Planar Graphs

November 9, 2016

1 Planar Graphs
Theorem 1.1 (Jordan Curve Theorem). A simple closed curve C in the plane R2 separates
the plane into two disjoint open sets.

• The Jordan Curve Theorem is not true for the Möbius band.

• A graph G is said to be embeddable in a plane, or planar, if it can be drawn in

the plane i such a way that no two edges cross each other. Such a drawing is called a
planar embedding of the graph.

• A planar graph G drawn in the plane is called a plane graph.

• A subdivision on an edge e with end-vertices u, v of a graph G is to add a new vertex

w on e, i.e., by adding the new vertex w to G and replacing the edge e with two edges,
one is incident with u, w and the other with v, w.

• A subdivision of a graph G is a graph obtained from G by a sequence of subdivisions

on some edges of G.

Proposition 1.2. A graph G is planar iff every subdivision of G is planar.

Proposition 1.3. A graph G is embeddable in the plane iff it is embeddable in the sphere.

Let G = (V, E) be a plane graph. We think ok G as the union V ∪ E, which is considered

to be a subspace of the plane R2 (or sphere S 2 ). The complement of G, R2 rG, is a collection
disconnected open sets of R2 (or of S 2 ), each is called a face of G. Each plane graph has
exactly one unbounded face, called the outer face. The numbers of vertices, edges, and
faces are denoted by v(G), e(G), and f (G) respectively.
The boundary of a face σ is the topological boundary of the open set σ in R2 (or S 2 ).
A face is incident with a vertex (an edge) if the vertex (edge) is contained in the boundary
of the face. Two faces are said to be adjacent if their boundaries contain a common edge.

Proposition 1.4. Let G be a planar graph. Let σ be a face of G in some planar embedding.
Then G admits a planar embedding whose outer face has the same boundary as that of σ.

1
 The proposition above means that outer face is not special. Any face can be an outer
face. For a plane graph G, we denote by f (G) the number of faces of G, by e(G) the number
of edges, and v(G) the number of vertices of G. For each face σ of G, let |σ| denote the
number of sides of the face σ. Then
X X
deg G (v) = 2e(G) = |σ|,
v∈V σ∈F

Theorem 1.5 (Face Cycle Theorem). Let G be a nonseparable graph and is neither K1 nor
K2 . Then each face of G is bounded by a cycle.

Proof. Consider an ear decomposition G0 , G1 , . . . , Gk , where G0 is a cycle, Gk = G, Gi =

Gi−1 ∪ Pi is nonseparable, Pi is a path, i = 1, . . . , k. We apply induction on k. For k = 0,
it is trivially true. Assume that it is true for Gk−1 and let σ denote the outer face of Gk−1 .
Now for Gk , then σ is divided into one bounded face σ1 and one unbounded face σ2 of Gk .
Let P = v0 e1 v1 · · · el vl be a closed path representing the boundary of σ, and let vi , vj be the
initial and terminal vertices of Pk with i < j. Let P10 denote the subpath of P from vi to vj ,
and P20 the subpath from vj to vi in the direction of P . Then the closed paths Pk−1 P10 and
P P2 bound the faces σ1 and σ2 respectively.

Corollary 1.6. Let G be a loopless 3-connected plane graph. Then the neighbor of every
vertex lie on a common cycle.

Proof. Fix a vertex v, the graph G r v is nonseparable. Let G be embedded in the plane,
and let σ denote the face of G r v in which the vertex v is located. Then by Theorem 1.5
the boundary of σ is a cycle, which contains all neighbors of v.

2 Duality
Definition 2.1. Let G be a plane graph. For each face σ of G, we choose a point σ ∗ inside
σ, and for each edge e bounding two faces σ, τ (the two faces may be the same), we draw
a topological path e∗ from σ ∗ to τ ∗ inside σ ∪ τ ∪ e, crossing the edge e once. The new
vertices σ ∗ and new edges e∗ form a graph embedded in the plane, called the dual graph
of G, denoted G∗ .

Proposition 2.2. The dual graph G∗ of a plane graph G is always connected. Moreover,

(a) Each cut edge of G corresponds to a loop of G∗ .

(b) Each loop of G corresponds to a cut edge of G∗ .

(c) If G is connected, then G∗∗ is isomorphic to G.

Proof. For two vertices σ ∗ , τ ∗ of G∗ , where σ, τ are two faces of G, there exists a topological
path from σ ∗ to τ ∗ , crossing finite number of edges e1 , . . . , ek of G, where each ei bounds two
faces σi−1 , σi , 1 ≤ i ≤ k. Then P = σ0∗ e∗1 σ1∗ · · · e∗k σk∗ is a path from σ ∗ to τ ∗ .
(a) Each cut edge e of G is on the boundary of one face σ of G, i.e., the outer face of G.
The corresponding edge e∗ is incident with the vertex σ ∗ twice in G∗ . So e∗ is a loop.

2
 Figure 1: The dual graph of a plane graph

(b) Each loop e of G encloses a face σ of G. The corresponding edge e∗ connects the part
of G∗ inside the loop e and the part of G∗ outside the loop e. So e∗ is a cut edge of G∗ .
(c) For each vertex v of G, let ej = vvj be the links and fk the loops of G at v, where
1 ≤ j ≤ m and 1 ≤ k ≤ n. Let σ0 denote the unbounded face of G.
Case 1: G has no edges at v. Then the conclusion is trivially true.
Case 2: G has only loops at v, i.e., G r v = ∅. Then the face σ0 of G is enclosed by
the closed walk
W = σ0∗ f1∗ σ1∗ f1∗ σ0∗ f2∗ σ2∗ · · · σ0∗ fn∗ σn∗ fn∗ σ0∗ .
in G∗ and contains the only vertex v.
Case 3: G has only links at v. Let Gi be the components of G r v, where 1 ≤ i ≤ l. We
may assume that the edges of G at v are ordered clockwise as e1 , . . . , em so that

[v, G1 ] = {e1 , . . . , em1 }, [v, G2 ] = [em1 +1 , . . . , em2 ], . . . , [v, Gl ] = {eml−1 +1 , . . . , eml },

where m0 = 0 and ml = m. Let σj denote the face of G between the edges ej and ej+1 ,
where mi−1 < j < mi and 1 ≤ i ≤ l. Then

Pi = σ0∗ e∗mi−1 +1 σm
∗
e∗ σ∗
i−1 +1 mi−1 +2 mi−1 +2
· · · e∗mi −1 σm
∗
e∗ σ ∗
i −1 mi 0

is a closed path (i.e. cycle) in G∗ , where 1 ≤ i ≤ l and m0 = 0. In particular, when

mi = mi−1 + 1, the closed path Pi is a loop. The face σ0 is enclosed by the closed walk
W 0 = P1 P2 · · · Pl in G∗ and contains the only vertex v. In particular, if n = 1, then W 0 is a
closed path and the face σ0 of G is enclosed by a cycle in G∗ .
Case 4: G has both links and loops at v. Then σ0 is enclosed by the closed walk
W = W 0 W and contains the only vertex v.
00

For each face F of G∗ , take an edge e∗ of F in G∗ , there exists an edge e crossing e∗ , then
F contains an end-vertex of the edge e in G. We have obtained a one-to-one correspondence
between the vertices of G∗∗ and the vertices of G.
By definition of plane dual graph, for each face F of G∗ , we may choose the vertex F ∗
of G∗∗ to be the only vertex of G inside F . Then G and G∗∗ have the same vertex set. Now
two vertices F1∗ , F2∗ of G∗∗ , which correspond to two faces F1 , F2 of G∗ , are adjacent in G∗∗
iff there exists an edge e∗ which corresponds an edge e of G with end vertices v1 , v2 .
Given two faces F1 , F2 of G∗ . If F1∗ , F2∗ are adjacent in G∗∗ , then F1 and F2 bound a
common edge e∗ of G∗ , since G∗∗ is the dual of G∗ . Thus there exists an edge e of G crossing
e∗ , since G∗ is the dual of G. It follows that the end-vertices v1 , v2 of e must be located in
the faces F1 , F2 respectively, say, vi ∈ Fi , i = 1, 2. Then Fi∗ = vi by the choices of Fi∗ inside
Fi , i = 1, 2. We have seen that F1∗ , F2∗ are already adjacent in G.

3
 Conversely, if F1∗ , F2∗ are adjacent in G by an edge e = v1 v2 with Fi∗ = vi , i = 1, 2. Then
there exists an edge e∗ of G∗ crossing the edge e of G. Since e is adjacent with F1∗ and F2∗
in G, it means that e∗ bounds the faces F1 , F2 of G∗ . Then by definition of dual the vertices
F1∗ , F2∗ are adjacent in G∗∗ .

Proposition 2.3. A simple plane graph G is a triangulation iff its dual graph G∗ is 3-regular.

Proposition 2.4. Let G be a connected plane graph with an edge e. If e is not a cut edge,
then (G r e)∗ ' G∗ /e∗ . If e is a link, then (G/e)∗ ' G∗ r e∗ .

Proof. Since e is not a cut edge, it must be on the boundary of two faces σ, τ of G. Deleting
the edge e results the faces σ, τ amalgamated into one face ρ. Let e1 , . . . , ek be the edges
bounding σ, and f1 , . . . , fl the edges bounding τ in G. Then in G∗ , the dual edges e∗1 , . . . , e∗k
are incident with the dual vertex σ ∗ , and f1∗ , . . . , fl∗ are incident with the dual vertex τ ∗ .
However in (G r e)∗ , the dual edges e∗1 , . . . , e∗k and f1∗ , . . . , fl∗ are incident with the dual
vertex ρ∗ , which is the result by contracting the edge e∗ in G∗ . Hence (G r e)∗ ' G∗ /e∗ .
Since G∗∗ ' G, we have (G∗ r e∗ )∗ ' G∗∗ /e∗∗ ' G/e. G∗ r e∗ ' (G/e)∗ .

Proposition 2.5. The dual graph G∗ of a nonseparable plane graph G is nonseparable.

Proof. It is trivial to verified it when G has only one or two vertices. We may assume that
G has at least three vertices. Then G has neither loops nor cut edges, so is G∗ . Actually,
both G and G∗ are 2-connected. We apply induction on the number of edges of G. It is
trivial when G has no edges, i.e., G is a single vertex. When G has some edges, take an edge
e. Then either G r e or G/e is nonseparable. Consequently, either (G r e)∗ or (G/e)∗ is
nonseparable by induction, i.e., either G∗ /e∗ or G∗ r e∗ is nonseparable. In the latter case,
G∗ is clearly nonseparable by adding the edge e∗ .
In the former case, i.e., G∗ /e∗ is nonseparable, we claim that G∗ is nonseparable. Suppose
G is separable with the separating vertex v ∗ . If v ∗ is not incident with e∗ , then G∗ /e∗ is
∗

separable, a contradiction. If v ∗ is incident with e∗ , let H denote the maximal nonseparable

component of G∗ that contains the edge e∗ . Then H cannot be just the edge e∗ , otherwise,
e∗ is a link of G∗ . Thus H/e∗ is nontrivial and is a nonseparable component of G∗ /e∗ , a
contradiction.

Proposition 2.6. Let G be a plane graph and G∗ its dual. Let C be a cycle of G. Let X ∗
be the vertices of G∗ that lie inside C. Then G∗ [X ∗ ] is connected.

Theorem 2.7. Let G be a plane graph and G∗ its dual graph.

(a) If C is a cycle of G, then C ∗ is a bond of G∗ .
(b) If B is a bond of G, then B ∗ is a cycle of G∗ .

Proof. (a) Let X ∗ be the set of vertices lie inside C and Y ∗ the set of vertices outside C.
Then both G∗ [X ∗ ] and G∗ [Y ∗ ] are connected. So C ∗ = [X ∗ , Y ∗ ] is a bond of G∗ .
(b) Let the edges of B be listed clockwise as e1 , e2 , . . . , ek , where ei is bounds the faces
σi−1 and σi , i = 1, . . . , k and σk = σ0 . Then σ0∗ e∗1 σ1∗ e∗2 σ2∗ · · · e∗k σk∗ form a cycle of G∗ .

Corollary 2.8. The cycle space of a plane graph G is isomorphic to the bond space of its
dual graph G∗ .

4
 Let D = (G, ω) be a plane digraph. Choose an orientation of the plane ~n. The directed
plane dual of D is the digraph D∗ = (G∗ , ω ∗ ), where ω ∗ is the orientation on G∗ such that
(a, a∗ , ~n) is in right system.

Theorem 2.9. Let D be a plane digraph and D∗ its directed plane dual digraph.
(a) If C is a directed cycle of D, then C ∗ is a directed bond of D∗ .
(b) If B is a directed bond of D, then B ∗ is a directed cycle of D∗ .

3 Euler Formula
Proposition 3.1. Let G be a plane graph. Let F (G) denote the set of faces of G. Then
X X
deg (v) = 2|E(G)| = |σ|,
v∈V (G) σ∈F (G)

where |σ| is the number of sides of the face σ.

Proof. The first equality follows from the fact that each edge two in the sum of vertex degrees.
The second equality follows from the fact that each edge contributes two sides either to one
face or to two faces, each for one face.

Theorem 3.2 (Euler’s Formula). For each connected plane graph G,

|V (G)| − |E(G)| + |F (G)| = 2.

Proof. We apply induction on the number of edges of G. When |E(G)| = 0, the graph G
must be a single vertex, and in this case the number of faces is one. Clearly,

|V (G)| − |E(G)| + |F (G)| = 1 − 0 + 1 = 2.

Assume that it is true for planar graphs with m edges, where m ≥ 1. We consider a connected
planar graph G with m+1 edges. The graph G may or may not have cycles. If G has no cycles,
i.e., G is a tree, then |E(G)| = |V (G|−1 by the tree formula, which is the Euler Formula since
|F (G)| = 1. If G has a cycle C, choose an edge e on C and remove e from G to obtain a new
planar graph G0 . Since e is on a cycle, G0 is still connected, but have less number of edges.
By induction we have |V (G0 )| − |E(G0 )| + |F (G0 )| = 2. Note that e bounds two faces, one is
inside the cycle C and the other is outside C. So |V (G0 )| = |V (G)|, |E(G0 )| = |E(G)| − 1,
and |F (G0 )| = |F (G)| − 1. We thus obtain |V (G)| − |E(G)| + |F (G)| = 2.
All planar embeddings of a planar graph G have the same umber of faces.

4 Kuratowski’s Theorem
• A subdivision of an edge e with end-vertices u, v of a graph G is to add a new vertex
w on e, i.e., by adding the new vertex w to G and replacing the edge e with two edges,
one is incident with u, w and the other with v, w.

5
 • A subdivision of a graph G is a graph obtained from G by a sequence of subdivisions
on some edges of G.

• A minor of a graph G is any graph obtained from G by means of a sequence of vertex

deletions, edge deletions, and edge contractions.
Let V (G) be partitioned into nonempty sets V0 , V1 , . . . , Vk . Let H be the graph obtained
from G by deleting V0 and shrinking G[Vi ] for 1 ≤ i ≤ k. Then any spanning subgraph
of H is a minor of G.

• The Peterson graph P5 has both K5 and K3,3 as its minors. However, P5 does contain
a subdivision of K3,3 , but does not contain any subdivision of K5 .

Figure 2: K5 , P5 , and subdivision of K3,3 embedded in P5

Theorem 4.1 (Kuratowski). A graph G is planar iff if G contains neither subdivision of

K5 nor subdivision of K3,3 .

Proof. Too long.

Exercises
Ch10: 10.1.3; 10.2.1; 10.2.2; 10.2.4; 10.2.9; 10.3.5.