---
title: 'M-Sets: Diverse Definitions in Mathematics'
url: https://www.emergentmind.com/topics/m-sets
type: topic
---

# M-Sets: Diverse Definitions in Mathematics

Searching arXiv for recent and relevant papers on “M-sets” and closely related usages.
In current mathematical usage, **M-set** is not a single standardized object but a family of distinct notions whose meanings depend strongly on context. In additive combinatorics it commonly abbreviates a **More Sums Than Differences** set; in lattice and polynomial theory it refers to a **maximal mediated set**; in categorical and semigroup-theoretic settings it denotes sets with a monoid action; and in other literatures it appears in multiset topology, discrete convex analysis, harmonic analysis, and complex dynamics [1509.01657], [1910.00502], [2011.11747], [1403.5642], [2403.07751], [2508.00182], [1810.06982].

## 1. Terminological range

The principal meanings attested in recent arXiv literature are summarized below.

| Usage | Core definition | Representative source |
|---|---|---|
| **MSTD set** | finite \(A\subseteq \mathbb Z\) with \(|A+A|>|A-A|\) | [1509.01657] |
| **Maximal mediated set** | largest \(\Delta\)-mediated lattice subset \(\Delta^*\) | [1910.00502] |
| **Topos of \(M\)-sets** | category \(\mathbf S^M\) of right \(M\)-sets for a monoid \(M\) | [2011.11747] |
| **M-set in multiset topology** | multiset represented by a count function \(C_M:X\to W\) | [1403.5642] |
| **\(M\)-convex set** | lattice set satisfying the discrete exchange axiom | [2403.07751] |
| **Walsh \(M\)-set** | set outside which a nontrivial Walsh null-series converges to \(0\) | [2508.00182] |

This terminological plurality reflects disciplinary separation rather than a shared universal definition. In several cases the initial “\(M\)” encodes a specific structural source—**more sums**, **mediated**, **monoid**, **multiset**, or **Murota-style \(M\)-convexity**—and the resulting theories are mathematically unrelated except at the level of nomenclature.

## 2. MSTD and generalized MSTD sets

In additive combinatorics, an **MSTD set** is a finite set \(A\subseteq \mathbb Z\) such that
\[
|A+A|>|A-A|.
\]
The generalized form compares two linear combinations with equal total weight:
\[
|sA-dA|>|\sigma A-\delta A| \qquad \text{with } s+d=\sigma+\delta.
\]
A central mechanism is the **fringe pair** formalism, in which the decisive contribution comes from the left and right fringes of a set inside an interval \([0,n]\). For a generalized MSTD fringe pair \((L,R;k)\), the inequality is forced at the level of truncated fringe sum/difference sets; a set is then made **rich** by arranging that the middle interval of the sumset is completely filled, so the global inequality is inherited from the fringes. This framework yields explicit and efficient constructions of \(k\)-generational and super \(k\)-generational examples, including the explicit choice
\[
L=\{0\},\qquad R=\{0,1,3\},
\]
and the concrete rich set
\[
A=\{0\} \cup [6k+1,12k+1] \cup (18k+2-\{0,1,3\}),
\]
which is \(k\)-generational and has only \(6k+5\) elements. The same paper proves that for any integer \(x\), a positive proportion of sets satisfy
\[
|sA-dA|-|\sigma A-\delta A|=x,
\]
and that for uniformly random \(A\subseteq[0,n]\),
\[
\lim_{n\to\infty}\Pr\!\left(1-\epsilon < \frac{\log|sA-dA|}{\log|\sigma A-\delta A|} < 1+\epsilon\right)=1.
\]
It also constructs a set with the then-largest known value
\[
\frac{\log|A+A|}{\log|A-A|}=\frac{\log 892}{\log 765}=1.02313,
\]
and develops **bi-MSTD** sets, for which both \(A\) and its complement in its convex hull are MSTD [1509.01657].

The earlier survey literature places these constructions in a broader probabilistic context. Under the uniform model on subsets of \(\{0,\dots,n-1\}\), a positive percentage of sets are MSTD; the survey records the lower bounds \(2\times 10^{-7}\) of Martin and O’Bryant and \(4\times 10^{-4}\) of Zhao. It also treats explicit constructions such as base expansion, Nathanson’s “one-point perturbation of a symmetric set,” \(P_n\)-sets, and the Miller–Orosz–Scheinerman and Zhao families. By contrast, if each element is chosen independently with probability \(p(n)\) where \(n^{-1}=o(p(n))\) and \(p(n)=o(1)\), then the probability of being difference-dominated tends to \(1\); the paper isolates the threshold behavior near \(p(n)\approx n^{-1/2}\). The same source formulates \(k\)-generational sum-dominance and proves that for every positive integer \(k\), a positive percentage of sets are \(k\)-generational, while no set is \(k\)-generational for all \(k\) [1107.2719].

A structural reduction due to Nathanson shows that every finite set of real numbers is Freiman isomorphic to a finite set of integers. As a consequence, there is no MSTD set \(A\) of real numbers with \(|A|\le 7\), and, up to Freiman isomorphism, there is exactly one MSTD set \(A\) of real numbers with \(|A|=8\), represented by
\[
\{0,2,3,4,7,11,12,14\}.
\]
The paper further proves that, up to affine isomorphism, this is the unique \(8\)-element real MSTD set [1609.04578].

The one-dimensional fringe philosophy also extends to higher dimensions. Explicit constructions in \(\mathbb Z^d\) use higher-dimensional fringe pieces along edges, faces, and corners together with a cross-shaped filled middle. The resulting theory gives generalized MSTD sets in every prescribed dimension, chains of simultaneous generalized inequalities, and \(k\)-generational sets satisfying
\[
|cA+cA|>|cA-cA| \quad \text{for every }1\le c\le k.
\]
The same work proves, under specified geometric hypotheses, that one cannot have
\[
|kA+kA|>|kA-kA| \quad \text{for all }k\in\mathbb N,
\]
thus separating finite-generation phenomena from indefinite persistence [2009.02758].

## 3. Maximal mediated sets

In lattice geometry and polynomial optimization, **M-sets** often mean **maximal mediated sets**. If \(\Delta\subset (2\mathbb Z)^n\) is a finite even set, a set \(L\subseteq \mathbb Z^n\) is **\(\Delta\)-mediated** if
\[
\Delta \subseteq L \subseteq \MidD(L) \cup \Delta,
\]
where
\[
\MidD(L)=\left\{\frac{\mathbf s+\mathbf t}{2}:\mathbf s,\mathbf t\in L\cap (2\mathbb Z)^n,\ \mathbf s\neq \mathbf t\right\}.
\]
The **maximal \(\Delta\)-mediated set** \(\Delta^*\) is the unique largest such set. Reznick’s existence theorem yields
\[
\Delta \cup \MidD(\Delta)\subseteq \Delta^* \subseteq \conv(\Delta)\cap \mathbb Z^n.
\]
This notion is central for the sums-of-squares problem of AGI-forms and nonnegative circuit polynomials: for a nonnegative simplicial AGI-form, or more generally a nonnegative circuit polynomial, the polynomial is SOS iff the distinguished inner exponent lies in the maximal mediated set of the Newton simplex. The paper further extends the criterion to simplex-supported SONC polynomials with several interior exponents \(Y\), showing
\[
f \text{ is SOS } \iff \text{every } \mathbf\beta\in Y \text{ satisfies } \mathbf\beta\in \Delta^*.
\]
For simplices, two extreme cases are isolated: the **\(M\)-simplex**, where \(\Delta^*=\Delta\cup\MidD(\Delta)\), and the **\(H\)-simplex**, where \(\Delta^*=\conv(\Delta)\cap\mathbb Z^n\). The same paper proves that MMS-preserving maps are exactly affine unimodular maps with even translation part, and that simplices containing the origin have isomorphic MMS precisely when the generated lattices have the same Hermite normal form up to column permutations. It also introduces the \(h\)-ratio \(H(\Delta)\) as a density statistic for \(\Delta^*\) inside the lattice points of \(\conv(\Delta)\), and reports extensive computations in dimension \(2\): for maximal degree \(150\), \(4,266,834\) simplices arising from \(886,297\) distinct lattices were tested, every one was either an \(M\)-simplex or an \(H\)-simplex, and the counts were \(4,250,533\) \(H\)-simplices versus \(16,301\) \(M\)-simplices [1910.00502].

## 4. \(M\)-sets as monoid actions and topoi

For a monoid \(M\), the category of right \(M\)-sets,
\[
\mathbf S^M,
\]
and the category of left \(M\)-sets,
\[
{}_M\mathbf S,
\]
are Grothendieck topoi. In this context an \(M\)-set is simply a set equipped with a monoid action, but the cited work studies the geometry of the resulting topos through its points and localisations. By Diaconescu’s theorem, points of \(\mathbf S^M\) correspond to **filtered left \(M\)-sets**. If \(A\) is such a filtered left \(M\)-set, the associated point is the geometric morphism \(p_A\) with inverse-image functor
\[
p_A^*(X)=X\otimes_M A.
\]

For finite monoids, the theory becomes idempotent-controlled. The paper proves that every filtered left \(M\)-set is of the form \(Me\) for an idempotent \(e\in M\), and hence
\[
(I(M))^{\mathrm{op}} \simeq \mathrm{Pts}(M),
\]
where \(I(M)\) is the category whose objects are idempotents and whose morphisms are \(fMe\). Isomorphism classes of points are therefore classified by Green’s \(\mathcal I\)-classes of idempotents:
\[
F_M \cong \mathrm{Idem}_{\mathcal I}(M).
\]
The paper then equips \(F_M\) with the order topology induced by
\[
e\le f \iff MeM\subseteq MfM.
\]
This topology is designed to be more informative than the classical SGA4 topology, which may be trivial for \(\mathbf S^M\). A parallel classification holds for localising subcategories:
\[
\mathrm{Loc}(\mathbf S^M)=II(M),
\]
where \(II(M)\) denotes the two-sided idempotent ideals of \(M\). For an idempotent \(e\) and an idempotent ideal \(\mathfrak m\), the paper gives the criterion
\[
p_{Me}\,\pitchfork\, \mathbf{Sh}(\mathfrak F_{\mathfrak m}) \iff e\in \mathfrak m.
\]
The resulting picture identifies points, open sets, and localisations of the topos directly with idempotents, Green relations, and idempotent ideals of the underlying finite monoid [2011.11747].

## 5. \(M\)-convex sets and quotient theory

In discrete convex analysis, an **\(M\)-convex set** is a nonempty set \(P\subseteq \mathbb Z^E\) satisfying the discrete exchange axiom: for all \(x,y\in P\) and any \(i\in \operatorname{supp}^+(x-y)\), there exists \(j\in \operatorname{supp}^-(x-y)\) such that
\[
x-e_i+e_j\in P,\qquad y+e_i-e_j\in P.
\]
All points of an \(M\)-convex set have the same coordinate sum, and bounded \(M\)-convex sets are in correspondence with integer-valued submodular functions and integral generalized permutohedra. The broader \(M^\natural\)-convex setting is realized as generalized polymatroids intersected with the integer lattice.

The paper “Quotients of M-convex sets and M-convex functions” develops a unified quotient theory that generalizes matroid quotients, polymatroid quotients, and strong maps of submodular functions. If \(P,Q\subseteq\mathbb Z^E\) are \(M\)-convex with associated submodular functions \(p,q\), then \(Q\) is a quotient of \(P\), written \(P\succcurlyeq Q\), if
\[
q(Y)-q(X)\le p(Y)-p(X)\qquad \forall\,X\subseteq Y\subseteq E.
\]
The main theorem gives **ten equivalent characterizations** of this relation. Among them are vertex containment for all permutation vertices, realization as the top and bottom layers of a common \(M^\natural\)-convex set, a deletion–contraction description on an extended ground set, an asymmetric exchange property, induction through a linking set, comparison in Green’s right preorder on a monoid of linking sets, a matroid-quotient lift criterion, and a flag \(M\)-convex formulation via generalized permutohedra. The same work initiates a quotient theory for \(M\)-convex functions, proving the implication chain
\[
\text{top/bottom} \implies \text{induction} \implies \text{exchange} \implies \text{minimizers},
\]
with equivalence when the rank drop is \(1\). It also shows that quotients are preserved under induction through linking sets and encoded order-theoretically by Green-type preorders on the linking-set monoid [2403.07751].

## 6. Additional meanings in topology, analysis, and dynamics

In **multiset topology**, an M-set is literally a multiset represented by a count function
\[
C_M:X\to W,
\]
with \(C_M(x)\) the multiplicity of \(x\). The theory defines inclusion, union, intersection, addition, subtraction, support set, power M-sets, and M-topologies in terms of multiplicity. On this basis the paper introduces **semi open M-sets (SOM)** and **semi closed M-sets (SCM)** via multiplicity inequalities relative to closure and interior, then develops **semi compactness**, **semi whole compactness**, **semi partial whole compactness**, and **semi full compactness**. Finite-intersection-property characterizations are established in the semi-open/semi-closed setting, and semi compactness is shown to behave well under subspaces [1403.5642].

In the **farthest point problem**, an \(M\)-compact set is a subset \(A\) of a normed space such that every maximizing sequence in \(A\) has a convergent subsequence. The paper studies uniquely remotal \(M\)-compact sets and proves that if \(A\) is uniquely remotal and \(M\)-compact with nonempty derived set, then the derived set is again uniquely remotal and \(M\)-compact. It also proves that every uniquely remotal set is singleton in a finite-dimensional strictly convex normed linear space, and that if a uniquely remotal \(M\)-compact set has compact derived set, then it is singleton [1605.04100].

In **complex dynamics**, the “M-set” attached to alternated Julia sets is a generalized Mandelbrot-type parameter object for the switching system
\[
z_{n+1}=\begin{cases}
z_n^2+c_1,& n \text{ even},\\
z_n^2+c_2,& n \text{ odd}.
\end{cases}
\]
Because \(c_1,c_2\in\mathbb C\), the parameter space is four-dimensional. The paper studies the connectedness locus \(CL\), disconnectedness locus \(DL\), and totally disconnectedness locus \(TDL\), using the auxiliary quartic
\[
Q(w)=\bigl(w^2+c_1\bigr)^2+c_2
\]
and critical-orbit boundedness criteria. The resulting M-set is therefore a four-dimensional fractal parameter space rather than a planar Mandelbrot set [1810.06982].

In **Walsh analysis**, an \(M\)-set is a set outside which a nontrivial Walsh series converges to \(0\). The cited paper constructs families of such sets in the \(d\)-dimensional Walsh system for rectangular, cubic, and iterated convergence. The sets are obtained by a recursive dyadic construction \(F=\bigcap_{s=1}^\infty F_s\), together with a quasimeasure \(\tau\) whose Fourier–Walsh series is a null-series on \(G^d\setminus F\). A notable stability property is that every nonempty open portion of the constructed \(F^{\boldsymbol\pi}\) is again an \(M\)-set. The same paper studies the coefficient scale of competing null-series and shows that if \(\psi_n=o(\tau_n)\), then \(\psi_n\equiv 0\), and it explains how symmetrizing the construction yields corresponding **\(U\)-sets** [2508.00182].

Taken together, these usages show that **M-set** is a context-sensitive mathematical label rather than a single transdisciplinary concept. Its meanings range from additive-combinatorial extremality and midpoint closure to monoid-action topoi, discrete exchange geometry, multiplicity-based topology, dyadic uniqueness theory, and generalized Mandelbrot parameter spaces.

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