Solving multiple equations in a mixed formulation in dolfinx

variational formulation
1

Hey,

I need to solve a set of 3 scalar equations and 1 vector equation using a mixed formulation in a monolithic solver. My approach works fine if I have one scalar equation and one vector equation (basically following the tutorials). I am using dolfinx version ‘0.6.0’, and extending the tutorial from the Poisson problem using a mixed formulation to my case through:

msh = create_unit_square(MPI.COMM_WORLD, 96, 96, CellType.quadrilateral)
U1 = ufl.FiniteElement("Lagrange", msh.ufl_cell(), 2)
R1 = ufl.FiniteElement("Lagrange", msh.ufl_cell(), 1)
ME = FunctionSpace(msh, R1 * U1 * R1 * R1)

Next, I define the trial and test functions as

u, r, th, phi = ufl.TrialFunctions(ME)
w, q, et, psi = ufl.TestFunctions(ME)

However I get the error:

ValueError: not enough values to unpack (expected 4, got 2)

Is this the appropriate approach to use a mixed formulation for a system of multiple equations? My full problem consists on the solution of 8 coupled equations, so I need to define a clear rationale for the implementation. Maybe the mixed form is not suitable for these problems? Are there any references I can learn from?

Thank you very much in advance.

2

I would recommend using the syntax:

U1 = ufl.FiniteElement("Lagrange", msh.ufl_cell(), 2)
R1 = ufl.FiniteElement("Lagrange", msh.ufl_cell(), 1)
ME = FunctionSpace(msh, ufl.MixedElement([R1, U1,  R1, R1]))

as multiplication of elements is confusing and misleading (as shown in your example), and is deprecated: Remove `*` overload in `FiniteElementBase` for creating `MixedElement` · Issue #60 · FEniCS/ufl · GitHub

3

Hi dokken, that sintax is not working anymore in Fenicsx v0.9.0. I would like to know when you plan to stop changing the interphase as it is cumbersome to maintain actual working code or even to follow the examples between different versions.

Thanks

4

While I understand your frustration, those interface changes are not done for the sake of doing them, but there is a rationale (typically, improving the codebase, or the performance, or supporting new features). For the specific case of the mixed formulation, please search the forum (or just look at the demos of 0.9.0) to find the new syntax, as this question has surely been asked several times.

Finally, as an administrator of the forum and as a member of the FEniCS steering council, I must remind you that we expect all users to follow the FEniCS code of conduct. The very short summary of it is “be respectful”, while your message comes across as very rude.

5

The new syntax is:

U1 = basix.ufl.element("Lagrange", msh.basix_cell(), 2)
R1 = basix.ufl.element("Lagrange", msh.basix_cell(), 1)
ME = dolfinx.fem.functionspace(msh, basix.ufl.mixed_element([R1,U1, R1, R1])

At every release, I try to list all the API changes in:

and I keep several tutorials covering base aspects of FEM up to date:

While I understand your frustration with the API changes, some of them are necessary to keep the code-base maintainable, performant and allowing new features, as @francesco-ballarin mentioned.

1 Like
6

I apologize if my commment has been misinterpreted. This comment is far from being rude and my intension was to get to the right documentation as many users has been asking about this, but those questions have been resolved for previous versions of FenicsX, as far as I was investigating. Thanks for sharing the links to the updated documentation and for the great work and effort you are investing into maintaining FenicsX.

Again apologies and thanks.