---
title: Ample Sets in Combinatorics and Design
url: https://www.emergentmind.com/topics/ample-sets
type: topic
---

# Ample Sets in Combinatorics and Design

In contemporary research usage, **ample sets** denotes at least two distinct technical notions. In combinatorics on the Boolean cube, an ample set is a family \(L\subseteq\{\pm1\}^E\) characterized by equality in the Dress–Pajor inequality and, equivalently, by a canonical \(\ell_1\)-geometric realization as a weakly convex cubical complex [2603.27835]. In design theory, the same phrase is also used for **large sets with multiplicity**, written \(LS(t,k,n;\mu)\), namely multisets of Steiner systems on a common point set such that every \(k\)-subset occurs in exactly \(\mu\) constituent systems [2007.09608]. These usages are mathematically unrelated, and the distinction is essential.

## 1. Terminological scope

The Boolean-cube notion originates in the theory of **lopsided sets**. Lopsided sets were introduced by Jim Lawrence in 1983 in connection with the intersection pattern of a convex set \(K\) with the orthants of \({\mathbb R}^E\), and Andreas Dress later called them **ample** sets [2603.27835]. In this setting, the ambient object is a family of sign vectors in \(\{\pm1\}^E\).

In design theory, by contrast, an ample set is a multiset of Steiner systems or related designs with a prescribed **multiplicity** parameter \(\mu\). The notation \(LS(t,k,n;\mu)\) is used for an ample set of Steiner systems \(S(t,k,n)\), and \(LH(n,g,k,t)\) for a large set of \(H\)-designs [2007.09608]. Here the ambient objects are designs on a finite point set rather than sign-vector families.

| Usage | Ambient object | Defining condition |
|---|---|---|
| Ample/lopsided set | \(L\subseteq\{\pm1\}^E\) | \(|L|=|\overline X(L)|\), equivalently \(|L|=|\underline X(L)|\) |
| Ample set in design theory | Multiset of \(S(t,k,n)\) | Every \(k\)-subset occurs in exactly \(\mu\) constituent systems |

A common source of confusion is that in the first usage “ample” is a structural equality condition, whereas in the second it is a uniform multiplicity condition.

## 2. Ample sets in the Boolean cube

Fix a finite ground set \(E\), with \(\#E=n\), and let \(L\subseteq\{\pm1\}^E\). For \(A\subseteq E\), define
\[
L_A:=\{\,s|_{E-A}: s\in L\},\qquad
L^A:=\{\,t\in\{\pm1\}^{E-A}: \text{every extension of }t\text{ to }E\text{ lies in }L\}.
\]
From these one obtains
\[
\overline X(L)=\{\,A\subseteq E : L_{E-A}=\{\pm1\}^A\},
\qquad
\underline X(L)=\{\,A\subseteq E : L^A\neq\varnothing\}.
\]
The simplices of \(\overline X(L)\) are the subsets of \(E\) **shattered** by \(L\), and those of \(\underline X(L)\) are **strongly shattered** [2603.27835].

Dress observed that every \(L\subseteq\{\pm1\}^E\) satisfies the combinatorial inequality
\[
\bigl|\underline X(L)\bigr| \;\le\; \bigl|L\bigr| \;\le\; \bigl|\overline X(L)\bigr|.
\]
An **ample** set is defined by equality on the right:
\[
L\text{ is ample}\quad\Longleftrightarrow\quad |L|=|\overline X(L)|.
\]
Equivalently, ampleness is the same as the “sparse” formulation
\[
|L|=|\underline X(L)|.
\]

This notion does not refer to cardinal largeness in any naive sense. Rather, it singles out those families for which the upper and lower combinatorial bounds collapse to equality. A plausible implication is that ampleness should be understood as a rigidity condition on the interaction between projections, extensions, and cube-combinatorics, rather than as a measure of size alone.

## 3. Cubihedra, weak convexity, and sign-vector structure

Let
\[
H(E)=[-1,+1]^E\subseteq{\mathbb R}^E
\]
be the standard \(n\)-cube, endowed with the \(\ell_1\)-metric
\[
d(r',r'')=\sum_{e\in E}|r'(e)-r''(e)|.
\]
For \(L\subseteq\{\pm1\}^E\), its **cubihedron** is
\[
|L|=\bigl\{\text{faces }F\subseteq H(E): F\cap\{\pm1\}^E\subseteq L\bigr\}.
\]
Its vertices are exactly the elements of \(L\), its edges are the pairs in \(L\) at Hamming distance \(1\), and its higher-dimensional faces are the subcubes all of whose vertices lie in \(L\) [2603.27835].

The decisive metric theorem states that for \(L\subseteq\{\pm1\}^E\), the following are equivalent: \(L\) is ample; \(|L|\subseteq({\mathbb R}^E,d)\) is **weakly convex**; \(|L|\) is **path–\(\ell_1\)-isometric** in \({\mathbb R}^E\); and every nonempty face of \(|L|\) is **gated** in the sense of Dress–Scharlau [2603.27835]. Here weak convexity means completeness together with **Menger-convexity**, namely that for any two distinct points \(x,y\in K\) there exists \(z\in K\setminus\{x,y\}\) such that
\[
d(x,y)=d(x,z)+d(z,y).
\]

This places ample sets among the cubical complexes that embed isometrically into \(\ell_1\)-spaces. The paper further shows that the cubihedra of ample sets endowed with the intrinsic \(\ell_1\)-metric are exactly the isometric subspaces of \(\ell_1\)-spaces, called **weakly convex sets** [2603.27835].

The face structure admits a sign-vector description in \(\{\pm1,0\}^E\). For a face \(F\subseteq H(E)\), its barycenter defines a sign vector \(t\in\{\pm1,0\}^E\). The set
\[
\Baryc(L)=\{\,t\in\{\pm1,0\}^E: F(t)\subseteq |L|\}
\]
satisfies
\[
\Baryc(L)=\uparrow\!\bigl(\Cocirc(L)\bigr)=\{\,t:\exists\,s\in L,\ t\preceq s\},
\]
where \(\Cocirc(L)=\Min(\Baryc(L))\) consists of the **cocircuits** of \(L\). The family \(\Baryc(L)\) satisfies a **signed-circuit axiom (SCA)**, and a set \(L\subseteq\{\pm1\}^E\) is ample if and only if its cocircuits satisfy SCA [2603.27835]. The paper explicitly frames this as an analogy with oriented matroids via covectors and cocircuits.

A further characterization concerns realizability. For a set \(K\subseteq{\mathbb R}^E\), let
\[
O(s)=\{\,r\in{\mathbb R}^E: r(e)s(e)\ge 0\ \forall e\in E\},
\qquad
L(K)=\{\,s\in\{\pm1\}^E: K\cap O(s)\neq\emptyset\}.
\]
Lawrence had shown that \(L(K)\) is ample whenever \(K\) is convex. The converse proved in the geometric paper is that \(L\subseteq\{\pm1\}^E\) is ample if and only if there exists a **weakly convex** set \(K\subseteq{\mathbb R}^E\) with \(L=L(K)\), and one may take \(K=|L|\subseteq H(E)\), compact and weakly convex [2603.27835].

## 4. Ample sets as large sets with multiplicity

In design theory, let \(Q\) be an \(n\)-element set. A Steiner system \(S(t,k,n)\) on \(Q\) is a collection \(\mathcal B\) of \(k\)-subsets such that each \(t\)-subset of \(Q\) lies in exactly one block. An **ample set** in this sense is written \(LS(t,k,n;\mu)\) and defined as a multiset \(\{S_1,\dots,S_M\}\) of Steiner systems \(S(t,k,n)\) on the same point set \(Q\), with the property that each \(k\)-subset \(B\subseteq Q\) occurs as a block in exactly \(\mu\) of the \(S_i\) [2007.09608].

Equivalently,
\[
M=\mu\cdot (n-t)!/(k-t)!/(n-k)!,
\]
and
\[
\forall\,B\in\binom{Q}{k},\quad |\{\,i:B\in S_i\,\}|=\mu.
\]
This is a direct generalization of an ordinary **large set**, which corresponds to multiplicity \(\mu=1\).

Existence is constrained first by the existence of the underlying Steiner system itself. The usual divisibility conditions require
\[
\binom{n-i}{t-i}/\binom{k-i}{t-i}\in{\mathbb Z}
\qquad (i=0,1,\dots,t-1).
\]
Hence \(LS(t,k,n;\mu)\) can exist only if \(S(t,k,n)\) exists. In the specific case of Steiner quadruple systems \(S(3,4,n)\), one needs
\[
n\equiv 2 \text{ or }4 \pmod 6,\qquad n\ge 4.
\]
For \(H\)-designs \(H(n,g,4,3)\), the necessary and sufficient conditions are
\[
gn\equiv 0 \pmod 2,\qquad g(n-1)(n-2)\equiv 0 \pmod 3,\qquad n\ge 4,\qquad (n,g)\neq (5,2)
\]
[2007.09608].

The design-theoretic literature emphasizes Steiner quadruple systems and related \(H\)-designs because explicit large sets are difficult to construct. The multiplicity formulation relaxes exact partitioning and replaces it by uniform coverage with parameter \(\mu\).

## 5. Constructions and existence theorems

The principal existence results in this direction are stated for both \(LS\) and \(LH\) families. For any integers \(g\ge 2\) and \(r\ge 2\) there exists an \(LH(2^r,g,4,3)\). For each \(\mu\ge 2\) there exists an \(LS(3,4,10;\mu)\). If there exists an \(LS(3,4,n;\mu)\) and a perpendicular array \(PA_2(2,n,n)\), then there exists an \(LS(3,4,2n;\mu)\); if \(LS(3,4,n;\mu)\) and \(PA_2(2,n,n)\) with \(\mu=(n-1)\) exist, then there is an \(LS(3,4,4n;\mu)\); and under the same hypotheses, for every \(m\ge 1\) there exists an \(LS(3,4,2^m n;\mu)\) [2007.09608].

| Ingredient | Output | Statement |
|---|---|---|
| \(OA(t,k,u)\) + \(LH(n,g,k,t)\) | \(LH(n,gu,k,t)\) | Theorem 4 |
| \(PA_1(k,n,n)\) + \(S(t,k,n)\) | \(LS(t,k,n;\mu)\) | Theorem 7 |
| \(LS(3,4,n;\mu)\) + \(PA_2(2,n,n)\) | \(LS(3,4,2n;\mu)\) | Theorem 16 |

The construction toolkit is heterogeneous. Orthogonal arrays \(OA(t,k,n)\) yield large sets of orthogonal arrays \(LOA(t,k,n)\) by a translation trick, and these feed into \(LH\)-constructions. Perpendicular arrays \(PA_1(k,n,n)\), equivalently \(k\)-homogeneous sets of permutations in \(S_n\), act on the blocks of a single Steiner system to produce \(LS(t,k,n;\mu)\). One-factorizations of the complete graph and Latin squares underlie the **doubling** constructions, while the **quadrupling** construction partitions \({\mathbb Z}_n\times{\mathbb Z}_4\) into four levels and organizes doubled systems on two-level slices together with inter-slice blocks of configuration \((1,1,1,1)\) [2007.09608].

The recursive theme is explicit. Starting from \(LS(3,4,n;\mu)\), one builds \(LS(3,4,2n;\mu)\), then \(LS(3,4,4n;\mu)\), and then continues by a Boolean \(SQS(2^m)\) gluing to reach \(LS(3,4,2^m n;\mu)\) for all \(m\ge 1\) [2007.09608]. The paper describes the general proof pattern as a partition of \(k\)-subsets by slice configurations, application of designs on each slice, addition of inter-slice blocks, and verification by counting that each relevant subset occurs exactly \(\mu\) times.

Several concrete examples are recorded. A computer-searched \(LS(4,5,11;2)\) on \(Q={\mathbb Z}_{11}\) consists of \(14\) Steiner \(S(4,5,11)\), with the other \(13\) obtained from one representative by explicitly listed coordinate permutations. From it one derives \(LS(3,4,10;2)\) by deleting a point and \(LS(5,6,12;2)\) by extension. An \(LS(3,4,10;3)\) is obtained by applying \(20\) explicit permutations to one \(S(3,4,10)\) derived from \(S(4,5,11)\). Known existence statements listed in the survey include \(LS(3,4,10;\mu)\) for all \(\mu\ge 2\), \(LS(4,5,11;2)\), \(LS(5,6,12;2)\), \(LS(3,4,14;720)\), and \(LS(3,4,20;9m)\) for all \(m\ge 1\) [2007.09608].

## 6. Related terminology and open directions

A third usage of the adjective appears in model theory as **very ampleness**. In that setting, a strongly minimal set \(X\) is called **very ample** if there exists a strongly minimal plane curve \(C\subseteq X^2\) whose generic type is very ample, and equivalently there is a definable very ample family of plane curves in \(X\) [2212.03774]. This is distinct from both the Boolean-cube notion of ample sets and the design-theoretic notion of large sets with multiplicity.

The model-theoretic results include: any strongly minimal set internal to an expansion of an algebraically closed field is very ample; if a strongly minimal set \(X\) is very ample and non-orthogonal to another strongly minimal set \(Y\), then \(X\) is internal to \(Y\); very ample strongly minimal sets admit very ample families of plane curves of all dimensions; and divisible strongly minimal groups are very ample [2212.03774]. These statements belong to stability theory and geometric model theory rather than to combinatorial design theory or Boolean-cube geometry.

For design-theoretic ample sets, the open problems explicitly listed are to construct \(PA_1(k,n,n)\) or \(PA_2(2,n,n)\) of minimal size for larger \(k,n\) to drive down \(\mu\); extend the quadrupling and recursion beyond \((t,k)=(3,4)\); find explicit small-\(n\) \(LS(t,k,n;\mu)\) for new parameter sets such as \(S(4,5,n)\) with multiplicity; and determine the exact minimal \(\mu(n,t,k)\) for which \(LS(t,k,n;\mu)\) exists [2007.09608].

Taken together, these lines of work show that **ample** is a term of strong local meaning rather than a universal concept. In one branch it identifies those Boolean-cube families whose combinatorics, cubical geometry, and sign-vector axioms coincide; in another it denotes uniformly multiplicative families of Steiner systems and \(H\)-designs; and in yet another it appears as part of the distinct model-theoretic notion of **very ampleness**.

Source: https://www.emergentmind.com/topics/ample-sets