Let $u\in \mathcal C^1_c(\mathbb R^n)$. I have that $$\int_{R_1}|u|^r\leq C\left(\int_{R_1}|\nabla u|^p\right)^{ar/p}\left(\int_{R_1}|u|^q\right)^{(1-a)r/q},$$ where $R_\rho=\{x\in \mathbb R^n\mid \rho\leq |x|<2\rho\}.$

Then it's written in my course that : Rescaling and multiplying by $\rho^{\gamma r}$ we get $$\int_{R_\rho}|x|^{\gamma r}|u|^r\leq C\left(\int_{R_\rho}|x|^{\alpha r}|\nabla u|^p\right)^{ar/p}\left(\int_{R_\rho}|x|^{\beta q}|u|^q\right)^{(1-a)r/q},$$ where $$\frac{1}{r}+\frac{\gamma }{n}=a\left(\frac{1}{p}-\frac{\alpha -1}{n}\right)+(1-a)\left(\frac{1}{q}+\frac{\beta }{n}\right),$$ $\gamma =a\varphi+(1-a)\beta $.


Could someone explain me how to get this ? I tried a substitution as $y=\rho^{\gamma r}x$, but no $|x|$ appear.

  • $\begingroup$ YOUR MISTAKE*: Don't Forget that $x\in R_1=\{x\in \mathbb R^n\mid 1\leq |x|<2 \} $ is set So putting $$y=\rho^{\gamma r}R_1$$ Does not make any sense as change of variables but rather you should put $$y=\rho^{\gamma r}x~~~\text{with}~~x\in R_1$$ $$R_\rho=\{x\in \mathbb R^n\mid \rho\leq |x|<2\rho\}.$$ $\endgroup$ – Guy Fsone Oct 15 '17 at 16:22
  • $\begingroup$ @GuyFsone: Thanks a lot for your answer. Unfortunately, after the substitution the $|x|^\gamma $ doesn't appear, no ? Because $dy=\rho^{\gamma r}dx.$ $\endgroup$ – user380364 Oct 15 '17 at 17:58
  • $\begingroup$ becarefull your jacobian should be raise to power n. check again $\endgroup$ – Guy Fsone Oct 15 '17 at 18:12
  • $\begingroup$ @GuyFsone: Yes indeed, thank :-) But I still no have $|x|$ that appear. $\endgroup$ – user380364 Oct 16 '17 at 8:52
  • $\begingroup$ you finally completly changed your post now the problem absolutely clear $\endgroup$ – Guy Fsone Oct 20 '17 at 18:25

Let $u : R_\rho \to \mathbb{R}$ and then rescale by defining $v : R_1 \to \mathbb{R}$ via $v(x) = u(\rho x)$. We apply your inequality to $v$ to get $$ \int_{R_1}|v|^r\leq C\left(\int_{R_1}|\nabla v|^p\right)^{ar/p}\left(\int_{R_1}|v|^q\right)^{(1-a)r/q}. $$ Now we change variables: $$ \int_{R_1} |v(x)|^r dx = \int_{R_1} |u(\rho x)|^r dx = \rho^{-n} \int_{R_\rho} |u(y)|^r dy, $$ $$ \int_{R_1} |v(x)|^q dx = \int_{R_1} |u(\rho x)|^q dx = \rho^{-n} \int_{R_\rho} |u(y)|^q dy, $$ and $$ \int_{R_1} |\nabla v(x)|^p dx = \int_{R_1} \rho^p |\nabla u(\rho x)|^p dx = \rho^{p-n} \int_{R_\rho} |\nabla u(y)|^p dy. $$

Now, for $y \in R_\rho$ we have that $\rho \le |y| \le 2 \rho$, and so $\rho^\delta \le |y|^\delta \le 2^\delta \rho^\delta$ for any $\delta >0$. Thus $$ \rho^{-n} \int_{R_\rho} |u(y)|^r dy \ge \frac{\rho^{-\gamma r-n}}{2^{\gamma r}} \int_{R_\rho} |y|^{\gamma r}|u(y)|^r dy, $$ $$ \rho^{-n} \int_{R_\rho} |u(y)|^q dy \le \rho^{-\beta q-n} \int_{R_\rho} |y|^{\beta q} |u(y)|^q dy, $$ and $$ \rho^{p-n} \int_{R_\rho} |\nabla u(y)|^p dy \le \rho^{p-n-\alpha p} \int_{R_\rho} |y|^{\alpha p}|\nabla u(y)|^p dy. $$

Combining all these shows that $$ \frac{\rho^{-\gamma r-n}}{2^{\gamma r}} \int_{R_\rho} |y|^{\gamma r}|u(y)|^r dy \le C \left( \rho^{p-n-\alpha p} \int_{R_\rho} |y|^{\alpha p}|\nabla u(y)|^p dy \right)^{ar/p} \left( \rho^{-\beta q-n} \int_{R_\rho} |y|^{\beta q} |u(y)|^q dy \right)^{(1-a)r/q}. $$

Now, if $\alpha,\beta,\gamma,p,q,r,a$ are chosen so that $$ (1-a)\left( \frac{\beta}{n} + \frac{1}{q} \right) +a\left(\frac{1}{p} + \frac{\alpha-1}{n} \right) = \frac{\gamma}{n} + \frac{1}{r} $$ then the $\rho$ terms exactly cancel out and we're left with $$ \int_{R_\rho} |y|^{\gamma r}|u(y)|^r dy \le 2^{\gamma r} C \left( \int_{R_\rho} |y|^{\alpha p}|\nabla u(y)|^p dy \right)^{ar/p} \left( \int_{R_\rho} |y|^{\beta q} |u(y)|^q dy \right)^{(1-a)r/q}. $$ I realize this isn't exactly what you've asked for, but I suspect there are two typos in your post. First, I suspect you want the power to be $\alpha p$ in the gradient integral, and second I suspect there's a sign error on the $\alpha-1$ term. At least, this is what I get when I adjust the power in the gradient term.

  • $\begingroup$ In fact it answer totally to my answer. Thank you very much :-) $\endgroup$ – user380364 Oct 21 '17 at 9:43

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.