In chapter 5 of Evans Partial differtial Equations (Section 5.5) he defines the trace operator for a bounded domain $\Omega$ with given smoothness as,

$T:H^1(\Omega)\rightarrow L^2(\partial\Omega)$

Where this operator is continuous i.e., $\|\gamma u\|_{L^2(\partial\Omega)}\leq C\|u\|_{H^1(\Omega)}$.

My question is does there exist a continuous trace operator for function in $L^2(\Omega)$. That is to say $\gamma: L^2(\Omega)\rightarrow L^2(\partial\Omega)$ s.t, $\|\gamma u\|_{L^2(\partial\Omega)}\leq C\|u\|_{L^2(\Omega)}$.

The trace operator is defined for continuous functions and then extended to functions in $H^1$. Since continuous functions are dense in $L^2$ I believe the result should hold.

Second question that I have is whether the trace operator has a continuous inverse i.e, is there a $T^{-1}:L^2(\partial\Omega)\rightarrow H^1(\Omega)$ that is also continuous. I can find a result for $\Omega=\mathbb{R}^d$ but not for bounded $\Omega$.

Thank you for any help.


1) No. The boundary of the domain in general has zero measure. Evaluating a $L^2$-function on a zero measure set is not well-defined. The continuation argument does not work, as the mapping $\gamma : H^1 \subset L^2(\Omega) \to L^2(\partial\Omega)$ is not bounded in the sense that there is no $C>0$ such that $\|\tau u\|_{L^2(\partial\Omega)}\le C \|u\|_{L^2(\Omega)}$ for all $u\in H^1$.

2) This is not true. The trace operator from $H^1$ to $L^2(\partial\Gamma)$ is compact in general. You have to use Sobolev-Slobodecki spaces to get invertibility.

  • $\begingroup$ Thank you for the help this clarifies a lot. I have started looking at the Sobolev-Slobodeck spaces. $\endgroup$ – CuriousCat Jan 15 '20 at 8:25

Here is another argument which shows that there cannot be a trace operator $\gamma$ for $L^2(\Omega)$ with

  • $\gamma(f) = f_{|\partial\Omega}$ for all $f \in C(\bar\Omega)$,
  • $\gamma \colon L^2(\Omega) \to L^2(\partial\Omega)$ is continuous.

First, we have $\gamma(\varphi) = 0$ for all $\varphi \in C_c^\infty(\Omega)$, since these functions are continuous and vanish on $\partial\Omega$. Since $C_c^\infty(\Omega)$ is dense in $L^2(\Omega)$, we get $\gamma(f) = 0$ for all $f \in L^2(\Omega)$. This, however, contradicts $\gamma(1) = 1$ (here we used that constant functions are continuous).


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.