- Email: [email protected]
Chemisorptional behaviour of Pd small supported particles depending on size and structure: TDS, SSIMS and TEM investigation
603
Surface Science 152/153 (1985) 603-614 North-Holland, Amsterdam
CHEMISORPTIONAL BEHAVIOUR OF Pd SMALL SUPPORTED PARTICLES DEPENDING ON SIZE AND STRUCTURE: TDS, SSIMS AND TEM INVESTIGATION E. GILLET,
S. CHANNAKHONE,
V. MATOLIN
and M. GILLET
hboratoire de Microscopic et distractions Electroniques, ERA 545, Facultk des Sciences et Techniques de Saint J&&ne, F- 13397 MarseiNe Cedex, France
Received
2 April 1984; accepted
for publication
17 May 1984
We study CO chemisorption on well defined Palladium particles supported on mica. Pd particles are prepared by vapour deposition in UHV; their structure and epitaxial orientation are investigated by TEM and TED. The as deposited particles show truncated triangular pyramid shapes with two epitaxial orientations on the mica substrate. When heated in CO + 0, atmosphere they undergo important changes in shape and orientation depending on their size. We can distinguish two populations of particles with stable structure: partictes having a mean diameter < 5 nm and particles with a mean diameter z 5 nm. Both showed a semi-cube-~tahedron morphology. CO chemisorption was studied by TDS and SSIMS and we show that CO molecules first are adsorbed on edge sites in linear bonding.
1. Introduction In recent years surface scientists have tried various integrated experiments to investigate basic problems in the mechanism of catalysis. Many experiments have been performed on single-crystal surfaces [l-6]; however it is not straightfo~ard to correlate the surface results with the catalytic behaviour of an actual catalyst. In many cases catalysts have specific properties that are related to the state of the catalyst metal, mainly to its dispersion and structure. The catalytic behaviour of supported catalysts often can be correlated to size of the small metallic particles, and it is well known that crystallographical structure and morphology are dependant on the particle size. So it is necessary to perform experiments with model catalysts exhibiting particles well defined in size, structure, and morphology. In the field of thin film growth it is well known that epitaxy can produce homogeneous, high-density populations of particles having the same orientation and shape [7,8], so it is attractive to use such a method to obtain small particles to serve as model catalysts. Experiments have been performed by some workers [9-131 who studied CO oxidation on Pd and Ni particles. We have worked in the same field. Our objective is to 0039-6028/85/$03.30 0 Elsevier Science Publishers (North-Holland Physics Publishing Division)
B.V.
604
E. Gillet et al. / Chemisorptional hehouiour
of Pd
study successive steps of the catalytic oxidation of CO on Pd and to correlate these successive steps to the size, morphology, and structure of the particles. This paper concerns the study of CO adsorption on small Pd particles. vapour deposited on mica. We give results obtained by thermal desorption spectroscopy (TDS) and static secondary ions mass spectroscopy (SSIMS) and try to interpret them using results from Transmission Electron Microscopy (TEM) investigations.
2. Experimental Experiments are performed in a UHV system composed of two chambers (preparation chamber and reaction chamber) which can be separated from one another and separately pumped to base pressure in the 10~7-10-H Pa range. The preparation chamber was used to prepare, by controlled vapour deposition, the small-particle sample and was equipped with standard equipment for surface investigations (low energy electron diffraction, Auger electron spectroscopy, reflection high energy electron diffraction and secondary ion mass spectrometry). The chemisorption and catalytic reactions were carried out in the reaction chamber. A special substrate holder permitted both the transfer process between the two chambers and the programmable Joule-effect heating. The temperature and heating rate were previously calibrated by attaching a thermocouple to the mica surface. Palladium particles were vapour deposited on a freshly cleaved mica substrate. Evaporations were performed in a residual vacuum of 7 X 10e7 Pa. The mica cleavage was degassed at 300°C for one hour before vapour deposition. During Pd deposition the substrate temperature was maintained at 3OO’C. The Pd source is a Knudsen-cell, the atomic flux of which was monitored by a quartz microbalance at a value of (l-2) x 10’” atoms cm-* s- ‘. Pd vapour exposure was determined in order to achieve an average particles size in range of 2-12 nm. Densities, morphologies, and structure are investigated before and after adsorption reaction by means of a transfer carbon replica for TEM observations. TDS experiments were performed in the reaction chamber using a VG quadrupole mass spectrometer with a multiple data-acquisition system. The pressure calibration was achieved assuming an initial sticking coefficient for CO on (111) plane equal to unity (s(O) = 1). SSIMS studies were performed using a l-300 amu range quadrupole mass spectrometer with an ion energy filter and a differentially pumped ion gun. The primary ion source provided an ion beam with energy up to 5 kV and an incident angle on the target of 45”. The mica substrate electrostatic charges were neutralized by a low energy electron beam. Ion beam energy and intensity were optimized in order to minimize damages caused by ion impingement during SSIMS analysis. So we used a (5-10) X lo-’ A cm-* intensity beam with energy of 600 eV and under these analysis conditions we verified by
E. GiNet et al. / Chemisorptional behaviour of Pd
605
TEM and mass spectrometry analysis that the structural, morphological, and chemical integrity of Pd particles were maintained during the experiment.
3. TDS results Thermal CO desorption spectra were obtained for various exposures at a constant temperature, T%= 423 K, and in a pressure range of lop4 to lop6 Pa. Exposures were increased by increments of 0.1 L up to complete saturation of the particles. Figs. la and lb show typical TDS sequences for various exposures. Figs. la and lb concern particles with a mean diameter respectively smaller than 5 nm and larger than 5 nm. It can be seen that TDS spectra exhibit two peaks in the case of particles with a diameter smaller than 5 nm. Peak 1 fills first and saturates at about 3 L exposure. Its relative population decreases as the particles size increases as it is shown in fig. 2, which represents the ratio of the area under the TDS peak 1 and the total area of the TDS spectrum at saturation versus the particle size. The total CO coverages are calculated for all the spectra and plotted as a function of exposure (fig. 3) for a (111) continuous, thin, Pd film (curve a) and for three Pd-particle populations with three different sizes. The coverage is expressed in CO molecules per Pd surface atom. We assume, from the results of section 4, that particles are half-cube-octahedra. Curves b, c, and d correspond to particles with mean diameters, respectively, of 6, 4 and 2.5 nm.
i
40
60
60
70 t(s)
00
90
40
50
60
70
t (8)
60
90
Fig. 1. (a) TDS spectra of CO adsorbed on Pd particles (mean diameter < 5 nm) for various exposures. (b) TDS spectra of CO adsorbed on Pd particles (mean diameter z 6 nm) for various exposures.
606
E. G&t
et ai. / ~hemlsorpiional
behaoiour
of Pd
From the curves, a, b, c and d (fig. 3) it is possible to calculate the initial sticking coefficients for each type of Pd sample. It is obvious that these initial sticking coefficients are abnormally high. In section 5. we shall give an interpretation in agreement with the TEM investigation.
1
k-+T‘3-T. .“-lp-___ 2
4
6
mean
particles
Fig. 2. Ratio of CO amount function of particle size.
------+10
diameter
adsorbed
.5 CO
6
12
14
1 nm
in TDS peak
1 to the total CO amount
1 EXPOSURE
J
_ Torrxs
1.5 .lO’
adsorbed,
as a
2
_
Fig., 3. CO coverage 0 versus exposure for Pd(l11) film (a) and for three samples diameters of 2.5 nm (b), 4 nm (c) and 6.5 nm (d). The straight line (e) corresponds coefficient s = 1,
mean particle to a sticking
607
E. Gillet et al. / Chemisorptional behaviour of Pd
4. SSIMS
results
Benninghoven [14] has shown, in 1973, that it is possible to investigate surface reactions by a so called static secondary ion mass spectroscopy and recently this SSIMS technique was used to study CO adsorption on various surfaces [15-191. It was shown that CO adsorption on a metal (M) gives rise to M,CO+ ion species, while M,C+ species are characteristic of dissociative CO adsorption. It was also demonstrated that the ratio c Pd,CO+/Pd,+
,
n
was for some conditions a linear function of the CO coverage. More recently Brown and Vickerman [15,16] proved that the secondary ion mass spectrum contains information about the adsorption state of the CO molecule: MCO+, M2CO+, and M&O+ being characteristic of the so called “linear”, “bridge” or “ three-fold” bonded CO molecule so their relative intensities are indicative of the adsorption mode. In the present work we give results of SSIMS studies performed on two sample series with particles in two size ranges, corresponding, respectively, to diameters larger than and smaller than 5 nm. 4.1. SSIMS results for large particles (diameter Before the SSIMS investigation, sphere of 0, + CO. (0,, 7 X lo-’ particles with stable morphology.
082
Q4
> 6 nm)
the particles were annealed in an atmoPa; CO, 7 X 10e6 Pa) at 570 K to obtain
‘PJ
Fig. 4. Variations of PdCO+/(PdCO+ + Pd,CO+ coverage on large Pd particles ( > 6 nm).
) (0).
and Pd,CO+
(0)
with relative
co
60X
E. Gillei
er al. /
Chemisorptional
behuvmu
of Pd
In fig. 4 we have plotted: PdCO+/(PdCO++
Pd,CO+)
and
Pd,CO+,
as a function of coverage e/8,,, (13, is the coverage at saturation). are deduced from TDS measurements. For low coverages the ratio PdCO+/(PdCO++
Coverages
Pd,CO+)
is about 1, suggesting that only linear bonds are present. (With our quadrupole mass spectrometer we cannot record Pd,CO+ intensities but it is evident that the low number of Pd2CO+ species indicates that the Pd,CO+ signal is null). For increasing coverage the PdCO +,‘( PdCO + + Pd ,CO + ) ratio decreases indicating the Pd(lll) surface [2]. 4.2. SSIMS
that “bridged”
bonds
results for small purticles (diameter
are formed as was observed
on
< 5 nm)
Here we give (fig. 5) SSIMS results concerning an experimental run performed on particles with a mean diameter of 4-5 nm. (This diameter was
C
n
i
1
3
Fig. 5. Variation of the area S under the TDP peaks (0) and of PdCO+/(PdCO+ + Pd2CO+ ) ratio (W), PdCO+/Pd+ +Pd,CO+/Pd: (A), PdC+/Pd+ (0), PdC+/Pd+ (o), during the ten adsorption TPD cycles A (on freshly prepared sample), C and during the thermal treatment B on Pd small particles (diameter < 5 nm).
E. Gillei et al. / Chemisorptional behaviour of Pd
609
measured after the full run.) The experimental run consisted of successively: 10 adsorption-desorption cycles on a freshly deposited sample (cycle A), 1 h annealing in 0, + CO atmosphere (cycle B) and 10 adsorption-desorption cycles (cycle C). All adsorption-desorption experiments are performed with CO saturation conditions (exposure 5L); annealing is done at 570 K in 1 x lo-’ Torr of 0, and 1 X lop8 Torr of CO. In fig. 5 we plot the value of S, the total quantity of adsorbed CO measured as the area under TPD peak, and of the ratios
observed at the beginning and at the end of ech experimental run. These ratios are indicative, respectively, of the CO coverage, 0,,, of the “linear” bondings, and of the presence of C on the surface (PdC+ and PdC:). After the cycle A we observe an important decrease of S, a constant coverage 0,,, and no carbon contamination. During the cycle B, both S and 0,,,, increase and C contamination is visible. After the cycle C we observe a decrease of both S and emax and an important increase of C contamination. The CO molecules are mainly lineary bonded for freshly deposited particles, but after annealing and stabilizing treatment in 0, + CO others type of bonding, probably “bridged”, appears.
5. TEM investigations Results concerning the complete investigation of the size, orientation, and morphology of small particles as a function of their growth conditions will be published elsewhere [20]. Here we give only some characteristic results that are needed for correlation with the preceding results obtained by TDS and SSIMS. Fig. 6a shows the typical appearance of “as-deposited” Pd particles. Most of the particles have triangular outlines corresponding to top-truncated particles. These particles have a (111) plane parallel to the substrate and two epitaxial orientations as can be seen from the electron diffraction pattern (fig. 6b): (III),
//
(Oof0ni~,
and
[IIOr~l//[IOIOti,,l;
(111)Pd // wunica and C11O,,I//[112OmicaI. When heated in the CO + 0, atmosphere, these Pd particles undergo an important change in shape and orientation. This change depends on their size and we give results for particles in two size domains.
5. I. particles with u diameter < 5 rzm Fig. 7a is a typical micrograph for particles with a diameter smaller than 5 nm. Most of the particles appear as half spheres and often exhibit hexagonal outlines and facets. Fig. 7b is the corresponding electron diffraction pattern. It shows, in addition to particles with the two (111) orientations, the existence of crystailites with two (110) orientations:
Using the dark-field image technique it is possible to detect the particles corresponding to various orientations. As examples, figs. 7c and 7d are dark-field electron micrographs performed with (110) and (220) reflections corresponding, respectively, to one (110) and (111) orientation. Most of particles are monocrystalline. Some of them exhibit double contrast in the (111) dark-field image; this contrast (so-called ” butterfly”) is characteristic of icosahedral particles and that are the smallest particles. Particles with a (Ill) or (110) orientation have the same appearance and the same contrast; their expected shape is a haif-cube-octahedron with either a (111) or a (110) plane parallel to the substrate. They result from a rebuilding of the “as deposited” tetrahedra. During annealing in the D2 + CO atmosphere. this rebuilding proceeds first by a spherical shape and, later, facets appear. Although it has not experimentally proved, some (110) facets can grow on the top of oriented (110) subo-octahedra. This can be expected because epitaxial particles have a strong tendency to be truncated by a plan parallel to the substrate plane.
Fig. 6 (a) Electron micrograph corresponding to (a).
of “as-deposited”
Pd particles.
(b) Electron
diffractton
pattern
E. Gillet et al. / Chemisorptional
5.2. Particles with a diameter
behaoiour of Pd
611
> 6 nm
Generally particles with a diameter > 6 nm keep their “as-deposited” (111) orientation. Their shape is not well defined; often they appear with rounded outlines or truncated triangular plates. For these particles we do not observe rebuilding as in the case of smaller particles, but only a change in their morphology.
Fig. 7. (a) Bright field electron micrograph of Pd particles annealed in 0, +CO atmosphere (see text for experimental conditions). (b) Electron diffraction pattern corresponding to (a). (c) Dark field electron micrograph with (220) reflection. Particles marked a-e are half cube-octahedra with (111) orientation. (d) Dark field electron micrograph with (111) reflection; “butterfly” contrasts marked I is due to icosahedral particles. Particles A-D are half-cube-octahedra with (110) orientation.
6.
Discussion
6.1. Correlation between TDS and TEM results TDS spectra exhibit two peaks for small particles. It is attractive to correlate this result with the particle shape that varies with the particle size. As seen < 5 nm) are half-cube-octahedra so they above, small particles (diameter exhibit two types of surface planes, (111) and (100). Large particles are often rounded or exhibit mainly (111) planes. However, this observation does not agree with the published heat of adsorption values which are approximately the same (= 146 kJ mol-‘) for CO adsorption on Pd(ll1) and Pd(lO0) planes, though different on the Pd(ll0) plane (= 168 kJ mall’) [21]. So we can expect that (111) and (100) facets have the same adsorption behaviour. As we have pointed out it is possible that (110) facets exist on the (110) oriented small particles and taking account of the relatively high value of the binding energy of CO on Pd(ll0) adsorption on this facet can give rise to the TDS peak observed for small particles. However, these facets are smail and their contribution to the total amount of adsorbed CO must be weak. Small particles (diameter < 5 nmf are faceted and we can expect that they have well defined edges. We assume that the particles are half cubotahedrons in the range size 3-6 nm and truncated spheres for sizes larger than 6 nm, in
.a a \
l
\
.6 +c z” .4”
mean
particles
diameter
/nm
Fig. 8. N,/N, ratio versus particle size (N, = number of edge atoms, N, = total number of surface atoms) for half-cube-octahedra in (110) orientation (curve a) and for half truncated spheres in (111) orientation; particle thickness = 2/3 of sphere radius (curve b). Crosses indicate the experimental values of relative amounts of CO adsorbed in TDS peak 1 (see text).
E. Gillet et al. / Chemisorptional behaviour of Pd
613
fig. 8 we have plotted, as a function of particle size, the ratio NJN, relative to the (110) oriented half octahedron (curve a) and the (111) oriented truncated half sphere (curve b) (N,, number of edge atoms; N,, total number of surface atoms). The hatched area represents the expected ratio N,/N, taking account of both the dispersion in shape and the relative proportion of particles in various orientations as determined by TEM. In fig. 8 we have also reported some values of the experimental ratio plotted in the fig. 2. It can be seen that the values of the relative population in the high temperature TDS peak (peak 1) observed for small particles is in a rather good agreement with the N,/N, ratios. So peaks 1 and peaks 2 in the TDS spectrum can be related respectively, to edge and face adsorption sites. It must be pointed out that some workers have suggested that edge and corners sites should have special properties either in chemisorption or in catalysis [22]. From TDS results we obtained initial sticking coefficients of CO molecule on small Pd particles which are abnormally high; that is to say the impingement rate of CO molecules per surface metal atom is larger than the calculated value derived from the kinetic theory of gases. To explain this behaviour we make the assumption that CO molecules impinging on the bare mica surface can join Pd particles by diffusion before they desorb. The net diffusion equation is resolve for an isolated Pd particle. This hypothesis is supported by the TEM observations which gave particles densities in a range of 1.5 x 1015-9 X 10” rnp2 and consequently mean distances between neighbouring particles L (L - N1’2) L > 10 nm. The value of h has been estimated from the well known formula: x = ( 07,)“2, where D is the diffusion mica, with D=
(u2/4)vd
coefficient
exp(-E,/kT),
and 7a the mean residence
TV= (l/v,)
time of CO on
exp(E_JkT),
Ed and E, are the diffusion and adsorption energies, and a the distance between two adjacent adsorption sites on mica. The va and vd vibration frequencies can be considered as equal. Ed and E, are not known for CO on mica, but generally one takes for gases on non metals E, - Ed < 0.1 eV which gives X < 10 nm. So X < L and each particle can be regarded as isolated on the substrate. A complete calculation of the diffusion flux fi will be published elsewhere [23], it shows that particles with a diameter of 4 nm can receive by such a diffusion process 2 to 10 times the direct flux. Taking account off, in the exposure calculations we find the initial sticking coefficient equal to unity for various particles samples. 6.2. Correlation between SSIMS The “freshly and orientation
and TEA4 results
deposited” particles undergo important changes both in shape when they are submitted to heating treatments in presence of
CO and 0. The “as-deposited” triangular particles are unstable; they can coalesce and are reconstructed in two successive stages. First they became spherical and in a second phase they develop facets, probably due to the gas environment. These two processes can be compared to the cycles A and B (fig. 5) of the SSIMS experiments, concerning particles with a diameter < 5 nm. On one hand the decrease in the total amount of adsorbed CO (cycle A) can be correlated with the change in the total exposed Pd surface by coalescence and rebuiIding, keeping a constant coverage 8. On other hand the increase for both S and 8 during the cycle B can be related to the particles facetting. At the end of cycle C we observe carbon contamination. This phenomenon has been generally attributed to CO dissociation and has seemed to be specific to the smallest particles [11,12]. In agreement with these results we did not observe PdC+ or PdC: ions during SSIMS investigations of large particles (diameter > 6 nm).
References [l] [2] (31 [4] [5] [6] [7] [S] [9] [IO] [ll] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23)
T. Engei, J. Chem. Phys. 69 (1978) 373. H. Conrad, G. Ertl and J. Kippers, Surface Sci. 76 (1978) 323. A.M. Bradshaw and F.M. Hoffmann, Surface Sci. 72 (1978) 513. A. Ortega. F.M. Hoffmann and A.M. Bradshaw, Surface Sci. 119 (1982) 79. M.P. Kiskinova and G.M. Bliznakov, Surface Sci. 123 (1982) 61. B.E. Nieuwenhuys, Surface Sci. 126 (1983) 307. M. Gillet and A. Renou, Surface Sci. 90 (1979) 91. F. Robinson and M. Gillet, Thin Solid Films 98 (1982) 179. M. Thomas, J.T. Dickinson, H. Poppa and G.M. Pound, J. Vacuum Sci. Technol. 568. M. Thomas, H. Poppa and G.M. Pound, Thin Solid Films 58 (1979) 273. D.L. Doering, J.T. Dickinson and H. Poppa, J. Catalysis 73 (1982) 91. D.L. Doering, H. Poppa and J.T. Dickinson, J. Catalysis 73 (1982) 104. H. Giassl, R. Kramer and K. Hayek, J. Catalysis 63 (1980) 167. A. Benninghoven, Surface Sci. 35 (1973) 427. A. Brown and J.C. Vickerman, Surface Sci. 124 (1983) 267. A. Brown and J.C. Vickerman, Surface Sci. 117 (1982) 154. A. Brown and J.C. Vickerman, J: Vacuum 31 (1981) 429. R.S. Bordoli, J.C. Vickerman and J. Wolstenholme, Surface Sci. 85 (1979) 244. G.J. Slusser and N. Winograd, Surface Sci. 95 (1980) 53. M. Gillet and S. Channakhone, J. Catalysis, submitted. H. Conrad. G. Ertl, J. Koch and E. Latta, Surface Sci. 43 (1974) 462. S. Ladas, H. Poppa and M. Boudart, Surface Sci. 102 (1981) 151. E. Gillet, S. Channakhone and V. Matolin, J. Catalysis, submitted.
15 (1978)
Surface Science 152/153 (1985) 603-614 North-Holland, Amsterdam
CHEMISORPTIONAL BEHAVIOUR OF Pd SMALL SUPPORTED PARTICLES DEPENDING ON SIZE AND STRUCTURE: TDS, SSIMS AND TEM INVESTIGATION E. GILLET,
S. CHANNAKHONE,
V. MATOLIN
and M. GILLET
hboratoire de Microscopic et distractions Electroniques, ERA 545, Facultk des Sciences et Techniques de Saint J&&ne, F- 13397 MarseiNe Cedex, France
Received
2 April 1984; accepted
for publication
17 May 1984
We study CO chemisorption on well defined Palladium particles supported on mica. Pd particles are prepared by vapour deposition in UHV; their structure and epitaxial orientation are investigated by TEM and TED. The as deposited particles show truncated triangular pyramid shapes with two epitaxial orientations on the mica substrate. When heated in CO + 0, atmosphere they undergo important changes in shape and orientation depending on their size. We can distinguish two populations of particles with stable structure: partictes having a mean diameter < 5 nm and particles with a mean diameter z 5 nm. Both showed a semi-cube-~tahedron morphology. CO chemisorption was studied by TDS and SSIMS and we show that CO molecules first are adsorbed on edge sites in linear bonding.
1. Introduction In recent years surface scientists have tried various integrated experiments to investigate basic problems in the mechanism of catalysis. Many experiments have been performed on single-crystal surfaces [l-6]; however it is not straightfo~ard to correlate the surface results with the catalytic behaviour of an actual catalyst. In many cases catalysts have specific properties that are related to the state of the catalyst metal, mainly to its dispersion and structure. The catalytic behaviour of supported catalysts often can be correlated to size of the small metallic particles, and it is well known that crystallographical structure and morphology are dependant on the particle size. So it is necessary to perform experiments with model catalysts exhibiting particles well defined in size, structure, and morphology. In the field of thin film growth it is well known that epitaxy can produce homogeneous, high-density populations of particles having the same orientation and shape [7,8], so it is attractive to use such a method to obtain small particles to serve as model catalysts. Experiments have been performed by some workers [9-131 who studied CO oxidation on Pd and Ni particles. We have worked in the same field. Our objective is to 0039-6028/85/$03.30 0 Elsevier Science Publishers (North-Holland Physics Publishing Division)
B.V.
604
E. Gillet et al. / Chemisorptional hehouiour
of Pd
study successive steps of the catalytic oxidation of CO on Pd and to correlate these successive steps to the size, morphology, and structure of the particles. This paper concerns the study of CO adsorption on small Pd particles. vapour deposited on mica. We give results obtained by thermal desorption spectroscopy (TDS) and static secondary ions mass spectroscopy (SSIMS) and try to interpret them using results from Transmission Electron Microscopy (TEM) investigations.
2. Experimental Experiments are performed in a UHV system composed of two chambers (preparation chamber and reaction chamber) which can be separated from one another and separately pumped to base pressure in the 10~7-10-H Pa range. The preparation chamber was used to prepare, by controlled vapour deposition, the small-particle sample and was equipped with standard equipment for surface investigations (low energy electron diffraction, Auger electron spectroscopy, reflection high energy electron diffraction and secondary ion mass spectrometry). The chemisorption and catalytic reactions were carried out in the reaction chamber. A special substrate holder permitted both the transfer process between the two chambers and the programmable Joule-effect heating. The temperature and heating rate were previously calibrated by attaching a thermocouple to the mica surface. Palladium particles were vapour deposited on a freshly cleaved mica substrate. Evaporations were performed in a residual vacuum of 7 X 10e7 Pa. The mica cleavage was degassed at 300°C for one hour before vapour deposition. During Pd deposition the substrate temperature was maintained at 3OO’C. The Pd source is a Knudsen-cell, the atomic flux of which was monitored by a quartz microbalance at a value of (l-2) x 10’” atoms cm-* s- ‘. Pd vapour exposure was determined in order to achieve an average particles size in range of 2-12 nm. Densities, morphologies, and structure are investigated before and after adsorption reaction by means of a transfer carbon replica for TEM observations. TDS experiments were performed in the reaction chamber using a VG quadrupole mass spectrometer with a multiple data-acquisition system. The pressure calibration was achieved assuming an initial sticking coefficient for CO on (111) plane equal to unity (s(O) = 1). SSIMS studies were performed using a l-300 amu range quadrupole mass spectrometer with an ion energy filter and a differentially pumped ion gun. The primary ion source provided an ion beam with energy up to 5 kV and an incident angle on the target of 45”. The mica substrate electrostatic charges were neutralized by a low energy electron beam. Ion beam energy and intensity were optimized in order to minimize damages caused by ion impingement during SSIMS analysis. So we used a (5-10) X lo-’ A cm-* intensity beam with energy of 600 eV and under these analysis conditions we verified by
E. GiNet et al. / Chemisorptional behaviour of Pd
605
TEM and mass spectrometry analysis that the structural, morphological, and chemical integrity of Pd particles were maintained during the experiment.
3. TDS results Thermal CO desorption spectra were obtained for various exposures at a constant temperature, T%= 423 K, and in a pressure range of lop4 to lop6 Pa. Exposures were increased by increments of 0.1 L up to complete saturation of the particles. Figs. la and lb show typical TDS sequences for various exposures. Figs. la and lb concern particles with a mean diameter respectively smaller than 5 nm and larger than 5 nm. It can be seen that TDS spectra exhibit two peaks in the case of particles with a diameter smaller than 5 nm. Peak 1 fills first and saturates at about 3 L exposure. Its relative population decreases as the particles size increases as it is shown in fig. 2, which represents the ratio of the area under the TDS peak 1 and the total area of the TDS spectrum at saturation versus the particle size. The total CO coverages are calculated for all the spectra and plotted as a function of exposure (fig. 3) for a (111) continuous, thin, Pd film (curve a) and for three Pd-particle populations with three different sizes. The coverage is expressed in CO molecules per Pd surface atom. We assume, from the results of section 4, that particles are half-cube-octahedra. Curves b, c, and d correspond to particles with mean diameters, respectively, of 6, 4 and 2.5 nm.
i
40
60
60
70 t(s)
00
90
40
50
60
70
t (8)
60
90
Fig. 1. (a) TDS spectra of CO adsorbed on Pd particles (mean diameter < 5 nm) for various exposures. (b) TDS spectra of CO adsorbed on Pd particles (mean diameter z 6 nm) for various exposures.
606
E. G&t
et ai. / ~hemlsorpiional
behaoiour
of Pd
From the curves, a, b, c and d (fig. 3) it is possible to calculate the initial sticking coefficients for each type of Pd sample. It is obvious that these initial sticking coefficients are abnormally high. In section 5. we shall give an interpretation in agreement with the TEM investigation.
1
k-+T‘3-T. .“-lp-___ 2
4
6
mean
particles
Fig. 2. Ratio of CO amount function of particle size.
------+10
diameter
adsorbed
.5 CO
6
12
14
1 nm
in TDS peak
1 to the total CO amount
1 EXPOSURE
J
_ Torrxs
1.5 .lO’
adsorbed,
as a
2
_
Fig., 3. CO coverage 0 versus exposure for Pd(l11) film (a) and for three samples diameters of 2.5 nm (b), 4 nm (c) and 6.5 nm (d). The straight line (e) corresponds coefficient s = 1,
mean particle to a sticking
607
E. Gillet et al. / Chemisorptional behaviour of Pd
4. SSIMS
results
Benninghoven [14] has shown, in 1973, that it is possible to investigate surface reactions by a so called static secondary ion mass spectroscopy and recently this SSIMS technique was used to study CO adsorption on various surfaces [15-191. It was shown that CO adsorption on a metal (M) gives rise to M,CO+ ion species, while M,C+ species are characteristic of dissociative CO adsorption. It was also demonstrated that the ratio c Pd,CO+/Pd,+
,
n
was for some conditions a linear function of the CO coverage. More recently Brown and Vickerman [15,16] proved that the secondary ion mass spectrum contains information about the adsorption state of the CO molecule: MCO+, M2CO+, and M&O+ being characteristic of the so called “linear”, “bridge” or “ three-fold” bonded CO molecule so their relative intensities are indicative of the adsorption mode. In the present work we give results of SSIMS studies performed on two sample series with particles in two size ranges, corresponding, respectively, to diameters larger than and smaller than 5 nm. 4.1. SSIMS results for large particles (diameter Before the SSIMS investigation, sphere of 0, + CO. (0,, 7 X lo-’ particles with stable morphology.
082
Q4
> 6 nm)
the particles were annealed in an atmoPa; CO, 7 X 10e6 Pa) at 570 K to obtain
‘PJ
Fig. 4. Variations of PdCO+/(PdCO+ + Pd,CO+ coverage on large Pd particles ( > 6 nm).
) (0).
and Pd,CO+
(0)
with relative
co
60X
E. Gillei
er al. /
Chemisorptional
behuvmu
of Pd
In fig. 4 we have plotted: PdCO+/(PdCO++
Pd,CO+)
and
Pd,CO+,
as a function of coverage e/8,,, (13, is the coverage at saturation). are deduced from TDS measurements. For low coverages the ratio PdCO+/(PdCO++
Coverages
Pd,CO+)
is about 1, suggesting that only linear bonds are present. (With our quadrupole mass spectrometer we cannot record Pd,CO+ intensities but it is evident that the low number of Pd2CO+ species indicates that the Pd,CO+ signal is null). For increasing coverage the PdCO +,‘( PdCO + + Pd ,CO + ) ratio decreases indicating the Pd(lll) surface [2]. 4.2. SSIMS
that “bridged”
bonds
results for small purticles (diameter
are formed as was observed
on
< 5 nm)
Here we give (fig. 5) SSIMS results concerning an experimental run performed on particles with a mean diameter of 4-5 nm. (This diameter was
C
n
i
1
3
Fig. 5. Variation of the area S under the TDP peaks (0) and of PdCO+/(PdCO+ + Pd2CO+ ) ratio (W), PdCO+/Pd+ +Pd,CO+/Pd: (A), PdC+/Pd+ (0), PdC+/Pd+ (o), during the ten adsorption TPD cycles A (on freshly prepared sample), C and during the thermal treatment B on Pd small particles (diameter < 5 nm).
E. Gillei et al. / Chemisorptional behaviour of Pd
609
measured after the full run.) The experimental run consisted of successively: 10 adsorption-desorption cycles on a freshly deposited sample (cycle A), 1 h annealing in 0, + CO atmosphere (cycle B) and 10 adsorption-desorption cycles (cycle C). All adsorption-desorption experiments are performed with CO saturation conditions (exposure 5L); annealing is done at 570 K in 1 x lo-’ Torr of 0, and 1 X lop8 Torr of CO. In fig. 5 we plot the value of S, the total quantity of adsorbed CO measured as the area under TPD peak, and of the ratios
observed at the beginning and at the end of ech experimental run. These ratios are indicative, respectively, of the CO coverage, 0,,, of the “linear” bondings, and of the presence of C on the surface (PdC+ and PdC:). After the cycle A we observe an important decrease of S, a constant coverage 0,,, and no carbon contamination. During the cycle B, both S and 0,,,, increase and C contamination is visible. After the cycle C we observe a decrease of both S and emax and an important increase of C contamination. The CO molecules are mainly lineary bonded for freshly deposited particles, but after annealing and stabilizing treatment in 0, + CO others type of bonding, probably “bridged”, appears.
5. TEM investigations Results concerning the complete investigation of the size, orientation, and morphology of small particles as a function of their growth conditions will be published elsewhere [20]. Here we give only some characteristic results that are needed for correlation with the preceding results obtained by TDS and SSIMS. Fig. 6a shows the typical appearance of “as-deposited” Pd particles. Most of the particles have triangular outlines corresponding to top-truncated particles. These particles have a (111) plane parallel to the substrate and two epitaxial orientations as can be seen from the electron diffraction pattern (fig. 6b): (III),
//
(Oof0ni~,
and
[IIOr~l//[IOIOti,,l;
(111)Pd // wunica and C11O,,I//[112OmicaI. When heated in the CO + 0, atmosphere, these Pd particles undergo an important change in shape and orientation. This change depends on their size and we give results for particles in two size domains.
5. I. particles with u diameter < 5 rzm Fig. 7a is a typical micrograph for particles with a diameter smaller than 5 nm. Most of the particles appear as half spheres and often exhibit hexagonal outlines and facets. Fig. 7b is the corresponding electron diffraction pattern. It shows, in addition to particles with the two (111) orientations, the existence of crystailites with two (110) orientations:
Using the dark-field image technique it is possible to detect the particles corresponding to various orientations. As examples, figs. 7c and 7d are dark-field electron micrographs performed with (110) and (220) reflections corresponding, respectively, to one (110) and (111) orientation. Most of particles are monocrystalline. Some of them exhibit double contrast in the (111) dark-field image; this contrast (so-called ” butterfly”) is characteristic of icosahedral particles and that are the smallest particles. Particles with a (Ill) or (110) orientation have the same appearance and the same contrast; their expected shape is a haif-cube-octahedron with either a (111) or a (110) plane parallel to the substrate. They result from a rebuilding of the “as deposited” tetrahedra. During annealing in the D2 + CO atmosphere. this rebuilding proceeds first by a spherical shape and, later, facets appear. Although it has not experimentally proved, some (110) facets can grow on the top of oriented (110) subo-octahedra. This can be expected because epitaxial particles have a strong tendency to be truncated by a plan parallel to the substrate plane.
Fig. 6 (a) Electron micrograph corresponding to (a).
of “as-deposited”
Pd particles.
(b) Electron
diffractton
pattern
E. Gillet et al. / Chemisorptional
5.2. Particles with a diameter
behaoiour of Pd
611
> 6 nm
Generally particles with a diameter > 6 nm keep their “as-deposited” (111) orientation. Their shape is not well defined; often they appear with rounded outlines or truncated triangular plates. For these particles we do not observe rebuilding as in the case of smaller particles, but only a change in their morphology.
Fig. 7. (a) Bright field electron micrograph of Pd particles annealed in 0, +CO atmosphere (see text for experimental conditions). (b) Electron diffraction pattern corresponding to (a). (c) Dark field electron micrograph with (220) reflection. Particles marked a-e are half cube-octahedra with (111) orientation. (d) Dark field electron micrograph with (111) reflection; “butterfly” contrasts marked I is due to icosahedral particles. Particles A-D are half-cube-octahedra with (110) orientation.
6.
Discussion
6.1. Correlation between TDS and TEM results TDS spectra exhibit two peaks for small particles. It is attractive to correlate this result with the particle shape that varies with the particle size. As seen < 5 nm) are half-cube-octahedra so they above, small particles (diameter exhibit two types of surface planes, (111) and (100). Large particles are often rounded or exhibit mainly (111) planes. However, this observation does not agree with the published heat of adsorption values which are approximately the same (= 146 kJ mol-‘) for CO adsorption on Pd(ll1) and Pd(lO0) planes, though different on the Pd(ll0) plane (= 168 kJ mall’) [21]. So we can expect that (111) and (100) facets have the same adsorption behaviour. As we have pointed out it is possible that (110) facets exist on the (110) oriented small particles and taking account of the relatively high value of the binding energy of CO on Pd(ll0) adsorption on this facet can give rise to the TDS peak observed for small particles. However, these facets are smail and their contribution to the total amount of adsorbed CO must be weak. Small particles (diameter < 5 nmf are faceted and we can expect that they have well defined edges. We assume that the particles are half cubotahedrons in the range size 3-6 nm and truncated spheres for sizes larger than 6 nm, in
.a a \
l
\
.6 +c z” .4”
mean
particles
diameter
/nm
Fig. 8. N,/N, ratio versus particle size (N, = number of edge atoms, N, = total number of surface atoms) for half-cube-octahedra in (110) orientation (curve a) and for half truncated spheres in (111) orientation; particle thickness = 2/3 of sphere radius (curve b). Crosses indicate the experimental values of relative amounts of CO adsorbed in TDS peak 1 (see text).
E. Gillet et al. / Chemisorptional behaviour of Pd
613
fig. 8 we have plotted, as a function of particle size, the ratio NJN, relative to the (110) oriented half octahedron (curve a) and the (111) oriented truncated half sphere (curve b) (N,, number of edge atoms; N,, total number of surface atoms). The hatched area represents the expected ratio N,/N, taking account of both the dispersion in shape and the relative proportion of particles in various orientations as determined by TEM. In fig. 8 we have also reported some values of the experimental ratio plotted in the fig. 2. It can be seen that the values of the relative population in the high temperature TDS peak (peak 1) observed for small particles is in a rather good agreement with the N,/N, ratios. So peaks 1 and peaks 2 in the TDS spectrum can be related respectively, to edge and face adsorption sites. It must be pointed out that some workers have suggested that edge and corners sites should have special properties either in chemisorption or in catalysis [22]. From TDS results we obtained initial sticking coefficients of CO molecule on small Pd particles which are abnormally high; that is to say the impingement rate of CO molecules per surface metal atom is larger than the calculated value derived from the kinetic theory of gases. To explain this behaviour we make the assumption that CO molecules impinging on the bare mica surface can join Pd particles by diffusion before they desorb. The net diffusion equation is resolve for an isolated Pd particle. This hypothesis is supported by the TEM observations which gave particles densities in a range of 1.5 x 1015-9 X 10” rnp2 and consequently mean distances between neighbouring particles L (L - N1’2) L > 10 nm. The value of h has been estimated from the well known formula: x = ( 07,)“2, where D is the diffusion mica, with D=
(u2/4)vd
coefficient
exp(-E,/kT),
and 7a the mean residence
TV= (l/v,)
time of CO on
exp(E_JkT),
Ed and E, are the diffusion and adsorption energies, and a the distance between two adjacent adsorption sites on mica. The va and vd vibration frequencies can be considered as equal. Ed and E, are not known for CO on mica, but generally one takes for gases on non metals E, - Ed < 0.1 eV which gives X < 10 nm. So X < L and each particle can be regarded as isolated on the substrate. A complete calculation of the diffusion flux fi will be published elsewhere [23], it shows that particles with a diameter of 4 nm can receive by such a diffusion process 2 to 10 times the direct flux. Taking account off, in the exposure calculations we find the initial sticking coefficient equal to unity for various particles samples. 6.2. Correlation between SSIMS The “freshly and orientation
and TEA4 results
deposited” particles undergo important changes both in shape when they are submitted to heating treatments in presence of
CO and 0. The “as-deposited” triangular particles are unstable; they can coalesce and are reconstructed in two successive stages. First they became spherical and in a second phase they develop facets, probably due to the gas environment. These two processes can be compared to the cycles A and B (fig. 5) of the SSIMS experiments, concerning particles with a diameter < 5 nm. On one hand the decrease in the total amount of adsorbed CO (cycle A) can be correlated with the change in the total exposed Pd surface by coalescence and rebuiIding, keeping a constant coverage 8. On other hand the increase for both S and 8 during the cycle B can be related to the particles facetting. At the end of cycle C we observe carbon contamination. This phenomenon has been generally attributed to CO dissociation and has seemed to be specific to the smallest particles [11,12]. In agreement with these results we did not observe PdC+ or PdC: ions during SSIMS investigations of large particles (diameter > 6 nm).
References [l] [2] (31 [4] [5] [6] [7] [S] [9] [IO] [ll] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23)
T. Engei, J. Chem. Phys. 69 (1978) 373. H. Conrad, G. Ertl and J. Kippers, Surface Sci. 76 (1978) 323. A.M. Bradshaw and F.M. Hoffmann, Surface Sci. 72 (1978) 513. A. Ortega. F.M. Hoffmann and A.M. Bradshaw, Surface Sci. 119 (1982) 79. M.P. Kiskinova and G.M. Bliznakov, Surface Sci. 123 (1982) 61. B.E. Nieuwenhuys, Surface Sci. 126 (1983) 307. M. Gillet and A. Renou, Surface Sci. 90 (1979) 91. F. Robinson and M. Gillet, Thin Solid Films 98 (1982) 179. M. Thomas, J.T. Dickinson, H. Poppa and G.M. Pound, J. Vacuum Sci. Technol. 568. M. Thomas, H. Poppa and G.M. Pound, Thin Solid Films 58 (1979) 273. D.L. Doering, J.T. Dickinson and H. Poppa, J. Catalysis 73 (1982) 91. D.L. Doering, H. Poppa and J.T. Dickinson, J. Catalysis 73 (1982) 104. H. Giassl, R. Kramer and K. Hayek, J. Catalysis 63 (1980) 167. A. Benninghoven, Surface Sci. 35 (1973) 427. A. Brown and J.C. Vickerman, Surface Sci. 124 (1983) 267. A. Brown and J.C. Vickerman, Surface Sci. 117 (1982) 154. A. Brown and J.C. Vickerman, J: Vacuum 31 (1981) 429. R.S. Bordoli, J.C. Vickerman and J. Wolstenholme, Surface Sci. 85 (1979) 244. G.J. Slusser and N. Winograd, Surface Sci. 95 (1980) 53. M. Gillet and S. Channakhone, J. Catalysis, submitted. H. Conrad. G. Ertl, J. Koch and E. Latta, Surface Sci. 43 (1974) 462. S. Ladas, H. Poppa and M. Boudart, Surface Sci. 102 (1981) 151. E. Gillet, S. Channakhone and V. Matolin, J. Catalysis, submitted.
15 (1978)