Discussion Forum

Hi All,

I am experiencing a very strange problem in one of my models. The model comprises three application modes: two stationary convection and diffusion equations and one electrostatic equation. The calculation domain is a square with 0.5mm. The model describes a gas discharge between parallel plate electrodes; one of the convection and diffusion modes describes electrons in the discharge, the other ions.

In one of the bondaries, the conditions are zero concentration for the electrons, convective flux for the ions and zero current. The purpose of using convective flux for ions is to eliminate a thin boundary layer which is very difficult and costly to resolve. For these boundary conditions, the model behaves rather oddly. For example, there is an extreme sensitivity to the mesh parameters. If I use a mapped mesh with 34 x 34 elements, I am not able to refine it further without loosing convergence, even for 35 x 35 elements and using the converged solution for 34 x 34 elements as an initial approximation.

So far, I have been using the same model but with zero concentration of ions on the above mentioned boundary. In that case, my (mapped) mesh contains 60 x 60 elements and I can increase the density of the mesh to 80 x 80 elements in a single step. The model behaves predictably.

It does not make sense to me how can convergence be lost simply by using a slightly more refined mesh. Any ideas? Thanks in advance.

Best regards,
Pedro Almeida