• Non ci sono risultati.

Quotients of projective Fra¨ıss´e limits

N/A
N/A
Protected

Academic year: 2021

Condividi "Quotients of projective Fra¨ıss´e limits"

Copied!
41
0
0

Testo completo

(1)

Quotients of projective Fra¨ıss´ e limits

(2)

The pseudo-arc

For X a topological space, U , V open coverings of X , say U refines V if

∀U ∈ U ∃V ∈ V U ⊆ V .

A continuum is a compact connected metric space. It is non-degenerate if it has more than one point.

A continuum X is chainable if every open cover of X is refined by a

chain: an open cover {U 1 , . . . , U n } such that U i ∩ U j 6= ∅ ⇔ |i − j| ≤ 1.

This is equivalent to ask that ∀ε ∈ R + , ∃f : X → [0, 1] continuous and

onto such that ∀t ∈ [0, 1] diam(f −1 ({t})) < ε.

(3)

The pseudo-arc

For X a topological space, U , V open coverings of X , say U refines V if

∀U ∈ U ∃V ∈ V U ⊆ V .

A continuum is a compact connected metric space. It is non-degenerate if it has more than one point.

A continuum X is chainable if every open cover of X is refined by a

chain: an open cover {U 1 , . . . , U n } such that U i ∩ U j 6= ∅ ⇔ |i − j| ≤ 1.

This is equivalent to ask that ∀ε ∈ R + , ∃f : X → [0, 1] continuous and

onto such that ∀t ∈ [0, 1] diam(f −1 ({t})) < ε.

(4)

The pseudo-arc

For X a topological space, U , V open coverings of X , say U refines V if

∀U ∈ U ∃V ∈ V U ⊆ V .

A continuum is a compact connected metric space. It is non-degenerate if it has more than one point.

A continuum X is chainable if every open cover of X is refined by a chain: an open cover {U 1 , . . . , U n } such that U i ∩ U j 6= ∅ ⇔ |i − j| ≤ 1.

This is equivalent to ask that ∀ε ∈ R + , ∃f : X → [0, 1] continuous and

onto such that ∀t ∈ [0, 1] diam(f −1 ({t})) < ε.

(5)

The pseudo-arc

For X a topological space, U , V open coverings of X , say U refines V if

∀U ∈ U ∃V ∈ V U ⊆ V .

A continuum is a compact connected metric space. It is non-degenerate if it has more than one point.

A continuum X is chainable if every open cover of X is refined by a chain: an open cover {U 1 , . . . , U n } such that U i ∩ U j 6= ∅ ⇔ |i − j| ≤ 1.

This is equivalent to ask that ∀ε ∈ R + , ∃f : X → [0, 1] continuous and

onto such that ∀t ∈ [0, 1] diam(f −1 ({t})) < ε.

(6)

The pseudo-arc

A continuum X is decomposable if there are proper subcontinua Y , Z such that

X = Y ∪ Z . Otherwise it is indecomposable.

A continuum is hereditarily decomposable (hereditarily

indecomposable) if all of its non-degenerate subcontinua are.

(7)

The pseudo-arc

A continuum X is decomposable if there are proper subcontinua Y , Z such that

X = Y ∪ Z . Otherwise it is indecomposable.

A continuum is hereditarily decomposable (hereditarily

indecomposable) if all of its non-degenerate subcontinua are.

(8)

The pseudo-arc

The pseudo-arc is the unique, up to homeomorphism, hereditarily indecomposable chainable continuum.

The pseudo-arc is the generic continuum: in the Polish space of all

subcontinua of [0, 1] N , its homeomorphism class is dense G δ .

(9)

The pseudo-arc

The pseudo-arc is the unique, up to homeomorphism, hereditarily indecomposable chainable continuum.

The pseudo-arc is the generic continuum: in the Polish space of all

subcontinua of [0, 1] N , its homeomorphism class is dense G δ .

(10)

Fra¨ıss´ e limits

Let K be a class of L-structures. K has

HP hereditary property if: for all A ∈ K and for all finitely generated substructure B of A, B is (isomorphic to some element) in K . JEP joint embedding property if: ∀A, B ∈ K , ∃C ∈ K such that A, B

are embeddable in K .

AP amalgamation property if: ∀A, B, C ∈ K , ∀e : A → B, f : A → C embeddings, ∃D ∈ K , ∃g : B → D, h : C → D embeddings, such that ge = hf .

Theorem. (Fra¨ıss´ e) L a countable language, K a non-empty countable set of finitely generated L-structures having HP, JEP and AP. Then there is an L-structure D, unique up to isomorphism, such that:

I card(D) ≤ ℵ 0

I up to isomorphism, K contains exactly the finitely generated substructures of D

I D is ultrahomogeneus: every isomorphism between finitely generated

substructures of D can be extended to an isomorphism of D

(11)

Fra¨ıss´ e limits

Let K be a class of L-structures. K has

HP hereditary property if: for all A ∈ K and for all finitely generated substructure B of A, B is (isomorphic to some element) in K . JEP joint embedding property if: ∀A, B ∈ K , ∃C ∈ K such that A, B

are embeddable in K .

AP amalgamation property if: ∀A, B, C ∈ K , ∀e : A → B, f : A → C embeddings, ∃D ∈ K , ∃g : B → D, h : C → D embeddings, such that ge = hf .

Theorem. (Fra¨ıss´ e) L a countable language, K a non-empty countable set of finitely generated L-structures having HP, JEP and AP. Then there is an L-structure D, unique up to isomorphism, such that:

I card(D) ≤ ℵ 0

I up to isomorphism, K contains exactly the finitely generated substructures of D

I D is ultrahomogeneus: every isomorphism between finitely generated

substructures of D can be extended to an isomorphism of D

(12)

Topological structures

L a first order language.

A topological L-structures is an L-structure A such that:

I A is a zero-dimensional, compact, second countable space

I Interpretations of relation symbols are closed

I Interpretations of function symbols are continuous

An epimorphism of topological L-structures A, B is a continuous surjection ϕ : A → B such that

I R B = ϕ × . . . × ϕ(R A ) for any relation symbol R

I f B (ϕ(x 1 ), . . . , ϕ(x n )) = ϕf A (x 1 , . . . , x n ) for any n-ary function symbol f

An isomorphism is a bijective epimorphism

(13)

Topological structures

L a first order language.

A topological L-structures is an L-structure A such that:

I A is a zero-dimensional, compact, second countable space

I Interpretations of relation symbols are closed

I Interpretations of function symbols are continuous

An epimorphism of topological L-structures A, B is a continuous surjection ϕ : A → B such that

I R B = ϕ × . . . × ϕ(R A ) for any relation symbol R

I f B (ϕ(x 1 ), . . . , ϕ(x n )) = ϕf A (x 1 , . . . , x n ) for any n-ary function symbol f

An isomorphism is a bijective epimorphism

(14)

Projective Fra¨ıss´ e families

∆ a family of topological L-structures, is a projective Fra¨ıss´ e family if:

I ∀D, E ∈ ∆, ∃F ∈ ∆, ∃F → D, F → E epimorphisms

I ∀C , D, E ∈ ∆, ∀ϕ 1 : D → C , ϕ 2 : E → C epimorphisms,

∃F ∈ ∆, ∃ψ 1 : F → D, ψ 2 : F → E epimorphisms such that

ϕ 1 ψ 1 = ϕ 2 ψ 2

(15)

Projective Fra¨ıss´ e limits

∆ a family of topological L-structures, D a topological L-structure.

D is a projective Fra¨ıss´ e limit of ∆ if:

I (projective universality) ∀D ∈ ∆, ∃ϕ : D → D epimorphism

I For any clopen partition of D there is an epimorphism of D to a structure in ∆ refining it

I (projective ultrahomogeneity) ∀D ∈ D, ∀ϕ 1 : D → D, ∀ϕ 2 : D → D

epimorphisms, ∃ψ : D → D isomorphism such that ϕ 2 = ϕ 1 ψ

(16)

Projective Fra¨ıss´ e limits

Theorem. A countable, non-empty class ∆ of finite topological L-structures has a projective Fra¨ıss´ e limit if and only if it is a projective Fra¨ıss´ e family.

Such a limit is unique up to isomorphism.

(17)

Construction of the limit

A fundamental sequence for ∆ is a sequence (D n ) in ∆ with epimorphisms π n : D n+1 → D n such that:

a) ∀D ∈ ∆, ∃n, ∃ϕ : D n → D epimorphism

b) ∀n, ∀E , F ∈ ∆, ∀ϕ 1 : F → E , ϕ 2 : D n → E epimorphisms,

∃m ≥ n, ∃ψ : D m → F epimorphism such that ϕ 1 ψ = ϕ 2 π m n , where π m n = π m−1 · . . . · π n : D m → D n .

For ∆ a countable non-empty class of finite topological L-structures, the existence of a fundamental sequence is equivalent to being a projective Fra¨ıss´ e family.

The projective Fra¨ıss´ e limit D of ∆ is the inverse limit of the sequence D 0

π

0

← D 1 π

1

← D 2 π

2

← D 3 π

3

← . . .

(18)

Construction of the limit

A fundamental sequence for ∆ is a sequence (D n ) in ∆ with epimorphisms π n : D n+1 → D n such that:

a) ∀D ∈ ∆, ∃n, ∃ϕ : D n → D epimorphism

b) ∀n, ∀E , F ∈ ∆, ∀ϕ 1 : F → E , ϕ 2 : D n → E epimorphisms,

∃m ≥ n, ∃ψ : D m → F epimorphism such that ϕ 1 ψ = ϕ 2 π m n , where π m n = π m−1 · . . . · π n : D m → D n .

For ∆ a countable non-empty class of finite topological L-structures, the existence of a fundamental sequence is equivalent to being a projective Fra¨ıss´ e family.

The projective Fra¨ıss´ e limit D of ∆ is the inverse limit of the sequence D 0

π

0

← D 1 π

1

← D 2 π

2

← D 3 π

3

← . . .

(19)

Linear graphs

Let L = {R}, R a binary relation symbol.

A linear graph is a reflexive, symmetric, connected relation such that each element has at most three neighbours (including itself) and, if there are more then two elements, exactly two of them have exactly two neighbours

• − • − • − • − . . . − • − • − •

(20)

Linear graphs

Let L = {R}, R a binary relation symbol.

A linear graph is a reflexive, symmetric, connected relation such that each element has at most three neighbours (including itself) and, if there are more then two elements, exactly two of them have exactly two neighbours

• − • − • − • − . . . − • − • − •

(21)

Theorem. (Irwin, Solecki) The class ∆ 0 of finite linear graphs is a projective Fra¨ıss´ e class.

Let (P, R P ) be the projective Fra¨ıss´ e limit of ∆ 0 . Then

I R P is an equivalence relation, whose classes have at most two elements.

I P/R P is a pseudo-arc.

This allowed Irwin and Solecki to recover some known and new properties

of the pseudo-arc.

(22)

Theorem. (Irwin, Solecki) The class ∆ 0 of finite linear graphs is a projective Fra¨ıss´ e class.

Let (P, R P ) be the projective Fra¨ıss´ e limit of ∆ 0 . Then

I R P is an equivalence relation, whose classes have at most two elements.

I P/R P is a pseudo-arc.

This allowed Irwin and Solecki to recover some known and new properties

of the pseudo-arc.

(23)

Question: What are the possible generalisations?

1.

I Change the class of finite topological structures

I Restrict the class of epimorphisms to some meaningful subclass

I Reformulate accordingly notions of projective universality and projective homogeneity

This has been succesfully undertaken by Irwin.

2. Study all possible quotients of projective Fra¨ıss´ e limits for projective

Fra¨ıss´ e families of finite topological structures in the language {R}, R a

binary relation symbol.

(24)

Question: What are the possible generalisations?

1.

I Change the class of finite topological structures

I Restrict the class of epimorphisms to some meaningful subclass

I Reformulate accordingly notions of projective universality and projective homogeneity

This has been succesfully undertaken by Irwin.

2. Study all possible quotients of projective Fra¨ıss´ e limits for projective

Fra¨ıss´ e families of finite topological structures in the language {R}, R a

binary relation symbol.

(25)

Question: What are the possible generalisations?

1.

I Change the class of finite topological structures

I Restrict the class of epimorphisms to some meaningful subclass

I Reformulate accordingly notions of projective universality and projective homogeneity

This has been succesfully undertaken by Irwin.

2. Study all possible quotients of projective Fra¨ıss´ e limits for projective

Fra¨ıss´ e families of finite topological structures in the language {R}, R a

binary relation symbol.

(26)

Note that if a projective Fra¨ıss´ e family contains a disconnected structure, then the projective Fra¨ıss´ e limit (D, R D ) contains non-empty clopen subsets invariant under R D . So if R D is an equivalence relation, D/R D is disconnected.

Thus it appears to exist a strong limitation in getting continua with approach 2:

Theorem.

1. If Γ is a projective Fra¨ıss´ e family and Γ 0 ⊆ Γ is such that

∀G ∈ Γ, ∃G 0 ∈ Γ 0 , ∃ϕ : G 0 → G epimorphism, then Γ 0 is a

projective Fra¨ıss´ e family, with the same projective Fra¨ıss´ e limit as Γ. 2. If ∆ is a projective Fra¨ıss´ e family of finite connected reflexive graphs

of unbounded diameter, then ∆ ⊆ ∆ 0 .

(27)

Note that if a projective Fra¨ıss´ e family contains a disconnected structure, then the projective Fra¨ıss´ e limit (D, R D ) contains non-empty clopen subsets invariant under R D . So if R D is an equivalence relation, D/R D is disconnected.

Thus it appears to exist a strong limitation in getting continua with approach 2:

Theorem.

1. If Γ is a projective Fra¨ıss´ e family and Γ 0 ⊆ Γ is such that

∀G ∈ Γ, ∃G 0 ∈ Γ 0 , ∃ϕ : G 0 → G epimorphism, then Γ 0 is a

projective Fra¨ıss´ e family, with the same projective Fra¨ıss´ e limit as Γ.

2. If ∆ is a projective Fra¨ıss´ e family of finite connected reflexive graphs

of unbounded diameter, then ∆ ⊆ ∆ 0 .

(28)

Theorem The compact metric spaces that can be obtained as quotients D/R D , where (D, R D ) is a projective Fra¨ıss´ e limit of a projective Fra¨ıss´ e family of finite topological {R}-structures with R D an equivalence relation are:

1. a Cantor space

2. a disjoint union of m singletons and n pseudo-arcs (m, n ∈ N, m + n > 0)

3. a disjoint union of n compact metric spaces X i , where

I

X

i

= P

i

∪ S

j ∈N

Q

ij I

P

i

a pseudo-arc

I

Q

ij

a Cantor space clopen in X

i I

j ∈N

Q

ij

dense in X

i

In particular, the only continua that can be obtained this way are the

point and the pseudo-arc.

(29)

Theorem The compact metric spaces that can be obtained as quotients D/R D , where (D, R D ) is a projective Fra¨ıss´ e limit of a projective Fra¨ıss´ e family of finite topological {R}-structures with R D an equivalence relation are:

1. a Cantor space

2. a disjoint union of m singletons and n pseudo-arcs (m, n ∈ N, m + n > 0)

3. a disjoint union of n compact metric spaces X i , where

I

X

i

= P

i

∪ S

j ∈N

Q

ij I

P

i

a pseudo-arc

I

Q

ij

a Cantor space clopen in X

i I

j ∈N

Q

ij

dense in X

i

In particular, the only continua that can be obtained this way are the

point and the pseudo-arc.

(30)

The spaces in the previous theorem can be obtained, for example, with the following projective Fra¨ıss´ e families:

1. (Cantor space): When there is no bound on the number of non-trivial components of the members in the family; for instance: Γ = class of all finite reflexive graphs

2. (m points and n pseudo-arcs): ∆ m n = class of all reflexive graphs with n + m connected components of which m are trivial and n are linear graphs

3. (n pseudo-arcs with a sequence of clopen Cantor sets converging to

each one of them): ∆ n = class of all reflexive graphs having

arbitrarily many connected components, exactly n of which

non-trivial, each of which a linear graph

(31)

The spaces in the previous theorem can be obtained, for example, with the following projective Fra¨ıss´ e families:

1. (Cantor space): When there is no bound on the number of non-trivial components of the members in the family; for instance:

Γ = class of all finite reflexive graphs

2. (m points and n pseudo-arcs): ∆ m n = class of all reflexive graphs with n + m connected components of which m are trivial and n are linear graphs

3. (n pseudo-arcs with a sequence of clopen Cantor sets converging to

each one of them): ∆ n = class of all reflexive graphs having

arbitrarily many connected components, exactly n of which

non-trivial, each of which a linear graph

(32)

The spaces in the previous theorem can be obtained, for example, with the following projective Fra¨ıss´ e families:

1. (Cantor space): When there is no bound on the number of non-trivial components of the members in the family; for instance:

Γ = class of all finite reflexive graphs

2. (m points and n pseudo-arcs): ∆ m n = class of all reflexive graphs with n + m connected components of which m are trivial and n are linear graphs

3. (n pseudo-arcs with a sequence of clopen Cantor sets converging to

each one of them): ∆ n = class of all reflexive graphs having

arbitrarily many connected components, exactly n of which

non-trivial, each of which a linear graph

(33)

The spaces in the previous theorem can be obtained, for example, with the following projective Fra¨ıss´ e families:

1. (Cantor space): When there is no bound on the number of non-trivial components of the members in the family; for instance:

Γ = class of all finite reflexive graphs

2. (m points and n pseudo-arcs): ∆ m n = class of all reflexive graphs with n + m connected components of which m are trivial and n are linear graphs

3. (n pseudo-arcs with a sequence of clopen Cantor sets converging to

each one of them): ∆ n = class of all reflexive graphs having

arbitrarily many connected components, exactly n of which

non-trivial, each of which a linear graph

(34)

Small Polish structures

A Polish structure is a pair (X , G ) where:

I G is a Polish group acting faithfully on a set X

I the stabilisers of all singletons are closed

This generalises the notion of a profinite structure.

(35)

Small Polish structures

A Polish structure is a pair (X , G ) where:

I G is a Polish group acting faithfully on a set X

I the stabilisers of all singletons are closed

This generalises the notion of a profinite structure.

(36)

For A ⊆ X , let G A be the pointwise stabiliser of A.

Let ~ a ∈ X and A, B ⊆ X finite. Let π A : G A → G A ~ a, g 7→ g ~ a.

~ a is nm-independent from B over A if π A −1 (G A∪B ~ a) is non-meagre in G A

(= π −1 A (G A ~ a)).

(X , G ) is nm-stable if there is no infinite sequence A 0 ⊆ A 1 ⊆ . . . of finite subsets of X and a ∈ X such that a is nm-dependent from A i +1

over A i .

Example. (Krupi´ nski) Let P be the pseudo-arc. Then (P, Homeo(P)) is

a small, not nm-stable, Polish structure.

(37)

For A ⊆ X , let G A be the pointwise stabiliser of A.

Let ~ a ∈ X and A, B ⊆ X finite. Let π A : G A → G A ~ a, g 7→ g ~ a.

~ a is nm-independent from B over A if π A −1 (G A∪B ~ a) is non-meagre in G A

(= π −1 A (G A ~ a)).

(X , G ) is nm-stable if there is no infinite sequence A 0 ⊆ A 1 ⊆ . . . of finite subsets of X and a ∈ X such that a is nm-dependent from A i +1

over A i .

Example. (Krupi´ nski) Let P be the pseudo-arc. Then (P, Homeo(P)) is

a small, not nm-stable, Polish structure.

(38)

For A ⊆ X , let G A be the pointwise stabiliser of A.

Let ~ a ∈ X and A, B ⊆ X finite. Let π A : G A → G A ~ a, g 7→ g ~ a.

~ a is nm-independent from B over A if π A −1 (G A∪B ~ a) is non-meagre in G A

(= π −1 A (G A ~ a)).

(X , G ) is nm-stable if there is no infinite sequence A 0 ⊆ A 1 ⊆ . . . of finite subsets of X and a ∈ X such that a is nm-dependent from A i +1

over A i .

Example. (Krupi´ nski) Let P be the pseudo-arc. Then (P, Homeo(P)) is

a small, not nm-stable, Polish structure.

(39)

A dendrite is a locally connected continuum that does not contain simple closed curves.

Let W be Wa˙zewski universal dendrite.

Then (W , Homeo(W )) is a small Polish structure. Problems.

I Is it nm-stable?

I Characterise dendrites D such that (D, Homeo(D)) is small.

(40)

A dendrite is a locally connected continuum that does not contain simple closed curves.

Let W be Wa˙zewski universal dendrite.

Then (W , Homeo(W )) is a small Polish structure.

Problems.

I Is it nm-stable?

I Characterise dendrites D such that (D, Homeo(D)) is small.

(41)

A dendrite is a locally connected continuum that does not contain simple closed curves.

Let W be Wa˙zewski universal dendrite.

Then (W , Homeo(W )) is a small Polish structure.

Problems.

I Is it nm-stable?

I Characterise dendrites D such that (D, Homeo(D)) is small.

Riferimenti

Documenti correlati

Nel caso specifico della Principessa daella Colchide possiamo dunque essere sicuri che un testo adatto ad un piccolo gruppo di spettatori comunicava e riprendeva in modo

Brine shrimps evidenced no toxic effects while oysters discriminated well among leachates: the tropical wood species showed similar or relatively lower toxic effects than the

For such an algebra, it is easy to show that the exponential map from the additive group to its multiplicative group of units is surjective... For such an algebra, it is easy to

Cohomological dimension of open subsets of the projective space..

The conjecture is known to be true in its full generality in dimension 1 by Gruson-Lazarsfeld-Peskine and Giaimo; in dimension 2, it is true for smooth surfaces by Lazarsfeld;

The conjecture is known to be true in its full generality in dimension 1 by Gruson-Lazarsfeld-Peskine and Giaimo; in dimension 2, it is true for smooth surfaces by Lazarsfeld;

and subcanonical curve in X, we find upper-bounds on δ which are, respectively, cubic, quadratic and linear polynomials in k ensuring the regularity of V δ (E(k)) at [s].

the second author proved that, over RCA 0 , FRA is equivalent to the statement that ST is well-quasi-ordered by homomorphism.) In this paper we exploit the fact that