Let $\Omega \subset \mathbb{R}^d$ be a open set, $L^2(\Omega)$ be the Lebesgue space, and $H^1(\Omega)$ be the Sobolev space.

Does it hold that

$(u,v)_{H^1} =0 \Rightarrow (u,v)_{L^2} =0$?


No, it doesn't hold. For $u,v\in C^1(\Omega)\cap H^1(\Omega)$ we have $(u,v)_{H^1}=(u,v)_{L^2}+(u',v')_{L^2}$. But the dot product can be negative too, so $(u,v)_{H^1}=0$ implies $(u,v)_{L^2}=-(u',v')_{L^2}$, which is not necessarily $0$.

For $d=1$, $\Omega=(0,1)$ we can consider $u(x)=e^{2 x}$ and $v(x)=e^{-\frac12 x}$. We get $$ (u,v)_{H^1}=\int_0^1e^{\frac32x}~dx+\int_0^1-e^{\frac32x}~dx=0$$ while $$ (u,v)_{L^2}=\int_0^1e^{\frac32x}~dx=\frac23\neq 0. $$ You can also see that the other direction is false too. If $(u,v)_{L^2}=0$ we can't deduce $(u,v)_{H^1}=0$. Here we consider $u(x)=x^2-\frac12$ and $v(x)=x$. We get $$ (u,v)_{L^2}=\int_0^1x^3-\frac12 x~dx=\frac14-\frac12\cdot\frac12=0 $$ while $$ (u,v)_{H^1}=0+\int_0^12x~dx=1\neq 0. $$ All in all, you see that orthogonality in $L^2$ and $H^1$ are not the same.


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.