4

The question is to show that if $T$ is normal, there exists a unitary operator $U$ such that $T^{*}=UT$. My guess is that we use the polar decomposition of $T$- into a product of a unitary and positive operator- in some way, but am not sure how to start.

Daniel Fischer
  • 206,697
Arundhathi
  • 1,453

1 Answers1

2

I think I figured it out after all.

Since $T$ is normal, there exists a unitary operator $V$ and a positive operator $P$ such that $T=VP$, and $P$, $V$ and $T$ commute with each other. $T=VP \Longrightarrow T^{*}=PV^{*}$.

$\Longrightarrow T^{*}V^{2}=PV^{*}VV=PV=T$ $\Longrightarrow T^{*}=T(V^{*})^{2}=(V^{*})^{2}T$

The last equality follows since $VT=TV$ and $T$ normal, by Fuglede-Putnam-Rosenblum, $VT^{*}=T^{*}V \Longrightarrow V^{*}T=TV^{*}$

Let $U=(V^{*})^{2}$

Then $U$ is unitary and satisfies $T^{*}=UT$

Arundhathi
  • 1,453
  • When is $U$ unique? Is it when $T$ is invertible? – Libertron Dec 08 '13 at 02:01
  • 2
    In the polar decomposition of $T$ as $T=UP$, with $U$ a unitary operator and $P$ a positive operator, $P$ is always uniquely determined as the positive square root of $T^{*}T$. So if $T$ is invertible clearly $U$ must be uniquely given by $U=TP^{-1}$. – Arundhathi Dec 09 '13 at 04:55