I have a question on a harmonic function and the boundary behavior.

Let $\mathbb{U} \subset \mathbb{C}$ be a unit disk. We denote by $\overline{\mathbb{U}}$ the closure of $\mathbb{U}$ in $\mathbb{C}$.

We have a reflected Brownian motion $X=(\{X_t\}_{t \ge0}, \{P_x\}_{x \in \overline{U}})$ on $\overline{\mathbb{U}}$. Let $B:=B(a,r)$ be an open disk centered at $a \in \mathbb{C}$ with radius $r>0$ and $C$ a closed disk such that $C \subset \mathbb{U} \cap B$.

$u(x):=P_{x}(\sigma_{C}>\tau_{B})$ is a harmonic function with respect to $X$ on $(\overline{\mathbb{U}}\cap B) \setminus C$, which satisfies $\lim_{x \to \partial C}u(x)=0$. Here, we define

\begin{align*} \sigma_C=\inf\{t>0 \mid X_t \in C\},\\ \tau_B=\inf\{t>0 \mid X_t \notin B\}. \end{align*}

In other words, $u$ is a positive harmonic function on $ (\mathbb{U} \cap B) \setminus C$ with the Neumann boundary condition on $\partial \mathbb{U} \cap B$ and the Dirichlet boundary condition on $\partial C$.


How $u(x)$ behave as $x \to \partial C$?

I am intersted in the rate of convergence of $\lim_{x \to \partial C}u(x)=0$.

Can we construct a nice positive harmonic function on $\mathbb{U} \setminus C$ with the Neumann boundary condition on $\partial \mathbb{U}$ and the Dirichlet boundary condition on $\partial C$?? If we know the behavior near $\partial C$, we should be able to obtain the behavior of $u$ near $\partial C$ by the boundary Harnack inequality.


Roughly: $u(x) \approx \operatorname{dist}(x,C)$ (in the sense that the ratio is bounded), except near the corners, where $u(x) \approx |x - x_0|^{(2\alpha / \pi) - 1} \operatorname{dist}(x,C)$, where $x_0$ is a corner point and $\alpha$ is the interior angle at $x_0$.

The easiest way to see this is to map your domain conformally into a square $[0,1] \times [0,1]$, so that the image $v$ of $u$ is harmonic, with Neumann boundary condition along vertical sides, Dirichlet condition $v = 0$ along the bottom side, and Dirichlet condition $v = 1$ along the top side. Then $v(x) = \operatorname{Im} x$, and everything boils down to the properties of conformal maps.

| cite | improve this answer | |
  • $\begingroup$ Thank you for your kind reply. However, our domain is $(U \cap B) \setminus C$, right? This is not a Jordan domain. How do you map our domain to the square? $\endgroup$ – sharpe Nov 11 '19 at 19:46
  • $\begingroup$ Sorry. I forgot to write the definition of $\tau_B$ and $\sigma_C$. $\endgroup$ – sharpe Nov 11 '19 at 19:55
  • $\begingroup$ Ah, I thought $C$ intersects $U$ as well. Then simply $u(x) \approx \operatorname{dist}(x, C)$ by the boundary Harnack inequality, right? $\endgroup$ – Mateusz Kwaśnicki Nov 11 '19 at 20:02
  • 1
    $\begingroup$ Yes (assuming that $K = C$ in your comment). BHI is a local result: in your case $u$ is a positive harmonic function in a $C^{1,1}$ domain $D = B(x_0, r_2) \setminus B(x_0, r_1)$ (where $C = \overline{B}(x_0, r_1)$ and $r_2 > r_1$ is small enough), which goes to zero continuously on $\partial C = \partial B(x_0, r_1)$. Then $u$ is necessarily comparable with the distance to $\partial C$. $\endgroup$ – Mateusz Kwaśnicki Nov 11 '19 at 20:32
  • 1
    $\begingroup$ If we only assume that $u$ bounded by $1$ in $D$, then, by comparison principle, $u(x) \le \log(|x - x_0| / r_1) / \log(r_2 / r_1)$, and this cannot be improved. For the function $u$ in the original question: things get complicated if $\partial C$ gets too close to either $\partial B$ (then $A$ in the upper bound explodes just as above) or $\partial U$ (then a similar $A$ in the lower bound goes to zero super-fast). $\endgroup$ – Mateusz Kwaśnicki Nov 11 '19 at 21:23

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.