Theorem 1.6.1 — the discrete pointwise ergodic theorem
ProvedPalmQueueing.Ergodic.discrete_pointwiseTheorem 1.6.1 (Discrete pointwise ergodic theorem). Let be a probability space and be a discrete ergodic flow on this space. For all ,
This is Birkhoff's theorem, which the book quotes rather than proves — though the proof it points to is its own: "A proof of this theorem based on Loynes' construction can be found in §2.2.5, Chapter 2", which is Theorem 2.2.1 of this series' mission II.
The in the display is part of the statement: the limit is asserted to exist almost surely and to equal , not merely to equal it wherever it happens to exist.
It is worth stating because the substrate is absent everywhere. Mathlib has the mean (von
Neumann, ) ergodic theorem in Analysis/InnerProductSpace/MeanErgodic but no pointwise one,
and the platform has neither. The attribution to Birkhoff belongs in the prose; the source cites
Baccelli's page, because that is where this mission read it.
Formalization Note. A discrete flow is, by the book's definition on p.46, "a bijective and measurable map from to itself, which preserves "; the statement carries Function.Bijective θ alongside Mathlib's Ergodic θ P0 (measure preserving and pre-ergodic).
import Mathlib import Definitions.Def_PalmQueueing_Ergodic_DiscreteFlow
/-!
# Theorem 1.6.1: the discrete pointwise ergodic theorem (§1.6.1, p.47)
-/
namespace PalmQueueing.Ergodic
open MeasureTheory Filter Topology
variable {Ω : Type*} [MeasurableSpace Ω]
/-- **Theorem 1.6.1 (Discrete pointwise ergodic theorem)** (p.47). Let `(Ω, F, P⁰)` be a
probability space and `θ` be a discrete ergodic flow on this space. For all `f ∈ L¹(P⁰)`,
`(1.6.1) ∃ lim_{N→∞} (1/N) Σ_{n=1}^{N} f ∘ θⁿ = E⁰[f], P⁰-a.s.`
**This is Birkhoff's theorem, which the book quotes rather than proves.** The proof it points to is
its own: "A proof of this theorem based on Loynes' construction can be found in §2.2.5, Chapter 2"
— which is Theorem 2.2.1 of mission `02a`, the queueing proof of the ergodic theorem.
The `∃` in the display is part of the statement: the limit is asserted to **exist** almost surely
and to equal `E⁰[f]`, not merely to equal it where it exists.
It is stated here because the substrate is absent everywhere: Mathlib has the *mean* (von Neumann,
`L²`) ergodic theorem in `Analysis/InnerProductSpace/MeanErgodic` but no pointwise one, and the
platform has neither. The `source` cites Baccelli's page, since that is where this mission read
it; the attribution to Birkhoff belongs in the prose.
`hbij` is the book's definition of a discrete flow (p.46): "a bijective and measurable map from
`Ω` to itself, which preserves `P⁰`". -/
theorem discrete_pointwise (P0 : Measure Ω) [IsProbabilityMeasure P0]
(θ : Ω → Ω) (hθ : Measurable θ) (hbij : Function.Bijective θ) (herg : Ergodic θ P0)
(f : Ω → ℝ) (hf : Integrable f P0) :
∀ᵐ ω ∂P0, Tendsto
(fun N : ℕ => (∑ n ∈ Finset.Icc 1 N, f (θ^[n] ω)) / (N : ℝ))
atTop (𝓝 (∫ ω, f ω ∂P0)) := by sorry
end PalmQueueing.Ergodic
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.