I'm reading Heinonen's book on metric measure spaces. He writes that for general domains $\Omega \subset \mathbb{R}^n$, $M^{1,p}(\Omega) \subset W^{1,p}(\Omega)$ where the former are Hajlasz-Sobolev spaces (defined in 5.4) and the latter are Sobolev spaces.

Later in 5.17, he remarks on a way to see that $M^{1,p}(\Omega)$ and $W^{1,p}(\Omega)$ are not equivalent by constructing a domain for which the Poincare inequality fails. I don't understand this comment though: if the Poincare inequality fails for $u \in W^{1,p}(\Omega)$, it will also fail for $M^{1,p}(\Omega)$ because $M^{1,p}(\Omega) \subset W^{1,p}(\Omega)$.

What is Heinonen trying to say when he's using the Poincare inequality to show non-equality of $M^{1,p}(\Omega)$ and $W^{1,p}(\Omega)$? Thanks!!


1 Answer 1


Your argument is not correct. If a property $P$ fails for $Y$ and $X\subset Y$, it does not follow that it fails for $X$. For example $X=\{0\}\subset\mathbb{R}=Y$ but there are many properties true for $X$ and not true for $Y$.

You always have $M^{1,p}(\Omega)\subset W^{1,p}(\Omega)$ for all $1\leq p\leq\infty$. However, the inclusion is usually strict.

The spaces $W^{1,p}(\Omega)$ and $M^{1,p}(\Omega)$ are equal if for example $\Omega$ is a bounded extension domain. There is in fact the following characterization:

Theorem. Let $1<p<\infty$. Then a bounded domain $\Omega\subset\mathbb{R}^n$ is a $W^{1,p}$-extension domain if and only if $M^{1,p}(\Omega)=W^{1,p}(\Omega)$ and there is $C>0$ such that \begin{equation} |B(x,r)\cap\Omega|\geq Cr^n \quad \text{for all $x\in\Omega$ and $r\leq\operatorname{diam}(\Omega)$.} \end{equation}

Example. Take a disc with a radius removed. The above condition for the measure is satisfied, but the domain is not an extension domain so $M^{1,p}$ cannot be equal to $W^{1,p}$.

For a good source for basic properties, see

P. Hajłasz, Sobolev spaces on metric-measure spaces. (Heat kernels and analysis on manifolds, graphs, and metric spaces (Paris, 2002)), 173--218, Contemp. Math. , 338, Amer. Math. Soc., Providence, RI, 2003.

The above theorem is from

P. Hajłasz, P. Koskela, H. Tuominen, Sobolev embeddings, extensions and measure density condition J. Funct. Anal. 254 (2008), 1217--1234.


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.