Z d Actions on the Cantor Set: Approximation, Rohlin Properties and - - PowerPoint PPT Presentation

z d actions on the cantor set approximation rohlin
SMART_READER_LITE
LIVE PREVIEW

Z d Actions on the Cantor Set: Approximation, Rohlin Properties and - - PowerPoint PPT Presentation

Z d Actions on the Cantor Set: Approximation, Rohlin Properties and Recursion Theory Michael Hochman Princeton University hochman @ math.princeton.edu AMS joint meeting, Washington DC, January 2009 Z d actions We want to understand the


slide-1
SLIDE 1

Zd Actions on the Cantor Set: Approximation, Rohlin Properties and Recursion Theory

Michael Hochman

Princeton University hochman@math.princeton.edu

AMS joint meeting, Washington DC, January 2009

slide-2
SLIDE 2

Zd actions

We want to understand the distribution of different types of dynamics in the space of actions.

slide-3
SLIDE 3

Zd actions

We want to understand the distribution of different types of dynamics in the space of actions. Denote the Cantor set by K = {0, 1}ℵ0 and the group of homeomorphisms of K by G = Homeo(K) This group is Polish in the topology of uniform convergence.

slide-4
SLIDE 4

The space of actions

The space of Zd actions on K is Ad = hom(Zd, G) ⊆ GZd and also Polish. Two actions are close if their generators are close in G.

slide-5
SLIDE 5

The space of actions

The space of Zd actions on K is Ad = hom(Zd, G) ⊆ GZd and also Polish. Two actions are close if their generators are close in G. G acts on Ad by conjugation: for ϕ = {ϕu}u∈Zd ∈ Ad and g ∈ G, the conjugation ϕg is the action (ϕg)u = g ◦ fu ◦ g−1 The orbit of ϕ under G is its conjugacy class.

slide-6
SLIDE 6

Classical results

Classical work has been done on the space of measure preserving automorphisms of a probability space.

slide-7
SLIDE 7

Classical results

Classical work has been done on the space of measure preserving automorphisms of a probability space.

Theorem (Rohlin / del Junco)

Let f be an aperiodic automorphism of a Lebesgue space. Then the conjugacy class of f in the group of automorphisms is dense, but it is meager.

Interpretation

At any finite resolution all aperiodic dynamics look the same, but no conjugacy class is too large. This is true also for measure-preserving actions of Zd.

slide-8
SLIDE 8

Background: The topological case in dimension d = 1

Theorem (Glasner-Weiss, 96)

There exists a Z-action f ∈ A1 with dense conjugacy class (weak Rohlin property). Remark: not every action f ∈ A1 has this property, with or without periodic points.

slide-9
SLIDE 9

Background: The topological case in dimension d = 1

Theorem (Glasner-Weiss, 96)

There exists a Z-action f ∈ A1 with dense conjugacy class (weak Rohlin property). Remark: not every action f ∈ A1 has this property, with or without periodic points.

Theorem (Kechris-Rosendal, 05)

There exists a homeomorphism f ∈ A1 with co-meager conjugacy class (strong Rohlin property).

slide-10
SLIDE 10

Background: The topological case in dimension d = 1

Theorem (Glasner-Weiss, 96)

There exists a Z-action f ∈ A1 with dense conjugacy class (weak Rohlin property). Remark: not every action f ∈ A1 has this property, with or without periodic points.

Theorem (Kechris-Rosendal, 05)

There exists a homeomorphism f ∈ A1 with co-meager conjugacy class (strong Rohlin property). What happens in d ≥ 2?

slide-11
SLIDE 11

Results

Theorem (H)

For d ≥ 2,

  • 1. Every f ∈ Ad has meager conjugacy class,
  • 2. There exists an f ∈ Ad with dense conjugacy class.

Thus Zd has the weak, but not the strong, Rohlin property.

slide-12
SLIDE 12

Results

Theorem (H)

For d ≥ 2,

  • 1. Every f ∈ Ad has meager conjugacy class,
  • 2. There exists an f ∈ Ad with dense conjugacy class.

Thus Zd has the weak, but not the strong, Rohlin property. Put another way, for d ≥ 2 the conjugation action of G on Ad is topologically transitive (i.e. it has dense orbits). From general considerations, the actions in Ad with dense conjugacy class form a residual set.

slide-13
SLIDE 13

Effective actions

Let

◮ m ∈ N, ◮ Ci a cylinder set in K, ◮ ui ∈ Zd.

The m-tuple (C1, u1), . . . , (Cm, um) is disjoint for an action f ∈ Ad if

m

  • i=1

fui(Ci) = ∅

slide-14
SLIDE 14

Effective actions

Let

◮ m ∈ N, ◮ Ci a cylinder set in K, ◮ ui ∈ Zd.

The m-tuple (C1, u1), . . . , (Cm, um) is disjoint for an action f ∈ Ad if

m

  • i=1

fui(Ci) = ∅ An action f ∈ Ad is effective if the family of disjoint tuples for f is recursively enumerable.

slide-15
SLIDE 15

The conjugation action

Theorem (H)

If d ≥ 2 and f ∈ Ad has dense conjugacy class, then f is not effective.

slide-16
SLIDE 16

The conjugation action

Theorem (H)

If d ≥ 2 and f ∈ Ad has dense conjugacy class, then f is not effective. In other words, although the conjugation action is topologically transitive, one cannot construct a transitive point.

slide-17
SLIDE 17

The conjugation action

Theorem (H)

If d ≥ 2 and f ∈ Ad has dense conjugacy class, then f is not effective. In other words, although the conjugation action is topologically transitive, one cannot construct a transitive point. Note: There does exist a dense family of effective systems in Ad for any d.

slide-18
SLIDE 18

Approximation of minimal actions

An action f ∈ Ad is minimal if every point x ∈ K has a dense f-orbit

slide-19
SLIDE 19

Approximation of minimal actions

An action f ∈ Ad is minimal if every point x ∈ K has a dense f-orbit The subspace Md ⊆ Ad consisting of minimal actions is Polish.

slide-20
SLIDE 20

Approximation of minimal actions

An action f ∈ Ad is minimal if every point x ∈ K has a dense f-orbit The subspace Md ⊆ Ad consisting of minimal actions is Polish.

Theorem

The effective actions are dense in M1.

slide-21
SLIDE 21

Approximation of minimal actions

An action f ∈ Ad is minimal if every point x ∈ K has a dense f-orbit The subspace Md ⊆ Ad consisting of minimal actions is Polish.

Theorem

The effective actions are dense in M1.

Theorem (H)

For d ≥ 2, the effective actions are nowhere dense in Md. .

slide-22
SLIDE 22

Shifts of finite type

A shift of finite type (SFT) is the subset of the form XL ⊆ {1, . . . , r}Zd consisting of all configurations omiting a finite set L of finite patterns. An SFT is effective and invariant under the Zd shift action S of translating configurations.

slide-23
SLIDE 23

Stability of shifts of finite type in the space of actions

Let X be an SFT and Y ⊆ X a closed, shift invariant subset without isolated points. Let f ∈ Ad be an action isomorphic to (Y, S|Y).

Proposition

For all g ∈ Ad sufficiently close to f, there is a factor map π : (K, g) → (Z, S|Z) for some closed subshift Z ⊆ X.

slide-24
SLIDE 24

Some proofs

Proposition (H)

An effective minimal subshift has Medvedev degree 0.

Theorem

For d ≥ 2, Md is nowhere dense in Ad.

Proof.

It suffices to show that there is some open subset of Md which does not contain effective minimal actions.

slide-25
SLIDE 25

Some proofs

Proposition (H)

An effective minimal subshift has Medvedev degree 0.

Theorem

For d ≥ 2, Md is nowhere dense in Ad.

Proof.

It suffices to show that there is some open subset of Md which does not contain effective minimal actions. Using [Simpson 08], let X be an SFT with non-trivial Medvedev degree. By Zorns lemma, there is a minimal subsystem X0 ⊆ X. It cannot have isolated points.

slide-26
SLIDE 26

Proof (continued)

Let g ∈ A2 be isomorphic to S|X0. If f is an effective minimal system sufficiently close to g, then (K, f) factors into (X, S|X), and the image is minimal and effective. But this contradicts the fact that X has non-trivial Medvedev degree.

slide-27
SLIDE 27

Open Problems

  • 1. Are there dynamical (rather than recursive) obstructions to

approximationof minimal actions by minimal effective ones (or by minimal SFTs)?

  • 2. Can every strongly irreducible action be approximated by

strongly irreducible effective actions?