Elsevier

Discrete Applied Mathematics

Volume 284, 30 September 2020, Pages 179-194
Discrete Applied Mathematics

Maximum nullity and zero forcing number on graphs with maximum degree at most three

https://doi.org/10.1016/j.dam.2020.03.027Get rights and content

Abstract

A dynamic coloring of a graph G starts with an initial subset FV(G) of colored vertices, while all the remaining vertices are non-colored. At each time step, a colored vertex with exactly one non-colored neighbor forces this non-colored neighbor to be colored. The initial set F is called a zero forcing set of G if, by iteratively applying the forcing process, every vertex in G becomes colored. The zero forcing number of G, denoted by F(G), is the cardinality of a minimum zero forcing set of G. The maximum nullity of G, denoted by M(G), is the largest possible nullity over all |V(G)| by |V(G)| real symmetric matrices A whose non-diagonal entries are non-zero if the corresponding vertices are adjacent in G and with no restriction for its diagonal entries. In this paper, we characterize all graphs G of order n, maximum degree at most three, and F(G)=3. Also we classify these graphs with their maximum nullity.

Introduction

Following the notation of [3], [9], for a graph G, we define the forcing process as follows: Let FV(G) be an initial set of vertices. At step zero, each vertex in F is colored and any other vertex is non-colored. For i1, at step i, each colored vertex v with exactly one non-colored neighbor forces its non-colored neighbor to be colored. At each step i, the set of newly get colored vertices is denoted by Vi. In particular V0=F. This process will continue until there is no more possible change. At the end of the process, the set of colored vertices is called the derived set. A zero forcing set of G is a set FV(G) of initially colored vertices if the corresponding derived set is equal to V(G), i.e., V(G)=Vi. For each vertex vV(G), the step that v gets colored is denoted by kF(v). Note that kF(v)=i if and only if vVi. The zero forcing number of a graph G, denoted by F(G), is the minimum possible cardinality for a zero forcing set of G and any zero forcing set of G with cardinality F(G) is named an F(G)-set. Throughout the paper, for brevity, we would rather use “forcing” instead of “zero forcing”. It is simple to see that F(G)=|V(G)| if and only if G is an empty graph i.e., a graph with no edge.

Let F be an F(G)-set. In the forcing process with initial vertex set F, any vertex forces at most one vertex while a vertex can be forced by different vertices. Therefore, V(G) can be partitioned to |F| chains (sequences) R1,,R|F|. For each of these chains, say Ri=(v0,v1,,vk), v0F and vj is forced by vj1 for each j[k]. The set {R1,,R|F|} is called a chain set with respect to F for G, which is not necessarily unique. For each j{0,,k1}, we define nextRi(vj)=vj+1 and for each j[k], we set prevRi(vj)=vj1. As a simple observation, one can see that each chain Ri=(v0,v1,,vk) induces a path Pi=v0v1vk in G, i.e, G[V(Ri)] is an induced path. To see that, note that if there exist two adjacent vertices vj1,vj2V(Ri), where j2>j1+1, then when vj1 forces vj1+1, its neighbor vj2 is non-colored which is impossible. A chain is called trivial if it has only one vertex.

For a graph G, a total forcing set is a forcing set of G inducing a subgraph with no isolated vertex. For simplicity, we write TF-set instead of total forcing set. The total forcing number of G, denoted by Ft(G), is the cardinality of a minimum TF-set in G. Total forcing set was first introduced by Davila [6] as a strengthening of forcing set which was originally introduced by the AIM-minimum rank group [14]. It is observed (for instance, see [6], [8]) that for a graph G with no isolated vertex, F(G)Ft(G)2F(G).

The forcing process and the forcing number were introduced in [5] and [14] to bound the minimum rank of a graph and hence its maximum nullity. Since then, the forcing number has gained a considerable attention in graph theory and has been related to many graph theoretic parameters. In general, study of forcing number is challenging for many reasons. First, it is difficult to compute, as it is known to be NP-hard [10], [18]. Further, many of the known bounds leave a wide gap for graphs in general. For example, the forcing number of a graph of order n can be as low as δ(G) (see [4]), and as high as nΔ(G)(Δ(G)+1) (see [2]).

Gentner and Rautenbach [12] proved that for a graph G with maximum degree Δ=Δ(G)3 and girth at least 5, we have F(G)(n2)n24log2n+6+2,moreover, they introduced two graphs G1 and G2 and proved F(G)n(Δ2)(Δ1) for any connected graph G{KΔ+1,KΔ,Δ,KΔ1,Δ,G1,G2}. For more details and the definitions of G1 and G2, see [12]. They also conjectured that F(G)n3+2 for any graph G with n vertices and Δ=3. In [13], an infinite family of graphs {Gn} with maximum degree 3 was introduced such that the forcing number of Gn is at least 49|V(Gn)|, a counter example to the Gentner–Rautenbach conjecture. Note that, at this point, the best upper bound for the forcing number of connected graphs with maximum degree three [2] is F(G)n2+1.The equality can be achieved for graphs K4 and K3,3. Davila and Henning [7] studied forcing sets and total forcing sets in claw-free cubic graphs. They proved F(G)<n2, where G is a connected claw-free cubic graph with n10 vertices. Akbari and Vatandoost [1] gave a partial answer to the problem of determining all graphs G with M(G)=F(G) posed by AIM Minimum Rank-Special Graphs Work Group [14], where M(G) is the maximum nullity of G (for the definition, see Section 1.1). Indeed, they characterized all cubic graphs with forcing number 3 and, as a corollary, they also showed that M(G)=3 for any graph G in this family.

Problem 1 [14]

Determine all graphs G for which M(G)=F(G).

In this paper, we characterize all graphs G with maximum degree at most three and forcing number 3. We will moreover investigate the maximum nullity of these graphs comparing to their forcing numbers, a partial answer to Problem 1.

For notation and terminology not presented here, we refer readers to [15]. We will use the notation Pn and Kn to respectively denote the path and the complete graph on n vertices. Also we use the standard notation [k]={1,2,,k}. Let G be a graph. For a path P=x0xm in G, we define xiP=xixm and Pxi=x0xi for each i[m].

Let Sn(R) be the set of all symmetric n by n matrices over the real numbers. For A=(aij)Sn(R), the graph of A, denoted by G(A), is a graph with vertex set {v1,,vn} and edge set vivj:aij0,1i<jn. It should be noted that the diagonal of A has no role in the definition of G(A). The set of symmetric matrices of graph G is S(G)={ASn(R):G(A)=G}. The minimum rank of a graph G, denoted by mr(G), is the minimum possible rank for a matrix in S(G) and, similarly, the maximum nullity of G, denoted by M(G), is the maximum nullity of symmetric matrices in S(G). Clearly, mr(G)+M(G)=n and M(G)1 for any graph G.

Johnson et al. [16] defined the graph of 2- parallel paths. A graph G, which is not a path, is said to be a graph of 2-parallel paths if there are two vertex disjoint induced paths covering all the vertices and G can be drawn in the plane so that these two paths are parallel horizontal lines if we forgot their vertices, and moreover the edges (drawn as segments, not curves) with ends in different paths do not cross. Particularly, union of two disjoint paths is a graph of 2-parallel paths.

As a generalization, in a natural way, we can define the graph of k-parallel paths for any integer k1 as follows: Simply, a 1-parallel path is just a path. For an integer k2, a graph G, which is not a graph of (k1)-parallel paths, is said to be a graph of k-parallel paths, if there exist k vertex disjoint induced paths covering all the vertices and G can be drawn in the plane in a way that these paths are parallel horizontal lines if we forgot their vertices, and moreover the edges (drawn as segments, not curves) whose ends are in different paths do not cross each other. Such a drawing is called a standard drawing of G, for example see Fig. 1. Note that a graph G may have several standard drawings. Here after, for a fixed standard drawing of a graph of k-parallel paths, the k paths used in the definition are called the parallel paths with respect to this drawing and any edge with end-points on different paths is called a segment.

It is clear that any graph of k-parallel paths has at least k vertices. Also, It is known that every planar graph can be drawn in the plane such that its edges are segment intersecting only at their end-points, see [11]. Consequently, every planar graph of order n is a graph of k-parallel paths for some kn. Clearly, a non-planar graphs of order n is not a graph of n-parallel paths. However, the following assertion indicates that the complete graphs of order n6 are not of k-parallel paths for each kn.

Observation 2

For n6, Kn is not a graph of k-parallel paths for each k[n].

Proof

For a contradiction, suppose that Kn is a graph of k-parallel paths where k[n]. Consider a standard drawing of Kn. Since Kn is a complete graph and the parallel paths are induced paths, then each of its parallel paths must be of order one or two. Furthermore, since Kn is not a planar graph, we know kn1. Consequently, there must be some parallel path isomorphic to P2. Clearly, there is only one such a P2 path, otherwise, the segments between two P2 paths intersect each others which is not possible. Therefore, there exist n2 parallel paths P1,,Pn2 isomorphic to P1 and exactly one parallel path Pn1 isomorphic to P2. Let V be all the vertices in Kn but a vertex from Pn1. In view of the standard drawing of Kn, the induced subgraph by V is a planar graph isomorphic to Kn1 which is not possible since n6.  

Section snippets

Main results

Akbari et al. [1] characterized all cubic graphs with forcing number 3 and proved that the maximum nullity of these graphs is 3 as well. In the next two results, as a generalization of their result, we characterize all the graphs with maximum degree at most three and F(G)=3.

Theorem 3

For a graph G with Δ(G)3, F(G)=3 if and only if G is a graph of 3-parallel paths. In particular, the left-most vertices of the parallel paths in any standard drawing of G form an F(G)-set.

As a consequence of Theorem 3, in

Known results and preliminary lemmas

This section is devoted to review some known results and to prove some key lemmas being used for the proofs of main results. We start with the following obvious observation.

Observation 7

[17]

For a graph G, F(G)=1 if and only if G is isomorphic to a path.

As stated before, the forcing number has been defined as a tool for studying the maximum nullity of graphs. The next proposition states that the maximum nullity of a graph G cannot exceed F(G).

Proposition 8

[14]

For any graph G, M(G)F(G).

Hence, the next theorem is a stronger

A sufficient condition for being of 3-parallel paths

Let G be a graph with Δ(G)3 whose vertex set can be partitioned into three induced vertex disjoint paths P1,P2,P3 satisfying the following properties:

  • 1.

    Each of P1 and P2 has at least two vertices.

  • 2.

    If P3 is a singleton, then its degree is at most two.

  • 3.

    For any edge xy with endpoints x and y on different paths Pi and Pj respectively, there is no edge xy such that xPi,yPj, and x<Pix and y<Pjy.

  • 4.

    For {i,j,k}=[3], if there are two edges ab and cd such that aPi,b,cPj, dPk, and c<Pjb, then there

Proof of main results

This section is mainly concerned with proving the main results stated in Section 2. In the following proposition, we present an upper bound for the forcing number of a graph of 3-parallel paths.

Proposition 32

For any graph G of 3-parallel paths, F(G)3. In particular, the left-most vertices of any standard drawing of G form a forcing set.

Proof

Consider a standard drawing of a graph G with parallel paths Q1,Q2,Q3, where their indices are set according to their position in this drawing, i.e., for i<j, the path Qi is

References (18)

There are more references available in the full text version of this article.

This work is a part of Elahe Rezaei-Sani’s PhD dissertation under the supervision of Meysam Alishahi. All the authors have the same contributions to this work.

View full text