John, I think what you are doing there is conceptually flawed. First, if you add constraints, you could rather use the corresponding continuous element. If you are planning to add them as Lagrange multipliers, like in a hybridized method, you should consider using integration on the face.
Quite generally, if you try to associate shape functions with points on your mesh, you are leaving the finite element framework. There might be particular cases, where this is necessary, but in my opinion it is worth thinking twice before doing anything like this. If you think you need to use this, I would be very interested in learning about your method. Best, Guido John Chapman wrote: > I am using discontinuous elements on a regular mesh. Given a degree > of freedom that I know lies on a given cell and face (technically has > support on that face, as this is dG) I would like to find out which > dof is its neighbour i.e. the equivalent dof from the cell adjoining > across the given face. > > It is not difficult to loop through all of the cells, then faces then > dofs and find the dofs with support on a given face. It is not > difficult to then extract the neighbouring cell and find which dofs > from that cell also lie on the face. However, what is the best way to > find which dof matches which dof. For example, in 2d using linear dG > elements each face has 2 dofs with support on that face from one cell, > and two from the neighbouring cell. I would like to know how they are > matched so I can add constraints. > > There are several very long ways to do this, but does anyone know of a > deal.ii function which can be given a cell, face and dof and return > the equivalent dof from the neighbouring cell? I could write my own > loop by either extracting the true location of the dof, or by using > knowledge of how the dofs are numbered (i.e. in x first, then y, then > z if it exists), but would prefer to use a pre-existing function. > > If any of this isn't clear, I can post some code, but I have avoiding > doing so now as what I have doesn't currently do what I want! > > Thanks, > > John > _______________________________________________ > dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
