From: Roy Stogner <roystgnr@ic...>  20080304 14:16:05

On Tue, 4 Mar 2008, Maxime Debon wrote: > Looking for tips to check the correctness of Ke matrix and Fe vector > in an handbook, I found this rule : A0 * u'' = (A0 * u')'  A0' * u' > > This latest transformation must be taken into account to adjust a > nonconstant coefficient to the diffusion term. It also explains why > the result was correct with a constant coefficient. Oh! I didn't even think of that. Yeah; if the coefficient is outside both derivatives then you get an extra term from it when integrating by parts. I'm more used to terms like (A0*u')' (which doesn't give you an extra term) than I am to A0*u''. Doublecheck and make sure that the latter is really what models your physical system. Often conservation models have a conservation statement: D/dt(conserved) = div(flux) and a constitutive relation: flux = A*grad(conserved) In which case A really does end up inside one but not both derivatives in the strong form of the resulting equation.  Roy 