4
$\begingroup$

if I have a densely defined closed linear operator $A$ and $A^* = -A$(same domain also closed). Is this sufficient that $AA^*$ and $A^*A$ are proper self-adjoint operators, assuming that we can also define both of them on a dense domain of our underlying Hilbert space. If this is true( I guess so), could anybody state the actual theorem that shows this or better: explain why the fact that the operators are closed and densely defined are important here?

$\endgroup$
  • $\begingroup$ Doesn't $\langle ABx,y\rangle=\langle Bx,A^*y\rangle=\langle x,B^*A^*y\rangle \implies (AB)^*=B^*A^*$? $\endgroup$ – blue Aug 2 '14 at 9:53
  • 1
    $\begingroup$ well, I thought that an operator $T$ is self-adjoint if the operator is closed, symmetric, densely-defined and $T-i$,$T+i$ are onto. I think your argument only shows symmetry. $\endgroup$ – user167575 Aug 2 '14 at 10:41
  • $\begingroup$ Since $A^*=-A$, $H=iA$ is self-adjoint and so is $H^2$. Now $AA^*$ and ${A^*}A$ are defined and coincide on the domain of $H^2$. $\endgroup$ – Urgje Aug 2 '14 at 12:10
  • $\begingroup$ @Urgje okay, you did this by using what $A^*$ actually is, in this example, thank you. Is the general case also true, if yes, could I encourage you to post this as an answer? I think this will be also more valuable to other people seeing this thread who may be more interested in the general case :-) $\endgroup$ – user167575 Aug 2 '14 at 12:12
4
$\begingroup$

Lemma: Let $B : \mathcal{D}(B)\subseteq X\rightarrow X$ be a symmetric linear operator on a Hilbert space $X$ for which $(Bx,x) \ge 0$. If $I+B$ is surjective, then $B$ is closed, densely-defined and selfadjoint.

Proof: Let $B$ as stated. Suppose that $B+I$ is surjective. To show that $B$ must be densely-defined, suppose that $y \perp \mathcal{D}(B)$ and write $y=(I+B)x$ for some $x\in \mathcal{D})$. Then $y \perp x$ gives $$ 0 = (y,x)=((I+B)x,x) \ge \|x\|^{2} \implies x = 0 \implies y = (I+B)x=0. $$ So the domain of $B$ must be dense. $B$ is closed because $(I+B)^{-1}$ is bounded and defined on all of $X$, owing to the fact that $$ \|x\|^{2} \le ((I+B)x,x) \le \|(I+B)x\|\|x\|,\;\;\; x \in \mathcal{D}(B). $$ Therefore $B^{\star}$ is also closed and densely-defined. Because $B$ is symmetric, then $B\preceq B^{\star}$ (meaning that $B^{\star}$ extends $B$.) To see that $B=B^{\star}$, suppose that $y \in \mathcal{D}(B^{\star})$. The proof is finished by showing that $y \in \mathcal{D}(B)$: to this end, note that there exists $z \in \mathcal{D}(B)$ such that $(I+B^{\star})y=(I+B)z$. So, $$ ((I+B)x,y)=(x,(I+B^{\star})y)=(x,(I+B)z)=((I+B)x,z),\;\;\; x \in \mathcal{D}(B). $$ Because $(I+B)$ is surjective, then $z=y \in \mathcal{D}(B)$. So $B=B^{\star}$. $\Box$

Theorem [von Neumann]: Let $X$ be a Hilbert space, and let $A : \mathcal{D}(A)\subseteq X\rightarrow X$ be a closed densely-defined linear operator. Define $A^{\star}A$ on the domain $$\mathcal{D}(A^{\star}A)=\{ x \in \mathcal{D}(A) : Ax \in \mathcal{D}(A^{\star})\}.$$ Then $A^{\star}A$ is a closed densely-defined selfadjoint linear operator. Because $A^{\star}$ is also closed and densely-defined, then $AA^{\star}$ is also a closed densely-defined selfadjoint linear operator.

Proof: Let $J : X\times X\rightarrow X\times X$ be defined as $J\langle x,y\rangle=\langle-y,x\rangle$. Let $\mathcal{G}(A)=\{ \langle x,Ax\rangle : x\in\mathcal{D}(A)\}\subset X\times X$ be the graph of $A$. Then it is well-known that $\mathcal{G}(A^{\star})=J(\mathcal{G}(A)^{\perp})=(J\mathcal{G}(A))^{\perp}$, where the orthogonal complement is taken in $X\times X$. Because $X\times X$ is a Hilbert space, then $X\times X = \mathcal{M}\oplus\mathcal{M}^{\perp}$ for any closed linear subspace $\mathcal{M}$ of $X\times X$. Therefore, $$ X\times X = \mathcal{G}(A)\oplus J\mathcal{G}(A^{\star}). $$ Hence, for every $a \in X$, there are unique $x \in \mathcal{D}(A)$, $y\in\mathcal{D}(A^{\star})$ such that $$ \langle a,0\rangle = \langle x,Ax\rangle + \langle -A^{\star}y,y\rangle, \\ a = x - A^{\star}y,\;\; y=-Ax. $$ So $x \in \mathcal{D}(A^{\star}A)$ and $(I+A^{\star}A)x=a$. That is, $I+A^{\star}A$ is surjective on its natural domain. Now the previous lemma applies to $B=A^{\star}A$. $\;\;\Box$

$\endgroup$
  • $\begingroup$ When and where did Von Neumann publish this theorem? How does one cite it? $\endgroup$ – Thibaut Demaerel Jul 13 '18 at 23:15

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.