Harris ergodicity gives -irreducibility from every point
ProvedMarkovChainCLT.exists_iterKernel_pos_of_harrisErgodicLet be a Markov kernel on a state space with invariant probability distribution , Harris ergodic in the mission's total-variation encoding: for every starting point . Then the chain is -irreducible from every point: for every measurable set with and every there is an with
This is -irreducibility with (Meyn and Tweedie 1993, Section 4.2), one of the standing hypotheses under which the drift and minorization theory of the source is stated. It follows from the pointwise convergence given by the total-variation convergence from every starting point.
Formalization Note The conclusion is stated for every , not just -almost every , because the mission's HarrisErgodic quantifies the total-variation convergence over every starting point; this is what makes the classical hypothesis available everywhere.
import Definitions.Def_MarkovErgodicity open MeasureTheory ProbabilityTheory Filter open scoped ENNReal NNReal Topology ProbabilityTheory
theorem MarkovChainCLT.exists_iterKernel_pos_of_harrisErgodic {X : Type*}
[MeasurableSpace X] (P : Kernel X X) [IsMarkovKernel P] (π : Measure X)
[IsProbabilityMeasure π] (hP : HarrisErgodic P π) (A : Set X) (hA : MeasurableSet A)
(hπA : 0 < π A) (x : X) :
∃ n : ℕ, 0 < (iterKernel P n) x A := by sorry