3114
J . Phys. Chem. 1992, 96, 31 14-3123
Differential Heat of Reoxidation of Reduced V20,/y-A1,03 Paul J. Andersen and Harold H. Kung* Ipatieff Laboratory and Department of Chemical Engineering, Northwestern University, Evanston, Illinois 60208 (Received: September 13, 1991)
Four V/y-A120, catalysts with coverages ranging from 1.4 to 8.2 V atoms/nm2 were examined with heat flow microcalorimetry, Raman spectroscopy, and chemical titration. Raman spectra showed that crystalline V2Os formed on the A1203surface between coverages of 6.1 and 8.2 V/nm2. The heats and rates of reoxidation of the reduced catalysts were measured in a static adsorption calorimetry system. Both the reoxidation rates and heats were strong functions of the degree of reduction. They decreased as the sample was reoxidized, and the variation depended on the reoxidation conditions. At a degree of reduction equivalent to 0.5 0 removed/V ion, the differential heats at 300 OC were about 170 kJ/mol of 0 (or 85 kJ/mol of V) for the two higher loading samples and about 200 kJ/mol of 0 (100 kJ/mol of V) for the two lower loading samples. However, at low extents of reduction, the situation was reversed and the multilayer samples had higher reoxidation heats such that the integral reoxidation heats were relatively constant independent of V loading. The rates of reduction by H2 and reoxidation by O2increased as the vanadia coverage increased. The variations in the reoxidation heats and rates with vanadia coverage were related to the structural characteristics of the samples.
Introduction It has been shown for a number of selective oxidation reactions, such as the oxidation of propene to acrolein over bismuth molybdatel and the oxidation of butane to maleic anhydride over vanadium phosphorous oxide,2 that lattice oxygen directly participates in the reaction. I8O2 isotope labeling experiments have shown that lattice oxygen is the proximate source of oxygen in the formation of C-O bonds. This has prompted the hypothesis that the metal-oxygen bond strength should be an important factor that determines the selectivity and possibly the activity of the catalyst. Various techniques have been used to characterize the oxygen binding strength or the metal-oxygen bond strength. They include determining the changes in the metal-oxygen vibrational frequency to indicate a change in the bond strength,) the determination of the desorption peak temperature in temperature-programmed desorption of oxygen to measure the oxygen desorption activation energy and/or the binding energy of oxygen: as well as by measuring the amount of oxygen vaporized from the solid as a function of temperature and then calculating the enthalpy of vaporization from the Clausius-Clapyron e q u a t i ~ n . ~ . ~ Adsorption calorimetry could directly provide the heat of oxygen removal from an oxide lattice, which is the thermodynamic quantity needed to estimate the energetics of the surface elementary steps in selective oxidation reactions. Auroux and Gravelle7found that the heats of adsorption of oxygen onto a series of Ag/Si02 ethylene epoxidation catalysts were directly related to the catalytic oxidation activities. Khalif et a1.8 used calorimetry to measure the heat of reoxidation of reduced V0,/Si02 catalysts as a function of the extent of catalyst reduction. They found that the reoxidation heat was initially constant at about 170 kJ/mol of 0 and then linearly decreased to 150 kJ/mol of 0 as the catalyst neared "complete" reoxidation. Unfortunately, the degree of reduction was not referenced to a well-defined catalyst state in
that study. Thus few conclusions could be drawn concerning the oxygen binding energy as a function of the degree of reduction. It is likely that, in their experiments, the reoxidation temperature (200 "C)was too low to oxidize all of the reduced vanadium ions to the +5 oxidation state. Subsequent studies on Bi-Mo-0: V-Mo-O,Io and Fe-Sb-0" catalysts provided more details concerning the degree of catalyst reduction with respect to a well-defined reference state. In this study, the heat of reoxidation of partially reduced vanadia supported on 7-A1203was determined. V/7-Al2O3 has been used in the partial oxidation of toluene,I2 and it was found that the catalytic properties depended strongly on the vanadium loading. As the vanadium loading was increased, both the toluene consumption rate (per vanadium) and the selectivity to benzaldehyde increased si@icantly until the equivalent of a monolayer coverage was reached, at which point these two quantities attained fairly constant values. This same behavior was observed for butadiene and 1-butenel3 oxidation over V/y-A1203 and for toluene14 and o - ~ y l e n e l ~oxidation *'~ over V/Ti02. For o-xylene oxidation, the monolayer catalyst was found to be more active and selective than pure V205. Vanadia is known to be well dispersed on the surface of 7-A1203. The structure of this layered phase changes with the vanadia loading. Two main types of surface species have been suggested.17s'8 One species is an isolated vanadyl group consisting of one terminal V=O bond and three oxygen atoms bridging the vanadium to the aluminum in the support. The other surface species is a two-dimensional disordered phase characterized by extensive networking of the vanadia species. This networked phase contains both V-0 terminal bonds and V-0-V groups. As the loading is increased, the amount of networked species increases with respect to the amount of isolated species, but both are present even at monolayer coverages. Both species can be identified with ~~
~
~
~~~
~~
(9) Bondareva, V. M.; Andrushkevich, T. V.; Pankratiev, Yu. D. React. Kinet. Catal. Left. 1986, 32 ( I ) , 171. (1) Peacock, J. M.; Parker, A. J.; Ashmore, P.G.; Hockey, J. A. J. Cufal. 1969, 1 5 , 398. (2) Pepera, M. A,; Callahan, J. L.; Desmond, M. J.; Milberger, E. C.; Blum, P. R.; Bremer, N. J. J. Am. Chem. SOC.1985, 107, 4883.
(3) Tarama, K.; Teranishi, S.; Yoshida, S.; Tamura, N. Proc. 3rd Int. Cong. Catal. Amsterdam 1964, 282. (4) Iwamoto, M.; Yoda, Y.; Tamazoe, N.; Seiyama, T. J. Phys. Chem. 1978, 82, 2564. ( 5 ) Sachtler. W.M. H.; Dorgelo, G. J. H.; Fahrenfort, J.; Voorhoeve, R. J. H. Proc. 4th Int. Cong. Catal. 1968, 355. (6) Sazonov, B. A.; Popovski, V. V.; Bcireskov, G. K. Kinet. Carol. 1967, 9, 255. (7) Auroux, A.; Gravelle, P. C. Thermochim. Acta 1981, 47, 333. (8) Khalif, V. A,; Aptekar, E. L.; Krylov, 0.V.; Ohlmann, G. Kinet.Katul. 1977, 18 ( 4 ) , 1055.
0022-3654/92/2096-3114$03.00/0
( I O ) Bondareva, V. M.; Andrushkevich, T. V.; Pankratiev, Yu. D.; Turkov, V . M. React. Kinet. Catal. Left. 1986, 32 (2), 387. ( 1 1) Veniaminov, S. A,; Malyshev., E. M.; Pankratiev, Yu. D.; Turkov, V.
M. React. Kiner. Catal. Lett. 1987, 33 (2), 363. (12) Jonson, B.; Rebenstorf, B.; Larsson, R.; Andersson, S. L. T.; Lundin, S. T. J. Chem. Soc., Faraday Trans. I 1986,82, 767. ( 1 3 ) Mori, K.; Miyamoto, A.; Murakami, Y. J. Chem. Soc., Faraday Trans. 1 1986, 82, 13. (14) Van Hengstum, A. J.; Van Ommen, J. G.; Bosch, H.; Gellings, P. J. Appl. Catal. 1983, 8, 369. ( 1 5) Wachs, I. E.; Saleh, R. Y.; Chan, S. S.;Chersich, C. C. Appl. Catal.
1985, 1 5 , 339.
(16) Saleh, R. Y.; Wachs, I. E. Appl. Cafal. 1987, 31, 87. ( 1 7 ) Went, G. T.; Oyama, S. T.; Bell, A. T. J. Phys. Chem. 1990,94,4240. (18) Eckert, H.; Wachs, I. E. J . Phys. Chem. 1989, 93, 6796.
0 1992 American Chemical Society
Reoxidation of Reduced Vz05/y-A1203 Raman spectroscopy. Other spectroscopic techniques such as 51V NMR,I8 ESR,19 and infrared spectroscopyz0support these conclusions. In this paper, the differential heats of reoxidation of a series of reduced V/y-Al,O, catalysts of different loadings are reported as a function of both the vanadium loading and the degree of catalyst reduction. These catalysts were also characterized with chemical titration to determine the changes in vanadium oxidation state upon reduction and with Raman spectroscopy for structural information.
Experimental Section Catatyst Synthesis. y-AlzO,was synthesized from the hydrolysis of aluminum isopropoxide (98+%, Aldrich). A doubly distilled HzO/(CH3)zCHOH solution was added dropwise to a stirred [(CH3)zCHO]3Al/(CH3)zCHOH solution held at 0 OC. The resulting precipitate was filtered and washed with HzO. The solid was dried a t 40, 80, and 120 "C for 24 h each. The dried solid was ground and then calcined at 550 OC for 24 h. An X-ray diffraction pattern was collected to confirm that the y phase resulted from this synthesis. The surface area was determined to be 181 m2/g with N2BET. The impurities in the y-Alz03were 21 ppm Na, 34 ppm Ca, 70 ppm Fe, 20 ppm Ti, and 0 ppm Zr as measured with ICP. V/y-AlZO3samples were prepared as follows: 3 g of the alumina was slurried in 50 mL of methanol (Anhydrous, Mallinckrodt) under flowing N2,and the appropriate amount of vanadium triisopropoxide (95-99%, Alfa) was injected into the reaction flask. The slurry was stirred at room temperature under N2 until the methanol evaporated. The resulting solid was dried in air at 120 O C . Finally, the solid was calcined at 520 OC in air. The final vanadium loading was determined with atomic absorption spectroscopy. The measurements reported here are for supported vanadia samples with loadings of 3.7, 8.2, 16.3, and 23.4 wt % Vz05,which are equivalent to vanadium coverages of 1.4, 2.9, 6.1, and 8.2 V atoms/nm2, respectively. Assuming that the average V-V distance is 3.4 & I 8 as in NH,V03, a vanadia monolayer coverage corresponds to 4.4 V/nm2. The monolayer coverage could also be defined as the point at which the vanadia coverage equals the coverage of vanadium ions on the (010) plane of VZOS, which is 9.7 V/nm2. calorimetric and Volumetric Measurements. Experiments were conducted in a volumetric system (Figure 1) connected to a Tian-Calvet heat flow calorimeter (SETARAM DSC- 111). The temperature of the sample, which was positioned in the calorimeter, was controlled with a PID temperature controller to within k0.02 "C. The volumetric system consisted of a gas handling manifold, a dosing volume, and a reactor. The background pressure of the gas manifold was 3 X lo4 Pa as measured with a Penning gauge. The system pressure was measured with MKS capacitance manometers. Small doses with pressures ranging from 0 to 1.33 H a could be introduced from volume Vz(14.07 mL) or volume (V, + V z )(27.77 mL). Larger doses with pressures ranging from 0 to 133 kPa could be introduced from volume (VI + Vz). The voltage signal from the pressure transducer was recorded digitally with a personal computer. Part of the dosing volume could be cooled to trap condensable reaction products such as gaseous HzO generated from the reduction of an oxide with H2. Cooling this portion increased the effective volume of the doser by 2.80 mL. The effective volume of the loaded reactor (V,) ranged from 7.25 mL at 300 OC to 6.83 mL at 500 O C . The calorimeter sensitivity was determined by measuring the calorimeter response to the reduction of CuO to Cu with HZ: CUO + Hz CU + HzO(g) (1) AH = 88.313 kJ/mol at 300 O C 2 1 = 91.826 kJ/mol at 500 OC +
(19) Busca, G.; Centi, G.; Marchetti, L.; Trifiro, F. Langmuir 1W, 2, 568. (20) Busca, G. Langmuir 1986, 2, 577.
The Journal of Physical Chemistry, Vol. 96, No. 7, 1992 3115
.rrh 4fTi L H,
r-
To Vacuum
Penning Gauge
O " CO
PI He
Control and Data Aquisition
-
\*\
v,
'Reactor
Calorimeter Figure 1. Adsorption calorimetry system. P I = 0-133 kPa capacitance manometer. P2 = 0-133 Pa or 0-1330 Pa capacitance manometer.
About 0.015 g of CuO (99.999+%, Aldrich) was pretreated in the calorimeter cell by heating it in vacuo at 500 OC. Afterward, the temperature was set to the desired value and the sample was reduced with small doses of Hz (ultrahigh purity, Linde). The heat generated from each dose was measured with the calorimeter, and the H2 consumptions were measured volumetrically. The sensitivity factor was calculated from eq 2. Experiments were C(area under thermogram for each dose) sensitivity = (molar reaction heat)(mol of H2 consumed) (2) carried out in which the H2 dose size was varied from 3 to 15 pmol/dose to determine the effect of the amount of heat generated on the experimental accuracy. In a typical catalyst reoxidation experiment, the sample (about 0.1 g of V/y-AlZ0,) was loaded into the 7-mm-0.d. fused silica calorimeter cell. About an equal volume of ground fused silica powder was loaded into the reference cell. The sample was heated to 500 O C under vacuum for 2 h to remove adsorbed water. O2 (101 kPa, extra dry, Linde) was admitted into the cell. After 30 min, the temperature was lowered to 400 O C , and then the system was evacuated for 30 min. A dose of H2 large enough to remove about 0.5 0 atom/V atom was admitted to the system. If the H2consumption had not stopped after 2 h, the sample temperature was raised to 500 O C at 2 OC/min and was held there until the H2 consumption stopped. After reduction, the system was evacuated a t 500 OC for 10 h to remove any adsorbed water generated in the reduction pretreatment. The temperature was lowered to 300 O C and the sample was reoxidized with small doses (11.5 pmol) of 02. The heat generated from each dose was measured with the calorimeter and the oxygen consumptions were measured volumetrically. When the reoxidation rate became immeasurably small, the temperature was raised to 500 OC, and the reoxidation was continued until the equilibrium oxygen pressure was >5 kPa. An empirical characterization of the reoxidation rate was estimated by measuring the time required for the oxygen pressure to fall to l / e of its initial value, t(l/e). The H2 reduction rate could also be estimated with this technique by trapping the H 2 0 produced from the oxide reduction and measuring the rate of pressure drop in the system. Since several factors make undetermined contributions to the observed reaction rate (expansion of dosing gas, reactant adsorption, lattice diffusion, etc.), a theoretical analysis was not attempted. Thus the choice of t ( 1/e)
Andersen and Kung
3116 The Journal of Physical Chemistry, Vol. 96, No. 7, 1992 TABLE I: Calorimeter Calibration Data: CuO Reduction with H2' redn no. of H2 mol of H2/mol of CuO sensitivity, temp, OC doses (h0.03) pV/mW 500 500 500 300
12 13 13 13
1.oo 1.oo 1.oo 0.96
8.44 8.55 8.46 8.64
H2dose size = 15 pmol, thermogram area = 1350 mJ did not imply a first-order reaction. Instead it was used as an empirical measure of relative reaction rates. Previous workers have characterized the reaction rates in similar experiments with other empirical parameters such as the time required for the calorimeter signal to reach certain values (maximum, half-maximum, or return to baseline) for a given dose22or the time required for 90% completion of the gas a d s o r p t i ~ n . * ~ - ~ ~ Chemical Analysis. Chemical analyses were performed following the method in ref 26. A catalyst sample was first reduced as described in the last section, and the extent of reduction was measured from the H2 consumption. Following the reduction, the sample was evacuated for 30 min at 500 OC. The reduced sample was transferred to a beaker and dissolved in 50 mL of 50% H2S04. This solution was diluted to 250 mL with H20. A 100-mL aliquot was titrated with 0.01 N KMn04 to determine the concentration of V3+ + V4+. This same solution was then titrated with 0.01 N Fe(NH4)2(S04)2to determine the concentration of V3+ V4+ + V5+. The endpoint was detected with diphenylamine-4-sulfonic acid prepared as in ref 27. A separate 100-mL aliquot was titrated as before with the Fe(NH4)2(S04)2 solution to determine the concentration of V5+. The fractions of V3+,V4+, and V5+could be calculated from the results of these three titrations. Raman Spectroscopy. Raman spectra were collected with the 514.5-nm line of a Coherent INOVA 70-2 Ar ion laser. The beam was passed through a laser monochrometer (Applied Photophysics) and directed onto the sample. The scattered radiation was collected with an f/1.4 camera lens (Nikon). The collected light was collimated and passed through an interference filter (Raman notch filter from Omega) to remove the elastically scattered light and then focused onto the entrance slit of a 320" focal length spectrograph equipped with a 2400 g/mm grating (HR320 from Instruments, S.A.). The detector was a liquid nitrogen cooled CCD (Thomson CSF TH7883-PM CCD in a system from Photometrics Ltd.) which had 576 X 384 pixels. Each pixel was 23 X 23 pm. This arrangement resulted in a dispersion of 0.55 cm-'/pixel. The Raman cell consisted of a 25-mm-diameter fused silica tube which had a fused silica window sealed to the tube at one end and a heated section in the middle. The other end of the tube was connected to a valve which could be connected to a gas handling system so that sample could be subjected to various pretreatments. The catalysts were pressed into self-supporting wafers of about 0.25 g and mounted in a sample holder attached to one end of a 6.4" stainless steel tube. The samples were heated in vacuo at 500 "C for 1 h to remove adsorbed HzO.The spectra were collected with the samples under dry O2 to avoid effects due to surface hydration which are known to alter the positions of certain Raman bands.28
+
Results Calorimeter Calibration. The molar differential heats of reduction as a function of the degree of reduction for three different (21) Handbook of Chemistry and Physics, 60th ed.; CRC Press: Cleveland, 1979. (22) Cardona-Martinez, J.; Dumesic, J. A. J. Catal. 1990, 125, 427. (23) Gatte, R. R.;Phillips, J. J . Catal. 1989, 116, 49. (24) Phillips, J.; Gatte, R. R. Thermochim. Acta 1989, 154, 13. (25) Gatte, R. R.; Phillips, J. Thermochim. Acta 1988, 133, 149. (26) Niwa, M.;Murakami, Y. J . Coral. 1982, 76.9. (27) Kolthoff, I. M.; Sandell, E. B. Textbook of Quantitative Inorganic Analysis, 3rd ed.;The MacMillan Co.: New York, 1952; p 474. (28) Chan, S . S.; Wachs, I. E.; Murell, L. L.; Wang, L.; Hall, W. K. J . Phys. Chem. 1984, 88, 5831.
Heat (kJ/mole H 2 )
Sample 1
200 -
150
-
50
1
O 0.2
Sample 2
-
Sample 3
'
L
0
+
I
0.4
0.8
0.8
1
0 Removed per Cu Figure 2. Differential heat of reduction of CuO with a function of the degree of reduction.
H2at 500 O C as
TABLE II: Effect of Dose Size on Calorimeter Accuracy: CuO Reduction with H2at 300 OC (S= 8.77 aV/mW)
H2 dose size, uno1 15 1.6 1.o 0.45
thermogram area per dose, mJ
f(l/e), s
measd molar heat, kJ/mol
1350 150 1 IO 45
31 17 13
90 h 2 94 =k 6 98 h 5 103 h 6
CuO samples at 500 "C are shown in Figure 2. These heats were calculated using the instrument sensitivity factor supplied by the manufacturer. The sensitivities calculated using eq 2 from these experiments and a similar experiment conducted at 300 O C are presented in Table I. The ratios of H2 consumed to CuO weighed into the calorimeter cell were unity, indicating that the reduction from CuO to Cu was complete. The peak areas from all the thermograms collected for each sample were summed to give the integral area from which the sensitivity could be calculated. The average measured calorimeter sensitivities were 8.48 pV/mW at 500 "C and 8.64 pV/mW at 300 "C. These numbers agreed very well with the Joule effect calibration supplied by SETARAM (8.47 pV/mW at 500 OC and 8.77 pV/mW at 300 "C), confirming the accuracy of the entire adsorption calorimetry experimental procedure. The sensitivity factors supplied with the instrument were used for all subsequent experiments. The effect of the quantity of heat generated per dose on the accuracy of the measurement is shown in Table 11. The rates of hydrogen consumption, as measured by t(l/e), were comparable for these four experiments. Also, the reduction rate did not change significantly with the degree of reduction. The data show that as the area of the thermogram decreased, the measured molar heat increased significantly. The molar heats determined from thermograms with areas below 150 mJ were higher than the literature value primarily due to the inaccurate baseline correction in the thermogram. Since the reoxidation rates of reduced V/ y-Al,O, were at least as fast as the rates of reduction of CuO, the accuracies in determining the areas of the thermograms in both experiments should be comparable. Therefore, the oxygen dose sizes in the reoxidation experiments were adjusted so that the resulting thermograms would have areas of at least 200 mJ. Raman Spectra. The Raman spectra of the catalysts are shown in Figure 3. The sharp feature at 1030 cm-I found in all spectra has been previously assigned to isolated, four-coordinated vanadyl groups, each of which has one terminal oxygen and three oxygens that bridge the vanadium to the support surface. This assignment
The Journal of Physical Chemistry, Vol. 96, No. 7, 1992 3117
Reoxidation of Reduced V20S/y-A1203 5!
Normalized Pressure
-
12--
1-
~-
I
'!
0.6
06
04
1100
800
500
~v (cm Figure 3. Raman spectra of V/y-AI2O3 collected in dry O2 (a) 1.4 V/nm2, (b) 2.9 V/nm2, (c) 6.1 V/nm2, and (d) 8.2 V/nm2.
was based on 51VNMR spectra,ls which indicated a vanadium coordination of 4, and from comparison with the Raman spectra of the vanadyl halides (VOX3, X = C1, F, etc.) which have C3, symmetry." The broad feature extending from approximately 700 to 1000 cm-l was thought to result from networked vanadia species." It appeared to be composed of two separate bands with peaks at about 780 and 900 ad. The relative intensities of these two bands changed as the loading was increased. For the 2.9 V/nm2 sample, the 900-cm-I band was more intense than the 780-cm-l band. Their intensities were roughly the same for the 6.1 V/nm2 sample, and the 780-cm-' band was more intense for the 8.2 V/nm2 sample. A comparison with spectra of aqueous vanadate solut i o n ~suggested . ~ ~ ~that ~ both ~ ~ bands ~ ~ could be assigned to a networked vanadia species. The 900-cm-' peak was due to vibration of V I 0 groups and the 780-cm-l peak to vibration of V-0-V groups in these species. Finally, in the 8.2 V/nm2 sample, a peak appeared at 998 cm-' which indicated the presence of V20, crystallites. Thus, the capacity of the support to effectively disperse the vanadia was exceeded between vanadia coverages of 6.1 and 8.2 V/nm2. If the monolayer coverage is considered to be 4.4 V/nm2, the vanadia formed multiple layers on the A1203surface before V205crystallites formed. On the other hand, if the monolayer coverage is considered to be 9.7 V/nm2, V205crystallites formed below the monolayer coverage. An XPS study of vanadia catalysts supported on Ti02,31which also effectively disperses vanadia, suggested that multilayers were formed. Thus it is possible that multilayers were formed on the V/y-A1203 samples as well. Reoxidation Heats. As described earlier, exposure of a reduced V/y-A1,03 sample to oxygen resulted in consumption of the oxygen dose and reoxidation of the oxide. The heats and rates of reoxidation of these samples can be expressed as functions of the degree of reduction by defining a reduction parameter, 4. The reduction parameter after the kth oxygen dose is defined with
(3)
where ml = mol of H2consumed in pretreatment, = mol of O2 consumed in dose i, and m3 = mol of V in sample. Figure 4 shows some typical pressure and calorimetric data to demonstrate the stability of the pressure gauge and the calorimeter ~
0 -
'
0
5
10
, ,
.-
\ I.-*
20
15
25
I
30
40
Time (minutes) Power (mW)
4
I
3.5
t
,
3t 2.5
-
a
I
I
0
__
~
0
5
10
15
20
25
30
- . ~
35
40
Time ( m i d Figure 4. Volumetric and calorimetric data for reoxidation at 300 O C of a reduced V/y-AI2O3catalyst with 6.1 V/nm2. Top: normalized O2 pressure as a function of time. (a) cp = 0.45, (b) cp = 0.25, and (c) cp = 0.22. Bottom: thermograms for the same reoxidation doses shown above.
baseline. These data are comparable to adsorption calorimetry data collected by other ~ ~ r k e r ~ . Figure ~ ~ - 4a~ shows ~ , ~the~ * ~ ~ O2pressure as a function of time at 300 OC for a series of doses admitted onto a reduced 6.1 V/nm2 sample. It shows that the rate of pressure drop decreased as 4 decreased, that is, as the catalyst was reoxidized. The corresponding thermograms for the same three doses are shown in Figure 4b. The thermogram for 4 = 0.22 is considerably broader than the other thermograms. As 4 approached 0.2, the reoxidation rate became so slow and the subsequent thermograms were so broad that it was very difficult to accurately measure the heat generated from additional doses. The differential heats and the reoxidation rates for this sample are shown in Figure 5 . The first few doses of oxygen (high 6)were consumed rapidly as indicated by the small values of t(l/e). As the reoxidation proceeded, t(l/e) increased, reflecting a decrease in the reoxidation rate. At 4 = 0.2, when the reoxidation rate was too slow for accurate measurements, the tem-
~~~
(29) Griffith, W. P.;Wickins, T. D. J . Chem. SOC.A 1966, 1087. (30) Griffith, W. P.;Lesniak, P. J . B. J . Chem. SOC.A 1969, 1066. (31) Bond, G . C.;Perez Zurita, J.; Flamerz, S. Appl. Coral. 1986, 27, 353.
35
(32) Gravelle, P.C.Catal. Rev. Sei. Eng. 1977, 16, 31. (33) Gravelle, P.C.Adu. Coral. 1972, 22, 191.
Andersen and Kung Heat (kJ/mole 0 atoms) I
m 200 4%
f
@*++
* * +
.-
,Ti
{ 400 60 /
4 200 !
+
+
I
,
,
,
+ -
Y
2.9
x
8.2
0 0
'0.3
0.2
0.1
0
0.5
0.4
;o Figure 5. Differential heats and rates of reaxidation of 6.1 V/nm2 V/ y-A1203at 300 and 500 OC.
0.2
w
0.5
0.4
0.3
Figwe 7. Differential heats of reoxidation at 500 O C for V/y-A120, catalysts after partial reoxidation at 300 O C . Heat (kJ/mole 0 atoms)
Heat (kJ/mole 0 atoms) 200
0.1
200
~
I
+ +
1501
150,
Y
+
n
0 "
x
0 i
x x
0 100'
1
I 50
I I
~
I
-
01 0
0.1
:a
+
14
1
f
2.9
1
-
'
f
L 0.1
0.2
0.3
I
81
---J '
---
I
ApL--d-L 0, 0 0.2 0.3 0.4 0.5
2.9
8.2
1
I 1
I
-2
0.4
0.5
q;
Ffi
for V/y-A1203
Figure 8. Differential heats of reoxidation at 500 OC for V/A1203 catalysts.
perature was increased to 500 O C to increase the reoxidation rate allowing further measurement of reoxidation heats. The differential heat profiles at 300 O C for four V/7-A1203 catalysts are shown in Figure 6. In each case, the experiment wap terminated when the reoxidation rate was very slow. These curves show the strong dependence of the reoxidation heats on the vanadium coverage and 4. At comparable values of 4, the reoxidationheats for the low coverage samples were 25-30 kJ/mol higha than the heats measured for the higher coverage samples. This trend was observed over the range of 4 measured. The reoxidation heats were also strong functions of the degree of reduction of the sample. At 300 O C , the heats decreased with decreasing 4. The same trends were observed when the temperature was subsequently raised to 500 O C to continue the reoxidation (Figure 7). The reoxidation heats were also measured by conducting the reoxidation entirely at 5 0 0 O C (Figure 8). Similar to Figures 6 and 7, a decrease in the reoxidation heat with increasing vanadia
coverage was observed when 4 was greater than 0.2. The difference in reoxidation heats between the high- and low-caverage sampks was also approximately 25-30 kJ/mol. However, the differential heats were noticeably weaker functions of 4 at 5 0 0 O C than at 300 "C, indicating either a more bomg"surface or the Occurrence of some type of averaging process. This will be discussed in more detail later. The differential heat profiles for the two types of experiments for the 6.1 V/nm2 sample are shown in Figure 9 for comparison. The integral heats for both experiments, which arc the areas unda these curves, should be identical after correction for the temperature dimrences. This correction was made by assuming that the reduced vanadia had the same enthalpy as Vz04and the oxidized vanadia had the same enthalpy as V,O+ Using the literature values of heat capacities and heats of formation of the bulk oxides, it was estimated that the reoxidation enthalpy was 15.74 k J / d of 0 lower at 500 O C than at 300 O C . This value was then subtracted from the differential heats at 300 O C . The
F i p c 6. Differential heats of reoxidation at 300 catalysts.
O C
The Journal of Physical Chemistry, Vol. 96, No. 7, 1992 3119
Reoxidation of Reduced V20S/yA1203
TABLE 111: Chemical Analysis of Reduced V/y-Al203' V
Heat (kJ/mole 0 atoms) LWW
coverage, V/nm2 1.4 A + * +
i
2.9 6.1
4J
% V4+ 80 79 59 61 66 63
% Vs+ 20 21 41 39 34 37
chem anal.b 0.40 0.395 0.295 0.305 0.33 0.3 15
H2mnsC 0.44 0.45 0.29 0.29 0.30 0.29
I%
+ 100
"Samples reduced in H2up to 500 OC. See text for procedure. bCalculated from chemical analysis data. 'Calculated from H2 consumption during sample pretreatment.
-
+
TABLE I V Stoichiometries of Reduction and Reoxidation under High O2 Pressure (Approximately 20 kPa)' V coverage. V/nm2 H, consb 0,consb H,/O,
50
~~
*
-
__
0 0
0.1
5oooc 30O/60O0C
, . - ~
03
0.2
1.4 2.9 6.1 8.2
-
~.
05
04
I-.
d
Figure 9. Comparison of the differential heats of reoxidation for 6.1 V/nm2 V/7-A1203for reoxidation at 300 "C followed by 500 OC and for
reoxidation entirely at 500 OC. 350
W e ) (seconds) +I+
300
2 9 V/nm2
Reox
~+ 29V/nmz
Red
I+
61V/nm2
Reox
6 1 V/nm2
Red
+
'
1
L-1
250 1
200 II I
150-
100
O1 0
I
0.05
0.1
0.15
0.2
0.25
io Figure 10. Rates of reduction with H2at 500 OC and reoxidation with O2 at 500 OC for V/7-A1203 catalysts.
resulting integral heats calculated from numerical integration of these differential heat profiles (from 4 = 0.05 to 0.5) were 61.73 kJ/mol of V for the 300/500 OC experiment, and 65.15 kJ/mol of V for the 500 OC experiment. The close agreement of these two values provided a check on the validity of the calorimetric measurements. Reductim/Reoxidatim Kinetics. The rates of reduction by H2 and reoxidation by O2 at 500 OC as indicated by the values of t ( l / e ) for the 2.9 and 6.1 V/nm2 samples are shown in Figure 10. As mentioned earlier, the reoxidation rates decreased with decreasing 4 and became too slow to measure before the samples were fully reoxidized. The point at which this occurred depended on the vanadium loading. For low-loading samples, the reoxidation rate slowed down a t a higher value of 4 (i.e., in a more reduced state) than higher loading samples. The rate of reduction also differed for different vanadium loadings. For the 6.1 V/nm2 sample, the reduction rate did not depend on 4 over the range
183 34 1 706 884
95 172 362 468
1.93 1.98 1.95 1.89
"Samples reduced in H2 up to 500 OC. See text for procedure. bpmol consumed/g of Reoxidation done at 500 OC in 120 kPa of 02. sample. where the rates were measured. However, for the 2.9 V/nm2 sample, the rate decreased with increasing 4. The point at which the reduction and reoxidation rates were equal for a given sample gave an estimate of the degree of reduction of the sample and the corresponding reaction rates if the steady state reaction of H2 and O2were carried out over these samples. The data shown in Figure 10 indicated that, at steady state, the 2.9 V/nm2 sample would be more reduced and the reduction and reoxidation would occur at a slower rate than the 6.1 V/nm2 sample. CbemicalAnalysis of Reduced V/y-A1203Catalysts. The results of the chemical analysis on the distribution of vanadium oxidation states are presented in Table 111. From these data, one could calculate 4 using eq 3 if one assumed that H2was consumed during the pretreatment only to reduce V5+ to V4+ by removal of 0.5 0 atom/V to form a H 2 0 molecule. These values are shown in Table 111. Also shown in Table I11 are values of 4 calculated directly from the amount of H2 consumed during the pretreatment. These independent measurements of 4 agreed reasonably well. However, there were relatively large uncertainties associated with the titration technique which made these measurements less quantitative than volumetric measurements. The data in Table I11 showed that the H2 pretreatment reduced the Vs+ to V4+. No V3+was detected. It is possible that V3+ions which might have been produced in the reduction pretreatment were oxidized to V4+ on exposure of the sample to air. Even if this happened, the amounts of V3+ were small because of the agreement between the 4 estimated from the two methods. Extent of Reoxidation by 02.Measurements were made to determine if the samples could be reoxidized to the same states as those before the H2 reduction, that is, after oxidation at 500 OC and heating in vacuo at 400 OC. The samples were reduced with the standard H2 pretreatment described earlier and then reoxidized with a large excess of O2(approximately 20 kPa) at 500 OC. Table IV shows the amounts of H2 and O2 consumed and the H2/02 ratios. Within experimental error, these ratios were 2, the value expected for a complete reduction/reoxidation cycle. The slight deviations from 2 might be due to small amounts of thermal reduction that occurred during the evacuation prior to the reduction step, but the deviations were within experimental uncertainties. These results showed that the effect due to H2 reduction could be completely reversed by O2treatment at 20 kPa and 500 OC. However, it is interesting to note that the samples were reoxidized to values of only C$ = 0.1-0.2when they were reoxidized during calorimetric measurements using I 1 33 Pa of O2 at 500 OC. The value of 4 at the end of the calorimetry experiment
3120 The Journal of Physical Chemistry, Vol. 96, No. 7 , 1992
Andersen and Kung Heat (kJ/mole 0 atoms) - .
V WVemgC, V/m2
H2 COns,' pmol/g
3oob
5W
500-HPd
H2/02'
1.4 6.1
21 1 911
45.5 248
16.6
400
104 441
2.02 2.06
O2cons, pmollg
200 I
150 -
opmol consumed/g of sample. *Total O2consumed at 300 OC,final O2 pressure 40.133 kPa. CTotalO2 consumed at 500 O C , final O2 pressure 10.133 P a . dTotal O2consumed at 500 OC, final O2 pressure 220 kPa. * Ratio of H2 consumed to total amount of O2consumed at 500 OC and final O2pressure 1 2 0 kPa.
...'.
I
m+.7!A
i
---
+
't
100
Heat (kJ/mole 0 .atoms) .. .- - - __. .__ _-
- +
1 I
0L 0
t
300°C
t
1
f
J
_'
€.000c - P O c
~
0.1
L
0.2
l
t
Reduced with CO
,
Reduced w i l h H 2
-
L 0.3
.
l 0.4
0.5
(-Q
6OO0C
100
-
12. Comparison of the differential heata of reoxidation for a 6.1 V/nm2 V/y-A1203catalyst after reduction with H2 a d reduction with
co.
I,.,,.:...... I
0
,-__ 0
figures that the profiles were essentially the same (within experimental uncertainties) independent of the reducing agent.
I
60
~
+
Reduced with H,
L -
0.1
0.2
0.3
0.4
0.5
io 11. Comparison of the differential heats of reoxidation for a 1.4 V/nm2 V/y-A120, catalyst after reduction with H2 and reduction with co.
depended on the vanadium coverage due to the variation in the reoxidation rate as described earlier. Further consumpion of O2 could be detected by exposing the samples to a large O2dose to bring the final preseure to 20 kPa. The amounts of O2consumed during 300 OC reoxidation by small O2d o w (final O2pressure I133 Pa), and then during 500 OC mxidation by small O2doses (final O2pressure I133 Pa), and finally during 500 OC reoxidation by large O2doses (final O2pressure L 20 @a)) for the 1.4 and 6.1 V/nmz samples are shown in Table V. Since the HJOZ r a t h afta the final 0 2 treatmmt wen 2,t h a data showed that t$ could be calculated by either eq 3 or eq 4,
(4)
where m l= total mol of O2consumed after O2P L 20 kPa, mu = mol of O2coasumed in dose i, and m3 = mol of V in sample, if the 02 pressure was raised to 20 kPa following the calorimetry experiment. R e d d a b of CO Reduced Smpks. In these experiments, there wae diffcultics in volumetrically measuring the CO consumption. Thus,to determine 4, the small O2doses used for the cakrimeery experiments were followed by a large O2dose to bring tbe fhd 02press~reto 20 k h and thus tbe sample wu completely reoxidized as described previously. This allowad the values of 4 to be calculated from eq 4. The differential heat profile for the 1.4 V/nm2 sample after CO reduction is shown in Figure 1 1. For comparison, the profie after H2 reduction is also shown. Figure 12 shows these two profiles for the 6.1 V/nmz sample. It can be Seen in these two
Disctu!3ioa Rehbetweem $t md tbc Exteat of Redoctb& To relate the measured calorimetric data to the catalytic behavior of these samples, it is necessary to provide a molecular interpretation to the defined parameter 4. The chemical analysis and the redox stoichiometry described earlier provide insight into the molecular transformations due to the H 2 and O2treatments. Treatment of a sample with H2 results in reduction of the vanadia and generation of gaseous H 2 0 . The chemical analyses indicate that the vanadia is predominantly reduced to the 4+ oxidation state. Haber et aL3' have suggested that one V-OH group forms on the vanadia phase for every oxygen removed from the vanadia by H 2 in the form of H,O. Using gravimetric measurements, they determined that about equal amounts of oxygen atoms per vanadium werc removed by reduction with either H2 or CO. In separatepulsed reoxidation experiments, V/y-A1203 samples reduced with H2were found to consistently consume 1.5 times as much oxygen as CO reduced samples. This suggested that some oxygen molecules were consumed to remove the V-OH groups formed during reduction on the H2 reduced sample. Unfortunately, few details were reported concerning the conditions which the samples were subjected to in bctwseo the reduction and mxidation treatments. F i i y , gravimetric measu~emcntss h e d that treatment in 13.3 kPa O2completely reoxidized the reduced samples. This last point was also obsmed h m that heating with 20 kpa of O2completely reoxidized the samples. In our experiments, the samples were heated at 500 OC under dynamic vacuum for 10 h following the reduction treatments. The H 2 / 0 2 ratios reported in Tables IV and V indicate that any hydroxyls which may have formed during the H2 reduction were either removed as H 2 0 during this evacuation treatment or during reoxidation. In the latter c8se,the presence of adsorbed hydroxyls should sisnificantly affect the heats of mxidation. Since reduction with CO could not generate any hydroxyls, om mild expact the heats of reoxidation to be quite different for samples reduced by the different reductants. The fact that the heats of reoxidation following the reduction by H2 and by (30 were essmtidy identical (Figures 11 and 12) suggests that the hydroxyls were removed (34) Habcr, J.; Kozlowska, A.; Kozlowski, R. J .
Cad.1986, 102, 52.
-
The Journal of Physical Chemistry, Vol. 96,No. 7, 1992 3121
Reoxidation of Reduced V20S/~-A1203
I
Heat (kJ/mole 0 atoms)
~~~~
_~
~~
195
H a , 4OO0C
Sample heated at 5OO0C
175
/ \ ,. A,
.. . . ..-. ... :. . ,'. ,. .. ...
.
*
" m
I '
I-
I "
155
02,5OO0C
02,3OO0C
m
' *
I
'
135 .>..
115
0 2 ,5OO0C 95
Legend
111 __
.
.
75
Fully Oxidized Partially Reduced
~
0
01
~- .._~ ~-
0.3
0.2
04
05
I
Reduced
Figure 13. Schematic representation of the proposed reoxidation mechanisms.
during evacuation in the pretreatment. Summarizing, the number of oxygen atoms removed from the supported vanadia phase by the H2 pretreatment (including the 10 h, 500 OC evacuation) can be calculated directly from the quantity of hydrogen consumed. Thus, $ can be interpreted as oxygen atoms removed per vanadium. Reoxidation Heats and Rates as a Function of $. One common feature observed in this study is that the rate of reoxidation decreases rapidly as $ decreases. At 300 OC,the rate under low O2pressures becomes too slow for measurement before the catalyst is fully reoxidized. In fact, a higher pressure oxygen treatment at 500 OC is needed for complete reoxidation. This behavior can be interpreted with the model summarized in Figure 13. Fully oxidized vanadia exists as a dispersed phase on yA1203. Upon reduction, the dispersion is reduced significantly, and particles of reduced vanadia are formed. When these reduced particles are reoxidized at 300 OC in low O2pressure, only the near surface region of the particle is oxidized due to the limited mobility of lattice oxygen at this temperature. When this outer shell becomes fully oxidized,the reoxidation rate stop due to the lack of available reduced vanadia on the particle surface. At a higher temperature of 500 OC,the mobility of the oxygen ions is significantly higher such that the various suboxides of vanadium are formed much more evenly throughout the reduced particle before additional doses are administered. Due to the decreasing density of surface sites which activate oxygen as the particles are reoxidized, a higher pressure of oxygen is needed for complete reoxidation at a reasonable rate. Except for the formation of particles after reduction and redispersion after reoxidation, this mechanism is similar to those commonly observed for oxidation of metal^.^^^^^ Redispersion of the vanadia upon reoxidation is supported by the fact that the Raman spectra of the catalysts after a reduction/reoxidation cycle are identical to those collected for a fresh sample. This fact also suggests that it is not likely that a vanadium aluminum mixed oxide compound was formed as a result of the redox cycle. The formation of particles of reduced vanadia from a dispersed phase of oxidized vanadia has been proposed by previous workers. Sobalik et observed that, after H2 reduction of V/y-A1203 catalysts, the intensities of infrared bands assigned to probe (35) See, e.&; Kubaschewski, 0.;Hopkins, B. E. Oxidarion of Metals and Alloys, 2nd ed.; Butterworths: London, 1962. (36) Sobalik, Z.; Kozlowski, R.;Haber, J. J. Cutal. 1991, 127, 665.
Figure 14. Effect of heat treatment on the reoxidation behavior of 6.1 V/nm2 V/y-A1203.
molecules adsorbed on the exposed A1203increased. Odriozola et al.37observed that the intensity of the vanadium XPS signal decreased after a V/y-A1203sample was reduced with NH3. In both cases, these effects were attributed to a decrease in the dispersion of the supported vanadia phase. In other words, the reduced vanadia phase agglomerated into particles exposing the A1203surface. A similar behavior was observed by Tops0e3*for V/Ti02 catalysts. At a high extent of reduction, the reoxidation heats are essentially constant for the 300 and the 500 OC reoxidations as shown in Figure 9. The lower heat at 500 OC is a result of the different heat capacities between the reactants and products. This difference can be compared with those reported for the oxidation of V204 to V20SrZ1which is 136.74kJ/mol of 0atom at 300 OC and 121.0 kJ/mol of 0atom at 500 OC. Within experimental uncertainties, the 15.74kJ/mol of 0atom difference is comparable to the 18-20 kJ/mol of 0 atom difference observed in our experiments. At lower values of 4, the differential heats of reoxidation decrease with decreasing $. However, the sharp decrease in the heat occurs at a much lower $ at 500 O C than at 300 "C.This can be understood with the model described. At 500 OC,the higher mobility of the oxygen ions results in a more uniform reoxidation of the vanadia particles which is assisted by the simultaneous redispersion of the vanadia. Thus an essentially constant differential heat is observed until the sample is fully oxidized. At 300 OC,the differential heat decreases rapidly as soon as the near surface region is nearly fully oxidized. Additional experiments were conducted to test the model of segregation into particles on reduction and redispersion on reoxidation. A 6.1 V/nm2 sample was pretreated and reoxidized at 300 OC as in the standard experiments. When the reoxidation rate became very slow (at $ = 0.21),the temperature was increased to 500 OC in the absence of oxygen for 3 h. The temperature was then lowered to 300 OC,and the reoxidation was continued. The results of this experiment are presented in Figure 14. After the heat treatment, the rate parameter, t ( l / e ) , decreased by a factor of 10, from 474 to 48 s, reflecting a large increase in the reoxidation rate, but the rate decreased rapidly as t(1 /e) increased rapidly to 152 and then 444 s in the subsequent O2doses. The points after the heat treatment are plotted with (37) Odriozola, J. A.; Soria, J.; Somorjai, G. A.; Heinemann, H.; Garcia de la Banda, J. F.; Lopez Granados, M.; Conesa, J. C. J . Phys. Chem. 1991, 95, 240.
(38)
Topme, N. J. Catal. 1991, 128, 499.
Andersen and Kung
3122 The Journal of Physical Chemistry, Vol. 96, No. 7, 1992 Heat (kJ/mole 0 atoms)
Heat (kJ/mole 0 atoms)
200 r-
150
-
100
-
~
*.
*
7
50
increasing degree of ' overall oxidation or increasing oxidation I temperature
-
t
-
I
, t
--
6
- Previoui 300/6OO0C Reoxidallon *
.
0' 0
0.1
. .__
I
Prevloua 6 0 0 ° C Reoxidation
1
I.
0.2
~'
0.3
0.4
0.5
Flgm 15. Differential heats of reoxidation for a 6.1 V/nm2 V/y-A1203 where the reoxidation temperature was cycled between 300 and 500 OC. The heats measured at 300 O C a n comcted to account for the difference in enthalpy between 300 and 500 O C .
the same symbol as the points before the heat treatment which had approximately the same reoxidation rate. According to the model the inaeese in the reoxidationrate after the beat treatment in vacuo results from a redistribution of oxygen through the reduced partick and redispersionof the vanadia phase at 500 OC. After the redispersion, there is a significant quantity of reduced vanadia on the catalyst surface which can be rapidly reoxidized with subsequent oxygen doses. When the same experiment was performed on the lowest coverage sample, similar results were obtained. This indicates that even with a vanadia coverage as low as 1.4V/nm2, the vanadia can agglomerate to form reduced particles. Thus interpretations of the behavior of a catalyst based on a model of isolated vanadia species, even at very low loadings, must be made with extreme caution. In addition to imrtasingthe rate of reoxidation by heating the sample to 500 OC in vacuo, the differential heat of reoxidation is also higher for the oxygen doses immediately after the heat treatment. This latter point is more evident in another experiment in which the mxidation temperature was cycled between 300 and 500 OC. F v 15 shows the results of this experiment. The heats measured with the standard procedure are shown for reference. Also, the measured heats at 300 OC were reduced by 15.74kJ/mol to account for the reoxidation heat difference between 300 and 500 oc. These results demonstrate that the measured differential heat depends not only on 4, the overall degree of reduction, but also on the dispersion of the vanadia phase on the A1203surface and the uniformity of the oxygen dispersion throughout the vanadia phase. Figure 16 shows conceptually the dependence of the heat at a given temperature on the overall degree of oxidation of a vanadia particle for different degrees of reduction at the near surface region, 4,. The size of the 4, region depends on the temperature of reoxidation, that is, on the rate of lattice oxygen diffusion. Curves A-F are for particles of increasing degree of oxidation in the bulk. They can also be used to represent the dependence on the temperature, which increases from A to F. Labeled on the curves are points that correspond to the various states of the sample in Figure 15. Briefly, the sample was first reoxidized at 300 OC and the heats followed curve A. The temperature was incteased to 500 OC at point 1 (4 = 0.31). Diffusion of lattice oxygen from the surface to the bulk was enhanced, the surface region became more reduced, and the sample moved to point 2. The differential heat was lower than that measured when
(ps, degree of oxidation of surface region Figure 16. Relationship between the reoxidation heat, the degree of reduction of acccssible vanadia, and the uniformity of the oxygen distribution in the support phase.
the entire experiment was coducted at 500 O C because the sample was at 500 OC only for a short time at this point. The oxygen ions continued to redistribute in the sample for the period at this temperature. When the temperature was returned to 300 OC at 4 = 0.26, the sample was at point 3 on curve C and a high reoxidation heat was registered. The rest of the experiment can likewise be explained. Rcoxklrtba Hcrb .ad Ratm u1 a b t h of the V.nrdir Coreraga As mentioned earlier, the high-coverage samples rtoxidized to a lower value of 4 than the lowcoverage samplcs under identical conditions (Figures 6 and 7). Since high pressures of oxygen could reoxidize fully all samples, the difference at low O2 pressures must be attributable to kinetics, i.e., the low-coverage samples reoxidize more slowly than the high coverage samples. A similar effect was reported by Kijenski et al.,39 who showed that the temperature at which the maximum reduction rate occurred in a temperature-programmed reduction experiment decreased by 10 O C as the vanadia coverage increased from l to 1.8 V/nm2. The kinetic difference may be explained by structural differen= between the samples. As indicated by the Raman spectra, the lowcovcrage samples contain small and well-separated patches of noncrystalline vanadia on the yA1203 surface. Each patch needs to be reduced and reoxidized separately. The higher coverage sample has an extensive two-dimensional vanadia network and possibly no isolated vanadia patches. This allows significant migration of oxygen through the vanadia phase inaeasig the rates of reduction and reoxidation substantially. The structuraldifferenasof the supported vanadia phase among the different samples may also be the reasan for the variation in the reoxidation heat with vanadia coverage. For the same 4, the major change in the reoxidation heats occurred as the vanadia coverage was increased from 2.9 to 6.1 V/nm2 (see Figures 6 and 7). Essentially the same reoxidation heats were obtained for the two low-coverage samples, which both have submonolayer coverages, and another heat for the two higher coverage samples, which both have vanadia multilayers. This sharp change in the reoxidation heats may be directly related to the formation of vanadia multilayers at a coverage of about 4.4 V/nm2. The difference in reoxidation heats for mondaya and multilayer samples could be related to differences in the ease of structural rearrangement of the vanadia phase. The low-coverage, highly ~~~~
~
(39) Kijenski, J.; Baikcr, A.; Glinski, M.; Dollenmeier, P.; Wokaun, A. J . Carol. 1986, 101. 1.
Reoxidation of Reduced V205/yA1203 dispersed samples have a significant portion of oxygen ions that bridge the vanadium ions to aluminum ions in the support, and terminal oxygen ions bonded only to vanadium ions. Both of these types of oxygen are likely to be much more difficult to remove (higher reoxidation heat) than those bridging two vanadium ions (lower reoxidation heat) because the vanadium ions are much more reducible than the aluminum ions. The Raman spectra of the higher coverage samples show the presence of a vanadia network with extensive V-0-V bonding. One would expect that such a network could accommodate structural changes due to oxidation and reduction much more readily than the highly dispersed phase which requires disrupting the more rigid V-0-A1 bonds. Finally, the measured integral reoxidation heat at 500 O C for the 6.1 V/nmz sample, 65.15 kJ/mol of V atoms, and the measured integral heat for the 2.9 V/nm2 sample, 67.81 kJ/mol of V atoms, are quite close to the heat of oxidation of V204to V205 at 500 OC reported in the literature, 60.50 kJ/mol of V.zl This indicates that the main effect of the support is to change the differential heat profile, not the integral heat. This can be seen most clearly in Figure 8. The heat of reoxidation for the higher coverage sample was initially lower than that for the lower coverage sample. At about 4 = 0.12, this relationship is reversed. As a result, the integral reoxidation heats for these two samples are about the same. Effect of the Reoxidation Heats and Rates on the Catalytic Behavior. For reactions in which the oxide catalysts undergo redox cycles and lattice oxygen are involved, it is expected that the heats and rates of reoxidation have direct influence on the observed catalytic properties of the oxides. If the rate-limiting step of the reaction involves removal of a lattice oxygen or reoxidation of a reduced oxide, the rates of these processes will determine the observed catalytic rate. One expects that the faster are these rates, the more active is the catalyst. If the rate-limiting step does not involve oxidation/reduction, the heats of reduction and reoxidation and the rates can still affect the selectivity because they will determine the relative rates of the competing surface processes some of which involve lattice oxygen and some do not. One expects that the higher the heat of reoxidation (that is, the lower the heat of reduction) and the slower this rate, the lower the rate of formation of oxygen-containing products or, alternatively, the lower the number of oxygen atoms being incorporated into the product. The data obtained in this study show that in order to understand the catalytic behavior of a catalyst using the simple argument
The Journal of Physical Chemistry, Vol. 96, No. 7, 1992 3123 above, it is necessary to know the state of the sample and the reaction condition: the overall degree of oxidation of the oxide, the degree of oxidation in the near surface region, the previous history of the sample, the reaction temperature, and oxygen partial pressure. This is because the heat and rate of reoxidation depend on these parameters, as shown in Figure 15. The data also show the strong dependence of the heat and rate of reoxidation on the degree of reduction of the sample and a weaker but still significant dependence on the vanadia loading. Thus at steady state, it is very probable that the degree of reduction of vanadia is different for samples of different loadings. If a V4+ ion is the site for dissociative adsorption of hydrocarbon and this step is rate limiting, catalysts of different loadings and thus different degrees of reduction at steady state would show very different activities. Work is continuing to elucidate these points.
Conclusions Hztreatment at 400 OC of V/yA1203 catalysts reduces the vanadium ions from the +5 to the +4 oxidation states. The vanadia phase formed particles upon reduction. These particles could be redispersed upon reoxidation at 500 O C . The heat of reoxidation of reduced V/y-A1203was a function of the degree of reduction, the dispersion of the vanadia phase, and the uniformity of the distribution of oxygen through the vanadia phase. Both the heats and rates of reoxidation decreased as the samples were reoxidized. The uniformity of reoxidation of reduced vanadia depended on the reoxidation conditions. At 300 O C , an oxidized shell formed on the ncar surface region of reduced vanadia particles and further oxidation was limited by diffusion of oxygen into the reduced portion of the particle. This resulted in large variations of the reoxidation heat with the degree of reduction. At 500 O C , the reduced particle reoxidized more evenly with a reoxidation heat essentially independent of the degree of reduction until the sample was nearly fully oxidized. The reoxidation heat decreased by 25-30 kJ/mol as the vanadia coverage increased from submonolayer to multilayers. The reoxidation rate was faster for higher vanadia coverage. Variations in the vanadia coverage alter the differential heat profiie but do not significantly affect the integral reoxidation heat. Acknowledgment. This work was supported by the National Science Foundation, Chemical and Thermal Systems Division.