0
$\begingroup$

Let $A \in \mathbb{R}^{n \times n}, B\in \mathbb{R}^{n \times n}$ are two positive semi-definite matrices and $A \leq B$ means $B - A$ is positive semi-definite. Also let $I \in \mathbb{R}^{m \times m}$ be the identity matrix. And $\otimes$ denotes the Kronecker product of matrix.

If $0 \leq A \leq B$, can I conclude that $(A \otimes I) \leq (B \otimes I)$?

$\endgroup$
0
$\begingroup$

Let $A \geq 0$. It is enough to show that $$ A \otimes 1_{\mathbb{R}^{m \times m}} \geq 0. $$ The following two approaches are based on https://math.stackexchange.com/q/1316606, where it has been shown that, for $B \in \mathbb{R}^{n \times n}$, we can identify $B \otimes 1_{\mathbb{R}^{m \times m}}$ with the $m \times m$ diagonal matrix with $B$ on the diagonal, which will be denoted by $\tilde{B}$.

  1. Let $v \in (\mathbb{R}^{n})^{m}$. Then we have $$ \langle \tilde{A} v, v\rangle = \langle \begin{pmatrix} Av_{1}\\ \vdots \\ Av_{m}\end{pmatrix}, v\rangle = \sum_{i=1}^{m} \langle A v_{i}, v_{i}\rangle \geq 0, $$ since $A \geq 0$. Hence, $\tilde{A} \geq 0$ and therefore also $A \otimes 1_{\mathbb{R}^{m \times m}} \geq 0$.

  2. We also can use the fact that $$ \det(B \otimes 1_{\mathbb{R}^{m \times m}}) = \det(\tilde{B}) = \det(B)^{m} $$ for all $B \in \mathbb{R}^{n \times n}$. With $$ \det(B \otimes 1_{\mathbb{R}^{m \times m}} - \lambda (1_{\mathbb{R}^{n \times n}} \otimes 1_{\mathbb{R}^{m \times m}})) = \det(\tilde{B} - \lambda 1_{(\mathbb{R}^{n \times n})^{m \times m}}) = \det(B-\lambda 1_{\mathbb{R}^{n \times n}})^{m} $$ for all $B \in \mathbb{R}^{n \times n}$ and $\lambda \in \mathbb{C}$ we see that $A \otimes 1_{\mathbb{R}^{m \times m}}$ and $A$ have the same eigenvalues. Hence, $A \otimes 1_{\mathbb{R}^{m \times m}}$ is positive. (Here we used the fact that a matrix is positive semidefinit if and only if all eigenvalues are non-negative.)

$\endgroup$

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.