J . Phys. Chem. 1989, 93, 42 13-42 18 difficult. The alternative procedure developed in the present work results in migration-stabilized fragments chemically anchored into accessible zeolite pores.
Acknowledgment. Acknowledgment is made to the donors of the Petroleum Research Fund, administered by the American
4213
Chemical Society, and to the Sandia-University Research Program (DOE), for partial support of this research. The operational funds for NSLS beam line X- 1 1 A are supported by DOE Grant DEAS0580Ei-10742. Registry No. [l], 12093-05-9; [3], 41853-19-4.
Extended X-ray Absorption Fine Structure Studies on the Structure Change of the Al,O,-Attached [Co"], Catalyst during a CO Oxidation Reaction Kiyotaka Asakura and Yasuhiro Iwasawa* Department of Chemistry, Faculty of Science, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113. Japan (Received: August 10, 1988; In Final Form: November 23, 1988)
The dynamic structure change of active Co sites in the AI2O3-attached [ C O ~ catalyst I]~ derived from C O ~ ( C O ) ~ / A I ~inO , the course of the catalytic CO oxidation with O2under atmospheric and high-pressure conditions was studied by means of in situ extended X-ray absorption fine structure (EXAFS). The C Oensemble ~ composed of four adjacent Co monomers (Co-0 = 0.197 nm) as a local structure was converted to the oxygen-bridged Co dimer structure with Co-0 = 0.196 nm and Co-Co = 0.330 nm in the reaction with 02.The bridged oxygen readily reacted with weakly adsorbed CO to produce C02, the original Co, ensemble structure composed of four adjacent Co monomers being recovered. The catalytic oxidation of CO with O2 at 273 K proceeds by a two-stage reaction mechanism involving the dissociative adsorption of O2 and the subsequent reaction with CO, in conjunction with the structure change between the Co monomer structure and the Co dimer structure.
The majority of studies on the reaction mechanisms for the solid catalysts has been carried out by means of the reaction kinetics, tracer techniques using an isotope in the reactants, and the identification of surface adsorbed species including reaction intermediates by means of IR, NMR, ESR, etc. However, in many of these studies the change of the surface structure of the active site and its dynamic change during the catalytic reaction, the knowledge of which is important for a complete understanding of catalytic reactions, are beyond the scope of the search. Recent developments in electron spectroscopies provide direct information on physical properties such as surface composition, surface structure, and electronic states, as well as on chemical phenomena such as adsorption and reaction on single-crystal surfaces.l However, these techniques require ultrahigh-vacuum (UHV) conditions, which are too different from real catalytic reaction conditions. This restriction prevents in situ observation of solid surfaces under reaction conditions. Some studies have shown that static surface structures and reaction mechanisms in vacuo are entirely different from dynamic ones under reaction conditions.z4 Thus, the direct observation of the active structures of solid surfaces during catalysis is of great importance for an understanding of the reaction mechanisms and the origin of catalysis and also to improve or to design better solid catalysts by this fundamental knowledge. EXAFS (extended X-ray absorption fine structure) has such prominent features that it can give direct information about the local structure around an X-ray absorbing atom without the requirements of long-range ordered structure and UHV condit i o n ~ . ~Thus ~ ~ - EXAFS ~ is a powerful tool for the determination ( I ) King, D. A.; Woodruff, D. P. The Chemical Physics of Solid Surfaces and Heterogeneous Catalysts; Elsevier: Amsterdam, 1982; Vol. 4.
(2) Zaera, F.; Gellman, A. J.; Somorjai, G. A. Acc. Chem. Res. 1986 19, 24. (3) Imbihl, R.; Cox, M . P.; Ertl, G . J . Chem. Phys. 1986, 84, 3519. (4) Wang, T.; Lee, C.; Schmidt, L. D.; Surf Sci. 1985, 163, 181. (5) Van't Blik, H. F. J.; Van Zon, J . B. A. D.; Huizinger, T.; Vis, J. C.; Koningsnerger, D. C.; Prins, R. J . Phys. Chem. 1983, 8 7 , 2264. (6) Nishimura, M.; Asakura, K.; Iwasawa, Y. Int. Congr. Catal., Proc. 9th 1988, 4 , 1842.
0022-3654/89/2093-4213$01.50/0
of surface structures of metal catalysts in the presence of reactant gas even at high pressures. Conventional supported metal catalysts usually contain many kinds of metal atoms with different environments and reactivities, most of which may not directly participate in the catalytic reaction. The EXAFS analysis provides averaged information about these structures. Inorganic oxide attached metal catalysts prepared by the reaction between the surface O H groups of an inorganic oxide such as Si02and A1203 and organometallic complexes possess well-defined surface structures showing almost uniform reactivity.lO*llIn this way one can overcome the equivocal and restricted spectroscopic information due to the surface heterogeneity and complexity of conventional catalysts. Recently we have carried out structure studies on a Mo dimer catalyst in the course of the oxidation of C 2 H 5 0 Hwith O2by means of in situ EXAFS.I2 We revealed that the molybdenum dimers changed their local structures by ca. 0.04 nm laterally and by ca. 0.01 nm vertically during the reaction: The catalytic C2H50H oxidation takes place in conjunction with the structural oscillation.I2 The A1203-attached Co(I1) cluster catalyst derived from C O , ( C O ) ~showed a much higher activity than the conventional impregnated or ion-exchanged Co/AI2O3 catalysts for the oxidation of CO with On the other hand, the SO2-attached 02.13914
(7) Sayers, D.E.; Stern, E. A.; Lytle, F. W. Phys. Rev. Lett. 1971, 27, 1204. (8) Lytle, F. W.; Via, G. H.; Sinfelt, J. H. J . Chem. Phys. 1977, 67, 3831. (9) Bart, J. C. J.; Vlaic, G.Adu. Catal. 1987, 35, I . (10) (a) Yermakov, Yu. I.; Kuznetov, B. N.; Zakharov, V. A. Catalysis by Supported Complexes; Elsevier: Amsterdam, 1981. (b) Ugo, R.;Psaro, R.; J . Mol. Catal. 1983, 20, 5 3 . (c) Basset, J. M.; Choplin, A. J . Mol. Catal. 1983, 21, 95. (d) Gates, B. C.; Guczi, L.; Knozinger, H. Metal Cluster Catalysrs; Elsevier: Amsterdam, 1986. ( I 1) (a) Iwasawa, Y. Tailored Metal Catalysts; D. Reidel Publishing Co.: Dordrecht, The Netherlands, 1985. (b) Iwasawa, Y. Adu. Catal. 1987, 35, 1..R7
(12) Iwasawa, Y.; Asakura, K.; Ishii, H.; Kuroda, H. Z . Phys. Chem. (Munich) 1985, 144, 105. (13) Iwasawa, Y.; Yamada, M.; Sato;Y.; Kuroda, H. J . Mol. Catal. 1984, 23, 95. (14) Yamada, Y.; Iwasawa, Y. Nippon Kagaku Kaishi 1984, 1049.
6 1989 American Chemical Society
4214
Asakura and Iwasawa
The Journal of Physical Chemistry, Vol. 93, No. 10, 1989 oc,j $oOC C'-OoC-
'co
r
CO CO CO
A1203
C04(C0)12 /A1203
313 K
02 293 K
513 K
,
( C )
\\
rCo-O =0.202 nm rCo-co=0.321 nm
rCo-co=0.249 n m
rCo-O E O . 1 9 7 nm
02/773 K
rco-o =O. 194 nm
rCo-co=O. 254 nm
Figure 1. Structure transformations of the Co species attached to A1203and SO,. I i q N7 Reservoir
Co(l1) catalyst prepared in a way similar to that of the A1203-attached catalyst showed little activity for the same reaction.l39l4 This difference in catalytic activity has been ascribed to the difference in the local Co(I1) s t r u ~ t u r e s . l 'In~ ~the ~ present paper we studied the detailed structure of the A1203-attached Co(11) oxide catalyst and the structural change during CO oxidation in the presence of 0, or CO a t atmospheric and high pressures by an in situ EXAFS technique.
Experimental Section Preparation of Catalysts. The catalysts were prepared in a similar way to that described p r e v i ~ u s l y . 'A1203 ~ ~ ~ ~and SiO, were pretreated a t 573 and 473 K, respectively, for 1.5 h in 0, to regulate the density of the surface OH groups of the supports, followed by evacuation for 1 h. The surface areas of A1203and Si02were 150 and 500 m2.g-I. C O ~ ( C Owas ) ~ deposited on the AI2O3or SiO, by a dry mixing m e t h ~ d . l ~The % ' ~Co loadings were 5 wt % for both supports. The samples were then treated at 3 13 K under vacuum to complete the reaction between the cobalt carbonyls and the support surfaces. The subsequent treatments for the structure transformations of surface Co species reported in previous paper^'^,^^ are summarized in Figure 1. The previous I R , UV-vis, and EXAFS studies showed that the deposited C O ~ ( C Owas ) ~ converted to C3"C O ~ ( C O )on , ~S i 0 2 (d) and disC O ~ ( C Oon ) ~A1203, ~ torted tetrahedral D2d or rectangular D4,, respectively. The C O , ( C O ) ~on~ AI2O3was completely decarbonylated with O2 to form the COO]^ cluster (b), which was transformed to the attached Co(I1) structure (c) by heating [COO], to 513 K under vacuum. The IR spectra of N O adsorbed on c as a function of the amount of adsorbed N O suggested a uniform environment around the Co(I1) species (c).I4 The SiO,-supported C O ~ ( C O (d) ) ~ ~was converted to the Co(I1) structure (e) by exposing it to O2 a t room temperature followed by evacuation up to 513 K. When [ C o o l 4 (b) on A1203and C O ~ ( C O )(d) , ~ on SiO, were treated with 0, at 773 K. the small spinel-like clusters [Co3O4In(f) were formed.I3*l4 The impregnation 5 wt % Co/A1203 catalyst was obtained by an impregnation method using an aqueous solution of Co(NO3), followed by drying a t 390 K for 2 h and calcining a t 773 K for 5 h in air. CO Oxidation Reaction. CO oxidation reactions were carried out in a closed circulating system a t 273 K and O2:CO = 6.4:3.2 kPa, using 0.287 g of catalyst. EXAFS Study. EXAFS spectra were measured a t BL-1OB of the Photon Factory in the National Laboratory for High Energy Physics (KEK-PF).I5 In order to examine the structure of Co sites under atmospheric and high-pressure conditions, we have made an in situ high-pressure EXAFS cell with two Be windows sealed with an O-ring, as shown in Figure 2, by which the EXAFS ( I 5) (a) Oyanagi, H.; Matsushita, T.;Ito, M.: Kuroda, H. KEK Rep. 1984, 83-30. (b) Nomura, M . K E K Rep. 1985,85-7.
Be window ~-
Figure 2. Chamber for the EXAFS measurements under vacuum and at high pressures.
measurements in the presence of 30 atm of gas were possible. The reactant gases were allowed into the cell through the valves in the upper lid. The sample could be cooled down to 100 K with liquid N2 and heated up to 733 K by using the resistive heater in the furnace. The upper lid and the outer cylinder were cooled with water. The Al,O,-attached [ C o o l 4 (b) was pressed into a self-supporting disk and put into the center of the furnace in the high-pressure EXAFS cell. The EXAFS spectrum was extracted by using a cubic spline method and normalized to the edge height. The k3-weighted EXAFS spectrum was Fourier transformed to r space, and the inversely Fourier-filtered data were analyzed by using a curve fitting method on the basis of the single-scattering plane-wave theory as expressed by eq where k is a wave vector, and k3X(k) = ~ k 2 N l S l ( k ) F l ( kexp(-2a?k2) ) sin (2kr,
+ di(k))/rf2
I
(1)
N,, S,(k), o,,and r, represent a coordination number of the ith shell, an amplitude reduction factor, a Debye-Waller-like factor, and an interatomic distance of the ith shell, respectively. We assumed S,(k) was independent of k because Si(k) consists of two factors-( 1) the finite lifetime of both core hole and photoelectron and (2) the effect of the passive electron-which have opposite k dependences.)* The S,(k) values for Co-0 and Co-Co were (16) Stern, E. A. Phys. Reu. B 1974, 10, 3027. ( 1 7) Teo, B. K. EXAFS Spectroscopy: Basic Principles and Data Analysis: Inorganic Chemistry Concepts 9; Springer-Verlag: Berlin, 1986. (IS) Teo. B. K.; Antonio, M. R.; Averill, B. A. J . A m . Chem. S O C .1983, 105. 3 1 5 1 .
The Journal of Physical Chemistry, Vol. 93, No. 10, 1989 4215
Structure Changes in an A1203/[Co11]4Catalyst TABLE I: Curve Fitting Analyses of EXAFS Data for the Standard Co Samples co-0
N
samde co coo Co304
6c 5.0b 5.44"
co-co
rlnm 0.213 0.2133' 0.192 0.195''
co-co
N
r/nm
N
rlnm
12Q 5.gb 12' 5.6b 6d
0.248 0.2509" 0.300 0.3017' 0.284 0.286d
8.3b 0.933"
0.336 0.33gd
20
m t i'\
101
/ I
n
1
Reference 20. b T h e coordination numbers for the Co-0 and the Co-Co were calculated from the S values of the Co-0 bonding in the COOand the CGC -o bonding in the Co metal, respectively. CReference 21. d T h e averaged values of Co,,,-0 and Cow,-0 or Cotet-Cotetand CO,,,-CO,~ are from ref 22. (I
TABLE 11: Curve Fitting Analyses for the EXAFS Data of Co Catalystso co-0
co-co
sample
N
rlnm
N
species b species c species e
2.3 2.3 2.1
0.202 0.197 0.194
1.5 0.321 not obsd not obsd
rlnm
" T h e errors in the EXAFS analyses were estimated to be 0.003 nm for the interatomic distances and 0.4 for the coordination numbers.
determined as 0.40 and 0.65 from the corresponding bondings of COO and Co metal. The theoretical phase shift (di(k)) and amplitude functions (F,(k)) were used for each shellla except for the analysis of [Co304],(f) where the empirical phase shift and amplitude functions directly obtained from the analysis of bulk c0304 were used. IR Specrroscopy. The IR spectra of the supported Co carbonyl clusters and of the adsorbed CO on species c were taken on a JEOL JIR- I O FT-IR spectrometer. The self-supporting sample was set in an IR quartz cell combined with a closed circulating system.
Results and Discussion Structures gf Surface Co Species Determined by EXAFS. Tables I and 11 show the curve fitting analyses of EXAFS data for some standard Co compounds and the Co catalysts, respectively. At first we examined the validity of the theoretical phase shift and amplitude functions by comparing the bond distances determined from the EXAFS analyses using the theoretical phase shift and the amplitude functions with those of the crystallographical data in Table I. A good agreement between the C o - 0 and Co-Co bond distances obtained from EXAFS and X-ray structure analysis was found, as shown in Table I . In the case of the Co-0 distances in c0304 having both tetrahedral and octahedral Co sites, the averaged value agreed well with the value obtained from EXAFS, as shown in Table I. The Co-Co distance in the second shell was determined to be 0.284 nm, corresponding to 0.286 nm of the X-ray structure analysis. The third shell in c0304 is composed of two kinds of Co-Co distances, i.e. T&oT&o(Co-Co distance: 0.349 nm; coordination number: 4) and T6Co-Oh-Co (Co-Co distance: 0.334 nm, coordination number: 12). The value determined from the EXAFS analysis was 0.336 nm, in agreement with the weighted average Co-Co distance of 0.338 nm from the crystallographic structure. Thus we concluded that the present EXAFS analyses using the theoretical parameters can estimate the Co-0 and Co-Co distances within an e r r o r of 0.003 nm. Figure 3 shows the EXAFS Fourier transforms of the A1203-supported Co catalysts derived from the CO,(CO)~together with those of the Si02-attached species (e). In the Fourier (19) (20) lishers: (21) (22)
Balandin, A . A . Adu. Catal. 1958, 60, 96. Wyckoff, R. W . G. Crystal Structures, 2nd ed.; lnterscience PubNew York, 1963; Vol. I . Greenwald, S . Acta Crystallogr.1953, 6, 396. Cossee, P.Thesis, Leiden, 1956.
r/
10-1 nm
Figure 3. Fourier transforms of the A1203-and Si02-attached Co species (1) species b, (2) species c, (3) species e.
transform for species b, a peak around 0.3 nm was assigned to C O C O bonding, which was confirmed by the curve-fitting analysis together with the determination of the bond distance to be 0.321 nm. This confirms the previous suggestion that the Co atoms exist in a cluster form like [COO],. The Co-0 bonding was found at 0.202 nm, which was quite shorter than the 0.213 nm of the COO bulk, indicating that the Co-0 bonds in [COO], (b) have a stronger or more covalent character than those in the COO bulk. The contraction may also be caused by the asymmetry in small clusters leading to the well-known phenomenon of surface contraction. The Co(I1) species (c) produced from species b showed no peak in the Fourier transform corresponding to a Co-Co bond as shown in part 2 of Figure 3. No Co-Co bond peak could be observed either in the EXAFS measurements of species c a t 30 K. The first peaks in part 2 of Figure 3 was attributed to C o - 0 bonding with a distance of 0.197 nm. These observations indicate that the [COO], clusters (b) under vacuum were broken into Co(I1) ensembles without a definite Co-Co bonding by heating to 5 13 K. The Co ions in the Co(I1) ensembles (c) are likely to be attached to the A1203surface, forming bonds with surface oxygen atoms ( C o - 0 distance: 0.197 i 0.003 nm; coordination number: 2.3 f 0.4) as suggested in the previous report.I3 The SO,-attached Co(I1) species (e) produced by the exposure of sample d to 0, a t room temperature followed by evacuation up to 5 13 K showed no Co-Co peak as shown in part 3 of Figure 3. O n l y one peak was observed due to a Co-0 bond, and its distance was 0.194 nm, which was nearly equal to 0.197 nm for species c. The coordination number of the C o - 0 bond was 2.1 f 0.4. Thus the Co(I1) ions in e on Si@, have a similar static local structure to those in c on AI2O3. When species b and d were oxidized with 0, at 773 K, spinel-like Co clusters (f) were obtained. Figure 4 shows the Fourier transforms of [Co304], (f) on A1203 and Si02together with that of the c030, bulk. Each peak in the Fourier transforms for species f appeared at the position corresponding to that of the Co304bulk, indicating the formation of Co304spinel clusters. Table 111 shows the coordination numbers and the interatomic distances for each shell of the A1203-and the SO2-supported [Co304],,clusters (f), obtained from the curve fitting analysis, where we adopted the empirically obtained phase shift and amplitude functions which were derived from the corresponding shells in the Co304crystal. The shorter interatomic distances for species f than for the Co304 crystal indicated that the [Co304],cluster framework is somewhat compressed at the surface. The much smaller coordination numbers as compared with those of the Co304bulk indicated the small size of the [Co3O4Inclusters on A1203and SiO,. The average size of the clusters was determined to be 1.3 nm by the transmission electron microscope (TEM) observation for the S O 2 attached [Co304],. No Co oxide species was observed by T E M in the case of the Al,O,-attached [Co304],,,suggesting a [Co304],,
4216
Asakura and lwasawa
The Journal of Physical Chemistry, Vol. 93, No. 10, 1989
TABLE Ill: EXAFS Analvses for the ICo,Oal. ( f )
co-0 sam~le fIAl20,
f/Si02 impreg" Co,04/Si02
co-co
co-co
co-Co
N
rJnm
N
r/nm
N
rJnm
N
rJnm
3.1 (0.3) 1.9 (0.3) 5.4 (0.3)
0.192 (0.002) 0.194 (0.003) 0.195 (0.002)
1.5 (0.3) 1.7 (0.3) 5.7 (0.3)
0.281 (0.003) 0.282 (0.003) 0.286 (0.002)
2.8 (0.5) 2.8 (0.5) 7.3
0.336 (0.003) 0.339 (0.003) 0.338 (0.003)
1.8 (0.7) 1.9 (0.7) IO (1.0)
0.488 (0.003) 0.491 (0.003) 0.494 (0.003)
(0.4)
Prepared from the impregnation of Si02 with Co(NO,), aqueous solution.
,
4-
3L
e,
n
5 =c
2-
.-c0,
a C
10 0
,
0'
I
I
3
2
r/
4 10-1 nm
I
5
Figure 5. Calculated and observed coordination numbers plotted against the first, second, and third Co-Co coordination shell. Curve 4 represents the coordination numbers observed for the first, second, and third Co-Co bonds in the SO2-attached [Co,04], cluster (f), and curves 1-3 represent those calculated on the basis of the shape models: (1) sphere, (2) three layers of (110) planes, (3) two layers of (1 10) planes.
r/'lO-' n m Figure 4. Fourier transforms of the attached [Co3041nand Co304 bulk: (1) Co,O,, (2) A1203-attached [Co30p],, (3) Si02-attached [Coj041n.
size below 1.O nm. It has been demonstrated that there are active oxygen atoms in the [Co304],clusters attached to Si02or A1203 and that the catalytic oxidation of CO with O2 on the [Co304], cluster proceeds by the following two-stage reaction processes: [C0304], + co [CO,O,-,], + c02 (2)
-
[C0304-xln
7202
+
[c03041,
(3)
When the observed amounts of active oxygen were fit to the stoichiometric redox equations ( 2 ) and (3), the values of n and for the species and 3 and x were established to be 5 and on Si02and A1203,respectively."-14 The fact that ( 1 ) [Co304], has two vacant sites on the coordinatively unsaturated Co atoms as indicated from the twin peaks of adsorbed NO at 1874 and 1800 cm-l, (2) the twin peaks are higher in wavenumber than the peaks of 1867 and 1782 cm-' observed with [CO"]~/AI,O~(c), suggesting the presence of Co(II1) adsorption sites in [Co304],, and that (3) there were 0.07-0.12 active oxygen atoms per Co atom has been well explained by a spinel (1 I O ) plane of a small [Co30p],c l u ~ t e r . ' ' ~The ' ~ shape of the [Co304],cluster (f) on Si02was evaluated by comparing the coordination numbers for Co-Co bondings of (f) with those of various shape models of [Co304], with a 1.3-nm particle size. The expected coordination numbers of the Co atoms were plotted when [Co304], was constituted by two or three cobalt layers of a spinel ( I I O ) plane or was present as sphere particles in Figure 5 , where also the coordination numbers determined by EXAFS for [Co304],/Si02 (f) are plotted against the first, second, and third Co shells. The data revealed that the [Co304], cluster has a raftlike structure consisting of three layers of the ( I 10) plane parallel to the surface. Since the size of the A1203-supported [Co304],was not determined by TEM, a detailed discussion of the shape of [Co304],clusters
I/
4
0
20 40 60 t / min Figure 6. CO oxidation with O2at 273 K on various cobalt catalysts with a catalyst loading of 0.287 g (5 wt % as Co/support): (1) Si0,-attached (2) Al,O,-attached [CO,~,],(f), (3) Al,O,-attached [ C O ] ~ [Co304], (0, oxide (c), (4) SiO,-attached Co oxide (e), ( 5 ) impregnation Co/AI2O, catalyst.
on A1203would not be valid. However, the fact that the coordination numbers of the surrounding Co atoms were similar to those for the Co atoms in the Si02-supported cluster, together with the fact that the size of a [Co3O4Incluster was less than 1.0 nm, suggests that the [Co304],cluster on A1203has a structure consisting of more than three layers (probably four layers) of the ( 1 10) plane. The [Co304], clusters may be attached more strongly
Structure Changes in an AI2O3/ [ C O I I ]Catalyst ~
The Journal of Physical Chemistry, Val. 93, NO. 10, 1989 4217 u
I
VI
-06 ;
, 6
I
I
I
8
I
12
10
k / 10 n m -1 Figure 8. Curve fitting for Co-Co bond in species c at 10 atm of 02:-, observed; - - -,calculated by using theoretical phase shift and amplitude
-
function.
0.330 n m
I
I
0.196
nm
i l i i l l l ~ ~ ~ o ~ i l f i l ~ IC1
14)
Reaction mechanism for CO oxidation with O2 on the A1203-attached[CO'~], cataIyst. Figure 9.
r/
10-1 nm
Figure 7. Fourier transforms of the Co species under various conditions: (1) species c under vacuum, (2) under 1 atm of O2introduced, (3) under 10 atm of O2introduced, (4) under 1 atm of CO introduced after the (5) after 30-min evacuation of the 10 atm exposure of c to IO atm of 02,
O2exposed sample. to A1,03 through surface oxygen atoms than to the SiOz surface, as reflected by the larger coordination number of the oxygen in the first shell of the [Co3O4Inon A1203. CO Oxidation Reactions. The smaller coordination numbers for Co-0 bonds in species f as compared with the averaged coordination number 5.4of the Co304 bulk indicate that coordinatively unsaturated Co atoms exist in [Co3O4In(0, probably at the cluster surface, which may generate a high activity for CO oxidation by providing CO adsorption sites with the subsequent reaction of adsorbed CO with the likewise unsaturated, active lattice oxygen as expressed in eq 2.14 As shown in Figure 6, the [Co304],species (f) supported on A1203and Si02were the most active catalysts for the oxidation of CO with O2among the various Co catalysts which oxidized all the gas-phase CO within 20 min.14 On the other hand, the impregnation c0304 catalyst, which was prepared by a usual impregnation method using C O ( N O ~ ) ~ aqueous solution, showed a negligible activity, as shown in Figure 6. (The impregnation catalyst included bulklike spinel c0304 catalyst particles as characterized with EXAFS and XRD.) The activity of the Si02-attached Co(I1) oxide (e) was also very low. In contrast to that, the AI20,-attached Co(I1) oxide (c) was found to be very active, as shown in Figure 6, though it had a similar structure to the Si02-attached Co(I1) (e) as demonstrated by EXAFS analysis. Direct Observation of Active Co Structures in Each Reaction Step by EXAFS. In order to clarify the origin (mechanism) of catalysis of the AI2O3-attached Co(I1) species (c), which has well-defined Co sites as compared with those of the [Co304],,on Al,03 or Si02 and which also has a high catalytic activity, we examined the structural change of the active Co sites in each step for the catalytic oxidations of CO with O2 by means of in situ EXAFS. Since adsorbed oxygen was found to react with CO over the catalyst (c) in the previous paper,14 the A120,-attached Co(I1) species (c) was first exposed to 02.The EXAFS of the Co(I1) species (c) measured under I atm of O2 exhibited a small new peak at 0.28 nm (the phase shift was uncorrected) in the Fourier transform as shown in Figure 7, part 2. This peak increased under
TABLE IV: Curve Fitting Results for the [CO"]~Species (c) under Various Conditions co-0 co-Co sample N rlnm N rlnm species c under vacuum 2.3 0.197 not obsd under 1 a t m of O2 2.5 0.196 0.4 0.330 under IO atm of O2 2.6 0.196 0.8 0.331 after interaction of species g 2.5 0.197 not obsd with 1 a t m of CO after 30-min evacuation of 2.4 0.196 not obsd 10 a t m O2exposed sample
"The errors in the EXAFS analyses were estimated to be 0.003 nm for the interatomic distances and 0.4 for the coordination numbers. 10 atm of 0, as shown in Figure 7, part 3. Figure 8 shows the comparison of the EXAFS oscillation obtained by inverse Fourier transformation over the range of 0.25-0.35 nm with that calculated by using the theoretical Co-Co phase shift and backscattering amplitude. From the good agreement between the filtered EXAFS oscillation and the calculated one, the new peak can be attributed to a Co-Co bonding with a bond length of 0.330 nm. This peak was lost by the introduction of CO at room temperature as shown in part 4 of Figure 7. It also disappeared by 30-min evacuation of the 02-exposed sample as shown in Figure 7 , part 5. No new peak appeared after the introduction of CO even at 10 atm. Curve fitting analyses of the EXAFS oscillation for the bond lengths and coordination numbers of the Co structures after these treatments are summarized in Table IV. The appearance of a Co-Co peak upon exposure of the A1203-attachedCo(I1) (c) to O2a t room temperature cannot be explained by the aggregation of the Co atoms into larger Co oxide particles. If this were the case, the aggregation of the Co should be accompanied by the cleavage of Co-0-AI bonds, but this is not possible solely by O2exposure at room temperature, since the Co-0-AI bonds in c were still stable at 700 K. Moreover, the disappearance of the Co-Co peak only by the 30-min evacuation of the 02-exposed sample a t room temperature also excludes the possibility of aggregation. The alternative explanation for the reversible Co-Co bond formation is as follows: It is most likely that the Co(I1) ions formed by the reaction of the [Cool4cluster (b) with surface OH groups of AI2o3l3are located adjacent to ensembles (c) as shown in Figure 9. each other, forming [CO'~],
4218
T h e Journal of Physical Chemistry, Vol. 93, No. IO, 1989
The definite Co-Co bonding of c was invisible in the EXAFS analysis, probably because of a large static or thermal disorder, which smears out the EXAFS o ~ c i l l a t i o n . 'Since ~ we could not observe the Co-Co peak in the EXAFS of the Co(I1) species (c) a t 30 K. the latter explanation may be less plausible. The 1602-'802 exchange reaction to 160180 on the Co", ensemble (c) occurred at a reaction rate comparable to that of the C O oxidation, suggesting that dissociative adsorption of 0, took place. These results demonstrate that the observation of a Co-Co peak in the presence of ambient 0, is due to the oxygen atom bridging of two adjacent Co atoms, which fixes the Co-Co separation at a definite distance, as shown in Figure 9. A similar phenomenon was observed with the Mo(V1) dimer catalyst, where no Mo-Mo peak was observed in the EXAFS Fourier transform above 200 K even though the t h o Mo atoms are located adjacently. A strong Mo-Mo peak appeared when two Mo atoms were bridged with an oxygen.12 The adsorption of oxjgen on the Co(l1) ensemble (c) to form the bridged oxygen was found to be weak: The adsorbed oxygen was completely desorbed by 30-min evacuation, as shown in the EXAFS analysis in part 5 of Figure 7 and Table IV. The further observation of EXAFS a t I O atm of O2confirms the reversible, weak adsorption of oxygen on c as shown in Figure 7, part 3, where the Co-Co peak due to the presence of adsorbed bridged oxygen atonis increased with an increase of 0, pressure. The weak bonding between Co and bridged oxygen is also suggested from little increase of the Co-0 coordination number under the high-pressure conditions in Table IV. Little increase in the C-0 coordination number may be due to the weak and longer Co-0 (bridged oxygen) bonding because longer and minor bondings are often hidden behind the stronger and shorter bondings in the EXAFS analyses, e.g. for V , 0 5 and MOO, where long V-0 (0.281 nm) and Mo-0 bondings (bridge, 0.225 and 0.223 nm) can hardly be seen in their Fourier transforms and little affect the curve-fitting analysis. Moreover, we could not find a change in the X-ray absorption near-edge structure (XANES) before and after the introduction of 0,. Since the XANES spectrum reflects the electronic structure of the Co( 11) species. the charge transfer from the Co(l1) to oxygen ma) not be large. The rapid loss of the Co-Co peak by the interaction with C O a:, shown in part 4 of Figure 7 together with the simultaneous formation of CO, demonstrates that the bridged oxygen easily rcacted with CO, structure g being converted to the original Co(I1) species (c). Consequently the catalytic oxidation of C O with 0, procecds by the repeated cqcles of the two-stage reaction steps a:, hhown in Figure 9.
Asakura and Iwasawa The high activity of the bridged oxygen toward C O oxidation can be explained by the weak bonding with Co atoms on the basis of the general rule that metal oxides that have a moderate strength of metal-oxygen bonding are most active for catalytic oxidation^.'^ The weak adsorption of oxygen may arise from the distortion of the oxygen-bridged Co dimer structure (g) where the strong Co-0-AI bonds prevent the transformation of g to a more stabilized structure with strong Co-0-Co bonds. This may be one of the most important roles of the support surface in catalysis of inorganic-oxide-attached metal catalysts. The adsorption of C O on species c was also very weak. A small C O band appeared at 2084 cm-' in the presence of 40 kPa of C O pressure, while it could not be observed under evacuated conditions. Thus the terminal C O weakly adsorbed on the Co(1I) is suggested to react with the bridged oxygen atom. The Co(I1) species (e) attached on SiO, showed no C d o peak in the Fourier transform by the introduction of O2 or C O at atmospheric and high pressures. It may be due to the larger separation of Co(I1) ions on S O 2 . The Si0,-attached Co(I1) monomers (e) were almost inactive for CO oxidation as compared with the A1203-attachedCo(I1) ensemble (c). Thus the high CO oxidation activity arises from the cooperative catalysis of the Co, ensemble. Conclusions
We have studied the structures of the highly active Co(I1) oxide catalyst (c) in the course of the C O oxidation reaction by means of an in situ EXAFS technique under atmospheric and highpressure conditions. The catalyst (c) was composed of Co, ensembles as active sites. The Co, species (c) reacted reversibly with O2to be converted to the oxygen-bridged Co dimer structure (g) with the subsequent reaction of the active bridged oxygen with weakly adsorbed CO on the coordinatively unsaturated C o ion of structure g. Consequently, the catalytic oxidation of C O with O2 readily proceeds by the two-stage mechanism as shown in Figure 9. On the other hand, the similar Si02-attached Co(I1) oxide catalyst (e) is not active because e with the larger Co-Co separation cannot have such oxygen-bridged Co dimer structures. The appropriate Co-Co distance in Co, ensembles may be an essential factor for the high cooperative catalysis.
Acknowledgment. We are grateful to Dr. M. Nomura and the staff of Photon Factory for their help in the EXAFS measurements (Proposal N o . 851 I O ) . We thank Prof. H. Kuroda and Dr. N. Kosugi for permission to use their EXAFS analysis program. Registry No. Co. 7440-48-4; CO, 630-08-0
