## On symmetry problems key words: symmetry, Pompeiu problem, Schiffer's conjecture, Helmholtz equations AMS sub ject classification: 35R30, 35J40, 35J05

### A.G. Ramm Mathematics Department, Kansas State University, Manhattan, KS 66506-2602, USA ramm@math.ksu.edu

Abstract
Several symmetry problems are discussed. These include the Pompeiu problem and similar conjectures for the heat and wave equations.

1 Introduction

There is a large literature on symmetry problems for PDE, in particular, for the Helmholtz equation. The most well-known and still open problems are the Pompeiu problem and Schiffer's conjecture, which are stated below. In this paper some old and new conjectures are formulated. One of them, Conjecture A, is proved under assumption that Conjecture P holds. An old conjecture H is proved. It is proved that if a solution $u$  to Helmholtz equation in a domain $D$  , homeomorphic to a ball, is spherically symmetric in some ball $B\left({x}_{0},a\right)\subset D$  , centered at a point ${x}_{0}\in D$  of radius $a>0$  , and $u{|}_{\partial D}=const$  , then $D$  is a ball (Lemma 1). An analog of Conjecture P is proved (Lemma 2). A symmetry question related to a wave equation, is discussed.
Let us call the Pompeiu problem (in the form we state it) Conjecture P. Historically Pompeiu has published a wrong result, not the one which is stated below as Conjecture P (see, e.g. , ).
Conjecture P: Consider a bounded domain $D\subset {\mathbb{R}}^{n}$  , $n=3,$  homeomorphic to a ball, with a smooth boundary $\Gamma$  , and let $N$  be the outer unit normal to $\Gamma$  . Assume that ${k}^{2}>0$  and $c\ne 0$  are constants, and the problem:
 $\begin{array}{c}{\nabla }^{2}u+{k}^{2}u=0\mathit{i}\mathit{n}D,\end{array}$ (1.1)
 $\begin{array}{c}{u}_{N}{|}_{\Gamma }=0,u{|}_{\Gamma }=c,\end{array}$ (1.2)
is solvable. Then $D$  is a ball.
Below we always assume that $D$  and $\Gamma$  satisfy the assumptions stated in Conjecture P, and that the following assumption holds:
A closed smooth surface $S$  is homeomorphic to a sphere and lies inside $D$  .
We do not repeat the above assumptions.
The usual formulation of the Pompeiu problem is the following ( see, e.g., , ):
Pompeiu problem: Assume the solvability of the following overdetermined problem:
${\nabla }^{2}v+{k}^{2}v=1\mathit{i}\mathit{n}D,{v}_{N}{|}_{\Gamma }=0,v{|}_{\Gamma }=0.$  Then $D$  is a ball.
This formulation is equivalent to Conjecture P. Indeed, let $v=u+{k}^{-2}$  . Then $u$  satisfies (1.1) and (1.2), and $c=-{k}^{-2}$  .
The Schiffer's conjecture is:
Conjecture S: If equation (1.1) has a solution which satisfies the following conditions:
 $\begin{array}{c}{u}_{N}{|}_{\Gamma }=c\ne 0,u{|}_{\Gamma }=0,\end{array}$ (1.3)
then $D$  is a ball.
This Conjecture and Conjecture P are still not proved.
Let us formulate a few more conjectures. Conjecture A, stated below, apparently has not been formulated in the literature.
Recall that the surface $S$  has been defined above.
Conjecture A: If equation (1.1) has a solution which satisfies the following conditions:
 $\begin{array}{c}\nabla u{|}_{S}=0,u{|}_{\Gamma }=c\ne 0,\end{array}$ (1.4)
then $D$  is a ball.
Conjecture H: Let
 $\begin{array}{c}{u}_{t}={\nabla }^{2}u,t>0,x\in D,u{|}_{t=0}=0,u{|}_{\Gamma }=c\ne 0.\end{array}$ (1.5)
Assume that all the isothermic surfaces ${S}_{\gamma }$  do not depend on time $t>0$  . Then $D$  is a ball.
This conjecture is stated in , p. 250. The parameter $\gamma \in \left(0,1\right]$  is chosen so that ${S}_{1}=\Gamma$  , and at $\gamma =0$  the surface ${S}_{\gamma }$  degenerates into a point. One can write:
 $\begin{array}{c}u{|}_{{S}_{\gamma }}={f}_{\gamma }\left(t\right),\end{array}$ (1.6)
where ${f}_{\gamma }$  does not depend on $x\in {S}_{\gamma }$  . If $\gamma <\nu$  , then we assume that ${D}_{\gamma }\subset {D}_{\nu }$  , where the boundary of ${D}_{\gamma }$  is ${S}_{\gamma }$  .
Another conjecture is formulated in , p.250. It reads:
”A thin elastic membrane of uniform areal density is stretched to a uniform tension and is fixed at the boundary $\Gamma$  , which is a simple closed curve. It is conjectured that if there is a solution to the wave equation ${u}_{tt}={\nabla }^{2}u$  such that $\nabla u$  vanishes along some simple closed curve $S$  within or on $\Gamma$  , then the drum (that is, the domain $D$  with boundary $\Gamma$  ) is a circular one.” The following is not clear in the above formulation: a) Does $\nabla u$  vanish on $S$  at all times or just at some moment of time?
b) What does it mean that $u$  is fixed at $\Gamma$  ? Does it mean that $u{|}_{\Gamma }=0$  or that $u{|}_{\Gamma }=\phi \left(s\right),$  where $\phi \left(s\right),s\in \Gamma$  , is a given function?
c) What are the initial conditions for $u$  ?
If the initial conditions or $\phi \left(s\right),s\in \Gamma$  , are arbitrary, then one cannot hope that $\nabla u{|}_{S}=0$  . Probably, the authors of , p.250, had in mind a solution of the type $u=sin\left(kt\right)\frac{sin\left(k|x|\right)}{|x|}$  in the three-dimensional case, for which $\nabla u=0$  on the spheres, centered at the origin, with radii which are positive roots of the equation $tan\left(kr\right)=kr$  .
This $u$  does not satisfy the zero initial condition ${u}_{t}=0$  at $t=0$  , and its value on the boundary depends on $t$  , so it is not fixed.
Let us consider the solution to the wave equation ${u}_{tt}={\nabla }^{2}u$  in $D×\left[0,\infty \right)$  , which satisfies the zero initial conditions $u={u}_{t}=0$  at $t=0$  , and the boundary condition $u{|}_{\Gamma }=c$  , where $c=const\ne 0$  .
Note that if $c=0$  , then $u=0$  , and no conclusions about the shape of the drum is possible.
One may write the solution to the above problem (with $c=1$  for definiteness), in the form:
$u=u\left(x,t\right)={\sum }_{j=1}^{\infty }\frac{1-cos\left(t\sqrt{{\lambda }_{j}}\right)}{\sqrt{{\lambda }_{j}}}{\int }_{\Gamma }{\phi }_{jN}ds{\phi }_{j}\left(x\right),$  where ${\phi }_{j}\left(x\right)$  are the eigenfunctions of the Dirichlet Laplacian in $D$  :
$\left({\nabla }^{2}+{\lambda }_{j}\right){\phi }_{j}=0\text{in}D,{\phi }_{j}{|}_{\Gamma }=0.$  Let us formulate a question W, where W stands for the wave equation.
Let
 $\begin{array}{c}{u}_{tt}={\nabla }^{2}u,\text{in}D×\left[0,\infty \right),u{|}_{t=0}=0,{u}_{t}{|}_{t=0}=0,u{|}_{\Gamma }=f\left(t\right),\end{array}$ (1.7)
where $f$  is a continuous bounded function.
Question W:
If (1.4) holds for all $t>0$  , where $S$  does not depend on $t>0$  , does it follow that $D$  is a ball ?
The following lemma will be used in Section 2:
Lemma 1. Let (1.1) holds, and $u{|}_{\Gamma }=c>0$  . Assume that the solution $u$  to (1.1) is spherically symmetric in a ball $B\left({x}_{0},a\right)\subset D,$  centered at ${x}_{0}$  and of radius $a$  . Then $D$  is a ball centered at ${x}_{0}$  . In Section 2 we prove Lemma 1 and Conjecture H. We also prove Conjecture A assuming that Conjecture P is true. We discuss question W and show that the zero initial conditions in (1.7) are not compatible with the assumption (1.4) if $S$  in (1.4) is assumed independent of time.

2 Proofs.

Proof of Lemma 1. Any solution to (1.1) is analytic in a neighborhood of any point of $D$  , where $D$  is a domain, i.e., an open set. By assumption one has:
 $\begin{array}{c}u={\sum }_{j=0}^{\infty }{c}_{j}{r}^{2j},r:=|x-{x}_{0}|,\end{array}$ (2.1)
where ${c}_{j}$  are constants and the series (2.1) converges in some disc $r<\epsilon$  . Substitute (2.1) into (1.1) and get ${c}_{j}=-\frac{{k}^{2}}{4{j}^{2}+2j}{c}_{j-1}$  . This implies that series (2.1) converges for all $r>0$  .
The condition ${\sum }_{j=0}^{\infty }{c}_{j}{r}^{2j}{|}_{\Gamma }=c$  implies that $r{|}_{\Gamma }=const$  , that is, $\Gamma$  is a spher, cenetered at ${x}_{0}$  . Lemma 1 is proved. $\square$  Remark. Equation (1.1) with ${k}^{2}<0$  and the first condition (1.2) imply that $u=0$  in $D$  . So, there is no analog of Conjecture P for ${k}^{2}<0$  .
This Remark implies, for example, that one cannot assume $c=0$  in problem (2.2) (see below).
Assuming that Conjecture P is true, we use Lemma 1 to prove that then Conjecture A is also true. In this proof we use the following statement:
If a function $u$  is analytic in a domain, containing D, and if $u\left(x\right)=u\left(r\right),r:=|x|$  in a ball $B\left({x}_{0},a\right)\subset D$  , centered at ${x}_{0}$  and of radius $a>0$  , then $D$  is a ball centered at ${x}_{0}$  .
The above statement forllows from Lemma 1.
Proof of Conjecture A. Let $\Omega \subset D$  be the domain with boundary $S$  . Then (1.1) holds in $\Omega$  , and (1.4) implies that ${u}_{N}{|}_{S}=0,u{|}_{S}=c$  , where $c=const\ne 0$  . Thus, if conjecture P is true, then $\Omega$  is a ball, and Lemma 1 implies that $D$  is a ball centered at ${x}_{0}$  . $\square$  Proof of Conjecture H. Take the Laplace transform of (1.5) and get:
 $\begin{array}{c}{\nabla }^{2}U-pU=0\text{in}D,U{|}_{\Gamma }=\frac{c}{p},U{|}_{{S}_{\gamma }}={F}_{\gamma }={F}_{\gamma }\left(p\right),\end{array}$ (2.2)
where $U:=U\left(x,p\right):={\int }_{0}^{\infty }u\left(x,t\right){e}^{-pt}dt$  . Since (by the assumption) all the isothermic surfaces ${S}_{\gamma }$  do not depend on time, and ${F}_{\gamma }$  does not depend on $x\in {S}_{\gamma }$  , one concludes from the last equation in (2.2) that
 $\begin{array}{c}{U}_{N}{|}_{{S}_{\gamma }}={G}_{\gamma }\left(p\right).\end{array}$ (2.3)
Indeed, taking values of $U$  on two surfaces ${S}_{\gamma }$  and ${S}_{\gamma +d\gamma }$  and calculating the normal derivative of $U$  , one gets ${U}_{N}{|}_{{S}_{\gamma }}=\frac{\partial {F}_{\gamma }\left(p\right)}{\partial \gamma }:={G}_{\gamma }\left(p\right)$  . From Lemma 1 and Lemma 2, stated below, it follows that $D$  is a ball. $\square$  Lemma 2. Equation ${\nabla }^{2}U-pU=0\text{in}{D}_{\gamma }$  with the boundary conditions $U{|}_{{S}_{\gamma }}={F}_{\gamma }\left(p\right)$  and ${U}_{N}{|}_{{S}_{\gamma }}={G}_{\gamma }\left(p\right)$  imply that ${S}_{\gamma }$  is a sphere.
This Lemma looks similar to Conjecture P, but the crucial difference is in the sign of the coefficients in front of the unknown function: this coefficient is ${k}^{2}>0$  in (1.1) and it is $-p<0$  in Lemma 2. Because of this difference we can prove Lemma 2, while Conjecture P is still open.
Proof of Lemma 2 is based on the arguments similar to the ones given in , pp.
306-309, and . We prove Lemma 2 at the end of this paper for convenience of the reader.
Discussion of Question W.
Taking the Laplace transform of (1.7) yields:
 $\begin{array}{c}{p}^{2}U-{\nabla }^{2}U=0\text{in}D×\left[0,\infty \right),\nabla U{|}_{S}=0,U{|}_{\Gamma }=F\left(p\right),\forall p>0.\end{array}$ (2.4)
The equation $\nabla U{|}_{S}=0,\forall p>0$  implies ${U}_{N}{|}_{S}=0,\forall p>0$  . Since $p>0$  , it follows from the differential equation (2.4) together with the boundary condition ${U}_{N}{|}_{S}=0$  that $U=0$  in $\Omega$  for all $p>0$  . This and the unique continuation principle for solutions to elliptic equations imply $U=0$  in $D$  . Therefore $u=0$  in $D$  for all $t>0$  . This is a contradiction, since $u=1$  on $\Gamma$  . Therefore the condition $\nabla U=0$  on $S$  for all $t>0$  is not compatible with the zero initial conditions for $u$  .
Proof of Lemma 2. Note that if one shows that for any unit vector $\mathbf{b}\mathbf{o}\mathbf{l}\mathbf{d}$  there is a plane $P$  , such that $D$  is symmetric with respect to $P$  , then $D$  is a ball.
Let us use the A.D.Alexandrov's method of moving plane (,). Let $\mathbf{b}\mathbf{o}\mathbf{l}\mathbf{d}$  be an arbitrary fixed unit vector and $T$  be a plane, orthogonal to $\mathbf{b}\mathbf{o}\mathbf{l}\mathbf{d}$  , and such that $D$  lies above $T$  . Let us move $T$  up, so that it is always orthogonal to $\mathbf{b}\mathbf{o}\mathbf{l}\mathbf{d}$  . At some position of $T$  it will become tangent to $\Gamma$  and then intersects $\Gamma$  . Denote by ${D}_{1}$  the part of $D$  below $T$  , reflect ${D}_{1}$  with respect to $T$  and denote by ${D}_{2}$  the resulting domain above $T$  . In the beginning, ${D}_{2}\subset D$  . Let ${\Gamma }^{\prime }$  denote the part of the boundary of ${D}_{2}$  above $T$  , and $F$  the boundary of ${D}_{2}$  which lie on $T$  . When $T$  is moving up, one of the two situations will occur:
1) ${\Gamma }^{\prime }$  touches $\Gamma$  at a point $M\in \Gamma$  , and ${D}_{2}\subset D$  , or 2) $T$  contains the normal to $\Gamma$  at a point $Q\in \Gamma$  .
Let us show that in both cases the ${D}_{2}$  has to coincide with $D\\overline{{D}_{1}}$  , so that the plane $T$  , which we denote by $P$  if either of the two situations occurs, is a symmetry plane for $D$  . Since $\mathbf{b}\mathbf{o}\mathbf{l}\mathbf{d}$  is arbitrary, one concludes that $D$  is a ball.
Thus, let us assume first that ${\Gamma }^{\prime }$  touches $\Gamma$  at a point $M$  and ${D}_{2}\subset D$  , and show that ${D}_{2}=D\\overline{{D}_{1}},$  that is, that $P$  is a symmetry plane for $D$  . Let $z$  be the coordinate along cartesian axis $\mathbf{b}\mathbf{o}\mathbf{l}\mathbf{d}$  , ${x}^{\prime }$  are cartesian coordiantes in the plane orthogonal to $\mathbf{b}\mathbf{o}\mathbf{l}\mathbf{d}$  , and $v\left({x}^{\prime },z\right):=U\left({x}^{\prime },z\right)-U\left({x}^{\prime },-z\right)$  for $\left({x}^{\prime },z\right)\in {D}_{2}$  . Then $v$  solves equation (2.2) in ${D}_{2}$  , $v=0$  on $F\subset P$  , $z=0$  on $P$  , $v<0$  on $\partial {D}_{2}\F:={\Gamma }^{\prime }$  by the maximum principle. This principle implies that $U$  attains its positive maximum on $\Gamma$  and non-negative minimum on $\Gamma$  : otherwise at a point ${x}_{0}\in D$  , at which $U$  attains its negative minimum one has ${\nabla }^{2}U\left({x}_{0}\right)\ge 0$  and $-pU\left({x}_{0}\right)>0$  , which contradicts differential equation (2.2). If $U$  would attain its positive maximum at a point ${x}_{0}\in D$  , then ${\nabla }^{2}U\left({x}_{0}\right)\le 0$  and $-pU\left({x}_{0}\right)<0$  , and again this contradicts differential equation (2.2). Thus $0\le U<\frac{c}{p}$  in the interior of $D$  . Therefore $v<0$  in the domain ${D}_{2}$  (which is an open set), and $v$  attains its maximum at the boundary of ${D}_{2}$  . At the point $M$  one has ${v}_{N}\left(M\right)>0$  by the Hopf lemma (see ), unless $v=0$  in ${D}_{2}$  , in which case ${D}_{2}=D\\overline{{D}_{1}},$  and $P$  is the symmetry plane for $D$  . On the other hand, a direct calculation, using the definition of $v$  , shows that ${v}_{N}\left(M\right)=0$  .
This proves that $P$  is the symmetry plane for $D$  .
Consider now the second situation. Let ${x}_{1}$  be the axis along the normal to $\Gamma$  at the point $Q$  , and $B\subset D$  be a ball of radius $R$  , which touches $\Gamma$  at the point $Q$  , centered at the point $O$  lying on the normal to $\Gamma$  passing through the point $Q$  . Consider a ball ${B}_{1}$  , centered at the point $Q$  of radius $0.25R$  , and let ${D}_{3}\subset {D}_{2}$  be the intersection of ${B}_{1}$  and ${D}_{2}$  . Let $w:=z\left({e}^{-a{r}^{2}}-{e}^{-a{R}^{2}}\right)$  , where $r=|x|$  in the coordinate system with the origin at $O$  . Then $w>0$  , $z>0$  , and $v<0$  in the interior of ${D}_{3}$  , so $v-w<0$  there. Let $s$  be a direction to $Q$  from the interior of ${D}_{3}$  , and $s$  is not tangential to the boundary of ${D}_{3}$  . Since $v\left(Q\right)=w\left(Q\right)=0$  , and $\psi :=v-w<0$  in the interior of ${D}_{3}$  , one has ${\psi }_{s}\ge 0$  , and if ${\psi }_{s}\left(Q\right)=0$  then ${\psi }_{ss}\ge 0$  . One checks that ${w}_{s}\left(Q\right)=0$  , ${w}_{ss}\left(Q\right)<0$  , ${v}_{j}\left(Q\right)=0$  and ${v}_{ij}\left(Q\right)=0$  , in particular, ${v}_{ss}\left(Q\right)=0$  . Thus, ${\psi }_{ss}\left(Q\right)=-{w}_{ss}\left(Q\right)<0$  . This contradiction shows that ${D}_{2}=D\\overline{{D}_{1}}$  , so that $P$  is a symmetry plane for $D$  . Thus $D$  is a ball. Lemma 2 is proved. $\square$  References

1. Problems in applied mathematics, SIAM, Philadelphia, 1990. (ed. M.Klamkin)
2. Ramm, A.G., Inverse Problems, Springer, New York, 2005.
3. Ramm, A.G., Shifrin, E., Symmetry problems in the elasticity theory problem for plane cracks of normal rapture, Prikl. Math. Mech., 69, No.1, (2005), 146-154.
4. Ramm, A.G., The Pompeiu problem, Applicable Analysis, 64, No.1-2, (1997), 19-26.
5. Serrin, J., A symmetry problem in potential theory, Archive Rational Mech. Anal., 43, (1971), 304 318.