---
title: Over-Saturated Sets in Extremal Combinatorics
url: https://www.emergentmind.com/topics/over-saturated-set
type: topic
---

# Over-Saturated Sets in Extremal Combinatorics

Searching arXiv for the cited papers and related terminology.
An over-saturated set is a set-system that lies beyond an extremal threshold for a forbidden configuration, so that one studies not merely whether the configuration must occur, but how many occurrences are forced. In the literature represented here, the term appears in two closely related senses. In finite posets, a subset of prescribed size is called minimally over-saturated, or supersaturated, when it minimizes the number of comparable pairs among all subsets of that size [1610.01521]. In Sperner theory, an oversaturated \(k\)-Sperner system is a family \(\mathcal F\subseteq\mathcal P(X)\) such that for every \(S\in\mathcal P(X)\setminus\mathcal F\), the number of \((k+1)\)-chains in \(\mathcal F\cup\{S\}\) is strictly larger than the number already present in \(\mathcal F\) [1402.5646]. A further manifestation arises in forbidden-intersection problems, where one fixes \(\ell\) and asks for the minimum number of pairs with intersection exactly \(t\) in an \(\ell\)-element family of \(k\)-subsets [2602.10292]. These formulations share a common extremal theme: once zero density of a forbidden relation becomes impossible, the central problem is to quantify the least unavoidable density.

## 1. Terminology and formal framework

For a monotone decreasing property \(\mathcal P\), a family \(\mathcal F\subseteq 2^{[n]}\) is \(\mathcal P\)-saturated if \(\mathcal F\subseteq\mathcal P\) and for every \(S\in 2^{[n]}\setminus\mathcal F\), the family \(\mathcal F\cup\{S\}\) fails to satisfy \(\mathcal P\) [1105.4453]. When \(\mathcal P\) is the \(k\)-Sperner property, this means that \(\mathcal F\) contains no chain
\[
F_1\subset F_2\subset\cdots\subset F_{k+1},
\]
but adjoining any outside set creates such a chain. The weak notion requires only that every outside set can be inserted into a \((k+1)\)-chain using members of \(\mathcal F\cup\{S\}\), while strong saturation requires both the \(k\)-Sperner condition and weak saturation [1105.4453].

Oversaturation strengthens this criterion. A family \(\mathcal F\subseteq\mathcal P(X)\) is an oversaturated \(k\)-Sperner system if, for every \(S\notin\mathcal F\), the number of \((k+1)\)-chains in \(\mathcal F\cup\{S\}\) is strictly larger than the number of \((k+1)\)-chains already present in \(\mathcal F\) [1402.5646]. The exposition also notes an equivalent viewpoint: one may start from any family \(\mathcal F\), possibly already containing chains, and require that adding any outside set creates strictly more new \((k+1)\)-chains than were before.

In poset language, let \((P,\le)\) be a finite poset and define, for \(S\subseteq P\),
\[
\mathrm{comp}(S)=\#\bigl\{\{x,y\}\subseteq S:x<y\text{ or }y<x\bigr\}.
\]
If \(m>\mathrm{width}(P)\), the basic supersaturation problem asks for the minimum of \(\mathrm{comp}(S)\) over all \(m\)-element subsets \(S\subseteq P\). A set achieving this minimum is called minimally over-saturated or supersaturated [1610.01521]. This formulation shifts attention from maximal forbidden-configuration avoidance to quantitative excess above the extremal boundary.

## 2. Over-saturation in finite posets

The general framework of Noel, Scott, and Sudakov studies supersaturation in finite posets via maximal chains and comparability digraphs [1610.01521]. A comparability digraph \(D=(V,E)\) orients each comparable pair exactly once. Given a distribution \(p\) on maximal chains of \(P\), one sets
\[
p_x=\Pr[x\in C], \qquad p_{x\to y}=\Pr[y\in C\mid x\in C].
\]
The key lemma in this framework bounds \(\mathrm{comp}(S)\) below using extremal values of \(p_x\) and \(p_{x\to y}\); in particular, if \(p_x\le\mu\) for all \(x\) and \(p_{x\to y}\le\alpha\) for all arcs, then
\[
\mathrm{comp}(S)\ge |S|-\frac{\mu}{\alpha}-1
\]
[1610.01521]. The method is a random-chain counting argument: the expected number of points of \(S\) on a random chain is compared against the expected number of comparable pairs of \(S\) that appear on that chain.

This framework yields explicit supersaturation theorems in several classical ranked posets. For the Boolean lattice \(P(n)=(\{0,1\}^n,\subseteq)\), if
\[
m=\sum_{r=0}^{k-1}\binom{n}{\left\lfloor \tfrac{n-k+1+2r}{2}\right\rfloor}
\quad\text{and}\quad |S|=m+t,
\]
then for sufficiently large \(n\),
\[
\mathrm{comp}(S)\ge t\binom{\lfloor (n+k)/2\rfloor}{k}
\]
[1610.01521]. Analogous statements hold for the subspace lattice \(V(q,n)\) and for the divisor poset \(\{0,1,2\}^n\), with \(q\)-binomial coefficients in the former and rank numbers \(\ell_i(n)\) in the latter. In each case, the lower bound has the form “excess size above the width” times a base comparable-pair constant.

The same paper uses these supersaturation bounds in a container-type lemma for posets. If every \(S\subseteq P\) with \(|S|>m\) satisfies \(\mathrm{comp}(S)\ge |S|d\), then every antichain is contained in \(T\cup f(T)\) for some small \(T\), where \(f\) is defined on subsets of size at most \(|P|/(2d+1)\) [1610.01521]. Iterated forms of this lemma are then used to count antichains and to analyze the largest antichain in \(p\)-random subsets.

## 3. Oversaturated \(k\)-Sperner systems

For a finite ground set \(X\), the parameter \(\osat(n,k)\) denotes the minimum size of an oversaturated \(k\)-Sperner family in \(\mathcal P(X)\) when \(|X|=n\), and \(\osat(k)=\lim_{n\to\infty}\osat(n,k)\) once stabilization is established by the standard atom-adding argument [1402.5646]. The main asymptotic problem is to determine the minimum order of magnitude of such families.

The state of the art summarized by Morrison, Noel, and Scott begins with Gerbner et al.’s lower bound
\[
\osat(n,k)=\osat(k)>2^{k/2-1}
\]
for every \(k\) and every \(n>k\), together with the earlier upper bound
\[
\osat(n,k)=\osat(k)=O\!\Bigl(\tfrac{\log k}{k}\,2^k\Bigr)
\]
[1402.5646]. Their improved construction shows that for all \(k\) and any ground set \(X\) with \(|X|\ge k^2+k\), there exists an oversaturated \(k\)-Sperner system \(\mathcal F\subseteq\mathcal P(X)\) with
\[
|\mathcal F|=O\!\bigl(k^5\,2^{k/2}\bigr).
\]
Combining this with the lower bound yields
\[
2^{k/2-1}<\osat(k)\le O\!\bigl(k^5\,2^{k/2}\bigr),
\qquad\text{hence}\qquad
\osat(k)=2^{(\frac12+o(1))k}
\]
[1402.5646]. Thus the correct exponential order is determined up to a polynomial factor.

The proof of the improved upper bound is based on a probabilistic covering construction. For each \(t\in[1,k^2+k]\), one finds small families \(\mathcal F_t,\mathcal G_t\subseteq\mathcal P(X)\) such that every pair \((F,G)\in\mathcal F_t\times\mathcal G_t\) satisfies \(|F|+|G|>k\), such that \(|\mathcal F_t|+|\mathcal G_t|=O(k^2 2^{k/2})\), and such that every \(t\)-subset \(S\subseteq X\) contains some \(F\in\mathcal F_t\) and is disjoint from some \(G\in\mathcal G_t\) [1402.5646]. One then grows each \(F\) upward and each \(G\) downward along chains of length at most \(k\), producing a global family of total size \(O(k^5 2^{k/2})\). The covering properties force the oversaturation condition for every \(S\notin\mathcal F\).

A recurring point of contrast is that the oversaturated problem is quantitatively sharper than classical saturation. The exposition explicitly notes that the oversaturated setting settles the correct exponential order, whereas for the corresponding saturation problem the minimum size behaves like \(2^{ck}\) for some unknown \(c\in[\tfrac12,1)\) [1402.5646].

## 4. Saturation, weak saturation, and flat antichains

Classical Sperner saturation provides the immediate backdrop for oversaturation. For \(1\le k\le n\), Gerbner et al. show the product-construction upper bound
\[
\mathrm{sat}(n,k)\le 2^{k-1},
\]
and an iterated doubling argument recovers the same estimate from \(\mathrm{sat}(n-1,k-1)\) [1105.4453]. For weak saturation, a covering-code argument gives
\[
\mathrm{wsat}(n,k)=O(\log k\,2^k),
\]
again independent of \(n\). On the lower-bound side, a covering argument implies
\[
\mathrm{wsat}(n,k)\ge 2^{n-k+1},
\]
and a more refined counting yields
\[
2^{k/2-1}\le \mathrm{wsat}(n,k)\le \mathrm{sat}(n,k)
\]
for all \(n\ge k\) [1105.4453]. The open problem formulated there asks whether \(\mathrm{sat}(n,k)=2^{k-1}\) for each fixed \(k\) and all large \(n\).

The same paper treats flat antichains, namely antichains consisting only of \(\ell\)-sets and \((\ell+1)\)-sets. If \(\mathcal F=\mathcal F_\ell\cup\mathcal F_{\ell+1}\) with \(\mathcal F_\ell\subseteq\binom{[n]}{\ell}\) and \(\mathcal F_{\ell+1}\subseteq\binom{[n]}{\ell+1}\), then \(\mathcal F\) is a saturating antichain if and only if
\[
\mathcal A(\mathcal F_{\ell+1})=\binom{[n]}{\ell}\setminus\mathcal F_\ell,
\qquad
\mathcal V(\mathcal F_\ell)=\binom{[n]}{\ell+1}\setminus\mathcal F_{\ell+1}
\]
[1105.4453]. For \(\ell=1\), Grütmüller, Hartmann, Kalinowski, Leck, and Roberts proved that every saturating family in \(\binom{[n]}2\cup\binom{[n]}3\) has size at least
\[
\binom{n}{2}-\Bigl\lfloor\frac{(n+1)^2}{4}\Bigr\rfloor,
\]
with equality characterized by a matching-construction [1105.4453].

These results do not define oversaturation directly, but they clarify the combinatorial environment in which over-saturation questions arise. A plausible implication is that saturation problems identify extremal obstructions, while oversaturation asks how rapidly forbidden structure accumulates once those obstructions are exceeded.

## 5. Supersaturation for forbidden intersections

A particularly sharp version of over-saturation appears in the Erdős–Sós forbidden-intersection problem. The generalized Johnson graph \(G(n,k,t)\) has vertex set \(\binom{[n]}{k}\) and edge set
\[
E=\bigl\{\{A,B\}:|A\cap B|=t\bigr\},
\]
so the extremal function
\[
ex(n,k,t)=\alpha(G(n,k,t))
\]
is the maximum size of a family \(\mathcal F\subset\binom{[n]}{k}\) with no two sets intersecting in exactly \(t\) points [2602.10292]. Frankl and Füredi showed that if \(k>2t+1\) and \(n\) is large, then
\[
ex(n,k,t)=\binom{n-t-1}{k-t-1}\sim \frac{n^{k-t-1}}{(k-t-1)!},
\]
whereas if \(k\le 2t+1\), then \(ex(n,k,t)=\Theta(n^t)\) [2602.10292].

The supersaturation function is
\[
S(n,k,t,\ell)
=\min_{\substack{\mathcal F\subset\binom{[n]}{k}\\|\mathcal F|=\ell}}
\bigl|\{\{A,B\}\subset\mathcal F:|A\cap B|=t\}\bigr|,
\]
equivalently the minimum number of edges in an induced \(\ell\)-vertex subgraph of \(G(n,k,t)\) [2602.10292]. A random \(\ell\)-subset of \(\binom{[n]}{k}\) has expected number of \(t\)-intersecting pairs approximately
\[
\frac{t!\binom{k}{t}^2}{2n^t}\,\ell^2,
\]
and the paper determines when this random prediction is asymptotically correct up to the leading constant.

The central threshold theorem states that for fixed \(k,t\), if \(k\ge 2t+1\) and
\[
n^{k-t-1}=o(\ell),\qquad \ell=o(n^{k-t}),
\]
then
\[
S(n,k,t,\ell)
=(1+o(1))\frac{t!}{2}\binom{k}{t}^2\frac{\ell^2}{n^t}
\]
as \(n\to\infty\) [2602.10292]. In particular, once \(\ell\gg ex(n,k,t)\), the minimum number of \(t\)-intersections in any \(\ell\)-family is of order \(\ell^2/n^t\).

The near-threshold regime is qualitatively different. Writing
\[
\ell=\binom{n-t-1}{k-t-1}+r,\qquad r=o(n^{k-t-1}),
\]
Theorem B gives, for \(k\ge 2t+3\),
\[
S(n,k,t,\ell)=\Theta(r\,n^{k-2t-1}),
\]
and Theorem C strengthens this to the exact formula
\[
S(n,k,t,\ell)
=r\binom{k}{t}\binom{n-k-t-1}{k-2t-1}
\]
when \(r=o(n^{k-2t-1})\) and \(n\) is large [2602.10292]. The extremal \(t\)-avoiding families here are \((t+1)\)-stars when \(k>2t+1\), while the \(k\le 2t+1\) regime is governed by design-like or Steiner-system constructions of size \(\Theta(n^t)\).

## 6. Structural methods, misconceptions, and open directions

Three proof paradigms recur across these over-saturation problems. The first is random-chain counting in posets, which furnishes lower bounds on forced comparable pairs and feeds directly into container lemmas [1610.01521]. The second is probabilistic covering, used in oversaturated \(k\)-Sperner systems to build small families that “witness” the effect of adding any outside set [1402.5646]. The third is sunflower or \(\Delta\)-system structure, together with Kruskal–Katona and Turán-type arguments, which drives the forbidden-intersection results of the Erdős–Sós setting [2602.10292].

In the intersection problem, the lower bounds are obtained by repeatedly applying a sunflower-structure theorem of Füredi or Jiang–Longbrake to peel off large \(k\)-partite subfamilies with strong regularity. Each piece either contains many \(t\)-intersections internally, via sunflower kernels of size \(t\), or else almost all of its sets share a common \((t+1)\)-kernel, producing a star-like structure [2602.10292]. Near the extremal threshold, the parameterization \(\ell=ex(n,k,t)+r\) permits a bootstrap argument: almost all sets lie in one \((t+1)\)-star, and the remaining \(r\) sets contribute \(\Theta(n^{k-2t-1})\) new edges each or yield the same order through mutual interaction.

A common misconception is to identify saturation and over-saturation as the same notion. The sources distinguish them sharply. Saturation requires that adding an outside element violates a forbidden-configuration condition; oversaturation requires a quantitative increase in the number of forbidden configurations [1402.5646]. Likewise, supersaturation in posets does not mean maximality at all: it is an extremal minimization problem above the width threshold [1610.01521]. A plausible implication is that “over-saturated set” is best understood as a family beyond the zero-density regime, with the precise meaning determined by the ambient structure and forbidden relation.

Several open problems remain. In Sperner theory, the conjecture \(\mathrm{sat}(n,k)=2^{k-1}\) for fixed \(k\) and large \(n\) is still unresolved [1105.4453]. In poset supersaturation, Noel, Scott, and Sudakov formulate conjectures for the \(r\)-ary vector poset \(\{0,1,\dots,r\}^n\), including a minimization conjecture for comparable pairs and a weak supersaturation conjecture; the exposition notes that Balogh, Petříčová, and Wagner disproved one of these minimization conjectures for all \(r\ge 2\), while many related cases remain open [1610.01521]. In the flat-antichain setting, the asymptotics for \(\ell\ge 2\) are tied to unknown Turán densities \(\pi_\ell\) [1105.4453]. These directions indicate that over-saturation is not a single theorem but a quantitative program in extremal set theory, linking exact thresholds, structure theorems, and counting methods across Boolean lattices, ranked posets, and forbidden-intersection graphs.

Source: https://www.emergentmind.com/topics/over-saturated-set