12
$\begingroup$

It's easy to see that, for $1\le p,q< \infty$ the spaces $L^p(\Bbb R)$ and $L^q(\Bbb R)$ of $p$-th and $q$-th power integrable functions on the real line are homeomorphic as topological spaces. In fact, the map $f(x)\mapsto sgn(f(x))|f(x)|^{q/p}$ provides an explicit homeomorphism $L^p(\Bbb R) \to L^q(\Bbb R)$.

However this argument cannot be applied to the case where $p$ or $q$ equals infinity. Thus, I'm asking whether there is a homeomorphism from $L^\infty(\Bbb R)$ to any (and hence every) $L^p(\Bbb R)$ with $p<\infty$.

$\endgroup$
  • 25
    $\begingroup$ Isn't $L^2(\mathbb{R})$ separable and $L^\infty(\mathbb{R})$ nonseparable? $\endgroup$ – Paul McKenney Feb 9 '15 at 22:38
  • 2
    $\begingroup$ A more interesting question would be: are $L^2({\bf R})$ and $C_0({\bf R})$ homeomorphic? $\endgroup$ – Yemon Choi Feb 9 '15 at 23:07
  • 19
    $\begingroup$ Any two separable infinite dimensional Frechet spaces are homeomorphic. This is due to M. I. Kadec. $\endgroup$ – Bill Johnson Feb 9 '15 at 23:25
  • 1
    $\begingroup$ The $q/p$ should be $p/q$ if you want it to take $L^p$ to $L^q$. $\endgroup$ – Robert Israel Feb 10 '15 at 2:24
  • 2
    $\begingroup$ @PaulMcKenney I like the use of the rhetorical question... $\endgroup$ – Yemon Choi Feb 10 '15 at 12:18
18
$\begingroup$

Paul is right. $L^2(\mathbb{R})$ is separable. (The rational simple functions ought to be one example of something ctbl. and dense.)

However, $L^{\infty}(\mathbb{R})$ isn't separable. By Jones' Lemma, if $L^{\infty}(\mathbb{R})$ were separable then any closed discrete (i.e., nonclustering) set must be of size less than continuum. But the characteristic functions $\chi_{[0,x]}$, $x\in[0,\infty)$, are pairwise of distance 1 from eachother.

$\endgroup$
  • 4
    $\begingroup$ This was already pointed out in comments $\endgroup$ – Yemon Choi Feb 10 '15 at 9:54
  • 2
    $\begingroup$ And I was clearly elaborating on that comment. $\endgroup$ – Michael Cotton Feb 10 '15 at 12:15
  • 4
    $\begingroup$ We need answers, since comments are likely to go away in the future. $\endgroup$ – Gerald Edgar Feb 10 '15 at 13:07
  • 7
    $\begingroup$ We need answers, since questions without answers constantly float back to the first page. $\endgroup$ – Steven Gubkin Feb 10 '15 at 14:29
  • 4
    $\begingroup$ Jones' Lemma seems like overkill here. Since the continuum-many functions $\chi_{[0,x]}$ are pairwise distance 1 in $L^\infty$, the continuum-many open balls $B(\chi_{[0,x]}, 1/2)$ are pairwise disjoint. Any dense set must contain points in every one of these balls, so a dense set in $L^\infty$ must have size at least continuum. $\endgroup$ – Nate Eldredge Jan 4 '17 at 20:14

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.