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

Strong odd coloring of sparse graphs

Hyemin Kwon, and Boram Park Department of Mathematics, Ajou University, Suwon-si, Gyeonggi-do, Republic of Korea. [email protected] Department of Mathematics, Ajou University, Suwon-si, Gyeonggi-do, Republic of Korea. [email protected]
Abstract

An odd coloring of a graph GG is a proper coloring of GG such that for every non-isolated vertex vv, there is a color appearing an odd number of times in NG(v)N_{G}(v). Odd coloring of graphs was studied intensively in recent few years. In this paper, we introduce the notion of a strong odd coloring, as not only a strengthened version of odd coloring, but also a relaxation of square coloring. A strong odd coloring of a graph GG is a proper coloring of GG such that for every non-isolated vertex vv, if a color appears in NG(v)N_{G}(v), then it appears an odd number of times in NG(v)N_{G}(v). We denote by χso(G)\chi_{so}(G) the smallest integer kk such that GG admits a strong odd coloring with kk colors. We prove that if GG is a graph with mad(G)207mad(G)\leq\frac{20}{7}, then χso(G)Δ(G)+4\chi_{so}(G)\leq\Delta(G)+4, and the bound is tight. We also prove that if GG is a graph with mad(G)3011mad(G)\leq\frac{30}{11} and Δ(G)4\Delta(G)\geq 4, then χso(G)Δ(G)+3\chi_{so}(G)\leq\Delta(G)+3.

1 Introduction

All graphs in this paper are finite. Let GG be a graph. For a vertex vv, let degG(v)\deg_{G}(v) be the degree of vv, NG(v)N_{G}(v) be the neighborhood of vv, and NG[v]:=NG(v){v}N_{G}[v]:=N_{G}(v)\cup\{v\}. Also, Δ(G)\Delta(G) is the maximum degree of GG. The girth g(G)g(G) of GG is the length of a shortest cycle in GG and the maximum average degree mad(G)\operatorname{mad}(G) of GG is the maximum of 2|E(H)||V(H)|\frac{2|E(H)|}{|V(H)|} over all non-empty subgraphs HH of GG.

For a positive integer kk, a proper kk-coloring of a graph is a function from the vertex set to {1,,k}\{1,\cdots,k\} such that adjacent vertices receive different colors. The minimum kk for which a graph GG has a proper kk-coloring is the chromatic number χ(G)\chi(G) of GG. A square coloring of a graph GG is a proper coloring of G2G^{2}, where G2G^{2} is the graph obtained from GG by adding edges joining vertices at distance at most 22 in GG. In 1977, Wegner conjectured the bound for χ(G2)\chi(G^{2}) when GG is planar.

Conjecture 1.1 ([24]).

Let GG be a planar graph. Then

χ(G2){7if Δ(G)=3,Δ(G)+5if 4Δ(G)7,3Δ(G)2+1if Δ(G)8.\chi(G^{2})\leq\begin{cases}7&\text{if $\Delta(G)=3$,}\\ \Delta(G)+5&\text{if $4\leq\Delta(G)\leq 7$,}\\ \left\lfloor\frac{3\Delta(G)}{2}\right\rfloor+1&\text{if $\Delta(G)\geq 8$.}\end{cases}

Thomassen[22] and Hartke et al.[17] independently proved Conjecture 1.1 when Δ(G)=3\Delta(G)=3, and it has attracted many researchers to study square coloring problems (see [1, 16, 18, 8, 5, 4, 13, 14, 12, 3] and Table 1 for some results on planar graphs).

Δ2\Delta\geq 2 Δ4\Delta\geq 4 Δ6\Delta\geq 6 Δ9\Delta\geq 9 Δ15\Delta\geq 15 Δ22\Delta\geq 22 Δ339\Delta\geq 339
g5g\geq 5 Δ+5\Delta+5[4] Δ+4\Delta+4[13] Δ+3\Delta+3[14]
g6g\geq 6 Δ+5\Delta+5[5] Δ+4\Delta+4[12] Δ+3\Delta+3[3]
g7g\geq 7 Δ+3\Delta+3[18]
g8g\geq 8 Δ+3\Delta+3[18]
Table 1: Some known bounds on χ(G2)\chi(G^{2}) for planar graphs GG with Δ(G)=Δ\Delta(G)=\Delta and g(G)=gg(G)=g.

The concept of odd coloring was introduced by Petruševski and Škrekovski [21] as not only a strengthening of proper coloring and but also a weakening of square coloring. For a positive integer kk, an odd kk-coloring of a graph GG is a proper kk-coloring of GG such that every non-isolated vertex vv has a color appearing an odd number of times in NG(v)N_{G}(v). The minimum kk for which GG has an odd kk-coloring is the odd chromatic number of GG, denoted by χo(G)\chi_{o}(G). Ever since the first paper by Petruševski and Škrekovski [21] appeared, there have been numerous papers [6, 7, 10, 15, 19, 20, 23, 9] studying various aspects of this new coloring concept across several graph classes. Recently, Dai, Ouyang, and Pirot [11] introduced a strengthening of odd coloring. For positive integers hh and kk, an hh-odd kk-coloring is a proper kk-coloring of a graph GG such that every vertex vv has min{degG(v),h}\min\{\deg_{G}(v),h\} colors each of which appears an odd number of times in NG(v)N_{G}(v). The χoh(G)\chi_{o}^{h}(G) is the minimum kk for which GG has an hh-odd kk-coloring. It is clear that χo(G)χoh(G)χ(G2)\chi_{o}(G)\leq\chi_{o}^{h}(G)\leq\chi(G^{2}) by definitions.

In this paper, we introduce a new concept that is a new generalization of odd coloring. For a positive integer kk, a strong odd kk-coloring (an SO kk-coloring for short) of a graph GG is a proper kk-coloring of GG such that for every non-isolated vertex vv, if a color appears in NG(v)N_{G}(v), then it appears an odd number of times in NG(v)N_{G}(v). The strong odd chromatic number of a graph GG, denoted by χso(G)\chi_{so}(G), is the minimum kk such that GG has an SO kk-coloring. See Figure 1 for an example. Like as odd coloring, strong odd coloring does not have hereditary, that is, there are graphs HH and GG such that HGH\subset G and χso(H)>χso(G)\chi_{so}(H)>\chi_{so}(G).

Refer to caption
Figure 1: A Petersen graph GG with χso(G)=6\chi_{so}(G)=6.

Every graph has a strong odd coloring since a square coloring is also a strong odd coloring. Moreover, it is clear from the definitions that for every graph GG,

χo(G)χso(G)χ(G2)(Δ(G))2+1.\chi_{o}(G)\leq\chi_{so}(G)\leq\chi(G^{2})\leq(\Delta(G))^{2}+1.

When GG is claw-free, that is, GG has no K1,3K_{1,3} as an induced subgraph, it holds that χso(G)=χ(G2)\chi_{so}(G)=\chi(G^{2}). However, χ(G2)χso(G)\chi(G^{2})-\chi_{so}(G) could be arbitrary large. For example, if G=Km,nG=K_{m,n}, then χ(G2)=m+n\chi(G^{2})=m+n and χso(G)4\chi_{so}(G)\leq 4. We remark that whereas χo(G)2Δ(G)+1\chi_{o}(G)\leq 2\Delta(G)+1 by [6], a linear bound of χso(G)\chi_{so}(G) in terms of Δ(G)\Delta(G) cannot be obtained in general. For example, if G=KnKnG=K_{n}\square K_{n}, then Δ(G)=2n2\Delta(G)=2n-2, the diameter of GG is two, and GG is claw-free, and thus χso(G)=χ(G2)=|V(G)|=n2=(Δ(G)+2)24\chi_{so}(G)=\chi(G^{2})=|V(G)|=n^{2}=\frac{(\Delta(G)+2)^{2}}{4}.

Refer to caption
Figure 2: A planar subcubic graph GG with χ(G2)=χso(G)=7\chi(G^{2})=\chi_{so}(G)=7

Regarding to subcubic planar graphs GG, we have χso(G)χ(G2)7\chi_{so}(G)\leq\chi(G^{2})\leq 7, the bound is tight since there is a subcubic planar graph GG such that χso(G)=χ(G2)=7\chi_{so}(G)=\chi(G^{2})=7. For a graph in Figure 2, it has maximum average degree 207\frac{20}{7}, and which is a tight example in the following our main result.

Theorem 1.2.

If GG is a graph with mad(G)207mad(G)\leq\frac{20}{7}, then χso(G)Δ(G)+4\chi_{so}(G)\leq\Delta(G)+4.

In [18], it was shown that χ(G2)Δ(G)+3\chi(G^{2})\leq\Delta(G)+3 if GG is a graph with Δ(G)4\Delta(G)\geq 4 and mad(G)83\operatorname{mad}(G)\leq\frac{8}{3} and therefore, we have the same bound for χso(G)\chi_{so}(G). We improve this as follows.

Theorem 1.3.

If GG is a graph with Δ(G)4\Delta(G)\geq 4 and mad(G)3011mad(G)\leq\frac{30}{11}, then χso(G)Δ(G)+3\chi_{so}(G)\leq\Delta(G)+3.

When Δ(G)=3\Delta(G)=3, we obtain the same result as long as GG has no 44-cycle as a subgraph.

Theorem 1.4.

If GG is a C4C_{4}-free subcubic graph with mad(G)3011mad(G)\leq\frac{30}{11}, then χso(G)6\chi_{so}(G)\leq 6.

Since (mad(G)2)(g(G)2)<4(\operatorname{mad}(G)-2)(g(G)-2)<4 for every planar graph GG, we obtain the following corollary from Theorems 1.2, 1.3, and 1.4. See also Table 1 to compare the results on square coloring of planar graphs with large maximum degree and large girth.

Corollary 1.5.

Let GG be a planar graph.

  • (i)

    If g(G)7g(G)\geq 7, then χso(G)Δ(G)+4\chi_{so}(G)\leq\Delta(G)+4.

  • (ii)

    If g(G)8g(G)\geq 8, then χso(G)Δ(G)+3\chi_{so}(G)\leq\Delta(G)+3.

The paper is organized as follows. In Section 2, we collect some basic observations on χso(G)\chi_{so}(G). Section 3 proves Theorem 1.2, and Section 4 proves both Theorems 1.3 and 1.4. We remain some remarks in Section 5.

2 Preliminaries

For SV(G)S\subset V(G), let GSG-S denote the graph obtained from GG by deleting the vertices in SS. If S={x}S=\{x\}, then denote GSG-S by GxG-x. For a positive integer kk, we use kk-vertex (resp. k+k^{+}-vertex and kk^{-}-vertex) to denote a vertex of degree kk (resp. at least kk and at most kk). We also use kk-neighbor (resp. k+k^{+}-neighbor and kk^{-}-neighbor) of a vertex vv to denote a kk-vertex (resp. k+k^{+}-vertex and kk^{-}-vertex) that is a neighbor of vv. For positive integers kk and dd, we call a kk-vertex with at least dd 22-neighbors a kdk_{d}-vertex. We also use kdk_{d}-neighbor of a vertex vv to call a neighbor of vv that is a kdk_{d}-vertex.

For an integer d2d\geq 2, let A1,,AdA_{1},\ldots,A_{d} be nonempty sets. A sequence (c1,,cd)(c_{1},\ldots,c_{d}) is called a system of odd representative for (A1,,Ad)(A_{1},\ldots,A_{d}) if it satisfies the following:

  • (i)

    ciAic_{i}\in A_{i} for each 1id1\leq i\leq d, and

  • (ii)

    if c=ctc=c_{t} for some t{1,,d}t\in\{1,\ldots,d\}, then cc appears an odd number of times in (c1,,cd)(c_{1},\ldots,c_{d}).

Lemma 2.1.

For an integer d2d\geq 2, let A1,,Ad1A_{1},\ldots,A_{d-1} be nonempty subsets of size at least 22. Then for every αAd\alpha\in A_{d}, there is a system of odd representative (β1,,βd1,α)(\beta_{1},\ldots,\beta_{d-1},\alpha) for (A1,,Ad)(A_{1},\ldots,A_{d}).

Proof.

It is enough to assume that |Ai|=2|A_{i}|=2 for each i{1,,d1}i\in\{1,\ldots,d-1\}. Fix αAd\alpha\in A_{d}. Then we define a loopless multigraph GG with V(G)=i=1dAiV(G)=\cup_{i=1}^{d}A_{i} and E(G)={A1,,Ad1}E(G)=\{A_{1},\ldots,A_{d-1}\}. Note that it is sufficient to find an orientation DD of GG such that degD(α)\deg_{D}^{-}(\alpha) is even and degD(x)\deg_{D}^{-}(x) is either odd or 0 for every xV(G){α}x\in V(G)\setminus\{\alpha\}, by taking the head of the arc Ai\overrightarrow{A_{i}} of DD as a representative of AiA_{i} for each i{1,,d1}i\in\{1,\ldots,d-1\}.

To find such orientation DD of GG, we will find an ordering v1,,v|V(G)|v_{1},\ldots,v_{|V(G)|} of the vertices of GG as follows. Let v1=αv_{1}=\alpha. If NG(vi){v1,,vi}=N_{G}(v_{i})-\{v_{1},\ldots,v_{i}\}=\emptyset, then take any vertex in V(G){v1,,vi}V(G)-\{v_{1},\ldots,v_{i}\} as vi+1v_{i+1}. If NG(vi){v1,,vi}N_{G}(v_{i})-\{v_{1},\ldots,v_{i}\}\neq\emptyset, then we take vi+1NG(vi){v1,,vi}v_{i+1}\in N_{G}(v_{i})-\{v_{1},\ldots,v_{i}\} and then choose an edge eie_{i} joining viv_{i} and vi+1v_{i+1}. Let LL be the set of such chosen edges eie_{i}, and note that LL induces a linear forest of GG.

For simplicity, let GiG_{i} be the spanning subgraph of GG induced by E(G){ejLj>i}E(G)\setminus\{e_{j}\in L\mid j>i\} for each i0i\geq 0. We will define an orientation DiD_{i} of GiG_{i} for each ii, as follows. First, we define an orientation D0D_{0} of G0G_{0} by A(D0)={(vi,vj)i>j,vivjE(G0)}A(D_{0})=\{(v_{i},v_{j})\mid i>j,v_{i}v_{j}\in E(G_{0})\}. Suppose that Di1D_{i-1} is defined. If Gi=Gi1G_{i}=G_{i-1}, then we define Di=Di1D_{i}=D_{i-1}. If GiGi1G_{i}\neq G_{i-1}, then E(Gi)=E(Gi1){ei}E(G_{i})=E(G_{i-1})\cup\{e_{i}\}, and so we define DiD_{i} so that A(Di)=A(Di1){ei}A(D_{i})=A(D_{i-1})\cup\{\overrightarrow{e_{i}}\}, where

ei={(vi,vi+1) if degDi1(vi)min{i1,1}(mod2),(vi+1,vi) otherwise.\overrightarrow{e_{i}}=\begin{cases}(v_{i},v_{i+1})&\text{ if }\deg_{D_{i-1}}^{-}(v_{i})\equiv\min\{i-1,1\}\pmod{2},\\ (v_{i+1},v_{i})&\text{ otherwise.}\end{cases}

It remains to show D:=D|V(G)|1D:=D_{|V(G)|-1} is a desired orientation, that is, satisfying the condition that degD(v1)\deg_{D}^{-}(v_{1}) is even and degD(vi)\deg_{D}^{-}(v_{i}) is either odd or 0 for every i2i\geq 2. Take a vertex viv_{i}. Suppose that the edge eie_{i} exists. By the definition of DiD_{i},

degD(vi)=degDi(vi)={degDi1(vi) if degDi1(vi)min{i1,1}(mod2),degDi1(vi)+1 otherwise.\deg_{D}^{-}(v_{i})=\deg_{D_{i}}^{-}(v_{i})=\begin{cases}\deg_{D_{i-1}}^{-}(v_{i})&\text{ if }\deg_{D_{i-1}}^{-}(v_{i})\equiv\min\{i-1,1\}\pmod{2},\\ \deg_{D_{i-1}}^{-}(v_{i})+1&\text{ otherwise.}\end{cases}

Hence it satisfies the condition. Suppose that the edge eie_{i} does not exist. If i=1i=1, then degG(v1)=0\deg_{G}(v_{1})=0 and so degD(v1)=0\deg_{D}^{-}(v_{1})=0. If i>1i>1, then NG(vi){v1,,vi}=N_{G}(v_{i})-\{v_{1},\ldots,v_{i}\}=\emptyset, and so it always holds that degD(vi){0,1}\deg_{D}^{-}(v_{i})\in\{0,1\}. Hence, it completes the proof. ∎

Throughout this paper, we use 𝒞\mathcal{C} to denote the set of all kk colors, when we consider a kk-coloring of a graph GG. For a (partial) coloring φ\varphi of a graph GG, let φ(X)={φ(v)vX}\varphi(X)=\{\varphi(v)\mid v\in X\} for XV(G)X\subset V(G).

In each proof of following lemmas, we will start a proof with defining a nonempty subset SV(G)S\subset V(G). We always let G=GSG^{\prime}=G-S, G′′=G2SG^{\prime\prime}=G^{2}-S, and φ\varphi be an SO (Δ(G)+c)(\Delta(G)+c)-coloring of GG^{\prime}, if it exists. We also define

Aφ(v)=𝒞φ(NG′′(v)).A_{\varphi}(v)=\mathcal{C}\setminus\varphi(N_{G^{\prime\prime}}(v)).

We call an element of Aφ(v)A_{\varphi}(v) an available color of vv under φ\varphi. Note that

|Aφ(v)|\displaystyle|A_{\varphi}(v)| =\displaystyle= |𝒞||φ(NG′′(v))|Δ(G)+cdegG′′(v).\displaystyle|\mathcal{C}|-|\varphi(N_{G^{\prime\prime}}(v))|\geq\Delta(G)+c-\deg_{G^{\prime\prime}}(v). (2.1)

When we color the vertices of SSS\setminus S^{*} for some SSS^{*}\subset S, we often use the same symbol φ\varphi to denote the resulting coloring, that is a coloring of GSG-S^{*}.

Lemma 2.2.

For an integer c3c\geq 3, let GG be a minimal graph with respect to |V(G)||V(G)| such that GG has no SO (Δ+c)(\Delta+c)-coloring, where Δ:=Δ(G)\Delta:=\Delta(G). For a 22-vertex xx, there is no SO (Δ+c)(\Delta+c)-coloring φ\varphi of GxG-x such that there is an available color of xx under φ\varphi and the neighbors of xx get distinct colors.

Proof.

Suppose that such coloring φ\varphi exists. Then by coloring xx with a color in Aφ(x)A_{\varphi}(x), we obtain an SO (Δ+c)(\Delta+c)-coloring of G. It is a contradiction to the fact that GG has no SO (Δ+c)(\Delta+c)-coloring. ∎

In all figures, we represent a vertex that has all incident edges in the figure as a filled vertex, and a hollow vertex may not have all incident edges in the figure.

Lemma 2.3.

For an integer c3c\geq 3, let GG be a minimal graph with respect to |V(G)||V(G)| such that GG has no SO (Δ+c)(\Delta+c)-coloring, where Δ:=Δ(G)\Delta:=\Delta(G). Then the following do not appear in GG:

  1. (i)

    a 11^{-}-vertex,

  2. (ii)

    a triangle v1v2v3v1v_{1}v_{2}v_{3}v_{1} such that v2v_{2} is a (c+1)(c+1)^{-}-vertex and v3v_{3} is a 22-vertex,

  3. (iii)

    two vertices with three common 22-neighbors,

  4. (iv)

    a dd1d_{d-1}-vertex with only (Δ+cd)(\Delta+c-d)^{-}-neighbors for 2dΔ2\leq d\leq\Delta.

Refer to caption
Figure 3: Illustrations of Lemma 2.3.
Proof.

In each proof, we consider an SO (Δ+c)(\Delta+c)-coloring φ\varphi of G=GSG^{\prime}=G-S for some SV(G)\emptyset\neq S\subset V(G), and then we finish the proof by coming up with an SO (Δ+c)(\Delta+c)-coloring of GG, which is a contradiction. See Figure 3.

(i) It is clear that GG has no isolated vertex. Suppose to the contrary that GG has a 11-vertex uu with neighbor vv. Let S={u}S=\{u\}. Since |Aφ(v)|c|A_{\varphi}(v)|\geq c by (2.1), we color uu with a color in Aφ(v)A_{\varphi}(v), which gives an SO (Δ+c)(\Delta+c)-coloring of GG.

(ii) Suppose to the contrary that GG has a triangle v1v2v3v1v_{1}v_{2}v_{3}v_{1} such that v2v_{2} is a (c+1)(c+1)^{-}-vertex and v3v_{3} is a 22-vertex. For an SO (Δ+c)(\Delta+c)-coloring φ\varphi of Gv3G-v_{3}, |Aφ(v3)|Δ+c(Δ+c1)1|A_{\varphi}(v_{3})|\geq\Delta+c-(\Delta+c-1)\geq 1 by (2.1). Since v1v2E(G)v_{1}v_{2}\in E(G^{\prime}), φ(v1)φ(v2)\varphi(v_{1})\neq\varphi(v_{2}). It is a contradiction by Lemma 2.2.

(iii) Suppose to the contrary that GG has two vertices xx and yy with three common 22-neighbors v1v_{1}, v2v_{2}, and v3v_{3}. Let S={v2,v3}S=\{v_{2},v_{3}\}. Then color v2v_{2} and v3v_{3} with the same color of v1v_{1} to obtain an SO (Δ+c)(\Delta+c)-coloring of GG.

(iv) Suppose to the contrary that GG has a dd1d_{d-1}-vertex uu with only (Δ+cd)(\Delta+c-d)^{-}-neighbors for some 2dΔ2\leq d\leq\Delta. Let v1,,vd1v_{1},\ldots,v_{d-1} be 22-neighbors of uu and vv^{\prime} be the other neighbor of uu. For each i{1,,d1}i\in\{1,\ldots,d-1\}, let wiw_{i} be the neighbor of viv_{i} other than uu. Note that viwjv_{i}\neq w_{j} for each i,j{1,,d1}i,j\in\{1,\ldots,d-1\} by (ii). Let S={u,v1,,vd1}S=\{u,v_{1},\ldots,v_{d-1}\}. Note that |Aφ(u)|Δ+c(Δ+d1)1|A_{\varphi}(u)|\geq\Delta+c-(\Delta+d-1)\geq 1, and Aφ(vi)Δ+c(Δ+1)=c12A_{\varphi}(v_{i})\geq\Delta+c-(\Delta+1)=c-1\geq 2 for each i{1,,d1}i\in\{1,\ldots,d-1\} by (2.1). We color uu with a color γ\gamma in Aφ(u)A_{\varphi}(u). Let XX be the set of all pairs (i,j)(i,j) such that wi=wjw_{i}=w_{j} and i<ji<j. Note that by (iii), each i{1,,d1}i\in\{1,\ldots,d-1\} appears at most once as an entry of an element of XX. Moreover, when (i,j)X(i,j)\in X, it holds that Aφ(vi)=Aφ(vj)A_{\varphi}(v_{i})=A_{\varphi}(v_{j}) and |Aφ(vi)|=|Aφ(vj)|c|A_{\varphi}(v_{i})|=|A_{\varphi}(v_{j})|\geq c by (2.1).

If c=3c=3 and XX\neq\emptyset, then X={(1,2)}X=\{(1,2)\}, d=3d=3, and so we can color v1v_{1}, v2v_{2} with distinct colors in Aφ(v1){γ}A_{\varphi}(v_{1})\setminus\{\gamma\} to obtain an SO (Δ+c)(\Delta+c)-coloring of GG. Suppose that c4c\geq 4 or X=X=\emptyset. Let Ai:=(Aφ(vi){φ(v)}){γ}A_{i}:=(A_{\varphi}(v_{i})\cup\{\varphi(v^{\prime})\})\setminus\{\gamma\} for each i{1,,d1}i\in\{1,\ldots,d-1\}, and Ad:={φ(v)}A_{d}:=\{\varphi(v^{\prime})\}. Since vv^{\prime} is a neighbor of viv_{i} in G2G^{2}, φ(v)Aφ(vi)\varphi(v^{\prime})\not\in A_{\varphi}(v_{i}). It follows that |Ai|=|Aφ(vi)|2|A_{i}|=|A_{\varphi}(v_{i})|\geq 2 for each i{1,,d1}i\in\{1,\ldots,d-1\}. If (i,j)X(i,j)\in X, then XX\neq\emptyset and so |Ai|=|Aφ(vi)|c4|A_{i}|=|A_{\varphi}(v_{i})|\geq c\geq 4, and then we redefine Ai:=AiA_{i}:=A^{\prime}_{i} and Aj:=AjA_{j}:=A^{\prime}_{j} for some disjoint subsets Ai,AjA^{\prime}_{i},A^{\prime}_{j} of AiA_{i} of size two. By applying Lemma 2.1, there is a system of odd representative (β1,,βd1,α)(\beta_{1},\ldots,\beta_{d-1},\alpha) for (A1,,Ad1,Ad)(A_{1},\ldots,A_{d-1},A_{d}), where α=φ(v)\alpha=\varphi(v^{\prime}). Then, coloring viv_{i} with βi\beta_{i} results in an SO (Δ+c)(\Delta+c)-coloring of GG. ∎

We finish the section by stating one famous theorem in graph coloring, called Brooks’ Theorem.

Theorem 2.4 ([2]).

For a graph GG, χ(G)Δ(G)+1\chi(G)\leq\Delta(G)+1. If GG has no component that is a complete graph or an odd cycle, then χ(G)Δ(G)\chi(G)\leq\Delta(G).

3 Strong odd (Δ+4)(\Delta+4)-coloring

In this section, we prove Theorem 1.2. Let GG be a minimal counterexample to Theorem 1.2. In each proof in the following, we always start with defining a nonempty set SV(G)S\subset V(G), and φ\varphi is an SO (Δ+4)(\Delta+{4})-coloring of G=GSG^{\prime}=G-S. We end up with an SO (Δ+4)(\Delta+{4})-coloring of GG.

The following is a list of reducible configurations, structures that never appear in GG. The configurations [C1]-[C6] are utilized in the final step to reach a contradiction.

  1. [C1]

    A 11^{-}-vertex.

  2. [C2]

    A dd1d_{d-1}-vertex for all 2d42\leq d\leq 4.

  3. [C3]

    A ddd_{d}-vertex for all d5d\geq 5.

  4. [C4]

    Two adjacent 313_{1}-vertices.

  5. [C5]

    A 33-vertex with two 313_{1}-neighbors.

  6. [C6]

    A 424_{2}-vertex with a 313_{1}-neighbor.

Note that [C1]-[C3] do not appear in GG by Lemma 2.3 (i) and (iv).

Suppose that a 22-vertex xx has a 33-neighbor, and φ\varphi is an SO (Δ+c)(\Delta+c)-coloring of GxG-x. Since |𝒞|=Δ+4|\mathcal{C}|=\Delta+4, it holds that Aφ(x)A_{\varphi}(x)\neq\emptyset by (2.1). Thus by Lemma 2.2, the colors of the neighbors of xx cannot be distinct colors. We often omit this explanation when we apply Lemma 2.2.

Lemma 3.1.

The graph GG has no two adjacent 313_{1}-vertices. [C4]

Refer to caption
Figure 4: An illustration of Lemma 3.1.
Proof.

Suppose to the contrary that GG has adjacent two 313_{1}-vertices u1u_{1} and u2u_{2}. For each i{1,2}i\in\{1,2\}, let viv_{i} be the 22-neighbor of uiu_{i}, uiu^{\prime}_{i} be the neighbor of uiu_{i} other than viv_{i} and u3iu_{3-i}, and wiw_{i} be the neighbor of viv_{i} other than uiu_{i}. See Figure 4. Note that v1v2v_{1}\neq v_{2} by Lemma 2.3 (ii). Thus u1u_{1}, u2u_{2}, v1v_{1}, and v2v_{2} are distinct, and we let S={u1,u2,v1,v2}S=\{u_{1},u_{2},v_{1},v_{2}\}. In addition, wiu3iw_{i}\neq u_{3-i} (equivalently, u3iviu^{\prime}_{3-i}\neq v_{i}) and wiv3iw_{i}\neq v_{3-i} for each i{1,2}i\in\{1,2\} by [C2]. Thus S{u1,u2,w1,w2}=S\cap\{u^{\prime}_{1},u^{\prime}_{2},w_{1},w_{2}\}=\emptyset.

Note that |Aφ(ui)|2|A_{\varphi}(u_{i})|\geq 2 and |Aφ(vi)|3|A_{\varphi}(v_{i})|\geq 3 for each i{1,2}i\in\{1,2\} by (2.1). We color u1u_{1}, u2u_{2}, v1v_{1} with a color in Aφ(u1)A_{\varphi}(u_{1}), Aφ(u2)A_{\varphi}(u_{2}), Aφ(v1)A_{\varphi}(v_{1}), respectively, so that colors of u1u_{1}, u2u_{2}, v1v_{1} are distinct. Then the resulting coloring is an SO (Δ+4)(\Delta+4)-coloring of Gv2G-v_{2}, which contradicts to Lemma 2.2. ∎

Lemma 3.2.

The graph GG has no path v1v2v3v4v_{1}v_{2}v_{3}v_{4} such that v1v_{1} is a 22-vertex, viv_{i} is a 33-vertex for each i{2,3,4}i\in\{2,3,4\}, where v4v_{4} is a common neighbor of v3v_{3} and vjv_{j} for some j{1,2}j\in\{1,2\}.

Refer to caption
Figure 5: Illustrations of Lemma 3.2.
Proof.

Suppose to the contrary that GG has a path path v1v2v3v4v_{1}v_{2}v_{3}v_{4} such that v1v_{1} is a 22-vertex and viv_{i} is a 33-vertex for each i{2,3,4}i\in\{2,3,4\}, where v4v_{4} is a common neighbor of v3v_{3} and either v1v_{1} or v2v_{2}. Other neighbors of viv_{i}’s are labeled as Figure 5. Let S={v1,v2,v3,v4}S=\{v_{1},v_{2},v_{3},v_{4}\}. By Lemma 2.3 (ii), S{w,w3,w4}=S\cap\{w,w_{3},w_{4}\}=\emptyset. Note that |Aφ(v1)|4|A_{\varphi}(v_{1})|\geq 4, |Aφ(v2)|3|A_{\varphi}(v_{2})|\geq 3, |Aφ(v3)|2|A_{\varphi}(v_{3})|\geq 2, and |Aφ(v4)|3|A_{\varphi}(v_{4})|\geq 3 by (2.1). Then it is possible to color viv_{i} with a color in Aφ(vi)A_{\varphi}(v_{i}) for each i{1,2,3,4}i\in\{1,2,3,4\} so that the colors of viv_{i}’s are distinct. It results in an SO (Δ+4)(\Delta+4)-coloring of GG. ∎

Lemma 3.3.

The graph GG has no 33-vertex with two 313_{1}-neighbors. [C5]

Refer to caption
Figure 6: An illustration of Lemma 3.3.
Proof.

Suppose to the contrary that GG has a 33-vertex uu with two 313_{1}-neighbors v1v_{1} and v2v_{2}. We follow the labeling of the vertices as Figure 6. Let S={u,v1,v2,v1,v2}S=\{u,v_{1},v_{2},v^{\prime}_{1},v^{\prime}_{2}\}. By Lemma 3.2, v1v2v^{\prime}_{1}\neq v^{\prime}_{2}, and so SS has five distinct vertices. Note that for each i{1,2}i\in\{1,2\}, vi{w3i,x3i}v_{i}\notin\{w_{3-i},x_{3-i}\} by Lemma 3.2. Also, uwiu\neq w_{i}, vivv^{\prime}_{i}\neq v by Lemma 2.3 (ii), and vi{w3i,x3i}v^{\prime}_{i}\notin\{w_{3-i},x_{3-i}\} by Lemma 2.3 (iv). Thus S{w1,w2,x1,x2,v}=S\cap\{w_{1},w_{2},x_{1},x_{2},v\}=\emptyset. Note that |Aφ(u)|2|A_{\varphi}(u)|\geq 2, |Aφ(vi)|2|A_{\varphi}(v_{i})|\geq 2, and |Aφ(vi)|3|A_{\varphi}(v^{\prime}_{i})|\geq 3 for each i{1,2}i\in\{1,2\} by (2.1).

First, we color uu with a color γ\gamma in Aφ(u)A_{\varphi}(u). For simplicity, let Ai=Aφ(vi){φ(v)}{γ}A_{i}=A_{\varphi}(v_{i})\cup\{\varphi(v)\}\setminus\{\gamma\} for each i{1,2}i\in\{1,2\}. Since φ(v)Aφ(vi)\varphi(v)\notin A_{\varphi}(v_{i}), |Ai|2|A_{i}|\geq 2 for each i{1,2}i\in\{1,2\}. Let A3={φ(v)}A_{3}=\{\varphi(v)\}. If x1x2x_{1}\neq x_{2}, then we apply Lemma 2.1 to obtain a system of odd representative (β1,β2,α)(\beta_{1},\beta_{2},\alpha) for (A1,A2,A3)(A_{1},A_{2},A_{3}), where α=φ(v)\alpha=\varphi(v). If x1=x2x_{1}=x_{2}, then |Aφ(vi)|3|A_{\varphi}(v_{i})|\geq 3 for each i{1,2}i\in\{1,2\} by (2.1), and so we choose βi\beta_{i} in Aφ(vi){γ}A_{\varphi}(v_{i})\setminus\{\gamma\} for each i{1,2}i\in\{1,2\} so that β1β2\beta_{1}\neq\beta_{2}. Then we color viv_{i} with βi\beta_{i} in AiA_{i} for each i{1,2}i\in\{1,2\}. We denote this coloring of G{v1,v2}G-\{v^{\prime}_{1},v^{\prime}_{2}\} by φ\varphi again. Then |Aφ(vi)|1|A_{\varphi}(v^{\prime}_{i})|\geq 1 for each i{1,2}i\in\{1,2\} by (2.1) and so we color v1v^{\prime}_{1} with a color in Aφ(v1)A_{\varphi}(v^{\prime}_{1}). Then the resulting coloring is an SO (Δ+4)(\Delta+4)-coloring of Gv2G-v^{\prime}_{2}, which contradicts to Lemma 2.2. ∎

Lemma 3.4.

The graph GG has no 424_{2}-vertex with a 313_{1}-neighbor. [C6]

Refer to caption
Figure 7: An illustration of Lemma 3.4.
Proof.

Suppose to the contrary that GG has a 44-vertex uu with two 22-neighbors v1v_{1}, v2v_{2} and one 313_{1}-neighbor v3v_{3}. We follow the labeling of the vertices as Figure 7. Let S={u,v1,v2,v3,x}S=\{u,v_{1},v_{2},v_{3},x\}. Note that xvix\neq v_{i} for each i{1,2}i\in\{1,2\} by Lemma 2.3 (ii), and so SS has five distinct vertices. For each i{1,2}i\in\{1,2\}, note that uyu\neq y and v3wiv_{3}\neq w_{i} by Lemma 2.3 (ii). Also, vi{w3,w3i,y}v_{i}\notin\{w_{3},w_{3-i},y\}, x{w1,w2,v}x\notin\{w_{1},w_{2},v\} by Lemma 2.3 (iv). Thus S{w1,w2,w3,y,v}=S\cap\{w_{1},w_{2},w_{3},y,v\}=\emptyset. Note that |Aφ(u)|1|A_{\varphi}(u)|\geq 1, |Aφ(vi)|3|A_{\varphi}(v_{i})|\geq 3 for each i{1,2}i\in\{1,2\}, |Aφ(v3)|2|A_{\varphi}(v_{3})|\geq 2, and |Aφ(x)|3|A_{\varphi}(x)|\geq 3 by (2.1).

First, we color uu with a color γ\gamma from Aφ(u)A_{\varphi}(u). For simplicity, let Ai=Aφ(vi){φ(v)}{γ}A_{i}=A_{\varphi}(v_{i})\cup\{\varphi(v)\}\setminus\{\gamma\} for each i{1,2,3}i\in\{1,2,3\}. Since φ(v)Aφ(vi)\varphi(v)\notin A_{\varphi}(v_{i}) for each i{1,2,3}i\in\{1,2,3\}, |Ai|3|A_{i}|\geq 3 for each i{1,2}i\in\{1,2\} and |A3|2|A_{3}|\geq 2. Let A4={φ(v)}A_{4}=\{\varphi(v)\}. We color v1,v2,v3v_{1},v_{2},v_{3} as follows.

If w1=w2=w3w_{1}=w_{2}=w_{3}, then A1=A2A_{1}=A_{2} and A3A1A_{3}\subset A_{1}, and so we color viv_{i}’s with a same color βA3{φ(v)}\beta\in A_{3}\setminus\{\varphi(v)\}. Suppose that it is not the case of w1=w2=w3w_{1}=w_{2}=w_{3}. Without loss of generality, let w2w3w_{2}\neq w_{3}. We take subsets A2A^{\prime}_{2} and A3A^{\prime}_{3} of A2A_{2} and A3A_{3}, respectively, of size exactly two. If w1=wjw_{1}=w_{j} for some j{2,3}j\in\{2,3\}, then |A1|4|A_{1}|\geq 4 and so we take a subset A1A^{\prime}_{1} of A1AjA_{1}\setminus A^{\prime}_{j} of size two. If w1wjw_{1}\neq w_{j} for each j{2,3}j\in\{2,3\}, then let A1=A1A^{\prime}_{1}=A_{1}. Now, we apply Lemma 2.1 to obtain a system of odd representative (β1,β2,β3,φ(v))(\beta_{1},\beta_{2},\beta_{3},\varphi(v)) for (A1,A2,A3,A4)(A^{\prime}_{1},A^{\prime}_{2},A^{\prime}_{3},A_{4}). We color viv_{i} with βi\beta_{i} for each i{1,2,3}i\in\{1,2,3\}. Then the resulting coloring is an SO (Δ+4)(\Delta+4)-coloring of GxG-x, which is a contradiction to Lemma 2.2. ∎

We complete the proof of Theorem 1.2 by discharging technique. Let μ(v):=degG(v)\mu(v):=\deg_{G}(v) be the initial charge of a vertex vv. Then vV(G)μ(v)20|V(G)|7\sum_{v\in V(G)}\mu(v)\leq\frac{20|V(G)|}{7}, since mad(G)207mad(G)\leq\frac{20}{7}. We let μ(v)\mu^{*}(v) be the final charge of vv after the following discharging rules:

  1. (R1)

    Each 3+3^{+}-vertex sends charge 37\frac{3}{7} to each of its 22-neighbors.

  2. (R2)

    Each 33-vertex without a 22-neighbor or 4+4^{+}-vertex sends charge 17\frac{1}{7} to each of its 313_{1}-neighbors.

Take a vertex vv. By [C1], degG(v)2\deg_{G}(v)\geq 2. If degG(v)=2\deg_{G}(v)=2, then vv has only 3+3^{+}-neighbors by [C2], and so μ(v)=2+372=207\mu^{*}(v)=2+\frac{3}{7}\cdot 2=\frac{20}{7} by (R1). Suppose that degG(v)=3\deg_{G}(v)=3. If vv has no 22-neighbor, then vv has at most one 313_{1}-neighbor by [C5], so μ(v)317=207\mu^{*}(v)\geq 3-\frac{1}{7}=\frac{20}{7} by (R2). Suppose that vv has a 22-neighbor. By [C2] and [C4], vv has exactly one 22-neighbor and two other 3+3^{+}-neighbors, which are either 4+4^{+}-vertices or 33-vertices having no 22-neighbors, and so μ(v)=337+172=207\mu^{*}(v)=3-\frac{3}{7}+\frac{1}{7}\cdot 2=\frac{20}{7} by (R1) and (R2). Suppose that degG(v)=4\deg_{G}(v)=4. If vv has at most one 22-neighbor, then μ(v)437173>207\mu^{*}(v)\geq 4-\frac{3}{7}-\frac{1}{7}\cdot 3>\frac{20}{7} by (R1) and (R2). If vv has at least two 22-neighbors, then vv has exactly two 22-neighbors and no 313_{1}-neighbor by [C2] and [C6], and so μ(v)=4372>207\mu^{*}(v)=4-\frac{3}{7}\cdot 2>\frac{20}{7} by (R1). If degG(v)5\deg_{G}(v)\geq 5, then it has at most (degG(v)1)(\deg_{G}(v)-1) 22-neighbors by [C3], and so μ(v)degG(v)37(degG(v)1)17>207\mu^{*}(v)\geq\deg_{G}(v)-\frac{3}{7}\cdot(\deg_{G}(v)-1)-\frac{1}{7}>\frac{20}{7} by (R1) and (R2).

From the fact that vV(G)μ(v)=vV(G)μ(v)20|V(G)|7\sum_{v\in V(G)}\mu(v)=\sum_{v\in V(G)}\mu^{*}(v)\leq\frac{20|V(G)|}{7}, we can conclude that every vertex has final charge exactly 207\frac{20}{7}. Then Δ=3\Delta=3, and V(G)=X2X31YV(G)=X_{2}\cup X_{3_{1}}\cup Y if we let X2X_{2} and X31X_{3_{1}} be the set of 22-, 313_{1}-vertices of GG, respectively, and let YY be the set of 33-vertices without 22-neighbors. Moreover, every 33-vertex in YY has exactly one 313_{1}-neighbor, and so each connected component of G[Y]G[Y] is 22-regular. Lastly, note that both X2X_{2} and X31X_{3_{1}} are independent sets by [C2] and [C4].

Refer to caption
Figure 8: An illustration for the proof of Lemma 3.5.

Now, we finish the proof by showing that G2G^{2} has a 77-coloring, which is also an SO coloring of GG. Let H:=G2[Y]H:=G^{2}[Y], the subgraph of G2G^{2} induced by YY. Note that Δ(H)5\Delta(H)\leq 5, since each vertex of HH has at most four neighbors with distance at most two in GG through only vertices in YY, and has one neighbor with distance at most two in GG through a vertex not in YY.

Lemma 3.5.

There is no connected component in HH that is isomorphic to K6K_{6}.

Proof.

Suppose to the contrary that there is a connected component KK in HH that is isomorphic to K6K_{6}. Then the vertices of KK form a cycle in GG, say K:v1v2v3v4v5v6v1K:v_{1}v_{2}v_{3}v_{4}v_{5}v_{6}v_{1}, and viv_{i} and vi+3v_{i+3} have a common 313_{1}-neighbor wiw_{i} for each i{1,2,3}i\in\{1,2,3\}. Note that viv_{i}’s and wjw_{j}’s are distinct. See Figure 8. For each i{1,2,3}i\in\{1,2,3\}, let xix_{i} be a 2-neighbor of wiw_{i}, xix^{\prime}_{i} be the neighbor of xix_{i} other than wiw_{i}. If x1=x2x_{1}=x_{2}, then x3x_{3} is a cut vertex of GG, which is a contradiction by permuting the colors of an SO 7-coloring of a connected component of Gx3G-x_{3}. Thus x1x_{1}, x2x_{2}, and x3x_{3} are distinct.

Let S=V(K){w1,w2,w3,x1,x2,x3}S=V(K)\cup\{w_{1},w_{2},w_{3},x_{1},x_{2},x_{3}\}. Note that |Aφ(wi)|6|A_{\varphi}(w_{i})|\geq 6, |Aφ(xi)|4|A_{\varphi}(x_{i})|\geq 4 for each i{1,2,3}i\in\{1,2,3\}, |Aφ(vj)|7|A_{\varphi}(v_{j})|\geq 7 for each j{1,,6}j\in\{1,\ldots,6\} by (2.1). First, we color wiw_{i} with a color in Aφ(wi)A_{\varphi}(w_{i}) for each i{1,2,3}i\in\{1,2,3\}, viv_{i} with a color in Aφ(vj)A_{\varphi}(v_{j}) for each j{1,3,4,5}j\in\{1,3,4,5\} so that the colors of seven vertices wiw_{i}’s and vjv_{j}’s are all distinct. Then we color v2v_{2} and v6v_{6} with the same color with w1w_{1}. We denote this resulting coloring of G{x1,x2,x3}G-\{x_{1},x_{2},x_{3}\} as φ\varphi again. For each i{1,2,3}i\in\{1,2,3\}, |Aφ(xi)|1|A_{\varphi}(x_{i})|\geq 1 by (2.1) and we color xix_{i} with a color in Aφ(xi)A_{\varphi}(x_{i}). It gives an SO 77-coloring of GG. ∎

From Lemma 3.5 and Theorem 2.4, there is a proper 55-coloring ϕH:V(H){1,,5}\phi_{H}:V(H)\rightarrow\{1,\ldots,5\}. We color each vertex in X31X_{3_{1}} with 66 or 77 so that two vertices having a common 22-neighbor have distinct colors, and then color each vertex xX2x\in X_{2} with a color not appeared in NG2(x)N_{G^{2}}(x). It is possible since |NG2(x)|6|N_{G^{2}}(x)|\leq 6. Then it gives a coloring G2G^{2} and completes the proof.

4 Strong odd (Δ+3)(\Delta+3)-coloring

Like as the previous section, in each proof, we always start with defining a nonempty set SV(G)S\subset V(G) and φ\varphi is always an SO (Δ(G)+3)(\Delta(G)+{3})-coloring φ\varphi of G=GSG^{\prime}=G-S.

4.1 Proof of Theorem 1.3

Let Δ(G)4\Delta(G)\geq 4, and let GG be a minimal counterexample to Theorem 1.3. We will show that the following do not appear in GG.

  1. [C1]

    A 11^{-}-vertex.

  2. [C2]

    A dd1d_{d-1}-vertex for all 2d32\leq d\leq 3.

  3. [C3]

    A ddd_{d}-vertex for all d4d\geq 4.

  4. [C4]

    A 22-vertex with only 33-neighbors in which at least one is in ZZ.

  5. [C5]

    A 434_{3}-vertex with a 313_{1}-neighbor.

Note that [C1]-[C3] do not appear in GG by Lemma 2.3 (i) and (iv). For simplicity, let

Z={vV(G)x is a 31-vertex with two 31-neighbors}.Z=\{v\in V(G)\mid x\text{ is a }3_{1}\text{-vertex with two $3_{1}$-neighbors}\}.

Let Δ:=Δ(G)\Delta:=\Delta(G). Let SS be a nonempty subset of V(G)V(G). Note that |𝒞|=Δ+37|\mathcal{C}|=\Delta+3\geq 7. If Δ(GS)4\Delta(G-S)\geq 4, then by the minimality of GG, there is an SO (Δ+3)(\Delta+3)-coloring of GSG-S. If Δ(GS)3\Delta(G-S)\leq 3, then GSG-S has an SO 77-coloring by Theorem 1.2, since mad(GS)3011<207mad(G-S)\leq\frac{30}{11}<\frac{20}{7}. Hence, there always exists an SO (Δ+3)(\Delta+3)-coloring of GSG-S.

Lemma 4.1.

The graph GG has no 22-vertex with two 33-neighbors in which one is in ZZ. [C4]

Refer to caption
Figure 9: An illustration of Lemma 4.1.
Proof.

Suppose to the contrary that there is a 22-vertex vv with 33-neighbors u,wu,w and uZu\in Z. Let v1,v2v_{1},v_{2} be the 313_{1}-neighbors of uu, v1v^{\prime}_{1} be the 2-neighbor of v1v_{1}, w1w_{1} be the other neighbor of v1v_{1}, and x1x_{1} be the neighbor of v1v_{1}^{\prime} other than v1v_{1}. See Figure 9. If we have an SO (Δ+3)(\Delta+3)-coloring of GvG-v such that the colors of uu and ww are distinct, then it is a contradiction to Lemma 2.2, since vv always has an available color from the facts that |𝒞|=Δ+37|\mathcal{C}|=\Delta+3\geq 7.

Let S={u,v,v1,v1}S=\{u,v,v_{1},v^{\prime}_{1}\}. By Lemma 2.3 (ii), vv1v\neq v^{\prime}_{1}, and so SS has four distinct vertices. By Lemma 2.3 (iv), ux1u\neq x_{1}, vw1v\neq w_{1}, and so S{w,v2,w1,x1}=S\cap\{w,v_{2},w_{1},x_{1}\}=\emptyset. Note that |Aφ(u)|Δ22|A_{\varphi}(u)|\geq\Delta-2\geq 2, |Aφ(v1)|1|A_{\varphi}(v_{1})|\geq 1, and |Aφ(v1)|2|A_{\varphi}(v^{\prime}_{1})|\geq 2 by (2.1). If we can color xx with a color in Aφ(x)A_{\varphi}(x) for each x{u,v1,v1}x\in\{u,v_{1},v^{\prime}_{1}\} so that their colors are distinct, then it is an SO (Δ+3)(\Delta+3)-coloring of GvG-v such that the colors of uu and ww are distinct, which contradicts to Lemma 2.2. Thus we may assume that Δ=4\Delta=4, Aφ(u)=Aφ(v1)={1,2}A_{\varphi}(u)=A_{\varphi}(v^{\prime}_{1})=\{1,2\}, Aφ(v1)={1}A_{\varphi}(v_{1})=\{1\} and φ(w1)=3\varphi(w_{1})=3. Then 3φ(NG[x1])φ(NG[v2])3\not\in\varphi(N_{G}[x_{1}])\cup\varphi(N_{G}[v_{2}]), φ(w)3\varphi(w)\neq 3. We color v1v^{\prime}_{1} and uu with the color 33 and v1v_{1} with the color 11. It is an SO 77-coloring of GvG-v such that the colors of uu and ww are distinct, which contradicts to Lemma 2.2. ∎

Lemma 4.2.

The graph GG has no 434_{3}-vertex with a 33-neighbor or a 434_{3}-neighbor. [C5]

Refer to caption
Figure 10: An illustration of Lemma 4.2.
Proof.

A 434_{3}-vertex does not have a 33-neighbor by Lemma 2.3 (iv). Also, Δ=4\Delta=4 by Lemma 2.3 (iv) again. Let u1u_{1} and u4u_{4} be adjacent 434_{3}-vertices. Let NG(ui)={u5i,vi,vi+1,vi+2}N_{G}(u_{i})=\{u_{5-i},v_{i},v_{i+1},v_{i+2}\} for each i{1,4}i\in\{1,4\}, and xix_{i} be the neighbor of viv_{i} other than u1,u4u_{1},u_{4} for each i{1,,6}i\in\{1,\ldots,6\}. See Figure 10. By Lemma 2.3 (ii), all viv_{i}’s are distinct.

Claim 4.3.

For each i{1,4}i\in\{1,4\}, there is no SO 77-coloring φ\varphi of G{vi,vi+1,vi+2}G-\{v_{i},v_{i+1},v_{i+2}\} such that the color of uiu_{i} is not equal to colors of xi,xi+1,xi+2x_{i},x_{i+1},x_{i+2}, satisfying that if xj=xkx_{j}=x_{k} for some distinct j,k{i,i+1,i+2}j,k\in\{i,i+1,i+2\}, then |Aφ(vj)|3|A_{\varphi}(v_{j})|\geq 3.

Proof.

Suppose that there is an SO 77-coloring of G{v1,v2,v3}G-\{v_{1},v_{2},v_{3}\} satisfying the condition of the claim. Suppose that x1=x2x_{1}=x_{2}. Then |Aφ(v1)|3|A_{\varphi}(v_{1})|\geq 3 and |Aφ(v2)|3|A_{\varphi}(v_{2})|\geq 3 by assumption. Also, |Aφ(v3)|1|A_{\varphi}(v_{3})|\geq 1 by (2.1), and therefore we can color viv_{i} with a color in Aφ(vi)A_{\varphi}(v_{i}) for each i{1,2,3}i\in\{1,2,3\} so that the colors are distinct, which is an SO 77-coloring of GG. Suppose that x1x_{1}, x2x_{2}, x3x_{3} are distinct. Let Ai:=𝒞φ(NG[xi])A_{i}:=\mathcal{C}-\varphi(N_{G^{\prime}}[x_{i}]) for each i{1,2,3}i\in\{1,2,3\}, and let A4={φ(v)}A_{4}=\{\varphi(v)\}. Note that |Ai|2|A_{i}|\geq 2 for each i{1,2,3}i\in\{1,2,3\}. Applying Lemma 2.1, there is a system of odd representative (β1,β2,β3,φ(v)\beta_{1},\beta_{2},\beta_{3},\varphi(v)) for (A1,A2,A3,A4)(A_{1},A_{2},A_{3},A_{4}). We color each viv_{i} with βi\beta_{i}, which is an SO 77-coloring of GG. ∎

Let S=NG(u1)NG(u4)S=N_{G}(u_{1})\cup N_{G}(u_{4}). Note that SS has eight distinct vertices. Also, |Aφ(vi)|3|A_{\varphi}(v_{i})|\geq 3 and |Aφ(uj)|4|A_{\varphi}(u_{j})|\geq 4 for each i{1,2,3,4,5,6}i\in\{1,2,3,4,5,6\} and j{1,4}j\in\{1,4\} by (2.1). First, suppose that xi,xi+1,xi+2x_{i},x_{i+1},x_{i+2} are distinct for some i{1,4}i\in\{1,4\}. Let i=1i=1. By Claim 4.3, we cannot color xx with a color in Aφ(x)A_{\varphi}(x) for each xNG[u4]x\in N_{G}[u_{4}] so that the colors of NG[u4]N_{G}[u_{4}] are distinct. Thus we may assume that Aφ(u1)=Aφ(u4)={1,2,3,4}A_{\varphi}(u_{1})=A_{\varphi}(u_{4})=\{1,2,3,4\} and Aφ(vj){1,2,3,4}A_{\varphi}(v_{j})\subset\{1,2,3,4\} for each j{4,5,6}j\in\{4,5,6\}, which implies that x4x_{4}, x5x_{5}, x6x_{6} are distinct. Then γAφ(v4)Aφ(v5)Aφ(v6)\gamma\in A_{\varphi}(v_{4})\cap A_{\varphi}(v_{5})\cap A_{\varphi}(v_{6}) for some γ\gamma and so we color v4v_{4}, v5v_{5}, v6v_{6} with γ\gamma, and then color u1u_{1} and u4u_{4} with distinct colors in {1,2,3,4}{γ}\{1,2,3,4\}\setminus\{\gamma\}. It is an SO 77-coloring of G{v1,v2,v3}G-\{v_{1},v_{2},v_{3}\}, which contradicts to Claim 4.3.

Secondly, suppose that xi,xi+1,xi+2x_{i},x_{i+1},x_{i+2} are not distinct for each i{1,4}i\in\{1,4\}. We may assume that x1=x2x_{1}=x_{2} and x4=x5x_{4}=x_{5}. By Lemma 2.3 (iii), x1x3x_{1}\neq x_{3} and x4x6x_{4}\neq x_{6}. Note that |Aφ(ui)|5|A_{\varphi}(u_{i})|\geq 5 for each i{1,4}i\in\{1,4\}, |Aφ(vj)|4|A_{\varphi}(v_{j})|\geq 4 for each j{1,2,4,5}j\in\{1,2,4,5\}, and |Aφ(vk)|3|A_{\varphi}(v_{k})|\geq 3 for each k{3,6}k\in\{3,6\} by (2.1). If |Aφ(vj)|5|A_{\varphi}(v_{j})|\geq 5 for some j{1,2,4,5}j\in\{1,2,4,5\}, say j=1j=1, then we color xx with a color in Aφ(x)A_{\varphi}(x) for each xNG[u4]x\in N_{G}[u_{4}] so that the colors of NG[u4]N_{G}[u_{4}] are distinct, which contradicts to Claim 4.3. Suppose that |Aφ(vj)|=4|A_{\varphi}(v_{j})|=4 for each j{1,2,4,5}j\in\{1,2,4,5\}. Then x1,x3,x4,x6x_{1},x_{3},x_{4},x_{6} are distinct vertices and so degG(x1)=2\deg_{G^{\prime}}{(x_{1})}=2. Therefore, φ(NG(x1)){φ(x3)}\varphi(N_{G^{\prime}}(x_{1}))\setminus\{\varphi(x_{3})\}\neq\emptyset. We color u1u_{1} with a color α1\alpha_{1} in φ(NG(x1)){φ(x3)}\varphi(N_{G^{\prime}}(x_{1}))\setminus\{\varphi(x_{3})\}. We color xx with a color in Aφ(x)A_{\varphi}(x) for each xNG[u4]{u1}x\in N_{G}[u_{4}]\setminus\{u_{1}\} so that the colors of NG[u4]N_{G}[u_{4}] are distinct. We denote this coloring of G{v1,v2,v3}G-\{v_{1},v_{2},v_{3}\} by φ\varphi again. Then |Aφ(v1)|,|Aφ(v2)|3|A_{\varphi}(v_{1})|,|A_{\varphi}(v_{2})|\geq 3 by the choice of α1\alpha_{1} and |Aφ(v3)|1|A_{\varphi}(v_{3})|\geq 1 by (2.1). It is a contradiction to Claim 4.3. ∎

We complete the proof of Theorem 1.3 by discharging technique. Let μ(v):=degG(v)\mu(v):=\deg_{G}(v) be the initial charge of a vertex vv, and let μ(v)\mu^{*}(v) be the final charge of vv after the discharging rules:

  1. (R1)

    Each 33-vertex in ZZ sends charge 311\frac{3}{11} to each of its 22-neighbors.

  2. (R2)

    Each 33-vertex not in ZZ sends charge 411\frac{4}{11} to each of its 22-neighbors.

  3. (R3)

    Each 4+4^{+}-vertex sends charge 511\frac{5}{11} to each of its 22-neighbors.

  4. (R4)

    Each 44-vertex with at most two 22-neighbors sends charge 111\frac{1}{11} to each of its 313_{1}-neighbors and 434_{3}-neighbors.

  5. (R5)

    Each 5+5^{+}- or 33-vertex without 22-neighbor sends charge 111\frac{1}{11} to each of its 313_{1}-neighbors and 434_{3}-neighbors.

Take a vertex vv. We will show that μ(v)3011\mu^{*}(v)\geq\frac{30}{11}. By [C1], degG(v)2\deg_{G}(v)\geq 2. Suppose that degG(v)=2\deg_{G}(v)=2. Then vv has only 3+3^{+}-neighbors by [C2]. If vv has no neighbor in ZZ, then μ(v)2+4112=3011\mu^{*}(v)\geq 2+\frac{4}{11}\cdot 2=\frac{30}{11} by (R2) and (R3). If vv has a neighbor in ZZ, then by [C4] vv also has a 4+4^{+}-neighbor and so μ(v)2+311+511=3011\mu^{*}(v)\geq 2+\frac{3}{11}+\frac{5}{11}=\frac{30}{11} by (R1) and (R3).

Suppose degG(v)=3\deg_{G}(v)=3. By [C2], it has at most one 22-neighbor. If vv has no 22-neighbor, then μ(v)31113=3011\mu^{*}(v)\geq 3-\frac{1}{11}\cdot 3=\frac{30}{11} by (R5). Suppose vv is a 313_{1}-vertex. If vZv\in Z, then by (R1), μ(v)=3311=3011\mu^{*}(v)=3-\frac{3}{11}=\frac{30}{11}. Suppose that vZv\not\in Z. Then vv has at least one 3+3^{+}-neighbor uu that is not a 313_{1}-vertex. Then uu is either a 33-vertex without 22-neighbors, a 44-vertex with at most two 2-neighbors by [C5], or a 5+5^{+}-vertex. By (R4) and (R5), μ(v)3411+111=3011\mu^{*}(v)\geq 3-\frac{4}{11}+\frac{1}{11}=\frac{30}{11}.

Suppose that degG(v)=4\deg_{G}(v)=4. Then it has at most three 22-neighbors by [C3]. If vv has at most two 22-neighbors, then by (R3) and (R4), μ(v)451121112=3111\mu^{*}(v)\geq 4-\frac{5}{11}\cdot 2-\frac{1}{11}\cdot 2=\frac{31}{11}. Suppose that vv has three 22-neighbors. Let uu be the 3+3^{+}-neighbor of vv. Then uu is neither a 33-vertex nor a 434_{3}-vertex by [C5] and so uu sends 111\frac{1}{11} to vv by (R4) and (R5). Then μ(v)45113+111=3011\mu^{*}(v)\geq 4-\frac{5}{11}\cdot 3+\frac{1}{11}=\frac{30}{11}. If degG(v)5\deg_{G}(v)\geq 5, then it has at most (degG(v)1)(\deg_{G}(v)-1) 22-neighbors and so μ(v)degG(v)511(degG(v)1)111>3011\mu^{*}(v)\geq\deg_{G}(v)-\frac{5}{11}\cdot(\deg_{G}(v)-1)-\frac{1}{11}>\frac{30}{11} by (R3) and (R5). Thus μ(v)3011\mu^{*}(v)\geq\frac{30}{11} for each vertex vv. Since μ(v)=μ(v)30|V(G)|11\sum\mu^{*}(v)=\sum\mu(v)\leq\frac{30|V(G)|}{11}, we have μ(v)=3011\mu^{*}(v)=\frac{30}{11} for each vertex vv. Thus vv is either 22-, 313_{1}-, or 434_{3}-vertex. Since Δ4\Delta\geq 4, there is a 434_{3}-vertex xx in GG. By [C3], there is a 3+3^{+}-neighbor yy of xx. However, yy is neither a 434_{3}-vertex nor a 33-vertex by [C5], which is a contradiction.

4.2 Proof of Theorem 1.4

Let GG be a minimal counterexample to Theorem 1.4. We will show that the following do not appear in GG.

  1. [C1]

    A 11^{-}-vertex.

  2. [C2]

    A dd1d_{d-1}-vertex for all 2d32\leq d\leq 3.

  3. [C3]

    A 313_{1}-vertex with two 313_{1}-neighbors.

Note that [C1] and [C2] do not appear in GG by Lemma 2.3 (i) and (iv).

Lemma 4.4.

The graph GG has no triangle consisting of 313_{1}-vertices.

Refer to caption
Figure 11: An illustration of Lemma 4.4.
Proof.

Suppose to the contrary that GG has a triangle v1v2v3v1v_{1}v_{2}v_{3}v_{1} consisting of only 313_{1}-vertices. Let viv^{\prime}_{i} be the 22-neighbor of viv_{i} and NG(vi)={vi,wi}N_{G}(v^{\prime}_{i})=\{v_{i},w_{i}\} for each i{1,2,3}i\in\{1,2,3\}. See Figure 11. Let S={v1,v2,v3,v1,v2,v3}S=\{v_{1},v_{2},v_{3},v^{\prime}_{1},v^{\prime}_{2},v^{\prime}_{3}\}. By Lemma 2.3 (ii), v1v^{\prime}_{1}, v2v^{\prime}_{2}, v3v^{\prime}_{3} are distinct, and so SS has distinct six vertices. Note that wiw_{i}’s are distinct by Lemma 2.3 (iv).

By (2.1), |Aφ(vi)|=5|A_{\varphi}(v_{i})|=5, and |Aφ(vi)|3|A_{\varphi}(v^{\prime}_{i})|\geq 3 for each i{1,2,3}i\in\{1,2,3\}. If we can choose a color in Aφ(x)A_{\varphi}(x) for each xSx\in S to color with distinct colors, then it results in an SO 66-coloring of GG. Thus Aφ(v1)=Aφ(v2)=Aφ(v3)A_{\varphi}(v_{1})=A_{\varphi}(v_{2})=A_{\varphi}(v_{3}) and so φ(w1)=φ(w2)=φ(w3)\varphi(w_{1})=\varphi(w_{2})=\varphi(w_{3}). Since |Aφ(vi)|3|A_{\varphi}(v^{\prime}_{i})|\geq 3 and φ(w1)=φ(w2)\varphi(w_{1})=\varphi(w_{2}), it follows that γAφ(v1)Aφ(v2)\gamma\in A_{\varphi}(v^{\prime}_{1})\cap A_{\varphi}(v^{\prime}_{2})\neq\emptyset for some γ\gamma. We color v1v^{\prime}_{1} and v2v^{\prime}_{2} with γ\gamma. For xNG[v3]x\in N_{G}[v_{3}], we color xx with a color in Aφ(x){γ}A_{\varphi}(x)\setminus\{\gamma\} so that all colors of the vertices in NG[v3]N_{G}[v_{3}] are distinct. It is an SO 66-coloring of GG. ∎

Lemma 4.5.

The graph GG has no 313_{1}-vertex with two 313_{1}-neighbors.[C3]

Refer to caption
Figure 12: An illustration of Lemma 4.5.
Proof.

Suppose to the contrary that GG has a vertex uZu\in Z. We follow the labeling of the vertices as Figure 12. Note that v1v^{\prime}_{1}, v2v^{\prime}_{2}, ww, w1w_{1}, and w2w_{2} are distinct since GG has no 44-cycle. Let S={u,v,v1,v2,v1,v2}S=\{u,v,v_{1},v_{2},v^{\prime}_{1},v^{\prime}_{2}\}. By Lemma 2.3 (ii), vviv\neq v^{\prime}_{i} for each i{1,2}i\in\{1,2\}, and so SS has six distinct vertices. Note that for each i{1,2}i\in\{1,2\}, uxiu\neq x_{i}, viwv_{i}\neq w by Lemma 2.3 (ii). Also, v{x1,x2,w1,w2}v\notin\{x_{1},x_{2},w_{1},w_{2}\}, vi{x1,x2,w1,w2,w}v^{\prime}_{i}\notin\{x_{1},x_{2},w_{1},w_{2},w\}, vix3iv_{i}\neq x_{3-i} by Lemma 2.3 (iv), and viw3iv_{i}\neq w_{3-i} by Lemma 4.4. Thus S{x1,x2,w1,w2,w}S\cap\{x_{1},x_{2},w_{1},w_{2},w\}\neq\emptyset.

In addition, ww, x1x_{1}, x2x_{2} are distinct by Lemma 2.3 (iv). Note that |Aφ(vi)|2|A_{\varphi}(v_{i})|\geq 2 for each i{1,2}i\in\{1,2\} by (2.1). Also, since ww is a 22-vertex in GG^{\prime}, |Aφ(v)|=3|A_{\varphi}(v)|=3 by (2.1).

Claim 4.6.

Aφ(v)Aφ(v1)Aφ(v2)=A_{\varphi}(v)\cap A_{\varphi}(v_{1})\cap A_{\varphi}(v_{2})=\emptyset.

Proof.

Suppose to the contrary that αAφ(v)Aφ(v1)Aφ(v2)\alpha\in A_{\varphi}(v)\cap A_{\varphi}(v_{1})\cap A_{\varphi}(v_{2}). We give a color α\alpha to vv, v1v_{1}, and v2v_{2}, and denote this coloring of G{u,v1,v2}G-\{u,v^{\prime}_{1},v^{\prime}_{2}\} by φ\varphi again. Note that |Aφ(vi)|1|A_{\varphi}(v^{\prime}_{i})|\geq 1 for each i{1,2}i\in\{1,2\} by (2.1). Next, we color viv^{\prime}_{i} with a color βi\beta_{i} in Aφ(vi)A_{\varphi}(v^{\prime}_{i}) for each i{1,2}i\in\{1,2\}.

If there is a color in 𝒞{φ(w1),φ(w2),φ(w),α,β1,β2}\mathcal{C}\setminus\{\varphi(w_{1}),\varphi(w_{2}),\varphi(w),\alpha,\beta_{1},\beta_{2}\}, then we color uu with the color and it is done. Thus {φ(w1),φ(w2),φ(w),α,β1,β2}=𝒞\{\varphi(w_{1}),\varphi(w_{2}),\varphi(w),\alpha,\beta_{1},\beta_{2}\}=\mathcal{C}. Moreover, Aφ(vi)={βi}A_{\varphi}(v^{\prime}_{i})=\{\beta_{i}\} and so φ(wi)φ(NG[xi])\varphi(w_{i})\notin\varphi(N_{G^{\prime}}[x_{i}]) for each i{1,2}i\in\{1,2\}. Coloring both v1v^{\prime}_{1} and uu with φ(w1)\varphi(w_{1}) gives an SO 66-coloring of GG. ∎

Claim 4.7.

There is a color αiAφ(vi)\alpha_{i}\in A_{\varphi}(v_{i}) for each i{1,2}i\in\{1,2\} such that |Aφ(v){α1,α2}|2|A_{\varphi}(v)\setminus\{\alpha_{1},\alpha_{2}\}|\geq 2.

Proof.

Recall that |Aφ(v1)|,|Aφ(v2)|2|A_{\varphi}(v_{1})|,|A_{\varphi}(v_{2})|\geq 2 and |Aφ(v)|=3|A_{\varphi}(v)|=3. Suppose that Aφ(vi)Aφ(v)A_{\varphi}(v_{i})\subset A_{\varphi}(v) for each i{1,2}i\in\{1,2\}. Then |Aφ(v1)Aφ(v2)||Aφ(v1)|+|Aφ(v2)||Aφ(v)|2+23=1|A_{\varphi}(v_{1})\cap A_{\varphi}(v_{2})|\geq|A_{\varphi}(v_{1})|+|A_{\varphi}(v_{2})|-|A_{\varphi}(v)|\geq 2+2-3=1. An element in Aφ(v1)Aφ(v2)A_{\varphi}(v_{1})\cap A_{\varphi}(v_{2}) belongs to Aφ(v1)Aφ(v2)Aφ(v)A_{\varphi}(v_{1})\cap A_{\varphi}(v_{2})\cap A_{\varphi}(v), a contradiction to Claim 4.6. Thus Aφ(vi)Aφ(v)A_{\varphi}(v_{i})\setminus A_{\varphi}(v)\neq\emptyset for some i{1,2}i\in\{1,2\}. We may assume that i=1i=1. We take α1Aφ(v1)Aφ(v)\alpha_{1}\in A_{\varphi}(v_{1})\setminus A_{\varphi}(v) and α2Aφ(v2){α1}\alpha_{2}\in A_{\varphi}(v_{2})\setminus\{\alpha_{1}\}, and so |Aφ(v){α1,α2}|2|A_{\varphi}(v)\setminus\{\alpha_{1},\alpha_{2}\}|\geq 2. ∎

We color viv_{i} with αi\alpha_{i} for each i{1,2}i\in\{1,2\}, where α1\alpha_{1} and α2\alpha_{2} satisfy Claim 4.7, and denote this coloring of G{u,v,v1,v2}G-\{u,v,v^{\prime}_{1},v^{\prime}_{2}\} by φ\varphi again. Then for each i{1,2}i\in\{1,2\}, |Aφ(vi)|1|A_{\varphi}(v^{\prime}_{i})|\geq 1 by (2.1), and

(§)if |Aφ(vi)|=1 then φ(wi)φ(NG[xi]).(\S)\qquad\text{if }|A_{\varphi}(v^{\prime}_{i})|=1\text{ then }\varphi(w_{i})\notin\varphi(N_{G^{\prime}}[x_{i}]).\qquad
Claim 4.8.

There is no SO 66-coloring of GvG-v that is an extension of φ\varphi such that the color of uu is not φ(w)\varphi(w).

Proof.

Suppose that we color v1v^{\prime}_{1}, v2v^{\prime}_{2}, and uu to obtain an SO 66-coloring φ\varphi^{*} of GvG-v satisfying φ(u)φ(w)\varphi^{*}(u)\neq\varphi(w). By the choice of α1\alpha_{1} and α2\alpha_{2}, there is at least one available color in Aφ(v)A_{\varphi^{*}}(v). It contradicts to Lemma 2.2. ∎

Let B=Aφ(v1)×Aφ(v2)B=A_{\varphi}(v^{\prime}_{1})\times A_{\varphi}(v^{\prime}_{2}). Note that BB\neq\emptyset. If there is a color γAφ(u){β1,β2}\gamma\in A_{\varphi}(u)\setminus\{\beta_{1},\beta_{2}\} for some (β1,β2)B(\beta_{1},\beta_{2})\in B, then coloring v1v^{\prime}_{1}, v2v^{\prime}_{2} and uu with β1\beta_{1}, β2\beta_{2} and γ\gamma, respectively, gives an SO 66-coloring of GvG-v such that γφ(w)\gamma\neq\varphi(w), which contradicts to Claim 4.8. Thus Aφ(u){β1,β2}=A_{\varphi}(u)\setminus\{\beta_{1},\beta_{2}\}=\emptyset, and so

𝒞\displaystyle\mathcal{C} =\displaystyle= {α1,α2,β1,β2,φ(w),φ(w1),φ(w2)} for every (β1,β2)B.\displaystyle\{\alpha_{1},\alpha_{2},\beta_{1},\beta_{2},\varphi(w),\varphi(w_{1}),\varphi(w_{2})\}\qquad\text{ for every }(\beta_{1},\beta_{2})\in B. (4.1)

Note that for every (β1,β2)B(\beta_{1},\beta_{2})\in B, βiφ(wi)\beta_{i}\neq\varphi(w_{i}) by the definition of Aφ(vi)A_{\varphi}(v^{\prime}_{i}) for each i{1,2}i\in\{1,2\} and {βi,φ(wi)}{α1,α2,φ(w)}=\{\beta_{i},\varphi(w_{i})\}\cap\{\alpha_{1},\alpha_{2},\varphi(w)\}=\emptyset for some i{1,2}i\in\{1,2\}.

Claim 4.9.

For every (β1,β2)B(\beta_{1},\beta_{2})\in B, it holds that |{β1,β2,φ(w1),φ(w2)}|=3|\{\beta_{1},\beta_{2},\varphi(w_{1}),\varphi(w_{2})\}|=3 and Aφ(v1)Aφ(v2){β1,β2,φ(w1),φ(w2)}A_{\varphi}(v^{\prime}_{1})\cup A_{\varphi}(v^{\prime}_{2})\subset\{\beta_{1},\beta_{2},\varphi(w_{1}),\varphi(w_{2})\}.

Proof.

By (4.1), |{β1,β2,φ(w1),φ(w2)}|3|\{\beta_{1},\beta_{2},\varphi(w_{1}),\varphi(w_{2})\}|\geq 3 for every (β1,β2)B(\beta_{1},\beta_{2})\in B. Suppose to the contrary |{β1,β2,φ(w1),φ(w2)}|=4|\{\beta_{1},\beta_{2},\varphi(w_{1}),\varphi(w_{2})\}|=4 for some (β1,β2)B(\beta_{1},\beta_{2})\in B. Without loss of generality, let {β1,φ(w1)}{α1,α2,φ(w)}=\{\beta_{1},\varphi(w_{1})\}\cap\{\alpha_{1},\alpha_{2},\varphi(w)\}=\emptyset. If |Aφ(v1)|2|A_{\varphi}(v^{\prime}_{1})|\geq 2, then we color v1v^{\prime}_{1}, v2v^{\prime}_{2}, and uu with a color in Aφ(v1){β1}A_{\varphi}(v^{\prime}_{1})\setminus\{\beta_{1}\}, β2\beta_{2}, and β1\beta_{1}, respectively, which gives an SO 66-coloring of GvG-v such that the color of uu is not φ(w)\varphi(w), a contradiction to Claim 4.8. Thus |Aφ(v1)|=1|A_{\varphi}(v^{\prime}_{1})|=1 and so by (§\S), φ(w1)φ(NG[x1])\varphi(w_{1})\notin\varphi(N_{G^{\prime}}[x_{1}]). Then we color v1v^{\prime}_{1}, v2v^{\prime}_{2}, and uu with φ(w1)\varphi(w_{1}), β2\beta_{2}, and φ(w1)\varphi(w_{1}), respectively, which gives an SO 66-coloring of GvG-v. Then the color of uu is not φ(w)\varphi(w) and so it is a contradiction to Claim 4.8. Thus |{β1,β2,φ(w1),φ(w2)}|=3|\{\beta_{1},\beta_{2},\varphi(w_{1}),\varphi(w_{2})\}|=3 and so {α1,α2,φ(w)}{β1,β2,φ(w1),φ(w2)}=\{\alpha_{1},\alpha_{2},\varphi(w)\}\cap\{\beta_{1},\beta_{2},\varphi(w_{1}),\varphi(w_{2})\}=\emptyset for every (β1,β2)B(\beta_{1},\beta_{2})\in B. Therefore each Aφ(vi)A_{\varphi}(v^{\prime}_{i}) is a subset of {β1,β2,φ(w1),φ(w2)}\{\beta_{1},\beta_{2},\varphi(w_{1}),\varphi(w_{2})\}. ∎

If φ(w1)=φ(w2)\varphi(w_{1})=\varphi(w_{2}), then for each i{1,2}i\in\{1,2\}, |Aφ(vi)|=1|A_{\varphi}(v^{\prime}_{i})|=1 by Claim 4.9 and so φ(w1)φ(NG[xi])\varphi(w_{1})\notin\varphi(N_{G^{\prime}}[x_{i}]) by (§\S). Then we color v1v^{\prime}_{1}, v2v^{\prime}_{2}, and uu with φ(w1)\varphi(w_{1}), and it is an SO 66-coloring of GvG-v such that the color of uu is not φ(w)\varphi(w), which contradicts to Claim 4.8. Thus φ(w1)φ(w2)\varphi(w_{1})\neq\varphi(w_{2}).

In addition, by Claim 4.9, since (φ(w2),φ(w1))B(\varphi(w_{2}),\varphi(w_{1}))\notin B,

()either φ(w2)Aφ(v1) or φ(w1)Aφ(v2).({\ddagger})\qquad\text{either }\varphi(w_{2})\not\in A_{\varphi}(v^{\prime}_{1})\text{ or }\varphi(w_{1})\not\in A_{\varphi}(v^{\prime}_{2}).
Claim 4.10.

For some i{1,2}i\in\{1,2\}, φ(wi)φ(NG[xi])\varphi(w_{i})\not\in\varphi(N_{G^{\prime}}[x_{i}]) and Aφ(v3i){φ(wi)}A_{\varphi}(v^{\prime}_{3-i})\neq\{\varphi(w_{i})\}.

Proof.

Suppose to the contrary that each φ(wi)\varphi(w_{i}) satisfies (P) or (Q), where (P) and (Q) are properties defined as follows: (P) φ(wi)φ(NG[xi])\varphi(w_{i})\in\varphi(N_{G^{\prime}}[x_{i}]) and (Q) Aφ(v3i)={φ(wi)}A_{\varphi}(v^{\prime}_{3-i})=\{\varphi(w_{i})\}.

By ({\ddagger}), φ(wi)\varphi(w_{i}) does not satisfy (Q) for some i{1,2}i\in\{1,2\}. Then φ(wi)\varphi(w_{i}) satisfies (P), and so |Aφ(vi)|2|A_{\varphi}(v^{\prime}_{i})|\geq 2 by (§\S). Then φ(w3i)\varphi(w_{3-i}) does not satisfy (Q). By the assumption, φ(w3i)\varphi(w_{3-i}) also satisfy (P) and so |Aφ(v3i)|2|A_{\varphi}(v^{\prime}_{3-i})|\geq 2 by (§\S). Since |Aφ(v1)Aφ(v2)|3|A_{\varphi}(v^{\prime}_{1})\cup A_{\varphi}(v^{\prime}_{2})|\leq 3 by Claim 4.9, there is γAφ(v1)Aφ(v2)\gamma\in A_{\varphi}(v^{\prime}_{1})\cap A_{\varphi}(v^{\prime}_{2}). By the definition of Aφ(vi)A_{\varphi}(v^{\prime}_{i}), γ{φ(w1),φ(w2)}\gamma\not\in\{\varphi(w_{1}),\varphi(w_{2})\}, and so Aφ(vi)={γ,φ(w3i)}A_{\varphi}(v^{\prime}_{i})=\{\gamma,\varphi(w_{3-i})\} for each i{1,2}i\in\{1,2\}. It is a contradiction to ({\ddagger}). ∎

By Claim 4.10, without loss of generality, we may let φ(w1)φ(NG[xi])\varphi(w_{1})\notin\varphi(N_{G^{\prime}}[x_{i}]) and Aφ(v2){φ(w1)}A_{\varphi}(v^{\prime}_{2})\setminus\{\varphi(w_{1})\}\neq\emptyset. We take β2Aφ(v2){φ(w1)}\beta_{2}\in A_{\varphi}(v^{\prime}_{2})\setminus\{\varphi(w_{1})\} and color v1v^{\prime}_{1}, v2v^{\prime}_{2}, and uu with φ(w1)\varphi(w_{1}), β2\beta_{2}, and φ(w1)\varphi(w_{1}), respectively, since φ(w1)φ(w2)\varphi(w_{1})\neq\varphi(w_{2}). It gives an SO 66-coloring of GvG-v and the color of uu is not φ(w)\varphi(w), which is a contradiction to Claim 4.8. ∎

We complete the proof of Theorem 1.4 by discharging technique. Let μ(v):=degG(v)\mu(v):=\deg_{G}(v) be the initial charge of a vertex vv, and μ(v)\mu^{*}(v) be the final charge of vv after the following discharging rules:

  1. (R1)

    Each 3+3^{+}-vertex sends charge 411\frac{4}{11} to each of its 22-neighbors.

  2. (R2)

    Each 33-vertex without 22-neighbors sends charge 111\frac{1}{11} to each of its 313_{1}-neighbors.

Take a vertex vv. We will show that μ(v)3011\mu^{*}(v)\geq\frac{30}{11}. By [C1], degG(v)2\deg_{G}(v)\geq 2. If degG(v)=2\deg_{G}(v)=2, then vv has only 3+3^{+}-neighbors by [C2], and so μ(v)=2+4112=3011\mu^{*}(v)=2+\frac{4}{11}\cdot 2=\frac{30}{11} by (R1). Suppose degG(v)=3\deg_{G}(v)=3. By [C2], it has at most one 22-neighbor. If vv has no 22-neighbor, then μ(v)31113=3011\mu^{*}(v)\geq 3-\frac{1}{11}\cdot 3=\frac{30}{11} by (R2). Suppose vv is a 313_{1}-vertex. By [C3], it has at most one 313_{1}-neighbor, and so vv has at least one 33-neighbor without 2-vetices. By (R1) and (R2), μ(v)3411+111=3011\mu^{*}(v)\geq 3-\frac{4}{11}+\frac{1}{11}=\frac{30}{11}. Thus μ(v)3011\mu^{*}(v)\geq\frac{30}{11} for each vertex vv. Since μ(v)=μ(v)30|V(G)|11\sum{\mu^{*}}(v)=\sum\mu(v)\leq\frac{30|V(G)|}{11}, we have μ(v)=3011\mu^{*}(v)=\frac{30}{11} for each vertex vv.

Let X0X_{0} be the set of 22-vertices, X1X_{1} be the the set of 313_{1}-vertices, and X2X_{2} be the the set of 33-vertices not in X1X_{1}. The final charge of each vertex is exactly 3011\frac{30}{11}, and it implies that G[X1]G[X_{1}] forms an induced matching, and X2X_{2} forms an independent set. We color each 22-vertex with a color 1, each 33-vertex in X2X_{2} with a color 22. We will color the vertices of X1X_{1} with colors 3,4,5,63,4,5,6 as follows. Let H=G2[X1]H=G^{2}[X_{1}], subgraph of G2G^{2} induced by X1X_{1}. Then Δ(H)4\Delta(H)\leq 4. Moreover, by the structure of GG, HH cannot have K5K_{5} as a connected component. By Theorem 2.4, there is a proper coloring φH\varphi_{H} of HH with the color set {3,4,5,6}\{3,4,5,6\}, and we color each vertex xX1x\in X_{1} with the color φH(x)\varphi_{H}(x). Then one can check that it is an SO coloring of GG, a contradiction.

5 Remarks

Regarding to Theorem 1.2, we know that the bound of the strong chromatic number of subcubic planar graph is tight by a graph in Figure 2. But we do not think the bound in Theorem 1.2 is not tight if a graph is large enough.

In odd coloring, Petruševski and Škrekovski [20] asked if χo(G)5\chi_{o}(G)\leq 5 for a planar graph GG. It is proved that if GG is a planar graph, then χo(G)8\chi_{o}(G)\leq 8 by Petr and Portier [20]. We can also ask it is possible to limit the strong odd chromatic number of all planar graphs to a constant.

Acknowledgements

This work was supported under the framework of international cooperation program managed by the National Research Foundation of Korea (NRF-2023K2A9A2A06059347).

References

  • [1] O. V. Borodin and A. O. Ivanova. List 2-facial 5-colorability of plane graphs with girth at least 12. Discrete Math., 312(2):306–314, 2012.
  • [2] R. L. Brooks. On colouring the nodes of a network. Proc. Cambridge Philos. Soc., 37:194–197, 1941.
  • [3] Y. Bu and C. Shang. List 2-distance coloring of planar graphs without short cycles. Discrete Math. Algorithms Appl., 8(01):1650013, 2016.
  • [4] Y. Bu and J. Zhu. Channel Assignment with r-Dynamic Coloring: 12th International Conference, AAIM 2018, Dallas, TX, USA, December 3–4, 2018. In Proceedings, pages 36–48, 2018.
  • [5] Y. Bu and X. Zhu. An optimal square coloring of planar graphs. J. Comb. Optim., 24(4):580–592, 2012.
  • [6] Y. Caro, M. Petruševski, and R. Škrekovski. Remarks on odd colorings of graphs. Discrete Appl. Math., 321:392–401, 2022.
  • [7] E.-K. Cho, I. Choi, H. Kwon, and B. Park. Odd coloring of sparse graphs and planar graphs. Discrete Math., 346(5):113305, 2023.
  • [8] D. W. Cranston. Coloring, list coloring, and painting squares of graphs (and other related problems). arXiv preprint, page arXiv:2210.05915, 2022.
  • [9] D. W. Cranston. Odd colorings of sparse graphs. arXiv preprint arXiv:2201.01455, 2022.
  • [10] D. W. Cranston, M. Lafferty, and Z.-X. Song. A note on odd colorings of 1-planar graphs. Discrete Appl. Math., 330:112–117, 2023.
  • [11] T. Dai, Q. Ouyang, and F. Pirot. New bounds for odd colourings of graphs. arXiv preprint arXiv:2306.01341, 2023.
  • [12] Z. Deniz. An improved bound for 2-distance coloring of planar graphs with girth six. arXiv preprint arXiv:2212.03831, 2022.
  • [13] Z. Deniz. On 2-distance (Δ+4{\Delta}+4)-coloring of planar graphs with girth at least five. arXiv preprint arXiv:2311.02201, 2023.
  • [14] W. Dong and B. Xu. 2-distance coloring of planar graphs with girth 5. J. Comb. Optim., 34:1302–1322, 2017.
  • [15] V. Dujmović, P. Morin, and S. Odak. Odd colourings of graph products. arXiv preprint arXiv:2202.12882, 2022.
  • [16] Z. Dvořák, R. Škrekovski, and M. Tancer. List-coloring squares of sparse subcubic graphs. SIAM J. Discrete Math., 22(1):139–159, 2008.
  • [17] S. G. Hartke, S. Jahanbekam, and B. Thomas. The chromatic number of the square of subcubic planar graphs. arXiv preprint arXiv:1604.06504, 2016.
  • [18] H. La. 2-distance list (Δ{\Delta}+ 3)-coloring of sparse graphs. Graphs Comb., 38(6):167, 2022.
  • [19] R. Liu, W. Wang, and G. Yu. 1-planar graphs are odd 13-colorable. Discrete Math., 346(8):113423, 2023.
  • [20] J. Petr and J. Portier. The odd chromatic number of a planar graph is at most 8. Graphs Comb., 39(2):28, 2023.
  • [21] M. Petruševski and R. Škrekovski. Colorings with neighborhood parity condition. Discrete Appl. Math., 321:385–391, 2022.
  • [22] C. Thomassen. The square of a planar cubic graph is 7-colorable. J. Comb. Theory, Ser. B, 128:192–218, 2018.
  • [23] T. Wang and X. Yang. On odd colorings of sparse graphs. Discrete Appl. Math., 345:156–169, 2024.
  • [24] G. Wegner. Graphs with given diameter and a coloring problem. Technical report, University of Dortmund, 1977.