2
$\begingroup$

I've been recently looking for a best possible Poincaré constant for a particular domains $\Omega$ (it's related to my previous question Unique weak solution to Helmholtz equation on a square) for $W_0^{1,2}(\Omega)$-functions.

If $\Omega = (0,1)^2$, than I was able able to prove that \begin{equation*} ||u||_2^2\leq \frac{1}{16}||\nabla u||_2^2,\quad\forall u\in W_0^{1,2}((0,1)^2). \end{equation*}

I was wondering what happens if $\Omega=B(0, 1)\subseteq\mathbb{R}^2$? I can circumsribe a square about the circle and I get that \begin{equation*} ||u||_2^2\leq \frac{1}{4}||\nabla u||_2^2,\quad\forall u\in W_0^{1,2}(B(0,1)), \end{equation*} but I believe there must be a clever way how to do it. I don't necessarily need the best possible constant (i.e. the smallest one), but I'm looking for some reasonable estimate (without some magic tricks) that gives a constant that is as small as possible.

Thank you for any help:)

$\endgroup$
6
$\begingroup$

The key to what you're looking for is Rayleigh's theorem, which states that $$ \lambda_1 = \min_{0 \neq u \in W^{1,2}_0(\Omega) } \frac{\int_\Omega |\nabla u|^2 } {\int_\Omega u^2}, $$ where $\lambda_1 >0$ is the principal eigenvalue of the Laplacian with Dirichlet boundary condition, i.e. $$ \begin{cases} -\Delta u = \lambda_1 u &\text{in } \Omega \\ u = 0 & \text{on }\partial \Omega. \end{cases} $$ Why is this useful? For any $u \in W^{1,2}_0(\Omega)$ such that $u \neq 0$ the Rayleigh theorem tells us that $$ \lambda_1 \int_\Omega u^2 \le \int_\Omega |\nabla u|^2, $$ but this inequality also trivially holds for $u=0$ and thus for all $u \in W^{1,2}_0(\Omega)$. Thus $1/\lambda_1$ is the best constant in the Poincaré inequality since the infimum is achieved by the solution to the Dirichlet problem. Now, the crucial feature of this is that for a ball, namely $\Omega = B(0,r)$, we can explicitly compute the eigenfunctions and eigenvalues of the Laplacian by using the classical PDE technique of separation of variables.

It's here where I will trail off and let you go discover the rest of the details on your own. If you get stuck you can either look in Strauss's book Partial Differential Equations: An Introduction, or you can hit up google. For instance, I think this link has what you're looking for. The key point is that you will find a connection with all sorts of beautiful classical mathematics related to Bessel functions and their zeroes (I can't help but throw out a reference to the tour-de-force by Watson, A Treatise on the Theory of Bessel Functions).

Let me know if you'd like more help or extra references. I'm happy to help!

$\endgroup$
  • $\begingroup$ Thank you for you help! I believe this will help me. I'll go through it. Thank you $\endgroup$ – user1321324 Dec 8 '15 at 10:21

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.