Difference sets and inverting the difference operator

For a setA of non-negative numbers, letD(A) (the difference set ofA) be the set of nonnegative differences of elements ofA, and letD ^k be thek-fold iteration ofD. We show that for everyk, almost every set of non-negative integers containing 0 arises asD ^k (A) for someA. We also give sufficient conditions for a setA to be the unique setX such that 0X andD ^k (X)=D ^k (A). We show that for eachm there is a setA such thatD(X)=D(A) has exactly 2 ^m solutionsX with 0X.This work was supported by grants DMS 92-02833 and DMS 91-23478 from the National Science Foundation. The first author acknowledges the support of the Hungarian National Science Foundation under grants, OTKA 4269, and OTKA 016389, and the National Security Agency (grant No. MDA904-95-H-1045).Lee A. Rubel died March 25, 1995. He is very much missed by his coauthors.     

An extremal problem related to probability

Summary We consider a walk from a stateA ₁ to a stateA _n+1 in which the probability of remaining atA _i isp _i , and the probability of progressing fromA _i toA _i+1 is 1 –p _i . The probabilityW _nk of reachingA _n+1 fromA ₁ in exactlyn + k steps can then be expressed as a polynomial of degreen + k in then variablesp ₁,,p _n . We determine the maximum value ofW _nk and the (unique) choice (p ₁,,p _n ) for which this extremum occurs.Dedicated to Professor Otto Haupt with best wishes on his 100th birthday     

Directeds-t numberings,Rubber bands,and testing digraphk-vertex connectivity

LetG=(V, E) be a directed graph andn denote |V|. We show thatG isk-vertex connected iff for every subsetX ofV with |X| =k, there is an embedding ofG in the (k–1)-dimensional spaceR ^k–1,fVR ^k–1, such that no hyperplane containsk points of {f(v)|vV}, and for eachvV–X, f(v) is in the convex hull of {f(w)| (v, w)E}. This result generalizes to directed graphs the notion of convex embeddings of undirected graphs introduced by Linial, Lovász and Wigderson in Rubber bands, convex embeddings and graph connectivity,Combinatorica 8 (1988), 91–102.Using this characterization, a directed graph can be tested fork-vertex connectivity by a Monte Carlo algorithm in timeO((M(n)+nM(k)) · (logn)) with error probability<1/n, and by a Las Vegas algorithm in expected timeO((M(n)+nM(k)) ·k), whereM(n) denotes the number of arithmetic steps for multiplying twon×n matrices (M(n)=O(n ^2.376)). Our Monte Carlo algorithm improves on the best previous deterministic and randomized time complexities fork>n ^0.19; e.g., for , the factor of improvement is >n ^0.62. Both algorithms have processor efficient parallel versions that run inO((logn)²) time on the EREW PRAM model of computation, using a number of processors equal to logn times the respective sequential time complexities. Our Monte Carlo parallel algorithm improves on the number of processors used by the best previous (Monte Carlo) parallel algorithm by a factor of at leastn ²/(logn)³ while having the same running time.Generalizing the notion ofs-t numberings, we give a combinatorial construction of a directeds-t numbering for any 2-vertex connected directed graph.     

Small transversals in hypergraphs

For each positive integerk, we consider the setA _k of all ordered pairs [a, b] such that in everyk-graph withn vertices andm edges some set of at mostam+bn vertices meets all the edges. We show that eachA _k withk2 has infinitely many extreme points and conjecture that, for every positive , it has only finitely many extreme points [a, b] witha. With the extreme points ordered by the first coordinate, we identify the last two extreme points of everyA _k , identify the last three extreme points ofA ₃, and describeA ₂ completely. A by-product of our arguments is a new algorithmic proof of Turán's theorem.     

Thep-intersection number of a complete bipartite graph and orthogonal double coverings of a clique

Thep-intersection graph of a collection of finite sets {S _i } _i=1 ⁿ is the graph with vertices 1, ...,n such thati, j are adjacent if and only if |S _i S _j |p. Thep-intersection number of a graphG, herein denoted _p (G), is the minimum size of a setU such thatG is thep-intersection graph of subsets ofU. IfG is the complete bipartite graphK _n,n andp2, then _p (K _{n, n} )(n ²+(2p–1)n)/p. Whenp=2, equality holds if and only ifK _n has anorthogonal double covering, which is a collection ofn subgraphs ofK _n , each withn–1 edges and maximum degree 2, such that each pair of subgraphs shares exactly one edge. By construction,K _n has a simple explicit orthogonal double covering whenn is congruent modulo 12 to one of {1, 2, 5, 7, 10, 11}.Research supported in part by ONR Grant N00014-5K0570.     

On construction ofk-wise independent random variables

A 0–1probability space is a probability space (, 2,P), where the sample space -{0, 1} ⁿ for somen. A probability space isk-wise independent if, whenY _i is defined to be theith coordinate or the randomn-vector, then any subset ofk of theY _i 's is (mutually) independent, and it is said to be a probability spacefor p ₁,p ₂, ...,p _n ifP[Y _i =1]=p _i .We study constructions ofk-wise independent 0–1 probability spaces in which thep _i 's are arbitrary. It was known that for anyp ₁,p ₂, ...,p _n , ak-wise independent probability space of size always exists. We prove that for somep ₁,p ₂, ...,p _n [0,1],m(n,k) is a lower bound on the size of anyk-wise independent 0–1 probability space. For each fixedk, we prove that everyk-wise independent 0–1 probability space when eachp _i =k/n has size (n _k ). For a very large degree of independence —k=[n], for >1/2- and allp _i =1/2, we prove a lower bound on the size of . We also give explicit constructions ofk-wise independent 0–1 probability spaces.This author was supported in part by NSF grant CCR 9107349.This research was supported in part by the Israel Science Foundation administered by the lsrael Academy of Science and Humanities and by a grant of the Israeli Ministry of Science and Technology.     

Two combinatorial problems on posets

Bialostocki proposed the following problem: Let nk2 be integers such that k|n. Let p(n, k) denote the least positive integer having the property that for every poset P, |P|p(n, k) and every Z _k -coloring f: P Z _k there exists either a chain or an antichain A, |A|=n and _aA f(a) 0 (modk). Estimate p(n, k). We prove that there exists a constant c(k), depends only on k, such that (n+k–2)²–c(k) p(n, k) (n+k–2)²+1. Another problem considered here is a 2-dimensional form of the monotone sequence theorem of Erdös and Szekeres. We prove that there exists a least positive integer f(n) such that every integral square matrix A of order f(n) contains a square submatrix B of order n, with all rows monotone sequences in the same direction and all columns monotone sequences in the same direction (direction means increasing or decreasing).     

Erd?s-Ko-Rado for three sets

Fix integers k?3 and n?3k/2. Let F be a family of k-sets of an n-element set so that whenever A,B,C∈F satisfy |A∪B∪C|?2k, we have A∩B∩C≠∅. We prove that with equality only when ?_F∈FF≠∅. This settles a conjecture of Frankl and Füredi [2], who proved the result for n?k²+3k.     

Edge-isoperimetric inequalities in the grid

The grid graph is the graph on [k] ⁿ ={0,...,k–1} ⁿ in whichx=(x _i ) ₁ ⁿ is joined toy=(y _i ) ₁ ⁿ if for somei we have |x _i –y _i |=1 andx _j =y _j for allji. In this paper we give a lower bound for the number of edges between a subset of [k] ⁿ of given cardinality and its complement. The bound we obtain is essentially best possible. In particular, we show that ifA[k] ⁿ satisfiesk ⁿ /4|A|3k ⁿ /4 then there are at leastk ^n–1 edges betweenA and its complement.Our result is apparently the first example of an isoperimetric inequality for which the extremal sets do not form a nested family.We also give a best possible upper bound for the number of edges spanned by a subset of [k] ⁿ of given cardinality. In particular, forr=1,...,k we show that ifA[k] ⁿ satisfies |A|r ⁿ then the subgraph of [k] ⁿ induced byA has average degree at most 2n(1–1/r).Research partially supported by NSF Grant DMS-8806097     

On-line coloring of perfect graphs

Lovász, Saks, and Trotter showed that there exists an on-line algorithm which will color any on-linek-colorable graph onn vertices withO(nlog^(2k–3) n/log^(2k–4) n) colors. Vishwanathan showed that at least (log ^k–1 n/k ^k ) colors are needed. While these remain the best known bounds, they give a distressingly weak approximation of the number of colors required. In this article we study the case of perfect graphs. We prove that there exists an on-line algorithm which will color any on-linek-colorable perfect graph onn vertices withn ^10k/loglogn colors and that Vishwanathan's techniques can be slightly modified to show that his lower bound also holds for perfect graphs. This suggests that Vishwanathan's lower bound is far from tight in the general case.Research partially supported by Office of Naval Research grant N00014-90-J-1206.     

Paley-Wiener type theorems on harmonic extensions ofH-type groups

Letn=vz be anH-type group, and letna be the harmonic semidirect product ofn withaR. LetNA be the corresponding simply connected Lie group. If dimv=m and dimz=k, denoteQ=m/2+k. We prove that the spherical Fourier transform is a topological isomorphism between thep-Schwartz spacel ^p (N,A),(0<p2), (0<p2), and the space of holomorphic rapidly decreasing functions on the strip {sC:|Re(s)|<Q/2} with =2/p–1.     

Smallest maximally nonhamiltonian graphs

A graphG ismaximally nonhamiltonian iffG is not hamiltonian butG + e is hamiltonian for each edgee inG ^c, i.e., any two non-adjacent vertices ofG are ends of a hamiltonian path. Bollobás posed the problem of finding the least number of edges,f(n), possible in a maximally nonhamiltonian graph of ordern. Results of Bondy show thatf(n) 3/2 n forn 7. We exhibit graphs of even ordern 36 for which the bound is attained. These graphs are the snarks,J _k, of Isaacs and mild variations of them. For oddn 55 we construct graphs from the graphsJ _k showing that in this case,f(n) = 3n + 1/2 or 3n + 3/2 and leave the determination of which is correct as an open problem. Finally we note that the graphsJ _k, k 7 are hypohamiltonian cubics with girth 6.     

On sets of integers with prescribed gaps

For a fixed setI of positive integers we consider the set of paths (p _o,...,p _k ) of arbitrary length satisfyingp _l –p _l–1I forl=2,...,k andp ₀=1,p _k =n. Equipping it with the uniform distribution, the random path lengthT _n is studied. Asymptotic expansions of the moments ofT _n are derived and its asymptotic normality is proved. The step lengthsp _l –p _l–1 are seen to follow asymptotically a restricted geometrical distribution. Analogous results are given for the free boundary case in which the values ofp ₀ andp _k are not specified. In the special caseI={m+1,m+2,...} (for some fixed m) we derive the exact distribution of a random m-gap subset of {1,...,n} and exhibit some connections to the theory of representations of natural numbers. A simple mechanism for generating a randomm-gap subset is also presented.     

Infinite spectra in the first order theory of graphs

The psectrum Spec(A) of a sentenceA is, roughly, the set of those a for whichA has a threshold function at or nearp=n ^–a . Examples are given ofA with infinite spectra and with spectra of order type ⁱ for arbitraryi.     

On the vector space of 0-configurations

Let α be a rational-valued set-function on then-element sexX i.e. α(B) εQ for everyB ⫅X. We say that α defines a 0-configuration with respect toA⫅2 ^x if for everyA εA we have α(B)=0. The 0-configurations form a vector space of dimension 2 ⁿ − |A| (Theorem 1). Let 0 ≦t<k ≦n and letA={A ⫅X: |A| ≦t}. We show that in this case the 0-configurations satisfying α(B)=0 for |B|>k form a vector space of dimension , we exhibit a basis for this space (Theorem 4). Also a result of Frankl, Wilson [3] is strengthened (Theorem 6).     

Addendum To Schrijver's Work On Minimum Permanents

Let denote the set of n×n binary matrices which have each row and column sum equal to k. For 2≤k≤n→∞ we show that is asymptotically equal to (k−1)^k−1k^2−k. This confirms Conjecture 23 in Minc's catalogue of open problems. * Written while the author was employed by the Department of Computer Science at the Australian National University.     

On the divisibility of homogeneous hypergraphs

We denote byK _k ,k, 2, the set of allk-uniform hypergraphsK which have the property that every element subset of the base ofK is a subset of one of the hyperedges ofK. So, the only element inK ₂ ² are the complete graphs. If is a subset ofK _k then there is exactly one homogeneous hypergraphH whose age is the set of all finite hypergraphs which do not embed any element of . We callH -free homogeneous graphsH _n have been shown to be indivisible, that is, for any partition ofH _n into two classes, oue of the classes embeds an isomorphic copy ofH _n . [5]. Here we will investigate this question of indivisibility in the more general context of-free homogeneous hypergraphs. We will derive a general necessary condition for a homogeneous structure to be indivisible and prove that all-free hypergraphs for K _k with 3 are indivisible. The-free hypergraphs with K _k ² satisfy a weaker form of indivisibility which was first shown by Henson [2] to hold forH _n . The general necessary condition for homogeneous structures to be indivisible will then be used to show that not all-free homogeneous hypergraphs are indivisible.This research has been supported by NSERC grant 69–1325.     

Lifts, Discrepancy and Nearly Optimal Spectral Gap*

We present a new explicit construction for expander graphs with nearly optimal spectral gap. The construction is based on a series of 2-lift operations. Let G be a graph on n vertices. A 2-lift of G is a graph H on 2n vertices, with a covering map π :H →G. It is not hard to see that all eigenvalues of G are also eigenvalues of H. In addition, H has n “new” eigenvalues. We conjecture that every d-regular graph has a 2-lift such that all new eigenvalues are in the range (if true, this is tight, e.g. by the Alon–Boppana bound). Here we show that every graph of maximal degree d has a 2-lift such that all “new” eigenvalues are in the range for some constant c. This leads to a deterministic polynomial time algorithm for constructing arbitrarily large d-regular graphs, with second eigenvalue . The proof uses the following lemma (Lemma 3.3): Let A be a real symmetric matrix with zeros on the diagonal. Let d be such that the l₁ norm of each row in A is at most d. Suppose that for every x,y ∈{0,1}ⁿ with ‹x,y›=0. Then the spectral radius of A is O(α(log(d/α)+1)). An interesting consequence of this lemma is a converse to the Expander Mixing Lemma. * This research is supported by the Israeli Ministry of Science and the Israel Science Foundation.     

On a Turán type problem of Erdös

Let L ^k be the graph formed by the lowest three levels of the Boolean lattice B _k , i.e.,V(L ^k )={0, 1,...,k, 12, 13,..., (k–1)k} and 0is connected toi for all 1ik, andij is connected toi andj (1i<jk).It is proved that if a graph G overn vertices has at leastk ^3/2 n ^3/2 edges, then it contains a copy of L ^k .Research supported in part by the Hungarian National Science Foundation under Grant No. 1812     

Ramsey partitions of integers and pair divisions

Ifk ₁ andk ₂ are positive integers, the partitionP = (₁,₂,..., _n ) ofk ₁+k ₂ is said to be a Ramsey partition for the pairk ₁,k ₂ if for any sublistL ofP, either there is a sublist ofL which sums tok ₁ or a sublist ofP –L which sums tok ₂. Properties of Ramsey partitions are discussed. In particular it is shown that there is a unique Ramsey partition fork ₁,k ₂ having the smallest numbern of terms, and in this casen is one more than the sum of the quotients in the Euclidean algorithm fork ₁ andk ₂.An application of Ramsey partitions to the following fair division problem is also discussed: Suppose two persons are to divide a cake fairly in the ratiok ₁k ₂. This can be done trivially usingk ₁+k ₂-1 cuts. However, every Ramsey partition ofk ₁+k ₂ also yields a fair division algorithm. This method yields fewer cuts except whenk ₁=1 andk ₂=1, 2 or 4.