1 Introduction

Only simple graphs are considered in this paper. Let G be a graph with vertex set V(G), edge set E(G), minimum degree \(\delta (G)\) and maximum degree \(\Delta (G)\) (for short, \(\Delta \)). For a vertex \(v\in V(G)\), let \(d_G(v)\) denote the degree of v in G. Set \(|G|=|V(G)|\) and \(||G||=|E(G)|\). A k-vertex, \(k^-\)-vertex, and \(k^+\)-vertex of G are a vertex with degree k, at most k, and at least k, respectively. A graph G is normal if it contains no isolated edges, and formal if it contains no leaves. A graph G is called planar if it can be embedded in the plane such that all edges intersect in their end-vertices. A plane graph is a particular drawing of a planar graph in the plane. For two nonnegative integers pq with \(p<q\), we use [pq] to denote the set of all integers between p and q (including p and q).

An edge-k-coloring of a graph G is a mapping \(\phi \) from the edge set E(G) to the color set \(\{1,2,\ldots ,k\}\) such that no two adjacent edges get same color. Here two edges are said to be adjacent if they share at least one common end vertex. The chromatic index \(\chi '(G)\) of the graph G is defined as the smallest integer k such that G admits an edge-coloring using k colors. Given an edge-k-coloring \(\phi \) of G and for a vertex \(v\in V(G)\), we use \(C_{\phi }(v)\) to denote the set of colors assigned to the edges incident with v. Suppose that xy are any pair of adjacent vertices in G. We say that \(\phi \) is neighbor-distinguishing if \(C_{\phi }(x)\ne C_{\phi }(y)\), strict neighbor-distinguishing if \(C_{\phi }(x)\not \subseteq C_{\phi }(y)\) and \(C_{\phi }(y)\not \subseteq C_{\phi }(x)\), and local neighbor-distinguishing if \(C_{\phi }(x)\not \subseteq C_{\phi }(y)\) and \(C_{\phi }(y)\not \subseteq C_{\phi }(x)\) whenever \(d_G(x),d_G(y)\ge 2\). The neighbor-distinguishing index \(\chi '_\textrm{a}(G)\) (strict neighbor-distinguishing index \(\chi '_\textrm{snd}(G)\), local neighbor-distinguishing index \(\chi '_\textrm{lnd}(G)\), respectively) of G is the smallest k such that G has a neighbor-distinguishing edge-k-coloring (a strict neighbor-distinguishing edge-k-coloring, a local neighbor-distinguishing edge-k-coloring, respectively).

As an easy observation, a graph G has a neighbor-distinguishing edge-coloring if and only if G is normal, and G has a strict neighbor-distinguishing edge-coloring if and only if G is formal. But the local neighbor-distinguishing edge-coloring is well defined for any graph G.

It is evident that \(\chi '_\textrm{snd}(G)\ge \chi '_\textrm{a}(G)\ge \Delta \) for any formal graph G. Moreover, the following propositions hold obviously.

Proposition 1

If G is a graph with \(\delta (G)\ge 2\), then \(\chi '_\textrm{lnd}(G)=\chi '_\textrm{snd}(G)\).

Proposition 2

If G is an \(r~(\ge 2)\)-regular graph, then \(\chi '_\textrm{lnd}(G)=\chi '_\textrm{snd}(G)=\chi '_\textrm{a}(G)\).

Zhang et al. [23] introduced the neighbor-distinguishing edge-coloring of graphs and proposed the following challenging conjecture.

Conjecture 1

Every normal graph G, other than a 5-cycle, has \(\chi '_\textrm{a}(G)\le \Delta +2\).

Akbari et al. [1] proved that every normal graph G satisfies \(\chi '_\textrm{a}(G)\le 3\Delta \). This result was gradually improved to \(\chi '_\textrm{a}(G)\le 2.5\Delta \) by Wang et al. [21], and to \(\chi '_\textrm{a}(G)\le 2\Delta +2\) by Vučković [17]. In 2005, using probabilistic analysis, Hatami [10] showed that every normal graph G with \(\Delta > 10^{20}\) has \(\chi '_\textrm{a}(G)\le \Delta +300\). Recently, this result was improved, by Joret and Lochet [13], to that \(\chi '_\textrm{a}(G)\le \Delta +19\) for a normal graph with sufficiently large \(\Delta \). Suppose that G is a normal planar graph. It was shown in [11] that if \(\Delta \ge 12\) then \(\chi '_\textrm{a}(G)\le \Delta +2\). Moreover, Wang and Huang [20] showed that if \(\Delta \ge 16\), then \(\Delta \le \chi '_\textrm{a}(G)\le \Delta +1\), and \(\chi '_\textrm{a}(G)=\Delta +1\) if and only if G contains adjacent \(\Delta \)-vertices. This result was improved in [19] to that if \(\Delta \ge 14\), then \(\Delta \le \chi '_\textrm{a}(G)\le \Delta +1\), and \(\chi '_\textrm{a}(G)=\Delta +1\) if and only if G contains adjacent \(\Delta \)-vertices.

The strict neighbor-distinguishing edge-coloring of graphs was studied in [24] (named there the Smarandachely adjacent vertex edge coloring). Let \(H_n\) (\(n\ge 2\)) denote the graph obtained from the bipartite graph \(K_{2,n}\) by inserting a 2-vertex into one edge. It is easy to show that \(\chi '_\textrm{snd}(H_n)=2n+1=2\Delta (H_n)+1\). Based on this fact, Gu et al. [8] raised the following conjecture.

Conjecture 2

Every connected formal graph G, different from \(H_{\Delta }\), has \(\chi '_\textrm{snd}(G)\le 2\Delta \).

Because \(\chi '_\textrm{snd}(K_{2,n})=2n=2\Delta (K_{2,n})\), the upper bound \(2\Delta \) in Conjecture 2 is sharp. Conjecture 2 remains open, but it was confirmed for graphs with \(\Delta \le 3\) in [8] and \(K_4\)-minor-free graphs in [9].

In this paper, we continue to study the strict neighbor-distinguishing edge-coloring of graphs, in particular, for the class of planar graphs. As a helpful tool, we consider its relaxed form, i.e., local neighbor-distinguishing edge-coloring of graphs. Our main results in this paper are stated as follows:

  • \(\chi '_\textrm{lnd}(G)\le 3\Delta -1\) for any simple graph G;

  • \(\chi '_\textrm{lnd}(G)\le \lceil 2.8\Delta \rceil +4\) for a planar graph G;

  • \(\chi '_\textrm{lnd}(G)\le 2\Delta +10\) for a planar graph G without 4-cycles;

  • \(\chi '_\textrm{lnd}(G)\le \Delta +23\) for a 3-connected planar graph G;

  • \(\chi '_\textrm{lnd}(G)\le \Delta +6\) for a Hamiltonian planar graph G.

2 An Upper Bound

Let G be a graph and \(\phi \) be a local neighbor-distinguishing edge-k-coloring of G. For the sake of briefness, \(\phi \) is called a k-LNDE-coloring of G. Two adjacent vertices u and v are exclusive in \(\phi \) if \(C_{\phi }(u)\not \subseteq C_{\phi }(v)\) and \(C_{\phi }(v)\not \subseteq C_{\phi }(u)\). To give an upper bound of the local neighbor-distinguishing index of a graph, we need to use the following result:

Lemma 2.1

([23]) For a cycle \(C_n\) with \(n\ge 3\),

$$\begin{aligned} \chi '_\textrm{a}(C_n)=\left\{ \begin{array}{ll} 3, &{} {\textrm{if}}\, n=3;\\ 5, &{} {\textrm{if}}\, n=5;\\ 4, &{} {\textrm{if}}\, n\ne 3,5. \end{array}\right. \end{aligned}$$

Theorem 2.2

Every graph with \(\Delta \ge 2\) has \(\chi '_\textrm{lnd}(G)\le 3\Delta -1\).

Proof

The proof is by induction on the edge number ||G||. If \(||G||\le 3\Delta -1\), then the result holds trivially since we can color the edges of G with distinct colors. Let G be a graph with \(||G||\ge 3\Delta \ge 6\). Without loss of generality, assume that G is connected. So, it follows that \(\Delta \ge 2\) and \(\delta (G)\ge 1\). In the following, we write simply \(K=3\Delta -1\) and let \(C=[1,K]\) denote the set of K colors.

First assume that \(\delta (G)=1\). Let v be a vertex adjacent to leaves \(x_1,\ldots ,x_l\) and \(2^+\)-vertices \(y_1,\ldots ,y_k\), where \(l\ge 1\) and \(k\ge 0\). Let \(H=G-x_1\). Then H is a graph with \(||H||<||G||\) and \(\Delta (H)\le \Delta \). By the induction hypothesis, H admits a K-LNDE-coloring \(\phi \) using the color set C. For \(i\in [1,k]\), since v and \(y_i\) are exclusive in \(\phi \), there exists a color \(r_i\in C_{\phi }(y_i){\setminus } C_{\phi }(v)\). Set \(R(v)=\{r_1,\ldots ,r_k\}\), which is called the second-level forbidden set of vertex v. Obviously, \(|R(v)|\le k\). Based on \(\phi \), we color \(vx_1\) with a color \(a\in C{\setminus } (C_{\phi }(v)\cup R(v))\). Since \(|C{\setminus } (C_{\phi }(v)\cup R(v))|\ge 3\Delta -1- |C_{\phi }(v)|-k\ge 3\Delta -1-(\Delta -1)-(\Delta -1) = \Delta +1\ge 3\), a exists and so the coloring is available. It is easy to check that the resultant coloring is a K-LNDE-coloring of G.

Next assume that \(\delta (G)\ge 2\). If \(\Delta =2\), then G is a cycle. By Lemma 2.1 and Proposition 2, \(\chi '_\textrm{lnd}(G)=\chi '_\textrm{a}(G)\le 5=3\Delta -1\). So assume that \(\Delta \ge 3\). The proof is split into two cases as follows, depending on the size of \(\delta (G)\).

Case I.\(\delta (G)=2\).

Let v be a 2-vertex with neighbors \(v_1,v_2\) such that \(d_G(v_1)\le d_G(v_2)\). Without loss of generality, we may suppose that \(d_G(v_2)\ge 3\) by the assumption that \(\Delta \ge 3\). The proof is split into two subcases as follows.

  • \(d_G(v_1)=2\). Let \(u_1\) be the neighbor of \(v_1\) other than v. If \(u_1=v_2\), then \(H=G-vv_1\) is a graph with \(||H||<||G||\) and \(\Delta (H)=\Delta \). By the induction hypothesis, H admits a K-LNDE-coloring \(\phi \) using the color set C. Based in \(\phi \), it suffices to color \(vv_1\) with some color in \(C\setminus C_{\phi }(v_2)\). If \(u_1\ne v_2\), then let \(H=G-v\), which has a K-LNDE-coloring \(\phi \) using the color set C by the induction hypothesis. We first color \(vv_2\) with \(a\in C{\setminus } (C_{\phi }(v_2)\cup R(v_2)\cup \{\phi (v_1u_1)\})\), where \(R(v_2)\) is the second-level forbidden set of vertex \(v_2\), as defined before. Then we color \(vv_1\) with \(b\in C{\setminus } (C_{\phi }(u_1)\cup C_{\phi }(v_2)\cup \{a\})\). For short, we write \(C^+_{\phi }(v_2)=C_{\phi }(v_2)\cup R(v_2)\) in the following discussion. Since \(|C{\setminus } (C^+_{\phi }(v_2)\cup \{\phi (v_1u_1)\})|\ge 3\Delta -1-2(d_G(v_2)-1)-1\ge \Delta \ge 2\) and \(|C{\setminus } (C_{\phi }(u_1)\cup C_{\phi }(v_2)\cup \{a\})|\ge 3\Delta -1-2\Delta \ge \Delta -1\ge 1\), both a and b exist and hence \(\phi \) is extended to G.

  • \(d_G(v_1)\ge 3\). Let \(H=G-v\), which admits a K-LNDE-coloring \(\phi \) using C. Based on \(\phi \), we color \(vv_1\) with \(a\in C\setminus (C^+_{\phi }(v_1)\cup C_{\phi }(v_2))\), and \(vv_2\) with \( b\in C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup \{a\})\). Since \(|C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2))|\ge 3\Delta -1-2(\Delta -1)-(\Delta -1)\ge 2\) and \(|C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup \{a\})|\ge 3\Delta -1-2(\Delta -1)-(\Delta -1)-1\ge 1\), ab exist and \(\phi \) is extended to G.

Case II. \(\delta (G)\ge 3\).

Take a vertex \(v\in V(G)\) with \(d_G(v)=\delta (G)\ge 3\). Let \(v_0, \ldots ,v_{k-1}\) be the neighbors of v in G, where \(k=d_G(v)\). Let \(H=G-v\). Then H is a graph with \(\delta (H)\ge 2\), \(\Delta (H)\le \Delta \), and \(||H||<||G||\). By the induction hypothesis, H admits a K-LNDE-coloring \(\phi \) using C. Let \(x_1, \ldots ,x_m\) be the neighbors of \(v_0\) in H, where \(m=d_G(v_0)-1\ge 2\). For \(i\in [1,m]\), there exists a color \(r_i\in C_{\phi }(x_i){\setminus } C_{\phi }(v_0)\) since \(v_0\) and \(x_i\) are exclusive in \(\phi \). Let \(R(v_0)=\{r_1, \ldots ,r_m\}\). Similarly, we can define \(R(v_1), \ldots ,R(v_{k-1})\). Let \(U_i=C^+_{\phi }(v_i)\) for \(i\in [0,k-1]\). Then \(|U_i|=|C_{\phi }(v_i)\cup R(v_i)|\le |C_{\phi }(v_i)|+|R(v_i)|\le (\Delta -1)+(\Delta -1)=2\Delta -2\).

To extend \(\phi \) to G, we design a coloring procedure as following.

Step 0. Color \(vv_0\) with a color \(c_0\in C{\setminus } (U_0\cup C_{\phi }(v_1))\), and then set \(B_0=\{c_0\}\).

Step 1. For \(i\in [1,k-1]\), we do the following operation, where all indices are taken modulo k:

  • If \(B_{i-1}\subseteq C_{\phi }(v_{i+1})\), then we color \(vv_i\) with a color \(c_i\in C\setminus (U_i\cup C_{\phi }(v_{i+1}))\); otherwise, we color \(vv_i\) with a color \(c_i\in C{\setminus } (U_i\cup B_{i-1})\).

  • Set \(B_i=B_{i-1}\cup \{c_i\}\).

Step 2. If \(i=k-1\), stop. Otherwise, set \(i=i+1\), then go to Step 1.

Let \(\pi \) denote the resultant edge-coloring of G after the above iterative process is ended. Let \(B=B_{k-1}\). Then \(B=C_{\pi }(v)\). We will show that \(\pi \) is a K-LNDE-coloring of G.

Claim 1. \(\pi \) is a proper edge-K-coloring of G.

Proof

We first prove the existence of the color \(c_i\) for \(i\in [0,k-1]\). In fact, since \(|C{\setminus } (U_0\cup C_{\phi }(v_1))|\ge |C|-|U_0|-|C_{\phi }(v_1)|\ge (3\Delta -1)-(2\Delta -2)-(\Delta -1) =2\), \(c_0\) exists. Assume that \(1\le i\le k-1\). If \(B_{i-1}\subseteq C_{\phi }(v_{i+1})\), then \(c_i\in C\setminus (U_i\cup C_{\phi }(v_{i+1}))\) by Step 1. Since \(|C{\setminus } (U_i\cup C_{\phi }(v_{i+1}))|\ge (3\Delta -1)-(2\Delta -2)-(\Delta -1)=2\), \(c_i\) exists. Otherwise, \(B_{i-1}\not \subseteq C_{\phi }(v_{i+1})\). By Step 1, \(c_i\in C{\setminus } (U_i\cup B_{i-1})\). Since \(|B_{i-1}|\le i\le k-1=d_G(v)-1=\delta (G)-1\le \Delta -2\), it follows that \(|C{\setminus } (U_i\cup B_{i-1})|\le (3\Delta -1)-(2\Delta -2)-(\Delta -2)=3\); thus, \(c_i\) exists. Next, we need to show that \(c_0,c_1,\ldots ,c_{k-1}\) are mutually distinct. Actually, this is true from the definition of \(B_i\)’s. Hence, \(\pi \) is a proper edge-K-coloring of G.\(\square \)

Claim 2. \(\pi \) is local neighbor-distinguishing.

Proof

It suffices to show that for any edge \(xy\in E(G)\), we have

$$\begin{aligned} C_{\pi }(x)\not \subseteq C_{\pi }(y)\ \textrm{and} \ C_{\pi }(y)\not \subseteq C_{\pi }(x)(*) \end{aligned}$$

By symmetry, we consider the following three possibilities.

Case 1. \(x,y\notin \{v,v_0,\ldots ,v_{k-1}\}\).

Note that \(\pi (e)=\phi (e)\) for each edge e incident with x or y in G. This implies that \(C_{\pi }(x)=C_{\phi }(x)\) and \(C_{\pi }(y)=C_{\phi }(y)\). Since \(C_{\phi }(x)\not \subseteq C_{\phi }(y)\) and \(C_{\phi }(y)\not \subseteq C_{\phi }(x)\), \((*)\) holds.

Case 2. \(y\notin \{v,v_0, \ldots ,v_{k-1}\}\) and \(x=v_i\) for some \(i\in [0,k-1]\).

By symmetry, suppose that \(i=0\), i.e., \(x=v_0\). Then \(C_{\pi }(y)=C_{\phi }(y)\) and \(C_{\pi }(v_0)=C_{\phi }(v_0)\cup \{c_0\}\). Since \(C_{\phi }(v_0)\not \subseteq C_{\phi }(y)\), it is immediate to derive that \(C_{\pi }(v_0)= C_{\phi }(v_0)\cup \{c_0\} \not \subseteq C_{\phi }(y) = C_{\pi }(y)\). Conversely, there is a color \(b\in R(v_0)\) such that \(b\in C_{\phi }(y){\setminus } C_{\phi }(v_0)\) and \(c_0\ne b\). This implies that \(C_{\pi }(y) \not \subseteq C_{\pi }(v_0)\). Hence, \((*)\) holds.

Case 3. \(x=v\) and \(y=v_i\) for some \(i\in [0,k-1]\).

Then \(C_{\pi }(v)=B\) and \(C_{\pi }(v_i)=C_{\phi }(v_i)\cup \{c_i\}\). Let \(i\in [0,k-1]\). Since \(d_G(v)=k=\delta (G)\le d_G(v_i)\), it is easy to derive that \(C_{\pi }(v_i)\not \subset C_{\pi }(v)=B\). Conversely, suppose that \(B\subset C_{\pi }(v_i)\). We discuss three possibilities to get a contradiction.

  • \(i=0.\) Since \(B_{k-2}=\{c_0, \ldots ,c_{k-2}\} \subset B \subset C_{\pi }(v_0)\), it follows that \(c_{k-1}\in C{\setminus } (U_{k-1}\cup C_{\phi }(v_0))\) by Step 1. This implies that \(c_{k-1}\notin C_{\phi }(v_0)\), which contradicts the assumption that \(c_{k-1}\in B_{k-1}=B\subset C_{\pi }(v_0)\).

  • \(i=1.\) Step 0 implies that \(c_0\in C{\setminus } (U_0\cup C_{\phi }(v_1))\). Thus, \(c_0\notin C_{\pi }(v_1)\). Since \(c_0\in B\), it follows that \(B\not \subset C_{\pi }(v_1)\), which is a contradiction.

  • \(i\in [2,k-1].\) Note that \(B_{i-2}=\{c_0, \ldots ,c_{i-2}\}\subset B\subset C_{\pi }(v_{i})\). By Step 1, \(c_{i-1}\in C\setminus (U_{i-1}\cup C_{\phi }(v_i))\), implying \(c_{i-1}\notin C_{\phi }(v_i)\), which contradicts the assumption that \(c_{i-1}\in B_{i-1}\subset B\subset C_{\pi }(v_i)\).

By Claims 1 and 2, \(\pi \) is a K-LNDE-coloring of G. \(\square \)

Theorem 2.3 follows immediately from Theorem 2.2 and Proposition 1:

Theorem 2.3

Every formal graph G has \(\chi '_\textrm{snd}(G)\le 3\Delta -1\).

3 General Planar Graphs

Assume that G is a plane graph. A cycle \(C^*\) of G is called a separating cycle if there exist at least one vertex in the interior and exterior of \(C^*\), respectively.

A bunch B(xym) of length \(m\ge 3\) with x and y as poles is defined as m paths \(Q_1, Q_2,\ldots ,Q_m\) having the following properties, as shown in 1:

  1. (a)

    Each \(Q_i\) has length 1 or 2 and joins x and y;

  2. (b)

    For each \(i\in [1,m-1]\), the cycle formed by \(Q_i\) and \(Q_{i+1}\) is not separating;

  3. (c)

    This sequence of paths is maximal, that is, there does not exist a path \(Q_0\) (or \(Q_{m+1}\)) that can be added to B(xym) with conditions (a) and (b) preserved.

Fig. 1
figure 1

a B(xym) with \(xy\notin E(G)\); b B(xym) with \(xy\in E(G)\)

If the length of \(Q_i\) is 2, say \(P_i=xz_iy\), then \(z_i\) is called a brother. If \(Q_i=xy\), then xy is called a parental edge. Assume that \(z_i\) exists. We further say that \(z_i\) is an external brother if \(i\in \{1,m\}\), an internal brother if \(i\in [2,m-1]\), and a strictly internal brother if \(i\in [3,m-2]\). It is easy to see that each internal brother is of degree 2, 3, or 4 and adjacent only to the poses and possibly to one or two of brothers.

Borodin et al. [4] introduced the concept of the bunch in a plane graph and established a structural theorem on plane graphs. For our purposes, we only give the following simplified version of their theorem.

Lemma 3.1

([4]) Every plane graph G with \(\delta (G)\ge 2\) contains one of the following configurations:

  • (A1) a k-vertex v, \(k\in [2,5]\), with neighbors \(v_1, \ldots ,v_{k}\) such that \(d_G(v_i)\le 25\) for all \(i\in [1,k-1]\), and \(d_G(v_1)+ \cdots +d_G(v_{k-1})\le 38\);

  • (A2) a bunch B(xym) with \(d_G(x)\ge 26\) and \(m\ge 0.2d_G(x)\).

Using Lemma 3.1, we can obtain an upper bound of the local neighbor-distinguishing index of planar graphs.

Theorem 3.2

If G is a planar graph, then \(\chi '_\textrm{lnd}(G)\le \lceil 2.8\Delta \rceil +4\).

Proof

The proof proceeds by induction on ||G||. If \(||G||\le \lceil 2.8\Delta \rceil +4\), then the result holds trivially. Let G be a planar graph with \(||G||\ge \lceil 2.8\Delta \rceil +5\ge 5\). Without loss of generality, suppose that G is connected and embedded in the plane. If \(\Delta \le 29\), then \(\chi '_\textrm{lnd}(G)\le 3\Delta -1\le \lceil 2.8\Delta \rceil +4\) by Theorem 2.2. So suppose that \(\Delta \ge 30\), which implies that \(||G||\ge \lceil 2.8\Delta \rceil +4\ge 88\). In what follows, let \(C=[1,K]\), where \(K=\lceil 2.8\Delta \rceil +4\), be the set of K colors.

If G contains a 1-vertex v, then the graph \(G-v\) admits a K-LNDE-coloring \(\phi \) using the color set C by the induction hypothesis. Similarly to the proof of Theorem 2.2, \(\phi \) can be extended to G.

So suppose that \(\delta (G)\ge 2\). By Lemma 3.1, G contains the configurations (A1) or (A2). Our proof is split into the following cases.

Case 1. G contains a k-vertex v, \(k\in [2,5]\), with neighbors \(v_1,\ldots ,v_{k}\) such that \(d_G(v_i)\le 25\) for all \(i\in [1,k-1]\), and \(d_G(v_1)+\cdots +d_G(v_{k-1})\le 38\).

Without loss of generality, assume that \(d_G(v_1)\le \cdots \le d_G(v_k)\). Depending on the size of k, we have some subcases to be considered below.

Case 1.1. \(k=2\).

Note that \(d_G(v_1)\le 25\). Let \(H=G-v\), which admits a K-LNDE-coloring \(\phi \) using C. Based on \(\phi \), we color \(vv_2\) with a color \(a\in C\setminus (C^+_{\phi }(v_2)\cup C_{\phi }(v_1))\), where \(C^+_{\phi }(v_2)=C_{\phi }(v_2)\cup R(v_2)\) as defined in 2, and \(vv_1\) with a color \(b\in C{\setminus } (C^+_{\phi }(v_1) \cup C_{\phi }(v_2)\cup \{a\})\). Since \(|C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1))|\ge \lceil 2.8\Delta \rceil +4 -2(d_G(v_2)-1)-(d_G(v_1)-1)\ge 2.8\Delta +7-2\Delta -25=0.8\Delta -18 \ge 6\) and \(|C{\setminus } (C^+_{\phi }(v_1) \cup C_{\phi }(v_2)\cup \{a\})|\ge \lceil 2.8\Delta \rceil +4 - 2(25-1)-(\Delta -1)-1\ge 1.8\Delta -44\ge 10\), both a and b exist and hence \(\phi \) is extended to G.

By Case 1.1, we may assume that \(d_G(v_i)\ge 3\) for all \(i\in [1,k]\) in the subsequent discussion.

Case 1.2. \(k=3\).

Note that \(d_G(v_2)\le 25\), and since \(d_G(v_1)+d_G(v_2)\le 38\), we derive that \(d_G(v_1)\le 19\). Let \(H=G-\{vv_1, vv_2\}\), which has a K-LNDE-coloring \(\phi \) using C. Suppose that \(\phi (vv_3)=1\). We color \(vv_2\) with \(a\in C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup \{1\})\) and \(vv_1\) with \(b\in C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup C_{\phi }(v_3)\cup \{a\})\). It is easy to calculate that \(|C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup \{1\})|\ge \lceil 2.8\Delta \rceil +4-2(\Delta -1)-(d_G(v_1)-1)-1 \ge 0.8\Delta +6-d_G(v_1)\ge 0.8\Delta +6-19\ge 11\) and \(|C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup C_{\phi }(v_3)\cup \{a\})|\ge \lceil 2.8\Delta \rceil +4-2(d_G(v_1)-1)-(d_G(v_2)-1)-\Delta -1\ge 1.8\Delta +6-(d_G(v_1)+d_G(v_2))-d_G(v_1) \ge 1.8\Delta +6-38-19\ge 3\). Hence, both a and b exist. Let \(\phi '\) denote the resultant coloring after \(vv_1\) and \(vv_2\) are colored. Obviously, \(\phi '\) is a proper edge-K-coloring of G. Note that \(C_{\phi '}(v)=\{1,a,b\}\), \(a\notin C_{\phi '}(v_1)\), \(b\notin C_{\phi '}(v_2)\), and \(a,b\notin C_{\phi '}(v_3)\). Since \(d_G(v_i)\ge 3\) for \(i\in [1,3]\), v is exclusive with each of its neighbors. Consequently, \(\phi '\) is a K-LNDE-coloring of G.

Case 1.3. \(k=4\).

Since \(d_G(v_1)+d_G(v_2)+d_G(v_3)\le 38\) and \(d_G(v_i)\ge 3\) for \(i\in [1,4]\), it follows that \(d_G(v_1)\le 12\), \(d_G(v_2)\le 17\), and \(d_G(v_3)\le 25\). Let \(H=G-\{vv_1, vv_2,vv_3\}\), which has a K-LNDE-coloring \(\phi \) using C and with \(\phi (vv_4)=1\). We color \(vv_3\) with \(a\in C{\setminus } (C^+_{\phi }(v_3)\cup C_{\phi }(v_1)\cup C_{\phi }(v_2)\cup \{1\})\), \(vv_2\) with \(b\in C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup C_{\phi }(v_3)\cup C_{\phi }(v_4)\cup \{a\})\), and \(vv_1\) with \(c\in C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup C_{\phi }(v_3)\cup C_{\phi }(v_4)\cup \{a,b\})\). It is easy to check that \(|C{\setminus } (C^+_{\phi }(v_3)\cup C_{\phi }(v_1)\cup C_{\phi }(v_2)\cup \{1\})|\ge 88-2(d_G(v_3)-1)-(d_G(v_2)-1)-(d_G(v_1)-1)-1= 91-d_G(v_3)-(d_G(v_1)+d_G(v_2)+d(v_3))\ge 91-25-38=28\), \(|C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup C_{\phi }(v_3)\cup C_{\phi }(v_4)\cup \{a\})|\ge \lceil 2.8\Delta \rceil +4-2(d_G(v_2)-1)-(d_G(v_1)-1)-(d_G(v_3)-1)-\Delta -1\ge 1.8\Delta +7-d_G(v_2)-(d_G(v_1)+d_G(v_2)+d_G(v_3))\ge 1.8\Delta +7-17-38\ge 6\), and \(|C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup C_{\phi }(v_3)\cup C_{\phi }(v_4)\cup \{a,b\})|\ge \lceil 2.8\Delta \rceil +4- 2(d_G(v_1)-1)-(d_G(v_2)-1)-(d_G(v_3)-1)-\Delta -2\ge 1.8\Delta +6-d_G(v_1)-(d_G(v_1)+d_G(v_2)+d_G(v_3))\ge 1.8\Delta +6-12-38\ge 10\). Hence, \(vv_1,vv_2,vv_3\) can be colored properly. Let \(\phi '\) denote the resultant coloring of G. It is easy to observe that \(C_{\phi '}(v)=\{1,a,b,c\}\), and \(b,c\notin C_{\phi '}(v_4)\), \(b,c\notin C_{\phi '}(v_3)\), \(a,c\notin C_{\phi '}(v_2)\), and \(a,b\notin C_{\phi '}(v_1)\). Since \(d_G(v_i)\ge 3\) for \(i\in [1,4]\), v is exclusive with each of its neighbors in \(\phi '\). Consequently, \(\phi '\) is a K-LNDE-coloring of G.

Case 1.4. \(k=5\).

Since \(d_G(v_1)+ \cdots +d_G(v_4)\le 38\) and \(d_G(v_i)\ge 3\) for \(i\in [1,5]\), it is immediate to deduce that \(d_G(v_1)\le 9\), \(d_G(v_2)\le 11\), \(d_G(v_3)\le 16\), and \(d_G(v_4)\le 25\). Let \(H=G-\{vv_1, vv_2,vv_3,vv_4\}\), which has a K-LNDE-coloring \(\phi \) using C such that \(\phi (vv_5)=1\). Define the sets \(M_4=\bigcup \limits _{i=1}^{4} C_{\phi }(v_i)\) and \(M_5=M_4\cup C_{\phi }(v_5)\). We have to consider two possibilities as follows.

Case 1.4.1. \(d_G(v_5)\ge 4\).

We color \(vv_4\) with \(a\in C{\setminus } (M_4\cup R(v_4)\cup \{1\})\), \(vv_3\) with \(b\in C{\setminus } ( M_4\cup R(v_3) \cup \{1,a\})\), \(vv_2\) with \(c\in C\setminus (M_5\cup R(v_2)\cup \{a,b\})\) and \(vv_1\) with \(d\in C\setminus (M_5\cup R(v_1) \cup \{a,b,c\})\). It is easy to calculate that \(|C{\setminus } (M_4\cup R(v_4)\cup \{1\})|\ge 88-2(d_G(v_4)-1)-(d_G(v_1)-1)-(d_G(v_2)-1)-(d_G(v_3)-1)-1=92-d_G(v_4)-(d_G(v_1)+d_G(v_2)+d_G(v_3)+d_G(v_4))\ge 92-25-38=29\), \(|C{\setminus } ( M_4\cup R(v_3) \cup \{1,a\})|\ge 88-2(d_G(v_3)-1)-(d_G(v_1)-1)- (d_G(v_2)-1)-(d_G(v_4)-1)-2=91-d_G(v_3)-(d_G(v_1)+d_G(v_2)+d_G(v_3)+d_G(v_4))\ge 91-16-38=37\), \(|C{\setminus } (M_5\cup R(v_2)\cup \{a,b\})|\ge \lceil 2.8\Delta \rceil +4-2(d_G(v_2)-1)-(d_G(v_1)-1)-(d_G(v_3)-1)-(d_G(v_4)-1)-d_G(v_5)-2 \ge 1.8\Delta + 7-d_G(v_2)-(d_G(v_1)+d_G(v_2)+d_G(v_3)+d_G(v_4))\ge 1.8\Delta +7-11-38\ge 12\), and \(|C{\setminus } (M_5\cup R(v_1) \cup \{a,b,c\})|\ge \lceil 2.8\Delta \rceil +4-2(d_G(v_1)-1)-(d_G(v_2)-1)-(d_G(v_3)-1)-(d_G(v_4)-1)-d_G(v_5)-3 \ge 1.8\Delta + 6-d_G(v_1)-(d_G(v_1)+d_G(v_2)+d_G(v_3) +d_G(v_4))\ge 1.8\Delta +6-9-38\ge 13\). Thus, the resultant coloring, denoted \(\phi '\), is a proper edge-K-coloring of G. Observe that \(C_{\phi '}(v)=\{1,a,b,c,d\}\), \(c,d\notin C_{\phi '}(v_5)\), \(c,b,d\notin C_{\phi '}(v_4)\), \(a,c,d\notin C_{\phi '}(v_3)\), \(a,b,d\notin C_{\phi '}(v_2)\), and \(a,b,c\notin C_{\phi '}(v_1)\). Since \(d_G(v_5)\ge 4\) and \(d_G(v_i)\ge 3\) for \(i\in [1,4]\), v is exclusive with each of its neighbors in \(\phi '\). Hence, \(\phi '\) is a K-LNDE-coloring of G.

Case 1.4.2. \(d_G(v_5)=3\).

It follows that \(d_G(v_i)=3\) for all \(i\in [1,4]\). It is evident that \(|M_5|\le 2\times 4+3=11\). We color \(vv_4\) with \(a\in C{\setminus } (M_5\cup R(v_4))\), \(vv_3\) with \(b\in C{\setminus } (M_5\cup R(v_3)\cup \{a\})\), \(vv_2\) with \(c\in C{\setminus } (M_5\cup R(v_2)\cup \{a,b\})\), and \(vv_1\) with \(d\in C{\setminus } (M_5\cup R(v_1)\cup \{a,b,c\})\). Then \(|C{\setminus } (M_5\cup R(v_4))|\ge 88-11-2=75\), \(|C{\setminus } (M_5\cup R(v_3)\cup \{a\})|\ge 88-11-2-1=74\), \(|C{\setminus } (M_5\cup R(v_2)\cup \{a,b\})|\ge 88-11-2-2 =73\), \(|C{\setminus } (M_5\cup R(v_1)\cup \{a,b,c\})|\ge 88-11-2-3=72\). It is easy to check that the extended coloring is a K-LNDE-coloring of G.

Case 2. G contains a bunch B(xym) with \(d_G(x)\ge 26\) and \(m\ge 0.2d_G(x)\).

Here we use directly the notation in the definition of B(xym), as shown in 1. Since \(d_G(x)\ge 26\), it follows that \(m\ge 6\). We need to deal with the following two subcases.

Case 2.1. There exist two adjacent vertices u and w such that \(3\le d_G(u)\le d_G(w)\le 4\).

Case 2.1.1. \(d_G(u)=3\).

Let st be the neighbors of u other than w. In view of the proof of Case 1.1, we may assume that \(d_G(s),d_G(t)\ge 3\). Let \(H=G-\{uw,us\}\), which has a K-LNDE-coloring \(\phi \) using C such that \(\phi (ut)=1\). We color us with \(a\in C{\setminus } (C^+_{\phi }(s)\cup C_{\phi }(w)\cup \{1\})\) and uw with \(b\in C{\setminus } (C^+_{\phi }(w)\cup C_{\phi }(s)\cup C_{\phi }(t)\cup \{a\})\). It is easy to check that \(|C{\setminus } (C^+_{\phi }(s)\cup C_{\phi }(w)\cup \{1\})|\ge \lceil 2.8\Delta \rceil +4-2(d_G(s)-1)-(d_G(w)-1)-1\ge 0.8 \Delta + 6 -d_G(w)\ge 0.8\Delta +6-4=26\) and \(|C{\setminus } (C^+_{\phi }(w)\cup C_{\phi }(s)\cup C_{\phi }(t)\cup \{a\})|\ge \lceil 2.8\Delta \rceil +4-2(d_G(w)-1)-(d_G(s)-1)-d_G(t)-1\ge 0.8\Delta +6-2\times 4\ge 22\). Hence, the resultant coloring \(\phi '\) is a proper edge-K-coloring of G. Since \(C_{\phi '}(u)=\{1,a,b\}\), \(a\notin C_{\phi '}(w)\), and \(b\notin C_{\phi '}(s)\cup C_{\phi '}(t)\), u is exclusive with each of its neighbors in \(\phi \). Thus, \(\phi \) is extended to G.

Case 2.1.2. \(d_G(u)=d_G(w)=4\).

Let stz be the neighbors of u other than w. By Case 2.1.1, assume that \(d_G(s),d_G(t),d_G(z)\ge 4\). Let \(H=G-\{uw,us\}\), which has a K-LNDE-coloring \(\phi \) using C such that \(\phi (ut)=1\) and \(\phi (uz)=2\). Since \(d_H(u)=2\), we see that \(1\notin C_{\phi }(z)\) and \(2\notin C_{\phi }(t)\). We color us with \(a\in C{\setminus } (C^+_{\phi }(s)\cup C_{\phi }(w)\cup \{1,2\})\) and uw with \(b\in C{\setminus } (C^+_{\phi }(w)\cup C_{\phi }(s) \cup \{1,2,a\})\). Since \(|C{\setminus } (C^+_{\phi }(s)\cup C_{\phi }(w)\cup \{1,2\})|\ge \lceil 2.8\Delta \rceil +4-2(d_G(s)-1)- (d_G(w)-1)- 2\ge 0.8\Delta +5-d_G(w)=0.8\Delta +5-4\ge 25\) and \(|C{\setminus } (C^+_{\phi }(w)\cup C_{\phi }(s) \cup \{1,2,a\})|\ge \lceil 2.8\Delta \rceil +4 -2(d_G(w)-1)- (d_G(s)-1)-3 \ge 1.8\Delta +4-2\times 4\ge 50\), uw and us can be colored properly. Denote by \(\phi '\) the resultant coloring. Noting that \(C_{\phi '}(u)=\{1,2,a,b\}\), \(1\notin C_{\phi '}(z)\), \(2\notin C_{\phi '}(t)\), \(a\notin C_{\phi '}(w)\), and \(b\notin C_{\phi '}(s)\), we obtain a K-LNDE-coloring of G.

Case 2.2. All strictly internal brothers are of degree 2 in G.

Let S denote the set of brothers \(z_i\)’s with \(d_G(z_i)=2\) in B(xym). Obviously, S contains all strictly internal brothers of B(xym). Since \(d_G(x)\ge 26\) and \(m\ge 0.2d_G(x)>5\), B(xym) has at least one strictly internal brother. Thus, \(s:=|S|\ge m-5\ge 1\). Let \(H=G-S\), which has a K-LNDE-coloring \(\phi \) using C. Let \(E_x=\{wx\,|\,w\in S\}\) and \(E_y=\{wy\,|\,w\in S\}\). For each edge \(e_x\in E_x\) and each edge \(e_y\in E_y\), we define a list assignment L as follows:

\(L(e_x)=C\setminus (C^+_{\phi }(x)\cup C_{\phi }(y))\), \(L(e_y)=C\setminus (C^+_{\phi }(y)\cup C_{\phi }(x))\).

First suppose that \(xy\notin E(G)\). Then \(s\ge m-4\ge 2\). It is easy to compute that \(|L(e_y)|\ge \lceil 2.8\Delta \rceil +4- 2(d_G(y)-s)-(d_G(x)-s)\ge 2.8\Delta +4 +3\,s-2d_G(y)-d_G(x)\ge 0.8\Delta +4+s+2\,s-d_G(x)\ge 0.8\Delta +4+s+2(m-4)-d_G(x)\ge 0.8\Delta -4+s+2\times (0.2d_G(x))-d_G(x) =0.8\Delta -4+s-0.6d_G(x)\ge 0.2\Delta -4+s\ge s+2\), and \(|L(e_x)|\ge \lceil 2.8\Delta \rceil +4- 2(d_G(x)-s)-(d_G(y)-s)\ge 1.8\Delta +4 +2\,s +s-2d_G(x)\ge 1.8\Delta +4+2\,s+(m-4)-2d_G(x)= 1.8\Delta +2\,s+m-2d_G(x)\ge 1.8\Delta +2\,s+0.2d_G(x)-2d_G(x)\ge 2\,s\).

Next suppose that \(xy\in E(G)\). In this case, \(s\ge m-5\ge 1\). Because \(xy\in E(G)\), we have \(\phi (xy)\in C_{\phi }(x)\cap C_{\phi }(y)\) and hence \(|C_{\phi }(x)\cap C_{\phi }(y)|\ge 1\). So, \(|L(e_y)|\ge \lceil 2.8\Delta \rceil +4- 2(d_G(y)-s)-(d_G(x)-s)+1\ge 2.8\Delta +5 +3\,s-2d_G(y)-d_G(x)\ge 0.8\Delta +5+s+2\,s-d_G(x)\ge 0.8\Delta +5+s+2(m-5)-d_G(x)\ge 0.8\Delta -5+s+2\times (0.2d_G(x))-d_G(x) =0.8\Delta -5+s-0.6d_G(x)\ge 0.2\Delta -5+s\ge s+1\), and \(|L(e_x)|\ge \lceil 2.8\Delta \rceil +4- 2(d_G(x)-s)-(d_G(y)-s)+1\ge 1.8\Delta +5 +2\,s +s-2d_G(x)\ge 1.8\Delta +5+2\,s+(m-5)-2d_G(x)= 1.8\Delta +2\,s+m-2d_G(x)\ge 1.8\Delta +2\,s+0.2d_G(x)-2d_G(x)\ge 2\,s\).

In each of the above two cases, we first color the edges in \(E_y\) with distinct colors in \(L(e_y)\) and then use \(C_y\) to denote the set of colors assigned to the edges in \(E_y\). Then we color the edges in \(E_x\) with distinct colors in \(L(e_x)\setminus C_y\). It is easy to check that the resultant coloring is a K-LNDE-coloring of G. \(\square \)

By Proposition 1, we have the following:

Theorem 3.3

If G is a formal planar graph, then \(\chi '_\textrm{snd}(G)\le \lceil 2.8\Delta \rceil +4\).

4 Planar Graphs Without 4-Cycles

For the class of planar graphs without 4-cycles, we can show that Conjecture is almost true (away from a constant). To achieve this goal, we need to apply the following structural lemma.

Lemma 4.1

([18]) Let G be a planar graph with \(\delta (G)\ge 2\) and without 4-cycles. Then G contains a k-vertex v, \(k\in [2,4]\), whose neighbors \(v_1,\ldots ,v_k\) satisfy one of the following conditions, assuming \(d_G(v_1)\le \cdots \le d_G(v_k):\)

  1. (1)

    \(k=2\) and \(d_G(v_1)\le 11\);

  2. (2)

    \(k=3\) and \(d_G(v_1)+d_G(v_2)\le 14\);

  3. (3)

    \(k=4\) and \(d_G(v_1)+d_G(v_2)+d_G(v_3)\le 15\).

Theorem 4.2

If G is a planar graph without 4-cycles, then \(\chi '_\textrm{lnd}(G)\le 2\Delta +10\).

Proof

The proof proceeds by induction on ||G||. If \(||G||\le 2\Delta +10\), then the result holds trivially. Let G be a connected planar graph with \(||G||\ge 2\Delta +11\ge 11\). If \(\Delta \le 11\), then \(\chi '_\textrm{lnd}(G)\le 3\Delta -1\le 2\Delta +10\) by Theorem 2.2. So suppose that \(\Delta \ge 12\). Again, let \(K=2\Delta +10\) and \(C=[1,K]\) denote a set of K colors. Hence, \(|C|=K=2\Delta +10\ge 34\).

First assume that \(\delta (G)=1\). Let u be a 1-vertex adjacent to a vertex v. Then \(d_G(v)\ge 2\) by the assumption. Let \(H=G-u\), which has a K-LNDE-coloring \(\phi \) using C. We color uv with a color \(a\in C{\setminus } C^+_{\phi }(v)\). Since \(|C{\setminus } C^+_{\phi }(v)|\ge 2\Delta +10-2(d_G(v)-1)\ge 2\Delta +10-2(\Delta -1)=12\), \(\phi \) is extended to G.

Next assume that \(\delta (G)\ge 2\). By Lemma 4.1, G contains a k-vertex v, \(k\in [2,4]\), whose neighbors \(v_1, \ldots ,v_k\) satisfy one of the conditions (1) to (3), where \(d_G(v_1)\le \cdots \le d_G(v_k)\). By the above proof, we may assume that \(d_G(v_i)\ge 2\) for all \(i\in [1,k]\).

Case 1. \(k=2\) and \(d_G(v_1)\le 11\).

Let \(H=G-v\), which admits a K-LNDE-coloring \(\phi \) using C. We have to discuss two possibilities.

  • \(d_G(v_1)=2\). Let y be the neighbor of \(v_1\) other than v. We color \(vv_2\) with \(a\in C{\setminus } (C^+_{\phi }(v_2)\cup \{\phi (yv_1)\})\) and \(vv_1\) with \(b\in C{\setminus } (C_{\phi }(y)\cup C_{\phi }(v_2)\cup \{a\})\). Since \(|C{\setminus } (C^+_{\phi }(v_2)\cup \{\phi (v_1y)\})|\ge 2\Delta +10-2(d_G(v_2)-1)-1\ge 11\) and \(|C{\setminus } (C_{\phi }(y)\cup C_{\phi }(v_2)\cup \{a\})|\ge 2\Delta +10-\Delta -(\Delta -1)-1=10\), \(\phi \) is extended to G.

  • \(d_G(v_1)\ge 3\). We color \(vv_2\) with \(a\in C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1))\) and \(vv_1\) with \(b\in C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup \{a\})\). Since \(|C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1))|\ge 2\Delta +10-2(\Delta -1)-10 \ge 2\) and \(|C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup \{a\})|\ge 2\Delta +10-2(d_G(v_1)-1)-\Delta \ge \Delta +10-2\times 10\ge 2\), \(\phi \) is extended to G.

Now, by Case 1, we may assume that \(d_G(v_i)\ge 3\) for all \(i\in [1,k]\) in the following two situations.

Case 2. \(k=3\) and \(d_G(v_1)+d_G(v_2)\le 14\).

It follows that \(d_G(v_1)\le 7\) and \(d_G(v_2)\le 11\). Let \(H=G-\{vv_1, vv_2\}\), which has a K-LNDE-coloring \(\phi \) using C such that \(\phi (vv_3)=1\). We color \(vv_2\) with \(a\in C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup \{1\})\) and \(vv_1\) with \(b\in C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup C_{\phi }(v_3)\cup \{a\})\). Since \(|C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup \{1\})|\ge 2\Delta +10-2(\Delta -1)-(d_G(v_1)-1)-1 \ge 12-(7-1)-1=5\) and \(|C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup C_{\phi }(v_3)\cup \{a\})|\ge 2\Delta +10-2(d_G(v_1)-1)-(d_G(v_2)-1)-\Delta -1\ge \Delta +12-d_G(v_1)- (d_G(v_1)+d_G(v_2)) \ge 24-7-14=3\), both a and b exist. Let \(\phi '\) denote the resultant coloring. Then \(C_{\phi '}(v)=\{1,a,b\}\), \(a\notin C_{\phi '}(v_1)\), and \(b\notin C_{\phi '}(v_2)\cup C_{\phi '}(v_3)\). Thus, v is exclusive with each of its neighbors and hence \(\phi \) is extended to G.

Case 3. \(k=4\) and \(d_G(v_1)+d_G(v_2)+d_G(v_3)\le 15\).

Then \(d_G(v_1)\le 5\), \(d_G(v_2)\le 6\), and \(d_G(v_3)\le 9\). Let \(H=G-\{vv_1, vv_2,vv_3\}\), which has a K-LNDE-coloring \(\phi \) using C such that \(\phi (vv_4)=1\). We color \(vv_3\) with \(a\in C{\setminus } (C^+_{\phi }(v_3)\cup C_{\phi }(v_1)\cup C_{\phi }(v_2)\cup \{1\})\), \(vv_2\) with \(b\in C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup C_{\phi }(v_3)\cup C_{\phi }(v_4)\cup \{a\})\), and \(vv_1\) with \(c\in C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup C_{\phi }(v_3)\cup C_{\phi }(v_4)\cup \{a,b\})\). Noting that \(|C{\setminus } (C^+_{\phi }(v_3)\cup C_{\phi }(v_1)\cup C_{\phi }(v_2)\cup \{a\})|\ge 34-2(d_G(v_3)-1)-(d_G(v_2)-1)-(d_G(v_1)-1)-1= 37-d_G(v_3)-(d_G(v_1)+d_G(v_2)+d_G(v_3))\ge 37-9-15=13\), \(|C{\setminus } (C^+_{\phi }(v_2)\cup C_{\phi }(v_1)\cup C_{\phi }(v_3)\cup C_{\phi }(v_4)\cup \{a\})|\ge 2\Delta +10-2(d_G(v_2)-1)-(d_G(v_1)-1)-(d_G(v_3)-1)-\Delta -1\ge \Delta +13-d_G(v_2)-(d_G(v_1)+d_G(v_2)+d_G(v_3))\ge \Delta +13-6-15\ge 4\), and \(|C{\setminus } (C^+_{\phi }(v_1)\cup C_{\phi }(v_2)\cup C_{\phi }(v_3)\cup C_{\phi }(v_4)\cup \{a,b\})|\ge 2\Delta +10- 2(d_G(v_1)-1)-(d_G(v_2)-1)-(d_G(v_3)-1)-\Delta -2\ge \Delta +12-d_G(v_1)-(d_G(v_1)+d_G(v_2)+d_G(v_3))\ge \Delta +12-5-15\ge 4\), \(vv_1,vv_2,vv_3\) can be properly colored. Let \(\phi '\) denote the resultant coloring. Since \(C_{\phi '}(v)=\{1,a,b,c\}\), \(a,b\notin C_{\phi '}(v_1)\), \(a,c\notin C_{\phi '}(v_2)\), \(c,b\notin C_{\phi '}(v_3)\), and \(b,c\notin C_{\phi '}(v_4)\), \(\phi '\) is a K-LNDE-coloring of G.

By Proposition 1, the following theorem holds automatically.

Theorem 4.3

If G is a formal planar graph without 4-cycles, then \(\chi '_\textrm{snd}(G)\le 2\Delta +10\).

5 Planar Graphs with [2, k]-Factors

For two positive integers \(k_1,k_2\) with \(k_2\ge k_1\), a spanning subgraph F of a graph G is called an \([k_1,k_2]\)-factor if \(k_1\le d_F(v)\le k_2\) for all \(v\in V(G)\). Tutte [15] showed that every 4-connected planar graph is Hamiltonian, i.e., it has a 2-connected [2, 2]-factor. By relaxing the 4-connected condition, Gao [7] showed that every 3-connected planar graph has a 2-connected [2, 6]-factor. Enomoto et al. [5] extended this result by showing that every 3-connected planar graph G with \(\delta (G)\ge 4\) has a 2-connected [2, 3]-factor. Both numbers 6 and 3 in these two results are best possible with respect to the required conditions.

The core \(G_{\Delta }\) of a graph G is the subgraph of G induced by \(\Delta \)-vertices.

Lemma 5.1

Let \(k\ge 3\). If a connected graph G has a connected [2, k]-factor, then G contains a connected [2, k]-factor F whose core is acyclic.

Proof

Let F be a connected [2, k]-factor of G with the least number of edges. We claim that the core of F is acyclic. Suppose to the contrary that \(F_{\Delta }\) contains a cycle C. Let \(e=xy\in E(C)\) be an arbitrary edge, and set \(F'=F-e\). Obviously, \(F'\) is a connected spanning subgraph of G. If \(v\in V(G){\setminus } \{x,y\}\), then \(d_{F'}(v)=d_F(v)\). If \(v\in \{x,y\}\), then \(d_{F'}(v)=d_F(v)-1\ge k-1\ge 2\). It follows that \(F'\) is a connected [2, k]-factor of G with \(||F'||<||F||\), which contradicts the choice of F. \(\square \)

The following result can be derived from Lemma 5.1 and the results in [5] and [7].

Corollary 5.2

Let G be a planar graph.

  1. (1)

    If G is 3-connected, then G contains a connected [2, 6]-factor F whose core is acyclic.

  2. (2)

    If G is 3-connected and \(\delta (G)\ge 4\), then G contains a connected [2, 3]-factor F whose core is acyclic.

The celebrated Vizing’s Theorem gives a tight upper bound for the chromatic index of a simple graph.

Theorem 5.3

([16]) Every simple graph G has \(\Delta \le \chi '(G)\le \Delta +1\).

A simple graph G is said to be Class 1 if \(\chi '(G)=\Delta \) and Class 2 if \(\chi '(G)=\Delta +1\).

Theorem 5.4

([6]) If the core of a simple graph G is a forest, then G is Class 1.

Theorem 5.5

([14]) If G is a planar graph with \(\Delta \ge 7\), then G is Class 1.

An edge-partition of a graph G is a decomposition of G into subgraphs \(G_1, \ldots ,G_m\) such that \(E(G)=E(G_1)\cup \cdots \cup E(G_m)\) and \(E(G_i)\cap E(G_j)=\emptyset \) for \(i\ne j\).

Suppose that H is a subgraph of a graph G. A restricted-strong edge-k-coloring of H on G is an edge-coloring \(\phi : E(H)\rightarrow [1,k]\) such that any two edges \(e_1,e_2\in E(H)\) having distance at most two in G get distinct colors. The restricted-strong chromatic index of H on G, denoted \(\chi '_\textrm{s}(H|_G)\), is the smallest integer k such that H has a restricted-strong edge-k-coloring on G.

Since all planar graphs G considered in Theorems 5.6 to 5.11 contain a [2, k]-factor, it follows that \(\delta (G)\ge 2\), which implies that \(\chi '_\textrm{lnd}(G)=\chi '_\textrm{snd}(G)\) by Proposition 1.

Theorem 5.6

Suppose that a connected graph G can be edge-partitioned into two graphs F and H such that F is a connected [2, k]-factor of G with \(k\ge 2\). Then \(\chi '_\textrm{snd}(G)\le \chi '(H)+\chi '_\textrm{s}(F|_G).\)

Proof

Note that \(2\le \delta (F)\le \Delta (F)\le k\) and \(\Delta (H)\le \Delta (G)-\delta (F)\le \Delta (G)-2\). Let \(\chi '(H)=l\) and \(\chi '_\textrm{s}(F|_G)=m\). Let \(\phi \) be an edge-l-coloring of H using the color set \(C_1=[1,l]\) and \(\pi \) be a restricted-strong-edge-coloring of F on G using the color set \(C_2=[l+1,l+m]\). We define an edge-coloring f of G as follows: \(f(e)=\phi (e)\) for \(e\in E(H)\), and \(f(e)=\pi (e)\) for \(e\in E(F)\). Obviously, f is a proper edge-\((l+m)\)-coloring of G using the color set \(C_1\cup C_2\). If we can show that f is strict neighbor-distinguishing, then it holds naturally that \(\chi '_\textrm{snd}(G)\le l+m=\chi '(H)+\chi '_\textrm{s}(F|_G)\). In fact, for any edge \(e=xy\in E(G)\), since \(d_F(x),d_F(y)\ge 2\), there exist an edge \(e_x\in E(F)\setminus \{e\}\) incident with x and an edge \(e_y\in E(F)\setminus \{e\}\) incident with y. Since \(e_x\) and \(e_y\) have the distance 2 in G, it follows that \(\pi (e_x)\notin C_{\pi }(y)\) and \(\pi (e_y)\notin C_{\pi }(x)\). Because \(C_1\cap C_2=\emptyset \), we have that \(\pi (e_x)\notin C_{f}(y)\) and \(\pi (e_y)\notin C_{f}(x)\), and so x and y are exclusive. \(\square \)

A family of graphs \(\mathcal G\) is called minor-closure if it is closed under deleting vertices, deleting edges, or contracting edges. Let \(\chi (G)\) denote the chromatic number of a graph G, which is defined as the smallest integer k for which the vertices of G can be colored using k colors such that no two adjacent vertices get same color. For a family of minor-closure graphs \(\mathcal{G}\), we define \(\chi (\mathcal{G})=\max \{\chi (G)\,|\,G\in \mathcal{G}\}\). It is easily seen that the family of planar graphs, denoted \(\mathcal{P}\), is minor-closure and \(\chi (\mathcal{P})\le 4\) by the Four-Color Theorem [2].

Lemma 5.7

([22]) If F is a subgraph of a planar graph G, then \(\chi '(F|_G)\le 4\chi '(F)\).

Combining Theorem 5.6 and Lemma 5.7, the following theorem holds automatically.

Theorem 5.8

Let G be a connected planar graph with a connected [2, k]-factor F, and \(H=G-E(F)\). Then \(\chi '_\textrm{snd}(G)\le \chi '(H)+4\chi '(F)\).

Theorem 5.9

Let G be a planar graph. Then the following statements (1) and (2) hold.

(1) If G is 3-connected, then \(\chi '_\textrm{snd}(G)\le \Delta +23\).

(2) If G is 3-connected and \(\delta (G)\ge 4\), then \(\chi '_\textrm{snd}(G)\le \Delta +11\).

Proof

(1) Since G is 3-connected, it follows from Corollary 5.2(1) that G has a connected [2, 6]-factor F whose core is acyclic. Let \(H=G-E(F)\). Then G is edge-partitioned into two subgraphs F and H. If \(\Delta (F)\le 5\), then \(\chi '(F)\le 5+1=6\) by Theorem 5.3. If \(\Delta (F)=6\), then \(\chi '(F)=6\) by Theorem 5.4. Hence, it always holds that \(\chi '(F)\le 6\). On the other hand, since \(\Delta (H)\le \Delta (G)-2\), we have \(\chi '(H)\le \Delta (G)-2+1=\Delta -1\) by Theorem 5.3. So, by Theorem 5.8, \(\chi '_\textrm{snd}(G)\le \chi '(H)+4\chi '(F)\le \Delta (H)+4\times 6\le \Delta -1+24=\Delta +23\). (2) By Corollary 5.2(2), G has a connected [2, 3]-factor F whose core is acyclic. Let \(H=G-E(F)\). Then G is edge-partitioned into two subgraphs F and H. If \(\Delta (F)=2\), then \(\chi '(F)\le 2+1=3\) by Theorem 5.3. If \(\Delta (F)=3\), then \(\chi '(F)=3\) by Theorem 5.4. Hence, we always have that \(\chi '(F)\le 3\). By Theorem 5.3, \(\chi '(H)\le \Delta (H)+1\le \Delta (G)-2+1=\Delta -1\). By Theorem 5.8, \(\chi '_\textrm{snd}(G)\le \chi '(H)+4\chi '(F)\le \Delta -1+4\times 3=\Delta +11\). \(\square \)

Theorem 5.10

If a planar graph G is Hamiltonian, then \(\chi '_\textrm{snd}(G)\le \Delta +6\).

Proof

Let \(C=v_0v_1\cdots v_{n-1}v_0\) be a Hamiltonian cycle of G, where \(n=|V(G)|\). Let \(H=G-E(C)\). Then \(H\cup C\) is an edge-partition of G with \(\Delta (H)\le \Delta -2\). Let \(\chi '(H)=k\). First we give an edge-k-coloring of H using the color set \(B_1=[1,k]\). Then we define an edge-7-coloring \(\pi \) of C using the color set \(B_2=[k+1,k+7]\) in two ways below:

  • Assume that n is even. Set \(M=\{v_0v_1,v_2v_3,\ldots ,v_{n-2}v_{n-1}\}\), and give a restricted-strong edge-4-coloring of M on G using the colors in \([k+1,k+4]\). Afterward, if \(n\equiv 0\) (mod 4), then we color alternatively \(v_1v_2,v_3v_4,\ldots ,v_{n-1}v_0\) with \(k+5\) and \(k+6\). If \(n\equiv 2\) (mod 4), then we color \(v_1v_2\) with \(k+7\), and then color alternatively \(v_3v_4,v_5v_6,\ldots ,v_{n-1}v_0\) with \(k+5\) and \(k+6\).

  • Assume that n is odd. Set \(M=\{v_0v_1,v_2v_3,\ldots ,v_{n-3}v_{n-2}\}\), and give a restricted-strong edge-4-coloring of M on G using \([k+1,k+4]\). Then we color \(v_{n-1}v_0\) with \(k+7\) and then color alternatively \(v_1v_2,v_3v_4,\ldots ,v_{n-2}v_{n-1}\) with \(k+5\) and \(k+6\).

Let f denote the resultant edge-\((k+7)\)-coloring of G formed by combining \(\phi \) and \(\pi \), using the color set \(B_1\cup B_2\). It is easy to inspect that f is proper, i.e., any two adjacent edges having distinct colors. It remains to show that f is strict neighbor-distinguishing. Let \(e=xy\) be an arbitrary edge of G. By the definition of M, at most one of x and y is not incident with any edge in M.

First assume that each of x and y is incident with an edge in M, respectively. Since M is a matching of G, there exist the unique \(e_x\in M\) incident with x and the unique \(e_y\in M\) incident with y. We have two possibilities as follows.

  • \(e_x=e_y\), that is, \(e=e_x=e_y\in M\), say \(e=v_iv_{i+1}\), where indices are taken modulo n. Then \(v_{i-1}v_i,v_{i+1}v_{i+2}\in E(C)\setminus M\). By the definition of \(\pi \), \(\pi (v_{i-1}v_i),\pi (v_{i+1}v_{i+2})\in [k+5,k+7]\) and \(\pi (v_{i-1}v_i)\ne \pi (v_{i+1}v_{i+2})\). Noting that \(\pi (v_{i-1}v_i)\notin C_{f}(v_{i+1})\) and \(\pi (v_{i+1}v_{i+2})\notin C_{f}(v_{i})\), \(v_i\) and \(v_{i+1}\) are exclusive in f.

  • \(e_x\ne e_y\). Since \(e_x\) and \(e_y\) have distance 2 in G, the definition of \(\pi \) implies that \(\pi (e_x)\ne \pi (e_y)\) and \(\pi (e_x),\pi (e_y)\in [k+1,k+4]\). So, \(\pi (e_x)\notin C_f(y)\) and \(\pi (e_y)\notin C_f(x)\), and henceforth x and y are exclusive. Next assume that x is not incident with any edge in M. Then \(x=v_{n-1}\). There are two subcases to be considered.

  • \(y\in \{v_0,v_{n-2}\}\), say \(y=v_0\). Because \(v_0v_1\in M\) satisfies \(\pi (v_0v_1)\in [k+1,k+4]\), and \(v_{n-2}v_{n-1}\in E(C){\setminus } M\) satisfies \(\phi (v_{n-2}v_{n-1})\in [k+5,k+7]\), it follows that \(\pi (v_0v_1)\notin C_f(v_{n-1})\) and \(\pi (v_{n-2}v_{n-1})\notin C_f(v_0)\), and therefore, x and y are exclusive.

  • \(y\in \{v_1,\ldots ,v_{n-2}\}\), say \(y=v_i\) for some \(i\in [1,n-2]\). Then exactly one of \(v_{i-1}v_i\) and \(v_iv_{i+1}\) belongs to M, whose color is \(a\in [k+1,k+4]\). Moreover, assuming that \(\pi (v_{n-1}v_0)=b\) and \(\pi (v_{n-2}v_{n-1})=c\), then \(b,c\in [k+5,k+7]\). Since \(a\notin C_f(v_{n-1})\) and at least one of b and c does not belong to \(C_f(v_i)\), x and y are exclusive.

The above analysis and Theorem 5.3 show that \(\chi '_\textrm{snd}(G)\le k+ 4+3\le \Delta -1+7=\Delta +6.\) \(\square \)

By Theorem 5.5 and the proof of Theorem 5.10, we can obtain the following better result.

Theorem 5.11

If G is a Hamiltonian planar graph with \(|G|\equiv 0\,\mathrm{(mod \ 4)}\) and \(\Delta \ge 9\), then \(\chi '_\textrm{snd}(G)\le \Delta +4\).

A Halin graph is a plane graph \(G=T\cup C\), where T is a plane tree with no vertex of degree two and at least one vertex of degree three or more, and C is a cycle connecting the pendant vertices of T in the cyclic order determined by the drawing of T.

Halin graphs are 3-connected, but any of their proper subgraphs is not. Bondy and Lovász [3] showed that Halin graphs are almost pancyclic with the possible exception of an even cycle. In particular, Halin graphs are Hamiltonian.

By Theorem 5.10, we have the following:

Corollary 5.12

If G is a Halin graph, then \(\chi '_\textrm{snd}(G)\le \Delta +6\).

6 Concluding Remarks

In this section, we are going to provide some open problems on the local neighbor-distinguishing edge-coloring of graphs. In contrast to Conjecture 2, we first put forward the following conjecture:

Conjecture 3

Every connected graph G, different from \(H_{\Delta }\), has \(\chi '_\textrm{lnd}(G)\le 2\Delta \). Observing the local neighbor-distinguishing index of \(K_{2,n}\), we see that the upper bound \(2\Delta \) in Conjecture 3 is tight if it were true.

Problem 1. Does every planar graph satisfy Conjecture \(3\,?\)

Problem 2. Is it true that there exists a constant c such that every planar graph G without 4-cycles has \(\chi '_\textrm{lnd}(G)\le \Delta +c\,?\)