Discussion Forum

I'm using COMSOL to generate a numerical Green's function for a piece of an electrostatic system. I have N points of interest (uniformly located along a line), so I run N simulations, in each simulation placing a unit point charge at a single point of interest, and I generate my Green's function by extracting the resulting potential at each of the other points. (So in the end I have a NxN matrix, with the (i,j)th entry telling me the potential at point i resulting from a unit charge at point j.)

My initial attempt ended up very "noisy," though. E.g. G(i,i) and G(i+1,i+1) can differ by 70% or so in random directions when the adjacent points are a nanometer apart and sitting in the middle of nowhere relative to the system geometry.

So my next step is to force a fine, rectangular mesh around my region of interest, under the assumption that a lot of this noise is coming from variation in auto-generated mesh element size. But does anybody have experience doing this sort of thing or recommendations for a better way?