XAS characterization and CO oxidation on δ-alumina supported La, Mn, Co and Fe oxides

δ-Al2O3 supported La, Mn, Co and Fe containing catalysts were prepared by impregnation of δ-Al2O3 with citrate-type precursors and calcination at 1073 K. The catalysts were characterized by X-ray diffraction (XRD), UV-Vis diffuse reflectance spectroscopy (DRS), X-ray absorption spectroscopy (XAS), and BET specific surface area determination.

XRD revealed the presence of δ-Al2O3 in all cases and, at 30 wt.% of metal loading, of other single (α-Mn2O3 and α-Fe2O3) or mixed (LaAlO3, LaMnO3, CoAl2O4 and LaFeO3) oxide phases.

XAS suggested the formation of some oxide phases also at lower loading.

In particular, all Mn and 10 wt.% La–Mn containing samples revealed the formation of α-Mn2O3, while the 30 wt.% La–Mn bimetallic sample showed the formation of LaMnO3 perovskite.

All Co containing samples revealed the presence of CoAl2O4 spinel.

Fe containing samples showed the formation of α-Fe2O3, while La–Fe containing ones, that of LaFeO3 perovskite.Catalytic tests of CO oxidation were performed in the temperature range 300–800 K. The sample containing 30 wt.% of La and Mn in the form of LaMnO3 perovskite dispersed on δ-Al2O3 was found the most active among all the examined catalysts.

Most of the Co containing catalysts were found active at RT too, but they deactivated rapidly.

None of the Fe-based samples was active at RT and these catalysts were found, on average, to be substantially less active than the Mn- and Co-based ones.


A great number of modern catalysts used for industrial applications consist of mixed metal oxides.

Among them, transition metals containing perovskite oxides represent an important and very studied class of materials, owing to their great activity and thermal stability in the catalytic oxidation of hydrocarbons.1–3

Perovskite oxides with general formula ABO3 have an ideal cubic structure with the larger A cation at the center of the cube 12-fold coordinated with oxygen anions and the smaller B atoms occupying the corners of the cube in a 6-fold coordination.4

Orthorhombic (e.g. LaFeO3) and rhombohedral structures (e.g. LaMnO3 and LaCoO3) are also known.

The world wide interest towards these materials and their great variety of properties is due to the fact that about 90% of the metallic elements in the Periodic Table can give perovskite-type structures.

To have a stable perovskite, the only rule to be respected is an ionic rule introduced by Goldschmidt,5 the so called tolerance factor (t, with 0.75 < t < 1) defined by the equation t = (rA + rO)/√2(rB + rO), where rA, rB and rO are the ionic radii of the cations A and B and of the oxygen anions, respectively.

The preparation method based on citrate-type precursors6,7 allows one to obtain perovskites with a few tens (m2 g−1) of specific surface area after a thermal treatment at 1073 K. This method is currently employed in our laboratory and many perovskite systems have been prepared and characterized for their solid state properties and catalytic activity towards alkane and CO oxidation.8–16

In order to obtain more suitable materials for catalytic purposes, the dispersion of an active phase on a support allows the preparation of high surface area materials and increases the possibility of contacting the active sites and reactant species.

To this purpose a project started in our laboratory with the aim of preparing and investigating the structural and catalytic properties of materials obtained by impregnation of a suitable support with La, Mn, Co and Fe containing citrate precursors and submitting the system to the same thermal treatments used to prepare the bulk perovskite.

The first support to be used was tetragonal zirconia, ZrO2, which resulted as scarcely reactive with the active phase, and therefore suitable for dispersing perovskite materials.17–19

In the present study the investigation was extended to the alumina, δ-Al2O3, support (hereafter labelled δ) which represented a more reactive material than zirconia, particularly with metal ions like cobalt and lanthanum, therefore being of great interest from the structural point of view.

To gain information on the catalytic properties of the examined systems in an oxidation reaction, all the samples were tested for the oxidation of carbon monoxide.

The most relevant results are reported.


Bimetallic La–Mn, La–Co and La–Fe containing catalysts with different metal loadings [10 and 30 wt.% calculated according to the ratio perovskite/(alumina + perovskite) (where perovskite is LaMO3 with M = Mn, Co or Fe)] were prepared by impregnation of δ-Al2O3 with the appropriate amount of citrate-type precursors.

As a comparison, monometallic La, Mn, Co and Fe containing supported materials, with the same amount of metal as in the bimetallic samples, were prepared too.

Hereafter, labels like LaM/δ (with M = Mn, Co or Fe) or M/δ (with M = La, Mn, Co or Fe) followed by a number in brackets, (10) or (30), referring to the metal loadings, were used to indicate the samples.

The δ-Al2O3 support (surface area = 119 m2 g−1) was obtained from a Pural NW CONDEA batch calcined at 1223 K.

Metals dispersion on the δ-Al2O3 support was performed by the citrate method.6,7,20

As already reported,7 the use of citrate-type precursors allows the formation of the perovskite phases in mild conditions, i.e. at lower temperature with respect to the temperature reached in other kinds of preparations.

Two solutions were added to an appropriate amount of δ-Al2O3: an aqueous solution of citric acid was added first and then a solution containing the metal nitrates.

The molar ratio between the citric acid and the overall metal nitrates was fixed at 1.

Under these conditions the different metals formed a citrate complex, that easily decomposes promoting the formation of the perovskite oxides.

The resulting solution was kept at 383 K until dryness and the samples, after grinding, were calcined at 1073 K for 5 h.

For a phase check, unsupported LaMO3 (with M = Mn, Co or Fe) perovskite-type oxides were prepared by the same citrate method.6,7

Chemical analyses for Mn, Co and Fe were performed by atomic absorption with a Varian SpectrAA-220 instrument.

The supported samples were very difficult to dissolve; the first attempt to dissolve these materials in a HCl solution failed, so a HF solution was used.

Also in this case, a slight problem related to incomplete solubility was revealed.

After the analyses the sample labels were not modified and maintained the nominal metal content.

Phase analysis was performed by XRD using a Philips PW 1729 diffractometer with Ni-filtered Cu Kα radiation.

Particle sizes (D) were evaluated by means of the Scherrer equation, D = Kλ/βcos θ.

K is a constant equal to 0.9; λ the wavelength of the X-ray used; β the effective linewidth of the X-ray reflection under observation, calculated by the expression β2 = B2 − b2 [where B is the full width at half maximum (FWHM) and b the instrumental broadening determined through the FWHM of the X-ray reflection at 2θ ≈ 28° of SiO2 with particles larger than 1000 Å]; θ the diffraction angle of the considered X-ray reflection.

BET surface area of the samples (SA, m2 g−1) was evaluated by nitrogen adsorption at 77 K in a vacuum glass apparatus.

Krypton was used instead of nitrogen for measurements of surface areas below 10 m2 g−1.

Measurements of UV-Vis DRS were performed on Co containing samples in the 200–2500 nm wavelength range with a Varian CARY 5E spectrometer.

XAS measurements at the M metal K-edge (M = Mn, Co or Fe) were performed for the supported samples, and for LaMO3, α-Mn2O3, CoAl2O4, Co3O4 and α-Fe2O3 as reference materials.

Collection energy ranges for the EXAFS spectra were: 6350–7400 eV for Mn (E0 = 6539 eV), 7550–8600 eV for Co (E0 = 7709 eV) and 6900–8200 eV for Fe (E0 = 7112 eV).

These measurements were collected at the beamline GILDA, ESRF, Grenoble (France) in the transmission mode.

The beamline monochromator was equipped with two Si(311) crystals; the two ion chambers were filled with Ar gas.

Powder samples were deposited on Millipore membranes or mixed to an appropriate amount of boron nitride (BN) and pressed into pellets.

During the data collection samples were held at the liquid nitrogen temperature.

The XANES part of the experimental signal was obtained by subtracting a linear pre-edge and normalizing to one in correspondence to the first EXAFS oscillation.

The EXAFS analysis was performed using the complete FEFF8 package.21

Fourier transforms (FTs) for the signal k3χ(k) were calculated in the range 2.5 < k < 12.0 Å−1 with a Kaiser window.

Structural information was obtained by fitting the FT(R) function in the 0.8–4.0 Å range.

Fits were performed using the theoretical phase and amplitude functions generated by the FEFF8 code.

These functions were calibrated by the EXAFS spectra of the reference compounds.

For all the fits, the calculated statistical errors related to the bond distances were smaller than the uncertainty (0.02 Å) attributed to the EXAFS technique; for the coordination numbers the inaccuracy was evaluated around 0.03.

The CO oxidation with O2 was studied in a flow system using 0.5 g of catalyst supported on a silica fritted disk internal to a silica reactor vertically positioned in a tubular electrical heater.

A Ni–NiCr thermocouple was positioned in correspondence to the middle of the catalyst bed.

The thermocouple signal was used both to monitor the reaction temperature and to drive an ASCON XS series proportional temperature programmer that powered the electrical heater.

The programmer was set to produce a linear temperature ramp of 1 K min−1 from 300 to 800 K.

The composition and the total flow rate of the reactants were 1% CO, 20% O2, balance He by volume and 100 cm3 STP.

The space velocity resulted 12 000 cm3 STP h−1g−1.

At time intervals of 20 min, that is every 20 K, 6 cm3 of effluent from the reactor were sampled and analyzed by gas-chromatography, using an Alltech CTR 1 column and a thermal conductivity detector.

The mass balance with respect to carbon was 100 ± 2%.

Before each run a standard pretreatment with oxygen in flow (100 cm3 STP) at 773 K for 1 h was carried out.

Kinetic parameters were evaluated on the basis of a first order kinetics with respect to CO and 0th pseudo-order with respect to oxygen because it was present in large excess.

The following first order kinetic constants were evaluated: where F is the reactant flow rate (mol CO h−1), W is the mass of catalyst (g), SA is the surface area of the catalyst (m2 g−1), molM are the moles of M per g of catalyst, with M = Mn, Fe, Co. X is the fractional conversion estimated by the CO2 produced.

Results and discussion

The colour of Mn, Co and Fe containing powder samples was brown, blue and ochre-yellow, respectively.

The δ-Al2O3 support showed particle sizes D ≈ 100 Å.

Phase analysis, performed by XRD, revealed the presence of δ-Al2O3 in all the samples.

The XRD patterns of the least concentrated (10% metal loading) supported samples did not reveal other phases than δ-Al2O3.

On the contrary, as shown in Figs. 1(a) and 1(b) where the pattern of the δ-Al2O3 support is also reported, the samples with the highest metal loading (30%) revealed the occurrence of other oxide phases in addition to δ-Al2O3: (i) α-Mn2O3 (D ≈ 500 Å), α-Fe2O3 (very weak reflections) and CoAl2O4 (D ≈ 100 Å) are present in the Mn(30)/δ, Fe(30)/δ and Co(30)/δ samples, respectively; (ii) LaAlO3 (D ≈ 250 Å), LaMnO3 (D ≈ 100 Å), LaFeO3 (D ≈ 130 Å) perovskites, and LaAlO3 (D ≈ 330 Å) are present in the La(30)/δ, LaMn(30)/δ, LaFe(30)/δ and LaCo(30)/δ samples, respectively.

In addition, for the samples with 30% of metal loading, XRD (not shown here) were also performed at the end of the catalytic tests, revealing that no other phases than those detected before, were formed after the catalytic process.

For the δ-Al2O3 support the BET surface area (SA) value was 119 m2 g−1; all the supported samples showed SA values in the 75–121 m2 g−1 range.

In Table 1 the phases detected by XRD and the SA values are reported for all the samples.

The nominal and experimental contents of transition metal (Mn, Co and Fe) are also reported in Table 1 for the supported samples.

For Co containing samples DRS spectra were collected with the aim of revealing the presence of Co2+ and/or Co3+ ionic species.

In fact, in Mn and Fe containing samples only trivalent ionic species were revealed, while in the case of Co containing samples the presence of Co2+ in the CoAl2O4 spinel phase was expected and the UV-Vis DRS is a particularly suitable technique to diagnose the presence of such ionic species.

Co2+ in tetrahedral coordination exhibits three different transition bands from the ground state: the 4A2(F) → 4T2(F) one which lies in the IR region, the 4A2(F) → 4T1(F) and 4A2(F) → 4T1(P) ones which appear as multiple absorptions in the near infrared and visible region respectively.22

Fig. 2 shows DRS spectra for the supported samples and for CoAl2O4 and Co3O4 used as reference compounds.

Both reference materials show typical d–d transition bands which appear as a multiple absorption in the 1000–1800 nm region (with maxima at about 1230, 1370 and 1510 nm), due to the 4A2(F) → 4T1(F) transition of Co2+ in tetrahedral coordination.23,24

In addition, CoAl2O4 shows a sharp and intense multiple d–d absorption band in the 500–700 nm range (with maxima at about 550, 580 and 625 nm), due to the 4A2(F) → 4T1(P) transition of tetrahedral Co2+,23,24 while Co3O4 reveals a broader and less intense absorption region in the 200–1000 nm range (with maximum at about 770 nm) due both to the 4A2(F) → 4T1(P) transition of tetrahedral Co2+, and to d–d transitions of octahedral Co3+ ionic species.23

All Co containing samples match the absorption features of CoAl2O4 but the most concentrated ones (30 wt.%) also show an absorption band starting at about 700 nm which can be attributed to a very small amount of Co3O4 oxide phase that is present together with the CoAl2O4 phase.

As to the XAS analysis, the high local sensitivity of this technique allows the observation of structural characteristics of the supported phases and to make a comparison between samples with low and high metal loading.

To this purpose it was useful to discuss results regarding Mn, Co and Fe containing samples separately.

Tables 2, 3 and 4 report the coordination numbers (N) and bond distances (R) obtained by fitting procedures for the samples and reference materials containing Mn, Co and Fe, respectively.

The FTs of the reference compounds were fitted by fixing the main crystallographic bond distances data (α-Mn2O3,25 LaMnO3,26 CoAl2O4,27 LaFeO328 and α-Fe2O329) and leaving free the E0 shift, the S20 term and the Debye–Waller factors.

In the fitting procedure of the supported samples Debye–Waller, E0 and S20 parameters were fixed to the value obtained from the fit of the reference compounds.

The coordination numbers (N) and bond distances (R) were left to vary in the fit.

The Debye–Waller factors were fixed to avoid the strong correlation with the coordination numbers.

These numbers are smaller than those expected, probably owing to a very disordered distribution of oxide phases on the surface of the δ-alumina and/or to the presence of small crystallites.

However, the coordination numbers generally increase with increasing the metals content, indicating a greater degree of particles aggregation at higher loading.

XAS for Mn containing samples

For all the samples the FTs, shown in Fig. 3, reveal a first coordination shell peak in the 0.8–1.9 Å range and peaks related to the following shells in the 1.9–4.0 Å range.

The first shell peak is due to oxygen nearest atoms, the following peaks can be attributed to farther Mn, O or eventually La atoms.

Fits (Fig. 3) were performed in the whole 0.8–4.0 Å region, including a multiple scattering (MS) contribution Mn–O–Mn for the LaMn(30)/δ sample in order to improve the quality of the fit.

Results, reported in Table 2, reveal the presence of a very spread α-Mn2O3 phase in all the samples except the LaMn(30)/δ one which showed the formation of a LaMnO3 perovskite phase on the surface of δ-Al2O3.

Evidence of the two different supported phases could be found in the different values of Mn–Mn bond distances (two distances at about 3.11 Å and 3.58 Å for α-Mn2O3 and a single distance at about 3.89 Å for LaMnO3) and in the presence of a Mn–La coordination shell (at about 3.30 Å) which was revealed for LaMnO3 but was obviously absent for α-Mn2O3.

XAS for Co containing samples

Fig. 4 shows first derivatives of the XANES spectra for the Co containing materials and the CoAl2O4 spinel as a reference compound.

No appreciable difference can be found in the shape of these derivatives indicating that Co is almost exclusively in the Co2+ ionic form which is typical for the spinel phase.

Such an evidence was supported also by the EXAFS results.

For all the samples the FTs, shown in Fig. 5, reveal two main peaks ranging in the 0.8–2.0 Å and 2.2–3.7 Å regions, respectively.

The first shell peak (centred on 1.5 Å) is due to oxygen nearest atoms, the following peak can be attributed to farther Co, Al or O atoms.

Fits (Fig. 5) were performed in the whole 0.8–3.7 Å region and revealed the presence of a very dispersed CoAl2O4 phase in all the samples.

In fact the typical bond lengths of the CoAl2O427 spinel phase were detected (see Table 3): two different Co–O distances at about 1.95 and 3.39 Å, respectively, a Co–Al distance at about 3.36 Å and a Co–Co one at about 3.51 Å.

In particular, the Co–O first shell bond distance of the supported samples, ranging in the 1.94–1.97 Å interval, agrees with the sum of the ionic radii of O2− (1.38 Å, in four-fold coordination) and Co2+ (0.58 Å in tetrahedral coordination) ions,30 which can be typically found in the CoAl2O4 spinel phase.

XAS for Fe containing samples

For all the samples the FTs, shown in Fig. 6, reveal a first coordination shell peak in the 0.8–2.0 Å range and peaks related to the following shells in the 2.2–4.0 Å range.

The first shell peak is due to oxygen nearest atoms, the following peaks can be attributed to farther Fe, O or eventually La atoms.

Fits (Fig. 6) were performed in the whole 0.8–4.0 Å region and revealed (see Table 4) two Fe–O bond distances in the 1.91–1.95 Å and 2.0–2.16 Å ranges, respectively.

In the Fe(10)/δ and Fe(30)/δ samples, the evidence of Fe–Fe bond distances in the 2.79–2.82 Å, 2.98–3.00 Å and 3.32–3.51 Å ranges could be attributed to the presence of a very spread α-Fe2O3 phase (Fe–Fe = 2.89, 2.97, 3.37 Å) on the δ-alumina surface.

In the LaFe(10)/δ and LaFe(30)/δ samples the evidence of Fe–La bond distances in the 3.25–3.37 Å and 3.65–3.75 Å ranges, and a Fe–Fe distance in the 3.73–3.82 Å range revealed the presence of an amorphous and very dispersed LaFeO3 perovskite phase (Fe–La = 3.35, 3.76 Å; Fe–Fe = 3.93 Å).

As we already reported in a previous work concerning LaFeO3 perovskite supported on zirconia,19 also in this case the fits of LaFe(10)/δ and LaFe(30)/δ samples improved remarkably by adding a further Fe–Fe contribution which resulted in a short bond distance (at about 3.06 Å) with a small coordination number (N = 0.5–0.9).

Such a contribution could suggest the presence of a small amount of very spread α-Fe2O3 which exhibits short Fe–Fe bond distances at 2.97 and 3.37 Å, respectively, and a longer one at 3.70 Å.

The simultaneous presence of different phases, mainly perovskite and some ferric oxide (very disordered and dispersed on the δ-alumina surface), is established also considering that the 3.70 Å Fe–Fe distance, belonging to α-Fe2O3, could affect the values of the Fe–Fe distance found in the LaFe/δ samples.

This resulting bond length is shorter than 3.93 Å, that is the value of the Fe–Fe bond distance in LaFeO3.

Therefore the fitted distance, found in the LaFe/δ supported samples, resulted in being an intermediate value between the Fe–Fe bond distances in the two oxide phases.

CO oxidation

The samples without transition metal ions (t.m.i.), namely the support δ-Al2O3, La(10)/δ and La(30)/δ were active only at high temperature (Fig. 7).

The contribution of the activity of support to those of the t.m.i. containing supported catalysts was therefore considered negligible.

Concerning the other catalysts, their activity data will be reported for each t.m.i. in terms of conversions (same catalyst mass: 0.5 g) and first order kinetic constants with respect to CO. The pseudo-order with respect to oxygen was assumed 0th, because this reactant was present in large excess in the reactant mixture.

The kinetic constants ks (normalized per m2 of surface area of the sample) and km (normalized per mol of t.m.i. in the sample) were calculated and their logarithms used to build up Arrhenius plots (not reported for brevity).

Acceptable straight lines were obtained up to 80% conversions data, but the Ea values changed in an erratic way, possibly due to temperature-dependent deactivation phenomena.

Some activity loss was in fact observed in experiments at constant reaction temperature and high time on stream.

Therefore, reasoning based on the kinetic constants should be regarded with prudence.

It is stressed that, for the supported samples, ks represent the areal activity of the active phase only if the latter completely covers the support.

In the following, we will tentatively use the comparison of the log ks and log km values of the unsupported and supported phases to have qualitative information on: (i) the coverage (surface area of the active phase)/(surface area of the sample) ratio, by comparison of log ks values; (ii) the dispersion (accessible active phase)/(accessible + bulk active phase) ratio, by comparison of log km values.

A word of caution is necessary because this approach holds under the assumption that the areal activities of the unsupported and supported active phases are similar. log ks and log km were determined at two temperatures, namely 400 and 500 K (Table 5) for all the examined catalysts.

Mn containing catalysts

Concerning the Mn containing catalysts, the sample LaMn(30)/δ, which has been shown to contain the LaMnO3 perovskite, is active at RT already and the conversion remains the highest among all the Mn containing samples, also at higher temperature (Fig. 8).

It is noted that decreasing the Mn loading the conversion decreases too, LaMn(10)/δ being less active than LaMn(30)/δ and Mn(10)/δ less active than Mn(30)/δ (Fig. 8).

Comparing the log km values for LaMn(30)/δ (containing supported LaMnO3) and LaMnO3 we can observe substantially higher values for LaMn(30)/δ (Table 5), indicating dispersion of the active phase on the supported sample.

Accordingly, the ks values of LaMnO3 and LaMn(30)/δ are in a similar range, suggesting high coverage of the support by the active phase.

Passing to the samples Mn(10)/δ, Mn(30)/δ, LaMn(10)/δ, in which manganese is present as α-Mn2O3, the log km and log ks values suggest, by comparison with the corresponding values for α-Mn2O3, that both coverage and dispersion are low.

Concerning the activity of the unsupported pure oxides in terms of log ks, α-Mn2O3 is by far more active than the perovskite LaMnO3.

α-Mn2O3 is also the most active catalyst among all the examined Mn containing samples.

Fe containing catalysts

Regarding the Fe containing materials, none of the catalysts is active at RT and Fig. 9 indicates that the Fe system is, on average, less active than the Mn system.

In terms of conversion, the catalyst Fe(30)/δ, containing iron as α-Fe2O3, is the most active one, while the sample containing the same active phase, but lower Fe loading, Fe(10)/δ, shows lower activity.

The samples containing Fe as LaFeO3, are, on average less active.

In fact LaFe(30)/δ is less active than Fe(30)/δ and LaFe(10)/δ less active than Fe(10)/δ.

This picture is consistent with the notably higher activity of α-Fe2O3 in comparison with LaFeO3 (see Fig. 9).

The log km values of α-Fe2O3, Fe(10)/δ and Fe(30)/δ (Table 5) indicate a considerable dispersion of the active phase in the supported samples.

The log ks values for the same samples suggest that the fraction of support surface covered by the active phase is high.

The log km values of LaFeO3, LaFe(10)/δ and LaFe(30)/δ indicate a notable dispersion of the active phase in the supported samples.

The log ks values suggest that the area of the supported active phase and that of the sample are similar.

Co containing catalysts

Regarding the Co containing catalysts, most of them display a peculiar behavior because they are active at RT already, but the conversion decreases with time on stream, notwithstanding the temperature is increasing, denoting a catalyst poisoning.

At a given temperature a minimum is reached followed by an activity increase up to 100% conversion (Fig. 10).

For Co containing catalysts, the poisoning effect was already observed31–34 and it can be attributed to a strong adsorption of the CO2 product as such, or, most probably, as carbonate species that at a given temperature start to decompose, thereby leading to the usual Arrhenius behavior.

An explanation of the poisoning effect related to the partial surface reduction of the catalyst and the difficulty of reoxidation can not be ruled out.

In terms of conversion and evaluating the activity only from the ascending part of the curves, the supported samples are substantially less active than pure Co3O4 (Fig. 10).

Co(10)/δ, that is not active at RT, displays the usual Arrhenius behavior, also shown by the reference compounds LaCoO3 and CoAl2O4 but in a very different temperature range (Fig. 11).

As shown before (Fig. 2), in the reflectance spectra of the samples LaCo(30)/δ and Co(30)/δ an absorption band in the 700 < λ < 1000 nm range, typical for Co3O4, was observed whereas this absorption was absent in Co(10)/δ.

Therefore, it seems quite reasonable to attribute the activity at RT and the unusual behavior of the samples LaCo(30)/δ and Co(30)/δ to the presence of some little amounts of Co3O4 on their surface.

In fact Co3O4 shows both activity at RT and the unusual poisoning effect (Fig. 10).

With an analogous reasoning the absence of activity at RT and the normal Arrhenius behavior of Co(10)/δ (Fig. 11) are ascribed to the absence of Co3O4 on this catalyst.

The activity of Co(10)/δ would be due to the CoAl2O4 phase which is present in all the δ-alumina supported and Co containing catalysts.

The behavior of LaCo(10)/δ, which shows RT activity and deactivates (Fig. 10), is not easily rationalized, because neither reflectance spectroscopy nor XAS demonstrated the presence of Co3O4 on this sample.

One could hypothesize that LaCo(10)/δ contains very small amounts of Co3O4 that escaped the experimental detection.

The log km and log ks values (Table 5) were determined from Arrhenius plots for which only the ascending part of the curves were considered.

It is recalled that none of the supported catalysts contains the perovskite phase LaCoO3, whereas the main constituent is the CoAl2O4 spinel.

Co(30)/δ and LaCo(30)/δ were found to contain also minor amounts of Co3O4.

Therefore the comparison between supported and reference materials can not include the LaCoO3 perovskite.

The log km and log ks values are difficult to interpret without considering the presence of minor amounts of Co3O4 together with the most abundant CoAl2O4 spinel phase.

In fact the latter phase has very small log ks values and the presence of Co3O4 is expected to boost the activity, as it is found, due to its high log ks.

Comparison of all the samples towards CO oxidation

Considering all the supported samples, the most active, in terms of either conversion or log km or log ks, is LaMn(30)/δ, which is already active at RT.

The XRD, XAS and DRS characterization showed that LaCoO3 never formed in Co containing supported samples.

In the absence of this datum the lower activity of LaCo(30)/δ in comparison with LaMn(30)/δ would have been hardly explainable, because LaCoO3 has an activity, in terms of log ks, higher than LaMnO3.

Finally we compare the activity of the unsupported pure compounds in terms of log ks.

Concerning the simple oxides, α-Mn2O3 is the most active one, followed by Co3O4 and then by α-Fe2O3.

α-Mn2O3 is also the most active among all the examined catalysts.

However, when supported on alumina, α-Mn2O3 is poorly dispersed under our preparation conditions.

Passing to the binary oxides, the resulting order of activity is: LaCoO3 > LaMnO3 > LaFeO3 > CoAl2O4.

It is also noted that, for either Mn- or Fe- or Co-based systems, the activity (log ks) of the simple oxide is higher than that of the corresponding binary oxide.

It can be observed that the order of the activities for simple and perovskite oxides found in the present study of CO oxidation does match that reported for the oxidation of hydrocarbons by Seiyama.2

This is possibly due to differences in the reaction mechanisms and/or reaction temperature range (considerably higher especially for methane oxidation).


XRD analysis for the δ-alumina supported La, Mn, Co and Fe containing samples showed only the presence of δ-alumina for materials with 10 wt.% of metal loading.

Materials with 30 wt.% of metal loading revealed the presence of other oxide phases depending on the composition.

UV-Vis DRS analysis, performed on Co containing samples, all blue-coloured, suggested the presence of the CoAl2O4 spinel phase in all the samples.

In addition, the most concentrated materials showed an absorption band typical for Co3O4 which was supposed to be present on the alumina surface, in a small amount, together with CoAl2O4.

XAS investigation gave really interesting results, particularly in the most diluted samples for which XRD gave no information.

In fact, the presence of different oxide phases on the alumina support was found also for materials containing 10 wt.% of metals.

In particular, at low loading of La and Mn on δ-Al2O3, and in spite of the simultaneous presence of both metals, only the most stable α-Mn2O3 phase was formed.

At higher loading (30 wt.%), the presence of Mn and La induced preferably the formation of the LaMnO3 perovskite.

δ-Al2O3 supported cobalt reacted with the support leading to the formation of the CoAl2O4 spinel phase in all the Co containing samples, also at the lowest loading, being CoAl2O4 the most stable phase in these conditions.

δ-Al2O3 supported iron led to the formation of α-Fe2O3; the simultaneous presence of Fe and La on the surface of the support, both at the highest and at the lowest metal loading, induced preferably the formation of the LaFeO3 perovskite, probably the most stable with respect to the LaMnO3 and LaCoO3 perovskite phases.

Materials with low metal loading showed the greatest degree of particles dispersion on the alumina surface.

As for the CO catalytic oxidation results, the LaMn(30)/δ sample displayed the highest activity (in terms of both conversion and log km or log ks) among all the investigated supported catalysts, demonstrating the advantage of dispersing the LaMnO3 perovskite on the alumina support.

In the case of the Fe-based system, the corresponding LaFe(30)/δ sample, containing the LaFeO3 perovskite, had lower activity in comparison with Fe(30)/δ which contained α-Fe2O3.

Concerning the Co-based system, the LaCoO3 perovskite never formed on the alumina surface because of the reaction between cobalt and alumina leading to the formation of the CoAl2O4 spinel phase that is poorly active.

The activity of the LaCo(30)/δ and Co(30)/δ samples also reflected the presence of minor amounts of highly active Co3O4.

Consequently, these samples were active also at RT, but deactivated rapidly.