1

Let $V$ be a finite dimensional inner product space and $T:V\to V.$

  • Prove Ker$T=($Im($T^*))^\bot$ and $($Ker$T^*)^\bot$ = Im$T$.
  • Deduce dimIm$T$ = dimIm$T^*$.

I have written a proof but I'm very unsure about some of the steps and would love another set of eyes:

Let's prove this by using two-sided containment.

First notice, if $KerT =\{0\}$ we have $ImT=ImT^* = V$ and therefore since $V = ImT^* \oplus (ImT^*)^\bot$, we have $(ImT^*)^\bot=\{0\}.$

Thus meaning we can assume we have $0 \neq v \in KerT$.

Using inner product we can receive 0 = $<T(v),v>$ = $<v,T^*(v)>$ thus implementing since $v \neq 0$ that $T^*(v) = 0$ and so $v \in KerT^*$.

We have $V = KerT^* \oplus (KerT^*)^\bot$ thus this means $v \in KerT^* \to v \notin (KerT^*)^\bot$.

This means we have $v \in (ImT^*)^\bot$ and therfore we have $KerT \subseteq (ImT^*)^\bot$.

For the other side, consider $0 \neq v \in (ImT^*)^\bot$, (which exists from the same reasons as the previous containment).

Since $V = ImT^* \oplus (ImT^*)^\bot$ we have $v \notin ImT^*$ and since $v \neq 0$ we have $v \in KerT^*$.

This implements we have $(ImT^*)^\bot \subseteq KerT$ and so we have $KerT = (ImT^*)^\bot$.

Then we have $ImT = (ImT^*)^* = (((ImT^*)^\bot)^\bot)^* = ((KerT)^\bot)^* = (KerT^*)^\bot$.

By dimension theoreom we have: $dimV = dimImT + dimKerT = dimImT^* + dimKerT^*$

Since $<T(v),v>$ = $<v,T^*(v)>$ = 0 $\to$ $T(v) = T^*(v) = 0$,
$\forall v \in KerT$ or $v \in KerT^*$ that satsifies $v \neq 0$

then we have $kerT = KerT^*$ thus implamenting $dimKerT = dimKerT^*$.

So we have $dimImT = dimImT^*$, as needed.

  • $<v,T^(v)>=0$ and $v \neq 0$ does not imply $T^(v) = 0$ – Anne Bauval May 26 '23 at 12:17
  • @AnneBauval It was a typo. Regarding your second comment, if the inner product is 0 then we must have $T^(v) = 0$ or $T^(v) \bot v$ at least that is what i was thinking – MathStudent101 May 26 '23 at 12:21
  • "and therefore since
  • $V = ImT^* \oplus (ImT^)^\bot$, we have $(ImT^)^\bot={0}$" can be replaced with "and therefore $(ImT^)^\bot=V^\perp={0}$". 2) $<v,T^(v)>=0\iff T^(v)\perp v$ by definition of $\perp$ but your claim that $T^(v)=0$ is wrong.

    – Anne Bauval May 26 '23 at 12:22
  • @AnneBauval yes i understand. Would it be alright to claim: $T^(v)$ $\bot$ $v$ implements $v \notin (KerT^)^\bot$ ? because that would allow the rest of the proof. – MathStudent101 May 26 '23 at 12:25
  • That is false too. – Anne Bauval May 26 '23 at 12:27
  • @AnneBauval then how can i deduce the correct answer? – MathStudent101 May 26 '23 at 12:28
  • " that would allow the rest of the proof": I don't think so. "$v \notin (KerT^)^\bot$. This means we have $v \in (ImT^)^\bot$" is wrong as well. I did not read the rest. – Anne Bauval May 26 '23 at 12:32
  • @AnneBauval thank you very much. Maybe if you could sketch an idea it would be of great help. – MathStudent101 May 26 '23 at 12:34
  • @AnneBauval yes seems like it. Is it possible to assume: $ImT = (KerT^)^\bot \iff KerT = (ImT^)^\bot$ ? – MathStudent101 May 26 '23 at 12:40
  • You cannot not assume they are equivalent. You must prove each of them. You can use one of them once proved (for every $T$) to deduce the other one. – Anne Bauval May 26 '23 at 12:44
  • @AnneBauval Ok thank you very much. – MathStudent101 May 26 '23 at 12:45