[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.3.1

Let $A=A_1\oplus A_2$. Show that

(1). $W(A)$ is the convex hull of $W(A_1)$ and $W(A_2)$; i.e., the smallest convex set containing $W(A_1)\cup W(A_2)$.

(2). $$\beex \bea \sen{A}&=\max\sed{\sen{A_1},\sen{A_2}},\\ \spr(A)&=\max\sed{\spr(A_1),\spr(A_2)},\\ w(A)&=\max\sed{w(A_1),w(A_2)}. \eea \eeex$$

Solution.

(1). We have $$\beex \bea W(A)&=\sed{x^*Ax;\sen{x}=1}\\ &=\sed{y^*A_1y+z^*A_2z;\sen{y}^2+\sen{z}^2=1}\\ &\supset W(A_1)\cup W(A_2), \eea \eeex$$ and $$\bex W(A)=\sed{\sen{y}^2 \sex{\frac{y}{\sen{y}}}^*A_1\frac{y}{\sen{y}} +\sen{z}^2 \sex{\frac{z}{\sen{z}}}^*A_2\frac{z}{\sen{z}}; \sen{y}^2+\sen{z}^2=1} \eex$$ is contained in any convex set containing $W(A_1)\cup W(A_2)$.

(2). $$\beex \bea \sen{Ax}^2&=\sen{\sex{A_1y\atop A_2z}}^2\quad\sex{x=\sex{y\atop z}}\\ &=\sen{A_1y}^2+\sen{A_2z}^2\\ &\leq \sen{A_1}^2\sen{y}^2+\sen{A_2}^2\sen{z}^2\\ &\leq \max\sed{\sen{A_1},\sen{A_2}}^2 \sex{\sen{y}^2+\sen{z}^2}\\ &=\max\sed{\sen{A_1},\sen{A_2}}^2 \sen{x}^2. \eea \eeex$$ $$\beex \bea &\quad Ax=\lm x\quad\sex{x\neq 0}\\ &\ra A_1y=\lm y,\quad A_2z=\lm z\quad\sex{x=\sex{y\atop z}}\\ &\ra |\lm|\leq\sedd{\ba{ll} \spr(A_1),&y\neq 0\\ \spr(A_2),&z\neq 0 \ea}\\ &\ra |\lm|\leq \max\sed{\spr(A_1),\spr(A_2)};\\ &\quad A_1y=\lm y\quad\sex{y\neq 0}\\ &\ra A\sex{y\atop 0}=\lm \sex{y\atop 0}\\ &\ra |\lm|\leq \spr(A);\\ &\quad A_2z=\lm z\quad\sex{z\neq 0}\\ &\ra |\lm|\leq \spr(A). \eea \eeex$$ $$\beex \bea w(A)&=\sup_{\sen{x}=1}\sev{\sef{x,Ax}}\\ &=\sup_{\sen{y}^2+\sen{z}^2=1} \sev{\sef{y,A_1y}+\sev{z,A_2z}}\\ &\leq \sup_{\sen{y}^2+\sen{z}^2=1} \sez{ \sen{y}^2w(A_1)+\sen{z}^2w(A_2) }\\ &\leq \max\sed{w(A_1),w(A_2)};\\ w(A_1)&=\sup_{\sen{y}=1}\sen{\sef{y,A_1y}}\\ &=\sup_{\sen{\sex{y\atop 0}}=1} \sev{\sef{\sex{y\atop 0},A\sex{y\atop 0}}}\\ &\leq w(A),\\ w(A_2)&\leq w(A). \eea \eeex$$

时间: 2024-10-10 02:39:52

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.3.1的相关文章

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.4.1

Let $x,y,z$ be linearly independent vectors in $\scrH$. Find a necessary and sufficient condition that a vector $w$ mush satisfy in order that the bilinear functional $$\bex F(u,v)=\sef{x,u}\sef{y,v}+\sef{z,u}\sef{w,v} \eex$$ is elementary. Solution.

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.3.7

For every matrix $A$, the matrix $$\bex \sex{\ba{cc} I&A\\ 0&I \ea} \eex$$ is invertible and its inverse is $$\bex \sex{\ba{cc} I&-A\\ 0&I \ea}. \eex$$ Use this to show that if $A,B$ are any two $n\times n$ matrices, then $$\bex \sex{\ba{c

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.5.5

Show that the inner product $$\bex \sef{x_1\vee \cdots \vee x_k,y_1\vee \cdots\vee y_k} \eex$$ is equal to the permanent of the $k\times k$ matrix $\sex{\sef{x_i,y_j}}$. Solution. $$\beex \bea &\quad \sef{x_1\vee \cdots \vee x_k,y_1\vee \cdots \vee y

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.5.1

Show that the inner product $$\bex \sef{x_1\wedge \cdots \wedge x_k,y_1\wedge \cdots\wedge y_k} \eex$$ is equal to the determinant of the $k\times k$ matrix $\sex{\sef{x_i,y_j}}$. Solution. $$\beex \bea &\quad \sef{x_1\wedge\cdots \wedge x_k,y_1\wedg

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.1.2

Let $X$ be nay basis of $\scrH$ and let $Y$ be the basis biorthogonal to it. Using matrix multiplication, $X$ gives a linear transformation from $\bbC^n$ to $\scrH$. The inverse of this is given by $Y^*$. In the special case when $X$ is orthonormal (

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.5.10

Every $k\times k$ positive matrix $A=(a_{ij})$ can be realised as a Gram matrix, i.e., vectors $x_j$, $1\leq j\leq k$, can be found so that $a_{ij}=\sef{x_i,x_j}$ for all $i,j$. Solution. By Exercise I.2.2, $A=B^*B$ for some $B$. Let $$\bex B=(x_1,\c

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.2.6

If $\sen{A}<1$, then $I-A$ is invertible, and $$\bex (I-A)^{-1}=I+A+A^2+\cdots, \eex$$ aa convergent power series. This is called the Neumann series. Solution.  Since $\sen{A}<1$, $$\bex \sum_{n=0}^\infty \sen{A}^n=\frac{1}{1-\sen{A}}<\infty. \ee

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.2.8

For any matrix $A$ the series $$\bex \exp A=I+A+\frac{A^2}{2!}+\cdots+\frac{A^n}{n!}+\cdots \eex$$ converges. This is called the exponential of $A$. The matrix $A$ is always invertible and $$\bex (\exp A)^{-1}=\exp(-A). \eex$$ Conversely, every inver

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.2.7

The set of all invertible matrices is a dense open subset of the set of all $n\times n$ matrices. The set of all unitary matrices is a compact subset of all $n\times n$ matrices. These two sets are also groups under multiplication. They are called th

[Bhatia.Matrix Analysis.Solutions to Exercises and Problems]ExI.2.5

Show that matrices with distinct eigenvalues are dense in the space of all $n\times n$ matrices. (Use the Schur triangularisation) Solution.  By the Schur triangularisation, for each matrix $A$, there exists a unitary $U$ such that $$\bex A=U\sex{\ba