I'm not entirely certain about how to tackle this problem.... I hope you ladies and gents can help :)

If $M\in M_{n\times n}(\mathbb{R})$ be such that every row has precisely tow non-zero entries, one is precisely equal to $1$ and the other is found in the diagonal and is strictly greater than one. Must $M$ be invertible?

My thoughts to date :)

I believe the answer to be yes; reasoning: Intuition: for $n\leq 2$ $M$ can be readily calculated directly.

Proof sketch idea: For arbitrary large $n$, I was thinking using the mini-max theorem to obtain a lower-bound on the smallest eigenvalue; and since all the non-zero entries are sufficiently large (at least 1); I would be done since then all eigenvalues must be strictly positive.... (But is the matrix Hermitian and how can I calculate this explicitly?)


Yes it is. Apply Gershgorin's Theorem.

  • 1
    $\begingroup$ Thanks I'll post my solution :) $\endgroup$ – AIM_BLB May 30 '15 at 3:27
  • $\begingroup$ Very pretty theorem :) $\endgroup$ – muaddib May 30 '15 at 3:37
  • 1
    $\begingroup$ I agree! Its one of the coolest an most visual I've every seen in linear algebra :) Thanks for the link :D $\endgroup$ – AIM_BLB May 30 '15 at 3:45

A corollary to Gershgorin's theorem states:

More generally: let $M\in M_{n\times n}(\mathbb{R})$ be such that: \begin{equation} (\forall i\in {1,..,n}) (\sum_{j\neq i} a_{i,j}) < a_{i,i}. \end{equation}

However, Gershgorin's theorem implies that each eigenvalue $\lambda$ of $M$ satisfies: \begin{equation} \lambda \in \cup_{i=1}^n Ball_{(\sum_{j\neq i} a_{i,j})}( a_{i,i}). \end{equation} However the first equation implies that $0\notin \cup_{i=1}^n Ball_{(\sum_{j\neq i} a_{i,j})}( a_{i,i})$ and in particular the disc in question lies in the first quadrant of the complex plain.

Therefore every eigenvalue $\lambda$ of $M$ satisfies: $min\{Re(\lambda),Im(\lambda)\} >0$ and in particular $M$ is in $GL_n(\mathbb{R})$.


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.