Polynomial cases of graph decomposition: A complete solution of Holyer’s problem

Let H be a fixed graph. A graph G has an H-decomposition if the edge set of G can be partitioned into subsets inducing graphs isomorphic to H. Let P_H denote the following decision problem: “Does an instance graph G admit H-decomposition?” In this paper we prove that the problem P_H is polynomial time solvable if H is a graph whose every component has at most 2 edges. This way we complete a solution of Holyer’s problem which is the problem of classifying the problems P_H according to their computational complexities.     

Mutually orthogonal graph squares

A decomposition ??={G₁, G₂,…,G_s} of a graph G is a partition of the edge set of G into edge‐disjoint subgraphs G₁, G₂,…,G_s. If G_i?H for all i∈{1, 2, …, s}, then ?? is a decomposition of G by H. Two decompositions ??={G₁, G₂, …, G_n} and ?={F₁, F₂,…,F_n} of the complete bipartite graph K_n,n are orthogonal if |E(G_i)∩E(F_j)|=1 for all i,j∈{1, 2, …, n}. A set of decompositions {??₁, ??₂, …, ??_k} of K_{n, n} is a set of k mutually orthogonal graph squares (MOGS) if ??_i and ??_j are orthogonal for all i, j∈{1, 2, …, k} and i≠j. For any bipartite graph G with n edges, N(n, G) denotes the maximum number k in a largest possible set {??₁, ??₂, …, ??_k} of MOGS of K_{n, n} by G. El‐Shanawany conjectured that if p is a prime number, then N(p, P_{p+ 1})=p, where P_{p+ 1} is the path on p+ 1 vertices. In this article, we prove this conjecture. © 2009 Wiley Periodicals, Inc. J Combin Designs 17: 369–373, 2009     

Edge decompositions into two kinds of graphs

Let H be an arbitrary graph and let K_1,2 be the 2-edge star. By a {K_1,2,H}-decomposition of a graph G we mean a partition of the edge set of G into subsets inducing subgraphs isomorphic to K_1,2 or H. Let J be an arbitrary connected graph of odd size. We show that the problem to decide if an instance graph G has a {K_1,2,H}-decomposition is NP-complete if H has a component of an odd size and H≠pK_1,2∪qJ, where pK_1,2∪qJ is the disjoint union of p copies of K_1,2 and q copies of J. Moreover, we prove polynomiality of this problem for H=qJ.     

Eccentric sequences and eccentric sets in graphs

By a graph we mean a finite undirected connected graph of order p, p ? 2, with no loops or multiple edges. A finite non-decreasing sequence S: s₁, s₂, …, s_p, p ? 2, of positive integers is an eccentric sequence if there exists a graph G with vertex set V(G) = {v₁, v₂, …, v_p} such that for each i, 1 ? i ? p, s, is the eccentricity of v₁. A set S is an eccentric set if there exists a graph G such that the eccentricity ρ(v₁) is in S for every v₁ ? V(G), and every element of S is the eccentricity of some vertex in G. In this note we characterize eccentric sets, and we find the minimum order among all graphs whose eccentric set is a given set, to obtain a new necessary condition for a sequence to be eccentric. We also present some properties of graphs having preassigned eccentric sequences.     

Odd‐angulated graphs and cancelling factors in box products

Under what conditions is it true that if there is a graph homomorphism G □ H → G □ T, then there is a graph homomorphism H→ T? Let G be a connected graph of odd girth 2k + 1. We say that G is (2k + 1)‐angulated if every two vertices of G are joined by a path each of whose edges lies on some (2k + 1)‐cycle. We call G strongly (2k + 1)‐angulated if every two vertices are connected by a sequence of (2k + 1)‐cycles with consecutive cycles sharing at least one edge. We prove that if G is strongly (2k + 1)‐angulated, H is any graph, S, T are graphs with odd girth at least 2k + 1, and ?: G□ H→S□T is a graph homomorphism, then either ? maps G□{h} to S□{t_h} for all h∈V(H) where t_h∈V(T) depends on h; or ? maps G□{h} to {s_h}□ T for all h∈V(H) where s_h∈V(S) depends on h. This theorem allows us to prove several sufficient conditions for a cancelation law of a graph homomorphism between two box products with a common factor. We conclude the article with some open questions. © 2008 Wiley Periodicals, Inc. J Graph Theory 58:221‐238, 2008     

On graph equivalences preserved under extensions

Let G be the set of finite graphs whose vertices belong to some fixed countable set, and let ≡ be an equivalence relation on G. By the strengthening of ≡ we mean an equivalence relation ≡_s such that G≡_sH, where G,H∈G, if for every F∈G, G∪F≡H∪F. The most important case that we study in this paper concerns equivalence relations defined by graph properties. We write G≡^ΦH, where Φ is a graph property and G,H∈G, if either both G and H have the property Φ, or both do not have it. We characterize the strengthening of the relations ≡^Φ for several graph properties Φ. For example, if Φ is the property of being a k-connected graph, we find a polynomially verifiable (for k fixed) condition that characterizes the pairs of graphs equivalent with respect to . We obtain similar results when Φ is the property of being k-colorable, edge 2-colorable, Hamiltonian, or planar, and when Φ is the property of containing a subgraph isomorphic to a fixed graph H. We also prove several general theorems that provide conditions for ≡_s to be of some specific form. For example, we find a necessary and sufficient condition for the relation ≡_s to be the identity. Finally, we make a few observations on the strengthening in a more general case when G is the set of finite subsets of some countable set.     

Resolutions of PG(5, 2) with point‐cyclic automorphism group

A t‐(υ, k, λ) design is a set of υ points together with a collection of its k‐subsets called blocks so that all subsets of t points are contained in exactly λ blocks. The d‐dimensional projective geometry over GF(q), PG(d, q), is a 2‐(q^d + q^d−1 + … + q + 1, q + 1, 1) design when we take its points as the points of the design and its lines as the blocks of the design. A 2‐(υ, k, 1) design is said to be resolvable if the blocks can be partitioned as ℛ = {R₁, R₂, …, R_s}, where s = (υ − 1)/(k−1) and each R_i consists of υ/k disjoint blocks. If a resolvable design has an automorphism σ which acts as a cycle of length υ on the points and ℛ^σ = ℛ, then the design is said to be point‐cyclically resolvable. The design associated with PG(5, 2) is known to be resolvable and in this paper, it is shown to be point‐cyclically resolvable by enumerating all inequivalent resolutions which are invariant under a cyclic automorphism group G = 〈σ〉 where σ is a cycle of length 63. These resolutions are the only resolutions which admit a point‐transitive automorphism group. Furthermore, some necessary conditions for the point‐cyclic resolvability of 2‐(υ, k, 1) designs are also given. © 2000 John Wiley & Sons, Inc. J Combin Designs 8: 2–14, 2000     

Multicolor Ramsey numbers involving graphs with long suspended paths

Let G₁,…,G_c be graphs and let H be a connected graph. Let H_n be a graph on n points which is homeomorphic to H. It is proved that if n is large enough, the Ramsey number r(G₁,…,G_c,H_n) has the form (X?1)(n?1)+T. Here X and T are two Ramsey-type functons involving G₁,…,G_c only. The properties of these functions are studied, leading to explicit evaluations in a number of cases.     

Decompositions of graphs into trees

Let θ be a family of graphs. By a θ-decomposition of a graph G we mean a partition λ of the edge set of G such that every F ? π spans in G a subgraph isomorphic to a graph in θ. In this paper we state the following conjecture: If T₁ and T₂ are two trees having relatively prime sizes then there exists c = c(T₁ T₂) such that every graph G satisfying the condition δ(G) ? c has a {T₁, T₂}-decom-position. We prove this conjecture for some special pairs of trees. In particular, we prove it in the following cases: (i) T₁ and T₂ are stars having relatively prime sizes; (ii) T₁ and T₂ are paths having relatively prime sizes; and. (iii) T₁ = T₂ - {v}, where v is a terminal vertex in T ₂.     

Embeddings of bipartite graphs

If G is a bipartite graph with bipartition A, B then let G_m,n(A, B) be obtained from G by replacing each vertex a of A by an independent set a₁, …, a_m, each vertex b of B by an independent set b₁,…, b_n, and each edge ab of G by the complete bipartite graph with edges a_ib_j (1 ≤ i ≤ m and 1 ≤ j ≤ n). Whenever G has certain types of spanning forests, then cellular embeddings of G in surfaces S may be lifted to embeddings of G_m,n(A, B) having faces of the same sizes as those of G in S. These results are proved by the technique of “excess-current graphs.” They include new genus embeddings for a large class of bipartite graphs.     

The minimum degree of Ramsey‐minimal graphs

We write H → G if every 2‐coloring of the edges of graph H contains a monochromatic copy of graph G. A graph H is G‐minimal if H → G, but for every proper subgraph H′ of H, H′ ? G. We define s(G) to be the minimum s such that there exists a G‐minimal graph with a vertex of degree s. We prove that s(K_k) = (k ? 1)² and s(K_a,b) = 2 min(a,b) ? 1. We also pose several related open problems. © 2006 Wiley Periodicals, Inc. J Graph Theory 54: 167–177, 2007     

On Threshold Probability for the Stability of Independent Sets in Distance Graphs

This paper considers the so-called distance graph G(n, r, s);its vertices can be identified with the r-element subsets of the set {1, 2,…,n}, and two vertices are joined by an edge if the size of the intersection of the corresponding subsets equals s. Note that, in the case s = 0, such graphs are known as Kneser graphs. These graphs are closely related to the Erd?s-Ko-Rado problem; they also play an important role in combinatorial geometry and coding theory.

We study properties of random subgraphs of the graph G(n, r, s) in the Erd?s-Rényi model, in which each edge is included in the subgraph with a certain fixed probability p independently of the other edges. It is known that if r > 2s + 1, then, for p = 1/2, the size of an independent set is asymptotically stable in the sense that the independence number of a random subgraph is asymptotically equal to that of the initial graph G(n, r, s). This gives rise to the question of how small p must be for asymptotic stability to cease. The main result of this paper is the answer to this question.

     

On the complexity of reconstructing H‐free graphs from their Star Systems

In the Star System problem we are given a set system and asked whether it is realizable by the multi‐set of closed neighborhoods of some graph, i.e. given subsets S₁, S₂, …, S_n of an n‐element set V does there exist a graph G = (V, E) with {N[v]: v∈V} = {S₁, S₂, …, S_n}? For a fixed graph H the H‐free Star System problem is a variant of the Star System problem where it is asked whether a given set system is realizable by closed neighborhoods of a graph containing no H as an induced subgraph. We study the computational complexity of the H‐free Star System problem. We prove that when H is a path or a cycle on at most four vertices the problem is polynomial time solvable. In complement to this result, we show that if H belongs to a certain large class of graphs the H‐free Star System problem is NP‐complete. In particular, the problem is NP‐complete when H is either a cycle or a path on at least five vertices. This yields a complete dichotomy for paths and cycles. Copyright © 2010 John Wiley & Sons, Ltd. 68:113‐124, 2011     

Packing graphs with odd and even trees

The Gyárfás-Lehel tree-packing conjecture asserts that any sequence T₁, T₂, …, T_n^?1 of trees with 1, 2, …, n - 1 edges packs into the complete graph K_n on n vertices. The present paper examines two conjectures that jointly imply the Gyárfás-Lehel conjecture: 1. For n even, any T₁, T₃, …, T_n^?1 pack into the half-complete graph Hn on n vertices.2. For n odd, any T₂, T₄, …, T_n^?1 pack into the half-complete graph Hn on n vertices. The H_n are uniquely defined by their degree sequences: H_n and H_n+1 are complements in K_n+1. It is shown that H_n and T_n+1 pack into H_n+2 if T_n+1 is a double star, unimodal triple star, interior-3 caterpillar, or scorpion. Hence Conjectures 1 and 2 are true for these specialized types of trees. The conjectures are also valid for all trees when n ≤ 9, so that the Gyárfás-Lehel conjecture holds for n ≤ 9.     

G-decomposition of Kn, where G has four vertices or less

The complete graph K_n, is said to have a G-decomposition if it is the union of edge disjoint subgraphs each isomorphic to G. The set of values of n for which K_n has a G-decomposition is determined if G has four vertices or less.     

1-Factorizations of cartesian products of regular graphs

Let G₁, G₂…, G_n be regular graphs and H be the Cartesian product of these graphs (H = G₁ × G₂ × … × G_n). The following will be proved: If the set {G₁, G₂…, G_n} has at leat one of the following properties: (*) for at leat one i ? {1, 2,…, n}, there exists a 1-factorization of G_i or (**) there exists at least two numbers i and j such that 1 ≤ i < j ≤ n and both the Graphs G_i and G_j contain at least one 1-factor, then there exists a 1-factorization of H. Further results: Let F be a cycle of length greater than three and let G be an arbitrary cubic graph. Then there exists a 1-factorization of the 5-regular graph H = F × G. The last result shows that neither (*) nor (**) is a necessary condition for the existence of a 1-factorization of a Cartesian product of regular graphs.     

Long cycles passing through a specified path in a graph

For a graph G and an integer k ≥ 1, let ς_k(G) = d_G(v_i): {v₁, …, v_k} is an independent set of vertices in G}. Enomoto proved the following theorem. Let s ≥ 1 and let G be a (s + 2)-connected graph. Then G has a cycle of length ≥ min{|V(G)|, ς₂(G) − s} passing through any path of length s. We generalize this result as follows. Let k ≥ 3 and s ≥ 1 and let G be a (k + s − 1)-connected graph. Then G has a cycle of length ≥ min{|V(G)|, − s} passing through any path of length s. © 1998 John Wiley & Sons, Inc. J. Graph Theory 29: 177–184, 1998     

Orthogonal factorizations of graphs

Let G be a graph with vertex set V(G) and edge set E(G). Let k₁, k₂,…,k_m be positive integers. It is proved in this study that every [0,k₁+…+k_m?m+1]‐graph G has a [0, k_i]₁^m‐factorization orthogonal to any given subgraph H with m edges. © 2002 Wiley Periodicals, Inc. J Graph Theory 40: 267–276, 2002     

A decomposition for strongly perfect graphs

A graph G is strongly perfect if every induced subgraph H of G contains a stable set that meets all the maximal cliques of H. We present a graph decomposition that preserves strong perfection: more precisely, a stitch decomposition of a graph G = (V, E) is a partition of V into nonempty disjoint subsets V₁, V₂ such that in every P₄ with vertices in both V_iapos;s, each of the three edges has an endpoint in V₁ and the other in V₂. We give a good characterization of graphs that admit a stitch decomposition and establish several results concerning the stitch decomposition of strongly perfect graphs.     

Covering the vertices of a graph by cycles of prescribed length

The main theorem of that paper is the following: let G be a graph of order n, of size at least (n² - 3n + 6)/2. For any integers k, n₁, n₂,…,n_k such that n = n₁ + n₂ +. + n_k and n_i ? 3, there exists a covering of the vertices of G by disjoint cycles (C_i) =l…k with |C_i| = ni, except when n = 6, n₁ = 3, n₂ = 3, and G is isomorphic to G₁, the complement of G₁ consisting of a C₃ and a stable set of three vertices, or when n = 9, n₁ = n₂ = n₃ = 3, and G is isomorphic to G₂, the complement of G₂ consisting of a complete graph on four vertices and a stable set of five vertices. We prove an analogous theorem for bipartite graphs: let G be a bipartite balanced graph of order 2n, of size at least n² - n + 2. For any integers s, n₁, n₂,…,n_s with n_i ? 2 and n = n₁ + n₂ + ? + n_s, there exists a covering of the vertices of G by s disjoint cycles C_i, with |C_i| = 2n_i.