Carathéodory conjecture

About 102 years old · traced to

Let α∈(0,1)\alpha\in(0,1) and let S⊂R3S\subset\mathbb{R}^{3} be a compact, connected, embedded surface without boundary of class C3,αC^{3,\alpha} which is convex, i.e. SS bounds a convex body in R3\mathbb{R}^{3}. Let ν ⁣:S→S2\nu\colon S\to\mathbb{S}^{2} be a unit normal vector field on SS, and for p∈Sp\in S let Ap ⁣:TpS→TpSA_{p}\colon T_{p}S\to T_{p}S denote the associated shape operator, whose eigenvalues κ1(p)\kappa_{1}(p) and κ2(p)\kappa_{2}(p) are the principal curvatures of SS at pp. A point p∈Sp\in S is called umbilic if

κ1(p)=κ2(p),\kappa_{1}(p)=\kappa_{2}(p),

equivalently if ApA_{p} is a multiple of the identity on TpST_{p}S. Then

#{p∈S: p is umbilic} ≥ 2.\#\{p\in S:\ p \text{ is umbilic}\}\ \geq\ 2 .
References

Primary source

Wikipedia

Additional references

  1. Wikipedia, Carathéodory conjecture, the article this problem comes from.

Progress summary

Refreshed
Claimed solved

A published proof covers the stated case, but a new AI-assisted claim of a counterexample has not yet been independently checked.

The conjecture asserts that every sufficiently smooth convex closed surface in three-dimensional Euclidean space has at least two umbilic points. It is attributed to Constantin Carathéodory, with an early formulation by Hans Hamburger in 1924.

Known results

  • Guilfoyle and Klingenberg (2008) report a proof for C3+αC^{3+\alpha} strictly convex surfaces, via complex points of Lagrangian sections of TS2TS^{2}.
  • An index-bound paper (2012) states that isolated umbilic indices are less than 22 and presents the C3+αC^{3+\alpha} case as resolved.
  • A 2025 exposition presents the same Lagrangian-section argument as proving the conjecture.

August 2026 AI-assisted counterexample claim

A circulating write-up, prepared with help from Claude, claims to disprove the conjecture by exploiting a distinction between smooth and analytic surfaces. No proof or counterexample has yet been independently verified in the retrieved sources, so this challenges but does not overturn the published result.

Current status (as of August 2026): The C3,αC^{3,\alpha} Euclidean case has a published proof, while a newly circulated AI-assisted counterexample claim remains unverified.

Sources

Solutions 0

No solutions have been posted yet.