1
$\begingroup$

Question is in the title. Let me elaborate: Suppose $u\in C_c^1(\mathbb R^n)$. Then we have the Sobolev inequality $$ \Vert u\Vert_{L^{2^*}(\mathbb R^n)}\leq C\Vert\nabla u\Vert_{L^2(\mathbb R^n)} $$ where $2^*=\frac{2n}{n-2}$ is the Sobolev conjugate of $2$, and $C$ depends only on $n$ in this formulation. It seems that this result can be used to prove the following: If $u\in W^{1,2}(\mathbb R^n)$, then $$ \Vert u\Vert_{L^{2^*}(\mathbb R^n)}\leq C\Vert\nabla u\Vert_{L^2(\mathbb R^n)}. $$ To see that the above follows from the Sobolev inequality, simply approximate $u$ in the $W^{1,2}-$norm and $L^{2^*}-$norm simultaneously by smooth compactly supported functions $u_m$, and use the Sobolev inequality.

In particular, this proves that if $u\in W_0^{1,2}(\Omega)$ for $\Omega\subset\mathbb R^n$ open, then we can write the same result as above with $\mathbb R^n$ replaced by $\Omega$, with $C$ independent of $\Omega$. This seems to improve, for instance Theorem 3 in section 5.6 in Evans book, for the specific Sobolev exponent $2^*$ (and analogously, $p^*$).

My question basically is a sanity check: is my sketch of the proof correct?

$\endgroup$
1
$\begingroup$

Yes, your proof is correct.

Indeed, you get $$\|u\|_{2^*} \le C_{n} \, \|\nabla u\|_2,$$ with $C_n$ depending only on the dimension $n$.

However, Theorem 3 also gives a bound of $u$ in $L^q(\Omega)$, with $2 \le q \le 2^*$. For this, you need Hölders inequality $$\|u\|_q \le \|u\|_{2^*} \, \|1\|_r \le \mu(\Omega)^{1/r} \, \|u\|_{2^*},$$ with $1/q = 1/2^* + 1/r$ and the volume $\mu(\Omega)$ is popping up.

To conclude, in $$\|u\|_{2^*} \le C_{n} \, \|\nabla u\|_2$$ the constant depends only on the dimension, but in $$\|u\|_{q} \le C_{\Omega,q,n} \, \|\nabla u\|_2$$ with $2 \le q \le 2^*$ you get an additional dependence on $\mu(\Omega)$ and $q$.

$\endgroup$

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.