CHAPTER 8. SUMMARY AND OUTLOOK 90 Under the operational conditions used in the present work the translation temperatures can be obtained from the Dopp

Transcription

1 Chapter 8 Summary and outlook In the present work reactive plasmas have been investigated by comparing experimentally obtained densities with the results from a simple chemical model. The studies have been performed in methane/oxygen rf discharges with a driving frequency of MHz in two discharge chambers with different coupling mechanisms (capacitively and inductively), pressures (10 and 100 Pa) and surface materials. These varying conditions gave the opportunity to study the influence of a wide range of external parameters on the chemical composition. The comparison with the hypothetical plasma composition calculated by the introduced model offers the possibility to estimate the shape of the EEPF under all discharge conditions used. Two different kind of plasma sources have been investigated: a capacitively coupled parallel-plate reactor at 100 Pa and an inductively coupled reactor (GEC reference cell) at 10 Pa. The ICP exhibits two different modes: one low-density mode sustained by the electrostatic field between the coil and a grounded electrode (E-mode), and a high density mode sustained by the induced azimuthal electric field (H-mode). The electron density in the CCP is positioned in between these two modes resulting in a wide range of accessible values: m m 3. One special feature of the ICP is the occurrence of a hysteresis between the two modes. In the present work a dependence of the transition point and the width of the hysteresis on the mixing ratio has been observed. The chemical composition has been investigated by means of absorption spectroscopy using tunable diode lasers. This technique is non-invasive and not affected by the typical problems in reactive plasmas, as e.g. the deposition/etching of surfaces, which are directly exposed to the discharge. Four spectrometers for different wavelength regions (760 nm μm) have been built up for the investigation of the two source gases (CH 4 and O 2 ), an intermediate radical (CH 3 ) and two stable products (CO and CO 2 ). The basic principles of this technique have been investigated in previous works in our group [9, 66, 79]. The sensitivity has been increased by using a frequency modulation technique and by enlarging the absorption path using amultiple reflection cell (Herriott type). 89

2 CHAPTER 8. SUMMARY AND OUTLOOK 90 Under the operational conditions used in the present work the translation temperatures can be obtained from the Doppler width of the absorption lines. In the CCP cell as well as in both ICP modes the translational temperature is around 300 K. The rotationalvibrational temperature can be calculated from a Boltzmann plot or by the temperature dependence of the line strength. At the maximum power of 60 W in the CCP cell a rotational-vibrational temperature of 360 K has been obtained in a previous work [9, 11] and a similar value (350 K) has been observed in the H-mode of the ICP cell. In the E-mode no increase above room temperature has been found. The observation of the chemical composition was focussed on the balance between the two stable carbon oxides. Completely different ratios have been observed in the CCP compared to the ICP cell. Whereas in the CCP the carbon dioxide molecule dominates the plasma composition, it plays only a minor role in the ICP and vice versa for the carbon monoxide. Additionally, inthe CCP the feed gases are nearly completely converted into carbon oxides under certain conditions, whereas the conversion in the ICP cell is only up to 34 %. Some additional measurements with a commercial Langmuir probe system have been performed, which have provided pieces of information on the electron density and energy distribution function. This technique was limited to very restricted regimes with sufficiently high electron densities (high methane content, high input power). For the CCP setup only an upper limit of m 3 (pure CH 4, 10 sccm, 100 Pa, 60 W) could be determined. In the E-mode of the ICP cell no measurements were possible due to the very low electron density. In the H-mode some measurements of the electron density have been obtained but only in very restricted regimes (high methane content) resulting in values from m 3 up to m 3 (H-mode, 10 Pa, W). The determination of the EEPF was disturbed for nearly all operational conditions. The EEPF is influencing nearly all processes in the discharge. A small change in the shape of the EEPF results in a completely different plasma composition. The accurate determination of the EEPF is therefore needed for the understanding of the plasma processes. Conventional methods for the experimental determination of the EEPF, like the Langmuir probe technique or Thomson-scattering, are often not applicable to reactive plasmas. Due to the manifold distortions these methods are not suitable to determine the shape of the EEPF in CH 4 /O 2 discharges correctly. In the present work another approach to achieve information on the electron components has been chosen: the densities obtained by absorption spectroscopy are compared with the results of a simple chemical model, which uses the EEPF as well as the electron density as input parameters. The model does not aim at a complete description of the overall plasma physics, but is a strongly simplified model, which describes more or less only the chemical reaction system. The sheaths are not taken into account at all and surface interactions are implemented by the sticking coefficients. The charged particles are

3 CHAPTER 8. SUMMARY AND OUTLOOK 91 treated as neutrals in this very simple approach, which means that neither self-consistent fields nor quasi-neutrality are included. This treatment can be justified by the low degree of ionization (< 10 6 ). The main purpose of the used model is to have an easily modifiable numerical tool with short execution time for every day and everyone's use. The EEPF can be calculated by an implemented Boltzmann solver or loaded from an external data file. An adequate EEPF can be estimated by comparing the calculated composition to the experimentally obtained densities. If a good agreement is achieved for reasonable electron densities, the selected EEPF is labelled as well fitting". Good reproduction of the species concentrations is obtained for all investigated discharge conditions. With a reasonable estimation for the electron density, which has been attained with the help of the Langmuir probe system and from the dissociation rate of methane, the dependence on the input power/electron density fits well for all species. The absolute number densities are partly over- or underestimated, but still within the error bars of the measurements. In addition, the influence of the flow rate, the pressure, the mixing ratio and the surface materials, which can be observed in the experiment, have been reconstructed with the model. Especially, the balance between the two carbon oxides is sensitive to small changes in the shape of the EEPF and only by these two molecules a differentiation between two EEPFs is possible in the CCP cell and the E-mode of the ICP cell. Another species (like oxygen), which can be predicted by the model, has to be measured for a clear differentiation of the EEPF in the H-mode of the ICP. Additional advantages of the model are the possibility to investigate the main production and depletion paths of the species and to predict species concentrations. The experimental CCP results for the exemplarily chosen mixing ratio of CH 4 :O 2 = 1:2 can be reproduced very well for both flow rates (2 and 10 sccm). In the E-mode of the ICP cell the carbon oxide densities are very sensitive to small changes in the EEPF, and also in the H-mode a range of a fitting EEPF can be estimated. For all operational discharge conditions the measured trends as well as the the relative densities are reproduced in dependence on the EEPF by the model introduced in the present work. The EEPF can be estimated for mixtures of CH 4 and O 2, where e.g. Langmuir probe measurements are not possible due to distortions and a low signal-to-noise ratio. Still, the experimental methods as well as the numerical model are far from being sophisticated. Both methods do not have spatial resolution, which introduces large error bars into the resulting densities. In the experiment an easy solution is not available due to the necessity of a multipass cell. One way out could be a decrease of the vertical dimension of the folded path through the mirrors or the choice of another type of multipass arrangement (as e.g. the White cell). In the future the cavity ringdown technique [52, 85] can support the density measurements but until today no multi-layer mirrors with a

4 CHAPTER 8. SUMMARY AND OUTLOOK 92 sufficiently high reflectivity (> %) for the desired wavelength regions (1-16 μm) are available. In addition, a reconstruction technique has to be applied, because only line-integrated values can be obtained by absorption spectroscopy. Together with radially resolved measurements the model has to be extended to at least one dimension, better two dimensions. A group of theoreticians at the IPP in Greifswald has developed a Particle-incell (PIC) code for the description of ECR methane plasmas [75]. A cooperation with this group has just started. In the near future their model will be adjusted to the operational conditions in our group, and time and spatial resolved quantities like the EEPF will be accessible. Further on, the implementation of oxygen is planned. This PIC-code is numerically very complicated and needs a long execution time (more than one day on a Linux cluster) and is not useful for daily practice. Therefore this PIC-code is not aimed to replace our simple approach, but can be very helpful for comparing and checking a 1- or 2-dimensional model to be developed for our setup. The variation of the input variables in the model leads to plausible estimations on the EEDF, if an adequate electron density is assumed. Of course it is necessary to check these assumptions experimentally. The Langmuir probe technique has to be improved in such a way that measurements of the electron density in discharges with a higher oxygen content are possible and reliable. The distortions impairing the determination of the EEDF might be also suppressed, but both improvements are out of the scope of the present work. Another possibility to obtain the EEPF at least in the H-mode (high electron density) could be a Thomson scattering setup, which is in principle able to measure the electron density and distribution function in reactive plasmas. At the moment, a system with a tunable Ti:S-laser is under construction in our group [58, 59]. A major disadvantage of the currently used model is the neglect of the quasi-neutrality condition. In a former work (Busch [9]) a fluid model has been built up, which implements the quasi-neutrality. This model was not able to describe all discharge conditions used in the present work and therefore had been skipped completely. In the future, this module should be rewritten and included again into the simplified description used in the present work resulting in a self-consistent plasma model. In addition, the chosen set of EEPFs is plausible but the interpretation of the resulting well fitting" EEPF is not easy. In section 3 the Maxwellian and Druyvesteyn distribution functions are described. Mathematically they differ in the exponent of the energy (see equations 3.6 and 3.7). If the dependence on the exponent of the energy in the description of the distribution function would be used as selective tool for the model, the resulting shape of the EEPF can directly refer to a Maxwellian EEPF (if the exponent equals one) or to a Druyvesteyn EEPF (if the exponent equals two). Everything in between can at least be interpreted as nearly Maxwellian or more Druyvesteyn-like. The cut-off energy has to be adjusted by the mean energy of the EEPF, and this could be implemented additionally.

5 CHAPTER 8. SUMMARY AND OUTLOOK 93 Another improvement would be an implementation of the changing plasma composition. At the moment, one EEPF is selected for one mixing ratio of the source gases and independently of the electron density. But in reality, the gas composition is drastically changed in the discharge and the EEPF is varying with the input power. The argumentation for the included diffusion constant is similar. They are also changing with the gas mixture, but in the present model they are kept constant.