fenics team mailing list archive

[Hans Petter Langtangen <hpl@xxxxxxxxx>]

Fri, 18 Jun Anders Logg wrote:
> On Fri, Jun 18, 2010 at 04:35:37PM +0200, Hans Petter Langtangen wrote:
> > Fri, 18 Jun Anders Logg wrote:
> > > 3. Is the normal flux + or - grad(u).n? I think of it as -grad(u).n.
> >
> > For k*inner(grad(u), grad(v))*dx, -k*grad(u).n is the right flux. It has
> > a minus sign. Some people use the word flux for grad(u).n, but that's
> > sloppy - normal derivative is then better.
> >
> > Hans Petter
> 
> I agree that -k*grad(u).n is the flux, but perhaps the question is
> whether to include the term "flux" at all in the Poisson demo. If it
> is included, then we should have a minus sign, both in the definition
> of g = -k*grad(u).n and in the form where we get -<g, v>*ds. But if we
> don't mention "flux" then it looks strange with two minus signs.

I agree - in the Poisson demo we skip "flux" and use "normal derivative"
instead and define it as grad(u).n.

Hans Petter