Let $T$ be an unbounded self-adjoint operator.

Does there exist, for any $\varphi$ normalized in the Hilbert space, a constant $k(\varphi)>0$ and a sequence of normalized $(\varphi_n)$ such that $$ \lim_{n \rightarrow \infty} \Vert \varphi-\varphi_n \Vert=0 $$ and $\Vert T \varphi_n \Vert \le k(\varphi).$

Somehow this looks strange, if you think of $\varphi \notin D(T)$ as then $$\Vert T\varphi \Vert"="\infty$$

all of a sudden, on the other hand, maybe you only have to choose $k$ in a suitable way.

  • $\begingroup$ Sorry, why can't we have $k(\varphi)=\|T\varphi\|$, and $\varphi_n=(1-\frac{1}{2^n})\varphi$? $\endgroup$ – fierydemon May 25 '20 at 23:26
  • 1
    $\begingroup$ @fierydemon: $\varphi$ is not assumed to be in the domain of $T$. $\endgroup$ – Nate Eldredge May 26 '20 at 0:06

No, this only holds when $\varphi \in D(T)$.

It's enough to assume that $T$ is closed and densely defined, so that $T^{**} =T$. Let $\psi \in D(T^*)$ be arbitrary. If the hypothesis holds then we have $$ |\langle \varphi, T^* \psi \rangle| = \lim_{n \to \infty} |\langle \varphi_n, T^* \psi \rangle| = \lim_{n \to \infty} |\langle T \varphi_n, \psi \rangle| \le k(\varphi) \|\psi\|$$ which is exactly the definition of $\varphi \in D(T^{**}) = D(T)$. One can also see that in fact $T \varphi_n \to T \varphi$ weakly.

Note the argument still goes through even if we only assume that $\varphi_n \to \varphi$ weakly instead of strongly.


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.