9966
J. Phys. Chem. 1995, 99, 9966-9973
Characterization of Cu(I1) Sites in CdSnOz Catalysts by Electron Spin Echo Envelope Modulation Spectroscopy Khalid Matar, Dongyuan Zhao, and Daniella Goldfarb" Chemical Physics Department, Weizmann Institute of Science, Rehovot 76100, Israel
Wan Azelee, Wayne Daniel, and Philip G. Harrison* Department of Chemistry, University of Nottingham, University Park, Nottingham NG7 2RD, U.K. Received: January 20, 1995; In Final Form: March 29, 1995@
The Cu(I1) sites in different preparations of tin oxide catalysts with low Cu(I1) contents were characterized by EPR spectroscopy and electron spin echo envelope modulation (ESEEM) spectroscopy. The catalysts were prepared by two methods: (a) coprecipitation of a mixed oxide gel from aqueous solutions containing both tin(1V) and copper(I1) ions and (b) by the sorption of Cu2+ cations onto tin(1V) oxide gel from aqueous solution. The samples were studied both before and after calcination. The EPR spectrum showed that from each type of preparation two major types of Cu(I1) species, termed A and B, were generated. Prior to any thermal treatment the major species in both preparations was A, whereas after calcination at 573-1073 K the major species was B. Whilst the EPR spectrum of species A showed that it is static (on the EPR time scale) both at 100 K and at ambient temperatures, species B showed dynamic effects above 100 K which we attribute to a dynamic Jahn-Teller effect. The immediate environment of the Cu(I1) was investigated in detail by following modulation from low-abundance 117.119Sn nuclei and from 'H nuclei in water and/or hydroxyl groups. In the latter we focused on the 'Hcombination harmonics generated in the two- and four-pulse ESEEM experiments. From these experiments we concluded that in species A the Cu(I1) is hydrated and situated on the external surface, coordinated either directly to a surface oxygen or via a hydrogen bond. In species B the Cu(I1) is well incorporated into the SnO2 lattice, it has very few protons in its vicinity, and some of the copper ions have an OH in their first coordination shell. This assignment was further substantiated by the inaccessibility of the Cu(I1) species in B to adsorbed ammonia. The major difference between the two preparations is the significant amount of species B in the coprecipitated material prior to calcination.
Introduction The promoting effect of Cu2+ cations on tin(1V) oxide catalysts for the oxidation of carbon monoxide by either dioxygen or nitric oxide is well documented,',2and such systems have been patented as catalysts for the control of noxious exhaust emissions from motor vehicle^.^ These materials have been prepared by two methods: (a) by the coprecipitation of a mixed oxide gel from aqueous solutions containing both tin(IV) and Cu(I1) ions and (b) by the sorption of Cu2+ cations onto tin(1V) oxide gel from aqueous solution. It might be expected, a priori, that the former would afford a catalyst in which the copper ions are distributed uniformly throughout the tin(1V) oxide particulate, whilst in the latter the copper ions would be solely at the surface of the oxide particles. In the former case, it is to be expected that thermal treatment will promote diffusion of the copper ions from the bulk lattice positions to grain boundaries. Conversely, thermal treatment of the oxide with the sorbed copper ions might lead to some incorporation of copper into the bulk lattice. Since this type of catalyst requires thermal activation in order to perform effectively, and operation is usually carried out at elevated temperatures, it is important to ascertain the behavior of the copper in the surfacelsolid state. Because of this, we have employed electron paramagnetic resonance (EPR) and electron spin echo envelope modulation (ESEEM) techniques to probe directly the environment and changes which are occurring specifically at the copper centers. @
Abstract published in Advance ACS Abstracts, May 15, 1995.
0022-365419512099-9966$09.00/0
EPR is one of the common methods for the characterization of paramagnetic cations in oxide catalysts. In orientationally disordered systems, such as metal oxide catalysts, the EPR spectrum of paramagnetic transition metal cations suffers from low resolution due to inhomogeneous broadening. Hence, superhyperfine splittings due to nearby nuclear spin, which provide information regarding the close environment of the paramagnetic ion, are not resolved. ESEEM is a well established method for measuring weak superhyperfine interactions in orientationally disordered solid^.^.^ For example, ESEEM has been successfully employed for the characterization of paramagnetic transition metal sites and complexes in zeolites and oxide In this study we have used EPR and ESEEM spectroscopies to characterize the Cu(I1) sites in CUI SnO2 catalysts obtained by both coprecipitation and sorption methods, and after controlled calcination. The state of hydration of the Cu2+ions was probed through the 'H modulation, whereas the incorporation of the Cu(I1) into the SnO2 structure was studied via the 1 7 7 . 1I9Sn modulation. We show that, although the natural abundance of 117.1'9Sn is rather low (16% total), it can be successfully used to examine the Cu(I1) environment. ESEEM Background. There are three different types of ESEEM experiments, the so-called two-, and fourpulselo experiments employing the sequences (n12-z-n-zecho), (n12- t-nI2 -T-nl2- z-echo), (n12- z-nl2 -TI2 -nTl2-nl2-z-echo), respectively. In the two-pulse ESEEM experiment the echo is measured as a function of the time interval 5 , and in the three- and four-pulse experiments it is recorded as a function of T. The two-pulse echo decay is
0 1995 American Chemical Society
Cu(I1) Sites in CdSnO2 Catalysts
J. Phys. Chem., Vol. 99, No. 24, 1995 9967 TABLE 1: Summary of the Samples Investigated and Their Corresponding EPR Parameters
a
sample
species
[Cu(H20)d2+ Cu2+(imp)Sn02
A
RII
AII,G
RI
2.4
119 2.09
2.38
122
AL, G
2.09
Cu2+(cop)-Sn02
Cu2+(cop)-Sn02(573K) Cu2'(imp)-Sn02(673-1073 K) B Cu2+(cop)-Sno2 Cu2+(cop)-Sn02(573- 1073 K) Cu'+(imp)-Sn02(773-1073 K) C Cu2+(cop)-Sn02(773- 1073 K) Cu?+(T,imp)-SnO2 D(A) Cu2+(T,imp)-Sn02(873K) E(B) F
2.47
70 2.086
44
2.43
68
2.086
44
2.38
115
2.10
2.46 2.40
65 2.086 95 2.086
basic frequency peaks due to partial cancellation of the hyperfine anisotropy. The combination harmonics in the two- and fourpulse ESEEM spectra can be used for direct determination of the anisotropic hyperfine component."-I4 The position of the sum harmonic frequency for a weak hyperfine interaction (2aiso T l v
423K
Figure 3. EPR spectra of Cu2+(cop)-Sn02(873 K) recorded at various
temperatures as indicated. similar to that of [Cu(H20)6]*+ in a 2: 1 waterlglycerol solution, as shown in the bottom trace of Figure la. The major difference is a slight decrease in gll (see Table 1). Calcination results in the appearance of a second Cu(I1) species, B, with significantly different hyperfine couplings and g values. The value of All decreased from 122 to 70 G, and the gl region becomes well resolved with an A 1 of 4 4 G. The amount of species A decreases with increasing calcination temperature, and above 773 K it cannot be detected any longer. The spectra of samples calcined at 773-1073 K reveal the existence of a third Cu(I1) species, C, with EPR parameters close to those of species B. The spectra of samples calcined at 873 and 1073 K show some contributions from other types of Cu(I1) with a broad spectral feature at g % 2.2. The EPR parameters of all species identified are listed in Table 1. The low-temperature EPR spectra of Cu2+(cop)-Sn02, prior to calcination and after calcination at 573-1073 K, are depicted in Figure Ib. The spectra of samples calcined above 673 K are essentially similar to those of the corresponding Cu*+(imp)SnO2 samples, indicating that coprecipitation from solutions containing Cu(I1) and Sn(1V) ions, followed by calcination, generates the same Cu(I1) species, B and C. The main difference between the impregnation and coprecipitation methods is the formation of small but significant amounts of the species B in uncalcined Cuzf(cop)-Sn02 and the larger amount of species B after calcination at 573 K. The relative amounts of species A and B in coprecipitated samples varied from preparation to preparation. Nonetheless, species B was always present. In both types of preparations, the transition from amorphous material to a crystalline SnOz with the rutile structure, which occurs at -873 K,I9 is associated with a significantly increased resolution in the gl region. The EPR spectra of Cu2+(T,imp)-Sn02, in which the Cuz+ ions were introduced by impregnation after calcination of the SnOr gel at 873 K, show that a multitude of Cu(I1) sites are present both in the as-prepared samples and after calcination (Figure 2 ) . The
Cu(I1) Sites in Cu/SnOz Catalysts
J. Phys. Chem., Vol. 99, No. 24, 1995 9969
1.o
Ib
"H
1.o
fresh
2800 G n
Figure 4. Two-pulse ESEEM waveforms of samples before calcination and after calcination at 573 and 873 K of (a) Cu2+(imp)-Sn02, (b) Cu2+(cop)-SnOz, and (c) Cu2+(T,imp)-SnOz, recorded at 3 150 G.
EPR parameters of the species that can be resolved, D, E, and F, are listed in Table 1. Note that the parameters of D and E are similar to those of A and B, respectively. The relative amount of species B is small compared to the other two preparations even after heating to 873 K. Whilst the spectrum of species A did not show any dynamic effects at low temperatures, the spectra of species B and C exhibit a considerable temperature dependence, as shown in Figure 3. At 100 K the hyperfine components are well resolved in both the gll and gl regions, but as the temperature increases, the lines broaden, the resolution decreases, and some shifts of peaks are observed. The latter is particularly evident at the lowfield edge of the spectrum. ESEEM Experiments. In order to get more information regarding the nature and location of the Cu(I1) species identified by the EPR spectra, we carried out ESEEM experiments. Figure 4a-c shows typical two-pulse ESEEM patterns of as-prepared and calcined Cu2+(imp)-Sn02, Cu2+(cop)-Sn02, and Cu2+(T,imp)-SnOz, recorded at an external magnetic field corresponding approximately to the gl position, where the echo intensity reaches a maximum. In all samples there is a considerable reduction in modulation depth with increasing calcination temperature. The modulation observed is mostly due to protons, and its depth increases in the following order: Cu2+(T,imp)-SnO2 > Cu2+(imp) -SnO2 > Cu2+(cop)-Sn02 in both as-prepared and calcined samples. This reflects the number of waterkydroxyl groups in the close vicinity of the Cu(II), which is always larger in Cu2+(T,imp)-Sn02. The two-pulse Fourier transform (FT) ESEEM spectra of freshly prepared Cu2+(imp)-Sn02 recorded at different magnetic fields along the EPR powder pattern are depicted in Figure Sa. These field dependent measurements are usually referred to as orientation selective experiments.*O The spectra consist of three peaks, one at 11-13.5 MHz, corresponding to the proton Larmor frequency, YH, and two sum harmonic peaks in the region of 2YH (22-27 MHz). The more intense peak, at 2% is due to protons with a weak dipolar coupling, referred to as matrix protons, whereas the peak shifted to higher frequencies is assigned to protons with a larger dipolar coupling. The weakly coupled protons are not directly coordinated to the Cu(11), whereas the more strongly coupled protons are part of the Cu(I1) ligands and are, therefore, closer to the Cu(I1). The
2700 G
,
0
.
10
,
20
.
,
30
.
I
~~
40
Frequency(MHz)
,
j
0
10
20
30
40
Frequency(MHz)
Figure 5. Two-pulse FT-ESEEM spectra of uncalcined (a) Cu2+(imp)SnO2 and (b) Cu2+(cop)-Sn02 recorded at different magnetic fields.
frequencies of all peaks decrease linearly with the magnetic field,
Ho,as expected. Furthermore, as HOapproaches the gll region, the intensity of the shifted peak decreases. Similar trends in 'the proton signals were also observed in the orientation selective two-pulse FT-ESEEM spectra of Cu2+(cop)-Sn02, shown in Figure 5b. There, however, the peaks are narrower than in Cu2+(imp)-SnO2 due to a slower echo decay (see Figure 4a,b). The spectra of Cu2+(cop)-SnOz exhibit an additional low-frequency peak at 4.4-4.8 MHz. In each of the spectra, the frequency of this peak is similar to the Larmor frequency of "'Sn and lI9Sn, which are I = 1/2 nuclei with a natural abundance of 7.7% and 8.6%, respectively. We, therefore, assign this peak to Sn nuclei in the vicinity of the Cu(I1). The difference in the Larmor frequencies of the two isotopes is very small and is not resolved in our spectra. The frequency of the IH shifted sum harmonic peak, assigned to Cu(I1) ligand protons, can be used to determine the Cu(I1)'H distance using eq 3. Due to the rather short phase memory time, the resolution of the combination peaks in the two-pulse spectra is not optimized. Better resolved combination peaks, leading to a more accurate determination of the above shift can be obtained from the four-pulse experiment. The four-phase waveform and the corresponding FT-ESEEM spectrum of fresh Cu2+(imp)-Sn02 recorded at 3150 G (t= 0.25 ps) are shown in Figure 6. In this experiment t was set to optimize the intensity of the 'H combination peaks and to suppress the corresponding fundamental frequencies. Figure 6 shows that, indeed, the sum harmonic peaks are significantly better resolved than in the two-pulse spectra, yielding a shift of 1.O MHz, from which a value of 0.26 nm for r was derived. This value is in good agreement with distance of the protons of equatorial water ligands in [C~(H20)6]*+.*~ The spectrum also shows a peak at -20 MHz and a weak corresponding harmonic at -40 MHz, which are due to an instrumental artifact. In order to substantiate the assignment of species A to hydrated Cu(II), we have compared the orientation selective
Matar et al.
9970 J. Phys. Chem., Vol. 99, No. 24, 1995
-1
k
' 2.5
0
I
I
I
I
1
I
0
10
20
30
40
50
Frequency (MHz) Figure 6. (top) Four-pulse ESEEM waveform of uncalcined Cu2+(imp)-SnO? recorded at 3150 G and 5 = 0.25 pus (only the modulation part is shown, arbitrarily normalized between 0 and -1). (bottom) The corresponding FT-ESEEM spectrum. The peaks marked with asterisks are due to instrumental artifacts. "H
~~
0
10
20
30
~
40
50
Frequency(MHz) Figure 7. Two-pulse FT-ESEEM spectra of a C U ( H ~ O )solution ~ ~ + in water/glycerol (2:1) recorded at different magnetic fields.
ESEEM spectra of the uncalcined samples with those of [Cu(H20)6I2+in a 2:1 water/glycerol solution, shown in Figure 7. The same three proton peaks appear, and the shift of the combination peak due to protons on the water ligands is also 1
MHz at 3100 G. Moreover, the relative intensity of this peak decreases as the field approaches the gll position in a similar fashion to the behavior observed in the Cu/SnO? samples. This field dependence can be explained in terms of the anisotropy of the modulation amplitudes. The latter vanishes when the magnetic field is parallel or perpendicular to the principal axis of the hyperfine interaction, which in our case can be assumed to be along the Cu-'H axis (see eq 2). The matrix protons can be well approximated as being spherically distributed around the Cu(I1) site. Therefore, the 2 v intensity ~ should not exhibit orientation dependence and as such can be used as a reference. The reduction in the relative intensity of the shifted IH combination peak as the gll position is approached is observed in both the spectra of [CU(H20)6]*' and Cu/SnO2 before calcination.21 Thus, we conclude that in the two Cu/ SnOz preparations the Cu(II) in species A has a structure similar to [Cu(H20)6I2+,although the number of water ligands may be lower than 6. A close look at the Cu/SnO2 and the [Cu(H20)6I2+ solution spectra shows that the relative intensity of the shifted peak in [CU(H20)6]*+ is somewhat larger. This may suggest that the number of water ligands in the CdSnOz samples is smaller or that not all Cu(I1) ions contributing to the echo are coordinated to water. The proton peaks in samples calcined at 573 K show the same field dependence as the as-prepared sample. Their intensities are, however, reduced, as seen from the shallower modulation depth (see Figure 4). This is consistent with the presence of species A in these samples as well. The effects of the calcination temperature on the two-pulse FT-ESEEM spectra of Cu2+(imp)-Sn02, Cu2+(cop)-Sn02, and Cu2+(T,imp)-Sn02 are shown in Figure 8a-c. In all spectra the relative intensity of the Larmor frequency peak of "7,119Sn (vsn) increases with increasing calcination temperature. Furthermore, the spectra of Cu2+(imp)-Sn02 and Cu2+(cop)-Sn02 calcined above 573 K exhibit a sum harmonics at 2vsn (Figure 8a,b). This peak appears in spectra recorded throughout the EPR powder pattem in both samples and has maximum intensity in the spectra of samples calcined at 873 K. When the modulation depth parameter, k, is small, contributions to the sum combination peaks from the presence of several nuclei can be neglected. This is particularly true for Sn modulation in which k is scaled-down by the natural abundance of 119.117Sn.22 In this case the relative intensities of the fundamental frequencies and the sum combination harmonic in the two-pulse FT-ESEEM spectrum should be 4:1, respectively (assuming overlapping vu, ~ p ) In . ~ the spectra of the samples calcined at 873 K the relative intensities of the vsn and 2vsn peaks are about 3:2; that is, the relative intensity of the 2vsn peak is significantly higher than expected. This indicates that the Cu-Sn distance is rather short, around 0.3 nm. When the dipolar interaction is large, the fundamental peaks are significantly broader than the sum harmonic peaks and are thus more susceptible to spectrometer dead time than the sum peaks. Simulations performed for one Sn nucleus at a distance of 0.36 nm showed a decrease of about 10% in the relative intensity of vsn with respect to 2vsn due to a dead time of 180 ns, whereas for a distance of 0.3 nm a decrease of about 30% was observed. We therefore conclude that considerable contributions to the vsn peak arise from I I9Sn nuclei in shells further away from the Cu(I1) in species B, whereas the 2vsn peak contains mostly contributions from 117.119Sn nuclei close to the Cu(I1). Although the vsn peak is apparent in all the spectra of Cu2+(T,imp)SnO2 calcined above 573 K, only a weak 2vsn could be detected in the spectrum recorded at low field, 2800 G, of Cu2'(T,imp)-
Cu(I1) Sites in Cu/SnO2 Catalysts
J. Phys. Chem., Vol. 99, No. 24, 1995 9971
I
a vu
b "5"
-
d\
873K 4
I
I
I
,
I
,
I
.
1
0
Frequency(MHz)
10
20
1
1
.
1
1
1
I
40
30
Frequency(M H z )
Frequency(MHz)
Figure 8. Two-pulse FT-ESEEM spectra of (a) Cu2+(imp)-Sn02, (b) Cu2+(cop)-Sn02, and (c) CuZ+(T,imp)-Sn02 as a function of calcination temperature. The ESEEM of (a) and (b) were recorded at 2900 G and that of ( c ) at 2950 G. The peaks marked with asterisks are due to instrumental artifacts.
Sn02(873 K). This indicates that the averaged Cu-Sn distance in this sample is longer than in the other two preparations. While the Sn peaks increase with calcination temperatures in both Cu2+(imp)-Sn02 and Cu2+(cop)-Sn02, the intensities of all proton peaks decrease considerably. We verified that the Sn peak indeed increases and that of the proton decreases by following the absolute intensity of the peaks prior to normalization. The spectra of samples calcined at 873 K and above also show a significant decrease in the intensity of the matrix proton peaks. Different batches of Cu2+(cop)-Sn02 showed different intensities of the matrix proton peaks. Nevertheless, they were always significantly less intense than in the uncalcined samples. These variations are attributed to different degrees of air exposures after calcination. Evacuation at 673 K further reduced the intensity of all proton signals, often leading to a complete elimination of the signals at YH and 2 Y H , as shown in Figure 9. The fundamental peak of the strongly coupled protons, which did resist the high-temperature evacuation, does not appear in these spectra because it is too broad and cannot be detected due to the spectrometer dead time. The broadening is a consequence of the large dipolar coupling. A comparison of the frequency of the 'H shifted sum combination peaks of [CU(H20)6]2+and uncalcined and calcined samples of Cu2+(imp)-Sn02 and Cu2+(cop)-Sn02 at similar fields and g values, listed in Table 2, shows that the normalized shift from 2 v ~ &VH, , i s larger in the samples calcined at 873 K. This observation indicates a shorter Cu-'H distance in the latter and further implies that it corresponds to a different type of proton, most probably a proton of an OH group rather than in a water molecule. Accordingly, we conclude that at least part of the Cu(I1) in species B and/or C has a directly coordinated OH group. This is further supported by orientation selective FT-ESEEM spectra of Cu2'(imp)-Sn02(873 K) and Cuzf(cop)-Sn02(873 K) in which the amplitude of the shifted 'H sum combination peak did not show a significant reduction
*
*Ti
b 2950 G (g=2.262)
*
I
I
I
I
I
0
10
20
30
40
,
a
1
50
Frequency(M Hz) Figure 9. Two-pulse FT-ESEEM spectra of (a) Cu2+(imp)-Sn02( 1073 K). (b) The same as (a) after evacuation at 673 K.
as the field approached the gll position. This is quite different from the dependence observed for the uncalcined samples and the [CU(H20)6]*+ solution. The behavior of Cu2+(T,imp)-Sn02 calcined at 873 K is significantly different from that of the corresponding Cu2+(imp)-Sn02(873 K) and Cu2+(cop)-Sn02(873 K). The proton
Matar et al.
9972 J. Phys. Chem., Vol. 99, No. 24, 1995
II
b
1
OH. 1
00
05
I5
10
00
05
10
15
20
-
25
H
r(i4
P P P P
1 (P8)
Figure 10. Two-pulse ESEEM waveforms of (a) Cu2+(cop)-Sn02(773 K) and Cu2+(T,imp)-Sn02(573 K) after adsorption of NH3 and (b) Cu2+(cop)-Sn02( 1073 K) and Cu2+(T,imp)-Sn02(573 K) after adsorption of ND3.
TABLE 2: Comparison of the Frequencies of the Shifted lH Sum Combination Peak in the Two-Pulse FT-ESEEM of Some of the Samples under Investigation Ho,
A,
Ha,
A.
sample G MHz A/VH G MHz A/VH 3100 1.05 0.08 3000 1.37 0.107 tC~(Hz0)61*' 3150 1.24 0.09 Cu2+(imp)-Sn02 Cu2+(cop)-Sn02 3080 1.00 0.08 Cu2+(imp)-Sn02(873 K) 3100 1.37 0.10 3000 2.1 0.169 Cu2+(cop)-Sn02(873 K) 3100 1.77 0.14 3000 2.0 0.152 modulation depth was significantly larger, and a decrease in the intensity of the shifted 'H sum combination peak was observed as the gll position was approached. This, together with the absence of the 2vsn peak, indicates that most of the Cu(I1) is located on the external surface of the oxide particles and is more accessible to water binding upon exposure to air after calcination. Furthermore, it is situated farther away from the Sn as compared to species B (or C). Experimental evidence for this conclusion was further obtained through adsorption of ammonia, as described in the next section. Adsorption of Ammonia. Samples of Cu2+(cop)-Sn02 calcined at 773 and 873 K were evacuated at 673 K and exposed to 400 Torr of ammonia at room temperature. The EPR spectra remained essentially invariant to the adsorption. Moreover, the ESEEM waveforms did not show any increase in modulation depth due to the adsorption, as shown in Figure loa. Cu2+has a high affinity for ammonia and readily coordinates the ammonia molecules when accessible. This has been shown in Cu2+exchanged zeolites A, X, and Y, where adsorption of ammonia generated the Cu(I1)-ammonia complex in the large cage of the zeolite.23 The latter was evident from clear I4N superhyperfine splittings in the EPR spectrum. The absence of strong proton modulation in Cu2+(cop)-Sn02 indicates that the Cu2+ is inaccessible to the ammonia and buried inside the Sn02 structure. We repeated this experiment with ND3 on a sample calcined at 1073 K. Due to its higher nuclear spin, the modulation induced by deuterium is deeper than that of protons, and it can therefore probe longer interaction distances.6 The resulting ESEEM is shown in Figure lob; a very shallow deuterium modulation depth is observed.
t:,,,3,,$:,,,P,,~,n,,P,CU ,,,,l,,l,,,,l,,,,,
1111/11 111111111111111
