Issue when applying MPC and dirichletbc to the same dofs

[Libinatos]

Hello, I am dealing with a problem where mpc and dirichletbc need to be applied to the same dofs. When I searched related posts, I found a post which made me confused:

@PBC

was trying to use dolfinx_MPC to apply periodic boundary conditions together with Dirichlet bcs, something like
u3 = u2 + u4, with u4=10.
However, the results show that only part of the Dirichlet BCS are applied. Can create_general_constraint be used with fem.dirichletbc together?

@dokken

That kind of behavior is currently not supported.

I started some work with this a while back, But I didn’t have time to finish the user interface for it.

@PBC

Thanks. Can I use dolfinx_MPC to apply MPC and assemble matrices? Then, once the Jacobian matrices and residual are obtained, then manually apply the Dirichlet conditions. Is it feasible?

@dokken

No. As you can see in the readme of dolfinx MPC, it computes K^TAK to eliminate dofs from the system, thus applying it afterwards won’t enforce it in a constraint.

If my understanding is correct, after computing K^TAK, the slave dofs (like u3 in the original post) will be eliminated, but the master dofs (like u4) will remain in the system. So why cannot we apply dirichletbc to the master dofs in the new system, e.g., following the standard lifting procedure?

[dokken]

@Libinatos

If my understanding is correct, after computing K^TAK, the slave dofs (like u3 in the original post) will be eliminated, but the master dofs (like u4) will remain in the system. So why cannot we apply dirichletbc to the master dofs in the new system, e.g., following the standard lifting procedure?

This is not possible as I have not had time to implement this into DOLFINx_MPC.

[Libinatos]

Got it. Thank you very much!