Polynomial access to a supporting face of the target vertex (conjectural)
OpenHirsch.polynomial_target_face_accessIs there a polynomial budget for reaching a supporting face containing a target vertex, without requiring that the target vertex itself be reached?
Precisely, there exist natural numbers , independent of all data, with the following property. Let
be bounded, and let be extreme points of . Suppose no row with nonzero normal is tight at both endpoints: for every with , at least one of
holds. Then there are a row , an extreme point of , and a vertex-edge walk from to such that
Only access to one nonzero supporting row of the target is required; need not equal , and the walk need not be monotone or non-revisiting. This is an explicitly conjectural local-routing subproblem for the polynomial Hirsch conjecture, not a known polynomial estimate. The conditional theorem Hirsch.diameter_bound_of_target_face_access gives its role in the decomposition: uniform access bounds imply diameter bounds with one additional polynomial degree. No values of are claimed here.
Formalization Note Ambient dimension and row count may be zero; distinct extreme points exclude the zero-dimensional case. Redundant rows, zero normals, and lower-dimensional polytopes are allowed. A row with zero normal cannot satisfy the conclusion. The supporting section need not be a facet. Walks allow stationary steps and are represented with exactly steps.
import Definitions.Def_Hirsch_model open scoped RealInnerProductSpace
namespace Hirsch
theorem polynomial_target_face_access :
∃ C k : ℕ, ∀ (d n : ℕ)
(a : Fin n → EuclideanSpace ℝ (Fin d)) (b : Fin n → ℝ),
Bornology.IsBounded (Hpoly a b) →
∀ u ∈ Set.extremePoints ℝ (Hpoly a b),
∀ v ∈ Set.extremePoints ℝ (Hpoly a b), u ≠ v →
(∀ i, a i ≠ 0 → ⟪a i, u⟫ ≠ b i ∨ ⟪a i, v⟫ ≠ b i) →
∃ (i : Fin n) (z : EuclideanSpace ℝ (Fin d)),
a i ≠ 0 ∧ ⟪a i, v⟫ = b i ∧
z ∈ Set.extremePoints ℝ (Hpoly a b) ∧ ⟪a i, z⟫ = b i ∧
∃ w : ℕ → EuclideanSpace ℝ (Fin d),
w 0 = u ∧ w (C * (n + d) ^ k) = z ∧
∀ j < C * (n + d) ^ k,
w j = w (j + 1) ∨ Adj (Hpoly a b) (w j) (w (j + 1)) := by sorry
end Hirsch