3.II.38C

Numerical Analysis
Part II, 2007

(a) Prove that all Toeplitz symmetric tridiagonal M×MM \times M matrices

A=[ab00bab00bab00ba]A=\left[\begin{array}{ccccc} a & b & 0 & \cdots & 0 \\ b & a & b & \ddots & \vdots \\ 0 & \ddots & \ddots & \ddots & 0 \\ \vdots & \ddots & b & a & b \\ 0 & \cdots & 0 & b & a \end{array}\right]

share the same eigenvectors (v(k))k=1M\left(v^{(k)}\right)_{k=1}^{M} with components vi(k)=sinkiπM+1v_{i}^{(k)}=\sin \frac{k i \pi}{M+1}, i=1,,Mi=1, \ldots, M, and eigenvalues to be determined.

(b) The diffusion equation

ut=2ux2,0x1,0tT\frac{\partial u}{\partial t}=\frac{\partial^{2} u}{\partial x^{2}}, \quad 0 \leqslant x \leqslant 1, \quad 0 \leqslant t \leqslant T

is approximated by the Crank-Nicolson scheme

umn+112μ(um1n+12umn+1+um+1n+1)=umn+12μ(um1n2umn+um+1n), for m=1,,M,\begin{aligned} u_{m}^{n+1}-\frac{1}{2} \mu\left(u_{m-1}^{n+1}-2 u_{m}^{n+1}+u_{m+1}^{n+1}\right) &=u_{m}^{n}+\frac{1}{2} \mu\left(u_{m-1}^{n}-2 u_{m}^{n}+u_{m+1}^{n}\right), \\ \text { for } \quad m &=1, \ldots, M, \end{aligned}

where μ=Δt/(Δx)2,Δx=1/(M+1)\mu=\Delta t /(\Delta x)^{2}, \quad \Delta x=1 /(M+1), and umnu_{m}^{n} is an approximation to u(mΔx,nΔt)u(m \Delta x, n \Delta t). Assuming that u(0,t)=u(1,t)=0,t,u(0, t)=u(1, t)=0, \forall t, \quad show that the above scheme can be written in the form

Bun+1=Cun,0n(T/Δt)1B u^{n+1}=C u^{n}, \quad 0 \leqslant n \leqslant(T / \Delta t)-1

where un=[u1n,,uMn]u^{n}=\left[u_{1}^{n}, \ldots, u_{M}^{n}\right]^{\top} and the real matrices BB and CC should be found. Using matrix analysis, find the range of μ\mu for which the scheme is stable. [Do not use Fourier analysis.]