INFINITE ERGODIC THEORY, EXERCISE SHEET #1 Exercise #1 (Measurable union) Let (X, B, m) be a probability space and C ⊂ B a sub-collection of sets. A set U = U (C) ∈ B is a measurable union of C if it satisfies • Covering property: For every C ∈ C, C ⊂ U modulo µ. • Saturation property: For every A ⊂ U, A ∈ B of positive µ-measure, there exists C 3 C ⊂ A such that µ(C) > 0. • As wa wram up convince yourself that the measurable union is unique up to a 0 measure set. A collection C is hereditary if B ⊂ C ∈ C measurable implies that B ∈ C. (1.1) Show that if C ⊂ B is hereditary then there exists C1 , C2 , ... ∈ C such that U := ∪∞ n=1 Cn is a measurable union for C. Hint: : Let 1 := sup {µ(C) : C ∈ C}, .... (1.2) Let T : (X, B, µ) a non singular transformation. Show that there exists wandering sets W1 , W2 , .... such that D(T ) = U(W(T )) = ∪∞ n=1 Wn and show that T −1 D(T ) ⊂ D(T ). Exercise #2 (Hopf Criteria for NST): Show that if (X, B, µ, T ) is a NST then (2.1) For any f ∈ L1 (X, µ), f ≥ 0 , # "∞ X µ k T̂ f = ∞ ⊂ C(T ). k=1 (2.2) For any f ∈ L1 (X, µ), f > 0 , "∞ # X µ k T̂ f = ∞ = C(T ). k=1 (2.3) Deduce the common Hopf’s criteria, If µ(X) < ∞ then "∞ # µ X k C(T )= T̂ 1 = ∞ k=1 Exercise #3. Kerngel’s positive-null decomposition: In this exercise you will show that if (X, B, µ, T ) is a NST then there exists a function h ∈ L1 (X, µ), h ≥ 0 such that T̂ h = h (h could be the zero function) and 1 INFINITE ERGODIC THEORY, EXERCISE SHEET #1 2 for any f ∈ L1 (X, µ) Ân f = n−1 1X k µ T̂ f − → 0 on [h = 0] . n k=0 ´ ´ Here T̂ is the transfer operator with respect to µ ( Tˆf gdµ = f g ◦ T dµ). The set N ull(T ) := [h = 0] is T -invariant and there exists no µ- absolutely continuous invariant probability on N ull(T ). (3.1) Show that Ân f −−−−→ 0 a.e. on D(T ) so we can assume WLOG that T is n→∞ conservative. (3.2) Let n o P := A ∈ B : ∃h ∈ L1 (X, µ), h ≥ 0, T̂ h = h, A ⊂ [h > 0] . Show that P is hereditary. If P consists only of negligible sets there is no absolutely continuous invariant measure. (3.3) Deduce that there exists hn ∈ L1 , hn ≥ 0, ´ hn dµ = 1, T̂ hn = hn and ∪∞ n=1 Bn Bn ∈ P such that Bn ⊂ [hn > 0] and = U (P). P −n (3.4) Show that setting h = n 2 hn then U(P) = [h > 0] and that for every H ∈ L1 (X, µ), H ≥ 0, if T̂ H = H then U(P)c ⊂ [H = 0]. (3.5) Deduce the result using Krengel’s stochastic ergodic Theorem. Exercise #4 (Inner functions) Let T : (R, m = Leb) → (R, m) be defined by T x = αx + β + N X k=1 pk tk − x with α, p1 , .., pn > 0 and β, t1 , ..., tn ∈ R. Recall Boole’s formula which states that T is non singular and for w = a + ib ∈ C, the functions 1 b = Im ϕw (x) = (x − a)2 + b2 x−w satisfy T̂ ϕw = ϕT w . (4.1.1) Show that T is conservative iff for some (hence for all) w ∈ C with Im(w) > 0, ∞ X ϕT n w (x) = ∞ for a.e. x ∈ R k=0 (4.1.2) Show that T is conservative iff for some (hence for all) w ∈ C with Im(w) > 0, ∞ X −1 Im = ∞. T n (w) k=0 INFINITE ERGODIC THEORY, EXERCISE SHEET #1 (4.2) Show that m ◦ T −1 = 1 αm 3 by • Show that bϕib −−−→ 1 uniformly on compact subsets. Hence for any b→∞ N A = [a, b] ⊂ R such that {tk }k=1 ∩ T −1 A = ∅, ˆ ˆ 1A ◦ T dx = lim bϕib 1A ◦ T dx b→∞ R R • Show that if T (ib) = u(b) + iv(b) with u, v : R → Rthen u(b) −−→ b − b→∞ v(b) −−→ b − b→∞ α and 0. • Deduce using Boole’s formula and the previous step that bT̂ ϕib −−−→ b→∞ 1 α uniformly on compact subsets and conclude the Theorem. (4.3) Show that if α > 1 then T is dissipative. Exercise #5: (Dissipative Bernoulli shifts and the Gurevic’-Oseledec Theorem). Let X = {0, 1}Z and T : X → X be the full shift (T x)i = xi+1 . For p 6= q, p, q ∈ (0, 1) define a product measure on X by n Y Pj (aj ) µ ([a]nk ) = j=k where (5.1) Calculate ( p, k ≥ 0, Pk (0) = 1 − Pk (1) = . q, k < 0 dµ◦T −1 dµ µ and T̂ n 1. Use this to show that (X, B, µ, T ) is totally dissipative (D(T ) = X). (5.2) Give another proof that the shift is dissipative by showing that ∀ > 0 there exists a set A ∈ B with µ (A ) > 1 − and ∞ X 1A ◦ T k < ∞, a.e x ∈ A . k=1 Hint: Think about what does the weak (or strong) law of large numbers say about P−1 P−1 1 1 N k=n xk versus n k=n T x k for N > n large. n Remark: The Gurevic’ Oseledec theorem says that a C 2 conservative hyperbolic transformation of a manifold has always a volume absolutely continuous invariant probability measure. Its proof is similar to (5.2) with use of the Markov partitions and existence of equilibrium measures developed by Sinai.