Défense de thèse de doctorat en Sciences physiques - Antoine FAUROUX

Experimental and theoretical study of a magnetron DC-PECVD acetylene discharge

Jury

• Prof. Yoann OLIVIER (département de Physique, UNamur), Président
• Prof. Stéphane LUCAS (Département de Physique, UNamur), Promoteur et Secrétaire
• Dr. Andreas PFLUG (Fraunhofer Institute for Surface Engineering and thin films, Germany)
• Dr. Stefan GROSSE (Robert Bosch GmbH, Germany)
• Prof. Erik NEYTS (NANOLab Center of Excellence, Universiteit Antwerpen)
  

Abstract

The deposition of DLC films from a low-pressure acetylene in magnetron reactors has continuously driven the interest of scientists for decades. It is used widely by industrials in large batch coaters to produce high added value protective coatings, or by experimentalists in more complex arrangements to develop processes that include amorphous hydrogenated carbon as a key component in novel and exciting applications, e.g. in electronics, energy storage, or even medicine. It is still only possible to resolve analytically the complex equations of the dynamics of cold temperature reactive plasmas for very simple cases, and gaining insight on the reactions and particle behaviour usually requires the use of numerical simulations. For very low-pressure discharges (below 1 Pa), the individual particle trajectory must be resolved, which implies the use of statistical Monte-Carlo approaches like PICMC. The high computational cost of such simulation limits the attainable powers, the simulation size and length, and the number of considered species. The object of the present study is therefore to find out if a realistic PIC-MC simulation of a magnetron assisted PECVD discharge in the case of the deposition of DLC from acetylene can be developed, and to see if it is possible to compare simulations' results in a constraining way with experiments.

We showed, that it is indeed possible to simulate this type of acetylenic discharge with a simple but self-coherent plasma chemistry model in a small 3D simulation box. With simulation times of up to 1.2 ms, an equilibrium could be attained for the densities and fluxes of all species. Even if ions are the main species created within the plasma, radicals accumulated to densities comparable to that of ions due to their slower diffusion speed. Differences in concentrations were observed across the simulation chamber between the thermally diffusing neutrals and the accelerated ions. Moreover, some species were created within the chamber from the reaction between species generated in the plasma bulk interacting with the background acetylene.

Those predictions were tested against mass spectrometric measurements in experiments with various acetylene ratios and discharge powers. The variations of species flux towards the substrate, with ratio were similar in both experiments and simulations. It was possible to match the linear evolution of the ions' flux with power between experiments and simulations. Moreover, the simulations showed a correlation between the electron density and all plasma products ones. This suggests that the power scaling of simulation's predictions should be possible until the point where the reactive species concentrations would become non-neglectable compared to the acetylene one.

Since the concentrations and fluxes varied throughout the chamber in this configuration, variation of the film characteristics and of the deposition rate were expected across the substrate. To test the simulations film deposition predictions, a dynamic surface chemistry model was set-up. It includes the creation of dangling bonds via the chemical sputtering due to the joint ion bombardment and hydrogen flux and the preferential deposition of radicals on dangling bonds. Since the equilibrium of the dangling bonds coverage was not obtained extrapolating the absorption of radicals was necessary. Despite this and the power extrapolation, simulations were able to give quantitatively accurate deposition rates and deposition profiles when compared to experiments. However, the hydrogen stoichiometry was higher in the simulations than the one measured with ERDA measurements. In order to improve the film deposition prediction several refining of the surface model would have to be included in future works. Some perspectives are given regarding the possible improvements of the current model, which could already be used for more complex configurations to reduce the need for the costly trial-and-error search for optimal deposition parameters.

Lien

 Cliquez ici pour participer à la présentation en ligne

Télecharger : vCal