---
title: Structure Theorem for Unions
url: https://www.emergentmind.com/topics/structure-theorem-for-unions
type: topic
---

# Structure Theorem for Unions

In extremal set theory, the expression *structure theorem for unions* commonly refers to structural descriptions of finite **union-closed families** that convert closure under unions into explicit frequency constraints on the ground elements. The central motivating problem is **Frankl’s union-closed sets conjecture**: every finite, nontrivial union-closed family should contain an element that belongs to at least half of its member sets. A decisive structural ingredient is the theorem of Falgas-Ravry for **separating** families, which organizes the family by element frequencies and produces canonical sets containing the most frequent elements; this framework is the basis of the small-family result of Bruhn and Schaudt, which verifies the conjecture for separating families up to an explicit threshold slightly above \(2m\), where \(m\) is the size of the universe [1508.05718].

## 1. Foundational framework

A family of sets \(\mathcal{A}\) is **union-closed** if for any \(A,B \in \mathcal{A}\), the union \(A \cup B\) also belongs to \(\mathcal{A}\). Its universe is
\[
U(\mathcal{A})=\bigcup_{A\in\mathcal{A}} A.
\]
For \(x \in U(\mathcal{A})\), the **frequency** of \(x\) is the number of sets of \(\mathcal{A}\) containing \(x\). A union-closed family is **separating** if for any two distinct elements \(x,y \in U(\mathcal{A})\), there exists a set in \(\mathcal{A}\) containing exactly one of \(x\) or \(y\) [1508.05718].

Frankl’s conjecture asserts that every finite, nontrivial union-closed family has an element of frequency at least \(|\mathcal{A}|/2\). The restriction to separating families is structurally natural because separation forces the family to distinguish every pair of ground elements through its member sets. In the small-family regime, this added rigidity is strong enough to produce canonical configurations and sharp counting arguments.

Later work broadened the framework around the conjecture by introducing notation that isolates containment and avoidance patterns. For \(B \subseteq [n]\),
\[
\mathcal{A}_B=\{A\in\mathcal{A}\mid A\cap B=B\}, \qquad
\mathcal{A}_{\underline{B}}=\{A\in\mathcal{A}\mid A\cap B=\emptyset\},
\]
so that one can compare how many member sets contain a prescribed block \(B\) and how many avoid it. This notation is central in generalized versions of the conjecture [2310.02482].

## 2. Falgas-Ravry’s structural theorem

The basic structure theorem begins by labeling the elements of the universe
\[
x_1,\dots,x_m
\]
in increasing order of frequency. Falgas-Ravry’s theorem states that there exist sets
\[
X_0,\dots,X_m \in \mathcal{A}
\]
such that
\[
x_i \notin X_i \quad \text{for } i=1,\dots,m,
\]
and
\[
\{x_{i+1},\dots,x_m\}\subset X_i \quad \text{for } i=0,\dots,m.
\]
Thus each \(X_i\) contains all sufficiently frequent elements while excluding a designated less frequent one [1508.05718].

This theorem supplies a canonical ladder of sets indexed by the frequency order. Its significance is twofold. First, it makes the frequency distribution visible inside the family itself: the tail \(\{x_{i+1},\dots,x_m\}\) must already occur as a common core of a member set. Second, it yields immediate lower bounds for highly frequent elements, because the most frequent elements appear across many of the sets \(X_i\).

An important corollary is that any separating union-closed family on \(m\) elements with at most \(2m\) member sets satisfies Frankl’s conjecture. In that form, the theorem functions as a structural certificate: if a counterexample exists among separating families, then its size must lie beyond the \(2m\) threshold.

## 3. The small-family theorem

Bruhn and Schaudt proved that the conjecture holds for separating union-closed families \(\mathcal{A}\) with universe size \(m\) whenever
\[
|\mathcal{A}| \le 2\left(m+\frac{m}{\log_2(m)-\log_2\log_2(m)}\right).
\]
Equivalently, every separating union-closed family in that range contains an element appearing in at least half of its member sets [1508.05718].

The proof combines the frequency-ordered structure above with a counting argument inspired by Knill. One constructs a minimal subset \(\hat{U}\) of the universe that **touches** every nonempty set of \(\mathcal{A}\), and then shows the existence of \(2^{|\hat{U}|}\) sets corresponding to all intersections with subsets of \(\hat{U}\). This produces a lower-complexity core inside the family. The argument is then balanced against frequency estimates coming from the separating structure.

The resulting threshold emerges from optimizing between an increasing term \(2^{k-1}\) and a decreasing term \(m/(k-2)\), where \(k\) is related to the size of such a minimal touching set. The significance of the theorem is precise but limited: it pushes the verified range beyond the classical \(2m\) barrier, yet the extra logarithmic term shrinks slowly as \(m\to\infty\). The supplied summary explicitly notes that this slow convergence toward \(2m\) indicates that substantially stronger bounds will require new ideas rather than a straightforward continuation of the same counting scheme [1508.05718].

## 4. Equivalent structural formalisms

A different structural direction replaces frequency counting by algebraic and order-theoretic descriptions of union-closed families. One basic duality is that \(\mathcal{F}\) is union-closed if and only if
\[
\mathcal{G}=\{[n]\setminus F : F\in\mathcal{F}\}
\]
is intersection-closed. Another is that \(\mathcal{F}\) is union-closed if and only if its complement in \(\mathcal{P}(n)\) is **simply rooted** [2208.03803].

A particularly useful cryptomorphic description uses **interior operators**. If \(\mathcal{F}\subseteq\mathcal{P}(n)\) is union-closed and \(\varnothing\in\mathcal{F}\), define
\[
T(X)=\bigcup\{F\in\mathcal{F}: F\subseteq X\}.
\]
Then \(T\) satisfies exclusivity \(T(X)\subseteq X\), monotonicity, and idempotence \(T(T(X))=T(X)\), and
\[
\mathcal{F}=\mathrm{Fix}(T)=\{X\subseteq[n]: T(X)=X\}.
\]
Conversely, every such interior operator determines a union-closed family. This establishes a bijection between union-closed families containing \(\varnothing\) and interior operators on \(\mathcal{P}(n)\).

The same viewpoint yields **congruence partitions** of \(\mathcal{P}(n)\): define \(X\sim Y\) if \(T(X)=T(Y)\). An equivalence relation is of this type exactly when
\[
A\sim B \implies (A\cap C)\sim(B\cap C)
\]
for all \(A,B,C\subseteq[n]\). Each congruence class is intersection-closed, and whenever \(A,B\) lie in the same class with \(A\subseteq B\), the whole interval
\[
[A,B]=\{X: A\subseteq X\subseteq B\}
\]
lies in that class as well. These reformulations do not resolve Frankl’s conjecture, but they replace the raw family by more rigid closure data, which is often more amenable to induction and decomposition.

## 5. Frequency hierarchies and generalized union conjectures

The structural program has also generated conjectural hierarchies extending Frankl’s original statement. If the ground set is ordered so that
\[
\#\{F\in\mathcal{F}:1\in F\}\ge \#\{F\in\mathcal{F}:2\in F\}\ge \cdots \ge \#\{F\in\mathcal{F}:n\in F\},
\]
then an iterated use of Frankl’s conjecture would imply
\[
\#\{F\in\mathcal{F}:k\in F\}\ge \frac{1}{2^k}\,|\mathcal{F}| \qquad \forall k\in[n].
\]
Nagel asked whether the stronger bound
\[
\#\{F\in\mathcal{F}:k\in F\}\ge \frac{1}{2^{k-1}+1}\,|\mathcal{F}|
\]
might hold for every union-closed family, and established partial structural evidence through fiber-counting lemmas and weak forms of the conjecture [2208.03803].

A second hierarchy is given by the conjectures
\[
\mathrm{UC}_x:\quad \exists B\in \binom{[n]}{n-x+1}\ \text{such that}\ |\mathcal{A}_B|\ge |\mathcal{A}_{\underline{B}}|.
\]
The extremal case \(x=n\) is exactly Frankl’s conjecture, because then \(B\) is a singleton. This family of statements interpolates between larger witness sets \(B\) and the original single-element claim. One structural implication proved in this framework is
\[
\mathrm{UC}_{n-1}\implies \mathrm{UC}_n,
\]
showing that a two-element witness statement would already force the classical conjecture [2310.02482].

The same paper develops strengthenings centered on **separating** families and the extremal role of power sets. In particular, it formulates the principle that in a separating union-closed family that is not a power set, the maximum element frequency should be strictly larger than \(|\mathcal{A}|/2\). This places power sets at the boundary of equality and treats them as the natural rigid models against which nontrivial families should be compared.

## 6. Weak structure theorems, implications, and limits

Even when the full conjecture remains out of reach, structural reformulations yield weaker but nontrivial substitutes. One example is the existence of large **up-sets** inside the ambient Boolean lattice that intersect a union-closed family in a controlled way: for every nontrivial union-closed \(\mathcal{F}\subseteq\mathcal{P}(n)\), there exists an up-set \(\mathcal{U}\subseteq\mathcal{P}(n)\) with
\[
|\mathcal{U}| \le 2^{n-1}
\quad\text{and}\quad
|\mathcal{F}\cap \mathcal{U}|\ge |\mathcal{U}|.
\]
More generally, for every \(t\in\mathbb{N}\), there exists an up-set \(\mathcal{U}\) with
\[
|\mathcal{U}| \le \left\lceil \frac{2^n}{t}\right\rceil
\quad\text{and}\quad
|\mathcal{F}\cap \mathcal{U}|\ge \frac{|\mathcal{F}|}{t}.
\]
These results do not identify an element of frequency \(|\mathcal{F}|/2\), but they isolate systematically large monotone regions of the Boolean lattice that must meet the family densely [2208.03803].

Another weak substitute concerns intersecting subfamilies. Nagel formulates the question whether every nontrivial union-closed family contains an intersecting subfamily of size at least a universal positive fraction of the whole family, and proves a quantitative frequency statement for arbitrary intersecting families:
\[
\#\{E\in\mathcal{G}: x\in E\}\ge \frac12\left[m+1+\sqrt{(m-1)(m-n)}\right]
\]
for some \(x\in[n]\), whenever \(\mathcal{G}\subseteq\mathcal{P}(n)\) is intersecting of size \(m\). Such statements suggest that part of the difficulty of Frankl’s conjecture lies in extracting intersecting or monotone substructure from a general union-closed family rather than in analyzing those substructures once found [2208.03803].

The present structural picture is therefore asymmetric. On one side, separating families admit rigid canonical sets, interior-operator descriptions, congruence partitions, and generalized witness conjectures. On the other, the strongest explicit small-family theorem in the supplied literature remains only slightly beyond \(2m\) for separating families. A plausible implication is that future advances will depend on blending these two viewpoints: the counting methods of the small-family theorem will likely need to interact with the more global closure structures revealed by interior operators, congruence classes, and the \(\mathrm{UC}_x\) hierarchy [1508.05718].

Source: https://www.emergentmind.com/topics/structure-theorem-for-unions