Redox properties of tungstated zirconia catalysts: Relevance to the activation of n-alkanes

Abstract

A tungstated zirconia catalyst containing approximately enough tungsten oxide to give a theoretical monolayer on the zirconia support was characterised by Raman spectroscopy, temperature-programmed reduction, EPR and IR spectroscopies. The data show that treatment of the catalyst in H₂ at temperatures as low as 470 K led to partial reduction; EPR spectra indicated the formation of W⁵⁺ centres in two locations: in the bulk of tungstate clusters dispersed on the zirconia surface and at the surfaces of these clusters and consequently accessible to gas-phase reactants. When 1% n-pentane in N₂ flowed through a bed of particles of the catalyst in an EPR tube at 523 K for 20 min, the following surface species formed: W⁵⁺ centres (shown by EPR); organic radicals (shown by EPR); and OH groups (shown by IR). The results indicate that n-pentane undergoes a homolytic C–H bond cleavage reaction such as the following

followed by one-electron transfer steps that yield surface W⁵⁺ ions, OH groups, and chain carriers in the catalytic isomerisation of n-pentane, inferred to be carbenium ions. These processes are considered to be non-catalytic redox initiation reactions that explain the promoter role of the tungsten in the catalyst.

---

Introduction

Sulfated zirconia (SZ) catalysts have attracted wide interest because of their high activities and selectivities for isomerisation of n-alkanes at low temperatures (300–423 K).^1–4 Catalysts related to these have recently found industrial application for isomerisation of C₅–C₆ alkanes.⁵ The high activities of these catalysts were attributed to superacidic properties of SZ,^6,7 but recent research indicated that the acid strength of SZ is much less than that of a superacid and similar to that of zeolite HY, for example.^8,9 Thus properties other than just acid strength, have been inferred.^10–12

Alkane isomerisations are acid-catalysed chain reactions proceeding by carbenium-ion (or carbenium-ion-like) intermediates. A key question is how these intermediates are formed, i.e., how the reaction is initiated. One possibility is protonation of an alkane to give a carbonium ion (as occurs by Olah-type chemistry in superacidic solutions¹³), which is readily converted into a carbenium ion. Another possibility is formation of a carbenium ion by H⁻ abstraction from the alkane on Lewis acid sites or by adsorption of an alkene onto a Brønsted acid site. As the catalysis occurs even when alkene impurities are removed from the feed, it has been inferred that alkenes may be formed by dehydrogenation of the alkane;^8,12,14–17 another key question is how this occurs.

Farcasiu et al.¹⁸ suggested that SZ is bifunctional, exhibiting both acidic and redox properties, and that alkane isomerisation might be initiated by oxidation of the alkane by sulfate, leading to carbenium ion chain carriers.¹⁸ Consistent with this suggestion, oxidizing properties of SZ have been observed,^19,20 and Ng and Horvath²¹ suggested that sulfate is reduced during n-butane isomerisation, either by n-butane or by H₂ product. Quantum-chemical calculations suggested that dehydrogenation processes are feasible on SZ, with Zr–O pairs being the redox sites.¹²

SZ catalysts suffer from the disadvantages of rapid deactivation and possibly from sulfur loss during reaction and regeneration. Platinum or metal oxides such as those of iron and/or manganese may be added to SZ catalysts to improve the activity, selectivity, and stability.^8,19,22–24

As an alternative to SZ, tungstated zirconia (WZ), reported by Hino and Arata⁶ to catalyse n-pentane isomerisation at low temperatures, has drawn wide attention recently.^25–28 Although tungstated oxides are typically markedly less active than sulfated oxides, they are appealing alternatives because they are typically more stable than sulfated oxides. WZ catalyses isomerisation of n-pentane and of n-hexane at low temperatures.⁶

The catalytic properties of WZ have been related to its ability to form surface W⁵⁺OH centres under reducing conditions at low temperatures. These centres were suggested to stabilise carbenium ion intermediates by delocalisation of the corresponding negative charge among several oxygen atoms.^28,29 The reduction of W⁶⁺ centres in WO_x clusters on WZ promoted with Pt (Pt/WZ) was inferred on the basis of H₂, O₂ , and CO titration data.²⁸ Hydrogen uptake exceeds CO chemisorption capacity, which suggests that hydrogen adatoms spill over onto WO_x clusters during H₂ chemisorption at room temperature. On the other hand, O₂ uptakes are much higher than H₂ or CO uptakes, which suggests that O₂ reoxidises reduced WO_x species. The reduction of W⁶⁺ in WZ and in Pt-promoted WZ has been monitored in temperature-programmed reduction (TPR) experiments,^27–30 which show that on WZ, higher WO_x surface densities lead to lower reduction temperatures, presumably because of the ability of WO_x to delocalise the net negative charge onto larger WO_x clusters. The addition of Pt leads to lower reduction temperatures, evidently because of spillover of hydrogen adatoms (formed by H₂ dissociation on Pt) onto condensed WO_x clusters. The accommodation of a proton by electron abstraction from a hydrogen atom and charge delocalisation on an extended WO_x network were proposed;^27–29 these steps could lead to structures suggestive of tungsten bronzes. The formation of tungsten bronzes has been demonstrated,^31–33 but there is still no direct spectroscopic evidence of the formation of W⁵⁺ centres in supported tungstate catalysts.

Our goal was to test for and monitor the formation of W⁵⁺ centres on WZ catalysts by EPR spectroscopy used in combination with TPR experiments. The formation of Brønsted acid sites by reduction at high temperatures was monitored by IR spectroscopy, and the strengths of the acidic sites were estimated on the basis of IR spectra of CO adsorbed at low temperature. Laser Raman spectroscopy was also used for structural characterisation of the WZ.

Experimental methods

Catalyst preparation

WZ samples were prepared by impregnation of amorphous Zr(OH)₄ (MEL Chemicals) with an aqueous solution of ammonium metatungstate ((NH₄)₆H₂W₁₂O₄₀·nH₂O) (Aldrich)).²⁵ The resultant suspension was refluxed overnight at 393 K and dried in an oven at 353 K prior to calcination. The samples were then calcined at 923 K in static air for 3 h.

The concentrations of tungsten in the solutions used to prepare samples were chosen to give tungsten loadings on the catalyst near and below the theoretical monolayer (monolayer capacity is ca. 19 wt.% WO₃ based on the molecular area of a WO₃ unit). Most experiments were carried out with catalysts containing 17 wt.% WO₃. A catalyst containing 19 wt.% WO₃ was used for low-temperature IR spectroscopy experiments. A tungstated zirconia sample containing x wt.% tungsten as WO₃ is denoted xWZ.

The WZ samples contained 6 or 17 wt.% WO₃ (and a sample used for IR spectroscopy contained 19 wt.% WO₃). The surface areas are summarised in Table 1.

Table 1 Properties of calcined catalyst samples

Catalyst	Loading (wt.% WO3)^a	Loading /molecule nm^−2	S BET/m^2 g^−1	Phase^b
a (g of WO3)/100 g sample. b Determined by Raman spectroscopy.
17WZ	17.7	3.47	132	tetragonal
6WZ	6.0	2.48	63	tetragonal, monoclinic

---

Raman spectroscopy

Raman spectra were obtained with a modified Dilor OMARS 89 spectrometer with a notch filter to suppress the Rayleigh line. The detector system was a thermoelectrically cooled CCD Camera (Princeton Instruments). An Ar⁺ ion laser (Spectra Physics, model series 2020) was used as an excitation source. The 488 nm line was selected for excitation. Spectra were recorded by use of the scanning multichannel technique^34,35 with a scanning time of 10 s per step and with a slit width of 150 μm, which gave a spectral resolution of 5 cm⁻¹. A laser power of 3 mW was employed for the samples measured under ambient conditions; the power was 30 mW for dehydrated samples. The samples were dehydrated in the Raman cell at 673 K for 20 min in flowing dry O₂.

Temperature-programmed reduction

TPR measurements were performed by use of a system equipped with a thermal conductivity detector. Sample powder (approximately 200 mg) was pretreated in the reactor in a flow of O₂ at 673 K for 1 h. The sample was then cooled to room temperature in a flow of dry N₂, followed by heating to 1073 K at a rate of 10 K min⁻¹ in a flow of 5 vol.% H₂ in N₂. The total gas flow through the sample was 40 ml min⁻¹. The parameters were chosen such that the K number defined by Monti and Baiker³⁶ had a value of 140 and was therefore in the permitted range. The reduction signal was calibrated by using the complete reduction of a standard CuO powder (Aldrich, 99.995%).

EPR spectroscopy

Measurements were carried out with a Varian E-line (E9) spectrometer equipped with a TE₁₀₄-mode double cavity and/or with a Bruker ESP 300E spectrometer. The spectra were recorded in X-band at room temperature (298 K) and at low temperature (80 K) with a microwave power of 10 mW; g-value determination was done with the WINEPR Simfonia Program from Bruker. The EPR tube was connected to a U-shaped reactor. Prior to each experiment, a sample was pretreated in the reactor in a flow of O₂ at 673 K for 1 h, and then transferred into the EPR tube. The reduction was carried out in a flow of H₂ at 473, 573, or 673 K for 1 h. A fresh sample was used for each reduction experiment.

IR spectroscopy

IR spectra were recorded with a Bruker IFS-66 IR spectrometer equipped with a liquid-N₂-cooled MCT detector. Self-supporting wafers (ca. 15–20 mg cm⁻²) were pretreated in a flow of dry O₂ (30 ml min⁻¹) at 673 K for 12 h in home-made IR cells.

The conditions of the high-temperature measurements were comparable to those used during a catalytic isomerisation reaction. Pretreatment and reduction were carried out in a high-temperature cell that allowed measurements at elevated temperatures. After pretreatment, each sample was cooled to room temperature in a flow of dry N₂ and then reduced for 1 h at 573 K in a flow of dry H₂ (20 ml min⁻¹). Spectra were then recorded at 573 K.

Low-temperature measurements were carried out to characterise surface OH groups formed by reduction. Since adsorbed water gives intense bands in the O–H stretching region, the samples were thoroughly evacuated after pretreatment to dehydrate them fully. To produce greater amounts of reduced species, the samples were reduced at 673 instead of 573 K. The reduction of a sample was carried out in the cell: (i) in a H₂ flow (20 ml min⁻¹) for 1 h, or (ii) by stepwise reduction in 4 kPa of D₂ for 20 min followed by evacuation at 673 K for 5 min. CO was adsorbed on the reduced samples at low temperature (85 K) and at increasing equilibrium pressures p in the range 0⩽p/kPa⩽25.

Results

Structural characterisation by Raman spectroscopy

The Raman bands characterising crystalline ZrO₂ appear in the range 150–700 cm⁻¹,³⁷ and the bands characterising the monoclinic and tetragonal phases provide a convenient basis for distinguishing the two phases. A summary of the ZrO₂ phase compositions determined by Raman spectroscopy is given in Table 1.

The region with energies ⩾700 cm⁻¹ is free of bands representing bulk crystalline ZrO₂ . This region of the spectrum allows detection of the high-frequency bands of oxide compounds with octahedrally and tetrahedrally coordinated tungsten.³⁸ The Raman spectrum of 17WZ in the region between 150 and 700 cm⁻¹ (Fig. 1a) is characteristic of poorly crystallised tetragonal ZrO₂ . Bands characteristic of monoclinic ZrO₂ were not observed. The presence of tetragonal ZrO₂ in 17WZ is in accord with reports showing that tungsten oxide inhibits the crystallisation of the thermodynamically more stable monoclinic phase and stabilises the high-temperature tetragonal modification of ZrO₂ when the synthesis is carried out by impregnation of amorphous Zr(OH)₄.^25,39

Fig. 1 Raman spectra in the low-frequency region of samples 17WZ (a) and 6WZ (b).

The spectrum of 6WZ (Fig. 1b) includes only two bands of tetragonal ZrO₂ (at 278 and 319 cm⁻¹) with relatively low intensities, whereas the bands representing the monoclinic phase are very intense in this spectrum. The stabilisation of the tetragonal phase was found to be less efficient at WO₃ loadings significantly lower than that corresponding to monolayer coverage, as was observed earlier.²⁵

Earlier work²⁵ showed that a WO_x overlayer consisting of tungstate clusters is formed on the surface of ZrO₂ without the formation of detectable three-dimensional WO₃ crystallites when WZ is synthesised by the impregnation method with loadings less than about 20 wt.% WO₃ and when it has a surface area of about 100 m² g⁻¹. Thus, we infer that the sur face of 17WZ was largely covered by a tungstate overlayer (which is denoted as WO_x to distinguish it from crystalline WO₃).

Raman spectra of the samples in the range 700–1100 cm⁻¹ are shown in Fig. 2 for the hydrated and dehydrated samples. Crystalline WO₃ is known to show intense and characteristic Raman bands at 807, 715 and 274 cm⁻¹,⁴⁰ which are absent from the spectra of 17WZ and 6WZ; this result indicates that microcrystalline WO₃ was not formed in these samples.

Fig. 2 Raman spectra in the high-frequency region of the samples 6WZ (a) and 17WZ (b) under ambient conditions (hydrated) and of 6WZ (c) and 17WZ (d) at 573 K in O₂ (dehydrated).

The Raman spectrum of dehydrated 17WZ (Fig. 2d) is characterised by a strong, sharp signal at 1020 cm⁻¹, assigned to W[double bond, length half m-dash]O stretching modes, and a broad band at 830 cm⁻¹, typical of W–O–W stretching modes.³⁸ The signal at 910 cm⁻¹ falls outside the typical range of W–O–W vibrations (800–870 cm⁻¹) and is attributed to W–O–Zr vibrations.²⁴ The presence of adsorbed water (Fig. 2b) leads to a shift of the band of the W[double bond, length half m-dash]O groups to lower frequencies, caused by their strong interaction with water by hydrogen bonding and resulting in a broad band pair at 952 and 976 cm⁻¹ (Fig. 2b). The band at 830 cm⁻¹ remained unchanged in the presence of water, whereas the band at 910 cm⁻¹ became hidden because of the overlap with the broad signals at 830 and 952 cm⁻¹. The presence of the broad band at 830 cm⁻¹ indicating W–O–W stretching modes shows that the WO_x surface structures consist of poorly defined oligomeric or polymeric species that are perturbed by their interactions with the ZrO₂ surface. The observed spectra are essentially identical to those reported²⁴ for analogously prepared materials. We therefore infer that the WO_x overlayer consists of condensed oligomeric and/or (probably three-dimensional) polymeric WO_x surface clusters that interact with tetragonal ZrO₂ and thereby cover the support surface.

The spectrum of sample 6WZ in the dehydrated state (Fig. 2c) exhibits a strong Raman band at 1011 cm⁻¹ with a shoulder at 1022 cm⁻¹, characterising W[double bond, length half m-dash]O vibrations, and two broad bands, at 830 and 874 cm⁻¹, attributed to W–O–W and W–O–Zr stretching vibrations, respectively. It has been reported^25,41 that the amount of tungsten affects the Raman shift of the W[double bond, length half m-dash]O vibration. On the basis of the results described above indicating bands of the W[double bond, length half m-dash]O mode at about 1020–1018 cm⁻¹ for large (possibly three-dimensional) clusters of WO_x on the support, we infer that the W[double bond, length half m-dash]O mode at 1011 cm⁻¹ can be assigned to smaller clusters of WO_x present at lower loadings.

Reduction behaviour of tungstated zirconia

TPR. TPR profiles characterising 17WZ and 6WZ are shown in Fig. 3. According to the literature,⁴² pure WO₃ exhibits three reduction peaks, namely, a shoulder 911 K (WO₃ → W₂₀O₅₈), a sharp peak at 1038 K (W₂₀O₅₈ → WO₂) and a peak at higher temperatures (WO₂ → W). Pure ZrO₂ does not show any detectable TPR peak at temperatures below 1270 K. Thus, we conclude that reduction peaks observed for WZ catalysts at temperatures below 1000 K can be attributed to the reduction of WO_x species interacting with the ZrO₂ support.

Fig. 3 TPR profiles of samples (a) 17WZ and (b) 6WZ.

The reduction in H₂ occurs at higher temperatures for 6WZ than for 17WZ (Fig. 3), probably because of a greater number of strong W–O–Zr bonds in samples with the lower coverage; the TPR profiles characterising 6WZ show a single reduction peak at 790 K with a broad shoulder at high temperature extending to ca. 950 K. In contrast, at higher tungsten loadings, the tungsten present in W–O–W groups is more easily reduced than that in W–O–Zr groups, consistent with reported results.²⁸

Data summarising the H₂ consumption and positions of the reduction peaks are given in Table 2. The observation of two signals for 6WZ (at 770 K, with a shoulder at 800 K) may be attributed to the presence of two WO_x species with slightly different reducibilities, consistent with the Raman spectrum (Fig. 2). The data indicate increasing H₂ consumption with increasing temperature (with an onset at about 900 K) for 17WZ. This observation may be explained by a further reduction of the tungsten species at the higher temperatures.

Table 2 Results of TPR experiments

Catalyst	Mass/mg	W content/mol	H2 consumption/mol	Temperature of reduction/K	H2:W molar ratio
17WZ	197	7.50 × 10^−5	1.1 × 10^−5	719	0.07
6WZ	176	—	<10^−5	766; 797	—

---

EPR. The EPR spectra of WZ samples were recorded at 80 and at 298 K, with the following results:

(a) Calcined samples: the calcined WZ samples exhibit an EPR signal at g = 4.24 (not shown), which can be assigned to isolated Fe³⁺ ions in octahedral or tetrahedral symmetry with strong rhombic distortion.^43,44

(b) Reduced samples: the EPR spectra of 17WZ after reduction in H₂ at 470, 570 and 670 K are shown in Fig. 4A–C and in Fig. 5.

Fig. 4 A, EPR spectra (X-band) of sample 17WZ after reduction at 470 K (spectra recorded at 298 (a) and 80 K (b)). B, EPR spectra (X-band) of sample 17WZ after reduction at 570 K (spectra recorded at 298 (a) and 80 K (b)). C, EPR spectra (X-band) of sample 17WZ after reduction at 670 K (spectra recorded at 298 (a) and 80 K (b)).

Fig. 5 EPR spectra (X-band) of sample 17WZ after reduction at 470 (a), 570 (b), and 670 K (c) (normalised gain) (spectra recorded at 80 K).

The spectra are complex. They can be analysed on the basis of (1) the evolution of the individual component signals as a function of the recording temperature, (2) the variation of the signal intensity as a function of the reduction temperature and (3) the magnetic parameters. Thus, three different component signals can be distinguished in the spectra. The first, present only when the sample had been reduced in H₂ at 470 K, was observed at g≈g_e and is temperature dependent, being barely observed at 298 K (Fig. 4Aa) but becoming clearly visible at 80 K (Fig. 4Ab). A possible explanation for this result is that it is an indication of nitrogen-containing paramagnetic impurities.

Ammonia used in the preparation of TiO₂ can lead to an Ostwald-type reaction:⁴⁵

with NO reacting further, as follows:

The paramagnetic species produced (NO, NO₂ and NO₂²⁻), especially when present simultaneously, lead to complex EPR patterns because of the hyperfine structures attributed to the presence of ¹⁵N atoms with nuclear spin I = 1. The signal at g≈g_e was no longer observed when the samples were reduced at 570 or 670 K, suggesting that the nitrogen-containing impurities were reduced to give N₂ at those temperatures. Because it is only a minor contribution to the complex EPR spectra of Fig. 4A–C, this signal is not discussed further.

From the change of signal intensity as a function of the reduction temperature (Fig. 5), a second EPR signal component can be distinguished, the intensity of which increased with increasing reduction temperature. The EPR signal clearly visible at 80 K was scarcely detected at 298 K, which implies that the corresponding paramagnetic species have a short T₁ relaxation time. The signal has a distinct g-tensor component at g_⊥ = 1.98. The other component can be taken as g_‖ = 1.93, although an orthorhombic symmetry or the presence of several different sites cannot be ruled out, particularly because of the overlap with the third signal at high magnetic field.

EPR signals with similar g values have been reported for ZrO₂. Torralvo et al.⁴⁶ observed a signal at g_⊥ = 1.981 and g_‖ = 1.956, which, because of its insensitivity to the presence of oxygen at room temperature, they assigned to Zr³⁺ ions located in the bulk of the material. Morterra et al.⁴⁷ observed a signal at g_⊥ = 1.978 and g_‖ = 1.953 (along with another g_⊥ = component at 1.975), which, because of the reversible broadening of the signal upon adsorption–evacuation of oxygen, they assigned to Zr³⁺ ions in a state of low co-ordination at the surface of ZrO₂. Consistent with these results, we assign the signal at g_⊥ = 1.98 and g_‖ = 1.93 to Zr³⁺ ions. Because the latter appear upon thermal reduction of our sample, we infer that the Zr³⁺ ions were likely located at the catalyst surface.

The third component signal was observed at a relatively high magnetic field (Fig. 5). It is inferred to be a composite signal because its shape changed as the temperature of reduction in H₂ increased. A weak anisotropic signal at g_⊥ = 1.84 and g_‖ = 1.51 was observed at 77 K after the sample had been reduced at 470 K (Fig. 5a). We suggest that this signal is associated with coordinatively unsaturated reduced species produced at the catalyst surface. The signal increased in intensity as the reduction temperature increased, leading to reduced species located more in the bulk of the tungstate clusters than at the surface. The signal observed for the samples reduced at 570 (Fig. 5b) and 670 K (Fig. 5c) is a superposition of several components, with the most intense at g_⊥ = 1.83 and g_‖ = 1.58, with other components appearing at g_⊥ = 1.84 and 1.81 and g_‖ = 1.64 and 1.51.

The composite signal was observed both at 80 and at 298 K (Fig. 4A–C), although it was slightly broadened at the latter temperature, corresponding to a decreased intensity; its detection was difficult for the sample reduced at 470 K in H₂ (Fig. 4Aa). The data also show that, other things being equal, the intensity of the signal increased with increasing tungsten loading in the sample.

There is little literature of EPR signals representing W⁵⁺ species in or on oxide matrices. Reduction of WO₃ by hydrogen, produced, for example, by reaction of Zn⁰ with HCl, leads to the formation of hydrogen tungsten bronzes, H_xWO₃, with the type of structure depending on the value of x.³¹ The observation of W⁵⁺ in such samples by EPR spectroscopy is usually difficult, either because of too low a concentration of spins or too high a conductivity.⁴⁸ Schirmer and Salje⁴⁹ succeeded in observing the EPR signal of W⁵⁺ species in H_xWO₃—but only when the sample was irradiated with light.

The EPR of tungsten impurities W⁵⁺ in various samples was reviewed by Bravo et al.,⁵⁰ who reported that for oxide matrices such as WO₃,⁴⁹ TiO₂,⁵¹ SnO₂,⁵² and GeO₂,⁵³ the g-tensor components of octahedral W⁵⁺ are in the range 1.4–1.8, substantially lower than those of the isoelectronic d¹ ions such as V⁴⁺ and Mo⁵⁺, because of the large value of the spin–orbit coupling constant of tungsten.⁵⁰

On the basis of the g values observed for W⁵⁺ ions in octahedral WO₆⁷⁻ complexes in various matrices^49–53 and the foregoing results, the composite signal observed in the g value range of 1.84–1.51 can be assigned to W⁵⁺ ions (Table 3).

Table 3 EPR parameters characterising W⁵⁺ ions in various oxide matrices and in the WO_x/ZrO₂

Oxide matrix	Complex	g ^b	gx ^a	g ^b	gy ^a	gz ^a	Ref.
a Orthorhombic g-tensor components. b Axial g-tensor components. c g-tensor components of W^5+ coordinatively unsaturated surface ions. d g-tensor components of the most abundant W^5+ ions; some of them may be in octahedral complexes.
WO3	WO6^7−	1.554	1.685	1.566	48
TiO2	WO6^7−	1.473	1.443	1.594	50
SnO2	WO6^7−	1.671	1.500	1.732	51
GeO2	WO6^7−	1.708	1.554	1.791	52
WOx/ZrO2	WOx^n^−	1.51^c	1.84^c	this work
WOx/ZrO2	WOx^n^−	1.58^d	1.83^d	this work

---

Because these W⁵⁺ ions are produced as a result of thermal reduction, it is likely that they are first formed at the surface of a catalyst and lead to W⁵⁺ ions in WO_xⁿ⁻ complexes, with x possibly <6. To test this hypothesis, oxygen was adsorbed on the sample at room temperature, and the sample was then purged with gaseous N₂ to minimise line broadening by paramagnetic O₂ molecules. The intensity of the W⁵⁺ signal decreased as the components at g_⊥ = 1.84 and g_‖ = 1.51 decreased, disappearing first as a new orthorhombic signal with g_z = 2.028, g_y = 2.008 and g_x = 2.002 appeared (not shown), characteristic of adsorbed O₂⁻ ions produced [italic v]ia the following electron transfer reaction:

These results suggest that the reacting W⁵⁺ ions are located at the surface of the catalyst, because they are accessible to oxygen, whereas the remaining ones are not. From the g_z value of O₂⁻ (2.028), it can be deduced⁵⁴ that the O₂⁻ species are stabilised not on W⁶⁺ but rather on the Zr⁴⁺ support ions. A similar effect has already been reported for WO₃/Al₂O₃ by Spiridonov et al.⁵⁵ and for WO₃/SiO₂ by Howe.⁵⁶

EPR characterisation of used n-pentane isomerisation catalyst. The formation of W⁵⁺ during the catalytic isomerisation of n-pentane has been postulated by Iglesia and coworkers,^27–29 although there is no reported experimental evidence to test the postulate directly. To provide such a test, a catalyst was characterised by EPR before and after catalysis of n-pentane isomerisation in a flow reactor that was an EPR tube. The 17WZ catalyst was pretreated with O₂ at 673 K in the EPR tube, and data were recorded after a mixture of 1 vol.% n-pentane in N₂ flowed over the catalyst particles at 523 K and atmospheric pressure for 20 min, undergoing catalytic isomerisation. EPR spectra were taken at 298 and at 80 K with the freshly pretreated and used catalyst samples.

Except for the Fe³⁺ signal mentioned above, no signal could be detected at either temperature after the O₂ pretreatment. However, after the catalytic reaction, a symmetric signal with a g value of 2.00 was observed when the spectrum was recorded at room temperature (Fig. 6a). An anisotropic signal with g_⊥ = 2.004 and g_‖ = 2.002 has been assigned to electrons trapped in oxygen vacancies in the bulk of ZrO₂.⁴⁶ Because the signal in the catalyst was formed as a consequence of the reaction, we infer that it represents surface species. Thus, we infer that the signal originates from organic radicals built up during the catalytic reaction.

Fig. 6 EPR spectra (X-band) of sample 17WZ after 20 min in a flow reactor for catalytic isomerisation of n-pentane: (a) measured at 298 K, (b) measured at 80 K.

When the spectrum of the used catalyst was recorded at 80 K, a signal was observed at g_⊥ = 1.84 and g_‖ = 1.51 (Fig. 6b), and this corresponds exactly to the signal of the surface W⁵⁺ ions produced upon reduction of the catalyst with H₂ at 470 K (Fig. 4Ab).

The appearance of organic radicals combined with the appearance of W⁵⁺ can be explained by a one-electron transfer from n-pentane or a radical precursor to the catalyst surface.

IR spectroscopy. It has been suggested^27–29 that reduction of tungstated ZrO₂ with H₂ at elevated temperatures leads to the formation of new OH groups associated with W⁵⁺ and that these OH groups or the W⁵⁺ sites may play a crucial role in the mechanism of the isomerisation of small alkanes. OH groups coordinated to W⁵⁺ should be visible in the IR spectrum, and thus we recorded spectra to search for them.

Low-temperature IR experiments were performed with a 19WZ catalyst. Since the hydrated samples showed broad OH bands characterising adsorbed water, the samples first had to be dehydrated to allow observation of new OH groups present after reduction. Fig. 7a shows results for dehydrated 19WZ after exposure to flowing O₂ (10 ml min⁻¹ at 673 K for 12 h) followed by evacuation to a pressure ⩽10⁻³ mbar at 673 K; the deformation mode of molecular water near 1610 cm⁻¹ was no longer observed, which shows that the sample was fully dehydrated. As a result (and consistent with the Raman spectra), the W[double bond, length half m-dash]O stretching vibration shifted from low wavenumbers in the hydrated state to 1022 cm⁻¹ (not shown) when the water was removed.

Fig. 7 IR spectra measured at 298 K of sample 19WZ: (a) dehydrated; (b) reduced at 673 K with H₂; and (c) reduced stepwise at 673 K with D₂ and evacuated at the same temperature.

Although the sample was completely dehydrated, there were still two weak bands in the O–H stretching region (Fig. 7a): a sharp band at 3738 cm⁻¹, characteristic of ZrOH groups, and a broad band near 3620 cm⁻¹. The latter is absent from the spectrum of pure ZrO₂. We therefore infer that it represents WOH groups.

To investigate the formation of OH groups during reduction in H₂, the sample was treated in the cell at 673 K in flowing H₂. The resulting spectrum recorded at 298 K (Fig. 7b) shows a broad signal characterising new O–H stretching vibrations in the 3640–3100 cm⁻¹ region, with several maxima, at 3620, 3560 and 3505 cm⁻¹. The latter two bands were not observed in the spectrum of the sample prior to reduction (and have not been reported previously). It is therefore inferred that these bands characterise OH groups formed during reduction. As indicated by the appearance of the deformation vibration of water at 1617 cm⁻¹, the reduction also leads to the formation of water that remains on the surface. The adsorbed water leads to the broad absorption between 3650 and 3100 cm⁻¹. The relatively low signal/noise ratio is a result of the inherently high absorption of the sample in the O–H stretching region; it can be improved by using D₂ rather than H₂ since the O–D stretching modes occur in a more favourable spectral region at lower frequency.

Reduction in D₂ was carried out under static conditions and repeated several times at a pressure of 4 kPa. After each reduction step in D₂ at 673 K for 20 min, the sample was evacuated for 5 min at the same temperature. This procedure was repeated until molecular D₂O was no longer formed during the D₂ treatment, as confirmed by the disappearance of the DOD deformation band at 1194 cm⁻¹ (not shown).

The spectrum in the O–D stretching region at 298 K of the D₂-treated sample is shown in Fig. 7c. It exhibits a weak band at 2758 cm⁻¹ and a broad band between 2730 and 2560 cm⁻¹ superimposed by two sharp bands, at 2630 and 2589 cm⁻¹, and a shoulder at 2673 cm⁻¹. The isotopically shifted O–D bands at 2758, 2673, 2630 and 2589 cm⁻¹ correspond to the O–H bands shown in Fig. 7b, with positions at 3738, 3620, 3560 and 3505 cm⁻¹. The ratio of O–H and O–D stretching frequencies was found to be 1.35, which is close to the theoretical isotope shift calculated according to the diatomic harmonic oscillator approximation (1.41).

The acid strength of the newly formed OD groups was tested by IR spectroscopy of adsorbed CO at low temperature (data not shown). CO interacts with OD groups [italic v]ia OD–CO hydrogen bonds at 85 K and CO pressures between 0 and 0.25 kPa. The resulting O–D stretching frequency shift, which is a measure of the hydrogen bond strength, could not be determined quantitatively, but is ⩽100 cm⁻¹, which shows that the acid strength of the O–D (or O–H) groups is less than that of the WZ samples prior to reduction. Hence, we infer that strongly acidic Brønsted sites were not formed by reduction in H₂.

The observation of OH groups in the catalyst at room temperature raises the question of whether they are stable at elevated temperature or whether they react to give desorbed water leading to the formation of exposed W⁵⁺ centres. To test these possibilities, the reduction of 17WZ with H₂ at 573 K in the IR-cell was followed by IR spectroscopy, with the spectra recorded at the treatment temperatures.

When the sample was heated in flowing H₂, large amounts of water were formed, as indicated by an intense broad band in the spectrum between 3000 and 3750 cm⁻¹ and a strong band at 1617 cm⁻¹ (not shown). Further heating led to the desorption of most of the water, but some molecularly adsorbed water was still detectable during heating, even when the temperature reached 573 K. Reduction at 573 K for as long as 1 h led to no further change of the spectrum. The spectrum of the reduced catalyst is rather similar to that of the oxidised catalyst, failing to show any new features that would indicate the presence of OH groups formed by the reduction at high temperature. It was not possible to determine whether acidic groups were present at elevated temperatures after reduction.

Discussion

It is widely agreed that the mechanism of n-pentane isomerisation catalysed by solid acids involves intermediate carbenium ions (or related surface species, which, for simplicity, we refer to simply as carbenium ions) that undergo skeletal isomerisation and/or cracking, depending on the reaction temperature⁵⁷ and the average lifetime^23,27–29 of branched carbenium ions on the catalyst surface. Several authors have postulated that the increase in isomerisation selectivity resulting from addition of H₂ to the feed to a flow reactor is a consequence of hydrogen storage on the tungstate overlayer until hydrogen transfer to the carbenium leads to product desorption.^23,27–29 Shorter lifetimes of the carbenium ions would permit the products to desorb after skeletal rearrangement but before β-scission (cracking) could occur.^23,27–29 Consistent with this interpretation, it was shown^27–29 that H₂ is indeed chemisorbed on the catalyst, and it was proposed that the subsequent reduction of the tungsten-containing phase leads to formation of a bronze-like structure with W⁵⁺ centres, along with Brønsted acid sites.

In this context, important results of the present work are (a) the observation (by EPR spectroscopy) that the tungsten phase is indeed reduced by treatment with H₂ at temperatures of 470–670 K and that W⁵⁺ is formed; (b) the observation (by IR spectroscopy) that this reduction also leads to the formation of OH groups; and (c) the observation (by IR spectroscopy) that the OH groups are Brønsted acids (although they are not very strong).

It has been proposed ([italic v]ide supra) that the effective delocalisation of the negative charge in the WO_x clusters of the catalyst may be responsible for the acidity.^28,29 It may therefore be inferred that the importance of the reduction with H₂ is related not only to the hydrogen transfer from the catalyst surface to carbenium ions but also possibly to the strength of Brønsted acid sites created by the reduction. Nonetheless, a key issue about the fate of the protons at high temperature remains to be resolved—the quality of the high-temperature IR spectra is not sufficient to demonstrate the existence of the OH groups. Perhaps the protons are not localised at high temperature and move rapidly in the WO_x lattice, as they do in tungsten bronzes formed by reduction of WO₃ with H₂.³³ Such a fast proton motion might explain the lack of a clearly visible absorption indicative of OH groups in the IR spectrum of a H₂-treated tungsten bronze⁵⁸ and in our high-temperature IR spectra. Alternatively, the protons might react to form water that desorbs, thus creating coordinatively unsaturated W⁵⁺ centres.

These results still leave open the question of how the alkane isomerisation catalysis is initiated, i.e., how carbenium ions are initially formed. Low-temperature IR spectra of adsorbed CO²⁵ show that the acidity of WZ is clearly less than that of zeolites such as HZSM-5, which are only moderately strong acids. Thus, we discount the possibility that WZ protonates alkanes to form carbonium ions, which would be precursors of carbenium ions, as in Olah superacid chemistry.¹⁴ Carbenium ions are evidently formed otherwise, e.g., by reactions involving alkane dehydrogenation, as mentioned in the Introduction.

The results reported here favour the suggestion that alkenes are formed by alkane dehydrogenation in a redox process involving organic radical intermediates, as follows: The same EPR signal, with g_⊥ = 1.84 and g_‖ = 1.51, was produced, either upon reduction of the catalyst with H₂ at 470 K (Fig. 4Ab) or by the catalytic reaction of n-pentane at 523 K (Fig. 6b). The essential difference is that, in the latter case, another symmetric signal at g = 2.00 was also observed, which can be assigned to an organic radical.

An implication of the observations is that the catalyst must be reduced until the carbenium ions (perhaps formed by alkene protonation on the surface) react (e.g., by isomerisation or cracking) and are desorbed. On the basis of semiempirical quantum chemical calculations,⁵⁹ Haber postulated that in activation of a C–H bond at the surface of a transition metal oxide having semiconducting properties, both parts of the cleaved C–H bond become attached to the surface oxide ions: the protons form OH groups; the hydrocarbon fragments form alkoxy groups. Simultaneously, the two electrons originating from this bond are injected into the conduction band of the solid, according to the following schematic representation:

As an alternative to this activation mechanism involving a two-electron transfer process and leading to delocalised electrons, an alternative mechanism involving a one-electron process and homolytic cleavage of R–H bonds can now be proposed on the basis of the present results:

Such a reaction (or a similar reaction involving oxygen atoms of the surface) leads to the formation of organic radicals, consistent with the EPR signal observed at g = 2.00 (Fig. 6a). Consistent with this suggestion, it is known⁶⁰ that the radicals H^• are strong reducing agents that can lead to the reduction of transition metal ions, consistent with the observation of W⁵⁺. Specifically, we suggest the following:

This suggestion accounts for the observations reported here, including the formation of OH groups, which may react to give water that desorbs, leading to coordinatively unsaturated W⁵⁺ ions in surface complexes WO_xⁿ⁻.

Our interpretation is consistent with the earlier suggestion that radical formation occurs during alkane isomerisation catalysed by SZ.^61,62 Although our data are consistent with the inference that organic radicals and W⁵⁺ formed from n-pentane and were involved in the initiation of catalytic isomerisation of n-pentane, we lack direct evidence that organic radicals played a role in the catalysis. Indirect evidence is provided by the results of Chen et al.,⁶³ who investigated the Lewis acid strengths of different SZ catalysts using benzene as a probe molecule and found a correlation between the radical formation from benzene and the isomerisation activity of the catalysts. However, the ionisation potential of n-pentane (10.35 eV⁶⁴) is greater than that of benzene (9.24 eV⁶⁵), and so it remains to be determined whether a similar correlation would be found for radical formation from n-pentane.

We emphasise that the redox process and radical formation (suggested here to be important in alkane isomerisation catalysis) are themselves not regarded as part of a catalytic cycle. Rather, we envision a one-electron transfer initiation process that leads to formation of the carbenium ion chain carriers in the catalytic cycle. Similarly, a role of cations such as Fe and Mn as redox initiators has been suggested for SZ catalysts containing these cations as promoters.⁶⁶

Conclusion

The results show that in a temperature range typically used for alkane isomerisation (470–670 K), partial reduction of the tungsten species in tungstated ZrO₂ takes place, giving surface W⁵⁺. OH groups associated with the presence of W⁵⁺ species were detected on the reduced catalyst at room temperature; their acid strengths appear to be less than those of OH groups on the oxidised²⁵ catalyst. It was not possible to determine whether these OH groups were present on the catalysts at high temperature or whether they were converted into water and desorbed. The formation of organic radicals during the alkane isomerisation reaction suggests the oxidation of the alkane by a redox process leading to radicals, along with W⁵⁺. A one-electron transfer from an organic radical is postulated to lead to carbenium ions that are chain carriers in the isomerisation catalysis. Thus, we infer that the redox properties of W⁶⁺ in oxidised tungsten-containing ZrO₂ samples may play a crucial role in the activation of alkanes and, consequently, in their catalytic behaviour; this role is suggested to be that of a redox initiator and not in itself catalytic.

Acknowledgements

The assistance of Bernard Morin in the EPR experiments is gratefully acknowledged. This work was supported financially by the Deutsche Forschungsgemeinschaft (Sonderforschungsbereich 338), the Fonds der Chemischen Industrie, and the Max-Buchner-Forschungsstiftung. P. Concepción acknowledges a grant from the Fonds der Chemischen Industrie. The international collaboration was made possible by support from the Alexander von Humboldt Foundation, the Max Planck Gesellschaft, and the BMBF (Max Planck Research Award to H. Knözinger). M. Che gratefully acknowledges the Alexander von Humboldt Foundation for a von Humboldt-Gay Lussac Award, and B. C. Gates thanks this foundation for an Alexander von Humboldt Senior Scientist Award.

References

1. K. Arata, Ad[italic v]. Catal., 1990, 37, 165.
2. K. Tanabe, H. Hattori and T. Yamaguchi, Crit. Re[italic v]. Surf. Chem., 1990, 1, 1.
3. A. Corma, Chem. Re[italic v]., 1995, 95, 559.
4. T. K. Cheung and B. C. Gates, Top. Catal., 1998, 6, 41.
5. C. Gosling, R. Rosin, P. Bullen, T. Shimizu and T. Imai, Petrol. Technol. Quart., Winter 1997/1998, 55..
6. M. Hino and K. Arata, J. Chem. Soc., Chem. Commun., 1988, 1259.
7. M. Hino and K. Arata, J. Chem. Soc., Chem. Commun., 1980, 851.
8. V. Adeeva, J. W. de Haan, J. Jänchen, G. D. Lei, V. Schünemann, L. J. M. van de Ven, W. M. H. Sachtler and R. A. van Santen, J. Catal., 1995, 151, 364.
9. F. Babou, G. Coudurier and J. C. Védrine, J. Catal., 1995, 152, 341.
10. T. K. Cheung, J. L. d'Itri and B. C. Gates, J. Catal., 1995, 153, 344.
11. T. K. Cheung, J. L. d'Itri, F. C. Lange and B. C. Gates, Catal. Lett., 1995, 31, 153.
12. Z. Hong, K. B. Fogash, R. M. Watwe, B. Kim, B. I. Masqueda-Jiménez, M. A. Natal-Santiago, J. M. Hill and J. A. Dumesic, J. Catal., 1998, 178, 489.
13. T. K. Cheung, F. C. Lange and B. C. Gates, Catal. Lett., 1995, 34, 351.
14. J. E. Tábora and R. J. Davis, J. Catal., 1996, 162, 125.
15. H. Liu, V. Adeeva, G. D. Lei and W. M. H. Sachtler, J. Mol. Catal. A, 1995, 100, 35.
16. V. Adeeva, G. D. Lei and W. M. H. Sachtler, Appl. Catal. A, 1994, 118, L11.
17. V. Adeeva, G. D. Lei and W. M. H. Sachtler, Catal. Lett., 1995, 33, 135.
18. D. Farcasiu, A. Ghenciu and Q. L. Li, J. Catal., 1996, 158, 116.
19. K. T. Wan, C. B. Khouw and M. E. Davis, J. Catal., 1996, 158, 311.
20. T. Riemer and H. Knözinger, J. Phys. Chem., 1996, 100, 6739.
21. F. T. T. Ng and N. Horváth, Appl. Catal. A, 1995, 123, L197.
22. E. Iglesia, S. L. Soled and G. M. Kramer, J. Catal., 1993, 144, 238.
23. K. Ebitani, J. Konishi and H. Hattori, J. Catal., 1991, 130, 257.
24. C.-Y. Hsu, C. R. Heimbuch, C. T. Armes and B. C. Gates, J. Chem. Soc., Chem. Commun., 1992, 1645.
25. M. Scheithauer, R. K. Grasselli and H. Knözinger, Langmuir, 1998, 14, 3019.
26. M. Scheithauer, T. K. Cheung, R. E. Jentoft, R. K. Grasselli, B. C. Gates and H. Knözinger, J. Catal., 1998, 180, 1.
27. E. Iglesia, D. G. Barton, S. L. Soled, S. Miseo, J. E. Baumgartner, W. E. Gates, G. A. Fuentes and G. D. Meitzner, Stud. Surf. Sci. Catal., 1996, 101, 533.
28. D. G. Barton, S. L. Soled, G. D. Meitzner, G. A. Fuentes and E. Iglesia, J. Catal., 1999, 181, 57.
29. D. G. Barton, S. L. Soled and E. Iglesia, Top. Catal., 1998, 6, 87.
30. J. G. Santiesteban, D. C. Calabro, W. S. Borghard, C. D. Chang, J. C. Vartuli, Y. P. Tsao, M. A. Natal-Santiago and R. D. Bastian, J. Catal., 1999, 183, 314.
31. O. Glemser and C. Naumann, Z. Anorg. Allg. Chem., 1951, 265, 288.
32. P. A. Sermon and G. C. Bond, J. Chem. Soc., Faraday Trans. 1, 1976, 72, 730.
33. M. A. Vannice, M. Boudart and J. J. Fripiat, J. Catal., 1970, 17, 359.
34. P. Knoll, R. Singer and W. Kiefer, Appl. Spectrosc., 1990, 44, 776.
35. D. Spielbauer, Appl. Spectrosc., 1995, 49, 923.
36. D. A. M. Monti and A. Baiker, J. Catal., 1983, 83, 323.
37. D. P. C. Thackeray, Spectrochim. Acta, Part A, 1974, 30, 549.
38. J. A. Horsley, I. E. Wachs, J. M. Brown, G. H. Via and F. D. Hardcastle, J. Phys. Chem., 1987, 91, 4014.
39. J. G. Santiesteban, J. C. Vartuli, S. Han, R. D. Bastian and C. D. Chang, J. Catal., 1997, 168, 431.
40. I. R. Beattie and T. R. Gilson, J. Chem. Soc. A, 1969, 2322.
41. D. Gazzoli, M. Valigi, R. Dragone, A. Marucci and G. Mattei, J. Phys. Chem. B, 1997, 101, 11129.
42. D. C. Vermaire and P. C. van Berge, J. Catal., 1989, 116, 309.
43. T. Castner, Jr., G. S. Newell, W. C. Holton and C. P. Slichter, J. Chem. Phys., 1960, 32, 668.
44. D. Loveridge and S. Parke, Phys. Chem. Glasses, 1971, 12, 19.
45. M. Che and C. Naccache, Chem. Phys. Lett., 1971, 8, 45.
46. M. J. Torralvo and M. A. Alario, J. Catal., 1984, 86, 473.
47. C. Morterra, E. Giamello, L. Orio and M. Volante, J. Phys. Chem., 1990, 94, 3111.
48. M. J. Sienko and H. Oesterreicher, J. Am. Chem. Soc., 1968, 90, 6568.
49. O. F. Schirmer and E. Salje, Solid State Commun., 1980, 25, 333.
50. D. Bravo, X. Ruiz, F. Díaz and F. J. López, Phys. Re[italic v]. B, 1995, 52, 3159.
51. T. T. Chang, Phys. Re[italic v]., 1966, 147, 264.
52. P. Mériaudeau, Y. Boudeville, P. de Montgolfier and M. Che, Phys. Re[italic v]. B, 1977, 16, 30.
53. D. P. Madaesi, M. Stapelbroek, R. B. Bossoli and O. R. Gilliam, J. Chem. Phys., 1982, 77, 3803.
54. M. Che and A. J. Tench, Ad[italic v]. Catal., 1983, 32, 1.
55. K. N. Spiridonov, O. V. Krylov and E. A. Fokina, Russ. J. Phys. Chem., 1983, 87, 3425.
56. R. F. Howe, J. Chem. Soc., Faraday Trans. 1, 1975, 71, 1689.
57. R. A. Comelli, S. A. Canavese and N. S. Figoli, Catal. Lett., 1998, 55, 177.
58. R. Schwarzmann and R. Birkenberg, Z. Naturforsch., B, 1971, 26, 1069.
59. J. Haber, Stud. Surf. Sci. Catal., 1997, 110, 1.
60. M. Che, B. Canosa and A. R. Gonzales-Elipe, J. Chem. Soc., Faraday Trans. 1, 1982, 78, 1043.
61. I. V. Bobricheva, I. A. Stavitsky, V. K. Yermolaev, N. S. Kotsarenko, V. P. Shmachkova and D. I. Kochubey, Catal. Lett., 1998, 56, 23.
62. D. Spielbauer, Dissertation, University of Munich, 1995..
63. F. R. Chen, G. Coudurier, J. F. Joly and J. C. Vedrine, J. Catal., 1993, 143, 616.
64. K. Watanabe, T. Nakayama and J. Mottl, J. Quant. Spectrosc. Radiat. Transfer, 1962, 2, 369.
65. S. Pignataro, A. Foffani, G. Innorta and G. Distefano, Z. Phys. Chem. (Munich), 1966, 49, 291.
66. F. C. Lange, T.-K. Cheung and B. C. Gates, Catal. Lett., 1996, 41, 95.

---

Footnote

† Present address: Instituto de Technología Química, UPV-CSIC, Avda los Naranjos s/n, Valencia 46022, Spain.

---