Shelah’s approachability ideal (part 1)

Given an infinite cardinal λ, Shelah defines an ideal I[λ] as follows.

Definition (Shelah, implicit in here). A set S is in I[λ] iff Sλ and there exists a collection {Dαα<λ}[P(λ)]<λ, and some club Eλ, so that for every δSE, there exists a cofinal subset Aδδ such that:

  • otp(Aδ)<δ (in particular, δ is singular);
  • for every γ<δ, there exists some α<δ such that AδγDα.

In other words, α<δDα contains all the initial segments of some small cofinal subset, Aδ, of δ.

Easy observations. 1) EωλI[λ];
2) if μ<κ<λ for all μ<λ, then EκλI[λ];
3) if Sλ is nonstationary, then SI[λ];
4) if ◻λ holds, then I[λ+]=P(λ+). (the definition of ◻λ may be found in here.)
Proof hints. 1) Let Dα:=[α]<ω for all α<λ.
2) Let Dα:=[α]<κ for all α<λ.
3) Take a club E which is disjoint from S.
4) If Cαα<λ+ is a ◻λ-sequence, then simply let Dα:={acc(Cα)}[Cα]<ω for all α<λ+, and restrict your attention to E:=acc(λ+). ◻

So, I[λ+] is rather trivial in the presence of ◻λ. On the other hand, it is a very sophisticated result of Mitchell, that I[λ+] may behave as the other extreme:

Theorem (Mitchell, 2009). Starting with a cardinal κ which is κ+-Mahlo, in some forcing extension, I[2] has the property that SE12 is nonstationary for every SI[2]. ◻

In other words, I[κ++] restricted to cofinality κ+ may coincide with nonstationary ideal NSκ+κ++. Next, let us ask what about I[κ++] restricted to cofinality κ? it turns out that we are back on track here, and this follows from the trivial fact that κ++ is a successor of a regular cardinal, together with the following less-trivial proposition:

Proposition (Shelah, 1991). E<λλ+I[λ+] for every regular cardinal λ.
Proof. Of course, we may assume that λ>0. For every ordinal α<λ+, fix an injection dα:αλ. Notice that for every α<δ<λ+:

  • i<λj[i,λ) such that dδ1[i]αdα1[j];
  • i<λj[i,λ) such that dδ1[j]αdα1[i].

It follows that for all α<δ<λ+, the following set is a club in λ: Cαδ:={i<λdδ1[i]α=dα1[i]}.

Next, for all α<λ+, let Dα:={dα1[i]γγα,i<λ}. Clearly, {Dαα<λ+}[P(λ+)]λ. We now fix an arbitrary limit ordinal δE<λλ+, and show that α<δDα contains all the initial segments of some small cofinal subset of δ.
Let u be a cofinal subset of δ of minimal order-type. In particular, |u|<λ, and hence C:=αuCαδ is a club in λ. Put i:=min(Csup(dδ[u])+1), and Aδ:=dδ1[i]. Then:

  • Aδδ;
  • uAδ, and hence sup(Aδ)=δ;
  • |Aδ|=|i|, and hence |Aδ|<λ;
  • for every αu, as iCαδ, we have Aδα=dδ1[i]α=dα1[i];
  • for every γ<δ, letting α:=min(uγ), we get that Aδγ=Aδαγ=dα1[i]γDα.

This completes the proof. ◻

We now arrive to the main result of today’s post:

Theorem (Shelah, 1993). If κ is a regular cardinal, and κ+<λ, then I[λ] contains a stationary subset of Eκλ.
Proof. We already know that EωλI[λ]. We also know that EκλI[λ] in the case λ=κ++. Thus, we shall assume that 0<κ<κ++<λ.
In an earlier post, we proved that E13 carries a club guessing sequence, and since here κ is regular and uncountable, the same argument shows that Eκκ++ carries a club-guessing sequence. Thus, let us fix such a club-guessing sequence C=CββEκκ++. Next, fix a large enough regular cardinal θ, and let M=Mαα<λ be an increasing sequence of elementary submodels of (Hθ,) such that for all α<λ:

  • |Mα|<λ;
  • κ+++α+1Mα+1;
  • C,λMα.

We shall now make a somewhat naive move, and simply let Dα:=MαP(λ) for all α<λ. Since Dα[P(λ)]<λ for all α<λ, the following set S is obviously in I[λ],S:={δEκλAδδ(otp(Aδ)<δ=sup(Aδ),γ<δα<δ[AδγDα])}.

On the other hand, the next statement is not entirely obvious.

Subclaim. S is stationary.
Proof. Given a club Dλ, we shall seek some δSD. Wlog, min(D)>κ.
Let N:=Nii<κ++ be a sequence of elementary submodels of (Hθ,) such that for all i<κ++:

  • |Ni|=κ++, with κ++Ni;
  • NjjiNi+1;
  • M,C,D,λNi;
  • Ni=j<iNj whenever i is a limit ordinal.

For all i<κ++, as D,λNi and sup(D)=λ, we get by elementarity that sup(NiD)=sup(Niλ). Recalling that D is closed, we infer that sup(Niλ)D for all i<κ++. Thus, it make sense to define a function f:κ++D by letting: f(i):=sup(Niλ),(i<κ++.)
By the second and fourth defining properties of N, we get that f is increasing and continuous, and that fτNτ+1 for all τ<κ++. Put ϵ:=sup(f[κ++]). Since κ+++ϵ+1Mϵ+1, we get (by elementarity) the existence of some strictly-increasing and cofinal function g:κ++ϵ that lies in Mϵ+1.
Since f and g are both continuous and cofinal in ϵ, a standard back-and-fourth argument yields the existence of a club cκ++ for which fc=gc. Since C is a club guessing sequence, we may find some βEκκ++ such that Cβc. Put δ:=f(β) and Aδ:=f[cβ]. Then δDEκλ, and Aδ is a cofinal subset of δ of order-type κ<min(D)δ. Thus, we are left with showing that Aδγα<δDα for all γAδ.
Fix γAδ, and let τ:=f1(γ). Then:

  • τ+1<β;
  • Aδγ=g[cβτ], and the latter belongs to Mϵ+1, since g,C,β,τMϵ+1 ;
  • Aδγ=(fτ)[cβ], and the latter belongs to Nτ+1 since fτ,C,βNτ+1.

Denote A:=Aδγ. Then AMϵ+1, and A,MNτ+1. Consequently: Nτ+1α<λ(AMα). It follows that there exists some α<sup(Nτ+1λ)=f(τ+1)<f(β)=δ such that AMα. In particular, Aδγα<δDα. ◻

So, S is an example of a stationary subset of Eκλ that belongs to I[λ]. ◻

1. Notice that in the above proof, we could have replaced κ++ with any regular cardinal μ<λ for which Eκμ carries a club-guessing sequence. In particular, the above proof shows:

Theorem (Shelah). If κ<κ+<μ<λ are all regular cardinals, then there exists a subset SEκλ such that:

  • SI[λ], and S reflects in the following sense:
  • {δEμλSδ is stationary} is stationary in λ. ◻

2. The proof may be adapted to show that if κ<κ+<λ are regular, then there exists a sequence CδδEκλ, and an enumeration {Dαα<λ}[P(λ)]<λ, such that for every club Dλ, there exists δacc(D) with:

  • otp(Cδ)=cf(δ)=κ;
  • Cδ is a club subset of Dδ;
  • for every γ<δ, there exists α<δ such that CδγDα.

This entry was posted in Blog, Expository and tagged , . Bookmark the permalink.

3 Responses to Shelah’s approachability ideal (part 1)

  1. Pingback: Shelah’s approachability ideal (part 2) | Assaf Rinot

  2. Pingback: Square principles | Assaf Rinot

  3. Mohammad says:

    I have a question. Assuming that the ideal is not trivial, is the forcing notion P(λ)/I[λ] interesting? What does forcing with it add?

Leave a Reply

Your email address will not be published. Required fields are marked *