SLIDE 1
Random elements of large groups: Discrete case Zoltn Vidnynszky - - PowerPoint PPT Presentation
Random elements of large groups: Discrete case Zoltn Vidnynszky - - PowerPoint PPT Presentation
Random elements of large groups: Discrete case Zoltn Vidnynszky Alfrd Rnyi Institute of Mathematics Toposym 2016 joint work with Udayan Darji, Mrton Elekes, Kende Kalina, Viktor Kiss The random graph, R = N , E R Edges: for
SLIDE 2
SLIDE 3
The random graph, R = N, ER Edges: for n, m ∈ N distinct let P((n, m) ∈ ER) = 1
2,
independently. Almost surely we obtain the same graph. Equivalently: for every disjoint, finite A, B ⊂ N there exists v ∈ N such that (∀x ∈ A)((x, v) ∈ ER) and (∀y ∈ B)((y, v) ∈ ER). If X, Y ⊂ R are finite and f : X → Y is an isomorphism then f extends to an automorphism of R. Every countable graph can be embedded into (R, ER).
SLIDE 4
The random graph, R = N, ER Edges: for n, m ∈ N distinct let P((n, m) ∈ ER) = 1
2,
independently. Almost surely we obtain the same graph. Equivalently: for every disjoint, finite A, B ⊂ N there exists v ∈ N such that (∀x ∈ A)((x, v) ∈ ER) and (∀y ∈ B)((y, v) ∈ ER). If X, Y ⊂ R are finite and f : X → Y is an isomorphism then f extends to an automorphism of R. Every countable graph can be embedded into (R, ER). Q, < If X, Y ⊂ Q are finite and f : X → Y is order preserving then f extends to an order preserving Q → Q map. Every countable linearly ordered set can be order preservingly embedded to Q.
SLIDE 5
Automorphism groups and genericity
S∞ is a Polish group with the pointwise convergence topology.
SLIDE 6
Automorphism groups and genericity
S∞ is a Polish group with the pointwise convergence topology. We are interested in the automorphism groups of countable structures ⇐ ⇒ closed subgroups of S∞.
- Definition. A property P of elements of Aut(A) is said to hold
generically if the set {f ∈ Aut(A) : P(f)} is co-meagre.
- Definition. If f, g ∈ Aut(A) we say that f and g are conjugate,
if there exists an h ∈ Aut(A) such that h−1fh = g. Note: if f, g ∈ Aut(A) then A, f ∼ = A, g ⇐ ⇒ (∃h ∈ Aut(A))(h−1fh = g).
- Definition. An automorphism is called generic if its conjugacy
class is co-meagre.
SLIDE 7
Conjugacy classes
“There are no infinite cycles and there are infinitely many cycles for every finite cycle length” holds generically in S∞ and Aut(R),
SLIDE 8
Conjugacy classes
“There are no infinite cycles and there are infinitely many cycles for every finite cycle length” holds generically in S∞ and Aut(R), in particular, there is a generic element in S∞.
SLIDE 9
Conjugacy classes
“There are no infinite cycles and there are infinitely many cycles for every finite cycle length” holds generically in S∞ and Aut(R), in particular, there is a generic element in S∞. (Kuske, Truss) There exist generic elements in Aut(Q) and Aut(R).
SLIDE 10
Conjugacy classes
“There are no infinite cycles and there are infinitely many cycles for every finite cycle length” holds generically in S∞ and Aut(R), in particular, there is a generic element in S∞. (Kuske, Truss) There exist generic elements in Aut(Q) and Aut(R). Kechris, Rosendal: Characterisation of the existence of generic elements of closed subgroups of S∞.
SLIDE 11
Measure
- Definition. (Christensen) Let (G, ·) be a Polish group and
B ⊂ G Borel. We say that B is Haar null if there exists a Borel probability measure µ on G such that for every g, h ∈ G we have µ(gBh) = 0. An arbitrary set S is called Haar null if S ⊂ B for some Borel Haar null set B.
- Definition. A property P of elements of Aut(A) is said to hold
almost surely if the set {f ∈ Aut(A) : P(f)} is co-Haar null.
SLIDE 12
Measure
- Definition. (Christensen) Let (G, ·) be a Polish group and
B ⊂ G Borel. We say that B is Haar null if there exists a Borel probability measure µ on G such that for every g, h ∈ G we have µ(gBh) = 0. An arbitrary set S is called Haar null if S ⊂ B for some Borel Haar null set B.
- Definition. A property P of elements of Aut(A) is said to hold
almost surely if the set {f ∈ Aut(A) : P(f)} is co-Haar null.
- Definition. A ⊂ G is called compact catcher if for every K ⊂ G
compact there exist g, h ∈ G so that gKh ⊂ A.
SLIDE 13
Measure
- Definition. (Christensen) Let (G, ·) be a Polish group and
B ⊂ G Borel. We say that B is Haar null if there exists a Borel probability measure µ on G such that for every g, h ∈ G we have µ(gBh) = 0. An arbitrary set S is called Haar null if S ⊂ B for some Borel Haar null set B.
- Definition. A property P of elements of Aut(A) is said to hold
almost surely if the set {f ∈ Aut(A) : P(f)} is co-Haar null.
- Definition. A ⊂ G is called compact catcher if for every K ⊂ G
compact there exist g, h ∈ G so that gKh ⊂ A. A is compact biter if for every K ⊂ G compact there exist a U open and g, h ∈ G so that U ∩ K = ∅, and g(U ∩ K)h ⊂ A.
SLIDE 14
Measure
- Definition. (Christensen) Let (G, ·) be a Polish group and
B ⊂ G Borel. We say that B is Haar null if there exists a Borel probability measure µ on G such that for every g, h ∈ G we have µ(gBh) = 0. An arbitrary set S is called Haar null if S ⊂ B for some Borel Haar null set B.
- Definition. A property P of elements of Aut(A) is said to hold
almost surely if the set {f ∈ Aut(A) : P(f)} is co-Haar null.
- Definition. A ⊂ G is called compact catcher if for every K ⊂ G
compact there exist g, h ∈ G so that gKh ⊂ A. A is compact biter if for every K ⊂ G compact there exist a U open and g, h ∈ G so that U ∩ K = ∅, and g(U ∩ K)h ⊂ A.
- Corollary. If A is compact biter then it is not Haar null.
SLIDE 15
Measure in S∞
- Theorem. (Dougherty, Mycielski) Almost all elements of S∞
have infinitely many infinite cycles and only finitely many finite cycles.
SLIDE 16
Measure in S∞
- Theorem. (Dougherty, Mycielski) Almost all elements of S∞
have infinitely many infinite cycles and only finitely many finite cycles. Therefore, almost all permutations included in the union of countably many conjugacy classes.
SLIDE 17
Measure in S∞
- Theorem. (Dougherty, Mycielski) Almost all elements of S∞
have infinitely many infinite cycles and only finitely many finite cycles. Therefore, almost all permutations included in the union of countably many conjugacy classes.
- Theorem. (Dougherty, Mycielski) All of these classes are Haar
positive, in fact, compact biters.
SLIDE 18
Measure and countable structures
- Definition. Let A be a structure, a ∈ A and X ⊂ A. We say
that A has the nice algebraic closure property (NAC) if for every finite A ⊂ A the {b : |{f(b) : f ∈ Stabp(A)}| < ∞} is finite.
SLIDE 19
Measure and countable structures
- Definition. Let A be a structure, a ∈ A and X ⊂ A. We say
that A has the nice algebraic closure property (NAC) if for every finite A ⊂ A the {b : |{f(b) : f ∈ Stabp(A)}| < ∞} is finite.
- Theorem. Let A be a countable structure.
A has NAC ⇔ almost every element of Aut(A) has finitely many finite cycles,
SLIDE 20
Measure and countable structures
- Definition. Let A be a structure, a ∈ A and X ⊂ A. We say
that A has the nice algebraic closure property (NAC) if for every finite A ⊂ A the {b : |{f(b) : f ∈ Stabp(A)}| < ∞} is finite.
- Theorem. Let A be a countable structure.
A has NAC ⇔ almost every element of Aut(A) has finitely many finite cycles, A has NAC ⇒ almost every element of Aut(A) has infinitely many infinite cycles.
SLIDE 21
Measure and countable structures
- Definition. Let A be a structure, a ∈ A and X ⊂ A. We say
that A has the nice algebraic closure property (NAC) if for every finite A ⊂ A the {b : |{f(b) : f ∈ Stabp(A)}| < ∞} is finite.
- Theorem. Let A be a countable structure.
A has NAC ⇔ almost every element of Aut(A) has finitely many finite cycles, A has NAC ⇒ almost every element of Aut(A) has infinitely many infinite cycles. R, Q has NAC, but this is not enough to characterize the positive conjugacy classes of Aut(R), Aut(Q).
SLIDE 22
Measure and Aut(Q)
f ∈ Aut(Q) extends to a ¯ f ∈ Homeo+(R).
- Definition. A + orbital (− orbital) of f is a maximal interval
I ⊂ R such that for every x ∈ I we have ¯ f(x) > x ( ¯ f(x) < x). Let Fix( ¯ f) = {x ∈ R : ¯ f(x) = x}.
SLIDE 23
Measure and Aut(Q)
f ∈ Aut(Q) extends to a ¯ f ∈ Homeo+(R).
- Definition. A + orbital (− orbital) of f is a maximal interval
I ⊂ R such that for every x ∈ I we have ¯ f(x) > x ( ¯ f(x) < x). Let Fix( ¯ f) = {x ∈ R : ¯ f(x) = x}.
- Proposition. f, g ∈ Aut(Q) are conjugate if and only if there
exists an order and rationality preserving isomorphism between Fix( ¯ f) and Fix(¯ g) so that the corresponding orbitals have the same sign.
SLIDE 24
Measure and Aut(Q)
- Theorem. For almost every element of Aut(Q)
between every two + orbitals (− orbitals) there is a − orbital (+ orbital) or a rational fixed point
SLIDE 25
Measure and Aut(Q)
- Theorem. For almost every element of Aut(Q)
between every two + orbitals (− orbitals) there is a − orbital (+ orbital) or a rational fixed point there are only finitely many rational fixed points.
SLIDE 26
Measure and Aut(Q)
- Theorem. For almost every element of Aut(Q)
between every two + orbitals (− orbitals) there is a − orbital (+ orbital) or a rational fixed point there are only finitely many rational fixed points.
- Theorem. This characterises the positive conjugacy classes, in
fact, every positive conjugacy class is compact biter.
SLIDE 27
Measure and Aut(Q)
- Theorem. For almost every element of Aut(Q)
between every two + orbitals (− orbitals) there is a − orbital (+ orbital) or a rational fixed point there are only finitely many rational fixed points.
- Theorem. This characterises the positive conjugacy classes, in
fact, every positive conjugacy class is compact biter. In particular, there are c many Haar positive conjugacy classes, and their union is almost everything.
SLIDE 28
Measure and Aut(Q)
- Theorem. For almost every element of Aut(Q)
between every two + orbitals (− orbitals) there is a − orbital (+ orbital) or a rational fixed point there are only finitely many rational fixed points.
- Theorem. This characterises the positive conjugacy classes, in
fact, every positive conjugacy class is compact biter. In particular, there are c many Haar positive conjugacy classes, and their union is almost everything.
SLIDE 29
Measure and Aut(R)
- Definition. Let v ∈ R and f ∈ Aut(R). Define
βf,v : N+ → {0, 1} as βf,v(n) = 1 ⇐ ⇒ (v, fn(v)) ∈ ER.
SLIDE 30
Measure and Aut(R)
- Definition. Let v ∈ R and f ∈ Aut(R). Define
βf,v : N+ → {0, 1} as βf,v(n) = 1 ⇐ ⇒ (v, fn(v)) ∈ ER.
- Proposition. (Truss) Suppose that f, g ∈ Aut(R) have only one
infinite cycle and no finite ones. Then f and g are conjugate if and only if βf,v = βg,w for some ( ⇐ ⇒ for every) v, w.
SLIDE 31
Measure and Aut(R)
- Definition. Let v ∈ R and f ∈ Aut(R). Define
βf,v : N+ → {0, 1} as βf,v(n) = 1 ⇐ ⇒ (v, fn(v)) ∈ ER.
- Proposition. (Truss) Suppose that f, g ∈ Aut(R) have only one
infinite cycle and no finite ones. Then f and g are conjugate if and only if βf,v = βg,w for some ( ⇐ ⇒ for every) v, w. Truss’ characterisation has an appropriate generalisation to every f, g ∈ Aut(R).
SLIDE 32
Measure and Aut(R)
- Definition. Let v ∈ R and f ∈ Aut(R). Define
βf,v : N+ → {0, 1} as βf,v(n) = 1 ⇐ ⇒ (v, fn(v)) ∈ ER.
- Proposition. (Truss) Suppose that f, g ∈ Aut(R) have only one
infinite cycle and no finite ones. Then f and g are conjugate if and only if βf,v = βg,w for some ( ⇐ ⇒ for every) v, w. Truss’ characterisation has an appropriate generalisation to every f, g ∈ Aut(R).
SLIDE 33
Measure and Aut(R)
- Theorem. Almost all elements of Aut(R) have the following
properties: for every disjoint, finite A, B ⊂ N there exists v ∈ N such that (∀x ∈ A)((x, v) ∈ ER) and (∀y ∈ B)((y, v) ∈ ER)
SLIDE 34
Measure and Aut(R)
- Theorem. Almost all elements of Aut(R) have the following
properties: for every disjoint, finite A, B ⊂ N there exists v ∈ N such that (∀x ∈ A)((x, v) ∈ ER) and (∀y ∈ B)((y, v) ∈ ER) and v ∈ the union of cycles generated by A ∪ B,
SLIDE 35
Measure and Aut(R)
- Theorem. Almost all elements of Aut(R) have the following
properties: for every disjoint, finite A, B ⊂ N there exists v ∈ N such that (∀x ∈ A)((x, v) ∈ ER) and (∀y ∈ B)((y, v) ∈ ER) and v ∈ the union of cycles generated by A ∪ B, there are only finitely many finite cycles.
SLIDE 36
Measure and Aut(R)
- Theorem. Almost all elements of Aut(R) have the following
properties: for every disjoint, finite A, B ⊂ N there exists v ∈ N such that (∀x ∈ A)((x, v) ∈ ER) and (∀y ∈ B)((y, v) ∈ ER) and v ∈ the union of cycles generated by A ∪ B, there are only finitely many finite cycles.
- Theorem. This characterises the positive conjugacy classes, in
fact, every positive conjugacy class is compact biter.
SLIDE 37
Measure and Aut(R)
- Theorem. Almost all elements of Aut(R) have the following
properties: for every disjoint, finite A, B ⊂ N there exists v ∈ N such that (∀x ∈ A)((x, v) ∈ ER) and (∀y ∈ B)((y, v) ∈ ER) and v ∈ the union of cycles generated by A ∪ B, there are only finitely many finite cycles.
- Theorem. This characterises the positive conjugacy classes, in
fact, every positive conjugacy class is compact biter. Again, there are c many Haar positive conjugacy classes, and their union is almost everything.
SLIDE 38
Measure and Aut(R)
- Theorem. Almost all elements of Aut(R) have the following
properties: for every disjoint, finite A, B ⊂ N there exists v ∈ N such that (∀x ∈ A)((x, v) ∈ ER) and (∀y ∈ B)((y, v) ∈ ER) and v ∈ the union of cycles generated by A ∪ B, there are only finitely many finite cycles.
- Theorem. This characterises the positive conjugacy classes, in
fact, every positive conjugacy class is compact biter. Again, there are c many Haar positive conjugacy classes, and their union is almost everything. Splitting lemma. If F ⊂ Aut(R) is finite set there exists a vertex v so that for every f, g ∈ F distinct we have f(v) = g(v).
SLIDE 39
Measure and Aut(R)
- Theorem. (Christensen) If A is a conjugacy invariant Haar
positive universally measurable set then A−1A contains a neighbourhood of the identity.
- Corollary. (Truss) For every f, g ∈ Aut(R) non-identity
elements, g is the product of four conjugates of f.
SLIDE 40
Questions
- 1. How many Haar positive conjugacy classes are there?
- 2. Is the union of the Haar null conjugacy classes Haar null?
SLIDE 41
Examples
- f Haar null classes is Haar null
C LC \ C NLC
- n
Zn ? ? ℵ0 Mn Z S∞; Aut(Q)? c
- ?
- f Haar null classes is not Haar null
C LC \ C NLC 2ω Z × 2ω Zω n Zn × (Z2 ⋉ Zω
3 )
HNN ×(Z2 ⋉ Zω
3 )
Zn × (Z2 ⋉ Qω
d )
ℵ0 ? Z × (Z2 ⋉ Zω
3 )
S∞ × (Z2 ⋉ Zω
3 )
c
- Aut(Q)ω × (Z2 ⋉ Zω
3 )
SLIDE 42
Examples
- f Haar null classes is Haar null
C LC \ C NLC – – – n Zn HNN ??? ℵ0 ??? Z S∞ c – – Aut(Q); Aut(R)
- f Haar null classes is not Haar null
C LC \ C NLC 2ω Z × 2ω Zω n Zn × (Z2 ⋉ Zω
3 )
HNN ×(Z2 ⋉ Zω
3 )
Zn × (Z2 ⋉ Qω
d )
ℵ0 ??? Z × (Z2 ⋉ Zω
3 )
S∞ × (Z2 ⋉ Zω
3 )
c – – Aut(Q) × (Z2 ⋉ Zω
3 )
SLIDE 43
Open problems
- Question. Are there natural examples of automorphism groups
with given cardinality of Haar positive conjugacy classes?
SLIDE 44
Open problems
- Question. Are there natural examples of automorphism groups
with given cardinality of Haar positive conjugacy classes?
- Question. Does there exist a Polish group such that it
consistently has κ many Haar positive conjugacy classes with ℵ0 < κ < c?
SLIDE 45
Open problems
- Question. Are there natural examples of automorphism groups
with given cardinality of Haar positive conjugacy classes?
- Question. Does there exist a Polish group such that it
consistently has κ many Haar positive conjugacy classes with ℵ0 < κ < c?
- Problem. Formulate necessary and sufficient model theoretic
conditions which characterise the measure theoretic behaviour
- f the conjugacy classes!
SLIDE 46