Modeling reactive magnetron sputtering: Fixing the parameter set

The total pressure was measured as a function of the oxygen gas flow during reactive magnetron sputtering of two different target materials, namely yttrium and aluminum. These experiments, generally referred to as hysteresis experiments, were repeated for different discharge currents, and formed the experimental data set to be fitted. Combining experimental input with simulations reduces the number of unknown parameters to two: the sticking coefficient of molecular oxygen on the target during sputtering, and the reaction rate coefficient defining the reaction between the implanted oxygen and the target material. Taking into account the error margins of three other parameters resulted into five degrees of freedom to fit the experimental data. These five fitting parameters were varied, and evaluated on their goodness-of-fit with the experiments. For this purpose, an algorithm was developed to find all the combinations of these five parameters that give acceptable fits. By simultaneously fitting to three different measurements, only a minor amount of the parameter sets showed a good fit. Relationships between the fitting parameters were found, while quantification of the hard-to-collect material dependent parameters was established. This allows a comparison to be made of the response of both target materials to their exposure to the argon/oxygen plasma.

► An extension of a model for reactive magnetron sputtering is proposed.
► An algorithm exploring the remaining fit freedom is developed.
► Quantification of the sticking coefficient and the bulk reaction rate of Al and Y.
► Agreement between the simulated and the experimental sputter yields of the oxides.