Theorem 2.5.4 — the Hewitt-Savage 0-1 law
ProvedDurrettProbability.hewitt_savage_zero_oneLet be independent and identically distributed, realized as the coordinates of the sequence space under the infinite product of copies of one law. An event is permutable when rearranging finitely many coordinates does not change it: for every bijection of fixing all but finitely many indices, . Then
This strengthens Kolmogorov's 0-1 law in the i.i.d. case. Every tail event is permutable, since a permutation of the first coordinates cannot affect an event determined by the coordinates beyond ; but the converse fails, and the standard witness is , which is permutable — because for large — but not a tail event. So the exchangeable σ-field is strictly larger than the tail σ-field, and the theorem says that for an i.i.d. sequence both are trivial.
The consequences are the useful part: for a random walk with i.i.d. steps, statements like "" or "the walk visits infinitely often" have probability or , with no further argument.
Formalization Note The sequence space is the countable product carrying the infinite product measure, so "i.i.d." is built into the setting rather than hypothesised. A permutable event is given by its invariance property directly; the fact that such events form a σ-field is not needed for the statement and is not asserted.
import Mathlib import Definitions.Def_DurrettProbability_Series open MeasureTheory ProbabilityTheory Filter
namespace DurrettProbability
theorem hewitt_savage_zero_one {S : Type*} [MeasurableSpace S] (μ : Measure S)
[IsProbabilityMeasure μ] (A : Set (ℕ → S)) (hA : IsPermutable A) :
(Measure.infinitePi (fun _ : ℕ => μ)) A = 0
∨ (Measure.infinitePi (fun _ : ℕ => μ)) A = 1 := by sorry
end DurrettProbabilityConfirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.