2
$\begingroup$

Let $H$ be Hilbert's space.

Motivated by my previous question about wildly discontinuous linear functionals, which may be interpreted as an attempt to classify dense hyperplanes in $H$, let me now go straight to the point:

Questions.

  1. Are there any significant differences among dense hyperplanes in $H$?

  2. If $L$ and $M$ are two dense hyperplanes in $H$, is there a unitary operator mapping $L$ to $M$?

  3. Assuming the answer to (2) is negative, how many orbits are there for the natural action of the unitary group $\mathscr U(H)$ on the set of dense hyperplanes?


Speaking about general (not necessarily closed or dense) subspaces of $H$, there are a few things one may say in that respect.

For example, not all such spaces may be described as the range of a bounded operator and, in particular, no dense hyperplane qualifies. This is because, if the range of such an operator has finite co-dimension, it must be closed (this follows easily from the Closed Graph Theorem).

The range of a compact operator does not contain any infinite dimensional closed subspace, so that is another property one could use to classify subspaces.

More Questions.

  1. Is there a necessary and sufficient condition, expressed in topological/analytical terms, characterizing the range of a bounded (resp. compact) operator among all subspaces of $H$?

  2. How many unitary equivalence classes of non-closed subspaces of $H$ are there? How many of these may be described in topological/analytical terms?

$\endgroup$
2
  • 2
    $\begingroup$ As to 2 and 3, the linear dual of $H$ has more elements than the unitary group of $H$. $\endgroup$ Commented Jan 15, 2021 at 0:34
  • $\begingroup$ Thanks for a very interesting comment! So there are indeed lots of different unitarily equivalence classes of dense hyperplanes. I'd be very curious to see examples of two such classed being distinguished by analytic properties! $\endgroup$
    – Black
    Commented Jan 16, 2021 at 14:01

1 Answer 1

1
$\begingroup$

I guess I have a simple answer to Question 4, in the compact case: An infinite dimensional subspace $E\subseteq H$ is the range of a compact operator iff there exists an orthogonal (as opposed to orthonormal) set $\{e_n\}_{n\in {\mathbb N}}\subseteq E$, such that $$ \lim_{n\to \infty }\Vert e_n\Vert = \infty , $$ and $$ E=\Big\{\xi \in \overline{\text{span}\{e_n\}}: \sum_{n=1}^\infty \big|\langle \xi , e_n\rangle \Big|^2<\infty \Big\}. $$ This follows easily from the Spectral Theorem for compact operators, and the fact that the range of a compact operator $T$ coincides with the range of $|T|$.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .