doi: / (

doi: 10.1063/1.369476(http://dx.doi.org/10.1063/1.369476)

JOURNAL OF APPLIED PHYSICS VOLUME 85, NUMBER 1 1 JANUARY 1999 Monodispersed Cr cluster formation by plasma-gas-condensation S. Yamamuro, a) K. Sumiyama, and K. Suzuki Institute for Materials Research, Tohoku University, Sendai 980-8577, Japan Received 7 July 1998; accepted for publication 28 September 1998 Nanometer-sized Cr clusters in the size range of 7.6 13 nm have been produced by a plasma-gas-condensation-type cluster deposition apparatus, which combines a grow-discharge sputtering technique with an inert gas condensation technique. We have studied influences of the Ar gas pressure, P Ar, and the Ar gas flow rate, V Ar, on the size distribution of Cr clusters by transmission electron microscopy. Monodispersed Cr clusters are formed at both low P Ar and low V Ar. At low P Ar, Cr clusters nucleate and grow only in the liquid-nitrogen-cooled growth region, and the deposition rate is rather low. At high P Ar, on the other hand, a large amount of Cr clusters are formed even near the sputtering source, and the nucleation and growth occur over a wide region between the sputtering source and the growth region. Under this condition, the deposition rate is relatively high. Consequently, the formation mechanism of the monodispersed clusters is similar to that of monodispersed colloidal particles: The nucleation and growth processes are definitely separated and the coagulation of growing particles is prohibited. In the present experiments, these conditions are effectively attained by using a carrier gas flow and liquid-nitrogen-cooling of the cluster growth region. 1999 American Institute of Physics. S0021-8979 99 03701-9 I. INTRODUCTION a CREST Fellow; electronic mail: yamamuro@snap8.imr.tohoku.ac.jp Nanostructured materials are topical in both fundamental and applied sciences, because their nanometer-scale compositional and structural heterogeneities give rise to several unique properties in contrast to homogeneous bulk materials. Granular solids 1 and nanocrystalline materials, 2 for example, exhibit fascinating magnetic, transport, and mechanical properties. These materials have been usually produced by precipitation from supersaturated solid solution or crystallization of amorphous alloys via annealing, or by consolidation of fine particles produced by an inert gas condensation. In these methods, however, it is difficult to control the nanostructural parameters precisely: For instance, the particle size and its dispersion cannot be changed independently in the granular solids. Assembling of free nanometer-sized clusters as building blocks is a good alternative process for constructing the nanostructured functional materials. 3 For this purpose, it is crucial to produce monodispersed clusters. There have been several reports on the formation of monodispersed metal, polymer, and composite particles from submicro to micrometer in size by using a colloidal method, 4,5 where surfactants are used for suppressing coagulation and coalescence growth. Controlling the concentration of dissolved spices and the temperature during chemical reaction is also effective to produce the monodispersed particles. In the conventional gas condensation method, however, there have been only a few reports on the monodispersed cluster formation, 6 because the particle formation process is rather complex. Recently, we have constructed an intensive and sizecontrollable cluster deposition system, i.e., a plasma-gascondensation-type PGC cluster deposition apparatus 7,8 which is similar to the one originally developed by Haberland et al. 9 The PGC method combines a grow-discharge sputtering technique with an inert gas condensation technique: Metal vapor is generated by sputtering at about 150 Pa pressure and condenses into clusters in a carrier gas flow. In contrast to conventional thermal evaporation sources, this method is advantageous to generating clusters consisting of refractive or low-vapor-pressure metals because any kind of metals can be vaporized by sputtering. In the present report, we demonstrate monodispersed Cr cluster formation using the PGC method, and elucidate its process by studying the influences of process parameters the Ar gas pressure and the Ar gas flow rate on the formed clusters with transmission electron microscopy TEM. II. EXPERIMENT Figure 1 shows a schematic drawing of the PGC apparatus. It mainly consists of three components: a sputtering chamber, a growth region, and a deposition chamber. The growth region was cooled with liquid nitrogen. Comparing with a previous apparatus, 7 we put an additional differentialpumping chamber prior to the deposition chamber to keep the deposition chamber in a better vacuum condition. A facing-target-type direct current dc sputtering source was used for vaporizing Cr targets, whose sizes are 70 mm in diameter. The distance between the two targets is 100 mm. The sputtering source, which is a combination of a hollowcathode discharge mode and a magnetron mode, was operated at high Ar gas pressure, P Ar, of 70 360 Pa. The magnetic field of about 20 mt was applied to the space between the two targets to get a high ionization rate of Ar gas and a high sputtering rate. A large amount of Ar gas the Ar gas flow rate, V Ar 1.7 10 6 2.0 10 5 m 3 /s was injected 0021-8979/99/85(1)/483/7/$15.00 483 1999 American Institute of Physics

484 J. Appl. Phys., Vol. 85, No. 1, 1 January 1999 Yamamuro, Sumiyama, and Suzuki FIG. 1. Schematic drawing of PGC-type cluster deposition apparatus. TMP, MBP, and CMP represent turbomolecular pump, mechanical booster pump, and compound molecular pump, respectively. continuously into the sputtering chamber from a gas inlet, and effectively evacuated by a mechanical booster pump MBP through a small nozzle whose typical diameter was 3.5 mm. Clusters were formed in a carrier gas flow, and ejected through the small nozzle and two skimmers by differential pumping, and then deposited onto a substrate fixed on a sample holder in the deposition chamber ( 1 10 2 Pa). The substrate temperature was at about 300 K during the deposition. If the nozzle diameter of the growth region is fixed, P Ar depends on V Ar. If the nozzle diameter is varied adequately, on the other hand, we can control P Ar and V Ar independently. The input power for sputtering was 300 W; the typical current and voltage values were 1 A and 300 V, depending strongly on P Ar. The microgrid, which consists of carbon-coated colodion film supported by a Cu grid, was used as a substrate for TEM observation. The Cr clusters were usually deposited in the deposition chamber with the weight thickness of 2 nm, which was measured by a quartz oscillator-type thickness monitor. To clarify the cluster formation region, we also deposited for 5 min at three positions: the sputtering chamber, and the entrance and the exit of the growth region. We observed cluster images using a 200 kv Hitachi HF-2000 TEM, and stored them as digital data with a slow scan charge coupled device camera installed in TEM. Using an image-analysis software Image-Pro PLUS: Media Cybernetics, we measured the cluster size distributions from these digitized images in the area of 350 350 nm 2. III. RESULTS In the inert gas condensation under the carrier gas flow, it has been pointed out that the cluster size strongly depends on V Ar or P Ar. 10 Thus, we first examined the influence of V Ar on the cluster size distribution. Figures 2 and 3 show the bright-field TEM images and the size distributions of Cr clusters produced at several V Ar values of 3.3 10 6 1.67 10 5 m 3 /s. In these experiments, P Ar also increases from 120 to 360 Pa with increasing V Ar from 3.3 10 6 to 1.67 10 5 m 3 /s. For the low V Ar of 3.3 10 6 6.7 10 6 FIG. 2. Bright-field TEM images of Cr clusters produced at several Ar gas flow rates, V Ar : a 3.3 10 6 m 3 /s, b 6.7 10 6 m 3 /s, c 1.0 10 5 m 3 /s, d 1.33 10 5 m 3 /s, and e 1.67 10 5 m 3 /s. The Ar gas pressure, P Ar, was also varied with V Ar. m 3 /s, uniform-sized Cr clusters were obtained. However, the size distribution suddenly becomes broad at V Ar 1.0 10 5 m 3 /s and broader with further increasing V Ar. At V Ar 1.0 10 5 m 3 /s, in particular, there are bimodal peaks locating at about 5 and 20 nm or more. These size distributions indicate that the clusters corresponding to each peaks are formed through different processes, because the cluster size distribution is crucially affected by its formation process. The mean cluster diameter and the standard deviation estimated from Figs. 3 a to 3 e and the other two samples are shown in Fig. 4 a. With increasing V Ar, the mean cluster diameter initially increases from 7.6 nm at V Ar 3.3 10 6 m 3 /s, and shows a maximum value of 13 nm at around V Ar 1.0 10 5 m 3 /s. The standard deviation shows a constant value of about 0.8 nm for V Ar 3.3 10 6 6.7 10 6 m 3 /s. This value is less than 10% of the mean cluster

J. Appl. Phys., Vol. 85, No. 1, 1 January 1999 Yamamuro, Sumiyama, and Suzuki 485 FIG. 4. a Mean cluster diameter and standard deviation estimated from Figs. 3 a to 3 e and other two samples, and b typical deposition rate, as a function of the Ar gas flow rate, V Ar. 10 5 m 3 /s with varying P Ar from 120 to 280 Pa. The monodispersed Cr clusters were prepared at both P Ar 120 and 170 Pa, whereas the size distribution becomes markedly broad at P Ar 280 Pa. The mean cluster diameter and the standard deviation evaluated from Figs. 6 a 6 c are shown in Fig. 7 a. With increasing P Ar from 120 to 280 Pa, the FIG. 3. Size distributions of Cr clusters estimated from Figs. 2 a to 2 e. diameter, showing a good monodispersivity. At V Ar 1.0 10 5 m 3 /s, however, the standard deviation rapidly increases with V Ar. Similar V Ar dependence of the mean cluster diameter and the standard deviation has been obtained in the previous experiments performed with a diode dc sputtering source. 7 Therefore, these features are characteristic for the PGC apparatus. Figure 4 b shows a typical deposition rate as a function of V Ar. The deposition rate shows a maximum value at around V Ar 1.3 10 5 m 3 /s. At V Ar 3.3 10 6 m 3 /s, the deposition rate is almost negligible, indicating that clusters are hardly formed under such low V Ar conditions. It has not been clear which process parameter affects the size distribution in the above mentioned results, because P Ar and V Ar varied simultaneously. Thus, we controlled them independently by varying the nozzle diameter of the growth region, and examined their effects on the cluster size distribution. Figures 5 and 6 show the TEM images and the size distributions of Cr clusters deposited at constant V Ar 1.0 FIG. 5. Bright-field TEM images of Cr clusters produced by varying the Ar gas pressure, P Ar : a 120 Pa, b 170 Pa, and c 280 Pa. The Ar gas flow rate, V Ar, was constant at 1.0 10 5 m 3 /s.

486 J. Appl. Phys., Vol. 85, No. 1, 1 January 1999 Yamamuro, Sumiyama, and Suzuki FIG. 6. Size distributions of Cr clusters estimated from Figs. 5 a to 5 c. mean cluster diameter increases from 8.6 to 13.2 nm and the standard deviation increases from 0.7 to 5.8 nm, indicating that small and monodispersed clusters can be formed at low P Ar. Figure 7 b shows a typical deposition rate as a function of P Ar, indicating that the deposition rate increases monotonically with P Ar. Thus, it can be said that the cluster FIG. 8. Bright-field TEM images of Cr clusters produced by varying the Ar gas flow rate, V Ar : a 5.0 10 6 m 3 /s, b 1.0 10 5 m 3 /s, c 1.42 10 5 m 3 /s, and d 2.0 10 5 m 3 /s. The Ar gas pressure, P Ar, was constant at 170 Pa. FIG. 7. a Mean cluster diameter and standard deviation estimated from Figs. 6 a to 6 c, and b typical deposition rate, as a function of the Ar gas pressure, P Ar, at constant Ar gas flow rate, V Ar, of 1.0 10 5 m 3 /s. formation is promoted at higher P Ar as reported in the conventional gas evaporation method. 11,12 Figures 8 and 9 show the TEM images and the size distributions of Cr clusters deposited at constant P Ar 170 Pa with varying V Ar from 5.0 10 6 to 2.0 10 5 m 3 /s. The monodispersed Cr clusters were formed at low V Ar ( 1.0 10 5 m 3 /s), whereas the cluster size distribution becomes broad at high V Ar ( 1.42 10 5 m 3 /s). The mean cluster diameter and the standard deviation estimated from Figs. 9 a to 9 d are shown in Fig. 10 a. AsV Ar increases from 5.0 10 6 to 2.0 10 5 m 3 /s, the mean cluster diameter does not vary much at around 10 nm, while the standard deviation increases monotonically from 0.54 to 3.5 nm. As shown in Fig. 10 b, the deposition rate reveals no simple V Ar dependence, however, it obviously decreases at low V Ar. For understanding the cluster formation process, it is necessary to clarify where the clusters nucleate and grow. Thus, we deposited the growing clusters onto the TEM microgids in the sputtering chamber and the growth region, and then observed them by TEM. Figure 11 shows the TEM images of the samples deposited at different positions and V Ar values. In these experiments, P Ar also increases from 70

J. Appl. Phys., Vol. 85, No. 1, 1 January 1999 Yamamuro, Sumiyama, and Suzuki 487 FIG. 11. Bright-field TEM images of Cr clusters deposited at different Ar gas flow rates, V Ar, of 1.7 10 6 1.0 10 5 m 3 /s and different positions. The samples were deposited near the sputtering target, at the entrance to the growth region, and the exit from the growth region. The Ar gas pressure, P Ar, was also varied with V Ar. FIG. 9. Size distributions of Cr clusters estimated from Figs. 8 a to 8 d. FIG. 10. a Mean cluster diameter and standard deviation estimated from Figs. 9 a to 9 d, and b typical deposition rate, as a function of the Ar gas flow rate, V Ar, at constant Ar gas pressure, P Ar, of 170 Pa. to 280 Pa with increasing V Ar from 1.7 10 6 to 1.0 10 5 m 3 /s. At V Ar 1.7 10 6 m 3 /s, the deposit forms island morphology both near the sputtering target and at the entrance to the growth region, while there is almost no deposit at the exit from the growth region. Under this condition, no clusters are observed at all the deposited positions. At V Ar 5.0 10 6 m 3 /s, there exist only the island morphology without any clusters at the positions of both the sputtering target and the entrance to the growth region. However, small clusters are observed at the exit from the growth region. These results indicate that the clusters nucleate and grow only in the liquid-nitrogen-cooled growth region. At V Ar 1.0 10 5 m 3 /s, large particles some of them exceed 100 nm in diameter are formed even near the sputtering target and at the entrance to the growth region. However, such large particles suddenly disappear at the exit from the growth region, probably because they drop out from the carrier gas stream owing to their high gravity. By contrast, a large number of small clusters are observed there. This indicates that the nucleation and growth occur in a wide region at high V Ar. IV. DISCUSSION In the colloidal method, a particle formation process is generally explained by a LaMer s figure, 13 which provides a concentration evolution of dissolved species as a function of the reaction time. There are incubation, nucleation, and growth stages in the particle formation process. In the incubation stage, the concentration of the dissolved species increases with the reaction time by adding the solute or changing the temperature. Since the concentration is lower than a critical supersaturation, the nucleation can be negligible in this stage. On reaching the critical supersaturation concentration, the clusters start to nucleate the nucleation stage. Since the formed nuclei grow by absorbing the dissolved atoms or molecules, the concentration of the dissolved species decreases rapidly, suppressing the nucleation probabil-

488 J. Appl. Phys., Vol. 85, No. 1, 1 January 1999 Yamamuro, Sumiyama, and Suzuki ity. After the concentration of the dissolved species becomes lower than the critical supersaturation concentration, the nucleation process almost stops and only the stable nuclei grow by diffusion of the dissolved species to the nuclei and coagulation of the growing particles the growth stage. In order to produce the monodispersed colloidal particles, the following two prerequisites must be satisfied. 4 First, the nucleation and growth stages must be separated distinctly. Since the nuclei grow stocastically also in the nucleation stage, the long nucleation stage leads to the broad size distribution. Thus, for the monodispersed particle formation, the nucleation must be finished in the early stage of the particle formation process, and then only the initially formed nuclei must be grown without further nucleation. Second, the coagulation of the growing particles must be prohibited because it significantly broadens the particle size distribution. It is important to control the nucleation and growth processes precisely for the monodispersed cluster formation in the gas condensation method as well as in the colloidal method. However, It is difficult to control them in the conventional gas evaporation method, because the particles nucleate and grow rapidly in a narrow region just above an evaporation source. 11,12 When we use the carrier gas flow, we can extend the nucleation and growth regions. 14 Indeed, the nucleation does not occur near the sputtering source at low P Ar in the present experiments as shown in Fig. 11. Hence, we can control the nucleation and growth processes independently in the present PGC apparatus. As shown in Fig. 11, it is revealed that the nucleation and growth are restricted only in the liquid-nitrogen-cooled growth region under the monodispersed cluster formation condition V Ar 5.0 10 6 m 3 /s, i.e., P Ar 170 Pa and it extends over a wide region under the broad-sized cluster formation condition V Ar 1.0 10 5 m 3 /s, i.e., P Ar 280 Pa. These results can be understood in terms of the first prerequisite mentioned above. At low P Ar, the sputtered metal atoms do not aggregate quickly into clusters, because of low probability of three-body collisions among metal and inert gas atoms. Thus, we can control the nucleation process effectively by liquid nitrogen cooling of the growth region, by which the supersaturation ratio of the sputtered metal vapor increases rapidly. This promotes the nucleation process and rapidly reduces the number density of sputtered metal atoms by absorption into the nuclei. Therefore, the nucleation process is terminated instantaneously, and the nucleation and the growth stages are clearly separated from each other. Hence, this leads to the monodispersed cluster formation. We have confirmed that the liquid nitrogen cooling of the growth region reduces a width of the size distribution. 15 At high P Ar, on the other hand, clusters nucleate and grow even near the sputtering source, where the formed clusters grow by both atom absorption and coagulation because the number densities of the sputtered metal atoms and the nuclei are significantly large near the target. Thus, it is difficult to control the nucleation process precisely. Moreover, the cluster formation process extends over the wide region from the sputtering source to the growth region. Hence, the nucleation and growth stages do not separate clearly, and the size distribution becomes broad. The second prerequisite mentioned above requires the low number density of nuclei i.e., clusters in the carrier gas flow. As shown in Figs. 4 b, 7 b, and 10 b, the deposition rates are low when we obtained the monodispersed clusters at low P Ar. The number of clusters deposited at the exit from the growth region at low P Ar 170 Pa is also smaller than that at high P Ar 280 Pa as shown in Fig. 11. These results indicate that the number density of nuclei is low under the monodispersed cluster formation condition. In fact, the number density of the sputtered metal atoms in the present experiments is roughly estimated to be about 10 20 atoms/m 3 with referring the sputtering rate at P Ar 133 Pa. 16 Since this value is 10 2 10 3 times smaller than the number density of Ar gas atoms, the number density of nuclei is also low. Therefore, the collisions among clusters or nuclei are hard to occur, preventing the coagulation of the clusters. At high P Ar, on the other hand, the deposition rate is higher than that at low P Ar. Figure 11 also shows that a large number of clusters are deposited at the exit from the growth region at high P Ar 280 Pa. These results indicate that the number density of nuclei is high. Therefore, the collisions among clusters or nuclei will be easy to occur, promoting the coagulation. Consequently, we conclude that the formation process of the monodispersed clusters in the gas phase is similar to that of the colloidal particle formation. The carrier gas flow rate also influences the dispersion of the cluster size distribution as shown in Fig. 10 a : The size distribution becomes broad with increasing V Ar. This result will be related to the state of the carrier gas stream. With increasing V Ar, the number density of Ar gas atoms is constant at the same P Ar, but the thermalization and cluster formation processes are disturbed in the faster carrier gas stream because these processes are stochastic. Therefore, the size distribution becomes broad, though the mean cluster size is not sensitive to V Ar. For a detailed understanding, it is necessary to perform a computer simulation of the gas stream including metal clusters formed in the sputtering chamber and the growth region. Moreover, the low temperature process of the sputtering technique will be also favorable to produce monodispersed clusters, because it suppresses the coalescence growth during the coagulation. V. CONCLUSION We have deposited nanometer-sized Cr clusters on TEM microgrids by the PGC method and observed them by TEM. The monodispersed Cr clusters were obtained in the size range of 7.6 13 nm by varying the Ar gas pressure, P Ar, and the Ar gas flow rate, V Ar. In order to clarify the cluster formation process, we studied the effects of these process parameters on the cluster size distributions. The size distribution becomes narrower with decreasing both P Ar and V Ar : The monodispersed clusters can be formed at low P Ar and low V Ar. At low P Ar, the nucleation and growth occur only in the liquid-nitrogen-cooled growth region, and the deposition rate is rather low. At high P Ar, on the other hand, a number of large clusters are formed even near the sputtering source, and the nucleation and growth region extend over a wide region from the sputtering source to the growth region.

J. Appl. Phys., Vol. 85, No. 1, 1 January 1999 Yamamuro, Sumiyama, and Suzuki 489 Under the latter condition, the deposition rate is relatively high. These results demonstrate that the monodispersed cluster formation process in the PGC method is analogous to the monodispersed colloidal particle formation: The nucleation and growth processes are separated definitely and the coagulation of growing clusters is prohibited. In the present experiments, these conditions are achieved effectively by using the carrier gas flow and liquid nitrogen cooling of the cluster growth process. ACKNOWLEDGMENT The authors thank Dr. M. Sakurai and Dr. T. J. Konno for useful comments. This work was supported by Core Research for Evolutional Science and Technology CREST of Japan Science and Technology Corporation JST. They are also indebted to Laboratory for Developmental Research of Advanced Materials in our Institute IMR for its support. 1 See, for example, B. Abeles, P. Sheng, M. D. Coutts, and Y. Arie, Adv. Phys. 24, 407 1975 ; W. L. McLean, Mater. Res. Soc. Symp. Proc. 195, 117 1990 ; C. L. Chien, Annu. Rev. Mater. Sci. 25, 129 1995. 2 See, for example, Y. Yoshizawa, S. Oguma, and K. Yamauchi, J. Appl. Phys. 64, 6044 1988 ; H. Gleiter, Prog. Mater. Sci. 33, 223 1989 ; R.W. Siegel, Annu. Rev. Mater. Sci. 21, 559 1991. 3 P. Melinon et al., Int. J. Mod. Phys. B 9, 339 1995. 4 T. Sugimoto, Adv. Colloid Interface Sci. 28, 65 1987. 5 See, for example, E. Matijevic, Annu. Rev. Mater. Sci. 15, 483 1985 ; Fine Particles Science and Technology: From Micro to Nanoparticles, edited by E. Pelizzetti Kluwer, Dordrecht, 1996. 6 M. Oda and N. Saegusa, Jpn. J. Appl. Phys., Part 2 24, L702 1985. 7 S. Yamamuro, M. Sakurai, K. Sumiyama, and K. Suzuki, AIP Conf. Proc. 416, 491 1998. 8 S. Yamamuro, M. Sakurai, K. Sumiyama, and K. Suzuki, Supramolecular Sci. in press. 9 H. Haberland, M. Karrais, M. Mall, and Y. Thurner, J. Vac. Sci. Technol. A 10, 3266 1992. 10 P. Pfau, K. Sattler, J. Mühlbach, R. Pflaum, and E. Recknagel, J. Phys. F 12, 2131 1982 ; H. Schaber and T. P. Martin, Surf. Sci. 156, 64 1985 ; F. Frank, W. Shulze, B. Tesche, J. Urban, and B. Winter, ibid. 156, 90 1985. 11 S. Yatsuya, S. Kasukabe, and R. Uyeda, Jpn. J. Appl. Phys. 12, 1675 1973. 12 C. G. Granqvist and R. A. Buhrman, J. Appl. Phys. 47, 2200 1976. 13 V. K. LaMer and R. H. Dineger, J. Am. Chem. Soc. 72, 4847 1950. 14 S. Iwama and K. Hayakawa, Nanostruct. Mater. 1, 113 1992. 15 S. Yamamuro, K. Sumiyama, and K. Suzuki unpublished. 16 S. Yatsuya, T. Kamakura, K. Yamauchi, and K. Mihama, Jpn. J. Appl. Phys., Part 2 25, L42 1986.