This problem is from Kunze Hoffman book. I think I go in the right direction to solve this but I miss some point to finish it. Can anyone help me?

Suppose $U$ is a self-adjoint unitary linear operator on a finite-dimensional inner product space $V$. Prove that we can find a subspace $W$ such that, if we express every $\alpha$ in $V\,$ in the form $\alpha = \beta + \gamma$, here $\beta$ in $W$ and $\gamma$ in $W^{\bot}$, then we have $U(\alpha) = \beta - \gamma$

If we denote $W = range(U+I)$, $W' = range(U - I)$,then must have $W \subset null(U-I)$ and $W \subset W^{'\bot}$. We can finish the proof if we can prove that $W' = null(U+I)$ or $W' = W^{\bot}$. But I can't prove this...



Hint: Geometrically this all wants to be the reflection through the subspace $W$. For any $\alpha$, define $$\beta:=\frac{\alpha+U\alpha}2\,, \quad \gamma:=\frac{\alpha-U\alpha}2$$ (Then the $\beta$'s will span $W={\rm ran}(U+I)$ and $\gamma$'s span all $W'={\rm ran}(U-I)$.) Prove that $\beta\perp\gamma$ using the hypothesis on $U$.

Ah, you also need that the whole $W$ is perpendicular to $W'$. But it's the same proof: $$\langle \alpha+U\alpha,\,\theta-U\theta\rangle=\langle \alpha,\theta\rangle+\langle U\alpha,\theta\rangle-\langle \alpha,U\theta\rangle- \langle U\alpha,U\theta\rangle = \dots$$

  • 1
    $\begingroup$ Thanks so much. It's unbelievably simple and clear. Great!!! $\endgroup$ – le duc quang Mar 11 '13 at 1:11

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.