0
現代数学解説
文献あり

Robustly Isomorphic Models

36
0

この文書はAlweさんの企画された Mathematical Logic Advent Calendar 2023 の15日目の記事です。まだ日本語を使わなくてすみません🙏。

I learned the following argument from Garrett Ervin.

Increasing continuous elementary chain

An elementary chain a sequence C={Mi:i<α} of models over a common first-order language L, indexed by ordinals i below α, such that for all ij<α, MiMj is an elementary extension.

We say C is increasing if for all i such that i+1<α, it holds that MiMi+1 is a proper extension.

We say C is continuous if for all limit ordinals β<α, it holds that Mβ=i<βMi is the union of its predecessors {Mi:i<β}.

Main theorem

Let M be an arbitrary model of size 0 over a countable language. The following hold:

  1. If κ is an infinite cardinal, and α is an ordinal such that α<κ+, then there exists an increasing continuous elementary chain {Mi:i<α} of elementary extensions MiM of size κ, such that the models {Mi:i<α} are all isomorphic.
  2. There exists an increasing continuous elementary chain {Mi:i<ω1} of elementary extensions MiM of size 0, such that the models {Mi:i<ω1} are all isomorphic.
  3. Assume CH or PFA . There exists an increasing continuous elementary chain {Mi:i<ω2} of elementary extensions MiM of size 1, such that the models {Mi:i<ω2} are all isomorphic.
  4. If κ is an infinite cardinal such that κ<κ=κ, then there exists an increasing continuous elementary chain {Mi:i<κ+} of elementary extensions MiM of size κ, such that the models {Mi:i<κ+} are all isomorphic.
    Assuming GCH , this holds for every regular cardinal κ.

In the cases (2), (3), and (4), the elementary chain {Mi:i<κ+} is maximal: The next model Mκ+=i<κ+Mi always has size κ+, so it cannot be isomorphic to the earlier models Mi of size κ.

It is currently an open question whether case (3) always holds in ZFC.

Ehrenfeucht–Mostowski models

The following lemma reveals the combinatorial content of the main theorem.

An increasing continuous chain of models C={Mi:i<α} is defined similarly as an increasing continuous elementary chain, except for all ij<α, we no longer require MiMj to be an elementary extension, but we instead assume MiMj is some extension (where the inclusion MiMj respects the function, relation, and constant symbols from the language).

Given an infinite cardinal κ and an ordinal α, the following are equivalent:

  1. For any model M of size 0 over a countable language, there exists an increasing continuous elementary chain {Mi:i<α} of elementary extensions MiM of size κ, such that the models {Mi:i<α} are all isomorphic.
  2. There exists an increasing continuous chain {(Ii,<):i<α} of linear orders (Ii,<) of size κ, such that the linear orders {(Ii,<):i<α} are all isomorphic.
1 ⇒ 2

Let M=(Q,<). Then the increasing continuous chain {Mi=(Ii,<):i<α} of linear orders is what we wanted.

To prove the implication (2 ⇒ 1), we need a theorem by Ehrenfeucht–Mostowski.

Skolem hull, see [1] §5.2

Every complete first-order theory T has an extension T into an expanded first-order language L(T)L(T), such that |T|=|T|, and T has built-in Skolem functions: For every formula ϕ(v,w¯) in the language of T, there is a function symbol fϕ(w¯) such that T contains the sentence w¯[(vϕ(v,w¯))ϕ(fϕ(w¯),w¯)].

For every complete first-order theory T with infinite models and every linear order (I,<), there is a model C of T with a sequence {ai:iI} of order indiscernibles: For every two increasing sequences i1<<in and j1<<jn in I and every formula ϕ(x1,,xn), we have Cϕ[ai1,,ain]ϕ[aj1,,ajn].

Given a complete first-order theory T with built-in Skolem functions, a model CT with order indiscernibles {ai:iI}, and a suborder I0I, the Skolem hull of I0={ai:iI0} is the L(T)-submodel EM(I0)C with universe {fϕC(a¯):fϕL(T),a¯I0}.

Ehrenfeucht–Mostowski, see [1] §5.2

Let T, CT, I={ai:iI}, and EM() be as in the definition above. For any I0,I1I, the following hold:

  1. EM(I0) is a model of T of size |I0|+|T|, and an elementary submodel of C.
  2. If I0I1, then EM(I0)EM(I1).
  3. More generally, every (I,<)-order-preserving f:I0I1 extends uniquely into an elementary embedding f¯:EM(I0)EM(I1).
  4. If f:I0I1 is an (I,<)-order-preserving bijection, then f extends uniquely into an isomorphism f¯:EM(I0)EM(I1).
2 ⇒ 1

Let ED(M)={ϕ(cm1,,cmn):Mϕ[m1,,mn]} be the elementary/complete diagram of M, so that ED(M) is a countable complete first-order theory extending the theory Th(M) of M, and every model NED(M) admits an elementary embedding jM,N:MN (mapping mcmN), such that if jN,N:NN is an elementary embedding between models of ED(M), then jN,NjM,N=jM,N (see [1] Lemma 2.3.3).

Let TED(M) be a countable complete first-order extension (in an expanded language) with built-in Skolem functions, and let CT, I={ai:iI}, and EM() be as in the Ehrenfeucht–Mostowski theorem, where I=i<αIi.

For every i<α, let Ii={ai:iIi}, and let Mi=EM(Ii). We can check that {Mi:i<α} is the elementary chain that we wanted:

  1. By the Ehrenfeucht–Mostowski theorem part (2), {Mi:i<α} forms an L(T)-elementary chain. So their reducts {MiL(M):i<α} form an L(M)-elementary chain.
  2. By the Ehrenfeucht–Mostowski theorem part (1), every model Mi is a model of ED(M) of size κ. So by identifying the model M with its isomorphic image MjM,M0(M)M0, the properties of ED(M) from before imply that MMi for all i<α under this identification.
  3. From the assumption that I is a sequence of order indiscernibles, one can prove that EM(J)I=J for all JI. This implies that since {Ii:i<α} is increasing, the chain {Mi:i<α} is also increasing.
  4. From the definition of Skolem hulls, one can prove that if {Ji:i<β} is a chain of subsets JiI (such that ij implies JiJj), then i<βEM(Ji)=EM(i<βJi). This implies that since {Ii:i<α} is continuous, the chain {Mi:i<α} is also continuous.
  5. By the Ehrenfeucht–Mostowski theorem part (4), since the linear orders {(Ii,<):i<α} are all isomorphic, the models {Mi:i<α} are also all isomorphic.

Robust linear orders

In light of the lemma from the previous section, the main theorem is equivalent to the following proposition:

Main proposition
  1. If κ is an infinite cardinal, and α is an ordinal such that α<κ+, then there exists an increasing continuous chain {(Ii,<):i<α} of linear orders of size κ which are all isomorphic.
  2. There exists an increasing continuous chain {(Ii,<):i<ω1} of linear orders of size 0 which are all isomorphic.
  3. Assume CH or PFA. There exists an increasing continuous chain {(Ii,<):i<ω2} of linear orders of size 1 which are all isomorphic.
  4. If κ is an infinite cardinal such that κ<κ=κ, then there exists an increasing continuous chain {(Ii,<):i<κ+} of linear orders of size κ which are all isomorphic.

In this section we will prove parts (1) and (2).

Note that in the notation above, we cannot have α>κ+, or else the increasing continuous property of the chain implies that Iκ+ always has size κ+>κ, so that Iκ+ cannot be isomorphic to previous linear orders of size κ.

Main proposition part 1

Assume α<κ+. Let β be a limit ordinal such that max{κ,α}β<κ+. For i<α, we can define the linear order
Ii=β{j+12:j<i},
such that (Ii,<) is a well-order, and β being a limit ordinal implies that (Ii,<)(β,<) for all i<α. It is easy to check that {Ii:i<α} is an increasing continuous chain of linear orders of size κ which are all isomorphic to β.

The next parts (2), (3), and (4) all concern the case α=κ+, so it is convenient to define a term for this case.

Robust linear order

We say that a linear order (I,<) of size κ is robust if there exists an increasing continuous chain {(Ii,<):i<κ+} of linear orders which are all isomorphic to (I,<).

Do robust linear orders always exist in every infinite cardinality?

Main proposition part 2

We will show that (Q,<) is robust of size 0.

For i<ω1, we define the linear order
(Ii,<)=((1+i)×Q,<lex).
This Ii is constructed by replacing every point in the linear order 1+i with a copy of Q. Then every Ii is a dense linear order without endpoints of size 0, so by Cantor's back-and-forth method (see [1] Theorem 2.4.1), we get an isomorphism (Ii,<)(Q,<).

It's easy to see that {Ii:i<ω1} forms an increasing continuous chain, witnessing robustness of (Q,<).

Baumgartner's theorem

In this section we'll see that the Proper Forcing Axiom (PFA) implies a robust linear order of size 1.

Baumgartner, see [2] Corollary 8.3

A set of reals XR is called κ-dense if for every nonempty open interval UR, there is |XU|=κ.

Assume PFA. Every two 1-dense sets of reals X and Y are isomorphic.

Main proposition part 3, assuming PFA

Let X be an 1-dense set of reals. We'll show that (X,<) is robust.

PFA implies |R|=2 ([3] Theorem 31.23), so we can pick 2 many distinct reals {xiR:i<ω2} that are not in X. For i<ω2, define
Ii=X{xj:j<i}.
Then every Ii is 1-dense since XIi is 1-dense and |Ii|=1, and so Baumgartner's theorem implies (Ii,<)(X,<).

It's easy to see that {Ii:i<ω2} is increasing continuous, which witnesses that (X,<) is robust of size 1.

I was given this argument in a Mathematics Stack Exchange answer by username Holo when I first asked about robust linear orders.

In an unpublished result , Itay Neeman has shown the consistency of “every two 2-dense sets of reals are isomorphic”.

Saturated linear orders

In this section, we will prove part (4) of the main proposition:

Ervin

If κ is an infinite cardinal such that κ<κ=κ, then there exists a robust linear order of size κ.

The assumption that κ<κ=κ is equivalent to the assumptions that κ is regular and 2<κ=κ.

The Continuum Hypothesis (CH) implies that κ=1 satisfies this condition, and so part (3) of the main proposition follows. More generally, the Generalized Continuum Hypothesis (GCH) implies that every infinite cardinal κ satisfies 2<κ=κ, which then implies there are robust linear orders of size any regular cardinal κ.

In the discussions below, we fix a regular cardinal κ satisfying 2<κ=κ.

(I,<) and J

(2κ,<lex) is the lexicographic ordering on binary strings of length κ.

Let I2κ be the set of all x2κ which are not eventually constant. The linear order (I,<) is the suborder of (2κ,<lex) induced by I.

Define JI to be the set of all elements of I of the form σ(01)κ=σ0101 for some σ2<κ.

Note that |I|=2κ and |J|=2<κ=κ.

Saturation lemma part 1

Let A,BI be two subsets of I such that |A|<κ, |B|<κ, and for all aA and bB there is a<b. Then there exists cJ such that a<c for all aA, and c<b for all bB.

We can define a supremum f=sup(A)2κ recursively, such that for all i<κ, f(i)2 is the smallest such that for all aA, there is a(i+1)lex(fi)f(i). It is not hard to show that for all aA, there is alexf, and for all g<lexf there exists some aA such that g<lexa.

We split into cases.

  1. If f=sup(A)A, then for all bB, since f<b, we can pick some γb<κ such that fγb<bγb. Then let γ=supbBγb so that γ<κ by regularity, and we see that g=(fγ)(1)κ satisfies gγ<bγ for all bB.
    Since fg, fAI, and gI, we have f<g, so that fδ<gδ for some γδ<κ. Letting c=(gδ)(01)κ, we have f<c, so that for all aA, aδfδ<cδ, and for all bB, cδ=gδ<bδ. By cJ, we've seen that c is what we wanted.
  2. Similarly if inf(B)B, then a desired cJ exists.
  3. Assume sup(A)A and inf(B)B.
    First we can show that f=sup(A) is eventually constantly 0. For every aA, fix γa<κ such that aγa<fγa, and let γ=supaAγa<κ by regularity. Then g=(fγ)(0)κ is such that gf, but a<g for all aA. Since f is a supremum, we must have f=g, and so f is eventually constantly 0 as claimed.
    Similarly h=inf(B) is eventually constantly 1.
    From A<B we know that fh. But since fh, we have f<h. So we can fix some δ<κ such that fδ<hδ, f=(fδ)(0)κ, and h=(hδ)(1)κ. Then c=(fδ)(01)κ satisfies f<c<h and cJ, so c is what we wanted.
Saturation lemma part 2, see [1] Theorem 4.3.20

Define a linear order (I,<) of size κ to be saturated if for every A,BI such that |A|<κ, |B|<κ, and a<b for all aA and bB, there exists cI such that a<c for all aA and c<b for all bB.

Every two saturated linear orders (L,<) and (K,<) of size κ are isomorphic.

Let LK={xi:i<κ}.

We can build a chain of partial isomorphisms {fi:i<κ} such that dom(fi)L and img(fi)K both have size <κ, fi is an isomorphism between (dom(fi),<L) and (img(fi),<K), and xidom(fi+1)img(fi+1).

  1. When i=0 let f0=.
  2. When i<κ is limit, let fi=j<ifj.
  3. When i+1 is a successor stage and xiLdom(fi), we let A=fi[(,xi)] and B=fi[(xi,)]. By saturation of K, there is cK such that A<c<B, and we can define fi+1=fi{(xi,c)}.
  4. The other cases are similar, and it's never hard to see that fi has the desired properties for all i<κ.

Finally, we get an isomorphism f=i<κfi from (L,<) to (K,<).

Main proposition part 4

We'll show the linear order (J,<) is robust of size κ.

Since |I|=2κ, we can pick κ+ many distinct elements {xiI:i<κ+} that are not in J. For i<κ+, define
Ii=J{xj:j<i}.
By the saturation lemma part 1, every Ii is saturated of size κ, because JIi. By the saturation lemma part 2, (Ii,<)(J,<).

It's easy to see that {Ii:i<κ+} is increasing continuous, which shows that (J,<) is robust of size κ.

This argument is due to Garrett Ervin in a private conversation.

From Shelah [4] Theorem 8.4.7, there exist saturated linear orders of size κ if and only if κ<κ=κ. For a simple argument when κ=ω, see [1] §4.3.

参考文献

[1]
David Marker, Model Theory: An Introduction, Graduate Texts in Mathematics, Springer, 2002
[2]
Stevo Todorcevic, Partition Problems in Topology, Contemporary Mathematics, American Mathematical Society, 1989
[3]
Thomas Jech, Set Theory: The Third Millennium Edition, Revised and Expanded, Springer Monographs in Mathematics, Springer, 2002
[4]
Saharon Shelah, Classification Theory: and the Number of Non-Isomorphic Models, Studies in Logic and the Foundations of Mathematics, North-Holland, 1990
投稿日:20231217
OptHub AI Competition

この記事を高評価した人

高評価したユーザはいません

この記事に送られたバッジ

バッジはありません。
バッチを贈って投稿者を応援しよう

バッチを贈ると投稿者に現金やAmazonのギフトカードが還元されます。

投稿者

duyaa
duyaa
6
276
中国人です。今はAlexander Kechris先生の弟子として、カリフォルニア工科大学で博士課程を受けています。

コメント

他の人のコメント

コメントはありません。
読み込み中...
読み込み中
  1. Ehrenfeucht–Mostowski models
  2. Robust linear orders
  3. Baumgartner's theorem
  4. Saturated linear orders
  5. 参考文献