Homogenize DirichletBC in dolfinx

dolfinx
1

Hi everyone,

I have a question regarding the DirichletBCs in dolfinx. In legacy FEniCS, there was the option to do the following

from fenics import *
import numpy as np


def boundary(x, on_boundary):
    return on_boundary


mesh = UnitSquareMesh(8, 8)
V = FunctionSpace(mesh, "CG", 1)

u = Function(V)
u.vector()[:] = 2.0  # set u to constant 2 function
u_hom = Function(V)
u_hom.vector()[:] = 2.0  # set u_hom to constant 2 function

bc = DirichletBC(V, 1.0, boundary)
bc_hom = DirichletBC(bc)  # copy the bc
bc_hom.homogenize()  # set the value to 0

bc.apply(u.vector())  # apply the inhomogeneous bc to u
bc_hom.apply(u_hom.vector())  # apply the homogeneous bc to u_hom

assert np.abs(u.vector().min() - 1.0) < 1e-14
assert np.abs(u_hom.vector().min() - 0.0) < 1e-14

Can I achieve the same in dolfinx? That means, given some (possibly inhomogeneous) DirichletBC, copy the boundary condition and set its value to 0?

I think I can achieve the copy part with

bc = dolfinx.fem.dirichletbc(...)
bc_copy = dolfinx.fem.DirichletBC(bc._cpp_object)

but trying to modify bc_copy.g, where the value of the DirichletBC is stored does fail as the property has no setter implemented. Of course it would also be sufficient to manually build a new value for the DirichletBC and re-create with dolfinx.fem.dirichletbc, in case all the inputs to the function are actually stored in the instance of the created dolfinx.fem.DirichletBC.

Any ideas or tips? Thanks a lot in advance,
Sebastian

2

Would approach this be an option for you?

bc_func = fem.Function(V)
bc_func.interpolate(my_bc_expression)
bc = fem.dirichletbc(bc_func, bc_dofs, V)
...
bc_func.x.array[:] = 0
3

Unfortunately not. I want to be able to do the following: Given some boundary condition object which is created in any feasible way, copy the boundary condition and set the value of the copied object to zero.

I noticed that I can use the method dolfinx.fem.DirichletBC.dof_indices to get the list of all dof indices (and the index where the local part ends and the global one starts). However, when I want to try to get the underlying function space with dolfinx.fem.DirichletBC.function_space, not the dolfinx.fem.FunctionSpace but the dolfinx.cpp.fem.FunctionSpace is returned. To give the BC a homogeneous value, I would need to be able to access the python object - or create a cpp Dirichlet boundary condition with a dolfinx.cpp.fem.Function which is kind of messy. Also I am not sure if this works robustly in parallel - so whether the dofs returned by dolfinx.fem.DirichletBC.dof_indices can be used as input for the dofs parameter in dolfinx.fem.dirichletbc (or only the local portion of dofs…)

4

I realized that the parameter alpha for setting the values of the right-hand side can, in principle, be used to achieve this. Setting this to alpha = 0.0 means that the values on the right-hand side will be set to zero.

However, this is not exactly what I was looking for. I would rather need a copy of the DirichletBC object with a different function value but the same dofs / boundaries that are affected.

1 Like
5

Why do you need an explicit copy of the object? Using alpha=0 (scale on main: dolfinx/python/dolfinx/fem/assemble.py at 2ac742c289cdc22acf04a329015a3589047d4291 · FEniCS/dolfinx · GitHub) is the most efficient way of achieving a homogenized bc.

A homogenised bc with the lifting approach means that you don’t need to call lifting at all, only set bc with alpha=0 to ensure that you get a homogenised bc.