4
$\begingroup$

I am curious whether there are any Liouville theorems for the following pde:

$$ - \Delta \phi(x) + a(x) \cdot \nabla \phi(x)=0 \qquad \mbox{in } R^N $$

where $a(x)$ is a smooth bounded vector field with $ |x| |a(x)| \rightarrow 0$ as $ |x| \rightarrow \infty$. As far as the solutions $ \phi$ I am assuming that $ \phi$ is smooth and the gradient satisfies a decay condition like $ | \nabla \phi(x)| \le C |x|^{-\sigma}$ for large $ x$ where $ \sigma>0$ is some small parameter.

As a second question, how about if we additionally assume $a(x)$ is divergence free.

Note if $ \phi$ went to zero at $\infty$ then we can get a Liouville theorem directly from the maximum principle. So one think I tried is taking a derivative of the equation to get a solution which decays... but I am not having much luck.

any comments or counterexamples would be welcomed. Craig

$\endgroup$

1 Answer 1

7
$\begingroup$

If $\Delta u + b(x) \cdot \nabla u = 0$ and $u$ is bounded, then $u$ is constant provided $|b| = O(1/(1+|x|))$. This follows from scaling, the Harnack inequality and the maximum principle.

Indeed, add a constant so that $\inf_{\mathbb{R}^n} u = 0$. Let $v(x) = u(Rx)$ for $R$ large. Then $v$ solves the equation $$\Delta v + Rb(Rx) \cdot \nabla v = 0.$$ By hypothesis, $R|b(Rx)|$ is uniformly bounded in $B_1-B_{1/4}$ for all large $R$, and for $R$ large we may assume by the maximum principle that $\inf_{\partial B_{1/2}}v < \epsilon$ for some $\epsilon$ small. The Harnack inequality, applied to $v$ in overlapping balls in the annulus $B_{3/4}-B_{1/4}$, implies that $v < C(n)\epsilon$ on $\partial B_{1/2}$, so by the maximum principle $u < C\epsilon$ in $B_{R/2}$. Since $\epsilon$ is arbitrarily small and this holds for all $R$ large, $u = 0$.

$\endgroup$
2
  • $\begingroup$ Thank you very much for the proof. Do you have any intuition on whether you think the Liouville theorem should hold under this gradient condition? $\endgroup$
    – Craig
    Nov 13, 2014 at 1:52
  • $\begingroup$ Why can we assume that $\inf_{\partial B_{1/2}}<\epsilon$ for $R$ large? I can’t understand how was the maximum principle used in this step. Can you elaborate? Thanks in advance. $\endgroup$
    – Feng
    Apr 14, 2021 at 8:24

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.