4
$\begingroup$

Let $J$ be a Jordan block matrix with blocks $J_1,\cdots,J_n$. I came up with some examples of $J$ and noticed that $\|J\|=\max_{i=1,\cdots,n}\|J_i\|$. Does this result always hold?

The norm I use here is the induced 2-norm. (I used Matlab for my examples. The norm would be the one that Matlab uses when I type norm(A).)

$\endgroup$
  • 1
    $\begingroup$ Which matrix norm are you using? $\endgroup$ – Eric Towers Apr 4 '16 at 5:20
  • $\begingroup$ No. Take the Frobenius norm. Then $\|J\|^2 = \sum_i \|J_j\|^2$. $\endgroup$ – copper.hat Apr 4 '16 at 5:33
  • $\begingroup$ Sorry I should have clarified. I used Matlab for my examples. The norm would be the one that Matlab uses when I type norm(A). $\endgroup$ – user327878 Apr 4 '16 at 5:44
  • $\begingroup$ That's the induced 2-norm, i.e. the largest singular value. $\endgroup$ – user1551 Apr 4 '16 at 5:49
5
$\begingroup$

The answer is yes. Let $$J = \left( {\begin{array}{*{20}c} \hfill {J_1} & \hfill {} & \hfill {} & \hfill {} \\ \hfill {} & \hfill {J_2} & \hfill {} & \hfill {} \\ \hfill {} & \hfill {} & \hfill \ddots & \hfill {} \\ \hfill {} & \hfill {} & \hfill {} & \hfill {J_n} \\ \end{array}} \right)$$ be your Jordan block matrix. Each Jordan block $J_i$ has a singular value decomposition $J_i = U_i \Sigma _i V_i ^* $, so we can write out a singular value decomposition for $J$ as $$J = \left( {\begin{array}{*{20}c} \hfill {U_1} & \hfill {} & \hfill {} & \hfill {} \\ \hfill {} & \hfill {U_2} & \hfill {} & \hfill {} \\ \hfill {} & \hfill {} & \hfill \ddots & \hfill {} \\ \hfill {} & \hfill {} & \hfill {} & \hfill {U_n} \\ \end{array}} \right)\left( {\begin{array}{*{20}c} \hfill {\Sigma _1} & \hfill {} & \hfill {} & \hfill {} \\ \hfill {} & \hfill {\Sigma _2} & \hfill {} & \hfill {} \\ \hfill {} & \hfill {} & \hfill \ddots & \hfill {} \\ \hfill {} & \hfill {} & \hfill {} & \hfill {\Sigma _n} \\ \end{array}} \right)\left( {\begin{array}{*{20}c} \hfill {V_1^* } & \hfill {} & \hfill {} & \hfill {} \\ \hfill {} & \hfill {V_2^* } & \hfill {} & \hfill {} \\ \hfill {} & \hfill {} & \hfill \ddots & \hfill {} \\ \hfill {} & \hfill {} & \hfill {} & \hfill {V_n^* } \\ \end{array}} \right).$$

The fact that we can express the singular value decomposition of $J$ in compatible block form with the Jordan blocks $J_i$ is essentially the only observation needed to answer your question.

$\left\| J \right\|_2 = \sigma _{\max}(J)$, so assume $\sigma_{max}(J)$ belongs to $\Sigma_k$ (it has to belong to some $\Sigma_i$ for $1\le i\le n$). Then, since $J_k=U_k\Sigma_k V_k^*$, we have that $$\|J\|_2=\sigma_{max}(J)=\|\Sigma_k\|_2=\|J_k\|_2=\max_{i=1,\cdots,n}\|J_i\|_2.$$

Note that this argument does not involve anything specific about the Jordan block structure. It remains valid provided $J$ has any diagonal block structure.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.