From: John Peterson <peterson@cf...>  20041206 15:57:35

Mark Blome writes: > > Hi everybody, > > I am currently developing a finite element geoelectric forward solver and I > am very pleased with the possibilities libmesh gives me for doing this. > For the equation system assembly function I ran across a problem for which I > cannot really find a solution. For the geolelectric problem I have to > introduce a source term associated to a node on the surface of the input mesh > (for the current electrode). The node is correctly marked with a boundary > marker and I can read it in without a problem (its a tetgen mesh). However I > cannot find a way to set the source term in the right hand side correctly > because I can not find out which one of the quadrature points is the one that > corresponds to the node that I marked in the mesh. Am I right that it is not > even guaranteed that one of the quadrature points will lie on the node where > my electrode is attached to? Would that mean that I have to interpolate the > source term from the node position to the nearest quadrature point to get the > correct right hand side value? Is there a way to force one of the quadrature > points to be on the node? Or are my thoughts just going in the wrong > direction ? Thanks for any help on this, Hi, You are correct that the Gaussian quadrature points do not lie on the nodes. There are the trapezoidal and Simpson quadrature rules whose points however do lie at the nodes. If the value of the source term is availabe at every node, you can use the finite element basis functions to interpolate a value at the quadrature point. I assume that it makes sense for your problem for the source term at a given point to be a weighted average of nearby points... John 