- Research article
- Open Access
Structural characterization of vanadium oxide catalysts supported on nanostructured silica SBA-15 using X-ray absorption spectroscopy
Chemistry Central Journalvolume 4, Article number: 3 (2010)
The local structure of vanadium oxide supported on nanostructured SiO2 (VxOy/SBA-15) was investigated by in situ X-ray absorption spectroscopy (XAS). Because the number of potential parameters in XAS data analysis often exceeds the number of "independent" parameters, evaluating the reliability and significance of a particular fitting procedure is mandatory. The number of independent parameters (Nyquist) may not be sufficient. Hence, in addition to the number of independent parameters, a novel approach to evaluate the significance of structural fitting parameters in XAS data analysis is introduced. Three samples with different V loadings (i.e. 2.7 wt %, 5.4 wt %, and 10.8 wt %) were employed. Thermal treatment in air at 623 K resulted in characteristic structural changes of the V oxide species. Independent of the V loading, the local structure around V centers in dehydrated VxOy/SBA-15 corresponded to an ordered arrangement of adjacent V2O7 units. Moreover, the V2O7 units were found to persist under selective oxidation reaction conditions.
Mixed transitions metal oxides (e.g. MoVNbTe oxides) are active in selective oxidation of propane to acrylic acid. In contrast to various binary oxides (e.g. MoO3 or V2O5), these mixed oxides exhibit a much higher selectivity. However, the origin of the promoting effect of, for instance, vanadium in mixed oxides is largely unknown. Hence, model systems are sought which enable conclusions on structure activity relationships of individual metal centers in active catalysts. For that, supported metal oxides possess two major advantages over bulk oxides. First, particular metal oxide structures which are not readily available for investigations under reaction conditions can be stabilized and studied on suitable support materials . Second, dispersed supported metal oxides simplify correlating the local structure around the metal centers with their catalytic performance. Distinguishing active metal centers at the surface from metal centers in the bulk of conventional oxide catalysts is no longer required.
VxOy supported on SBA-15 (nanostructured SiO2)  constitutes a suitable model system to investigate the role of vanadium during selective oxidation catalysis [3–6]. Structural characterization of VxOy supported on SiO2 has been subject of many spectroscopic studies including IR [[7–9], XPS [6, 10, 11], Raman [10–15], UV-VIS [11, 13–16] and EXAFS [13, 17–22]. A recent review of spectroscopic investigations and structural characteristics of various supported vanadium oxides has been presented by Weckhuysen and Keller . It is assumed, that the structure of supported vanadium oxide depends on both amount of vanadium and degree of hydration . Hence, most studies were performed on VxOy/SiO2 samples exhibiting low vanadium loading (< 10 wt %). At these loadings a monolayer of supported VxOy species is assumed and crystalline V2O5 is not detectable. Under ambient conditions the structure of hydrated vanadium oxide supported on SiO2 resembles that of V2O5 [10, 13, 18]. Thermal treatment in oxygen results in dehydration of the vanadium oxide species. This dehydrated state has been proposed to consist of isolated VO4 tetrahedrons bond to the SiO2 support [13, 17, 24, 25]. However, V2O7 dimers or further extended structures supported on SiO2 have not been excluded . In total, the structure of dehydrated vanadium oxide species supported on SiO2 remains under debate.
XAS is particularly suitable to study supported catalysts under reaction conditions. The average valence, for instance, can be readily obtained by comparison with known reference compounds. Elucidating the geometric structure, however, is often more difficult. In the conventional approach theoretical XAFS scattering amplitudes and phases are calculated for a suitable model structure. Subsequently, a sum of theoretical XAFS functions is refined to the experimental data. Structural parameters like coordination numbers, nearest neighbor distances, and disorder parameters may be determined. However, more often than not, the number of potential parameters exceeds the number of "independent" parameters. The upper limit may be calculated from Fourier theory and must not be exceeded. Nonetheless, it appears that even refinements employing a much smaller number of freely varied parameters may yield ambiguous structural results. The often used Nyquist criteria may not be sufficient to deem a fitting procedure reliable. Basically, one pair of strongly correlated parameters suffices to render a seemingly good agreement between experimental data and theoretical model structure meaningless. While this case may be clearly indicated by the correlation matrix of the refinement, other pitfalls may be less obvious. Hence, procedure are sought that enable evaluating the significance of each fitting parameter individually.
Here, we have performed in situ XAS investigations of VxOy supported on SBA-15 in the hydrated and dehydrated state. The same materials were already carefully characterized by several standard techniques (i.e. physisorption, TEM, IR, Raman, UV-Vis, and XPS) and the results of these studies have been described in Ref [9, 10, 14]. In particular, using UV-Vis and Raman spectroscopy, Hess et al. showed that the catalysts are similar to other systems previously described in the literature. However, the results obtained could not unequivocally determine the local structure around the V species on SiO2, as it was also the case in previous studies. Our approach focused on elucidating the local structure around the vanadium centers in the dehydrated state of VxOy-SBA-15 model catalysts with different vanadium loadings. A detailed XAFS data analysis, in particular of higher V-V distances, was performed together with a detailed evaluation of the significance of the fitting parameters employed. This procedure permitted detailed conclusions on the extended local structure of the vanadium oxide species supported on SBA-15.
Results and discussion
Local structure of dehydrated VxOy/SBA-15 - Comparison to V oxide references
Characterization of pore structure and surface area, and optical spectroscopic investigations of the same model catalysts studied here have been previously described [9, 10, 14]. After surface functionalisation and ion exchange to introduce the V precursor, the materials were calcined at 823 K in air. Calcination results in decomposition of both precursor and functionalisation agent. Preparation, functionalisation, and thermal treatment also have been described in Ref [9, 10, 14]. The authors stated that residuals of the functionalisation agent were no longer detectable (i.e. IR, Raman, UV-Vis, and XPS) in the material obtained. A brief summary of the N2 physisorption analysis described in Ref  is given in Table 1.
Here, we have performed a detailed XAFS investigation of samples with different V loadings in the hydrated and dehydrated state. In particular, we wanted to analyze the contribution of higher scattering shells to the XAFS signal and possibly reveal the presence of V nearest neighbors in the local structure of vanadium oxide species supported on SiO2. A detailed XAFS analysis of higher shells in the FT(χ(k)*k3) has been largely neglected in the corresponding literature.
During thermal treatment of as-prepared hydrated VxOy/SBA-15 in oxygen (20% in He) a loss of water and a distinct change in structure were observed. After thermal treatment dehydrated VxOy/SBA-15 was cooled to 293 K in oxygen in He without exposure to air or water. No changes in XAFS spectra were observed during cooling. The EXAFS χ(k)*k3 of dehydrated VxOy/SBA-15 with different V loadings are depicted in Figure 1. The usable spectral ranged extended from 2.7 Å through 11.0 Å. The V K edge XANES spectra and the FT(χ(k)*k3) of dehydrated VxOy/SBA-15 samples measured at 293 K are shown in Figure 2. FT(χ(k)*k3) are not phase shift corrected. Thus, the distances in the FT(χ(k)*k3) are shifted by ~0.4 Å to lower values compared to crystallographic distances. Compared to vanadium oxide references, the overall XANES region of dehydrated VxOy/SBA-15 resembled best those of NH4VO3, Mg2V2O7, and Na3VO4 (Figure 3(a)). In the local structure of these references vanadium centers are tetrahedrally coordinated by four oxygen atoms. Compared to the XANES spectrum of dehydrated VxOy/SBA-15, NH4VO3 and Mg2V2O7 exhibit very similar pre-edge peak heights in their XANES spectra (i.e. 0.65). Conversely, the pre-edge peaks in the XANES of Na3VO4 and Mg3V2O8 are much higher than that of dehydrated VxOy/SBA-15 (Figure 3(a)).
In Figure 3(b) the FT(χ(k)*k3) of NH4VO3, Mg2V2O7, and Na3VO4 are compared to that of dehydrated VxOy/SBA-15. The first V-O peak in the FT(χ(k)*k3) at ~1.4 Å (not phase shift corrected) for all references shown corresponds to a VO4 tetrahedron in the respective structures. Apparently, the spectra of NH4VO3 and Mg2V2O7 most closely resemble that of dehydrated VxOy/SBA-15. In the FT(χ(k)*k3) of Na3VO4 and Mg3V2O8 the first V-O peak is significantly higher than in the FT(χ(k)*k3) of dehydrated VxOy/SBA-15. In contrast to the FT(χ (k)*k3) of Mg3V2O8 which exhibits a significant amplitude at distances above 2 Å, the FT(χ (k)*k3) of Na3VO4, NH4VO3, Mg2V2O7, and dehydrated VxOy/SBA-15 show little amplitude at higher distances. Moreover, looking at the differences between the FT(χ (k)*k3) of Na3VO4, NH4VO3, and Mg2V2O7, the latter appears to yield the best agreement with that of dehydrated VxOy/SBA-15. In all reference the low amplitude of the FT(χ (k)*k3) at R > 2 Å is characteristic of the local structure around the V centers. It is not caused by an increased amount of disorder. In total, based on comparing the XANES and FT(χ (k)*k3) of dehydrated VxOy/SBA-15 to those of potential references, NH4VO3 and Mg2V2O7 have been identified as suitable references to serve as model systems for a more detailed structural analysis.
Before we discuss the details of analyzing the XAFS data of dehydrated VxOy/SBA-15, a suitable analysis procedure for the higher V-V contributions in the XAFS spectra of references NH4VO3 and Mg2V2O7 was sought. As an example and to reduce the number of tables here, the application of confidence limits and F parameter to distinguish analysis fitting procedures is described for two refinements of a suitable model structure to the experimental FT(χ (k)*k3)of dehydrated VxOy/SBA-15. The model structure consisted of a tetrahedral coordination of the V center by four oxygen atoms at ~1.7 Å (NH4VO3), two vanadium atoms at distances at ~3.4 Å (NH4VO3) and 3.6 Å (Mg2V2O7), one oxygen atom at ~2.9 Å (Mg2V2O7), and one Si atom at ~2.8 Å (Table 2). Experimental FT(χ(k)*k3) of Mg2V2O7 and NH4VO3 and the corresponding XAFS refinements are shown in Figure 4. Deviations between the theoretical and experimental spectrum of Mg2V2O7 in the range from 2 - 4 Å are caused by the number of Mg neighbors and nearly linear multiple-scattering paths in Mg2V2O7 that contribute in this range. These are not sufficiently accounted for by the simplified refinement procedure. The results of the XAFS refinement for dehydrated VxOy/SBA-15, NH4VO3, and Mg2V2O7 are summarized in Table 2. Apparently, the distorted VO4 tetraeder in NH4VO3 required two different V-O distances to be included in the refinement, while Mg2V2O7 and dehydrated VxOy/SBA-15 exhibited a similar single V-O distance. It seems that the distortion in the VO4 units of crystalline reference Mg2V2O7 could not be resolved by the XAFS analysis procedure employed. Accordingly, a lower σ2 was obtained (0.0012 Å2) for NH4VO3 compared to those of dehydrated VxOy/SBA-15 and Mg2V2O7 (0.0075 Å2 and 0.0059 Å2, respectively). Contrarily, a single V-V distance at 3.47 Å (CN = 2) sufficed for NH4VO3 (consisting of chains of VO4 units), while two V-V distances had to be included for Mg2V2O7 (consisting of adjacent V2O7 units) and dehydrated VxOy/SBA-15. In all three cases, a similar σ2 parameter for the V-V contributions of about 0.014 Å2 was obtained.
The corresponding confidence limits and significance parameters F are given in Table 3. In fitting procedure #1 two V-O distances in the first V-O shell were allowed to vary independently (both with a CN of 2 and the same σ2). Moreover, E0 was also allowed to vary in fitting procedure #1. Because of Nind = 18 and Nfree = 11 refinement procedure #1 would be taken as reliable according to the Nyquist criteria. However, confidence limits of the first V-O distance of ± 0.1 Å and an F parameter of 0.7 for both distance and σ2 were obtained with procedure #1. Moreover, E0 exhibited a confidence limit of ± 9.2 and F = 0.9. Apparently, fitting procedure #1 already exceeds the number of meaningful parameters and yield ambiguous structural parameters. On the one hand, the reduced amplitude of the FT(χ(k)*k3) of dehydrated VxOy/SBA-15 compared to that of references consisting of undistorted VO4 units (Figure 3) suggested the presence of more than one V-O distance, including a short "vanadyl" V = O distance. On the other hand, however, the resolution in the experimental FT(χ(k)*k3) and the available degree of freedom did not permit refining more than one V-O distance in the procedure used. Therefore, the fitting procedure was modified. In the following, E0 was kept invariant in the refinement and only one V-O distance at ~1.75 Å was used. In contrast to procedure #1, procedure #2 yielded reasonable confidence limits and acceptable F parameters.
The σ2 parameter of the V-O contribution at 2.9 Å exhibited a rather high confidence limit and F = 0.7. Apparently, both V-O and V-Si neighbors in the distance range from 2.5 Å to 2.9 Å are required for a good refinement of the model structure to the experimental data. This is indicated by the confidence limits and F parameters calculated for the corresponding distances (Table 3). Nevertheless, the high σ2 obtained for the V-Si contribution and the rather low σ2 obtained for the V-O at 2.9 Å indicate a certain ambiguity of the corresponding fitting results. The reason may be a considerable static disorder and, thus, a broadened V-Si distance distribution. Hence, calculating and evaluating confidence limits and F tests permitted to arrive at a meaningful and reliable fitting procedure. In that, the approach employed appears to be superior to only calculating the Nyquist criteria. In total, procedure #2 worked very well for XAFS data analysis of dehydrated VxOy/SBA-15 and Mg2V2O7. In contrast, the local structure around V centers in NH4VO3 was best described by assuming two different V-O distances in the first coordination shell and only one V-V distance at 3.47 Å (CN = 2) (Table 2). A V-O distance at 2.8 Å was found to be insignificant.
Local structure of dehydrated VxOy/SBA-15 - XAFS refinement of "VO4" based model structures
After having identified two suitable references as model structures for XAFS refinements to the experimental FT(χ(k)*k3) of dehydrated VxOy/SBA-15 (Figure 3(b)), the XAFS analysis approach chosen shall be described in more detail. In addition to using confidence limits and F tests as introduced above, the suitable XAFS fitting procedure was developed stepwise as outlined in the following.
First, we started with an often repeated assumption from the literature. DR-UV-Vis or Raman measurements revealed that dehydration of VxOy/SBA-15 resulted in a characteristic change from a distorted square pyramidal to a distorted tetrahedral coordination [10, 13, 16]. The local structure of vanadium oxide species supported on SiO2 was assumed to correspond to isolated VO4 units. Hence, in a first tetrahedron approach the theoretical XAFS function of a VO4 tetrahedron consisting of two slightly different V-O distances was refined to the FT(χ(k)*k3) of dehydrated VxOy/SBA-15 (Figure 5, top) together with the Fourier transformed χ(k)*k3 of the individual scattering paths). Because of the similar height of the pre-edge peak in the XANES (Figure 3(a)) and the first V-O peak in the FT(χ(k)*k3), phases and amplitudes employed in the refinement were calculated using the model structure of NH4VO3 (ICSD 1487 ) and Mg2V2O7 (ISCD 2321 ). Figure 5, top shows a good agreement between theoretical and experimental FT(χ(k)*k3) of dehydrated VxOy/SBA-15 for the first V-O peak below 2 Å. Naturally, the amplitude between 2 Å and 4 Å in the FT(χ(k)*k3) could not be accounted for. Hence, a structural model assuming only isolated VO4 species cannot adequately describe the local structure around the V centers in dehydrated VxOy/SBA-15.
Therefore, we assumed that higher coordination shells around the vanadium centers significantly contribute to the FT(χ(k)*k3) of dehydrated VxOy/SBA-15. These shells have to be included in the refinement. A seemingly expected contribution may arise from silicon backscatterers in the SiO2 support at distances of less than 3.0 Å. This has been previously proposed by Keller et al. . Thus, in extension of the tetrahedron approach a silicon atom at a V-Si distance of 2.8 Å was included in the theoretical model. In the corresponding "O3V-O-Si" unit a Si-O distance of 1.62 Å is assumed (inset in Figure 5, bottom)), as it is found in various silicates. The result of the XAFS refinement of the "O3V-O-Si" model to the FT(χ(k)*k3) of dehydrated VxOy/SBA-15 is depicted in Figure 5, bottom). The additional Si backscatterer resulted in a better agreement between theoretical and experimental FT(χ(k)*k3) at distances of about 2.4 Å (not phase shift corrected). The resulting V-Si distance amounted to 2.54 Å, comparable to the distance obtained by Keller et al. (2.61 Å ). However, it can be easily seen from Figure 5, bottom) that the amplitude in the FT(χ(k)*k3) of dehydrated VxOy/SBA-15 between 2.4 and 4.0 Å is still not accounted for.
Local structure of dehydrated VxOy/SBA-15 - XAFS refinement of "V2O7" based model structures
Figure 5 shows that an "isolated VO4" model did not properly describe the local structure between 2 Å and 4 Å around vanadium centers in dehydrated VxOy/SBA-15. Hence, we assumed that at least "V2O7 dimers" would be needed to achieve a good agreement between theoretical and experimental XAFS FT(χ(k)*k3). V2O7 units are present in the structures of the references NH4VO3 and Mg2V2O7 whose spectra resembled best the XANES and EXAFS spectra of dehydrated VxOy/SBA-15 (Figure 3). Therefore, a V-V scattering path at 3.4 Å was included in the model used in the XAFS refinement. This distance corresponds to the shortest V-V distance between two corner-sharing VO4 tetrahedrons in "V2O7 dimers" of NH4VO3 and Mg2V2O7. The result of the corresponding XAFS refinement is shown in. Apparently, a structural model based on isolated V2O7 dimers was equally unsuited to describe the local structure around V centers in dehydrated VxOy/SBA-15. The agreement between theoretical and experimental FT(χ(k)*k3) in the range from 2 to 4 Å is still not sufficient (Figure 6). Also, adding a V-Si distance to this "isolated V2O7 dimer model" only resulted in a minor improvement of the refinement.
Figure 5 and Figure 6 clearly show that neither an "isolated VO4" model nor an "isolated V2O7" model properly describe the local structure of the majority of V centers in dehydrated VxOy/SBA-15. Hence, in the next step an ordered arrangement of neighboring V2O7 units was assumed. Because of their similar XANES and EXAFS spectra, we again referred to NH4VO3 and Mg2V2O7 as references. V2O7 units form chains in NH4VO3 with one V-V distance. Conversely, V2O7 units are neighboring but more separated in Mg2V2O7 resulting in two distinct V-V distances (ICSD 2321 ). Accordingly, two additional scattering paths were added to the previous "isolated V2O7" model. These two paths correspond to V-O (2.8 Å) and V-V (3.6 Å) distances between two neighboring V2O7 units in the structure of Mg2V2O7. The result of the corresponding XAFS refinement to the FT(χ(k)*k3) of dehydrated VxOy/SBA-15 is shown in Figure 7 together with the various V-O and V-V distances used. Apparently, assuming neighboring V2O7 units in an ordered arrangement supported on SBA-15 yielded a good agreement between theoretical and experimental FT(χ(k)*k3) of dehydrated VxOy/SBA-15 over the extended R range from 1 Å to 4 Å. The structural and fitting parameters obtained from the XAFS refinement to the experimental FT(χ(k)*k3) of dehydrated VxOy/SBA-15 and Mg2V2O7 are given in Table 2. The similar V-O distances, V-V distances, and σ2 parameters of dehydrated VxOy/SBA-15 and Mg2V2O7 corroborate our choice of model system to describe the local structure around V centers dehydrated VxOy/SBA-15.
Schematic structural representation of dehydrated VxOy/SBA-15
A schematic structural representation of the ordered arrangement of V2O7 units in dehydrated VxOy/SBA-15 is depicted in Figure 8. In contrast to previous results on low loaded (< 1 V/nm2) VxOy/SiO2 samples [17, 21] we conclude that isolated VO4 units are not the major vanadium oxide species present on the dehydrated VxOy/SBA-15 samples studied here. From the different loadings studied, only the 2.7 wt % VxOy/SBA-15 sample possessed a vanadium content of less than 1 V/nm2. The three dehydrated VxOy/SBA-15 samples exhibited only minor differences in their XANES spectra (Figure 2), FT(χ(k)*k3) (Figure 9), and XAFS fitting results (Table 4). Hence, independent of the V loading in the range 2.7 - 10.8 wt % the local structure of the majority of V centers in dehydrated VxOy/SBA-15 is described best by an ordered arrangement of neighboring V2O7 units (Table 4, Figure 9). Decomposition of the decavanadate precursor during calcinations of the as-prepared materials will most likely result in the formation of dehydrated VxOy/SBA-15. Any description of the formation mechanism of ordered V2O7 units on the surface of SiO2, however, is beyond the scope of this work. Exposure of the calcined material to ambient conditions apparently results in re-hydration and formation of the hydrated VxOy species supported on SBA-15. This transformation already suggests a reversible hydration-re-hydration behavior of vanadium oxide species supported on SBA-15 which should be the subject of further studies.
The presence of non-monomeric VxOy species in dehydrated VxOy/SBA-15 samples was also recently concluded based on NEXAFS studies combined with theoretical calculations . Eventually, structural evolution and catalytic activity of dehydrated VxOy/SBA-15 were studied by combined in situ XAS-MS under selective propene oxidation reaction condition. Onset of catalytic activity was detected at about 573 K. The in situ XAS data measured indicated, that the characteristic ordered arrangement of V2O7 dimers in the local structure of dehydrated VxOy/SBA-15 persisted under catalytic reaction conditions. A more detailed analysis of structure activity correlations of VxOy/SBA-15 under selective oxidation reaction conditions will be presented elsewhere.
Oxygen and silicon atoms of the SiO2 support are not depicted in the schematic representation depicted in Figure 8. In particular Si atoms in the topmost layer of SiO2 belong to the second coordination sphere of the V centers. Previous reports have indicated that V-Si distances may contribute to the experimental FT(χ(k)*k3) of dehydrated VxOy/SBA-15 . Therefore, a single V-Si scattering path was included in the refinement of the "neighboring V2O7" model described above (Figure 7). The structural parameters and refinement details are given in Table 2 and Table 4. Comparing fit residuals, confidence limits, and F parameters a significant improvement was visible. Apparently, both local structure in VxOy species and interaction with SiO2 support are required to describe the FT(χ(k)*k3) of dehydrated VxOy/SBA-15 samples.
Limitations of XAFS analysis of dehydrated VxOy/SBA-15
Eventually, the limitations of the XAFS analysis of dehydrated VxOy/SBA-15 presented here should be discussed. XAFS is not a very sensitive technique with respect to distinguishing and identifying additional minority species. Experimental XAFS spectra are clearly dominated by the signal of the majority phase. Hence, the presence of minority vanadium oxide species in dehydrated VxOy/SBA-15 with concentrations of less than ~5% cannot be excluded. Only if the contribution of additional phases amounts to more than ~5-10%, distortion of the FT(χ(k)*k3) and deviation from the model structure assumed will be detectable. This holds in particular, if these minority species happen to be less ordered than the majority phase.
Moreover, XAFS is an averaging technique. Certainly, higher shells should be properly taken into account and various references should be measured for comparison. Even then, however, it may remain difficult to unambiguously distinguish between mixtures of various species or structures. Hence, alternative scenarios with different vanadium oxide species need to be considered and discussed. An equal mixture of isolated VO4 and neighboring V2O7 units, for instance, may exhibit a XAFS FT(χ(k)*k3) similar to that presented here. For two reasons this assumption is not very likely. First, the V-O distances in the first "VO4" shell of the two species would have to be the same. Otherwise a strong reduction in amplitude of the first V-O peak in the FT(χ(k)*k3) caused by destructive interference would be discernible. Second, isolated VO4 would not contribute to the FT(χ(k)*k3) in the range from 2 Å to 4 Å. Thus, reduction in amplitude and much higher σ2 parameters compared to reference Mg2V2O7 would be detectable. Both are not observed in the EXAFS analysis of dehydrated VxOy/SBA-15 presented here (Figure 3(b), Table 2). A similar statement holds for isolated V2O7 units or a less ordered arrangement of neighboring V2O7 units. Both would result in a reduction in FT(χ(k)*k3) amplitude because of missing contribution in the 2-4 Å range or destructive interference caused by a broadened distribution of distances, respectively. Moreover, for isolated or less ordered V2O7 units a single V-V distance would suffice to describe the experimental XAFS spectrum. This was also not observed in the XAFS analysis procedure employed. Eventually, higher V-Si distances may have to be considered in addition to a V-Si distance of ~2.5 Å (Table 2). However, a significant contribution of V-Si distances at more 3.0 Å range in the FT(χ(k)*k3) would require a highly ordered arrangement of VxOy species on the SiO2 support and a very narrow distance distribution. This seems to be unlikely.
In total, assuming a structural arrangement of vanadium centers in dehydrated VxOy/SBA-15 that has already been established for reference vanadium oxides (i.e. Mg2V2O7) is simple and results in a good agreement with experimental data. More complex and artificially constructed arrangements of VxOy species supported on SiO2 may be conceivable but appear to be less likely. Both XANES and EXAFS analysis corroborate a local structure around the majority of V centers in dehydrated VxOy/SBA-15 similar to the ordered arrangement of neighboring V2O7 dimers in the structure of Mg2V2O7.
Local structure of hydrated VxOy/SBA-15 Comparison to V oxide references
The EXAFS χ(k)*k3 of hydrated VxOy/SBA-15 (as-prepared) with different V loadings are depicted in Figure 10. The usable spectral ranged extended from 2.7 Å through 10.5 Å. The V K edge XANES spectra and the FT(χ(k)*k3) of hydrated VxOy/SBA-15 are shown in Figure 11. The Fourier transformed χ(k)*k3 and the V K near edge spectra of hydrated VxOy/SBA-15 are compared to those of vanadium oxide references in Figure 12. The XANES spectrum of hydrated VxOy/SBA-15 resembles that of MgV2O6, [H3N(CH2)4]6V10O28, and V2O5 (Figure 12(a)). In these vanadium oxide references vanadium centers exhibit a distorted octahedral or distorted square pyramidal coordination. The XANES spectra of hydrated VxOy/SBA-15, V2O5, MgV2O6, and [H3N(CH2)4]6V10O28 show a similar height of the pre-edge peak. Because the pre-edge peak height is determined by the coordination of the vanadium centers , hydrated VxOy/SBA-15 also appears to exhibit a distorted square pyramidal coordination of V centers. This has also been observed by Bell et al.  and others.
The range of potential model structures describing the local structure of hydrated VxOy/SBA-15 can be further narrowed when comparing the corresponding FT(χ(k)*k3) (Figure 12(b)). Considering peak positions and relative peak heights in the FT(χ(k)*k3), it appears that from the references available the FT(χ(k)*k3) of V2O5 resembles that of hydrated VxOy/SBA-15. Because of the lower intensity in the FT(χ(k)*k3) hydrated VxOy/SBA-15 may possess a more disordered structure compared to that of crystalline V2O5. This is in good agreement with a detailed comparison of the corresponding XANES spectra. On the one hand, the positions of the various peaks in the XANES of hydrated VxOy/SBA-15 are similar to that of V2O5 (Figure 12(a)). On the other hand, the lower peak intensities are also indicative of a disordered V2O5 like structure of the vanadium oxide species in hydrated VxOy/SBA-15.
Local structure of hydrated VxOy/SBA-15 XAFS refinement of "V2O5" based model structure
Comparison of the XANES and FT(χ(k)*k3) of hydrated VxOy/SBA-15 to those of various references identified V2O5 as most suitable model structure for a detailed EXAFS analysis. Ammonium decavanadate decomposes to V2O5 during treatment in air at temperatures above 773 K. Calcination of the materials studied here will most likely result in formation of the dehydrated species as described above. Re-hydration upon exposure to ambient conditions resulted in vanadium oxide species supported on SBA-15 with a local structure similar to that of V2O5. A detailed discussion of the underlying formation mechanisms is beyond the scope of this work. Therefore, a theoretical XAFS function calculated on the basis of a V2O5 model structure (ICSD 60767 ) was refined to the experimental FT(χ(k)*k3) of hydrated VxOy/SBA-15. Details of the XAFS refinement procedure and the structural parameters obtained are given in Table 5. In addition, the V2O5 model structure was refined to the FT(χ(k)*k3) of V2O5 to validate the procedure chosen. Good agreement between the theoretical XAFS function of a V2O5 model structure and the FT(χ(k)*k3) of hydrated VxOy/SBA-15 and V2O5 was obtained (Figure 13).
As described above, the validity of the XAFS analysis approach chosen was evaluated by calculating confidence limits and F parameters (Table 5). The model structure employed corresponds to the local structure around V centers in bulk V2O5 (Table 6). In fitting procedure #1 there V-O distances (1.6 Å, 1.8 Å, and 2.0 Å) and two σ2 (one for R = 1.6 Å and one for all other V-O distances) in the first V-O shell were allowed to vary independently. Additionally, three V-V distances (3.1 Å, 3.4 Å, and 3.6 Å) with the same σ2 were refined. Moreover, E0 was also allowed to vary in fitting procedure #1. Again because of Nind = 18 and Nfree = 10 refinement procedure #1 would be considered reliable according to the Nyquist criteria. Reasonable confidence limits and F = 0 were calculated for the V-V distances and σ2(V-V) parameter. However, rather high confidence limits of the V-O distances of ± 0.05 Å and F parameters of 0.8 for both σ2(V-O) parameters were obtained with procedure #1. Moreover, E0 exhibited a confidence limit of ± 17.0 and F = 0.9. Hence, fitting procedure #1 clearly exceeds the number of meaningful parameters. Therefore, the fitting procedure was modified and the number of free parameters was reduced. E0 was kept invariant again in the refinement, two V-O distances at ~1.6 Å and 1.9 Å, and one σ2(V-O) parameter were used. In contrast to procedure #1, procedure #2 yielded reasonable confidence limits (e.g. ± 0.02 for V-O distances) and acceptable F parameters (mostly F = 0).
Table 5 indicates a small increase in the various V-O and V-V distances from V2O5 to hydrated VxOy/SBA-15. Intercalation of water in hydrated VxOy/SBA-15 may be accompanied by increasing nearest neighbor distances. More pronounced, though, is the increase in the disorder parameter σ2 for the V-O and V-V scattering scattering paths used in the XAFS refinement for hydrated VxOy/SBA-15 (Table 5). In particular, V-V contributions are strongly damped in the FT(χ (k)*k3) of hydrated VxOy/SBA-15 indicating an increased disorder in the local structure of hydrated VxOy species supported on SBA-15 compared to bulk V2O5.
The structural similarity between hydrated vanadium oxide species supported on SiO2 and V2O5 has previously been observed by Raman spectroscopy . Evidently, the local structure of hydrated VxOy/SBA-15 used here is very similar to other materials previously described in the literature. Dehydration should therefore result in a similar structure of the dehydrated phase. In addition to 10.8 wt % VxOy/SBA-15, samples with lower loadings of 2.7 wt % and 5.4 wt % V were measured (Figure 10and Figure 11) and analyzed according to the procedure described above. Very similar results were obtained for the hydrated state of the low-loading samples compared to 10.8 wt % hydrated VxOy/SBA-15. Apparently, in the range of V loadings from ~3 to 11 wt % the local structure of both hydrated and dehydrated VxOy/SBA-15 is largely independent of the amount of vanadium oxide supported on SBA-15.
X-ray absorption spectroscopy is a very suitable technique for studying the local structure of dispersed metals or metal oxides on various support materials. Conventional XAFS analysis consists of finding a suitable model structure and fitting the corresponding theoretical XAFS functions to the experimental data. Because the number of potential parameters often exceeds the number of "independent" parameters, evaluating the reliability and significance of a particular fitting procedure is mandatory. Therefore, the number of independent parameters (Nyquist) alone is not sufficient. Here, we have employed confidence limits and F parameters to identify suitable analysis procedures. The local structure of vanadium oxide supported on nanostructured SiO2 (SBA-15) was investigated. Three samples with different vanadium loadings (i.e. 2.7 wt %, 5.4 wt %, and 10.8 wt %) were employed. Thermal treatment in air at 623 K resulted in characteristic structural changes of the V oxide species. The local structure of dehydrated VxOy/SBA-15 was described best by assuming a model structure consisting of an ordered arrangement of neighboring V2O7 units. This is in good agreement with recent NEXAFS studies and theoretical calculations that also concluded the presence of V-V bonds in the VxOy species supported on SBA-15 . Moreover, the local structures of both hydrated and dehydrated VxOy/SBA-15 were found to be independent of the V loading over the range employed. With respect to XAFS data previously presented in the literature and interpreted in terms of isolated VO4 units it can be suggested that including contributions of higher shells would also lead to conclude polymeric VxOy units. Comparing the influence of surface properties and structure of various support materials will be the subject of future work. Eventually, onset of catalytic activity in selective propene oxidation was detected at about 573 K. In situ XAS data measured under reaction conditions indicated, that the characteristic ordered arrangement of V2O7 dimers in the local structure of dehydrated VxOy/SBA-15 persisted. Future studies on VxOy species supported on SBA-15 as model systems for vanadium based selective oxidation catalysts will have to take the presence of V2O7 species rather than isolated VO4 units into account.
Silica SBA-15 was prepared according to literature procedures . Details of preparation and characterization of the same catalysts were described elsewhere [9, 10, 14]. Briefly, SBA-15 was functionalized by adding 3-aminopropyltrimethoxysilane (APTMS) to a suspension of SBA-15 in toluene at 338 K. The suspension was stirred for 12 hours. The contents were filtered, washed and finally stirred in 0.3 M HCl for 12 hours. The contents was filtered again, washed with water and dried in air overnight (functionalized SBA-15). The vanadium oxides supported on SBA-15 were prepared by adding appropriate amounts of butylammonium decavanadate  to a suspension of functionalized SBA-15 in water. The resulting powder was calcined at 823 K for 12 hours. The results of the N2 physisorption analysis of the SBA-15 and the SBA-15 supported vanadium oxide samples are given in and have been discussed in detail previously together with detailed structural characterization . Importantly, in the presence of vanadium oxide the hexagonal structure of SBA-15 is preserved, the mesopores remain accessible to reactants, and the vanadia species are located inside the pores of SBA-15.
The vanadium oxides supported on SBA-15 obtained are denoted as hydrated VxOy/SBA-15 (as-prepared) or dehydrated VxOy/SBA-15 (after thermal treatment). In addition to VxOy/SBA-15 several vanadium oxide reference compounds with an average valance of +5 were measured. These were either used as-purchased (V2O5 (Alfa Aesar 99.8%), NH4VO3 (Riedel de Haën, 99.5%) and Na3VO4 (Alfa Aesar, 99.9%)) or were prepared according to literature procedures (Mg2V2O7, MgV2O6, Mg3V2O8 , and [H3N(CH2)4]6V10O28 ).
X-ray absorption spectroscopy (XAS)
In situ transmission XAS experiments were performed at the V K edge (5.465 keV) at beamline E4 at the Hamburg Synchrotron Radiation Laboratory, HASYLAB, using a Si(111) double crystal monochromator. The energy range used for V K near edge scans (XANES) and extended XAFS scans (EXAFS) was 5.4-5.7 keV (~3 min/scan) and 5.4-6.0 keV (~20 min/scan), respectively. For in situ and ex situ XAFS measurements samples were mixed with BN and PE, respectively, and pressed into self-supporting pellets (5 mm and 13 mm in diameter, respectively). In order to obtain an edge jump, Δ μx, at the V K below 1.0, 2 mg of 10.8 wt % and 5.4 wt % VxOy/SBA-15, 3 mg of 2.7 wt % VxOy/SBA-15, ~1 mg for bulk vanadium oxides (NH4VO3, V2O5, [H3N(CH2)4]6V10O28, Na3VO4) diluted with BN (~15 mg), and 3-6 mg for bulk vanadium oxides (Mg3V2O8, MgV2O6) diluted with PE (~200 mg) were employed. Transmission XAS measurements were performed in an in situ cell described previously . Dehydration of VxOy/SBA-15 was conducted in 20% O2 and He (total flow 30 ml/min) in a temperature range from 293 K to 623 K at a heating rate of 5 K/min and a holding time of 30 min at 623 K. Reaction tests were performed in 5% propene and 6% O2 in He in the temperature range from 293 K to 723 K (5 K/min, total flow 30 ml/min). The gas atmosphere was analyzed using a noncalibrated mass spectrometer in a multiple ion detection mode (QMS200 from Pfeiffer). Ex situ XAFS measurements were performed in He atmosphere at room temperature.
X-ray absorption fine structure (XAFS) analysis was performed using the software package WinXAS v3.2 . Background subtraction and normalization were carried out by fitting linear polynomials to the pre-edge and 3rd degree polynomials to the post-edge region of an absorption spectrum, respectively. The extended X-ray absorption fine structure (EXAFS) χ(k) was extracted by using cubic splines to obtain a smooth atomic background μ0 (k). The FT(χ(k)*k3), often referred to as pseudo radial distribution function, was calculated by Fourier transforming the k3-weighted experimental χ(k) function, multiplied by a Bessel window, into the R space. EXAFS data analysis was performed using theoretical backscattering phases and amplitudes calculated with the ab-initio multiple-scattering code FEFF7 . EXAFS refinements were performed in R space simultaneously to magnitude and imaginary part of a Fourier transformed k3-weighted and k1-weighted experimental χ (k) using the standard EXAFS formula . This procedure strongly reduces the correlation between the various XAFS fitting parameters. Structural parameters allowed to vary in the refinement were (i) disorder parameter σ2 of selected single-scattering paths assuming a symmetrical pair-distribution function and (ii) distances of selected single-scattering paths. Coordination numbers (CN), E0 shifts, and amplitude reduction factor S02 were kept invariant in the final fitting procedures. Correlations of specific parameters to reduce the number of free running parameters and to improve the stability of the refinement are described below.
The statistical significance of the fitting procedure employed was carefully evaluated in three steps. First, the number of independent parameters (Nind) was calculated according to the Nyquist theorem Nind = 2/π*ΔR* Δk + 2. In all cases the number of free running parameters in the refinements was well below Nind. Second, confidence limits were calculated for each individual parameter. In the corresponding procedure, one parameter was successively varied by a certain percentage (i.e. 0.05% for R and 5% for σ2) and the refinement was restarted with this parameter kept invariant. The parameter was repeatedly increased or decreased until the fit residual exceed the original fit residual by more than 5%. Eventually, the confidence limit of the parameter was obtained from linear interpolation between the last and second last increment for an increase in fit residual of 5%. This procedure was consecutively performed for each fitting parameter. Third, a so-called F test was performed to assess the significance of the effect of additional fitting parameters on the fit residual. The corresponding procedure was adopted from the well-known library "Numerical Recipes in C" where it is described in detail . In short, one parameter was varied by a certain percentage (i.e. between 2 and 8% for R and between 10 and 80% for σ2) and the refinement was restarted with this parameter kept invariant. Subsequently, the difference between experimental and theoretical function (i.e. magnitude and imaginary part of FT(χ(k)*k3) for a refinement in R space) was calculated and compared to that of the original refinement. The corresponding F parameter ranges between 0.0 and 1.0, where F = 1.0 indicates an insignificant change in the fit residual, while F = 0.0 indicates a highly significant change in fit residual. The iterative procedure was terminated when the corresponding F parameter was below 0.7. Fit parameters with F = 0.8 or higher are most likely strongly correlated and may be statistically insignificant. These parameters should be kept invariant in the refinement. Eventually, this procedure was also consecutively performed for each fitting parameter.
Ressler T, Walter A, Huang ZD, Bensch W: Structure and properties of a supported MoO3-SBA-15 catalyst for selective oxidation of propene. J Catal. 2008, 254: 170-179. 10.1016/j.jcat.2007.12.012.
Zhao DY, Huo QS, Feng JL, Chmelka BF, Stucky GD: Nonionic Triblock and Star Diblock Copolymer and Oligomeric Surfactant Syntheses of Highly Ordered, Hydrothermally Stable, Mesoporous Silica Structures. J Am Chem Soc. 1998, 120: 6024-6036. 10.1021/ja974025i.
Hess C: Characterization of the synthesis and reactivity behavior of nanostructured vanadia model catalysts using XPS and vibrational spectroscopy. Surf Sci. 2006, 600: 3695-3701. 10.1016/j.susc.2005.11.063.
Hess C, Drake IJ, Hoefelmeyer JD, Tilley TD, Bell AT: Partial Oxidation of Methanol Over Highly Dispersed Vanadia Supported on Silica SBA-15. Catal Lett. 2005, 105: 1-8. 10.1007/s10562-005-7997-x.
Liu YM, Cao Y, Yi N, Feng WL, Dai WL, Yan SR, He HY, Fan KN: Vanadium oxide supported on mesoporous SBA-15 as highly selective catalysts in the oxidative dehydrogenation of propane. J Catal. 2004, 224: 417-428. 10.1016/j.jcat.2004.03.010.
Liu W, Lai SY, Dai H, Wang S, Sun H, Au CT: Oxidative dehydrogenation of n-butane over mesoporous VOx/SBA-15 catalysts. Catal Lett. 2007, 113: 147-154. 10.1007/s10562-007-9023-y.
Keller DE, Visser T, Soulimani F, Koningsberger DC, Weckhuysen BM: Hydration effects on the molecular structure of silica-supported vanadiumoxide catalysts: A combined IR, Raman, UV-vis and EXAFS study. Vib Spectrosc. 2007, 43: 140-151. 10.1016/j.vibspec.2006.07.005.
Venkov TV, Hess C, Jentoft FC: Redox Properties of Vanadium Ions in SBA-15-Supported Vanadium Oxide: An FTIR Spectroscopic Study. Langmuir. 2007, 23: 1768-1777. 10.1021/la062269n.
Hess C, Wild U, Schlögl R: The mechanism for the controlled synthesis of highly dispersed vanadia supported on silica SBA-15. Microporous and Mesoporous Materials. 2006, 95: 339-349. 10.1016/j.micromeso.2006.06.010.
Hess C, Tzolova-Müller G, Herbert R: The Influence of Water on the Dispersion of Vanadia Supported on Silica SBA-15: A Combined XPS and Raman Study. J Phys Chem C. 2007, 111: 9471-9479. 10.1021/jp0713920.
Hess C: Direct correlation of the dispersion and structure in vanadium oxide supported on silica SBA-15. J Catal. 2007, 248: 120-123. 10.1016/j.jcat.2007.02.024.
Oyama ST, Went GT, Lewis KB, Bell AT, Somorjai GA: Oxygen Chemisorption and Laser Raman Spectroscopy of Unsupported and Silica-Supported Vanadium Oxide Catalysts. J Phys Chem B. 1989, 93: 6786-6790. 10.1021/j100355a041.
Gao X, Bare SR, Weckhuysen BM, Wachs IE: In Situ Spectroscopic Investigation of Molecular Structures of Highly Dispersed Vanadium Oxide on Silica under Various Conditions. J Phys Chem B. 1998, 102: 10842-10852. 10.1021/jp9826367.
Hess C, Hoefelmeyer JD, Tilley TD: Spectroscopic Characterization of Highly Dispersed Vanadia Supported on SBA-15. J Phys Chem B. 2004, 108: 9703-9709. 10.1021/jp037714r.
Schraml-Marth M, Wokaun A: Spectroscopic Investigation of the Structure of Silica-supported Vanadium Oxide Catalysts at Submonolayer Coverages. J Chem Soc Faraday Trans. 1991, 87 (16): 2635-2646. 10.1039/ft9918702635.
Gao X, Wachs IE: Investigation of Surface Structures of Supported Vanadium Oxide Catalysts by UV-vis-NIR Diffuse Reflectance Spectroscopy. J Phys Chem B. 2000, 104: 1261-1268. 10.1021/jp992867t.
Bronkema JL, Bell AT: Mechanistic Studies of Methanol Oxidation to Formaldehyde on Isolated Vanadate Sites Supported on MCM-48. J Phys Chem C. 2007, 111: 420-430. 10.1021/jp0653149.
Olthof B, Khodakov A, Bell AT, Iglesia E: Effects of Support Composition and Pretreatment Conditions on the Structure of Vanadia Dispersed on SiO2, Al2O3, TiO2, ZrO2, and HfO2. J Phys Chem B. 2000, 104: 1516-1528. 10.1021/jp9921248.
Keller DE, Koningsberger DC, Weckhuysen BM: Molecular Structure of a Supported VO4 Cluster and Its Interfacial Geometry as a Function of the SiO2, Nb2O5, and ZrO2 Support. J Phys Chem B. 2006, 110: 14313-14325. 10.1021/jp060749h.
Keller DE, Koningsberger DC, Weckhuysen BM: Elucidation of the molecular structure of hydrated vanadium oxide species by X-ray absorption spectroscopy: correlation between the V ⋯ V coordination number and distance and the point of zero charge of the support oxide. Phys Chem Chem Phys. 2006, 8: 4814-4824. 10.1039/b609757j.
Keller DE, Airaksinen SMK, Krause AO, Koningsberger DC, Weckhuysen BM: Atomic XAFS as a Tool To Probe the Reactivity of Metal Oxide Catalysts: Quantifying Metal Oxide Support Effects. J Am Chem Soc. 2007, 129: 3189-3197. 10.1021/ja0667007.
Keller DE, Weckhuysen BM, Koningsberger DC: Application of AXAFS Spectroscopy to Transition-Metal Oxides: Influence of the Nearest and Next Nearest Neighbour Shells in Vanadium Oxides. Chem--Eur J. 2007, 13: 5845-5856. 10.1002/chem.200601128.
Weckhuysen BM, Keller DE: Chemistry, spectroscopy and the role of supported vanadium oxides in heterogeneous catalysis. Catal Today. 2003, 78: 25-46. 10.1016/S0920-5861(02)00323-1.
Tanaka T, Yamashita H, Tsuchitani R, Funabiki T, Yoshida S: X-Ray Absorption (EXAFS/XANES) Study of Supported Vanadium Oxide Catalysts. J Chem Soc Faraday Trans. 1988, 84 (9): 2987-2999. 10.1039/f19888402987.
Keller DE, de Groot FMF, Koningsberger DC, Weckhuysen BM: ËO4 Upside Down: A New Molecular Structure for Supported VO4 Catalysts. J Phys Chem B. 2005, 109: 10223-10233. 10.1021/jp044539l.
Hawthorne FC, Calvo C: The crystal chemistry of the M+VO3 (M+ = Li, Na, K, NH4, Tl, Rb, and Cs) pyroxenes. J Solid State Chem. 1977, 22 (2): 157-170. 10.1016/0022-4596(77)90033-0.
Gopal R, Calvo C: Crystal Structure of Magnesium Divanadate, Mg2V2O7. Acta Crystallogr, Sect B. 1974, 30 (10): B2491-2493. 10.1107/S0567740874007400.
Cavalleri M, Hermann K, Knop-Gericke A, Hävecker M, Herbert R, Hess C, Oestereich A, Döbler J, Schlögl R: Analysis of silica-supported vanadia by X-ray absorption spectroscopy: Combined theoretical and experimental studies. J Catal. 2009, 262: 215-223. 10.1016/j.jcat.2008.12.013.
Wong J, Lytle FW, Messmer RP, Maylotte DH: K-edge absorption spectra of selected vanadium compounds. Phys Rev B. 1984, 30: 5596-5610. 10.1103/PhysRevB.30.5596.
Enjalbert R, Galy J: A Refinement of the Structure of V2O5. Acta Crystallogr, Sect C. 1986, C42: 1467-1469. 10.1107/S0108270186091825.
Roman P, Aranzabe A, Luque A, Gutierrez-Zorilla JM: Preparation and solid state characterization of some alkylammonium decavanadates. Crystal structure of the hexakis(n-hexylammonium) decavanadate dihydrate. Mater Res Bull. 1991, 26: 731-740. 10.1016/0025-5408(91)90062-Q.
Gao X, Ruiz P, Xin Q, Guo X, Delmon B: Preparation and characterization of three pure magnesium vanadate phases as catalysts for selective oxidation of propane to propene. Catal Lett. 1994, 23: 321-337. 10.1007/BF00811367.
Ressler T, Jentoft RE, Wienold J, Günter MM, Timpe O: In Situ XAS and XRD Studies on the Formation of Mo Suboxides during Reduction of MoO3. J Phys Chem B. 2000, 104: 6360-6370. 10.1021/jp000690t.
Ressler T: WinXAS: a Program for X-ray Absorption Spectroscopy Data Analysis under MS-Windows. J Synch Rad. 1998, 5: 118-122. 10.1107/S0909049597019298.
Rehr JJ, Booth CH, Bridges F, Zabinsky SI: X-ray-absorption fine structure in embedded atoms. Phys Rev B. 1994, 49: 12347-12350. 10.1103/PhysRevB.49.12347.
Ressler T, Brock SL, Wong J, Suib SL: Multiple-Scattering EXAFS Analysis of Tetraalkylammonium Manganese Oxide Colloids. J Phys Chem B. 1999, 103: 6407-6420. 10.1021/jp9835972.
Teukolsky A, Vetterling WT, Flannery BP: Numerical Recipes: The Art of Scientific Computing. 2007, Cambridge: Cambridge University Press, Third
The Hamburg Synchrotron Radiation Laboratory, HASYLAB, is acknowledged for providing beamtime for this work. This research was supported by the Deutsche Forschungsgemeinschaft (DFG). The authors acknowledge support from the Cluster of Excellence "Unifying Concepts in Catalysis" (DFG). C.H. thanks the DFG for an Emmy Noether fellowship (DFG).
The authors declare that they have no competing interests.
AW participated in XAFS data measurements and analysis, and drafted the manuscript. RH performed the sample preparation and N2 adsorption analysis. CH participated in design and coordination of the study, sample preparation, and N2 adsorption analysis. TR participated in XAFS data measurements and analysis, and design and coordination of the study. All authors read and approved the final manuscript.
Authors’ original submitted files for images
Below are the links to the authors’ original submitted files for images.