Introduction à Un Domaine de Recherche

Generic Actions of Finitely Generated

Groups in Diff ∞
vol(M)

Disheng XU
École Normale Supérieure de Paris, SI-2011
Under the supervision of
Prof. Artur AVILA, Paris 6 and 7
October 20, 2013
Contents
1 Basic Concepts of Dynamical Systems 2

2 Differential Dynamical Systems 3

3 Group Actions and Homogeneous Dynamics 5

4 My Domain of Research 7

Appendix A Lyapunov Exponents and Nonuniformly Hyperbolic Dy-

namical Systems 8

1
 My research focuses on Differential dynamical systems. Before getting into more
details about my research field, I start by introducing some basic concepts.

1 Basic Concepts of Dynamical Systems

Definition 1.1. A measure-preserving dynamical system (X, B, µ, T ) is a proba-
bility space with a measure-preserving transformation T on it, i.e., for each A ∈
B, µ(T −1 A) = µ(A).

Definition 1.2. The measure-preserving transformation T as defined above is called

ergodic if for every T −invariant set A ∈ B, either µ(A) = 0 or µ(A) = 1.

Example 1. An irrational rotation on R/Z, is preserving Lebesgue measure. More-

over, it is an ergodic transformation.

Theorem 1.1. (Birkhoff-Khinchin) Let f be measurable, E(|f |) < ∞, and T be a

measure-preserving map. Then with probability 1:
n−1
1X
lim f (T k x) = E(f |C),
n→∞ n
k=0

where E(f |C) is the conditional expectation given the σ-algebra C of invariant sets of
T.

Corollary 1.1. (Pointwise Ergodic Theorem): In particular, if T is also ergodic, the

C is the trivial σ-algebra, and thus with probability 1:
n−1
1X
lim f (T k x) = E(f ).
n→∞ n
k=0

Example 2. Suppose f is a measurable 1−periodic function on R, θ is an irrational

number, Then for almost every x ∈ R :
n−1 Z 1
1X
lim f (x + kθ) = f (x)dx
n→∞ n 0
k=0

Definition 1.3. A topological dynamical system is defined as a complete metric space

X, which together with a continuous transformation T of that space.
Given a topological dynamical system (X, T ), if for every two open subsets U, V ,
there exists an integer n for which T n (U ) ∩ V 6= ∅, we call (X, T ) topologically tran-
sitive. If every point has a dense orbit, then we say that (X, T ) is minimal.

2
 In fact, T is transitive if and only if there exists a point in X which has a dense
orbit. Another good remark here is that T is minimal if and only if every T -invariant
closed set is either empty or X itself.
The central problem in dynamical systems is to consider whether a dynamcial
system is ergodic or a topological dynamical system is transitive or minimal. In par-
ticular, we consider these questions at the same time for conservative diffeomorphism
(area preserving diffeomorphism).

2 Differential Dynamical Systems

Definition 2.1. We say that a diffeomorphism f (C r , 0 ≤ r ≤ ∞) on a smooth
manifold M is conservative if it preserves a fixed volume form ω on M . The space of
C r conservative diffeomorphism on M is noted as Diffrvol (M).
Example 3. All the area preserving diffeomorphisms on surfaces are conservative.
A symplectic diffeomorphism f of a symplectic manifold (M, ω) (here we denote the
symplectic form ω) are conservative since they preserve the volume form ω n if M has
dimension 2n.
Corollary 2.1. If a volume preserving diffeomorphism is ergodic, then almost every
point has a dense orbit.
Corollary 2.1 can be proved by using Birkhoff ergodic theorem at almost every
point and indicate functions of arbitrary nonempty open set.
It would be very difficult for us to study an arbitrary area preserving diffeomor-
phism, hence we try to study generic behaviors in Diff∞vol (M).

Definition 2.2. We say a set in Diffrvol (M),0 ≤ r ≤ ∞ is residual if it is a countable

intersection of open dense sets. We call a property is generic in Diffrvol (M)if it holds
on a residual set.
Here is an example:
Theorem 2.1. (Oxtoby-Ulam)[1] For any generic homeomorphism f ∈ Diff0vol (M ),
Diff0vol (M ) denotes the group of homeomorphisms which preserve volume of M , the
invariant measure v is ergodic.
The theorem Oxtoby-Ulam is established mainly for the reason that the C 0 topol-
ogy is so weak that we can use a homeomorphism sufficiently chaos to approach an
arbitrary homeomorphism. The KAM theory tells us when r ≥ 4, the same conclusion
is false for Diffrvol (M), and that the generic transitivity is not true when r ≥ 4.
Suppose f ∈ Diff4vol (M), with an elliptic fixed point x. We can prove that, there
exists a neighborhood U of x and a C 4 diffeomorphism h from U into a neighborhood
V of the origin in R2 such that,
(hf h−1 )(r, θ) = (r, θ + α0 + α1 r) + o(krk3 + kθk3 )

3
 where (r, θ) the polar coordinates of a point in h(U ∩ f −1 (U )).
Definition 2.3. Under the above assumptions , if α0 6= 0, ± π2 , π, ± 2π
3
, α1 6= 0, we say
that the elliptic fixed point x of diffeomorphism f is non-degenerate.
Theorem 2.2. (Kolmogorov-Arnold-Moser)[2]1 Let f be a volume-preserving diffeo-
morphism of class C r , r ≥ 4 of a surface M . If x is a non-degenerate elliptic fixed
point, then for every  > 0 there exist an arbitrarily small neighborhood U of x and a
set U0 ⊂ U with the following properties:
• U0 is a union of f-invariant simple closed curves of class C r−1 containing x0 in
their interior.
• The restriction of f to each of these curves is topologically equivalent to an
irrational rotation.
• Denoting by λ the measure associated with the volume form of M , we have

λ(U − U0 ) ≤ λ(U ).

Notice that if g is a diffeomorphism close to f enough in Diffrvol (M), then there

exists a unique nondegenerate ellipitic fixed point xg of g close to x. As a corollary of
KAM theorem, for every such g, there exists a g−invariant disc Dg which has positive
measure and nonempty interior. In fact, we can choose Dg such that

∂(Dg ) = γg

where γg is one of the g−invariant circle.

Corollary 2.2. Let M be a compact surface without boundary, ω a volume form on
M and Diff rvol (M) the space of C r diffeomorphisms of M that preserve ω, endowed
with the C r topology. Then the set of f ∈ Diff rvol (M) which are ergodic (with respect
to the probability measure λ associated to ω) is not dense in Diff rvol (M) for r ≥ 4.
For 1 ≤ r ≤ 3 (The KAM theory is true for all r > 3), the problem of generic
ergodicity is still open, we get some partial results under certain assumptions of
hyperbolicity.
Definition 2.4. Let M be a manifold and f : M → M a diffeomorphism, f is called
uniformly hyperbolic or an Anosov diffeomorphism if for each x ∈ M there exists
a splitting of the tangent space Tx M = E s (x) ⊕ E u (x) and there are constants C > 0,
λ ∈ (0, 1) such that for all n ∈ N we have:

kDf n (v)k ≤ Cλn kvk , ∀v ∈ E s (x)

Df −n (v) ≤ Cλ−n kvk , ∀v ∈ E u (x)

1
The strongest version of Theorem KAM is due to Herman

4
 It is not hard to prove that Anosov diffeomorphism forms an open set in Diff r (M),
for all r ≥ 1. Moreover, there exists (un)stable manifolds tangent to (un)stable
distribution at all points in M .

Example 4. Let M = R2 /Z2 be a torus , f is a linear toral transformation on M ,

2 1

f (x) = A · x, A =
1 1

Then f is Anosov diffeomorphism, since the two eigendirections of A form (un)stable

distributions of f . And there exists a neighborhood U ⊂ Diff r (M) of f such that
∀g ∈ U , g is Anosov.

By Hopf argument [9][8], we know when r > 1, all volume preserving Anosov
diffeomorphisms are ergodic. But it is still unknown for r = 1.
We finish our disccussion about the measure-theoretical behavior of generic dif-
feomorphisms in Diffrvol . Now we will consider the topological dynamical behavior.
An early but remarkable result is:

Theorem 2.3. (Pugh[13] [14], closing lemma) For any generic diffeomorphism f ∈
Diff1vol , the periodic points of f are dense in M .

And an even better result is:

Theorem 2.4. (Generic transitivity [10]) Any generic diffeomorphism f ∈ Diff 1v is

transitive: there is a Gδ and dense subset G ⊂ M such that the forward orbits of any
x ∈ G is dense in M .

A remark here is that, the conclusions above are open for arbitrary r > 1, it might
even be extremely difficult. As what we mentioned before, generic transitivity is false
for r > 3.

3 Group Actions and Homogeneous Dynamics

In this section, we will generalize our conception of dynamical systems.

Definition 3.1. Let A be a set of diffeomorphisms (not necessarily volume-preserving)

on a manifold M , and Γ the semigroup (or group) generated by A. The action by
Γ is transitive if there exists a point which has dense orbit, is minimal if all points
have dense orbits, and is ergodic with respect to a volume v if Γ preserves v and any
G-invariant set has full measure or 0 measure.

Here are some examples: When M in the previous definition is a Lie group G,
or a homogeneous space G/H, A is defined as {Lgi : gi ∈ G}, where Lg means the
action by left multiplication of g. The volume v is always chosen as the homogeneous

5
measure on G/H. Then the problem whether Γ−action is transitive corresponds to
the behavior of Γ · H in G. Obviously, when G is not abelian, if A contains only one
element f , then the Γ−action is a special case in section 2. However f -action maynot
be transitive for general G/H, this is one reason why we consider Γ−actions.

Definition 3.2. If there exists a probability measure µ on A ( when A is finite, we

1
can give each element the same weight |A| ), we call a measure ν on M is µ-stationary
R R
if for all continuous function φ, φ(x)dν(x) = φ(g · x)dν(x)dµ(g).

Remark 3.1. There always exists stationary measure, the question is when there is
only one or very few stationary measures? When Γ preserves a volume v and there
exists an unique stationary measure, then it must be v and it implies easily that the
Γ action is ergodic and minimal. That explains why we care about uniqueness of
stationary measure.

Here is an example of stationary measure given by Furstenberg [3] and Margulis[4].

Example 5. Suppose A ⊂ SL(2, R), with a probability measure µ on A, acts on

projective space P1 , such that the support of µ is Zariski dense in SL(2, R). Then
there exists an unique stationary measure ν and ν is non-atomness. Moreover, if
Z
log kgk dµ(g) < ∞,

then the Lyapunov exponent is positive, i.e.:

Z
1
lim log kbn bn−1 ...b1 k > 0,
n→∞ n

where bn , ..., b1 is the random product of the i.i.d elements under the law µ.

Recently Y. Benoist and F. Quint proved a hard theorem about stationary mea-
sure:

Theorem 3.2. [12] Let G be a real simple Lie group, and Λ a lattice of G. Suppose µ
is a probability measure with compact support on G, and Supp(µ) generates a Zariski
dense subgroup of G. Then every µ-stationary probability without atoms on G/Λ is
G-invariant.

In other words, Benoist-Quint theorem provides a classification for non-atomness

µ−stationary measures.

6
4 My Domain of Research
My research interest lies mainly in generic actions of finitely generated groups Diff∞
vol (M),
studying properties like transitivity, minimality, ergodicity and uniqueness of station-
ary measure.
When the finitely generated group G ⊂ Diff∞ vol (M) is a cyclic group, we know from
KAM theory that the G-action cannot be generically transitive and ergodic. Thus
we consider the case when number of generator of G is bigger than 1, for which we
only get the result about transitivity now.

Theorem 4.1. [5] Suppose that M is a smooth surface. Then there exists a residual
set R ⊂ Diff rvol (M ) × Diff rvol (M )(r ∈ N ∪ {∞}) with the product C r topology such that
if (f, g) ∈ R, then the iterated function system IFS(f, g) is transitive.

Here IFS means iterated function systems, which means in fact the semigroup
generated by f, g. The main techniques used in the theorem are KAM theory, discus-
sions about hyperbolic fixed points, and some usual techniques in topology of surface
such as prime -ends.
Unfortunately, it still remains very hard to prove that all points have dense or-
bits simply by improving the theorem, and this paper does not have any measure
theoretical argument.
Now we consider the measure theoretical properties fof G−actions. On one hand,
Benoist-Quint shows us how to prove the uniqueness of of non-atomness stationary
measure under the condition of f, g acting on homogeneous space. Given an positive
lower bound for the Lyapunov exponents of the random product of elements in G, it
is possible to extend Benoist-Quint theorem partially to the case of diffeomorphisms.
On the other hand, with small Lyapunov exponents, the following two theorems
help to illustrate some interesting phenomena. Although it is limited to analytical
cases, we can generalize it to C k cases by similar arguments.

Definition 4.1. Let R : Γ → Diff(V ) be a homomorphism of a (discrete) group Γ,

and V a compact manifold. We say that R is recurrent if for all x ∈ V , there exists
a sequence (γi )i≥1 of distinct elements in Γ such that R(γi )(x) converges to x.

Theorem 4.2. (Ghys)[6] Let Γ be a group containing a free non abelian subgroup
and S a generating part of Γ. Assume that M is a real analytic compact manifold
and Diff w (M ) the group of its analytical diffeomorphisms, with its natural topology.
Then there exists a neighbourhood U of identity in Diff w (M ) such that every homeo-
morphism R : Γ → Diff w (M ) which maps S in U is recurrent.

Theorem 4.3. (Rebelo)[7] There is an open neighbourhood U of identity in Diff w (S 1 )

with the following property: if G is a non solvable subgroup of Diff w (S 1 ) and G admits
a finite set of generators contained in U, then there exists a ( local and nowhere zero)
vector field X defined in a neighbourhood of any point p ∈ S 1 , provided that p is not

7
a periodic point for G. Yet this vector field X defines a local flow which is contained
in the C ∞ -closure of the restriction of elements in G to this neighbourhood.

The previous two theorems tell us that it is useful to obtain some elements near
identity from small Lyapunov exponents condition, but how to obtain it is a quite
interesting problem. One possible method is to use Pesin theory (nonuniformly
hyperbolic) since it provides some useful distortion estimations. Due to limited space,
we only present in this paper the important Oseledec Theorem and the definition of
nonuniformly hyperbolic systems, see the Appendix.
For the rest of my research, I hope to gain new understanding and results about
generic actions of finitely generated groups Diff∞vol (M) by using tools like stationary
measure, Lyapunov exponents (Pesin theorey), KAM theory, etc.

Appendix A Lyapunov Exponents and Nonuniformly

Hyperbolic Dynamical Systems
Instead of prescribing bounds for the expansion and contraction of vectors, the theory
of nonuniformly hyperbolic dynamical systems measures the infinitesimal asymptotic
exponential relative behavior of points by the Lyapunov exponent.

Definition A.1. For a diffeomorphism f : M → M the forward Lyapunov exponent

of a vector v at a point x is defined by
1
X + (x, v) := lim log kDx f n (v)k .
n→∞ n

At any x this takes only finitely many values X1+ (x) < ... < Xp++ (x) (x) that de-
termine vector subspaces Vi+ (x) := {v ∈ Tx M |X + (x, v) ≤ Xi+ (x)}of Tx M which are
nested:
{0} = V0+ (x) ( ... ( Vp+ (x) (x) = Tx M.

Definition A.2. The multiplicity of Xi+ (x) is defined as ki+ (x) := dim Vi+ (x) −
+
dim Vi−1 (x); both of functions Xi+ and ki+ are f -invariant. The sum of the positive
Lyapunov exponents is X X
(x) := ki+ (x)Xi+ (x)
Xi+ (x)>0

Taking the corresponding limits as n → −∞ yields the backwards Lyapunov

exponents X − , for which the corresponding results hold.

Theorem A.1. (The Oseledec Multiplicative Ergodic Theorem) For a smooth dif-
feomorphism of a compact manifold M and any invariant Borel probability measure

8
almost every point x is Lyapunov-Perron regular, i.e.,

p+ (x) = p− (x) := p(x),

p(x)
Tx M = ⊕i=1 Ei (x), where Ei (x) := Vi+ (x) ∩ Vi− (x),
1
lim log kDx f n (v)k = Xi+ (x) = −Xi− =: Xi (x) uniformly in {v ∈ Ei (x)| kvk = 1},
n→±∞ n
p(x)
1 n
X
lim log |det Dx f | = Xi (x) dim Ei (x).
n→±∞ n
i=1

Definition A.3. A diffeomorphism f is said to have nonzero exponents on an in-

variant set Λ if for each x ∈ Λ there is an s = s(x) such that

X1 (x) < ... < Xs (x) < 0 < Xs+1 (x) < ... < Xp(x) (x). (1)

An f -invariant Borel probability measure is said to be hyperbolic if equation (1)

holds for almost every x ∈ M .

Definition A.4. A Borel set Λ ⊂ M is called nonuniformly partially hyperbolic (in

the broad sense) if there exist

• a Df -invariant decomposition Tx M = E 1 (x)⊕E 2 (x) for x ∈ Λ (i.e., Df E i (x) =

E i (f (x)) for i = s, u

• f -invariant Borel functions λ1 , λ2 : Λ → R+ (i.e., λi ◦ f = λi for i = 1, 2) such

that either λ1 < min{1, λ2 } or λ2 > max{1, λ1 } with the following properties:
For every 0 > 0 there are an f -invariant Borel function  : Λ → (0, 0 ) and
positive Borel functions C, K on Λ such that for every x ∈ Λ we have

∠(E 2 (x), E 1 (x)) ≥ K(x),

kDfxn vk ≤ C(x)(λ1 (x))n kvk , v ∈ E 1 (x), n ∈ N,

Dfx−n v ≤ C(x)(λ2 (x))−n kvk , v ∈ E 2 (x), n ∈ N,
and for every n ∈ Z and x ∈ Λ we have

C(f n (x) ≤ C(x)e(x)|n| , K(f n (x)) ≥ K(x)e(x)|n| .

We say that Λ is nonuniformly (completely) hyperbolic if λ1 < 1 < λ2 .

We see easily that for dim M = 2, if for almost every x ∈ M , Lyapunov exponent
for f is nonzero at point x, then there exists a full measure Λ which is nonunformly
hyperbolic.

9
References
[1] J. Oxtoby and S. Ulam, Measure-preserving homeomorphisms and metrical tran-
sitivity, Ann of Math. 42 (1941), 874 - 920.

[2] M. Herman, Sur les courbes invariantes par les difféomorphismes de l’anneau,
Astérisque 103-104, (1983).

[3] H. Furstenberg, Noncommuting random products, Transactions of the American

Mathematical Society, 1963.

[4] IY. Gol’dsheid, GA. Margulis, Lyapunov indices of a product of random matrices,
Russian Mathematical Surveys, 1989.

[5] A. Koropecki, M. Nassiri, Transitivity of generic semigroups of area-preserving

surface diffeomorphisms, Mathematische Zeitschrift (2010) 266: 707 - 718.

[6] E. Ghys, Sur les groupes engendrés par des difféomorphismes proches de l’identitié
, Sociedade Brasileira de Matemática (1993), Mat., Vol. 24, N. 2, 137 -178.

[7] J C. Rebelo, Ergodicity and rigidity for certain subgroups of Diff ω (S 1 ), Annales
Scientifiques de l’É.N.S. 4e série, tome 32, No.4(1990), p. 433 - 453.

[8] Anosov, D.V. and Ja. Sinai, Certain smooth ergodic systems, Uspehi Mat. Nauk
22 1967 no. 5 (137), 107 - 172.

[9] Hopf, E. Statistik der geodtischen Linien in Mannigfaltigkeiten negativer Krm-

mung. Ber. Verh. Sächs. Akad. Wiss. Leipzig 91, (1939). 261 - 304.

[10] C. Bonatti and S. Crovisier, Récurrence et généricité, Invent. Math. 158 (2004),
33 - 104.

[11] M.-C. Arnaud, C. Bonatti and S. Crovisier, Dynamiques symplectiques

génériques, preprint IMB 363 (2004), to appear in Ergod. Th. & Dynam. Sys.

[12] Y. Benoist, J-F. Quint, Mesures stationnaires et fermés invariants des espaces
homogènes, Annals of mathematics 174 (2011), 1111 - 1162.

[13] C.Pugh The closing lemma, Amer.J.Math. 89(1967),956-1009.

[14] C.Pugh An improved closing lemma and a general density theorem, Ergod. Th.&
Dynam. Sys. 3(1983),261-314.

10