Theorem 5 (Furstenberg–Kesten) — n⁻¹ log [Y₁⋯Y_n]_ij converges with probability one and in mean to a finite λ independent of i, j
ProvedKingmanSubadditive.PositiveMatrices.theorem_5furstenberg-kestenlyapunov-exponentp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1random-matricessubadditive-ergodic-theorem
Let be random matrices () on a probability space, and let
Suppose that
- the elements of the matrices are strictly positive;
- their logarithms have finite expectations: for all and all ;
- the sequence is stationary (its joint law is invariant under the shift ).
Then there is a single finite random variable with such that for every the limit
exists with probability one and in mean (); in particular it does not depend on or .
This is (a variant of) the Furstenberg–Kesten theorem on products of random matrices; Kingman derives it as a corollary of his subadditive ergodic theorem. The quantity is the top Lyapunov exponent of the product.
Formalization Note. One random variable serves every entry. The logarithm is applied only to strictly positive numbers because positivity of the is a hypothesis; the paper's index is 0 : Fin k.
Preamble
import Mathlib import Definitions.Def_KingmanSubadditive_PositiveMatrices_Model open MeasureTheory Filter Topology
Formal statement
namespace KingmanSubadditive.PositiveMatrices
/-- **Theorem 5** (Furstenberg–Kesten), p. 891 (Kingman, *Subadditive ergodic theory*, Ann.
Probab. 1(6):883–899 (1973)). Suppose that the elements of the random `k × k` matrices `Y_n` are
strictly positive, that their logarithms have finite expectations, and that the sequence
`(Y_n)` is stationary. Then the finite limit `λ = lim_{n→∞} n⁻¹ log [X_n]_ij`, (2.2.2), with
`X_n = Y₁ Y₂ ⋯ Y_n`, exists with probability one and in mean, and does not depend on `i` or `j`.
**Formalization Note.** One random variable `lam` (`λ` is a Lean keyword) serves every
`(i, j)`: that is "does not depend on i or j". "Finite" is `lam : Ω → ℝ`; "in mean" is
`E|n⁻¹ log [X_n]_ij − λ| → 0` with `λ` integrable. `Real.log` is the genuine logarithm because
every entry of `X_n`, `n ≥ 1`, is strictly positive. Stationarity is joint-law stationarity of
`(Y_n)_{n≥1}`, positivity holds for every outcome, `k ≥ 1`. -/
theorem theorem_5 {Ω : Type*} [MeasurableSpace Ω] (P : Measure Ω) [IsProbabilityMeasure P]
{k : ℕ} [NeZero k] (Y : ℕ → Ω → Matrix (Fin k) (Fin k) ℝ) (hY : Hypotheses P Y) :
∃ lam : Ω → ℝ, Integrable lam P ∧ ∀ i j : Fin k,
(∀ᵐ ω ∂P, Tendsto (fun n : ℕ => Real.log (X Y n ω i j) / n) atTop (𝓝 (lam ω))) ∧
Tendsto (fun n : ℕ => ∫ ω, |Real.log (X Y n ω i j) / n - lam ω| ∂P) atTop (𝓝 0) := by sorry
end KingmanSubadditive.PositiveMatrices
Source
Kingman, Subadditive ergodic theory, Ann. Probab. 1(6):883–899 (1973), DOI 10.1214/aop/1176996798, p. 891, Theorem 5
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.