1

Let $M$ be a closed Riemannian manifold and $D$ a first-order self-adjoint elliptic differential operator.

Then we can define the first Sobolev space $H^1(M)$ as the completion of $C_c^\infty(M)$, the compactly supported smooth functions on $M$, with respect to the norm $$\|u\|_{H^1}=(\|u\|^2_{L^2}+\|Du\|^2_{L^2})^{1/2}.$$

Clearly, we have $\|u\|_{L^2}\leq\|u\|_{H^1}$, so the identity map $\text{id}$ on $C_c^\infty(M)$ extends continuously to a map $$\widetilde{\text{id}}\colon H^1(M)\to L^2(M).$$

Question: Is $\widetilde{\text{id}}$ an injection?

The fact that $H^1(M)$ is lies inside $L^2(M)$ is clear from other definitions of Sobolev space (e.g. here), but it should also be possible to see this from the above definition, although I don't quite see it.

geometricK
  • 5,115

1 Answers1

2

Let $u\in H^1(M)$. If $(u_n)$ is a sequence in $C_c^\infty(M)$ such that $u_n\to u$ in $H^1(M)$, then by definition $\widetilde{\mathrm{id}}(u)=\lim_{n\to\infty}u_n$ in $L^2(M)$, where existence and independence of this limit from the chosen sequence $(u_n)$ are standard exercises.

Therefore, to show that $\widetilde{\mathrm{id}}$ is injective, it suffices to show that whenever $(u_n)$ is an $H^1$-Cauchy sequence in $C_c^\infty(M)$ such that $u_n\to 0$ in $L^2(M)$, then $\|u_n\|_{H^1}\to 0$. In particular, $Du_n\to v$ for some $v\in L^2(M)$. Since $D$ is assumed to be symmetric on $C_c^\infty(M)$, it is closable (with closure $D^{\ast\ast}$). Hence $u_n\to 0$ and $Du_n\to v$ in $L^2(M)$ together imply $v=0$. Thus $\|u_n\|_{H^1}\to 0$.

Some remarks are in order: A quadratic form $q\colon D(q)\subset H\to H$ is called closable if the inclusion $D(q)\subset H$ extends to an injective linear map from the completion of $D(q)$ with respect to the norm $\|u\|_q=(\|u\|_H^2+q(u))^{1/2}$ into $H$. What I have shown above is essentially that if $q$ is of the form $q(u)=\|Su\|_H^2$ for some closable operator $S$ on $D(q)$, then $q$ is closable.

Note that I did not use that the manifold $M$ is closed nor that $D$ is elliptic, and instead of self-adjointness of $D$ it suffices to assume that $D$ is symmetric on $C_c^\infty(M)$, which is usually easier to verify than self-adjointness (although it is true that every symmetric elliptic differential operator on $C^\infty(M)$ is essentially self-adjoint if the manifold is closed).

MaoWao
  • 16,366
  • 2
  • 21
  • 43