Number of distinguishing colorings and partitions

https://doi.org/10.1016/j.disc.2020.111984Get rights and content

Abstract

A vertex coloring of a graph G is called distinguishing (or symmetry breaking) if no non-identity automorphism of G preserves it, and the distinguishing number, shown by D(G), is the smallest number of colors required for such a coloring. This paper is about counting non-equivalent distinguishing colorings of graphs with k colors. A parameter, namely Φk(G), which is the number of non-equivalent distinguishing colorings of a graph G with at most k colors, is shown here to have an application in calculating the distinguishing number of the lexicographic product and the X-join of graphs. We study this index (and some other similar indices) which is generally difficult to calculate. Then, we show that if one knows the distinguishing threshold of a graph G, which is the smallest number of colors θ(G) so that, for kθ(G), every k-coloring of G is distinguishing, then, in some special cases, counting the number of distinguishing colorings with k colors is very easy. We calculate θ(G) for some classes of graphs including the Kneser graph K(n,2). We then turn to vertex partitioning by studying the distinguishing coloring partition of a graph G; a partition of vertices of G which induces a distinguishing coloring for G. There, we introduce Ψk(G) as the number of non-equivalent distinguishing coloring partitions with at most k cells, which is a generalization to its distinguishing coloring counterpart.

Introduction

Breaking symmetries of graphs via vertex coloring is a subject initiated by Babai’s work [3] in 1977. There, he introduced an asymmetric coloring of a graph, and proved that a tree has an asymmetric coloring with two colors if all vertices have the same degree α2, where α can be an arbitrary finite or infinite cardinal. The concept was later called distinguishing coloring in the literature, since the appearance of [1] by Albertson and Collins in 1996.

This paper is about counting non-equivalent distinguishing colorings and partitions of a given graph with k colors. Considering this number in case of 2 colors is as old as symmetry breaking in graphs; in 1977 Babai [3] tried to count distinguishing colorings of infinite trees, while in 1991 Polat and Sabidussi [19] tried to count essentially different asymmetrising sets in finite and infinite trees which are distinguishing (coloring) partitions with 2 cells in our terminology (see Section 4).

A vertex coloring of a graph G is called distinguishing if it is only preserved by the identity automorphism; in this case, we say that the coloring breaks all the symmetries of G. By a k-distinguishing coloring we mean a distinguishing coloring which uses exactly k colors. The distinguishing number of a graph G, denoted D(G), is the smallest number d such that there exists a distinguishing vertex coloring of G with d colors. A graph G is called d-distinguishable if there exists a distinguishing vertex coloring with d colors [1]. The distinguishing number of some important classes of graphs are as follows: D(Kn)=n, D(Kn,n)=n+1, D(Pn)=2 for n2, D(C3)=D(C4)=D(C5)=3 while D(Cn)=2 for n6 [1].

A whole wealth of results on the subject has already been generated. Among many, we can only mention a few, only those that have essentially important results, or those that have introduced new indices based on distinguishing colorings. For a connected finite graph G, it was independently proved by Collins and Trenk [5] and Klavžar et al. [18] that D(G)Δ+1, where Δ is the largest degree of G. Equality holds if and only if G is a complete graph KΔ+1, a balanced complete bipartite graph KΔ,Δ, or C5. Collins and Trenk [5] also mixed the concept of distinguishing colorings with proper vertex colorings to introduce the distinguishing chromatic number χD(G) of a graph G. It is defined as the minimum number of colors required to properly color the vertices of G such that this coloring is only preserved by the trivial automorphism. They also showed that, for a finite connected graph G, we have χD(G)2Δ(G) and that equality holds only if G is isomorphic to KΔ,Δ or C6.

Symmetry breaking can also happen via other kinds of graph colorings. An analogous index for an edge coloring, namely the distinguishing index D(G), has been introduced by Kalinowski and Pilśniak in [16] as the minimum number of the required colors in an asymmetric edge-coloring of a connected graph GK2. Moreover, they showed that D(G)Δ(G) for a finite connected graph G, unless G is isomorphic to C3, C4 or C5. Another analogous index is introduced by Kalinowski, Pilśniak and Woźniak in [17]; the total distinguishing number D(G) is the minimum number of required colors in an asymmetric total coloring of G.

To generalize some results from the finite case to the infinite ones, Imrich, Klavžar and Trofimov [14] considered the distinguishing number for infinite graphs. They showed that for an infinite connected graph G we have D(G)n, where n is a cardinal number such that the degree of any vertex of G is not greater than n. Most symmetry breaking concepts have their relative counterparts in the infinite case, however, there are some (such as the Infinite Motion Conjecture) that arise only when we consider infinite graphs. As an instance, one can take a look at [13] by Imrich et al. which contains comparisons of some distinguishing indices of connected infinite graphs.

It was also interesting to know the distinguishing number for product graphs. For example, Bogstad and Cowen [4] showed that for k4, every hypercube Qk of dimension k, which is the Cartesian product of k copies of K2, is 2-distinguishable. It has also been shown by Imrich and Klavžar in [15] that the distinguishing number of Cartesian powers of a connected graph G is equal to two except for K22,K32,K23. Meanwhile, Imrich, Jerebic and Klavžar [12] showed that Cartesian products of relatively prime graphs whose sizes are close to each other can be distinguished with a small number of colors. Moreover, Estaji et al. in [8] proved that for every pair of connected graphs G and H with |H||G|<2|H||H|, we have D(GH)2. Gorzkowska, Kalinowski and Pilśniak proved a similar result for the distinguishing index of the Cartesian product [10].

The lexicographic product was a subject of symmetry breaking via vertex and edge coloring by Alikhani and Soltani in [2], where they showed that under some conditions on the automorphism group of a graph G, we have D(G)D(Gk)D(G)+k1, where Gk is the kth lexicographic power of G, for any natural number k. As well, they showed that if G and H are connected graphs, then D(H)D(GH)|V(G)|D(H).

Coloring is not the only mean of symmetry breaking in graphs. For example, one might break the symmetries of a graph via a more general tool such as vertex partitioning. Ellingham and Schroeder introduced distinguishing partition of a graph as a partition of the vertex set that is preserved by no nontrivial automorphism [7]. Here, unlike coloring, some graphs have no distinguishing partition. Anyhow, for a graph G that admits a distinguishing partition, one may think of the minimum number of required cells in a distinguishing partition of the vertex set. Here, we show this index by DP(G).

In this paper, we introduce some further indices related to symmetry breaking of graphs by studying the number of non-equivalent distinguishing colorings of a graph with k colors and some other similar quantities. This is motivated by the problem of evaluating the distinguishing number of a lexicographic product or the X-join of some graphs, which we consider in Section 5.

The paper is organized as follows. In Section 2, we consider the number of non-equivalent distinguishing colorings of a graph G with (exactly) k colors, namely Φk(G) (and φk(G)) and, we calculate these indices for some simple types of graphs. Afterwards in Section 3, we introduce the distinguishing threshold as a dual index to the distinguishing number. It is shown that calculations of some indices introduced in Sections 2 Non-equivalent distinguishing colorings, 4 Non-equivalent distinguishing partitions are easier, in some cases, when we know the distinguishing threshold. Moreover, in Section 4, we consider the number of non-equivalent distinguishing coloring partitions of a graph G with (exactly) k cells, namely Ψk(G) (and ψk(G)) and, we calculate them in some special cases. Additionally in Section 4, some other auxiliary indices are also introduced. Then, we present an application of one of the indices introduced here, namely Φk(G), in Section 5. We finally conclude the paper by shedding some lights on the future investigations in Section 6.

Here, we use the standard notation and terminology of graph theory, which can be found in [6]. We only remind that the set of neighbors of a vertex v in G is denoted by N(v), while N[v] stands for the set N(v){v}.

Section snippets

Non-equivalent distinguishing colorings

Two colorings c1 and c2 of a graph G are called equivalent if there is an automorphism α of G such that c1(v)=c2(α(v)) for all vV(G).

The number of non-equivalent distinguishing colorings of a graph G with {1,,k} as the set of admissible colors is shown by Φk(G), while the number of non-equivalent k-distinguishing colorings of a graph G with {1,,k} as the set of colors is shown by φk(G). When G has no distinguishing colorings with exactly k colors, we put φk(G)=0. It is also clear that ΦD(G)(G

Distinguishing threshold

For any graph G, we define the distinguishing threshold θ(G) to be the minimum number t such that for any kt, any arbitrary coloring of G with k colors is distinguishing. For example θ(Kn)=θ(Kn¯)=n and θ(Km,n)=m+n. Note, also, that for an asymmetric graph G, we have θ(G)=D(G)=1. Moreover, we always have θ(G)D(G).

Let G be a graph on n vertices. If any two distinct vertices u,vG have different set of neighbors other than themselves, then any (n1)-coloring of G has to be distinguishing because

Non-equivalent distinguishing partitions

In this section, we turn our attention to the case of different distinguishing partitions of graphs. Let G be a graph and let P1 and P2 be two partitions of the vertices of G. We say P1 and P2 are equivalent if there is a non-trivial automorphism of G which maps P1 onto P2. The number of non-equivalent partitions of G, with at most k cells, is called the partition number of G and is denoted by Πk(G).

Meanwhile, a distinguishing coloring partition of a graph G is a partition of the vertices of G

Distinguishing lexicographic products

In this section we provide an important application of one of the indices introduced in this paper, namely Φk(G). We start by recalling some preliminaries to the topic of lexicographic product of graphs.

Let X be a graph. The X-join of {Yx|xV(X)}, is the graph Z with V(Z)={(x,y):xX,yYx}and E(Z)={(x,y)(x,y):xxE(X) or else x=x and yyE(Yx)}.Whenever, for all xX, we have YxY, for a fixed graph Y, the graph Z is called the lexicographic product of X and Y and we write Z=XY.

We remind the

Conclusion

We have seen in Section 5 that counting the number of non-equivalent distinguishing colorings, Φ, has an application in finding the distinguishing number of lexicographic products. Moreover, other indices have shown to have deep interactions with each other and also with Φ. It should be noted that calculating these indices are not always easy. Even when the automorphism group is very small and simple, counting non-equivalent distinguishing colorings or distinguishing (coloring) partitions faces

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

We owe a great debt to professor Wilfried Imrich, who proposed several problems, which led to this paper, during his visit to Shiraz University.

References (20)

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

Cited by (0)

View full text