Hi James,

I will let David or Lawrence answer your question properly, as I'm not sure about the current status of the Real function space in Firedrake.  (There is an open pull request here: https://github.com/firedrakeproject/firedrake/pull/909, but this is somewhat out of date)

However, you might be able to solve your problem an alternative way, using the nullspace feature, as described here: http://www.firedrakeproject.org/solving-interface.html#solving-singular-systems.

Best,
Andrew

On 19 April 2017 at 15:09, James Jackaman <wb816921@live.reading.ac.uk> wrote:

Dear all,


I'm currently working on an Firedrake implementation which utilises Lagrange multipliers, in the spirit of the Poisson equation with pure Neumann boundary conditions where the average value of the solution is constrained by a Lagrange multiplier.

I have attempted to translate the Fenics demo, see https://fenicsproject.org/olddocs/dolfin/1.6.0/python/demo/documented/neumann-poisson/python/documentation.html, into Firedrake with no success, see neumann-poisson.py in the attached gist.

I believe there may be a problem with the function space of real constants FunctionSpace(mesh,"R",0), see real_solver.py in the attached gist which returns u=0 instead of u=1.

How can I modify my example codes in the gist such that they behave as expected, or alternatively how can I implement the Poisson equation with pure Neumann boundary conditions using Lagrange multipliers in Firedrake?


Gist: https://gist.github.com/j130792/64d12e84cf5ede2b327d1859a36dc36d


Thanks for your help,

James