Hi Roy,
> Use Lagrange finite elements, but instead of approximating your > integrals with Gaussian quadrature, approximate them with nodal > quadrature (QTrap in libMesh IIRC). Each quadrature point falls on a > node, and only one Lagrange shape function is non-zero at that node, > so you only get contributions to the equation for that shape function > so your mass matrix is diagonal. Okay, thanks. I will try this. What should be the errors like when I use QTrap instead of Gaussian quadratures? Do I need to make a finer mesh? Thanks, Harshad On Wed, Dec 16, 2015 at 5:39 PM, Roy Stogner <royst...@ices.utexas.edu> wrote: > > On Wed, 16 Dec 2015, Harshad Sahasrabudhe wrote: > > Thanks. Yes, my goal is to have a non-diagonal mass matrix. I want >> to take the easiest path possible for doing that using LibMesh. By >> mass lumping with a Lagrange basis, do you mean having cell centered >> finite volume using a 0th order Lagrange basis in finite element? >> > > No, even simpler a change than that: > > Use Lagrange finite elements, but instead of approximating your > integrals with Gaussian quadrature, approximate them with nodal > quadrature (QTrap in libMesh IIRC). Each quadrature point falls on a > node, and only one Lagrange shape function is non-zero at that node, > so you only get contributions to the equation for that shape function > so your mass matrix is diagonal. > --- > Roy > ------------------------------------------------------------------------------ _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
