Consider some $n\times n$ matrix $\mathbf{A}$; denote its dominant (a.k.a. leading) eigenvalue $\lambda_{A,d}$. Consider another matrix $${B} = \left[ \begin{array}[cc] \ \mathbf{A} & \vec{c} \\ \vec{r}^{\top} & k \end{array}\right]$$

where $\vec{c}$ and $\vec{r} $ are column vectors from $\mathbb{R}^N$ and $k\in\mathbb{R}$. Denote the dominant eigenvalue of B by $\lambda_{B,d}$.

Does it then follow that $\text{Re}(\lambda_{A,d})\leq \text{Re}(\lambda_{B,d})$? I ran some simulations that seem to suggest this is true, but I am having trouble proving this.

Note: I've been considering the case where $\text{Re}(\lambda_{A,d})>0$. I don't think it matters, but if it does and/or makes your life easier, then please do feel free to impose that.

Thanks so much for taking the time to at least read this!


$$ \begin{pmatrix} 1 & -1\\ 2 & -1 \end{pmatrix} $$ The eigenvalues are $i$ and $-i$, whose real parts are zero, that is less than 1.

Notice: what you said is true for positive definite hermitian matrices

  • $\begingroup$ Hi Exodd! Thanks for getting back to me so quickly on this, I really appreciate it. I'm not quite sure I understand the first part of your explanation (I assumed A was just an arbitrarily-chosen, real valued matrix). Also, is what I said necessarily true for matrices that are not positive definite and/or not hermitian? Thanks again! $\endgroup$ Sep 13 '20 at 0:16
  • $\begingroup$ @cluelessmathematician The matrix from the beginning of the answer is a matrix $B$, the matrix constituted by the left upper cell of $B$ is the matrix $A$. Next are presented eigenvalues of $B$ and $A$. $\endgroup$ Sep 23 '20 at 11:56
  • $\begingroup$ Hi Alex, thanks for responding to this. Unless I am missing something, I still don't see how this helps me prove what I am asking, since it's just an example. $\endgroup$ Sep 23 '20 at 20:49
  • $\begingroup$ @cluelessmathematician It is a COUNTER-example. It proves that what you are trying to prove is wrong $\endgroup$
    – Exodd
    Sep 24 '20 at 7:47
  • $\begingroup$ Yikes, you're right. Sorry about that, Exodd! $\endgroup$ Sep 24 '20 at 20:03

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.