Abstract
DP-coloring is a generalization of list coloring in simple graphs. Many results in list coloring can be generalized in those of DP-coloring. Kim and Ozeki showed that every planar graph without k-cycles where \(k=3,4,5,\) or 6 is DP-4-colorable. Recently, Kim and Yu extended the result on 3- and 4-cycles by showing that every planar graph without triangles adjacent to 4-cycles are DP-4-colorable. Xu and Wu showed that every planar graph without 5-cycles adjacent simultaneously to 3-cycles and 4-cycles is 4-choosable. In this paper, we extend the results on 3-, 4-, and 5-cycles as follows. Let G be a planar graph without pairwise adjacent 3-, 4-, and 5-cycle. We prove that each precoloring of a 3-cycle of G can be extended to a DP-4-coloring of G. As a consequence, each planar graph without pairwise adjacent 3-, 4-, and 5-cycle is DP-4-colorable.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Every graph in this paper is finite, simple, and undirected. Embedding a graph G in the plane, we let V(G), E(G), and F(G) denote the vertex set, edge set, and face set of G. For \(U \subseteq V(G),\) we let G[U] denote the subgraph of G induced by U. For \(X, Y \subseteq V(G)\) where X and Y are disjoint, we let \(E_G(X,Y)\) be the set of all edges in G with one endpoint in X and the other in Y.
The concept of choosability was introduced by Vizing [9] and by Erdős et al. [5], independently. A k-assignmentL of a graph G assigns a list L(v) (a set of colors) with \(|L(v)|= k\) to each vertex v. A graph G is L-colorable if there is a proper coloring f where \(f(v)\in L(v).\) If G is L-colorable for every k-assignment L, then we say G is k-choosable.
Dvořák and Postle [4] introduced a generalization of list coloring in which they called correspondence coloring. But following Bernshteyn et al. [3], we call it DP-coloring.
Definition 1
Let L be an assignment of a graph G. We call H a cover of G if it satisfies all the followings:
- (i)
The vertex set of H is \(\bigcup _{u \in V(G)}(\{u\}\times L(u)) =\{(u,c): u \in V(G), c \in L(u) \};\)
- (ii)
\(H[u\times L(u)]\) is a complete graph for every \(u \in V(G);\)
- (iii)
For each \(uv \in E(G),\) the set \(E_H(\{u\}\times L(u), \{v\}\times L(v))\) is a matching (may be empty).
- (iv)
If \(uv \notin E(G),\) then no edges of H connect \(\{u\}\times L(u)\) and \(\{v\}\times L(v).\)
Definition 2
An (H, L)-coloring of G is an independent set in a cover H of G with size |V(G)|. We say that a graph is DP-k-colorable if G has an (H, L)-coloring for every k-assignment L and every cover H of G. The DP-chromatic number of G, denoted by \(\chi _{DP}(G),\) is the minimum number k such that G is DP-k-colorable.
If we define edges on H to match exactly the same colors in L(u) and L(v) for each \(uv \in E(G),\) then G has an (H, L)-coloring if and only if G is L-colorable. Thus DP-coloring is a generalization of list coloring. This also implies that \(\chi _{DP}(G) \ge \chi _l(G).\) In fact, the difference of these two chromatic numbers can be arbitrarily large. For graphs with average degree d, Bernshteyn [2] showed that \(\chi _{DP}(G) = \Omega (d /\log d),\) while Alon [1] showed that \(\chi _l(G) = \Omega (\log d).\)
Dvořák and Postle [4] showed that \(\chi _{DP}(G) \le 5\) for every planar graph G. This extends a major result on list coloring by Thomassen [8]. On the other hand, Voigt [10] gave an example of a planar graph which is not 4-choosable (thus not DP-4-colorable). It is of interest to obtain sufficient conditions for planar graphs to be DP-4-colorable. Kim and Ozeki [6] showed that every planar graph without k-cycles is DP-4-colorable for each \(k =3,4,5,6.\) Kim and Yu [7] extended the result on 3- and 4-cycles by showing that every planar graph without triangles adjacent to 4-cycles is DP-4-colorable.
Let \({\mathcal {A}}\) denote the family of planar graphs G without pairwise adjacent 3-, 4-, and 5-cycle. In this paper, we extend the results on 3-, 4-, and 5-cycles as follows.
Theorem 1.1
Let \(G \in {\mathcal {A}}.\) Then each precoloring of a 3-cycle can be extended to a DP-4-coloring of G.
The following corollary is immediate.
Corollary 1.2
Every planar graph without pairwise adjacent 3-, 4-, and 5-cycle is DP-4-colorable.
Corollary 1.2 generalizes the aforementioned result by Kim and Yu [7] and the following result by Xu and Wu [11].
Theorem 1.3
[11] Every planar graph without 5-cycles adjacent simultaneously to 3-cycles and 4-cycles is 4-choosable.
2 Preliminaries
First, we introduce some notations and definitions. A k-vertex (\(k^+\)-vertex, \(k^-\)-vertex, respectively) is a vertex of degree k (at least k, at most k, respectively). The same notations are applied to faces.
A \((d_1,d_2,\dots ,d_k)\)-facef is a face of degree k where vertices on f have degree \(d_1,d_2,\dots ,d_k\) in a cyclic order. A \((d_1,d_2,\dots ,d_k)\)-vertexv is a vertex of degree k where faces incident to v have degree \(d_1,d_2,\dots ,d_k\) in a cyclic order. Note that some face may appear more than one time in the order. We say xy is a chord in a cycle C if \(x,y\in V(C)\) but \(xy\in E(G)- E(C).\) An internal chord of C is a chord inside C while external chord of C is a chord outside C
A graph C(m, n) is obtained from a cycle \(x_1x_2\ldots x_{m+n-2}\) with an internal chord \(x_1x_m.\) A graph C(l, m, n) is obtained from a cycle \(x_1x_2\ldots x_{l+m+n-4}\) with internal chords \(x_1x_l\) and \(x_1x_{l+m-2}.\) A graph C(m, n, p, q) can be defined similarly. We use int(C) and ext(C) to denote the sets of vertices inside and outside a cycle C, respectively. The cycle C is a separating cycle if int(C) and ext(C) are not empty. We use B(f) to denote a boundary of a face f. It is straightforward to see that if f is a \(5^-\)-face, then B(f) is a cycle.
3 Structures
Proof of Theorem 1.1
Let G be a minimal counterexample to Theorem 1.1 with |V(G)| minimized and a precolored 3-cycle \(C_0.\)\(\square \)
Lemma 3.1
G has no separating 3-cycles.
Proof
Suppose to the contrary that there exists G contains a separating 3-cycle C. Note that C is not necessary \(C_0.\) By symmetry, we assume \(V(C_0) \subseteq V(C) \cup int(C).\) By the minimality of G, a precoloring of \(C_0\) can be extended to \(V(C) \cup int(C).\) After C is colored, then again the coloring on C can be extended to ext(C). Thus we have a DP-4-coloring of G, a contradiction. \(\square \)
Since \(C_0\) is not a separating 3-cycle by Lemma 3.1, we may assume that \(C_0\) is the boundary of the outer face D of G in the remaining of the paper.
Definition 3
Let H be a cover of G with a list assignment L. Let \(G'= G-F\) where F is an induced subgraph of G. A list assignment \(L'\) is a restriction of L on \(G'\) if \(L'(u) = L(u)\) for each vertex in \(G'.\) A graph \(H'\) is a restriction of H on \(G'\) if \(H'= H[\{\{v\} \times L(v): v \in V(G')\}].\) Assume \(G'\) has an \((H',L')\)-coloring with an independent set \(I'\) in \(H'\) such that \(|I'|= |V(G)|-|V(F)|.\)
A residual list assignment\(L^*\) of F is defined by
for each \(x \in V(F).\)
A residual cover \(H^*\) is defined by \(H^*=H[\{\{x\} \times L^*(x): x \in V(F)\}].\)
From above definitions, we have the following fact.
Lemma 3.2
Assume \(I'\) is a \((H',L')\)-coloring of \(G'.\) A residual cover \(H^*\) is a cover of F with an assignment \(L^*. \) Furthermore, if F is \((H^*,L^*)\)-colorable, then G is (H, L)-colorable.
Proof
One can check from the definitions of a cover and a residual cover that \(H^*\) is a cover of F with an assignment \(L^*.\)
Suppose that F is \((H^*,L^*)\)-colorable. Then \(H^*\) has an independent set \(I^*\) with \(|I^*|= |F|.\) It follows from Definition 3 that no edges connect \(H^*\) and \(I'.\) Additionally, \(I'\) and \(I^*\) are disjoint. Altogether, we have that \(I=I'\cup I^*\) is an independent set in H with \(|I|=(|V(G)|-|V(F)|)+|V(F)|=|V(G)|.\) Thus G is (H, L)-colorable. \(\square \)
Lemma 3.3
Every vertex not on \(C_0\) has degree at least 4.
Proof
Suppose to the contrary that G has a vertex x not on \(C_0\) with degree at most 3. Let L be a 4-assignment and let H be a cover of G such that G has no (H, L)-coloring. By the minimality of G, the subgraph \(G'=G-x\) an \((H',L')\)-coloring where \(L'\) (and \(H'\)) is a restriction of L (and H, respectively) on \(G'.\) Thus there is an independent set \(I'\) with \(|I'|= |G'|\) in \(H'.\) Consider a residual list assignment \(L^*\) on x. Since \(|L(x)| = 4\) and \(d(x)\le 3\), we obtain \(|L^*(x)|\ge 1.\) Clearly, \(\{(x,c)\}\) where \(c \in L^*(x)\) is an independent set in \(G[\{x\}].\) Thus \(G[\{x\}]\) is \((H^*,L^*)\)-colorable. It follows from Lemma 3.2 that G is (H, L)-colorable, a contradiction. \(\square \)
Lemma 3.4
-
(a)
A 5-cycle has no chords.
-
(b)
A bounded 3-face f is not adjacent to a 4-face g.
-
(c)
If bounded 3-faces f and g are adjacent, then \(B(f)\cup B(g) = C(3,3).\)
-
(d)
If a bounded 3-face f is adjacent to a 5-face g, then \(B(f)\cup B(g) = C(3,5).\)
-
(e)
If C(3, 5) is obtained from a 6-cycle C with a chord, then C has exactly one chord.
-
(f)
If bounded 3-faces f and g are adjacent, then f is not adjacent to a bounded \(5^-\)-face.
Proof
- (a)
Let \(C= rstuv\) be a 5-cycle. Suppose that rt is a chord. Then we have three pairwise adjacent cycles rst, rtuv, and rstuv, contrary to \(G \in {\mathcal {A}}.\)
- (b)
Let \(B(f)=uvw\) and \(B(g)=vwxy.\) Suppose that \(u = x\) or y. We have that \(d(v) =2\) or \(d(w)=2,\) contrary to Lemma 3.3. Thus \(x\ne u \ne y.\) We obtain a 5-cycle uwxyv with a chord vw, contrary to (a).
- (c)
Let \(B(f)=uvw\) and \(B(g)=uvx.\) Since both f and g are bounded, we have that \(w \ne x.\) Thus \(B(f)\cup B(g) = C(3,3).\)
- (d)
Let \(B(f)=uvz\) and \(B(g)=uvwxy\). If \(z\ \in \{w,x,y\},\) then a 5-cycle B(g) has a chord, contrary to (a).
- (e)
Let C be a cycle uzvwxy with a chord uv. Suppose to the contrary that C contains another chord st. By (a) and (b), st is not a chord in a 5-cycle uvwxy. By symmetry, we may assume \(s=z\) and \(t=x\) or \(t=y.\) Then we have three pairwise adjacent cycles uvz, uvwxy, and xyuz or uvzy, contrary to \(G \in {\mathcal {A}}.\)
- (f)
Let \(B(f)=uvw\) and \(B(g)=vwx\). By (c), \(u \ne x.\) Suppose f is adjacent to a k-face h where \(k\le 5.\) By (b), \(h =3\) or 5.
\(B(h)=uvz.\)
By (c), \(z \ne w.\) Suppose to the contrary that \(z =x.\) Then \(d(v)=3,\) contrary to Lemma 3.3. Thus \(z \ne x.\) Altogether, we have three pairwise adjacent 3-cycles uvw, uvxw, zuwxv, contrary to \(G \in {\mathcal {A}}.\)
\(B(h)=rstuv.\)
By (a), \(w \notin \{r,s,t\}.\) By (e), \(x \notin \{r,s,t\}.\) Altogether, we have three pairwise adjacent cycles uvw, uvxw, and rstuv, contrary to \(G \in {\mathcal {A}}.\)
Thus h is not a \(5^-\)-face. \(\square \)
Lemma 3.4 (f) yields this immediate consequence.
Corollary 3.5
For \(k\ge 4,\) a k-vertex v in G is incident to at most \(k-2\) 3-faces.
Lemma 3.6
Let \(C(l_1,\ldots ,l_k)\) be obtained from a cycle \(C=x_1 \ldots x_m\) with k internal chords sharing a common endpoint \(x_1\) such that \(V(C)\cap V(C_0)= \emptyset .\) Suppose \(x_2\) or \(x_m\) is not the endpoint of any chord in C. If \(d(x_1) \le k+2,\) then there exists \(i \in \{2,3,\ldots ,m\}\) such that \(d(x_i) \ge 5.\)
Proof
By symmetry, let \(x_m\) be not an endpoint of any chord in C. Suppose to the contrary that \(d(x_i) \le 4\) for each \(i =2,3,\ldots ,m.\) Let L be a 4-assignment and let H be a cover of G such that G has no (H, L)-coloring. By the minimality of G, the subgraph \(G'=G-\{x_1,\ldots , x_m\}\) admits an \((H',L')\)-coloring where \(L'\) (and \(H'\), respectively) is a restriction of L (and H, respectively) in \(G'.\) Thus there is an independent set \(I'\) with \(|I'|= |G'|\) in \(H'.\)
Consider a residual list assignment \(L^*\) on F. Since \(|L(v)|= 4\) for every \(v\in V(G),\) we have \(|L^*(x_1)|\ge 3\) and \(|L^*(v)| \ge 3\) for each \(v\in V(C)\) with an edge \(x_1v\) and \(|L^*(x_i)| \ge 2\) for each of the remaining vertices \(x_i\) in V(C). Let \(H^*\) be an residual cover of F. Since \(x_m\) is not an endpoint of a chord in C, we can choose a color c from \(L^*(x_1)\) such that \(|L^*(x_{m})-\{c':(x_1,c)(x_{m},c')\in E(H^*)\}|\ge 2.\) By choosing colors of \(x_2\), \(x_3,\dots , x_{m}\) in this order, we obtain an independent set \(I^*\) with \(|I^*|=m=|F|.\) Thus F is \((H^*,L^*)\)-colorable. It follows from Lemma 3.2 that G is (H, L)-colorable, a contradiction. \(\square \)
Corollary 3.7
For each C(3, 5) such that \(V(C(3,5))\cap V(C_0)= \emptyset ,\) there exists a vertex with degree at least 5.
Proof
Let C(3, 5) be obtained from a 6-cycle \(C= x_1 \ldots x_6\) with a chord \(x_1x_3\) such that \(V(C)\cap V(C_0)= \emptyset .\) By Lemma 3.4 (e), C has no other chords. The proof is complete by Lemma 3.6. \(\square \)
Lemma 3.8
Let \(F=C(3,5,3), C(3,5,5)\) or C(5, 3, 5) be obtained from a cycle \(C=x_1\ldots x_m\) with two internal chords sharing an endpoint \(x_1\) such that \(V(C)\cap V(C_0)= \emptyset .\) If \(d(x_1)=5,\) then there exists \(i\in \{2,\ldots ,m\}\) with \(d(x_i) \ge 5.\)
Proof
By Lemma 3.6, it suffices to show that \(x_2\) or \(x_m\) is not an endpoint to a chord in C.
Let \(F = C(3,5,3).\) It follows from Lemma 3.4 (e) that \(x_2\) is not adjacent to \(x_4,x_5,\) or \(x_6.\) If \(x_2\) is adjacent to \(x_7,\) then we have separating 3-cycle \(x_1 x_2 x_7,\) contrary to Lemma 3.1. Thus \(x_2\) is not an endpoint of any chord of C.
Let \(F = C(3,5,5).\) Suppose there exists a chord e of C where \(e=x_2x_i,\) for otherwise we have the desired condition. If \( x_i= x_9,\) then we have separating 3-cycle \(x_1 x_2 x_9,\) contrary to Lemma 3.1. It follows from Lemma 3.4 (e) that \(i \notin \{4,5,6\}.\) Then \(x_i = x_7\) or \(x_8.\) By Lemma 3.4 (a), \(x_9\) is not adjacent to \(x_6\) or \(x_7.\) Thus \(x_9\) is not an endpoint of any chord of C.
Let \(F_1= C(5,3,5).\) Suppose there exists a chord e of C where \(e=x_2x_i,\) for otherwise we have the desired condition. If \( x_i= x_9,\) then we have separating 3-cycle \(x_1 x_2 x_9,\) contrary to Lemma 3.1. It follows from Lemma 3.4 (e) that \(i \notin \{4,5,6\}.\) Then \(x_i= x_7\) or \(x_8.\) By Lemma 3.4 (a), \(x_9\) is not adjacent to \(x_6\) or \(x_7.\) Thus \(x_9\) is not an endpoint of any chord of \(C'.\)\(\square \)
Corollary 3.9
Let v be a 5-vertex with incident bounded faces \(f_1,\ldots ,f_5\) in a cyclic order. Let \(F = B_1 \cup B_2 \cup B_3\) where \(B_i\) denote \(B(f_i)\) and \(V(F)\cap V(C_0)= \emptyset .\) If \((d(f_1),d(f_2),d(f_3))=(3,5,3)\) or (3, 5, 5), or \((d(f_1),d(f_2),d(f_3),d(f_4))= (5,3,5,3),\) then there exists \(w \in V(F)\) such that \(d(w) \ge 5\) and \(w \ne v.\)
Proof
By Lemma 3.8, it suffices to show that \(F= C(3,5,3),\)C(3, 5, 5), or C(5, 3, 5).
\((d(f_1),d(f_2),d(f_3))=(3,5,3).\)
Let \(B_1 = rsv,\)\(B_2= vstuw,\) and \(B_3=vwx.\) It follows from Lemma 3.4 (d) that \(V(B_1) \cap V(B_2) = \{s, v\}\) and \(V(B_2) \cap V(B_3) = \{v,w\}.\) If \(r=x,\) then \(d(v)=3,\) contrary to Lemma 3.3. Thus \(V(B_1) \cap V(B_3) =\{v\}.\) Altogether we have \(F=C(3,5,3).\)
\((d(f_1),d(f_2),d(f_3))=(3,5,5).\)
Let \(B_1= rsv,\)\(B_2= vstuw,\) and \(B_3=vwxyz.\) It follows from Lemma 3.4 (d) that \(V(B_1) \cap V(B_2) = \{s,v\}.\) If \(r = z,\) then \(d(v) = 3,\) contrary to \(d(v)=5.\) It follows from Lemma 3.4 (a) that neither r nor s is in \(V(B_3).\) Thus \(V(B_1) \cap V(B_3)= \{v\}.\) Now consider \(V(B_2) \cap V(B_3).\) It follows from Lemma 3.4 (a) that neither s nor u is in \(V(B_3).\) Similarly, neither x nor z is in \(V(B_2).\) Then \(V(B_2) \cap V(B_3) =\{t=y, v,w\}\) or \(\{v,w\}.\) Note that \(r\ne z,\) for otherwise rsv is a separating cycle, contrary to Lemma 3.1. Moreover, \(r \notin \{x,y\},\) otherwise the cycle vwxyz has a chord, contrary to Lemma 3.4 (a). Thus \(V(B_1) \cap V(B_3) = \{v\}.\) If \(V(B_2) \cap V(B_3) =\{t=y, v,w\}\), then we have three adjacent pairwise cycles rsv, stzv, stuwv, contrary to \(G \in {\mathcal {A}}.\) Thus \(V(B_2) \cap V(B_3)= \{v, w\}.\) Altogether we have \(F=C(3,5,5).\)
\((d(f_1),d(f_2),d(f_3),d(f_4))= (5,3,5,3).\)
Let \(B_1= rstuv,\)\(B_2= uvw,\)\(B_3= vwxyz,\) and \(B_4=vpz.\) It follows from Lemma 3.4 (e) that \(V(B_1) \cap V(B_2) = \{u,v\}\) and \(V(B_2) \cap V(B_3) = \{v,w\}.\) Consider \(V(B_1) \cap V(B_3).\) It follows from Lemma 3.4 (a) that neither r nor u is in \(V(B_3).\) Similarly, neither w nor z is in \(V(B_2).\) By Lemma 3.4 (a) that neither r nor u is in \(V(B_3).\)
Suppose \(\{s,t\} \subseteq V(B_1) \cap V(B_3).\)
Then \(\{x,y\} \subseteq V(B_1) \cap V(B_3)\) and \(s=y\) and \(t=x.\) Consequently, we have three adjacent pairwise cycles uvw, uvwt, rstuv, contrary to \(G \in {\mathcal {A}}.\)
Suppose \(s \in V(B_1) \cap V(B_3)\) but \(t \notin V(B_1) \cap V(B_3).\)
Then \(\{v,s\} = V(B_1) \cap V(B_3).\) Consequently \(s=x\) or \(s=y.\) In the former case, we have three adjacent pairwise cycles uvw, stuw, vwxyz, contrary to \(G \in {\mathcal {A}}.\) In the later case, we have three adjacent pairwise cycles pvz, srvz, vwxyz, contrary to \(G \in {\mathcal {A}}.\)
Suppose \(t \in V(B_1) \cap V(B_3)\) but \(s \notin V(B_1) \cap V(B_3).\)
Then \(\{v,t\} = V(B_1) \cap V(B_3).\) Consequently \(t=x\) or \(t=y.\) In the former case, we have three adjacent pairwise cycles uvw, uvwt, vwxyz, contrary to \(G \in {\mathcal {A}}.\) In the later case, we have three adjacent pairwise cycles uvw, uwxy, vwxyz, contrary to \(G \in {\mathcal {A}}.\)
Thus \(V(B_1) \cap V(B_3) = \{v\}.\) Altogether we have \(B_1 \cup B_2 \cup B_3 = C(5,3,5).\)
\(\square \)
Corollary 3.10
Let v be a 6-vertex with consecutive incident faces \(f_1,\ldots ,f_6.\) Let \(F = B_1 \cup B_2 \cup B_3\cup B_4\) where \(B_i\) denote \(B(f_i)\) and \(V(F)\cap V(C_0)= \emptyset .\) If \((d(f_1),d(f_2),d(f_3),d(f_4))= (3,5,3,5),\) then there exists \(w \in V(F)\) such that \(w \ne v\) and \(d(w) \ge 5.\)
Proof
By Lemma 3.6, it suffices to show that \(F= C(3,5,3,5).\) Similar to the proof of corollary 3.9, one can show that \(B_1 \cup B_2 \cup B_3 = C(3,5,3)\) and \(B_2 \cup B_3 \cup B_4 = C(5,3,5).\) Let \(V(B_1) = \{t,u,v\}\) and \(V(B_4) =\{v,w,x,y,z\}\) where \(t\in V(B_2)\) and \(w \in V(B_3).\) It only remains to show that \(u \notin \{x,y,z\}.\) If \(u =x\) or y, then the cycle vwxyz has a chord vx or vy, contrary to Lemma 3.4 (a). If \(u=z,\) then vtz is a separating cycle, contrary to Lemma 3.1. Thus \(F= C(3,5,3,5).\)\(\square \)
4 Discharging Process
We are now ready to present a discharging procedure that will complete the proof of Theorem 1.1. Let each vertex \(v\in V(G)\) have an initial charge of \(\mu (v)=2d(v)-6\), each face \(f\ne D\) has an initial charge of \(\mu (f)=d(f)-6\) and \(\mu (D)=d(D)+6=9\). By Euler’s Formula, \(\sum _{x\in V\cup F}\mu (x)=0\). Let \(\mu ^*(x)\) be the charge of \(x\in V\cup F\) after the discharge procedure. We prove that \(\mu ^*(x)\ge 0\) for all \(x\in V\cup F\) and \(\mu ^*(D)>0\) to get a contradiction.
Let \(w(v \rightarrow f)\) be the charge transferred from a vertex v to an incident face f. We say that v is a flaw vertex if v is a \((3,5,3,5^+)\)-vertex. The discharging rules are as follows.
(R1) Let f be a 3-face.
(R1.1) For a 4-vertex v not in \(C_0,\)
(R1.2) For a \(5^+\)-vertex v not in \(C_0,\)
(R2) Let f be a 4-face.
For a \(4^+\)-vertex v not in \(C_0,\)\(w(v \rightarrow f) = \frac{1}{2}\).
(R3) Let f be a 5-face.
(R3.1) For a 4-vertex v not in \(C_0,\)
(R3.2) For a 5-vertex v not in \(C_0,\)
(R3.3) For a \(6^+\)-vertex v not in \(C_0,\)
(R4) The outerface D gets \(\mu (v)\) from each incident vertex v and gives 2 to each 4- or 5-face or 3-face sharing exactly one vertex with D, \(\frac{12}{5}\) to each 3-face sharing one edge with D.
It suffices to check that each \(x\in V(G)\cup F(G)\) has nonnegative final charge and D has positive final charge. By (R4), we have \(\mu ^*(v)=0\) for each \(v \in V(C_0).\) Thus we only consider a vertex v not on \(C_0.\)
Let v be a vertex with neighbors \(v_1,v_2,\ldots , v_{d(v)}\) in a cyclic order. Let \(f_1,f_2,\ldots ,f_{d(v)}\) be incident faces of v in a cyclic order with \(v_i\) and \(v_{i+1}\) incident to \(f_i\) where \(i+1\) is taken in modulo d(v). Thus v is a \((d(f_1),d(f_2),\ldots ,d(f_{d(v)})\)-vertex.
Case 1v is a 4-vertex but v is not a flaw vertex.
It follows from Lemma 3.4 (b) that a 3-face is not adjacent to a 4-face, and from Corollary 3.5 that v has at most two incident 3-faces. Thus it suffices to consider that v is a \((4^+ ,4^+ ,4^+ ,4^+ )\)-, (3, 5, 4, 5)-, \((3,5^+,4,6^+)\)-, \((3,5^+,5^+,5^+)\)-, \((3,3, 5^+,5^+)\)-, or \((3,5^+,3,5^+)\)-vertex.
v is \((4^+ ,4^+ ,4^+ ,4^+ )\)-vertex.
Then v sends charge at most \(\frac{1}{2}\) to each incident face by (R2) and (R3.1). Thus \(\mu ^*(v)\ge \mu (v) - 4\times \frac{1}{2} = 0.\)
v is a (3, 5, 4, 5)-vertex.
Then \(w(v \rightarrow f_1) \le 1\) by (R1.1), \(w(v \rightarrow f_2) = w(v \rightarrow f_4) = \frac{1}{5}\) by (R3.1), and \(w(v \rightarrow f_3) = \frac{1}{2}\) by (R2). Thus \(\mu ^*(v)\ge \mu (v) -2\times \frac{1}{5}- 1 -\frac{1}{2}> 0.\)
v is a \((3,5^+,4,6^+)\)-vertex.
Then \(w(v \rightarrow f_1) \le 1\) by (R1.1), \(w(v \rightarrow f_2) \le \frac{1}{3}\) by (R3.1), and \(w(v \rightarrow f_3) = \frac{1}{2}\) by (R2). Thus \(\mu ^*(v)\ge \mu (v) - 1 -\frac{1}{2}-\frac{1}{3}> 0.\)
v is a \((3,5^+,5^+,5^+)\)-vertex.
Then \(w(v \rightarrow f_1) \le 1\) by (R1.1) and \(w(v \rightarrow f_i) \le \frac{1}{3}\) for \(2\le i \le 4\) by (R3.1). Thus \(\mu ^*(v)\ge \mu (v) - 1 - 3\times \frac{1}{3} = 0.\)
v is a \((3,3,5^+,5^+)\)-vertex.
It follows from Lemma 3.4 (b) that \(f_3\) and \(f_4\) are \(6^+\)-faces. Thus \(\mu ^*(v) \ge \mu (v) - 2\times 1= 0\) by (R1.1).
v is a \((3,5^+,3,5^+)\)-vertex.
Since v is not a flaw vertex, \(f_2\) and \(f_4\) are \(6^+\)-faces. Thus \(\mu ^*(v) \ge \mu (v) - 2\times 1= 0\) by (R1.1).
Case 2v is a flaw vertex, that is v is a \((3,5,3,5^+)\)-vertex.
Each adjacent vertex of v is a 4-vertex.
Then \(\max \{w(v \rightarrow f_1), w(v \rightarrow f_3)\} \le 1\) by (R1.1) and \(w(v \rightarrow f_2)=w(v \rightarrow f_4)=0\) by (R3.1). Thus \(\mu ^*(v) \ge \mu (v) - 2 \times 1= 0\).
v is adjacent to exactly one \(5^+\)-vertex.
Then \(\max \{w(v \rightarrow f_1), w(v \rightarrow f_3)\} \le 1,\)\(\min \{w(v \rightarrow f_1), w(v \rightarrow f_3)\} \le \frac{4}{5}\) by (R1.1), and \(\max \{w(v \rightarrow f_2),w(v \rightarrow f_4)\} = \frac{1}{10}\) by (R3.1). Thus \(\mu ^*(v) \ge \mu (v) - 1-\frac{4}{5}-2\times \frac{1}{10} = 0\).
v is adjacent to at least two \(5^+\)-vertices and incident to a \((4,5^+,5^+)\)-face.
Then \(\max \{w(v \rightarrow f_1), w(v \rightarrow f_3)\} \le 1,\)\(\min \{w(v \rightarrow f_1), w(v \rightarrow f_3)\} = \frac{3}{5}\) by (R1.1), and \(\max \{w(v \rightarrow f_2),w(v \rightarrow f_4)\} = \frac{1}{5}\) by (R3.1). Thus \(\mu ^*(v) \ge \mu (v) - 1-\frac{3}{5}-2\times \frac{1}{5}= 0.\)
v is adjacent to at least two \(5^+\)-vertices but neither \(f_1\) nor \(f_3\) is a \((4,5^+,5^+)\)-face.
Then \(f_1\) and \(f_3\) are \((4,4,5^+)\)-faces. It follows that \(w(v \rightarrow f_1)= w(v \rightarrow f_3)\} = \frac{4}{5}\) by (R1.1) and \(\max \{w(v \rightarrow f_2),w(v \rightarrow f_4)\} \le \frac{1}{5}\) by (R3.1). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{4}{5}-2\times \frac{1}{5} = 0.\)
Case 3v is a 5-vertex but is not a (3, 5, 3, 5, 5)-vertex. It follows from Corollary 3.5 that v is incident to at most three 3-faces. Since a 3-face is not adjacent to a 4-face [Lemma 3.4 (b)], we may assume v is a \((3^+,4^+,4^+, 4^+, 4^+)\)-, \((3,3,4^+, 3^+, 4^+)\)-, \((3,6^+,3, 6^+, 5^+)\)-, \((3,5,3, 6^+, 6^+)\)-, \((3,6^+,3, 5, 5)\)-, or \((3,5,3, 5,6^+)\)-vertex.
v is a \((3^+,4^+,4^+, 4^+, 4^+)\)-vertex.
It follows that \(w(v \rightarrow f_1) \le \frac{7}{5}\) by (R1.2), (R2), (R3.2).
Suppose v is a \((3^+,5^+,5^+, 5^+, 5^+)\)-vertex.
Then each incident 5-face is adjacent to a \(5^+\)-face. Consequently, \(\max \{w(v \rightarrow f_2),\ldots , w(v \rightarrow f_5)\} \le \frac{3}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - \frac{7}{5}-4\times \frac{3}{5} > 0.\)
Suppose there exists \(i \in \{2,\ldots ,5\}\) such that \(d(f_i)=4.\)
Then \(w(v \rightarrow f_i) \le \frac{1}{2}\) by (R2) and \(\max \{w(v \rightarrow f_2),\ldots , w(v \rightarrow f_5)\} \le \frac{7}{10}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - \frac{7}{5}-\frac{1}{2}-3\times \frac{7}{10} = 0.\)
v is a \((3,3,4^+, 3^+, 4^+)\)-vertex.
It follows from Lemma 3.4 (f) that \(f_3\) and \(f_5\) are \(6^+\)-faces. Since both \(f_1\) and \(f_2\) are 3-faces, each of common incident vertices is not a flaw vertex. Then \(\max \{w(v \rightarrow f_1),w(v \rightarrow f_3)\} \le \frac{6}{5}\) and \(w(v \rightarrow f_4) \le \frac{7}{5}\) by (R1.2). Thus \(\mu ^*(v) \ge \mu (v) -2\times \frac{6}{5}- \frac{7}{5} > 0.\)
v is a \((3,6^+,3, 6^+, 5^+)\)- or a \((3,5,3, 6^+, 6^+)\)-vertex.
If v is a \((3,6^+,3, 6^+, 5^+)\)-vertex, then \(\max \{w(v \rightarrow f_1),w(v \rightarrow f_3)\} \le \frac{7}{5}\) by (R1.2) and \(w(v \rightarrow f_5) \le \frac{7}{10}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{7}{5}-\frac{7}{10} > 0.\) The proof is similar for v is a \((3,5,3, 6^+, 6^+)\)-vertex.
v is a \((3,6^+,3, 5, 5)\)-vertex.
Then \(\max \{w(v \rightarrow f_1),w(v \rightarrow f_3)\} \le \frac{7}{5}\) by (R1.2) and \(\max \{w(v \rightarrow f_4),w(v \rightarrow f_5)\} \le \frac{3}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{7}{5}-2\times \frac{3}{5} = 0.\)
v is a \((3,5,3, 5,6^+)\)-vertex.
If \(f_3\) is a \((4^+,5,5^+)\)-face, then \(f_2\) or \(f_4\) is incident to at least two \(5^+\)-vertex. If \(f_3\) is a (4, 4, 5)-vertex, then applying Lemma 3.7 to \(f_3\) and \(f_4\) yields that \(f_4\) is incident to at least two \(5^+\)-vertex. Consequently, \(\max \{w(v \rightarrow f_1),w(v \rightarrow f_3)\} \le \frac{7}{5}\) by (R1.2) and \(\max \{w(v \rightarrow f_2),w(v \rightarrow f_4)\} \le \frac{7}{10}\) and \(\min \{w(v \rightarrow f_2),w(v \rightarrow f_4)\} \le \frac{2}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{7}{5}-\frac{7}{10}-\frac{2}{5} > 0.\)
Case 4v is a (3, 5, 3, 5, 5)-vertex.
\(f_1\) and \(f_3\) are (4, 4, 5)-faces.
Applying Corollary 3.9 to \(f_1,f_2,f_3,\) we have that \(f_2\) is incident to at least two non-adjacent \(5^+\)-vertices (including v). Then \(\max \{w(v \rightarrow f_1),w(v \rightarrow f_3)\} \le \frac{7}{5}\) by (R1.2) and \(w(v \rightarrow f_2) \le \frac{3}{10}\) by (R3.2).
Suppose \(f_4\) is incident to exactly one \(5^+\)-vertex.
Applying Corollary 3.9 to \(f_3,f_4,f_5,\) we have that \(f_5\) is incident to at least two non-adjacent \(5^+\)-vertices. Then \(w(v \rightarrow f_4) \le \frac{3}{5}\) and \(w(v \rightarrow f_5) \le \frac{3}{10}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{7}{5}-2\times \frac{3}{10}-\frac{3}{5} = 0.\)
Suppose \(f_4\) and \(f_5\) are incident to at least two \(5^+\)-vertices. Then \(\max \{w(v \rightarrow f_4),w(v \rightarrow f_5)\} \le \frac{2}{5}\) by (R3.2) Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{7}{5}-\frac{3}{10}-2\times \frac{2}{5} > 0.\)
\(f_1\) or \(f_3\) is a \((5,5^+,5^+)\)-face.
By symmetry, we assume \(f_1\) is a \((5,5^+,5^+)\)-face. It follows that \(f_2\) and \(f_5\) are incident to at least two \(5^+\)-vertices. Then \(w(v \rightarrow f_1) \le 1, w(v \rightarrow f_3) \le \frac{7}{5}\) by (R1.2) and \(w(v \rightarrow f_4) \le \frac{3}{5}, \max \{w(v \rightarrow f_2),w(v \rightarrow f_5)\} \le \frac{2}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - \frac{7}{5}-1-\frac{3}{5}-2\times \frac{2}{5} > 0.\)
\(f_1\) and \(f_3\) are \((4,5,5^+)\)-faces with incident \(5^+\)-vertex x and y, respectively, where \(x \ne v \ne y.\)
Then \(\max \{w(v \rightarrow f_1) , w(v \rightarrow f_3)\} \le \frac{6}{5}\) by (R1.2).
Suppose x and y are not incident to \(f_2.\)
It follows that \(f_4\) and \(f_5\) are incident to at least two \(5^+\)-vertices. Consequently, \(w(v \rightarrow f_2) \le \frac{7}{10}, \max \{w(v \rightarrow f_4),w(v \rightarrow f_5)\} \le \frac{2}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{6}{5}-\frac{7}{10}-2\times \frac{2}{5} > 0.\)
Suppose x and y are incident to \(f_2\).
Then \(w(v \rightarrow f_2) = \frac{3}{10}, \max \{w(v \rightarrow f_4),w(v \rightarrow f_5)\} \le \frac{3}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{6}{5}-\frac{3}{10}-2\times \frac{3}{5} > 0.\)
Suppose x is incident to \(f_2\) but y is not.
Then \(f_4\) is incident to at least two \(5^+\)-vertices. Consequently, \(\max \{w(v \rightarrow f_2),w(v \rightarrow f_4)\} \le \frac{2}{5}, w(v \rightarrow f_3) \le \frac{3}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{6}{5}-2\times \frac{2}{5} -\frac{3}{5}> 0.\)
\(f_1\) is a (4, 4, 5)-face but \(f_3\) is a \((4,5,5^+)\)-face with its two incident \(5^+\)-vertices are also incident to \(f_2.\)
It follows that \(w(v \rightarrow f_1) \le \frac{7}{5}, w(v \rightarrow f_3) \le \frac{6}{5}\) by (R1.2) and \(w(v \rightarrow f_2) \le \frac{2}{5}\) by (R1.2). Applying Corollary 3.9 to \(f_1,f_5,f_4,\) we have that \(f_4\) or \(f_5\) is incident to at least two \(5^+\)-vertices.
Suppose \(f_4\) is incident to at least two non-adjacent \(5^+\)-vertices.
Then \(w(v \rightarrow f_4) \le \frac{3}{10}, w(v \rightarrow f_5) \le \frac{3}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - \frac{7}{5}-\frac{6}{5}-\frac{2}{5}-\frac{3}{10}-\frac{3}{5} > 0.\) The proof is similar for \(f_5\) is incident to at least two non-adjacent \(5^+\)-vertices.
Suppose a \(5^+\)-vertex u incident to \(f_4\) or \(f_5\) is adjacent to v.
From assumption on \(f_1\) and \(f_3,\) we have that u is not incident to \(f_1\) and is not incident to \(f_3.\) It follows that u is incident to \(f_4\) and \(f_5.\) Then \(\max \{w(v \rightarrow f_4,w(v \rightarrow f_5)\} \le \frac{2}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - \frac{7}{5}-\frac{6}{5}-3\times \frac{2}{5} > 0.\)
\(f_1\) is a (4, 4, 5)-face but \(f_3\) is a \((4,5,5^+)\)-face with its two \(5^+\)-vertices are also incident to \(f_4.\)
It follows that \(w(v \rightarrow f_1) \le \frac{7}{5}, w(v \rightarrow f_3) \le \frac{6}{5}\) by (R1.2). Applying Corollary 3.9 to \(f_5,f_1,f_2,\) we have that \(f_2\) or \(f_5\) is incident to at least two \(5^+\)-vertices.
Suppose \(f_2\) is incident to a \(5^+\)-vertex u where \(u \ne v.\)
We have that u is not incident to \(f_1\) and \(f_3\) by assumption on \(f_1\) and \(f_3.\) This implies u is not adjacent to v. Then \(w(v \rightarrow f_2) \le \frac{3}{10},\)\(w(v \rightarrow f_4) \le \frac{2}{5},\) and \(w(v \rightarrow f_5) \le \frac{3}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - \frac{7}{5}-\frac{6}{5}-\frac{3}{10}-\frac{2}{5}-\frac{3}{5} >0.\)
Suppose \(f_5\) is incident to a \(5^+\)-vertex u where \(u \ne v.\)
If u is not incident to \(f_4,\) then u is not adjacent to v. It follows that \(w(v \rightarrow f_2) \le \frac{7}{10},\)\(w(v \rightarrow f_4) \le \frac{2}{5},\) and \(w(v \rightarrow f_5) \le \frac{3}{10}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - \frac{7}{5}-\frac{6}{5}-\frac{7}{10}-\frac{2}{5}-\frac{3}{10} =0.\)
If u is also incident to \(f_4,\) then \(f_4\) is a \((4^+,4^+,5,5^+,5^+)\)-vertex. It follows that \(w(v \rightarrow f_2) \le \frac{7}{10},\)\(w(v \rightarrow f_4) \le \frac{3}{10}\) and \(w(v \rightarrow f_5) \le \frac{2}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - \frac{7}{5}-\frac{6}{5}-\frac{7}{10}-\frac{3}{10}-\frac{2}{5} =0.\)
Case 5v is a 6-vertex.
From Lemma 3.4 (a) that a 3-face is not adjacent to a 4-face, it suffices to consider v is a \((3,3,k_3,k_4,k_5,k_6)\)-, \((3^+,4^+,3^+,4^+,4^+,4^+)\)-, \((3^+,4^+,4^+,3^+,4^+,4^+)\)-, (3, 5, 3, 5, 3, 5)-, or a \((3, 5^+,3,5^+,3,6^+)\)-vertex.
v is a \((3,3,k_3,k_4,k_5,k_6)\)-vertex.
It follows from Lemma 3.4 (b) that \(f_3\) and \(f_6\) are \(6^+\)-faces. Thus \(\mu ^*(v) \ge \mu (v) - 4\times \frac{7}{5}>0\) by (R1.2), (R2), and (R3.3).
v is a \((3^+,4^+,3^+,4^+,4^+,4^+)\)-vertex (or \((3^+,4^+,4^+,3^+,4^+,4^+)\)-vertex, respectively).
Then v sends charge at most \(\frac{7}{5}\) to \(f_1\) and \(f_3\) (or \(f_4,\) respectively) by (R1.2), (R2), (R3.3), and v sends charge at most \(\frac{4}{5}\) to each of the remaining incident faces by (R2), (R3.3). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{7}{5}-4\times \frac{4}{5}= 0.\)
v is a (3, 5, 3, 5, 3, 5)-vertex.
Suppose at least two incident 5-faces of v, say \(f_2\) and \(f_4,\) are incident to at least two \(5^+\)-vertices.
Then \(\max \{w(v \rightarrow f_1),w(v \rightarrow f_3),w(v \rightarrow f_5)\} \le \frac{7}{5}\) by (R1.2), \(\max \{w(v \rightarrow f_2),w(v \rightarrow f_4)\} = \frac{2}{5}\) and \(w(v \rightarrow f_6) \le \frac{4}{5}\) by (R3.3). Thus \(\mu ^*(v) \ge \mu (v) - 3\times \frac{7}{5}-2\times \frac{2}{5}-\frac{4}{5}>0.\)
Suppose two incident 5-faces of v, say \(f_2\) and \(f_4,\) are incident to exactly one \(5^+\)-vertex.
Applying Lemma 3.10 to \(f_1, f_2, f_3, f_4,\) we obtain that \(f_1\) is incident to two \(5^+\) vertices, say \(v_1\) and v, that are also incident to \(f_6.\) Applying Lemma 3.10 to \(f_2, f_3, f_4,f_5,\) we obtain that \(f_5\) is incident to two \(5^+\) vertices, say \(v_5\) and v, that are also incident to \(f_6.\) Then \(\max \{w(v \rightarrow f_1),w(v \rightarrow f_5)\} \le \frac{6}{5}\) and \(w(v \rightarrow f_3) \le \frac{7}{5}\) by (R1.2). Moreover, \(\max \{w(v \rightarrow f_2),w(v \rightarrow f_4)\} \le \frac{4}{5}\) and \(w(v \rightarrow f_6) \le \frac{2}{5}\) by (R3.2). Thus \(\mu ^*(v) \ge \mu (v) - 2\times \frac{6}{5}-\frac{7}{5}-2\times \frac{4}{5}-\frac{2}{5}>0.\)
Case 6v is a d-vertex with \(d\ge 7.\)
v is a \((3,3,k_3,\ldots , k_d)\)-vertex.
It follows from Lemma 3.4 that \(f_3\) and \(f_d\) are two \(6^+\)-faces. Thus \(\mu ^*(v) \ge \mu (v) - (d-2)\times \frac{7}{5}=2d-6- (d-2)\times \frac{7}{5}>0\) by (R1.2), (R2), and (R3.3).
v has no adjacent incident 3-faces.
It follows that v is incident to at most \(\frac{d}{2}\) 3-faces. Since v sends charge at most \(\frac{7}{5}\) to each of its incident 3-faces by (R1.2) and v sends charge at most \(\frac{4}{5}\) to each of the remaining incident faces by (R2) and (R3.3), we have \(\mu ^*(v) \ge \mu (v)-\frac{d}{2}\times \frac{7}{5}-\frac{d}{2}\times \frac{4}{5} =(2d-6)-d\times \frac{11}{10}>0.\)
Let f be a face in G. Let \(V(f)\cap V(D)\ne \emptyset \). If \(d(f)=3\), then f gets \(\frac{12}{5}\) from D when f shares an edge with D, 2 from D when f shares exactly one vertex with D. Note that each vertex of f in \(int(C_0)\) sends at least \(\frac{1}{2}\) to f. It follows that \(\mu ^*(f)\ge -3+\min \{\frac{12}{5}+\frac{1}{2}, 2+\frac{1}{2}\times 2\}=0\). If \(d(f)\in \{4,5\}\), then it gains 2 from D. Thus \(\mu ^*(f)\ge d(f)-6+2\ge 0\). If \(d(f)=6\), then \(\mu ^*(f)=\mu (f)=0\). If \(d(f)\ge 7\), then \(\mu ^*(f)\ge (k-6)-k\times \frac{k-6}{k}=0\). Thus we may assume that \(V(f)\cap V(D)=\emptyset \) for the remaining of the paper. Let f be a \(5^-\)-face with vertices \(v_1,v_2,\ldots , v_{d(v)}\) in a cyclic order.
Case 7f is a 3-face.
f is a (4, 4, 4)-face or each vertex of f is not a flaw vertex.
Then \(\mu ^*(f) = \mu (f) +3\times 1=0\) by (R1.1).
f is a \((4,4,5^+)\)-face with exactly one incident flaw vertex, say \(v_1.\)
Then \(w(v_1 \rightarrow f)=\frac{4}{5},\)\(w(v_2 \rightarrow f)= 1\) by (R1.1) and \(w(v_3 \rightarrow f)= \frac{6}{5}\) (R1.2). Thus \(\mu ^*(f) =\mu (f) +\frac{4}{5}+1+\frac{6}{5}=0.\)
f is a \((4,4,5^+)\)-face with \(v_1\) and \(v_2\) are flaw vertices.
Then \(w(v_1 \rightarrow f)=w(v_2 \rightarrow f)= \frac{4}{5}\) by (R1.1) and \(w(v_3 \rightarrow f)= \frac{7}{5}\) (R1.2). Thus \(\mu ^*(f) = \mu (f) +2\times \frac{4}{5}+\frac{7}{5}=0.\)
f is a \((4,5^+,5^+)\)-face and \(v_1\) is a flaw vertex.
Then \(w(v_1 \rightarrow f)=\frac{3}{5}\) by (R1.1) and \(w(v_2 \rightarrow f)=w(v_3 \rightarrow f)= \frac{6}{5}\) by (R1.2) Thus \(\mu ^*(f) = \mu (f) +\frac{3}{5}+2\times \frac{6}{5}=0\).
Case 8f is a 4-face.
We obtain \(\mu ^*(f) \ge \mu (f) +4\times \frac{1}{2}= 0\) by (R2).
Case 9f is a 5-face.
f is incident to at least three \(5^+\)-vertices.
It follows that each of its incident 4-vertex is adjacent to at least one \(5^+\)-vertex. Then each of these 4-vertices sends charge at least \(\frac{1}{10}\) to f by (R3.1) and each \(5^+\)-vertex sends charge at least \(\frac{3}{10}\) to f by (R3.2) and (R3.3). Thus \(\mu ^*(f)\ge \mu (f)+2\times \frac{1}{10}+ 3\times \frac{3}{10}> 0.\)
f is a \((4,5^+,4,5^+,4)\)-face.
Since \(v_1\) and \(v_5\) are adjacent to at least one \(5^+\)-vertex and \(v_3\) is adjacent to at least two \(5^+\)-vertices, we have \(\min \{w(v_1 \rightarrow f),w(v_5 \rightarrow f)\} \ge \frac{1}{10}\) and \(w(v_3\rightarrow f)\ge \frac{1}{5}\) by (R3.1). We have \(\min \{w(v_2 \rightarrow f),w(v_4 \rightarrow f)\} \ge \frac{1}{10}\) by (R3.2) and (R3.3). Thus \(\mu ^*(f)\ge \mu (f)+ 2\times \frac{1}{10}+\frac{1}{5}+2\times \frac{3}{10}= 0.\)
f is a \((4,4,4,5^+,5^+)\)-face
Since \(v_1\) and \(v_3\) are adjacent to at least one \(5^+\)-vertex, we have \(\min \{w(v_1 \rightarrow f),w(v_3 \rightarrow f)\} \ge \frac{1}{10}\) by (R3.1). We have \(\min \{w(v_2 \rightarrow f),w(v_4 \rightarrow f)\} \ge \frac{2}{5}\) by (R3.2) and (R3.3). Thus \(\mu ^*(f)\ge \mu (f)+2\times \frac{1}{10}+ 2\times \frac{2}{5}= 0.\)
f is a \((4,4,4,4,6^+)\)-face.
Since \(v_1\) and \(v_4\) are adjacent to at least one \(5^+\)-vertex, we have \(\min \{w(v_1 \rightarrow f),w(v_4 \rightarrow f)\} \ge \frac{1}{10}\) by (R3.1). We have \(w(v_5 \rightarrow f) = \frac{4}{5}\) by (R3.3). Thus \(\mu ^*(f)\ge \mu (f)+2\times \frac{1}{10}+ \frac{4}{5}= 0.\)
f is a (4, 4, 4, 4, 5)-face with at least one adjacent \(5^+\)-face \(f_i.\)
Let a 4-vertex \(v_i\) be in \(f_i.\) It follows that a 4-vertex \(v_i\) is not a flaw vertex. Then \(w(v_i \rightarrow f) = \frac{1}{3}\) by (R3.1), \(\min \{w(v_1 \rightarrow f),w(v_4 \rightarrow f)\} \ge \frac{1}{10}\) by (R3.1), and \(w(v_5\rightarrow f) = \frac{3}{5}\) by (R3.2), Thus \(\mu ^*(f)\ge \mu (f)+\frac{1}{3}+\frac{1}{10}+ \frac{3}{5}> 0.\)
f is a (4, 4, 4, 4, 5)-face with five adjacent \(4^-\)-faces and a 4-vertex \(v_i\) which is not a flaw vertex.
Then \(w(v_i \rightarrow f) = \frac{1}{3}\) by (R3.1) and \(w(v_5\rightarrow f) = \frac{7}{10}\) by (R3.2). Thus \(\mu ^*(f)\ge \mu (f)+ \frac{1}{3}+\frac{7}{10}> 0.\)
f is a (4, 4, 4, 4, 5)-face with five adjacent \(4^-\)-faces and four flaw vertices.
It follows that each adjacent face of f is a 3-face. Let \(f_4\) be a 3-face incident to \(v_4\) and \(v_5\) and let \(f_5\) be a 3-face incident to \(v_1\) and \(v_5.\) Applying Corollary 3.9 to \(f, f_4,\) and \(f_5,\) we have that \(f_4\) or \(f_5\) is a \((4,5,5^+)\)-face. By symmetry, let \(f_5\) be a \((4,5,5^+)\)-face. Consequently, \(v_1\) is adjacent to at least two \(5^+\)-vertices and \(v_4\) is adjacent to at least one \(5^+\)-vertex. It follows that \(w(v_1\rightarrow f) =\frac{1}{5}\) and \(w(v_4\rightarrow f)\ge \frac{1}{10}\) by (R3.1). We have \(w(v_5 \rightarrow f) = \frac{7}{10}\) by (R3.2). Thus \(\mu ^*(f)\ge \mu (f)+ \frac{1}{5}+\frac{1}{10}+\frac{7}{10}= 0.\)
f is a (4, 4, 4, 4, 4)-face.
Applying Corollary 3.7 to f and its adjacent 3-face, we have that each adjacent 3-face of f is a \((4,4,5^+)\)-face. This implies that each incident flaw vertex of f is adjacent to at least two \(5^+\)-vertex. If \(v_i\) is a flaw vertex, then \(w(v_i \rightarrow f) \ge \frac{1}{5},\) otherwise \(w(v_i \rightarrow f) \ge \frac{1}{3}\) by (R3.1). Thus \(\mu ^*(f)\ge \mu (f)+ 5\times \frac{1}{5}= 0.\)
Case 10 Consider the outerface D.
Let \(f_3', f'\) be the number of 3-faces sharing exactly one edge with D, 3-faces sharing exactly one vertex with D or 4-or 5-faces sharing vertices with D, respectively. Let \(E(C_0, V(G)-C_0)\) be the set of edges between \(C_0\) and \(V(G)-C_0\) and let \(e(C_0, V(G)-C_0)\) be its size. Then by (R4),
So we may consider that each edge \(e\in E(C_0,V(G)-C_0)\) gives a charge of 2 to D. Since each \(5^-\)-face is a cycle, it contains two edges in \(E(C_0,V(G)-C_0)\). It follows that \(e(C_0,V(G)-C_0)-f_3'-f'\ge 0\). Note that \(f_3'\le 3\). Thus \(\mu ^*(D)>0.\)
This completes the proof.
References
Alon, N.: Degrees and choice numbers. Random Struct. Algorithms 16, 364–368 (2000)
Bernshteyn, A.: The asymptotic behavior of the correspondence chromatic number. Discrete Math. 339, 2680–2692 (2016)
Bernshteyn, A., Kostochka, A., Pron, S.: On DP-coloring of graphs and multigraphs. Sib. Math. J 58, 28–36 (2017)
Dvořák, Z., Postle, L.: Correspondence coloring and its application to list-coloring planar graphs without cycles of lengths \(4\) to \(8\). J. Comb. Theory Ser. B 129, 38–54 (2018)
Erdős, P., Rubin, A.L., Taylor, H.: Choosability in graphs. In: Proceedings, West Coast Conference on Combinatorics, Graph Theory and Computing, Arcata, CA., Sept. 5–7, Congressus Numerantium, vol. 26 (1979)
Kim, S.-J., Ozeki, K.: A sufficient condition for DP-\(4\)-colorability. Discrete Math. 341, 1983–1986 (2018)
Kim, S.-J., Yu, X.: Planar graphs without \(4\)-cycles adjacent to triangles are DP-\(4\)-colorable. Graphs Comb. 35(3), 707–718 (2019)
Thomassen, C.: Every planar graph is \(5\)-choosable. J. Comb. Theory Ser. B 62, 180–181 (1994)
Vizing, V.G.: Vertex colorings with given colors. Metody Diskret Anal. 29, 3–10 (1976). (in Russian)
Voigt, M.: List colourings of planar graphs. Discrete Math. 120, 215–219 (1993)
Xu, R., Wu, J.L.: A sufficient condition for a planar graph to be \(4\)-choosable. Discrete Appl. Math. 224, 120–122 (2017)
Acknowledgements
We would like to thank anonymous referees for comments which are helpful for improvement in this paper. The first author is supported by Development and Promotion of Science and Technology talents project (DPST). The second author is supported by the Commission on Higher Education and the Thailand Research Fund under Grant RSA6180049.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Xueliang Li.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Sittitrai, P., Nakprasit, K. Every Planar Graph Without Pairwise Adjacent 3-, 4-, and 5-Cycle is DP-4-Colorable. Bull. Malays. Math. Sci. Soc. 43, 2271–2285 (2020). https://doi.org/10.1007/s40840-019-00800-1
Received:
Revised:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40840-019-00800-1