Paper 2, Section II, E

Numerical Analysis | Part II, 2015

(a) The boundary value problem Δu+cu=f-\Delta u+c u=f on the unit square [0,1]2[0,1]^{2} with zero boundary conditions and scalar constant c>0c>0 is discretised using finite differences as

ui1,jui+1,jui,j1ui,j+1+4ui,j+ch2ui,j=h2f(ih,jh),i,j=1,,m\begin{array}{r} -u_{i-1, j}-u_{i+1, j}-u_{i, j-1}-u_{i, j+1}+4 u_{i, j}+c h^{2} u_{i, j}=h^{2} f(i h, j h), \\ i, j=1, \ldots, m \end{array}

with h=1/(m+1)h=1 /(m+1). Show that for the resulting system Au=bA u=b, for a suitable matrix AA and vectors uu and bb, both the Jacobi and Gauss-Seidel methods converge. [You may cite and use known results on the discretised Laplace operator and on the convergence of iterative methods.]

Define the Jacobi method with relaxation parameter ω\omega. Find the eigenvalues λk,l\lambda_{k, l} of the iteration matrix HωH_{\omega} for the above problem and show that, in order to ensure convergence for all hh, the condition 0<ω10<\omega \leqslant 1 is necessary.

[Hint: The eigenvalues of the discretised Laplace operator in two dimensions are 4(sin2πkh2+sin2πlh2)4\left(\sin ^{2} \frac{\pi k h}{2}+\sin ^{2} \frac{\pi l h}{2}\right) for integers k,lk, l.]

(b) Explain the components and steps in a multigrid method for solving the Poisson equation, discretised as Ahuh=bhA_{h} u_{h}=b_{h}. If we use the relaxed Jacobi method within the multigrid method, is it necessary to choose ω1\omega \neq 1 to get fast convergence? Explain why or why not.

Typos? Please submit corrections to this page on GitHub.