arXiv++ Combinatorics

Given an $m\times n$ binary matrix $M$ with $|M|=p\cdot mn$ (where $|M|$ denotes the number of 1 entries), define the discrepancy of $M$ as $\mbox{disc}(M)=\displaystyle\max_{X\subset [m], Y\subset [n]}\big||M[X\times Y]|-p|X|\cdot |Y|\big|$. Using semidefinite programming and spectral techniques, we prove that if $\mbox{rank}(M)\leq r$ and $p\leq 1/2$, then $$\mbox{disc}(M)\geq Ω(mn)\cdot \min\left\{p,\frac{p^{1/2}}{\sqrt{r}}\right\}.$$ We use this result to obtain a modest improvement of Lovett's best known upper bound on the log-rank conjecture. We prove that any $m\times n$ binary matrix $M$ of rank at most $r$ contains an $(m\cdot 2^{-O(\sqrt{r})})\times (n\cdot 2^{-O(\sqrt{r})})$ sized all-1 or all-0 submatrix, which implies that the deterministic communication complexity of any Boolean function of rank $r$ is at most $O(\sqrt{r})$.

The goal of this expository note is to give a short, self-contained proof of nearly optimal lower bounds for the second largest eigenvalue of the adjacency matrix of regular graphs.

The positive discrepancy of a graph $G$ of edge density $p=e(G)/\binom{v(G)}{2}$ is defined as $$\mbox{disc}^{+}(G)=\max_{U\subset V(G)}e(G[U])-p\binom{|U|}{2}.$$ In 1993, Alon proved (using the equivalent terminology of minimum bisections) that if $G$ is $d$-regular on $n$ vertices, and $d=O(n^{1/9})$, then $\mbox{disc}^{+}(G)=Ω(d^{1/2}n)$. We greatly extend this by showing that if $G$ has average degree $d$, then $\mbox{disc}^{+}(G)=Ω(d^{\frac{1}{2}}n)$ if $d\in [0,n^{\frac{2}{3}}]$, $Ω(n^2/d)$ if $d\in [n^{\frac{2}{3}},n^{\frac{4}{5}}]$, and $Ω(d^{\frac{1}{4}}n/\log n)$ if $d\in \left[n^{\frac{4}{5}},(\frac{1}{2}-\varepsilon)n\right]$. These bounds are best possible if $d\ll n^{3/4}$, and the complete bipartite graph shows that $\mbox{disc}^{+}(G)=Ω(n)$ cannot be improved if $d\approx n/2$. Our proofs are based on semidefinite programming and linear algebraic techniques. An interesting corollary of our results is that every $d$-regular graph on $n$ vertices with ${\frac{1}{2}+\varepsilon\leq \frac{d}{n}\leq 1-\varepsilon}$ has a cut of size $\frac{nd}{4}+Ω(n^{5/4}/\log n)$. This is not necessarily true without the assumption of regularity, or the bounds on $d$. The positive discrepancy of regular graphs is controlled by the second eigenvalue $λ_2$, as $\mbox{disc}^{+}(G)\leq \frac{λ_2}{2} n+d$. As a byproduct of our arguments, we present lower bounds on $λ_2$ for regular graphs, extending the celebrated Alon-Boppana theorem in the dense regime.

Consider the partially ordered set on $[t]^n:=\{0,\dots,t-1\}^n$ equipped with the natural coordinate-wise ordering. Let $A(t,n)$ denote the number of antichains of this poset. The quantity $A(t,n)$ has a number of combinatorial interpretations: it is precisely the number of $(n-1)$-dimensional partitions with entries from $\{0,\dots,t\}$, and by a result of Moshkovitz and Shapira, $A(t,n)+1$ is equal to the $n$-color Ramsey number of monotone paths of length $t$ in 3-uniform hypergraphs. This has led to significant interest in the growth rate of $A(t,n)$. A number of results in the literature show that $\log_2 A(t,n)=(1+o(1))\cdot α(t,n)$, where $α(t,n)$ is the width of $[t]^n$, and the $o(1)$ term goes to $0$ for $t$ fixed and $n$ tending to infinity. In the present paper, we prove the first bound that is close to optimal in the case where $t$ is arbitrarily large compared to $n$, as well as improve all previous results for sufficiently large $n$. In particular, we prove that there is an absolute constant $c$ such that for every $t,n\geq 2$, $$\log_2 A(t,n)\leq \left(1+c\cdot \frac{(\log n)^3}{n}\right)\cdot α(t,n).$$ This resolves a conjecture of Moshkovitz and Shapira. A key ingredient in our proof is the construction of a normalized matching flow on the cover graph of the poset $[t]^n$ in which the distribution of weights is close to uniform, a result that may be of independent interest.

In 1993, Fishburn and Graham established the following qualitative extension of the classical Erdős-Szekeres theorem. If $N$ is sufficiently large with respect to $n$, then any $N\times N$ real matrix contains an $n\times n$ submatrix in which every row and every column is monotone. We prove that the smallest such $N$ is at most $2^{n^{4+o(1)}}$, greatly improving the previously best known double-exponential upper bound, and getting close to the best known lower bound $n^{n/2}$. In particular, we prove the following surprising sharp transition in the asymmetric setting. On one hand, every $8n^2\times 2^{n^{4+o(1)}}$ matrix contains an $n\times n$ submatrix, in which every row is mononote. On the other hand, there exist $n^{2}/6\times 2^{2^{n^{1-o(1)}}}$ matrices containing no such submatrix .

The goal of this paper is to show the existence (using probabilistic tools) of configurations of lines, boxes, and points with certain interesting combinatorial properties. (i) First, we construct a family of $n$ lines in $\mathbb{R}^3$ whose intersection graph is triangle-free of chromatic number $Ω(n^{1/15})$. This improves the previously best known bound $Ω(\log\log n)$ by Norin, and is also the first construction of a triangle-free intersection graph of simple geometric objects with polynomial chromatic number. (ii) Second, we construct a set of $n$ points in $\mathbb{R}^d$, whose Delaunay graph with respect to axis-parallel boxes has independence number at most $n\cdot (\log n)^{-(d-1)/2+o(1)}$. This extends the planar case considered by Chen, Pach, Szegedy, and Tardos.

Given a family $\mathcal{B}$ of axis-parallel boxes in $\mathbb{R}^d$, let $τ$ denote its piercing number, and $ν$ its independence number. It is an old question whether $τ/ν$ can be arbitrarily large for given $d\geq 2$. Here, for every $ν$, we construct a family of axis-parallel boxes achieving $$τ\geq Ω_d(ν)\cdot\left(\frac{\log ν}{\log\log ν}\right)^{d-2}.$$ This not only answers the previous question for every $d\geq 3$ positively, but also matches the best known upper bound up to double-logarithmic factors. Our main construction has further implications about the Ramsey and coloring properties of configurations of boxes as well. We show the existence of a family of $n$ boxes in $\mathbb{R}^{d}$, whose intersection graph has clique and independence number $O_d(n^{1/2})\cdot \left(\frac{\log n}{\log\log n}\right)^{-(d-2)/2}.$ This is the first improvement over the trivial upper bound $O_d(n^{1/2})$, and matches the best known lower bound up to double-logarithmic factors. Finally, for every $ω$ satisfying $\frac{\log n}{\log\log n}\ll ω\ll n^{1-\varepsilon}$, we construct an intersection graph of $n$ boxes with clique number at most $ω$, and chromatic number $Ω_{d,\varepsilon}(ω)\cdot \left(\frac{\log n}{\log\log n}\right)^{d-2}.$ This matches the best known upper bound up to a factor of $O_d((\log w)(\log \log n)^{d-2})$.

We prove that the maximum number of edges in a 3-uniform linear hypergraph on $n$ vertices containing no 2-regular subhypergraph is $n^{1+o(1)}$. This resolves a conjecture of Dellamonica, Haxell, Luczak, Mubayi, Nagle, Person, Rödl, Schacht and Verstraëte. We use this result to show that the maximum number of edges in a $3$-uniform hypergraph on $n$ vertices containing no immersion of a closed surface is $n^{2+o(1)}$. Furthermore, we present results on the maximum number of edges in $k$-uniform linear hypergraphs containing no $r$-regular subhypergraph.

An $r$-uniform hypergraph $H$ is semi-algebraic of complexity $\mathbf{t}=(d,D,m)$ if the vertices of $H$ correspond to points in $\mathbb{R}^{d}$, and the edges of $H$ are determined by the sign-pattern of $m$ degree-$D$ polynomials. Semi-algebraic hypergraphs of bounded complexity provide a general framework for studying geometrically defined hypergraphs. The much-studied semi-algebraic Ramsey number $R_{r}^{\mathbf{t}}(s,n)$ denotes the smallest $N$ such that every $r$-uniform semi-algebraic hypergraph of complexity $\mathbf{t}$ on $N$ vertices contains either a clique of size $s$, or an independent set of size $n$. Conlon, Fox, Pach, Sudakov, and Suk proved that $R_{r}^{\mathbf{t}}(n,n)<\mbox{tw}_{r-1}(n^{O(1)})$, where $\mbox{tw}_{k}(x)$ is a tower of 2's of height $k$ with an $x$ on the top. This bound is also the best possible if $\min\{d,D,m\}$ is sufficiently large with respect to $r$. They conjectured that in the asymmetric case, we have $R_{3}^{\mathbf{t}}(s,n)<n^{O(1)}$ for fixed $s$. We refute this conjecture by showing that $R_{3}^{\mathbf{t}}(4,n)>n^{(\log n)^{1/3-o(1)}}$ for some complexity $\mathbf{t}$. In addition, motivated by results of Bukh and Matoušek and Basit, Chernikov, Starchenko, Tao and Tran, we study the complexity of the Ramsey problem when the defining polynomials are linear, that is, when $D=1$. In particular, we prove that $R_{r}^{d,1,m}(n,n)\leq 2^{O(n^{4r^2m^2})}$, while from below, we establish $R^{1,1,1}_{r}(n,n)\geq 2^{Ω(n^{\lfloor r/2\rfloor-1})}$.

Given positive integers $k\leq d$ and a finite field $\mathbb{F}$, a set $S\subset\mathbb{F}^{d}$ is $(k,c)$-subspace evasive if every $k$-dimensional affine subspace contains at most $c$ elements of $S$. By a simple averaging argument, the maximum size of a $(k,c)$-subspace evasive set is at most $c |\mathbb{F}|^{d-k}$. When $k$ and $d$ are fixed, and $c$ is sufficiently large, the matching lower bound $Ω(|\mathbb{F}|^{d-k})$ is proved by Dvir and Lovett. We provide an alternative proof of this result using the random algebraic method. We also prove sharp upper bounds on the size of $(k,c)$-evasive sets in case $d$ is large, extending results of Ben-Aroya and Shinkar. The existence of optimal evasive sets has several interesting consequences in combinatorial geometry. We show that the minimum number of $k$-dimensional linear hyperplanes needed to cover the grid $[n]^{d}\subset \mathbb{R}^{d}$ is $Ω_{d}\big(n^{\frac{d(d-k)}{d-1}}\big)$, which matches the upper bound proved by Balko, Cibulka, and Valtr, and settles a problem proposed by Brass, Moser, and Pach. Furthermore, we improve the best known lower bound on the maximum number of incidences between points and hyperplanes in $\mathbb{R}^{d}$ assuming their incidence graph avoids the complete bipartite graph $K_{c,c}$ for some large constant $c=c(d)$.

In this paper we study the fundamental problem of finding small dense subgraphs in a given graph. For a real number $s>2$, we prove that every graph on $n$ vertices with average degree at least $d$ contains a subgraph of average degree at least $s$ on at most $nd^{-\frac{s}{s-2}}(\log d)^{O_s(1)}$ vertices. This is optimal up to the polylogarithmic factor, and resolves a conjecture of Feige and Wagner. In addition, we show that every graph with $n$ vertices and average degree at least $n^{1-\frac{2}{s}+\varepsilon}$ contains a subgraph of average degree at least $s$ on $O_{\varepsilon,s}(1)$ vertices, which is also optimal up to the constant hidden in the $O(.)$ notation, and resolves a conjecture of Verstraëte.

We prove that if $G$ is an abelian group and $H_1x_1,\dots,H_{k}x_k$ is an irredundant (minimal) cover of $G$ with cosets, then $$|G:\bigcap_{i=1}^{k}H_{i}|=2^{O(k)}.$$ This bound is the best possible up to the constant hidden in the $O(\cdot)$ notation, and it resolves conjectures of Pyber (1996) and Szegedy (2007). We further show that if $G$ is an elementary $p$-group for some large prime $p$, and $H_1,\dots,H_k$ is a sequence of hyperplanes with many repetitions, then the bound above can be improved. As a consequence, we establish a substantial strengthening of the recently solved Alon-Jaeger-Tarsi conjecture: there exists $α>0$ such that for every invertible matrix $M\in\mathbb{F}_p^{n\times n}$ and any set of at most $p^α$ forbidden coordinates, one can find a vector $x\in\mathbb{F}_p^{n}$ such that neither $x$ nor $Mx$ have a forbidden coordinate.

For a positive integer $t$, let $F_t$ denote the graph of the $t\times t$ grid. Motivated by a 50-year-old conjecture of Erdős about Turán numbers of $r$-degenerate graphs, we prove that there exists a constant $C=C(t)$ such that $\mathrm{ex}(n,F_t)\leq Cn^{3/2}$. This bound is tight up to the value of $C$. One of the interesting ingredients of our proof is a novel way of using the tensor power trick.

We present several results in extremal graph and hypergraph theory of topological nature. First, we show that if $α>0$ and $\ell=Ω(\frac{1}α\log\frac{1}α)$ is an odd integer, then every graph $G$ with $n$ vertices and at least $n^{1+α}$ edges contains an $\ell$-subdivision of the complete graph $K_t$, where $t=n^{Θ(α)}$. Also, this remains true if in addition the edges of $G$ are properly colored, and one wants to find a rainbow copy of such a subdivision. In the sparser regime, we show that properly edge colored graphs on $n$ vertices with average degree $(\log n)^{2+o(1)}$ contain rainbow cycles, while average degree $(\log n)^{6+o(1)}$ guarantees rainbow subdivisions of $K_t$ for any fixed $t$, thus improving recent results of Janzer and Jiang et al., respectively. Furthermore, we consider certain topological notions of cycles in pure simplicial complexes (uniform hypergraphs). We show that if $G$ is a $2$-dimensional pure simplicial complex ($3$-graph) with $n$ $1$-dimensional and at least $n^{1+α}$ 2-dimensional faces, then $G$ contains a triangulation of the cylinder and the Möbius strip with $O(\frac{1}α\log\frac{1}α)$ vertices. We present generalizations of this for higher dimensional pure simplicial complexes as well. In order to prove these results, we consider certain (properly edge colored) graphs and hypergraphs $G$ with strong expansion. We argue that if one randomly samples the vertices (and colors) of $G$ with not too small probability, then many pairs of vertices are connected by a short path whose vertices (and colors) are from the sampled set, with high probability.

Recently, the first two authors proved the Alon-Jaeger-Tarsi conjecture on non-vanishing linear maps, for large primes. We extend their ideas to address several other related conjectures. We prove the weak Additive Basis conjecture proposed by Szegedy, making a significant step towards the Additive Basis conjecture of Jaeger, Linial, Payan, and Tarsi. In fact, we prove it in a strong form: there exists a set $A\subset\mathbb{F}_p^*$ of size $O(\log p)$ such that if $B\subset\mathbb{F}_p^{n}$ is the union of $p$ linear bases, then $A\cdot B=\{a\cdot v:a\in A, v\in B\}$ is an additive basis. An old result of Tomkinson states that if $G$ is a group, and $\{H_{i}x_{i}:i\in [k]\}$ is an irredundant coset cover of $G$, then $|G:\bigcap_{i\in [k]} H_{i}|\leq k!,$ and this bound is the best possible. It is a longstanding open problem whether the upper bound can be improved to $e^{O(k)}$ in case we restrict cosets to subgroups. Pyber proposed to study this question for abelian groups. We show that somewhat surprisingly, if $G$ is abelian, the upper bound can be improved to $e^{O(k\log \log k)}$ already in the case of general coset covers, making the first substantial improvement over the $k!$ bound. Finally, we prove a natural generalization of the Alon-Jaeger-Tarsi conjecture for multiple matrices.

A recent result of Alon, Ben-Eliezer and Fischer establishes an induced removal lemma for ordered graphs. That is, if $F$ is an ordered graph and $\varepsilon>0$, then there exists $δ_{F}(\varepsilon)>0$ such that every $n$-vertex ordered graph $G$ containing at most $δ_{F}(\varepsilon) n^{v(F)}$ induced copies of $F$ can be made induced $F$-free by adding/deleting at most $\varepsilon n^2$ edges. We prove that $δ_{F}(\varepsilon)$ can be chosen to be a polynomial function of $\varepsilon$ if and only if $|V(F)|=2$, or $F$ is the ordered graph with vertices $x<y<z$ and edges $\{x,y\},\{x,z\}$ (up to complementation and reversing the vertex order). We also discuss similar problems in the non-induced case.

Let $D_2$ denote the $3$-uniform hypergraph with $4$ vertices and $2$ edges. Answering a question of Alon and Shapira, we prove an induced removal lemma for $D_2$ having polynomial bounds. We also prove an Erdős-Hajnal-type result: every induced $D_2$-free hypergraph on $n$ vertices contains a clique or an independent set of size $n^{c}$ for some absolute constant $c > 0$. In the case of both problems, $D_2$ is the only nontrivial $k$-uniform hypergraph with $k\geq 3$ which admits a polynomial bound.

Let $\mathcal{F}\subset 2^{[n]}$ be a set family such that the intersection of any two members of $\mathcal{F}$ has size divisible by $\ell$. The famous Eventown theorem states that if $\ell=2$ then $|\mathcal{F}|\leq 2^{\lfloor n/2\rfloor}$, and this bound can be achieved by, e.g., an `atomic' construction, i.e. splitting the ground set into disjoint pairs and taking their arbitrary unions. Similarly, splitting the ground set into disjoint sets of size $\ell$ gives a family with pairwise intersections divisible by $\ell$ and size $2^{\lfloor n/\ell\rfloor}$. Yet, as was shown by Frankl and Odlyzko, these families are far from maximal. For infinitely many $\ell$, they constructed families $\mathcal{F}$ as above of size $2^{Ω(n\log \ell/\ell)}$. On the other hand, if the intersection of any number of sets in $\mathcal{F}\subset 2^{[n]}$ has size divisible by $\ell$, then it is easy to show that $|\mathcal{F}|\leq 2^{\lfloor n/\ell\rfloor}$. In 1983 Frankl and Odlyzko conjectured that $|\mathcal{F}|\leq 2^{(1+o(1)) n/\ell}$ holds already if one only requires that for some $k=k(\ell)$ any $k$ distinct members of $\mathcal{F}$ have an intersection of size divisible by $\ell$. We completely resolve this old conjecture in a strong form, showing that $|\mathcal{F}|\leq 2^{\lfloor n/\ell\rfloor}+O(1)$ if $k$ is chosen appropriately, and the $O(1)$ error term is not needed if (and only if) $\ell \, | \, n$, and $n$ is sufficiently large. Moreover the only extremal configurations have `atomic' structure as above. Our main tool, which might be of independent interest, is a structure theorem for set systems with small 'doubling'.

One of the central questions in Ramsey theory asks how small can be the size of the largest clique and independent set in a graph on $N$ vertices. By the celebrated result of Erdős from 1947, the random graph on $N$ vertices with edge probability $1/2$, contains no clique or independent set larger than $2\log_2 N$, with high probability. Finding explicit constructions of graphs with similar Ramsey-type properties is a famous open problem. A natural approach is to construct such graphs using algebraic tools. Say that an $r$-uniform hypergraph $\mathcal{H}$ is \emph{algebraic of complexity $(n,d,m)$} if the vertices of $\mathcal{H}$ are elements of $\mathbb{F}^{n}$ for some field $\mathbb{F}$, and there exist $m$ polynomials $f_1,\dots,f_m:(\mathbb{F}^{n})^{r}\rightarrow \mathbb{F}$ of degree at most $d$ such that the edges of $\mathcal{H}$ are determined by the zero-patterns of $f_1,\dots,f_m$. The aim of this paper is to show that if an algebraic graph (or hypergraph) of complexity $(n,d,m)$ has good Ramsey properties, then at least one of the parameters $n,d,m$ must be large. In 2001, Rónyai, Babai and Ganapathy considered the bipartite variant of the Ramsey problem and proved that if $G$ is an algebraic graph of complexity $(n,d,m)$ on $N$ vertices, then either $G$ or its complement contains a complete balanced bipartite graph of size $Ω_{n,d,m}(N^{1/(n+1)})$. We extend this result by showing that such $G$ contains either a clique or an independent set of size $N^{Ω(1/ndm)}$ and prove similar results for algebraic hypergraphs of constant complexity. We also obtain a polynomial regularity lemma for $r$-uniform algebraic hypergraphs that are defined by a single polynomial, that might be of independent interest. Our proofs combine algebraic, geometric and combinatorial tools.

Given a $d$-dimensional tensor $T:A_1\times\dots\times A_d\rightarrow \mathbb{F}$ (where $\mathbb{F}$ is a field), the $i$-flattening rank of $T$ is the rank of the matrix whose rows are indexed by $A_{i}$, columns are indexed by $B_{i}=A_1\times\dots\times A_{i-1}\times A_{i+1}\times\dots\times A_{d}$ and whose entries are given by the corresponding values of $T$. The max-flattening rank of $T$ is defined as $\text{mfrank}(T)=\max_{i\in [d]}\text{frank}_{i}(T)$. A tensor $T:A^{d}\rightarrow\mathbb{F}$ is called semi-diagonal, if $T(a,\dots,a)\neq 0$ for every $a\in A$, and $T(a_{1},\dots,a_{d})=0$ for every $a_{1},\dots,a_{d}\in A$ that are all distinct. In this paper we prove that if $T:A^{d}\rightarrow\mathbb{F}$ is semi-diagonal, then $\text{mfrank}(T)\geq \frac{|A|}{d-1}$, and this bound is the best possible. We give several applications of this result, including a generalization of the celebrated Frankl-Wilson theorem on forbidden intersections. Also, addressing a conjecture of Aharoni and Berger, we show that if the edges of an $r$-uniform multi-hypergraph $\mathcal{H}$ are colored with $z$ colors such that each colorclass is a matching of size $t$, then $\mathcal{H}$ contains a rainbow matching of size $t$ provided $z>(t-1)\binom{rt}{r}$. This improves previous results of Alon and Glebov, Sudakov and Szabó.