When every geodesic minimizes for the same distance

← All digests

Family 344 · Differential geometry
Importance 5/5 · Provisional editorial rating
Sources consulted 2026-10-26

1. The question

On a round sphere, a traveler following a great circle takes a shortest route until reaching the antipode. Every starting point and every direction have the same cutoff distance: half a great circle, equal to the sphere’s diameter.

Could a different Riemannian geometry have this same property? The metric Blaschke conjecture says that the possibilities are exactly the standard compact rank-one symmetric spaces: round spheres, real, complex, and quaternionic projective spaces, and the Cayley plane, with arbitrary overall scale.

Family 344 claims the full metric classification. Its hypothesis compares two numbers: the global injectivity radius and the diameter. Their equality forces not merely the topology of a model but its exact Riemannian metric up to scale.

This is a strong distinction. Many geometries can share a topological space, and periodic geodesics alone do not determine a round metric. The crucial extra requirement is that every geodesic remain minimizing all the way to the common diameter.

2. Background and essential definitions

Diameter and the loss of shortest paths

For a Riemannian manifold $(M,g)$, let $d_g(p,q)$ be the shortest-path distance. Its diameter is

$$\operatorname{diam}(M,g)=\sup_{p,q\in M}d_g(p,q).$$

At a point $p$, the exponential map $\exp_p$ sends a tangent vector $v$ to the point reached after following the geodesic with that initial velocity for time one. The pointwise injectivity radius $\operatorname{inj}_p$ is the largest radius on whose open tangent ball this map is a diffeomorphism onto its image.

Inside that radius, radial geodesics give unique shortest paths from $p$. At the cut locus, uniqueness or the local regularity of this description fails. The global injectivity radius is

$$\operatorname{inj}(M,g)=\inf_{p\in M}\operatorname{inj}_p(M,g).$$

The manuscript calls a connected closed positive-dimensional manifold Blaschke if

$$\operatorname{inj}(M,g)=\operatorname{diam}(M,g)=D.$$

Closed means compact without boundary. The equality implies that every unit-speed geodesic minimizes through length $D$, is uniquely minimizing for lengths less than $D$, and cannot minimize beyond $D$. At exactly $D$, uniqueness is not asserted.

The model spaces

Real projective space parametrizes real lines through the origin; complex and quaternionic projective spaces use lines over the corresponding number systems. Their standard metrics are highly symmetric. The Cayley plane is an exceptional sixteen-dimensional analogue associated with the octonions.

Together with spheres, these form the compact rank-one symmetric spaces. “Symmetric” means that geodesic reflection at each point is a local isometry. The projective metrics need not have constant sectional curvature: in the usual normalization, complex and quaternionic projective spaces have tangent-plane curvatures ranging between two values.

An isometry preserves all distances and the metric. A diffeomorphism preserves only the smooth manifold structure. The claimed theorem provides the stronger conclusion. Multiplying the metric by a positive constant is allowed because it rescales both injectivity radius and diameter by the same factor.

3. Development history and previous partial results

Blaschke’s original question arose from surfaces whose geodesics reconverge in a uniform way. Green settled the surface problem in 1963. The subsequent theory, developed in Besse’s book and summarized in McKay’s corrected survey, related the minimizing condition to smooth cut loci, closed geodesics, and fibrations of tangent spheres by great spheres. Bott and Samelson’s topological restrictions explain why the familiar symmetric models appear. Green, McKay’s corrected survey.

The metric sphere and real-projective cases were established through work of Berger, Kazdan, Weinstein, and Yang. Volume comparisons played a central role. Berger also proved a sharp lower bound for volume in terms of injectivity radius, using weighted estimates for Jacobi fields and a scalar inequality of Kazdan. Berger’s isoembolic inequality.

For the projective models, different partial conclusions needed careful separation. Knowing the cohomology ring, knowing the total volume at a fixed geodesic period, and knowing the metric are three different achievements. Yang established the model-volume statement for Blaschke manifolds of complex-projective homotopy type. Smooth incidence geometry and classification results identified the smooth type in projective-plane cases. Other equal-volume results required additional topological hypotheses. Yang.

There were also metric rigidity results under extra curvature assumptions or agreement with a model on part of the manifold. A September 2026 preprint by Song proves rigidity assuming an adapted complex structure on the entire tangent bundle. That entire-extension condition is additional to being Blaschke. Song.

The new manuscript claims to remove such extra geometric hypotheses. It combines a sharp lower volume bound with a matching upper bound, developing an intrinsic argument for all higher quaternionic dimensions rather than assuming the topology needed by a pre-existing equal-volume theorem.

4. What the new manuscript claims

Every connected closed smooth Riemannian manifold of positive dimension satisfying $\operatorname{inj}=\operatorname{diam}$ is, up to multiplying its metric by a positive constant, isometric to one of the standard models:

• A round sphere, including the circle.
• Real, complex, or quaternionic projective space.
• The Cayley plane.

All quaternionic projective dimensions are included. There is no assumed sectional-curvature bound, no hypothesis that the manifold is already diffeomorphic to the model, and no adapted-complex-structure assumption.

The paper states an intermediate comparison. Classical Blaschke structure assigns a symmetric model $(M_0,g_0)$ of the appropriate type. Scale it to have the same diameter as $(M,g)$. Then

$$\operatorname{Vol}(M,g)\geq\operatorname{Vol}(M_0,g_0),$$

with equality if and only if the two metrics are isometric.

The remaining argument supplies the opposite inequality in each model type, using classical inputs where their hypotheses apply and a new quaternionic calculation where needed. Equality then triggers metric rigidity.

Under the classical structure theory used in the paper, every unit geodesic is simple and closed with prime period $2D$. “Prime” means the first return period, not a repeated traversal. The volume comparisons therefore use an equivalent normalization whether they fix the diameter or the prime period.

5. Why the advance matters

Importance: 5/5. If correct, the theorem resolves a classical global rigidity problem by characterizing a short list of exact metrics through a common minimizing-distance condition.

The claim is stronger than a topological classification. It rules out every nonstandard deformation of the metric that might preserve the global equality. It also covers the quaternionic families uniformly, where topology, characteristic classes, and geodesic geometry interact in a particularly intricate way.

The proposed lower-bound argument is notable because it extracts the necessary comparison from the cut-time structure without assuming the model’s parallel curvature eigenspaces. A proof that simply imposed that geometric splitting would have assumed much of the desired rigidity.

The rating concerns these stated consequences and methods. It does not constitute an independent assessment of the proof.

6. The main idea

Read the geometry of shortest-path failure

For each point $p$, send a unit tangent direction $v$ to $\exp_p(Dv)$, its cut point. Classical Blaschke theory gives a smooth sphere fibration: several directions may reach the same cut point, and each fiber is a great sphere in the tangent sphere.

The fiber dimension identifies the model type. In the intermediate projective cases, the normal directions along a geodesic split numerically into two groups: those associated with Jacobi fields vanishing again at the cut time, and the remaining directions. This is information about vanishing and dimensions, not yet a parallel splitting of curvature.

A Jacobi field describes the infinitesimal separation of neighboring geodesics. Determinants built from such fields measure how volume expands in polar coordinates. They are the analytic bridge from geodesic behavior to total volume.

Obtain a sharp lower volume bound

Normalize $D=\pi/2$, so the prime period is $\pi$. In a standard projective model, the radial volume density has the form

$$w(t)=(\sin t)^m\left(\frac{\sin 2t}{2}\right)^k, \qquad 0<t<D.$$

Here $k$ is the dimension of the great-sphere fiber at the cut locus, and $m=n-1-k$ for real dimension $n$. For complex projective space $k=1$, for quaternionic projective space $k=3$, and for the Cayley plane $k=7$.

The new comparison represents vanishing Jacobi fields as Lagrangian planes in a symplectic vector space. A fixed matrix chart separates an endpoint determinant into a regular block and a Schur-complement block. A Schur complement is the matrix remaining after one block has been algebraically eliminated; it isolates the part sensitive to the cut-time intersection.

Both blocks are estimated by scalar interval integrals. A logarithmic version of Kazdan’s weighted cancellation combines them over the closed geodesic. Reflection symmetry cancels boundary contributions, while carefully matched half-period weights retain the right multiplicities.

A nonnegative strict-convexity remainder keeps track of equality. Integrating over all unit tangent directions gives the model lower bound. If equality holds, the radial volume densities agree with the model everywhere. The paper then invokes Szabó’s classification of compact simply connected harmonic manifolds to identify the metric. Harmonicity here means that polar volume density depends only on radius, not direction.

Obtain the quaternionic upper bound from integral topology

The opposite bound has a different source. For quaternionic type in dimension $4r$, consider the incidence space of pairs of points separated by the cut distance. Its normal four-plane bundle records the great-three-sphere families of directions.

The nonintersection pattern of distinct direction planes constrains a Pontryagin class of the tangent bundle. Pontryagin classes are cohomological invariants of real vector bundles. The signature theorem relates their combinations to the signed intersection form, and the manuscript uses this relation to determine a coefficient in every quaternionic rank.

Next pass to the space of oriented geodesics, forgetting their marked starting points. The unit tangent bundle becomes a circle bundle over this space. Its Euler class records how those circles twist, and the contact volume expresses the manifold’s volume through a top intersection power of that class.

Integral Gysin sequences and closed-geodesic Morse theory connect this intersection number to the earlier incidence coefficient. The calculation retains primitive integral information, including checks over prime fields. Rational cohomology alone would lose the lattice normalization needed for the sharp numerical bound.

Finally, an algebraic intersection calculation gives volume at most that of quaternionic projective space. Combined with the lower bound, this forces equality. The other model types are completed using the appropriate classical volume and topological inputs, with their assumptions checked in the manuscript.

7. Scope, limitations, and open questions

The hypothesis is the exact global equality of injectivity radius and diameter. The theorem does not classify manifolds where those quantities are merely close, where the equality holds at only one point, or where all geodesics are closed without the minimizing condition.

No claim is made about a quantitative stability estimate. To conclude that near-equality forces closeness to a model, one would need to specify the notion of closeness and probably additional compactness assumptions.

The conclusion includes projective spaces with nonconstant sectional curvature, so “every Blaschke manifold is round” would be misleading. Round spheres are one part of the classification, not its entirety.

The manuscript relies on established Blaschke structure theory and harmonic-manifold rigidity, then supplies its claimed determinant comparison and quaternionic upper bound. This digest reports those dependencies and conclusions without auditing the arguments or independently establishing the claimed resolution.

8. Sources and further reading

McKay’s corrected survey provides an accessible map of the historical distinctions between topology, volume, and metric rigidity. The primary papers linked above give representative milestones. Optional background appears under cut locus, Jacobi field, and symmetric space.

Source manuscripts

• M1. OpenAI, The metric Blaschke theorem (September 23, 2026).

Sources consulted on 26 October 2026 at repository revision 1581d4c. Download the manuscript citations.