Ergodicity, Recurrence and Mixing
2.8 Proof of Weak-Mixing Equivalences
Exercise 2.7.12.WriteT(k) for the k-fold Cartesian productT × · · · ×T. Prove(29) thatT(k) is ergodic for allk2 if and only ifT(2) is ergodic.
Exercise 2.7.13.Let T be an ergodic endomorphism of Td. The following exponential error rate for the mixing property(30),
f1, UTnf2 −
f1
f2
S(f1)S(f2)θn
for some θ < 1 depending on T and for a pair of constants S(f1), S(f2) depending onf1, f2∈C∞(Td), is known to hold.
(a) Prove an exponential rate of mixing for the map Tn : T → T defined byTn(x) =nx (mod 1).
(b) Prove an exponential rate of mixing for the automorphism ofT2 defined byT : (xy)→ y
x+y
.
(c) Could an exponential rate of mixing hold for all continuous functions?
2.8 Proof of Weak-Mixing Equivalences 55
1
n|Jk∩[0, n)|k1 n
n−1
i=0
ai−→0
asn→ ∞for eachk1, so eachJk has zero density. We will construct the setJ by taking a union of segments of each setJk. Since each of the sets Jk has zero density, we may inductively choose numbers 0< 1< 2<· · · with the property that
1
n|Jk∩[0, n)| 1
k (2.33)
fornk and anyk1. Define the setJ by J =
∞ k=0
Jk∩[k, k+1) .
We claim two properties for the setJ, namely
• an−→
n /∈J0 asn→ ∞;
• J has density zero.
For the first claim, note thatJk∩[k,∞)⊆J by (2.32), so if J nk
thenn /∈Jk, and soan 1k. This shows thatan−→
n /∈J0 as claimed.
For the second claim, notice that ifn∈[k, k+1) then again by (2.32)J∩ [0, n)⊆Jk∩[0, n) and so
1
n|J ∩[0, n)| 1 k by (2.33), showing thatJ has density zero.
(2) =⇒ (1): The sequence (an) is bounded, so there is some R > 0 withanRfor alln1. For eachk1 chooseNk so that
J nNk =⇒ an< 1 k and so that
nNk =⇒ 1
n|J∩[0, n)| 1 k. Then fornkNk,
1 n
n−1 i=0
ai = 1 n
⎛
⎜⎝
Nk−1 i=0
ai+
Nki<ni∈J,
ai+
i /∈J, Nki<n
ai
⎞
⎟⎠
< 1 n
RNk+R|J∩[0, n)|+n1 k
2R+ 1
k ,
showing (1).
(3) ⇐⇒ (1): This is clear from the characterization (2) of property (1).
Proof of Theorem 2.36. Properties (1), (6) and (7) are equivalent by Lemma2.41applied withan=|μ(A∩T−nB)−μ(A)μ(B)|.
(6) =⇒ (3): Given sets A1, B1, A2, B2 ∈ B, property (6) gives sets J1
andJ2 of density zero with μ
A1∩T−nB1
−→
n /∈J1μ(A1)μ(B1) and
μ
A2∩T−nB2
−→
n /∈J2μ(A2)μ(B2).
LetJ=J1∪J2; this still has density zero and lim
Jn→∞(μ×μ)
(A1×A2)∩(T×T)−n(B1×B2)
−(μ×μ)(A1×A2)·(μ×μ)(B1×B2)
= lim
Jn→∞μ(A1∩T−nB1)·μ(A2∩T−nB2)
−μ(A1)μ(A2)μ(B1)μ(B2)
= 0,
soT×T is weak-mixing since the measurable rectangles generateB×B. (3) =⇒ (1): IfT×T is weak-mixing, then property (1) holds in particular for subsets ofX×X of the form A×X and B×X, which shows that (1) holds forT, soT is weak-mixing.
(1) =⇒ (4): Let (Y,BY, ν, S) be an ergodic system and assume thatT is weak-mixing. For measurable setsA1, B1∈BandA2, B2∈BY,
1 N
N−1 n=0
(μ×ν)
A1×A2∩(T×S)−n(B1×B2)
= 1 N
N−1 n=0
μ(A1∩T−nB1)ν(A2∩S−nB2)
= 1 N
N−1 n=0
μ(A1)μ(B1)ν(A2∩S−nB2)
+1 N
N−1 n=0
μ(A1∩T−nB1)−μ(A1)μ(B1)&
ν(A2∩S−nB2). (2.34)
2.8 Proof of Weak-Mixing Equivalences 57
By the characterization in (2.31) and ergodicity ofS, the expression on the right in (2.34) converges to
μ(A1)μ(B1)ν(A2)ν(B2).
The second term in (2.34) is dominated by 1
N
N−1 n=0
μ(A1∩T−nB1)−μ(A1)μ(B1)
which converges to 0 sinceT is weak-mixing. It follows that 1
N
N−1 n=0
(μ×ν)
A1×A2∩(T×S)−n(B1×B2)
−→μ(A1)μ(B1)ν(A2)ν(B2)
soT×S is ergodic by the characterization in (2.31).
(4) =⇒ (2): Let (Y,BY, ν, S) be the ergodic system defined by the identity map on the singletonY ={y}. ThenT×Sis isomorphic toT, so (4) shows thatT is ergodic. Invoking (4) again now shows thatT×T is ergodic, proving (2).
(2) =⇒ (7): We must show that 1
N
N−1 n=0
μ(A∩T−nB)−μ(A)μ(B)2−→0
as N → ∞, for everyA, B ∈ B. Letμ2 denote the product measure μ×μ on (X×X,B⊗B). By the ergodicity ofT×T,
1 N
N−1 n=0
μ
A∩T−nB
= 1
N
N−1 n=0
μ2
(A×X)∩(T×T)−n(B×X)
−→μ2(A×X)·μ2(B×X) =μ(A)μ(B) and
1 N
N−1 n=0
μ
A∩T−nB2
= 1
N
N−1 n=0
μ2
(A×A)∩(T×T)−n(B×B)
−→μ2(A×A)·μ2(B×B) =μ(A)2μ(B)2. It follows that
1 N
N−1 n=0
μ
A∩T−nB
−μ(A)μ(B)&2
= 1 N
N−1 n=0
μ
A∩T−nB2
+μ(A)2μ(B)2
−2μ(A)μ(B)1 N
N−1 n=0
μ
A∩T−nB
→2μ(A)2μ(B)2−2μ(A)2μ(B)2= 0, so (7) holds.
(2) =⇒ (5): Suppose thatf is a measurable eigenfunction forT, so UTf =λf
for someλ∈S1. Define a measurable function onX×X by g(x1, x2) =f(x1)f(x2);
then
UT×Tg(x, y) =g(T x, T y) =λλg(x, y) =g(x, y)
so by ergodicity of T×T, g (and hencef) must be constant almost every- where.
All that remains is to prove that (5) =⇒ (2), and this is considerably more difficult. There are several different proofs, each of which uses a non- trivial result from functional analysis(31). Assume thatT×T is not ergodic, so there is a non-constant function f ∈ L2μ2(X×X) that is almost every- where invariant underT×T. We would like to have the additional symmetry property f(x, y) =f(y, x) for all (x, y)∈X ×X. To obtain this additional property, consider the functions
(x, y)→f(x, y) +f(y, x) and
(x, y)→i(f(x, y)−f(y, x)).
Notice that if both of these functions are constant, thenf must be constant. It follows that one of them must be non-constant. So without loss of generality we may assume thatf satisfies f(x, y) = f(y, x). We may further suppose (by subtracting
fdμ2) that
fdμ2 = 0. It follows that the operator F onL2μ defined by
(F(g)) (x) =
X
f(x, y)g(y) dμ(y)
2.8 Proof of Weak-Mixing Equivalences 59
is a non-trivial self-adjoint compact(32) operator, and so by Theorem B.3 has at least one non-zero eigenvalue λ whose corresponding eigenspace Vλ
is finite-dimensional. We claim that the finite-dimensional spaceVλ⊆L2μ is invariant underT. To see this, assume thatF(g) =λg. Then
λg(T x) =
X
f(T x, y)g(y) dμ(y)
=
X
f(T x, T y)g(T y) dμ(y) (sinceμisT-invariant)
=
X
f(x, y)g(T y) dμ(y),
sincef isT×T-invariant, soF(g◦T) =λ(g◦T) and thusg◦T ∈Vλ. It follows that UT restricted to Vλ is a non-trivial linear map of a finite-dimensional linear space, and therefore has a non-trivial eigenvector. Since
fdμ2 = 0,
any such eigenvector is non-constant.
2.8.1 Continuous Spectrum and Weak-Mixing
A more conventional proof of the difficult step in Theorem2.36, which may be taken to be (5) =⇒ (1), proceeds via the Spectral theorem (Theorem B.4) in the following form.
Alternative proof of (5) =⇒ (1) in Theorem 2.36.Definition2.35is clearly equivalent to the property that
Nlim→∞
1 N
N−1 n=0
|UTnf, g − f,1 · 1, g|= 0
for anyf, g∈L2μ, and by polarization this is in turn equivalent to
Nlim→∞
1 N
N−1 n=0
|UTnf, f − f,1 · 1, f|= 0
for any f ∈ L2μ (see Exercise 2.7.8 and page 441). By subtracting
Xfdμ fromf, it is therefore enough to show that iff ∈L2μ has
Xfdμ= 0, then 1
N
N−1 n=0
|UTnf, f|2−→0
asN → ∞. By (B.1), it is enough to show that for the non-atomic measureμf
onS1,
1 N
N−1 n=0
S1
zndμf(z)
2−→0 (2.35)
asN → ∞. Sincezn=z−nforz∈S1the product in (2.35) may be expanded to give
1 N
N−1 n=0
S1
zndμf(z) 2 = 1
N
N−1 n=0
S1
zndμf(z)·
S1
w−ndμf(w)
= 1 N
N−1 n=0
S1×S1
(z/w)ndμ2f(z, w) (by Fubini)
=
S1×S1
1 N
N−1 n=0
(z/w)n
dμ2f(z, w).
The measureμf is non-atomic so the diagonal set{(z, z)|z∈S1} ⊆S1×S1 has zeroμ2f-measure. Forz=w,
1 N
N−1 n=0
(z/w)n= 1 N
1−(z/w)N 1−(z/w)
−→0
as N → ∞, so the convergence (2.35) holds by the dominated convergence
theorem (Theorem A.18).
Exercises for Sect. 2.8
Exercise 2.8.1.Is the hypothesis that the sequence (an) be bounded neces- sary in Lemma2.41?
Exercise 2.8.2.Give an alternative proof of (1) =⇒ (5) in Theorem2.36 by proving the following statements:
(1) Any factor of a weak-mixing transformation is weak-mixing.
(2) A complex-valued eigenfunctionf ofUT has constant modulus.
(3) Iff is an eigenfunction ofUT, thenx→arg (f(x)/|f(x)|) is a factor map from (X,B, μ, T) to (T,BT, mT, Rα) for someα.
Exercise 2.8.3.Show the following converse to Exercise2.5.6: if a measure- preserving system (Y,BY, ν, S) is not totally ergodic then there exists a measure-preserving system (X,B, μ, T) and a K > 1 with the property that (Y,BY, ν, S) is measurably isomorphic to the system
(X(K),B(K), μ(K), T(K)) constructed in Exercise2.5.6.