Prev: 6.3 Performance

7. High-Pressure Chemical Vapor Deposition

Chemical vapor deposition is based on the principle, that a heated substrate activates a chemical reaction of gaseous reactants, which leads to the deposition of a solid film. The reaction kinetics of the film formation are determined by the balance between the diffusivity of the reactants and the reaction rates of the involved chemical reactions. Since typical process pressures for chemical vapor deposition are in the range of some 10Torr, the transport of the gaseous reactants is determined by diffusion. The profile evolution is determined by the competition between supply of reactants by diffusion and consumption of the reactants by chemical reactions. This chapter will give an insight into modeling of these competing processes which are characteristic for deposition processes based on CVD.

In recent years a variety of three-dimensional simulators has been presented. In Section 2.1 they have been listed according to the surface propagation algorithm they use. Here, the programs are recapitulated in order to check the process models they have implemented.

SAMPLE-3D [65] includes macroscopic models for flux determined etching and deposition simulations based on a facet motion algorithm for the surface propagation. A similar segment-based approach [6] uses a redistribution model for reactive particles for simulating LPCVD. In the level set methods developed at UC Berkeley [68] and Stanford University [27] low-pressure processes are implemented by various combinations of analytic distribution functions for particle fluxes. Other two- or three-dimensional topography simulators working with similar macroscopic models are presented in [18] and [21].

A Monte Carlo approach combines results from the three-dimensional reactor scale Monte Carlo particle transport simulator SIMSPUD with the feature scale simulator SIMBAD, thus resulting in interpolated three-dimensional deposition profiles [69]. This approach is very closely related to the manufacturing equipment since it incorporates experimental data such as target erosion profiles. Monte Carlo models combining particle transport simulation and surface reaction kinetics for plasma etching processes are also well established for multiple dimensions [7][26].

The simulation programs listed so far share a severe disadvantage. All these simulators only model processes at pressures below 1Torr. In this regime the Knudsen number as the ratio between the gas mean free path and the characteristic feature length is well above unity. Therefore the molecules inside the feature travel in line-of-sight and the particles undergo only very few collisions. Gas phase reactions mainly occur in the reactor space above the feature. Non-uniform film thickness is mainly caused by shadowing effects in the particle transport.

For pressures above 1Torr the gas mean free path becomes relatively small (Knudsen number $K_n \ll$ 1) and the ratio between molecule-molecule and molecule-surface collisions increases drastically. Particle transport becomes determined by species effective diffusivities in the gas mixture. Unfortunately only one simulator is available to model this type of mass transport and only a two-dimensional implementation can be found. Two-dimensional simulations of such low Knudsen number transport CVD processes are originally reported in [49] and extended in [35]. The three-dimensional expansion of the related simulator, EVOLVE-3D [34], combines the facet motion algorithm proposed by Scheckler [64] with the chemistry and reaction model of EVOLVE but modeling for the three-dimensional expansion is restricted to the ballistic transport part.

The goal of the high-pressure model introduced in this chapter is to close the gap of missing high-pressure models in three-dimensional simulators. Therefore the two-dimensional continuum transport and reaction model was extended to a fully three-dimensional model for the feature scale simulation of arbitrary, multiple chemistry, high-pressure CVD processes in the continuum transport regime.

The key issues for the model which will be presented in detail in this chapter are surface extraction from the three-dimensional cellular structure (Section 7.1.1), meshing of the reactor domain above the feature (Section 7.1.2), combined diffusion/reaction simulation (Section 7.1.3) and surface propagation (Section 7.1.4). Time-step control (Section 7.1.5) and the overall process control (Section 7.1.6) conclude the presentation of the CVD model.

Section 7.2 demonstrates the implementation of the model to the deposition of tungsten which is applied to a Ti/TiN/W plug-fill process and for various geometries and process conditions in Section 7.3. Finally, the link between reactor and feature scale simulations will be shown (Section 7.4).

Prev: 6.3 Performance