In rank at least two a closed local geodesic on a sphere lifts to a great circle
DisprovedHarmonicBuilding.closedSphericalLocalGeodesicLiftsOfRankGeTwoLet be a conical Euclidean building of type and dimension with cone point , and let be a -periodic curve lying on the sphere of radius about which turns at constant angular speed :
Then there are an orthonormal pair in the model apartment , an element of the rotational Weyl group , and a positive integer with
Role. This is the building-theoretic content of Section 4, stated with no reference to harmonic maps. Rescaling by carries into the space of directions , whose metric is the angle at , so the hypothesis says exactly that the rescaled curve is a closed local geodesic of speed there --- Lemmas 4.1 and 4.2. Since is a spherical building of type , composing with the anisotropy map and reparametrizing by arclength gives a closed billiards path of minimal period , the integer counting how many times the circle wraps; and a billiards path lifts to a great circle, uniquely once its value at one parameter is prescribed, so closing up downstairs becomes equivariance under a single .
Everything after this is already machine-checked: iterating the equivariance times makes the great circle periodic, a great circle closes only at multiples of , and Lagrange gives , whence .
On the rank hypothesis. is not cosmetic. In rank one the model apartment is a line and admits no orthonormal pair, so the conclusion cannot hold; the configuration is excluded instead, and by a genuinely different argument --- in a tree the image of the circle would have to be a single point, and subharmonicity of the squared distance to the cone point then forces the map to be constant. That is the content of the companion statement for rank one, and it uses harmonicity, which the present statement does not.
import Definitions.Def_frame_2026_harmonic_building_conical import Definitions.Def_metric_geodesic_angle import Definitions.Def_spherical_great_circle
namespace HarmonicBuilding
universe v
theorem closedSphericalLocalGeodesicLiftsOfRankGeTwo
{N : ℕ} (hN : 2 ≤ N) (C : EuclideanCoxeterData N) (M : ConicalBuildingModel.{v} N C)
(gamma : ℝ → M.carrier) (alpha L delta : ℝ)
(hL : 0 < L) (halpha : 0 < alpha) (hdelta : 0 < delta)
(hsphere : ∀ theta : ℝ, dist (gamma theta) M.cone.conePoint = L)
(hper : ∀ theta : ℝ, gamma (theta + 2 * Real.pi) = gamma theta)
(hspeed : ∀ theta1 theta2 : ℝ, |theta1 - theta2| ≤ delta →
MetricGeometry.comparisonAngle M.cone.conePoint (gamma theta1) (gamma theta2)
= alpha * |theta1 - theta2|) :
∃ (v1 v2 : ModelEuclideanSpace N) (w : C.weyl) (m : ℕ),
‖v1‖ = 1 ∧ ‖v2‖ = 1 ∧ inner ℝ v1 v2 = 0 ∧ 0 < m ∧
∀ s : ℝ, SphericalGeometry.greatCirclePath v1 v2 (s + 2 * Real.pi * alpha / m)
= (w : OrthogonalGroup N) (SphericalGeometry.greatCirclePath v1 v2 s) := by sorry
end HarmonicBuilding