Ergodic measures are extreme invariant measures
ProvedFamousTheorems.ergodic_iff_extreme_invariantErgodic measures are the extreme invariant measures. Let be measurable and an -invariant probability measure. Then is ergodic for if and only if is an extreme point of the convex set of -invariant probability measures.
An ergodic measure cannot be written as a nontrivial convex combination of other invariant measures. This is the basis of the ergodic decomposition, which writes every invariant measure as an integral of ergodic ones. It is used for existence of ergodic measures via the Krein–Milman theorem.
Formalization note. Mathlib's Ergodic.iff_mem_extremePoints. The convex set is {ν | MeasurePreserving f ν ν ∧ IsProbabilityMeasure ν}, and extreme points are taken for convex combinations with coefficients in (Set.extremePoints ENNReal), which agrees with the usual notion on probability measures. Ergodic f μ includes the invariance of .
import Mathlib
namespace FamousTheorems
open MeasureTheory
theorem ergodic_iff_extreme_invariant {X : Type*} {m : MeasurableSpace X} {μ : Measure X} {f : X → X} [IsProbabilityMeasure μ] :
Ergodic f μ ↔
μ ∈ Set.extremePoints ENNReal {ν : Measure X | MeasurePreserving f ν ν ∧ IsProbabilityMeasure ν} := by sorry
end FamousTheorems