4
$\begingroup$

I am given two $n\times n$ complex matrices $A$ and $B$. $A> 0$ ($A$ is positive definite) and $B$ is Hermitian. By the properties of positive definite matrices, it follows that: $A=T^*T$ for some invertible matrix $T$. My question is: why is $(T^{-1})^{*}B(T^{-1})\geq 0 $? (i.e why is $(T^{-1})^{*}B(T^{-1})$ positive semidefinite?).

I tried the following: for any $x\in \mathbb{C}^{n}$: $$x^*(T^{-1})^{*}B(T^{-1})x=\left \langle B(T^{-1})x,(T^{-1})x \right \rangle=\left \langle By,y \right \rangle$$ where $y=(T^{-1})x$. I need to prove that $\left \langle By,y \right \rangle\geq 0$, but I can't see how since $B$ is Hermitian and not positive semidefinite. Any help? Thanks.

$\endgroup$
2
$\begingroup$

It is not true. Take for example $A=I_n$, $B=-I_n$. Then you can also take $T=I$, and your assertion claims that $-I_n\geq0$, which is of course false.

$\endgroup$
0
$\begingroup$

Since $T^{-1}$ is invertible, $(T^{-1})^*BT$ is positive semi-definite if and only if so is $B$. Indeed, $x^*(T^{-1})^*BTx=(T^{-1}x)^*BT^{-1}x$, and if $B$ is positive semi-definite, $x^*(T^{-1})^*BTx\geq 0$ for all $x$. Conversely, if $(T^{-1})^*BT$ is positive semi-definite then for a $x$, we can find $x'$ such that $T^{-1}x'=x$. Then $(x')^*(T^{-1})^*BTx'=x^*Bx\geq 0$.

$\endgroup$

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.