According to wikipedia, a totally positive matrix (https://en.wikipedia.org/wiki/Totally_positive_matrix) always have non-negative eigenvalues. But, is it necessary for a matrix to be totally positive to have non-negative eigenvalues?

I'm looking for the necessary and sufficient condition for a real square matrix (not necessarily symmetric) to have non-negative eigenvalues.

  • $\begingroup$ You may want to look up diagonal dominance and Gershgorin's circle theorems. $\endgroup$ – bcf Jun 3 '16 at 2:18
  • $\begingroup$ @bcf : Thank you. Gershgorin's circle theorem looks like a good starting point. $\endgroup$ – Sachin Jun 3 '16 at 3:00

No it is not necessary. Just consider the matrix $$ A= \begin{bmatrix} 2 & -1\\ -1 & 2 \end{bmatrix} $$ which is not totally positive but it is still has non-negative eigenvalues $\{1,3\}.$

  • $\begingroup$ Is there any necessary condition then? $\endgroup$ – Sachin Jun 3 '16 at 5:35
  • $\begingroup$ Yes, you can get many necessary conditions. e.g. for a square matrix $A,$ $det(A)\geq 0$ or $Tr(A) \geq 0$ are necessary conditions (although these are not sufficient) $\endgroup$ – A. Ray Jun 3 '16 at 5:44

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.