I have a question regarding the treatment of Robin boundary conditions in a multigrid solver. I am solving the Poisson equation in $\Omega=(0,1)^2$ with Robin boundary conditions on the boundary, $$- \Delta u = f \text{ in } \Omega$$ and $$\frac{\partial u}{\partial n} - au= g \text{ on } \partial \Omega.$$ I use a finite difference discretization with central differences (a standard five point stencil) for the interior points, and eliminate the boundary conditions so that, for example, at a point $x_C$ on the left boundary with east, south and north neighbors $x_E, x_S, x_N$ the resulting equation is

$$\frac{1}{h^2}\left( 4u(x_C)-u(x_S)-u(x_N)-2u(x_E)\right) - 2\frac{a}{h}u(x_C) = \frac{2}{h}g(x_C)+f(x_C).$$

In chapter 5 of the book 'Multigrid' of Trottenberg et. al., it is explained that if this approach is followed for a Neumann problem ($a=0$) and the full-weight restriction operator $R$ is modified on the boundary points, e.g., on a vertical boundary the modified stencil is $$ \frac{1}{16}\left[\begin{array}{cc} 2 &2 \\ 4&4 \\2&2 \end{array}\right]_{h}^{2h} $$ then a relaxation method can be applied directly on the equations for the boundary points with the eliminated boundary condition (no need of transferring the equation and boundary condition separately). I am using the interpolation operator $I=4R^T$ and Galerkin coarse grid systems.

My question is: Is this multigrid approach correct (modified FW operator on the boundaries + relaxation on the equation with eliminated boundary conditions and Galerkin coarse systems) also for more general Robin boundary conditions, provided a good relaxation method is used?

Thanks in advance.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.