2
$\begingroup$

The spectral theorem says that if $A$ is a self-adjoint operator on a Hilbert space $H$ with compact inverse, then the eigenvectors of $A$ form a complete orthonormal basis of $H$. Furthermore each eigenvalue is real and $|\lambda_n|\to\infty$ as $n\to\infty$.

Now let us slightly change the conditions of the theorem. Let $A$ be a self-adjoint operator with increasing sequence of eigenvalues $0<\lambda_1<\lambda_2<...$ Is it true that the eigenvectors of $A$ form a complete orthonormal basis of $H$?

$\endgroup$
  • $\begingroup$ Do you mean "eigenvalue" in the strict sense of $Ax = \lambda x$, or do you mean it in a more general sense (e.g. $A - \lambda I$ fails to be invertible)? $\endgroup$ – Omnomnomnom May 28 '18 at 14:49
  • $\begingroup$ @Omnomnomnom In the strict sense: $Ax=\lambda x$. $\endgroup$ – Oleg May 28 '18 at 14:50
  • $\begingroup$ I suspect that there exists a modification of the $\phi(t) \mapsto t\phi(t)$ map described here which has an increasing sequence of eigenvalues whose eigenvectors fail to span the space. $\endgroup$ – Omnomnomnom May 28 '18 at 14:59
4
$\begingroup$

Let $H=\ell^2(\mathbb N)\oplus L^2[0,1]$. Write $\{e_n\}$ for the canonical basis of $\ell^2(\mathbb N)$. Now define $A$ by $$ Ae_n=\lambda_ne_n, $$ and on the second component let $(Af)(t)=tf(t). $ Now the eigenvectors of $A$ are $\{e_n\}$, which do not span $H$; in fact, their orthogonal complement is infinite-dimensional. You can tweak the second component to make the spectrum of $A$ almost anything you want.

$\endgroup$
  • $\begingroup$ Martin, thank you very much for a very nice example! It means that this compact inverse condition in spectral theorem cannot be replaced by my suggested condition. $\endgroup$ – Oleg May 28 '18 at 15:08
  • $\begingroup$ Indeed. Basically, you cannot learn much structure from just the eigenvalues. $\endgroup$ – Martin Argerami Jun 4 '18 at 19:01

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.