"Bill Greene" wrote in message <email@example.com>... > >I need to refine my mesh a number of times before I can arrive at the desired accuracy. > > I suggest taking a look at the adaptmesh function. It does automatic > h-refinement of the mesh and often yields more accurate solutions with > fewer elements than the assempde function. Its input arguments are very > similar to assempde. > > Bill
Thanks for the input. I realized that the real reason for the loss in accuracy comes from the 'pdegrad' function. It reduces the convergence of my program from second order to first order since it is a part of the iterative scheme that my program uses. Is there a more accurate method to find the gradient? Thanks for your help.