On the two-dimensional modelling of horizontal CVD reactors and its limitations

In order to clarify under what conditions two-dimensional theoretical models can realistically account for the actual growth rate distribution in horizontal CVD reactors, a systematic numerical study was performed, and the results were compared with published experimental data for the MOCVD of gallium arsenide. Three models of increasing complexity were studied. It is shown that two-dimensional models can produce realistic predictions only for reactors with large width-to-height (aspect) ratios, that are operated at subcritical Rayleigh number,i.e., under dominant forced flow conditions. But even then, depending on the carrier gas, thermal (Soret) diffusion must be included in the model. Furthermore, velocity corrections for finite aspect ratios must be made, and buoyancy effects can be significant in the entrance region. On the other hand, the results are relatively insensitive to axial diffusion, temperature dependence of transport properties (versus averaged properties), and an assumed Poiseuille velocity profile (versus a rigorous solution of the velocity field). Yet, boundary layer approximations are prone to give unrealistic results at the low Reynolds numbers typical for CVD operations.