5
$\begingroup$

Let $L=\Delta + c$ in 3 dimensions, where $c$ is a positive constant.

I met this modified mean value property of a solution $u$ of $Lu=0$ as $$u(\xi)=\frac{\sqrt{c}\rho}{sin(\sqrt{c}\rho)}\frac{1}{4\pi \rho^2}\int_{\partial B(\xi,\rho)} u(x)d\sigma(x)$$ where $sin(\sqrt{c}\rho) \ne 0$, and $\sigma$ denotes the surface measure.

I'm trying to prove it. The following is what I did.

First I figured out that $$K(x,\xi)=-\frac{cos(\sqrt{c}r)}{4\pi r}, r=|x-\xi|$$ is a fundamental solution of $L$ with pole $\xi$. Then I tried to apply the second Green identity (which is also true for $L$ for sure) $$\int_{\Omega}(LG) u-\int_{\Omega}G (Lu)=\int_{\partial \Omega} \frac{\partial G}{\partial \nu} u- G \frac{\partial u}{\partial \nu}$$ For fixed $\rho$, let $G=K+\frac{cos(\sqrt{c}\rho)}{4\pi \rho}$ and $\Omega=B_{\rho}(\xi)$, so that $G$ vanishes on the boundary of $\Omega$, and $LG=\delta_{\xi}+c\frac{cos(\sqrt{c}\rho)}{4\pi \rho}$, where $\delta_{\xi}$ denotes the Dirac mass at $\xi$. Then the identity above becomes $$u(\xi)+c\frac{cos(\sqrt{c}\rho)}{4\pi \rho}\int_{B_{\rho}(\xi)}u(x)dx=\frac{\sqrt{c}\rho sin(\sqrt{c}\rho)+cos(\sqrt{c}\rho)}{4 \pi \rho^2} \int_{\partial B(\xi,\rho)} u(x)d\sigma(x)$$

Then no matter how hard I tried, I just could not get the desired form.

I also tried to take the derivative with respect to $\rho$ of the desired form, and I got a complicated form which doesn't seem to be zero. Then I took the second derivative, and things got worse...

Can anyone either give the right reference about the solution of this operator $L$, or help me figure out how to obtain the desired form? Thanks!

$\endgroup$
1
$\begingroup$

Ah, it is one of these cases when you better know your formula in all dimensions. It is $$ \int_{|x-y|=\rho} u(x) d\sigma(x) = (2\pi)^{n/2} \cdot \frac{J_{n/2-1}(\sqrt{c}\rho)}{(\sqrt{c}\rho)^{n/2-1}} \cdot u(y) $$
One of the ways to prove it: $$ \Delta u + c u = 0, $$ $$ \int_{|x-y|\leq R} \Delta u(x) + c u(x) dx = 0. $$ Apply Green's formula to the first integrand, $$ \int_{|x-y|= R} \frac{\partial u}{\partial r}(x) d\sigma(x) + \int_{|x-y|\leq R} c u(x) dx = 0, $$ (here $\frac{\partial u}{\partial r}(x) = \nabla u(x)\cdot x/|x|$). $$ R^{n-1}\int_{|x|=1} \frac{\partial u}{\partial r}(y-Rx) d\sigma + c \int_{|x-y|\leq R} u(x) dx = 0. $$
Differentiate w.r. to $R$ $$ R^{n-1}\int_{|x|=1} \frac{\partial^2 u}{\partial r^2}(y-Rx) d\sigma +(n-1)R^{n-2} \int_{|x|=1} \frac{\partial u}{\partial r}(y-Rx) d\sigma + c R^{n-1} \int_{|x|=1} u(y-Rx) d\sigma = 0, $$ $$ \int_{|x|=1} \frac{\partial^2 u}{\partial r^2}(y-Rx) d\sigma +\frac{n-1}{R}\int_{|x|=1} \frac{\partial u}{\partial r}(y-Rx) d\sigma + c \int_{|x|=1} u(y-Rx) d\sigma = 0. $$ $$ \frac{d^2}{dR^2}\int_{|x|=1} u (y-Rx) d\sigma +\frac{n-1}{R} \frac{d}{dR}\int_{|x|=1} u(y-Rx) d\sigma + c \int_{|x|=1} u(y-Rx) d\sigma = 0. $$ So $v(R) = \int_{|x|=1} u (y-Rx) d\sigma $ is a solution of $$ v''(R) + \frac{n-1}{R} v'(R) + c v(R)=0. $$ This equation is one of the forms of Bessel's equation $$ x^2 y''+ xy' + (x^2-\alpha^2) y = 0. $$ In particular, it is not difficult to show that $$ v(R) = C_1 \frac{J_{n/2-1}(R\sqrt{c})}{R^{n/2-1}}, $$ here $J_\nu$ is the Bessel function of the first kind. To get the right constant just send $R\to 0$: on the one hand $$ \lim_{R\to 0} v(R) = \omega_{n-1} \cdot u(y), $$ on the other $$ \lim_{R\to 0} v(R) = C_1\lim_{R\to 0} \frac{J_{n/2-1}(R\sqrt{c})}{R^{n/2-1}} = C_1 \frac{(\sqrt{c}/2)^{n/2-1}}{2\pi^{n/2}} \omega_{n-1}. $$ So $C_1 = (2\pi)^{n/2} c^{1/2-n/4} u(y)$.

$\endgroup$
  • $\begingroup$ any references for MVP for the above formula and even better for more general pdes eg. elliptic pdes if it still works for some types? $\endgroup$ – OOESCoupling Dec 24 '16 at 17:28
  • $\begingroup$ @ThomasKojar See Ch 18 in Titchmarsh's Eigenfunction expansions .... part 2 . He considers the case of Laplace + potential. There is a local formula for elliptic operators which is called Moiseev's MVP . It can be found in the book of Il'in Spectral theory of differential operators $\endgroup$ – almaz Dec 24 '16 at 18:38
  • $\begingroup$ It generalizes for many elliptic and parabolic pdes: 1)"Spherical Means for Pdes" edited by K. Karl Karlovich Sabelfeld, I. A. Shalimopva 2)"Spherical and plane integral operators for PDEs : construction, analysis, and applications", Karl K. Sabelfeld, Irina A. Shalimova, $\endgroup$ – OOESCoupling Dec 25 '16 at 19:59
  • $\begingroup$ @ThomasKojar This is good to know. Thanks for the reference ! $\endgroup$ – almaz Dec 27 '16 at 17:29

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.