Track new results across pure and applied mathematics.
Search papers, labs, and topics across Lattice.
The authors construct the first bounded, convex, non-spherical domains with real-analytic boundaries in dimensions 3, 4, 6, 8, 10, and 14 that admit nonconstant solutions to the overdetermined Helmholtz problem $\Delta u + u = 0$ with $u=1$ and $\nabla u = 0$ on the boundary. Utilizing conformal mappings in 3D and rank-two compact Lie algebra representations in higher dimensions, the boundary value problems are reduced to planar systems and certified via rigorous computer-assisted contraction mappings. This decisively refutes both the Schiffer and Pompeiu conjectures even within the class of convex domains, dismantling long-held assumptions regarding the spectral rigidity of Euclidean balls.
Decades-old conjectures that Euclidean balls are uniquely characterized by overdetermined spectral PDE data fail even under strict convexity, with non-spherical, real-analytic counterexamples established in dimensions up to 14.
In dimensions $3,4,6,8,10,14$ we construct bounded convex non-ball domains with real-analytic spherical boundaries. Each admits a nonconstant solution of $Δu+u=0$ with $u=1$ and $\nabla u=0$ on the boundary. Green's identity makes their indicator Fourier transforms vanish on the unit sphere, disproving the Schiffer and Pompeiu conjectures even within the convex class in these dimensions. The three-dimensional domain is axisymmetric, invariant under $O(2)$ on the last two coordinates and reflection of the first, and is obtained by lifting a conformal map of its meridian section. In dimensions $n=2m+2$, $m=1,2,3,4,6$, the domains lie in the rank-two compact Lie algebras $\mathfrak{u}(2)$, $\mathfrak{so}(4)$, $\mathfrak{su}(3)$, $\mathfrak{so}(5)$ and $\mathfrak g_2$. Harish-Chandra's radial part formula reduces these cases to planar Helmholtz problems, and Kostant's theorem reduces convexity to the Cartan section. Exact solutions near dyadic approximations follow from computer-assisted contractions in weighted polynomial coefficient spaces, with interval bounds for finite blocks and analytic bounds for every infinite tail. A second, strictly star-shaped but nonconvex, example in dimension three is retained as an independent result. To our knowledge, apart from the earlier version of this work, these are the first counterexamples in dimensions at least three and the first known convex counterexamples in any dimension.