An experimentally verified useful new model for reactive sputtering is presented. By considering the total system (target erosion, gas injection, chamber wall deposition, reactive gas gettering at all surfaces, etc.) during deposition it is possible to evaluate quite simple relationships between processing parameters. We have expanded earlier treatments to include these phenomena. The model involves that gettering of the reactive gas takes place at the target and at the walls opposite to the target. Arguments are also presented for how the sputtered materials (elemental target atoms and the formed compound) contribute to the formation of the surface composition of the walls opposite to the sputtering electrode. The mass flow of the reactive gas has been chosen as the independent parameter in this presentation. Results for partial pressure and sputter rate are presented. The theoretical values are compared with experimental results from reactive sputtering of TiN. It is also pointed out that the calculated values agree extremely well with results presented in the literature by several other authors.

The electrical, optical, and mechanical properties of a compound film depend strongly on the composition of the film. Therefore, it is interesting to study a wide variety of compositions of many new compound materials. Reactive sputtering is a widely used technique to produce compound thin films. With this technique it is possible to fabricate thin films with different compositions. However, it has not yet, to any great extent, been possible to predict the composition of the sputtered film. In this article we will present a model that enables us to predict both sputtering rate and film composition during reactive sputtering. The results point out that there exists a very simple linear relationship between processing parameters for maintaining constant thin-film composition in the reactive sputtering process. Based on these results, it is possible for the first time to combine information of both sputtering rate and film composition into the same graphical representation. Access to this new and simple graphical representation may eliminate much of the ‘‘trial and error’’ work that earlier has been associated with the reactive sputtering process.

No abstract

Reactive sputtering is a very complex and nonlinear process. There are many parameters involved. Normally it is not possible to vary a single parameter independently of the others. It is therefore very difficult to characterize the process based on experimental observations. A better understanding of the reactive sputtering mechanism is needed. We have suggested a simple model for the reactive sputtering process. This model is primarily based on well-known gas kinetics, transferred to this application. With this model it is possible to theoretically predict different processing conditions and actually study the influence of a change in an individual parameter value. The results may then be used to predict optimal experimental conditions. With this technique it is also possible to study means of affecting the well-known hysteresis effect. This article is specially devoted to explain the width of the hysteresis region and how it is affected by the sputtering intensity. Experimental results are presented that support the validity of the proposed model. It was found that the hysteresis width increases in proportion to the rf power fed to the sputtering target. It is thus not possible to eliminate the hysteresis effect by any kind of variation of the sputtering intensity.The effect of the hysteresis effect on the composition limitation of the deposited films will also be described.

At first sight one might assume that it is unlikely to influence the sputtering yield of a specific ion/substrate combination by any external means. However, we have found that such an influence may well be introduced. The sputtering yield is predominantly determined by the ion/substrate momentum transfer efficiency and the energy of the incoming ion. Sputter erosion of, e.g., carbon atoms by argon ions from a carbon substrate exhibits a very low sputtering yield. Due to the difference in masses between carbon and argon much of the momentum is transferred into the bulk of the carbon substrate. This situation could be changed by simultaneous codeposition of Pt atoms onto the carbon substrate surface during the argon sputtering. Keeping the argon flux at a level well above what is needed to sputter remove all the deposited Pt atoms the following effect occurs. Some of the deposited Pt atoms will be forward implanted by the energetic argon ions into the near surface region of the carbon substrate. Collision between both argon (M=40) and carbon (M=12) atoms with the implanted Pt atom (M=195) can result in reflection of some of the light atoms. Therefore implanted Pt atoms will act as effective reflection centers pushing the collision cascade to take place closer to the surface region, thereby contributing to an effective increase of the number of collision cascades in this region. This reflection effect will result in a substantial increase of the sputtering yield of carbon atoms. Experimental verification of this effect for the (Ar+C+Pt) and the (Ar+Si+Pt) systems will be presented. It will be also shown that this sputter amplification process can be predicted from computer simulations using the t-dyn version of the Monte Carlo trim code.