6
$\begingroup$

Let $A,B \in \mathcal{M}_n (\mathbb{R})$ two matrices so that:

a) $AB^2=B^2 A$ and $BA^2=A^2 B$

b) $\text{rank}(AB-BA)=1$.

Prove: $$(\det(A-B)+\det(A+B) )^2≥4\det(A^2-B^2 )$$


This is a solution:

Denote $C=AB-BA$. Then $rank(C)=1$ so $C=p\cdot q^T$ where $p,q$ are column vectors and $tr(C)=0$. This proves that $C^2=p\cdot q^T\cdot p\cdot q^T=tr(C)C=0.$

If $D=A^2-B^2$ then a) implies that $CD=DC$ and $D$ commutes with $A,B$.

As a consequence $tr(CD^{-1})=0$.

If $\det(D)=0$ we have nothing to prove. Else $D$ is invertible and $(CD^{-1})^2=0$.

We have $(A-B)(A+B)=A^2-B^2+AB-BA=D+C$. We would like to prove that $\det(A-B)(B-A)=\det(A^2-B^2)$.

For this define $ f(t)=\det(AB-BA+t(A^2-B^2))$

and see that $f(t)=\det(D)\det(CD^{-1}+tI)=\det(A^2-B^2)\cdot t^n$.

Replace $t=1$ in the previous relation to get $\det(AB-BA+A^2-B^2)=\det(A^2-B^2)$ and we are done.

But i don't understand why $tr(CD^{-1})=0$ ??

$\endgroup$
1
$\begingroup$

Because $(CD^{-1})^2=0$, then $CD^{-1}$ only has zero eigenvalue. There should be a mistake in the order of proof.

$\endgroup$
  • 1
    $\begingroup$ The "As a consequence" part should come after the discussion of whether $D$ is invertible. $\endgroup$ – ggttttll Apr 10 '15 at 6:23
1
$\begingroup$

Note $D^{-1}$ also commutes with $C$, so

$CD^{-1}CD^{-1}=C^2D^{-1}D^{-1}=0$

Let $\lambda_i$ be any eigenvalue of $CD^{-1}$. Then $\lambda_i^2$ is eigenvalue of $(CD^{-1})^2$. So $\lambda_i^2=0$ and $\lambda_i=0$.

Thus $Tr(CD^{-1})=\sum \limits_{k=1}^n\lambda_k=0$.

$\endgroup$

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.