In this work we have studied three natural aluminosilicates reported in previous works ^{20,21,22,23,24} : the certified sample VL1 with a proposed content of SiO₂ (1.16 wt.%), Al₂O₃ (37.38 wt.%), Fe2O3 (35.72 wt.%), TiO₂ (3.15 wt.%) LOI (22.54 %)20; the certified sample SLB1 with a proposed composition of: SiO₂ (1.93 wt. %), Al₂O₃ (45.50 wt.%), Fe₂O₃ (24.03 wt.%), TiO₂ (1.75 wt.%) LOI (26.79 %) ²¹ , both samples analyzed before and after being used in catalytic tests in oxidative dehydrogenation of propane to propene; the third material is a natural catalyst (ULA1) composed ²² by Fe (62.55 wt.%), Al (0.40 wt.%), Si (14.20 wt.%), Ti (0.10 wt.%); which was analyzed before and after being used as a catalyst in hydrocarbon synthesis. The samples used in catalytic reactions were placed into the XPS spectrometer, after being exposed to atmospheric conditions. For identification purposes a sample studied before or after a catalytic test will be designated sample BT or sample AT respectively. The samples were analyzed in the temperature range from 523 to 773 K. Ultra pure oxides samples: Fe₂O₃, SiO₂, Al₂O₃ were also analyzed as well as a mechanical mixture (mixture M) with a 4:2:1 weight proportion (equivalent to 57.14 wt.% Fe₂O₃, 28.57 wt.% Al₂O₃, and 14.28 wt.% SiO₂).

We also include previous results of an ultra pure sample of MgO for comparison of the Auger structures and because in its spectrum the structures assigned by Wagner ¹ and Fuggle ² can be clearly observed. In all the tables and in figure 5h we include our results, which differ to those of Wagner but agree to those of Fuggle.

A VSW spectrometer with a preparation chamber and Ar⁺ ion etching facilities was used for this work in conjunction with a Leybold LHS 10S spectrometer whose specifications can be found elsewhere^16,18. Vacuum in the spectrometers, during measurements, was in the 10^-9mbar range. Its hemispherical analyzer was operated at constant pass energy of 22.4 eV. Non-monochromatic Al Kα radiation was employed as X-ray photon source with a constant 300 watts power. The C1s binding energy at 285.0 eV of adventitious carbon was used, whenever possible, as an acceptable binding energy reference; however, when the intensity of this peak was very low (after ion etching), the Al 2p level binding energy in Al₂O₃ (74.6 eV) or the Si 2p binding energy in SiO₂ (103.8 eV) were used as internal energy references, depending on the sample under study. The mechanical mixture and ULA1 sample exhibited a differential charge effect ^18,24 which leads to inaccuracies on determining the binding energies, but this can be solved taking internal references like the Si 2p binding energy (103.8 eV) for the peaks originating from the sample physical region consisting of SiO₂, the O 1s binding energy (530.3 eV) for the oxygen bonded to iron for those XPS peaks rising from the sample physical region consisting of Fe₂O₃ , and the Al 2p level (74.6 eV) for the peaks originating from the sample region consisting of Al₂O₃

Samples were mounted as reported previously ^22,23,24 . Each sample, depending on the spectrometer used, was subjected to a sequence of sample treatments performed in the spectrometer preparation chamber, as shown in table 1. Ion bombardment was done with 3 KeV Ar⁺ ions in 15 minutes cycles. Annealing was done in vacuum at 773 K for approximately 6 hours. Both treatments, the hydrogen reduction (in 1x 10^-5 mbar of H₂) and re-oxidation (in 1x 10^-5 mbar of O₂), were performed in the spectrometer preparation chamber at 773 K for a period between 4-20 hours depending on the sample under study.

Due to their hygroscopic nature, all the samples (excepting VL1 BT, SLB1BT, SiO₂ and Al₂O₃) were heated at 423 K before introducing them into the analysis chamber in order to preserve its vacuum. For this reason, VL1 BT and SLB1 BT samples were calcined due to the high content of interstitial water. SiO2 and Al2O3 were also calcined before XPS analysis. The calcination period done at 773 K lasted between 5 and 12 hours depending on the sample.

The procedures to exploit quantitatively the spectra are given in detail in Mendialdua et al. ²⁵ and other works of our lab.

According to its composition this sample should have a N_Al/N_Fe ≈ 3 bulk concentration ratio, while the XPS surface analysis gives a 3.9 ratio for the SLB1 BT reoxidized sample, indicating an Al surface enrichment.

Oxygen Auger spectra corresponding to transitions KL₁V to KVV (KL₁L₂₃ to KL₂₃L₂₃) for SLB1, Al₂O₃ and Fe2O3 are shown in figure 1, where the ordinate scale has been normalized to one for the maximum value of each spectrum. The oxygen Auger region for SLB1 does not exhibit, on the high energy side of the KL₂₃L₂₃ transition, the obvious structure shown by Al₂O₃ in this region. The KL₁L₂₃ transition peak has a less defined structure than in Al₂O₃ and Fe₂O₃.

Fig. 1 Fig. 1 O KLL Auger spectra from: a) Fe2O3, b) Al2O3, c) SLB1 sample.

Fig. 2 Fig. 2 Comparison of oxygen Auger region for sample SLB1, obtained from a linear combination of spectra for Al2O3 and Fe2O3 using the proper coefficients obtained from a) the surface NAl/NFe ratio, b) the bulk NAl/NFe ratio, with c) experimental oxygen Auger region for sample SLB1.

One might think that SLB1 is a mixture of Al₂O₃ and Fe₂O₃; however, a linear combination of spectra (see figure 2, with the ordinate scale normalized to one) for Al₂O₃ and Fe₂O₃, using the proper coefficients obtained from the bulk and surface N_Al/N_Fe ratios, does not correspond to the SLB1 spectrum obtained experimentally. Comparing the values of the Auger parameters and their shifts for SLB1 with those of Al₂O₃ and Fe₂O₃ (see tables 1 and 2) it can be seen that the values of α´ and ∆α´ are close to those of SiO₂ despite its composition; the values of β´ and ∆ β´ are closer to those of Al₂O₃. Analogously, the values of the hole-hole correlation energies U(2p2p) and U(2s2p) are close to those of Al2O3. However, the less defined KL₁L₂₃ structure indicates a valence band structure different to Al₂O₃ and Fe₂O₃.

It appears that SLB1 is composed of iron aluminate predominating over the aluminum and iron oxides. This is consistent with the inertness of this sample to different sample treatments in contrast with the reactivity exhibited by Fe₂O₃, whether alone or in a mechanical mixture with SiO₂ and Al₂O₃, under the same conditions; this supports the fact that iron in SLB1 is in a higher stability state than that shown in Fe₂O₃. However, SLB1 experiences a transformation once it is subjected to catalytic tests in oxidative dehydrogenation reactions of propane into propene ²⁶ , since its modified Auger parameter (see table 1) has a value close to that of alumina.

According to its composition, this sample has a N_Al/N_Fe = 1.7 bulk concentration ratio, while the XPS data gives a surface concentration ratio N_Al/N_Fe = 1.0 indicating Al depletion on the surface region. No carbonates were detected by XPS in the C 1s region and the intensity of the CO region is very low, indicating a very small contribution of this oxygen in the Auger region.

A comparison of the oxygen Auger region for VL1, Al₂O₃, and Fe₂O₃ is shown in figure 3. A suitable linear combination of the Al₂O₃ and Fe₂O₃ spectra according to the NAl/NFe ratios is shown in figure 4. From this figure, it is deduced that this sample is not an Al₂O₃, and Fe₂O₃ mixture. It is important to point out (see figure 5d) that the O1s peak shape does not correspond to that expected for the N_Al/N_Fe surface ratio. Its modified Auger parameter (1040.8) (see table 1, column 7) is completely different from that of alumina, iron oxide, lying between the two.

Fig. 3 Fig. 3 O KLL Auger spectra of a) Al2O3, b) Fe2O3, c) sample VL1

Table 1

(*) values obtained from an estimation of the KL1L2 position in the oxygen bonded to iron. AR: sample in the as received condition, BT : before catalytic tests, AT : after catalytic tests.

The spectral structure (see figure 5) does not permit obtaining two values for the Auger parameter, as can be done for ULA1 and mechanical mixture (MM) (see figures 5f and 5g respectively). The spectral region for the KL₁L₂₃ transition does not show practically any structure, contrary to what it is observed for Al₂O₃ and Fe₂O₃. This evidence indicates the presence of some aluminate (and/or ferrate) on the sample’s surface, different to that present on sample SLB1, also shown by the difference in the Auger parameters (see table 1) as well as in table 2, where it can be observed that the values of ∆α´ and ∆β´ differ from those of Fe₂O₃, Al₂O₃ and SiO₂; the hole-hole repulsion energies U(2p2p) and U(2s2p) are also very differ ent from those three samples. On the other hand, it can also be pointed out that VL1 responds differently to catalytic tests than SLB1.

Fig. 4 Fig. 4 Comparison of the oxygen Auger region for sample VL1, obtained from a linear combination of spectra Al2O3 and Fe2O3 using the proper coefficients given by a) the surface NAl/NFe ratio, b) the bulk NAl/NFe ratio, with c) experimental oxygen Auger region for sample VL1.

Fig. 5 Fig. 5 O1s XPS spectral regions for samples: a) Fe2O3, b) Al2O3, c) SiO2 AR, d) VL1, e) SLB1, f) ULA1 AT, g) mixture M AR, h) MgO.

The oxygen Auger spectral region for Fe₂O₃, SiO₂, Al₂O₃ and mixture M is shown in figure 6. The structures related to the KL₁L₂₃ region present in SiO₂, Al₂O₃ and Fe₂O₃ are also present in the mechanical mixture. In addition, the KL₂₃L₂₃ transition of oxygen bonded to both iron and silicon allows us to define two Auger parameters for this sample; one in the SiO₂ region and the other in the Fe₂O₃ region.

The mechanical mixture exhibits a differential charge effect between the silicon and iron physical regions, which was easily observed in the O 1s XPS peak; the C 1s peak spectrum for this sample has a FWHM remarkably superior than in Fe₂O₃, due to the differential charge effect. However, the low intensity of this spectrum, that indicates the low content of carbon in this sample, did not permit peak fitting. No carbonate signal was detected and the CO contribution was insignificant. The differential charge effect is also present in the Auger spectra, but does not affect the Auger parameters determination as expected. The spectrum g in figure 7 permits, by an approximate spectral decomposition, the determination of the energy values of the O KL₂L₃ transition in the corresponding region of SiO₂ and Fe₂O₃. These values are presented in table 1.

Table 2 Table 2 Auger parameter shifts and hole-hole correlation energies for oxygen in the samples studied

Fig. 6 Fig. 6 O KLL Auger spectra from: a) Fe2O3, b) Al2O3, c) SiO2 and d) mixture M.

Fig. 7 Fig. 7 Oxygen KLL Auger Spectral regions for samples: a) Fe2O3, b) Al2O3, c) SiO2AR, d) VL1, e) SLB1, f) ULA1, AT g) Mixture M AR, h) MgO

Fig. 8 Fig. 8 A comparison among the oxygen Auger spectra for samples: a) SiO2, b) Fe2O3, c) ULA1, d) mixture M.

ULA1 also exhibits a differential charge effect. This effect should be taken into account when comparing Auger spectra from samples exhibiting differential charge effect and those, which have a homogenous charge effect (see figure 8). In this case ULA1 and mixture M are considered as having a homogenous charge effect to obtain a unique kinetic energy scale. This implies that the KL_23.23 transition peak position from oxygen bonded to silicon (or to iron; one of the two) are fictitious depending on which value of the charge effect is used, since for these samples there are two values for the charge referencing procedure.

The C 1s spectrum, not shown here, for this sample shows a FWHM remarkably larger to that observed in Fe₂O₃, in corre spondence to the differential charge effect present on this sample ²³ . No carbonates were found in this sample and the CO contribution was very low.

In figure 8 the oxygen Auger spectrum of ULA1 AT (spectrum c) is compared to those of SiO₂ (spectrum a), Fe₂O₃ (spectrum b) and Mixture M (spectrum d); the differential charge effect is corrected in spectrum c according to its value in the iron region. A good correspondence is present among the peaks similar to previous reports16, that show mainly Fe₂O₃ and SiO₂ , with no iron silicate present. Structure C1 in figure 8 (spectrum c) and figure 7 (spectrum f) does not appear in the corresponding spectrum of the mechanical mixture (figure 8 spectrum d and figure 7 spectrum g), without resorting to a peak decomposition, probably due to the fact that in the latter sample there is a contribution from the oxygen bonded to Al₂O₃.

Structures A, B, C, D in figure 7 have been assigned by Wagner ¹ to the transitions: KL₁L₂ (¹P₃), KL₁L₃ (³P), KL₂L₂ (¹S₀), and KL L (¹D ). Structure C ascribed to the KL L (1S ) tran sition appears, according to Wagner, in ionic compounds like LiF and Na₂CrO₄. This structure is present in all the compounds studied in this report except VL1 and SLB1, and it is clear in the more covalent SiO₂ than in MgO where the structure is less sharp.

Structure E in figure 7h has been attributed, by Fuggle et al. ² , to a KL23M transition (where m indicates a hole in a metal like orbital) for adsorbed oxygen on metals. However, this explanation could be extended to lattice oxygen, assuming hybridization between the anionic and cationic levels. Wagner et al. ¹ attribute this structure to OHgroups or H²O present. This is supported by the 20 eV energy difference between structure E and C. The energy for structure (E) does not correspond to that of OHspecies, which appears at energies between 508 eV and 510 eV and even less (503 eV) in organic compounds. If this structure corresponds to another oxygen species it should appear in the O 1s XPS peak, with low intensity due to the energy dependence of λ. Likewise, correlation should exist between this component´s intensity and that of the Auger structure in the KVV part of the spectrum. Figures 7h and 5h show that such correlation does not exist, making Fuggle´s explanation the most probable, thus this structure´s intensity will depend on both the degree of hybridization and the involved levels occupation. In figure 7 spectrum e, representing the SLB1 oxygen Auger region, compared with VL1 (spectrum d) especially in the KL₁L₂(¹P) and KL₁L₃(³P) transition regions, shows differences which indicate different chemical environments. Their Auger parameters are listed in table 1.

The energy, for the different samples, of the final state ion with two holes in level L₂₃ are listed in table1 column 6; the highest value corresponds to SiO₂ and the lowest to Fe₂O₃.

Mixture M and ULA1 show the values corresponding to the presence of silicon and iron oxides. From the final state ion energy the following sequence can be established: Fe₂O₃ < SLB1 < Al₂O₃< SiO₂. Thus, the oxygen in Fe₂O₃ has a more polarizable medium, leading to more ionic bonding, with higher charge density around oxygen, than in SLB1, etc.

Likewise, insofar as the character of the bonding to be less ionic and the covalent character increase, the difference KVV – KL₁L₂₃ (table 1 column 4) increases also. The lowest value is obtained by Fe₂O₃ and the highest by SiO₂.

In the same way, in table 2 we can observe that the values of ∆α´ and ∆β´ follow the same tendency, i.e. the sample with the highest final ion energy (SiO₂) should have the lowest value of ∆ α´ and ∆ β´ and the one that has the lowest final ion energy (Fe₂O₃) will exhibit the highest values for ∆α´ and ∆β´, as in fact it happens. These results are coherent with the fact that SiO₂ shows the highest values for the hole-hole correlation energies U(2s,2p) and U(2p,2p) of the final ion state, and Fe₂O₃ have the lowest values for these quantities. The values of these energies for the samples Al₂O₃, VL1, SLB1 (ULA1 and the Mixture with its two set of values) are laid between those extremes.

It can be established from tables 1 and 2 and from figures 7 and 8, that the sample sequence according to the values of the Auger parameter α´ is opposed to that obtained from the values of the correlation energy U(2p,2p). In addition the sequence found by using the parameter β´ is contrary to that given by the correlation energy U(2s,2p), as it should be in accordance with the definitions of these quantities ¹³ .

Sample ordering according to α´ values is not the same as that obtained using the β´parameter; since these parameters are independent of both the charge effect and the work function, thus any significant difference (beyond the experimental uncertainties), should arise from the different chemical environment effect on the levels involved in the transitions: KL_23.23 for α´ and KL₁L₂₃ for β´. In this sense, it is interesting to point out that the sample sequence obtained according to the values of the difference KVV-KLV is nearly identical to that found by the values of the energy U(2s,2p). That difference increases as the ionicity of the oxygen bond with the corresponding cation decreases, that it is to say as the electron density on the oxygen initial state decreases, the final state ion energy increases ³ .