I'm studying Elliptic Partial Differential Equations by Q. Han and F. Lin. In Lemma 1.41 is given the elliptic equation $D_j(a_{ij}D_i u)=0$ where the coefficient matrix $(a_{ij})$ is constant positive definite and elliptic (with the bound $0<\lambda\leq\Lambda$). There is an energy estimate of its weak solution $u\in C^1(B_1)$ which reads $$\int_{B_\rho}u^2\leq c\rho^n\int_{B_1}u^2$$ where $c=c(\lambda,\Lambda)$. Is there any reason that the case $1/2<\rho\leq 1$ is trivial? I can hardly find out why it is evident (to the authors). If so, why does this inequality holds in this case?

  • $\begingroup$ Does the maximum principle hold for this PDE? $\endgroup$ – user7530 Aug 17 '13 at 15:19
  • $\begingroup$ Yes, it does. Does this fact explain anything? $\endgroup$ – Sunghan Kim Aug 17 '13 at 16:33
  • $\begingroup$ I just added the reference in the content. Thanks for your advice. $\endgroup$ – Sunghan Kim Aug 18 '13 at 0:44

The authors neglect to mention that $c$ depends on dimension $n$ as well. It should have been $c=c(\lambda,\Lambda,n)$. Indeed, if you look at the proof on page 23, you'll see that

  • in Method 1, they use a linear change of variables. This contributes the Jacobian determinant, which depends on $n$.
  • in Method 2, they choose $k$ depending on $n$, and use a constant that depends on $k$

Now that the dependency of $c$ on $n$ is admitted, we can take $c\ge 2^n$ to satisfy the inequality for $\rho\in [1/2,1]$.

  • 1
    $\begingroup$ I was also thinking that $c$ must depend on $n$ as well as $\lambda$ and $\Lambda$. Moreover, the inequality (in Method 1) $$\int_{B_\rho}|u|^2\leq \rho^n\|u\|_{L^\infty(B_{1/2})}$$ is also incorrect; it is a H\"{o}lder inequality so we need to multiply the right side by $\omega_n=|B_1|$. Thanks for your help. $\endgroup$ – Sunghan Kim Aug 19 '13 at 23:58

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.