Let $A\in M_n(\mathbb{R})$ be a real symmetric matrix that is strictly diagonal dominant (sdd) with positive diagonal values, that is, $$\forall i, a_{ii} = |a_{ii}|>\sum_{j\neq i}|a_{ij}|.$$ I know that a sdd matrix is invertible and that if $\lambda_i$ is an eigenvalue of $A$ then we have that $$\forall i,|\lambda_i-a_{ii}|\leq \sum_{j\neq i}|a_{ij}|.$$

I guess, that we could simply diagonalize $A = PDP^{t}$ with $D= \text{diag}(\lambda_1,\lambda_2,\cdots,\lambda_n).$ Then if we can show that $A$ is strictly diagonal dominant (with positive diagonal values) iff $D$ is strictly diagonal dominant (with positive diagonal values) then we are done. I am not sure how to proceed with this part of the proof.


You can do it explicitly: let $ 0 \neq x= (x_1,\ldots, x_n)^t \in \mathbb R^n$, and consider \begin{align*} x^t A x &= \sum^n_{i=1}\sum^n_{j=1} a_{ij}x_ix_j \\ &= \sum^n_{i=1} x_i^2a_{ii} + 2 \sum_{i=1}^n \sum^{n}_{j=i+1} x_i x_j a_{ij}\\ &\ge \sum^n_{i=1} x_{i}^2a_{ii} - \sum_{i=1}^n \sum^{n}_{j=i+1} (x_i^2 + x_j^2) \lvert a_{ij} \rvert \\ &= \sum_{i=1}^n \left[\left(a_{ii} - \sum_{i\neq j} \lvert a_{ij}\rvert \right)x_i^2\right] > 0, \end{align*} since $\left(a_{ii} - \sum_{i\neq j} \lvert a_{ij}\rvert \right) > 0$ for all $i$, and $x_i^2>0$ for at least one $i$.


From your observations, $$ \forall i, |\lambda_i - a_{ii}| < a_{ii} $$ with $a_{ii}>0$.

We need to show that $\lambda_i > 0$.

if $\lambda_i \leq 0$, i.e, $\lambda_i = -|\lambda_i|$, $$ |\lambda_i - a_{ii}| = |-1 \times (|\lambda_i| + a_{ii})| = |\lambda_i| + a_{ii} $$


$$ |\lambda_i| + a_{ii} \geq a_{ii} $$

Therefore, $\lambda_i>0$ by contradiction.


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.