Prove or disprove, if $A\in\mathbb R^{n×n}$ is positive semidefinite and symmetric then $\sqrt{A_{ii}A_{jj}}\ge |A_{ij}|$, $\forall 1\le i,j\le n$

linear algebrapositive-semidefinite

Positive semidefinite means that for any $x$, we have $x^tAx\ge 0$. There exists decomposition like Cholesky's and other ways to know how a matrix is positive semidefinite like that their eigenvalues must all be greater or equal to zero. Don't know if any of this facts would help me prove the assumption.

I tried with a counter example, which I couldn't find, so I´m thinking this is true but don't know how to prove this. I'd appreciate any help.

Best Answer

Because $A$ is PSD, we have $v^\top A v \ge 0$ for any vector $v \in \mathbb{R}^n$. In particular, if $v$ has zeros in all entries except $v_i$ and $v_j$ in entries $i$ and $j$, then $0 \le v^\top A v = \begin{bmatrix}v_i & v_j \end{bmatrix} \begin{bmatrix}A_{ii} & A_{ij} \\ A_{ji} & A_{jj}\end{bmatrix} \begin{bmatrix}v_i \\ v_j \end{bmatrix}$ which shows that $\begin{bmatrix}A_{ii} & A_{ij} \\ A_{ji} & A_{jj}\end{bmatrix}$ is also PSD, and must have a nonnegative determinant. [This is Brian Borchers's hint.]