- Email: [email protected]
Generation of supported metal clusters of controlled size
Journal of Molecular Catalysis, 20 (1983)
GENERATION SIZE
OF SUPPORTED
279
279 - 287
METAL
CLUSTERS
OF CONTROLLED
R. S. BOWLES Department
of Chemical Engineering,
Princeton University, Princeton, NJ 08544
(U.S.A.)
S. B. PARK, N. OTSUKA and R. P. ANDRESt School of Chemical Engineering,
Purdue University, West Lafayette, IN 47907
(U.S.A.)
Summary A versatile source capable of generating extremely small metal clusters and depositing these clusters on clean substrates is described. The metal clusters are grown as aerosol particles of controlled size entrained in an inert gas flow. Model cluster-support combinations have been synthesized by expansion of the aerosol through an orifice into a vacuum, separation of the clusters from metal atoms in the flow, and impingment of the clusters on to clean substrates. Silver clusters containing approximately 50 atoms (d z 10 A) have been grown and deposited on amorphous carbon films. The size distribution and the structure of these supported clusters have been studied by TEM. All TEM measurements were made after the samples were exposed to air. Even at low coverages and room temperature, cluster aggregates were observed on the carbon films. Aggregate particles smaller than approximately 40 A in diameter exhibited lattice images characteristic of bulk silver. Aggregate particles larger than approximately 40 A in diameter exhibited lattice images characteristic of Ag,O.
Introduction The scientific and technological potential of a technique for synthesizing supported metal clusters of controlled size is obvious. Perhaps the most interesting clusters for study are those whose diameters are less than cu. 20 A. It is in this size range that the electronic properties of metal clusters are likely to be a strong function both of cluster size and of cluster-support interaction. The most common technique for synthesizing model cluster-support combinations in which the clusters have diameters less than 20 A is by vacuum evaporation of metal atoms onto a cooled substrate. In particular, Hamilton
TAuthor to whom correspondence
should be addressed. @ Elsevier Sequoia/Printed in The Netherlands
280
and his coworkers have used this method to grow metal clusters containing only a few atoms [l]. Unfo~una~ly, clusters produced by this technique exhibit a broad size distribution. A second approach is to first form clusters in the gas phase and then deposit the clusters themselves on the substrate. One method of accomplishing this is simply by evaporation into a cold inert gas. Kimoto and his coworkers 12 f and Granqvist and Buhrman [3] have used this method to produce metal clusters. In order to produce a narrow size dis~ibution of small clusters with this technique, there must be rapid cooling of the condensing vapor and immediate dilution. This ideal is only approached in practice. Sattler and his coworkers [ 41 and Yokozeki and Stein [ 51 have developed similar sources optimized for small clusters. Unfortunately, these sources still yield a relatively broad size d~tribution and do not permit facile control of mean cluster size. Research at Princeton and Purdue [6 - 91 has led to the development of a technique for producing small metal clusters that permits facile control of cluster size in the range 5 -500atoms. This technique, which we have termed the Multiple Expansion Cluster Source (MECS) has to date been used to produce clusters of Cu, Ag, Au and Ni. Taytelbaum [6] has shown that the MECS technique yields a cluster size distribution which is as small as is theoretically possible with a condensation type process, and Griffin [7] has shown that mean cluster size can easily be controlled by expanding the aerosol of small metal clusters, single metal atoms and inert carrier gas from the MECS into a vacuum, thereby arresting cluster growth. Recently Bowles [9] has designed and built a Free Jet Deceleration Filter (FJDF) which strips the single metal atoms from the small clusters and permits synthesis of model cluster-support materials. This combination of MECS and FJDF techniques was used in the present study.
Experimental
MECSIFJDFuppamtus In principle, the MECS/FJDF apparatus can (a) produce a continuous stream of metal clusters with a narrow size dis~bution and controlled mean size, (b) determine the cluster size distribution and (c) deposit these clusters on various substrates. The apparatus is shown schematically in Fig. 1, where four distinct zones, i.e., oven, quench zone, condensation reactor and filter/detector, are illustrated. The function of the vaporization region or oven is to maintain set initial concentrations of metal monomer and dinner (cluster nuclei). These species are entrained in a helium carrier and expand through a sonic orifice in the oven wall. The resulting free jet flow passes into the quench region where it is mixed with room-temperature helium.
281
METAL POOL
/
OVEN
/
“=I
cD~~N
1
FILTER
AND
DETECTOR
1
Fig. 1. Schematic layout of MECS/FJDF apparatus. The majority of the flow in this apparatus is helium, introduced in the oven and the quench zone, Most of this flow is pumped by a mechanical pump attached to the condensation reactor. A small fraction is pumped by a diffusion pump attached to the main vacuum chamber.
The quench zone is critical to the operation of the MECS. Ideally, the hot gas stream from the oven mixes with cool quench gas on a time scale that is short compared to the time between monomer-dimer collisons. The flow leaving the quench zone consists of a dilute but highly supersaturated mixture of metal atoms and dimer species entrained in helium. This flow passes through a second sonic orifice separating the quench zone and the condensation reactor. The condensation reactor, except for a short entrance region, is an isothermal, constant-velocity, subsonic, fast flow reactor. Cluster growth occurs by monomer accretion onto existing nuclei (original metal dimer). It is a simple matter to freeze cluster growth by sampling from the center line of the conden~tion reactor through a third sonic orifice into a modest vacuum, thereby forming a low density free jet. Varying the residence time in the condensation reactor varies the size of the sampled clusters. As each of the four zones is separated by a sonic expansion, the pressure drops by a factor of two or more between each region. With this constraint, operation of each zone is independent. For example, the residence time in the condensation reactor may be varied independently of metal evaporation rate in the oven. The third sonic expansion which leads from the ~onden~tion reactor into the main vacuum chamber, is especially critical for the operation of the filter/detector. The supersonic jet formed by this expansion is sampled by a
282
from oven
Fig. 2. Schematic
MAIN CHAMBER layout of FJDF deceleration cell.
deceleration cell shown schematically in Fig. 2. This combination of low density free jet and specially designed deceleration cell forms what we have termed a Free Jet Deceleration Filter (FJDF) [ 91. Both metal clusters and metal monomer are accelerated to a uniform supersonic velocity by the helium carrier gas. On passing into the deceleration cell they gradually lose this initial velocity via collisions with the stopping gas in the cell. Nguyen and Andres [lo] have developed a theoretical description of this deceleration process. The distance required to decelerate a given species is directly proportional to its mass, is inversely proportional to its collision cross-section for momentum transfer and is inversely proportional to the cell pressure. Cell pressure is easily controlled by bleeding additional stopping gas into the cell. As the cell pressure is increased, first the metal atoms and then the clusters themselves no longer reach the film thickness monitor (FTM) positioned at the rear of the cell. By varying the cell pressure it is feasible both to strip the metal monomer from the flow and to determine the mean cluster size. An example of FJDF data with helium as stopping gas is shown in Fig. 3. These data are for copper; however, the size resolution for silver is even sharper. The mean stopping pressure for a 50 atom Ag cluster is 0.75 torr with helium as the stopping gas.
Preparation of supported metal clusters Clean amorphous carbon substrates less than 20 A thick were prepared by vacuum evaporation of spectroscopically pure carbon onto air cleaved NaCl crystals. The carbon film was floated on deionized water and mounted on
283
.an Elurtw rtoppinp prrrrur..
4.
*
B
SB
100
CELL
150
PRESSURE
*
200
”
.*
25-a
380
350
Fig. 3. Sample of FJDF data. Note (1) quantitative separation of metal monomer and metal clusters, and (2) attenuation of cluster flux with pressure chaxacteristic of narrow cluster size distribution.
carbon-coated microgrids having holes 0.1 - 1 pm in diameter. The substrates were mounted on a rotatable plate in the deceleration cell. The film thickness monitor was also mounted on this plate (Fig. 2). Each substrate was coated with Ag clusters of a predetermined size by the following procedure. First, the FTM was positioned to intercept the cluster flow. After the cluster size was checked by determining the mean stopping pressure of the clusters, the cell pressure was adjusted below this level but high enough to strip all the Ag atoms from the flow, and the cluster flux was measured. Then the substrate was rotated into position and exposed to the cluster flow for a predetermined time. After the substrates were exposed they were transferred in air to the TEM. Often samples were exposed to air for as much as a week before they were studied in the TEM. A JEM 200 CX transmission electron microscope was used to observe the clusters. This microscope had a resolution limit of 1.4 a under tilting beam conditions (line resolution) and of 3.2 A under axial illumination conditions (point resolution). Results and discussion Two sizes of Ag clusters, nomin~ly Ag, and Ag,, were deposited on the carbon substrates. Unfortunately, individual Ag, clusters could not be resolved by the TEM. Thus, we focus discussion primarily on the Ag,, samples. Figure 4 is a composite of four photomicrographs of a single Ag,,oncarbon sample (cu. 10 l2 clusters cmP2). The upper two micrographs were taken using bright-field imaging. The one on the left, which is of a region
284
Fig. 4. Composite micrograph of low coverage Agsc-oncarbon sample. Left micrograph are of unexposed carbon film. Right micrographs are of carbon film on which Agsc clusters have been deposited.
of substrate that was unexposed to clusters, exhibits no particles. The one on the right, which is of an exposed region, exhibits many particles. The smallest particles that can be resolved in this micrograph have diameters of 10 - 12 A as expected for Ag,,. Many particles, however, are larger. In fact, particles with diameters ranging from 10 to 100 A were observed for the sample used in Fig. 4. This result contradicts what was expected from our preparation conditions. The number and size of these aggregate particles is very sensitive to coverage and substrate variations. It is felt that they are formed by cluster migration on the carbon film. Once the substrate has been exposed to air, however, no further migration or aggregation is observed. The lower micrographs in Fig. 4 show diffraction patterns from unexposed and exposed areas of the sample. In addition to the diffuse scattering seen with the unexposed carbon film, the diffraction pattern from the exposed area exhibits a weak continuous ring, whose position corresponds to the 111 reflection of an Ag crystal. Distinct diffraction spots are also evident.
285
The strength of these spots suggests that they are from the largest particles seen in the bright field micrograph. Their position corresponds to reflections from Ag,O (cubic) crystals. One can estimate the number of AgzO particles by counting the diffraction spots and dividing by the expected number of spots per particle (cu. 2 - 3). Comparing this number with the particle size distribution observed with bright-field imaging and assuming that all of the largest particles are AgzO, we estimate that all particles with diameters greater than about 40 A are Ag,O. This conclusion is supported by photomicrographs of lattice images taken under tilting beam conditions. Because of the randomness of their orientations, few particles in a given field of view exhibited 111 lattice fringes. Nevertheless, all the large particles ( > 40 A) that showed fringes exhibited
Fig. 5. Lattice images of small particles. (a) Ag particle with perfect crystal structure. (b) Ag particle with composite structure. (c) AgzO particle with perfect crystal structure.
286
the 2.7 A spacing characteristic of AgzO and all the small particles (G4OA) that showed fringes exhibited the 2.3 A spacing characteristic of silver. Lattice images provide additional info~at~on on the internal structure of the particles. The lattice fringes from Ag,O particles were always perfect as shown in Fig. 5(c), indicating that these particles consist of single crystals. On the other hand, silver particles sometimes had nearly perfect lattice fringes (Fig. 5(a)) and sometimes had distorted lattice fringes such as would be expected for aggregates of small (-10 8) clusters with various orientations (Fig. 5(b)). The reason the large particles are oxidized while the small particles are not is not known. I-t may be due to size-dependent electronic properties of these small Ag particles or it may be due to variations in the thermal history of different size particles when exposed to air. As particles of about 10 a in diameter were the smallest that could be resolved with either bright-field or dark-field imaging, results with the Agson-carbon samples are limited. Particles ranging from 10 to 50 A were observed in bright-field micrographs. Diffraction patterns from the exposed areas exhibited both a very weak ring corresponding to the 111 reflection of Ag and occasional weak spots inside the ring. Metal deposition on each carbon substrate was controlled. Thus, a rough qualitative comparison between metal deposited and metal observed by the TEM is possible. Two results emerge. First, the large particles seen in Fig. 4 are probably twodimensional aggregates of Ag,, clusters rather than threedimensional spherical aggregates. Second, much of the metal deposited on Ags samples was not observed in the bright-field micrographs. Conclusions Small silver clusters (nomin~ly Ag, and Ag,e) were grown using an MECS/FJDF apparatus and were deposited on amorphous carbon substrates. TEM photomicrographs taken after exposure of these Ag-on-carbon samples to air showed a wide particle size distribution. This result contradicts what was expected from the preparation conditions of controlled initial cluster size and low coverage deposition. The mechanism for formation of particles larger than the initial clusters is not fully understood. TEM evidence in the case of Ag,, indicates that the large particles are likely the result of migration and ~lome~t~on of the AgsO clusters on the substrate. This conclusion is drawn from the following observations: (a) many composite particles were observed by lattice imaging; (b) a twodimensional raft model for the particles was indicated by material balance conside~tions; and (c) the particle size distribution was nearly uniform in diameter space. These experiments are preliminary. Further experiments are planned to provide more information on the Ag, samples, on possible substrate modifications to lower cluster mobility and on the thermal stability of various model cluster-support systems.
287
Acknowledgements This work was supported in part by NSF MRL Grant DMR-80-20249. The electron microscope was provided through NSF Grant DMR-78-09025.
References 1 J. F. Hamilton, D. R. Preuss and G. P. Apai, Surf. Sci., 106 (1981.) 146. Y. Kamiya, M. Nonoyarna and R. Uyeda, Jpn. J. Appl. Phys., 2 (1963) 2 K. Kimoto, 702. 3 C. G. Granqvist and R. A. Buhrman, J. Appl. Phys., 47 (1976) 2200. J. Phys. E, 13 (1980) 673. 4 K. Sattler, J. Muhlbach, E. Recknagel and A. Reyer-Flotte, 5 A. Yokozeki and G. D. Stein, J. Appl. Phys., 49 (1978) 2224. Ph. D. Thesis, Princeton University (1977). 6 M. P. Taytelbaum, University (1979). 7 G. L. Griffin, Ph. D. Thesis, Princeton 8 R. S. Bowles, J. J. Kolstad, J. M. Calo and R. P. Andres, Surf, Sci., 106 (1981) 117. University (1983). 9 R. S. Bowles, Ph. D. Thesis, Princeton 10 T. K. Nguyen and R. P. Andres, in S. S. Fisher (ed.), Rarefield Gas Dynamics, Progress in Astronautics and Aeronautics, 74: Part I., American Institute of Aeronautics and kstronautics, New York, N.Y. 1981, p. 627.
GENERATION SIZE
OF SUPPORTED
279
279 - 287
METAL
CLUSTERS
OF CONTROLLED
R. S. BOWLES Department
of Chemical Engineering,
Princeton University, Princeton, NJ 08544
(U.S.A.)
S. B. PARK, N. OTSUKA and R. P. ANDRESt School of Chemical Engineering,
Purdue University, West Lafayette, IN 47907
(U.S.A.)
Summary A versatile source capable of generating extremely small metal clusters and depositing these clusters on clean substrates is described. The metal clusters are grown as aerosol particles of controlled size entrained in an inert gas flow. Model cluster-support combinations have been synthesized by expansion of the aerosol through an orifice into a vacuum, separation of the clusters from metal atoms in the flow, and impingment of the clusters on to clean substrates. Silver clusters containing approximately 50 atoms (d z 10 A) have been grown and deposited on amorphous carbon films. The size distribution and the structure of these supported clusters have been studied by TEM. All TEM measurements were made after the samples were exposed to air. Even at low coverages and room temperature, cluster aggregates were observed on the carbon films. Aggregate particles smaller than approximately 40 A in diameter exhibited lattice images characteristic of bulk silver. Aggregate particles larger than approximately 40 A in diameter exhibited lattice images characteristic of Ag,O.
Introduction The scientific and technological potential of a technique for synthesizing supported metal clusters of controlled size is obvious. Perhaps the most interesting clusters for study are those whose diameters are less than cu. 20 A. It is in this size range that the electronic properties of metal clusters are likely to be a strong function both of cluster size and of cluster-support interaction. The most common technique for synthesizing model cluster-support combinations in which the clusters have diameters less than 20 A is by vacuum evaporation of metal atoms onto a cooled substrate. In particular, Hamilton
TAuthor to whom correspondence
should be addressed. @ Elsevier Sequoia/Printed in The Netherlands
280
and his coworkers have used this method to grow metal clusters containing only a few atoms [l]. Unfo~una~ly, clusters produced by this technique exhibit a broad size distribution. A second approach is to first form clusters in the gas phase and then deposit the clusters themselves on the substrate. One method of accomplishing this is simply by evaporation into a cold inert gas. Kimoto and his coworkers 12 f and Granqvist and Buhrman [3] have used this method to produce metal clusters. In order to produce a narrow size dis~ibution of small clusters with this technique, there must be rapid cooling of the condensing vapor and immediate dilution. This ideal is only approached in practice. Sattler and his coworkers [ 41 and Yokozeki and Stein [ 51 have developed similar sources optimized for small clusters. Unfortunately, these sources still yield a relatively broad size d~tribution and do not permit facile control of mean cluster size. Research at Princeton and Purdue [6 - 91 has led to the development of a technique for producing small metal clusters that permits facile control of cluster size in the range 5 -500atoms. This technique, which we have termed the Multiple Expansion Cluster Source (MECS) has to date been used to produce clusters of Cu, Ag, Au and Ni. Taytelbaum [6] has shown that the MECS technique yields a cluster size distribution which is as small as is theoretically possible with a condensation type process, and Griffin [7] has shown that mean cluster size can easily be controlled by expanding the aerosol of small metal clusters, single metal atoms and inert carrier gas from the MECS into a vacuum, thereby arresting cluster growth. Recently Bowles [9] has designed and built a Free Jet Deceleration Filter (FJDF) which strips the single metal atoms from the small clusters and permits synthesis of model cluster-support materials. This combination of MECS and FJDF techniques was used in the present study.
Experimental
MECSIFJDFuppamtus In principle, the MECS/FJDF apparatus can (a) produce a continuous stream of metal clusters with a narrow size dis~bution and controlled mean size, (b) determine the cluster size distribution and (c) deposit these clusters on various substrates. The apparatus is shown schematically in Fig. 1, where four distinct zones, i.e., oven, quench zone, condensation reactor and filter/detector, are illustrated. The function of the vaporization region or oven is to maintain set initial concentrations of metal monomer and dinner (cluster nuclei). These species are entrained in a helium carrier and expand through a sonic orifice in the oven wall. The resulting free jet flow passes into the quench region where it is mixed with room-temperature helium.
281
METAL POOL
/
OVEN
/
“=I
cD~~N
1
FILTER
AND
DETECTOR
1
Fig. 1. Schematic layout of MECS/FJDF apparatus. The majority of the flow in this apparatus is helium, introduced in the oven and the quench zone, Most of this flow is pumped by a mechanical pump attached to the condensation reactor. A small fraction is pumped by a diffusion pump attached to the main vacuum chamber.
The quench zone is critical to the operation of the MECS. Ideally, the hot gas stream from the oven mixes with cool quench gas on a time scale that is short compared to the time between monomer-dimer collisons. The flow leaving the quench zone consists of a dilute but highly supersaturated mixture of metal atoms and dimer species entrained in helium. This flow passes through a second sonic orifice separating the quench zone and the condensation reactor. The condensation reactor, except for a short entrance region, is an isothermal, constant-velocity, subsonic, fast flow reactor. Cluster growth occurs by monomer accretion onto existing nuclei (original metal dimer). It is a simple matter to freeze cluster growth by sampling from the center line of the conden~tion reactor through a third sonic orifice into a modest vacuum, thereby forming a low density free jet. Varying the residence time in the condensation reactor varies the size of the sampled clusters. As each of the four zones is separated by a sonic expansion, the pressure drops by a factor of two or more between each region. With this constraint, operation of each zone is independent. For example, the residence time in the condensation reactor may be varied independently of metal evaporation rate in the oven. The third sonic expansion which leads from the ~onden~tion reactor into the main vacuum chamber, is especially critical for the operation of the filter/detector. The supersonic jet formed by this expansion is sampled by a
282
from oven
Fig. 2. Schematic
MAIN CHAMBER layout of FJDF deceleration cell.
deceleration cell shown schematically in Fig. 2. This combination of low density free jet and specially designed deceleration cell forms what we have termed a Free Jet Deceleration Filter (FJDF) [ 91. Both metal clusters and metal monomer are accelerated to a uniform supersonic velocity by the helium carrier gas. On passing into the deceleration cell they gradually lose this initial velocity via collisions with the stopping gas in the cell. Nguyen and Andres [lo] have developed a theoretical description of this deceleration process. The distance required to decelerate a given species is directly proportional to its mass, is inversely proportional to its collision cross-section for momentum transfer and is inversely proportional to the cell pressure. Cell pressure is easily controlled by bleeding additional stopping gas into the cell. As the cell pressure is increased, first the metal atoms and then the clusters themselves no longer reach the film thickness monitor (FTM) positioned at the rear of the cell. By varying the cell pressure it is feasible both to strip the metal monomer from the flow and to determine the mean cluster size. An example of FJDF data with helium as stopping gas is shown in Fig. 3. These data are for copper; however, the size resolution for silver is even sharper. The mean stopping pressure for a 50 atom Ag cluster is 0.75 torr with helium as the stopping gas.
Preparation of supported metal clusters Clean amorphous carbon substrates less than 20 A thick were prepared by vacuum evaporation of spectroscopically pure carbon onto air cleaved NaCl crystals. The carbon film was floated on deionized water and mounted on
283
.an Elurtw rtoppinp prrrrur..
4.
*
B
SB
100
CELL
150
PRESSURE
*
200
”
.*
25-a
380
350
Fig. 3. Sample of FJDF data. Note (1) quantitative separation of metal monomer and metal clusters, and (2) attenuation of cluster flux with pressure chaxacteristic of narrow cluster size distribution.
carbon-coated microgrids having holes 0.1 - 1 pm in diameter. The substrates were mounted on a rotatable plate in the deceleration cell. The film thickness monitor was also mounted on this plate (Fig. 2). Each substrate was coated with Ag clusters of a predetermined size by the following procedure. First, the FTM was positioned to intercept the cluster flow. After the cluster size was checked by determining the mean stopping pressure of the clusters, the cell pressure was adjusted below this level but high enough to strip all the Ag atoms from the flow, and the cluster flux was measured. Then the substrate was rotated into position and exposed to the cluster flow for a predetermined time. After the substrates were exposed they were transferred in air to the TEM. Often samples were exposed to air for as much as a week before they were studied in the TEM. A JEM 200 CX transmission electron microscope was used to observe the clusters. This microscope had a resolution limit of 1.4 a under tilting beam conditions (line resolution) and of 3.2 A under axial illumination conditions (point resolution). Results and discussion Two sizes of Ag clusters, nomin~ly Ag, and Ag,, were deposited on the carbon substrates. Unfortunately, individual Ag, clusters could not be resolved by the TEM. Thus, we focus discussion primarily on the Ag,, samples. Figure 4 is a composite of four photomicrographs of a single Ag,,oncarbon sample (cu. 10 l2 clusters cmP2). The upper two micrographs were taken using bright-field imaging. The one on the left, which is of a region
284
Fig. 4. Composite micrograph of low coverage Agsc-oncarbon sample. Left micrograph are of unexposed carbon film. Right micrographs are of carbon film on which Agsc clusters have been deposited.
of substrate that was unexposed to clusters, exhibits no particles. The one on the right, which is of an exposed region, exhibits many particles. The smallest particles that can be resolved in this micrograph have diameters of 10 - 12 A as expected for Ag,,. Many particles, however, are larger. In fact, particles with diameters ranging from 10 to 100 A were observed for the sample used in Fig. 4. This result contradicts what was expected from our preparation conditions. The number and size of these aggregate particles is very sensitive to coverage and substrate variations. It is felt that they are formed by cluster migration on the carbon film. Once the substrate has been exposed to air, however, no further migration or aggregation is observed. The lower micrographs in Fig. 4 show diffraction patterns from unexposed and exposed areas of the sample. In addition to the diffuse scattering seen with the unexposed carbon film, the diffraction pattern from the exposed area exhibits a weak continuous ring, whose position corresponds to the 111 reflection of an Ag crystal. Distinct diffraction spots are also evident.
285
The strength of these spots suggests that they are from the largest particles seen in the bright field micrograph. Their position corresponds to reflections from Ag,O (cubic) crystals. One can estimate the number of AgzO particles by counting the diffraction spots and dividing by the expected number of spots per particle (cu. 2 - 3). Comparing this number with the particle size distribution observed with bright-field imaging and assuming that all of the largest particles are AgzO, we estimate that all particles with diameters greater than about 40 A are Ag,O. This conclusion is supported by photomicrographs of lattice images taken under tilting beam conditions. Because of the randomness of their orientations, few particles in a given field of view exhibited 111 lattice fringes. Nevertheless, all the large particles ( > 40 A) that showed fringes exhibited
Fig. 5. Lattice images of small particles. (a) Ag particle with perfect crystal structure. (b) Ag particle with composite structure. (c) AgzO particle with perfect crystal structure.
286
the 2.7 A spacing characteristic of AgzO and all the small particles (G4OA) that showed fringes exhibited the 2.3 A spacing characteristic of silver. Lattice images provide additional info~at~on on the internal structure of the particles. The lattice fringes from Ag,O particles were always perfect as shown in Fig. 5(c), indicating that these particles consist of single crystals. On the other hand, silver particles sometimes had nearly perfect lattice fringes (Fig. 5(a)) and sometimes had distorted lattice fringes such as would be expected for aggregates of small (-10 8) clusters with various orientations (Fig. 5(b)). The reason the large particles are oxidized while the small particles are not is not known. I-t may be due to size-dependent electronic properties of these small Ag particles or it may be due to variations in the thermal history of different size particles when exposed to air. As particles of about 10 a in diameter were the smallest that could be resolved with either bright-field or dark-field imaging, results with the Agson-carbon samples are limited. Particles ranging from 10 to 50 A were observed in bright-field micrographs. Diffraction patterns from the exposed areas exhibited both a very weak ring corresponding to the 111 reflection of Ag and occasional weak spots inside the ring. Metal deposition on each carbon substrate was controlled. Thus, a rough qualitative comparison between metal deposited and metal observed by the TEM is possible. Two results emerge. First, the large particles seen in Fig. 4 are probably twodimensional aggregates of Ag,, clusters rather than threedimensional spherical aggregates. Second, much of the metal deposited on Ags samples was not observed in the bright-field micrographs. Conclusions Small silver clusters (nomin~ly Ag, and Ag,e) were grown using an MECS/FJDF apparatus and were deposited on amorphous carbon substrates. TEM photomicrographs taken after exposure of these Ag-on-carbon samples to air showed a wide particle size distribution. This result contradicts what was expected from the preparation conditions of controlled initial cluster size and low coverage deposition. The mechanism for formation of particles larger than the initial clusters is not fully understood. TEM evidence in the case of Ag,, indicates that the large particles are likely the result of migration and ~lome~t~on of the AgsO clusters on the substrate. This conclusion is drawn from the following observations: (a) many composite particles were observed by lattice imaging; (b) a twodimensional raft model for the particles was indicated by material balance conside~tions; and (c) the particle size distribution was nearly uniform in diameter space. These experiments are preliminary. Further experiments are planned to provide more information on the Ag, samples, on possible substrate modifications to lower cluster mobility and on the thermal stability of various model cluster-support systems.
287
Acknowledgements This work was supported in part by NSF MRL Grant DMR-80-20249. The electron microscope was provided through NSF Grant DMR-78-09025.
References 1 J. F. Hamilton, D. R. Preuss and G. P. Apai, Surf. Sci., 106 (1981.) 146. Y. Kamiya, M. Nonoyarna and R. Uyeda, Jpn. J. Appl. Phys., 2 (1963) 2 K. Kimoto, 702. 3 C. G. Granqvist and R. A. Buhrman, J. Appl. Phys., 47 (1976) 2200. J. Phys. E, 13 (1980) 673. 4 K. Sattler, J. Muhlbach, E. Recknagel and A. Reyer-Flotte, 5 A. Yokozeki and G. D. Stein, J. Appl. Phys., 49 (1978) 2224. Ph. D. Thesis, Princeton University (1977). 6 M. P. Taytelbaum, University (1979). 7 G. L. Griffin, Ph. D. Thesis, Princeton 8 R. S. Bowles, J. J. Kolstad, J. M. Calo and R. P. Andres, Surf, Sci., 106 (1981) 117. University (1983). 9 R. S. Bowles, Ph. D. Thesis, Princeton 10 T. K. Nguyen and R. P. Andres, in S. S. Fisher (ed.), Rarefield Gas Dynamics, Progress in Astronautics and Aeronautics, 74: Part I., American Institute of Aeronautics and kstronautics, New York, N.Y. 1981, p. 627.