The following is the beginning of a proof of the Poincaré inequality:

enter image description here

Here are my questions:

  • How is the limiting argument made so that "it is sufficient to prove the inequality for $u\in C_c^\infty(\Omega)$?
  • What does "The inequality is invariant under rotations and translations" mean? Could anyone write it out explicitly?
  • $\begingroup$ For the first question because it still applies by approximation, taking the limit under the integral sign. For the second question, I think which the reference is to the Lebesgue measure. $\endgroup$ – Andrew Sep 5 '16 at 18:12
  • 2
    $\begingroup$ Id say the "proving it for $u \in C^\infty_c$ is enough" part is the most important idea of all functional analysis. $\endgroup$ – reuns Sep 5 '16 at 20:36
  • $\begingroup$ @user1952009 You really are right! in every book I find ... is sufficient to prove that..ok, why? $\endgroup$ – Andrew Sep 5 '16 at 21:21

1) Suppose you prove the inequality for all smooth $u$ with compact support in $\Omega$. Let $u \in H^1_0(\Omega)$. Then by density of $C_c^\infty(\Omega)$ in $H_0^1$, there exists a sequence $\phi_k \in C_c^\infty(\Omega)$ converging to $u$ in the $H^1_0(\Omega)$ norm. Therefore $$\lim_{k\to \infty} \int_\Omega \phi_k^2 \, dx = \int_\Omega u^2 \, dx \ \ \ \text{ and } \ \ \ \lim_{k\to \infty} \int_\Omega |D\phi_k|^2 \, dx = \int_\Omega |Du|^2 \, dx.$$ Since the estimate holds for each term $\phi_k$ in the sequence, with the same uniform constant $C$, it must hold for $u$ as well. This is an argument by density or approximation.

2) A rotation and translation is a transformation of the form $y=Ax + b$, where $b \in \mathbb{R}^n$ and $A$ is an orthogonal matrix ($A^TA=AA^T=I$, and $\text{det}(A)=1$). Suppose we change variables and set $v(y)=u(x)=u(A^T(y-b))$. Then $dy = |\text{det}(A)|dx = dx$ and $Dv(y) = ADu(x)$. Since $A$ is orthogonal, $|Dv(y)|=|Du(x)|$ and we have $$\int_\Omega u(x)^2 \,dx = \int_{\Omega'} v(y)^2 \, dy \ \ \ \text{ and } \ \ \ \int_\Omega |Du(x)|^2 \,dx = \int_{\Omega'} |Dv(y)|^2 \, dy,$$ where $\Omega' = A\Omega + b$. Since both sides of the estimate are invariant under rotations and translations, we may as well assume the bounded direction lies in the $x_n$ direction between $0 < x_n < a$.


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.