0
$\begingroup$

Let us consider $A$ a translation invariant lower semi-bounded operator on $L^2(\mathbb{R}^n)$ with domain $D(A)$ and with empty discrete spectrum (I exclude bound states). I have the following intuition : since the operator is invariant by translation, the information should be "at infinity" and the bottom of its (essential) spectrum should not depend on any finite region. More specifically, I'm wondering if the following equality holds :

$\text{inf} \hspace{0.1cm} \sigma(A) = \underset{\psi \in D(A) \\ \| \psi \|=1}{\text{inf}} \langle \psi|A \psi \rangle = \underset{\psi \in D(A) \cap \{\psi = 0 \text{ on } B_R\} \\ \| \psi \|=1}{\text{inf}} \langle \psi|A \psi \rangle$ $\hspace{1cm}$ where $B_R$ is the ball of radius $R>0$.

However, I didn't find such a result in the books I looked through, so I was wondering if anyone knew wether this is true or not. I think it holds for example for the Laplace operator, for instance $\sigma(\frac{-d}{dx^2}) = [0, +\infty)$ on $L^2(\mathbb{R})$ but also on $L^2(\mathbb{R}\setminus [-R,R])$ if we add Neumann or Dirichlet boundary conditions.

$\endgroup$
4
  • $\begingroup$ Some observations: You can always replace $A$ by $A\wedge n$ for $n$ sufficiently large without changing the bottom of the spectrum, so one may assume that $A$ is bounded. Bounded translation-invariant operators are of the form $\mathcal F^{-1}M_f \mathcal F$, where $M_f$ is the operator of multiplication with the bounded function $f$. The spectrum of $A$ is the same as the spectrum of $M_f$, and the bottom of the spectrum of $M_f$ is the essential infimum of $f$. Furthermore, a function $\psi$ is an approximate eigenfunction of the bottom of the spectrum if [contd.] $\endgroup$
    – MaoWao
    Commented Jun 24 at 11:19
  • $\begingroup$ [contd.] it is "localized around the minimizers of $f$" (some care has to be taken if $f$ is not continuous). $\endgroup$
    – MaoWao
    Commented Jun 24 at 11:21
  • $\begingroup$ Thanks for your comment. I hear what you're saying but if $\psi$ is localized in Fourier space, it will be "spread" in configuration space and might be non-zero on the ball of radius $R$. How does one overcome that ? $\endgroup$
    – Hugo
    Commented Jun 24 at 13:41
  • $\begingroup$ I don't know, that's why it's not an answer. ;) Probably you'll also have to quantify how "localized" $\psi$ needs to be in Fourier space to not run in a contradiction with Paley-Wiener (so you don't want $\psi$ to be compactly supported). $\endgroup$
    – MaoWao
    Commented Jun 24 at 14:00

0

You must log in to answer this question.