Let A∈Rn×n with n>2 and define Spec(A)={λ∈C∣A−λI is not invertible }.
The QR algorithm for computing Spec(A) is defined as follows. Set A0=A. For k=0,1,… compute the QR factorization Ak=QkRk and set Ak+1=RkQk. (Here Qk is an n×n orthogonal matrix and Rk is an n×n upper triangular matrix.)
(a) Show that Ak+1 is related to the original matrix A by the similarity transformation Ak+1=QˉkTAQˉk, where Qˉk=Q0Q1⋯Qk is orthogonal and QˉkRˉk is the QR factorization of Ak+1 with Rˉk=RkRk−1⋯R0.
(b) Suppose that A is symmetric and that its eigenvalues satisfy
∣λ1∣<∣λ2∣<⋯<∣λn−1∣=∣λn∣
Suppose, in addition, that the first two canonical basis vectors are given by e1=∑i=1nbiwi, e2=∑i=1nciwi, where bi=0,ci=0 for i=1,…,n and {wi}i=1n are the normalised eigenvectors of A.
Let Bk∈R2×2 be the 2×2 upper left corner of Ak. Show that dH(Spec(Bk),S)→0 as k→∞, where S={λn}∪{λn−1} and dH denotes the Hausdorff metric
dH(X,Y)=max{x∈Xsupy∈Yinf∣x−y∣,y∈Ysupx∈Xinf∣x−y∣},X,Y⊂C
[Hint: You may use the fact that for real symmetric matrices U,V we have dH(Spec(U),Spec(V))⩽∥U−V∥2⋅]