8
$\begingroup$

Why is the dual of the Sobolev space $H_0^1(\Omega)$ denoted $H^{-1}(\Omega)$ ?

For a positive integer $k$, $H^k(\Omega)=W^{k,2}(\Omega)$. What is the motivation behind the $-1$ exponent?

$\endgroup$
  • 3
    $\begingroup$ I think the answer comes from Sobolev of real exponent, defined with Fourier transform. $\endgroup$ – Davide Giraudo Apr 20 '12 at 19:38
5
$\begingroup$

Definition Let $s\in\mathbb R$. $H^s(\mathbb R^n)$ is the vector space which consists of elements $u\in\mathcal S'(\mathbb R^n)$ such that $\widehat u$ is measurable and $(1+|\xi|^2)^{\frac s2}\widehat u \in L^2(\mathbb R^n)$. We endow $H^s(\mathbb R^n)$ with the inner product \begin{equation} \langle u,v\rangle_{H^s(\mathbb R^n)}:=\int_{\mathbb R^n}(1+|\xi|^2)^s\widehat u(\xi) \overline{\widehat v (\xi)}d\xi \end{equation} and we denote
\begin{equation} \lVert u\rVert_{H^s(\mathbb R^n)} :=\left(\int_{\mathbb R^n}(1+|\xi|^2)^s|\widehat u(\xi)|^2d\xi \right)^{\frac 12} \end{equation} the corresponding norm.

Let $s\in \mathbb R$ and $v\in H^{-s}(\mathbb R^n)$. If $u\in H^s(\mathbb R^n)$, the map $\xi\mapsto \widehat u(\xi)\widehat v(-\xi)$ is in $L^1(\mathbb R^n)$.

Indeed, for all $\xi\in \mathbb R^n$, $\widehat u(\xi)\widehat v(-\xi)=\widehat u(\xi) (1+|\xi|^2)^{\frac s2}\widehat v(-\xi)(1+|\xi|^2)^{-\frac s2}$, and the RHS is the product of two map of $L^2(\mathbb R^n)$. By Cauchy-Schwarz inequality, we have
$$\left|\int_{\mathbb R^n}\widehat u(\xi)\widehat v(-\xi)d\xi\right|\leq \lVert u\rVert_{H^s(\mathbb R^n)} \lVert v\rVert_{H^{-s}(\mathbb R^n)}.$$ Therefore, if $v\in H^{-s}(\mathbb R^n)$, the map $L_v$ defined by \begin{equation} u\in H^s(\mathbb R^n)\mapsto L_v(u)=(2\pi)^{-n}\int_{\mathbb R^n}\widehat u(\xi)\widehat v(-\xi)d\xi =\int_{\mathbb R^n}\widehat u(\xi)\overline{\mathcal F}v(\xi)d\xi \end{equation} is a continuous linear functional on $H^s(\mathbb R^n)$ (hence an element of $(H^s(\mathbb R^n))'$ and $\lVert L_v\rVert _{(H^s(\mathbb R^n))'}\leq (2\pi)^{-n}\lVert v\rVert_{H^{-s}(\mathbb R^n)}$. We can define $L\colon H^{-s}(\mathbb R^n)\to (H^s(\mathbb R^n))'$ by $L(v)=L_v$.

Theorem The map $L$ defined above is linear, bijective, and bicontinuous. It allow us to indentify the dual space of $H^s(\mathbb R^n)$ by $H^{-s}(\mathbb R^n)$.

Sketch of proof:

  • The fact that $L$ is linear comes from linearity of the inverse Fourier transform and the integral. Continuity follows from $\lVert L_v\rVert _{(H^s(\mathbb R^n))'}\leq (2\pi)^{-n}\lVert v\rVert_{H^{-s}(\mathbb R^n)}$.
  • $L$ is surjective. Indeed, let $T\in (H^s(\mathbb R^n))'$. We can show that there exists a constant $C$ such that for all $\varphi\in \mathcal S(\mathbb R^n)$ $$|\langle T,\varphi\rangle|\leq C\lVert (1+|\xi|^2)^{s/2}\mathcal F^{-1}\varphi\rVert_{L^2(\mathbb R^n)}.$$ We can see that $(1+|\xi|^2)^{-s/2}\widehat T$ is a linear functional over $L^2(\mathbb R^n)$ and we apply Riesz representation theorem.
  • Injectivity needs only the property $\mathcal F^{-1}\mathcal F(\varphi)=\varphi$ for all $\varphi\in\mathcal S(\mathbb R^n)$.
  • The inverse map is continuous using Banach isomorphism theorem.

When $\Omega$ is a arbitrary open subset of $\mathbb R^n$ we have to use charts (and $\Omega$ needs to be regular enough, since the Sobolev of integer order, defined in a classical way may be not equal to the corresponding one, even if $\Omega$ is bounded), but at least it shows the idea.

$\endgroup$

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.