System for In Situ Characterization of Nanoparticles Synthesized in a Thermal Plasma Process

Transcription

1 Plasma Chemistry and Plasma Processing, Vol. 25, No. 5, October 2005 ( 2005) DOI: /s System for In Situ Characterization of Nanoparticles Synthesized in a Thermal Plasma Process X. Wang, 1 J. Hafiz, 1 R. Mukherjee, 1 T. Renault, 1,2 J. Heberlein, 1 S. L. Girshick, 1 and P. H. McMurry 1,3 Received November 1, 2004; revised February 15, 2005 We have designed a particle diagnostic system that is able to measure particle size and charge distributions from low stagnation pressure ( 746 Pa) and high temperature ( K) environments in near real time. This system utilizes a sampling probe interfaced to an ejector to draw aerosol from the low pressure chamber. Particle size and charge distributions are measured with a scanning mobility particle sizer. A hypersonic impactor is mounted in parallel with the scanning mobility particle sizer to collect particles for off-line microscopic analysis. This diagnostic system has been used to measure size and charge distributions of nanoparticles (Si, Ti, Si Ti N, etc.) synthesized with our thermal plasma reactor. We found that the mean particle size increases with operating pressure and reactant flow rates. We also found that most particles from our reactor are neutral for particles smaller than 20 nm, and that the numbers of positively and negatively charged particles are approximately equal. KEY WORDS: Nanoparticles; plasma synthesis; particle diagnostics; size distribution; charge distribution. 1. INTRODUCTION The need to measure aerosol size distributions in low pressure environments that may also be at elevated temperatures is encountered in aerosol synthesis reactors, semiconductor processing equipment, etc. A variety of approaches have been used for such measurements, including laser light scattering, (1 3) sampling from the exhaust of turbomolecular pumps, (4,5) mobility analysis at low pressure, (6 9) and particle beam mass spectrometry (PBMS). (10 12) There are limitations to each of these methods. While laser light scattering is fast and can provide information on 1 Department of Mechanical Engineering, University of Minnesota, 111 Chruch St. S.E., Minneapolis, MN 55455, U.S.A. 2 Current address: Thermal Dynamics Corporation, 82 Benning St., West Lebanon, NH 03784, U.S.A. 3 To whom correspondence should be addressed. wxl@me.umn.edu /05/ / Springer Science+Business Media, Inc.

2 440 Wang et al. aerosols throughout a reactor, the light scattering signals are complex convolutions of particle size, shape, concentration and refractive index. It is not trivial to deconvolute these parameters to obtain accurate information on particle size distributions. Measurements made in the exhaust stream of a vacuum pump are subject to changes in size distributions that occur due to deposition, coagulation or nucleation as the aerosol travels through the pump. Mobility analyzers that operate at pressures as low as 200 Pa have been reported, but these instruments are constrained to measurements of particle size distributions in the 3 20 nm diameter range. The PBMS was designed to sample from reactors at 133 Pa and can measure size distributions of particles in the nm diameter range. A drawback of the PBMS is that it is a prototype instrument that requires more expertise to operate than commercially available instrumentation. In this paper we describe a sampling scheme that combines an air ejector with a scanning mobility particle sizer (13) to measure particle size and charge distributions from high temperature and low pressure environments. This system is similar to the one described previously. (14,15) We made two major modifications, First, a new type of ejector (Air-Vac, UV143H) is used. This modification enables us to sample particles from the supersonic jet at a chamber pressure of 266 Pa, while the old system only works at chamber pressures of 2666 Pa or higher. Second, a bypass route was added to the bipolar charger, which enables us to measure the charge distribution of aerosol sampled from the reactor. In brief, this system utilizes an ejector-dilutor system to sample particles from the hot vacuum zone through a water-cooled sampling probe and to adjust particle concentrations to a suitable level for measurements with the scanning mobility particle sizer. (13) The scanning mobility particle sizer consists of a differential mobility analyzer (DMA) (16) to select particles of a given mobility, and an ultrafine condensation particle counter (UCPC) (17) to detect them. Size distributions are determined by carrying out UCPC measurements for a range of mobilities. Both a regular DMA (16) and a Nano-DMA (18) are used in our scanning mobility particle sizer to measure size distributions of particles in the nm diameter range with a time resolution of 2 min. We refer these measurements as in situ meaning that the particle measurement system samples particles directly from the aerosol jet. The measurements are near real time because the aerosol residence time from the sampling inlet to the measurement instruments is typically less than 5 s. A hypersonic impactor (19) with cut size of about 6 nm silicon particles samples the aerosol in parallel with the scanning mobility particle sizer. Particles are collected on TEM support grids for off-line measurements of morphology and chemical composition. We report here on the performance of this

3 System for In Situ Characterization of Nanoparticles 441 Fig. 1. Schematic diagram of hypersonic plasma particle deposition apparatus. nanoparticle characterization system and its use for measuring size and charge distributions of nanoparticles produced by our hypersonic plasma particle deposition (HPPD) apparatus. (15,20 22) The experimental apparatus for particle synthesis and deposition with the HPPD process is shown in Fig. 1. Vapor phase reactants are injected into a DC thermal plasma and dissociate due to the high temperature (about 4000 K) in the reaction zone. The reactant mixture then expands from a static pressure of about 67 kpa through a converging nozzle to a deposition chamber maintained at approximately 266 Pa. Rapid cooling during expansion drives nucleation of nanoparticles in the nozzle. Particles are then accelerated to high velocities in the nozzle and in the downstream free expansion region. Finally particles impact on substrate 1 hypersonically and form a nanostructured film. We refer to this process as high rate deposition. In a complementary process, substrate 1 is removed and particles are passed through an aerodynamic lens assembly to form a tightly collimated particle beam. (23,24) The particle beam can be used to directly write micropatterns with characteristic dimensions of a few tens of microns on the computer controlled substrate 2. We refer to this process as high definition deposition. The size and charge distributions of the synthesized particles are of primary importance in the HPPD process. Many mechanical properties of

4 442 Wang et al. the nanostructured films and micropatterns are directly related to particle size. Electrical charges carried by particles enable the use of electrostatic forces to increase impaction velocity, enhance deposition efficiency and manipulate the deposition pattern. Furthermore, knowledge of particle size and charge distributions provides insights into the formation process for these nanoparticles, which in turn helps us to optimize operating conditions such as reactant flow rates, operating pressures and plasma conditions. Therefore, it is desirable to measure particle size and charge distributions in real time. It is not straightforward to measure size and charge distributions of particles under our experimental conditions. The normal operating pressure in the deposition chamber is low ( 266 Pa background pressure and 3600 Pa stagnation pressure in the jet); the temperature at the sampling location is high ( K); and the particle concentration is high ( cm 3 ). We show in the following sections that our particle diagnostic system can successfully overcome these difficulties and measure quantitatively particle size and charge distributions in near real time. 2. EXPERIMENTAL SETUP A schematic of the particle diagnostic system is shown in Fig. 2. Particles exiting the nozzle are sampled through a water-cooled molybdenum Fig. 2. Schematic diagram of the particle diagnostic system.

5 System for In Situ Characterization of Nanoparticles 443 probe connected to a two-stage ejector. For the data reported in this paper the probe was located 2 cm downstream of the nozzle, where the high rate deposition of films takes place, but it can be moved to sample particles from different axial locations or be rotated to sample from off-axis locations. A high flow rate (123.5 slm) of particle-free nitrogen is fed to the ejector to create the reduced pressure that draws aerosol from the sampling probe and delivers it to atmospheric pressure where measurements are carried out using the scanning mobility particle sizer and the hypersonic impactor. The scanning mobility particle sizer is used to measure particle size distributions and charge state, and the hypersonic impactor is used to collect nanoparticle samples for analysis by electron microscopy. The sampling probe is a short molybdenum capillary with an inner diameter of 3 mm. The first half length of the probe is cooled by water to prevent melting. Note that the sampling is subisokinetic, which results in modest enrichment of the larger particles in the sampled flow. The ejector is a two-stage Venturi tube that can sample aerosol from a 746 Pa stagnation pressure. Although our reaction chamber operates at about 266 Pa, a normal shock forms in front of the sampling probe and the static pressure behind the shock (and just in front the sampling probe) rises back to about 3600 Pa, (22) which enables us to sample particles from the chamber. The ejectors have simple flow paths that we believe will lead to minimal perturbation on sampled particle size distributions. The sampling extraction system was designed to minimize changes in sampled aerosol size distributions due to coagulation prior to measurement. The aerosol residence time from the sampling inlet to the point where nitrogen dilution occurs inside the ejector is less than 10 ms. Particle coagulation is negligible for aerosol concentrations up to cm 3 in such a short time. The primary nitrogen dilution flow reduces concentrations to 10 8 cm 3 during the 0.29 s transport time to the point where secondary dilution occurs, again helping to suppress coagulation. The pressure-dependent dilution factor of the ejector is obtained by calibrating the flowrate through the sampling probe as a function of chamber pressure and nitrogen flowrate. A portion of the flow leaving the ejector passes through a laminar flowmeter prior to further dilution by clean nitrogen. The dilution factor of this second dilution stage can be adjusted by controlling the aerosol flow through the laminar flowmeter or the dilution nitrogen flow so that particle concentrations are brought to the range that can be measured with minimal coincidence by the UCPC (17) (about 10 4 cm 3 ), which counts particles individually. Particle coagulation is negligible during the 4 s transport time from the second dilution stage to the UCPC due to the low concentration. The data reported in this paper was corrected for both ejector and second stage nitrogen dilutions.

6 444 Wang et al. Prior to entering the scanning mobility particle sizer particles either pass through a 210 Po bipolar charger, or through a dummy charger that is geometrically identical to the bipolar charger but that does not contain a source of ionizing radiation. Particles that pass through the bipolar charger are brought to a well-defined stationary charge distribution (25) from which the total (neutral and charged particles) size distribution can be inferred. The 210 Po sources used in our system had an initial activity of 0.5 mci and were replaced every 90 days. The efficacy of a bipolar charger is determined by the product of the concentration of positive and negative ions that it produces, N, and the time to which particles are exposed to these ions as they flow through the charger, t. Based on rates of ion production and recombination from this source and on the aerosol residence time in the source, we estimate that the Nt product for this charger is equal to or greater than ions/cm 3 s. This Nt value is large enough to ensure particles passing through the charger will achieve the stationary charge distribution, independent of their initial charge state. (26) The dummy route is used to measure the unaltered charge distribution of the aerosol as it is delivered to the scanning mobility particle sizer, in an effort to obtain information about the unaltered charge distribution of the sampled aerosol. The possibility that this charge distribution undergoes changes during transport through the sampling and dilution system is discussed later. When using the scanning mobility particle sizer to measure charge distributions, either a positive or a negative classifying voltage is applied to the DMA when aerosol is sampled through the dummy charger. By using the UCPC to measure concentrations for a range of DMA classifying voltages, it is possible to infer the relative abundances of positively and negatively charged particles as a function of size. To obtain accurate size distribution data, it is necessary to know the size dependent particle losses in the sampling line (including the ejector) and the DMAs, the fraction of particles that are charged and the UCPC counting efficiency. In this work, particle transport losses by diffusion and inertial deposition were accounted for. (27) Both the regular DMA and the Nano-DMA operate at 1.5 slm aerosol flow and 15 slm sheath flow. The size-dependent particle losses inside the regular DMA are estimated by a 13-meter effective diffusion length, (28) and the losses in the Nano-DMA are estimated from the experimental data provided by Chen et al. (18) Since half of the sampling probe length is water cooled, there is a large temperature gradient inside the probe. Thermophoretic losses might be important. However, there is currently no model available to estimate the thermophoretic losses in such under-developed and high temperature gradient flows. Hence we do not account for thermophoretic losses for the data reported in this paper, Fortunately, as thermophoretic losses

7 System for In Situ Characterization of Nanoparticles 445 Fig. 3. Fuchs stationary charge distribution in N 2 at atmospheric pressure. The curves are obtained by assuming the listed properties of charging agents. (25,29) It approximates the charge distribution that particles acquire after passing through the bipolar charger. q is the number of charges on a particle, d p is the diameter of the particle, and M I+,M I,Z I+,Z I are mass and electrical mobility of positive and negative ions. are independent of size for the particle Knudsen numbers in our experiment, neglecting them does not skew the size distribution. Particles can be assumed to achieve a stationary charge distribution after passing through a bipolar charger. (25) The stationary charge distribution in nitrogen is shown in Fig. 3. (29) Note that the fractions of doubly charged particles are at least about one order of magnitude lower than the singly charged particles for particles smaller than 100 nm, which is approximately the size range of synthesized particles in our experiments. Therefore, the effect of multiply charged particles is not considered in the size distribution inversion in this paper. The counting efficiency of the UCPC was obtained by comparing concentrations of DMA classified monodisperse particles by the UCPC and by an electrometer (30) located in parallel. Particle transport efficiencies through the DMAs and the counting efficiency of the UCPC are shown in Fig. 4. Also shown is a curve of particle transport efficiency in the sampling line under our typical operating conditions (266 Pa background reactor chamber pressure, 1.5 slm aerosol flow through the laminar flowmeter and 38.5 slm clean nitrogen in the second stage dilution). A hypersonic impactor is used to collect representative samples for analysis by electron microscopy. (19) Under our experimental conditions, the pressure ratio across the 0.34 mm diameter orifice is about 390. Hence the sampled aerosol expands supersonically through the orifice. The particle laden hypersonic free jet impinges against the collection plate. The cut size

8 446 Wang et al. Fig. 4. The size dependent counting efficiency of the UCPC and particle transport efficiencies in the sampling line, the regular DMA and the Nano-DMA at typical operating conditions. of this impactor is about 6 nm for silicon particles when the orifice-to-plate distance is about 4 orifice diameters. (31) Figure 5(a) and (b) show pictures taken when we were synthesizing and analyzing titanium particles. Figure 5(a) shows an argon plasma before hydrogen and reactants were injected. Figure 5(b) shows a bow shock covering the sampling probe. The bow shock is made visible by photoemission from titanium particles. 3. RESULTS 3.1. Results of Particle Size Distribution Measurements The effects on particle size distributions of varying chamber pressure, reactant flow rates, and the method used to inject reactants are addressed in this section. Figure 6 shows silicon particle size distributions at different chamber pressures with the same reactant flow rate. The first point to note is that at normal operating chamber pressure (266 Pa), most particles are smaller than 10 nm. Second, it is clear that particles grow to larger sizes as the chamber pressure increases. Similar results were obtained for particles of other materials (Ti, Si Ti N, etc.). We believe that particle size increases with pressure because both particle concentrations and transport

9 System for In Situ Characterization of Nanoparticles 447 Fig. 5. Photographs of the plasma-expansion process: (a) Argon plasma before hydrogen and reactants were injected; (b) Particle sampling. The sampling probe is covered by a bow shock containing titanium particles. Fig. 6. Silicon particle size distributions at different chamber pressures with a constant reactant flow rate. n is the particle number concentration. times increase with increased pressure. Hence more particles coagulate at higher pressures, and particles become larger. Figure 7 shows the dependence of particle size distributions on reactant flow rates for a fixed chamber pressure. A background size

10 448 Wang et al. Fig. 7. Dependence of silicon particle size distributions on reactant (SiCl 4 ) flow rate at a fixed chamber pressure. distribution was measured before reactants were injected. Note that for this particular run, virtually no particles were detected when the SiCl 4 flow rate equaled 20 sccm. This is presumably because the reactant saturation ratio was too low for particles to nucleate. For 30 and 40 sccm SiCl 4, bimodal distributions were observed. Also note that the mode particle size increases as reactant flow rates increase. This observation is consistent with a previous aerosol simulation on a similar system, which shows that higher initial reactant concentrations result in larger particles due to faster condensation and coagulation. (32) In our system, we are able to inject different reactants separately or premixed into the reaction zone as shown in Fig. 8 for Si Ti N synthesis. Without premixing, the reactants are injected separately from two injection ports; with premixing, reactants are mixed before they are injected. As is seen in Fig. 9, the particle size distributions are not significantly different for these two cases. However, significant differences in film properties were found, as is shown in the X-ray diffraction (XRD) spectra in Fig. 10. Note that we obtained multiple silicide peaks in the film with premixed reactants, while these peaks were not observed in the unmixed films. It seems that when reactants are injected separately, they do not mix well in the reacting zone. Silicon and titanium particles nucleate separately before they react further with nitrogen. However, when SiCl 4 and TiCl 4

11 System for In Situ Characterization of Nanoparticles 449 Fig. 8. Schematic of different reactant injection methods: (a) Un-premixed; (b) premixed. Fig. 9. Si-Ti-N particle size distributions with reactants injected premixed or un-premixed. are premixed, they dissociate in the hot plasma gas and co-nucleate during expansion cooling to form silicide phases. The fact that the particle size distributions in the two cases are so similar is surprising, and currently unexplained Results of Particle Charge Distribution Measurements As was mentioned earlier, when particles pass through a bipolar charger, they achieve a stationary charge distribution at atmospheric pressure (Fig. 3). (25) This known charge fraction can be used to obtain the total particle (positive + negative + neutral) size distribution from the measured distributions of charged (positive or negative) particles. On the other hand, if the bipolar charger is bypassed, the SMPS measures the

12 450 Wang et al. Fig. 10. X-ray diffraction (XRD) spectra for unmixed and premixed Si-Ti-N films. charged particles directly from the reactor. The particle charge distribution (size dependent charge fraction) can be obtained by dividing the size distribution of these charged particles by the aforementioned total size distribution. Figure 11 shows the size distributions of total particles and particles that acquired charges within the reactor during titanium nanoparticle synthesis. The size dependent fractions of charged particles are shown in Fig. 12. Note that the positive and negative particles have almost identical size distributions in the size range we measured. Furthermore, the fractions of charged particles are very small, especially for smaller sizes. This indicates that most particles from the reactor are neutral for sizes less than 20 nm. Particles in a plasma gain charges by attachment of electrons and ions and by thermionic emission of electrons. When emission is not important, the equilibrium charge is negative due to the higher flux of electrons than positive ions. However, when electron emission is significant, the equilibrium charge may be positive. (33) Therefore, it is not surprising that our measured charge distributions have almost equal fractions of positive and negative particles. It is not yet clear which charging mechanisms

13 System for In Situ Characterization of Nanoparticles 451 Fig. 11. Size distributions of positively and negatively charged titanium particles directly from the reactor and total particle size distributions. Fig. 12. Fraction of particles that obtained +1 and 1 charge within the reactor. are important in our system. Furthermore, we have made the questionable assumption that the particle charge distribution does not change during transport. An equilibrium calculation shows that the concentrations of positive (Ti + ) and negative (electron) charge agents are on the order of cm 3, when 40 sccm TiCl 4 is disassociated in an Ar-H 2 plasma at

14 452 Wang et al K and 266 Pa. Assuming the electron and ion concentrations are frozen before the dilution in the ejector, the Nt value is ions/cm 3 sfrom the sampling inlet to the ejector dilution (with a residence time 10 ms). This Nt value might have been sufficiently high to affect the charge state of the sampled aerosol. Therefore, a detailed study of the role of possible charging mechanisms (diffusion, thermionic emission, secondary electron emission, etc.) and the electron-ion-particle interaction dynamics is still required to quantitatively explain the measurements. 4. SUMMARY A diagnostic system was developed to characterize nanosize particles synthesized with a thermal plasma process. This system can measure particle size and charge distributions in near real time from a low pressure and high temperature environment. It can also collect particles for subsequent microscopic analysis. The effects of chamber pressure ( Pa) and precursor flow rates (20 80 sccm) on particle size distributions were investigated. It was found that the particle mean size increased with the chamber pressure due to longer residence times for coagulation at higher chamber pressures. At lower chamber pressures (<1333 Pa), most particles were smaller than 10 nm. The mean size also increased with reactant flow rates. SMPS measurements for particles of both positive and negative polarity were carried out with and without the bipolar charger upstream of the DMA to determine the charge distribution of the sampled aerosol. It was found that while most sub-20 nm particles from the reactor were neutral, the populations of positively and negatively charged particles were almost the same. These results are of foremost interest in the modeling and understanding of nucleation and coagulation phenomena in our thermal plasma process. ACKNOWLEDGMENTS This work was supported by NSF (Grant No. DIM ). REFERENCES 1. G. S. Selwyn, Jpn. J. Appl. Phys. Part 1 32(6B), 3068 (1993). 2. Y. Watanabe and M. Shiratani, Jpn. J. Appl. Phys. Part 1 32(6B), 3074 (1993). 3. L. Boufendi, W. Stoffels, and E. Stoffels, in: Dusty Plasmas: Physics, Chemistry and Technological Impacts in Plasma Processing (A. Bouchoule, ed.). John Wiley & Sons Ltd, New York B. R. Forsyth and B. Y. H. Liu, Aerosol Sci. Technol. 36(5), 515 (2002). 5. B. R. Forsyth and B. Y. H. Liu, Aerosol Sci. Technol. 36(5), 526 (2002).

15 System for In Situ Characterization of Nanoparticles T. Seto, T. Nakamoto, K. Okada, M. Adachi, Y. Kuga, and K. Takeuchi, J. Aerosol Sci. 28(2), 193 (1997). 7. K. S. Seol, Y. Tsutatani, R. P. Camata, J. Yabumoto, S. Isornura, Y. Okada, K. Okuyama, and K. Takeuchi, J. Aerosol Sci. 31(12), 1389 (2000). 8. K. S. Seol, Y. Tsutatani, T. Fujimoto, Y. Okada, K. Takeuchi, and H. Nagamoto, J. Vac. Sci. Technol. B. (19(5)), 1998 (2001). 9. T. Makino, N. Suzuki, Y. Yamada, T. Yoshida, T. Seto, and N. Aya, Appl. Phys. A (69 [Suppl.]), S243 (1999). 10. P. J. Ziemann, P. Liu, D. B. Kittelson, and P. H. McMurry, J. Aerosol Sci. 26(5), 745 (1995). 11. P. J. Ziemann, P. Liu, S. Nijhawan, D. B. Kittelson, P. H. McMurry, and S. A. Campbell, 41st Annual Technical Meeting of the Institute of Environmental Sciences, Anaheim, CA, 1995n. 12. S. Nijhawan, P. H. McMurry, M. T. Swihart, S.-M. Suh, S. L. Girshick, S. A. Campbell, and J. E. Brockmann, J. Aerosol Sci. 34, 691 (2003). 13. S. C. Wang and R. C. Flagan, Aerosol Sci. Technol. 13, 230 (1990). 14. N. P. Rao, S. L. Girshick, J. V. R. Heberlein, P. H. McMurry, S. Jones, D. Hansen, and B. Micheel, Plasma Chem. Plasma Process. 15(4), 581 (1995). 15. N. P. Rao, N. Tymiak, J. Blum, A. Neumann, H. J. Lee, S. L. Girshick, P. H. McMurry, and J. V. R. Heberlein, J. Aerosol Sci. 29 (5/6), 707 (1998). 16. E. O. Knutson and K. T. Whitby, J. Aerosol Sci. 6, 443 (1975). 17. M. R. Stolzenburg and P. H. McMurry, Aerosol Sci. Technology 14, 48 (1991). 18. D.-R. Chen, D. Y. H. Pui, D. Hummes, H. Fissan, F. Quant, and G. Sem, J. Aerosol Sci. 29(5/6), 497 (1998). 19. J. Fernandez de la Mora, S. V. Hering, N. P. Rao, and P. H. McMurry, J. Aerosol Sci. 21(2), 169 (1990). 20. F. Di Fonzo, A. Gidwani, M. H. Fan, A. Neumann, D. I. Iordanoglou, J. V. R. Heberlein, P. H. McMurry, S. L. Girshick, N. Tymiak, W. W. Gerberich, and N. P. Rao, Appl. Phys. Lett. 77(6), 910 (2000). 21. S. L. Girshick, J. V. R. Heberlein, P. H. McMurry, W. W. Gerberich, D. I. Iordanoglou, N. P. Rao, A. Gidwani, N. Tymiak, F. D. Fonzo, M. H. Fan, and D. Neumann, in: Innovative Processing of Films and Nanocrystalline Powders (K.-L. Choy, ed.) Imperial College Press, London, A. Gidwani, Ph.D. Thesis, Department of Mechanical Engineering, University of Minnesota, Minneapolis, 55455, (2003). 23. P. Liu, P. J. Ziemann, D. B. Kittelson, and P. H. McMurry, Aerosol Sci. Technol. 22(3), 293 (1995a). 24. P. Liu, P. J. Ziemann, D. B. Kittelson, and P. H. McMurry, Aerosol Sci. Technol. 22(3), 314 (1995b). 25. N. A. Fuchs, Geofisica pura e applicata. 56, p. 185 (1963). 26. B. Y. H. Liu and D. Y. H. Pui, J. Aerosol Sci. 5, 465 (1974). 27. W. C. Hinds, Aerosol Technology, Properties, Behavior, and Measurement of Airborne Particles. John Wiley & Sons Inc, New York, A. Reineking and J. Porstendorfer, Aerosol Sci. Technol. 5, 483 (1986). 29. A. Wiedensohler and H. J. Fissan, Aerosol Sci. Technol. 14, p. 358 (1991). 30. B. Y. H. Liu and D. Y. H. Pui, J. Aerosol Sci. 6, 249 (1975). 31. N. P. Rao, B. Micheel, D. Hansen, C. Fandrey, M. Bench, S. L. Girshick, J. V. R. Heberlein, and P. H. McMurry, J. Mater. Res. 10(8), 2073 (1995). 32. S. L. Girshick and C. P. Chiu, Plasma Chem. Plasma Process. 9(3) (1989). 33. J. Goree, Plasma Sources Sci. Technol. 3, 400 (1994).