I have a asymmetric diagonally dominant matrix $A=M+k*I$, where $M$ is bounded, non-negative and asymmetric, $k$ could be arbitrarily large so that $A$ is always diagonally dominant, and a bounded strictly positive semi definite matrix $B$.

I am interested in the range of absolute value of eigenvalues of $C=A^{-1}B$. My conjecture is that $$|\lambda_i|\in(0,\epsilon)$$ for any $\epsilon>0$, provided we let $k$ to infinity. However, due to $M$ matrix is asymmetric, we cannot conclude that A is strictly positive semi-definite.

Any idea on proving my conjecture, or showing that my conjecture is wrong?

  • $\begingroup$ This is not true. If $B$ is singular, so is $C$ and hence some $\lambda_i$ is zero. $\endgroup$ – user1551 May 18 '18 at 6:21
  • $\begingroup$ B is strictly PSD, so all eigenvalue are strictly positive $\endgroup$ – M.Z. May 18 '18 at 20:01

Use the Gershgorin theorem to say that all eigenvalues of $A$ lie in the disk $\{z\in\mathbb{C}:|z-k|\leq \max_{1\leq i\leq n}\sum_{j=1}^{n}|M_{i,j}|\}$, so that the smallest possible absolute value of any eigenvalue of $A$ is $k-\max_{1\leq i\leq n}\sum_{j=1}^{n}|M_{i,j}|$. Then the eigenvalues of $A^{-1}$ lie in the disk $\{z\in\mathbb{C}:|z|\leq (k-\max_{1\leq i\leq n}\sum_{j=1}^{n}|M_{i,j}|)^{-1}\}$. Clearly as $k\rightarrow\infty$, this radius shrinks to zero, and because $B$ is bounded, this means $A^{-1}B\rightarrow0.$ By continuity of the eigenvalues, all eigenvalues must also tend to 0 as $k\rightarrow\infty$.


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.