1

Good evening, the question is the following: what is the "natural" subspace of $L^2(\mathbb{R}^d)$ that provides the domain of the operator $H=-\Delta+|x|^2$? I mean, $H$ is an unbounded operator on $L^2$ and it is surely defined on the dense subset of $L^2$ given by the Schwartz functions, $\mathcal{S}$. However, it may be difficult to prove that $H$ is self-adjoint on $\mathcal{S}$ (as in the definition of self-adjointness for unbounded operators). I guessed that the domain of $H$ should sound like this: $D(H)$ is the set of those functions $f\in L^2(\mathbb{R}^d)$ s.t.

  1. the partial derivatives of $f$ exist a.e. and they are $L^2$ functions;
  2. the same for the second derivatives;
  3. some further conditions on the decay of $f$, $\nabla f$ and $\Delta f$ in order to integrate by parts and "cancel the boundary term".

This would prove easily at least that $H$ is symmetric. The self-adjointness may still be problematic.

Thank you in advance.

1 Answers1

1

The Hermite operator is essentially self-adjoint on $D(H)=C_c^\infty(\mathbb{R}^d)$: If $\psi\in D(H^\ast)$, then $$ \langle \phi, H^\ast \psi\rangle=\langle H\phi,\psi\rangle=-\int (\Delta \phi)\psi+\int|x|^2\phi\psi $$ for $\phi\in C_c^\infty(\mathbb{R}^d)$.

Thus $\Delta \psi\in L^2_{\mathrm{loc}}$ (distributional Laplacian) and $H^\ast\psi=-\Delta \psi+|x|^2\psi$. By elliptic regularity, $\psi\in W^{2,2}_{\mathrm{loc}}$. It is not hard to see that in fact $D(H^\ast)=\{\psi\in L^2\cap W^{2,2}_{\mathrm{loc}}\mid -\Delta \psi+|x|^2\psi\in L^2\}$.

For essential self-adjointness it is sufficient to prove that $\ker(H^\ast+1)=\{0\}$. For that purpose let $\psi\in \ker(H^\ast+1)$ and $\phi\in C_c^\infty(\mathbb{R}^d)$ with $0\leq \phi\leq 1$, $\phi=1$ on $B_1(0)$ and $\operatorname{supp}\phi\subset B_2(0)$. Set $\phi_R(x)=\phi(x/R)$.

As $\psi\in W^{2,2}_{\mathrm{loc}}$ we have $\psi\phi_R\in W^{2,2}$. In particular, $$ \langle H^\ast(\psi\phi_R),\psi\phi_R\rangle=\int|\nabla(\phi_R\psi)|^2+|x|^2(\phi_R\psi)^2\geq 0. $$ A little computation with the Leibniz rule shows $$ \langle (H^\ast+1)(\psi\phi_R),\psi\phi_R\rangle=\mathrm{Re}\langle \phi_R(H^\ast+1)\psi,\phi_R\psi\rangle+\int |\nabla \phi_R|^2\psi^2=\int|\nabla \phi_R|^2\psi^2. $$ Thus $$ \int_{B_R(0)}\psi^2\leq \langle (H^\ast+1)(\phi_R \psi),\phi_R\psi\rangle=\int|\nabla \phi_R|^2\psi^2\leq \|\nabla \phi_R\|_\infty^2\int \psi^2. $$ Letting $R\to \infty$, we obtain $\psi=0$.

Note that this argument works not only for $-\Delta+|x|^2$, but more generally for $-\Delta+V$ with lower bounded $V\in L^2_{\mathrm{loc}}$.

MaoWao
  • 16,366
  • 2
  • 21
  • 43
  • Thank you! Is there any difference if one defines $H$ on $\mathcal{S}(\mathbb{R}^d)$ instead of $\mathcal{C}^\infty_0(\mathbb{R}^d)$? – UnusualMathem Dec 27 '20 at 15:21
  • No, $H$ is also essentially self-adjoint on $\mathcal{S}(\mathbb{R}^d)$ (because it is contained in the domain of the closure of $H|_{C_c^\infty}$, which is $D(H^\ast)$ described in my answer). – MaoWao Dec 27 '20 at 16:27
  • Fine, thank you :D – UnusualMathem Dec 27 '20 at 18:00
  • @UnusualMathem It appears that you have not approved any of the answers to your questions so far. Do you know that you can click on the tick mark to the left of the answer to approve it? If you don't do this all your questions will be left as unanswered questions. – Kavi Rama Murthy Apr 06 '21 at 10:02