Chromium Oxide Supported on Different Al2O3 Supports: Catalytic

Dec 18, 1998 - The tendency of Cr3+ to form aggregates such as Cr2O3 was induced by the support. The features of the support exert an influence upon c...
1 downloads 0 Views 110KB Size
396

Ind. Eng. Chem. Res. 1999, 38, 396-404

MATERIALS AND INTERFACES Chromium Oxide Supported on Different Al2O3 Supports: Catalytic Propane Dehydrogenation Liliana R. Mentasty, Osvaldo F. Gorriz, and Luis E. Cadus* INTEQUI, Instituto de Investigaciones en Tecnologı´a Quı´mica (UNSL-CONICET), Casilla de Correo 290, 5700 San Luis, Argentina

Chromium oxide was supported on different commercial aluminas. Their characteristics were assessed by different techniques (X-ray diffraction, X-ray photoelectron, laser Raman, and electron paramagnetic resonance spectroscopies, and Brunauer-Emmett-Teller surface, temperature-programmed reduction, and temperature-programmed desorption measurements). The variation of the Cr3+/Cr6+ ratio was accounted for by the interaction between the chromium phase and the support. The tendency of Cr3+ to form aggregates such as Cr2O3 was induced by the support. The features of the support exert an influence upon chromium stabilization both in its oxidation states 6+ and 3+ and in the coordination of surface chromia species. The total acid sites and the oxidized chromium have a marked effect on activity and selectivity at initial operation times. Introduction Supported chromium ions exhibit important catalytic properties and have been studied mainly on alumina, silica, and zircornia.1-8 It has been accepted that the catalytic properties of these systems are determined by the species formed by the chromium oxide. Consequently, research interest has been focused on the molecular structures and on the chromium oxide surface species. The reviewed studies have mostly used nonaggressive chromium phase precursors with low support interaction. According to Vuurman et al.,9 the molecular structure of the chromium(VI) oxide surface species is determined by surface coverage and calcination temperature, because under ambient conditions the surface structure depends on the point of zero charge at the surface net pH of the hydrated oxide. Increasing surface coverage causes a decrease of the net surface pH as well as formation of more polymerized chromium oxide species. When chromium nitrate is used as the precursor up to a surface coverage of 9% CrO3/Al2O3, the chromium oxide is stabilized by the alumina support in the 6+ oxidation state after calcination.9 In previous studies10,11 the behavior of alumina supports was observed using different precursors. The surface net potential was decreased by means of chromium carboxylates, and attention was focused on the relation between surface species and the type of molecule used. The obtained results indicated that weak metal support interaction might be the major cause of surface species polymerization. The interaction between the chromia phase and the support surface, which stabilizes different oxidation states and coordinations of the chromia species, defines the surface architecture. Part of the Cr3+ is incorporated * To whom correspondence should be addressed. Fax: 54 652 26711. E-mail: [email protected].

in the vacant octahedral sites of the spinel surface. The larger extension of this phenomenon with chromic acid as the precursor might be responsible for the formation of surface-inactive species with the resultant loss of catalytic activity. Besides, deactivation between cycles might depend on the catalyst stabilization by the incorporation of Cr3+ in the support structure. When carboxylates were used as precursors, it was noted that the surface array of monomeric species in a polymeric species environment stabilized the catalyst architecture, also conferring catalytic stability between cycles. The use of the same precursor in all catalysts might permit one to compare the behavior of the different characteristics of supports during the formation of chromia species, independently from the contribution of the precursor to the surface net potential. The precursor characteristics might favor a strong metal support interaction that defines anchoring points for the subsequent formation of catalytically active centers. That is to say, chromium would initially deposit on numerous centers on which greater loadings subsequently accumulate.10 The anchoring sites have been described in the literature.12 In the case of silica gel, the oxygen atoms of the chromic anhydride evidently coordinate with the protons of the hydroxyl groups, forming a compound of the conjugated acid type. The reaction of CrO3 with the surface hydroxyl groups of aluminum oxide results in a compound of the type aluminum chromate. The amount of chemically bound CrO3 increased with increasing surface area, and the degree of reduction declined accordingly. However, since aluminum chromate has a strong Lewis acid, namely, aluminum oxide, as its substrate, it has a higher oxidizing power than pure aluminum chromate and initiates an electron shift in the CrO3 monolayers not bound chemically to the surface, thus increasing their reducibility.

10.1021/ie9802562 CCC: $18.00 © 1999 American Chemical Society Published on Web 12/18/1998

Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999 397 Table 1. Chromium Weight Content catalyst

% Cr2O3

% Cr

8Cr/A 8Cr/B 8Cr/C

12.71 13.34 13.05

8.70 9.13 8.93

Table 2. Characteristics of the Supports Used in This Work support

crystalline phase

acidity of Al2O3 (au)

acidity/ m2

SBET (m2/g)

jrpore (Å)

A B C

γ γ R-θ

144.5 65.2 57.3

0.71 0.68 0.88

202 96 65

32 56 76

Table 3. Surface Properties of Catalysts catalyst

SBET (m2/g)

% coke

Cr2p3/2/Al2p XPS

8Cr/A 8Cr/B 8Cr/C

156 72 61

11.3 5.0 3.7

0.25 0.27 0.38

The purpose of the present work is to study the influence of surface acidity and the texture of different aluminas on the characteristics of supported chromium oxide and on catalytic performance. Experimental Section Catalyst Preparation. The Cr2O3/Al2O3 catalysts were prepared with equal Cr contents from three highpurity commercial aluminas, A and B are Rhone Poulenc and C is Ketjen, defined by their specific surface in m2/g: 202 (A), 96 (B), and 65 (C). The aluminas were impregnated with a chromic acid normalized solution by the method of continuous adsorption. This method consists of loading the support in the form of pellets in a continuous reactor, where the chromic acid solution is recirculated by means of a peristaltic pump at constant flow and ambient temperature for 30 min. Immediately after impregnation, the catalysts were dried overnight to 323 K in a vacuum oven and then calcined to 923 K for 7 h in an air atmosphere. The catalysts are referred to as 8Cr/A, 8Cr/B, and 8Cr/ C. The final percent of Cr concentrations were determined by X-ray fluorescence (Table 1). The catalytic activity was measured according to the method described in a previous paper.13 Catalyst Characterization. The supports and catalysts were characterized by different techniques. Results are shown in Tables 2 and 3, respectively. Specific surface areas (SBET, m2/g) by the BrunauerEmmett-Teller (BET) method and pore volume distribution of all samples were determined from nitrogen adsorption-desorption isotherms at 77 K. A Micrometrics Accusorb 2100E was used. The metal content of the catalysts was determined by X-ray fluorescence with a Philips PW 1400 X-ray spectrometer. X-ray diffraction (XRD) patterns were obtained by using a Rigaku diffractometer operated at 30 kV and 20 mA employing Ni-filtered Cu KR radiation (λ ) 0.154 18 nm). Temperature-programmed reduction (TPR) studies were performed in a TPR conventional unit. This apparatus consists of a gas handling system with mass flow controllers (Matheson), a tubular reactor, a linear temperature programmer (Omega, model CN 2010), a PC for data retrieval, a furnace, and various cold traps.

Samples of ca. 100 mg were first oxidated in a 30 mL/ min flow of 20 vol % O2 in He at 877 K for 30 min and then cooled at room temperature. After that, helium was admitted at room temperature to remove the oxygen. The samples were subsequently contacted with a 30 mL/ min flow of 5 vol % H2 in N2 and heated, at a rate of 10 K/min, to a final temperature of 1000 K, while the hydrogen consumption was monitored by a thermal conductivity detector after removal of the water formed. The amount of hydrogen consumed was evaluated by integration of the TPR peak areas up to 100 min of previous calibration. Acidity measurements were determined by ammonia temperature-programmed desorption (TPD) by using a conventional flow system with a thermal conductivity detector. The adsorption step was carried out in pure ammonia flow for 30 min at room temperature (299 K). Then, the samples were swept with helium for 30 min, and finally the desorption step was performed from room temperature to 823 K at a heating rate of 10 K/min. Continuous voltages from the detector cell and reactor thermocouple were converted to digital signals, amplified with a data acquisition workstation, and stored in a PC. Acidity was calculated by dividing the total area by the BET surface area of the sample. Spectroscopic Measurements. (a) Laser Raman Spectroscopy. Raman spectra were recorded from powdered samples, with a Jasco TRS600SZP multichannel monochromatic spectrometer. The samples were pressed into self-supporting wafers. The Raman spectra were recorded at room temperature and with the 514.5 nm line of an argon ion laser as the excitation source. The laser power at the catalyst wafers was 200 mW unless stated otherwise. The samples were placed under a microscope, making it possible to point the laser beam at any desired particle. (b) X-ray Photoelectron Spectroscopy (XPS). XPS spectra were obtained with a Shimadzu spectrometer employing a Mg KR X-ray excitation source (hν ) 1253.6 eV). The sample, as fine powder, was pressed. The reported binding energies have been referenced to C1s as 284.6 eV. Referencing was performed both before and after a complete set of spectra had been obtained. The spectra were analyzed in the sequence Cr2p, C1s, O1s, Al2p, and C1s. Ratios of atomic concentration in the outer layers of the samples were expressed as the corresponding XPS area ratios by using the effective ionization cross section of ejected electrons tabulated by Scofield and the formulas given by Seah and Dench.14 (c) Electron Paramagnetic Resonance (EPR). EPR measurements were made at room temperature on a Bruker spectrometer operated at X-band frequencies. A Klystron frequency of 9.7 GHz and 100 kHz magnetic field modulation were used. The spectrometer was equipped with an on-line computer for data treatment. Catalytic Test. The activity was measured in a flow apparatus at atmospheric pressure. The equipment has previously been described in detail.13 The catalyst (usually 1.0 g, 0.5-0.85 mm particle diameter), was placed on a fritted disk sealed inside a steel reactor. The height of the catalyst bed was around 10 mm. The reactor temperature was controlled within (1 K by a commercial device. Isothermal conditions were approached by feeding the reactant diluted in an inert gas. The analysis of reactants and products was carried out with a gas chromatograph (Hewlett-Packard GC 5780) with an activated alumina packed column and employ-

398 Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999

Figure 1. X-ray patterns of fresh catalysts.

ing nitrogen as the carrier gas. A flame ionization detector (FID) was used. The system allowed the separation of hydrogen, methane, ethane, ethylene, propane, and propene, in this order. The areas of the gas chromatographic peaks were evaluated by an integrator (Hewlett-Packard, model 3390A). The operation conditions were the following: contact time τ ) 2.5 g s cm-3; propane molar fraction yp ) 0.2 using N2 as the diluent, at 833 K constant temperature. Mass differences between fresh and coked catalyst were determined in order to estimate the amount of produced coke during the reaction time. The catalyst regeneration after each dehydrogenation run was performed in situ in air flow at 723 K for 2 h. Reference to “after catalytic test” means “after five complete cycles”. Results BET Specific Surface and Pore Volume Distribution. The selected supports showed different specific surfaces and pore volume distribution (Table 2). Support A exhibited the largest specific surface and the lowest rpore, close to the micropore limit, indicating a high percentage of microporous structure. On the other hand, supports B and C showed low values for surface area and high rpore, which are typical of mesoporous structures. No changes in SBET or pore volume distribution by effects of the thermal treatment on the supports were observed. The results for the specific surface of impregnated catalysts are shown in Table 3. In all cases the surface decreased with respect to that of the corresponding support. In 8Cr/A and 8Cr/B the surface decreased by 25%, and it decreased by 6% for 8Cr/C. X-ray Fluorescence. The analyses of the final chromium percent content in catalysts 8Cr/A, 8Cr/B, and 8Cr/C are presented in Table 1. The chromium concentration was approximately the same in all samples (about 9% in chromium weight). X-ray Diffraction. The X-ray diffractograms of the supports presented diffraction lines assigned to the following: supports A and B, γ-Al2O3 (JCPDS 29-36); support C, R-Al2O3 and corundum θ-Al2O3 (JCPDS 431484 and 35-121, respectively), as shown in Table 2. The XRD results of the catalysts are shown in Figure 1. The crystalline phases of the supports could be observed for all catalysts. The patterns corresponding to Cr2O3 crystallites were not observed in catalyst 8Cr/

Figure 2. Acidity distribution from NH3 TPD measurements.

A, but they were observed in 8Cr/B and 8Cr/C. The obtained results for 8Cr/A might be explained either by the dispersion or by the smaller size of the Cr2O3 crystals on the catalyst surface. Determination of Acidity by NH3 TemperatureProgrammed Desorption (TPD). The acidities of supports A, B, and C were determined by NH3 TPD. The results are shown in Figure 2. All of the supports showed only one desorption peak. The total acidity was determined by integrating the area below the peak (Table 2). The distributions of the acidity potentials were similar for the three selected supports. The total acidity was highest for the support with the highest specific surface. The results, expressed in terms of acidity/m2BET, indicate similar values for supports A and B, with support C being the one with the highest acidity per surface unit. Temperature-Programmed Reduction (TPR). Figure 3 shows the spectra of all of the catalysts before and after being subjected to reaction conditions. Apparently, there was a single reducing stage which included the bulk Cr. The variations in the reduction Tmax were small, in both fresh and regenerated catalysts. In all cases, the reduction Tmax values were higher for regenerated catalysts, keeping the sequence shown by fresh catalysts. These shifts might indicate a different Crsupport interaction. The maximum temperature of reduction for each catalyst is shown in Table 4. Laser Raman Spectroscopy. Raman spectroscopy is a powerful technique that permits one to obtain structural information about solids, especially oxides of heavy metals on aluminas or silicas. Because the vibrations of these oxides have frequencies below 1100 cm-1, they are normally observed in the infrared spectra by strong absorption of the support. Alumina and silica are poor in Raman scatterers, which makes Raman spectroscopy widely used for detecting these vibrations. Figure 4 shows the Raman spectra of catalysts 8Cr/ A, 8Cr/B, and 8Cr/C, in the region between 1264 and 514 cm-1. Low fluorescence was observed in all of the samples because of a diminution of sites available for impurities caused by high metal loadings on the catalyst. Catalyst 8Cr/A exhibited the most intense bands at 1000, 960, 896, and 850 cm-1. The 550 cm-1 band was weak, and the 600 cm-1 one was not observed in this catalyst. The ∼1000 cm-1 band was assigned to polymeric surface species on alumina. The ∼896 cm-1 band was assigned to hydrated dichromate species (Cr2O72-),

Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999 399

Figure 3. TPR profiles of fresh and regenerated catalysts. Table 4. Tmax Reduction from TPR fresh catalyst

Tmax (°C)

regenerated catalyst

Tmax (°C)

8Cr/A 8Cr/B 8Cr/C

373.5 352.5 348.6

8Cr/A 8Cr/B 8Cr/C

376.4 364.4 356.0

Figure 4. Raman spectra of fresh catalysts.

and the ∼960 and ∼850 cm-1 bands were indicative of the presence of hydrated trichromate species (Cr3O102-) on the alumina surface. The ∼960, ∼896, and ∼850 cm-1 bands were consequently assigned to νs(CrO2), νs(CrO3), and νs(OCrO), respectively.9 The Raman spectra of Cr2O3/Al2O3 catalysts with high Cr loadings presented a very sharp band at 550 cm-1. This band, weak in 8Cr/A but very intense in 8Cr/B and 8Cr/C, was assigned9 to metal-oxygen vibration of distorted octahedrally coordinated chromium atoms in crystalline Cr2O3. The Raman spectra showed that, although at low loadings the support stabilizes the chromium in its oxidation state 6+, at high loadings a portion of Cr is in the form of Cr3+, forming Cr2O3 crystals. Catalysts 8Cr/B and 8Cr/C also presented the characteristic bands at ∼1000 and ∼600 cm-1, which were

assigned to νs(CrdO) and νs(Cr-O-Cr) of surface polymeric species on the alumina. XPS. Analyses by XPS were performed on fresh catalysts (before catalytic test), on catalysts after the catalytic test, and on regenerated catalysts (removal of coke by air current). Table 3 summarizes the results obtained for the Cr2p3/2/Al2p surface atomic ratios of fresh catalysts. The values of the Al2p binding energy were approximately the same in all of the samples and did not change with analysis time, indicating that the sample was not electrically charged. The Cr/Al surface atomic ratios were similar (0.25 and 0.27) for 8Cr/A and 8Cr/B. In 8Cr/C, the number of Cr atoms per Al atom was considerably higher than that obtained for the other two catalysts (0.38). After the catalytic test, the Cr/Al atomic ratio increased for both 8Cr/B and 8Cr/C (0.35 and 0.41, respectively). EPR. Figure 5 shows the results of EPR analysis for fresh, coked, and regenerated 8Cr/A, 8Cr/B, and 8Cr/ C. Under the analysis conditions used (central field at 3200 G and scanned field of 3000 G), it was only possible to analyze central phase β, which corresponds to Cr3+ in chromia aggregates and is characterized by g ) 1.993, and the phase γ signal attributed to Cr5+ with a characteristic g ) 1.967. The chromia-phase β signal was similar in all catalysts, both fresh and coked. Slight differences in shape and width though not in amplitude were found between fresh and coked 8Cr/C. The width and amplitude of the chromia-phase β signal were similar in all regenerated catalysts. In regenerated 8Cr/A and 8Cr/B the amplitude and width of this resonance were lower than those of their fresh and coked counterparts. In regenerated 8Cr/C, this signal did not show width changes but its amplitude was larger than that of coked catalyst. All fresh and regenerated catalysts showed a weak resonance band known as phase γ and attributed to Cr5+. The amplitude of this resonance depended on the type of catalyst and its treatment. It was similar for fresh 8Cr/A and 8Cr/B, with a lower value for 8Cr/C.

400 Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999

Figure 5. EPR spectra for fresh, used, and regenerated catalysts.

Coked catalysts did not present the chromium 5+ signal. Catalysts 8Cr/A showed a narrow ∆Hpp ) 2 G signal which cannot be attributed to Cr5+. In regenerated catalysts the amplitude of the Cr5+ signal in chromia-phase γ 8Cr/A doubled that of 8Cr/B and approximately tripled that of 8Cr/C. The chromia-phase β signal ratio of catalysts 8Cr/A, 8Cr/B, and 8Cr/C was 1:0.8:0.47. The decrease of the chromia-phase β signal intensity in 8Cr/B and 8Cr/C might be explained by the formation of Cr2O3 observed by XRD and Raman though not detected by EPR because of experimental conditions. Evaluation of the Catalytic Activity through the Propane Dehydrogenation Reaction. Catalysts 8Cr/ A, 8Cr/B, and 8Cr/C were subjected to propane dehydrogenation tests. In the catalytic tests, the starting catalyst is in its most oxidized state. The last pretreatment before the catalytic test is calcination in an air atmosphere. Between cycles, an air current at 723 K is flown through the catalytic bed. The obtained results can be attributed to two phenomena: one associated with catalytic activity, which distinguishes each catalyst, and another one associated with selectivity, which distinguishes 8Cr/B and 8Cr/C from 8Cr/A. The data for conversion and selectivity against time on stream (5-160 min) of the catalysts for the fifth dehydrogenation run are shown in Figure 6. Even though the conversion values at 40 min time on stream were similar (Table 5), the starting values were very different, being highest for 8Cr/A. At longer times, the conversion values were similar for 8Cr/A and 8Cr/B and slightly higher than those of 8Cr/C. Selectiv-

ity to propylene was markedly different for the different catalysts at short times. Catalysts 8Cr/B and 8Cr/C were highly selective, with similar values, while the values for 8Cr/A were lower. The conversion values per surface unit X %/m2BET are presented in Figure 7. Catalyst 8Cr/C showed the highest conversion values in the entire operation time range. These values were slightly higher than those of 8Cr/B, and those of both 8Cr/C and 8Cr/B were higher than those of 8Cr/A. The selectivity to propylene of catalysts 8Cr/B and 8Cr/C was relatively high (approximately 75%) and remained similar for the entire time on stream range. In comparison, 8Cr/A exhibited lower selectivity at short times but a tendency to asymptotically approach 8Cr/B and 8Cr/C at longer times. Discussion Architecture of the Catalytic Surface. The use of the same precursor in all catalysts permits one to compare the ways in which the support characteristics affect the formation of chromia species, independently from the contribution of the precursor to the surface net potential. The precursor characteristics enable a strong metal-support interaction that defines anchoring points for the subsequent formation of catalytically active dehydrogenation centers. That is to say, chromium initially deposits on numerous centers on which higher chromium loadings subsequently accumulate.1 Although support A is microporous and supports B and C are mesoporous, both 8Cr/A and 8Cr/B exhibit a decrease of the BET specific surface (SBET) of ap-

Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999 401

Figure 6. Conversion and selectivity of the catalysts. Table 5. Catalytic Results for Propane Dehydrogenation 40 min 8Cr/A 8Cr/B 8Cr/C

100 min

X%

S%

X%

S%

42.7 44.4 40.3

58.8 71.3 76.7

27.9 26.2 27.6

68.1 75.1 76.4

Figure 7. Conversion per specific surface area of the catalysts.

proximately 25% and catalyst 8Cr/C loses 6% of its SBET. It might be stated that this surface decrease is due to pore clogging by the chromia phase. However, taking into account that A and B show very different pore sizes (32 and 56 Å, respectively) and that the decrease of SBET is similar in both cases, this does not seem to be a feasible explanation. Rather, and considering that the aluminas remain unaltered in their original crystalline phases, it can be speculated that the decrease in the specific surface is due to sintering. However, this possibility should be discarded because the SBET and pore volume distribution of the calcined supports at the catalyst preparation temperature do not show significant changes. Because the effects of pore clogging and support sintering are discarded, the decrease of the catalyst SBET

values might be attributed to a sintering effect fostered by chromium. It is important to note that supports A and B, despite presenting different textural characteristics and equal predominant crystalline phase γ, exhibit similar acidity per surface unit. On the other hand, although B and C have similar textural characteristics (SBET ) 96 and 65 m2/g, respectively, with mesoporosity in both cases), they show different crystalline phases and also different values of acidity per surface unit. It might be suggested that the different crystalline phases γ and θ determine at least in part the acidic features of the support. It is possible to assume that a change in the surface architecture during the catalytic test favors the metalsupport interaction. Our results might indicate that the solid system tends to become stable thanks to a redispersion of the chromia phase. This phenomenon might take place in the ordering 8Cr/B:8Cr/C:8Cr/A, with 8Cr/A being initially the most stable. The metalsupport interaction, using the same chromium support, could be caused by surface acidity, thus determining both dispersion and the type of species of the chromia phase. In fact, the chromium loading is enough to form a monolayer on the alumina surface. Yet, the XPS results show that the Cr2p/Al2p surface atomic ratios are well below 1. The decrease of dispersion in the sequence 8Cr/A:8Cr/B:8Cr/C, which is the same sequence obtained for Tmax (TPR), might be caused by the strength of the precursor anchoring to the support. Supports A and B have similar acidities per specific surface unit. Similarly, it could be assumed that they possess a similar number of chromium anchoring centers per surface unit. However, since SBET for B is lower than that for A, the dispersion per surface unit must be similar for A and B. This is corroborated by XPS through Cr2p/Al2p global atomic ratios. Support C shows a higher acidity per surface unit than A and B. In the same line of reasoning as that used before, it is possible to assume that the anchoring centers of C are more numerous than those of A and B. XPS shows that the Cr2p/Al2p ratio increases in C with respect to A and B, indicating a change in the dispersion per surface unit.

402 Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999

On the catalyst surface, the Cr precursor-support interaction is initially high and markedly above the Cr precursor-adsorbed Cr precursor interaction. After a certain time has elapsed, Cr precursor-support and Cr precursor-adsorbed Cr precursor interactions become equal, and finally the latter becomes stronger. The results obtained by TPR show only one peak for all of the catalysts, indicating that there is a single reduction stage. The maximum temperatures vary for the different catalysts and for the same fresh and used catalysts. The reducibility follows the ordering 8Cr/A . 8Cr/B = 8Cr/C for both fresh and used catalysts. In all used with subsequent regeneration catalysts, there is an increase of the maximum reduction temperature as compared with fresh catalysts. The maximum reduction temperature could be an acceptable indicator of the difficulty in reducing supported chromia species and therefore of the Cr-alumina interaction. The XRD results confirm that the chromia-support phase interaction decreases in the above-described ordering. The low intensity of the XRD patterns for 8Cr/A might indicate that Cr, due to its array, is stabilized by the support, which keeps it in its maximum oxidation number and thus prevents it from becoming Cr3+. The Tmax of reduction before and after the catalytic test indicates that catalyst 8Cr/A is the most stable in regards to surface characteristics. The small Tmax variation might indicate that the metal-support interaction is not considerably changed by the catalytic test and subsequent regeneration. Although A and B exhibit similar acidities per surface unit, in B the greater chromium accumulation due to smaller specific surface per anchoring point might facilitate redispersion of the greater amount of Cr available, thus increasing the Tmax of TPR. The XPS results after the catalytic test corroborate the TPR data, showing that dispersion is highest in that catalyst with highest Tmax variation. It has been reported in the literature9 that, at high values of chromium loadings (close to those used in this work), a very intense Raman band at 550 cm-1 is observed. This band is originated by a vibration of distorted octahedrically coordinated chromium atoms in a Cr2O3 crystalline structure, and it becomes more intense with higher chromium loadings. It might seem strange for the Raman band to be weak in 8Cr/A as compared to 8Cr/B and 8Cr/C. However, this behavior can be explained by the Raman spectroscopy data and by our discussion in the above paragraph. XRD analyses reveal the presence of Cr2O3 crystallites. The discrepancy between Raman and XRD data for Cr2O3 detection is due to different technique sensitivities. XRD only detects crystals bigger than 40 Å while Raman spectroscopy has excellent sensitivity for metal oxide crystallites. Both 8Cr/B and 8Cr/C show surface polymeric species which include Cr2O3 (550 cm-1). In 8Cr/A the bands for polymeric species (1000 and 600 cm-1) are not detected, and that corresponding to Cr3+ polymers is weak. Catalyst 8Cr/A shows intense bands assigned to tri- and dichromates, with predominance of the former, while these bands are weak in 8Cr/B. In 8Cr/C, the trichromate band is weak and that corresponding to dichromates is not detected. The relative intensity of the 550 cm-1 band is higher in 8Cr/C than in 8Cr/B. Because no monolayer is formed although the chromium is enough to form multilayers, it might be

assumed that chromium accumulates on the anchoring centers. Taking into account the chromium dispersion per surface unit and the total surface of the catalysts, the relative height of the catalytic centers can be inferred to be 0.50, 1.00, and 0.84 for 8Cr/A, 8Cr/B, and 8Cr/C, respectively. All of the catalysts have similar chromium loadings. The specific surface of support A, together with a strong chromia-support interaction, hinders the formation of polymeric species. This interaction is favored by the size of the chromium agglomerates in the catalytic centers. On the other hand, the formation of polymers is favored by the accumulation of chromium on the same catalytic centers. Thus, weaker chromia-support interaction probably favors redispersion of the chromium phase. At low loadings, Cr is stabilized by the support in the form of Cr6+, while at high loadings, a portion of chromium is in the form of Cr3+ (in part as Cr2O3 crystals). When chromium carboxylates are used as precursors, Cr3+ species are observed even at low Cr loadings.11 EPR measurements at ambient temperature permit one to observe only that part of Cr3+ which is not in the Cr2O3 crystalline structures detected by Raman. From the analysis of Cr3+ of phase β (g ) 1.993) it results that in coked 8Cr/C the Cr3+ species are in an environment of greater symmetry than that in the fresh one. Regenerated 8Cr/A and 8Cr/B show weaker Cr-Cr interactions than fresh ones, while this interaction increased in regenerated 8Cr/C. As compared to fresh and coked catalysts, the concentration of Cr3+ decreased in regenerated 8Cr/A and 8Cr/B and increased in 8Cr/ C. During regeneration, the support favors the oxidation of part of Cr3+ to Cr6+, which probably acts as a diluent decreasing the Cr-Cr interaction. It is worth noticing that the intensity of the Cr3+ signal shows no increase in coked catalysts. This might indicate the formation of Cr2O3, which contains Cr3+, not observable by EPR at ambient temperature. It could be suggested that, in each Cr3+ cycle, polymeric species turn into other kinds of species observable by EPR at ambient temperature. Such transformations are apparently reversible. The Raman spectra show that in 8Cr/B and 8Cr/C the concentration of Cr2O3 is higher than that in 8Cr/A, which could be an indicator of the dynamic limitations of catalyst 8Cr/A in the above-proposed equilibrium. It cannot be discarded that in all cases the grafted Cr6+ species are those which form first. It is likely that these species distribute on the surface of the alumina, anchoring through Al-O-Cr bridges and constituting monomeric species. It was observed by UV-vis-nearIR diffuse reflectance spectroscopy that the maximum quantity of these species is 1.5 wt % CrO3 and the same for all of the studied samples.15 It can be postulated that there is a second type of Cr6+ species that are not chemically bound to the support but which interact under the effect of temperature (reaction temperature, for instance). In an oxygen environment and at 673 K, it is surprising that Cr3+ does not reach the oxidation state 6+. A feasible explanation is that more severe conditions are necessary to oxidize the Cr3+ present in the polymeric structures of Cr2O3, but not all of the Cr3+ is part of them. It is proposed16 that part of the Cr3+ is incorporated in the vacant Al3+ sites, which is facilitated by the similar ionic radii and charges of Cr3+ and Al3+. These species are occluded in the support and not easily

Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999 403

oxidized. This could explain why Cr3+ can be detected by EPR after regeneration. By default it can be inferred that the higher concentration of Cr6+ follows the ordering 8Cr/A:8Cr/B:8Cr/C. Additionally, Cr5+ can be observed on zircornia in a square-pyramidal configuration together with Cr6+ mononuclear species.7 Obviously, the concentration of Cr5+ is related to that of Cr6+. All fresh and regenerated catalysts show phase γ of Cr5+ while coked catalysts do not. The concentration sequence, which for fresh catalysts is 8Cr/A = 8Cr/B > 8Cr/C, becomes 8Cr/A . 8Cr/B = 8Cr/C in regenerated catalysts. This indicates that there is a rearrangement of the catalyst surface after the catalytic cycle. The above observations show that catalyst 8Cr/A, because of its high specific surface and elevated number of anchoring centers, behaves like a catalyst at low chromium loadings. It could also be concluded that the tendency of Cr3+ to form aggregates such as Cr2O3 is induced by the support. It is clear that the described features of the support exert an influence upon chromium stabilization both in its oxidation state 6+ or 3+ and in the coordination of surface chromia species. Catalytic Performance. The catalytic performance indicates that it is possible to distinguish catalysts by their behavior in propane conversion and selectivity to propylene. The above-described surface characteristics and the published findings on the catalytic behavior of each involved phase permit one to suggest an explanation for the results obtained here. Two phenomena might occur: one associated with the propane conversion which distinguishes each catalyst and the other one associated with selectivity which distinguishes 8Cr/B and 8Cr/C from 8Cr/A. The supports exhibit decreasing total acidity in the order A:B:C, even though the acidity per surface unit is higher in C and lower, with similar values, in A and B. Lugo et al.17 report that, at similar chromium loadings, the dehydrogenation catalytic activities of Cr/ Al2O3 and Cr/SiO2 are similar. These authors suggest that probably in both cases the active sites are the same and, therefore, there would be no significant electronic effects of the support on the chrome. However, we have observed that the consequences of the support acidity do affect the acid-base characteristics of the catalyst and could consequently act on its catalytic performance. It has been reported in the literature15 that in the dehydrogenation of isobutane there exists an increment of activity when Cr3+ oxide species are present, even at concentrations not detectable by XRD. The authors suggest that these Cr3+ species are more active than those formed in the reaction environment by the reduction of Cr6+. Besides, the Cr3+ species in Cr2O3 might be less active than those in amorphous Cr3+ oxide. The appearance of Cr2O3 polymeric species takes place only after the alumina surface starting anchoring centers have been covered. A number of publications suggest that, independently from the starting oxidation state or the technique used for chromium addition, after a stable catalytic performance is reached, activity is only a function of the total chromium content. In the opinion of many authors,6,18,19 the dehydrogenation active species are Cr3+ ions, while others suggest7.20 they are Cr2+ and Cr3+ or coordinativeley unsaturated Cr2+. There are indications that the catalyst is rapidly reduced to Cr3+ and/or Cr2+ and that Cr6+ does not remain for a long time on stream. To summarize, it could be stated that

(i) there exists a direct relation between catalytic activity and chromium dispersion and (ii) the activity of the reduced chromium species changes with the nature of the chromium species. By comparison of 8Cr/B and 8Cr/C, it can be observed that there is a lineal dependence between catalytic activity per unit of specific surface and dispersion per unit of surface measured after the XPS test. The good correlation obtained for 8Cr/B and 8Cr/C is acceptable because both catalysts possess similar surface descriptions of chromium species. This dependence is not observed in 8Cr/A because the expected X %/m2 value is higher than the experimental one. The deviation of 8Cr/A can be accounted for by the original description of the surface architecture. It could be suggested that 8Cr/A does not form a considerable amount of Cr3+ species, either as amorphous oxide or as Cr2O3. It has already been mentioned that its behavior could be compared to that observed for catalysts at low loadings. Therefore in this catalyst, Cr3+ species would form at the expense of Cr6+. Even though there are some Cr3+ species in 8Cr/A, part of them would be inactive in the Al-O-Cr bridges. Both total conversion and selectivity are affected by total acid sites, which promote the formation of coke, and by the oxidized chromium, which promote cracking and combustion reactions. The effect of these two factors is strong at initial operation times, leading to high conversion and low selectivity values. This effect gradually decreases while the reaction is in progress because of the reduction of Cr6+ and the blocking of acid sites by coke, which also undoubtedly affects the catalyst active phase. The selectivity to propylene of 8Cr/A tends to asymptotically approximate to that of 8Cr/B and 8Cr/C, as a function of time on stream. Conclusions In each Cr3+ cycle, polymeric species turn into other kinds of species observable by EPR at ambient temperature. Such transformations are apparently reversible. Cr3+ is incorporated in the vacant Al3+ sites, which is facilitated by the similar ionic radii and charges of Cr3+ and Al3+. These species are occluded in the support and are not easily oxidized. This could explain why Cr3+ can be detected by EPR after regeneration. The 8Cr/A catalyst, because of its high specific surface and elevated number of anchoring centers, behaves like a catalyst at low chromium loadings. It could also be concluded that the tendency of Cr3+ to form aggregates such as Cr2O3 is induced by the support. The described features of the support exert an influence upon chromium stabilization both in its oxidation state 6+ or 3+ and in the coordination of surface chromia species. Cr3+ species would form at the expense of Cr6+. Even though there are some Cr3+species in 8Cr/A, part of them would be inactive in the Al-O-Cr bridges. The total acid sites and the oxidized chromium have a marked effect on activity and selectivity at initial operation times. The selectivity to propylene of 8Cr/A tends to asymptotically approximate to that of 8Cr/B and 8Cr/C, as a function of time on stream. Acknowledgment The authors are grateful to Japan International Cooperation Agency for the grant of the Laser Raman Spectrometer to CENACA. Financial support is also

404 Ind. Eng. Chem. Res., Vol. 38, No. 2, 1999

acknowledged to CONICET and to Universidad Nacional de San Luis. Literature Cited (1) McDaniel, M. P. The State of Cr(VI) on the Phillips Polymerization Catalyst. II. The Reaction between Silica and CrO2Cl2. J. Catal. 1982, 76, 17. (2) McDaniel, M. P. The State of Cr(VI) on the Phillips Polymerization Catalyst. III. The Reaction between CrO3/Silica and HCl. J. Catal. 1982, 76, 29. (3) McDaniel, M. P. The State of Cr(VI) on the Phillips Polymerization Catalyst. IV. Saturation Coverage. J. Catal. 1982, 76, 37. (4) Ghiotti, G.; Garrone, E.; Della Gatta, G.; Fubini, B.; Giamello, E. The Chemistry of Silica-Supported Chromium Ions: Calorimetric and Spectroscopic Study of Nitric Oxide Adsorption. J. Catal. 1983, 80, 249. (5) Myers, D. L.; Lunsford, J. H. Silica-Supported Chromium Catalysts for Ethylene Polymerization: The Active Oxidation States of Chromium. J. Catal. 1986, 99, 140. (6) Grunert, W.; Saffert, W.; Feldhaus, R.; Anders, K. Reduction and Aromatization Activity of Chromia-Alumina Catalysts. J. Catal. 1986, 99, 149. (7) Cimino, A.; Cordischi, D.; De Rossi, S.; Ferraris, G.; Gazzoli, D.; Indovina, V.; Minelli, M.; Occhiuzzi, M.; Valigi, M. Studies on Chromia/Zirconia Catalysts. Y. Preparation and Characterization of the System. J. Catal. 1991, 127, 744. (8) Vuurman, M. A.; Wachs, I. E.; Stufkens, D. J.; Oskam, A. Characterization of Chromium Supported on Al2O3, ZrO2, TiO2, and SiO2 Under Dehydrated Conditions. J. Mol. Catal. 1993, 80, 209. (9) Vuurman, M. A.; Hardcastle, F. D.; Wachs, I. E. Characterization of CrO3/Al2O3 Catalysts Under Ambient Conditions: Influence of Coverage and Calcination Temperature. J. Mol. Catal. 1993, 84, 193. (10) Carra´, S.; Forni, L.; Vintani, C. Kinetics and Mechanism in Catalytic Dehydrogenation of n-Butane over Chromia-Alumina. J. Catal. 1967, 9, 154.

(11) Gorriz, O. F.; Cadus, L. E. Supported Chromium Catalysts using Metal Carboxylate Complexes: Dehydrogenation of Propane. Appl. Catal. Accepted for publication. (12) Miesserov, K. G. Nature of Active Sites of Supported Chromic Oxide Polymerization Catalysts. J. Catal. 1970, 22, 340. (13) Gorriz, O. F.; Arrua, L. A.; Cadus, L. E.; Rivarola, J. B. Catalytic Dehydrogenation of Propane to Propylene. Lat. Am. Appl. Res. 1989, 19, 31. (14) Seah, M. P.; Dench, W. A. Quantitative Electron Spectroscopy of Surfaces: A Standard Data Base for Electron Inelastic Mean Free Paths in Solids. Surf. Int. Anal. 1979, 1, 2. (15) Cavani, F.; Koutyrev, M.; Trifiro, F.; Bartolini, A.; Ghisletti, D.; Iezzi, R.; Santucci, A.; Del Piero, G. Chemical and Physical Characterization of Alumina-Supported Chromia-Based Catalysts and Their Activity in Dehydrogenation of Isobutene. J. Catal. 1996, 158, 236. (16) Weckhuysen, B. M.; Verberckmoes, A. A.; Buttiens, A. L.; Schoonheydt, R. A. Diffuse Reflectance Spectroscopy Study of the Thermal Genesis and Molecular Structure of Chromium-Supported Catalysts. J. Phys. Chem. 1994, 98, 579. (17) Lugo, H. J.; Lunsford, J. H. The Dehydrogenation of Ethane over Chromium Catalysts. J. Catal. 1985, 91, 155. (18) Grunert, W.; Shpiro, E. S.; Feldhaus, R.; Anders, K.; Antoshin, G. V.; Minachev, Kh. M. Reduction and Aromatization Activity of Chromia-Alumina Catalysts. J. Catal. 1986, 100, 138. (19) Hakuli, A.; Kytokivi, A.; Krause, Y.; Suntola, T. Initial Activity of Reduced Chromia/Alumina Catalysts in n-Butane Dehydrogenation Monitored by On-Line FT-IR Gas Analysis. J. Catal. 1996, 161, 393. (20) Gorriz, O. F.; Corte´s Corbera´n, V.; Fierro, J. L. Propane Dehydrogenation and Coke Formation on Chromia-Alumina Catalysts: Effect of Reductive Pretreatments. Ind. Eng. Chem. Res. 1992, 31, 2670.

Received for review April 27, 1998 Revised manuscript received September 24, 1998 Accepted November 5, 1998 IE9802562