306 CHAPTER 12. THE CONSTRUCTION OF MEASURES

Definition 12.8.6 The above mapping T is ergodic if the only invariant sets have measure0 or 1.

If the map, T is ergodic, the following corollary holds.

Corollary 12.8.7 In the situation of Theorem 12.8.5, if T is ergodic, then

g(ω) =∫

f (ω)dµ

for a.e. ω .

Proof: Let g be the function of Theorem 12.8.5 and let R1 be a rectangle in R2 = Cof the form [−a,a]× [−a,a] such that g−1 (R1) has measure greater than 0. This set isinvariant because the function, g is invariant and so it must have measure 1. Divide R1 intofour equal rectangles, R′1,R

′2,R′3,R′4. Then one of these, renamed R2 has the property that

g−1 (R2) has positive measure. Therefore, since the set is invariant, it must have measure1. Continue in this way obtaining a sequence of closed rectangles, {Ri} such that thediameter of Ri converges to zero and g−1 (Ri) has measure 1. Then let c = ∩∞

j=1R j. Weknow µ

(g−1 (c)

)= limn→∞ µ

(g−1 (Ri)

)= 1. It follows that g(ω) = c for a.e. ω . Now

from Theorem 12.8.5,

c =∫

cdµ = limn→∞

1n

∫Sn f dµ =

∫f dµ.

12.9 Product MeasuresLet (X ,S ,µ) and (Y,T ,ν) be two complete measure spaces. In this section consider theproblem of defining a product measure, µ×ν which is defined on a σ algebra of sets ofX ×Y such that (µ×ν)(E×F) = µ (E)ν (F) whenever E ∈S and F ∈ T . I found thefollowing approach to product measures in [47] and they say they got it from [50].

Definition 12.9.1 Let R denote the set of countable unions of sets of the form A×B, whereA ∈S and B ∈ T (Sets of the form A×B are referred to as measurable rectangles) andalso let

ρ (A×B) = µ (A)ν (B) (12.9.30)

More generally, define

ρ (E)≡∫ ∫

XE (x,y)dµdν (12.9.31)

whenever E is such that

x→XE (x,y) is µ measurable for all y (12.9.32)

and

y→∫

XE (x,y)dµ is ν measurable. (12.9.33)

306 CHAPTER 12. THE CONSTRUCTION OF MEASURESDefinition 12.8.6 The above mapping T is ergodic if the only invariant sets have measureOor 1.If the map, T is ergodic, the following corollary holds.Corollary 12.8.7 In the situation of Theorem 12.8.5, if T is ergodic, thens(@) = | f(o)dufor ae. @.Proof: Let g be the function of Theorem 12.8.5 and let R; be a rectangle in R* = Cof the form [—a,a] x [—a,a] such that g~'(R,) has measure greater than 0. This set isinvariant because the function, g is invariant and so it must have measure |. Divide R; intofour equal rectangles, R),R5,R,R/,. Then one of these, renamed R2 has the property thatg! (R2) has positive measure. Therefore, since the set is invariant, it must have measure1. Continue in this way obtaining a sequence of closed rectangles, {R;} such that thediameter of R; converges to zero and g~'(R;) has measure 1. Then let c = M1) Rj. Weknow uw (g7!(c)) = lim, +. (g~! (R;)) = 1. It follows that g(@) = c for a.e. @. Nowfrom Theorem 12.8.5,c= [edu =jim~ [Stan = f fan. 1:no nN12.9 Product MeasuresLet (X,.%,) and (Y,.7,v) be two complete measure spaces. In this section consider theproblem of defining a product measure, Lt x V which is defined on a o algebra of sets ofX x Y such that (ux v)(E x F) = (E£)v(F) whenever FE € .Y and F € J. I found thefollowing approach to product measures in [47] and they say they got it from [50].Definition 12.9.1 Let & denote the set of countable unions of sets of the form A x B, whereA€SYandBe J (Sets of the form A x B are referred to as measurable rectangles) andalso letp(AxB)=u(A)v(B) (12.9.30)More generally, definep(E)= / / Xe (x,y) dudv (12.9.31)whenever E is such thatx— Xz (x,y) is uw measurable for all y (12.9.32)andyo / Xe (x,y) du is v measurable. (12.9.33)