Dust particle formation in silane plasmas

Dust can be found anywhere: in the kitchen, in the car, in space… Not surprisingly we also see dust in commercial and laboratory plasmas. Dust can be introduced in the plasma, but it can also grow there by itself. In the microelectronics industry, contamination of the processing plasma by dust is an unacceptable phenomenon. People have put a lot of effort fighting the dust formation in plasmas. However, recent discoveries have proven that particles can also be used in the production of more efficient and stable solar cells and in single electron microelectronics devices. Particles that are used there have crystalline structure and a typical size of several nanometres. Since visible light has a wavelength of more than 400 nanometres, these particles cannot be detected optically; hence, new diagnostics are needed. In most cases, the active gas used in the discharges is silane – SiH4. This work is devoted to the study of the particle formation in silane containing plasmas. The commonly accepted mechanism of dust formation in a low-pressure silane plasma consists of several stages. First, silane radicals produce crystallite clusters of several nanometres in diameter. These clusters, when they reach a critical concentration, approximately equal to the concentration of ions in the plasma, start to agglomerate to form particles of 20 – 50 nm. These bigger dust formations no longer possess the crystallite structure of the initial clusters. All further growth is happening on the surface of the agglomerates. In the first part of this thesis we show that by following the behaviour of the fourth harmonic of the rf current enables us to identify phases of the particle formation. As a result we obtain a temperature dependence of the agglomeration time, which is the time before the nano-crystallites start to agglomerate. In agreement with previous observations, we show that a higher gas temperature slows down the particle formation process. In our experiments we have obtained an exponential dependence of the agglomeration time on temperature, which has not been reported before. Next, the agglomeration time dependence on the gas flow is studied. We present a semi-empirical particle formation model, explaining both flow and temperature dependence of the agglomeration time. Depending on the experimental conditions, several possible scenarios of the flow dependence have been proposed, many of which are also demonstrated to exist in the experimental results. We also point out the issue of the extra silane consumption from the vacuum volume, which strongly affects the speed of the particle formation and has not been addressed before. Then we give a general theory of the photodetachment experiment already developed before, with, however, a not yet addressed problem of the applicability. We use a small dielectric probe inside a microwave cavity to create a relatively big local perturbation of the medium inside the cavity. We then apply the principles of the microwave resonance technique to this situation and compare the calculated response of the cavity with the experimental values. Results of this test allow us to conclude that the medium perturbations caused by the extra electrons created during the photodetachment are small enough not to affect the results of the discussed theory for Summary 88 the laser induced photodetachment combined with the microwave resonance measurement technique. A new set-up has been built applying the microwave resonance technique to the fast measurements of the average electron density during the particle formation. The obtained results reveal several features that have not been observed before. The electron density in a silane containing discharge experiences a peak at around 300 µs after the discharge ignition. The electron density peak value in the silane/helium/argon mixture is higher that that in pure argon, which can easily be explained by a lower ionisation potential of silane compared to the ones of helium and of argon. However, the electronegativity of silane radicals quickly results in a decrease of the electron density, which results in the fact that, already 2 ms after the plasma ignition, the electron density in silane mixture is lower than in a pure argon discharge at the same rf power. In parallel with the electron density measurements we have also performed electrical measurements of the electrode self-bias and the rf harmonic amplitudes. We show, that during the first milliseconds of the discharge a noticeable decrease of the electron density (about 60 %) does not result in a significant change of the measured third harmonic of the rf current amplitude and the rf electrode self-bias voltage. We also study the electron density evolution during the particle agglomeration. As was already measured before, the particle agglomeration results in a drop of the electron density up to one order of magnitude. The change in the electrical parameters, however, is less drastic, although it is still possible to correlate the moment of the onset of the agglomeration with the evolution of the harmonic amplitudes of the rf current. At 1200C gas temperature we also looked at the evolution of the electron density for the first twenty-five milliseconds after the discharge ignition. We show that at 1200C the electron density drops about two times faster than in the experiments at 200C, indicating a higher electron loss rate at higher gas temperatures. Finally, we also briefly discuss the issue of plasma bursts, a phenomenon when the plasma suddenly spreads out of the discharge chamber, switching to another regime of operation, which results in a major increase in the electrode self-bias voltage, but not in the average electron density in the discharge chamber.