Table of contents

Various microwave-sustained, atmospheric-pressure plasma torches have been developed, investigated and applied over the last few decades. To avoid some of their shortcomings, we have designed a novel torch termed TIAGO (Torche à Injection Axiale sur Guide d'Ondes, in French). Its main advantages are simplicity, smooth impedance matching (low sensitivity to changes in operating conditions) and a short gas channel to prevent vapour condensation. Furthermore, it is possible to arrange TIAGOs in arrays to form a compact torch system which can be supplied, with equal distribution of power between plasma flames, from a single waveguide. This unique feature makes the new device particularly suitable when high gas throughputs or sequential processing are required. We describe the design, electrodynamic characteristics and experimental investigation of various torch arrangements based on the TIAGO principle, operated at 2.45 GHz with powers of a few hundred watts up to 2-3 kW per nozzle.

The scaling laws of pulsed plasma thrusters operating in the predominantly electromagnetic acceleration mode (EM-PPT) are investigated theoretically and experimentally using gas-fed pulsed plasma thrusters. A fundamental characteristic velocity that depends on the inductance per unit length and the square root of the capacitance to the initial inductance ratio is identified. An analytical model of the discharge current predicts scaling laws in which the propulsive efficiency is proportional to the EM-PPT performance scaling number, defined here as the ratio of the exhaust velocity to the EM-PPT characteristic velocity. The importance of the effective plasma resistance in improving the propulsive performance is shown. To test the validity of the predicted scaling relations, the performance of two gas-fed pulsed plasma thruster designs (one with coaxial electrodes and the other with parallel-plate electrodes), was measured under 70 different operating conditions using an argon plasma. The measurements demonstrate that the impulse bit scales linearly with the integral of the square of the discharge current as expected for an electromagnetic accelerator. The measured performance scaling is shown to be in good agreement with the theoretically predicted scaling. Normalizing the exhaust velocity and the impulse-to-energy ratio by the EM-PPT characteristic velocity collapses almost all the measured data onto single curves that uphold the general validity of these scaling laws. [12pt]This paper is dedicated to the memory of Dr Daniel Birx

Concentrations of NO, NO₂, NO₃, N₂O₅, and O₃ were measured by classical absorption spectroscopy in dielectric barrier discharges in flowing O₂/NO_x and N₂/O₂/NO_x mixtures. The results of measurements in different parts of the discharge chamber and in its exhaust are compared to a numerical zero-dimensional kinetic model and good agreement is found. The experimentally found upper limit of the NO_x concentration allowing ozone production is confirmed by the kinetic calculations for both gas mixtures. The rotational temperature of different nitrogen bands was measured by high-resolution emission spectroscopy. The results are explained on the basis of a simplified model and related to the gas temperature in the microdischarge channel and the surrounding gas.

Helicon plasmas were produced using a very high radio frequency (RF) (98 MHz) source which is commonly used for radio communication. As in the plasmas generated at lower frequency such as 13.56 MHz, E-H-W mode transitions were observed. In the helicon mode, a bright column was produced at the central part of the cylindrical plasma. Experiments using Nagoya type III antennae with different leg lengths showed that the antenna having the shortest legs produced the highest density, which suggests the importance of the antenna ring section. A study of the operation window of the helicon plasma at 98 MHz demonstrated that the H-W mode transition occurs at lower magnetic field at higher gas pressure with the RF power fixed. It was also observed that the transition arises at lower magnetic field at higher RF power with the gas pressure fixed, indicating the correlation between the three important operating parameters on helicon plasma generation and sustainment.

The substrate temperature was measured during plasma immersion ion implantation (PIII) of nitrogen as a function of the pulse bias voltage and the repetition frequency. The variation of the equilibrium frequency for temperatures between 150 and 500 °C and pulse voltages between -5 and -30 kV was investigated. Using these data, the relative dose per pulse for different voltages was obtained and its voltage dependence compared with different models for the plasma sheath expansion during PIII. A higher plasma density than measured for a static plasma without pulses, due to the interaction of secondary electrons with the plasma, must be assumed. Good agreement with dose measurements of N implanted in Si was also observed. For high pulse frequencies above 1 kHz a deviation was observed, clearly showing that depletion of the ions from the plasma during the pulses leads to reduced average plasma density at high repetition rates.

A model for a finite length cylindrical electronegative (EN) plasma is constructed by obtaining an approximate solution when negative ions exist over the entire plasma, but with ion flux to the walls dominating internal recombination flux. The profiles obtained in this manner are useful for making calculations even when the recombination flux competes with the diffusion flux for ion loss. The result is that there is an approximately parabolic profile in the smaller dimension (radius or axial length) with a flat central region and a similar edge profile in the larger dimension. The results are extended to include plasmas in which the electronegative region does not occupy the entire cylinder by introducing a single length scale that defines a smaller EN cylinder embedded in the larger plasma. Single types of positively and negatively charged ions and electrons are assumed to be the principal charged species. The plasma thus formed is embedded in a full model of an attaching feedstock gas with all relevant species and reaction rates included. The complete model can be considered as a generalized global model which includes one additional spatial parameter and one additional equation. The resulting equations are readily solved as in previous global models. The results are compared with experiments in an oxygen discharge, obtaining reasonably good agreement for the measurable parameters.

Considering a simple model of a spiral antenna inductive discharge in the collisionless heating regime, the basis of the equivalence of the stochastic and surface impedance methods is demonstrated. Formulae for the power per unit area, for each approach, are obtained in the regime of small normalized RF frequency, w, and shown to have the same functional dependence on w; the numerical coefficients agree to within a factor of 0.6. A study of the azimuthal electric field, E_θ, appropriate to the surface impedance method shows that although generally nonmonotonic, within the vicinity of the coil E_θ falls exponentially. The decay length δ_c very closely matches the skin depth of the stochastic method for a range of electron densities and RF frequencies. Thus the assumption that an exponential decay for E_θ in the stochastic method is a very good approximation for the purpose of calculating the collisionless heating is confirmed.

Space and time resolved relative atomic density distributions of nitrogen have been measured for the first time at a single filament within a dielectric barrier discharge (DBD) reactor with submillimetre radial dimensions. Two-photon-Absorption Laser-Induced Fluorescence (TALIF) spectroscopy of atomic nitrogen using radiation at λ = 206.7 nm is applied to a DBD with fast rising voltage amplitudes. The decay time of the atomic nitrogen density depends strongly on the position within the discharge and the distance from the dielectric where the lifetime is maximum. Admixed oxygen leads to an increase of the N density decay by an order of magnitude even at small fractions.

Isotope effects in an electron cyclotron resonance ion source are investigated in mixtures of ¹⁵N and ¹⁴N. The ratio of the extracted currents I(¹⁵N^{q +} )/I(¹⁴N^{q +} ) increases with increasing charge state q. This so-called `isotope anomaly' proves to be dependent on both the mixture ratio of the gases as well as the microwave power fed into the discharge. It is shown that the often-utilized `ion cooling' model cannot solely explain the observed isotope effects as long as the temperature is assumed to be equal for all ion species. Since we observed the occurrence of low-frequency noise in correlation with the isotope anomaly, we suggest that the ions are additionally heated by ion Landau damping. Due to the mass dependence of this process, the lighter ion species will be heated predominantly. This results in an enhanced transport of lighter ions out of the plasma and a better confinement for the heavier component, which, consequently, may be more effectively ionized to higher charges states.

A modification of the plasma density and the dc ambipolar potential profile by the ponderomotive force has been observed, for the first time, in an inductively coupled plasma (ICP) at low gas pressure. This ponderomotive effect was observed only at a low excitation frequency where the rf Lorentz force acting on electrons in the skin layer prevails over the rf electric field force. It is shown that this ponderomotive effect in an ICP is possible only at a condition of an anomalous skin effect when the classical formula for the ponderomotive force is no longer valid.

The design of an ion extraction system often requires the help of computer codes. This article describes an experimental setup based on a helicon source equipped with a three-grid extraction system and that is used as an ion source. A numerical approach based on a particles in cells (PIC) method is introduced. The results obtained with the ion source for different gases are compared with the simulation and are used in order to validate the computer code. The numerical code is the first component of an ion source design package that deals with grid erosion and ion beam current distribution on the target.

The evolution of the plasma obtained from magnetic probes and other electrical measurements in a coaxial gun with a Ti central electrode (cathode) used for coating studies in a mbar N₂ atmosphere is given. The results indicate that the currently used snowplough models adequately describe the cinematic of the plasma current sheet only if an additional mass to that of the gas between electrodes is included in the plasma sheet. The need to include extra mass is taken as evidence of relevant erosion of the central electrode, in accordance with the production of substantial coatings in similar devices. Evidence that sizeable portions of the discharge current remain attached near the end of the electrode system is presented, and some of the implications for the use of these devices for coating purposes are discussed.

Plasma instabilities have been studied in low-pressure inductive processing discharges with SF₆ and Ar/SF₆ mixtures, i.e. attaching gases. Oscillations are seen in charged particle density, electron temperature and plasma potential using electrostatic probe and optical emission measurements. For SF₆, instability onset in pressure and driving power has been explored for gas pressures between 2.5 and 100 mTorr and absorbed powers between 150 and 900 W. For pressures above 20 mTorr, increasing power is required to obtain the instability with increasing pressure, with the frequency of the instability increasing with pressure, mainly lying between 1 and 100 kHz. For Ar/SF₆ mixtures, we observe a similar low power transition, with an upper transition to a stable inductive mode. The instability windows become smaller as the argon partial pressure increases. For Ar/SF₆ mixtures, we observe a significant effect of the matching network. We improve a previously developed volume-averaged (global) model to describe the instability. We consider a cylindrical discharge containing time varying electrons, positive ions, negative ions, and time invariant excited states. The driving power is applied to the discharge through a conventional L-type capacitive matching network, and we use realistic models for the inductive and capacitive energy deposition. The particle and energy balance equations are integrated, considering quasi-neutrality in the plasma volume and charge balance at the walls, to produce the dynamical behaviour. As pressure or power is varied to cross a threshold, the instability is born at a Hopf bifurcation, with relaxation oscillations between higher and lower density states. The model qualitatively agrees with experimental observations, and also shows a significant influence of the matching network.

Using an emissive probe the distribution of plasma potential V_p in the bulk plasma of a dc Magnetron discharge has been determined for a range of argon pressures (0.26, 0.53 and 0.78 Pa) and cathode voltages (between -236 and -338 V). The results reveal a large axial variation in the space potential in the confined plasma, with ΔV_p~25 V over a distance of 5 cm, from plasma to sheath-edge. By combining the derived electric field with the modelled magnetic field, the distribution of single-particle drifts have been found, namely the electron E∧B, ∇B and curvature drift speeds. The predicted E∧B drift speeds (with values up to about 1.5×10⁵ m s^-1) are typically two to three times higher than the ∇B and curvature drifts. The Hall current channel is a broad region extending from above the `racetrack' down to a position close to the axis, 6 cm from the cathode. The calculated total Hall current is approximately five times the discharge current. Using a simple model of the discharge, in which there is no spatial variation in electron current density J_e, the gyrofrequency to collision frequency ratio averaged over the plasma bulk is found to be ω/ν≈7.7±4.2. In an extension to the model, a possible distribution of electron current throughout the plasma is considered, which allows the determination of ω/ν locally in the bulk. Using this method, the maximum value of ω/ν is found to be about 25, however both models indicate that cross-field electron transport occurs more rapidly than from a classical prediction.

The ionized metal physical vapour deposition (IMPVD) process is being developed to produce metal seed layers and diffusion barriers in deep contacts and vias for microelectronics fabrication. An IMPVD reactor is typically an antenna excited system where transmission line effects may produce asymmetric ion fluxes to the target and hence an asymmetric distribution of sputtered metal species in the reactor. A possible result is a non-uniform metal deposition on the wafer. In this paper, a three-dimensional model for an IMPVD reactor is employed to examine the consequences of asymmetric excitation and irregular sputter tracks on species' densities and fluxes. It was found that for typical conditions for Al IMPVD severe asymmetries in electron temperature and electron density profiles produced by a poorly optimized antenna are not reflected in the metal fluxes to the substrate. The metal species have improved symmetry due to charge exchange of the buffer gas ions to the metal and the higher mobility of the metal ions relative to the buffer gas ions. The symmetry and uniformity of the metal species above the wafer significantly improve when increasing the aspect ratio of the plasma region or increasing the pressure due to there being more diffusional transport. However, this improvement is accompanied by a decrease in the magnitude of metal fluxes to the wafer. Irregular sputter tracks combined with rotation of the target were also investigated.

The chemistry of argon, argon/nitrogen and argon/nitrogen/acetylene expanding thermal plasmas is investigated in order to unravel the role of plasma species in the fast deposition (up to 40 nm s^-1) of hydrogenated amorphous carbon nitride (a-C:H:N) films. The precursor dissociation is determined and the emission from the different plasmas is compared in order to distinguish possible mechanisms for species production and excitation.

The chemistry of an argon/nitrogen thermal plasma expanding through a graphite nozzle is investigated in order to unravel the role of plasma species in the deposition (1-3 nm s^-1 rate) of non-hydrogenated amorphous carbon nitride (a-C:N) films. The spectral emission of plasma species is studied and is used in combination with mass spectrometry and nozzle temperature measurements to distinguish possible mechanisms of species production and excitation.

Three plasma diagnostic methods, tunable infrared diode laser absorption spectroscopy, optical emission spectroscopy and microwave interferometry have been used to monitor concentrations of transient and stable molecules, CH₃, CH₄, C₂H₂, C₂H₆, and of electrons in capacitively coupled CH₄-H₂-Ar radiofrequency (RF) plasmas (f_RF = 13.56 MHz, p = 100 Pa, ϕ_total = 66 sccm) for various discharge power values (P = 10-100 W) and gas mixtures. The degree of dissociation of the methane precursor varied between 3% and 60%. The methyl radical concentration was found to be in the order of 10¹² molecules cm^-3 and the electron concentration in the order of 10¹⁰ cm^-3. The methyl radical concentration and the concentrations of the stable C-2 hydrocarbons, C₂H₂ and C₂H₆, produced in the plasma, increased with discharge power. The fragmentation rates of the methane precursor and conversion rates to the measured C-2 hydrocarbons were estimated in dependence on discharge power. Radial distributions of the electron and methyl radical concentrations, and of the gas temperature were measured for the first time simultaneously in the plasma region between the discharge electrodes. The measurements allow us to draw qualitative conclusions on the main chemical processes and the plasma chemical reaction paths.