Theorem 2.4 — facial/graphical irreducibility and indecomposability of m-topical maps
ProvedTopicalIrreducibility.Relations.irreducibility_relationsFix a finite dimension . Let be an m-topical map (continuous, order-preserving, homogeneous, with ). Then:
- if is facially irreducible or graphically irreducible, then is indecomposable;
- if is m-convex and indecomposable, then is graphically irreducible;
- if is subadditive and indecomposable, then is facially irreducible.
In symbols,
Here , and abbreviate facial irreducibility, graphical irreducibility and indecomposability, respectively. Facial irreducibility means no nontrivial proper closed face of the cone is invariant; graphical irreducibility means the Gaubert–Gunawardena graph , with an arc when as , is strongly connected; The vector has every entry equal to , and is the indicator vector of . Indecomposability means that for every nonempty proper some coordinate has .
The theorem organizes the irreducibility notions for nonlinear Perron–Frobenius theory: indecomposability is the weakest, it coincides with graphical irreducibility for m-convex maps, and all three coincide for subadditive maps (which are m-convex by Lemma 1.4).
Formalization Note The same arbitrary m-topical appears in all three clauses; m-convexity and subadditivity are premises of clauses 2 and 3 only. The dimension is arbitrary, including and , where there is no nonempty proper subset of and every clause holds.
Differences from the source:
- generalization: the dimension-zero case is not discussed → every n : Nat, including n = 0; reason: there are no nonempty proper coordinate subsets or vertices at n = 0, so all three predicates hold; every positive-dimensional source instance is retained.
- encoding: real nonnegative coordinate vectors indexed by 1,...,n, and subsets of these indices → Fin n → ℝ≥0 with its product topology and pointwise operations; subsets are Finset (Fin n); reason: the coordinates are relabeled from 1-based to 0-based; the type represents the entire closed real cone, and positive NNReal scalars represent all positive real scalars.
- encoding: the limit of each outside coordinate ray is finite → for each outside coordinate i there exists B : ℝ≥0 bounding f(1 + t e_J)_i for every t : ℝ≥0; reason: order preservation makes each full nonnegative-real ray monotone, so boundedness is equivalent to convergence to a finite limit; restricting t to the nonnegative reals changes no limit at positive infinity.
- encoding: the directed graph is strongly connected → every ordered pair is related by Relation.ReflTransGen of the source arc relation; reason: this is directed reachability by possibly empty paths, preserving the usual singleton-vertex convention and the source arc orientation.
import Mathlib import Definitions.Def_TopicalIrreducibility_Relations_MTopical import Definitions.Def_TopicalIrreducibility_Relations_Irreducibility open scoped NNReal
namespace TopicalIrreducibility.Relations
/-- Theorem 2.4 (Lins, p. 4). Let `f : ℝⁿ_{≥0} → ℝⁿ_{≥0}` be m-topical.
(1) facially or graphically irreducible ⇒ indecomposable;
(2) m-convex and indecomposable ⇒ graphically irreducible;
(3) subadditive and indecomposable ⇒ facially irreducible. -/
theorem irreducibility_relations {n : ℕ} (f : StdCone n → StdCone n) (hf : IsMTopical f) :
((FaciallyIrreducible f ∨ GraphicallyIrreducible f) → Indecomposable f) ∧
(IsMConvex f → Indecomposable f → GraphicallyIrreducible f) ∧
(IsSubadditive f → Indecomposable f → FaciallyIrreducible f) := by sorry
end TopicalIrreducibility.Relations
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.