The structure of supported and unsupported

Applied Catalysis A: General 285 (2005) 151–162 www.elsevier.com/locate/apcata

The structure of supported and unsupported vanadium oxide under calcination, reduction and oxidation determined with XAS Geert Silversmit a,*, Jeroen A. van Bokhoven b,1, Hilde Poelman a, Ad M.J. van der Eerden b, Guy B. Marin c, Marie-Franc¸oise Reyniers c, Roger De Gryse a b

a Ghent University, Department of Solid State Sciences, Krijgslaan 281 S1, B-9000 Gent, Belgium Utrecht University, Department of Inorganic Chemistry and Catalysis, Sorbonnelaan 16, 3584 CA Utrecht, The Netherlands c Ghent University, Department of Chemical Engineering, Krijgslaan 281 S5, B-9000 Gent, Belgium

Received 11 January 2005; received in revised form 15 February 2005; accepted 15 February 2005 Available online 19 March 2005

Abstract A vanadium oxide powder catalyst supported on titanium oxide (VOx/TiO2(anatase)) was investigated with in situ X-ray absorption spectroscopy (XAS). This system is applied in industry for partial oxidation processes. The structural changes of the supported vanadium oxide due to different treatments were determined: during a reduction–oxidation cycle at 623 K after calcination in air and after heating in inert atmosphere. Unsupported crystalline V2O5 powder was used as a reference system. After calcination in air of the VOx/TiO2(anatase) a V2O5 bulk crystal structure appears, while an octahedral co-ordination with a V O vanadyl bond is present after heating in inert atmosphere. The redox cycle performed on fully oxidized supported vanadium oxide induces similar structural changes as for unsupported V2O5: a reduction to V4+ with a slightly distorted V–O octahedron and a re-oxidation to the initial V2O5 structure. However, when the supported vanadium oxide is reduced after heating in inert, a valence lower than V4+ is obtained and a different structure is found: a symmetrical V–O octahedron is present with V–O and V–V distances identical to the Ti–O and Ti–Ti distances in TiO2(anatase or rutile). After re-oxidation the V2O5 bulk structure is again obtained for the supported vanadium oxide. # 2005 Elsevier B.V. All rights reserved. Keywords: Supported vanadium oxide; Titanium oxide; Anatase; EXAFS; XANES; V2O5; VOx/TiO2; Calcination; Reduction; Oxidation; In situ gas treatments

1. Introduction Vanadium oxides are important catalysts in various industrial processes, for instance in the oxidation of SO2 to SO3, in the partial oxidation of hydrocarbons like o-xylene to phthalic anhydride, or n-butane to maleı¨c anhydride, and in de-NOx reactions [1,2]. Supported vanadium oxide catalysts were developed to improve mechanical strength, * Corresponding author. Tel.: +32 9 264 43 71; fax: +32 9 264 49 96. E-mail addresses: [email protected] (G. Silversmit), [email protected] (J.A. van Bokhoven). 1 Present address: Institute for Chemical and Bioengineering, Federal Institute of Technology (ETH), CH-8093 Zurich, Switzerland. 0926-860X/$ – see front matter # 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.apcata.2005.02.018

thermal stability and lifetime. Although the supports are inert in the catalytic reaction, the structure and composition of the support material can influence the activity and selectivity of the supported vanadium oxide species. For instance, the partial oxidation of o-xylene to phthalic anhydride is superior with the TiO2 anatase support over other oxide supports such as SiO2, Al2O3 and MgO [3,4]. The structure of the supported vanadium oxide on anatase at room temperature depends on the preparation method and the V2O5 loading. The VOx monolayer is suggested to be the active phase in the selective oxidation of o-xylene. The capability of the TiO2 anatase support to stabilize the VOx monolayer is assumed to account for its superiority [4].

152

G. Silversmit et al. / Applied Catalysis A: General 285 (2005) 151–162

The catalytic performance of a vanadium oxide catalyst depends on temperature, vanadium cation oxidation state, and the vanadium oxide structure. In order to examine the changes in structure and oxidation state of the supported vanadium oxide under catalytically relevant conditions, in situ reduction and oxidation measurements at reaction temperature are needed. As extended X-ray absorption finestructure spectroscopy (EXAFS) probes the local coordination of a specific element and can be applied in situ, at high temperature and in different gas atmospheres, it is an appropriate tool to examine the VOx/TiO2 system in reducing and oxidizing atmospheres. As supported vanadium oxides have better catalytic performances, their behaviour during catalytic reactions should be different from that of unsupported vanadium oxides. Determination of the structural changes of supported vanadium oxides during oxidizing and reducing treatments compared to unsupported V2O5 can thus help to reveal the superiority of supported vanadium oxides and the role of the support. In this work the structural changes of a supported VOx/ TiO2(anatase) powder catalyst were determined with in situ X-ray absorption spectroscopy (XAS) under different treatments: heating in inert, calcination in air and during a reduction–oxidation cycle. The structural changes were compared with those of unsupported V2O5.

2. Experimental Except for one sample, all V K-edge and Ti K-edge absorption spectra were collected at station D44 of the DCI storage ring at LURE (Laboratoire pour l’Utilisation du Rayonnement Electromagne´ tique) in Orsay (France) using a Si(1 1 1) double crystal monochromator with an energy resolution of 4 104. The storage ring was filled with 1.85 GeV positrons with a current of 310–260 mA. All measurements were performed in transmission mode using ion chambers filled with air. The monochromator was

detuned to 50% of the maximum intensity at the V K-edge (5465 eV) to minimise the presence of higher harmonics. To reduce noise, six scans were averaged for the V K-edge spectra and three scans for the Ti K-edge spectra of the VOx/ TiO2(anatase) catalyst. For the V2O5 V K-edge spectra three scans were averaged. The counting time was 1–5 s per data point, depending on the energy region. The V K-edge absorption spectra were recorded in five energy regions: 5365–5455 eV in 10 eV steps and 1 s counting time, 5455– 5485 eV in 0.3 eV steps and 2 s, 5485–5535 eV in 0.5 eV steps and 2 s, 5535–6000 eV in 3 eV step and 4 s, and 6000– 6505 eV in 5 eV steps and 5 s counting time. The XAS spectra of several materials (V and Ti foil, TiO2 (rutile and anatase), VO2, V6O13, NH4VO3, VOSO43H2O) were recorded at room temperature (RT) in air as references together with the VOx/TiO2(anatase) powder catalyst (hereafter named VTiO(AirRT)) and V2O5 powder (V2O5 (AirRT)). Crystallographic parameters like bond distances and co-ordination numbers for the reference compounds can be found in Table 1. The XANES measurement on the VOx/TiO2(anatase) calcined in air at 773 K was performed on the DUBBLE CRG beamline (BM26A) of the ESRF storage ring. The current during measurement was 80–70 mA during a 16bunch fill mode. The V K-edge XANES spectra were recorded in transmission mode in air at room temperature using a Si(1 1 1) double crystal monochromator and ionisation chambers filled with Ar/He mixtures. The vertical focussing mirror after the monochromator was used for the harmonic rejection. No EXAFS spectra were taken for this sample. The reference spectra were also recorded on this beamline for the XANES analysis. The structural changes of the V–O co-ordination in V2O5 and in the VOx/TiO2(anatase) under heating, oxidation and reduction were determined with XAS. The experimental conditions for all XAS measurements are summarized in Tables 2 and 3. The V2O5 was heated from RT to 623 K in an inert atmosphere (He, 1 bar, 100 ml/min) (named V2O5(HeatHe)). A reduction/oxidation cycle was given to

Table 1 Bonding type, bond distances (R), and co-ordination numbers (N) for the reference compounds (V6O13 has three different V crystal sites) ˚) Formal valence Compound Bond type N R (A M–O co-ordination (M V or Ti)

Reference

V2O5

V O V–O

1 4

1.59 1.78, 1.88 (2), 2.02

Distorted square pyramidal ‘‘VO5’’

5

[6]

V6O13

V–O

6

1.77, 1.88 (2), 1.96, 1.99, 2.07 1.66, 1.76, 1.90 (2), 2.09, 2.28 1.64, 1.92 (2), 1.93, 1.98, 2.26

Distorted octahedral ‘‘VO6’’

4.33

[7]

VO2 NH4VO3 VOSO43H2O

V–O V–O V O V–O

6 4 1 5

1.76, 1.86, 1.89, 2.01, 2.02, 2.06 1.65, 1.67, 1.80 (2) 1.56 2.01, 2.03, 2.05, 2.08, 2.28

Distorted octahedral ‘‘VO6’’ Tetrahedral ‘‘VO4’’ Distorted octahedral ‘‘VO6’’

4 5 4

[8] [9] [10]

TiO2(anatase)

Ti–O Ti–Ti

6 8

1.93 (4), 1.98 (2) 3.03 (4), 3.78 (4)

Octahedral

4

[11]

TiO2(rutile)

Ti–O Ti–Ti

4 2

1.94 (4), 1.99 (4) 2.96 (2)

Octahedral

4

[11]

G. Silversmit et al. / Applied Catalysis A: General 285 (2005) 151–162 Table 2 Experimental conditions applied to the V2O5 powder before and during Xray absorption data collection Treatment name

Treatment

Measuring conditions

1: V2O5(AirRT) 2: V2O5(HeatHe)

None RT to 623 K (10 K/min) in He, 35 min 6 h reduction with 5% H2 in He at 623 K 10 h oxidation with 5% O2 in He at 623 K

Air, RT He, 623 K

3: V2O5(He,RED) 4: V2O5(He,OX)

5% H2 in He, 623 K 5% O2 in He, 623 K

the V2O5(HeatHe) (V2O5(He,RED); V2O5(He,OX)). Two heat treatments were given to the VOx/TiO2(anatase): a heating in inert (VTiO(HeatHe)) and a calcination in air (RT to 773 K) (VTiO(CalcAir)). The reduction/oxidation behaviour of the supported vanadium oxide after heating in inert (VTiO(He,RED); VTiO(He,OX)) and in air (VTiO(Air,RED); VTiO(Air,OX)) was studied. Reduction and oxidation were performed at 623 K with 100 ml/min flows of 5% H2 in He and 5% O2 in He, respectively. The spectra were collected under flowing reducing or oxidizing atmospheres. Before measurements, the samples were pretreated during several hours with the corresponding gas flow. The samples were pressed as self-supported pellets in the rectangular hole of a stainless steel sample holder, the thickness of the pellets was chosen to give an absorption (mx) of 2.5 at the V K absorption edge for optimal signal to noise ratio. For the in situ measurements, an EXAFS cell equipped with beryllium windows was used [5]. Typically 5 mg of V2O5 or VOx/TiO2 powder was pressed with 40 mg of BN as inert binder. The VOx/TiO2(anatase) powder system studied here is a catalyst used in industry for partial oxidation reactions and is made by impregnation. As determined with EDX, the VOx/ TiO2(anatase) powder has 70.6 at.% oxygen (43.8 wt.%), 27.5 at.% titanium (51.0 wt.%), 1.4 at.% vanadium (2.8 wt.%), and 0.5 at.% Sb (2.5 wt.%). With XPS the following concentrations were found: 51.4 at.% oxygen, 15.7 at.% titanium, 23.6 at.% carbon, 3.6 at.% nitrogen, Table 3 Experimental conditions applied to the VOx/TiO2(anatase) powder catalyst before and during X-ray absorption data collection Treatment name

Treatment

Measuring conditions

1: 2: 3: 4:

None RT to 623 K (10 K/min) in He RT to 773 K (10 K/min) in air 7 h reduction with 5% H2 in He at 623 K 7 h oxidation with 5% O2 in He at 623 K 6 h reduction with 5% H2 in He at 623 K 5 h oxidation with 5% O2 in He at 623 K

Air, RT He, 623 K Air, RT 5% H2 in He, 623 K 5% O2 in He, 623 K 5% H2 in He, 623 K 5% O2 in He, 623 K

VTiO(AirRT) VTiO(HeatHe) VTiO(CalcAir) VTiO(He,RED)

5: VTiO(He,OX) 6: VTiO(Air,RED) 7: VTiO(Air,OX)

153

2.9 at.% vanadium and 2.3 at.% Sb. The BET surface area is 7 m2/g. VO2 (99%) and V6O13 (99.5%) powders were purchased from Alfa Aesar, NH4VO3 (99.99%), VOSO43H2O (99.99%), TiO2-rutile (99.9%) and BN (99%) powders were purchased from Sigma-Aldrich and TiO2-anatase (99%) and V2O5 (96%) powders from Merck. The EXAFS data reduction and analysis were performed with the XDAP code [12]. The pre-edge background was subtracted using a modified Victoreen curve [13] and the atomic background, m0, using a cubic spline routine [14]. The spectra were normalised by division of the absorption data by the edge step at 50 eV above the edge position. The fit parameters were determined by multiple shell fitting in Rspace, by applying the difference file technique using Fourier transformation [15]. Phase shifts and backscattering amplitudes used to calculate the V–O EXAFS contributions were obtained from Na3VO4, as described in [16]. The V–V contribution was calculated with FEFF 8.0 [17] on a V bcc ˚ . The accuracy was checked on cluster with a radius of 7 A the V-foil measurement. There was good agreement up to ˚ 1 and an EXAFS fit up to the sixth shell (6.3 A ˚ ) was 15.6 A 2 possible. In all fits Ds was confined between 0.006 and 0.006 and DE0 between 5 and +5 eV. The Fourier transform of the EXAFS signal can be taken with different k-weightings (multiplication by kn). Elements with low mass will scatter mainly at low k-weightings while heavy elements will also scatter significantly at higher kweightings. This k-weighting can thus be used to distinguish between low (here oxygen) and high (here vanadium) Z scatterers. In general, the following procedure was used to fit the EXAFS data for all the unknown structures. First, the XANES data were compared with the references. The structure of the compound with the best resemblance was then used as a first guess. Only the V–O contributions were taken into account, which were then optimised in k1weigthing. When a satisfactory fit was obtained for the V–O shell, V–V contributions were added where necessary. The optimisation for the V–V shell was performed in k3weighting. Then a full optimisation for all contributions (V– O and V–V) was carried out in k2-weighting and the results were always checked in k1- and k3-weighting. From the fit on the V-foil and V2O5 reference systems we estimate the accuracy of the derived parameters to be 20% for N, two significant figures for E0 and Ds2 and the error for R smaller ˚ . However, the errors for the supported system can than 0.1 A be bigger due to the small absorption signal. No reference spectrum could be measured simultaneously with the sample under study, so that the co-ordinate axes of all absorption spectra have calibrated monochromator energy values. The main-edge of the normalised data is taken at mx = 0.5. The position of the pre-edge is determined by the maximum intensity of the pre-edge peak. The pre-edge intensities are determined as illustrated in Fig. 1. First the main-edge under the pre-edge is interpolated with a polynomial spline, then the pre-edge is isolated by subtracting this spline from the data, finally

154


Fig. 1. The pre-edge region of V2O5, illustrating the isolation of the preedge by interpolating the background with a polynomial spline.

the surface area S under the isolated pre-edge is taken as the P pre-edge intensity. S is calculated numerically as S ¼ i ðxiþ1 xi Þ 12 ðyi þ yiþ1 Þ, with yi the intensity of the isolated pre-edge peak at X-ray energy xi. Different background spline interpolations were tried, until the resulting subtracted pre-edge peak had a shape as close as possible to a Gaussian. For the example in Fig. 1, S was found to be 1.8 and the variation of the pre-edge intensity was smaller than 0.1 (normalised units eV) when different acceptable background spline interpolations were tried. First the XANES spectra of the vanadium oxide reference materials will be discussed. Then the XANES spectra of the heat treatments and redox cycles on V2O5 and VOx/ TiO2(anatase) will be presented. Finally, the EXAFS analysis based on the XANES observations will be given. Values of V–O co-ordination numbers in the EXAFS fits are often fixed based on the XANES results.

3. Results 3.1. XANES spectra 3.1.1. Vanadium oxide references Wong et al. [18] have shown that the main-edge energy position of vanadium compounds contains information about the valence of the V cation, and the pre-edge intensity about the V co-ordination. Practically, the position of both pre-edge peak and main-edge shifts towards higher energies with increasing valence of the vanadium cation, but they

Fig. 2. Normalised absorption (mx) spectra for the V–O reference compounds recorded at RT in air.

shift to a different extent. Symmetrical vanadium–ligand co-ordinations with inversion symmetry, such as regular ‘‘VO6’’ octahedra, have very small pre-edge intensities, while co-ordinations with no inversion symmetry, like distorted ‘‘VO6’’ units or ‘‘VO4’’ tetrahedrons, have significant to large pre-edge intensities. These relations are illustrated on the XANES spectra of the reference compounds. The XANES spectra of the measured reference compounds are given in Fig. 2 and their pre-edge features in Table 4. Because no absolute energy reference was available, the difference, D, between main- and pre-edge position is taken as an estimate for the valence of the V cations. In the vanadium oxide series V2O5, V6O13, VO2 the valence of the V cation decreases from 5+ to 4+ and the corresponding D from 9.9 to 8.2 eV (Table 4). The reference compounds NH4VO3 and VOSO43H2O with valences V5+ and V4+ respectively, have D values 10 and 8.6 eV, in agreement with the ones for the vanadium oxides. The accuracy for the value of D is 0.3 eV. NH4VO3 has an almost regular tetrahedral ‘‘VO4’’ co-ordination and a very strong pre-edge intensity. The nearest V–O co-ordination sphere in V2O5 is a strongly distorted square pyramidal ‘‘VO5’’ unit, with one short ˚ perpendicular to the plane vanadyl bond (V O) at 1.59 A

Table 4 Pre-edge properties of the XANES spectra for the V–O reference compounds from Fig. 2 Compound

Threshold (eV)

Pre-edge peak position (eV)

Main-edge position (eV)

D (eV)

Pre-edge intensity

V2O5 V6O13 VO2 NH4VO3 VOSO43H2O

5470.6 5470.3 5468.2 5469.7 5469.7

5471.5 5471.3 5470.3 5470.6 5470.7

5481.4 5480.8 5478.5 5480.6 5479.3

9.9 9.5 8.2 10.0 8.6

1.8 1.6 1.3 2.5 0.8

D presents the difference between main-edge and pre-edge peak position. Corresponding V cation valences and V–O co-ordinations can be found in Table 1.


155

Fig. 3. Normalised absorption (mx) spectra for the V2O5 treatments (for experimental conditions see Table 2).

formed by the four other oxygen neighbours. A large preedge intensity is present. The distorted octahedral coordinations in V6O13 and VO2 have smaller intensities. In V6O13 three different V crystal sites are present (Table 1). VOSO43H2O has the smallest pre-edge intensity. This compound has a slightly distorted ‘‘VO6’’ co-ordination ˚ and a long V–O bond at with a short vanadyl bond at 1.56 A ˚ 2.28 A both perpendicular to the more or less symmetrical square plane of the ‘‘VO6’’ unit. 3.1.2. Heating Fig. 3 displays the normalised XANES spectra for all the V2O5 treatments and Fig. 4 for the VOx/TiO2 treatments. Changes in the XANES spectra due to the applied treatments are clearly present. The pre-edge features for the XANES spectra of all V2O5 treatments are summarized in Table 5, and for the VOx/TiO2(anatase) in Table 6. Comparison of the relative position of the pre-edge with those for the reference compounds from Table 4 gives the qualitative vanadium valence estimates in Tables 5 and 6. The corresponding V–O co-ordinations based on the XANES observations are given in Fig. 5. Heating of V2O5 to 623 K in He (V2O5(HeatHe)) reduces the vanadium oxide without loss of pre-edge intensity. The difference in main-edge and pre-edge position (D) decreases slightly. This points to a reduction during heating. There are only small XANES signals for this sample, so a large disorder is present. The XANES of VTiO(AirRT) resembles mostly the one of VOSO43H2O, but the pre-edge is more intense. A ‘‘VO6’’ co-ordination should thus be present and this ‘‘VO6’’ unit is slightly more distorted than in VOSO43H2O. This reference compound has a formal valence of 4+, the edge positions (D) for the VOx/TiO2 catalyst at room temperature also suggest a 4+ valence. When the supported vanadium oxide catalyst is heated from RT to 623 K in He (VTiO(HeatHe)), the XANES signal changes so this heating influences the supported

Fig. 4. Normalised absorption (mx) spectra for the VOx/TiO2(anatase) treatments (for experimental conditions see Table 3).

vanadium oxide layer. The difference between main- and pre-edge positions now increases to the value of V2O5, pointing to an oxidation of the supported vanadium oxide during heating in He. The pre-edge intensity decreases only slightly, so the octahedral ‘‘VO6’’ co-ordination is preserved at 623 K. The XANES spectrum for the supported vanadium oxide after calcination in air (VTiO(CalcAir)) resembles the one for V2O5(AirRT). As this calcination was recorded at another beamline, the pre-edge features cannot be compared with the references from Table 2, but pre-edge intensity and edge position do correspond with the ones of V2O5 measured on this beamline. 3.1.3. Redox behaviour A reduction–oxidation cycle is given to V2O5(HeatHe). After reduction with H2 the pre-edge intensity diminishes, suggesting a fairly symmetric octahedral co-ordination for V2O5(He,RED). The D value becomes smaller, most likely the vanadium is reduced to V4+. The XANES of V2O5(He,OX) and V2O5(AirRT) are the same since preedge intensity and D retake the values of V2O5 (Fig. 3, Tables 4 and 5). As the re-oxidation of the reduced sample restores the V2O5 structure, the effect of the H2-reduction on V2O5 powder at 623 K is reversed by a re-oxidation at the same temperature.

156


Table 5 Spectral properties of the XANES spectra for the V2O5 treatments (for conditions see Table 2) V2O5 sample

Threshold (eV)



D (eV)

Estimated valence

Pre-edge intensity

Estimated co-ordination

V2O5(AirRT) V2O5(HeatHe) V2O5(He,RED) V2O5(He,OX)

5470.7 5469.5 5469.5 5470.3

5471.7 5470.9 5470.7 5471.3

5481.8 5480.3 5479.8 5481.2

10.1 9.4 9.1 9.9

5 4.33 4 5

1.8 1.9 1.0 1.8

As V2O5 As V2O5 Symmetric ‘‘VO6’’ As V2O5

D presents the difference between main-edge and pre-edge peak position.

Table 6 Spectral properties of the XANES spectra for the VOx/TiO2(anatase) treatments (for conditions see Table 3) VOx/TiO2 sample

Threshold (eV)



D (eV)

Estimated valence

Pre-edge intensity

Estimated co-ordination

VTiO(AirRT) VTiO(HeatHe) VTiO(He,RED) VTiO(He,OX) VTiO(Air,RED) VTiO(Air,OX)

5469.7 5468.5 5467.6 5469.7 5469.5 5469.5

5470.7 5470.3 5470.1 5470.9 5470.9 5470.9

5479.9 5480.3 5477.8 5481.4 5479.9 5481.4

9.2 10.0 7.7 10.5 9.0 10.5

4 5