5
$\begingroup$

$\newcommand{\v}[1]{\boldsymbol{#1}}$For an $u\in H^1(\Omega) = W^{1,2}(\Omega)$ where $\Omega$ is Lipschitz, we have

$$ {\|u - \frac{1}{\left|\Omega\right|} \int_{\Omega} u\|}_{L^2}\leq C {\|\nabla u \|} $$

and the Poincare constant is basically a multiple of diameter of the domain.

However in $\mathbb{R}^3$, the only similar result for $\mathbf{curl}$-square integrable vector fields $\v{u}$ would be:

$$ {\|\v{u} - \frac{1}{\left|\Omega\right|} \int_{\Omega} \v{u}\|}_{L^2} \leq {\|\mathbf{curl} \ \v{u}\|} $$

if $\v{u}$ is divergence free.

If not, suppose $\v{u}$'s divergence is not well-defined, then we could have:

$$ {\|\v{u} - \frac{1}{\left|\Omega\right|} \int_{\Omega} \v{u}\|}_{L^2(\Omega)} \leq C_1 \left(\|\v{u}\|_{L^2(\Omega)}^2+{\|\mathbf{curl} \ \v{u}\|}^2 \right)^{1/2} $$

Since we could still do Helmholtz decomposition $\v{u} = \v{w}+\nabla p$, qualitatively speaking, locally in the a compact subdomain $K\subset \Omega$, if the $\v{u}$ is more irrotational(that gradient field $\nabla p$ is dominant), then $C_1$ is closed to $1$, if the $\v{u}$ is more weakly-solenoidal($\v{w}$ is dominant), then the $C_1$ is more closed to $\sqrt{1+d^2}$.

I am curious if we could decompose the domain $\Omega$ into different parts, such that we estimate the Poincare constant locally and then put together to get a sharper bound?

$\endgroup$
3
  • $\begingroup$ I don't quite grok your paragraph starting "Since we could..." Isn't $w$ the divergence free part? Then when $w$ is dominant, shouldn't you get closer to the divergence free case, and the reverse when $\nabla p$ is dominant? $\endgroup$ Sep 11, 2012 at 11:09
  • $\begingroup$ BTW, have you seen arxiv.org/abs/1208.6045 ? $\endgroup$ Sep 11, 2012 at 11:12
  • $\begingroup$ @Willie Wong, thanks for the comment, the main reason I am doing this is because I want to circumvent the divergence part, since for general $H(\mathbf{curl})$ vector fields, the divergence is not defined, and $w$ is in the weakly divergence free, such that I could get a Poincare inequality solely for $H(\mathbf{curl})$ vector fields not having the divergence free constraint. Thanks for the heads up of the uniformity paper as well, I will look into it. $\endgroup$
    – Shuhao Cao
    Sep 11, 2012 at 16:02

1 Answer 1

1
$\begingroup$

I don't understand what you are trying to say in your third inequality, as $C_1$ is always $1$. Consider the minimization problem $$ I_u=\inf_{k\in\mathbb R} \left\| u -k \right\|_{L^2(\mathbb \Omega)}^2 $$ On one hand, it is clear that $I_u \leq \left\| u \right\|_{L^2(\mathbb \Omega)}^2$, since this upper bound corresponds to $k=0$. On the other hand, since $$ \left\| u -k \right\|_{L^2(\mathbb \Omega)}^2 = k^2 \left|\Omega\right| -2 k \int_\Omega u ~\text d x+ \int_\Omega u^2 \text d x $$ the infimum is a minimum corresponding to $k= \frac 1 {\left|\Omega\right|} \int_\Omega u \text d x$ thus $$ \left\| u -\frac 1 {\left|\Omega\right|} \int_\Omega u ~\text d x \right\|_{L^2(\mathbb \Omega)} \leq \left\| u \right\|_{L^2(\mathbb \Omega)}, $$ and equality is attained for any function with zero average. Surely you meant something else?

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.