Paper 1, Section II, E

Numerical Analysis
Part II, 2021

Let ARn×nA \in \mathbb{R}^{n \times n} with n>2n>2 and define Spec(A)={λCAλI\operatorname{Spec}(A)=\{\lambda \in \mathbb{C} \mid A-\lambda I is not invertible }\}.

The QR algorithm for computing Spec(A)\operatorname{Spec}(A) is defined as follows. Set A0=AA_{0}=A. For k=0,1,k=0,1, \ldots compute the QR\mathrm{QR} factorization Ak=QkRkA_{k}=Q_{k} R_{k} and set Ak+1=RkQkA_{k+1}=R_{k} Q_{k}. (Here QkQ_{k} is an n×nn \times n orthogonal matrix and RkR_{k} is an n×nn \times n upper triangular matrix.)

(a) Show that Ak+1A_{k+1} is related to the original matrix AA by the similarity transformation Ak+1=QˉkTAQˉkA_{k+1}=\bar{Q}_{k}^{T} A \bar{Q}_{k}, where Qˉk=Q0Q1Qk\bar{Q}_{k}=Q_{0} Q_{1} \cdots Q_{k} is orthogonal and QˉkRˉk\bar{Q}_{k} \bar{R}_{k} is the QR factorization of Ak+1A^{k+1} with Rˉk=RkRk1R0\bar{R}_{k}=R_{k} R_{k-1} \cdots R_{0}.

(b) Suppose that AA is symmetric and that its eigenvalues satisfy

λ1<λ2<<λn1=λn\left|\lambda_{1}\right|<\left|\lambda_{2}\right|<\cdots<\left|\lambda_{n-1}\right|=\left|\lambda_{n}\right|

Suppose, in addition, that the first two canonical basis vectors are given by e1=i=1nbiwi\mathbf{e}_{1}=\sum_{i=1}^{n} b_{i} \mathbf{w}_{i}, e2=i=1nciwi\mathbf{e}_{2}=\sum_{i=1}^{n} c_{i} \mathbf{w}_{i}, where bi0,ci0b_{i} \neq 0, c_{i} \neq 0 for i=1,,ni=1, \ldots, n and {wi}i=1n\left\{\mathbf{w}_{i}\right\}_{i=1}^{n} are the normalised eigenvectors of AA.

Let BkR2×2B_{k} \in \mathbb{R}^{2 \times 2} be the 2×22 \times 2 upper left corner of AkA_{k}. Show that dH(Spec(Bk),S)0d_{\mathrm{H}}\left(\operatorname{Spec}\left(B_{k}\right), S\right) \rightarrow 0 as kk \rightarrow \infty, where S={λn}{λn1}S=\left\{\lambda_{n}\right\} \cup\left\{\lambda_{n-1}\right\} and dHd_{\mathrm{H}} denotes the Hausdorff metric

dH(X,Y)=max{supxXinfyYxy,supyYinfxXxy},X,YCd_{\mathrm{H}}(X, Y)=\max \left\{\sup _{x \in X} \inf _{y \in Y}|x-y|, \sup _{y \in Y} \inf _{x \in X}|x-y|\right\}, \quad X, Y \subset \mathbb{C}

[Hint: You may use the fact that for real symmetric matrices U,VU, V we have dH(Spec(U),Spec(V))UV2]\left.d_{\mathrm{H}}(\operatorname{Spec}(U), \operatorname{Spec}(V)) \leqslant\|U-V\|_{2} \cdot\right]