This paper proves that planar graphs lacking pairwise adjacent 3-, 4-, 5-, and 6-cycles are 4-choosable, extending previous results by Xu and Wu to a broader class of graphs.
Contribution
It establishes a new sufficient condition for 4-choosability in planar graphs, generalizing earlier work by removing adjacency restrictions among certain cycles.
Findings
01
Planar graphs without pairwise adjacent 3-, 4-, 5-, and 6-cycles are 4-choosable.
02
The result broadens the class of graphs known to be 4-choosable.
03
Improves upon previous conditions requiring non-adjacency of 5-cycles to 3- and 4-cycles.
Abstract
Xu and Wu proved that if every 5-cycle of a planar graph G is not simultaneously adjacent to 3-cycles and 4-cycles, then G is 4-choosable. In this paper, we improve this result as follows. If G is a planar graph without pairwise adjacent 3-,4-,5-, and 6-cycle, then G is 4-choosable.
Equations6
u∈V(G)∑μ(u)+f∈F(G)∑μ(f)=−12.
u∈V(G)∑μ(u)+f∈F(G)∑μ(f)=−12.
μ∗(D)
μ∗(D)
=9+2v∈C0∑(d(v)−2)−2×3−512f3′−2f′
=3−52f3′+2(e(C0,V(G)−C0)−f3′−f′)
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Graph Theory Research · Graph Labeling and Dimension Problems · graph theory and CDMA systems
Full text
Planar graphs without pairwise adjacent 3-,4-,5-, and 6-cycle are 4-choosable
Pongpat Sittitrai Kittikorn Nakprasit
Abstract
Xu and Wu proved that if every 5-cycle of a planar graph G
is not simultaneously adjacent to 3-cycles and 4-cycles,
then G is 4-choosable. In this paper, we improve this result as follows.
If G is a planar graph without pairwise adjacent 3-,4-,5-, and 6−cycle,
then G is 4-choosable.
1. Introduction
Every graph in this paper is finite, simple, and undirected.
The concept of choosability was introduced by Vizing in 1976 [12]
and by Erdős, Rubin, and Taylor in 1979 [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)∈L(v).
If G is L-colorable for any k-assignment L, then we say G
is k-choosable.
It is known that every planar graphs is 4-colorable [1, 2].
Thomassen [11] proved that every planar graph is 5-choosable.
Meanwhile, Voight [13] presented an example of non 4-choosable planar graph.
Additionally, Gutner [8] showed that determining
whether a given planar graph 4-choosable is NP-hard.
Since every planar graph without 3-cycle always has a vertex of
degree at most 3, it is 4-choosable.
More conditions for a planar graph to be 4-choosable are investigated.
It is shown that a planar graph is 4-choosable if it has no
4-cycles [10], 5-cycles [15], 6-cycles [7],
7-cycles [6], intersecting 3-cycles [16],
intersecting 5-cycles [9], or
3-cycles adjacent to 4-cycles [3, 4].
Xu and Wu [14] proved that if every 5-cycle of a planar graph G
is not simultaneously adjacent to 3-cycles and 4-cycles,
then G is 4-choosable. In this paper, we improve this result as follows.
Theorem 1.1**.**
If G is a planar graph without pairwise adjacent 3-,4-,5-, and 6−cycle,
then G is 4-choosable.
2. Preliminaries
First, we introduce some notations and definitions.
Let G be a plane graph. We use V(G),E(G), and F(G) for the vertex set, the edge set,
and the face set respectively.
We use B(f) to denote a boundary of a face f.
A wheelWn is an n-vertex graph formed
by connecting a single vertex (hub) to all vertices (external vertices) of an (n−1)-cycle.
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 (d1,d2,…,dk)-facef is a face of degree k
where vertices on f have degree d1,d2,…,dk in a cyclic order.
A (d1,d2,…,dk)-vertexv is a vertex of degree k
where faces incident to v have degree d1,d2,…,dk in a cyclic order.
Note that some face may appear more than one time in the order.
A face is called poor, semi-rich, and rich, respectively
if it is incident to no 5+-vertices, exactly one 5+-vertex,
and at least two 5+-vertices, respectively.
An extreme face is a bounded face that shares a vertex with the unbounded face.
An inner face is a bounded face that is not an extreme face.
A (3,5,3,5+)-vertex v is called a flaw 4-vertex
if v is incident to a poor 5-face and all incident faces of v are inner faces.
We say xy is a chord in an embedding cycle C if x,y∈V(C) but xy∈E(G)−E(C).
An internal chord is a chord inside C
while external chord is a chord outside C.
A triangular chord is a chord e such that two edges in C and e form a 3-cycle.
A graph C(m,n) is obtained from
a cycle x1x2…xm+n−2 with an internal chord x1xm.
A graph C(l,m,n) is obtained from
a cycle x1x2…xl+m+n−4 with internal chords x1xl and x1xl+m−2.
A graph C(m,n,p,q) can be defined similarly.
We use int(C) and ext(C) to denote the graphs induced by
vertices inside and outside a cycle C, respectively.
The cycle C is a separating cycle if int(C) and ext(C) are not empty.
Let L be a list assignment of G and let H be an induced subgraph of G.
Suppose G−H has an L-coloring ϕ on G−H where L is restricted to G−H.
For a vertex v∈H, let L′′(v) be a set of colors used on the neighbors of v by ϕ.
We define a residual list assignmentL′ of H by L′(v)=L(v)−L′′(v).
One can see that if G−H has an an L-coloring ϕ and H has an L′-coloring,
then G has an L-coloring.
The following is a fact on list colorings that we use later.
Lemma 2.1**.**
[5]** Let L be a 2-assignment.
A cycle Cn is L-colorable if and only if
n is even or L does not assign the same list to all vertices.
Let A denote the family of planar graphs
without pairwise adjacent 3-,4-,5-, and 6−cycle.
Next, we explore some properties of graphs in A which are helpful in a proof of the main results.
Lemma 2.2**.**
*Every graph G in A does not contain each of the followings:
(1) C(3,3,4),
(2) C(3,3,5),
(3) C(3,4,4−),
(4) C(4,3,5).
(5) W5 that shares exactly one edge with a 6−-cycle.*
Proof.
Let C(l,m,n) be obtained from a cycle x1x2…xl+m+n−4
with internal chords x1xl and x1xl+m−2.
(1) Suppose G contains C(3,3,4).
Then we have four pairwise adjacent cycles x1x2x3,x1x2x3x4,x1x3x4x5x6, and
x1x2x3x4x5x6, contrary to G∈A.
(2) Suppose G contains C(3,3,5).
Then we have four pairwise adjacent cycles x1x3x4,x1x2x3x4,x1x4x5x6x7, and
x1x3x4x5x6x7, contrary to G∈A.
(3) Suppose G contains C(3,4,3).
Then we have four pairwise adjacent cycles x1x2x3,x1x3x4x5,x1x2x3x4x5, and
x1x2x3x4x5x6, contrary to G∈A.
Suppose G contains C(3,4,4).
Then we have four pairwise adjacent cycles x1x2x3,x1x3x4x5,x1x2x3x4x5, and
x1x3x4x5x6x7, contrary to G∈A.
(4) Suppose G contains C(4,3,5).
Then we have four pairwise adjacent cycles x1x4x5,x1x2x3x4,x1x2x3x4x5, and
x1x4x5x6x7x8, contrary to G∈A.
(5) Let the hub of W5 be q and let external vertices be r,s,u, and v in a cyclic order.
Suppose there is a cycle uvw.
Then we have four pairwise adjacent cycles vwu,vwuq,vwusq, and
vwusqr, contrary to G∈A.
Suppose there is a cycle uvwx.
Then we have four pairwise adjacent cycles usq,usqv,usqrv, and
usqvwx, contrary to G∈A.
Suppose there is a cycle uvwxy.
Then we have four pairwise adjacent cycles uqv,uqrv,uqsrv, and
uqvwxy, contrary to G∈A.
Suppose there is a cycle uvwxyz.
Then we have four pairwise adjacent cycles uvq,uvqs,uvqrs, and
uvwxyz, contrary to G∈A.
∎
Lemma 2.3**.**
If C is a 6-cycle with a triangular chord,
then C has exactly one chord.
Proof.
Let C=tuvxyz with a chord tv. Suppose to the contrary that C has another chord e.
By symmetry, it suffices to assume that e=ux,uy,tx,ty, or xz.
If e=ux, then we have four pairwise adjacent cycles tuv,tuxv,tvxyz, and tuvxyz,
contrary to G∈A.
If e=uy, then we have four pairwise adjacent cycles tuv,uvxy,tvxyz, and tuvxyz,
contrary to G∈A.
If e=tx, then we have four pairwise adjacent cycles tuv,tuvx,tvxyz, and tuvxyz,
contrary to G∈A.
If e=ty, then we have four pairwise adjacent cycles tuv,tvxy,tvxyz, and tuvxyz,
contrary to G∈A.
If e=xz, then we have four pairwise adjacent cycles tuv,tvxz,tvxyz, and tuvxyz,
contrary to G∈A.
Thus C has exactly one chord.
∎
3. Structure
To prove Theorem 1.1, we prove a stronger result as follows.
Theorem 3.1**.**
If G∈A with a 4-assignment L,
then each precoloring of a 3-cycle in G can be extended to an L-coloring of G.
If G does not contain a 3-cycles, then G is 4-choosable as stated above.
So we consider (G,C0) and a 4-assignment L where C0 is a precolored 3-cycle
as a minimal counterexample to Theorem 3.1. Embed G in the plane.
Lemma 3.2**.**
G* has no separating 3-cycles.*
Proof.
Suppose to the contrary that there exists G contains a separating 3-cycle C.
By symmetry, we assume V(C0)⊆V(C)∪int(C).
By the minimality of G, a precoloring of C0 can be extended to V(C)∪int(C).
After C is colored, then again the coloring of C can be extended to ext(C).
Thus we have an L-coloring of G, a contradiction.
∎
So we may assume that a minimal counterexample (G,C0)
has no separating 3-cycles
and C0 is the boundary of the unbounded face D of G in the rest of this paper.
Lemma 3.3**.**
Each vertex in int(C0) has degree at least four.
Proof.
Suppose otherwise that there exists a 3−-vertex v in int(C0).
By the minimality of (G,C0), (G−v,C0) has an L-coloring.
One can see that the residual list L′(v) is not empty.
Thus we can color v and thus extend a coloring to G, a contradiction.
∎
Lemma 3.4**.**
*For faces in G, each of the followings holds.
(1) The boundary of a bounded 6−-face is a cycle.
(2) If a bounded k1-face f and a bounded k2-face g are adjacent
where k1+k2≤8,
then B(f)∪B(g)=C(k1,k2).
(3) If a bounded 4-face f and a bounded 5-face g are adjacent, then B(f)∪B(g) is C(4,5)
or a configuration as in Figure 1 where tuy is C0.
(4) If bounded 5-faces f and g are adjacent, then B(f)∪B(g) is C(5,5)
or a configuration as in Figure 2.*
Proof.
(1) One can observe that a boundary of a 5−-face is always a cycle.
Consider a bounded 6-face f. If B(f) is not a cycle, then a boundary closed walk is in a form of
uvwxywu. By Lemma 3.3, u or x has degree at least 4.
Consequently, uvw or xyw is a separating 3-cycle, contrary to Lemma 3.2.
(2) It suffices to show that such f and g share exactly two vertices.
Let B(f)=uvw and B(g)=vwx. If u=x, then f or g is the unbounded face, a contradiction.
Let B(f)=uvw and B(g)=vwxy. If u=x or y,
then d(w)=2 or d(v)=2, contrary to Lemma 3.3.
Let B(f)=uvw and B(g)=vwxyz. If u=x or z,
then d(v)=2 or d(w)=2, contrary to Lemma 3.3.
If u=y, then vyz or wxy is a separating 3-cycle, contrary to Lemma 3.2.
Let B(f)=stuv and B(g)=uvwx. If s=w, then d(v)=2, contrary to Lemma 3.3.
If s=x, then utx or vwx is a separating 3-cycle, contrary to Lemma 3.2.
The remaining cases are similar.
(3) Let B(f)=stuv and B(g)=uvwxy.
It suffices to show that V(B(f))∩V(B(g))={u,v} or {u,v,x} where x=s or t.
If t=w, then svw or uvw is a separating 3-cycle, contrary to Lemma 3.2.
If t=x, then tuy is C0, otherwise tuy is a separating cycle, contrary to Lemma 3.2.
The remaining cases are similar.
(4) Let B(f)=rstuv and B(g)=uvwxy.
It suffices to show that V(B(f))∩V(B(g))={u,v} or {u,v,x=s}.
If r=w, then d(v)=2, contrary to Lemma 3.3.
If B(f)∩B(g)={u,v,r=x},
then vwx,uvxy,uvwxy, and stuvwx are four pairwise adjacent cycles, contrary to G∈A.
If B(f)∩B(g)={u,v,r=x,s=y},
then rvx,rvuz,rvuts, and rstuvs are four pairwise adjacent cycles, contrary to G∈A.
then uts or vwx is a separating 3-cycle, contrary to Lemma 3.2.
If B(f)∩B(g)={u,v,r=y}, then ruv is a separating 3-cycle, contrary to Lemma 3.2.
If B(f)∩B(g)={u,v,s=w},
then rvw,tuvw,uvwxy, and rwxyuv are four pairwise adjacent cycles, contrary to G∈A.
The remaining cases are similar.
∎
Lemma 3.5**.**
*If a k-vertex v is incident to bounded faces f1,…,fk
in a cyclic order and di is a degree of a face fi for each i∈{1,…,k},
then each of the followings holds.
(1) (d1,d2,d3)=(3,3,4).
(2) (d1,d2,d3)=(3,3,5).
(3) (d1,d2,d3)=(3,4,4−).
(4) (d1,d2,d3)=(4,3,5).
(5) Let H be W5 such that a hub and each two vertices of consecutive external vertices
form a boundary of an inner 3-face. Then H is not adjacent to a boundary of a 6−-face
other than these 3-faces.*
Proof.
Let F=B1∪B2∪B3 where Bi denote B(fi).
(1) Suppose (d1,d2,d3)=(3,3,4).
Let B1=rsv,B2=vst, and B3=vtxy.
It follows from Lemma 3.4 (2) that V(B1)∩V(B2)={s,v}
and V(B2)∩V(B3)={t,v}.
If r=x, then stx or vxy is a separating 3-cycle, contrary to Lemma 3.2.
If r=y, then d(v)=3, contrary to Lemma 3.3.
Thus V(B1)∩V(B3)={v}.
Altogether we have F=C(3,3,4), contrary to Lemma 2.2(1).
(2) Suppose (d1,d2,d3)=(3,3,5).
Let B1=rsv,B2=vst, and B3=vtxyz.
It follows from Lemma 3.4 (2) that
V(B1)∩V(B2)={s,v} and V(B2)∩V(B3)={t,v}.
We have C=stxyzv is a 6-cycle with a triangular chord tv.
If r∈{x,y,z}, then C has another chord, contrary to Lemma 2.3.
Thus V(B1)∩V(B3)={v}.
Altogether we have F=C(3,3,5), contrary to Lemma 2.2(2).
(3) Suppose (d1,d2,d3)=(3,4,3).
Let B1=rsv,B2=vstu, and B3=vuw.
It follows from Lemma 3.4 (2) that V(B1)∩V(B2)={s,v}
and V(B2)∩V(B3)={u,v}.
If r=w, then d(v)=3, contrary to Lemma 3.3. Thus V(B1)∩V(B3)={v}.
Altogether we have F=C(3,4,3), contrary to Lemma 2.2(3).
Suppose (d1,d2,d3)=(3,4,4).
Let B1=rsv,B2=vstu, and B3=uvxy.
It follows from Lemma 3.4 (2) that V(B1)∩V(B2)={s,v}
and V(B2)∩V(B3)={u,v}. If r=x, then d(v)=3, contrary to Lemma 3.3.
If r=y, then vuy is a separating 3-cycle, contrary to Lemma 3.2.
Thus V(B1)∩V(B3)={v}.
Altogether we have F=C(3,4,4), contrary to Lemma 2.2(3).
(4) Suppose (d1,d2,d3)=(4,3,5).
Let B1=qrsv,B2=vst, and B3=vtxyz.
It follows from Lemma 3.4 (2) that
V(B1)∩V(B2)={s,v} and V(B2)∩V(B3)={t,v}.
We have C=stxyzv is a 6-cycle with a triangular chord tv.
If {q,r} and {x,y,z} are not disjoint, then C has another chord, contrary to Lemma 2.3.
Thus V(B1)∩V(B3)={v}.
Altogether we have F=C(4,3,5), contrary to Lemma 2.2(2).
(5) Let v be a hub and let w,x,y,z be external vertices of H in the cyclic order.
Suppose to the contrary that H is adjacent to a face f with
B(f)=wxq,wxqr,wxqrs, or wxqrst.
Now we have {w,x}⊆V(H)∩V(B(f)).
By Lemma 2.2(5), V(H)∩V(B(f))={w,x}.
If q=y, then d(x)=3, contrary to Lemma 3.3.
If r=y, then vwxqyz is a 6-cycle with four triangular chords, contrary to Lemma 2.3.
If s=y, then vxw,vxwz,vxwzy, and vxqryz are four pairwise adjacent cycles,
contrary to G∈A.
If t=y, then vxw,vxwz,vxwzy, and vxqrsy are four pairwise adjacent cycles,
contrary to G∈A.
The remaining cases lead to similar contradictions.
Thus f is not a 6−-face.
∎
Let C(m,n) in int(C0) be obtained from a cycle C=x1…xm+n−2 with a chord x1xm and d(x1)≤5.
If C has at most one additional chord e and e is not incident to x1,
then there exists i∈{2,…,m+n−2} with d(xi)≥5.
Corollary 3.7**.**
*If v is a flaw vertex, then we have the followings.
(1) v is incident to exactly one poor 5-face.
(2) Each 3-face that is incident to v is a semi-rich face.*
Proof.
Let v be incident to inner faces f1,f2,f3,f4 in a cyclic order where f1 and f3 are 3-faces,
f2 is a poor 5-face, and f4 is a 5+-face. By Lemma 3.4,
B(f1)∪B(f2) and B(f2)∪B(f3) are C(3,5).
It follows from Lemma 3.6 that
some vertex in B(f1)∪B(f2) and in B(f2)∪B(f3) has degree at least 5.
Observe that some vertex in B(f1) and in B(f3) has degree at least 5 since f2 is a poor face.
(1) If f4 is also a poor 5-face, then f1 is a poor face, contrary to the observation above.
(2) By observation above, f1 and f3 are not poor 3-faces.
Since f2 is a poor face, we obtain that f1 and f3 are not rich faces.
∎
Lemma 3.8**.**
If H in Figure 2 is in int(C0) and contains a 5−-vertex v,
then there is another vertex of H with degree at least 5 in G.
Proof.
First, we show that H is an induced subgraph.
Suppose to the contrary that there is an edge e joining vertices in V(H) such that e∈/E(H).
If e=ty, then sty or tuy is a separating 3-cycle.
If e=ux, then stu is a separating 3-cycle.
If e=sv, then rsv is a separating 3-cycle.
If e=rw, then rvw is a separating 3-cycle.
All consequences contradicts Lemma 3.2. Thus H is an induced subgraph.
Suppose to the contrary that d(v)≤5 but each of remaining vertices has degree at most 4.
By minimality, G−H has an L-coloring where L is restricted to G−H.
Consider a residual list assignment L′ on H.
Since L is a 4-assignment,
we have ∣L′(s)∣=4,∣L′(u)∣≥3, and ∣L′(v)∣,∣L′(r)∣,∣L′(t)∣,∣L′(y)∣,∣L′(w)∣≥2.
We begin by choosing a color c from L′(u) such that
∣L′(y)−c∣≥2.
Then we choose colors of v,r,w,t,s, and y in this order,
we obtain an L′-coloring on H.
Thus we can extend an L-coloring to G, a contradiction.
∎
Corollary 3.9**.**
Let v be a k-vertex in int(C0)
with consecutive incident faces f1,…,fk where k≤5.
If f1 and f2 are inner 5−-faces,
then there exists w∈B(f1)∪B(f2) such that w=v and d(w)≥5.
Proof.
The result follows directly from Lemmas 3.4, 3.6, and 3.8.
∎
Corollary 3.10**.**
If v is a 5-vertex in which each incident face is an inner 5−-face,
then v is incident to at least three rich faces.
Proof.
Suppose to the contrary that v is incident to at most two rich faces.
Consequently, v is incident to two adjacent semi-rich faces f and g.
But d(f)≤5 and d(g)≤5 while all vertices in B(f)∪B(g) except v has degree 4.
This contradicts Corollary 3.9.
∎
Lemma 3.11**.**
Let C(l1,…,lk) in int(C0) be obtained from a cycle C=x1…xm
with k internal chords sharing a common endpoint x1.
Suppose x1 is not incident to other chords
while x2 or xm is not incident to any chord.
If d(x1)≤k+3, then there exists i∈{2,3,…,m} such that d(xi)≥5.
Proof.
By symmetry, we assume xm is not an endpoint of any chord in C.
Suppose to the contrary that d(xi)≤4 for each i=2,3,…,m.
By the minimality of G,
the subgraph G−{x1,…,xm} has an L-coloring
where L is restricted to G−{x1,…,xm}.
Consider a residual list assignment L′ on x1,…,xm.
Since L is a 4-assignment,
we have ∣L′(x1)∣≥3 and ∣L′(v)∣≥3 for each v∈V(C) with an edge x1v
and ∣L′(xi)∣≥2 for each of the remaining vertices xi in V(C).
Since xm is not an endpoint of a chord in C, we can choose a color c from L′(x1) such that
∣L′(xm)−c∣≥2.
By choosing colors of x2, x3,…,xm in this order,
we obtain an L′-coloring on G′.
Thus we can extend an L-coloring to G, a contradiction.
∎
Corollary 3.12**.**
Let v be a 6-vertex with consecutive incident faces f1,…,f6
and let F=B1∪B2∪B3∪B4 where Bi denote B(fi).
If f1…f4 are bounded faces and (d(f1),d(f2),d(f3),d(f4))=(5,3,5,3),
then there exists w∈V(F)−{v} with d(w)≥5.
Proof.
By Lemma 3.11, it suffices to show that F=C(5,3,5,3).
Let cycles B1=vqrst,B2=vtu,B3=vuwxy, and B4=vyz.
Using Lemma 3.4, we have that V(B1)∩V(B2)={v,t},V(B2)∩V(B3)={v,u}, and V(B3)∩V(B4)={v,y}.
It suffices to show that V(B1)∩V(B3)={v}=V(B4)∩(V(B1)∪V(B2)).
Suppose to the contrary that V(B1)∩V(B3)={v}.
Consider a 6-cycle vtuwxy with a triangular chord uv.
If s=u,w,x, or y, then vtuwxy has another chord, contrary to Lemma 2.3.
Thus s∈/V(B1)∩V(B3). Similarly each of q,w, and y is not in V(B1)∩V(B3).
The only remaining possibility is that r=x. Suppose this holds.
Then vyz,vyxq,vyxwu, and vyrstu are four pairwise adjacent cycle, contrary to G∈A.
Thus V(B1)∩V(B3)={v} which implies B1∪B2∪B3=C(5,3,5).
As a consequence, we have vqrstu and vtuwxy are 6-cycles with a triangular chord.
If there is a vertex b∈V(B4)∩(V(B1)∪V(B2)) such that b=v,
then vqrstu or vtuwxy has another chord, contrary to Lemma 2.3.
This completes the proof.
∎
Corollary 3.13**.**
Let v be a 4-vertex incident to four inner 3-faces.
If all four neighbors of v are 5−-vertices,
then at least three of them are 5-vertices.
Proof.
Let w,x,y,z be neighbor of v in a cyclic order.
Let cycles B1=vwx and B2=vxy.
Note that w and y are not adjacent, otherwise vwy is a separating 3-cycle,
contrary to Lemma 3.2.
Similarly, x and z are not adjacent.
Suppose to the contrary that there are at least two 4-vertices among w,x,y, and z.
If those two 4-vertices are not adjacent, say w and y,
then B1∪B2 contradicts Lemma 3.6.
Thus we assume that w and x are 4-vertices.
Let H be the graph induced by v and its neighbors.
By minimality of G, the graph G−H has an L-coloring
where L is restricted to G−H.
Consider a residual list assignment L′ on H.
Since L is a 4-assignment, we have ∣L′(y)∣,∣L′(z)∣≥2,∣L′(w)∣,∣L′(x)∣≥3, and ∣L′(v)∣=4.
It suffices to assume that equalities holds for these list sizes.
We aim to show that H has an L′-coloring,
and thus an L-coloring can be extended to G, a contradiction.
CASE 1. There is a color t in L′(v)−(L′(y)∪L′(z)).
We begin by choosing t for v.
Each of the residual lists of w,x,y,z now has sizes at least 2. By Lemma 2.1, an even cycle is 2-choosable, thus H has an L′-coloring.
CASE 2. L′(v)−(L′(y)∪L′(z))=∅.
This implies L′(y)∩L′(z)=∅.
Choose t∈L′(v)−L′(w) for v.
If t∈L′(y), then t∈/L′(z) and we can color y,x,z, and w in this order,
otherwise we can color z,y,x, and w in this order.
Thus H has an L′-coloring.
This contradiction completes the proof.
∎
Let the initial charge of a vertex u in G be μ(u)=2d(u)−6
and the initial charge of a face f in G be μ(f)=d(f)−6.
Then by Euler’s formula ∣V(G)∣−∣E(G)∣+∣F(G)∣=2 and by the Handshaking lemma, we have
[TABLE]
Now we design the discharging rule transferring charge
from one element to another to provide a new charge μ∗(x)
for all x∈V(G)∪F(G). The total of new charges remains −12.
If the final charge μ∗(x)≥0 for all x∈V(G)∪F(G),
then we get a contradiction and complete the proof.
Before we establish a discharging rule, some definitions are required.
A graph C(3,3,3) in int(C0 is called a trio .
A vertex that is not in any trio is called a good vertex.
We call a vertex v incident to a face f in a trio T
a bad (worse, worst, respectively) vertex of f if v is incident to exactly one
(two, three, respectively) 3-face(s) in T.
We call a face f in a trio T a bad (worse,worst, respectively) face
of a vertex v if v is a bad (worse, worst, respectively) vertex of f in T.
For our purpose, we regard an external vertex of W5
as a worse vertex of its incident 3-faces in W5.
Let w(v→f) be the charge transferred from
a vertex v to an incident face f.
From now on, a vertex v is in int(C0) unless stated otherwise.
The discharging rules are as follows.
(R1) Let f be an inner 3-face that is not adjacent to another 3-face.
(R1.1) For a 4-vertex v,
w(v→f)={109,1,if v is flaw,otherwise.
(R1.2) For a 5+-vertex v,
w(v→f)={56,1,if f is a (4,4,5+)-face,otherwise.
(R2) Let f be an inner 3-face that is adjacent to another 3-face.
(R2.1) For a 4-vertex v,
w(v→f)=⎩⎨⎧21,1,32,if v is a hub of W5,if f is a good, bad, or worse face of v,if f is a worst face of v.
(R2.2) For a 5-vertex v,
w(v→f)=⎩⎨⎧1,45,23,if f is a good or worst face of v,if f is a worse face of v,if f is a bad face of v.
(R2.3) For a 6+-vertex v,
w(v→f)={1,23,if f is a good or worst face of v,if f is a bad or worse face of v.
(R3) Let f be an inner 4-face.
(R3.1) For a 4-vertex v, let w(v→f)=31.
(R3.2) For a 5+-vertex v,
w(v→f)={1,32,if f is a (4,4,4,5+)-face,if f is rich.
(R4) Let f be an inner 5-face.
(R4.1) For a 4-vertex v,
w(v→f)=⎩⎨⎧51,31,0,if v is flaw and f is a poor 5-face,if v is incident to at most one 3-face,otherwise.
(R4.2) For a 5+-vertex v,
w(v→f)=⎩⎨⎧1,32,t1,if f is a (4,4,4,4,5+)-face adjacent to five 3-faces,if f is a (4,4,4,4,5+)-face adjacent to at least one 4+-face other than f,if f is a rich face with t incident 5+-vertices.
(R5) Let f be an inner 3-face incident to a 4-vertex v
that is incident to four 3-faces.
If f is adjacent to a 7+-face g, we let w(g→f)=81.
(R6) The unbounded face D gets μ(v) from each incident vertex.
(R7) Let f be an extreme face.
w(x→f)=⎩⎨⎧25,2,2,21,0,if f is a 3-face that B(f) shares an edge with C0 and x=D,if f is a 4- or 5-face and x=D,if f is a 3-face that B(f) shares exactly one vertex with C0 and x=D,if f is a 3-face and x is a vertex in int(C0),otherwise.
(R8) After (R1) to (R7),
redistribute the total of charges of 3-faces in
the same cluster of at least three adjacent inner 3-faces (trio or W5) equally among
its 3-faces.
It remains to show that resulting μ∗(x)≥0
for all x∈V(G)∪F(G).
Let v be a k-vertex incident to faces f1,…,fk in a cyclic order.
By (R6), we only consider v in int(C0).
Consider the following cases.
(1)
v is a 4-vertex.
(1.1)
A vertex v is incident to a 3-face that is adjacent to another 3-face.
(1.1.1)
v is incident to at least two consecutive 3-faces.
If v is incident to four 3-faces,
then v is a hub of W5.
Thus μ∗(v)≥μ(v)−4×21=0 by (R2.1) and (R7).
If v is incident to exactly three 3-faces, say f1,f2, and f3,
then f4 is a 6+-face by Lemma 3.5(1) , (2).
Thus μ∗(v)≥μ(v)−3×32=0 by (R2.1) and (R7).
If v is incident to exactly two 3-faces, say f1 and f2,
then f3 and f4 are 6+-faces by Lemma 3.5(1), (2).
Thus μ∗(v)≥μ(v)−2×1=0 by (R2.1) and (R7).
(1.1.2)
v has no adjacent incident 3-faces.
Let f1 be a 3-face adjacent to another 3-cycle.
If follows from Lemma 3.5(1) and (2) that f2 and f4 are 6+-faces.
Then w(v→f1)≤1 by (R2.1) and (R7),
and w(v→f3)≤1 by (R2.1), (R3.1), (R4.1), and (R7).
Thus μ∗(v)≥μ(v)−2×1=0.
(1.2)
v is not incident to a 3-face that is adjacent to another 3-face
and v is adjacent to at most one 3-face.
Using the fact that w(v→fi)≤1 for a 3-face fi by (R1.1) and (R7),
and w(v→fi)≤31 for each 4+-face fi by (R3.1), (R4.1), and (R7),
we obtain that μ∗(v)≥μ(v)−1−3×31=0.
(1.3)
v is not incident to a 3-face that is adjacent to another 3-face
and v is adjacent to at least two 3-faces.
Consequently, v is incident to exactly two 3-faces, say f1 and f3.
It follows from Lemma 3.5(3) that f2 and f4 are 5+-faces.
If f2 is an extreme face, then w(v→f2)=0 by (R7),
w(v→f4)=0 by (R4.1) and (R7), and
w(v→fi)≤1 for i=1 and 3 by (R1.1) and (R7).
Thus μ∗(v)≥μ(v)−2×1=0.
Now assume that all incident faces of v are inner faces.
If v is flaw, then v is incident to exactly one poor 5-face,
say f2 by Corollary 3.7(1).
w(v→fi)=109 for i=1 and 3 by (R1.1),
w(v→f2)≤51 and w(v→f4)=0
by (R4.1) and (R7).
Thus μ∗(v)≥μ(v)−2×109−51=0.
If v is not flaw, then w(v→fi)=1 for i=1 and 3 by (R1.1)
and w(v→fi)=0 for i=2 and 4 by (R4.1).
Thus μ∗(v)=μ(v)−2×1=0.
(2)
A 5-vertex v is incident to a 3-face that is adjacent to another 3-face.
(2.1)
v has at least two consecutive incident 3-faces.
If v is incident to four 3-faces say f1,f2,f3, and f4,
then one can see that B(f1)∪B(f2)∪B(f3)∪B(f4)=C(3,3,3,3).
But C(3,3,3,3) contains four pairwise adjacent cycles that contradict G∈A.
Thus v is incident to at most three consecutive 3-faces.
If v incident to consecutive three 3-faces say f1,f2, and f3,
then f4 and f5 are 6+-faces by Lemma 3.5(1) and (2).
Thus μ∗(v)=μ(v)−3×1>0 by (R2.2) and (R7).
If v incident to exactly two consecutive 3-faces say f1 and f2,
then f3 and f5 are 6+-faces by Lemma 3.5(1) and (2).
Consequently, w(v→fi)≤45 for i=1 and 2,
and w(v→f4)≤23 by (R1.2), (R3.2), (R4.2), and (R7).
Thus μ∗(v)≥μ(v)−2×45−23=0.
(2.2)
v has no adjacent incident 3-faces.
Let f1 be a 3-face adjacent to another 3-face.
If follows from Lemma 3.5(1) and (2) that f2 and f5 are 6+-faces.
By (R2.2) and (R7), w(v→f1)≤23.
If neither f3 nor f4 are 3-faces,
then w(v→fi)≤1 for i=3 and 4 by (R3.2), (R4.2), and (R7).
Thus μ∗(v)≥μ(v)−23−2×1>0.
Now assume that f3 is a 3-face.
By the condition of (2.2), f4 is a 4+-face
which implies w(v→f4)≤1 by (R3.2), (R4.2), and (R7).
If f3 is adjacent to another 3-face, then f4 is a 6+-face by Lemma 3.5(1) and (2).
Moreover, w(v→f3)≤23 by (R2.2) and (R7).
Thus μ∗(v)≥μ(v)−2×23>0.
If f3 is not adjacent to another 3-face, then w(v→f3)≤1 by (R2.2) and (R7).
Thus μ∗(v)≥μ(v)−23−2×1>0.
(3)
A 5-vertex v is not incident to a 3-face that is adjacent to another 3-face
and v is incident to at least one 6+-face.
Consequently, v is incident to at most two 3-faces.
(3.1)
v is incident to at least two 6+-faces.
Recall that w(v→fi)≤56 for each 3-face fi by (R1.2) and (R7),
and w(v→fi)≤1 for each k-face fi where k=4,5 by (R3.2), (R4.2), and (R7).
If v is incident to t3-faces,
then there are at most 3−t faces f with d(f)=4 or 5.
Thus μ∗(v)≥μ(v)−t×56−(3−t)×1>0 by t≤3.
(3.2)
v is incident to exactly one 6+-face and incident to at most one 3-face.
If v has no incident 3-faces,
then v has all incident faces f except one 6+-face has d(f)=4 or 5.
Thus μ∗(v)≥μ(v)−4×1=0 by (R3.2), (R4.2), and (R7).
Assume v is incident to exactly one 3-face, say f1. By Lemma 3.5(3), v is not a (3,4,4,4,6+)- or a (3,4,4,6+,4)-face.
Consequently, v has at least one incident 5-face fj.
Moreover, fj is adjacent to at least one 4+-face.
We have w(v→f1)≤56 by (R1.2) and (R7),
w(v→fj)≤32 by (R4.2) and (R7),
and w(v→fi)≤1 for each remaining k-face fi
where k=4,5 by (R3.2), (R4.2), and (R7).
Thus μ∗(v)≥μ(v)−56−32−2×1>0.
(3.3)
v is incident to exactly one 6+-face and incident to exactly two 3-faces,
say f1 and f3.
It follows from Lemma 3.5(3) and (4) that v is either
a (3,5,3,5,6+)-, (3,5,5,3,6+)- or (3,5,4,3,6+)-vertex.
Assume v is a (3,5,3,5,6+)- or (3,5,5,3,6+)-vertex.
Applying Corollary 3.9 to B(f2)∪B(f3),v has an incident 5-face fj which is rich or extreme.
Recall that w(v→fi)≤56 for each 3-face fi by (R1.2) and (R7),
w(v→fj)≤21 by (R4.2) and (R7),
and w(v→fi)≤1 for the remaining 5-face fi by (R4.2) and (R7).
Thus μ∗(v)≥μ(v)−2×56−21−1>0.
Assume v is a (3,5,4,3,6+)-vertex.
Applying Corollary 3.9 to B(f1)∪B(f2),
we obtain that f1 or f2 is rich or extreme.
In the former case, w(v→f1)≤1 by (R1.2) and (R7),
and w(v→f2)≤32 by (R4.2) and (R7).
In the latter case, w(v→f1)≤56 by (R1.2) and (R7),
and w(v→f2)≤21 by (R4.2) and (R7).
Combining with w(v→f3)≤1 by (R3.2) and (R7) and
w(v→f4)≤56 by (R1.2) and (R7),
we have μ∗(v)≥μ(v)−2×1−32−56>0 or
μ∗(v)≥μ(v)−2×56−21−1>0.
(4)
A 5-vertex v is not incident to a 3-face that is adjacent to another 3-face
and v is not incident to a 6+-face.
Consequently, v is incident to at most two 3-faces.
Using Corollary 3.10,
we have that v has at least three incident faces that are rich or extreme.
(4.1)
v has no incident 3-faces.
If f has an extreme face fi,
then w(v→fi)=0 by (R7)
and w(v→fi)≤1 for each remaining fi by (R3.2), (R4.2), and (R7).
Thus μ∗(v)≥μ(v)−4×1=0.
If f has t rich faces,
then μ∗(v)≥μ(v)−t×32−(5−t)×1≥0
by (R3.2), (R4.2), (R7), and t≥3.
(4.2)
v is incident to exactly one 3-face, say f1.
It follows from Lemma 3.5(3) that v has at most two incident 4-faces.
(4.2.1)
v has no incident 4-faces.
We have that w(v→f1)≤56 by (R1.2) and (R7)
and w(v→fi)≤32 for each 5-face fi by (R4.2) and (R7).
Thus μ∗(v)≥μ(v)−56−4×32>0.
(4.2.2)
v has exactly one incident 4-face.
It follows from Lemma 3.5(4) that v is a (3,5,4,5,5)-face.
Recall that w(v→f1)≤56 by (R1.2) and (R7),
w(v→f3)≤1 by (R3.2) and (R7),
and w(v→fi)≤32 for each remaining fi by (R4.2) and (R7).
If f3 is rich, then w(v→f3)≤32 by (R3.2) and (R7).
Thus μ∗(v)≥μ(v)−56−4×32>0.
If f3 is not rich, then f2 and f4 are rich by Corollary 3.9.
Consequently, w(v→fi)≤21 for i=2 or 4 by (R4.2) and (R7).
Thus μ∗(v)≥μ(v)−56−1−2×21−32>0.
(4.2.3)
v has exactly two incident 4-faces.
It follows from Lemma 3.5(3) and (4) that v is a (3,4,5,5,4)- or a (3,5,4,4,5)-face.
Moreover, v has at least three incident rich faces by Corollary 3.10.
Consequently, we have (i) f1 and at least one 4-face fi are rich,
(ii) f1 and two 5+-faces are rich, (iii) a 4-face and two 5-faces are rich, or
(iv) two 4-faces and a 5-face are rich.
Recall that w(v→f1)≤56 by (R1.2) and (R7),
w(v→fi)≤1 for each 4-face fi by (R3.2) and (R7),
and w(v→fi)≤32 for each 5-face fi by (R4.2) and (R7).
Additionally, w(v→f1)≤1 if f1 is rich by (R1.2) and (R7),
w(v→fi)≤32 for each rich 4-face fi by (R3.2) and (R7),
and w(v→fi)≤21 for each rich 5-face fi by (R4.2) and (R7).
If f1 and a 4-face fi are rich,
then μ∗(v)≥μ(v)−2×1−3×32=0.
If f1 and two 5+-faces are rich,
then μ∗(v)≥μ(v)−1−2×1−2×21=0.
If a 4-face and two 5+-faces are rich,
then μ∗(v)≥μ(v)−56−1−32−2×21>0.
If two 4-faces and a 5-face are rich,
then μ∗(v)≥μ(v)−56−3×32−21>0.
(4.3)
v is incident to exactly two 3-faces, say f1 and f3.
It follows from Lemma 3.5(3) and (4) that v has no incident 4-faces.
This implies v is a (3,5,3,5,5)-vertex. Recall that w(v→fi)≤56 for each 3-face fi by (R1.2) and (R7),
and w(v→fi)≤32 for each 5-face fi by (R4.2) and (R7).
Additionally, w(v→f1)≤1 if f1 is rich by (R1.2) and (R7),
and w(v→fi)≤21 for each rich 5-face fi by (R4.2) and (R7).
If each incident 5-face is rich,
then μ∗(v)≥μ(v)−2×56−3×21>0.
If f2 is not rich, then f1 and f3 are rich by Corollary 3.9.
Consequently, f4 and f5 are also rich.
Thus μ∗(v)≥μ(v)−3×1−2×21=0.
If f4 is not rich, then f3 and f5 are rich by Corollary 3.9.
Consequently, f2 is also rich.
Thus μ∗(v)≥μ(v)−56−1−32−2×21>0.
The case that f5 is not rich is similar.
(5)
A 6-vertex v is incident to a 3-face that is adjacent to another 3-face.
(5.1)
v is incident to at least two consecutive 3-faces.
Let f1,…,fk be consecutive 3-faces.
Similar to Case (2.1), we have k≤3.
It follows from Lemma 3.5(1) and (2) that v is a (3,3,6+,k4,k5,6+)-
or (3,3,3,6+,k5,6+)-face.
Since w(v→fi)≤23 for each 5−-face fi by (R2.3), (R3.2), (R4.2), and (R7),
Thus μ∗(v)≥μ(v)−4×23=0.
(5.2)
v has no adjacent incident 3-faces.
Let f1 be a 3-face adjacent to another 3-face.
It follows from Lemma 3.5(1) and (2) that f2 and f6 are 6+-faces.
Similar to Case 5.1, we obtain that μ∗(v)≥μ(v)−4×23=0.
(6)
A 6-vertex v is not incident to a 3-face that is adjacent to another 3-face.
Consequently, v is incident to at most three 3-faces.
(6.1)
v is incident to at least one 6+-face.
Recall that w(v→fi)≤56 for each 3-face fi by (R1.2) and (R7),
and w(v→fi)≤23 for each k-face fi where k=4 or 5
by (R3.2), (R4.2), and (R7).
Thus μ∗(v)≥μ(v)−t×56−(5−t)×1>0 where t≤3 is
the number of incident 3-faces.
(6.2)
v has no incident 6+-face.
(6.2.1)
v has no incident 3-faces.
By (R3.2), (R4.2), and (R7), we have μ∗(v)≥μ(v)−6×1=0.
(6.2.2)
v has exactly one incident 3-face, say f1.
It follows from Lemma 3.5(3) that v is not a (3,4,4,4,4,4)-vertex.
Consequently, v has s5-faces where t≥1.
Note that each incident face of v is adjacent to another 4+-face.
It follows that w(v→fi)≤32 for each 5-face fi by (R4.2) and (R7).
Recall that w(v→f1)≤56 by (R1.2) and (R7),
and w(v→fi)≤1 for each 4-face f.
Thus μ∗(v)≥μ(v)−56−s×32−(5−s)×1>0.
(6.2.3)
v has exactly two incident 3-faces.
Consequently, v is a (3,k2,3,k4,k5,k6)- or (3,k2,k3,3,k5,k6)-vertex.
Assume v is a (3,k2,3,k4,k5,k6)-face.
Then k2=5 by Lemma 3.5(3).
This implies k4=k6=5 by Lemma 3.5(4).
Since v is a (3,5,3,5,4+,5)-vertex,
we have w(v→fi)≤56 for i=1 and 3 by (R1.2) and (R7),
w(v→fi)≤1 for i=2 and 5 by (R3.2),(R4.2) and (R7),
and w(v→fi)≤32 for i=4 and 6 by (R4.2) and (R7).
Thus μ∗(v)≥μ(v)−2×56−2×1−2×32>0.
Assume v is a (3,k2,k3,3,k5,k6)-vertex.
It follows from Lemma 3.5(4) that {k2,k6}={4,5}.
If k2=k6=4, then k3=k5=5 by Lemma 3.5(3).
Consequently, we may assume that v is a (3,4,5,3,5,4)- and (3,5,5,3,5,5)-vertex.
Recall that w(v→fi)≤56 for i=1 and 4 by (R1.2) and (R7),
w(v→fi)≤1 for each 4-face fi by (R3.2) and (R7).
and w(v→fi)≤32 for each 5-face fi by (R4.2) and (R7).
Thus a (3,4,5,3,5,4)-vertex has
μ∗(v)≥μ(v)−2×56−2×1−2×32>0,
and a (3,5,5,3,5,5)-vertex has
μ∗(v)≥μ(v)−2×56−4×32−1>0.
(6.2.4)
v has exactly three incident 3-faces.
Consequently, v is a (3,5,3,5,3,5)-vertex by Lemma 3.5(3).
It follows from Corollary 3.12 that
two incident 5-faces of v, say f2, and f4, are rich.
This implies w(v→fi)≤56 for i=1,3 and 5 by (R1.2) and (R7),
w(v→fi)≤21 for i=2 and 4,w(v→f6)≤1 by (R4.2) and (R7).
Thus μ∗(v)≥μ(v)−3×56−2×21−1>0.
(7)
v is a k-vertex where k≥7.
(7.1)
A vertex v is incident to a 3-face that is adjacent to another 3-face.
Then v is incident to at least two 6+-faces by Lemma 3.5(1) and (2).
Thus μ∗(v)≥μ(v)−(k−2)×23>0 by (R2.3), (R3.2), (R4.2), and (R7).
(7.2)
A vertex v is not incident to a 3-face that is adjacent to another 3-face.
Consequently v is incident to t3-faces where t≤k/2.
Thus μ∗(v)≥μ(v)−t×56−(k−t)×1>0 by (R1.2), (R3.2), (R4.2), and (R7).
(8)
An inner 3-face f is not adjacent to another 3-face.
If f has no incident flaw 4-vertices,
then μ∗(f)≥μ(f)+3×1=0 by (R1.1) adn (R1.2).
If f has an incident flaw vertex,
then f is a (4,4,5+)-face by Corollary 3.7(2).
Recall that w(v→f)≥109 for an incident 4-vertex v by (R1.1),
and w(v→f)≥56 for an incident 5+-vertex v by (R1.2).
Thus μ∗(f)≥μ(f)+2×109+56=0.
(9)
An inner 3-face f is adjacent to another 3-face.
Note that we use only (R2) to calculate a new charge.
(9.1)
A face f is not in a trio.
Then μ∗(f)≥μ(f)+3×1=0.
(9.2)
A face f is in a trio T but not in W5 formed by four inner 3-faces.
Let f1,f2, and f3 be 3-faces in the same trio T.
Define μ(T):=μ(f1)+μ(f2)+μ(f3)=−9 and
μ∗(T):=μ∗(f1)+μ∗(f2)+μ∗(f3).
By (R8), it suffices to prove that μ∗(T)≥0.
(9.2.1)
A worst vertex is a 5+-vertex.
Then μ∗(T)≥μ(T)+9×1=0.
(9.2.2)
A worst vertex is a 4-vertex and each worse vertex is a 4-vertex.
Then two bad vertices are 5+-vertices by Corollary 3.9.
Thus μ∗(T)≥μ(T)+3×32+2×23+4×1=0.
(9.2.3)
A worst vertex is a 4-vertex and one of worse vertices is a 5-vertex.
Then Corollary 3.9 yields that the other worse vertex
or at least one bad vertex is a 5+-vertex.
Thus μ∗(T)≥μ(T)+3×32+4×45+2×1=0 or
μ∗(T)≥μ(T)+3×32+2×45+23+3×1=0, respectively.
(9.2.3)
A worst vertex is a 4-vertex and one of worse vertices is a 6+-vertex.
Then μ∗(T)≥μ(T)+3×32+2×23+4×1=0.
(9.3)
A face f is in W5 formed by four inner 3-faces incident to v.
Let f1,f2,f3, and f4 be 3-faces in the same W5.
Define μ(W5):=μ(f1)+μ(f2)+μ(f3)+μ(f4)=−12 and
μ∗(W5):=μ∗(f1)+μ∗(f2)+μ∗(f3)+μ∗(f4).
By (R8), it suffices to prove that μ∗(W5)≥0.
Note that each 3-face in W5 is adjacent to a 7+-face by Lemma 3.5(5).
Thus W5 always obtains 4×81 from four 7+-faces by (R5).
(9.3.1)
Each vertex of W5 is a 5−-vertex.
Then at least three of them are 5-vertices by Corollary 3.13.
Thus μ∗(W5)≥μ(W5)+6×45+2×1+4×21+4×81=0.
(9.3.2)
Exactly one vertex of W5 is a 6+-vertex.
Then one of the remaining vertices is a 5+-vertex by Corollary 3.9.
Thus μ∗(W5)=μ(W5)+2×23+2×45+4×1+4×21+4×81=0.
(9.3.3)
At least two vertices of W5 are 6+-vertices.
Then μ∗(W5)≥μ(W5)+4×23+4×1+4×21+4×81>0.
(10)
f is an inner 4-face.
We claim that f is a (4+,4+,4+,5+)-face.
Suppose to the contrary that f is a (4,4,4,4)-face.
By the minimality of G,
there is an L-coloring of G−B(f) where L is restricted to G−B(f).
After the coloring, each vertex of B(f) has at least two legal colors.
By Lemma 2.1, we can extend an L-coloring to G, a contradiction.
If f is a (4,4,4,5+)-face, then μ∗(f)≥μ(f)+3×31+1=0 by (R3).
If f is a (4+,4+,5+,5+)- or (4+,5+,4+,5+)-face,
then f is a rich face and thus
μ∗(f)≥μ(f)+2×31+2×32=0 by (R3).
(11)
f is an inner 5-face.
(11.1)
f is a poor 5-face, that is f is a (4,4,4,4,4)-face.
It follows from Lemma 3.5(2) that
each incident 4-vertex of f is incident to at most two 3-faces.
If an incident vertex v of f is incident to at most one 3-face,
then w(v→f)=31 by (R4.1),
otherwise v is a flaw vertex by Lemma 3.5(2) and (3)
which implies w(v→f)=51 by (R4.1).
Consequently, f gains at least 51 from each incident vertex. Thus μ∗(f)≥μ(f)+5×51=0.
(11.2)
f is a (4,4,4,4,5+)-face.
(11.2.1)
f is adjacent to at least one 4+-face g.
It follows from (R4.2) that w(v→f)=32 for an incident 5+-vertex v of f.
Consider a 4-vertex u∈V(B(f))∩V(B(g)).
It follows from Lemma 3.5(2) that u is incident to at most one 3-face.
Consequently, w(u→f)=31 by (R4.1).
Thus μ∗(f)≥μ(f)+32+31=0.
(11.2.2)
f is adjacent to five 3-faces.
Then μ∗(f)=μ(f)+1=0 by (R4.2).
(11.3)
f is a rich face with t incident 5+-vertices.
Then μ∗(f)≥μ(f)+t×t1=0 by (R4.2).
(12)
f is an inner 6+-face.
If f is a 6-face, then μ∗(f)=μ(f)=0.
If f is a k-face where k≥7,
then μ∗(f)≥μ(f)−k×81>0 by (R5).
(13)
f is an extreme face.
(13.1)
f is a 3-face that shares exactly one vertex, say u, with C0.
It follows from (R7) that w(D→f)=2 and
w(v→f)=21 for each incident vertex v in int(C0).
Thus μ∗(f)=μ(f)+2+2×21=0.
(13.2)
f is a 3-face that shares an edge with C0.
It follows from (R7) that w(D→f)=25
and w(v→f)=21 for an incident vertex v in int(C0).
Thus μ∗(f)=μ(f)+25+21=0.
(13.3)
f is a 4- or 5-face.
Then μ∗(f)≥μ(f)+2≥0 by (R7).
(14)
D is the unbounded face.
Let f3′,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(C0,V(G)−C0) be the set of edges between C0 and V(G)−C0
and let e(C0,V(G)−C0) be its size. Then by (R6) and (R7),
[TABLE]
So we may consider that each edge e∈E(C0,V(G)−C0) gives a charge of 2 to D.
Since each 5−-face is a cycle, it contains two edges in E(C0,V(G)−C0).
It follows that e(C0,V(G)−C0)−f3′−f′≥0. Note that f3′≤3. Thus μ∗(D)>0.
Bibliography16
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] K. Appel, W. Haken, Every planar map is four colorable. I. Discharging, Illinois J. Math. 21(3)(1977) 429-490.
2[2] K. Appel, W. Haken, J. Koch, Every planar map is four colorable. II. Reducibility, Illinois J. Math. 21(3)(1977) 491-561.
4[4] P.P. Cheng, M. Chen, Y.Q. Wang, Planar graphs without 4 4 4 -cycles adjacent to triangles are 4 4 4 -choosable, Discrete Math. 339(2016) 3052-3057.
5[5] P. Erdős, A.L. Rubin, H. Taylor, Choosability in graphs, in: Proceedings, West Coast Conference on Combinatorics, Graph Theory and Computing, Arcata, CA., Sept. 5-7, in: Congr. Numer., vol. 26, 1979.
6[6] B. Farzad, Planar graphs without 7 7 7 -cycles are 4 4 4 -choosable, SIAM J. Discrete Math. 23(2009) 1179-1199.
7[7] G. Fijavž, M. Juvan, B. Mohar, R. Škrekovski, Planar graphs without cycles of specific lengths, European J. Combin. 23(2002) 377-388.
8[8] S. Gutner, The complexity of planar graph choosability, Discrete Math. 159(1996) 119-130.