betaFeedback

WorksMathematics

A large ultrafilter number at ℵω\aleph_ω

Abstract

Gitik constructed a model where ℵω\aleph_ω is strong limit, 2ℵω=ℵω+22^{\aleph_ω}=\aleph_{ω+2} and every strongly uniform ultrafilter on ℵω\aleph_ω has character ℵω+2\aleph_{ω+2}. We show that in fact every uniform ultrafilter on ℵω\aleph_ω has character ℵω+2\aleph_{ω+2} in that model.

Introduction

An ultrafilter DD on an infinite set XX is uniform if ∣A∣=∣X∣|A|=|X| for every A∈DA\in D. For an ideal II on XX and A,B⊆XA,B\subseteq X, write A⊆IBA\subseteq_I B if A∖B∈IA\setminus B\in I. A base of DD modulo II is a family B⊆D\mathcal{B}\subseteq D such that for every A∈DA\in D there is some B∈BB\in\mathcal{B} such that B⊆IAB\subseteq_I A. Set

χI(D)=min⁡{∣B∣:B is a base of D modulo I},uI(κ)=min⁡{χI(D):D is uniform on κ}.\chi_I(D)=\min\{|\mathcal{B}|:\mathcal{B}\text{ is a base of }D\text{ modulo }I\},\qquad u_I(\kappa)=\min\{\chi_I(D):D\text{ is uniform on }\kappa\}.

So when II is the trivial ideal {∅}\{\varnothing\} these recover the usual character χ(D)\chi(D) and ultrafilter number u(κ)u(\kappa), and in general we have χI(D)≤χ(D)\chi_I(D)\leq\chi(D) and uI(κ)≤u(κ)u_I(\kappa)\leq u(\kappa).

We shall mainly be concerned with the case when XX is a singular cardinal κ\kappa and I=[κ]<κI=[\kappa]^{<\kappa}; here we also denote χI(D),uI(κ)\chi_I(D),u_I(\kappa) by χ∗(D),u∗(κ)\chi^*(D),u^*(\kappa) respectively. At one point we shall also consider the slightly more general case I=[X]<κI=[X]^{<\kappa} for some set XX; here we also denote χI(D)\chi_I(D) by χ<κ(D)\chi_{<\kappa}(D).

Garti and Shelah [3] initiated the study of ultrafilter numbers at singular cardinal; in particular they showed the consistency of κ+=u∗(κ)=u(κ)<2κ\kappa^+=u^*(\kappa)=u(\kappa)<2^\kappa relative to a supercompact cardinal. Garti, Gitik and Shelah [2] achieved this at ℵω\aleph_\omega and lowered the large cardinal hypothesis to the optimal one, a measurable cardinal κ\kappa with o(κ)=κ++o(\kappa)=\kappa^{++}. There have since been many further results on u(κ)u(\kappa) for singular κ\kappa [1, 5, 8].

A natural question is whether u∗(κ)>κ+u^*(\kappa)>\kappa^+ or even u(κ)>κ+u(\kappa)>\kappa^+ is consistent for a strong limit singular κ\kappa. Gitik [4] made important progress in this direction: starting from a model of GCH with a (κ,κ++)(\kappa,\kappa^{++})-extender, he constructed a model that satisfies the following:

  • (i) ℵω\aleph_\omega is a strong limit and 2ℵω=ℵω+22^{\aleph_\omega}=\aleph_{\omega+2};

  • (ii) every strongly uniform ultrafilter DD on ℵω\aleph_\omega satisfies χ∗(D)=ℵω+2\chi^*(D)=\aleph_{\omega+2}.

Here an ultrafilter DD on a singular cardinal κ\kappa is called strongly uniform if there is a strictly increasing cofinal sequence of regular cardinals ⟨λi:i<cf⁡(κ)⟩\langle\lambda_i:i<\operatorname{cf}(\kappa)\rangle below κ\kappa such that

{i<cf⁡(κ):∣A∩λi∣=λi} is unbounded in cf⁡(κ)(A∈D).\{i<\operatorname{cf}(\kappa):|A\cap\lambda_i|=\lambda_i\}\text{ is unbounded in }\operatorname{cf}(\kappa)\qquad(A\in D).

The exact value of u∗(ℵω)u^*(\aleph_\omega) in this model was left open, and Gitik suggested that it might be ℵω+1\aleph_{\omega+1}. Here we shall show that in fact it is ℵω+2\aleph_{\omega+2}, thus answering the above question in the positive. This is a simple consequence of two facts: a dichotomy about uniform ultrafilters on ℵω\aleph_\omega and a lower bound for the character of a regular ultrafilter under suitable pcf hypotheses.

Lemma 1. Let DD be a uniform ultrafilter on κ=ℵω\kappa=\aleph_\omega. Then either DD is (ω,κ)(\omega,\kappa)-regular, or DD is Rudin–Keisler above a strongly uniform ultrafilter on κ\kappa.

Recall that DD is (θ,κ)(\theta,\kappa)-regular if there is a family of κ\kappa many sets in DD such that the intersection of any θ\theta many of them is empty.

Lemma 2. Suppose ⟨λi:i<ω⟩\langle\lambda_i : i < \omega\rangle are strictly increasing regular cardinals cofinal in κ=ℵω\kappa= \aleph_\omega. Also suppose Λ>κ\Lambda> \kappa and the bounding number of (∏i<ωλi,<∗)(\prod_{i<\omega}\lambda_i,<^*) is at least Λ\Lambda. If DD is an ultrafilter on a set XX and DD is (θ,κ)(\theta,\kappa)-regular for some θ<κ\theta< \kappa, then χ<κ(D)≥Λ\chi_{<\kappa}(D) \geq\Lambda.

We only need this Lemma 2 for the case X=κX=\kappa, but we feel it better to distinguish between the underlying set XX and the regularity parameter κ\kappa.

Granted these two lemmas, we can prove u∗(ℵω)=ℵω+2\mathfrak{u}^*(\aleph_\omega)=\aleph_{\omega+2} in Gitik’s model as follows. Let DD be a uniform ultrafilter on ℵω\aleph_\omega. By Lemma 1 it is either (ω,κ)(\omega,\kappa)-regular or Rudin–Keisler above a strongly uniform ultrafilter UU. In the latter case it is easy to see that χ∗(D)≥χ∗(U)\chi^*(D)\geq\chi^*(U), and χ∗(U)=ℵω+2\chi^*(U)=\aleph_{\omega+2} by Gitik’s result, so χ∗(D)=ℵω+2\chi^*(D)=\aleph_{\omega+2}. In the former case, by standard facts from pcf theory there is some increasing sequence (λi:i<ω)(\lambda_i:i<\omega) of regular cardinals below ℵω\aleph_\omega that carries a scale of length ℵω+2\aleph_{\omega+2} (Gitik [4] also gave such a set explicitly), so in particular the bounding number of (∏i<ωλi,<∗)(\prod_{i<\omega}\lambda_i,<^*) is ℵω+2\aleph_{\omega+2}, and we apply Lemma 2 with Λ=κ++\Lambda=\kappa^{++}, X=κX=\kappa and θ=ω\theta=\omega.

It remains to prove the two lemmas.

Proof of Lemma 1. Our proof resembles that of Kanamori [6 Theorem 2.5]. First note that whenever a uniform ultrafilter WW on κ\kappa is not strongly uniform, we can project it as follows. The sequence (ℵi:i<ω)(\aleph_i:i<\omega) fails to witness strong uniformity, so there exists A∈WA\in W for which ∣A∩ℵi∣<ℵi|A\cap\aleph_i|<\aleph_i for all large enough i<ωi<\omega; by discarding an initial segment of AA, we may assume this is true for every ii, and moreover A∩ℵ0=∅A\cap\aleph_0=\varnothing. Take ℵ−1=0\aleph_{-1}=0. We can define an injection a:A→κa:A\to\kappa that maps A∩[ℵi,ℵi+1)A\cap[\aleph_i,\aleph_{i+1}) into [ℵi−1,ℵi)[\aleph_{i-1},\aleph_i) for each ii; note that a(x)<xa(x)<x. Extend aa to e:κ→κe:\kappa\to\kappa by letting ee be identity outside AA; it has the same projection effect as aa.

Now suppose DD is uniform and does not project to any strongly uniform ultrafilter on κ\kappa. Recursively for ξ<κ\xi<\kappa define fξ:κ→κf_\xi:\kappa\to\kappa and WξW_\xi as follows:

  1. f0=id⁡f_0=\operatorname{id}.

  2. Wξ=(fξ)∗DW_\xi=(f_\xi)_*D.

  3. Apply the above construction to WξW_\xi to get Aξ∈WξA_\xi\in W_\xi, aξ:Aξ→κa_\xi:A_\xi\to\kappa and eξ:κ→κe_\xi:\kappa\to\kappa, and let fξ+1=eξ∘fξf_{\xi+1}=e_\xi\circ f_\xi.

  4. For limit η\eta, let fη(x)f_\eta(x) be the eventual value of the non-increasing sequence (fξ(x):ξ<η)(f_\xi(x):\xi<\eta).

We need to verify that WξW_\xi is uniform in order for the recursion to go through. For ξ<κ\xi<\kappa let Hξ(x)={η<ξ:fη(x)≠fη+1(x)}H_\xi(x)=\{\eta<\xi:f_\eta(x)\ne f_{\eta+1}(x)\} be the finite set of stages where the value at xx changes. We claim that fξ(x)f_\xi(x) and Hξ(x)H_\xi(x) together recover xx; this is because the maps aηa_\eta are all injective. Thus uniformly for y<κy<\kappa we have

∣fξ−1(y)∣≤∣ξ∣<ω≤max⁡{ℵ0,∣ξ∣}<κ.|f_\xi^{-1}(y)|\leq|\xi|^{<\omega}\leq\max\{\aleph_0,|\xi|\}<\kappa.

Consequently, if C⊆κC\subseteq\kappa has size strictly smaller than κ\kappa then so does fξ−1[C]f_\xi^{-1}[C]. This shows Wξ=(fξ)∗DW_\xi=(f_\xi)_*D is uniform.

Now for ξ<κ\xi<\kappa let Xξ=fξ−1[Aξ]∈DX_\xi=f_\xi^{-1}[A_\xi]\in D. Note that x∈Xξx\in X_\xi if and only if fξ(x)>fξ+1(x)f_\xi(x)>f_{\xi+1}(x), so no xx can belong to infinitely many XξX_\xi’s. The sequence (Xξ:ξ<κ)(X_\xi:\xi<\kappa) then witnesses DD is (ω,κ)(\omega,\kappa)-regular. □

Proof of Lemma 2. Our proof resembles Solomon’s proof of b≤u\mathfrak{b}\leq\mathfrak{u} [7]. Suppose B⊆D\mathcal{B}\subseteq D has size smaller than Λ\Lambda. We are going to find a set Y⊆XY\subseteq X that splits every B∈BB\in\mathcal{B}, i.e., ∣B∩Y∣≥κ|B\cap Y|\geq\kappa and ∣B∩Yc∣≥κ|B\cap Y^c|\geq\kappa, so B\mathcal{B} cannot be a base modulo [X]<κ[X]^{<\kappa}.

By assumption there is a sequence (Xα:α<κ)(X_\alpha:\alpha<\kappa) such that each XαX_\alpha is in DD and the intersection of any θ\theta of them is empty; in other words, for each x∈Xx\in X the set r(x):={α<κ:x∈Xα}r(x):=\{\alpha<\kappa:x\in X_\alpha\} has size strictly smaller than θ\theta. Note that every set A∈DA\in D has size at least κ\kappa. Temporarily fix B∈BB \in\mathcal{B} and n<ωn < \omega. The set B∩XαB \cap X_\alpha has size at least κ=ℵω\kappa= \aleph_\omega, so we can choose Eα⊆B∩XαE_\alpha\subseteq B \cap X_\alpha with size ℵn\aleph_n. Let Zα=⋃x∈Eαr(x)Z_\alpha= \bigcup_{x \in E_\alpha} r(x), so ∣Zα∣≤ℵn⋅θ<ℵω|Z_\alpha| \le\aleph_n \cdot\theta< \aleph_\omega. Let (λi:i<ω)(\lambda_i : i < \omega) be as in the assumption; we may assume λ0>θ\lambda_0 > \theta. For i>ni > n, we can choose h(i)<λih(i) < \lambda_i above every ordinal in ⋃α<λi−1Zα∩λi\bigcup_{\alpha< \lambda_{i-1}} Z_\alpha\cap\lambda_i.

As B∈BB \in\mathcal{B} and n<ωn < \omega vary, we get fewer than Λ\Lambda many such functions hh, so by assumption they are <∗<^*-bounded by some g∈∏i<ωλig \in\prod_{i<\omega} \lambda_i; we may assume λi−1<g(i)<λi\lambda_{i-1} < g(i) < \lambda_i for i>0i > 0, so in particular gg is strictly increasing. Define

Y={x∈X:g−1[r(x)] has a largest element and it is even}.Y = \{x \in X : g^{-1}[r(x)] \text{ has a largest element and it is even}\}.

Let EαB,nE^{B,n}_\alpha, ZαB,nZ^{B,n}_\alpha and hB,nh^{B,n} be as above. Take N>nN > n such that hB,n(i)<g(i)h^{B,n}(i) < g(i) for all i≥Ni \ge N. We claim that for all i≥Ni \ge N and x∈Eg(i)B,nx \in E^{B,n}_{g(i)} we have max⁡g−1[r(x)]=i\max g^{-1}[r(x)] = i. Clearly i∈g−1[r(x)]i \in g^{-1}[r(x)] because EαB,n⊆XαE^{B,n}_\alpha\subseteq X_\alpha. Now suppose j>ij > i and g(j)∈r(x)g(j) \in r(x); we have g(i)<λi≤λj−1g(i) < \lambda_i \le\lambda_{j-1}, so g(j)∈Zg(i)B,n∩λjg(j) \in Z^{B,n}_{g(i)} \cap\lambda_j and g(j)<hB,n(j)g(j) < h^{B,n}(j), a contradiction.

Considering the even and odd i>Ni > N respectively and x∈Eg(i)B,nx \in E^{B,n}_{g(i)}, we see that both B∩YB \cap Y and B∩YcB \cap Y^c have size at least ℵn\aleph_n for any nn, hence at least ℵω\aleph_\omega, as desired. □

AI disclosure: no human intelligence was involved in writing this paper. All ideas are due to GPT 6 Astra.

References

Department of Mathematics, University of California, Irvine, Irvine, CA 92697

Email address: f [email protected]

References

  1. [1]Tom Benhamou and Sittinon Jirattikansakul. A small ultrafilter number at every singular cardinal. Acta Mathematica Hungarica, 171(1):12–38, 2023.arxiv.org/abs/2302.07311
  2. [2]Shimon Garti, Moti Gitik, and Saharon Shelah. Cardinal characteristics at ℵω. Acta Mathematica Hungarica, 160(2):320–336, 2020.
  3. [3]Shimon Garti and Saharon Shelah. The ultrafilter number for singular cardinals. Acta Mathematica Hungarica, 137(4):296–301, 2012.
  4. [4]Moti Gitik. On spectrum of strongly uniform ultrafilters over a singular cardinal, 2019. https://www.math.tau.ac.il/~gitik/str-spectrum2.pdf.DOI
  5. [5]Radek Honzik and Šárka Stejskalová. Small u(κ) at singular κ with compactness at κ⁺⁺. Archive for Mathematical Logic, 61(1):33–54, 2022.
  6. [6]Akihiro Kanamori. Weakly normal filters and irregular ultrafilters. Transactions of the American Mathematical Society, 220:393–399, 1976.DOI
  7. [7]R. C. Solomon. Families of sets and functions. Czechoslovak Mathematical Journal, 27(4):556–559, 1977.DOI
  8. [8]Toshimichi Usuba. Monotonicity of the ultrafilter number function. Archive for Mathematical Logic, 65(4):545–568, 2026.arxiv.org/abs/2501.14988

Paper details

vlnc:W_crmamesw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents