Depositing Aluminium Oxide: A Case Study of Reactive Magnetron Sputtering

Abstract

The deposition of aluminium oxide from an aluminium target in DC mode is used in this chapter to illustrate the different aspects of the reactive magnetron sputter process. The choice for this combination of target material and reactive gas is given by the fact that the well-known hysteresis behaviour, described in the previous chapter, is clearly demonstrated for this combination. Two key aspects of the process are responsible for this behaviour. The strong difference in the molecular sputter yield and density for Al and Al2O3 results in a dramatic decrease of the deposition rate when the target is poisoned [1]. Similarly the ion induced secondary electron emission coefficient of aluminium changes strongly when reacting with oxygen to form aluminium oxide. Hence, the absolute discharge voltage drops in the order of 100V when the target becomes poisoned [2]. The previous chapter describes the “Berg” model for reactive magnetron sputtering. Several experimental trends can be explained using this model. The model includes most important processes accounting for its success. In short the model describes the gettering of the reactive gas by the target material which influences the reactive gas partial pressure. At low reactive gas flow, the reactive gas is almost completely gettered and hence the target condition remains metallic. When on increasing the reactive gas flow the maximum amount of reactive gas which can be gettered is reached, the reactive gas partial pressure increases and the target becomes completely poisoned. Depending on the experimental conditions this change from metallic to poisoned mode can occur abruptly. Although the model described in the previous chapter explains quite well the hysteresis behaviour some experiments described in this chapter are difficult to understand from this model. This finds its origin to the opinion of the authors in the description of the poisoning mechanism of the target. Indeed, in the “Berg” model, the reaction between the target material and the reactive gas is described by chemisorption. This is in first order a correct approach for the reaction on the substrate but is not a complete description for the target process. Indeed, during magnetron sputtering the target is bombarded by ions from the plasma, including reactive gas ions. The ion energy is defined by the discharge voltage and is in the order of 400 eV. Hence, the ions become implanted in the target at a depth in the order of a few nanometers. Two major conclusions can be drawn. First, when reacting with the target material, the reactive ion implantation results in a much thicker layer than modelled in the “Berg” model. Secondly, the reaction mechanism becomes more complicated and hence this could influence the description of the dynamics of reactive magnetron sputtering. These two major conclusions will be addressed in the final part of this chapter where an model is presented describing the reactive ion implantation in an analytical way.