1 Introduction

All graphs considered in this paper are finite and simple. A plane graph is a particular drawing of a planar graph on the Euclidean plane so that any two edges intersect only at their ends. For a plane graph G, we denote its vertex set, edge set, and face set by V(G), E(G), and F(G), respectively. For \(x\in V(G)\cup F(G)\), let \(d_G(x)\) denote the degree of x. Let \(N_G(v)\) be the set of neighbors of v. Then, \(d_G(v)=|N_G(v)|\). Let \(\Delta (G)=\max \{d_G(v)|v\in V(G)\}\) be the maximum degree of G and \(\delta (G)=\min \{d_G(v)|v\in V(G)\}\) be the minimum degree of G. In G, a vertex of degree k, at least k, or at most k is called a k-vertex, \(k^{+}\)-vertex, or \(k^{-}\)-vertex, respectively. We use \(d_k^G(v)\), \(d_{k^+}^G(v)\), and \(d_{k^-}^G(v)\) to denote the number of k-vertices, \(k^{+}\)-vertices, or \(k^{-}\)-vertices adjacent to v, respectively.

A proper edge k-coloring of G is a mapping  \(\phi \): \(E(G)\rightarrow \{1, 2,\ldots , k\}\) such that \(\phi (e)\ne \phi (e') \) for any two adjacent edges e and \(e'\). The chromatic index, denoted by \(\chi '(G)\), is the smallest integer k such that G has a proper edge k-coloring. Let \(C_\phi (v)= \{\phi (xv)|xv\in E(G)\}\) denote the set of colors assigned to edges incident with vertex v. We call two adjacent vertices u and vconflict under an edge coloring \(\phi \) if \(C_\phi (u)=C_\phi (v)\). A proper edge k-coloring \(\phi \) of G is adjacent vertex distinguishing, or a k-avd-coloring in short, if any two adjacent vertices do not conflict. That is \(C_{\phi }(u)\ne C_{\phi }(v)\) for any \(uv\in E(G)\). The adjacent vertex distinguishing chromatic index, denoted by \(\chi _a'(G)\), is the smallest integer k such that G has a k-avd-coloring.

Note that if G contains an isolated edge, then G has no avd-coloring. A graph G is normal if it has no isolated edges. The adjacent vertex distinguishing coloring is variously known as adjacent strong edge coloring (Zhang et al. [11]) and 1-strong edge colorings (Akbari et al. [1]). It follows from the definition that \(\chi _a'(G)\ge \chi '(G)\ge \Delta (G)\). The well-known Vizing’s theorem (1964) asserts that \(\chi '(G)\le \Delta (G)+{1}\) for any simple graph G. However, there exists infinitely many graphs G such that \(\chi _a'(G)>\Delta (G)+{1}\). For example, for a cycle \(C_n\) of length n, \(\chi _a'(C_n)=4=\Delta (C_n)+{2}\) if \(n\not \equiv 0(\text{ mod }\ 3)\) and \(n\ne 5\), and \(\chi _a'(C_5)=5=\Delta (C_5)+{3}\).

Zhang et al. [11] completely determined \(\chi _a'(G)\) for some special graphs. Based on these results, they proposed the following conjecture.

Conjecture 1

Let G be a normal connected graph and \(G\ne C_5\). Then, \(\chi _a'(G)\le \Delta (G)+{2}\).

Balister et al. [2] confirmed Conjecture 1 for all bipartite graphs and graphs with \(\Delta (G)\le 3\) and proved that \(\chi _a'(G)\le \Delta (G)+O(\log k)\), where k is the vertex chromatic number of the normal graph G. Hatami [4] showed that every normal graph G with \(\Delta (G)\ge 10^{20}\) has \(\chi _a'(G)\le \Delta (G)+300\) by the probabilistic method. Wang and Wang [8] characterized \(\chi _a'(G)\) for \(K_4\)-minor free graphs G with \(\Delta (G)\ge 5\). For more results about the adjacent vertex distinguishing edge coloring with the small maximum degree, the readers may refer to [5, 7, 9].

Bu et al. [3] confirmed Conjecture 1 for planar graphs with girth at least 6. Yan et al. [10] showed that if G is a normal planar graph with girth \(g(G)\ge 5\) and \(G\ne C_5\), then \(\chi _a'(G)\le \Delta (G)+2\). Huang et al. [6] showed that if G is a connected planar graph without 3-cycles and maximum degree at least 12, then \(\chi _a'(G)\le \Delta (G)+1\). In this paper, we will prove the following result:

Theorem 1

Let G be a normal planar graph without 4-cycles. Then, \(\chi _a'(G)\le \max \{9,\Delta (G)+1\}\).

To obtain our main result, we need to introduce some notations. Let G be a plane graph. For \(k=2\) or 3, a k-vertex v is called bad in G if \(d_k^{G}(v)=1\); it is called good in G if \(d_k^{G}(v)=0\). We use \(d_{kb}^{G}(v)\) or \(d_{kg}^{G}(v)\) to denote the number of bad k-vertices or good k-vertices in G adjacent to a vertex v, respectively. Let \(t_G(v)\) denote the number of 3-faces incident with a vertex v. An edge uv is called light k-edge if \(d_G(u)=d_G(v)=k\) with \(k=2\) or 3. For \(f\in F(G)\), we use b(f) to denote the boundary walk of f, and \(f=[v_0v_1\cdots v_{k-1}]\) if \(v_0\), \(v_1\),..., \(v_{k-1}\) are the boundary vertices of f in cyclic order. Let \(d_k^G(f)\) be the number of k-vertices incident with f. Similarly, we can define \(d_{k^+}^G(f)\), \(d_{k^-}^G(f)\), \(d_{kg}^G(f)\), and \(d_{kb}^G(f)\). A 3-face \([v_0v_1v_2]\) is called a \((d_0,d_1,d_2)\)-face if \(d_G(v_i)=d_i\) for \(i=0,1,2\). A 5-face f is called bad if \(d_{2b}^G(f)=2\). A 5-face f is called special if \(d_{2g}^G(f)=2\). A 6-face f is called special if \(d_2^G(f)=3\) and \(d_{2b}^G(f)=2\). A 7-face is called special if \(d_{2b}^G(f)=4\).

2 Proof of Theorem 1

Theorem 1 will be proved by contradiction. For any graph G, set \(n_i(G)=|\{v\in V(G)|d_G(v)=i\}|\) for \(i=1,2,\ldots , \Delta (G)\). We denote that a graph H is “smaller” than G if \((|E(H)|,n_t(H),n_{t-1}(H),\ldots ,n_2(H),n_1(H))\) precedes \((|E(G)|,n_t(G),n_{t-1}(G),\ldots ,n_2(G),n_1(G))\) with respect to the standard lexicographic order, where \(t=\max \{\Delta (G),\Delta (H)\}\). Let G be the “smallest” counterexample of Theorem 1. Obviously, G is connected. First, we shall prove the following claims on the characterizations of G. Then, we will derive a contradiction using the discharging method.

For two positive integers pq with \(p<q\), we use [pq] to denote the set of all integers between p and q (including p and q). Let \(T(G)=\max \{9,\Delta (G)+1\}\). Then, \(T(G)\ge 9\). We may assume that \(\Delta (G)\ge 4\) by the result in [2].

Claim 1

There is no edge uv with \(d_G(u)=1\) and \(d_G(v)\le 5\).

Proof

Assume to the contrary that G contains an edge uv with \(d_G(u)=1\) and \(d_G(v)\le 5\). Let \(H=G-u\). Then, H is connected and hence normal as \(\Delta (G)\ge 4\). By the smallest of G, H has a T(G)-avd-coloring \(\phi \) with the color set C. Since \(|C|\ge 9\) and \(d_G(v)\le 5\), we can color uv with a color in \(C\setminus C_{\phi }(v)\), such that v does not conflict with its neighbors. \(\square \)

Remark 1

Claim 1 implies that if e is an arbitrary edge of G, then \(H=G-e\) is a normal subgraph, and therefore, \(\chi '_a(H)\le T(H)\le T(G)\) by the smallest of G.

Claim 2

Let v be a k-vertex with \(2\le k\le 4\). Then, \(d_k^G(v)\le 1\).

Proof

We only need to proof the case that \(k=4\). (For \(k=2\) and 3, the proof is similar and simple.) Assume to the contrary that \(d_4^G(v)\ge 2\). Let \(v_1, v_2,v_3, v_4\) be the neighbors of v with \(d_G(v_1)=d_G(v_2)=4\). Let \(u_i\ (i=1,2,3)\) be the neighbors of \(v_1\) other than v; \(w_i\ (i=1,2,3)\) be the neighbors of \(v_2\) other than v. Then, \(G-vv_1\) has a T(G)-avd-coloring \(\phi \) by Remark 1. Let \(\phi (vv_i)=i-1\) for \(i\in [2,4]\), and \(\phi (v_1u_i)=a_i\) for \(i=1,2,3\). The proof of this claim is divided into four cases, depending on the value of \(s=|\{a_1,a_2,a_3\}\setminus \{1,2,3\}|\).

  1. (i)

    Suppose that \(s=1\), say \(a_1=1\), \(a_2=2\), and \(a_3=4\). Suppose that \(vv_1\) cannot be colored without creating a conflict. That is, colors \(\{5,6,\ldots ,9\}\) must be occurred on the vertices \(\{v_2,v_3,v_4,u_1,u_2,u_3\}\). Without loss of generality, we may assume that \(C_\phi (v_i)=\{1,2,3,i+3\}\) for \(i=2,3,4\), \(C_\phi (u_1)=\{1,2,4,8\}\), and \(\{1,2,4,9\}\in \{C_{\phi }(u_2), C_{\phi }(u_3)\}\). Note that \(vv_2\) can be recolored with a color \(\alpha \in \{6,7,8,9\}\) such that \(v_2\) does not conflict with its neighbors. Let \(\beta \in \{6,7\}\setminus \{\alpha \}\). If \(\{1,2,4,\beta \}\notin \{C_{\phi }(u_2), C_{\phi }(u_3)\}\), then we can recolor \(vv_2\) with \(\alpha \) and color \(vv_1\) with \(\beta \). If \(\{1,2,4,\beta \}\in \{C_{\phi }(u_2), C_{\phi }(u_3)\}\), then we can recolor \(v_1u_1\) with a color \(\alpha _1\in \{3,5,6,7\}\) such that \(u_1\) does not conflict with its neighbors, and color \(vv_1\) with 9.

  2. (ii)

    Suppose that \(s=0\). If \(vv_2\) can be recolored without creating a conflict, turn to the case that \(s=1\). So we may assume that \(vv_2\) cannot be recolored without creating a conflict, say \(\phi (v_2w_i)=i+3\) and \(C_{\phi }(w_i)=\{4,5,6,i+6\}\) for \(i=1,2,3\). We can recolor \(v_2w_1\) with a color \(\alpha \in \{2,3,8,9\}\) such that \(w_1\) does not conflict with its neighbors, and recolor \(vv_2\) with a color in \(\{8,9\}\setminus \{\alpha \}\), turn to the case that \(s=1\).

  3. (iii)

    Suppose that \(s=2\), say \(a_1\in [1,3]\), \(a_2=4\), and \(a_3=5\). Suppose that \(vv_1\) cannot be colored without creating a conflict. That is, \(C_\phi (v_i)=\{1,2,3,i+4\}\) for \(i=2,3,4\), and \(\{a_1,4,5,9\}\in \{C_{\phi }(u_i)| i=1,2,3\}\); or \(C_\phi (v_i)=\{1,2,3,i+4\}\) for \(i=2,3\), and \(\{a_1,4,5,8\},\{a_1,4,5,9\}\in \{C_{\phi }(u_i)| i=1,2,3\}\).

  • Suppose that \(C_\phi (v_i)=\{1,2,3,i+4\}\) for \(i=2,3,4\), and \(\{a_1,4,5,9\}\in \{C_{\phi }(u_i)| i=1,2,3\}\). If \(\{a_1,4,5,7\}\notin \{C_{\phi }(u_i)|i=1,2,3\}\), then we can recolor \(vv_2\) with a color \(\alpha \in \{4,5,8,9\}\) such that \(v_2\) does not conflict with its neighbors, and color \(vv_1\) with 7. So we may assume that \(\{a_1,4,5,7\}\in \{C_{\phi }(u_i)|i=1,2,3\}\). Similarly, we can show that \(\{a_1,4,5,8\}\in \{C_{\phi }(u_i)|i=1,2,3\}\). Now, we can recolor \(vv_3\) with a color \(\beta \in \{4,5,8,9\}\) such that \(v_3\) does not conflict with its neighbors, and color \(vv_1\) with 6.

  • Suppose that \(C_\phi (v_i)=\{1,2,3,i+4\}\) for \(i=2,3\), and \(\{a_1,4,5,8\},\{a_1,4,5,9\}\in \{C_{\phi }(u_i)| i=1,2,3\}\). Note that \(\{a_1,4,5,6\}\) and \(\{a_1,4,5,7\}\) is not the color set of \(u_1,u_2\), and \(u_3\); otherwise, the proof is similar as the above arguments. It is observed that \(vv_2\) can be recolored with a color \(\alpha \in \{4,5,8,9\}\) such that \(v_2\) does not conflict with its neighbors. If \(\{2,3,\alpha ,7\}\ne C_{\phi }(v_4)\), then we can recolor \(vv_2\) with \(\alpha \) and color \(vv_1\) with 7. So we may assume that \(\{2,3,\alpha ,7\}=C_{\phi }(v_4)\). Then, we can recolor \(vv_3\) with a color \(\beta \in \{4,5,8,9\}\) such that \(v_3\) does not conflict with its neighbors, and color \(vv_1\) with 6.

  1. (iv)

    Suppose that \(s=3\), say \(a_1=4\), \(a_2=5\), and \(a_3=6\). Suppose that \(vv_1\) cannot be colored without creating a conflict. That is, \(C_\phi (v_i)=\{1,2,3,i+5\}\) for \(i=2,3\), and \(C_{\phi }(v_4)=\{1,2,3,9\}\) or \(\{4,5,6,9\}\in \{C_{\phi }(u_i)|i=1,2,3\}\).

  • Suppose that \(C_{\phi }(v_4)=\{1,2,3,9\}\). If \(\{4,5,6,9\}\notin \{C_{\phi }(u_i)|i=1,2,3\}\), then we can recolor \(vv_2\) with a color in \(\{4,5,6,8\}\) such that \(v_2\) does not conflict with its neighbors, and color \(vv_1\) with 9. So we may assume that \(\{4,5,6,9\}=\{C_{\phi }(u_1)\}\). Similarly, we can show that \(\{4,5,6,7\},\{4,5,6,8\}\in \{C_{\phi }(u_i)|i=2,3\}\). We can recolor \(vv_2\) with a color in \(\{4,5,6,9\}\) such that \(v_2\) does not conflict with its neighbors, recolor \(v_1u_1\) with a color in \(\{1,2,3,7\}\) such that \(u_1\) does not conflict with its neighbors, and color \(vv_1\) with 8.

  • Suppose that \(\{4,5,6,9\}\in \{C_{\phi }(u_i)|i=1,2,3\}\), say \(\{4,5,6,9\}=C_{\phi }(u_1)\). Note that \(\{\{4,5,6,7\},\{4,5,6,8\}\}\ne \{C_{\phi }(u_2), C_{\phi }(u_3)\}\); otherwise, the proof is similar as the above arguments. It is observed that \(vv_i\) can be recolored with a color \(\alpha _i\in \{4,5,6,9\}\) such that \(v_i\) does not conflict with its neighbors for \(i=2,3\). If \(\{4,5,6,8\}\) or \(\{2,3,\alpha _2,8\}\) is not the color set of \(u_2,u_3\), and \(v_4\), then we can recolor \(vv_2\) with \(\alpha _2\) and color \(vv_1\) with 8. If \(\{4,5,6,7\}\) or \(\{1,3,\alpha _3,7\}\) is not the color set of \(u_2,u_3\), and \(v_4\), then we can recolor \(vv_3\) with \(\alpha _3\) and color \(vv_1\) with 7. Hence, by the above notation, we may assume that \(C_{\phi }(u_2)=\{4,5,6,8\}\) and \(C_{\phi }(v_4)=\{1,3,\alpha _3,7\}\). Then, \(C_{\phi }(u_3)\ne \{4,5,6,7\}\). We can recolor \(vv_4\) with a color in \(\{4,5,6,8,9\}\setminus \{\alpha _3\}\) such that \(v_4\) does not conflict with its neighbors, and color \(vv_1\) with 7.

\(\square \)

Claim 3

Let v be a k-vertex of G with \(k\le 4\). Then, \(d_i^G(v)=0\), where \(i<k\).

Proof

Assume to the contrary that G contains a k-vertex v adjacent to a i-vertex u with \(i<k\le 4\). Then, \(G-vu\) has a T(G)-avd-coloring \(\phi \) using the color set \(C=[1,T(G)]\) by Remark 1. By Claim 2, \(d_k^G(v)\le 1\) and \(d_i^G(u)\le 1\). Note that \(|C_{\phi }(v)|\le k-1\le 3\) and \(|C_{\phi }(u)|\le i-1\le 2\). Since \(T(G)\ge 9\), we can color uv with a color in \([1,T(G)]\setminus (C_{\phi }(v)\cup C_{\phi }(u))\) such that neither v nor u conflicts with its neighbors. \(\square \)

Claim 4

For each vertex \(v\in V(G)\), if \(d_1^G(v)\ge i\), then \(d_{i+1}^G(v)=0\) for \(i=1,2\).

Proof

Let \(v_1, v_2,\ldots , v_k\) be the neighbors of v with \(d_G(v_1)=1\).

  1. (i)

    First, we show that \(d_2^G(v)=0\) when \(d_1^G(v)\ge 1\). Assume to the contrary that \(d_2^G(v)\ge 1\), say \(d_G(v_2)=2\) with \(N_G(v_2)=\{u,v\}\). Let H be the graph obtained from \(G-v_2u\) by adding a new vertex \(v_2'\) and a new edge \(uv_2'\). Then, \(|E(H)|=|E(G)|\) and \(n_2(H)<n_2(G)\), i.e., H is smaller than G. Hence, H has a T(G)-avd-coloring \(\phi \) by the smallest of G. Now, we need to find the T(G)-avd-coloring \(\psi \) of G. Let \(\psi (e)=\phi (e)\) for \(e\in (E(H)\cap E(G))\setminus \{vv_1,vv_2\}\).

Suppose that \(d_G(u)\ne 2\). Let \(\psi (v_2u)=\phi (v_2'u)\), \(\psi (vv_2)\in \{\phi (vv_1),\phi (vv_2)\}\setminus \{\psi (v_2u)\}\), and \(\psi (vv_1)\in \{\phi (vv_1),\phi (vv_2)\}\setminus \{\psi (vv_2)\}\). Suppose that \(d_G(u)=2\) with \(N_G(u)=\{v_2,u_1\}\). Let \(\psi (vv_2)\in \{\phi (vv_1),\phi (vv_2)\}\setminus \{\psi (uu_1)\}\), and \(\psi (vv_1)\in \{\phi (vv_1),\phi (vv_2)\}\setminus \{\psi (vv_2)\}\). Then, we can color \(v_2u\) properly.

  1. (ii)

    Next, we show that \(d_3^G(v)=0\) when \(d_1^G(v)\ge 2\). Suppose that \(d_G(v_2)=1\). Assume to the contrary that \(d_3^G(v)\ge 1\), say \(d_G(v_3)=3\) with \(N_G(v_3)=\{u_1,u_2,v\}\). Let H be the graph obtained from \(G-\{v_3u_1,v_3u_2\}\) by adding a new vertex \(v_3'\) and two new edges \(v_3'u_1\) and \(v_3'u_2\). Then, \(|E(H)|=|E(G)|\) and \(n_3(H)<n_3(G)\), i.e., H is smaller than G. Hence, H has a T(G)-avd-coloring \(\phi \) by the smallest of G. Now, we need to find the T(G)-avd-coloring \(\psi \) of G. Let \(\psi (e)=\phi (e)\) for \(e\in (E(H)\cap E(G))\setminus \{vv_1,vv_2,vv_3\}\).

Suppose that \(d_G(u_1)\ne 3\) and \(d_G(u_2)\ne 3\). Let \(\psi (v_3u_i)=\phi (v_3'u_i)\) for \(i=1,2\), \(\psi (vv_3)\in \{\phi (vv_1),\phi (vv_2),\phi (vv_3)\}\setminus \{\psi (v_3u_1),\psi (v_3u_2)\}\), \(\{\psi (vv_1),\psi (vv_2)\}=\{\phi (vv_1),\phi (vv_2),\phi (vv_3)\}\setminus \{\psi (vv_3)\}\). By Claim 2, \(d_3^G(v_3)\le 1\). Suppose that \(d_G(u_1)=3\) and \(d_G(u_2)\ne 3\). We remove the color of \(v_3'u_1\) and set \(\alpha \in C_{\phi }(u_1)\setminus \{\phi (v_3'u_2)\}\). Let \(\psi (v_3u_2)=\phi (v_3'u_2)\), \(\psi (vv_3)\in \{\phi (vv_1),\phi (vv_2),\phi (vv_3)\}\setminus \{\psi (v_3u_2), \alpha \}\), \(\{\psi (vv_1),\psi (vv_2)\}=\{\phi (vv_1),\phi (vv_2),\phi (vv_3)\}\setminus \{\psi (vv_3)\}\). Then, we can color \(v_3u_1\) properly. \(\square \)

Claim 5

There is no 5-cycle incident with three 2-vertices in G; and there is no 5-cycle incident with two bad 2-vertices and a bad 3-vertex in G simultaneously.

Proof

Let \(C=v_0v_1v_2v_3v_4v_0\) be the 5-cycle of G with \(d_G(v_0)=d_G(v_1)=2\). Then, \(G-v_0v_1\) has a T(G)-avd-coloring \(\phi \).

First, we show that G does not contain any 5-cycle incident with three 2-vertices. Assume to the contrary that C is incident with three 2-vertices. By Claim 2, we may assume that \(d_G(v_3)=2\). If \(\phi (v_0v_4)\ne \phi (v_1v_2)\), we can color \(v_0v_1\) properly to obtain a T(G)-avd-coloring of G, a contradiction. Assume that \(\phi (v_0v_4)=\phi (v_1v_2)\). Then, \(\phi (v_3v_4)\ne \phi (v_1v_2)\). We can exchange the colors of \(v_1v_2\) and \(v_2v_3\) and color \(v_0v_1\) properly.

Next, we show that G does not contain any 5-cycle incident with two bad 2-vertices and a bad 3-vertex simultaneously. Assume to the contrary that C is incident with bad 3-vertex. By Claim 3, \(d_G(v_3)=3\) with \(d_3^G(v_3)=1\). Let x be the neighbor of \(v_3\) other than \(v_2\) and \(v_4\). Then, \(d_G(x)=3\). Let \(x_1\) and \(x_2\) be the neighbors of x other than \(v_3\). If \(\phi (v_0v_4)\ne \phi (v_1v_2)\), we can color \(v_0v_1\) properly to obtain a T(G)-avd-coloring of G, a contradiction. So we may assume that \(\phi (v_0v_4)=\phi (v_1v_2)=1\). Then, we can assume that \(\phi (v_2v_3)=2\) and \(\phi (v_3v_4)=3\). We remove the color of \(v_3x\). Note that \(\{\{1,2\},\{1,3\}\}\setminus \{\phi (xx_1),\phi (xx_2)\}\ne \emptyset \), say \(\{\phi (xx_1),\phi (xx_2)\}\ne \{1,2\}\). Then, we can exchange the colors of \(v_0v_4\) and \(v_3v_4\) and color \(v_0v_1\) and \(v_3x\) properly. \(\square \)

Claim 6

G does not contain any configuration as shown in Fig. 1.

Proof

Assume to the contrary that G contains \(G_i\) as a subgraph as shown in Fig. 1, where \(i\in [1,7]\). We will use the labels of graphs as shown in Fig. 1. Note that we first remove the colors of all light 2-edges and light 3-edges. After recoloring some edges of \(G_i\) for \(i\in [1,7]\) so that the color sets of each pair of end vertices of light edges are distinct, then we can color them properly to deduce a T(G)-avd-coloring of G.

Case 1G contains \(G_1\) as a subgraph.

Let H be the graph obtained from \(G-vv_1\) by adding a new vertex \(v_1'\) and a new edge \(vv_1'\). Then, \(|E(H)|=|E(G)|\) and \(n_2(H)<n_2(G)\), i.e., H is smaller than G. Thus, H has a T(G)-avd-coloring \(\phi \). Now, we need to construct a T(G)-avd-coloring \(\psi \) of G. First, let \(\psi (e)=\phi (e)\) for \(e\in E(H)\cap E(G)\).

Let \(\alpha =\phi (v_1u_1)\) when \(d_G(u_1)\ne 2\). When \(d_G(u_1)=2\), let x be the second neighbor of \(u_1\) other than \(v_1\). Then, \(d_G(x)\ne 2\) by Claim 2. We remove the color of \(v_1u_1\) and let \(\alpha =\phi (u_1x)\). If \(\phi (vv_1')\ne \alpha \), then set \(\psi (vv_1)=\phi (vv_1')\). So we may assume that \(\phi (vv_1')= \alpha \). If \(\phi (v_2u_2)\ne \alpha \), then we can recolor \(vv_2\) with \(\alpha \) and color \(vv_1\) with \(\phi (vv_2)\). If \(\phi (v_2u_2)=\alpha \), then we can exchange the color of \(vu_2\) and \(v_2u_2\) and color \(vv_1\) with \(\psi (vu_2)\).

Fig. 1
figure 1

The configurations used in Claim 6. In this figure solid vertices have no other vertices adjacent to them other than those showed. In \(G_2\), we further assume that \(d_G(v_3)\le 2\)

Case 2G contains \(G_2\) or \(G_3\) as a subgraph.

Then, \(G-v_1v_2\) admits a T(G)-avd-coloring \(\phi \) by Remark 1. Suppose that \(v_1v_2\) cannot be colored without creating a conflict. That is, \(\{\phi (vv_1),\phi (v_1u_1)\}=\{\phi (vv_2),\phi (v_2u_2)\}\).

  1. (i)

    Suppose that G contains \(G_2\) as a subgraph. If \(d_G(v_3)=1\) or \(v_3\) is a good 2-vertex in G, then we can recolor \(vv_3\) with a color in \(\{\phi (vv_1),\phi (vv_2)\}\setminus \{\phi (v_3u_3)\}\), say \(\phi (vv_1)\), and color \(vv_1\) with \(\phi (vv_3)\). So we may assume that \(v_3\) is a bad 2-vertex in G. Let \(N_G(v_3)=\{v,u_3\}\) and \(N_G(u_3)=\{v_3,u_3'\}\). Then, \(d_G(u_3')\ne 2\) by Claim 2. We can recolor \(vv_3\) with a color in \(\{\phi (vv_1),\phi (vv_2)\}\setminus \{\phi (u_3u_3')\}\), say \(\phi (vv_1)\), and color \(vv_1\) with \(\phi (vv_3)\).

  2. (ii)

    Suppose that G contains \(G_3\) as a subgraph. Note that \(\{\phi (vv_3),\phi (v_3u_3)\}\ne \{\phi (vv_4),\phi (v_4u_4)\}\), say \(\phi (v_3u_3)\ne \phi (vv_4)\). We can recolor \(vv_3\) with a color in \(\{\phi (vv_1),\phi (vv_2)\}\setminus \{\phi (v_3u_3)\}\), say \(\phi (vv_1)\). Then, we can recolor \(vv_1\) with \(\phi (vv_3)\) and color \(v_1v_2\), \(v_3v_4\) properly.

Case 3G contains \(G_i\) as a subgraph, where \(i\in [4,7]\).

Then, \(G-v_1v_2\) admits a T(G)-avd-coloring \(\phi \) by Remark 1. Suppose that \(v_1v_2\) cannot be colored without creating a conflict. Without loss of generality, we may assume \(\phi (v_2v_3)= \phi (v_0v_1)=1\).

  1. (i)

    G contains \(G_4\) as a subgraph. We can exchange the color of \(v_2v_3\) and \(v_3v_4\) and color two light edges \(v_1v_2\) and \(v_4v_5\) properly.

  2. (ii)

    G contains \(G_5\) as a subgraph. If \(\phi (v_4v_5)\ne 1\), then we can exchange the colors of \(v_0v_1\) and \(v_0v_5\). So we may assume that \(\phi (v_4v_5)=1\). If \(\phi (v_3v_4)\ne \phi (v_0v_5)\), then we can recolor \(v_2v_3\), \(v_3v_4\), and \(v_4v_5\) with \(\phi (v_3v_4)\), 1, \(\phi (v_3v_4)\), respectively. Hence, assume that \(\phi (v_3v_4)=\phi (v_0v_5)=2\). Let \(\alpha =\phi (x_1y_1)\) when \(d_G(y_1)\ne 2\). When \(d_G(y_1)=2\), let \(y_1'\) be the second neighbor of \(y_1\) other than \(x_1\). Then, \(d_G(y_1')\ne 2\) by Claim 2. We remove the color of \(x_1y_1\) and let \(\alpha =\phi (y_1y_1')\). If \(\alpha \ne 1\), then we can recolor \(v_3x_1\), \(v_2v_3\) with 1, \(\phi (v_3x_1)\), respectively. If \(\alpha =1\), then we can recolor \(v_3v_2\), \(v_3v_4\), \(v_4v_5\), \(v_5v_0\), \(v_0v_1\), \(v_3x_1\) with \(\phi (v_3x_1)\), 1, 2, 1, 2, 2, respectively.

  3. (iii)

    G contains \(G_6\) as a subgraph. We remove the color of \(v_4v_5\). If \(\phi (v_5v_6)\ne 1\), then we can exchange the colors of \(v_2v_3\) and \(v_3v_4\). So we may assume that \(\phi (v_5v_6)=1\). If \(\phi (v_3v_4)\ne \phi (v_0v_6)\), then we can recolor \(v_0v_1\), \(v_0v_6\), and \(v_5v_6\) with \(\phi (v_0v_6)\), 1, \(\phi (v_0v_6)\), respectively. Hence, assume that \(\phi (v_3v_4)=\phi (v_0v_6)=2\). Let \(\alpha =\phi (x_1y_1)\) when \(d_G(y_1)\ne 2\). When \(d_G(y_1)=2\), let \(y_1'\) be the second neighbor of \(y_1\) other than \(x_1\). Then, \(d_G(y_1')\ne 2\) by Claim 2. We remove the color of \(x_1y_1\) and let \(\alpha =\phi (y_1y_1')\). If \(\alpha \ne 1\), then we can recolor \(v_3x_1\), \(v_2v_3\) with 1, \(\phi (v_3x_1)\), respectively. Therefore, \(\alpha =1\). We can recolor \(v_3v_4\), \(v_5v_6\), \(v_6v_0\), \(v_0v_1\), \(v_3x_1\), with \(\phi (v_3x_1)\), 2, 1, 2, 2, respectively.

  4. (iv)

    G contains \(G_7\) as a subgraph. We remove the color of \(v_5v_6\). If \(\phi (v_4v_5)\ne 1\), then we can exchange the colors of \(v_0v_1\) and \(v_0v_6\). So we may assume that \(\phi (v_4v_5)=1\). If \(\phi (v_3v_4)\ne \phi (v_0v_6)\), then we can recolor \(v_2v_3\), \(v_3v_4\), and \(v_4v_5\) with \(\phi (v_3v_4)\), 1, \(\phi (v_3v_4)\), respectively. Hence, assume that \(\phi (v_3v_4)=\phi (v_0v_6)=2\). Let \(\alpha =\phi (x_1y_1)\) when \(d_G(y_1)\ne 2\). When \(d_G(y_1)=2\), let \(y_1'\) be the second neighbor of \(y_1\) other than \(x_1\). Then, \(d_G(y_1')\ne 2\) by Claim 2. We remove the color of \(x_1y_1\) and let \(\alpha =\phi (y_1y_1')\). If \(\alpha \ne 1\), then we can recolor \(v_3x_1\), \(v_2v_3\) with 1, \(\phi (v_3x_1)\), respectively. Therefore, \(\alpha =1\). We can recolor \(v_2v_3\), \(v_3v_4\), \(v_4v_5\), \(v_6v_0\), \(v_0v_1\), \(v_3x_1\) with \(\phi (v_3x_1)\), 1, 2, 1, 2, 2, respectively.

\(\square \)

Claim 7

Every light 2-edge must be incident with some 5-cycle.

Proof

Assume to the contrary that there exists some light 2-edge uv which is not incident with any 5-cycle. Let \(N_G(u)=\{v,u'\}\) and \(N_G(v)=\{u,v'\}\). If \(u'=v'\), then \(G-uv\) admits a T(G)-avd-coloring \(\phi \). Note that \(\phi (u'u) \ne \phi (v'v)\). We can color uv properly to obtain a T(G)-avd-coloring of G, a contradiction. If \(u'\ne v'\), then it must be incident with two \(6^+\)-cycle. Let H be the graph obtained by contract edge uv and denote the new vertex by w. Then, H is a planar graph without 4-cycles. Thus, H has a T(G)-avd-coloring \(\phi \). Now, we can construct the T(G)-avd-color \(\psi \) of G: \(\psi (e)=\phi (e)\) for \(e\in E(H)\cap E(G)\), \(\psi (uu')=\phi (wu')\), \(\psi (vv')=\phi (wv')\), and color uv properly. \(\square \)

By the proof of Claim 7, we can deduce that every light 2-edge is not incident with 3-cycle.

Claim 8

Let v be a 5-vertex of G.

  1. (1)

    \(d_{3^-}^G(v)\le 1\);

  2. (2)

    v is not adjacent to bad 2-vertex;

  3. (3)

    If \(d_{3^-}^G(v)=1\), then v is not incident with any 3-cycle.

Proof

Let \( v_1,v_2,\ldots ,v_5\) be the neighbors of v with \(d_G(v_1)=k\le 3\). Let \(u_1,u_2,\ldots ,v_{k-1}\) be the neighbors of \(v_1\) other than v. Then, \(G-vv_1\) has a T(G)-avd-coloring \(\phi \) with color set \(C=[1,T(G)]\) by Remark 1. Note that \(T(G)\ge 9\). Suppose that \(\phi (vv_i)=i-1\) for \(i\in [2,5]\).

  1. (1)

    Assume to the contrary that \(d_{3^-}^G(v)\ge 2\). Then, \(d_5^G(v)\le 3\). Note that \(k\ge 2\) by Claim 1.

Suppose that \(k=2\). Let \(\alpha =\phi (v_1u_1)\) if \(d_G(u_1)\ne 2\). When \(d_G(u_1)=2\), \(N_G(u_1)=\{y,v_1\}\). We can recolor \(v_1u_1\) with a color in \([1,4]\setminus C_{\phi }(u_1)\), and let \(\alpha =\phi (u_1y)\). Then, we can color \(vv_1\) with a color in \([5,9]\setminus \{\alpha \}\) such that v does not conflict with its neighbors.

Suppose that \(k=3\). By Claim 2, \(d_3^G(v_1)\le 1\). Let \(\alpha _i=\phi (v_1u_i)\) if \(d_3^G(v_1)=0\). When \(d_3^G(v_1)=1\), let \(d_G(u_1)=3\). We can recolor \(v_1u_1\) with a color in \([1,4]\setminus C_{\phi }(u_1)\) and set \(\alpha _1\in C_{\phi }(u_1)\setminus C_{\phi }(v_1)\), \(\alpha _2=\phi (v_1u_2)\). Without loss of generality, assume that \(\alpha _1\), \(\alpha _2\in [1,6]\). Firstly, \(vv_1\) can be colored with any color in \(\{7,8,9\}\). Secondly, we can recolor \(vv_2\) with a color \(\beta \in [5,9]\setminus C_\phi (v_2)\) such that \(v_2\) does not conflict its neighbors, and color \(vv_1\) with a color in \([7,9]\backslash \{\beta \}\). Hence, we have at least 4 different ways to recolor or color the edges incident to v. Note that v has at most 3 conflict vertices. So we can extend \(\phi \) to G, a contradiction.

  1. (2)

    Assume to the contrary that v is adjacent to a bad 2-vertex \(v_1\).

Let \(N_G(v_1)=\{u_1,v\}\) and \(N_G(u_1)=\{v_1, x\}\). By Claim 7, \(v_1u_1\) is incident with some 5-cycle, say \(xv_2\in E(G)\). Let \(\alpha =\phi (u_1x)\). If \(\alpha \in [1,4]\), then we can color \(vv_1\) with a color in [5, 9] such that v does not conflict with its neighbors. So we may assume that \(\alpha =5\). Suppose that \(vv_1\) cannot be colored with a color in [6,9] so that \(C_{\phi }(v)\ne C_{\phi }(v_i)\) for \(i=2,3,4,5\), say \(C_\phi (v_i)=\{1,2,3,4,i+4\}\). If \(\{2,3,4,6,a\}\notin \{C_{\phi }(y)| yv_2\in E(G)\ \mathrm{and} ~y\ne x\}\) for some \(a\in [7,9]\), then we can recolor \(vv_2\) with a and color \(vv_1\) with a color in \([7,9]\setminus \{a\}\). So we may assume that \(\{\{2,3,4,6,a\}|a=7,8,9\}=\{C_{\phi }(y)| yv_2\in E(G)\ \mathrm{and} ~y\ne x\}\). We exchange the colors of \(u_1x\) and \(v_2x\) and recolor \(vv_2\) with a color \(b\in [7,9]\) such that \(v_2\) does not conflict with x. Then, we can color \(vv_1\) with a color in \([7,9]\setminus \{b\}\).

  1. (3)

    Assume to the contrary that v is incident with some 3-cycle when \(d_{3^-}^G(v)=1\).

Case 1 Suppose that \(v_1\) is incident with some 3-cycle, say \(v_1v_2\in E(G)\).

  1. (i)

    Suppose that \(d_G(v_1)=2\). Let \(\phi (v_1v_2)=\alpha \). If \(\alpha \subseteq [1,4]\), then we can color \(vv_1\) with a color in [5, 9] such that v does not conflict with its neighbors. So we may assume that \(\alpha =5\). Then, we can color \(vv_1\) with a color in [6, 9] such that v does not conflict with its neighbors.

  2. (ii)

    Suppose that \(d_G(v_1)=3\) with \(N_G(v_1)=\{u_1,v,v_2\}\). Recall that \(d_3^G(v_1)\le 1\) by Claim 2. When \(d_3^G(v_1)=0\), let \(\phi (v_1u_1)=a_1\) and \(\phi (v_1v_2)=a_2\). When \(d_3^G(v_1)=1\), \(d_G(u_1)=3\) by Claim 8 (1). We can recolor \(v_1u_1\) with a color in \([1,4]\setminus C_{\phi }(u_1)\), and let \(a_1\in C_{\phi }(u_1)\setminus \{\phi (v_1v_2)\}\) and \(a_2=\phi (v_1v_2)\).

  • Suppose that \(a_1\), \(a_2\in [1,4]\). Then, we can color \(vv_1\) with a color in [5,9] such that v does not conflict with its neighbors.

  • Suppose that \(a_1\in [1,4]\), \(a_2=5\). Then, we can color \(vv_1\) with a color in [6,9] such that \(C_{\phi }(v)\ne C_{\phi }(v_i)\) for \(i=3,4,5\). Note that \(a_2=5\), then \(C_{\phi }(v)\ne C_{\phi }(v_2)\).

  • Suppose that \(a_1=5\) and \(a_2\in [2,4]\), say \(a_2=2\). Suppose that \(vv_1\) cannot be colored without creating a conflict. That is, \(C_\phi (v_i)=\{1,2,3,4,4+i\}\) for \(i=2,3,4,5\). We can recolor \(vv_2\) with a color in \(\alpha \in \{5,7,8,9\}\) such that \(v_2\) does not conflict with its neighbors. Then, we can color \(vv_1\) with a color in \(\{7,8,9\}\backslash \{\alpha \}\).

  • Suppose that \(a_1=5, a_2=6\). Suppose that \(vv_1\) cannot be colored without creating a conflict. That is, \(C_\phi (v_i)=\{1,2,3,4,4+i\}\) for \(i=3,4,5\). Then, we can exchange the colors of \(vv_2\) and \(v_1v_2\) and color \(vv_1\) with a color in \(\{7,8,9\}\) such that \(C_{\phi }(v)\ne C_{\phi }(v_2)\).

Case 2 Suppose that \(v_1\) is not incident with any 3-cycle, say \(v_2v_3\in E(G)\).

Subcase 2.1\(d_G(v_1)=2\).

Let \(\alpha =\phi (v_1u_1)\) if \(d_G(u_1)\ne 2\). When \(d_G(u_1)=2\), \(N_G(u_1)=\{y,v_1\}\). We can recolor \(v_1u_1\) with a color in \([1,4]\setminus C_{\phi }(u_1)\), and let \(\alpha =\phi (u_1y)\). If \(\alpha \in [1,4]\), then we can color \(vv_1\) with a color in [5,9] such that v does not conflict with its neighbors. So we may assume that \(\alpha =5\). Suppose that \(vv_1\) cannot be colored without creating a conflict. That is, \(C_\phi (v_i)=\{1,2,3,4,i+4\}\) for \(i=2,3,4,5\). We can recolor \(vv_2\) with color a color \(\beta \in \{5,7,8,9\}\) such that \(v_2\) does not conflict with its neighbors. Then, we can color \(vv_1\) with a color in \(\{7,8,9\}\backslash \{\beta \}\).

Subcase 2.2\(d_G(v_1)=3\) with \(N_G(v_1)=\{u_1,u_2,v\}\).

By Claim 2, \(d_3^G(v_1)\le 1\). When \(d_3^G(v_1)=0\), let \(\phi (v_1u_1)=a_1\) and \(\phi (v_1u_2)=a_2\). When \(d_3^G(v_1)=1\), say \(d_G(u_1)=3\). We can recolor \(v_1u_1\) with a color in \([1,4]\setminus C_{\phi }(u_1)\), and let \(a_1\in C_{\phi }(u_1)\setminus \{\phi (v_1u_2)\}\) and \(a_2=\phi (v_1u_2)\). Without loss of generality, assume that \(a_1,a_2\in [1,6]\). Suppose that \(vv_1\) cannot be colored without creating a conflict. That is, \(C_\phi (v_2)=\{1,2,3,4,7\}\), \(\{1,2,3,4,8\},\{1,2,3,4,9\}\in \{C_{\phi }(v_i)|i=3,4,5\}\). Then, \(d_G(v_2)=5\).

  1. (i)

    Suppose that \(C_{\phi }(v_3)\in \{\{1,2,3,4,8\},\{1,2,3,4,9\}\}\), say \(C_{\phi }(v_3)=\{1,2,3,4,8\}\). We can recolor \(vv_2\) with a color in \(\beta \in \{5,6,8,9\}\) such that \(v_2\) does not conflict with its neighbors. If \(\beta \in \{5,6\}\), then we can color \(vv_1\) with a color in \(\{8,9\}\) such that v does not conflict with its neighbors. Suppose that \(\beta \in \{8,9\}\). If \(C_{\phi }(v_5)\ne \{2,3,4,8,9\}\), then we can color \(vv_1\) with a color in \(\{8,9\}\setminus \{\beta \}\). So we may suppose that \(C_\phi (v_5)=\{2,3,4,8,9\}\). Then, we can recolor \(vv_3\) with a color in \(\gamma \in \{5,6,7,9\}\) such that \(v_3\) does not conflict with its neighbors, and color \(vv_1\) with a color in \(\{7,9\}\setminus \{\gamma \}\).

  2. (ii)

    Suppose that \(C_{\phi }(v_4)=\{1,2,3,4,8\}\) and \(C_{\phi }(v_5)=\{1,2,3,4,9\}\).

  • Suppose that \(vv_2\) can be recolored with a color \(\beta \in \{5,6,8,9\}\) such that \(v_2\) does not conflict with its neighbors. If \(\beta \in \{5,6\}\), then we can color \(vv_1\) with a color in \(\{8,9\}\) such that v does not conflict with \(v_3\). Suppose that \(\beta \in \{8,9\}\). If \(C_{\phi }(v_3)\ne \{2,3,4,8,9\}\), then we can color \(vv_1\) with a color in \(\{8,9\}\setminus \{\beta \}\). So we may suppose that \(C_\phi (v_3)=\{2,3,4,8,9\}\). Then, we can recolor \(vv_2\) with \(\beta \), recolor \(vv_3\) with a color in \(\gamma \in \{1,5,6,7\}\) such that \(v_3\) does not conflict with its neighbors, and color \(vv_1\) with a color in \(\{1,5,6,7\}\setminus \{a_1,a_2,\gamma \}\).

  • Suppose that \(vv_2\) cannot be recolored with a color \(\beta \in \{5,6,8,9\}\) such that \(v_2\) does not conflict with its neighbors. That is, \(\{\{2,3,4,7,i\}|i=5,6,8,9\}=\{C_{\phi }(x)|x\in N_G(v_2)\}\). First, assume that \(C_{\phi }(v_3)\in \{\{2,3,4,7,i\}|i=5,6\}\), say \(C_{\phi }(v_3)=\{2,3,4,5,7\}\). We can recolor \(vv_2\) with 5, recolor \(vv_3\) with a color \(\beta \in \{1,6,8,9\}\) such that \(v_3\) does not conflict with its neighbors, and color \(vv_1\) with a color in \(\{8,9\}\setminus \{\beta \}\). Next, assume that \(C_{\phi }(v_3)\in \{\{2,3,4,7,i\}|i=8,9\}\), say \(C_{\phi }(v_3)=\{2,3,4,7,8\}\). We can recolor \(vv_2\) with 8, recolor \(vv_3\) with a color \(\beta \in \{1,5,6,9\}\) such that \(v_3\) does not conflict with its neighbors, and color \(vv_1\) with a color in \(\{1,5,6,9\}\setminus \{a_1,a_2,\beta \}\). \(\square \)

Claim 9

Let v be a 6-vertex of G.

  1. (1)

    \(d_{3^-}^G(v)\le 2\);

  2. (2)

    If \(d_{2^-}^G(v)\ge 1\), then \(d_{3^-}^G(v)\le 1\);

  3. (3)

    If vertex v is adjacent to a \((2,6,k_1)\)-cycle, then \(d_4^G(v)=0\);

  4. (4)

    If v is incident with a \((3,6,k_1)\)-cycle, then \(d_3^G(v)=1\). Hence, v is not incident with (3, 3, 6)-cycle.

Proof

Let \(v_1, v_2,\ldots , v_6\) be the neighbors of v with \(d_G(v_1)=k\le 3\). Let \(u_1, u_2,\ldots ,u_{k-1}\) be the neighbors of \(v_1\) other than v. Then, \(G-vv_1\) admits a T(G)-avd-coloring \(\phi \) with the color set \(C=[1,T(G)]\) by Remark 1. Suppose that \(\phi (vv_i)=i-1\) for \(i\in [2,6]\).

  1. (1)

    Assume to the contrary that \(d_{3^-}^G(v)\ge 3\). Then, \(d_G(v_2)\), \(d_G(v_3)\le 3\). Recall that \(k\le 3\). By Claim 2, \(d_k^G(v_1)\le 1\). When \(d_k^G(v_1)=0\), let \(\phi (v_1u_1)=a_1\) and \(\phi (v_1u_{k-1})=a_2\). When \(d_k^G(v_1)=1\) with \(k=2\), say \(d_G(u_1)=2\), let \(C_{\phi }(u_1)=\{a_1,a_2\}\). When \(d_k^G(v_1)=1\) with \(k=3\), say \(d_G(u_1)=3\). We recolor \(v_1u_1\) with a color in \([1,5]\setminus C_{\phi }(u_1)\), and let \(a_1\in C_{\phi }(u_1)\setminus \{\phi (v_1u_2)\}\) and \(a_2=\phi (v_1u_2)\). Without loss of generality, assume that \(a_1,a_2\in [1,7]\).

Firstly, \(vv_1\) can be colored with any color in \(\{8,9\}\). Next, for each \(i\in \{2,3\}\), \(v_i\) has at most one possible conflict vertex by Claim 2. So \(vv_i\) can be recolored with \(b_i\in \{6,7,8,9\}\setminus C_\phi (v_i)\) such that \(v_i\) does not conflict with its neighbors. Then, we can color \(vv_1\) with a color in \(\{8,9\}\backslash \{b_i\}\). Hence, we have at least \(2+1\times 2=4\) different ways to recolor or color the edges incident to v. Since v has at most three conflict vertices, we can extend \(\phi \) to G, a contradiction.

  1. (2)

    Assume to the contrary that \(d_{3^-}^G(v)\ge 2\) when \(d_{2^-}^G(v)\ge 1\). Then, \(k\le 2\) and \(d_G(v_2)\le 3\). When \(d_G(v_1)=2\) and \(d_G(u_1)\ne 2\), let \(\alpha =\phi (v_1u_1)\). When \(d_G(v_1)=2\) and \(d_G(u_1)=2\), \(N_G(u_1)=\{y,v_1\}\). By Claim 2, \(d(y)\ne 2\). We can recolor \(v_1u_1\) with a color in \([1,5]\backslash C_\phi (u_1)\), and let \(\alpha =\phi (u_1y)\). Without loss of generality, assume that \(\alpha \in [1,6]\).

Firstly, \(vv_1\) can be colored with any color in \(\{7,8,9\}\). Next, we can recolor \(vv_2\) with a color \(b_1\in \{6,7,8,9\}\backslash C_\phi (v_2)\) such that \(v_2\) does not conflict with its neighbors and color \(vv_1\) with a color in \(\{7,8,9\}\backslash \{b_1\}\). Hence, we have at least 3+2=5 different ways to recolor or color the edges incident to \(v_1\). Since v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  1. (3)

    Assume to the contrary that \(d_4^G(v)\ge 1\) when v is adjacent to a \((2,6,k_1)\)-cycle. Then, \(k=2\) and \(v_1v_2\in E(G)\). Then, \(d_G(v_2)\ge 5\) by Claims 9 (2) and 3. Let \(\phi (v_1v_2)=a\in [1,6]\). Assume that \(d_G(v_3)=4\) with \(N_G(v_3)=\{x_1,x_2,x_3,v\}\). Firstly, \(vv_1\) can be colored with any color in \(\{7,8,9\}\). Next, when \(d_4^G(v_3)=0\), we can recolor \(vv_3\) with a color \(b\in [6,9]\setminus C_{\phi }(v_3)\) and color \(vv_1\) with a color in \([7,9]\setminus \{b\}\). When \(d_4^G(v_3)=1\), say \(d_G(x_1)=4\), we remove the color of \(v_3x_1\). Then, we can recolor \(vv_3\) with a color \(b\in \{6,7,8,9\}\backslash C_\phi (v_3)\) such that \(v_3\) does not conflict with \(x_1\), color \(v_3x_1\) properly and color \(vv_1\) with a color in \(\{7,8,9\}\backslash \{b\}\). Hence, we have at least 3+2=5 different ways to recolor or color the edges incident to \(v_1\). Since v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  2. (4)

    Assume to the contrary that \(d_3^G(v)\ge 2\) when v is incident with some \((3,6,k_1)\)-cycle C. Then, \(k=3\) and \(N_G(v_1)=\{u_1,v_2,v\}\), say \(v_1v_2\in E(G)\).

Case 1\(d_G(v_2)=3\) with \(N_G(v_2)=\{x,v_1,v\}\).

We can recolor \(v_1v_2\) with a color in \([2,5]\setminus \{\phi (v_1u_1),\phi (v_2x)\}\). Let \(\phi (v_1u_1)=a_1\) and\(\phi (v_2x)=a_2\).

  1. (i)

    Suppose that at most one of \(a_1\) and \(a_2\) belongs to [1, 5], say \(a_1,a_2\in [1,6]\). We can color \(vv_1\) with a color in [7,9], or we can recolor \(vv_2\) with 7 and color \(vv_1\) with a color in \(\{8,9\}\). Hence, we have at least 3+2=5 different ways to recolor or color the edges incident to v. Since v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  2. (ii)

    Suppose that neither \(a_1\) nor \(a_2\) belongs to [1,5], say \(a_1=6\) and \(a_2=7\). We can color \(vv_1\) with a color in [7,9], or we can recolor \(vv_2\) with 9 and color \(vv_1\) with a color in \(\{7,8\}\). So we have at least 3+2=5 different ways to recolor or color the edges incident to v. While v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

Case 2\(d_G(v_2)\ne 3\).

Suppose that \(d_G(v_3)=3\) with \(N_G(v_3)=\{x_1,x_2\}\). Recall that \(d_3^G(v_3)\le 1\) by Claim 2. When \(d_G(u_1)\ne 3\), let \(\phi (v_1u_1)=a_1\) and \(\phi (v_1v_2)=a_2\). When \(d_G(u_1)=3\), we can recolor \(v_1u_1\) with a color in \([1,5]\setminus (\{\phi (v_1v_2\}\cup C_{\phi }(u_1))\). Let \(a_2=\phi (v_1v_2)\) and \(a_1\in C_{\phi }(u_1)\setminus \{a_2\}\).

  1. (i)

    Suppose that at most one of \(a_1\) and \(a_2\) belongs to [1,5], say \(a_1,a_2\in [1,6]\). Firstly, \(vv_1\) can be colored with any color in \(\{7,8,9\}\). Next, we can recolor \(vv_3\) with a color \(\alpha \in \{6,7,8,9\}\backslash \{C_\phi (v_3)\}\) such that \(v_3\) does not conflict with its neighbors, and color \(vv_1\) with a color in \(\{7,8,9\}\backslash \{\alpha \}\). Thus, we have at least 3+2=5 different ways to recolor the edges incident to v. While v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  2. (ii)

    Suppose that neither \(a_1\) nor \(a_2\) belongs to [1,5], say \(a_1=6\), \(a_2=7\). Firstly, \(vv_1\) can be recolored with 8 or 9. Secondly, we can exchange the colors of \(v_1v_2\) and \(vv_2\) and color \(vv_1\) with 8 or 9. Finally, we can recolor \(vv_3\) with a color \(\alpha \in \{6,7,8,9\}\backslash \{C_\phi (v_3)\}\) such that \(v_3\) does not conflict with its neighbors, and color \(vv_1\) with a color in \(\{8,9\}\backslash \{\alpha \}\). Thus, we have at least 2+2+1=5 different ways to recolor the edges incident to v. While v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction. \(\square \)

Claim 10

Let v be a 7-vertex of G.

  1. (1)

    If \(d_1^G(v)\ge 1\), then \(d_{3^-}^G(v)\le 2\).

  2. (2)

    If \(d_{2^-}^G(v)\ge 1\), then \(d_{3^-}^G(v)\le 3\).

  3. (3)

    Suppose that v is incident with a (3, 3, 7)-cycle. Then, \(d_{3^-}^G(v)\le 2\).

Proof

Let \(v_1, v_2,\ldots , v_7\) be the neighbors of v with \(d(v_1)=k\le 3\). Let \(u_1, u_2,\ldots ,u_{k-1}\) be the neighbors of \(v_1\) other than v. Then, \(G-vv_1\) admits a T(G)-avd-coloring \(\phi \) with the color set \(C=[1,T(G)]\) by Remark 1. Suppose that \(\phi (vv_i)=i-1\) for \(i\in [2,7]\). \(\square \)

Remark 2

Suppose that \(v_i\) is a \(3^-\)-vertex. If \(d_G(v_i)\le 2\) or \(d_G(v_i)=3\) with \(d_3^G(v_i)=0\), then we can recolor \(vv_i\) with a color \(b_i\in \{7,8,9\}\setminus C_\phi (v_i)\) such that \(v_i\) does not conflict with its neighbors. If \(d_G(v_i)=3\) with \(d_3^G(v_i)=1\), let \(x_i\) be the 3-vertex adjacent to \(v_i\). We can recolor \(v_ix_i\) with a color in \([1,6]\setminus (C_\phi (v_i)\cup C_\phi (x_i))\) and recolor \(vv_i\) with a color \(b_i\in \{7,8,9\}\setminus C_\phi (v_i)\) so that \(v_i\) does not conflict with \(x_i\). Hence, we can recolor \(vv_i\) with some color \(b_i\in \{7,8,9\}\) without creating a conflict.

  1. (1)

    Assume to the contrary that \(d_{3^-}^G(v)\ge 3\) when \(d_1^G(v)\ge 1\). Then, \(d_G(v_1)=1\), and \(d_G(v_2)\), \(d_G(v_3)\le 3\). Firstly, \(vv_1\) can be colored with any color in \(\{7,8,9\}\). Secondly, for each \(i\in [2,3]\), \(vv_i\) can be recolored with \(b_i\in \{7,8,9\}\) without creating a conflict by Remark 2. Then, we can color \(vv_1\) with a color in \(\{7,8,9\}\backslash \{b_i\}\). Hence, we have at least \(3+1\times 2=5\) different ways to recolor the edges incident to v. While v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  2. (2)

    Assume to the contrary that \(d_{3^-}^G(v)\ge 4\) when \(d_{2^-}^G(v)\ge 1\). Then, by Claim 10 (1), \(d_G(v_1)=2\), and \(d_G(v_2)\), \(d_G(v_3)\), \(d_G(v_4)\le 3\). Let \(u_1\) be the neighbor of \(v_1\) other than v. Let \(\phi (v_1u_1)=\alpha \) when \(d_G(u_1)\ne 2\). When \(d_G(u_1)=2\), let y be the neighbor of \(u_1\) other than \(v_1\). We can recolor \(v_1u_1\) with a color in \([1,6]\setminus \{C_{\phi }(u_1y)\}\), and let \(\phi (u_1y)=\alpha \). Without loss of generality, assume that \(\alpha \in [1,7]\). Firstly, \(vv_1\) can be colored with 8 or 9. Secondly, for each \(i\in [2,4]\), \(vv_i\) can be recolored with \(b_i\in \{7,8,9\}\) without creating a conflict by Remark 2. Then, we can color \(vv_1\) with a color in \(\{8,9\}\backslash \{b_i\}\). Hence, we have at least \(2+1\times 3=5\) different ways to recolor or color the edges incident to v. While v has at most 3 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  3. (3)

    Assume to the contrary that \(d_{3^-}^G(v)\ge 3\) when \(k_1=3\). Let \(d_G(v_2)=3\) with \(N_G(v_2)=\{x,v_1,v\}\) and \(d_G(v_3)\le 3\). We recolor \(v_1v_2\) with a color in \([1,6]\setminus \{\phi (v_1u_1),\phi (v_2x)\}\). Let \(\phi (v_1u_1)=a_1\) and \(\phi (v_2x)=a_2\).

  • Suppose that \(a_1,a_2\in [1,6]\). We can color \(vv_1\) with a color in [7,9], or we can recolor \(vv_2\) with 7 and color \(vv_1\) with a color in \(\{8,9\}\). Hence, we have at least 3+2=5 different ways to recolor or color the edges incident to v. Since v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  • Suppose that \(a_1\in [1,6]\) and \(a_2=7\). Firstly, \(vv_1\) can be recolored with 8 or 9. Secondly, \(vv_2\) can be recolored with 8, and \(vv_1\) can be colored with 7 or 9. Finally, \(vv_2\) can be recolored with 9, and \(vv_1\) can be colored with 7. Hence, we have at least 2+2+1=5 different ways to recolor or color the edges incident to v. Since v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  • Suppose that \(a_2\in [1,6]\) and \(a_1=7\). Firstly, \(vv_1\) can be recolored with 8 or 9. Secondly, \(vv_1\) can be colored with 8, and \(vv_2\) can be recolored with 7 or 9. Finally, \(vv_1\) can be colored with 9, and \(vv_2\) can be recolored with 7. Hence, we have at least 2+2+1=5 different ways to recolor or color the edges incident to v. Since v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  • Suppose that \(a_1=7\) and \(a_2=8\). Firstly, \(vv_1\) can be recolored with 8 or 9. Secondly, \(vv_2\) can be recolored with 7, and \(vv_1\) can be colored with 1 or 9. Finally, \(vv_1\) can be colored with 8, and \(vv_2\) can be recolored with 9. Hence, we have at least 2+2+1=5 different ways to recolor or color the edges incident to v. Since v has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction. \(\square \)

Claim 11

Let v be a k-vertex of G with \(k\ge 8\).

  1. (1)

    If \(d_1^G(v)\ge i\) for \(i=1,2\), then \(d_k^G(v)\ge d_{(i+1)^-}^G(v)+1\).

  2. (2)

    If \(d_2^G(v)\ge 1\), then \(d_k^G(v)\ge 1\).

Proof

  1. (1)

    We will show the case that \(i=2\). (The proof of \(i=1\) is similar.) Let \(t=d_{3^-}^G(v)\). Assume to the contrary that \(d_k^G(v)\le t\). Let \(v_1,v_2,\ldots ,v_k\) be the neighbors of v with \(d_G(v_1)=d_G(v_2)=1\), \(d_G(v_3)\le \cdots \le d_G(v_t)\le 3\). Then, \(H=G-vv_1\) has a T(H)-avd-coloring \(\phi \) using the color set \(C=[1,T(G)]\) by Remark 1. Let \(\phi (vv_i)=i-1\) for \(2\le i\le k\).

Firstly, \(vv_1\) can be colored with k or \(k+1\). Secondly, \(vv_2\) can be recolored with k and \(vv_1\) can be colored with \(k+1\). Finally, for \(3\le i\le t\), \(vv_i\) can be recolored with a color \(b_i \in \{1,k,k+1\}\) without creating a conflict by Remark 2. For \(b_i\in \{k,k+1\}\), we can color \(vv_1\) with a color in \(\{k,k+1\}\backslash \{b_i\}\). For \(b_i=1\), we can recolor \(vv_2\) with k and color \(vv_1\) with \(k+1\). This implies that there are at least \(2+1+(t-2)=t+1\) different ways to recolor or color the edges incident with v. While v has at most t conflict vertices, we can extend \(\phi \) to G, a contradiction.

  1. (2)

    Similarly, it is easy to show (2) is true. \(\square \)

Claim 12

Let v be a k-vertex of G and \(v_1,v_2,\ldots ,v_k\) be the neighbors of v in the clockwise order. If \(d_2^G(v)\ge 3\), then there are no two consecutive 2-vertices with at least one of them is bad, where \(v_k\) and \(v_1\) are considered to be consecutive.

Proof

Assume to the contrary that there exist some \(i(1\le i\le k)\), say \(i=1\), such that \(d_G(v_1)=d_G(v_2)=2\) and \(v_1\) is a bad 2-vertex in G. Let \(d_G(v_j)=2\) by \(d_2^G(v)\ge 3\). Let \(u_l\) be the second neighbor of \(v_l\) other than v for \(l=1,2,j\), and \(y_1\) be the second neighbor of \(u_1\) other than \(v_1\). Note that by Claim 5, no 5-cycle is incident with both \(v_1\) and \(v_2\). Recall that by the proof of Claim 7, \(y_1\ne v\). Now, the proof is split into the following two cases.

Case 1 \(v_2\) is a good 2-vertex in G.

Suppose that \(v_1\) and \(v_2\) are incident with some 6-cycle \(vv_1u_1y_1u_2v_2v\). Then, \(H=G-v_1u_1\) admits a T(G)-avd-coloring \(\phi \) using the color set \(C=[1,T(G)]\) by Remark 1. Assume that \(\phi (vv_i)=i\) for \(i\in [1,k]\). If \(\phi (u_1y_1)\ne 1\), then \(v_1u_1\) can be colored properly. So we may suppose that \(\phi (u_1y_1)=1\). If \(\phi (v_2u_2)\ne 1\), then we can exchange the colors of \(vv_1\) and \(vv_2\) and color \(v_1u_1\) properly. So \(\phi (v_2u_2)=1\). If \(\phi (u_2y_1)\ne 2\), then we can recolor \(u_1y_1\), \(u_2y_1\), \(v_2u_2\) with \(\phi (u_2y_1)\), 1, \(\phi (u_2y_1)\) and color \(u_1v_1\) properly. Suppose that \(\phi (u_2y_1)=2\). When \(d(u_j)\ne 2\), set \(\alpha =\phi (v_ju_j)\). When \(d(u_j)=2\), let \(y_j\) be the second neighbor of \(u_j\) other than \(v_j\). Recolor \(v_ju_j\) with a color in \([3,k]\backslash C_\phi (u_j)\), and set \(\alpha =\phi (u_jy_j)\). Recolor \(vv_j\) with a color \(\beta \in \{1,2\}\backslash \{\alpha \}\). If \(\beta =1\), then we can recolor \(vv_1\) with j and color \(v_1u_1\) properly. If \(\beta =2\), then we can recolor \(vv_2\), \(v_2u_2\), \(u_2y_1\), \(y_1u_1\) with j, 2, 1, 2 and color \(v_1u_1\) properly.

Suppose that no 6-cycle is incident with both \(v_1\) and \(v_2\). Let H be the graph obtained by adding a new vertex w to \(G-\{v_1u_1,v_2u_2\}\) and joining w to both \(u_1\) and \(u_2\). Then, H contains no 4-cycles, \(|E(H)|=|E(G)|\), \(\Delta (H)=\Delta (G)\), and \(n_2(H)<n_2(G)\). By the minimality of G, H has a T(G)-avd-coloring \(\phi \) using the color set \(C=[1,T(G)]\). Assume that \(\phi (vv_i)=i\) for \(i\in [1,k]\), \(\phi (y_1u_1)=a\), \(\phi (wu_2)=b\). Note that \(a\ne b\), Since \(\phi \) is the avd-coloring. Now, we need to find the T(G)-avd-coloring \(\psi \) of G. First, let \(\psi (e)=\phi (e)\) for \(e\in E(H)\cap E(G)\). If \(a\ne 1\) and \(b\ne 2\), then we can color \(v_1u_1\) properly and color \(v_2u_2\) with b. If \(a=1\) or \(b=2\), then \(b\ne 1\) or \(a\ne 2\), respectively. We can recolor or color \(v_2u_2\), \(vv_1\) and \(vv_2\) with b, 2, 1, respectively, and color \(v_1u_1\) properly.

Case 2\(v_2\) is a bad 2-vertex, say \(d_G(u_2)=2\) and \(N_G(u_2)=\{v_2,y_2\}\).

By Claim 6, no 6-cycle and 7-cycle are incident with both \(v_1\) and \(v_2\). Let \(H=(G-\{v_1u_1,v_2u_2\})\cup \{u_1u_2\}\). Then, H contains no 4-cycles, \(|E(H)|<|E(G)|\), and \(\Delta (H)=\Delta (G)\). Since H is smaller than G, H has a T(G)-avd-coloring \(\phi \) using the color set \(C=[1,T(G)]\). Assume that \(\phi (vv_i)=i\) for \(i\in [1,k]\), \(\phi (u_1y_1)=a\), \(\phi (u_2y_2)=b\). Note that \(a\ne b\), since \(\phi \) is the avd-coloring. Now, we need to find the T(G)-avd-coloring \(\psi \) of G. First, let \(\psi (e)=\phi (e)\) for \(e\in E(H)\cap E(G)\). If \(a\ne 1\) and \(b\ne 2\), then we can color \(v_1u_1\) and \(v_2u_2\) properly. If \(a=1\) or \(b=2\), then \(b\ne 1\) or \(a\ne 2\), respectively. We can recolor \(vv_1\) and \(vv_2\) with 2, 1, respectively, and color \(v_1u_1\), \(v_2u_2\) properly. \(\square \)

3 Discharging Methods

Let H be the graph obtained from G by removing all 1-vertices. Then, H is a planar graph without 4-cycles. We show the relationship between \(d_G(v)\) and \(d_H(v)\) for \(v\in V(H)\) by claims 1, 9-11 in Table 1. Hence, we can deduce that \(\delta (H)\ge 2\). Therefore, H is normal. Other properties of H are collected in the following.

Claim 13

  1. (1)

    Let v be a k-vertex of H with \(2\le k\le 4\). Then, \(d_k^H(v)\le 1\) and \(\sum \limits _{2\le i\le 4, i\ne k}d_i^H(v)=0\).

  2. (2)

    Let v be a 5-vertex of H. Then, \(d_{2b}^H(v)=0\) and \(d_{3^-}^H(v)\le 1\). Furthermore, if \(d_{3^-}^H(v)=1\), then \(t_H(v)=0\).

  3. (3)

    Let v be a 6-vertex of H. Then,

    1. (3.1)

      \(d_{3^-}^H(v)\le 2\).

    2. (3.2)

      If \(d_2^H(v)\ge 1\), then \(d_{3^-}^H(v)\le 1\).

    3. (3.3)

      If v is incident with \((2,6,k_1)\)-cycle, then \(d_{4}^H(v)=0\).

    4. (3.4)

      If v is incident with \((3,6,k_1)\)-cycle, then \(d_{3}^H(v)=1\).

  4. (4)

    Let v be a 7-vertex of H. Then,

    1. (4.1)

      If \(d_2^H(v)\ge 1\), then \(d_{3^-}^H(v)\le 3\).

    2. (4.2)

      If v is incident with (3, 3, 7)-cycle, then \(d_{3}^H(v)\le 2\).

  5. (5)

    Let v be a k-vertex of H with \(d_2^H(v)\ge 1\) and \(k\ge 8\), then \(d_{5^+}^H(v)\ge 1\).

  6. (6)

    Each 3-face in H must be a \((2,6^+,6^+)\)-face, \((3,6^+,6^+)\)-face, \((3,3,7^+)\)-face, \((4,4,7^+)\)-face, or \((4^+,5^+,5^+)\)-face.

  7. (7)

    Each f of H is incident with at most \(\lfloor \frac{2}{3}d_H(f)\rfloor \)\(4^-\)-vertices.

  8. (8)

    For each vertex v of H, \(t_H(v)\le \lfloor \frac{d_H(v)}{2}\rfloor \).

Table 1 The relation between \(d_G(v)\) and \(d_H(v)\)

Proof

By Table 1 and Claims 1, 2, 10, the statements (1) and (5) are obviously true.

  1. (2)

    Suppose that \(d_H(v)=5\). When \(d_G(v)=5\), statement (2) is true by Claim 8. When \(d_G(v)\ge 7\), \(d_1^G(v)\ge 2\). By Claim 4, we have \(d_2^G(v)+d_{3}^G(v)=0\). Assume that \(d_G(v)=6\). Then, \(d_1^G(v)=1\). By Claim 9, \(d_2^G(v)+d_{3}^G(v)=0\). Thus, by Table 1, \(d_{3^-}^H(v)=0\) when \(d_G(v)\ge 6\). Hence, statement (2) is also true for \(d_G(v)\ge 6\) by Table 1.

  2. (3)

    Suppose that \(d_H(v)=6\). When \(d_G(v)=6\), statement (3) is true by Claim 9. When \(d_G(v)=7\), \(d_1^G(v)=1\). By Claims 4 and  10, \(d_2^G(v)=0\) and \(d_{3^-}^G(v)\le 2\), which implies that \(d_2^H(v)=0\) and \(d_{3}^H(v)\le 1\) by Table 1. Thus, statement (3) holds. When \(d_G(v)\ge 8\), \(d_1^G(v)\ge 2\). By Claim 4, we have \(d_2^G(v)+d_{3}^G(v)=0\), which implies that \(d_{3^-}^H(v)=0\) by Table 1. Hence, statement (3) holds.

  3. (4)

    Suppose that \(d_H(v)=7\). When \(d_G(v)=7\), statement (4) holds by Claim 10. When \(d_G(v)\ge 8\), then \(d_1^G(v)\ge 1\). By Claim 4, we have \(d_2^G(v)=0\), which implies that \(d_2^H(v)=0\) by Table 1. Hence, statement (4.1) holds. By Claim 6, G does not contain \(G_2\) as subgraph, which implies that v is not incident with any (3, 3, 7)-cycle. So statement (4.2) holds.

  4. (6)

    Let \(f=[v_0v_1v_2]\) be a \((d_0,d_1,d_2)\)-face in H with \(d_0\le d_1\le d_2\). Suppose that \(d_0=2\), then by the proof of Claim 7, \(d_1\ne 2\). By Claim 13 (1) and (2), \(d_1\ge 6\), say f is a \((2,6^+,6^+)\)-face.

  5. (i)

    Suppose that \(d_0=3\). If \(d_1=3\), then \(d_2\ge 7\) by Claim 13 (1), (2), and (3.4). So f is a \((3,3,7^+)\)-face. Suppose that \(d_1\ne 3\). By Claim 13 (1), (2), we have \(d_1\ge 6\), say f is a \((3,6^+,6^+)\)-face.

  6. (ii)

    Suppose that \(d_0\ge 4\). If f is neither the \((4^+,5^+,5^+)\)-face nor the \((4,4,7^+)\)-face, then f must be the (4, 4, 5)-face or (4, 4, 6)-face by Claim 13 (1). Now, we need to show that H does not contain (4, 4, 5)-face and (4, 4, 6)-face. Assume to the contrary that H contains a (4, 4, 5)-face or a (4, 4, 6)-face. Then, \(d_G(v_0)=d_G(v_1)=4\). Then, \(H'=G-v_0v_1\) admits a T(G)-avd-coloring \(\phi \) using the color set \(C=[1,T(G)]\) by Remark 1. Note that \(T(G)\ge 9\). If \(C_{\phi }(v_0)\ne C_{\phi }(v_1)\), then we can color \(v_0v_1\) properly to obtain a T(G)-avd-coloring of G, a contradiction. So we may assume that \(C_{\phi }(v_0)=C_{\phi }(v_1)=\{1,2,3\}\), \(\phi (v_2v_0)=1\), and \(\phi (v_2v_1)=2\). If \(d_1^G(v_2)\ge 2\), then let \(u_1\) and \(u_2\) be the 1-vertices adjacent to \(v_2\). Note that \(\{\phi (v_2u_1),\phi (v_2u_2)\}\setminus \{3\}\ne \emptyset \), say \(\phi (v_2u_1)\ne 3\). We can exchange the color of \(v_2u_1\) and \(v_2v_0\) and color \(v_0v_1\) properly to achieve a T(G)-avd-coloring of G, a contradiction. So \(d_1^G(v_2)\le 1\), say \(d_G(v_2)=5\), 6, or 7.

  • Suppose that \(d_G(v_2)=5\). First, we can recolor \(v_2v_0\) with any color in \([4,9]\setminus C_{\phi }(v_2)\). Next, we can recolor \(v_2v_1\) with any color in \([4,9]\setminus C_{\phi }(v_2)\). Hence, there are at least 3+3=6 different ways to recolor the edges incident with \(v_2\). While \(v_2\) has at most 3 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  • Suppose that \(d_G(v_2)=6\). Note that color 3 may not belong to \(C_{\phi }(v_2)\). Without loss of generality, assume that \(C_{\phi }(v_2)\subseteq [1,7]\). Firstly, \(v_2v_0\) can be recolored with 8 or 9. Secondly, \(v_2v_1\) can be recolored with 8 or 9. Finally, \(v_2v_0\) can be recolored with 8, and \(v_2v_1\) can be recolored with 9. Hence, there are at least 2+2+1=5 different ways to recolor the edges incident with \(v_2\). While \(v_2\) has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  • Suppose that \(d_G(v_2)=7\). Then, \(d_1^G(v_2)=1\) by the previous discussion. Let \(u_1\) be the 1-vertex adjacent to \(v_2\). If \(\phi (v_2u_1)\ne 3\), then we can exchange the color of \(v_2u_1\) and \(v_2v_0\) and color \(v_0v_1\) properly to achieve a T(G)-avd-coloring of G, a contradiction. Suppose that \(\phi (v_2u_1)=3\). Without loss of generality, assume that \(C_{\phi }(v_2)=[1,7]\). Firstly, \(v_2v_0\) can be recolored with 8 or 9. Secondly, \(v_2v_1\) can be recolored with 8 or 9. Finally, \(v_2v_0\) can be recolored with 8, and \(v_2v_1\) can be recolored with 9. Hence, there are at least 2+2+1=5 different ways to recolor the edges incident with \(v_2\). While \(v_2\) has at most 4 conflict vertices, we can extend \(\phi \) to G, a contradiction.

  1. (7)

    By Claim 13 (1), it is easy to check that statement (7) holds.

  2. (8)

    By the fact that H contains no 4-cycles, statement (8) holds.

\(\square \)

Using Euler’s formula \(|V(H)|-|E(H)|+|F(H)|=2\), we have :

$$\begin{aligned} \sum \limits _{v\in V(H)}(d_H(v)-4)+\sum \limits _{f\in F(H)}(d_H(f)-4)=-8 \end{aligned}$$

In order to complete the proof, we make use of discharging method. First, we define an initial charge function \(w(x)=d_H(x)-4\) for every \(x\in V(H)\cup F(H)\). Next, we design some discharging rules and redistribute weights accordingly. Once the discharging is finished, a new charge function \(w'\) is produced. However, the sum of all charge is kept fixed when the discharging is in progress. Nevertheless, we can show that \(w'(x)\ge 0\) for all \(x\in V(H)\cup F(H)\). This leads to the following obvious contradiction:

$$\begin{aligned} 0\le \sum \limits _{x\in V(H)\cup F(H)}w'(x)=\sum \limits _{x\in V(H)\cup F(H)}w(x)=-8<0 \end{aligned}$$

and hence demonstrates that the counterexample cannot exist.

We define the following discharging rules:

  • R1 Let f be a 3-face incident with \(5^{+}\)-vertex v. If f is \((4,4,7^+)\)-face, then v sends \(\frac{1}{2}\) to f; otherwise, v sends \(\frac{1}{d_{5^+}^H(f)}\) to f.

  • R2 Every 5-vertex sends \(\frac{1}{2}\) to its adjacent \(3^-\)-vertex.

  • R3 Let v be a \(6^+\)-vertex adjacent to a \(3^-\)-vertex u.

    • R3.1 If u is bad 2-vertex, then v sends \(\frac{5}{6}\) to u.

    • R2.2 If u is good 2-vertex with \(t_H(u)=1\), then v sends \(\frac{3}{4}\) to u.

    • R3.3 If u is good 2-vertex with \(t_H(u)=0\), then v sends \(\frac{1}{2}\) to u.

    • R3.4 If \(d_H(u)=3\), then v sends \(\frac{1}{3}\) to u.

  • R4 Every \(5^{+}\)-face sends \(\frac{1}{2}\) to each its incident good 2-vertex, \(\frac{7}{12}\) to each incident bad 2-vertex, \(\frac{1}{6}\) to each incident bad 3-vertex, and \(\frac{1}{6}\) to each adjacent \((4,4,7^+)\)-face.

  • R5 Every \(6^{+}\)-face sends \(\frac{1}{3}\) to each incident bad 5-face through every light 2-edge.

  • R6 Let f be a \(5^{+}\)-face, and f is not a special k-face for \(k=5,6,7\).

    • R6.1 If \(d_H(f)\ge 6\), then f sends \(\frac{1}{12}\) to each incident \(5^+\)-vertex.

    • R6.2 Suppose that \(d_H(f)=5\) with \(d_2^H(f)\le 1\). Then, f sends \(\frac{1}{12}\) to each incident \(5^+\)-vertex.

    • R6.3 Suppose that \(d_H(f)=5\) with \(d_{2b}^H(f)=2\), say \(u_1\) be the bad 2-vertex incident with f. Let u be the \(5^+\)-vertex adjacent to \(u_1\). If f is not adjacent to \((4,4,7^+)\)-face, then f sends \(\frac{1}{12}\) to u.

Remark 3

By Claim 13 (1), \(d_4^H(v)=1\) for any 4-vertex v. So a \(5^+\)-face f sends at most \(\frac{1}{6}d_4^H(f)\) to adjacent \((4,4,7^+)\)-faces in total.

Now, we will show that \(w'(x)\ge 0\) for all \(x\in V(H)\cup F(H)\) in the following two lemmas.

Lemma 1

\(w'(f)\ge 0\) for all \(f\in F(H)\).

Proof

  1. (1)

    Suppose that \(d_H(f)=3\). Then, \(w(f)=-1\). Recall that by Claim 13 (6), \(d_{5^+}^H(f)\ge 1\). If f is a \((4,4,7^+)\)-face, then \(w'(f)=-1+\frac{1}{2}+\frac{1}{6}\times 3=0\) by R1 and R4. Otherwise, \(w'(f)=-1+d_{5^+}^H(f)\times \frac{1}{d_{5^+}^H(f)}=0\) by R1.

  2. (2)

    Suppose that \(d_H(f)=5\). Then, \(w(f)=1\). By Claim 5, \(d_2^H(f)=d_2^G(f)\le 2\). Suppose that \(d_2^H(f)\le 1\). Note that \(d_{4^-}^H(f)\le 3\) by Claim 13 (7). So \(d_3^H(f)+d_4^H(f)\le 2\). Thus, \(w'(f)\ge 1-\frac{1}{2}d_2^H(f)-\frac{1}{6}d_3^H(f)- \frac{1}{6}d_4^H(f)-\frac{1}{12}d_{5^+}^H(f)\ge 1-\frac{1}{2}- \frac{1}{6}\times 2-\frac{1}{12}\times 2=0\) by R4 and R6.2. Suppose that \(d_2^H(f)=2\), say \(d_{2g}^H(f)=2\) or \(d_{2b}^H(f)=2\). If \(d_{2g}^H(f)=2\), then f is special and \(d_3^H(f)+d_4^H(f)=0\) by Claim 13 (1). Thus, \(w'(f)=1-\frac{1}{2}\times 2=0\) by R4. If \(d_{2b}^H(f)=2\), then f must be adjacent to a \(6^+\)-face by the fact that H contains no 4-cycles. By Claims 5 and  13 (1), \(d_4^H(f)\le 1\) and f is not incident with bad 3-vertex. If f is not adjacent to \((4,4,7^+)\)-face, then \(w'(f)=1-\frac{7}{12}\times 2+\frac{1}{3}-\frac{1}{12}\times 2=0\) by R4 and R6.3. If f is adjacent to \((4,4,7^+)\)-face, then \(w'(f)=1-\frac{7}{12}\times 2+\frac{1}{3}-\frac{1}{6}=0\) by R4 and R6.3.

  3. (3)

    Suppose that \(d_H(f)=k\ge 6\). By Claim 13 (7), we have \(d_{4^-}^H(f)\le \lfloor \frac{2}{3}k\rfloor \). Thus, by R4–R6 and Remark 3, we have

$$\begin{aligned} w'(f)\ge & {} (k-4)-\frac{1}{2}d_{2g}^H(f)-\frac{7}{12}(d_2^H(f)-d_{2g}^H(f))- \frac{1}{6}d_{3b}^H(f)\nonumber \\&-\frac{1}{6}d_{4}^H(f)-\frac{1}{3}\times \frac{1}{2}(d_2^H(f)-d_{2g}^H(f))-\frac{1}{12}d_{5^+}^H(f)\nonumber \\\ge & {} \frac{11}{12}k-4-\frac{8}{12}d_{2}^H(f)-\frac{1}{12}d_3^H(f)- \frac{1}{12}d_4^H(f)+\frac{1}{4}d_{2g}^H(f) \end{aligned}$$
(1)
$$\begin{aligned}\ge & {} \frac{11}{12}k-4-\frac{8}{12}d_{4^-}^H(f) \end{aligned}$$
(2)
  • Suppose that \(k\ge 8\). Then, \(w'(f)\ge 0\) by (2) and \(d_{4^-}^H(f)\le \lfloor \frac{2}{3}k\rfloor \).

  • Suppose that \(k=7\). If \(d_{4^-}^H(f)\le 3\), then \(w'(f)\ge \frac{11}{12}\times 7-4-\frac{8}{12}\times 3=\frac{5}{12}\) by (2). Suppose that \(d_{4^-}^H(f)=4\). If \(d_{2b}^H(f)\le 2\), then by (1), we have \(w'(f)\ge \frac{11}{12}\times 7-4-\frac{8}{12}d_{2}^H(f)-\frac{1}{12}d_{3}^H(f)-\frac{1}{12}d_{4}^H(f)+\frac{1}{4}d_{2g}^H(f)\ge \frac{29}{12}-\frac{8}{12} d_{2b}^H(f)-\frac{5}{12}(d_{2g}^H(f)+d_3^H(f)+d_4^H(f))\ge \frac{29}{12}-\frac{1}{4}d_{2b}^H(f)-\frac{5}{12}d_{4^-}^H(f)\ge \frac{29}{12}-\frac{1}{4}\times 2-\frac{5}{12}\times 4=\frac{1}{4}\). If \(d_{2b}^H(f)=4\), then f is special. Thus, \(w'(f)\ge (7-4)-\frac{7}{12}\times 4-\frac{1}{3}\times 2=0\) by R4–R6.

  • Suppose that \(k=6\). By Claim 6, \(d_2^H(f)\le 3\). If \(d_{2}^H(f)=3\) and \(d_{2b}^H(f)=2\), then f is special. Thus, \(w'(f)\ge (6-4)-\frac{7}{12}\times 2-\frac{1}{2}-\frac{1}{3}\times 1=0\) by R4–R6. If \(d_{2g}^H(f)=3\), then \(d_3^H(f)+d_4^H(f)=0\) by Claim 13 (1). Thus, \(w'(f)\ge (6-4)-\frac{1}{2}\times 3-\frac{1}{12}\times 3=\frac{1}{4}\) by R4 and R6. Suppose that \(d_{2}^H(f)=2\). By Claim 13 (7), \(d_3^H(f)+d_4^H(f)\le 2\). Thus, \(w'(f)\ge (6-4)-\frac{7}{12}\times 2-\frac{1}{6}\times 2-\frac{1}{3}\times 1-\frac{1}{12}\times 2=0\) by R4–R6 and Remark 3. Suppose that \(d_{2}^H(f)\le 1\). By Claim 13 (7), we have \(d_{4^-}^H(f)\le 4\). Thus, \(w'(f)\ge (6-4)-\frac{1}{2}d_2^H(f)-\frac{1}{6}d_{3}^H(f)-\frac{1}{6} d_4^H(f)-\frac{1}{12}d_{5^+}^H(f)\ge 2-\frac{1}{2}-\frac{1}{6}\times 3-\frac{1}{12}\times 2=\frac{5}{6}\) by R4–R6 and Remark 3.

\(\square \)

Lemma 2

\(w'(v)\ge 0\) for all \(v\in V(H)\).

Proof

Let v be a k-vertex in H. Let \(v_0,v_1,\ldots ,v_{k-1}\) be the neighbors of v in the clockwise order. For each \(i\in [0,k-1]\), let \(f_i\) be the face with \(vv_i\) and \(vv_{i+1}\) as its boundary edges, where the indices are taken modular k.

  1. (1)

    Suppose that \(d_H(v)\)=2. Then, \(w(v)=-2\). If \(d_2^H(v)=1\), by Claim 13 (1), (2), and (6), \(d_{6^+}^H(v)=1\) and v is incident with two \(5^+\)-faces. Then, \(w'(v)=-2+\frac{5}{6}+\frac{7}{12}\times 2=0\) by R3 and R4. Suppose that \(d_2^H(v)=0\). We have \(d_{5^+}^H(v)=2\) by Claim 13 (1). If \(t_H(v)=0\), then \(w'(v)=-2+\frac{1}{2}\times 2+\frac{1}{2}\times 2=0\) by R2–R4. If \(t_H(v)=1\), then \(w'(v)=-2+\frac{3}{4}\times 2+\frac{1}{2}=0\) by R2–R4.

  2. (2)

    Suppose that \(d_H(v)=3\). Then, \(w(v)=-1\). If v is good, by Claim 13 (1), we have \(d_{5^+}^H(v)=3\). Then, \(w'(v)=-1+\frac{1}{3}\times 3=0\) by R2 and R3. If v is bad, by Claim 13 (1), we have \(d_{5^+}^H(v)=2\). Note that v is incident with at least two \(5^+\)-face. Thus, \(w'(v)=-1+\frac{1}{3}\times 2+\frac{1}{6}\times 2=0\) by R2–R4.

  3. (3)

    Suppose that \(d_H(v)=4\). Then, \(w'(v)=w(v)=0\)

  4. (4)

    Suppose that \(d_H(v)=5\). Then, \(w(v)=1\). Note that v sends at most \(\frac{1}{2}\) to each incident 3-face by R1 and Claim 13 (6). If \(d_{3^-}^H(v)=1\), then \(t_H(v)=0\) and \(d_{2b}^H(v)=0\) by Claim 13 (2). Thus, \(w'(v)=1-\frac{1}{2}=\frac{1}{2}\) by R2. If \(d_{3^-}^H(v)=0\), then \(t_H(v)\le 2\) by Claim 13 (8). Thus, \(w'(v)=1-\frac{1}{2}\times 2=0\) by R1.

  5. (5)

    Suppose that \(d_H(v)=6\). Then, \(w(v)=2\). Note that \(t_H(v)\le 3\) by Claim 13 (8), and v sends at most \(\frac{1}{2}\) to each incident 3-face by R1 and Claim 13 (6). First, assume that \(d_2^H(v)=0\). Then, \(d_3^H(v)\le 2\) by Claim 13 (3). If v is incident with a \((3,6,k_1)\)-face, then \(d_3^H(v)=1\) by Claim 13 (3). Thus, \(w'(v)=2-\frac{1}{3}-\frac{1}{2}\times 3=\frac{1}{6}\) by R1 and R3. If v is not incident with any \((3,6,k_1)\)-face, then \(t_H(v)\le 2\) when \(d_3^H(v)\ge 1\). Thus, \(w'(v)=2-\max \{\frac{1}{3}\times 2+\frac{1}{2}\times 2,\frac{1}{2}\times 3\}=\frac{1}{3}\) by R1 and R3. Next, assume that v is adjacent to a 2-vertex u. Then, \(d_2^H(v)=1\) and \(d_3^H(v)=0\) by Claim 13 (3). If \(d_2^H(u)=0\) and \(t_H(u)=0\), then \(w'(v)\ge 2-\frac{1}{2}-\frac{1}{2}\times 3=0\) by R1 and R3. If \(d_2^H(u)=0\) and \(t_H(u)=1\), then v is incident with a \((2,6,k_1)\)-face \(f'\). Hence, \(d_4^H(v)=0\) by Claim 13 (3). So v sends \(\frac{1}{3}\) to each incident 3-face \(f\ne f'\). Thus, \(w'(v)\ge 2-\frac{3}{4}-\frac{1}{2}-\frac{1}{3}\times 2=\frac{1}{12}\) by R1 and R3. If \(d_2^H(u)=1\), then \(t_H(v)\le 2\) since u is not incident with 3-face. Thus, \(w'(v)\ge 2-\frac{5}{6}-\frac{1}{2}\times 2=\frac{1}{6}\) by R1 and R3.

  6. (6)

    Suppose that \(d_H(v)=7\). Then, \(w(v)=3\) and \(t_H(v)\le 3\) by Claim 13 (8). By Claim 13 (4), \(d_{3^-}^H(v)\le 3\) when \(d_2^H(v)\ge 1\).

  • Assume that \(d_2^H(v)=0\). If v is incident with a (3,3,7)-face, then \(d_3^H(v)\le 2\) by Claim 13 (4). Thus, \(w'(x)=3-\frac{1}{3}\times 2-1-\frac{1}{2}\times 2=\frac{1}{3}\) by R1 and R3. So suppose that v is not incident with (3,3,7)-face. Then, \(d_3^H(v)\le 7-t_H(v)\). Recall that v sends at most \(\frac{1}{2}\) to each incident 3-face by R1 and Claim 13 (6). Thus, \(w'(x)\ge 3-\frac{1}{3}(7-t_H(v))-\frac{1}{2}t_H(v)=\frac{2}{3}-\frac{1}{6}t_H(v)\ge \frac{2}{3}-\frac{1}{6}\times 3=\frac{1}{6}\) by R1 and R3.

  • Assume that \(d_2^H(v)\ge 1\), then \(d_3^H(v)\le 2\). By Claim 6, v is not incident with (3,3,7)-face. So v sends at most \(\frac{1}{2}\) to each incident 3-face by R1 and Claim 13 (6). If \(d_2^H(v)=1\), then \(w'(v)\ge 3-\frac{5}{6}-\frac{1}{3}\times 2-\frac{1}{2}\times 3=0\) by R1 and R3. So suppose that \(d_2^H(v)\ge 2\). By Claim 6, v is not incident with \((2,7,k_1)\)-face. Hence, v sends \(\frac{1}{2}\) to each adjacent good 2-vertex. If \(d_{2g}^H(v)\ge 2\), then \(t_H(v)\le 2\). Thus, \(w'(v)=3-\frac{1}{2}\times 2-\frac{5}{6}-\frac{1}{2}\times 2=\frac{1}{6}\) by R1 and R3. If \(d_{2g}^H(v)=1\), then \(t_H(v)\le 2\) when \(d_2^H(v)=2\), and \(t_H(v)\le 1\) when \(d_2^H(v)=3\). Thus, \(w'(v)\ge 3-\frac{1}{2}-\max \{\frac{5}{6}+\frac{1}{3}+\frac{1}{2}\times 2, \frac{5}{6}\times 2+\frac{1}{2}\}=\frac{1}{3}\) by R1 and R3. If \(d_{2g}^H(v)=0\), then \(t_H(v)\le 2\) when \(d_2^H(v)=2\), and \(t_H(v)\le 1\) when \(d_2^H(v)=3\). Thus, \(w'(v)\ge 3-\max \{\frac{5}{6}\times 2+\frac{1}{3}+\frac{1}{2}\times 2, \frac{5}{6}\times 3+\frac{1}{2}\}=0\) by R1 and R3.

  1. (7)

    Suppose that \(d_H(v)=k\ge 8\).

Case 1\(d_2^H(v)=0\).

If v is incident with (3, 3, k)-face, say \(f_2\), then by Claim 6, v is incident with exactly one (3, 3, k)-face. Thus, \(d_3^H(v)\le k+1-t_H(v)\). For \(i=1,3\), it is easy to check that \(d_H(f_i)\ge 5\) and \(f_i\) is not a special face. Further, when \(d_H(f_i)=5\), then \(d_2^H(f_i)\le 1\) by Claim 13 (1) and \(d_2^H(v)=0\). So \(f_i\) sends \(\frac{1}{12}\) to v by R6 for \(i=1,3\). Hence, \(w'(v)\ge (k-4)-\frac{1}{3}d_3^H(v)-1-\frac{1}{2}(t_H(v)-1)+\frac{1}{12}\times 2\ge \frac{2}{3}k-\frac{1}{6}t_H(v)-\frac{14}{3} \ge \frac{2}{3}k-\frac{1}{6}\lfloor \frac{k}{2}\rfloor -\frac{14}{3}\ge 0\) by R1 and R3. If v is not incident with (3, 3, k)-face, then \(d_3^H(v)\le k-t_H(v)\). Thus, \(w'(v)\ge (k-4)-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v) \ge \frac{2}{3}k-\frac{1}{6}t_H(v)-4\ge \frac{2}{3}k-\frac{1}{6}\lfloor \frac{k}{2}\rfloor -4\ge \frac{2}{3}\).

Case 2\(d_2^H(v)\ge 1\).

Then, \(d_{5^+}^H(v)\ge 1\) by Claim 13 (5). By Claim 6, v is not incident with any (3, 3, k)-face. Hence, \(d_3^H(v)\le k-d_2^H(v)-t_H(v)\). If v is incident with some \((2,k,k_1)\)-face, then \(d_2^H(v)=1\) by Claim 6. Thus, \(w'(v)\ge (k-4)-\frac{3}{4}-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v)\ge k-4-\frac{3}{4}-\frac{1}{3}(k-1-t_H(v))-\frac{1}{2}t_H(v)=\frac{2}{3}k-\frac{1}{6}t_H(v)-\frac{53}{12}\ge \frac{1}{4}\) by R1 and R3. So assume that v is not incident with any \((2,k,k_1)\)-face. Then, \(d_2^H(v)+t_H(v)\le k\). If \(d_{2b}^H(v)=0\), then \(w'(v)\ge (k-4)-\frac{1}{2}d_2^H(v)-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v)=\frac{2}{3}k-\frac{1}{6}d_2^H(v)-\frac{1}{6}t_H(v)-4\ge \frac{2}{3}k-\frac{1}{6}k-4\ge 0\) by R1 and R3. Hence, assume that \(d_{2b}^H(v)\ge 1\). Now, the proof is split into following two cases, depending on the value of \(d_2^H(v)\).

Subcase 2.1\(d_2^H(v)\ge 3\).

Remark 4

By Claim 12, if \(v_i\) is a bad 2-vertex, then \(d_H(v_{i+1})\ge 3\) and \(d_H(v_{i-1})\ge 3\), where the indices are taken modular k. So \(d_{2b}^H(v)\le \lfloor \frac{k}{2}\rfloor \). Further, if \(d_{2b}^H(v)<\frac{k}{2}\), then \(d_{3^+}^H(v)\ge d_{2b}^H(v)+1\); if \(d_{2b}^H(v)=\frac{k}{2}\), then \(d_{3^+}^H(v)\ge d_{2b}^H(v)\). This implies that \(d_2^H(v)=k-d_{3^+}^H(v)\le k-d_{2b}^H(v)\). By Claim 6 and Claim 12, \(f_i\) and \(f_{i-1}\) are not special when \(v_i\) is a bad 2-vertex. By the fact that G contains no 4-cycles, we have \(\max \{d_H(f_{i-1}),d_H(f_i)\}\ge 6\). So at least one of \(f_i\) and \(f_{i-1}\) sends \(\frac{1}{12}\) to v by R6.

Recall that \(d_3^H(v)\le k-d_2^H(v)-t_H(v)\). Since v is not incident with any \((2,k,k_1)\)-face, we have \(t_H(v)\le \lfloor \frac{k-d_2^H(v)}{2}\rfloor \). By R1, R3, and Remark 4, we have

$$\begin{aligned} w'(v)= & {} (k-4)-\frac{5}{6}d_{2b}^H(v)-\frac{1}{2}(d_2^H(v)-d_{2b}^H(v))-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v)+\frac{1}{12}d_{2b}^H(v)\nonumber \\\ge & {} \frac{2}{3}k-\frac{1}{4}d_{2b}^H(v)-\frac{1}{6}d_2^H(v)-\frac{1}{6}t_H(v)-4 \nonumber \\\ge & {} \frac{2}{3}k-\frac{1}{4}d_{2b}^H(v)-\frac{1}{6}d_2^H(v)-\frac{1}{6}\times \frac{k-d_2^H(v)}{2}-4 \nonumber \\\ge & {} \frac{7}{12}k-\frac{1}{4}d_{2b}^H(v)-\frac{1}{12}(k-d_{2b}^H(v))-4 \nonumber \\\ge & {} \frac{1}{2}k-\frac{1}{6}d_{2b}^H(v)-4 \end{aligned}$$
(3)
$$\begin{aligned}\ge & {} \frac{1}{2}k-\frac{1}{6}\times \lfloor \frac{k}{2}\rfloor -4 \end{aligned}$$
(4)
  1. (i)

    Suppose that \(k\ge 10\). Then, \(w'(v)\ge \frac{1}{6}\) by (4).

  2. (ii)

    Suppose that \(k=9\). If \(d_{2b}^H(v)\le 3\), then \(w'(v)\ge 0\) by (3). So suppose that \(d_{2b}^H(v)=4\). Then, \(d_{3^+}^H(v)=5\), \(d_{5^+}^H(v)\ge 1\), and \(t_H(v)\le 1\) by Claim 13 (5) and Remark 4. Thus, by R1, R3, and Remark 4, we have \(w'(v)\ge (9-4)-\frac{5}{6}\times 4-\frac{1}{3}\times 4-\frac{1}{2}+\frac{1}{12}\times 4=\frac{1}{6}\).

  3. (iii)

    Suppose that \(k=8\). Then, \(w(v)=4\), \(d_{2b}^H(v)\le 4\) and \(d_{5^+}^H(v)\ge 1\).

  • Suppose that \(d_{2b}^H(v)=4\), then \(d_3^H(v)\le 3\) and \(t_H(v)=0\). Thus, \(w'(v)\ge 4-\frac{5}{6}\times 4-\frac{1}{3}d_3^H(v)+\frac{1}{12}\times 4\ge 0\) by R1, R3, and Remark 4.

  • Suppose that \(d_{2b}^H(v)=3\). If \(v_0\), \(v_2\), and \(v_4\) are bad 2-vertices, then \(d_H(v_i)\ge 3\) for \(i=1,3,5,7\). Moreover, if \(d_H(v_6)=2\), then \(t_H(v)=0\); if \(d_H(v_6)\ge 3\), then \(t_H(v)\le 1\). It is easy to check that both \(f_1\) and \(f_2\) send \(\frac{1}{12}\) to v by R6, since \(d_H(f_0)\ne 3\) and \(d_H(f_3)\ne 3\). Hence, \(w'(v)\ge 4-\frac{5}{6}\times 3-\frac{1}{3}\times 4-\frac{1}{2}+\frac{1}{12}\times 4=0\) by R1, R3, and Remark 4. So assume that \(v_0\), \(v_2\), and \(v_5\) are bad 2-vertices. Then, \(d_H(v_i)\ge 3\) for \(i=1,3,4,6,7\) and \(d_H(f_j)\ne 3\) for \(j=0,1,2,4,5,7\). If at least one of \(f_3\) and \(f_6\) is not a 3-face, say \(d_H(f_3)\ne 3\), then it is easy to check that both \(f_1\) and \(f_2\) send \(\frac{1}{12}\) to v by R6. Hence, \(w'(v)\ge 4-\frac{5}{6}\times 3-\frac{1}{3}\times 4-\frac{1}{2}+\frac{1}{12}\times 4=0\) by R1, R3, and Remark 4. So assume that \(d_H(f_3)=d_H(f_6)=3\). If at least one of \(v_3\), \(v_4\), \(v_6\), and \(v_7\) is a 4-vertex, then \(d_3^H(v)\le 2\) by Claim 13 (1). Hence, \(w'(v)\ge 4-\frac{5}{6}\times 3-\frac{1}{3}\times 2-\frac{1}{2}\times 2+\frac{1}{12}\times 3=\frac{1}{12}\) by R1, R3, and Remark 4. Therefore, none of \(v_3\), \(v_4\), \(v_6\), and \(v_7\) is a 4-vertex. So for \(j=0,1,2,4,5,7\), \(f_j\) is not special, and \(f_j\) is not adjacent to \((4,4,7^+)\)-face when \(d_H(f_j)=5\). Then, \(f_j\) sends \(\frac{1}{12}\) to v by R6. Hence, \(w'(v)\ge 4-\frac{5}{6}\times 3-\frac{1}{3}\times 3-\frac{1}{2}\times 2+\frac{1}{12}\times 6=0\) by R1 and R3.

  • Suppose that \(d_{2b}^H(v)=2\). Recall that v is not incident with any \((2,k,k_1)\)-face. Then, it is easy to check that \(d_{2g}^H(v)+t_H(v)\le 3\) by Remark 4. Thus, \(w'(v)\ge 4-\frac{5}{6}\times 2-\frac{1}{2}d_{2g}^H(v)-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v)+\frac{1}{12}\times 2 \ge \frac{5}{2}-\frac{1}{2}d_{2g}^H(v)-\frac{1}{3}(8-2-d_{2g}^H(v)-t_H(v))-\frac{1}{2}t_H(v)=\frac{1}{2}-\frac{1}{6}(d_{2g}^H(v)+t_H(v))\ge \frac{1}{2}-\frac{1}{6}\times 3=0\) by R1, R3, and Remark 4.

  • Suppose that \(d_{2b}^H(v)=1\). Then, it is easy to check that \(d_{2g}^H(v)+t_H(v)\le 5\) by Remark 4. Thus, \(w'(v)\ge 4-\frac{5}{6}-\frac{1}{2}d_{2g}^H(v)-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v)+\frac{1}{12} \ge \frac{13}{4}-\frac{1}{2}d_{2g}^H(v)-\frac{1}{3}(8-1-d_{2g}^H(v)-t_H(v))-\frac{1}{2}t_H(v)=\frac{11}{12}-\frac{1}{6}(d_{2g}^H(v)+t_H(v))\ge \frac{11}{12}-\frac{1}{6}\times 5=\frac{1}{12}\) by R1, R3, and Remark 4.

Subcase 2.2\(1\le d_2^H(v)\le 2\).

Then, \(d_{2b}^H(v)\le d_2^H(v)\le 2\). Recall that \(d_3^H(v)\le k-d_2^H(v)-t_H(v)\), since v is not incident with (3, 3, k)-face by Claim 6. Since v is not incident with any \((2,k,k_1)\)-face, we have \(t_H(v)\le \lfloor \frac{k-d_2^H(v)}{2}\rfloor \). By R1 and R3, we have

$$\begin{aligned} w'(v)= & {} (k-4)-\frac{5}{6}d_{2b}^H(v)-\frac{1}{2}(d_2^H(v)-d_{2b}^H(v))-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v) \nonumber \\\ge & {} \frac{2}{3}k-\frac{1}{3}d_{2b}^H(v)-\frac{1}{6}d_2^H(v)-\frac{1}{6}t_H(v)-4 \nonumber \\\ge & {} \frac{2}{3}k-\frac{1}{3}d_{2b}^H(v)-\frac{1}{6}d_2^H(v)-\frac{1}{6}\times \frac{k-d_2^H(v)}{2}-4 \nonumber \\\ge & {} \frac{7}{12}k-\frac{1}{3}d_{2b}^H(v)-\frac{1}{12}d_2^H(v)-4 \end{aligned}$$
(5)
$$\begin{aligned}\ge & {} \frac{7}{12}k-\frac{29}{6} \end{aligned}$$
(6)
  1. (i)

    Suppose that \(k\ge 9\). Then, \(w'(v)\ge \frac{5}{12}\) by (6).

  2. (ii)

    Suppose that \(k=8\). If \(d_{2b}^H(v)\le 1\), then \(w'(v)\ge \frac{7}{12}\times 8-\frac{1}{3}\times 1-\frac{1}{12}\times 2-4=\frac{1}{6}\) by (5). So suppose that \(d_{2b}^H(v)=2\). Then, \(d_{2g}^H(v)=0\) and \(t_H(v)\le 3\). Without loss of generality, assume that \(v_0\) and \(v_i\) are a bad 2-vertex for \(i\in [1,7]\). Then, \(\max \{d_H(f_0),d_H(f_7)\}\ge 6\) and \(\max \{d_H(f_{i-1}),d_H(f_i)\}\ge 6\) by the fact that G contains no 4-cycles. If \(d_H(v_1)\ge 3\) and \(d_H(v_7)\ge 3\), then \(f_0\) or \(f_7\) is a non-special \(6^+\)-face by Claim 6. Similarly, \(f_{i-1}\) or \(f_i\) is a non-special \(6^+\)-face. Thus, \(w'(v)\ge (8-4)-\frac{5}{6}\times 2-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v)+\frac{1}{12}\times 2\ge \frac{5}{2}-\frac{1}{3}(8-2-t_H(v))-\frac{1}{2}t_H(v)=\frac{1}{2}-\frac{1}{6}t_H(v)\ge 0\) by R1, R3, and R6. So suppose that \(d_H(v_1)=2\) or \(d_H(v_7)=2\), say \(d_H(v_1)=2\). Then, \(d_H(v_7)\ge 3\) and \(d_H(v_2)\ge 3\). If \(t_H(v)\le 2\), then \(w'(v)\ge (8-4)-\frac{5}{6}\times 2-\frac{1}{3}d_3^H(v)-\frac{1}{2}t_H(v)\ge \frac{7}{3}-\frac{1}{3}(8-2-t_H(v))-\frac{1}{2}t_H(v)=\frac{1}{3}-\frac{1}{6}t_H(v)\ge 0\) by R1 and R3. Thus, \(t_H(v)=3\), say \(d_H(f_j)=3\) for \(j=2,4,6\). If at least one of \(v_2\), \(v_3,\cdots ,v_7\) is a 4-vertex, then \(d_3^H(v)\le 2\) by Claim 13 (1). Hence, \(w'(v)\ge (8-4)-\frac{5}{6}\times 2-\frac{1}{3}\times 2-\frac{1}{2}\times 3=\frac{1}{6}\) by R1 and R3. Therefore, none of \(v_2\), \(v_3,\cdots ,v_7\) is a 4-vertex. So for \(j=1,7\), \(f_j\) is not special, and \(f_j\) is not adjacent to \((4,4,7^+)\)-face when \(d_H(f_j)=5\). Then, \(f_j\) sends \(\frac{1}{12}\) to v by R6. Hence, \(w'(v)\ge (8-4)-\frac{5}{6}\times 2-\frac{1}{3}\times 3-\frac{1}{2}\times 3+\frac{1}{12}\times 2=0\) by R1 and R3. \(\square \)