Langmuir 1987, 3, 259-265 diffusion flame is currently in progress.
Conclusions A technique for extracting soot or other particulates from flames has been devised that affords a detailed examination of particle morphology and its regional variation within an annular diffusion flame. The method is based on subjecting a cold probe surface for exposure times as short as 30 ms to precisely defined regions of the flame. The transit times achieved are as low as 12 ms. Several thermophoretic probes exposed to the flames are examined by TEM for best image quality to reveal particle morphology. The morphological features so derived provide not only valuable qualitative information on particle agglomeration, surface growth, and oxidation but also quantitative data on primary particle size as a function of flame coordinates. Observations with a series of probes consistently show an increasing primary particle size with height above the burner in the low and intermediate regions of the flame. The size then decreases through oxidation as the particles move to the tip of the flame. Our studies further show that, a t intermediate heights in the flame, the primary particle size and the degree of agglomeration peak on the radius of maximum soot volume fraction (r = r,). Both TEM and SEM observations with two different probes show that the primary particle size and the state of agglomeration are substantially reduced on the oxygen-rich size (r 2 r,). A similar trend is present for the fuel-rich side (r < r,) but the observations of this region are less reliable. The technique described can be used for the study of a variety of flame-generated particles and affords important supplementary information to the data yielded by light scattering observations. Appendix Some simplifying assumptions can lead to an estimate of the quench time T ~ Le., , the time for a typical soot
259
particle to traverse from the flame to the probe through the thermal boundary layer of thickness 6,. The thermophoretic velocity driving the particles that are in the free molecular regime is given by
where K is the thermophoretic velocity coefficient, v the kinematic viscosity of the host gas mixture, and T the local temperature. We can approximate
where T, is the temperature of the probe, T the temperature of the host gas, T, the average thermak boundary layer temperature, v, the kinematic viscosity of the gas mixture at Ta,and 6, the thermal boundary layer thickness. The quench time is then given by
Acknowledgment. The instructions of Michael Sosnowski in the operation of the Philips 420 EM are acknowledged with deep appreciation. Roger L. Follansbee made major contributions to the design of the probe control system and was responsible for its construction. Professor Harry Kolsky generously provided high-speed photographic equipment and guidance in its use. This research was sponsored by the Center for Fire Research of the National Bureau of Standards under Grant NB83NADA4025. Registry No. Ethene, 74-85-1.
Formation of Ultrafine @-SiliconCarbide Powders in an Argon Thermal Plasma Jet? Peter C. Kong and E. Pfender” Department of Mechanical Engineering, University of Minnesota, Minneapolis, Minnesota 55455 Received August 13, 1986. I n Final Form: November 25, 1986 Ultrafine @-siliconcarbide powders are synthesized in a thermal plasma jet, generated with a swirl-stabilized plasma torch, using methane and silicon monoxide as reactants. High temperatures (>lOOOO K) combined with ultrarapid quench rates (>lo6K/s) lead to a high degree of supersaturation, resulting in homogeneous nucleation of ultrafine Sic particles. The maximum conversion of Si0 to Sic determined by thermogravimetric analyses is 97.3%. Particle size analyses show a bimodal distribution with the majority of the particles falling in a size range from 2 to 40 nm. Larger particles with sizes greater than 80 nm are also observed. Triangular and hexagonal Sic single-crystalplatelets are observed throughout this work, and the formation of these particles is discussed in this paper.
Introduction The plasma utilized for these experiments is classified as a thermal or equilibrium plasma which may be produced by means of electric ~ C orS rf discharges. Thermal plasm= provide high temperatures (>15000 K) and a chemically +Presented a t the symposium on “Fine Particles: High Temperature Synthesis”, 60th Colloid and Surface Science Symposium, June 15-18, 1986, Atlanta, GA; G. Mulholland, Chairman.
0743-7463/87/2403-0259$01.50/0
pure environment in contrast to conventional chemical flames. Over the past two decades thermal plasmas have gained wide recognition because of their potential for material processing. Thermal plasmas are used in extractive metallurgy and mineral e~ploitation,l-~ decomposition of in(1)Sayce, I. G. Adv. Extr. Metall. Refin. Proc. Int. Symp. 1971, 241. (2) Rykalin, N. N., Pure A p p l . Chem. 1976, 48,179.
0 1987 American Chemical Society
260 Langmuir, Vol. 3, No. 2, 1987
dustrial wastes,6 ceramic and metallic powder spray coatings,',* plasma sintering of ceramic^,^ single-crystal growth,'O ultrafine powder prod~ction,"-'~spheroidization of material^,'^^'^ coal gasification and coal desulfurization,16-22and plasma synthesis of refractory powd e r ~ . ~ ~ - ~ ~ Thermal plasma synthesis of refractory powders offers several advantages: (A) In a thermal plasma there exists a high-temperature reaction zone (>4000 K) which may induce fast reaction kinetics so that the overall reaction time becomes very short (in the order of milliseconds). (B) High-purity powders may be produced, which is particularly true with rf-generated thermal plasmas. (C) Uniform submicron particles may be produced with clean surfaces. These particles are particularly suitable for pressureless sintering. (D) The rapid quench rate in thermal plasmas may lead to the formation of metastable phases (e.g., pWC1-A. Conventionally, S i c is produced in the Acheson furn a ~ e Wei . ~ ~et al.34" reportedly synthesized P-SiC powders
(3) Hamblyn, S. M. L. Miner. Sci. Eng. 1977,9 (3), 151. (4) Thorpe, M. L. Adv. Extr. Metall. Rejin., Proceed. Int. Symp. 1971, 257. ( 5 ) Warren, I. H.; Shimizu, H. Trans.-Can. Inst. Min.Metall. Min. SOC.N. S. 1965,68,170. (6) Stokes, C. S. Chem. Eng. 1965,12,191. (7) Stokes, C. S.; Knipe, W. W. Ind. Eng. Chem. 1960,54, (2), 287. (8) Waldie, B. Chem. Eng. (Rugby, Engl.) 1972,No. 259, 92. (9) Johnson, L. D.; Rizzo, R. A. Am. Ceram. SOC.Bull. 1980,59,467. (10) Savitsky, E. M.; Burkanov, G. S. J. Cryst. Growth 1978,43,457. (11) Excell, S. F.; Roggen, R.; Billot, J.; Lux, B. Proc.-Electrochem, SOC.1973, 1965 (Fine Particles Symposium). (12) Wilks, P. H.; Lacroix, D. R. Proc.-Electrochem. SOC.1973,179 (Fine Particles Symposium). (13)Akashi, K.; Ohno, S.; Ishizuka, R.; Yoshida, T. Proceed. Int. Symp. Plasma Chem., 3rd 1977,3, paper S-5-6. (14) Fine Particles, Second International Conference; Kuhn, W. E., Ed.; Electrochemical Society, 1974. (15) Cheney, R. F. Mater. Res. SOC.Symp. Proc. 1984,30, 163. (16) Sheer, C.; Korman, S.; Dougherty, T. J. Proc. Int. Symp. Plasma Chem., 4th 1979,I , 227. (17) Sheer, C.; Korman, S. Adu. Chem. Ser. 1974,131, 42. (18) Schulze, R. A. Chem. Ind. (London) 1968,9,1539. (19) Korman, S.; Dougherty, T. J.; Sheer, C. Proc. Jnt. Symp. Plasma Chem., 4th 1979, I , 295. (20) Muller, R.; Peuckert, C. Proc. Int. Symp. Plasma Chem.,6th 1983, 1, 270. (21) Gauvin, W. H.; Mum, R. J.; Grosdidier de Matons, P.; Stuart, P. Proc. Int. Symp. Plasma Chem., 6th 1983,I , 276. (22) Szymanski, A.; Plotczyk, W. W.; Resztak, A,; Huczko, A. Proc. Int. Symp. Plasma Chem., 6th 1983,1, 294. (23) Matsumoto, 0.; Asakura, T.; Konuma, M.; Kanzaki, Y. High Temp. Sci. 1978, I O , 175. (24) Matsumoto, 0.; Shirato, Y.; Miyazaki, M. J . Electrochem. SOC. Jpn. 1968,36 (4), 219. (25) Kuhn, W. E. J . Electrochem. SOC.1963,110, 298. (26) Matsumoto, 0.; Miyazaki, T. High Temp. Sci. 1973,5, 40. (27) Kuhu, W. E., Proc.-Electrochem. SOC.1973,81 (Fine Particles Symposium). (28) Boudin, E.; Fauchais, P. Proc. Int. Symp. Plasma Chem., 3rd 1977,invited paper, Paper G-1. (29) Matsumoto, 0.;Yaguchi, Y. Proc. Int. Symp. Plasma Chem., 3rd 1977,3, Paper S-4-2. (30) Matsumoto, 0.; Konuma, M.; Kawazki, Y. Proc. Int. Symp. Plasma Chem., 4th 1979,271. (31) Kong, P.; Young, R.; Zuzuki, M.; Pfender, E. J . Plasma Chem. Plasma Process. 1983,3 (l),115. (32) Yoshida, T.; Kawasaki, A,; Nakagawa, K.; Akashi, K. J . Mater. Sci. 1979,14, 1624. (33) Knippenberg, W. F. Philips Res. Rep. 1963,18(30),Chapters 1 and 2.
Kong and Pfender
by carbothermic reduction of silica at temperatures slightly above 1550 "C. These preparative procedures for Sic require long reaction times and the particles produced are large agglomerates with a wide range of particle sizes from 0.8 to 85 pm. Recent experiments have shown the feasibility of the synthesis of S i c by thermal cracking of organosilanes (R,SiC14-, or R,SiH4-,)36-40 or by a reaction of methane and silane in thermal plasmas.41 All these reactions occur in the gas phase. Other attempts to synthesize S i c in thermal plasmas involving either or t ~ osolid ~ ~ reactants have met with partial success only. In principle the kinetics of a solid/gas or solid/solid reaction is slow compared to a gas/gas reaction even when thermodynamic considerations are favorable. Solid/gas reactions generally require a longer time before any substantial product formation takes place. A gas/gas reaction benefits only partially from the high-energy content of the plasma, whereas a solid/gas or solid/solid reaction may require the high-energy content of the plasma. In the latter case solid particles are melted and subsequently evaporated after entering the plasma. The reaction between silicon oxide and methane exemplifies this aspect. In a series of runs silicon monoxide has been used as the solid reactant, injected simultaneously with CH, into an arc plasma. The silicon monoxide directly vaporizes in the plasma and then reacts with carbon-bearing species from decomposed CHI to form Sic. For product analysis, X-ray diffraction is used for phase identification, TEM for particle size measurements and for determination of particle morphologies. Thermal gravimetric analysis (TGA) using an electrobalance has been employed to determine the total carbide phase and free carbon in the sample.
Thermochemistry of SiO/CH4 Reactions A computer program, SOLGASMIX-PV,~~ has been used for calculating the chemical equilibria in a system containing a gaseous phase, solid solutions, liquid solutions, and compounds of invariable and variable compositions. The calculations comprise system compositions, within certain constraints, having the minimum free energy. The constraints refer to the conservation of mass of each element present at either constant pressure or volume. In a system containing Si, 0, C, and H, about 38 species are found over a temperature range from 500 to 6000 K. Figure l a shows the equilibrium composition for the stoichiometric reaction between S i 0 and CHI. A t high temperatures, Si0 is completely reduced by C(g) to Si(g). As the temperature decreases, the concentrations of C(g) and Si(g) drop rapidly due to the formation of other stable gas-phase intermediates. Some of the important species formed in the temperature range from 3000 to 5000 K are Si(g), C2H(g),SiC2(g),and Si2C(g). These species interact (34) Harris, L. A.; Kennedy, C. R.; Wei, C. T.; Jeffers, F. P. J . Am. Ceram. SOC.1984,67,C-121. (35) Wei, C. T. J. Am. Ceram. SOC.1983,66, C-111. (36) Kato, A.; Okabe, Y.; Hojo, J. J. Jpn. SOC.Powder Powder Metall. 1982,27 ( l ) , 32. (37) Inukai, T. J. Jpn. SOC. Powder Powder Metall. 1982,27(8), 249. (38) de Pous, 0.; Mollard, F.; Lux, B. Proc. Int. Symp. Plasma Chem. 3rd 1977,3, S-4-7. (39) Salenges, R. M. Ind. Eng. Chem. Prod. Res. Deu. 1977,16, 230. (40) Canteloup, J.; Mocellin, A. Spec. Ceram. 1975,6 , 209. (41) Vogt, G. J.; Hollabaugh, C. M.; Hull, D. E.; Newkirk, L. R.; Petrovic, J. J. Proc. Mater. Res. SOC. Symp. 1984,30, 283. (42) Sayce, I. G.; Selton, B. Spec. Ceram. 1972,5, 157. (43) Kuhn, W. E. Ultrafine Particles; Wiley: New York, 1963; p 156. (44) Ando, Y.; Ohkohchi, M. J . Cryst. Growth 1982,60, 147. (45) Besmann, T. M. Oak Ridge Natl. Lab., [Rep.]ORNL-TM (U.S.) 1977,ORNL-TM-5775.
,
~
~
Langmuir, Vol. 3, No. 2, 1987 261
Ultrafine @-Silicon Carbide Powder Formation 2.0
1.8
1.6
1.4
I .2
-001
=
1.0
0.8
0.6
0.4
‘ I
0.2 I
(
1
1
1
5
1
T
0.0 1000
0
2000
3000
4000
5000
[io3K1
6
T (W TEMPERATURE [lo0 K1 2
1 2 r 3. 12
30
28
26
24
22
10
a Y
9
s
6 (I
5 -74
-a -
2
3
3.5
1/T
4 [104K-’l
I
4.5
- 9-1 0
34
30
26
22
T[lOOKI
Figure 1. (a) Equilibrium diagram of the Si0 + 2CH4reaction. (b) Free energy of the reactions involved in Sic formation. (c) Gas-phase reactions leading to SiC(s)formation (see listing of possible reactons). (d) Partial pressures of Sic with and without SiC(s)precipitation.
via gas-phase reactions to form SiC(s). Below 3000 K silicon carbide precipitates as a stable product by chemical vapor deposition (CVD). Reactions leading to silicon carbide formation are listed below and the corresponding free energy of these reactions is shown in Figure lb.
-- -++ + + + -
Si(g,l) + C(g,s)
SiC(s)
(a)
Si2C(g)
SiC(s)
Si(g)
(b)
SiC2(g)
SiC(s)
C(s)
(C)
SizC(g) SizC(g)
SiCz(g)
SiCz(g) S i k ) + C,H(g)
C(s)
Si2C(g) + C2H(g)
3SiC(s)
(d) (e)
2SiC(s)
(f)
SiC(s) + C(s) + l/zHz(g)
(g)
-
Si(g)
2SiC(s)
2SiC(s) + C(s) + 1/2H2(g) (h)
The first four reactions are unimportant, as discussed
elsewhere.46 Reactions e-h are gas phase and their large negative free energies favor SiC(s) formation. All these gas-phase reactions involve two important reactive intermediate subcarbides, namely, Sicz and Si2C. In the temperature range from 3500 to 3000 K, equilibrium composition calculations predict no SiC(s) precipitation from these gas-phase reactions. Figure ICshows a plot of log K vs. 1/T for these reactions. When the temperature drops from 3500 to 1800 K, the system is highly supersaturated with the gas-phase reactive intermediates and SiC(s) precipitates directly from the reactions. One interesting result deduced from the calculation is that the S i c vapor is not involved in the nucleation to SiC(s). Calculations in the temperature region from 3500 to 2000 K, suppressing SiC(s) formation, show (Figure Id) an insignificant increase of the partial pressures of Sic. This demonstrates that (46) Kong, P.; Huang, 1986, PS-14, 357.
T.T.;Pfender, E. IEEE Trans. Plasma Sci.
262 Langmuir, Vol. 3, No. 2, 1987
Kong and Pfender
Ar in
injec
c
I
I
ction tube quenching
collection
chamber
-cooling
I1
+ v
I TIME
Figure 3. Thermogravimetric analysis of the product.
+gas exhaust flan
Figure 2. Plasma torch reactor. SiC(s) is formed b y chemical rather than b y physical vapor deposition. Details of the C V D reactions for SiC(s) formation i n a fast q u e n c h e d plasma will be discussed elsewhere.47 Equilibrium calculations provide an overall picture of the possible reactions and of equilibrium compositions of the final products. T h e y cannot describe the actual kinetics and the controlling mechanisms of a particular reaction, b u t t h e y present useful guidelines for a chemical system.
Experimental Setup and Procedures A dc plasma jet reactor has been used for the synthesis of S i c ultrafine powders. This reactor can be operated either in a transferred or in a nontransferred arc mode. The reactor has been used extensively for the production of fine powders of transition-metal carbides and nitrides and it is capable of processing In a plasma jet, whether both powders and gas confined or totally free, the temperature decays over a relatively short distance and the corresponding temperature gradients are very steep. The steep temperature gradients (>lo4K/m) combined with the high velocities (-100 m/s) of the plasma lead to quenching rates in excess of lo6 K/s. It is the fast quenching rate that is responsible for producing ultrafine particles in the plasma. A schematic of the plasma jet reactor is shown in Figure 2. The reactor consists of a swirl-stabilized plasma torch and a quench/collection chamber. Reactants are injected radially into the plasma jet through a two-nozzle injection ring, immediately downstream of the anode. The reaction occurs in the plasma jet and the products are allowed to cool by adiabatic expansion in the collection chamber. The plasma torch is operated with Ar at atmospheric pressure and the arc current and voltage assume values around 900 A and 25 V, respectively. The arc voltage may be substantially higher when a diatomic gas such as hydrogen or nitrogen is added to the plasma fluid or by increasing the torch gas flow rates. The plasma jet temperatures may rise as a consequence of this higher power dissipation. In this experiment, silicon monoxide is first produced by hydrogen reduction of silica particles in a confined plasma jet. Then the S i 0 powder is used as a reactant in conjunction with CH4 to form S i c powder in an atmospheric pressure plasma jet, using argon as the base gas. 147) Kong, P., to be published in, Proceedings of the Materials Research Society, Spring, 1987. (48) Ronsheim, P.; Mama, A,; Christensen, A. N. Plasma Chem. Plasma Process. 1981, 1, (2), 135. (49) Mitrofanov, B.; Pfender, E.; Ronsheim, P.; Toth, L. E. Mater. Sci. Eng. 1981,48, 21. (50) Ronsheim, P. A. Ph.D. Thesis, University of Minnesota, 1981.
Unlike silicon dioxide, silicon monoxide has a high vapor pressure. The temperature difference between melting and boiling for silicon monoxide is less than 200 'C; Le., when a silicon monoxide particle is injected into the plasma, and is subjected to the intense heating in the plasma, the particle will rapidly evaporate. Due to this rapid heating, silicon monoxide can be easily dissociated or reduced by carbon to release silicon vapor particles for the reaction. Therefore, silicon monoxide is well suited for synthesizing silicon carbide. For the synthesis of Sic, a free plasma jet is used. The reaction occurs in the plasma jet within the collection chamber and the products condense on the walls after leaving the plasma jet. For improving particle heating, the injection ring is equipped with a short graphite projection tube which extends into the fringes of the plasma jet. The projection tube allows the particles to penetrate deeper into the high-temperature region of the plasma core. Inside this region, particles evaporate much faster than particles in the boundary layer. A series of experiments, performed with a fixed silicon monoxide mass injection rate of 1.27 g/min and methane mass injection rates varying between 0.14 and 2.41 g/min, show that the color of the product varies from one experiment to another, depending on the amount of methane used for the reaction. For the lowest limit of 0.14 g/min, the color of the powder appears yellowish brown, and above 1.02 g/min of methane, the powders are black because of an excess amount of free carbon present in the product.
Results and Discussion In the reaction product, the free carbon and t h e unreacted silicon monoxide a r e amorphous and, therefore, X - r a y diffraction cannot determine t h e relative a m o u n t of each phase in the sample. This, in turn, will hinder t h e determination of total carbide produced i n the reaction. Therefore, a thermogravimetric analysis has been performed to determine the total silicon carbide yield in each experiment. Figure 3 shows a typical thermogravimetric curve for the oxidation of the products derived from a plasma j e t experiment. The weight loss represents the weight of free carbon in the sample while the weight gain refers to t h e oxidation of S i c a n d / o r S i 0 and S i (if present). Free carbon begins to oxidize around 330 "C, b u t the oxidation rate is m u c h faster at t e m p e r a t u r e s above 550 "C. The carbide oxidation begins above 750 "C. Results of the thermogravimetric analyses a r e plotted i n Figure 4 and comparisons a r e m a d e with theoretical yield curves predicted at different temperatures. T h e m a x i m u m conversion occurs at a C H 4 / S i 0 mole ratio of 2.23. The shape of the experimental curve is similar to t h e theoretical yield curve at 2500 K and this suggests that the reaction proceeds at temperatures above 2500 K. The fluctuation of t h e data remains within 4% for identical runs and i t is d u e to t h e difficulty of reproducing the same powder feed rate into the plasma in different runs. E v e n
Langmuir, Vol. 3, No. 2, 1987 263
Ultrafine p-Silicon Carbide Powder Formation .
100.
.
,
.
. .
.
.
.
,
. . . .
1
\\\
a8
0
1.1
211
16
24
28
C H 4 1 S i O Mole ratio
PARTICLE SIX
nm
Figure 4. Experimental and theoretical yield for Sic formation.
PLASMA JET REACTION
I i; PARTICLE SIZE
nm
Figure 6. (a) Particle size distribution derived from X-ray diffraction profile D in Figure 5. (b) Particle size distribution derived from X-ray diffraction profile G in Figure 5.
60
50
40
30
20
28
Figure 5. X-ray diffraction profiles of Si0 products.
+ CHI reaction
small fluctuations of the powder feed rate will change the CH,/SiO ratio resulting in corresponding fluctuations of the carbide conversion. The experiment requires a CH,/SiO ratio of 2.23 for a complete reaction. This value is 11.5% higher than the theoretical value for a stoichiometric reaction. In the reaction between silicon monoxide and methane, free silicon precipitation is always observed when the CH4/Si0 ratio is less than 2.23. Some typical X-ray traces of the reaction products obtained in an unconfined plasma jet are represented in Figure 5. The CH4/Si0 mole ratio increases in alphabetical order from A to G. At low methane flow rates in the system (Figure 5A), the carbide formation is low and liquid silicon droplets nucleate from the gas phase and they are subsequently quenched into solids. This is evident as silvery silicon metal spheres which are always observed as a byproduct when the CH4/Si0 mole ratio falls below the value for a stoichiometric reaction. The X-ray traces from Figure 5, traces B-E, demonstrate that when the CH,/SiO ratios approach values slightly below or above those required for a stoichiometric reaction, almost a single-phase product of Sic is obtained. The maximum yield of Sic obtained in these experiments is 97.3% (trace D) as determined by ther-
mogravimetric analysis. When high methane flow rates are used (traces F and G), the yield of carbide drops substantially due to heavy precipitation of free carbon from the reaction. The free carbon obtained is amorphous. One interesting result observed from this series of experiments is that at low carbon to silicon ratios, the particle size is generally larger and the reaction is incomplete. At higher carbon to silicon ratios, the reaction is more complete, but the average particle size decreases. This is clearly demonstrated in Figure 5. There are several reasons for this variation. First, at lower methane flow rates the plasma jet is not quenched as severely as in the case of high methane flows. Therefore, the average temperature in the reaction zone is higher. Second, because of the low methane concentration in the system, the plasma reaction zone will not be highly supersaturated with the gas-phase reaction products. Under the condition of a low degree of supersaturation, there is a likelihood of generating variations in the concentration of the nutrient molecules in the plasma which will affect the growth of particles. By coupling several factors, such as higher temperatures, lower quench rates, lower degree of supersaturation, and variations in concentration of the gas-phase products, heterogeneous nucleation becomes more favorable. Heterogeneous nucleation produces multiple-phase products with larger particle sizes and the reaction is usually incomplete. This conjecture is confirmed by the X-ray profile (Figure 5A). When the CH,/SiO ratio is approaching the value for a stoichiometric reaction, both homogeneous and heterogeneous nucleations of particles occur at the same time with homogeneous nucleations playing a dominating role in the product formation. The successive broadening of the X-ray profiles (Figure 5B-D) due to increasing CH,/SiO ratios is indicative of the process. This is further supported by TEM observations of a wider distribution
264 Langmuir, Vol. 3, No. 2, 1987
Kong and Pfender
a
Figure 7. TEM micrographs of SIC polycrystals.
of the much finer particles compared to the larger particles. Figure 6a shows a particle size distribution profile of sample D, Figure 5. There are two particle size ranges in this distribution. The first one is in the lower submicron range and bas an average particle size of 36 nm. Homogeneous nucleation is believed to be responsible for the formation of these particles. A second peak, with an average particle size around 96 nm, is in the higher submicron range. These larger submicron particles are most likely generated by heterogeneous nucleation, condensation, and conglomeration of the much smaller particles. Heterogeneous nucleations will form larger single crystals with specific orientations. The condensation process will favor the formation of particle conglomerates. Collisions among solid particles will result in loose agglomerates which may subsequently sinter in flight in the plasma resulting in larger polycrystals. The sintering of particle agglomerates into larger polycrystals has much less significance in terms of its contribution to the overall number of larger particles, since the dwell time of particles in the plasma is extremely short Therefore, the major mechanisms for the formation of larger polycrystals are probably condensation and heterogeneous nucleation. The size of particles decreases even further into the lower submicron range a t very high methane flow rates (Figure 5G). The particle size distribution of sample G, Figure 5, is shown in Figure 6b. This distribution shows ultrafine particles with an average particle size of approximately 10 nm. A smaller amount of slightly larger ultrafine particles with an average size of 38 nm is also observed. The severe quenching of the plasma by very high methane flows will lower the temperature of the reaction zone in the plasma jet. These conditions enhance the degree of supersaturation of the vapor species. Under high supersaturations and small fluctuations of the concentration gradients of nutrient particles, homogeneous nucleation of solid particles from the gas phase will become favorable. The narrow range of the particle size, 9.6 i 4.4 nm, strongly suggests that particles are nucleated by purely homogeneous processes. In a chemical vapor reaction, whether single crystals or polycrystals form from the reaction will depend on the degree of supersaturation of chemical vapors. In a reaction system which has an extremely high degree of supersaturation tending toward a homogeneous nucleation, all growing crystal faces will have a large number of kinks or growth sites and thus a corresponding growth rate. Consequently, particles with isotropic growth will form polycrystals, resulting in spherical shapes (Figure 78). Single crystalsalso have spherical shapes as a whole, yet they are often faceted. In the case of very high supersaturation, the formation of anisotropically grown single crystals by vapor-phase reactions is basically limited. Particles formed by homogeneous nucleation are extremely small (Figure 7a) and they tend to aggregate. Particle aggregates can be sintered in fight to form large fused polycrystals (Figure 7b). But this type of fused polycrystals constitute only a minor fraction of the total particles.
.A,
% 4
Figure 8. TEM micrographs of Sic single-crystal platelets.
Throughout this work, spherically faceted single crystals have never been observed. Only thin single-crystal platelets with specific shapes have been frequently observed. Some of the thin single-crystal platelets are equiaxial triangles and truncated triangles, hexagons and irregular hexagons, tetrahedrons and truncated tetrahedrons, and rhombuses. Examples of these thin single-crystal platelets are shown in Figure 8a-h. These particles, triangles, hexagons, and tetrahedrons all have the same orientation. The basal plane of the crystals are the (111)-type planes as determined by electron diffraction. The electron diffraction patterns in Figure 8 parts c and d correspond to the truncated triangle in part a and the regular hexagon in part b. The patterns show (111)-typeorientations. The transparency of these crystals indicates that the particles are very thin. The uniform intensity shown on the electron micrographs suggests that these particles have uniform thickness and the crystal planes are perfectly formed. Most of the particles observed are larger than 100 nm, although the growth of large single crystals from a gasphase reaction under superfast quenching conditions is not likely. Figure 8g shows the evidence of epitaxial growth on a substrate particle. Originally the particle is a truncated triangle and the edges are growth fronts. The edges of a crystal are less perfect than the crystal face and always contain defects. During growth, the edges will outgrow the crystal face, thus forming thin platelets. If the particle drifts to a region in the plasma where the temperature is relatively high and the supersaturation of the vapor is relatively low, epitaxial nucleation on the growing front can occur simultaneously. Note the nonuniformity of
Langmuir, Vol. 3, No.2, 1987 265
Ultrafine 8-Silicon Carbide Powder Formation a
Figure 9. TEM micrographs of Sic single crystals. a
-
b
__J
2Bnm
24~nm
Figure 10. TEM bright and dark field images of truncated Sic tetrahedrons.
thickness and width of the growth front which implies severe fluctuations of the temperature and vapor density gradients. The single-crystal platelet in Figure 8h shows only one growing front, suggesting fluctuations in the temperature and vapor density gradients in the plasma. The three-dimensional particle in Figure 9c is a tetrahedron hounded by four (111)-type planes. The electron beam is in the (111)direction with reference to the basal plane and this gives rise to the (22O)-type diffraction spots (Figure 9b). This particle is the result of a fast epitaxial growth on the basal plane. In order to have a fast epitaxial growth on the crystal face, the particle must be in a plasma region where the degree of chemical vapor supersaturation is extremely high and the density fluctuation of the vapor must be low. The rhombus in Figure 9a is prohahly formed by a collision between two equilateral triangles. The dark line dividing the rhombus is the grain boundary. If the rhombus is formed by the growth of one edge of an equilateral triangle in a very narrow region of extremely high vapor density and supersaturation, no grain boundary should be observed. Figure 10a shows an apex truncated tetrahedron. The upper and basal planes are the (111)planes of the crystal. Three beveled surfaces forming the side planes and layer structures on the beveled edge are clearly visible. These steps suggest that the particle is formed by a two-dimensional epitaxial nucleation and growth process. A band of stacking faults running across the (111)surface is indicated by white arrows. By measuring the width of the
stacking fault band (Figure lod), the thickness of the crystal is estimated to be about 15 nm. From the thickness and the width of the beveled edge of the particle, the angle of inclination of the edge is -70'. Therefore, the beveled planes are identified as the (111)-type planes. The growth of this particle occurs by epitaxial nucleation on the (111) face. Conditions favorable for this type of growth exist when the temperature of the growth area is not too high and the degree of supersaturation of the vapor is just high enough to promote nucleation and growth on the face of the crystal. If the crystal face has defects such as dislocations and stacking faults in addition to kinks and steps, then the crystal face will outgrow the edges. The particle in Figure 10 does possess these features, stacking faults and kinky-stepped faces for two-dimensional growth. Under high-angle tilting, 29O away from the basal plane along the (110) direction, the upper (111)plane and the stacking fault become very prominent. The layer on the beveled edge thus becomes visible (Figure lob). Since the beveled planes make an angle of -70" with the basal (111) plane, these planes must belong to one of the (111)planes. When tilting the crystal along the (110) direction, one of the (111)planes will he aligned with the optical axis, thus showing up in dark field imaging (Figure 1Oc). If this particle stays in the plasma long enough, it will grow into a complete tetrahedron which is the same as that shown in Figure 9c.
Conclusions (1)Synthesis of ultrafine j3-Sic powders in a thermal plasma is feasible by a reaction between CHI and SiO. (2) The reaction represents a high-temperature gasphase reaction which occurs in a temperature range from 2500 to 4000 K. (3) T w o important gas-phase reactive intermediates (Sic2and Si$) are involved in SiC(s) formation hy a CVD process. (4) The experimental yield (97.3%) of SiC(s) at a CH,/SiO mole ratio of 2.32 is in good agreement with theoretical predictions, obtained from thermodynamic equilibrium calculations. (5)Homogeneous nucleation is the predominant process for the formation of ultraline powders while heterogeneous processes seem to be involved in the formation of larger single-crystal platelets.
Acknowledgment. This project has been supported by NSF under Grant NSF/ENG 8200 628. Registry No. SIC, 409-21-2.