From: Roy Stogner <roy@st...>  20080812 22:59:05

On Tue, 12 Aug 2008, Nasser Mohieddin Abukhdeir wrote: > I'm trying to implement Dirichlet boundary conditions in a 1D > problem using DiffSystem. There isn't really an example to cover this, > so I just want to verify what I came up with makes sense, where two > variables "B" and "lB" should be zero at the domain boundaries x=0,1 > (note, using numerical jacobians for now): At first glance this looks correct, but it may be overkill. It's actually not necessary to do anything special in 1D. You might not think that a 1D element has "sides" that you can integrate over, but in recent (0.6.0 or so?) libMesh versions we now have a "NodeElem" object whose job is to let you do just that: Dirichlet boundary conditions (or any boundary conditions) expressed as quadrature loops on sides now work correctly even when the side is a single point. I'd add a 1D case to example 18 to make this clear, but 1D incompressible flow is so dull that people would wonder why the option was even there.  Roy 