5
$\begingroup$

We know by mean value property that harmonic functions satisfies the equalities \begin{equation} u(x) = \dfrac{1}{|B_r|}\int_{B_r}f dx = \dfrac{1}{|\partial B_r|}\int_{ \partial B_r}udS. \end{equation} I'm wondering if we do not have harmonic functions, instead we have $\Delta u =f$. Is there some equality like above? For example adding some term depending on $f$ or its intergral on the right hand side?

$\endgroup$
1
  • 4
    $\begingroup$ There is indeed such an equality. This is usually called $Green's\;representation\;formula$. $\endgroup$ – Etienne Jun 13 '13 at 18:56
3
$\begingroup$

Consider the problem in $\mathbb{R}^n$. If $$ u(x) = \frac{1}{|\partial B(x,r)|}\int_{ \partial B(x,r)}u(y) \,dS(y) ,\tag{1} $$ then $u$ has to be harmonic. Please refer to the proof in Evans' book about mean value property. For if define $$ \phi(r) = \frac{1}{|\partial B(x,r)|}\int_{ \partial B(x,r)}u(y)\,dS(y) = \frac{1}{|\partial B(0,1)|}\int_{ \partial B(0,1)}u(x+rz)\,dS(z), $$ and $$ \phi'(r) = \frac{1}{|\partial B(0,1)|}\int_{ \partial B(0,1)}\nabla u(x+rz)z\,dS(z) \\ = \frac{1}{|\partial B(x,r)|}\int_{ \partial B(x,r)} \nabla u(y) \cdot n\,dS(y)\\ =\frac{1}{|\partial B(x,r)|} \int_{ B(x,r)} \Delta u(y) \,dy = \frac{1}{|\partial B(x,r)|} \int_{ B(x,r)} f(y) \,dy.$$

If $f\neq 0$, $\phi'(r)\neq 0$, and $\phi(r)$'s value depends on $r$. If (1) holds, then $u(x)$'s value is not unique, contradiction. If $u$ is not harmonic we have to find something else.


If $f$ is radially symmetric with respect to the point $p$, then you have $u$ is radially symmetric with respect to point $p$. We can set $$u(x) = v(|x-p|) = v(r), \quad \text{ and }\quad f(x) = f(|x-p|) = h(r),$$ then $$ \Delta u = v''(r) + \frac{n-1}{r}v'(r) = h(r), $$ Assuming $u(x)$ is smooth everywhere, then $$ v(r) = v(0) + \int^r_0\tau^{1-n}\int^{\tau}_0 \xi^{n-1}h(\xi) d\xi, $$ and translates to $$ u(x) = u(p) + \int^{|x-p|}_0\frac{1}{|\partial B(p,\tau)|} \left(\int^{\tau}_0 \Big(\int_{\partial B(p,r)} \Delta u(y)dy\Big)dr \right) d\tau, $$ which is a formula similar to mean value formula for we are integrating on $(n-1)$-spheres layer by layer around a point $p$, also with an unspecified boundary value.


Lastly, you can represent the solution using Green's formula if boundary value is given in a domain, please refer to Evans' PDE book Chapter 2 Theorem 12.

$\endgroup$

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.