6
$\begingroup$

We consider block matrices $$\mathcal A = \begin{pmatrix} 0 & A\\A^* & 0 \end{pmatrix}$$ and $$\mathcal B = \begin{pmatrix} 0 & B\\C & 0 \end{pmatrix}.$$

Then we define the new matrix $$T(t) = \begin{pmatrix} \mathcal A+t & \mathcal B \\ \mathcal B^* & \mathcal A-t\end{pmatrix}.$$

Numerical experiments seem to show that the eigenvalues of $[0,\infty) \ni t\mapsto T(t)$ have the property that their absolute values are monotonically increasing in $t \ge 0.$ However, I do not have a proof of this, does anybody know how this follows? (The eigenvalues of $T(t)$ seem to come in pairs $\pm \lambda$ with $\lambda = \lambda(t) \ge 0$, i.e. $+\lambda(t)$ is increasing, while $-\lambda(t)$ is decreasing.

To illustrate the effect, consider

$$T(t)=\begin{pmatrix} t & 1& 0& 2\\ 1 & t & 0& 0\\ 0 & 0& -t & 1\\ 2& 0 & 1 & -t \end{pmatrix},$$ then the eigenvalues of $T(t)$ are $$ \pm 1 \mp \sqrt{2+t^2}.$$

Please let me know if anything is unclear.

$\endgroup$

1 Answer 1

6
$\begingroup$

The idea is to apply a unitary congruence $$U=\dfrac{1}{\sqrt{2}}\begin{pmatrix}I&-I\\I&I\end{pmatrix}.$$ I consider here $\mathcal{B}$ to be hermitian, $\mathcal{B}=\mathcal{B}^*$, whereas the general case may be 'different'.

So $$R=UT(t)U^*=\begin{pmatrix}\mathcal{A}-\mathcal{B}&tI\\tI&\mathcal{A}+\mathcal{B}\end{pmatrix}.$$ Similarly $$ R-\lambda I=\begin{pmatrix}-\lambda I&(A-B)&tI&0\\(A^*-C)&-\lambda I&0&tI\\tI&0&-\lambda I &(A+B)\\0&tI&(A^*+C)&-\lambda I\end{pmatrix}. $$ Using the well known determinant formula for block matrices with a commuting off-diagonal block, you obtain that the eigenvalues $\lambda$ satisfy $$\det\left(\begin{pmatrix}-\lambda I&F\\G&-\lambda I\end{pmatrix}\begin{pmatrix}-\lambda I&2A-F\\2A^*-G&-\lambda I\end{pmatrix}-t^2I\right)=0.$$ Equivalently, $$\det\begin{pmatrix}(\lambda^2-t^2)I +F(2A^*-G)&-2\lambda A\\-2\lambda A^*&(\lambda^2-t^2)I+G(2A-F)\end{pmatrix}=0.$$

The monotonicity of the eigenvalues follows and $-\lambda$ is also an eigenvalue since $$\begin{pmatrix}Q&X\\Y&Z\end{pmatrix}$$ is unitarily congruent to $$\begin{pmatrix}Q&-X\\-Y&Z\end{pmatrix}.$$

If $\mathcal{B}$ is not hermitian the block matrix $$T(t)=\begin{pmatrix}tI&A&0&B\\A^*&tI&C&0\\0&C^*&-tI&A\\B^*&0&A^*&-tI\end{pmatrix}$$ may not be similar to $-T(t)$ for $2\times 2$ blocks, $t=2$, $A=\begin{pmatrix}1&0\\0&1\end{pmatrix}, B=\begin{pmatrix}0&0\\0&5\end{pmatrix}$ and $C= \begin{pmatrix}0&0\\3&0\end{pmatrix}.$

The property seems to hold for $T(t)\in \mathbb{M}_4(\mathbb{C})$, consider the characteristic polynomial of $T(t)$ and $-T(t)$.

$\endgroup$
5
  • $\begingroup$ Thanks, so indeed it seems to hinge on $B$ being hermitian?-Even though the result might not... $\endgroup$
    – Sascha
    Mar 16, 2022 at 17:08
  • $\begingroup$ Welcome. Yes this is for $\mathcal{B}$ hermitien, (in general no clue). You could give an example in the general case maybe this helps. $\endgroup$
    – Toni Mhax
    Mar 16, 2022 at 17:22
  • $\begingroup$ thanks, yes, i added a slightly more general example. $\endgroup$
    – Sascha
    Mar 16, 2022 at 17:35
  • $\begingroup$ First your problem seems for $n\times n$ block matrices so you may want to keep the initial statement and give the example in $2 \times 2$ like you did. $\endgroup$
    – Toni Mhax
    Mar 16, 2022 at 17:51
  • $\begingroup$ thanks, indeed the initial statement was more general, but maybe this scenario sheds some light on what we are missing? :) $\endgroup$
    – Sascha
    Mar 16, 2022 at 18:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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