2
$\begingroup$

Consider the Laplace equation in a ball $B(r) \subset \mathbb{R}^n$ of radius $r$: $$ \begin{cases} -\Delta u &= 0, \quad \text {in} \quad B(r), \\ \ \ \ \ \ \, u&= g, \quad \text {in}\quad \partial B(r). \end{cases} $$ Suppose $g$ can be extended to a function $\overline g$ defined in $B((1+\sigma)r)$ such that $\overline g \in C^{1,\alpha}(B((1+\sigma)r))$ for some $\alpha, \sigma >0$ and $g=\overline g$ on $\partial B(r)$.

Now my question is that is it possible to extend $u$ to a function $\overline u$ defined in some ball $B((1+\delta)r)=:B_\delta$, for $\delta >0$, such that $\overline u$ is a subsolution in $B_\delta$. Moreover, if possible, I would also like to have $|\overline u_{x_i}|$ to be a subsolution in $B_\delta$, for every $i=1, \dots, n$.

$\endgroup$
2
$\begingroup$

One can build a distributional subsolution which is a Lipschitz extension of $u$ by making an extension with a positive jump in radial derivative across the boundary. Say $u$ is harmonic on $B_1$ and we want to subharmonically extend to $B_2$.

Let $v$ be a harmonic function on $B_2-B_1$ with boundary data $g$ on $\partial B_1$ and, say, constant on $\partial B_2$. Since the boundary data are $C^{1,\alpha}$ we have that $|\nabla u|,|\nabla v| < C$ globally. Let $$w(x) = \begin{cases} u(x), \quad x \in B_1 \\ v(x) + C(|x|^2-1), x \in B_2 - B_1 \end{cases}$$

Then $w$ is a Lipschitz extension of $v$ with a positive jump in the radial derivative across $\partial B_1$. Let $w_r^+$ denote the radial derivative from outside and $w_r^-$ denote the radial derivative from inside.

We can now show that $w$ is distributionally subharmonic. Since it is harmonic in $B_1$ and subharmonic outside $B_1$, we just need to check on $\partial B_1$. Take any ball $B_{\eta}$ centered on $\partial B_1$ and let $\nu$ be the outer normal to $\partial B_{\eta}$. We compute (using that $w$ is harmonic in $B_1$) $$\int_{\partial B_{\eta}} w_{\nu} = \int_{\partial B_{\eta} \cap B_1^c} w_{\nu} - \int_{B_{\eta} \cap \partial B_1} w_r^-.$$ Since $$0 < \int_{B_{\eta} \cap B_1^c} \Delta w = \int_{\partial B_{\eta} \cap B_1^c} w_{\nu} - \int_{ B_{\eta} \cap \partial B_1} w_r^+,$$ and $w_r^- \leq w_r^+$ we conclude that $$\int_{\partial B_{\eta}} w_{\nu} \geq 0$$ showing the desired property.

$\endgroup$
1
  • $\begingroup$ Thanks! I guess something similar might work also for the partial derivatives of $u$. $\endgroup$ Apr 8 '14 at 17:49

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.