3
$\begingroup$

Let $A, B$ be closable, unbounded, linear operators on a Hilbert space $H$. Suppose further that the operator $A + B$, defined on the intersection of domains $D(A) \cap D(B)$, is also closable.

I would like to show that $\overline{A + B} \supseteq \overline{A} + \overline{B}$.

This is problem 2.8 from Teschl's book Mathematical Methods in Quantum Mechanics.

I have gotten stuck, but I have been trying to rely on the fact that, if $A$ is closable, then the graph $\Gamma(\overline{A})$ is equal to $\overline{\Gamma(A)}$. So if we take $\psi \in D(\overline{A}) \cap D(\overline{B})$. We can get sequences $\{\psi_n\} \subseteq D(A)$, $\{\psi'_n\} \subseteq D(B)$ so that $(\psi_n, A\psi_n) \to (\psi, \overline{A} \psi)$ and $(\psi'_n, B\psi'_n) \to (\psi, \overline{B} \psi)$, where the convergence is in the standard norm on $H^2$, $\|(\psi, \varphi) \|_{H^2} := \|\psi\| + \|\varphi\|$. The problem is that I'd really like to just have one single sequence in $D(A) \cap D(B)$ converging to $\psi$, but I haven't been able to show that this happens.

Hints or solutions are greatly appreciated!

$\endgroup$
  • 1
    $\begingroup$ Do you intend $A$ and $B$ to be linear operators? This was the first difficulty with the obvious solution... $\endgroup$ – Eric Towers May 10 '16 at 20:46
  • $\begingroup$ @EricTowers---Yes, A and B are linear maps. I will edit the question to emphasize this. $\endgroup$ – JZS May 10 '16 at 20:48
2
$\begingroup$

I think that exercise is false.

Let $A=\frac{d^2}{dx^2}$ on all finite linear combinations of $\{ 1,x,x^2,x^3,\cdots \}$ and let $B=\frac{d^2}{dx^2}$ on all finite linear combinations of $\{ \sin(\pi x),\sin(2\pi x),\sin(3\pi x),\cdots\}$. Both of these operators are densely-defined; and they're both closable because they're both restrictions of the same closed operator $C=\frac{d^2}{dx^2}$ on $\mathcal{D}(C)=H^2[0,1]$.

  • $\mathcal{D}(A+B)=\{0\}$, which means that $A+B$ is trivially closed on its domain $\{0\}$. This is because the linear space of polynomials has nothing in common with finite linear combinations of $\sin(n\pi x)$ terms, other than the $0$ function.

  • The closure of $B$ has domain $$ \mathcal{D}(\overline{B})=\{ f \in H^2[0,1] : f(0)=f(1)=0 \}. $$ This is because $\{ \sin(n\pi x) \}_{n=1}^{\infty}$ is an orthonormal basis of $\{ f \in H^2[0,1] : f(0)=f(1)=0 \}$.

  • The domain of $A$ contains polynomials that vanish at $0$ and $1$, which means $\mathcal{D}(A)\cap\mathcal{D}(\overline{B}) \ne \{0\}$. So $\mathcal{D}(\overline{A})\cap\mathcal{D}(\overline{B}) \ne \{0\}$, which proves that $$ \overline{A}+\overline{B} \not \subseteq \overline{A+B}. $$

$\endgroup$

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.