4.II.25J

Probability and Measure
Part II, 2007

Let (E,E,μ)(E, \mathcal{E}, \mu) be a measure space with μ(E)<\mu(E)<\infty and let θ:EE\theta: E \rightarrow E be measurable.

(a) Define an invariant set AEA \in \mathcal{E} and an invariant function f:ERf: E \rightarrow \mathbb{R}.

What is meant by saying that θ\theta is measure-preserving?

What is meant by saying that θ\theta is ergodic?

(b) Which of the following functions θ1\theta_{1} to θ4\theta_{4} is ergodic? Justify your answer.

On the measure space ([0,1],B([0,1]),μ)([0,1], \mathcal{B}([0,1]), \mu) with Lebesgue measure μ\mu consider

θ1(x)=1+x,θ2(x)=x2,θ3(x)=1x\theta_{1}(x)=1+x, \quad \theta_{2}(x)=x^{2}, \quad \theta_{3}(x)=1-x

On the discrete measure space ({1,1},P({1,1}),12δ1+12δ1)\left(\{-1,1\}, \mathcal{P}(\{-1,1\}), \frac{1}{2} \delta_{-1}+\frac{1}{2} \delta_{1}\right) consider

θ4(x)=x\theta_{4}(x)=-x

(c) State Birkhoff's almost everywhere ergodic theorem.

(d) Let θ\theta be measure-preserving and let f:ERf: E \rightarrow \mathbb{R} be bounded.

Prove that 1n(f+fθ++fθn1)\frac{1}{n}\left(f+f \circ \theta+\ldots+f \circ \theta^{n-1}\right) converges in LpL^{p} for all p[1,)p \in[1, \infty).