4
$\begingroup$

Let $(a_n)$ be a sequence that tends to zero and $0<a_n<1$. Why is then the mapping $$ \ell_2 \rightarrow \ell_2,\ (x_1,x_2,\ldots)\mapsto ((1-a_1)x_1,(1-a_2)x_2,\ldots)$$ not compact ?

Someone said I should look at its inverse, but that didn't helpt me either...

$\endgroup$
6
$\begingroup$

Prove it directly. The bounded sequence $$e_n=(\underbrace{0 \ldots 1}_{n\text{th place}} 0 \ldots )$$

is mapped to a sequence with no Cauchy subsequences.

$\endgroup$
6
$\begingroup$

For any non-negative $n$, let $e (n)$ be the sequence such that $e (n)_i = 0$ if $i \neq n$, and $e (n)_n = 1$. These sequences are eigenvectors for the operator $T$, as:

$$T e(n) = (1-a_n) e(n).$$

This implies that $1-a_n$ belongs to the spectrum $\sigma(T)$ of $T$ for all $n$. Hence, $\sigma (T)$ has an accumulation point at $1$, which can't happen if $T$ is compact.

Edit: actually, you can show the following stronger properties:

  • $I-T$ is compact (which also implies that $T$ is non-compact);
  • $\sigma (T) = \{1-a_n: \ n \in \mathbb{N}^*\} \cup \{1\}$.
$\endgroup$
  • $\begingroup$ How does $I - T$ being compact imply that $T$ is non-compact? $\endgroup$ – csss Nov 28 '16 at 21:17
  • $\begingroup$ @csss: the space is infinite-dimensional. Then $I-T$ being compact imply that $1$ is a not a pole of finite multiplicity for the resolvent of $T$, whence $T$ is not compact. $\endgroup$ – D. Thomine Dec 9 '16 at 16:14
4
$\begingroup$

The idea which consists of looking at the inverse works, because an invertible linear bounded operator between two infinite dimensional Banach spaces is never compact.

  • We check that the inverse operator $T^{-1}$ is defined as $$T^{-1}((x_n)_{n\geq 1})=\left(\frac 1{1-a_n}x_n\right)_{n\geq 1},$$ which is bounded from $\ell^2$ to $\ell^2$.
  • If $S_1$ is a compact operator and $S_2$ is bounded then $S_1S_2$ and $S_2S_1$ are compact.
  • If $T$ were compact, then the identity operator would be compact. But it's not possible because the unit ball is not strongly compact.
$\endgroup$
1
$\begingroup$

I think looking at bounded sequence $$e_n=(0, \ldots ,\underbrace{\frac{1}{1-a_n}}_{n\text{th place}} ,0, \ldots )$$ is more direct.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy