Polynomial edge refinement of circuit walks in dimension at least four
OpenHirsch.polynomial_edge_refinement_of_circuit_walks_dim_ge_fourThis is the high-dimensional remainder of circuit-to-edge refinement for the polynomial Hirsch conjecture.
There exist constants such that the following holds. Let and let be a bounded H-polytope cut out by an irredundant, strictly feasible family of linear inequalities. Whenever and are vertices of joined by a padded maximal row-circuit walk of length , there is a padded vertex-edge walk of from to of length
The replacement walk need not visit the original circuit intermediates, which may fail to be vertices. It need not be monotone in any linear functional, and it need not preserve every already visited facet.
In ambient dimension at most three, Klee's theorem already supplies the linear graph-diameter bound , which is a polynomial circuit-to-edge overhead. The hypothesis therefore isolates the remaining content of the unrestricted circuit-to-edge refinement statement. Constant or dimension-only overhead per circuit step is already false for polygons; the present statement still allows polynomial row-dependent overhead.
Formalization Note The Lean statement uses the same circuit-walk and irredundancy vocabulary as Hirsch.polynomial_edge_refinement_of_circuit_walks, with the additional hypothesis . Walks are padded: each step is either stationary or an edge, and the budget is an exact natural-number length.
import Definitions.Def_Hirsch_circuit_model set_option autoImplicit false open scoped RealInnerProductSpace open Hirsch
theorem Hirsch.polynomial_edge_refinement_of_circuit_walks_dim_ge_four :
∃ C k : ℕ, ∀ (d n : ℕ)
(a : Fin n → EuclideanSpace ℝ (Fin d)) (b : Fin n → ℝ),
4 ≤ d →
Bornology.IsBounded (Hirsch.Hpoly a b) →
Hirsch.RowPresentationIrredundant a b → Hirsch.StrictlyFeasibleRows a b →
∀ u ∈ Set.extremePoints ℝ (Hirsch.Hpoly a b),
∀ v ∈ Set.extremePoints ℝ (Hirsch.Hpoly a b),
∀ L : ℕ, Hirsch.RowCircuitWalk a b L u v →
∃ w : ℕ → EuclideanSpace ℝ (Fin d),
w 0 = u ∧ w (C * (n + d) ^ k * L) = v ∧
∀ j < C * (n + d) ^ k * L,
w j = w (j + 1) ∨
Hirsch.Adj (Hirsch.Hpoly a b) (w j) (w (j + 1)) := by sorry