Documentation

Dynamics.Absorption

Geometric convergence from a uniform absorption block #

The access hypothesis of finite_absorption (∀ a, ∃ n, K.iterate n f a < 1) follows from reachability: exists_iterate_lt_one_of_reachable turns a chain of transitions of positive probability ending where f vanishes into a finite time at which f is below one in expectation.

theorem Dynamics.Kernel.geometric_blocks {α : Type u_1} [Fintype α] (K : Kernel α) (f : α → ℝ) (q : ℝ) (hq : 0 ≤ q) (m : ℕ) (hblock : ∀ (a : α), K.iterate m f a ≤ q * f a) (n : ℕ) (a : α) :
K.iterate (n * m) f a ≤ q ^ n * f a

If a nonnegative survival observable contracts every m rounds, its expectation has a geometric bound at multiples of m.

theorem Dynamics.Kernel.geometric_blocks_tendsto {α : Type u_1} [Fintype α] (K : Kernel α) (f : α → ℝ) (hf : ∀ (a : α), 0 ≤ f a) (q : ℝ) (hq : 0 ≤ q) (hq1 : q < 1) (m : ℕ) (hblock : ∀ (a : α), K.iterate m f a ≤ q * f a) (a : α) :
Filter.Tendsto (fun (n : ℕ) => K.iterate (n * m) f a) Filter.atTop (nhds 0)

The block bound tends to zero whenever the contraction factor is below one.

theorem Dynamics.Kernel.iterate_antitone {α : Type u_1} [Fintype α] (K : Kernel α) (f : α → ℝ) (h : ∀ (a : α), K.apply f a ≤ f a) (a : α) :
Antitone fun (n : ℕ) => K.iterate n f a

An observable satisfying K f ≤ f has antitone finite-time expectations.

theorem Dynamics.Kernel.iterate_succ_lt_one {α : Type u_1} [Fintype α] (K : Kernel α) {f : α → ℝ} (hf : ∀ (a : α), f a ≤ 1) {s t : α} (hst : 0 < (K s).weight t) {n : ℕ} (ht : K.iterate n f t < 1) :
K.iterate (n + 1) f s < 1

One transition of positive probability, from s to a state t where an observable bounded by 1 is below one in expectation after n steps, puts it below one after n + 1 steps from s.

theorem Dynamics.Kernel.exists_iterate_lt_one_of_reachable {α : Type u_1} [Fintype α] (K : Kernel α) {f : α → ℝ} (hf : ∀ (a : α), f a ≤ 1) {s t₀ : α} (h : Relation.ReflTransGen (fun (s t : α) => 0 < (K s).weight t) s t₀) (h₀ : f t₀ = 0) :
∃ (n : ℕ), K.iterate n f s < 1

Reachability gives access. If t₀ is reachable from s along transitions of positive probability and an observable f ≤ 1 vanishes at t₀, then the expectation of f drops below one at some finite time from s. This is the access hypothesis of finite_absorption for a survival indicator f.

theorem Dynamics.Kernel.exists_uniform_block {α : Type u_1} [Fintype α] [Nonempty α] (K : Kernel α) (f : α → ℝ) (hf : ∀ (a : α), f a = 0 ∨ f a = 1) (hstep : ∀ (a : α), K.apply f a ≤ f a) (haccess : ∀ (a : α), ∃ (n : ℕ), K.iterate n f a < 1) :
∃ (m : ℕ), 0 < m ∧ ∃ (q : ℝ), 0 ≤ q ∧ q < 1 ∧ ∀ (a : α), K.iterate m f a ≤ q * f a

Finiteness turns state-dependent access into a uniform contraction block.

theorem Dynamics.Kernel.finite_absorption {α : Type u_1} [Fintype α] [Nonempty α] (K : Kernel α) (f : α → ℝ) (hf : ∀ (a : α), f a = 0 ∨ f a = 1) (hstep : ∀ (a : α), K.apply f a ≤ f a) (haccess : ∀ (a : α), ∃ (n : ℕ), K.iterate n f a < 1) (a : α) :
Filter.Tendsto (fun (n : ℕ) => K.iterate n f a) Filter.atTop (nhds 0)

An accessible absorbing target in a finite chain has vanishing survival probability. f is its complement indicator; hstep expresses absorption and haccess access.