Ergodicity, Recurrence and Mixing
2.3 Ergodicity
(b) Can you prove this starting with the weaker assumption that the upper densityd(A) is positive, and reaching the same conclusion?
Exercise 2.2.3.(a) Let (X,d) be a compact metric space and letT :X →X be a continuous map. Suppose that μ is a T-invariant probability measure defined on the Borel subsets ofX. Prove that forμ-almost everyx∈X there is a sequencenk → ∞withTnk(x)→xas k→ ∞.
(b) Prove that the same conclusion holds under the assumption that X is a metric space, T : X → X is Borel measurable, and μ is a T-invariant probability measure.
2.3 Ergodicity
Ergodicity is the natural notion of indecomposability in ergodic theory(15). The definition of ergodicity for (X,B, μ, T) means that it is impossible to split X into two subsets of positive measure each of which is invariant un- derT.
Definition 2.13.A measure-preserving transformation T : X → X of a probability space (X,B, μ) isergodic if for any∗ B∈B,
T−1B=B =⇒ μ(B) = 0 orμ(B) = 1. (2.2) When the emphasis is on the map T : X → X, and we are studying differentT-invariant measures, we will also say thatμis an ergodic measure forT. It is useful to have several different characterizations of ergodicity, and these are provided by the following proposition.
Proposition 2.14.The following are equivalent properties for a measure- preserving transformationT of (X,B, μ).
(1)T is ergodic.
(2)For any B∈B,μ(T−1BB) = 0 implies that μ(B) = 0 orμ(B) = 1.
(3)ForA∈B,μ(A)>0 implies that μ(∞
n=1T−nA) = 1.
(4)ForA, B∈B,μ(A)μ(B)>0 implies that there existsn1 with μ(T−nA∩B)>0.
(5)Forf :X →Cmeasurable, f ◦T =f almost everywhere implies that f is equal to a constant almost everywhere.
In particular, for an ergodic transformation and countably many sets of positive measure, almost every point visits all of the sets infinitely often under iterations by the ergodic transformation.
∗A setB∈BwithT−1B=Bis calledstrictly invariantunderT.
Proof of Proposition2.14.(1) =⇒ (2): Assume thatTis ergodic, so the implication (2.2) holds, and letB be analmost invariant measurable set—
that is, a measurable setB withμ
T−1BB
= 0. We wish to construct an invariant set from B, and this is achieved by means of the following limsup construction. Let
C= ∞ N=0
∞ n=N
T−nB.
For anyN 0,
B ∞
n=N
T−nB⊆ ∞
n=N
BT−nB
andμ(BT−nB) = 0 for alln1, sinceBT−nB is a subset of
n−1 i=0
T−iBT−(i+1)B,
which has zero measure. LetCN =∞
n=NT−nB; the setsCN are nested, C0⊇C1⊇ · · ·,
andμ(CNB) = 0 for eachN. It follows thatμ(CB) = 0, so μ(C) =μ(B).
Moreover,
T−1C= ∞ N=0
∞ n=N
T−(n+1)B= ∞ N=0
∞ n=N+1
T−nB=C.
ThusT−1C=C, so by ergodicityμ(C) = 0 or 1, so μ(B) = 0 or 1.
(2) =⇒ (3): Let A be a set with μ(A) >0, and let B =∞
n=1T−nA.
ThenT−1B⊆B; on the other handμ T−1B
=μ(B) soμ(T−1BB) = 0.
It follows that μ(B) = 0 or 1; since T−1A ⊆ B the former is impossible, soμ(B) = 1 as required.
(3) =⇒ (4): LetAandB be sets of positive measure. By (3), μ
∞
n=1
T−nA
= 1, so
0< μ(B) =μ ∞
n=1
B∩T−nA
∞
n=1
μ
B∩T−nA .
It follows that there must be somen1 withμ(B∩T−nA)>0.
2.3 Ergodicity 25
(4) =⇒ (1): LetA be a set withT−1A=A. Then 0 =μ(A∩XA) =μ(T−nA∩XA) for alln1 so, by (4), either μ(A) = 0 orμ(XA) = 0.
(2) =⇒ (5): We have seen that if (2) holds, thenT is ergodic. Let f be a measurable complex-valued function onX, invariant underT in the stated sense. Since the real and the imaginary parts off must also be invariant and measurable, we may assume without loss of generality thatf is real-valued.
Fixk∈Zandn1 and write
Akn={x∈X |f(x)∈[kn,k+1n )}.
ThenT−1AknAkn ⊆ {x∈X |f◦T(x)=f(x)}, a null set, so by (2) μ(Akn)∈ {0,1}.
For each n, X is the disjoint union
k∈ZAkn. It follows that there must be exactly onek=k(n) withμ(Ak(n)n ) = 1. Thenf is constant on the set
Y = ∞ n=1
Ak(n)n
andμ(Y) = 1, sof is constant almost everywhere.
(5) =⇒ (2): Ifμ(T−1BB) = 0 thenf =χBis aT-invariant measurable function, so by (5)χB is a constant almost everywhere. It follows thatμ(B)
is either 0 or 1.
Proposition 2.15.Bernoulli shifts are ergodic.
Proof. Recall the measure-preserving transformation σ defined in Exam- ple 2.9 on the measure space X = {0,1, . . . , n}Z with the product mea- sureμ. LetB denote aσ-invariant measurable set. Then given anyε∈(0,1) there is a finite union of cylinder sets A with μ(AB) < ε, and hence with|μ(A)−μ(B)|< ε. This meansAcan be described as
A={x∈X|x|[−N,N]∈F}
for some N and some finite set F ⊆ {0,1, . . . , n}[−N,N] (for brevity we write [a, b] for the interval of integers [a, b]∩Z. It follows that forM >2N,
σ−M(A) ={x∈X |x|[M−N,M+N]∈F},
where we think ofx|[M−N,M+N] as a function on [−N, N] in the natural way, is defined by conditions on a set of coordinates disjoint from [−N, N], so
μ(σ−MAA) =μ(σ−MA∩XA) =μ(σ−MA)μ(XA) =μ(A)μ(XA).
(2.3) SinceB isσ-invariant,μ(Bσ−1B) = 0. Now
μ(σ−MAB) =μ(σ−MAσ−MB)
=μ(AB)< ε, soμ(σ−MAA)<2εand therefore
μ(σ−MAA) =μ(Aσ−MA) +μ(σ−MAA)<2ε. (2.4) Therefore, by (2.3) and (2.4),
μ(B)μ(XB)<(μ(A) +ε) (μ(XA) +ε)
=μ(A)μ(XA) +εμ(A) +εμ(XA) +ε2
< μ(A)μ(XA) + 3ε <5ε.
Sinceεwas arbitrary, this implies thatμ(B)μ(XB) = 0, soμ(B) = 0 or 1
as required.
More general versions of this kind of approximation argument appear in Exercises2.7.3and2.7.4.
Proposition 2.16.The circle rotationRα:T→Tis ergodic with respect to the Lebesgue measuremT if and only if αis irrational.
Proof.Ifα∈Q, then we may writeα= pq in lowest terms, soRαq =IT is the identity map. Pick any measurable setA⊆Twith 0< mT(A)<1q. Then
B =A∪RαA∪ · · · ∪Rqα−1A
is a measurable set invariant underRαwithmT(B)∈(0,1), showing thatRα
is not ergodic.
If α /∈ Q then for any ε > 0 there exist integers m, n, k with m = n and|mα−nα−k|< ε. It follows thatβ= (m−n)α−klies withinεof zero but is not zero, and so the set{0, β,2β, . . .}considered inTisε-dense (that is, every point ofTlies withinεof a point in this set). Thus (Zα+Z)/Z⊆T is dense.
Now suppose thatB⊆Tis invariant underRα. Then for anyε >0 choose a functionf ∈C(T) withf−χB1< ε. By invariance ofB we have
f◦Rnα−f1<2ε for alln. Since f is continuous, it follows that
f◦Rt−f12ε
2.3 Ergodicity 27
for allt∈R. Thus, sincemTis rotation-invariant, f−
f(t) dt 1
=
(f(x)−f(x+t)) dt dx f(x)−f(x+t)dxdt2ε
by Fubini’s theorem (see Theorem A.13) and the triangle inequality for inte- grals. Therefore
χB−μ(B)1χB−f1+ f−
f(t) dt 1
+
f(t) dt−μ(B) 1
<4ε.
Since this holds for every ε >0 we deduce that χB is constant and there- foreμ(B) ∈ {0,1}. Thus for irrational α the transformation Rα is ergodic
with respect to Lebesgue measure.
Proposition 2.17.The circle-doubling mapT2 :T →T from Example 2.4 is ergodic (with respect to Lebesgue measure).
Proof.By Example 2.8, T2 and the Bernoulli shift σ on X = {0,1}N to- gether with the fair coin-toss measure are measurably isomorphic. By Propo- sition2.15 the latter is ergodic, and it is clear that measurably isomorphic systems are either both ergodic or both not ergodic.
Ergodicity (indecomposability in the sense of measure theory) is a uni- versal property of measure-preserving transformations in the sense that ev- ery measure-preserving transformation decomposes into ergodic components.
This will be shown in Sects. 4.2 and 6.1. In contrast the natural notion of indecomposability in topological dynamics—minimality—does not permit an analogous decomposition (see Exercise 4.2.3).
In Sect. 2.1 we pointed out that in order to check whether a map is measure-preserving it is enough to check this property on a family of sets that generates the σ-algebra. This is not the case when Definition 2.13 is used to establish ergodicity (see Exercise2.3.2). Using a different character- ization of ergodicity does allow this, as described in Exercise2.7.3(3).
Exercises for Sect. 2.3
Exercise 2.3.1.Show that ergodicity is not preserved under direct products as follows. Find a pair of ergodic measure-preserving systems (X,BX, μ, T) and (Y,BY, ν, S) for whichT×S is not ergodic with respect to the product measureμ×ν.
Exercise 2.3.2.Define a mapR:T×T→T×TbyR(x, y) = (x+α, y+α) for an irrational α. Show that for any set of the form A×B with A, B measurable subsets ofT(such a set is called ameasurable rectangle) has the property of Definition2.13, but the transformationRis not ergodic, even ifα is irrational.
Exercise 2.3.3.(a) Find an arithmetic condition onα1andα2that is equiv- alent to the ergodicity ofRα1×Rα2 :T×T→T×Twith respect tomT×mT. (b) Generalize part (a) to characterize ergodicity of the rotation
Rα1× · · · ×Rαn:Tn →Tn with respect tomTn.
Exercise 2.3.4.Prove that any factor of an ergodic measure-preserving sys- tem is ergodic.
Exercise 2.3.5.Extend Proposition2.14by showing that for eachp∈[1,∞] a measure-preserving transformationT is ergodic if and only if for any Lp functionf,f◦T =f almost everywhere implies thatf is almost everywhere equal to a constant.
Exercise 2.3.6.Strengthen Proposition2.14(5) by showing that a measure- preserving transformation T is ergodic if and only if any measurable func- tionf :X→Rwithf(T x)f(x) almost everywhere is equal to a constant almost everywhere.
Exercise 2.3.7.LetXbe a compact metric space and letT :X →Xbe con- tinuous. Suppose thatμis aT-invariant ergodic probability measure defined on the Borel subsets ofX. Prove that forμ-almost everyx∈X and everyy in the support of μthere exists a sequence nk ∞ such thatTnk(x)→y as k → ∞. Here the support Supp(μ) of μ is the smallest closed subset A ofX withμ(A) = 1; alternatively
Supp(μ) =X
O⊆Xopen, μ(O)=0
O.
Notice that X has a countable base for its topology, so the union is still aμ-null set (see p. 406).