Modelling of Reactive Sputter Deposition of Oxynitrides

Sören Berg1, Tomas Nyberg1, *, Diederik Depla2
1Solid State Electronics Division, The Angstrom Laboratory, Uppsala University, Box 534, 75121,Uppsala, Sweden and 2Department of Solid State Sciences, Ghent University, Krijgslaan, 281(S1), 9000Ghent, Belgium

Abstract

Reactive sputter deposition is frequently carried out in a mixture of argon and oxygen or nitrogen to obtain oxides and nitrides. The behavior of such "single-reactive gas" processes has been explained theoretically and verified by numerous industrial thin film deposition applications. However, mixing two reactive gases with the argon sputtering gas in order to carry out reactive sputter deposition of oxy-nitride films is a far more complicated process. A first order simple process model for such a mixed process is presented. Modelling indicates that altering the supply of one gas will not only cause a change of the partial pressure of this gas but also may significantly change the partial pressure of the other reactive gas. Moreover, different reactivities of the reactive gases result in stoichiometries that are very different from the relative reactive gas supplies. This linked behavior between the reactive gases may cause severe process control problems. In a second part of the chapter, a more advanced model is presented, which includes reactive ion implantation and knock-on implantation. First this model is tested vs. experimental results published in literature. Based on the good agreement in the noticed trends, the time dependence of the poisoning behavior of the target is discussed. Finally, the influence of the deposition profile on the hysteresis behavior, and the composition of the oxynitride is discussed, showing the importance of a complete model.

Keywords: : Sputtering, reactive sputtering, oxynitrides, process modelling, chemical reactivity, target poisoning, process control, hysteresis.

* Address correspondence to Tomas Nyberg: Angström Laboratory, Uppsala University, Box 534, 751 21 Uppsala, Sweden; Tel: 46-(0)-18-471-3164; Fax: 46-(0)-18-55 50 95; E-mail: tomas.nyberg@angstrom.uu.se

INTRODUCTION

Reactive sputtering is a well known technique to deposit oxides and nitrides from different metal targets. Processes involving a single elemental metal target and a gas mixture of argon and one reactive gas (e.g. oxygen or nitrogen) have been

extensively studied and well described in the literature. Normally this type of reactive sputtering process exhibits a hysteresis behaviour that may cause complications concerning processing stability. It should be understood that despite the relative complicated processing behaviour the "one metal one reactive gas" sputtering process is the simplest reactive sputtering process to describe. A simple model has been developed that describes the main features of this process [1]. Results from this model may predict the target sputter erosion rate and the partial pressure of the reactive gas as a function of the supply of the reactive gas. Typical results are shown in Figs. 1 and 2.

Figure 1)

Schematic of processing curve for partial pressure of reactive gas vs. gas supply using constant sputtering current.

Figure 2)

Schematic of processing curve for target sputter erosion vs. supply of reactive gas using constant sputtering current.

The S-shaped processing curves can only be obtained if the partial pressure of the reactive gas is used as control parameter in a feedback control system. The solid lines define the processing curves if the supply of the reactive gas will be used as control parameter. The dotted segment will be reached if the partial pressure of the reactive gas is used as control parameter. The width of the S-shaped curves defines the hysteresis width of the process. It should be understood that if the supply of the reactive gas is used as the input control parameter it will not be possible to reach the processing points on the S-shaped curves inside the hysteresis width. This gives a fundamental limit for what compositions that can be obtained for the deposited films.

Replacing the single element metal target with 2 different single element targets or an alloy target consisting of 2 metals significantly changes the processing conditions. Since the two metals generally have different reactivity to the reactive gas it will be necessary to adjust the supply of the reactive gas to a level where the less reactive metal will fully react with this gas. This normally forces the process into a low rate sputtering mode. This low rate may sometimes be as low as one magnitude lower than what is possible to obtain if a single element target was used.

It is also possible to use one single element target and add 2 reactive gases to the argon processing gas. By this technique it is possible to reactive sputter deposit e.g. oxy-nitrides. Both the nitrogen and the oxygen have to react with the sputtered metal atoms. Also for this system the difference in reactivity of these gases will give new restrictions in the processing control. In addition a complication in controlling the partial pressures of the 2 reactive gases will appear. Varying the partial pressure of one of these gases will affect the partial pressure of the other gas. Therefore quite advanced feed-back control systems are needed to obtain full control of such a reactive sputtering process [2].

BASIC MODELLING OF SPUTTERING WITH TWO REACTIVE GASES

In this chapter we will describe somewhat in detail the mechanisms responsible for the processing behaviour for a reactive sputter deposition process carried out from sputtering one single element target in a mixture of argon and 2 reactive gases (oxygen/nitrogen). The purpose of the presentation is to clarify what kind of additional complications will arise that will not be present when sputtering one single element target in a mixture of argon and one reactive gas. We believe that the simplest way to illustrate the differences may be to outline a simplified model of the process. From this model it will be possible to predict processing behaviour caused by the different involved processing parameters. As a first approximation we will follow the description earlier presented in reference [1].

We will assume a system consisting of one single element target having an area At. Power may be applied to this target generating an argon ion current density J evenly distributed onto the target surface. All sputtered material will be collected and evenly distributed at the collecting surface having area Ac. Two reactive gases (e.g. oxygen and nitrogen) will also be present in the chamber having partial pressures pO and pN respectively. It will be assumed that these partial pressures are so small so that they do not give rise to any contribution to the ion current bombarding the target. Thus only argon ions are responsible for the sputter erosion from the target. There will be a probability that the gases will react with the pure metal atoms at the At and Ac surfaces. These probabilities will be denoted aO and aN for oxygen and nitrogen respectively. For simplicity we will assume the same probabilities at both surfaces. It is far too complicated to include all possible chemical reactions that may occur during processing. However, we will consider one possible effect. It is well known that oxygen normally is more reactive than nitrogen. In fact it may be possible that the oxygen also may react with the nitride and convert it into oxide. This can only happen at surfaces where nitrides have been formed. The probability for this to happen will be denoted aON.

We assume the following to happen in the processing chamber. Due to the presence of oxygen in the chamber some fractions ØtO and ØcO of the surfaces At and Ac respectively will be covered to oxide. The fractions covered by nitride will be denoted ØtN and ØcN.

Based on the assumptions above it is possible to define a number of equations that together determine the expected processing behaviour. The outline in Fig. 3 will serve to illustrate a simplified model of this type of reactive sputtering process.

Figure 3)

Schematic of incoming gas fluxes and ion current to the different area fractions at the target.

We will assume steady state conditions.

The relation between the partial pressure p and the flux of molecules F (number/ unit area and time) that this pressure generates at all surfaces will be:

The flux of sputter eroded metal atoms out from the target will be denoted RM.

Where J denotes the current density of argon ions and e denoted the elemental electronic charge and Ym denotes the partial sputtering yield of metal by the argon ions. Oxide and nitride are formed at the target surface. We will assume that sputtering of these compounds will...