arXiv subjects

Dhruv Mubayi

Publications and source records attributed to Dhruv Mubayi.

At least 127 records · Page 7Linked to original sources

arXiv2011

Two-part set systems

The two part Sperner theorem of Katona and Kleitman states that if $X$ is an $n$-element set with partition $X_1 \cup X_2$, and $\cF$ is a family of subsets of $X$ such that no two sets $A, B \in \cF$ satisfy $A \subset B$ (or $B \subset A$) and $A \cap X_i=B \cap X_i$ for some $i$, then $|\cF| \le {n \choose \lfloor n/2 \rfloor}$. We consider variations of this problem by replacing the Sperner property with the intersection property and considering families that satisfiy various combinations of these properties on one or both parts $X_1$, $X_2$. Along the way, we prove the following new result which may be of independent interest: let $\cF, \cG$ be families of subsets of an $n$-element set such that $\cF$ and $\cG$ are both intersecting and cross-Sperner, meaning that if $A \in \cF$ and $B \in \cG$, then $A \not\subset B$ and $B \not\subset A$. Then $|\cF| +|\cG| < 2^{n-1}$ and there are exponentially many examples showing that this bound is tight.

math.CO↗

arXiv2011

On independent sets in hypergraphs

The independence number of a hypergraph H is the size of a largest set of vertices containing no edge of H. In this paper, we prove new sharp bounds on the independence number of n-vertex (r+1)-uniform hypergraphs in which every r-element set is contained in at most d edges, where 0 < d < n/(log n)^{3r^2}. Our relatively short proof extends a method due to Shearer. We give an application to hypergraph Ramsey numbers involving independent neighborhoods.

math.CO↗

arXiv2011

New lower bounds for the independence number of sparse graphs and hypergraphs

We obtain new lower bounds for the independence number of $K_r$-free graphs and linear $k$-uniform hypergraphs in terms of the degree sequence. This answers some old questions raised by Caro and Tuza \cite{CT91}. Our proof technique is an extension of a method of Caro and Wei \cite{CA79, WE79}, and we also give a new short proof of the main result of \cite{CT91} using this approach. As byproducts, we also obtain some non-trivial identities involving binomial coefficients.

math.CO↗

arXiv2011

The Turán number of $F_{3,3}$

Let $F_{3,3}$ be the 3-graph on 6 vertices, labelled abcxyz, and 10 edges, one of which is abc, and the other 9 of which are all triples that contain 1 vertex from abc and 2 vertices from xyz. We show that for all $n \ge 6$, the maximum number of edges in an $F_{3,3}$-free 3-graph on $n$ vertices is $\binom{n}{3} - \binom{\lfloor n/2 \rfloor}{3} - \binom{\lceil n/2 \rceil}{3}$. This sharpens results of Zhou and of the second author and Rödl.

math.CO↗

arXiv2010

The de Bruijn-Erdos Theorem for Hypergraphs

Fix integers $n \ge r \ge 2$. A clique partition of ${[n] \choose r}$ is a collection of proper subsets $A_1, A_2, \ldots, A_t \subset [n]$ such that $\bigcup_i{A_i \choose r}$ is a partition of ${[n] \choose r}$. Let $\cp(n,r)$ denote the minimum size of a clique partition of ${[n] \choose r}$. A classical theorem of de Bruijn and Erd\H os states that $\cp(n, 2) = n$. In this paper we study $\cp(n,r)$, and show in general that for each fixed $r \geq 3$, \[ \cp(n,r) \geq (1 + o(1))n^{r/2} \quad \quad \mbox{as}n \rightarrow \infty.\] We conjecture $\cp(n,r) = (1 + o(1))n^{r/2}$. This conjecture has already been verified (in a very strong sense) for $r = 3$ by Hartman-Mullin-Stinson. We give further evidence of this conjecture by constructing, for each $r \ge 4$, a family of $(1+o(1))n^{r/2}$ subsets of $[n]$ with the following property: no two $r$-sets of $[n]$ are covered more than once and all but $o(n^r)$ of the $r$-sets of $[n]$ are covered. We also give an absolute lower bound $\cp(n,r) \geq {n \choose r}/{q + r - 1 \choose r}$ when $n = q^2 + q + r - 1$, and for each $r$ characterize the finitely many configurations achieving equality with the lower bound. Finally we note the connection of $\cp(n,r)$ to extremal graph theory, and determine some new asymptotically sharp bounds for the Zarankiewicz problem.

math.CO↗

arXiv2010

The de Bruijn-Erdos Theorem for hypergraphs

Fix integers $n \ge r \ge 2$. A clique partition of ${[n] \choose r}$ is a collection of proper subsets $A_1, A_2, ..., A_t \subset [n]$ such that $\bigcup_i{A_i \choose r}$ is a partition of ${[n] \choose r}$. Clique partitions are related to design theory, coding theory, projective geometry, and extremal combinatorics. Let $\cp(n,r)$ denote the minimum size of a clique partition of ${[n] \choose r}$. A classical theorem of de Bruijn and Erd\H os states that $\cp(n, 2) = n$ and also determines the extremal configurations. In this paper we study $\cp(n,r)$, and show in general that for each fixed $r \geq 3$, \[\cp(n,r) \geq (1 + o(1))n^{r/2} \quad \quad {as}n \to \infty.\] We conjecture $\cp(n,r) = (1 + o(1))n^{r/2}$, and prove this conjecture in a very strong sense for $r = 3$ by giving a characterization of optimal clique partitions of ${[n] \choose 3}$ for infinitely many $n$. Precisely, when $n = q^2 + 1$ and $q$ is a prime power, we show \[ \cp(n,3) = n\sqrt{n-1} \] and characterize those clique partitions achieving equality. We also give an absolute lower bound $\cp(n,r) \geq {n \choose r}/{q + r - 1 \choose r}$ when $n = q^2 + q + r - 1$, and for each $r$ characterize the finitely many configurations achieving equality with the lower bound. Finally we note the connection of $\cp(n,r)$ to extremal graph theory, and determine some new asymptotically sharp bounds for the Zarankiewicz problem.

math.CO↗

arXiv2010

Toward a Hajnal-Szemeredi theorem for hypergraphs

Let $H$ be a triple system with maximum degree $d>1$ and let $r>10^7\sqrt{d}\log^{2}d$. Then $H$ has a proper vertex coloring with $r$ colors such that any two color classes differ in size by at most one. The bound on $r$ is sharp in order of magnitude apart from the logarithmic factors. Moreover, such an $r$-coloring can be found via a randomized algorithm whose expected running time is polynomial in the number of vertices of $\cH$. This is the first result in the direction of generalizing the Hajnal-Szemerédi theorem to hypergraphs.

math.CO↗

arXiv2010

Almost all triple systems with independent neighborhoods are semi-bipartite

The neighborhood of a pair of vertices $u,v$ in a triple system is the set of vertices $w$ such that $uvw$ is an edge. A triple system $\HH$ is semi-bipartite if its vertex set contains a vertex subset $X$ such that every edge of $\HH$ intersects $X$ in exactly two points. It is easy to see that if $\HH$ is semi-bipartite, then the neighborhood of every pair of vertices in $\HH$ is an independent set. We show a partial converse of this statement by proving that almost all triple systems with vertex sets $[n]$ and independent neighborhoods are semi-bipartite. Our result can be viewed as an extension of the Erd\H os-Kleitman-Rothschild theorem to triple systems. The proof uses the Frankl-Rödl hypergraph regularity lemma, and stability theorems. Similar results have recently been proved for hypergraphs with various other local constraints.

math.CO↗

arXiv2010

Books vs Triangles

A book of size b in a graph is an edge that lies in b triangles. Consider a graph G with n vertices and \lfloor n^2/4\rfloor +1 edges. Rademacher proved that G contains at least \lfloor n/2\rfloor triangles, and Erdos conjectured and Edwards proved that G contains a book of size at least n/6. We prove the following "linear combination" of these two results. Suppose that α\in (1/2, 1) and the maximum size of a book in G is less than αn/2. Then G contains at least α(1-α) n^2/4 - o(n^2) triangles as n approaches infinity. This is asymptotically sharp. On the other hand, for every α\in (1/3, 1/2), there exists β>0 such that G contains at least βn^3 triangles. It remains an open problem to determine the largest possible βin terms of α. Our proof uses the Ruzsa-Szemeredi theorem.

math.CO↗

arXiv2009

Almost all cancellative triple systems are tripartite

A triple system is cancellative if no three of its distinct edges satisfy $A \cup B=A \cup C$. It is tripartite if it has a vertex partition into three parts such that every edge has exactly one point in each part. It is easy to see that every tripartite triple system is cancellative. We prove that almost all cancellative triple systems with vertex set [n] are tripartite. This sharpens a theorem of Nagle and Rodl on the number of cancellative triple systems. It also extends recent work of Person and Schacht who proved a similar result for triple systems without the Fano configuration. Our proof uses the hypergraph regularity lemma of Frankl and Rodl, and a stability theorem for cancellative triple systems due to Keevash and the second author.

math.CO↗

arXiv2009

Counting substructures III: quadruple systems

For various quadruple systems F, we give asymptotically sharp lower bounds on the number of copies of F in a quadruple system with a prescribed number of vertices and edges. Our results extend those of Furedi, Keevash, Pikhurko, Simonovits and Sudakov who proved under the same conditions that there is one copy of $F$. Our proofs use the hypergraph removal Lemma and stability results for the corresponding Turan problem proved by the above authors.

math.CO↗

arXiv2009

Counting substructures I: color critical graphs

Let $F$ be a graph which contains an edge whose deletion reduces its chromatic number. We prove tight bounds on the number of copies of $F$ in a graph with a prescribed number of vertices and edges. Our results extend those of Simonovits, who proved that there is one copy of $F$, and of Rademacher, Erd\H os and Lovász-Simonovits, who proved similar counting results when $F$ is a complete graph. One of the simplest cases of our theorem is the following new result. There is an absolute positive constant $c$ such that if $n$ is sufficiently large and $1 \le q < cn$, then every $n$ vertex graph with $n$ even and $n^2/4 +q$ edges contains at least $q(n/2)(n/2-1)(n/2-2)$ copies of a five cycle. Similar statements hold for any odd cycle and the bounds are best possible.

math.CO↗

arXiv2009

Counting substructures II: triple systems

For various triple systems $F$, we give tight lower bounds on the number of copies of $F$ in a triple system with a prescribed number of vertices and edges. These are the first such results for hypergraphs, and extend earlier theorems of Bollobás, Frankl, Füredi, Keevash, Pikhurko, Simonovits, and Sudakov who proved that there is one copy of $F$. A sample result is the following: Füredi-Simonovits and independently Keevash-Sudakov settled an old conjecture of Sós by proving that the maximum number of triples in an $n$ vertex triple system (for $n$ sufficiently large and even) that contains no copy of the Fano plane is $p(n)={n/2 \choose 2}n.$ We prove that there is an absolute constant $c$ such that if $n$ is sufficiently large and $1 \le q \le cn^2$, then every $n$ vertex triple system with $p(n)+q$ edges contains at least $6q({n/2 \choose 4}+(n/2 -3){n/2 \choose 3}$$ copies of the Fano plane. This is sharp for $q\le n/2-2$. Our proofs use the recently proved hypergraph removal lemma and stability results for the corresponding Turán problem.

math.CO↗

arXiv2009

Biclique Coverings and the Chromatic Number

Consider a graph $G$ with chromatic number $k$ and a collection of complete bipartite graphs, or bicliques, that cover the edges of $G$. We prove the following two results: \medskip \noindent $\bullet$ If the bicliques partition the edges of $G$, then their number is at least $2^{\sqrt{\log_2 k}}$. This is the first improvement of the easy lower bound of $\log_2 k$, while the Alon-Saks-Seymour conjecture states that this can be improved to $k-1$. \medskip \noindent $\bullet$ The sum of the orders of the bicliques is at least $(1-o(1))k\log_2 k$. This generalizes, in asymptotic form, a result of Katona and Szemerédi who proved that the minimum is $k\log_2 k$ when $G$ is a clique.

math.CO↗

arXiv2009

Coloring H-free Hypergraphs

Fix $r \ge 2$ and a collection of $r$-uniform hypergraphs $\cH$. What is the minimum number of edges in an $\cH$-free $r$-uniform hypergraph with chromatic number greater than $k$. We investigate this question for various $\cH$. Our results include the following: An $(r,l)$-system is an $r$-uniform hypergraph with every two edges sharing at most $l$ vertices. For $k$ sufficiently large, the minimum number of edges in an $(r,l)$-system with chromatic number greater than $k$ is at most $c(k^{r-1}\log k)^{l/(l-1)}$, where $$c<...$$ This improves on the previous best bounds of Kostochka-Mubayi-Rödl-Tetali \cite{KMRT}. The upper bound is sharp aside from the constant $c$ as shown in \cite{KMRT}. The minimum number of edges in an $r$-uniform hypergraph with independent neighborhoods and chromatic number greater than $k$ is of order $\tilde k^{r+1/(r-1)}$ as $k \to \infty$. This generalizes (aside from logarithmic factors) a result of Gimbel and Thomassen \cite{GT} for triangle-free graphs. Let $T$ be an $r$-uniform hypertree of $t$ edges. Then every $T$-free $r$-uniform hypergraph has chromatic number at most $p(t)$, where $p(t)$ is a polynomial in $t$. This generalizes the well known fact that every $T$-free graph has chromatic number at most $t$. Several open problems and conjectures are also posed.

math.CO↗

arXiv2008

Coloring Simple Hypergraphs

Fix an integer $k \ge 3$. A $k$-uniform hypergraph is simple if every two edges share at most one vertex. We prove that there is a constant $c$ depending only on $k$ such that every simple $k$-uniform hypergraph $H$ with maximum degree $\D$ has chromatic number satisfying $$χ(H) <c (\frac{\D}{\log \D})^{\frac{1}{k-1}}.$$ This implies a classical result of Ajtai-Komlós-Pintz-Spencer-Szemerédi and its strengthening due to Duke-Lefmann-Rödl. The result is sharp apart from the constant $c$.

math.CO↗

arXiv1999

Multiple vertex coverings by specified induced subgraphs

Given graphs H_1,...,H_k, we study the minimum order of a graph G such that for each i, the induced copies of H_i in G cover V(G). We prove a general upper bound of twice the sum of the numbers m_i, where m_i is one less than the order of H_i. When k=2 and one graph is an independent set of size n, we determine the optimum within a constant. When k=2 and the graphs are a star and an independent set, we determine the answer exactly.

math.CO↗

arXiv1999

Edge-bandwidth of graphs

The edge-bandwidth of a graph is the minimum, over all labelings of the edges with distinct integers, of the maximum difference between labels of two incident edges. We prove that edge-bandwidth is at least as large as bandwidth for every graph, with equality for certain caterpillars. We obtain sharp or nearly-sharp bounds on the change in edge-bandwidth under addition, subdivision, or contraction of edges. We compute edge-bandwidth for cliques, bicliques, caterpillars, and some theta graphs.

math.CO↗