0
$\begingroup$

Consider the weighted Sobolev-type space $$ W_\alpha:=\{f\in L^2(0,\infty):\hbox{id}^\alpha\cdot f'\in L^2(0,\infty)\}. $$ Are there any known embeddings? Ideally, I am looking for an embedding of the above space into $L^p(0,\infty)$ for some $p>2$, some $\alpha_0$, and all $\alpha \in [0,\alpha_0]$. The only result I am aware of is that $W_2$ is not embedded in $L^\infty(0,\infty)$, since $\min\{0,\log\}\in W_2\setminus L^\infty(0,\infty)$.

$\endgroup$
3
  • 1
    $\begingroup$ If $\alpha<1/2$, then $f'\in L^1(0,1)$, so there are no problems near $x=0$ (and weaker assumptions would suffice of course to get $f\in L^p$), and for large $x$ we never have any problems since $f,f'\in L^2$ already imply that $f\in L^p$ for all $p\ge 2$. $\endgroup$ Commented Oct 4, 2017 at 18:44
  • $\begingroup$ @ChristianRemling This is true: one certainly has $f\in W^{1,1}_{loc}$ and, as you observe, even a bit more for certain ranges of $\alpha$. But how does this help to say anything, near 0, about $f$ (rather than $f'$)? Can you please elaborate on your "of course"? Are you thinking of some interpolation inequality between $L^2(0,\infty)$ and $\{f\in L^2(0,\infty):f'\in L^1(0,\infty)$? $\endgroup$ Commented Oct 5, 2017 at 19:18
  • 1
    $\begingroup$ These $f$ are locally absolutely continuous, so $f(x)=f(1)-\int_x^1 f'(t)\, dt$, and if now $f'\in L^1(0,1)$ (as it will be when $\alpha<1/2$, by CS), then it follows that $f$ is bounded (in fact, continuous on $[0,\infty)$). $\endgroup$ Commented Oct 5, 2017 at 20:28

0

You must log in to answer this question.

Browse other questions tagged .