Table of contents

Volume 3

Number 3, August 1994

Previous issue Next issue

(CHATEAU DE BONAS, CASTERA-VERDUZAN, FRANCE, 30 AUGUST TO 3 SEPTEMBER 1993)

239

, , and

A review is presented of the phenomena associated with particles in low pressure plasmas. Dust particles which are typically micrometers in diameter have been observed by laser light scattering in various low-pressure, radiofrequency-excited plasmas. Experiments have been designed so that the origin of the dust material is unambiguous and, to some extent, quantitative. The processes involved in the appearance of the mesoscopic dust particles are outlined and compared with our experimental observations. The source material and its required generation rate, nucleation, charging, growth mechanisms, growth rates, and saturation mechanisms are discussed. The mutual influences of dust and plasma, particularly the role of geometric and circuit boundary conditions in laboratory plasmas, are described.

246

, , , , and

The growth of particles in a radiofrequency (RF) (13.56 MHz) plasma at pressures from 25 to 200 mTorr in mixtures of CF4, CF2Cl2 and argon has been studied experimentally. A planar configuration was used, with a silicon wafer on the powered electrode. The electron density has been measured with microwave resonance spectroscopy using a cylindrical cavity surrounding the plasma. The same geometry has been used to measure the density of various species of negative ions by detecting the extra electrons created by laser-induced photo detachment. It appears that the negative ion density is much larger in the case of CF2Cl2 than in the case of CF4. There seems to be hardly any dependence of the negative ion concentration on the CF2Cl2/Ar partial pressure ratio in the range where powder growth occurs. However, the attachment rate to chlorine is found to be much higher than to fluorine. Furthermore the gas phase discharge chemistry has been studied using infrared absorption spectroscopy. Both a tunable diode laser system, and a Fourier transform spectrometer have been applied. The CF2 concentration appears to decrease strongly when powder growth occurs. The SiF4 concentration then has a maximum. The results indicate that the presence of chlorine in the plasma feed gas is essential. In CF4 no particle formation is detected. The wafer surface is blackened during powder formation. SEM inspection indicates that this is caused by micromasking. Considering all the information, we arrive at the conclusion that particle growth is initiated by micromasking at the Si surface combined with a highly directional etching process. Due to residual isotropic etching the particles are released from the surface and enter the plasma, where they start coalescing and growing under the influence of CF2 polymerization.

252

, , and

After having distinguished five successive steps in the temporal evolution of a powder-forming SiH4 radiofrequency glow discharge, we examine the initial mechanism by which silicon clusters start growing up to the point where they suddenly aggregate into multiply charged particles and modify the discharge regime. This 'induction' period can be much longer than the diffusion time of positive ions and neutral radicals, which implies that cluster growth involves negative ions (anions). We provide a review of basic data concerning anions in SiH4 plasmas and analyse mass spectrometric data showing that anion-molecule reactions SinH2n+1-+SiH4 to Sin+1H2n+3-+H2 at relatively low rate (about 10-12 cm3 s-1), and fast exothermic anion-radical reactions SinHm-+SiHm' to Sin+1Hm+m'-2q-+qH2 at Langevin rates (about 10-9 cm3 s-1), initiate clustering. The effective anion lifetime involves a succession of dissociative attachment to SiH4, detachment or recombination, and attachment to neutral radicals or clusters competing with diffusion out of the plasma. Anion-molecule and anion-radical cluster reactions at Langevin rates probably dominate the cluster growth kinetics below 100 Si atoms whereas anion-neutral and neutral-neutral condensation at size-scaling collision rates govern the subsequent homogenous nucleation regime. At the end of the nucleation period (up to 104 Si atoms) the fraction of singly charged clusters can reach 50%. The reduction of powder formation upon gas heating is attributed to a decrease of the rate of non-dissociative attachment to radicals and neutral clusters.

262

and

The growth of particle size has been measured in a low-pressure argon-silane plasma using high-resolution transmission electronic microscopy. The results show that formation and growth of dust particles is an homogeneous process; the first generation size distribution is monodispersed; and the growth kinetics reveals a three-step process from molecular ions to large particles. Together with measurements of particle concentration obtained by laser light scattering, these measurements give a clear indication that the growth proceeds through three successive steps: (i) 'rapid' formation of crystalline clusters (as shown by dark-field high-resolution transmission electron microscopy) with concentrations of up to 1010 cm-3; (ii) formation of aggregates, of diameters up to 50 nm, by coagulation (during coagulation the particle concentration decreases dramatically); and (iii) growth of the particles with a constant concentration by surface deposition of SiHx radicals, whilst the numerical density remains constant. Laser-induced particle explosive evaporation has been performed using a XeCl (308 nm) laser. This experiment allowed detection of nanocrystallites and also the beginning of their coagulation and gave clear evidence of the temperature effect on particle formation.

268

and

The production, suspension and transport of fluorocarbon particulates in capacitively coupled radiofrequency discharges are studied using in situ laser light scattering and ex situ chemical analysis. The time evolution of the spatial distribution of suspended particles is obtained by two-dimensional imaging of the scattered light. The chemistry of the discharge is varied by use of a range of pure fluorocarbon gases and mixtures with argon, oxygen and hydrogen-containing molecules. Addition of hydrogen to a fluorocarbon discharge increases the rate of formation of particles although these powders are found by Fourier transform infrared measurements to contain negligible amounts of hydrogen. Particle formation rates correlate with polymer deposition rates and are independent of apparatus history. It is proposed that this is a clear example of gas phase rather than surface processes leading to particle nucleation and growth.

273

and

Particle formation has been studied in Ar sputtering plasmas and CCl2F2-Ar reactive ion etching plasmas from Si and SiO2 substrates by laser light scattering, scanning and transmission electron microscopy and optical emission plasma diagnostics. Particles nucleate and grow continuously, and are swept out into the exhaust under a wide variety of plasma conditions. Within a more limited range of values for pressure and flow rate, particles grow large enough in the plasma so as to form a particle cloud suspended above the substrate. Stability and position of the cloud depend on the process conditions. When particles become visible by light scattering, at a diameter of about 200 nm, they are essentially spherical and monodisperse in size, but the size distribution becomes much wider as the average particle size increases with increasing discharge time. In sputtering, particles smaller than about 100 nm are quite porous, have a somewhat more irregular shape and exhibit a spherulitic (spherically columnar) mode of growth. Optical emission spectroscopy of the plasma and compositional analysis of the particles indicate that in all cases, Si atoms are responsible for particle nucleation and growth. These observations are discussed in terms of possible mechanisms for generation and transport of plasma species in the discharge, particle nucleation and growth, particle transport and particle cloud formation.

278

, , , and

Experimental identification of the precursors and processes leading to particles is essential for understanding particulate contamination in deposition plasmas. We have investigated particle formation in radiofrequency silane plasmas using light scattering (elastic and inelastic) and quadrupole ion mass spectrometry as complementary plasma diagnostics. Negative ions reach high masses (at least 500 amu) and are the only elementary species with a residence time on the scale of the powder formation time. Furthermore, a negative-ion polymerization scheme shows that the densities of high-mass anions are strongly diminished at kilohertz power modulation frequencies, at which reduced powder production is also observed. We conclude that negative ions are the particle precursors and that initial clusters grow by negative-ion polymerization in silane plasmas. In situ light scattering techniques are described to determine particle size, number density and refractive index self-consistently. Novel, visible photoluminescence measurements from particles suspended in the plasma are also reported. These diagnostics demonstrate that particle evolution proceeds by an agglomeration phase and that the particle properties are different from those of the bulk material early in particle development.

286

and

Formation processes of particulates in radiofrequency power-modulated SiH4 discharges have been studied using two kinds of laser light scattering, a modified Langmuir probe and absorption methods. The results suggest that particulates are formed by three phases of nucleation, rapid growth and growth saturation. In the nucleation phase, the formation may be caused by short-lifetime radicals such as SiH2, SiH or Si. In the rapid growth phase, SiHn+ (n=0-3) ions and/or SiH3 radicals seem to make an important contribution to particulate growth, if only direct influx of the particles onto a particulate is taken into account. In the growth saturation phase, the saturation may be explained by taking into account the decrease in enhancement factor of the ion-collecting areas of particulates and the increase in their loss due to the reduction of ambipolar field in the periphery of the discharge column. Power modulation in SiH4 radiofrequency discharges is found to be practically very effective for decreasing the sizes of particulates.

292

and

Dust generation in plasma reactors used for PECVD is a general limiting effect which occurs when trying to obtain high deposition rates in the fabrication of thin films. In such dust-forming processes, for instance silane discharges, very high concentrations of submicrometre sized particulates are readily produced. The theoretical approach and the modelling of these dusty dense plasmas suggest that they have very peculiar properties with spectacular effects concerning the plasma equilibrium and the behaviour of the particulate cloud. Well characterized dusty dense plasma situations have been obtained in argon-silane or in pure argon RF discharges and experimental data obtained in these situations are reported here, in connection with the theoretical predictions. In terms of plasma properties the drastic modification of the free electron population, induced by the presence of the particles, is one of the most important results, with significant effects on the chemical equilibrium of the plasma. In terms of the particle cloud behaviour the strong electrostatic interaction between the negatively charged particulates is shown to be an order of magnitude higher than their kinetic energy and this particle cloud has to be described as a 'Coulomb liquid'. An overview of our experimental studies of these effects in a dusty dense plasma situation is given, including the most recent results.

302

, , and

Particulate generation has been studied during reactive-ion etching of oxide wafers in C2F6-CHF3 and CF4-CHF3 plasmas using both a commercial etch tool and the GEC reference cell modified to resemble the commercial tool. Under certain discharge process conditions, copious amounts of submicrometre-sized particles are shown to form due to plasma interactions with the oxide substrate. In the commercial tool, particles were detected only by a downstream particle flux monitor, whereas in the reference cell, particles were observed by both in situ laser light scattering and downstream monitoring. In the commercial tool, wafers etched to end-point were shown by post-process surface analysis to be contaminated by submicrometre-sized columnar structures. Previous reports of similar such columnar structures formed during reactive-ion etching of oxide films have attributed the phenomenon to polymer micromasking. However, the results of this study clearly contradict this conclusion and suggest that the presence of columnar oxide etch residues is linked to process-induced particulate contamination. Laser light scattering measurements were made in the reference cell during reactive-ion etching of blanket oxide wafers and used to help clarify the complex processes of particulate nucleation, growth and deposition during oxide etching. Polarization coagulation of spherical particles formed in the reference cell is shown to occur, presumably in the high-field regions of the sheath, forming filamentous rod-like particle aggregates. The implications of this observation for wafer contamination are explored.

310

, , and

Particle growth and subsequent extinction in plasma reactors for surface processing has been investigated in a parallel-plate radiofrequency discharge at 13.56 MHz in Ar with (-CF2CF2-)n and its fragments as an impurity. The space- and time-variation of the radiofrequency plasma structure is mainly studied simultaneously with the particle size and density by both spatiotemporally resolved optical emission spectroscopy and Mie scattering of laser light. A correlation is demonstrated between the particle size/density and the radiofrequency plasma structure.

314

, , and

The characteristics of plasma parameters and their spatial structures in a parallel-plates RF discharge in argon modified by the injection of small carbon particles have been investigated. A larger peak-to-peak fluctuation of space potential in the bulk plasma region was observed by an emissive probe measurement together with a decrease in the self-bias voltage. The electron energy distribution function in the plasma bulk as measured by an energy analyser also showed the presence of a higher energy tail. The enhancement of an argon atomic line emission in the midgap was consistent with the above results, showing the transition of the discharge mode from the gamma to alpha regimes. Spatial and temporal behaviours of particle density and size were measured by a newly proposed ellipsometric detection method of the laser Mie scattered light. Slow shift of the peak position of density and size distributions towards the sheath edge of the powered electrode was observed.

320

, , , , and

In situ Fourier transform infrared spectroscopy has been used to study particulate formation in a CCl2F2/Ar RF discharge. Strong absorption bands at 1000-1100 cm-1 have been found and attributed to C-F and Si-F absorption. Furthermore continuous extinction due to Rayleigh and Mie scattering has been observed. The relative intensities of C-F, Si-F and scattering signals vary with plasma conditions. There are several experimental indications that the clusters are formed on the surface and ejected into the plasma. An SEM study of the substrate surface has allowed us to establish the mechanism for the particulate production in this discharge.

325

, , , , and

Sources of particles in a closed-coupled electron cyclotron resonance plasma source used for polysilicon etch included flaking of a residual film deposited on chamber surfaces and shedding of material from the electrostatic wafer chuck. A large, episodic increase in the number of particles added to a wafer in a clean system is observed more frequently for a plasma-on than for a gas-only source condition. For film-forming process conditions, particles were added to wafers by a residual film, which was observed to fracture and flake away from chamber surfaces. The presence of a plasma, especially when radiofrequency bias is applied to the wafer, caused more particles to be ejected from the walls and added to wafers than the gas-only condition; however, no significant influence was observed with different microwave powers. A study of the effect of electrode temperatures on particles added showed that thermophoretic and gravitational forces are not significant for this electron cyclotron resonance operating configuration. Particles originating from the electrostatic chuck were observed to be deposited on wafers in much larger numbers in the presence of the plasma as compared with gas-only conditions, implying the existence of a large ion drag force.

334

, , , and

This paper studies asymmetric particle clouds confined within electrostatic particle traps in the presence of a molecular drag force. The drag force, due to flow of neutral Ar gas around a particle, is proportional to the gas flow, which can be controlled by a mass flow controller. Our system has a component of the drag force parallel to the wafer electrode surface. The particle clouds associated with a cube and a disc are seen to be highly asymmetric in the direction of the drag force, as would be expected. The asymmetry increases as gas flow increases. In order to visualize the clouds quantitatively, we have introduced a technique called spot scanning in which we move the laser beam slowly with a serpentine path through the particle cloud. The forward scattered light is recorded by a charge-coupled device camera onto videotape. The image of the scattered light on a frame of the videotape is a nearly circular red spot. Using commercially available computer software, the red spots can be used to determine the boundary of the cloud as well as to find contours of constant power density scattered from the cloud.

340

Particles or 'dust' in etching or deposition plasmas are an important cause of product yield loss and equipment down-time. Traditional methods of particle control are only partially effective in plasma processing. This is because formation and transport of particles are strongly influenced by plasma electrical and chemical properties. Particle control in plasma processing requires understanding of these effects and their relation to aspects of tool and process design. Laser light scattering has been used to monitor the behaviour of particles for a wide range of plasma tools. This method provides information on location and transport of particles. In some variations, light scattering may also be used for particle size determination. Results have been obtained in sputter, etch and deposition tools of planar diode and magnetron-enhanced designs, Some results have also been obtained in electron cyclotron resonance tools and radiofrequency inductive tools. From this database, differences and common elements are observed for the behaviour of particles. The particle trapping phenomenon is often observed. Particle traps have an important bearing on wafer contamination. Traps cause particles to accumulate into localized regions during plasma operation, only to be suddenly released at the end of the process, thereby contaminating the wafer. Spatially resolved optical emission may be used to map the location and intensity of particle traps. This method also provides a semi-quantitative comparison with two-dimensional modelling studies. It may also be used to optimize grooved electrode design for particle contamination control.

348

, , , , and

The effects of plasma processing conditions on the microstructural properties of silicon powders are presented. Hydrogenated nanophase silicon powders were prepared using low-pressure and low-temperature square wave modulated RF plasma (13.56 MHz) using pure silane gas. Plasma parameters such as pressure, RF power, plasma modulation frequency, and gas flow rate were varied. In situ analysis by quadrupolar mass spectroscopy and ex situ analysis of the silicon powders by Fourier transform infrared spectroscopy (FTIR) and thermal desorption spectrometry of hydrogen were performed. The thermal desorption spectrometry results show the fundamental differences between the concentrations of hydrogen weakly and strongly bonded in silicon powders as compared to amorphous silicon films. The FTIR analysis also determined the microstructural characteristics of powders and hence their volume/surface ratio. This parameter was determined from the balance of Pj probabilities of having one of the Hj-Si-Si4-j bond arrangements in the powder particles. These results reveal an increase in hydrogen content and a reduction in volume/surface ratio as the modulation frequency of RF power increases. In consequence, higher compactness of silicon powders is associated with long particle residence times inside the plasma as a result of ion bombardment. TEM analysis indicated a considerable dispersion of particle size and some degree of structure of the silicon powder characterized by intergrain linkage. We point out the dominant presence of hydrogen on the particle surfaces (external voids), which may cause the high reactivity of grains, increasing the degree of intergrain linkage.

355

, , and

The effects of particles on helium-diluted silane RF discharges are studied using a power modulation method for various Values of relevant parameters. Compared with CW discharge cases, close correlation is clearly found between the particle growth, the self-bias voltage and the phase shift between the current and voltage of the RF discharge. Total particle number and optical emission intensity in the bulk plasma region increase monotonically after RF power-on. With these increases, the magnitude of self-bias voltage and the current-voltage phase shift decrease considerably to their minimum values and then slightly increase to their quasi-steady values. The decreases can be explained by the fact that particles in plasmas behave as very heavy negative ions. The increases may be related to the increase in the diffusion rate of electrons. It is also confirmed through this study that the modulation is very effective in the suppression of particle growth. In particular, for a duty cycle of 20%, the discharge parameters are close to those for a pure He discharge. This tendency is consistent with the result that no particles can be observed for this duty cycle.

359

, , , and

In order to avoid particle contamination of substrates during the plasma-enhanced chemical vapour deposition of amorphous silicon, we have investigated the dependence on temperature and gas flow conditions of the dynamics of submicrometre particles in discharges in silane. We report on light scattering experiments in an RF-powered plasma reactor with heatable parallel electrodes. The motion and trapping of particles was found to be strongly influenced by the gas flow conditions and the temperature gradient in the discharge. To understand this behaviour the equation of motion of a particle was solved in one dimension including gravitation, electric force on a 'dressed' particle, friction force and thermophoresis. The validity of the friction force is verified by observing experimentally the sinking of particles without electric force after switching off the discharge. Results for calculated particle positions and velocities as a function of time are presented for the cases of particles with radii of 0.1 and 0.15 mu m. When the discharge is on, the light particle is trapped at the sheath boundary. In the case of a temperature gradient (heating the lower electrode to 80 degrees C) the lighter particles are driven upward by thermophoresis and may be trapped at the sheath edge. Finally, a strong downward-directed gas flow (feeding gas through the upper electrode) of 30 cm s-1 prohibits the trapping of particles at the lower sheath boundary in agreement with the experimental observation. Electrode heating, together with proper gas flow, is found in our experiments to completely avoid the trapping of particles in the interelectrode space.

369

, , , , , , and

A particle analyser is described that simultaneously detects and characterizes <0.02 to >10 mu m diameter particles independent of particle composition in real time. No previous instrument has been able to perform these functions simultaneously. Our design uses pulsed laser ablation of particles followed by time-of-flight mass spectrometric analysis of the resulting ions. The ion intensity is related to particle size at least for small particles. Thus particle size information is obtained concurrently with the other information.

373

and

Potential particle-forming reactions of hydrogenated silicon cations with silane and disilane, in the presence and absence of water, are summarized. The reactions are studied in the ion trap of a Fourier transform mass spectrometer at 10-7-10-5 Torr. Cations react sequentially with silane to produce larger silicon-containing ions. In general, reaction rates decrease with increasing number of silicon atoms in the ion; no ions containing more than six silicon atoms are formed in reactions with pure silane. Reaction with disilane produces larger silicon-containing ions; reaction rates decrease with increasing number of silicon atoms in the cation. The largest ion produced is larger than for the silane system but still contains less than nine silicon atoms. There is insufficient reactivity for hydrogenated silicon cations in pure silane or disilane to produce macroscopic particles. Mixtures of silane or disilane containing 7% water, however, significantly enhance the growth of ions. In the limit of our measurements, the reactions in both silane and disilane do not terminate.

381

, , , and

Microscopic cauliflowers have been observed in a surprisingly diverse range of dusty plasmas. Their microstructure, as analysed by electron microscopy, is consistent with growth by ballistic deposition rather than diffusion limited aggregation. The morphology of the grains supports the inference from dust growth kinetics that they form by accretion of positive ions rather than neutral radicals. The dense, amorphous structure is capped by a fractal surface whose texture is concisely described by a recursion based on the modified midpoint method. The surface texture may be reconstructed by ion bombardment, providing a quantitative link between growth kinetics and roughness through a Mullins-Sekerka stability analysis of Laplacian growth.

388

Particle nucleation in a thermal plasma reactor occurs as a high-temperature gas undergoes a cooling trajectory. Cooling leads to formation of supersaturated vapours, which causes either homogeneous or ion-induced nucleation. Detailed models have been developed for homogeneous nucleation in a plasma reactor, including discrete-sectional models and moment-type models. The discrete-sectional models are capable of incorporating size-dependent cluster property data as these become available. Calculations in which a one-dimensional cooling trajectory was assumed in a discrete-sectional code indicate that cooling rates play a key role in determining the final particle size. Moment-type models are more computationally efficient, and have been coupled to two-dimensional reactor transport models. A two-dimensional model was compared with experimental results for synthesis of iron particles over a range of conditions in which the volume-mean particle size ranged from roughly 20 nm to 70 nm, with reasonable quantitative agreement for particle size versus reactant feed rate. The major weakness of current models for particle formation in thermal plasmas is the lack of an adequate understanding of ion-induced nucleation. Additionally, there is considerable need for well-characterized experiments in which particle sizes are determined by probe sampling or light scattering.

395

A negative ion plasma is produced by introducing a small amount of SF6 gas into a low-temperature ( approximately=0.2 eV) potassium plasma produced in a Q machine. The density ratio of negative to positive ions is continuously varied in the range up to more than 0.9999, where there appears a remarkable decrease in electron shielding for potential variations, yielding a clear effect on plasma collective phenomena. By introducing fullerene (C60)particles into the Q-machine plasma, we can produce a plasma including large negative C60 ions. This ultrafine particle plasma might prove very attractive in fields of materials science.

400

Several models that predict the charge of particles in a plasma are reviewed. The simplest is based on orbit-limited probe theory. This basic model can be improved by adding several effects: charge reduction at high dust densities, electron emission, ion trapping and fluctuations. The charge is reduced at high dust densities, when a significant fraction of the charge in the plasma resides on the particles, depleting the plasma. Electron emission due to electron impact or ultraviolet exposure can cause a particle to have a positive charge, which has useful implications for plasma processing, since particles are confined in a discharge only if they have a negative charge. Ion trapping occurs due to ion-neutral collisions within the attractive Debye sphere of a negatively charged particle. Trapped ions reduce the net electric force on a particle. A particle's charge fluctuates because the currents collected from the plasma consist of discrete charges arriving at the particle at random intervals. The root mean square fractional fluctuation level varies as 0.5(N)- 12 / where (N)=(Q)/e is the mean number of electron charges on the particle.

407

, and

We present some results from numerical, fluid and particle models of non-thermal low pressure RF discharges contaminated by dust particles. These models have been used (1) to show the effect of the presence of dust particles on the electrical characteristics of an RF discharge, (2) to study the charge and floating potential of dust particles in a low pressure plasma from situations where the particles can be considered isolated to situations where they interact electrostatically and (3) to study the effect of the electrode geometry on the spatial distribution of dust particles in an RF discharge. The results confirm the existence of electrostatic traps close to the plasma-sheath boundary whose shape is very sensitive to the electrode configuration.

418

, , and

The transport of particles ('dust') in low-pressure electrical glow discharges is of interest with respect to contamination of semiconductor wafers during plasma etching and deposition. The distribution of dust particles in these reactors is determined by a variety of forces, the most important being electrostatic, viscous ion drag, gravitational, thermophoretic and neutral fluid drag. In this paper we present results from a series of computer models to predict the spatial distribution of dust particles in capacitively coupled electrical glow discharges considering these forces. The results are parametrized over power deposition, gas flow and particle size. We find that the spatial distribution of dust depends on the spatial dependence of the sheaths and plasma potential in bulk plasma which in turn depend upon the electrical topography of the surfaces. Experimentally observed 'dome' and 'ring' distributions of dust particles are computationally reproduced for specific combinations of discharge power particle size and substrate topography.

426

and

We review our recent efforts to develop and apply computational models that can predict fluid, thermal and particle transport in semiconductor process equipment such as that used for chemical vapour deposition or plasma etching. The purpose of this work is to supply equipment designers and operators with models that allow them to optimize process conditions and to develop tool designs that reduce particle contamination levels. The algorithms for predicting particle transport are briefly described. A Lagrangian approach is used in this work when both particle inertia and applied forces are important, while a Eulerian approach is used when both particle Brownian motion and applied forces are important. As an example, a commercial finite-element code is used to calculate the fluid and thermal fields in a simple geometry representative of real single-wafer processing tools, namely axisymmetric flow between a showerhead and a parallel plate separated by a small gap. Using the calculated velocity field, both Lagrangian and Eulerian particle transport formulations give the same particle collection efficiency for terminal-velocity-dominated deposition when particle inertia can be neglected. Although plasma-induced forces on the particles are not treated in detail, we discuss how models for these forces can be incorporated into the Lagrangian and Eulerian framework as they become available.

433

, , and

Once particles are formed or injected into plasmas used for materials processing, such as in plasma etching, plasma-assisted chemical vapour deposition or sputtering plasma systems, the nature of particle transport will largely determine whether a processing surface will be contaminated. We investigate the situation in which the particle density is low enough to ignore particle-particle and particle-plasma interactions. Emphasis is placed on obtaining expressions for the forces experienced by particles. These expressions depend on the local plasma condition: plasma density, electron temperature, positive ion directed and random kinetic energies, electric field and ion mass. We apply a model of an electron cyclotron resonance discharge to prediction of the existence and nature of particle trapping. Model predictions indicate that a high-density source such as an electron cyclotron resonance discharge is unlikely to trap particles mainly because of the large ion drag force sweeping particles out of the discharge. Finally, we present a model of particle heating in discharges. Under typical radiofrequency discharge conditions, particles are generally predicted to be near the neutral gas temperature in the discharge. We have conducted experiments and found results in agreement with these predictions. However, under conditions typically encountered in high-density plasma sources such as an electron cyclotron resonance source, the model predicts that particles may be heated to temperatures of two to three times room temperature.

442

, , and

Spatially resolved optical emission from an argon discharge is used to detect regions of enhanced emission over a grooved electrode designed to trap and channel particles. A groove extends the entire length of the electrode and is aligned with the optical detection axis. Thus we measure the integrated line-of-sight emission inside, above and next to the groove. Enhanced emission is seen and shown to be pressure-dependent for a given groove design. At low pressure (<100 mTorr) a single 'bright' spot is noted above the centre of the groove. This spot splits into two with increasing separation as the pressure increases. Laser light scattering detection of suspended particles shows correlated splitting of a single trapping region at low pressure into two traps at higher pressure. A two-dimensional radiofrequency discharge model is applied to the grooved electrode. The model consists of solving the electron, ion and continuity equations, the electron energy balance and Poisson's equation over the two-dimensional domain. The drift-diffusion approximation is used for electron and ion fluxes. Model results of the ionization rate are in reasonable agreement with experimental measurements. The resulting potential profiles from the model solution may be used to analyse particle trap locations.

448

, , , and

One observes in radiofrequency-heated vacuum chambers that dust, if present or being produced within the chamber, may float in layers close to both the upper and lower electrodes. Important forces on the dust are the electric force, gravity, plasma drag and the thermophoretic force, which is caused by temperature gradients in the background neutral gas in the vacuum chamber. We here discuss the thermophoretic force and show that the normally adopted formula for this force, which is computed on the assumption of an infinite plasma in all directions from the dust, must be modified when close to the plasma chamber walls. Taking into account the closeness of the plasma walls, we find that the thermophoretic force will be reduced out to many neutral gas molecular collision lengths from the wall, compared with the results from the standard formula. This modification of the thermophoretic force should be of importance for the force equilibrium and stability of dust in the dust layers observed in so many dust levitation experiments.