42.03.10 · mathematical-logic / set-theory-forcing

Club Sets, Stationary Sets, and Fodor's Lemma

shipped3 tiersLean: none

Anchor (Master): Kunen 2011 *Set Theory* (College Publications) Ch. II §6 and the Silver-theorem and ◊-application exercises; Jech 2003 *Set Theory* 3e (Springer) Ch. 8 (club filter, Fodor, Solovay splitting), Ch. 23 (stationary sets in [λ]^{<κ}, the generalised club filter for proper forcing), and Ch. 8/17 for the Mahlo and weakly compact characterisations; Devlin 1984 *Constructibility* (Springer) Ch. II for ◊ and Suslin trees built on the stationary-set machinery

Intuition Beginner

Imagine the positions below some enormous infinite stopping-point, laid out in a long line that climbs past every counting number and far beyond. We want a way to say that a collection of these positions is "almost all of them" — big enough that you cannot avoid it no matter how you travel up the line. The clean way to do this is to single out the well-behaved collections first and call those the unmissable ones.

A collection of positions is closed and unbounded, or a club, when it has two features. First, it never runs out: however high you climb, there is always a club position still higher. Second, it is closed: if a run of club positions piles up toward some ceiling that is still below the very top, that ceiling is itself a club position. A club is a thick, gap-free net stretched all the way up the line.

Now here is the key move. A collection is called stationary when it cannot dodge a single club: every club net catches at least one of its positions. Stationary collections are the ones that are "not negligible" — they keep showing up no matter which thick net you throw. They need not be thick themselves; they only need to be impossible to avoid.

The headline tool is a pressing-down rule. Suppose you have a stationary collection and you tag each of its positions with some strictly earlier position. Then a huge sub-collection — still stationary, still unmissable — all gets tagged with the very same earlier position. You cannot keep spreading the tags out; they bunch up. That bunching is the engine behind much of the arithmetic of the infinite.

Visual Beginner

A club is a thick, gap-free net climbing the whole line of positions; a stationary collection is one that every such net must touch. The picture contrasts a club with a stationary-but-thin collection.

   line of positions          a CLUB net            a STATIONARY set
   (climbs past every          (thick, reaches       (thin, scattered,
    counting number)           the top, no gaps)      but unavoidable)

      top ===                    | === reaches top      *   <- always lands
       |                         |    and is closed          on some club rung
       |                         | === under pile-ups   *
       |                         |                            *
       |                         | ===                   *
      ...                        | ===                  *  ...

The club net on the left is dense and runs to the very top; the scattered marks on the right are sparse, yet wherever you lay a club net the marks meet it. The next table shows the difference in plain words.

collection	thick / reaches the top?	can a club net miss it?
a club	yes, dense and unbounded	no — it is a net
a stationary set	not necessarily	no — it dodges nothing
a small "bounded" set	no, stops partway up	yes — easily missed

The key idea: clubs are the thick unmissable nets, stationary sets are the collections no net can dodge, and bounded sets that stop partway up are the negligible ones.

Worked example Beginner

Let us see by hand why the limit positions below the first uncountable stopping-point form a club. Write the stopping-point as the first position that comes after an uncountable run of earlier positions; call a position a limit if it has no single immediate predecessor (it sits just above an endless run of earlier positions, the way the first infinite position sits above all the counting numbers).

Step 1. Are the limit positions unbounded — do they keep appearing however high we climb? Pick any position $p$ . Climb one step, then another, taking a counting-number's worth of steps, each landing higher. The position these steps pile up toward has no immediate predecessor, so it is a limit, and it sits above $p$ . So there is always a limit position higher than $p$ .

Step 2. Are the limit positions closed — does a pile-up of limits land on a limit? Take a run of limit positions creeping up toward some ceiling below the top. The ceiling has earlier limit positions arbitrarily close below it, so it cannot have a single immediate predecessor either. So the ceiling is itself a limit position.

Step 3. Both features hold, so the limit positions form a club: a thick, gap-free, top-reaching net.

Step 4. Now a contrast. Consider only the positions below some fixed height $h$ (with $h$ well short of the top). This collection stops at $h$ ; it is bounded. The club of limit positions has members above $h$ , so this bounded collection misses that club entirely.

What this tells us: "club" packages exactly the two features — keeps appearing, and survives pile-ups — that make a collection a thick unmissable net, while a collection that stops partway up is negligible and dodgeable.

Check your understanding Beginner

Formal definition Intermediate+

Fix a regular uncountable cardinal $κ$ throughout; ordinals, cofinality $cf$ , and the regular/singular split are as in 42.03.04, and $cf (κ) = κ > ω$ is used repeatedly. We work with subsets of $κ$ , viewing $κ$ as the set of ordinals below it carrying the order topology, in which the limit points of a set $A \subseteq κ$ are the ordinals $γ < κ$ with $sup (A \cap γ) = γ$ .

A set $C \subseteq κ$ is closed if it contains all its limit points below $κ$ : whenever $γ < κ$ is a limit ordinal with $sup (C \cap γ) = γ$ , then $γ \in C$ . It is unbounded if $sup C = κ$ . A club (closed unbounded) set is one that is both. The club filter is $$ \mathcal{C}_\kappa = { X \subseteq \kappa : X \supseteq C \text{ for some club } C \subseteq \kappa }, $$ the collection of sets containing a club. Its members are said to hold on a club, or for almost all $α < κ$ .

A set $S \subseteq κ$ is stationary if $S \cap C \neq = \emptyset$ for every club $C \subseteq κ$ ; equivalently $S$ is not in the dual ideal of $C_{κ}$ . The non-stationary ideal is $$ \operatorname{NS}\kappa = { A \subseteq \kappa : A \cap C = \varnothing \text{ for some club } C } = { A : \kappa \setminus A \in \mathcal{C}\kappa }, $$ so a set is stationary precisely when it is not in $NS_{κ}$ . Every club is stationary, and every set containing a club is stationary, but stationary sets can be far thinner than clubs.

Given a sequence $⟨ C_{α} : α < κ ⟩$ of subsets of $κ$ , the diagonal intersection is $$ \bigtriangleup_{\alpha < \kappa} C_\alpha = { \xi < \kappa : \xi \in \textstyle\bigcap_{\alpha < \xi} C_\alpha }. $$ A filter $F$ on $κ$ is normal if it is closed under diagonal intersection: $C_{α} \in F$ for all $α$ implies $△_{α} C_{α} \in F$ . A function $f$ defined on $S \subseteq κ$ is regressive if $f (α) < α$ for every $α \in S$ with $α > 0$ .

Counterexamples to common slips Intermediate+

"Two stationary sets always meet." No: by Solovay splitting (below) a stationary set decomposes into $κ$ disjoint stationary sets, any two of which are disjoint. Stationary sets meet every club, not every other stationary set.
"A stationary set is unbounded but not conversely, and that is the whole difference." Unboundedness is necessary but far from sufficient. The successor ordinals below $κ$ are unbounded yet non-stationary: the limit ordinals form a club they entirely miss. Stationarity is meeting every club, a much stronger demand.
"The club filter is closed under arbitrary intersections because closedness is." Closedness is preserved by any intersection, but unboundedness is not: $⋂_{α < κ} (α, κ)$ is empty. The filter is only $κ$ -complete — closed under intersections of fewer than $κ$ clubs — which is exactly why diagonal intersection (a controlled $κ$ -fold operation) is the right closure to demand instead.
"Fodor's lemma needs only that $S$ is unbounded." It needs $S$ stationary. On the non-stationary unbounded set of successor ordinals the regressive map $α + 1 \mapsto α$ is injective, constant on no unbounded set at all — the hypothesis of stationarity is doing real work.

Key theorem with proof Intermediate+

The organising fact is that clubs are not merely a filter but a normal one: closed under the diagonal intersection that single-index intersection cannot reach. Normality is the precise strength that makes Fodor's pressing-down lemma true, and pressing-down is the tool every later application invokes.

Theorem (the club filter is $κ$ -complete and normal). Let $κ$ be regular uncountable. If $⟨ C_{α} : α < κ ⟩$ are clubs in $κ$ , then for every $λ < κ$ the intersection $⋂_{α < λ} C_{α}$ is a club, and the diagonal intersection $△_{α < κ} C_{α}$ is a club ^{[Kunen Ch. II §6]}.

Proof. For the $< κ$ intersection, fix $λ < κ$ and put $D = ⋂_{α < λ} C_{α}$ . Closedness is immediate: a limit point of $D$ is a limit point of each $C_{α}$ , hence in each $C_{α}$ , hence in $D$ . For unboundedness, fix $β_{0} < κ$ ; build an increasing sequence $⟨ β_{n} : n < ω ⟩$ where $β_{n + 1} < κ$ is chosen above $β_{n}$ so that each $C_{α}$ ( $α < λ$ ) has a member in the interval $(β_{n}, β_{n + 1})$ — possible because each $C_{α}$ is unbounded and $λ < κ = cf (κ)$ keeps the sup of these $λ$ -many choices below $κ$ . Let $γ = sup_{n} β_{n}$ ; then $cf (κ) > ω$ gives $γ < κ$ , and each $C_{α}$ meets cofinally in $γ$ , so $γ \in C_{α}$ for all $α < λ$ , whence $γ \in D$ and $γ > β_{0}$ .

For the diagonal intersection $D = △_{α} C_{α}$ , closedness: let $γ < κ$ be a limit point of $D$ and let $α < γ$ . Every $ξ \in D$ with $α < ξ < γ$ satisfies $ξ \in C_{α}$ (as $α < ξ$ ), and these $ξ$ are cofinal in $γ$ , so $γ$ is a limit point of $C_{α}$ and thus $γ \in C_{α}$ ; as $α < γ$ was arbitrary, $γ \in ⋂_{α < γ} C_{α}$ , i.e. $γ \in D$ . Unboundedness: given $β_{0}$ , set $β_{n + 1} =$ some element of $⋂_{α \leq β_{n}} C_{α}$ above $β_{n}$ (a $< κ$ intersection of clubs, club by the first part, hence unbounded). Then $γ = sup_{n} β_{n} < κ$ lies in $C_{α}$ for every $α < γ$ (each such $α$ is below some $β_{n}$ , and the tail $β_{n + 1}, β_{n + 2}, \dots \in C_{α}$ is cofinal in $γ$ with $γ$ closed in $C_{α}$ ), so $γ \in D$ . $□$

Bridge. Normality is the foundational reason the club filter behaves like a measure of "almost all" rather than a bookkeeping convenience: single-index intersections of $κ$ -many clubs can be empty, and the diagonal intersection is exactly the repair that survives. This is exactly the property Fodor's lemma converts into pressing-down, the way König's cofinality bound of 42.03.04 converts into the gimel calculus — the diagonal intersection is dual to a regressive function, in that $△_{α} C_{α}$ collects the ordinals that have already entered every earlier club, while a regressive $f$ sends each ordinal strictly backward. It builds toward Solovay splitting and Fodor below, and the central insight is that a normal filter cannot be pushed back: putting these together, if $f$ were regressive and non-constant on every stationary piece, the sets ${f = ν}$ would be non-stationary, their complements clubs, and the diagonal intersection of those clubs a club on which $f (ξ) \geq ξ$ , contradicting regressivity. This appears again in 42.03.06, where $♢$ guesses every subset on a stationary set, and in the Silver-theorem proof of 42.03.04, where the normal filter on $cf (κ)$ transfers GCH instances upward.

Exercises Intermediate+

Exercise 4 (medium, symbolic).

Prove that the intersection of two clubs $C, D \subseteq κ$ is a club, directly (without invoking the general $< κ$ theorem).

Hint

Closedness is automatic. For unboundedness, alternate: pick a point of $C$ , then of $D$ above it, then of $C$ , building an interleaved $ω$ -chain.

Answer

Closed: a limit point of $C \cap D$ is a limit point of both $C$ and $D$ , hence in both, hence in $C \cap D$ . Unbounded: fix $β_{0} < κ$ . Define $β_{2 n + 1} \in C$ above $β_{2 n}$ and $β_{2 n + 2} \in D$ above $β_{2 n + 1}$ , using unboundedness of $C$ and $D$ ; each step stays below $κ$ . Let $γ = sup_{n} β_{n} < κ$ (as $cf (κ) > ω$ ). The subsequence $β_{1}, β_{3}, \dots \in C$ is cofinal in $γ$ so $γ \in C$ ; likewise $β_{2}, β_{4}, \dots \in D$ gives $γ \in D$ . Hence $γ \in C \cap D$ with $γ > β_{0}$ . Rubric: full credit for the interleaving construction and the use of closedness on both cofinal subsequences.

Exercise 5 (medium, symbolic).

Show that the club filter is not an ultrafilter: exhibit a set $S \subseteq κ$ such that neither $S$ nor $κ ∖ S$ contains a club. (Use that a stationary co-stationary set exists, e.g. via Solovay splitting or directly.)

Hint

Split the limit ordinals by their own cofinality: $S = {α < κ : cf (α) = ω}$ when $κ \geq ℵ_{2}$ .

Answer

Take $κ \geq ℵ_{2}$ and $S = {α < κ : cf (α) = ω}$ , $T = {α < κ : cf (α) = ω_{1}}$ . Both are stationary: given any club $C$ , its $ω$ -th limit point (the sup of an $ω$ -chain in $C$ ) has cofinality $ω$ and lies in $C$ , so $C \cap S \neq = \emptyset$ ; similarly taking an $ω_{1}$ -chain in $C$ (possible as $C$ is club and $κ \geq ℵ_{2}$ ) yields a point of cofinality $ω_{1}$ in $C$ , so $C \cap T \neq = \emptyset$ . Since $S$ and $T$ are disjoint and both stationary, neither contains a club: a club inside $S$ would be disjoint from the stationary $T$ , impossible. Hence neither $S$ nor $κ ∖ S \supseteq T$ contains a club, so $C_{κ}$ is not an ultrafilter. Rubric: full credit for exhibiting two disjoint stationary sets by the cofinality split and concluding neither contains a club.

Exercise 6 (hard, symbolic).

Prove Fodor's pressing-down lemma: if $S \subseteq κ$ is stationary and $f : S \to κ$ is regressive, then there is a stationary $S^{'} \subseteq S$ on which $f$ is constant.

Hint

Suppose not: each fibre $f^{- 1} (ν)$ is non-stationary, so misses a club $C_{ν}$ . Diagonal-intersect the $C_{ν}$ and derive a contradiction with regressivity.

Answer

Suppose for contradiction that for every $ν < κ$ the fibre $S_{ν} = {α \in S : f (α) = ν}$ is non-stationary; pick a club $C_{ν}$ with $C_{ν} \cap S_{ν} = \emptyset$ . The diagonal intersection $C = △_{ν < κ} C_{ν}$ is a club (Key theorem), so $S \cap C$ is stationary, in particular nonempty; fix $α \in S \cap C$ with $α > 0$ . Since $α \in C = △_{ν} C_{ν}$ , we have $α \in C_{ν}$ for every $ν < α$ . By regressivity $f (α) < α$ , so with $ν = f (α) < α$ we get $α \in C_{f (α)}$ ; but $α \in S$ and $f (α) = f (α)$ put $α \in S_{f (α)}$ , so $α \in C_{f (α)} \cap S_{f (α)} = \emptyset$ , a contradiction. Hence some fibre $S_{ν}$ is stationary, and $f \equiv ν$ there. Rubric: full credit for assembling the clubs $C_{ν}$ , diagonal-intersecting, choosing $α$ in the stationary $S \cap C$ , and deriving the contradiction via $ν = f (α) < α$ .

Exercise 7 (hard, symbolic).

Let $f : κ \to κ$ and let $C_{f} = {γ < κ : f [γ] \subseteq γ}$ be the set of closure points of $f$ . Prove $C_{f}$ is a club, and deduce that for any countable family of functions $κ \to κ$ the common closure points form a club.

Hint

For unboundedness, iterate $f$ and take the $ω$ -limit. For closedness, a limit of closure points is a closure point.

Answer

Closed: if $γ = sup (C_{f} \cap γ)$ with $γ < κ$ a limit, then for $ξ < γ$ pick $δ \in C_{f} \cap γ$ with $ξ < δ$ ; $f (ξ) < δ < γ$ as $δ$ is a closure point, so $f [γ] \subseteq γ$ and $γ \in C_{f}$ . Unbounded: given $β_{0}$ , set $β_{n + 1} = sup ({β_{n} + 1} \cup f [β_{n}]) < κ$ (a $\leq ∣ β_{n} ∣ < κ$ supremum, below $κ$ by regularity). Then $γ = sup_{n} β_{n} < κ$ satisfies $f [γ] = ⋃_{n} f [β_{n}] \subseteq ⋃_{n} β_{n + 1} = γ$ , so $γ \in C_{f}$ with $γ > β_{0}$ . For a countable family ${f_{k} : k < ω}$ , $⋂_{k} C_{f_{k}}$ is a $< κ$ intersection of clubs (indeed a countable one, and $cf (κ) > ω$ ), hence a club by the Key theorem; equivalently it is the club of simultaneous closure points obtained by the same iteration interleaving all $f_{k}$ . Rubric: full credit for both club properties of $C_{f}$ and the reduction of the countable case to $κ$ -completeness.

Advanced results Master

The club filter $C_{κ}$ and its dual ideal $NS_{κ}$ form a normal $κ$ -complete filter/ideal pair on every regular uncountable $κ$ ; the substantive theorems describe how stationary sets behave under partition, how the apparatus lifts to large cardinals, and how it powers the cardinal arithmetic and inner-model results of the surrounding chapter.

Theorem 1 (Fodor's pressing-down lemma). A regressive function on a stationary $S \subseteq κ$ is constant on a stationary subset of $S$ ^{[Kunen Ch. II §6]}. Equivalently, $NS_{κ}$ is a normal ideal: the diagonal union of $NS_{κ}$ -sets is $NS_{κ}$ , and a function regressive on a positive set is bounded on a positive set. Fodor's lemma is the combinatorial core of the chapter; every reflection and guessing argument below routes through it.

Theorem 2 (Solovay's stationary splitting). Every stationary $S \subseteq κ$ is the disjoint union of $κ$ pairwise-disjoint stationary subsets ^{[Jech Ch. 8]}. Consequently $NS_{κ}$ is nowhere $κ^{+}$ -saturated in the crude sense, and the Boolean algebra $P (κ) / NS_{κ}$ is atomless on every stationary set. The proof on $S \cap {α : cf (α) = ω}$ assigns to each $α$ a cofinal $ω$ -sequence, applies Fodor to a regressive choice of its entries to find $κ$ -many stationary fibres, and refines; the general case reduces to this by partitioning according to cofinality.

Theorem 3 (Silver's theorem via the normal filter). If $κ$ is singular with $cf (κ) > ω$ and ${μ < κ : 2^{μ} = μ^{+}}$ is stationary in $κ$ along a continuous cofinal sequence, then $2^{κ} = κ^{+}$ ^{[Jech Ch. 8]}. The hypothesis $cf (κ) > ω$ is precisely what makes the club filter on $cf (κ)$ normal, so Fodor applies and the GCH instances below $κ$ on the stationary set transfer upward through a generic ultrapower; this is the stationary-set engine inside the singular-cardinals result of 42.03.04, and its failure for $cf (κ) = ω$ is why $\neg SCH$ is engineered at $ℵ_{ω}$ .

Theorem 4 ( $♢$ and Suslin trees). Jensen's principle $♢$ asserts a sequence $⟨ A_{α} : α < ω_{1} ⟩$ with $A_{α} \subseteq α$ such that ${α : X \cap α = A_{α}}$ is stationary for every $X \subseteq ω_{1}$ ^{[Devlin Ch. II]}. $♢$ holds in $L$ , implies CH, and constructs a Suslin tree by sealing every potential uncountable antichain on the stationary set of levels where the $♢$ -sequence guesses it. The stationarity of the guessing set is exactly what guarantees every candidate is caught; this is the application that the present unit makes available to 42.03.06.

Theorem 5 (Mahlo and weakly compact reflection). A regular uncountable $κ$ is Mahlo if the inaccessible cardinals below $κ$ form a stationary set, equivalently every club in $κ$ contains an inaccessible; $κ$ is weakly compact iff it is inaccessible and every stationary subset of $κ$ reflects — for some $α < κ$ , $S \cap α$ is stationary in $α$ ^{[Jech Ch. 8]}. Stationary reflection, ineffability (every $♢$ -like sequence has a stationary homogeneous set), and the Jónsson/Rowbottom partition properties stratify the large cardinals between weak compactness and measurability, all phrased in the stationary-set language of this unit; these feed directly into 42.03.09.

Synthesis. The foundational reason the club filter governs the infinite combinatorics of $κ$ is that it is normal: $κ$ -completeness alone would leave the diagonal operation uncontrolled, and normality is exactly the closure that makes Fodor's pressing-down lemma true. This is the central insight tying the chapter together — pressing-down is dual to the diagonal intersection, and the same normality that proves it powers Solovay splitting, Silver's transfer, and $♢$ -guessing. The stationary/non-stationary dichotomy generalises the bounded/unbounded dichotomy of 42.03.02 the way cardinal exponentiation generalises Cantor's diagonal in 42.03.04: bounded sets are negligible, clubs are conegligible, and stationary sets are the genuinely large remainder $NS_{κ}$ cannot swallow. Putting these together, the singular-cardinals theorem of 42.03.04 and the inner-model combinatorics of 42.03.06 are two faces of one apparatus — Silver transfers GCH up a stationary set, $♢$ guesses subsets along a stationary set — while the large-cardinal hierarchy of 42.03.09 is stratified by how strongly stationary sets reflect. The bridge is that "almost all $α < κ$ " is made precise by the normal filter, and every theorem here reads as a statement that some construction succeeds on a set the non-stationary ideal cannot reach.

Full proof set Master

Proposition 1 (the club filter is a $κ$ -complete filter). For regular uncountable $κ$ , $C_{κ}$ is a proper filter (it omits $\emptyset$ ) closed under intersections of fewer than $κ$ of its members.

Proof. $κ \in C_{κ}$ since $κ$ is a club in itself; $C_{κ}$ is upward closed by definition. For closure under $< κ$ intersections it suffices, by the superset definition, to intersect clubs: given clubs $⟨ C_{α} : α < λ ⟩$ with $λ < κ$ , $D = ⋂_{α} C_{α}$ is closed (a limit point of $D$ is a limit point of each $C_{α}$ , hence in each) and unbounded — given $β_{0}$ , build $β_{n + 1} < κ$ above $β_{n}$ with each $C_{α}$ meeting $(β_{n}, β_{n + 1})$ (the $λ$ -many witnesses have sup $< κ$ by regularity), and $γ = sup_{n} β_{n} < κ$ lies in every $C_{α}$ by closedness, so $γ \in D$ . Properness: every club is unbounded hence nonempty, so $\emptyset \in / C_{κ}$ . $□$

Proposition 2 (normality). If $⟨ C_{α} : α < κ ⟩$ are clubs then $△_{α} C_{α}$ is a club; hence $C_{κ}$ is normal.

Proof. Write $D = △_{α} C_{α} = {ξ : \forall α < ξ ξ \in C_{α}}$ . Closed: let $γ < κ$ be a limit point of $D$ and fix $α < γ$ . The elements $ξ \in D$ with $α < ξ < γ$ satisfy $ξ \in C_{α}$ , and they are cofinal in $γ$ , so $γ \in C_{α}$ by closedness of $C_{α}$ ; as $α < γ$ was arbitrary, $γ \in D$ . Unbounded: given $β_{0}$ , use that each $E_{δ} := ⋂_{α \leq δ} C_{α}$ is a club (Proposition 1) to pick $β_{n + 1} \in E_{β_{n}}$ above $β_{n}$ ; then $γ = sup_{n} β_{n} < κ$ , and for each $α < γ$ there is $n$ with $α \leq β_{n}$ , so the cofinal tail $β_{n + 1}, β_{n + 2}, \dots$ lies in $C_{α}$ , giving $γ \in C_{α}$ ; thus $γ \in D$ . $□$

Proposition 3 (Fodor's pressing-down lemma). If $S \subseteq κ$ is stationary and $f : S \to κ$ is regressive, then $f$ is constant on a stationary subset of $S$ .

Proof. If not, then for each $ν < κ$ the set $S_{ν} = {α \in S : f (α) = ν}$ is non-stationary, so there is a club $C_{ν}$ with $S_{ν} \cap C_{ν} = \emptyset$ . By Proposition 2 the diagonal intersection $C = △_{ν < κ} C_{ν}$ is a club, so $S \cap C$ is stationary, hence meets $κ ∖ {0}$ ; pick $α \in S \cap C$ , $α > 0$ . As $α \in C$ , $α \in C_{ν}$ for all $ν < α$ ; in particular for $ν = f (α) < α$ (regressivity), $α \in C_{f (α)}$ . But $α \in S$ and $f (α) = f (α)$ give $α \in S_{f (α)}$ , so $α \in S_{f (α)} \cap C_{f (α)} = \emptyset$ , absurd. Hence some $S_{ν}$ is stationary. $□$

Proposition 4 (stationarity is preserved by club intersection and equivalent reformulations). A set $S \subseteq κ$ is stationary iff $S \cap C$ is stationary for every club $C$ iff $κ ∖ S \in / C_{κ}$ ; moreover if $S$ is stationary and $C$ is a club then $S \cap C$ is stationary.

Proof. If $S$ is stationary and $C$ a club, then for any club $D$ the set $C \cap D$ is a club (Proposition 1), so $S \cap (C \cap D) \neq = \emptyset$ , i.e. $(S \cap C) \cap D \neq = \emptyset$ ; hence $S \cap C$ is stationary, giving the forward direction of the first equivalence (take $D = κ$ for the converse). The second equivalence is the definition of $NS_{κ}$ : $S$ meets every club iff no club avoids $S$ iff $κ ∖ S$ contains no club iff $κ ∖ S \in / C_{κ}$ . $□$

Proposition 5 (Solovay splitting on the $ω$ -cofinal part). Let $κ$ be regular uncountable and $S \subseteq {α < κ : cf (α) = ω}$ stationary. Then $S$ splits into $κ$ disjoint stationary sets.

Proof (structure). For each $α \in S$ fix an increasing cofinal $⟨ a_{n}^{α} : n < ω ⟩$ with limit $α$ . For fixed $n$ , the map $α \mapsto a_{n}^{α}$ is regressive on $S$ , so by Fodor it is constant, say $= β_{n}$ , on a stationary $S_{n} \subseteq S$ . If for some $n$ the value $β_{n}$ can be taken to range over $κ$ -many stationary fibres we are done; otherwise, were every "column" to use boundedly many values, the sequences $⟨ a_{n}^{α} ⟩_{n}$ would take fewer than $κ$ values overall, forcing two distinct $α, α^{'} \in S$ with identical cofinal sequences and hence $α = α^{'}$ , a contradiction once $S$ is stationary (so of size $κ$ ). Choosing, for each target value $ξ < κ$ , the stationary set ${α \in S : a_{n (ξ)}^{α} = ξ}$ along a coordinate $n (ξ)$ where the values spread to size $κ$ , and disjointifying, exhibits $κ$ pairwise-disjoint stationary subsets of $S$ . The full argument (refining to genuine disjointness across all $ξ$ simultaneously) is Solovay's; the load-bearing step is the Fodor reduction of each coordinate. $□$

Proposition 6 (a normal filter concentrates on limit ordinals; non-triviality of $NS_{κ}$ ). The club $Lim (κ)$ of limit ordinals belongs to $C_{κ}$ , so every stationary set has stationary intersection with $Lim (κ)$ ; consequently no normal filter on $κ$ extending $C_{κ}$ contains a regressive function's level set, and the successor ordinals are non-stationary.

Proof. $Lim (κ)$ is a club (Exercise 3): unbounded via $β \mapsto β + ω$ and closed since a limit of limit ordinals is a limit ordinal. Hence $Lim (κ) \in C_{κ}$ , and by Proposition 4 any stationary $S$ has $S \cap Lim (κ)$ stationary. The successor ordinals form the complement $κ ∖ Lim (κ)$ of a club, hence lie in $NS_{κ}$ . Finally, on $Lim (κ)$ the identity is not regressive but any regressive $f$ is, by Proposition 3, constant on a stationary set — so its level sets cannot all be non-stationary, which is the normality statement re-expressed. $□$

Connections Master

Cofinality, cardinal exponentiation, and the singular cardinals hypothesis 42.03.04 is the direct prerequisite and the first consumer: its proof of Silver's theorem — that SCH cannot first fail at a singular $κ$ of uncountable cofinality — runs entirely on the club filter and Fodor's lemma on $cf (κ)$ proved here. That unit states "the stationary set of approximating cardinals transfers GCH upward" and defers the machinery; this unit supplies it, and the dependence is exactly the normality that fails at $cf (κ) = ω$ , explaining why $\neg$ SCH is built at $ℵ_{ω}$ .
Martin's Axiom and large cardinals 42.03.09 is where the stationary-set apparatus becomes the language of the large-cardinal hierarchy: a Mahlo cardinal is one whose inaccessibles form a stationary set, a weakly compact cardinal is characterised by stationary reflection, and ineffable, Ramsey, and measurable cardinals are stratified by progressively stronger stationary partition and guessing properties. The normal filter of this unit is the finite-level shadow of the normal measures that define measurable cardinals there, and the generalised club filter on $[λ]^{< κ}$ introduced for proper forcing is the same construction relativised to small subsets.
The constructible universe $L$ and GCH 42.03.06 consumes the stationary-set theory through Jensen's $♢$ and $□$ principles: $♢$ guesses every subset of $ω_{1}$ on a stationary set of levels and so builds a Suslin tree, and these combinatorial principles hold in $L$ precisely because the fine structure produces club-many well-behaved levels. The ordinals indexing $L_{α}$ are the ordinals carrying the club filter analysed here, and the stationarity of the guessing set is the non-negotiable hypothesis that makes the $♢$ -construction catch every candidate.
Ordinals, transfinite induction, and recursion 42.03.02 supplies the substrate: clubs, stationary sets, and the diagonal intersection are all defined by reference to the order type and limit structure of $κ$ as a von Neumann ordinal, and the closure points of a function — the prototypical club — are produced by exactly the transfinite iteration that unit establishes. The bounded/unbounded dichotomy of the ordinal line is the degenerate $cf = ω$ shadow of the stationary/non-stationary dichotomy refined here.

Historical & philosophical context Master

The closed unbounded filter and its dual ideal emerged from the study of regressive functions in the 1950s. Géza Fodor proved the pressing-down lemma in 1956, Eine Bemerkung zur Theorie der regressiven Funktionen (Acta Sci. Math. Szeged 17), building on earlier partial results of Paul Erdős, Géza Fodor, and Alfréd Tarski ^{[Kunen Ch. II §6]}; the statement that a regressive function on a stationary set is constant on a stationary set had appeared in weaker forms (the "free set" and "regressive function" problems of the Hungarian combinatorial school) and Fodor's theorem gave it its definitive normal-filter formulation. The recognition that the club sets form a normal $κ$ -complete filter, and that stationary sets are its positive sets, organised a scattered body of results into the calculus used today.

Robert Solovay proved the stationary splitting theorem in the late 1960s, establishing that every stationary subset of a regular uncountable cardinal divides into $κ$ disjoint stationary pieces ^{[Jech Ch. 8]} and thereby that the non-stationary ideal is far from prime; the result underlies the modern theory of saturated ideals and generic ultrapowers. The apparatus reached its widest application through Jack Silver's 1974 theorem on the singular cardinals problem and through Ronald Jensen's fine-structure principles $♢$ and $□$ , isolated in his 1972 The fine structure of the constructible hierarchy (Ann. Math. Logic 4) and developed in Keith Devlin's Constructibility ^{[Devlin Ch. II]}, where $♢$ 's stationary guessing builds a Suslin tree and refutes the naive hope that CH settles the Suslin problem. Paul Mahlo's 1911 study of the cardinals now bearing his name, defined by the stationarity of the inaccessibles below them, predates the filter formulation and was later recast in exactly the stationary-set language of Kunen ^{[Kunen Ch. II §6]} and Jech ^{[Jech Ch. 8]}, from which the present treatment is drawn.

Bibliography Master

@book{kunen2011set,
  author    = {Kunen, Kenneth},
  title     = {Set Theory},
  series    = {Studies in Logic},
  volume    = {34},
  publisher = {College Publications},
  year      = {2011}
}

@book{jech2003set,
  author    = {Jech, Thomas},
  title     = {Set Theory},
  edition   = {3},
  series    = {Springer Monographs in Mathematics},
  publisher = {Springer},
  year      = {2003}
}

@book{devlin1984constructibility,
  author    = {Devlin, Keith J.},
  title     = {Constructibility},
  series    = {Perspectives in Mathematical Logic},
  publisher = {Springer},
  year      = {1984}
}

@article{fodor1956bemerkung,
  author  = {Fodor, G{\'e}za},
  title   = {Eine Bemerkung zur Theorie der regressiven Funktionen},
  journal = {Acta Scientiarum Mathematicarum (Szeged)},
  volume  = {17},
  year    = {1956},
  pages   = {139--142}
}

@article{jensen1972fine,
  author  = {Jensen, Ronald B.},
  title   = {The fine structure of the constructible hierarchy},
  journal = {Annals of Mathematical Logic},
  volume  = {4},
  year    = {1972},
  pages   = {229--308}
}

@article{mahlo1911ueber,
  author  = {Mahlo, Paul},
  title   = {{\"U}ber lineare transfinite Mengen},
  journal = {Berichte {\"u}ber die Verhandlungen der K{\"o}niglich S{\"a}chsischen Gesellschaft der Wissenschaften zu Leipzig},
  volume  = {63},
  year    = {1911},
  pages   = {187--225}
}

Prerequisites

42.03.04

Tier anchors

beginner: Hrbacek-Jech 1999 *Introduction to Set Theory* 3e (Marcel Dekker) Ch. 11-12 (cofinality, closed unbounded and stationary sets read informally before the alephs are pushed hard); Halmos 1960 *Naive Set Theory* (Van Nostrand) §17-19 (well-orderings and limit-ordinal pictures by hand, for the bottom-rung intuition of "happening cofinally often")
intermediate: Kunen 2011 *Set Theory* (College Publications) Ch. II §6 (closed unbounded sets in a regular uncountable κ, the club filter, the closure of the club filter under intersections of fewer than κ clubs and under diagonal intersection — normality — stationary sets and the non-stationary ideal, Fodor's pressing-down lemma with proof); Jech 2003 *Set Theory* 3e (Springer) Ch. 8 (the closed unbounded filter, stationary sets, Fodor's theorem, Solovay's splitting theorem, the Mahlo and weakly compact reflection picture)
master: Kunen 2011 *Set Theory* (College Publications) Ch. II §6 and the Silver-theorem and ◊-application exercises; Jech 2003 *Set Theory* 3e (Springer) Ch. 8 (club filter, Fodor, Solovay splitting), Ch. 23 (stationary sets in [λ]^{<κ}, the generalised club filter for proper forcing), and Ch. 8/17 for the Mahlo and weakly compact characterisations; Devlin 1984 *Constructibility* (Springer) Ch. II for ◊ and Suslin trees built on the stationary-set machinery

References

Kunen, K. — Set Theory · College Publications, Studies in Logic 34 (2011, revised edition). Chapter II §6 develops the infinite combinatorics of a regular uncountable cardinal κ. A set C ⊆ κ is closed unbounded (club) if it is unbounded (sup C = κ) and closed in the order topology (sup(C ∩ β) ∈ C whenever sup(C ∩ β) < κ, equivalently C contains the sup of every bounded subset of itself). Kunen proves the club filter is κ-complete: the intersection of fewer than κ clubs is again club (closedness is preserved by arbitrary intersection; unboundedness survives a <κ intersection by a back-and-forth argument building an increasing ω-chain whose limit lies in every club, using cf(κ) > ω). He proves normality: the club filter is closed under the diagonal intersection △_{α<κ} C_α = {ξ < κ : ξ ∈ ⋂_{α<ξ} C_α}, which is club whenever each C_α is club. A set S ⊆ κ is stationary if it meets every club; the stationary sets are the C-positive sets of the club filter, and their complements form the non-stationary ideal NS_κ, a κ-complete normal ideal. Fodor's pressing-down lemma is proved: if S is stationary and f : S → κ is regressive (f(α) < α for all α > 0 in S), then f is constant on a stationary subset of S; equivalently the club filter is normal in the sense that a regressive function on a stationary set cannot be injective on a club's worth of points. Applications include the splitting of stationary sets, the reflection picture for Mahlo cardinals, and the proof of Silver's theorem on the singular cardinals hypothesis (cross-referenced to the cardinal-arithmetic chapter).
Jech, T. — Set Theory · Springer Monographs in Mathematics, 3rd millennium edition (2003). Chapter 8 (Stationary Sets) develops the closed unbounded filter on a regular uncountable κ, proves κ-completeness and normality (closure under diagonal intersection), defines stationary sets as those meeting every club and the non-stationary ideal NS_κ as the dual normal ideal, and proves Fodor's theorem (every regressive function on a stationary set is constant on a stationary set). Jech proves Solovay's splitting theorem: every stationary subset of a regular uncountable κ is the disjoint union of κ pairwise-disjoint stationary sets, so NS_κ is nowhere κ^+-saturated in this elementary form. The chapter develops the generalised notion of club and stationary subsets of [λ]^{<κ} (the κ-club filter on the set of small subsets of λ), which underpins the theory of proper forcing and the reflection principles, and connects stationarity to Mahlo cardinals (κ is Mahlo iff the inaccessibles below κ form a stationary set) and to weakly compact cardinals (Π^1_1-indescribability and the tree property, characterised through stationary reflection). Silver's theorem (Chapter 8) that the GCH/SCH cannot first fail at a singular of uncountable cofinality is proved using the stationarity of the set of approximating points and the normality of the club filter on cf(κ).
Devlin, K. — Constructibility · Perspectives in Mathematical Logic, Springer (1984). Chapter II develops Jensen's combinatorial principle ◊ (diamond): there is a sequence ⟨A_α : α < ω_1⟩ with A_α ⊆ α such that for every X ⊆ ω_1 the set {α < ω_1 : X ∩ α = A_α} is stationary. ◊ holds in the constructible universe L and implies the continuum hypothesis; its strength over CH is precisely the stationary-set guessing, and Devlin uses it to construct a Suslin tree (a tree of height ω_1 with no uncountable chain or antichain), the canonical application that the club/stationary apparatus of the present unit makes possible. The construction proceeds by recursion along ω_1 using the ◊-sequence to anticipate, on a stationary set of levels, every potential uncountable antichain and seal it off, the stationarity guaranteeing the anticipations actually catch every candidate.

Estimated time

beginner: 20m
intermediate: 50m
master: 90m