. B, 1997, 133, 15. 16 R. L. Chin and D. M. Hercules, J. Phys. Chem., 1982, 86, 360. 50 A. M. Hilmen, D. Schanke and A. Holmen, Stud. Surf. Sci. Catal., 17 I. Alstrup, I. Chokendorft, R. Candia, B. S. Clausen and 1997, 107, 237. H. Topsøe, J. Catal., 1982, 77, 397. 51 Exxon Research & Engineering Co., US Pat. 5 227 407, 1993. 18 C. Wivel, B. S. Clausen, R. Candia, S. Mørup and H. Topsøe, 52 Exxon Research & Engineering Co., Eur. Pat. 0 339 923 B1, 1989. J. Catal., 1984, 87, 497. 53 N. N. Greenwood and A. Earnshaw, Chemistry of the Elements, 19 P. Arnoldy and F. E. Massoth, J. Catal., 1985, 93, 38. Pergamon, Oxford, 1984, p. 280. 20 M. A. Stanick, M. Houalla and D. Hercules, J. Catal., 1987, 54 S. E. Colley, R. G. Copperthwaite and G. J. Hutchings, Catal. 103, 151. Today, 1991, 9, 203. 21 E. Garbowski, M. Guenin, M. Marion and M. Primet, Appl. Catal., 1990, 64, 209. 22 H. C. Tung, C. Yeh and C. T. Hong, J. Catal., 1990, 122, 211. Paper 8/04261F J. Mater. Chem., 1998, 8(11), 2533–2539 25