Laver Ultrafilters Explained

• Laver ultrafilters are nonprincipal ultrafilters on ω whose associated forcing notion exhibits the Laver property, controlling function growth in extensions.
• They are characterized via combinatorial and ideal-theoretic frameworks, employing partition bounds and level-counting techniques for deep structural insights.
• Their relationships with rapid P-points and hereditarily rapid ultrafilters reveal nuanced implications in forcing, cardinal invariants, and set-theoretic models.

Laver ultrafilters are a class of nonprincipal ultrafilters $\mathcal{U}$ on $\omega$ (the natural numbers) characterized by the property that the associated Laver forcing $\mathbb{L}_{\mathcal{U}}$ possesses the Laver property. Their introduction refines the landscape of combinatorial ultrafilter classes, clarifying structural distinctions among $P$-points, rapid ultrafilters, and frameworks based on ideals such as those of Baumgartner and Yorioka. Laver ultrafilters exhibit rich connections with measure-theoretic, combinatorial, and forcing-theoretic properties, and their existence and structure are sensitive to the ambient set-theoretic universe (Horvath et al., 1 Feb 2026).

1. Laver Property and Forcing Construction

A forcing notion $\mathbb{P}$ is said to have the Laver property if for every $\mathbb{P}$-name $\dot{f}$ for a function in $\omega^\omega$ and every ground-model $g\in\omega^\omega$ such that

$$\Vdash_{\mathbb{P}}\ \forall n\ (\dot{f}(n) < g(n)),$$

there is a ground-model sequence of finite sets $S = \langle S(n) : n\in\omega\rangle$ with $S(n)\subseteq g(n)$ and $|S(n)| \leq n+1$, such that

$$\Vdash_{\mathbb{P}}\ \forall n\ (\dot{f}(n) \in S(n)).$$

For a nonprincipal ultrafilter $\mathcal{U}$, the associated Laver forcing $\mathbb{L}_{\mathcal{U}}$ consists of trees $T\subseteq \omega^{<\omega}$ with a stem such that for every node $s$ extending the stem, the set of immediate successors $\mathrm{succ}_T(s)$ lies in $\mathcal{U}$. Ordering is by end-extension (i.e., reverse inclusion with fixed stem) (Horvath et al., 1 Feb 2026).

An ultrafilter $\mathcal{U}$ is a Laver ultrafilter if $\mathbb{L}_{\mathcal{U}}$ has the Laver property. This property admits combinatorial, ideal-theoretic, and level-counting characterizations, which clarify the operational content beyond the forcing language.

2. Combinatorial and Ideal-Theoretic Characterizations

Partition Characterization

$\mathcal{U}$ is Laver if for every sequence of finite partitions $\langle \mathcal{P}_n : n\in\omega \rangle$, there exists $x\in\mathcal{U}$ such that for all $n$,

$$\left| \{ P \in \mathcal{P}_n : x \cap P \neq \emptyset \} \right| \leq n+1.$$

A uniform bound with any unbounded function $h(n)$ in place of $n+1$ remains equivalent by a thinning argument (Tree-Coding Lemma, Fact 2.2).

Level-Counting and Ideals

Given $A\subseteq 2^\omega$, define $$\mathrm{level}_A(n) = |\{ x{\restriction} n : x\in A \}|$$. An ultrafilter $\mathcal{U}$ is Laver if for every $F: \omega\to 2^\omega$ and every non-decreasing unbounded $f: \omega\to\omega$, there is $x\in\mathcal{U}$ such that for all $n$,

$\mathrm{level}_{F[x]}(n) \leq f(n)+1.$

Let $$\mathcal{H} = \{f\in\omega^\omega: f \text{ non-decreasing, unbounded, } f(n)\leq n\}$$, and for $f\in\mathcal{H}$, define the ideal

$$\mathcal{I}_f = \left\{ A\subseteq 2^\omega : \forall d>0,\, \mathrm{level}_A(n)\leq^* (1/d)\,f(n) \right\}.$$

An ultrafilter $\mathcal{U}$ is Laver if and only if $\mathcal{U}$ is an $\mathcal{I}_f$-ultrafilter for every $f\in\mathcal{H}$ (Baumgartner-style characterization, Proposition 2.5).

3. Structural Relationships with Other Ultrafilter Classes

Laver ultrafilters form a combinatorially robust class situated between rapid $P$-points and more general hereditarily rapid or measure-zero ultrafilters. The following relationships are established (Horvath et al., 1 Feb 2026):

Property/Class	Laver Ultrafilter Position	Reference
Rapid $P$-points	Proper subset	Proposition 3.3
Hereditarily Rapid	Proper superset	Proposition 3.2
Measure Zero	Proper superset	Corollary 3.6
Yorioka Ideals	Contained	Corollary 3.6
Scattered	Not necessary	Theorem 3.8 (under MA)

• Downward Rudin–Keisler closure holds: if $\mathcal{U}$ is Laver and $\mathcal{V}\leq_{\mathrm{RK}} \mathcal{U}$, then $\mathcal{V}$ is also Laver (Fact 3.1).
• Every Laver ultrafilter is hereditarily rapid.
• Every rapid $P$-point is a Laver ultrafilter, but not all Laver ultrafilters are $P$-points.
• Sums of Laver ultrafilters (e.g., $\mathcal{U}-\sum_i \mathcal{V}_i$, with all Laver) are again Laver, so some Laver ultrafilters are not $P$-points.
• Every Laver ultrafilter is a $\mathcal{Y}_f^0$-ultrafilter (hence measure zero, hence nowhere dense).
• Under $\mathrm{MA}(\sigma$-linked), there exists a Laver ultrafilter which is not scattered.

4. Existence, Consistency, and Model-Theoretic Results

The generic existence of Laver ultrafilters can be parametrized by the minimal size $\mathfrak{ge}(\mathrm{Laver})$ of a filter base sufficient for diagonalization over all $\mathcal{I}_f$ for $f\in\mathcal{H}$. The cardinal relations are summarized as follows:

Cardinal Invariant	Bound Regarding $\mathfrak{ge}(\mathrm{Laver})$
$\mathrm{cov}(\mathcal{M})$	Lower bound ($\le$)
$\mathrm{non}(\mathcal{NA})$	Lower bound ($\le$)
$\mathrm{non}(\mathcal{E})$, $\mathfrak{d}$	Upper bound (max)
$\mathrm{non}(\mathcal{SN})$	Upper bound ($\le$)

Hence,

$$\mathrm{cov}(\mathcal{M}),\;\mathrm{non}(\mathcal{NA}) \leq \mathfrak{ge}(\mathrm{Laver}) \leq \min\left\{ \mathrm{non}(\mathcal{SN}),\; \max\{\mathrm{non}(\mathcal{E}), \mathfrak{d}\} \right\}.$$

In classic forcing models:

• Cohen model: $\mathrm{cov}(\mathcal{M}) = \mathfrak{c}$, so Laver ultrafilters exist generically.
• Random, Sacks models: $$\mathrm{non}(\mathcal{SN}) \leq \omega_1 < \mathfrak{c}$$, so Laver ultrafilters do not exist generically.
• Silver model: No Laver ultrafilters exist (Theorem 4.10).
• Mathias, Laver, Miller models: No rapid ultrafilters, thus no Laver ultrafilters.

Importantly, it is consistent that there are no $P$-points but Laver ultrafilters can exist generically. This is achieved by an iteration interleaving Shelah's $P$-point-destroying forcing $\mathbb{Q}(\mathcal{U})$ and a slalom-adding forcing $\mathbb{Q}_b$, ensuring that $\mathrm{non}(\mathcal{NA}) = \mathfrak{c}$ but no $P$-points survive (Theorem 4.13).

5. Proof Strategies and Technical Frameworks

Key tools for proofs regarding Laver ultrafilters leverage pure decision/fusion (Judah–Shelah methodology) for $\mathbb{L}_{\mathcal{U}}$:

• A fusion sequence of trees is constructed using partition-guided successors at each node and thinned by the combinatorial property.
• Ground-model slaloms $c(n)$ of size $\leq n^3$ can be produced, capturing relevant $\mathbb{P}$-names.
• Lower bounds for $\mathfrak{ge}(\mathrm{Laver})$ are established via standard arguments for Martin numbers in ccc posets and Pawlikowski’s slalom characterizations for null almost disjoint families.
• Non-existence in the Silver model is shown using case analysis on Silver-generic partitions, applying density and fusion techniques.
• Absence of $P$-points with generic existence of Laver ultrafilters is realized by interleaved iterations involving $\omega^\omega$-bounding and slalom-adding forcings.

6. Outstanding Problems and Research Directions

Fundamental questions remain regarding the structure and existence spectrum of Laver ultrafilters:

1. Does $\mathrm{MA}$ alone imply the existence of a hereditarily rapid, countable-closed ultrafilter that is not Laver?
2. Is it consistent that $$\max\{\mathrm{non}(\mathcal{NA}), \mathrm{cov}(\mathcal{M})\} < \mathfrak{ge}(\mathrm{Laver})$$?
3. Do Laver ultrafilters exist in the random-real model (as opposed to merely their generic existence)?

A plausible implication is that model-dependent hierarchies among combinatorial ultrafilter classes are more intricate than previously recognized, and their behavior under specific set-theoretic hypotheses is not yet fully classified (Horvath et al., 1 Feb 2026).