This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

On the δ\delta-chromatic numbers of the Cartesian products of graphs

Wipawee Tangjai Department of Mathematics, Faculty of Science, Mahasarakham University, Maha Sarakham 44150, Thailand [email protected] Witsarut Pho-on Department of Mathematics, Faculty of Science, Srinakharinwirot University, Sukhumvit 23, 10110 Bangkok, Thailand [email protected]  and  Panupong Vichitkunakorn Division of Computational Science, Faculty of Science, Prince of Songkla University, Songkla 90110, Thailand [email protected]
Abstract.

In this work, we study the δ\delta-chromatic number of a graph which is the chromatic number of the δ\delta-complement of a graph. We give a structure of the δ\delta-complements and sharp bounds on the δ\delta-chromatic numbers of the Cartesian products of graphs. Furthermore, we compute the δ\delta-chromatic numbers of various classes of Cartesian product graphs, including the Cartesian products between cycles, paths, and stars.

Keyword: delta-complement graph, chromatic number, Cartesian product, coloring
MSC: 05C07, 05C15, 05C35, 05C38, 05C69

1. Introduction

The concept of δ\delta-complement was introduced in 2022 [5]. Their research focused on exploring various intriguing characteristics of these graphs, including properties like δ\delta-self-complementary, adjacency, and hamiltonicity. In 2023, Vichitkunakorn et al. [7] introduced the term δ\delta-chromatic number of a graph GG which refers to the chromatic number of the δ\delta-complement of GG. They established a Nordhaus-Gaddum bound type relation between the chromatic number and the δ\delta-chromatic number across various parameters: the clique number, the number of vertices and the degrees of vertices. The given bounds are sharp and the classes of graphs satisfying those bounds are given [7]. In this study, we present a more detailed outcome concerning the δ\delta-chromatic number of the Cartesian product of graphs.

In 1957, Sabidussi [6] showed that the chromatic number of the Cartesian product graphs is equal to the maximum chromatic number between such two graphs. A lot of subsequent research has been exploring different types of chromatic numbers of the Cartesian product graphs such as list chromatic number [2], packing chromatic number [3] and bb-chromatic number [1, 4].

We first recall some basic notations and definitions needed in this article. Let GG be a graph. For a subset UU of V(G)V(G), G[U]G[U] denotes the subgraph induced by UU. A vertex coloring cc of GG is a proper coloring if each pair of adjacent vertices has distinct colors. The chromatic number of GG, denoted by χ(G)\chi(G), is the minimum number of colors needed so that (G,c)(G,c) is properly colored. For each vertex uV(G)u\in V(G), we use notation dG(u)d_{G}(u) for the degree of uu in GG. Throughout this article, we let PnP_{n} be a path with nn vertices, KnK_{n} be a complete graph with nn vertices and CnC_{n} be a cycle with nn vertices. We let S1,nS_{1,n} be a star with nn pendants. For graphs GG and HH, the Cartesian product of GG and HH, denoted by GHG\square H, is a graph where V(GH)=V(G)×V(H)V(G\square H)=V(G)\times V(H) and uvE(GH)uv\in E(G\square H) if either x=xx=x^{\prime} and yyE(H)yy^{\prime}\in E(H) or y=yy=y^{\prime} and xxE(G)xx^{\prime}\in E(G) for u=(x,y)u=(x,y) and v=(x,y)v=(x^{\prime},y^{\prime}).

In this work, we give a structure of the δ\delta-complement of the finite Cartesian products of graphs. Sharp bounds on the δ\delta-chromatic number (the chromatic number of δ\delta-complement) of the finite Cartesian products of graphs are also given. In addition, we determine the specific value of the δ\delta-chromatic numbers of various classes of the Cartesian product of well-known graphs such as cycle, path, and star.

2. Preliminary results

In this section, we review some basic definitions and previous results.

Definition 1 ([5]).

The δ\delta-complement of a graph GG, denoted GδG_{\delta}, is a graph obtained from GG by using the same vertex set and the following edge conditions: uvE(Gδ)uv\in E(G_{\delta}) if

  1. (1)

    d(u)=d(v)d(u)=d(v) in GG and uvE(G)uv\notin E(G), or

  2. (2)

    d(u)d(v)d(u)\neq d(v) in GG and uvE(G)uv\in E(G).

Definition 2 ([7]).

A δ\delta-chromatic number χδ(G)\chi_{\delta}(G) of a graph GG is the chromatic number of GδG_{\delta}.

Results on the δ\delta-chromatic numbers of some important graphs are χδ(Pn)=n22\chi_{\delta}(P_{n})=\left\lceil\frac{n-2}{2}\right\rceil for n5n\geq 5 [7], χδ(Cn)=n2\chi_{\delta}(C_{n})=\left\lceil\frac{n}{2}\right\rceil [7], and χδ(Wn)=1+χδ(Cn)=1+n2\chi_{\delta}(W_{n})=1+\chi_{\delta}(C_{n})=1+\left\lceil\frac{n}{2}\right\rceil.

Theorem 3 ([6]).

Let GG and HH be graphs. We have χ(GH)=max{χ(G),χ(H)}\chi(G\square H)=\max\{\chi(G),\chi(H)\}.

Theorem 4 ([7]).

For n4n\geq 4, let GG be a graph with nn vertices. Let d1,,dmd_{1},\dots,d_{m} be all distinct values of the degrees of the vertices in GG. Partition V(G)V(G) into non-empty sets Vd1,Vd2,,VdmV_{d_{1}},V_{d_{2}},\dots,V_{d_{m}}. We have

max1im{|Vdi|}χ(G)χδ(G)(m+n2)2\max_{1\leq i\leq m}\{|V_{d_{i}}|\}\leq\chi(G)\cdot\chi_{\delta}(G)\leq\left(\frac{m+n}{2}\right)^{2}

and

2max1im{|Vdi|}χ(G)+χδ(G)m+n.2\cdot\sqrt{\max_{1\leq i\leq m}\{|V_{d_{i}}|\}}\leq\chi(G)+\chi_{\delta}(G)\leq m+n.

3. Structure of the δ\delta-complements of Cartesian products

This section contains the structure of the δ\delta-complement of the Cartesian product of graphs.

The following theorem shows that the edge set of the δ\delta-complements of the Cartesian product contains the edge set of the Cartesian product of the δ\delta-complements of graphs. It is a fundamental result that will be used throughout what follows.

Theorem 5.

For graphs GG and HH, we have (GH)δ=(V,E)(G\square H)_{\delta}=(V,E) where V=V(GH)V=V(G\square H) and E=E(GδHδ)SE=E(G_{\delta}\square H_{\delta})\cup S where S={uv:u=(u1,u2)V(GH) and v=(v1,v2)V(GH) where u1v1,u2v2 and dGH(u)=dGH(v)}S=\{uv:u=(u_{1},u_{2})\in V(G\square H)\text{ and }v=(v_{1},v_{2})\in V(G\square H)\text{ where }u_{1}\neq v_{1},u_{2}\neq v_{2}\text{ and }d_{G\square H}(u)=d_{G\square H}(v)\}.

Proof.

()(\Longrightarrow) Let u=(u1,u2)u=(u_{1},u_{2}) and v=(v1,v2)v=(v_{1},v_{2}) be distinct vertices in (GH)δ(G\square H)_{\delta} where uvE((GH)δ)uv\in E((G\square H)_{\delta}). It follows that either

  • uvE(GH)uv\in E(G\square H) and dGH(u)dGH(v)d_{G\square H}(u)\neq d_{G\square H}(v), or

  • uvE(GH)uv\not\in E(G\square H) and dGH(u)=dGH(v)d_{G\square H}(u)=d_{G\square H}(v).

In case uvE(GH)uv\in E(G\square H) and dGH(u)dGH(v)d_{G\square H}(u)\neq d_{G\square H}(v), without loss of generality, we suppose that u1=v1u_{1}=v_{1}, u2v2u_{2}\neq v_{2} and u2v2E(H)u_{2}v_{2}\in E(H). Since dG(u1)=dG(v1)d_{G}(u_{1})=d_{G}(v_{1}), it follows that dH(u2)dH(v2)d_{H}(u_{2})\neq d_{H}(v_{2}). Thus u2v2E(Hδ)u_{2}v_{2}\in E(H_{\delta}). Hence uvE(GδHδ)uv\in E(G_{\delta}\square H_{\delta}). In case uvE(GH)uv\not\in E(G\square H) and dGH(u)=dGH(v)d_{G\square H}(u)=d_{G\square H}(v), we have u1v1u_{1}\neq v_{1} and u2v2u_{2}\neq v_{2}. Thus uvSuv\in S.

()(\Longleftarrow) Let uvE(GδHδ)Suv\in E(G_{\delta}\square H_{\delta})\cup S. Consider uvE(GδHδ)uv\in E(G_{\delta}\square H_{\delta}). Without loss of generality, we suppose that u1=v1u_{1}=v_{1}, u2v2u_{2}\neq v_{2} and u2v2E(Hδ)u_{2}v_{2}\in E(H_{\delta}). If dGH(u)=dGH(v)d_{G\square H}(u)=d_{G\square H}(v), then dH(u2)=dH(v2)d_{H}(u_{2})=d_{H}(v_{2}). Thus u2v2E(H)u_{2}v_{2}\not\in E(H). Hence uvE(GH)uv\not\in E(G\square H). Since dGH(u)=dGH(v)d_{G\square H}(u)=d_{G\square H}(v), we have uvE((GH)δ)uv\in E((G\square H)_{\delta}). If dGH(u)dGH(v)d_{G\square H}(u)\neq d_{G\square H}(v), then dH(u2)dH(v2)d_{H}(u_{2})\neq d_{H}(v_{2}). Thus u2v2E(H)u_{2}v_{2}\in E(H) and uvE(GH)uv\in E(G\square H). Since dGH(u)dGH(v)d_{G\square H}(u)\neq d_{G\square H}(v), we have uvE((GH)δ)uv\in E((G\square H)_{\delta}). Now, we consider uvSuv\in S. We have that u1v1u_{1}\neq v_{1} and u2v2u_{2}\neq v_{2}. So uvE(GH)uv\notin E(G\square H). Since dGH(u)=dGH(v)d_{G\square H}(u)=d_{G\square H}(v), it follows that uvE((GH)δ)uv\in E((G\square H)_{\delta}). ∎

Corollary 6.

Let GG and HH be graphs. We have (GH)δ=GδHδ(G\square H)_{\delta}=G_{\delta}\square H_{\delta} if and only if for any u=(u1,u2)u=(u_{1},u_{2}) and v=(v1,v2)v=(v_{1},v_{2}) in V(GH)V(G\square H) where u1v1u_{1}\neq v_{1} and u2v2u_{2}\neq v_{2}, we have dGH(u)dGH(v)d_{G\square H}(u)\neq d_{G\square H}(v).

In general, we have the following theorem for a finite Cartesian product of graphs.

Theorem 7.

For graphs G1,,GkG_{1},\dots,G_{k}, we have (G1Gk)δ=(V,E)(G_{1}\square\cdots\square G_{k})_{\delta}=(V,E) where V=V(G1Gk)V=V(G_{1}\square\cdots\square G_{k}) and E=E((G1)δ(Gk)δ)SE=E((G_{1})_{\delta}\square\cdots\square(G_{k})_{\delta})\cup S such that SS is the set of uvuv where u=(u1,,uk)Vu=(u_{1},\dots,u_{k})\in V, v=(v1,,vk)Vv=(v_{1},\dots,v_{k})\in V, there are at least two indices ii that uiviu_{i}\neq v_{i}, and dG1Gk(u)=dG1Gk(v)d_{G_{1}\square\cdots\square G_{k}}(u)=d_{G_{1}\square\cdots\square G_{k}}(v).

Proof.

It is well-known that two vertices u=(u1,,uk)u=(u_{1},\dots,u_{k}) and (v1,,vk)(v_{1},\dots,v_{k}) in G1GkG_{1}\square\dots\square G_{k} are adjacent if and only if there is exactly one ii such that uiviu_{i}\neq v_{i} and uiviE(Gi)u_{i}v_{i}\in E(G_{i}). The rest of the proof follows similar arguments as in Theorem 5. ∎

The following three results are applications of Theorem 7.

Theorem 8.

(G1Gk)δ=(G1)δ(Gk)δ(G_{1}\square\cdots\square G_{k})_{\delta}=(G_{1})_{\delta}\square\cdots\square(G_{k})_{\delta} if and only if there are at most one ii such that GiK1G_{i}\neq K_{1}.

Proof.

From Theorem 7, we need to show that S=S=\emptyset if and only if there are at most one ii such that GiK1G_{i}\neq K_{1}.

Suppose that there are iji\neq j such that GiK1G_{i}\neq K_{1} and GjK1G_{j}\neq K_{1}. Choose u=(u1,,uk)u=(u_{1},\dots,u_{k}) and v=(v1,,v2)v=(v_{1},\dots,v_{2}) such that uiviu_{i}\neq v_{i}, ujvju_{j}\neq v_{j}, dGi(ui)=dGi(vi)d_{G_{i}}(u_{i})=d_{G_{i}}(v_{i}), dGj(uj)=dGj(vj)d_{G_{j}}(u_{j})=d_{G_{j}}(v_{j}) and u=vu_{\ell}=v_{\ell} for all {i,j}\ell\not\in\{i,j\}. So dG1Gk(u)=dG1Gk(v)d_{G_{1}\square\cdots\square G_{k}}(u)=d_{G_{1}\square\cdots\square G_{k}}(v). Then uvSuv\in S. Hence SS\neq\emptyset.

The converse is obvious. ∎

Corollary 9.

(GH)δ=GδHδ(G\square H)_{\delta}=G_{\delta}\square H_{\delta} if and only if G=K1G=K_{1} or H=K1H=K_{1}.

4. Bounds on the δ\delta-chromatic numbers of Cartesian products

In this section, we provide some exact numbers and bounds on the δ\delta-chromatic numbers of some common graphs.

Theorem 10.

Let G1,,GkG_{1},\dots,G_{k} be graphs. We have

max{χδ(G1),,χδ(Gk)}χδ(G1Gk).\max\{\chi_{\delta}(G_{1}),\dots,\chi_{\delta}(G_{k})\}\leq\chi_{\delta}(G_{1}\square\cdots\square G_{k}).
Proof.

The proof follows directly from Theorem 3 and 7. ∎

Theorem 11.

Let GG and HH be graphs. If any positive degree difference of vertices in GG is not equal to that of in HH, then

χδ(GH)nmax(H)max(χδ(G),m(H))\chi_{\delta}(G\square H)\leq n_{\max}(H)\cdot\max(\chi_{\delta}(G),m(H))

where nmax(H)n_{\max}(H) denotes the maximum number of vertices of the same degree in HH and m(H)m(H) is the number of different degrees in HH. Furthermore, the bound is sharp.

Proof.

By Theorem 5 and the assumption that any positive degree difference of vertices in GG is not equal to that of in HH, the edges in SS are uvuv where u=(u1,u2)u=(u_{1},u_{2}), v=(v1,v2)v=(v_{1},v_{2}) such that u1v1u_{1}\neq v_{1}, u2v2u_{2}\neq v_{2}, dG(u1)=dG(v1)d_{G}(u_{1})=d_{G}(v_{1}) and dH(u2)=dH(v2)d_{H}(u_{2})=d_{H}(v_{2}). We partition V(H)V(H) according to vertex degree into W1,W2,,Wm(H)W_{1},W_{2},\dots,W_{m(H)}. Write Wj={hj,1,hj,2,,hj,nj}W_{j}=\{h_{j,1},h_{j,2},\dots,h_{j,n_{j}}\} for 1jm(H)1\leq j\leq m(H).

Define p=max(χδ(G),m(H))p=\max(\chi_{\delta}(G),m(H)). Let c0:V(G){1,2,,χδ(G)}c_{0}:V(G)\to\{1,2,\dots,\chi_{\delta}(G)\} be a proper coloring of GδG_{\delta}. We define a coloring c:V(G)×V(H){1,2,,nmax(H)p}c:V(G)\times V(H)\to\{1,2,\dots,n_{\max}(H)\cdot p\} as

c(g,hj,k)=f(g,j)+(k1)p,c(g,h_{j,k})=f(g,j)+(k-1)p,

for k=1,,njk=1,\dots,n_{j}, where f(g,j){1,2,,p}f(g,j)\in\{1,2,\dots,p\} and f(g,j)c0(g)+j1(modp)f(g,j)\equiv c_{0}(g)+j-1\pmod{p}. The first copy of GG in W1W_{1} gets the original coloring c0c_{0}, while we keep adding pp to the coloring of each other copy of GG in W1W_{1}. In other WjW_{j}, we perform different cyclic permutations modulo pp to c0c_{0} and assign it to the first copy of GG in WjW_{j}. See Table 1 for an example. We see that the vertices in the same copy of GG received a coloring equivalent to c0c_{0} and a cyclic permutation modulo pp up to an additive constant (k1)p(k-1)p for some k=1,,njk=1,\dots,n_{j}. For a fixed gV(G)g\in V(G), the vertices in the same copy of HH, written in the form (g,hj,k)(g,h_{j,k}) where 1jm(H)1\leq j\leq m(H) and 1knj1\leq k\leq n_{j}, received distinct colors because jpj\leq p and knmax(H)k\leq n_{\max}(H).

Lastly, any endpoints of an edge in SS are of the form (g,hj,k)(g,h_{j,k}) and (g,hj,k)(g^{\prime},h_{j,k^{\prime}}) where ggg\neq g^{\prime} and kkk\neq k^{\prime}, which received different colors as kkk\neq k^{\prime}. The sharpness of the bound appears in Theorem 13. ∎

h1,1h_{1,1} h1,2h_{1,2} h2,1h_{2,1} h3,1h_{3,1} h3,2h_{3,2} h4,1h_{4,1} h4,2h_{4,2}g1g_{1} 1 5 2 3 7 4 8 g2g_{2} 3 7 4 1 5 2 6 g3g_{3} 1 5 2 3 7 4 8 g4g_{4} 2 6 3 4 8 1 5 g5g_{5} 3 7 4 1 5 2 6
Table 1. An example of a coloring in the proof of Theorem 11 where χδ(G)=3\chi_{\delta}(G)=3, m(H)=4m(H)=4 and nmax(H)=2n_{\max}(H)=2.
Corollary 12.

Let GG be a graph with χδ(G)2\chi_{\delta}(G)\geq 2. If dG(v)dG(u)+1d_{G}(v)\neq d_{G}(u)+1 for all u,vV(G)u,v\in V(G), then χδ(GP3)2χδ(G)\chi_{\delta}(G\square P_{3})\leq 2\chi_{\delta}(G).

Theorem 13.

For n5n\geq 5, we have χδ(CnP3)=2χδ(Cn)=2n2\chi_{\delta}(C_{n}\square P_{3})=2\chi_{\delta}(C_{n})=2\left\lceil\frac{n}{2}\right\rceil.

Proof.

Let P3=v1v2v3P_{3}=v_{1}v_{2}v_{3}. It is easy to see that χδ(Cn)=χ(Cn¯)=n2\chi_{\delta}(C_{n})=\chi(\overline{C_{n}})=\left\lceil\frac{n}{2}\right\rceil. Since CnC_{n} is regular, it also follows that dCnP3(u,v1)=dCnP3(w,v3)d_{C_{n}\square P_{3}}(u,v_{1})=d_{C_{n}\square P_{3}}(w,v_{3}) for all u,wV(Cn)u,w\in V(C_{n}). Since (u,v1)(u,v_{1}) is not adjacent to (w,v3)(w,v_{3}) in CnP3C_{n}\square P_{3}, it follows that each vertex in the first copy of CnC_{n} is adjacent to all the vertices in the third copy of CnC_{n} in (CnP3)δ(C_{n}\square P_{3})_{\delta}. Hence the colors used in the two copies do not coincide. Thus χδ(CnP3)2χδ(Cn)\chi_{\delta}(C_{n}\square P_{3})\geq 2\chi_{\delta}(C_{n}). By Corollary 12, we can conclude that χδ(CnP3)=2χδ(Cn)\chi_{\delta}(C_{n}\square P_{3})=2\chi_{\delta}(C_{n}). ∎

Example 14.

For m3m\geq 3, we have χδ(S1,mP3)2(m+1)\chi_{\delta}(S_{1,m}\square P_{3})\leq 2(m+1).

Proof.

Let G=S1,mG=S_{1,m} and H=P3H=P_{3}. Theorem 11 gives the desired upper bound. ∎

The bound in Example 14 is not sharp. When G=K1HG=K_{1}\vee H is a join of a singleton and a regular graph HH, the following theorem gives an improved upper bound on χδ(GP3)\chi_{\delta}(G\square P_{3}) in terms of χδ(H)\chi_{\delta}(H). Examples of the graph GG include stars S1,m=K1NmS_{1,m}=K_{1}\vee N_{m} (in Theorem 17), wheels Wm=K1CmW_{m}=K_{1}\vee C_{m}, and windmills K1mKnK_{1}\vee mK_{n}.

Theorem 15.

Let HH be a kk-regular graph. Let G={u}HG=\{u\}\vee H be the join of a singleton and HH. Suppose |V(H)|3|V(H)|\geq 3 and χδ(H)2\chi_{\delta}(H)\geq 2. If |V(H)|>k+2|V(H)|>k+2, then χδ(GP3)2χδ(H)\chi_{\delta}(G\square P_{3})\leq 2\chi_{\delta}(H).

Proof.

Let rV(H)r\in V(H). Note that dG(u)=|V(H)|d_{G}(u)=|V(H)|. Since dGP3(r,vi)=dG(r)+dP3(vi)k+3<dG(u)+1dGP3(u,vj)d_{G\square P_{3}}(r,v_{i})=d_{G}(r)+d_{P_{3}}(v_{i})\leq k+3<d_{G}(u)+1\leq d_{G\square P_{3}}(u,v_{j}) for i,j{1,2,3}i,j\in\{1,2,3\}, we have (r,vi)(r,v_{i}) and (u,vj)(u,v_{j}) are adjacent in (GP3)δ(G\square P_{3})_{\delta} if and only if i=ji=j.

Let HδiH_{\delta}^{i} be the ii-th copy of HδH_{\delta} in GδG_{\delta} for i=1,2,3i=1,2,3. Next, we construct a proper coloring cc as follows. We trivially color Hδ1H_{\delta}^{1}, as a copy of HδH_{\delta}, by a χδ(H)\chi_{\delta}(H)-coloring. Since each vertex in Hδ1H_{\delta}^{1} is adjacent to any vertices in Hδ3H_{\delta}^{3}, it requires 2χδ(H)2\chi_{\delta}(H) colors for Hδ1H^{1}_{\delta} and Hδ3H^{3}_{\delta}. We color Hδ3H^{3}_{\delta} using c(r,v3)=c(r,v1)+χδ(H)c(r,v_{3})=c(r,v_{1})+\chi_{\delta}(H). For i=1,3i=1,3, we notice that a vertex (r,vi)V(Hδi)(r,v_{i})\in V(H^{i}_{\delta}) and (s,v2)V(Hδ2)(s,v_{2})\in V(H^{2}_{\delta}) are adjacent if and only if r=sr=s. We let c(r,v2)=c(r,v1)+1c(r,v_{2})=c(r,v_{1})+1 if 1c(r,v1)χδ(H)11\leq c(r,v_{1})\leq\chi_{\delta}(H)-1; otherwise, c(r,v2)=1c(r,v_{2})=1. Lastly, we color (u,v1)(u,v_{1}), (u,v2)(u,v_{2}) and (u,v3)(u,v_{3}) by χδ(H)+1\chi_{\delta}(H)+1, χδ(H)+2\chi_{\delta}(H)+2 and 11, respectively. This gives a proper coloring of (GP3)δ(G\square P_{3})_{\delta} with 2χδ(H)2\chi_{\delta}(H) colors. ∎

The sharpness of the bound in Theorem 15 will be shown in Theorem 17.

5. The δ\delta-chromatic numbers of the Cartesian products of some graphs

In this section, we give the exact values of the δ\delta-chromatic numbers of the Cartesian products of stars and paths.

Theorem 16.

χδ(S1,mS1,n)=mn\chi_{\delta}(S_{1,m}\square S_{1,n})=mn for m,n3m,n\geq 3.

Proof.

Let V(S1,n)={0,1,,k}V(S_{1,n})=\{0,1,\dots,k\} where dS1,k(0)=kd_{S_{1,k}}(0)=k for k=n,mk=n,m. An mnmn-coloring on (S1,mS1,n)δ(S_{1,m}\square S_{1,n})_{\delta} is

c(i,j)={j+1if i=0,(i1)n+jif 1im and 1jn,(i+1)nif 1i<m and j=0,n+2if i=m and j=0,c(i,j)=\begin{cases}j+1&\text{if }i=0,\\ (i-1)n+j&\text{if }1\leq i\leq m\text{ and }1\leq j\leq n,\\ (i+1)n&\text{if }1\leq i<m\text{ and }j=0,\\ n+2&\text{if }i=m\text{ and }j=0,\end{cases}

as shown in Fig. 1. In addition, the set {(i,j):1im,1jn}\{(i,j):1\leq i\leq m,1\leq j\leq n\} forms an mnmn-clique in (S1,mS1,n)δ(S_{1,m}\square S_{1,n})_{\delta}. ∎

112233\ddotsn+1n+12n2n1122\ddotsnn3n3nn+1n+1n+2n+2\ddots2n2n4n4n2n+12n+12n+22n+2\ddots3n3n\dotsn+2n+2mnn+1mn-n+1mnn+2mn-n+2\ddotsmnmn
Figure 1. A proper mnmn-coloring of (S1,mS1,n)δ(S_{1,m}\square S_{1,n})_{\delta}. The vertices of the same degree in S1,mS1,nS_{1,m}\square S_{1,n} are indicated by the same color and are pairwise adjacent in (S1,mS1,n)δ(S_{1,m}\square S_{1,n})_{\delta}. Each double line denotes the edges connecting the corresponding vertices between two copies of S1,nS_{1,n}. Note that the blue and the green will have the same degree when m=nm=n.
Theorem 17.

χδ(S1,mPn)=mn22\chi_{\delta}(S_{1,m}\square P_{n})=m\left\lceil\frac{n-2}{2}\right\rceil for m3m\geq 3 and n3n\geq 3.

Proof.

Let V(S1,m)={0,1,,m}V(S_{1,m})=\{0,1,\dots,m\} where dS1,m(0)=md_{S_{1,m}}(0)=m and Let V(Pn)={1,2,,n}V(P_{n})=\{1,2,\dots,n\} where dPn(1)=dPn(n)=1d_{P_{n}}(1)=d_{P_{n}}(n)=1.

When n=3n=3, Theorem 15 gives χδ(S1,mP3)2m\chi_{\delta}(S_{1,m}\square P_{3})\leq 2m. Since the set {(i,j)V((S1,mP3)δ):1im,j=1,3}\{(i,j)\in V((S_{1,m}\square P_{3})_{\delta}):1\leq i\leq m,j=1,3\} forms a 2m2m-clique in (S1,mP3)δ(S_{1,m}\square P_{3})_{\delta}, we get χδ(S1,mP3)=2m\chi_{\delta}(S_{1,m}\square P_{3})=2m.

When n=4n=4, Theorem 11 with G=P4G=P_{4} and H=S1,mH=S_{1,m} gives χδ(S1,mP4)=χδ(P4S1,m)2m\chi_{\delta}(S_{1,m}\square P_{4})=\chi_{\delta}(P_{4}\square S_{1,m})\leq 2m. The set {(i,j)V((S1,mP4)δ):1im,j=1,4}\{(i,j)\in V((S_{1,m}\square P_{4})_{\delta}):1\leq i\leq m,j=1,4\} forms a 2m2m-clique in (S1,mP4)δ(S_{1,m}\square P_{4})_{\delta}. Hence χδ(S1,mP4)=2m\chi_{\delta}(S_{1,m}\square P_{4})=2m.

When n5n\geq 5, we let k=n22k=\lceil\frac{n-2}{2}\rceil. A coloring is

c(i,j)={i+(k1)m0im and j=1,i1im and j=n,i+(j/21)m0im and 2jn1 where (i,j)(0,2),(0,3),kmi=0 and j=2,3,nc(i,j)=\begin{cases}i+(k-1)m&0\leq i\leq m\text{ and }j=1,\\ i&1\leq i\leq m\text{ and }j=n,\\ i+(\lfloor j/2\rfloor-1)m&0\leq i\leq m\text{ and }2\leq j\leq n-1\text{ where }(i,j)\neq(0,2),(0,3),\\ km&i=0\text{ and }j=2,3,n\\ \end{cases}

as shown in Fig. 2. We thus have a proper kmkm-coloring of (S1,mPn)δ(S_{1,m}\square P_{n})_{\delta}. In addition, the set {(i,j)V((S1,mPn)δ):1im and 2jn1 and j is even}\{(i,j)\in V((S_{1,m}\square P_{n})_{\delta}):1\leq i\leq m\text{ and }2\leq j\leq n-1\text{ and }j\text{ is even}\} forms a clique of size mn22m\left\lceil\frac{n-2}{2}\right\rceil in (S1,mPn)δ(S_{1,m}\square P_{n})_{\delta}. ∎

(k1)m(k-1)m(k1)m+1(k-1)m+1(k1)m+2(k-1)m+2\ddotskmkmkmkm1122\ddotsmmkmkm1122\ddotsmm(k1)m(k-1)m(k1)m+1(k-1)m+1(k1)m+2(k-1)m+2\ddotskmkmkmkm1122\ddotsmm
Figure 2. A proper kmkm-coloring of (S1,mPn)δ(S_{1,m}\square P_{n})_{\delta} where k=n22k=\lceil\frac{n-2}{2}\rceil. The vertices of the same degree in S1,mPnS_{1,m}\square P_{n} are indicated by the same color and are pairwise adjacent except for the pairs with a dashed line.

The following lemma is crucial for proving Theorem 19.

Lemma 18.

For n6n\geq 6 and k8k\geq 8, we have

2n22+2k22+1<(n2)(k2)2.2\left\lceil\frac{n-2}{2}\right\rceil+2\left\lceil\frac{k-2}{2}\right\rceil+1<\left\lceil\frac{(n-2)(k-2)}{2}\right\rceil.
Proof.

Suppose 2n22+2k22+1(n2)(k2)22\left\lceil\frac{n-2}{2}\right\rceil+2\left\lceil\frac{k-2}{2}\right\rceil+1\geq\left\lceil\frac{(n-2)(k-2)}{2}\right\rceil.

Case 1.

nn and kk are even.

We have

n+k3\displaystyle n+k-3 (n2)(k2)2,\displaystyle\geq\frac{(n-2)(k-2)}{2},
2n+2k6\displaystyle 2n+2k-6 nk2n2k+4.\displaystyle\geq nk-2n-2k+4.

Thus n4k10k4<6n\leq\frac{4k-10}{k-4}<6 when k8k\geq 8, which is a contradiction.

Case 2.

nn is odd and kk is even.

We have

n+k2\displaystyle n+k-2 (n1)(k2)2,\displaystyle\geq\frac{(n-1)(k-2)}{2},
2n+2k4\displaystyle 2n+2k-4 nk2nk+2.\displaystyle\geq nk-2n-k+2.

Thus n3k6k492n\leq\frac{3k-6}{k-4}\leq\frac{9}{2}, which is not possible when k8k\geq 8. The same argument can be applied when nn is even and kk is odd.

Case 3.

nn and kk are odd.

n+k1\displaystyle n+k-1 (n1)(k1)2,\displaystyle\geq\frac{(n-1)(k-1)}{2},
2n+2k4\displaystyle 2n+2k-4 nknk+1.\displaystyle\geq nk-n-k+1.

Thus n3k5k3195n\leq\frac{3k-5}{k-3}\leq\frac{19}{5}, which is not possible when k8k\geq 8.

Therefore 2n22+2k22+1<(n2)(k2)2.2\left\lceil\frac{n-2}{2}\right\rceil+2\left\lceil\frac{k-2}{2}\right\rceil+1<\left\lceil\frac{(n-2)(k-2)}{2}\right\rceil.

Theorem 19.

For 6nk6\leq n\leq k, we have

χδ(PnPk)=(n2)(k2)2.\chi_{\delta}(P_{n}\square P_{k})=\left\lceil\frac{(n-2)(k-2)}{2}\right\rceil.
Proof.

Let VdV_{d} be the set of vertices of degree dd in PnPkP_{n}\square P_{k}. The vertex set of PnPkP_{n}\square P_{k} can be partitioned into V2,V3V_{2},V_{3} and V4V_{4}. We note that V(PnPk)=V((PnPk)δ)=V(Pn)×V(Pk)V(P_{n}\square P_{k})=V((P_{n}\square P_{k})_{\delta})=V(P_{n})\times V(P_{k}). Let (i,j)V(PnPk)(i,j)\in V(P_{n}\square P_{k}) for i=1,,ni=1,\dots,n and j=1,,kj=1,\dots,k. We have that V3={(i,j):i=1,n and 2jk1}{(i,j):j=1,k and 2in1}V_{3}=\{(i,j):i=1,n\text{ and }2\leq j\leq k-1\}\cup\{(i,j):j=1,k\text{ and }2\leq i\leq n-1\} and V4={(i,j):2in1 and 2jk1}V_{4}=\{(i,j):2\leq i\leq n-1\text{ and }2\leq j\leq k-1\}. Thus |V2|=4|V_{2}|=4, |V3|=2(n+k4)|V_{3}|=2(n+k-4) and |V4|=(n2)(k2)|V_{4}|=(n-2)(k-2). The vertices (i,j)(i,j) and (i,j)(i^{\prime},j^{\prime}) are adjacent in PnPkP_{n}\square P_{k} if and only if |ii|+|jj|=1|i-i^{\prime}|+|j-j^{\prime}|=1. Thus

  • if dPnPk(i,j)=dPnPk(i,j)d_{P_{n}\square P_{k}}(i,j)=d_{P_{n}\square P_{k}}(i^{\prime},j^{\prime}), then the vertices (i,j)(i,j) and (i,j)(i^{\prime},j^{\prime}) are adjacent in (PnPk)δ(P_{n}\square P_{k})_{\delta} if and only if |ii|+|jj|2|i-i^{\prime}|+|j-j^{\prime}|\geq 2,

  • if dPnPk(i,j)dPnPk(i,j)d_{P_{n}\square P_{k}}(i,j)\neq d_{P_{n}\square P_{k}}(i^{\prime},j^{\prime}), then the vertices (i,j)(i,j) and (i,j)(i^{\prime},j^{\prime}) are adjacent in (PnPk)δ(P_{n}\square P_{k})_{\delta} if and only if |ii|+|jj|=1|i-i^{\prime}|+|j-j^{\prime}|=1.

Since each pair of vertices (i,j),(i,j)V4(i,j),(i^{\prime},j^{\prime})\in V_{4} with |ii|+|jj|2|i-i^{\prime}|+|j-j^{\prime}|\geq 2 are adjacent, it follows that

χδ(PnPk)ω((PnPk)δ)(n2)(k2)2.\chi_{\delta}(P_{n}\square P_{k})\geq\omega((P_{n}\square P_{k})_{\delta})\geq\left\lceil\frac{(n-2)(k-2)}{2}\right\rceil.

We note that

(n2)(k2)2={(n2)(k2)2 if n or k is even,k22(n2)+n22 if n and k are odd.\left\lceil\frac{(n-2)(k-2)}{2}\right\rceil=\begin{cases}\frac{(n-2)(k-2)}{2}&\text{ if }n\text{ or }k\text{ is even},\\ \left\lfloor\frac{k-2}{2}\right\rfloor(n-2)+\left\lceil\frac{n-2}{2}\right\rceil&\text{ if }n\text{ and }k\text{ are odd}.\end{cases}

We color V4V_{4} by a coloring c0c_{0} defined by

c0(i,j)={(i2)k22+j22 if i=2,,n1 and j=2,,2k22+1,(n2)k22+i22 if k is odd and j=k1,i=1,,n2.c_{0}(i,j)=\begin{cases}(i-2)\left\lfloor\frac{k-2}{2}\right\rfloor+\left\lfloor\frac{j-2}{2}\right\rfloor&\text{ if }i=2,\dots,n-1\text{ and }j=2,\dots,2\left\lfloor\frac{k-2}{2}\right\rfloor+1,\\ (n-2)\left\lfloor\frac{k-2}{2}\right\rfloor+\left\lfloor\frac{i-2}{2}\right\rfloor&\text{ if }k\text{ is odd and }j=k-1,i=1,\dots,n-2.\end{cases}

The coloring c0c_{0} uses (n2)(k2)2\left\lceil\frac{(n-2)(k-2)}{2}\right\rceil colors. Since the coloring in V4V_{4} has at most 2 vertices with the same color and they are adjacent in PnPkP_{n}\square P_{k}, which are not adjacent in (PnPk)δ(P_{n}\square P_{k})_{\delta}, the coloring c0c_{0} on (PnPk)δ[V4](P_{n}\square P_{k})_{\delta}[V_{4}] is proper. The case 6k76\leq k\leq 7 can be verified. Now, we suppose that k8k\geq 8. We color V3V_{3} using 2n22+2k222\left\lceil\frac{n-2}{2}\right\rceil+2\left\lceil\frac{k-2}{2}\right\rceil colors. We color the vertices V3V_{3} in pair of consecutive vertices (except possibly the last one in a block) clockwise starting from location (1,2)(1,2) to (2,1)(2,1). Each color in V3V_{3} needs to avoid at most 2n22+2k2212\left\lceil\frac{n-2}{2}\right\rceil+2\left\lceil\frac{k-2}{2}\right\rceil-1 colors of the other vertices in V3V_{3} and two neighbors per each color in V4V_{4}, i.e., we have to avoid 2n22+2k22+12\left\lceil\frac{n-2}{2}\right\rceil+2\left\lceil\frac{k-2}{2}\right\rceil+1 colors. By Lemma 18, there is a remaining color in V4V_{4} that is available to assign to the considered vertex. Since each vertex in V2V_{2} has degree 5 in (PnPk)δ(P_{n}\square P_{k})_{\delta} and (n2)(k2)2>5\left\lceil\frac{(n-2)(k-2)}{2}\right\rceil>5, we can color V2V_{2}. This completes the proof. ∎

655330764201764201975318975318699880944228
Figure 3. A proper 10-coloring of (P6P7)δ(P_{6}\square P_{7})_{\delta}. The vertices in V2V_{2}, V3V_{3} and V4V_{4} are shown in red, blue and black, respectively. The vertices in each ViV_{i} for i=2,3,4i=2,3,4 are pairwise adjacent in (P6P7)δ(P_{6}\square P_{7})_{\delta} except for the pairs with a dashed line.

6. Conclusion

We give a structure of (G1Gk)δ(G_{1}\square\cdots\square G_{k})_{\delta} associated with (G1)δ(Gk)δ(G_{1})_{\delta}\square\cdots\square(G_{k})_{\delta} and the necessary and sufficient condition that both graphs are equal. We also give sharp bounds on the δ\delta-chromatic number of GHG\square H with a class of graphs achieving such bound. The δ\delta-chromatic number of the Cartesian product of several classes of well-known graphs are also given.

Acknowledgement

The authors thank Rasimate Maungchang for his invaluable comments. This research project was financially supported by Mahasarakham University.

References

  • [1] R. Balakrishnan, S.F. Raj, and T. Kavaskar. bb-chromatic number of cartesian product of some families of graphs. Graphs and Combinatorics, page 511–520, 2014.
  • [2] M. Borowiecki, S. Jendrol’, D. Král’, and J. Miškuf. List coloring of cartesian products of graphs. Discrete Mathematics, 306(16):1955–1958, 2006.
  • [3] B. Brešar, S. Klavžar, and D. F. Rall. On the packing chromatic number of cartesian products, hexagonal lattice, and trees. Discrete Applied Mathematics, 155(17):2303–2311, 2007.
  • [4] C. Guo and M. Newman. On the bb-chromatic number of cartesian products. Discrete Applied Mathematics, 239:82–93, 2018.
  • [5] A. Pai, H. A Rao, S. D’Souza, P. G. Bhat, and S. Upadhyay. δ\delta-complement of a graph. Mathematics, 10(8):1203, 2022.
  • [6] G. Sabidussi. Graphs with given group and given graph-theoretical properties. Canadian Journal of Mathematics, 9:515–525, 1957.
  • [7] P. Vichitkunakorn, R. Maungchang, and W. Tangjai. On nordhaus-gaddum type relations of δ\delta-complement graphs. Heliyon, 9(6):e16630, 2023.