Hello, Guys,

I am using FVM to discretize the Laplace equation on unstructured meshes (cell-center based) and found out that it is not accurate to compute gradient across the interface by two point finite differencing. If I use least square or Green's theorem to compute the gradient, the matrix becomes very complex. Is there any way to increase the accuracy of gradient computation across the interface? Thanks in advance.

Stein

if you go to the CFD-wiki, i have written descrisation of poission eq. for FVM, you can use the same for Laplace. Have a look.

I got it. Nice page, thanks.