ROCKY MOUNTAIN

JOURNAL OF MATHEMATICS
Volume 37, Number 2, 2007

TOPOLOGICAL TRANSITIVITY AND

MIXING NOTIONS FOR GROUP ACTIONS

GRANT CAIRNS, ALLA KOLGANOVA AND ANTHONY NIELSEN

ABSTRACT. This paper surveys six notions of dynamical

transitivity and mixing, in the context of group actions on
topological spaces. We discuss the relations between these
notions, and the manner in which they are inherited by
subgroups, by taking products, and when passing to the
induced action on hyperspace, i.e., the space of compact
subsets. The focus of the paper is on the fact that certain
standard notions, which are equivalent in the classical theory
of the dynamics of flows and the iteration of single maps,
are distinct for general group actions. The paper examines
how the notions coalesce (a) for actions of abelian groups and
(b) for chaotic actions.

0. Introduction. Consider an action of an infinite group G on

a Hausdorff topological space M . This paper surveys six notions of
dynamical transitivity and mixing for the action of G. We don’t
assume any particular topology on G, but we assume that the action
is “continuous” in the sense that, for each group element g, the
corresponding map g : M → M is a homeomorphism.
The fundamental transitivity and mixing notions are:

Definition 1. The action of G on M is:

(a) topologically transitive if, for every pair of nonempty open subsets
U and V of M , there is an element g ∈ G such that gU ∩ V = ∅.
(b) strongly topologically mixing if for any pair of nonempty open
subsets U and V of M , the set {g ∈ G; gU ∩ V = ∅} is finite.
(c) topologically k-transitive for k ∈ N, if the induced action of G on
the k-fold Cartesian product M k is topologically transitive. Topological
2-transitivity is also called weak topological mixing.

Key words and phrases. Group action, mixing, topological transitivity.

Received by the editors on September 4, 2005.

c
Copyright 2007 Rocky Mountain Mathematics Consortium

371
372 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

We emphasize that we are not assuming any topology on G. For

actions of topological groups, it is more natural, and in keeping with
tradition, to define strong topologically mixing by the condition: for any
pair of nonempty open subsets U, V of M , the set {g ∈ G; gU ∩V = ∅}
is compact.
For brevity, we will drop the adjective topological wherever there is
no risk of confusion. In particular, transitivity will mean topological
transitivity, and not the group theoretic sense of point-transitivity.
Apart from the above fundamental notions, there are also two related
notions that we will consider in this paper:

Definition 2. The action of G on M is:

(a) totally transitive if every subgroup of finite index is transitive on
M.
(b) elastic if for every n ∈ N and any finite collection of nonempty
open sets U, V1 , . . . , Vn , there exists g ∈ G such that gU ∩ Vi = ∅, for
all i ∈ {1, . . . , n}.

The above notions of transitive, and strongly and weakly mixing

actions are classical, see [19, 20, 24]. The notions of totally transitive
and elastic actions are less common; they are both generalizations of
corresponding notions of the dynamics of flows and the iterates of a
single map. The term “totally transitive” is reasonably well established.
A continuous map f : M → M is said to be totally transitive if, for
all natural numbers k, the kth iterate f k is transitive, see [3, 8, 15,
34]. When f is a homeomorphism, this is the same as demanding
that the finite index subgroups of the group f  = {f k ; k ∈ Z} are
transitive. The notion of an elastic action has appeared occasionally
in the literature, but it doesn’t usually have a separate name since it
coincides with weak mixing for flows and single maps, see Theorem 2
below. For single maps, elastic was termed “strongly transitive” by
Banks in [7]. However the terminology “strongly transitive” is already
employed to mean two distinct things in topological dynamics, see [10,
23, 30]. For this reason we have felt it necessary to introduce the term
“elastic”.
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 373

For flows and single maps, the three notions k-transitive for k ∈ N,
weak mixing and elastic are all equivalent, while we will see that,
for general group actions, the three conditions are distinct. The aim
of this paper is to give an essentially self-contained discussion of the
relations between the six conditions: strongly mixing, k-transitive for
all k, weakly mixing, elastic, totally transitive and transitive. We also
examine their inheritance properties, and the special cases of actions
of abelian groups and actions which are chaotic in the sense that they
are transitive and the points with finite orbit form a dense set. See
Theorems 1, 2 and 3. The general conclusion is that, for abelian groups,
the relations between the six conditions have the same equivalences
as they do for flows and single maps, while the assumption that an
action (of a not necessarily abelian group) is chaotic has an even greater
combining effect on the conditions.
The sections of this paper are: Section 1, Brief review of transitivity,
Section 2, Logical implications between the notions, Section 3, Exam-
ples, Section 4, Inheritance of notions (a) under semi-conjugacy, (b)
from and by subgroups, (c) when taking products, (d) when passing to
hyperspace, Section 5, Actions of abelian groups and Section 6, Chaotic
actions.

Important note. Throughout this paper, we consider a continuous

action of an infinite group G on a Hausdorff space M . We denote the
image of x ∈ M under g ∈ G simply by gx. We denote the orbit
{gx; g ∈ G} of x ∈ M by Gx. We denote by id the identity element of
G. For brevity, instead of saying that the action of G on M is transitive,
elastic, etc., we will simply say that G is transitive, elastic, etc.

1. Brief review of transitivity. For general properties of tran-

sitivity in the case of a single map, see [2, 5, 25]. The following
elementary lemma is quite useful, see [20, Remark 9.10] for further
equivalent conditions.

Lemma 1. For the action of G on M , the following conditions are

equivalent:
(a) G is transitive,
(b) every nonempty G-invariant open subset of M is dense,
374 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

(c) every G-invariant subset U ⊆ M is either dense or nowhere dense,

(d) M does not possess two disjoint G-invariant nonempty open
subsets.

Notice that if there exists a point x ∈ M such that the orbit Gx

of x is dense in M , then G is transitive. For the converse, one needs
some additional hypotheses on M . Recall that a subset A ⊆ M is said
to be meager (or of first category) if A is the union of countably many
nowhere dense subsets of M . (For more information on meager sets, see
the exercises in [9, Chapter IX.5].) A Gδ set is a subset of M that can
be written as the intersection of countably many open sets. A subset
A ⊆ M is said to be residual, or generic, if its complement is meager.
A topological space M is a Baire space (or a space of second category)
if every countable intersection of open dense subsets of M is dense in
M . Every locally compact Hausdorff space is a Baire space, as is every
complete metric space; this latter fact is Baire’s category theorem. A
good brief account of Baire spaces is given in [9, Chapter IX.5.3]. A
key fact for us is that Baire spaces are not themselves meager, and
more generally:

Lemma 2. In Baire spaces, meager sets have empty interiors.

The next result is well known, see [20, Theorem 9.22], and is often
presented for continuous maps of spaces, see for example, [29, Prop.
I.11.4]:

Proposition 1. If G is transitive and M is a second countable Baire

space M , then there exists a point x ∈ M with dense orbit in M and,
in fact, the set of points with dense orbit is a residual set.

Proof. If M is second countable, then there exists a countable open

base {Ui ; i ∈ Z}. Let Vi = G(Ui ) for each i, and set V = ∩Vi . The
sets Vi are dense, by Lemma 1, and so V is residual and hence dense,
as M is a Baire space. We claim that each element of V has a dense
orbit. Indeed, if W ⊆ M is open, one has Ui ⊆ W for some i. Thus,
if x ∈ V , we have x ∈ Vi and it follows that gx ∈ Ui for some g ∈ G.
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 375

Recall that a topological space M is dense in itself if it contains no

isolated points. The following lemma was shown for compact metric
spaces in [24]; their short argument also applies to any second countable
Baire space M which is dense in itself. We present the result in slightly
more generality:

Lemma 3. If G is transitive and M is dense in itself, then for every

pair of nonempty open subsets U, V of M , the set {g ∈ G; gU ∩V = ∅}
is infinite.

Proof. Let U, V be as in the statement of the lemma, and let k ∈ N.

We will show that the set {g ∈ G; gU ∩ V = ∅} has at least k
elements. First note that there exist k pair-wise disjoint nonempty
subsets V1 , . . . , Vk ⊆ V . Indeed, as M is Hausdorff and dense in
itself, there exist nonempty disjoint open subsets V1 and W1 of V .
Similarly, there exist nonempty disjoint open subsets V2 and W2 of
W1 , and nonempty disjoint open subsets V3 and W3 of W2 , and so on.
Let U1 = U . As the action is transitive, there exists g1 ∈ G such that
g1 U1 ∩V1 = ∅. Let U2 = U1 ∩g1−1 V1 . Then there exists g2 ∈ G such that
g2 U2 ∩ V2 = ∅. As V1 ∩ V2 = ∅, we have g2 = g1 . Let U3 = U2 ∩ g2−1 V2 .
There exists g3 ∈ G such that g3 U3 ∩ V3 = ∅. As V1 ∩ V3 = ∅ and
V2 ∩ V3 = ∅, we have g3 = g1 and g3 = g2 . Continuing is this way, we
obtain k distinct elements g1 , . . . , gk ∈ {g ∈ G; gU ∩ V = ∅}.

An action of G on M is said to be non-wandering if the set {g ∈

G; g = id , gU ∩ U = ∅} is not empty for all nonempty open sets U . A
transitive flow on an infinite space M is non-wandering if and only if M
is dense in itself [16, Proposition II.4.10]. Similarly, by Lemma 3, every
transitive action of a group on a dense in itself Hausdorff space is non-
wandering. However, there are transitive non-wandering group actions
on spaces which are not dense in themselves. For example, consider the
usual action of the infinite dihedral group D∞ on Z, equipped with the
discrete topology; D∞ is generated by a translation x → x + 1 and a
reflection x → −x. Obviously, this action is non-wandering, but every
point of Z is an isolated point. Provided the Hausdorff space M has at
least two elements, it is easy to see that M is dense in itself if M admits
an elastic action of some group. The same is true if M admits a weakly
376 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

mixing action, see [24]. However, it is easy to construct examples of

totally transitive actions on spaces which are not dense in themselves;
see the third action in Example 2 below, and consider a group with
discrete topology.
Lemma 3 has a useful extension which we will use later:

Lemma 4. If G is elastic on M , then for all nonempty open subsets

U, V1 , . . . , Vk of M , the set {g ∈ G; gU ∩ Vi = ∅ for all i = 1, . . . , k}
is infinite.

Proof. As we have just remarked, if G is elastic, then M is dense

in itself. Consider nonempty open subsets U, V1 , . . . , Vk . Since G is
transitive, the set SU,U = {g ∈ G; gU ∩U = ∅} is infinite, by Lemma 3.
Let

S = {g ∈ G; gU ∩ Vi = ∅ for all i = 1, . . . , k}

and assume that S is finite: S = {h1 , . . . , hn }, say. Let g ∈ SU,U and

 = gU ∩ U . As G is elastic, there exists h ∈ G such that hU
U  ∩ Vi = ∅
for all i. One has hgU ∩hU ∩Vi = ∅ and so hg, h ∈ S. Then hg = hi and
h = hj for some i, j ∈ {1, . . . , k} and g = h−1
j hi . Thus, the set SU,U
contains at most n2 elements, which gives the required contradiction.

Now recall:

Definition 3. A subset A ⊆ M of a topological space M is said to

have the Baire property if there are meager sets B, C and an open set
U such that A = B ∪ (U \C).

It is easy to see that the class of sets having the Baire property is a
σ-algebra. We state the following simple lemma without proof:

Lemma 5. Suppose that M is a Baire space and that A ⊆ M has

the Baire property, with A = B ∪ (U \C) where B, C are meager and U
is open. Then A is meager if and only if U = ∅.
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 377

For more information on sets with the Baire property, see [32], [33,
Chapter 4], [9, Chapter XII.8] and Exercise 6 of [9, Chapter IX.5].
Note that for sets A with the Baire property, the presentation A =
B ∪ (U \C) is not unique. Obviously one can suppose in general that
B ∩ U = ∅ and that C ⊆ U , but this still doesn’t eliminate the lack of
uniqueness. In particular, there exist sets A with the Baire property
for which one can write A = B ∪ (U \C) and A = B  ∪ (U  \C  ), where
B, B  , C, C  are meager and U, U  are open, and U  = U . For example,
the interval A = (0, 1) can be written as ∅∪A, and as {1/2}∪U , where
U = (0, (1/2)) ∪ ((1/2), 1). Nevertheless, there is a “canonical” way to
write a set with the Baire property, as the following lemma shows:

Lemma 6. Suppose that M is a topological space and that A ⊆ M

has the Baire property. Then there are unique sets B, C, U such that
the following conditions hold :
(a) B, C are meager, U is open and A = B ∪ (U \C),
(b) B ∩ U = ∅ and C ⊆ U ,
(c) if B  , C  , U  verify the analogous conditions to parts (a) and (b),
then U  ⊆ U , B ⊆ B  and C ⊆ C  .

Proof. Consider all the possible ways of writing A = B  ∪ (U  \C  ),

where B  , C  , U  verify the analogous conditions to part (a) and (b),
and let U be the union of the U  , and B, C be the intersection of
B  , C  respectively. It is easy to verify that B, C satisfy the required
properties.

We will require the following:

Lemma 7. Suppose that G acts on M and that A ⊆ M is a G-

invariant set with the Baire property. Write A = B ∪ (U \C), where
B, C, U have the properties of Lemma 6. Then the open set U is G-
invariant.

Proof. Let g ∈ G. As A is G-invariant, we have A = gA =

gB ∪ (gU \gC). Thus, by Lemma 6, gU ⊆ U . For the same reason,
g −1 U ⊆ U , and thus gU = U .
378 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

Definition 4. We say that an action of G on M is topologically

ergodic if every G-invariant set with the Baire property is either meager
or residual.

The following result is probably well known, but we could not find its
proof in the literature. It was stated without proof by Oxtoby in [33].

Proposition 2. A continuous action on a Baire space is transitive

if and only if it is topologically ergodic.

Proof. Suppose that G acts continuously on a Baire space M .

Suppose first that the action is not transitive. By Lemma 1 (d), M
possesses two nonempty disjoint G-invariant open subsets, U1 , U2 . As
U1 , U2 are open, they have the Baire property. Moreover, by Lemma 5,
neither of them is meager and consequently, as each is contained in the
complement of the other, neither of them is residual. Hence, the action
is not topologically ergodic.
Conversely, if the action is not topologically ergodic, then M possesses
two disjoint G-invariant subsets, A1 , A2 , each having the Baire prop-
erty, such that neither is meager. As the Ai have the Baire property,
we can write Ai = Bi ∪ (Ui \Ci ), where Bi , Ci , Ui have the properties of
Lemma 6. Since the sets Ai are not meager, the open sets Ui are not
empty, by Lemma 5. As A1 , A2 are disjoint, one has U1 ∩ U2 ⊆ C1 ∪ C2 .
In particular, the open set U1 ∩ U2 is meager and hence empty by
Lemma 2. Thus, the Ui are disjoint and nonempty and, moreover, they
are G-invariant by Lemma 7. Hence, the action is not transitive.

Transitivity is a topological version of the measure theoretic notion

of ergodicity. We finish this brief review by describing this connection.
Recall that an action of a group G on a measure space M by measure
preserving transformations is ergodic if M does not contain two disjoint
G-invariant measurable sets each of positive measure. The following
classical result provides a useful source of transitive actions, cf. [29,
Proposition II.2.6]:

Proposition 3. On a measure space in which nonempty open sets

have positive measure, every ergodic action is transitive.
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 379

Proof. Suppose that a group G acts on a measure space M by measure

preserving transformations. If the action is not transitive, then by part
Lemma 1 (d), M possesses two nonempty disjoint G-invariant open
subsets, U1 , U2 . As the Ui are open and nonempty, they have positive
measure. So the action is not ergodic.

2. Logical implications between the notions. Most of the

implications in the following theorem are generalizations of known
results for flows or single maps. The only exception is the result that
weak mixing implies total transitivity. For actions of abelian groups
this is a consequence of well known simple arguments, see Section 5.
However, for non-abelian groups, one requires a different approach.

Theorem 1. On second countable Baire spaces, one has the following

implications:

strongly mixing

k-transitive for all k 



weakly mixing   elastic
  


totally transitive

transitive
and the “second countable Baire” hypothesis is only used in the proof
that weak mixing implies total transitivity.

Proof. It is obvious that k-transitivity for all k implies both weak

mixing and elasticity. It is also obvious that total transitivity implies
transitivity. So there remain 3 implications to prove:
380 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

Strongly mixing ⇒ k-transitive for all k. Indeed, let k ∈ N and sup-

pose the action of G is strongly mixing, and let U1 , . . . , Uk , V1 , . . . , Vk
be nonempty open subsets of M . Let Ai = {g ∈ G; gUi ∩ Vi = ∅} for
i ∈ {1, . . . , k} and A = ∪i Ai . Since the action is strongly mixing, A is
finite and so there exists g ∈ G\A. Then gUi ∩ Vi = ∅ for all i and so
G is k-transitive.
Elastic ⇒ totally transitive. Suppose that H is a subgroup of finite
index in G. Consider a left transversal {g1 , g2 , . . . , gk } of H in G, i.e.,
suppose that gi H are the distinct left cosets of H. Let U and V be
nonempty open sets of M . Since G is elastic, there exists g ∈ G such
that gU ∩ gi V = ∅, for all i ∈ {1, . . . , k}. Notice that g = gi h for some
i and some h ∈ H. Then gi hU ∩ gi V = ∅ implies hU ∩ V = ∅. Hence
H is transitive. So G is totally transitive.
Weakly mixing ⇒ totally transitive. We will prove the contrapositive,
i.e., if the action of a group G on a second countable Baire space M is
not totally transitive, then G is not weakly mixing. That is, we suppose
that G acts transitively, and that the induced action of a finite index
subgroup H is not transitive; we will show that the action of G is not
weakly mixing.
For each x ∈ M , let Vx denote the interior of the closure of the H-
orbit of x; Vx = Int (Hx). Let D denote the set of points x for which
Gx is dense in M . By Proposition 1, D is nonempty.

Lemma 8. For each x ∈ D, we have:

(a) Vx is H-invariant, i.e., for all h ∈ H, hVx = Vx ,
(b) x ∈ Vx ,
(c) Vx = Hx,
(d) For all g ∈ G, if Vx ∩ Vgx = ∅, then Vgx = Vx .

Proof. Let x ∈ D. Part (a) is immediate from the definition of Vx .

(b) Consider a left transversal {g1 = id, g2 , . . . , gk } of H in G. We
have Gx = ∪ki=1 gi Hx, and so

k
 k

M = Gx = gi Hx = gi Hx.
i=1 i=1
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 381

Since M is a Baire space and M is the finite union of the homeomorphic

sets gi Hx, the set g1 Hx = Hx must have nonempty interior, i.e., Vx is
nonempty. Since Vx is a nonempty open subset of Hx and the H-orbit
of x is dense in Hx, there exists h ∈ H such that hx ∈ Vx . Thus, by
part (a), x ∈ Vx .
(c) It follows from (a) and (b) that Hx ⊆ Vx and so Hx ⊆ Vx .
Conversely, by definition, Vx ⊆ Hx and so Vx ⊆ Hx.
(d) If Vx ∩ Vgx = ∅, then Vx ∩ Vgx is a nonempty open subset of Hx.
Thus, as the H-orbit of x is dense in Hx, there exists h ∈ H such that
hx ∈ Vgx . Then, by part (a), Hx ⊆ Vgx and so Hx ⊆ Vgx = H(gx),
by part (c). Thus, taking interiors, Vx ⊆ Vgx . Since x ∈ D, we have
gx ∈ D. So the previous argument gives Vgx ⊆ Vg−1 gx = Vx . Hence,
Vx = Vgx .

Returning to the proof of Theorem 1, let x ∈ D. As H is not

transitive, by hypothesis, Hx = M . As M \Hx is open, there exists
g ∈ G such that gx ∈ M \Hx. If G is weakly mixing, then there exists
f ∈ G such that f Vx ∩Vx = ∅ and f Vx ∩Vgx = ∅. But then, by Lemma
8 (d), one would have f Vx = Vx and f Vx = Vgx . But then Vx = Vgx ,
which is impossible as gx ∈ Vgx by Lemma 8 (b) and gx ∈ / Vx , by our
choice of g.

3. Examples. In this section we give some simple examples which

show that the implications of Theorem 1 are strict.

Example 1. Transitive ⇒ totally transitive. The action by multi-

plication on R of the group R∗ of nonzero reals is transitive, but the
induced action of the positive reals R+
∗ is not.

Example 2. Totally transitive ⇒ weakly mixing nor elastic. We

give three examples. First, the linear action of SL(2, Z) on R2 is
totally transitive [13], but it is neither weakly mixing nor elastic.
Indeed, consider a small disc U centered at the origin, and small discs
V1 , V2 centered on the standard basis vectors e1 , e2 . Because SL(2, Z)
preserves area, there is no matrix A ∈ SL(2, Z) for which AU ∩ V1 and
AU ∩ V2 are both nonempty.
382 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

Second, for an action of an abelian group, consider an irrational

rotation Rα on the circle S 1 . The cyclic group of homeomorphisms
n
generated by Rα is totally transitive since Rα and every iterate Rα =
Rnα with n ∈ Z\{0}, is transitive. However, being an isometry, Rα is
clearly neither weakly mixing nor elastic.
Our third example emphasizes just how weak the totally transitive
hypothesis is. Let G be an infinite topological group which is simple
as an abstract group, for instance G = SO(3, R). The action of G on
itself by left-translation is obviously transitive, and is neither weakly
mixing nor elastic. Moreover, it is totally transitive since G has no
nontrivial finite index subgroups. Indeed, recall that if G had a finite
index subgroup H, with left transversal {g1 , g2 , . . . , gl }, then its core,
Core (H) = ∩li=1 gi Hgi−1 , is the largest normal subgroup of G that is
contained in H. Each of the conjugates gi Hgi−1 has the same index in G
as H and, since the intersection of finitely many finite index subgroups
has finite index (by Poincaré lemma [36]), Core (H) would also have
finite index. Thus, G would have a finite index normal subgroup,
contradicting the assumption that G is simple.

Example 3. Elastic ⇒ weakly mixing. Let G be the group of all

orientation preserving homeomorphisms of R. It is clear that G is
elastic, but it is not weakly mixing, since no element of G can reverse
the order of two disjoint subintervals.

Example 4. Weakly mixing ⇒ elastic. For n > 2 the linear

action of SL(n, Z) on Rn is weakly mixing [13], but it is not elastic.
Indeed, arguing as in Example 2, consider a small disc U centered at
the origin and small discs V1 , . . . , Vn centered on the standard basis
vectors e1 , . . . , en . Because SL(n, Z) preserves volume, there is no
matrix A ∈ SL(n, Z) for which the AU ∩ Vi are all nonempty.

Example 5. Weakly mixing + elastic ⇒ k-transitive for all k. Let

G be the group of homeomorphisms of R. It is clear that G is elastic
and weakly mixing, but it is not 3-transitive, since G cannot perform
all permutations of three disjoint subintervals.
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 383

Example 6. k-transitive for all k ⇒ strongly mixing. It is easy to

see that the group of homeomorphisms of Rn for n ≥ 2, or in fact any
manifold of dimension n ≥ 2, is k-transitive for all k but not strongly
mixing. For more examples, see subsection 4.3 and Section 6.

Example 7. Strongly mixing examples. Every infinite group G

has a strongly mixing action on a compact metric space. Indeed, let
M = {0, 1}G be equipped with the product topology. The natural
action of G on {0, 1}G is given by

g(f )(x) = f (g −1 x),

for all g, x ∈ G and f : G → {0, 1}. It is well known and not difficult
to see that this action is strongly mixing [24].

4. Inheritance of notions.

4.1 Inheritance under semi-conjugacy. Recall that if G acts on two

topological spaces M1 and M2 , a continuous map f : M1 → M2
is called a semi-conjugacy if f is surjective and G-equivariant, i.e.,
gf (x) = f (gx) for all g ∈ G, x ∈ M1 . If f is a semi-conjugacy, then it
is obvious that if the action of G on M1 is strongly mixing, k-transitive
for all k, weakly mixing, elastic or transitive, then the action of G on
M2 also enjoys the same property. Of course, properties of the action
on M2 are not in general inherited by the action on M1 .

4.2 Inheritance from and by subgroups. Suppose that H is a subgroup

of G. We first consider what properties of the action of H are inherited
by G.

Proposition 4. If H is a subgroup of G, then

(a) H strongly mixing ⇒ G strongly mixing, provided H has finite
index in G,
(b) H k-transitive for all k ⇒ G k-transitive for all k,
(c) H weakly mixing ⇒ G weakly mixing,
(d) H elastic ⇒ G elastic,
384 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

(e) H totally transitive ⇒ G totally transitive,

(f) H transitive ⇒ G transitive.

Proof. Parts (b), (c), (d) and (f) are obvious. For (e), note that if
H is totally transitive, and K is a finite index subgroup of G, then
H ∩ K has finite index in H and so H ∩ K is transitive and thus K is
transitive.
For (a), choose a right transversal {g1 = id, g2 , . . . , gl } of H in G;
G = Hg1 ∪ · · · ∪ Hgl .
Let U and V be nonempty open subsets of M and consider the set

{g ∈ G; gU ∩ V = ∅} = {g ∈ Hgi ; gU ∩ V = ∅}
i=1,... ,l

= {hgi ; h ∈ H and h(gi U ) ∩ V = ∅}
i=1,... ,l

= {h ∈ H; h(gi U ) ∩ V = ∅}gi .
i=1,... ,l

As H is strongly mixing, the sets {h ∈ H; h(gi U ) ∩ V = ∅} are finite.

So {g ∈ G; gU ∩ V = ∅} is finite; thus G is strongly mixing.

We now consider what properties of the action of G are inherited

by H.

Proposition 5. If H is a finite index subgroup of G, then

(a) G strongly mixing ⇒ H strongly mixing,
(b) G k-transitive for all k ⇒ H k-transitive for all k,
(c) G weakly mixing ⇒ H weakly mixing,
(d) G elastic ⇒ H elastic,
(e) G totally transitive ⇒ H totally transitive,
(f) G transitive ⇒ H transitive.

Proof. Parts (a) and (e) are obvious; part (a) is given in [24]. Parts
(b) and (d) follow from the following:
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 385

Lemma 9. If G is kl-transitive, then every subgroup of index l is

k-transitive.

Proof. Consider a left transversal of the index l subgroup H in G:

{g1 , g2 , . . . , gl }. Let U1 , . . . , Uk and V1 , . . . , Vk be nonempty open sets.
If G is kl-transitive, there exists g ∈ G such that
gUj ∩ gi Vj = ∅,
for all i ∈ {1, . . . , l} and j ∈ {1, . . . , k}. Notice that g = gi h for some
gi and some h ∈ H. Then gi hUj ∩ gi Vj = ∅ implies hUj ∩ Vj = ∅.
Hence, H is k-transitive.

Parts (c) and (d) are well known. For (c), let G be the group
of homeomorphisms of R, and let H be the subgroup of orientation
preserving homeomorphisms; see Examples 3 and 4. For (f), let G be
the group R∗ of nonzero reals, acting by multiplication on the real line,
and let H be the positive reals R+∗ , see Example 1.

Remark 1. The assumption that H has finite index is not required in

part (a); it suffices to assume that H is infinite.

4.3 Inheritance when taking products. The inheritance of dynamical

properties under the taking of products is a traditional problem, see
[19, 24], [16, Section II.4]. It is obvious that if infinite groups G and
H act on spaces M and N , and if both actions are k-transitive for all
k, respectively weakly mixing, respectively elastic, respectively totally
transitive, respectively transitive, then the obvious product action of
G × H on M × N also has the same property. Notice however that,
except in the trivial case of actions on singleton sets, strong mixing is
never preserved by taking products. Indeed, for nonempty open sets
U1 , U2 ⊆ M, V1 , V2 ∈ N , one has:
{(g, h) ∈ G × H; ((g, h)(U1 × V1 )) ∩ (U2 × V2 ) = ∅}
= {g ∈ G; gU1 ∩ V1 = ∅}
× H ∪ G × {h ∈ H; hU2 ∩ V2 = ∅}.
Provided M or N has at least two elements, one of the last two sets is
infinite; so the action of G × H on M × N is not strongly mixing. Note
386 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

that this gives a simple means of constructing examples of actions that

are k-transitive for all k but not strongly mixing.
We now turn to actions on products in a different sense; we consider
actions of a group G on spaces M and N and we consider the diagonal
action of G on M × N defined by g(x, y) = (g(x), g(y)) for all (x, y) ∈
M × N . There are examples where G is weakly mixing on M and
N , but the diagonal action on M × N is not transitive [28]. Note
that the group of all orientation preserving homeomorphisms of R, see
Example 3, is elastic, but its diagonal action on R2 is not transitive.
On the other hand, the following proposition holds (parts (a) and (f)
were stated in [24] for compact spaces):

Proposition 6. Suppose that G acts on M and on a dense in itself

Hausdorff space N . If the action of G on M is strongly mixing, then:
(a) G strongly mixing on N ⇒ G strongly mixing on M × N ,
(b) G k-transitive for all k on N ⇒ G k-transitive for all k on M ×N ,
(c) G weakly mixing on N ⇒ G weakly mixing on M × N ,
(d) G elastic on N ⇒ G elastic on M × N ,
(e) G totally transitive on N ⇒ G totally transitive on M × N ,
(f) G transitive on N ⇒ G transitive on M × N .

Proof. Assume that the action of G on M is strongly mixing and

consider nonempty open sets U1 , U2 ⊆ M , V1 , V2 ∈ N .
(a) As G is strongly mixing on N , the sets {g ∈ G; gU1 ∩ U2 = ∅}
and {g ∈ G; gV1 ∩ V2 = ∅} are finite and so {g ∈ G; gU1 ∩ U2 =
∅ or gV1 ∩ V2 = ∅} is finite. Hence the diagonal action of G on M × N
is also strongly mixing.
(f) The set {g ∈ G; gU1 ∩ U2 = ∅} is finite and if G is transitive on
N , the set {g ∈ G; gV1 ∩ V2 = ∅} is infinite, by Lemma 3. So there
exists g ∈ G with gU1 ∩ U2 = ∅ and gV1 ∩ V2 = ∅. Hence, G acts
transitively on M × N .
(d) Consider nonempty open sets U, U1 , . . . , Uk ⊆ M , V, V1 , . . . , Vk ∈
N . As G is strongly mixing on M , the set ∪i {g ∈ G; gU ∩ Ui = ∅}
is finite while, by Lemma 4, the set {g ∈ G; gV ∩ Vi = ∅ for all i} is
infinite. Thus G is elastic on M × N .
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 387

(e) Suppose that G is totally transitive on N , and let H be a

finite index subgroup of G. Then H is strongly mixing on M , by
Proposition 5, and H is transitive on N ; so by (f), H is transitive on
M × N . Thus, G is totally transitive on M × N .
(b), (c) Suppose the action of G on N is k-transitive, i.e., G is
transitive on N k . As G is strongly mixing on M , we have from (a)
that G is strongly mixing on M k . Thus, by (f), G is transitive on
(M × N )k , i.e., G is k-transitive on M × N .

Note that the above proposition gives examples which show that
Proposition 4 (a) fails without the finite index hypothesis. Indeed,
suppose that G is strongly mixing M . Then G2 is not strongly mixing
on M 2 , as we remarked above, but the diagonal action of G is strongly
mixing on M 2 by Proposition 6(a). That is, on M 2 , the diagonal
subgroup {(g, g); g ∈ G} ≤ G2 is strongly mixing, but the group G2
isn’t.

Remark 2. If G is k-transitive for all k on M , then obviously the

diagonal action of G on M 2 is also k-transitive for all k. More precisely,
if G is nk-transitive on M , then the action of G on M n is k-transitive.

4.4 Inheritance when passing to hyperspace. Let (M, d) be a metric

space. For each x ∈ M and ε > 0, let Bε (x) denote the open ball of
radius ε centered at x, and for each K ⊆ M and ε > 0, let Dε (K)
denote the union of all Bε (x) with x ∈ K. Let K(M ) be the set of
all nonempty compact subsets of M . For each pair K, L ∈ K(M ), the
Hausdorff distance from K to L is
h(K, L) = inf{ε; K ⊆ Dε (L), L ⊆ Dε (K)}.
It is easy to verify that (K(M ), h) is a metric space, called the hy-
perspace of (M, d). If (M, d) is connected, compact, or complete then
(K(M ), h) has the corresponding property, see [22, 31].
Let f : M → M be a continuous function, and let fˆ : K(M ) → K(M )
be the image function which takes A ∈ K(M ) to f (A). Using a
modification of the proof of Heine’s theorem on uniform continuity
[4, Theorem 4-24], it is not difficult to see that fˆ is continuous [31,
Corollary 4.8]. (For more information about the properties of fˆ, see
388 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

[14]). Let Φ be a continuous action of a group G on a metric space

(M, d), and let Φ̂ be the induced image action of G on K(M ). The
subspace of (K(M ), h) consisting of the singleton subsets of M is an
isometric copy (M, d), and the action of G on this subspace is a copy
of the continuous action Φ.
It is easy to see that in (K(M ), h), the open ball of radius ε centered
at K is

Bε (K) = {L ∈ K(M ); L ⊆ Dε (K), K ⊆ Dε (L)}.

In particular, if L ⊆ K, then L ∈ Bε (K) if and only if K ⊆ Dε (L).

Notice that for each K ∈ K(M ) and ε > 0 there is a finite (therefore
compact) subset Aε ⊆ K such that K ⊆ Dε (Aε ), that is, such that
Aε ∈ Bε (K). If x ∈ M , the open ball Bε ({x}) in K(M ) is

{A ∈ K(M ); A ⊆ Dε ({x}), {x} ⊆ Dε (A)}

= {A ∈ K(M ); A ⊆ Dε ({x})},

that is, just the nonempty compact subsets contained in the open ball
Bε (x) of (M, d). Similarly, for points x1 , . . . , xn ∈ M , the open ball
Bε ({x1 , . . . , xn }) in K(M ) is

{A ∈ K(M ); A ⊆ Dε ({x1 , . . . , xn }), {x1 , . . . , xn } ⊆ Dε (A)}

  
= A ∈ K(M ); A ⊆ Bε (xi ), xi ∈ Dε (A), ∀ i ,
i
  
= A ∈ K(M ); A ⊆ Bε (xi ), A ∩ Bε (xi ) = ∅, ∀ i .
i

The following result generalizes results for the case of a single map
in [7, 37], see [17, 21, 28] for related results.
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 389

Proposition 7. On second countable complete metric spaces, one

has the following implications:

Φ strongly mixing ⇐⇒ Φ̂ strongly mixing

Φ k-transitive for all k ⇐⇒ Φ̂ k-transitive for all k


Φ̂ weakly mixing Φ̂ elastic


Φ̂ totally transitive

Φ̂ transitive

Φ weakly mixing Φ elastic


Φ totally transitive

Φ transitive

Proof. It is obvious that the properties of Φ̂ are passed onto Φ. So in

view of Theorem 1, it remains to establish 3 things:
Φ strongly mixing ⇒ Φ strongly mixing. Consider a pair of open
balls Bε (A), Bε (B) of (K(M ), h). Cover A by n balls Bε/2 (xi ) of
(M, d) with S = {x1 , . . . , xn } ⊆ A and B by n balls Bε/2 (yi ) with
T = {y1 , . . . , yn } ⊆ B. Since Φ is strongly mixing, for each i the set
390 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

of g such that gBε/2 (xi ) ∩ Bε/2 (yi ) = ∅ is finite. Therefore, all but
finitely many g ∈ G satisfy gBε/2 (xi ) ∩ B(ε/2) (yi ) = ∅, i ∈ {1, . . . , n}.
Consider one such g. For i ∈ {1, . . . , n} let xi ∈ Bε/2 (xi ) and yi =
gxi ∈ Bε/2 (yi ), and put S  = {x1 , . . . , xn }, T  = gS  = {y1 , . . . , yn }.
Then
ε ε
h(A, S  ) ≤ h(A, S) + h(S, S  ) < + = ε,
2 2
and, likewise, h(B, T  ) < ε. Since T  = gS  , we have gBε (A) ∩ Bε (B) =
∅.
Φ k-transitive for all k ⇒ Φ̂ k-transitive for all k. Consider 2k
balls of (K(M ), h): Bε (Ai ), Bε (Bi ), i ∈ {1, . . . , k}. Cover each Ai and
Bi by n balls of (M, d), all of radius ε/2 and with centers xij ∈ Ai ,
yij ∈ Bi . Since Φ is kn-transitive, some g ∈ G satisfies gBε/2 (xij ) ∩
Bε/2 (yij ) = ∅, for all i ∈ {1, . . . , k}, j ∈ {1, . . . , n}. Arguing as in the
previous paragraph, this g also satisfies gBε (Ai ) ∩ Bε (Bi ) = ∅, for all
i ∈ {1, . . . , k}.
Φ̂ transitive ⇒ Φ elastic. Suppose Bε (x), Bε (x1 ), . . . , Bε (xn ) are
open balls of (M, d). We must find a g such that gB( x) ∩ Bε (xi ) =
∅, for all i ∈ {1, . . . , n}. Since Φ̂ is transitive there is a g such
that gBε ({x}) ∩ Bε ({x1 , . . . , xn }) = ∅ in K(M ). That is, from the
discussion immediately before the proposition, there is A ∈ K(M ) with
A ⊆ Bε (x), gA ⊆ ∪i Bε (xi ) and gA ∩ Bε (xi ) = ∅, for all i ∈ {1, . . . , n}.
But this also means gBε (x) ∩ Bε (xi ) = ∅, for all i ∈ {1, . . . , n}, as
required.

5. Actions of abelian groups. Furstenberg showed that for

flows, weakly mixing implies k-transitive for all k [19], see also [16,
Proposition II.4.12]. Petersen showed that for actions of abelian groups,
elastic implies weakly mixing [35]. Combining these ideas, one obtains:

Theorem 2. For actions of abelian groups, one has the following

implications:
⎧ ⎫
⎨ k-transitive for all k ⎬
strongly totally
=⇒ weakly mixing =⇒ =⇒ transitive,
mixing ⎩ ⎭ transitive
elastic

where the conditions inside the braces are equivalent.

 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 391

Remark 3. Notice that we do not impose the “second countable Baire”

hypothesis, as we did for Theorem 1. Indeed, the “second countable
Baire” hypothesis was only used in the proof of Theorem 1 to show
that weak mixing implies total transitivity, and we don’t require this
here.

Proof. Given Theorem 1, we need only establish two things:

Weakly mixing ⇒ k-transitive for all k. Consider k pairs of nonempty
open sets Ui , Vi , i ∈ {1, . . . , k}. If the action is weak mixing, there is
some g2 ∈ G such that

g2 U2 ∩ U1 = ∅ and g2 V2 ∩ V1 = ∅.

Next, there is some g3 such that

g3 U3 ∩ (g2 U2 ∩ U1 ) = ∅ and g3 V3 ∩ (g2 V2 ∩ V1 ) = ∅,

and so on, up to k. Let

U = gk Uk ∩ · · · ∩ g2 U2 ∩ U1 = ∅,

and define V similarly. There is some g ∈ G such that gU ∩ V = ∅.

Setting g1 = id, we have for all i ∈ {1, . . . , k},

ggi Ui ∩ gi Vi = gi gUi ∩ gi Vi = gi (gUi ∩ Vi ) = ∅.

Thus gUi ∩ Vi = ∅, for all i. So the action is k-transitive for all k.

Elastic ⇒ weakly mixing. Consider nonempty open sets U1 , U2 , V1 , V2 .
By elasticity, there is some g1 such that

g1 U1 ∩ U2 = ∅ and g1 U1 ∩ V2 = ∅,

and next, there is some g such that

g(U1 ∩ g1−1 U2 ) ∩ V1 = ∅ and g(U1 ∩ g1−1 U2 ) ∩ (U1 ∩ g1−1 V2 ) = ∅.

Then gU1 ∩ V1 = ∅ and

gg1−1 U2 ∩ g1−1 V2 = g1−1 gU2 ∩ g1−1 V2 = g1−1 (gU2 ∩ V2 ) = ∅,

392 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

which gives gU2 ∩ V2 = ∅. Therefore Φ is weakly mixing.

This completes the proof of Theorem 2.

As discussed at the beginning of subsection 4.3, if G is strongly mixing

on M , then the action of G × G on M 2 is k-transitive for all k, but not
strongly mixing; in particular, there are such actions of abelian groups.
An action of an abelian group which is totally transitive but not
weakly mixing was given in Example 2. The group in Example 1 (whose
action is transitive but not totally transitive) is abelian.
We now turn to the inheritance properties. In Proposition 5 we saw
that if H is a finite index subgroup of G, then G weakly mixing doesn’t
necessarily imply H weakly mixing. However, this does hold if G is
abelian, as was observed in [24]. This can be expressed in a somewhat
more useful way: if G has an abelian subgroup A whose induced action
is weakly mixing, then every finite index subgroup H of G is weakly
mixing. Indeed, if A is abelian and weakly mixing, then by Theorem 2,
A is elastic, so G is elastic by Proposition 4, and thus H is elastic, by
Proposition 5, and H is weakly mixing, again by Theorem 2.
For products, notice that from Theorem 2 and Remark 2, if the action
of an abelian group G is weakly mixing on M , then the action of G on
M 2 is also weakly mixing; this was proved in [24].
For the induced action on hyperspace, the next result follows imme-
diately from Theorem 2 and Proposition 7:

Proposition 8. For actions of abelian groups on metric spaces, one

has the following implications:
Φ strongly mixing, Φ̂ strongly mixing
⇓
 
Φ̂k-transitive ∀ k, Φ̂ elastic, Φ̂ weakly mixing, Φ̂ totally transitive
Φk-transitive ∀ k, Φ elastic, Φ weakly mixing, Φ̂ transitive
⇓
Φ totally transitive
⇓
Φ transitive,
where the conditions inside the braces are equivalent.
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 393

6. Chaotic actions.

Definition 5. The action of G on M is chaotic if it is transitive and

the set of points in M whose orbit under G is finite is a dense subset
of M .

A group G has a faithful chaotic action on some Hausdorff topological

space M if and only if G is residually finite [11]. (Curiously, it is also
true that a finitely generated group is residually finite if and only if it is
the group of isometries of a metric compactum [18]). Chaotic actions
enjoy the usual “sensitivity to initial conditions” property [11]. Every
compact triangulable manifold of dimension greater than 1 admits a
weakly mixing chaotic action of every countably generated free group
[12].
The following result is a generalization of a result by Peter Stacey for
single maps, see [5].

Theorem 3. For chaotic actions on second countable Baires, one

has:
⎧ ⎫
⎪ k-transitive for all k ⎪
⎨ ⎬
weakly mixing
strongly mixing =⇒ =⇒ transitive,
⎪
⎩ elastic ⎪
⎭
totally transitive

where the conditions inside the braces are equivalent.

Proof. In view of Theorem 1, it suffices to show that if the action of

G is totally transitive and chaotic, then G is k-transitive for all k. We
argue by induction: we assume that G is totally transitive, chaotic and
k-transitive for some k ≥ 1, and we will show that G is (k+1)-transitive.
Let U1 , . . . , Uk+1 and V1 , . . . , Vk+1 be nonempty open sets of M . By
the inductive hypothesis, there exists g ∈ G such that gUi ∩ Vi = ∅
for all 1 ≤ i ≤ k. As G is chaotic, for each 1 ≤ i ≤ k there exists
xi ∈ Ui ∩ g −1 Vi such that Gxi is finite. Consider the intersection H of
the stabilizer subgroups of the points xi :

H = {h ∈ G; h(xi ) = xi , for all 1 ≤ i ≤ k}.

394 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

H has finite index and so, as G is totally transitive, there exists h ∈ H

such that hUk+1 ∩ g −1 Vk+1 = ∅. Then ghUk+1 ∩ Vk+1 = ∅, and
ghUi ∩ Vi = ∅ for all 1 ≤ i ≤ k since ghxi = gxi ∈ Vi .

Example 8. Consider the linear action of SL(n, Z) on the torus

Tn = Rn /Zn , for n ≥ 2. Recall that a subgroup generated by a
hyperbolic matrix is weakly mixing on Tn , see [19]. Moreover, for
each m ∈ N, the image in Tn of the points of the form (1/m)x, where
x ∈ Zn , is a finite SL(n, Z)-invariant set. It follows that the action
of SL(n, Z) on Tn is chaotic. Thus, by Theorem 3, the action is k-
transitive for all k. However, the action is clearly not strongly mixing.
(A totally transitive chaotic function which is not strongly mixing was
given in [15]).
Now let M denote the disjoint union of two copies of Tn , and let G
be the direct product of SL(n, Z) and the two element group {id, τ }.
There is an obvious action of G on M ; SL(n, Z) acts linearly on
each connected component of M and τ is the homeomorphism that
interchanges the two components. This action is clearly transitive (and
chaotic), but it is not totally transitive.

Turning now to the inheritance properties, it is immediate from

Proposition 5 and Theorem 3 that finite index subgroups of chaotic
weakly mixing groups are also chaotic weakly mixing. In particular, by
Theorem 3, if G is chaotic and weakly mixing on M , then for every point
x ∈ M with finite orbit, the action on M of the stabilizer subgroup
Gx = {g ∈ G; gx = x} is k-transitive for all k.
For products, notice that, from Theorem 3 and Remark 2, if G is
totally transitive and chaotic on M , then the action of G on M 2 is also
totally transitive and chaotic. Notice however, that for a chaotic action
of G on M , the action of G on M 2 may fail to be chaotic; indeed, if the
action of G on M is chaotic but not totally transitive, as in Example 8,
then G is not weakly mixing on M by Theorem 3, and so G is not
transitive on M 2 .
For the induced action on hyperspace, the next result follows imme-
diately from Theorem 3 and Proposition 7:
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 395

Proposition 9. For chaotic actions on second countable complete

metric spaces, one has:
 
Φ strongly mixing, Φ̂ strongly mixing
⇓
⎧ ⎫
⎨ Φ̂k-transitive ∀ k, Φ̂ elastic, Φ̂ weakly mixing ⎬
Φk-transitive ∀ k, Φ elastic, Φ weakly mixing
⎩ ⎭
Φ̂ totally transitive, Φ̂ transitive, Φ totally transitive
⇓
Φ transitive
where the conditions inside the braces are equivalent.

Remark 4. We do not know whether the second countable Baire

hypothesis is necessary in Theorems 1 and 3 and in Proposition 9.

REFERENCES

1. E. Akin, The general topology of dynamical systems, Grad. Studies in Math.,

Amer. Math. Soc., Providence, RI, 1993.
2. E. Akin, J. Auslander and K. Berg, When is a transitive map chaotic? in
Convergence in ergodic theory and probability (V. Bergelson, P. March and J.
Rosenblatt, eds.), (Columbus, OH, 1993), de Gruyter, Berlin, 1996, pp. 25 40.
3. Ll. Alsedà, M.A. del Rı́o and J.A. Rodrı́guez, A note on the totally transitive
graph maps, Internat. J. Bifur. Chaos Appl. Sci. Engrg. 11 (2001), 841 843.
4. T.M. Apostol, Mathematical analysis: A modern approach to advanced calcu-
lus, Addison-Wesley Publ., Reading, MA, 1957.
5. J. Banks, Regular periodic decompositions for topologically transitive maps,
Ergodic Theory Dynam. Systems 17 (1997), 505 529.
6. , Topological mapping properties defined by digraphs, Discrete Contin.
Dyn. Syst. 5 (1999), 83 92.
7. , Chaos in hyperspace, La Trobe University Math. Research Paper 95-14,
1995.
8. F. Blanchard, B. Host and A. Maass, Topological complexity, Ergodic Theory
Dynam. Systems 20 (2000), 641 662.
9. N. Bourbaki, General topology, Chapters 5 10, Springer-Verlag, Berlin, 1989
(reprint of the 1966 ed.).
10. O. Bratteli, G.A. Elliott and D.W. Robinson, Strong topological transitivity
and C ∗ -dynamical systems, J. Math. Soc. Japan 37 (1985), 115 133.
11. G. Cairns, G. Davis, D. Elton, A. Kolganova and P. Perversi, Chaotic group
actions, Enseign. Math. 41 (1995), 123 133.
396 G. CAIRNS, A. KOLGANOVA AND A. NIELSEN

12. G. Cairns and A. Kolganova, Chaotic actions of free groups, Nonlinearity 9

(1996), 1015 1021.
13. G. Cairns and A. Nielsen, On the dynamics of the linear action of SL(n, Z)
on Rn , Bull. Austral. Math. Soc. 71 (2005), 359 365.
14. J.J. Charatonik, Recent results on induced mappings between hyperspaces of
continua, Topology Proc. 22 (1997), 103 122.
15. A. Crannell, A chaotic, non-mixing subshift, Discrete Contin. Dynam. Sys-
tems, added vol. I (1998), 195 202.
16. J. de Vries, Elements of topological dynamics, in Mathematics and its
applications, Kluwer Acad. Publ., Dordrecht, 1993.
17. A. Edalat, Dynamical systems, measures, and fractals via domain theory,
Inform. and Comput. 120 (1995), 32 48.
18. A.V. Egorov, Residual finiteness of groups, and topological dynamics, Mat.
Sb. 191 (2000), 53 66.
19. H. Furstenberg, Disjointness in ergodic theory, minimal sets, and a problem
in Diophantine approximation, Math. Systems Theory 1 (1967), 1 49.
20. W.H. Gottschalk and G.A. Hedlund, Topological dynamics, Amer. Math. Soc.
Colloq. Publ., vol. 36, Amer. Math. Soc., Providence, RI, 1955.
21. H. Haase, Chaotic maps in hyperspaces, Real Anal. Exchange 21 (1995/96),
689 695.
22. A. Illanes and S.B. Nadler, Jr., Hyperspaces, Marcel Dekker, New York, 1999.
23. A. Kameyama, Topological transitivity and strong transitivity, Acta Math.
Univ. Comenian. (N.S.) 71 (2002), 139 145.
24. H.B. Keynes and J.B. Robertson, On ergodicity and mixing in topological
transformation groups, Duke Math. J. 35 (1968), 809 819.
25. S. Kolyada and L. Snoha, Some aspects of topological transitivity a survey,
in Iteration theory (Ludwig Reich, Jaroslav Smı́tal and György Targonski, eds.),
(ECIT 94, Opava, Czech. Repub.) Grazer Math. Ber., vol. 334, Karl-Franzens-Univ.
Graz, Graz, 1997, pp. 3 35.
26. S.C. Koo, Recursive properties of transformation groups in hyperspaces,
Math. Systems Theory 9 (1975), 75 82.
27. Kazimierz Kuratowski and Andrzej Mostowski, Set theory, 2nd ed., North
Holland Publ., Amsterdam, 1976.
28. K. Lau and A. Zame, On weak mixing of cascades, Math. Systems Theory 6
(1972/73), 307 311.
29. R. Mañé, Ergodic theory and differentiable dynamics, Ergeb. Math. Gren-
zgeb., Springer-Verlag, Berlin, 1987.
30. G. Manzini and L. Margara, A complete and efficiently computable topolog-
ical classification of D-dimensional linear cellular automata over Zm , Theoret.
Comput. Sci. 221 (1999), 157 177.
31. E. Michael, Topologies on spaces of subsets, Trans. Amer. Math. Soc. 71
(1951), 152 182.
32. J.C. Morgan, II, Point set theory, Marcel Dekker, New York, 1990.
 TOPOLOGICAL TRANSITIVITY AND MIXING NOTIONS 397

33. J.C. Oxtoby, Measure and category, 2nd ed., Grad. Texts in Math., Springer-
Verlag, New York, 1980.
34. J. Peters and T. Pennings, Chaotic extensions of dynamical systems by
function algebras, J. Math. Anal. Appl. 159 (1991), 345 360.
35. K.E. Petersen, Disjointness and weak mixing of minimal sets, Proc. Amer.
Math. Soc. 24 (1970), 278 280.
36. D.J.S. Robinson, A course in the theory of groups, Springer-Verlag, New
York, 1993.
37. H. Román-Flores, A note on transitivity in set-valued discrete systems, Chaos
Solitons Fractals 17 (2003), 99 104.

E-mail address: G.Cairns@latrobe.edu.au

E-mail address: Alla.Kolganova@vanguard.com.au

E-mail address: Anthony.Nielsen@latrobe.edu.au