arXiv++ Combinatorics

We survey the history of the capset problem in the context of related results on progression-free sets, discuss recent progress, and mention further directions to explore.

We show that for a finite, nonempty subset $A$ of a group, the quotient set $A^{-1}A:=\{a_1^{-1}a_2\colon a_1,a_2\in A\}$ has size $|A^{-1}A|\ge\frac53\,|A|$, unless $A$ is densely contained in a coset, or in a union of two cosets of a finite subgroup.

We show that a finite zero-sum-free sequence $α$ over an abelian group has at least $c|α|^{4/3}$ distinct subsequence sums, unless $α$ is "controlled" by a small number of its terms; here $|α|$ denotes the number of terms of $α$, and $c>0$ is an absolute constant.

Suppose that $A$ is a finite, nonempty subset of a cyclic group of either infinite or prime order. We show that if the difference set $A-A$ is ``not too large'', then there is a nonzero group element with at least as many as $(2+o(1))|A|^2/|A-A|$ representations as a difference of two elements of $A$; that is, the second largest number of representations is, essentially, twice the average. Here the coefficient $2$ is the best possible. We also prove continuous and multidimensional versions of this result, and obtain similar results for sufficiently dense subsets of an arbitrary abelian group.

We show that a zero-sum-free sequence of length $n$ over an abelian group spans at least $2n$ distinct subsequence sums, unless it possesses a rigid, easily-described structure.

Let $A$ be a finite, nonempty subset of an abelian group. We show that if every element of $A$ is a sum of two other elements, then $A$ has a nonempty zero-sum subset. That is, a (finite, nonempty) sum-full subset of an abelian group is not zero-sum-free.

We give a comprehensive description of the sets $A$ in finite cyclic groups such that $|2A|<\frac94|A|$; namely, we show that any set with this property is densely contained in a (one-dimensional) coset progression. This improves earlier results of Deshouillers-Freiman and Balasubramanian-Pandey.

Let $p\ge 3$ be a prime, $S\subseteq\mathbb F_p^2$ a nonempty set, and $w\colon\mathbb F_p^2\to\mathbb R$ a function with $\mathrm{supp}\, w=S$. Applying an uncertainty inequality due to András Biró and the present author, we show that there are at most $\frac12|S|$ directions in $\mathbb F_p^2$ such that for every line $l$ in any of these directions, one has $$ \sum_{z\in l} w(z) = \frac1p\sum_{z\in\mathbb F_p^2} w(z), $$ except if $S$ itself is a line and $w$ is constant on $S$ (in which case all, but one direction have the property in question). The bound $\frac12|S|$ is sharp. As an application, we give a new proof of a result of Rédei-Megyesi about the number of directions determined by a set in a finite affine plane.

We establish a number of uncertainty inequalities for the additive group of a finite affine plane, showing that for $p$ prime, a nonzero function $f\colon\mathbb F_p^2\to\mathbb C$ and its Fourier transform $\hat f\colon\widehat{\mathbb F_p^2}\to\mathbb C$ cannot have small supports simultaneously. The "baseline" of our investigation is the well-known Meshulam's bound, which we sharpen, for the particular groups under consideration, taking into account not only the sizes of the support sets $\mathrm{supp}\,f$ and $\mathrm{supp}\,\hat f$, but also their structure. Our results imply in particular that, with some explicitly classified exceptions, one has $|\mathrm{supp}\,f||\mathrm{supp}\,\hat f|\ge3p(p-2)$; in comparison, the classical uncertainty inequality gives $|\mathrm{supp}\,f||\mathrm{supp}\,\hat f|\ge p^2$.

We show that a non-empty subset of an abelian group with a small edge boundary must be large; in particular, if $A$ and $S$ are finite, non-empty subsets of an abelian group such that $S$ is independent, and the edge boundary of $A$ with respect to $S$ does not exceed $(1-γ)|S||A|$ with a real $γ\in(0,1]$, then $|A| \ge 4^{(1-1/d)γ|S|}$, where $d$ is the smallest order of an element of $S$. Here the constant $4$ is best possible. As a corollary, we derive an upper bound for the size of the largest independent subset of the set of popular differences of a finite subset of an abelian group. For groups of exponent $2$ and $3$, our bound translates into a sharp estimate for the additive dimension of the popular difference set. We also prove, as an auxiliary result, the following estimate of possible independent interest: if $A \subset \mathbb Z^n$ is a finite, non-empty downset then, denoting by $w(a)$ the number of non-zero components of the vector $a\in A$, we have \[\frac1{|A|} \sum_{a\in A} w(a) \le \frac12\, \log_2 |A|.\]

Consider the projections of a finite set $A\subset R^n$ onto the coordinate hyperplanes. How small can the sum of the sizes of these projections be, given the size of $A$? In a different form, this problem has been studied earlier in the context of edge-isoperimetric inequalities on graphs, and it is can be derived from the known results that there is a linear order on the set of $n$-tuples with non-negative integer coordinates, such that the sum in question is minimised for the initial segments with respect to this order. We present a new, self-contained and constructive proof, enabling us to obtain a stability result and establish algebraic properties of the smallest possible projection sum. We also solve the problem of minimising the sum of the sizes of the one-dimensional projections.

As an easy corollary of Kneser's Theorem, if $A$ is a subset of the elementary abelian group ${\mathbb Z}_5^n$ of density $5^{-n}|A|>0.4$, then $3A={\mathbb Z}_5^n$. We establish the complementary stability result: if $5^{-n}|A|>0.3$ and $3A\ne{\mathbb Z}_5^n$, then $A$ is contained in a union of two cosets of an index-$5$ subgroup of ${\mathbb Z}_5^n$. Here the density bound $0.3$ is sharp. Our argument combines combinatorial reasoning with a somewhat non-standard application of the character sum technique.

We define the discrete norm of a complex $m\times n$ matrix $A$ by $$ \|A\|_Δ:= \max_{0\neξ\in\{0,1\}^n} \frac{\|Aξ\|}{\|ξ\|}, $$ and show that $$ \frac c{\sqrt{\log h(A)+1}}\,\|A\| \le \|A\|_Δ\le \|A\|, $$ where $c>0$ is an explicitly indicated absolute constant, $h(A)=\sqrt{\|A\|_1\|A\|_\infty}/\|A\|$, and $\|A\|_1,\|A\|_\infty$, and $\|A\|=\|A\|_2$ are the induced operator norms of $A$. Similarly, for the \emph{discrete Rayleigh norm} $$ \|A\|_P := \max_{\substack{0\neξ\in\{0,1\}^m \\ 0\neη\in\{0,1\}^n}} \frac{|ξ^tAη|}{\|ξ\|\|η\|} $$ we prove the estimate $$ \frac c{\log h(A)+1}\,\|A\| \le \|A\|_P \le \|A\|. $$ These estimates are shown to be essentially best possible. As a consequence, we obtain another proof of the (slightly sharpened and generalized version of the) converse to the expander mixing lemma by Bollobas-Nikiforov and Bilu-Linial.

We study extremal properties of the function $$ F(x) := \min\{k\|x\|^{1-1/k}\colon k\ge 1\},\ x\in[0,1], $$ where $\|x\|=\min\{x,1-x\}$. In particular, we show that $F$ is the pointwise largest function of the class of all real-valued functions $f$ defined on the interval $[0,1]$, and satisfying the relaxed convexity condition $$ f(tx_1+(1-t)x_2) \le tf(x_1)+(1-t)f(x_2)+|x_2-x_1|, \ x_1,x_2,t\in[0,1] $$ and the boundary condition $\max\{f(0),f(1)\}\le 0$. As an application, we prove that if $A$ and $S$ are subsets of a finite abelian group $G$, such that $S$ is generating and all of its elements have order at most $m$, then the number of edges from $A$ to its complement $G\setminus A$ in the directed Cayley graph induced by $S$ on $G$ is $$ \partial_S(A) \ge \frac{1}{m} |G| F(|A|/|G|). $$

For non-negative integers $r\ge d$, how small can a subset $C\subset F_2^r$ be, given that for any $v\in F_2^r$ there is a $d$-flat passing through $v$ and contained in $C\cup\{v\}$? Equivalently, how large can a subset $B\subset F_2^r$ be, given that for any $v\in F_2^r$ there is a linear $d$-subspace not blocked non-trivially by the translate $B+v$? A number of lower and upper bounds are obtained.

Let $G$ be a finite abelian abelian group of exponent $m\ge 2$. For subsets $A,S\subset G$, denote by $\partial_S(A)$ the number of edges from $A$ to its complement $G\setminus A$ in the directed Cayley graph, induced by $S$ on $G$. We show that if $S$ generates $G$, and $A$ is non-empty, then $$ \partial_S(A) \ge \frac{e}m\,|A|\ln\frac{|G|}{|A|}. $$ Here the coefficient $e=2.718...$ is best possible and cannot be replaced with a number larger than $e$. For homocyclic groups $G$ of exponent $m$, we find an explicit closed-form expression for $\partial_S(A)$ in the case where $S$ is a "standard" generating subset of $G$, and $A$ is an initial segment of $G$ with respect to the lexicographic order, induced by $S$ on $G$. Namely, we show that in this situation $$ \partial_S(A) = |G|\,ω_m(|A|/|G|), $$ where $ω_2$ is the Takagi function, and $ω_m$ for $m\ge 3$ is an appropriate generalization thereof. This particular case is of special interest, since for $m\in\{2,3,4\}$ it is known to yield the smallest possible value of $\partial_S(A)$, over all sets $A\subset G$ of given size. We give this classical result a new proof, somewhat different from the standard one. We also give a new, short proof of the Boros-Pales inequality $$ω_2(\frac{x+y}2) \le \frac{ω_2(x) + ω_2(y)}2 + \frac12\,|y-x|, $$ establish an extremal characterization of the Takagi function as the (pointwise) maximal function, satisfying this inequality and the boundary condition $\max\{ω_2(0),ω_2(1)\}\le 0$, and obtain similar results for the 3-adic analog $ω_3$ of the Takagi function.

Let A and B be subsets of an elementary abelian 2-group G, none of which are contained in a coset of a proper subgroup. Extending onto potentially distinct summands a result of Hennecart and Plagne, we show that if |A+B|<|A|+|B|, then either A+B=G, or the complement of A+B in G is contained in a coset of a subgroup of index at least 8, whence |A+B| is at least 7/8 |G|. We indicate conditions for the containment to be strict, and establish a refinement in the case where the sizes of A and B differ significantly.

We show that for a graph $G$ with the vertex set $V$ and the largest eigenvalue $λ_{\max}(G)$, letting $$ M(G) := \max_{X,Y \subset V} \frac{e(X,Y)}{\sqrt{|X||Y|}} $$ (where $e(X,Y)$ denotes the number of edges between $X$ and $Y$), we have $$ M(G) \le λ_{\max}(G) \le \big(\frac14 \log|V| + 1 \big) \M(G). $$ Here the lower bound is attained if $G$ is regular or bi-regular, whereas the logarithmic factor in the upper bound, conjecturally, can be improved --- although we present an example showing that it cannot be replaced with a factor growing slower than $(\log |V|/\log\log|V|)^{1/8}$. Further refinements are established, particularly in the case where $G$ is bipartite.

We prove that the sizes of the maximal dissociated subsets of a given finite subset of an abelian group differ by a logarithmic factor at most. On the other hand, we show that the set $\{0,1\}^n\seq\Z^n$ possesses a dissociated subset of size $\Ome(n\log n)$; since the standard basis of $\Z^n$ is a maximal dissociated subset of $\{0,1\}^n$ of size $n$, the result just mentioned is essentially sharp.

We prove that there is an absolute constant $c>0$ with the following property: if $Z/pZ$ denotes the group of prime order $p$, and a subset $A\subset Z/pZ$ satisfies $1<|A|<p/2$, then for any positive integer $m<\min\{c|A|/\ln|A|,\sqrt{p/8}\}$ there are at most $2m$ non-zero elements $b\in Z/pZ$ with $|(A+b)\setminus A|\le m$. This (partially) extends onto prime-order groups the result, established earlier by S. Konyagin and the present author for the group of integers. We notice that if $A\subset Z/pZ$ is an arithmetic progression and $m<|A|<p/2$, then there are exactly $2m$ non-zero elements $b\in Z/pZ$ with $|(A+b)\setminus A|\le m$. Furthermore, the bound $c|A|/\ln|A|$ is best possible up to the value of the constant $c$. On the other hand, it is likely that the assumption $m<\sqrt{p/8}$ can be dropped or substantially relaxed.