Let $X$ be a Hilbert space, $T\in B(X)$ and $\lambda$ be a scalar such that $|\lambda|=\lVert T \rVert$. Prove that $Im(\lambda I - T)+Ker(\lambda I-T)$ is dense in $X$.

Since $Ker(\lambda I -T)=Im(\bar{\lambda}I-T^*)^\perp$ where $T^*$ is the adjoint of $T$, we need to show $Im(\lambda I-T)+Im(\bar{\lambda} I-T^*)^\perp$ is dense. My attempt is to assume it is not dense. Hence we can find something outside its closure. However, I am kind of stuck at this point.

I have proved that if $T^*x = \bar{\lambda}x$, then $Tx = \lambda x$. I don't know if this helps.


Suppose that $x$ is orthogonal to this space, and let's also normalize, $\|x\|=1$. Then, in particular, $x\in R(T-\lambda)^{\perp}=N(T^*-\overline{\lambda})$, so $$ \overline{\lambda} \langle x, Tx \rangle = \langle x, TT^* x\rangle = \|T^*x\|^2 = |\lambda|^2 $$ and thus $\langle x, Tx \rangle =\lambda$. On the other hand, $$ |\langle x, Tx \rangle | \le \|Tx\| \le |\lambda| \quad( = |\langle x, Tx \rangle |) . $$ Equality in the Cauchy-Schwarz inequality means that the vectors are linearly dependent, so $Tx = cx$, and only $c=\lambda$ works here.

So, in conclusion $x\in N(T-\lambda)$, but we also assumed that $x$ is orthogonal to this space, so $x=0$. (This whole argument assumes that $\lambda\not= 0$; of course, the claim is trivial if $T=0$.)


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.