Alec Kriebel

Radial orthogonality does not imply dual 1-conformal flatness

Read the paper (PDF, 4 pages) · Download PDF · Source and verification (ZIP) · Zenodo record

Plain-English explanation · The result · Read on this page · Citation

Orthogonal wavefronts need not force flatness

Choose a point in a curved space and follow the geodesics leaving it. At each endpoint, take all directions perpendicular to the arriving geodesic. Saying that these directions are integrable means they fit together locally into hypersurfaces, much like wavefronts surrounding the starting point.

For a Riemannian metric with its Levi–Civita connection, the classical first variation of energy always supplies these hypersurfaces. The question studied here asks whether the same integrability condition forces a particular flatness property of the dual statistical connection.

The product of a unit sphere and a line gives a negative answer. Its sphere directions have curvature, while a plane mixing a sphere direction with the line has zero curvature. These two facts prevent every possible statistical 1-conformal flattening, even though the radial hypersurfaces exist around every center.

Original abstract

We give a negative answer to Kurose's Question 3(e) in the 1998 collection Open Problems in Affine Differential Geometry and Related Topics. For the product S²(1) × ℝ with its product metric and Levi–Civita connection, every radial orthogonal distribution in that question is integrable, but the dual statistical structure is nowhere locally 1-conformally flat. Integrability follows from the first variation of energy; nonflatness follows from two curvature evaluations. A conformal metric and projective connection change also give an example with nowhere zero cubic tensor. The same obstruction works in every dimension at least three. The proof is independent of computation. Historical priority is unresolved; no claim of a first resolution is made.

The precise result

Let M = S²(1) × ℝn−2, for any integer n ≥ 3, with the product of the unit round sphere metric and the Euclidean metric. Give it the Levi–Civita connection ∇. For every center p and every admitted geodesically convex neighborhood U, let γq be the unique affinely parametrized geodesic from p to q. The radial field on U ∖ {p} is

Dpq = {w ∈ TqM : hq(γ̇q(1), w) = 0}.

Every such field is integrable, but the dual statistical structure is nowhere locally 1-conformally flat. In particular, no nonempty open subset admits the required flattening. The base example is self-dual: its dual connection ∇* equals ∇, and its cubic tensor ∇h is zero.

Here a statistical structure has a positive-definite metric h, a torsion-free connection ∇, and a totally symmetric cubic tensor ∇h. Its dual is defined by Xh(Y,Z) = h(∇XY,Z) + h(Y,∇*XZ). A statistical 1-conformal change, in Kurose’s convention, is

ĥ = eφh,   ∇̂XY = ∇XY − h(X,Y) gradhφ.

The flatness condition concerns this transformed connection. It is distinct from ordinary conformal flatness of the metric. The theorem gives an explicit negative answer to Kurose’s Question 3(e), printed on page 126 of Furuhata–Matsuzoe–Urakawa’s 1998 problem collection.

Why the two properties separate

For the Levi–Civita connection, define the endpoint energy Ep(q) = ½ ∫₀¹ h(γ̇q,γ̇q) du. First variation gives dEp(w) = h(γ̇q(1),w). Away from the center this differential is nonzero, so Dp = ker dEp. Its integral hypersurfaces are the regular energy levels. This works for every center and every convex neighborhood in the question.

On the other hand, a flat statistical 1-conformal change would require an endomorphism A at each point with

R(X,Y)Z = h(Y,Z)A(X) − h(X,Z)A(Y).

Take orthonormal sphere directions e₁,e₂ and a Euclidean direction e₃. The sphere-plane equation R(e₁,e₂)e₂ = e₁ forces A(e₁) = e₁. The mixed-plane equation R(e₁,e₃)e₃ = 0 forces A(e₁) = 0. The contradiction occurs at every point and already excludes arbitrary A, before imposing the differential conditions that a conformal potential would satisfy. Additional Euclidean directions leave the argument unchanged.

An example with nowhere zero cubic tensor

The paper also removes the restriction to self-dual examples. On S²(1) × ℝ, write h₀ for the product metric and ∇⁰ for its Levi–Civita connection. Using the standard statistical (−1)-conformal deformation credited to Matsuzoe, set

h₁ = eth₀,   ∇¹XY = ∇⁰XY + dt(X)Y + dt(Y)X.

This connection is torsion-free, and its cubic tensor is symmetric with C(∂t,∂t,∂t) = −3et, so it is nonzero everywhere. The new structure is non-self-dual. The projective connection change preserves the unparametrized connecting geodesics, and the conformal metric change preserves orthogonality. Thus the radial fields agree with the original fields and remain integrable for every center on the corresponding convex neighborhoods.

Its dual is (∇¹)*XY = ∇⁰XY − h₀(X,Y)∂t. Any further statistical 1-conformal change with potential ψ becomes ∇⁰XY − h₀(X,Y) gradh₀(t + ψ). The same curvature obstruction excludes its flatness. The deformation therefore retains the negative answer while making the cubic tensor nowhere zero.

Scope and limitations

The claims concern positive-definite metrics and local statistical 1-conformal flatness. The radial field excludes its center and uses unique geodesics in an admitted convex neighborhood. No global foliation across a center or cut locus, dimension-two classification, or indefinite-metric extension is asserted. The analytic proof supplies the universal statements; the finite computations in the supporting package supplement it.

Historical priority remains unresolved. The classical first-variation mechanism, Kurose’s conformal criterion, and Matsuzoe’s deformation are credited ingredients. Relevant historical sources remain uninspected, including Kurose’s 1999 affine-realization paper, Binder–Simon’s 2000 problem list and the full content of Kurose’s 2016 talk. The paper makes no first-resolution claim and does not assert that the question remained open at publication.

Read the paper

Open PDF in a full window · Download PDF · Zenodo original PDF

The manuscript is dated 26 September 2026 and prints version 1.0.1. The record and its current files were updated on 10 October 2026, separately from that original publication date. These downloads reproduce the current published files for this record, with the manuscript PDF unchanged.

Original files

  1. SHA256SUMS.txt (170 bytes) · Zenodo original
  2. source-and-verification.zip (192,319 bytes) · Zenodo original
  3. paper.pdf (65,130 bytes) · Zenodo original

The ZIP includes the manuscript source, an exact rational checker, a separate symbolic checker, source and priority notes, and reproducibility instructions. The written proof supplies the every-center and all-dimensional conclusions.

Original manuscript, notes and page text are licensed CC BY 4.0. Attribution should name Alec Kriebel and retain the source citations and AI disclosure. Original Python verification and packaging scripts are MIT-licensed, as specified in the archive’s LICENSES.md. Cited third-party publications retain their own rights; their source PDFs are linked rather than redistributed.

Cite this paper

Kriebel, Alec. Radial orthogonality does not imply dual 1-conformal flatness. 26 September 2026. Version 1.0.1 [Preprint]. Zenodo. doi:10.5281/zenodo.22983519.

Download BibTeX

Background and prior work

Status and AI use

The candidate argument was supplied as AI-generated work. The manuscript, supporting computations, source audit and separate adversarial AI reviews were prepared with OpenAI Codex. This is an unrefereed, AI-assisted preprint; AI reviews are neither external human peer review nor proof-assistant formalization. The author is responsible for the claims.

Collaboration

Independent verification, a substantive extension, or development of a journal version would be welcome. Collaboration and authorship would be based on actual mathematical contributions.