6.

Planarity

Let G(V, E) be a graph with V = {v1 , v2 , . . ., vn } and E = {e1 , e2 , . . ., em}. Let S be any
surface (like the plane, sphere) and P = {p1 , p2 , . . ., pn } be a set of n distinct points of S,
pi corresponding to vi , 1 ≤ i ≤ n. If ei = v j vk , draw a Jordan arc Ji on S from p j to pk such
that Ji does not pass through any other pi . Then P ∪ {J1 , J2 , . . ., Jm } is called a drawing of
G on S, or a diagram representing G on S. The pi are called the points of the diagram and
Ji , the lines of the diagram.
An embedding of a graph G on a surface S is a diagram of G drawn on the surface such
that the Jordan arcs representing any two edges of G do not intersect except at a point
representing a vertex of G.
A graph is planar if it has an embedding on the plane. A graph which has no embedding
on the plane is nonplanar. That is, a graph G is said to be planar if there exists some
geometric representation of G which can be drawn on a plane such that no two of its edges
intersect and a graph that cannot be drawn on a plane without a crossover between its edges
is called nonplanar.
In order that a graph G is nonplanar, we have to show that of all possible geometric
representations of G, none can be embedded in a plane. Equivalently, a geometric graph G
is planar if there exists a graph isomorphic to G that is embedded in a plane.
An embedding of a planar graph G on a plane is called a plane representation of G.
Figure 6.1 shows three diagrams of the same graph which is planar. The two graphs in
Figure 6.2 represent the same planar graph.

Fig. 6.1

Fig. 6.2
Graph Theory 135

6.1 Kuratowski’s Two Graphs

The complete graph K5 and the complete bipartite graph K3,3 are called Kuratowski’s
graphs, after the polish mathematician Kasimir Kurtatowski, who found that K5 and K3,3
are nonplanar.
Theorem 6.1 The complete graph K5 with five vertices is nonplanar.
Proof Let the five vertices in the complete graph be named v1 , v2 , v3 , v4 , v5 . Since in a
complete graph every vertex is joined to every other vertex by means of an edge, there is a
cycle v1 v2 v3 v4 v5 v1 that is a pentagon. This pentagon divides the plane of the paper into two
regions, one inside and the other outside, Figure 6.3(a).
Since vertex v1 is to be connected to v3 by means of an edge, this edge may be drawn
inside or outside the pentagon (without intersecting the five edges drawn previously). Sup-
pose we choose to draw the line from v1 to v3 inside the pentagon, Figure 6.3(b). In case we
choose outside, we end with the same argument. Now we have to draw an edge from v2 to
v4 and another from v2 to v5 . Since neither of these edges can be drawn inside the pentagon
without crossing over the edge already drawn, we draw both these edges outside the pen-
tagon, Figure 6.3(c). The edge connecting v3 and v5 cannot be drawn outside the pentagon
without crossing the edge between v2 and v4 . Therefore v3 and v5 have to be connected with
an edge inside the pentagon, Figure 6.3(d). Now, we have to draw an edge between v1 and
v4 and this cannot be placed inside or outside the pentagon without a crossover. Thus the
graph cannot be embedded in a plane. q

Fig. 6.3
136 Planarity

Theorem 6.2 The complete bipartite graph K3,3 is nonplanar.

Proof The complete bipartite graph has six vertices and nine edges. Let the vertices be
u1 , u2 , u3 , v1 , v2 , v3 . We have edges from every ui to each vi , 1 ≤ i ≤ 3. First we take the
edges from u1 to each v1 , v2 and v3 . Then we take the edges between u2 to each v1 , v2 and
v3 , Figure 6.4(a). Thus we get three regions namely I, II and III. Finally we have to draw
the edges between u3 to each v1 , v2 and v3 . We can draw the edge between u3 and v3 inside
the region II without any crossover, Figure 6.4(b). But the edges between u3 and v1 , and
u3 and v2 drawn in any region have a crossover with the previous edges. Thus the graph
cannot be embedded in a plane. Hence K3,3 is nonplanar. q

Fig. 6.4

We observe that the two graphs K5 and K3,3 have the following common properties.

1. Both are regular.

2. Both are nonplanar.
3. Removal of one edge or a vertex makes each a planar graph.
4. K5 is a nonplanar graph with the smallest number of vertices, and K3,3 is the nonplanar
graph with smallest number of edges.

Thus both are the simplest nonplanar graphs.

The following result given independently by Fary [77] and Wagner [260] implies that
there is no need to bend edges in drawing a planar graph to avoid edge intersections.

Theorem 6.3 (Fary) Every triangulated planar graph has a straight line representation.

Proof The proof is by induction on the number of vertices. The result is obvious for
n = 4. So, let n ≥ 5 and assume that the result is true for all planar graphs with fewer than
n vertices. Let G be a plane graph with n vertices.
First, we show that G has an edge e belonging to just two triangles. For this, let x be any
vertex in the interior of a triangle T and choose x and T such that the number of regions
inside T is minimal.
Graph Theory 137

Let y be a neighbour of x, and the edge xy lies inside T , and let xy belong to three triangles
xyz1 , xyz2 and xyz3 . Then one of these triangles lies completely inside another. Assume that
z3 lies inside xyz1 . Then z3 and xyz1 contradict the choice of x and T (Fig. 6.5).

Fig. 6.5

Thus there is an edge e = xy lying in just two triangles xyz1 and xyz2 . Contracting xy to
a vertex u, we get a new graph G0 with a pair of double edges between u and z1 , and u and
z2 . Remove one each of this pair of double edges to get a graph G00 which is a triangulated
graph with n − 1 vertices. By the induction hypothesis, it has a straight line representation
H 00 . The edges of G00 correspond to uz1 , uz2 in H 00 . Divide the angle around u into two parts
in one of which the pre-images of the edges adjacent to x in G lie, and in the other, the
pre-images of the edges adjacent to y in G. Thus u can be pulled apart to x and y, and the
edge xy is restored by a straight line to get a straight line representation of G. q

6.2 Region
A plane representation of a graph divides the plane into regions (also called windows,
faces or meshes). A region is characterised by the set of edges (or the set of vertices)
forming its boundary. We note that a region is not defined in a nonplanar graph, or even in
138 Planarity

a planar graph not embedded in a plane. Thus a region is a property of the specific plane
representation of a graph and not of an abstract graph.

Fig. 6.6

Infinite region: The portion of the plane lying outside a graph embedded in a plane such
as region 4 in the graph given in Figure 6.6, is infinite in its extent. Such a region is called
an infinite, unbounded, outer or exterior region for that particular plane representation.
Like other regions, the infinite region is also characterised by a set of edges (or vertices).
Clearly, by changing the embedding of a given planar graph, we can change the infinite
region. Consider the graphs of Figure 6.7. In the two embeddings of the same graph, finite
region v1 v3 v5 in (a) is infinite in (b).

Fig. 6.7

Embedding on a sphere: A graph is spherical if it can be embedded on the surface of

a sphere.
The following result implies that a planar graph can often be embedded on the surface
of a sphere, so that there is no distinction between finite and infinite region.

Theorem 6.4 A graph can be embedded on the surface of a sphere if it can be embedded
in a plane.

Proof Consider the stereographic projection of a sphere on the plane. Put the sphere on
the plane and call the point of contact as SP (south-pole). At point SP, draw a straight line
perpendicular to the plane, and let the point where this line intersects the surface of the
sphere be called NP (north-pole).
Graph Theory 139

Fig. 6.8

Now, corresponding to any point p on the plane, there exists a unique point p0 on the
sphere, and vice versa, where p0 is the point where the straight line from point p to point
NP intersects the surface of the sphere. Thus there is a one-one correspondence between
the points of the sphere and the finite points on the plane, and points at infinity in the plane
corresponding to the point NP on the sphere.
Therefore from this construction, it is clear that any graph that can be embedded in a
plane can also be embedded on the surface of the sphere, and vice versa. q

We now have the following result.

Theorem 6.5 A planar graph can be embedded in a plane such that any specified region,
specified by the edges forming it, can be made the infinite region.

Proof A planar graph embedded on the surface of the sphere divides the surface of the
sphere into different regions. Each region of the sphere is finite, the infinite region on the
plane having been mapped onto the region containing the point NP. Clearly, by suitably
rotating the sphere, we can make any specified region map onto the infinite region on the
plane. Hence the result. q

Remark Thinking in terms of the regions on the sphere, we observe that there is no real
difference between the infinite region and the finite regions on the plane. Therefore, when
we discuss the regions in a plane representation of a graph, we include the infinite region.
Also, since there is no essential difference between an embedding of a planar graph on a
plane, or on a sphere (a plane can be regarded as the surface of the sphere of infinitely large
radius), the term plane representation of a graph is often used to include spherical as well
as plane embedding.

6.3 Euler’s Theorem

The following important result due to Euler gives a relation between the number of vertices,
edges, regions and the components of a planar graph.

Theorem 6.6 If G is a planar graph with n vertices, m edges, f regions and k components,
then

n − m + f = k + 1. (6.6.1)
140 Planarity

Proof We construct the graph G by the addition of successive edges starting from the
null graph Kn . For this starting graph, k = n, m = 0, f = 1, so that (6.6.1) is true.
Let Gi−1 be the graph at the start of ith stage and Gi be the graph obtained from Gi−1 by
addition of the ith edge e. If e connects two components of Gi−1, then f is not altered, m is
increased by 1 and k is reduced by 1, so that (6.6.1) holds for Gi as it holds for Gi−1. If e
joins two vertices of the same components of Gi−1, k is unaltered, m is increased by 1 and
f is increased by 1, so that again (6.6.1) holds for Gi . q

The following relation between the number of vertices, edges and regions is the discov-
ery of Euler [75] and is also called as Euler’s formula for planar graphs.

Theorem 6.7 (Euler) In a connected planar graph with n vertices, m edges and f regions
(faces), n − m + f = 2.
(The proof can also be deduced from Theorem 6.6 by taking k = 1, as the connected
graph has one component.)

Proof Without loss of generality, assume that the planar graph is simple. Since any
simple planar graph can have a plane representation such that each edge is a straight line,
any planar graph can be drawn such that each region is a polygon (a polygon net). Let
the polygon net representing the given graph consist of f regions. Let k p be the number of
p-sided regions. Since each edge is on the boundary of exactly two regions,

3k3 + 4k4 + 5k5 + . . . + rkr = 2m, (6.7.1)

where kr is the number of polygons with r edges.

Also, k3 + k4 + k5 + . . . + kr = f . (6.7.2)

The sum of all angles subtended at each vertex in the polygon net is 2π n. (6.7.3.)
Now, the sum of all interior angles of a p-sided polygon is π (p − 2) and the sum of the
exterior angles is π (p +2). The expression in (6.7.3) is the total sum of all interior angles of
f − 1 finite regions plus the sum of the exterior angles of the polygon defining the infinite
region. This sum is
π (3 − 2)k3 + π (4 − 2)k4 + . . .+ π (r − 2)kr + 4π

= π [3k3 + 4k4 + . . .+ rkr + 2(k3 + k4 + . . .+ kr )] + 4π

= π (2m − 2 f ) + 4π = 2π (m − f + 2). (6.7.4)

Equating (6.7.3) and (6.7.4) we get
2π (m − f + 2) = 2nπ , so that f = m − n − 2. q

Definition: Let φ be a region of a planar graph G. We define the degree of φ , denoted

by d(φ ), as the number of edges on the boundary of φ .
Graph Theory 141

Second proof of Theorem 6.7 We use induction on m, the number of edges. If m = 0,

then G is K1 , a graph with 1 vertex and 1 region. So, n − m + f = 1 − 0 + 1 = 2, and the result
is true. If m = 1, then the number of vertices in G is either one or two, the first possibility
occurring when the edge is a loop. These two possibilities give rise respectively to two
regions and one region.

Fig. 6.9

1−1+2, in the loop case ,
Therefore, n − m + f =
2−1+1, in the non loop case .

Thus, n − m + f = 2, and again the result is true.

Assume that the result is true for all connected planar graphs with fewer than m edges.
Let G have m edges.
Case I Suppose G is a tree. Then m = n − 1 and f = 1, because a planar representation of
a tree has only one region. Thus, n − m + f = n − (n − 1) + 1 = 2, and the result holds.
Case II Suppose G is not a tree. Then G has cycles. Let C be a cycle in G. Let e be an
edge of C. The graph G − e has one edge less than the graph G. Also the number of vertices
in G − e and G are same. Since removing e coalesces two regions in G into one in G − e,
therefore G − e has one region less than that in G. Thus by induction hypothesis in G − e,
we have n − (m − 1) + ( f − 1) = 2, so that n − m + f = 2. Hence the result. q

The next result gives an upper bound for the edges of a simple planar graph.

Theorem 6.8 Let G be a simple planar graph with n vertices and m edges, where n ≥ 3.
Then m ≤ 3n − 6.

Proof First assume that the planar graph G is connected. If n = 3, then since G is simple,
therefore G has at most three edges. Thus m ≤ 3, that is, m ≤ 3 × 3 − 6, and the result is true.
Now, let n = 4. First let G be a tree so that m = n − 1. Since n ≥ 4, obviously we have
n − 1 ≤ 3n − 6.
Now, let G be not a tree. Then G has cycles. Clearly, there is a cycle in G, all of whose
edges lie on the boundary of the exterior region of G. Then, since G is simple, we have
d(φ ) ≥ 3, for each region φ of G.

Let b = ∑ d (φ ), where F denotes the set of all regions of G.

φ ∈F
142 Planarity

Since each region has at least three edges on its boundary, therefore we have b ≥ 3 f ,
where f is the number of regions of G. However, when we sum to get b, each edge of G
is counted either once or twice (twice when it occurs as a boundary edge for two regions).
Therefore, b ≤ 2m so that 3 f ≤ b ≤ 2m. In particular, we have 3 f ≤ 2m.
Using Euler’s formula n − m + f = 2, we get m ≤ 3n − 6.
Let G be not connected and let G1 , G2 , . . ., Gt be its connected components. For each
i, 1 ≤ i ≤ t , let ni and mi denote the number of vertices and edges in Gi . Since each Gi is a
planar simple graph, by the above argument, we have mi ≤ 3ni − 6, for each i, 1 ≤ i ≤ t .
t t
Also, n = ∑ ni and m = ∑ mi .
i=1 i=1

t t t
Hence, m = ∑ mi ≤ ∑ (3ni − 6) = 3 ∑ ni − 6t ≤ 3n − 6. q
i=1 i=1 i=1

The following result gives the existence of a vertex of degree less than six in a simple
planar graph.
Theorem 6.9 If G is a simple planar graph, then G has a vertex v of degree less than 6.

Proof If G has only one vertex, then this vertex has degree zero. If G has only two
vertices, then both vertices have degree at most one.
Let n ≥ 3. Assume degree of every vertex in G is at least six.
Then, ∑ d (v) ≥ 6n.
v∈V (G)

We know ∑ d (v) ≥ 2m.

v∈V (G)

Thus, 2m ≥ 6n so that m ≥ 3n.

This is not possible because, by Theorem 6.8, we have m ≤ 3n − 6. Thus we get a
contradiction. Hence G has at least one vertex of degree less than 6. q

Corollary 6.1 K5 is nonplanar.

Proof Here n = 5 and m = 10. So, 3n − 6 = 15 − 6 = 9. Thus m > 3n − 6. Therefore K5 is

nonplanar.

Corollary 6.2 K3, 3 is nonplanar.

Proof Since K3, 3 is bipartite, it contains no odd cycles, and so no cycle of length three.
It follows that every region of a plane drawing of K3, 3 if it exists, has at least four boundary
edges. We have d(φ ) ≥ 4, for each region φ of G.
Graph Theory 143

Let b = ∑ d (φ ), where F denotes the set of all regions of G. Since each region has at
φ ∈F
least four edges on its boundary, we have b ≥ 4 f , where f is the number of regions of G.
Now, when we sum up to get b, each edge of G is counted either once or twice, and so
b ≤ 2m. Thus, 4 f ≤ b ≤ 2m, so that 2f ≤ m.
9
For K3, 3 , we have m = 9, and so 2 f ≤ 9 giving f ≤ . But by Euler’s formula, f =
2
m − n + 2 = 9 − 6 + 2 = 5, a contradiction. Hence K3, 3 is nonplanar. q

Euler’s formula gives a necessary condition for connected planar graphs.

Theorem 6.10 For a connected simple planar graph with n vertices and m edges, and
g (n − 2)
girth g, we have m ≤ .
g−2

Proof Let G be a connected simple planar graph with n vertices, m edges and girth g.
Let F be the set of regions and d(φ ) be the degree of the region φ . Then ∑ d (φ ) = 2m. Let
φ ∈F
f be the number of regions, that is |F| = f . As g is the length of the smallest cycle in G,
therefore g is the smallest degree of a region. So, ∑d (φ ) ≥ g f . Thus, 2m ≥ g f . Now, using
φ ∈F
g (n − 2)
Euler’s formula, f = m − n + 2, we have 2m ≥ g(m − n + 2), so that m ≤ .
g−2

Corollary 6.3 The Peterson graph P is nonplanar.

Proof The girth

 of the
 Perterson graph P is 5, n(P) = 10, and m(P) = 15. Thus, if P were
10 − 2
planar, 15 = 5 , which is not true. Hence P is nonplanar. q
5−2

Corollary 6.4 Any graph containing a homeomorph of a K5 or K3, 3 is nonplanar. This

follows from the fact that any homeomorph of a nonplanar graph is nonplanar.

The following result characterises nonplanar 3-connected graphs with minimum edges.

Theorem 6.11 A nonplanar connected graph G with minimum number of edges that
contains no subdivision of K5 or K3, 3 is simple and 3-connected.

Proof Since G has minimum number of edges, G is minimal with respect to these prop-
erties. The minimality of G ensures that it is simple. Also, a graph is planar if and only if
each of its blocks is planar. This implies that a minimal nonplanar graph is a block.
Suppose such a graph has a two vertex cut S = {u, v}. Then G − S is disconnected. Let
G1 be one of its components and G2 be the union of all of its other components.
Let H1 = < V (G1 ) ∪ S > + e and H2 = < V (G2 ) ∪ S > + e, where e = uv.
Clearly, these are nonempty graphs with fewer edges than G. Since G is nonplanar, at
least one of H1 and H2 is nonplanar. If both are planar, let H̃1 be a plane embedding of H1
and let φ be a region of H̃1 containing the edge e.
144 Planarity

A plane embedding H̃2 of H2 inside φ can then be obtained with the edge e of both H̃1 and
H̃2 coinciding. Then H̃1 ∪ H̃2 − e is a plane embedding of G, contradicting our hypothesis.
Suppose H1 is nonplanar. Since m(H1 ) < m(G), therefore H1 contains a subdivision K of
either K5 or K3, 3 according to the hypothesis. Clearly e ∈ E , since otherwise K ⊂ G, contra-
dicting the assumption. Now, replacing e of K by a u − v path in H2 , we get a homeomorph
of K contained in G. This is again a contradiction and shows that G has no 2-vertex cut.
Hence G is 3-connected. q

The following result is used in proving Kuratowski’s theorem.

Theorem 6.12
a. If G|e contains a subdivision of K5 , then G contains a subdivision of K5 or K3, 3 .
b. If G|e contains a subdivision of K3, 3 , then G contains a subdivision of K3, 3 .

Proof Let G0 = G|e be a graph obtained by contracting the edge e = xy of G. Let w be the
vertex of G0 obtained by contracting e = xy.
a. Let G|e contains a subdivision of K5 , say H . If w is not a branch vertex of H , then G
also contains a subdivision of K5 , obtained by expanding w back into the edge xy, if
necessary (Fig. 6.10).

Fig. 6.10

Assume w is a branch vertex of H and each of x, y is incident in G to two of the four

edges incident to w in H . Let u1 and u2 be the branch vertices of H that are at the
other ends of the paths leaving w on edges incident to x in G. Let v1 , v2 be the branch
vertices of H that are at the other ends of the paths leaving w on edges incident to y
in G (Fig. 6.11).

By deleting the u1 − u2 path and v1 − v2 path from H , we obtain a subdivision of K3, 3

in G, in which y, u1 , u2 are branch vertices for one partite set, and x, v1 , v2 are branch
vertices of the other.
Graph Theory 145

Fig. 6.11

b. Let G|e contain a subdivision of K3, 3 , say H . If w is not a branch vertex of H , then G
also contains a subdivision of K3, 3 , obtained by expanding w back into the edge xy, if
necessary (Fig. 6.12).

Fig. 6.12

Now, assume that w is a branch vertex in H and at most one of the edges incident to
w in H is incident to x in G. Then w can be expanded into xy to lengthen that path and
y becomes the corresponding branch vertex of K3, 3 in G (Fig. 6.13). q

Fig. 6.13

6.4 Kuratowski’s Theorem

This theorem was independently given by Kuratowski [144] and Frink and Smith. In 1954,
Dirac and Schuster [69] found a proof that was slightly shorter than the original proof. The
proof given here is due to Thomassen [241].
146 Planarity

Theorem 6.13 (Kuratowski) A graph is planar if and only if it does not have any
subdivision of K5 or K3, 3 .

Proof
Necessity Let G be a planar graph. Then any of its subgraphs is neither K5 nor K3, 3 nor
does it contain any subdivision of K5 or K3, 3 .

Sufficiency It is enough to prove sufficiency for 3-connected graphs. Let G be a 3-

connected graph with n vertices. We prove that the 3-connected graph G either contains
a subdivision of K5 or K3, 3 or has a convex plane representation. This we prove by using
induction on n. Since G is 3-connected, therefore n ≥ 4. For n = 4, G = K4 and clearly has
a plane representation.
Now, let n ≥ 5. Assume the result to be true for all 3-connected graphs with fewer than n
vertices. Since G is 3-connected, G has an edge e such that G|e is 3-connected. Let e = xy.
If G|e contains a subdivision of K5 or K3, 3 , then G also contains a subdivision of K5 or K3, 3 .
Therefore, let G|e = H have a convex plane representation. Let z be the vertex obtained by
contraction of e = xy. The plane graph obtained by deleting the edges incident to z has a
region containing z (this may be the exterior region). Let C be the cycle of H − z bounding
this region.
Since we started with a convex plane representation of H , we have straight segments
from z to all its neighbours. Let x1 , x2 , . . ., xk be the neighbours of x in that order on C.
If all the neighbours of y belong to a single segment from xi to xi+1 on C, then we obtain
a convex plane representation of G by putting x at z in H , and putting y at a point close to z
in the wedge formed by x xi and x xi+1 .
If all the neighbours of y do not belong to any single segment xi xi+1 on C (1 ≤ i ≤ k,
xk+1 = x1 ), then we have the following cases (Fig. 6.14).
a. y shares three neighbours with x. In this case C together with these six edges involving
x and y form a subdivision of K5 .
b. y has two u, v in C that are in different components of the subgraph of C obtained by
deleting xi and xi+1 , for some i. In this case, C together with the paths uyv, xi xxi+1 and
xy form a subdivision of K3, 3 . q

Fig. 6.14
Graph Theory 147

6.5 Geometric Dual

Let G be a plane graph. The dual of G is defined to be the graph G∗ constructed as follows.
To each region f of G there is a corresponding vertex f ∗ of G∗ and to each edge e of G there
is corresponding edge e∗ in G∗ such that if the edge e occurs on the boundary of the two
regions f and g, then the edge e∗ joins the corresponding vertices f ∗ and g∗ in G∗ . If the
edge e is a bridge, i.e., the edge e lies entirely in one region f , then the corresponding edge
e∗ is a loop incident with the vertex f ∗ in G∗ . For example, consider the graph shown in
Figure 6.15.

Fig. 6.15

Theorem 6.14 The dual G∗ of a plane graph is planar.

Proof Let G be a plane graph and let G∗ be the dual of G. The following construction of
G∗ shows that G∗ is planar.
Place each vertex f k∗ of G∗ inside its corresponding region f i . If the edge ei lies on the
boundary of two regions f j and f k of G, join the two vertices f ∗j and f k∗ by the edge e∗i ,
drawing so that it crosses the edge e exactly once and crosses no other edge of G (Fig.
6.16). q

Fig. 6.16

Remarks Clearly, there is one-one correspondence between the edges of plane graph G
and its dual G∗ with one edge of G∗ intersecting one edge of G.
148 Planarity

1. An edge forming a self-loop in G gives a pendant edge in G∗ (An edge incident on a

pendant vertex is called a pendant edge).

2. A pendant edge in G gives a self loop in G∗ .

3. Edges that are in series in G produce parallel edges in G∗ .

4. Parallel edges in G produce edges in series in G∗ .

5. The number of edges forming the boundary of a region f i in G is equal to the degree
of the corresponding vertex f i∗ in G∗ .

6. Considering the process of drawing a dual G∗ from G, it is evident that G is a dual

of G∗ . Therefore, instead of calling G∗ a dual of G, we usually say that G and G∗ are
dual graphs.

7. Let n, m, f , r and µ denote the number of vertices, edges, regions, rank and nullity of
a connected plane graph G and let n∗ , m∗ , f ∗ , r∗ and µ ∗ be the corresponding numbers
in G∗ . Then n∗ = f , m∗ = m, f ∗ = n. We have r∗ = n∗ − 1, µ ∗ = m∗ − n∗ + 1, r = n − 1,
µ = m − n + 1. So, r∗ = f − 1, µ ∗ = m − f + 1, r = n − 1, µ ∗ = m − n + 1.

Using Euler’s formula, n − m + f = 2 or f = m − n + 2, we have r∗ = m − n − 2 − 1 =

m − n + 1 = µ and µ ∗ = m − f + 1 = n + f − 2 − f + 1 = n − 1 = r.

We now have the following result.

Theorem 6.15 The edge e is a loop in G if and only if e∗ is a bridge in G∗ .

Proof Let the edge e be a loop in a plane graph G. Then it is the edge on the common
boundary of two regions on which, say f , lies within the area of the plane surrounded by e
with the other, say g, lying outside the area. Thus, from definition of G∗ , e∗ is the only path
from f ∗ to g∗ in G∗ . Thus e∗ is a bridge in G∗ .
Conversely, let e∗ be a bridge in G∗ , joining vertices f ∗ and g∗ . Thus e∗ is the only path
in G∗ from f ∗ to g∗ . This implies, again from the definition of G∗ , that the edge e in G
completely encloses one of the regions f and g. So e is a loop in G. q

Remark The occurrence of parallel edges in G∗ is easily described. Given two regions f
and g of G, there are k parallel edges between f ∗ and g∗ if and only if f and g have k edges
on their common boundary.

Note We have defined the dual of a plane graph instead of a planar graph. The reason
for this is that different plane drawings G1 and G2 of the same planar graph G may result in
non-isomorphic duals G∗1 and G∗2 .

2-isomorphism Two graphs G(V, E)and H(W, F) are said to be 2-isomorphic if there
exists a bijection φ : E → F such that both φ and φ −1 preserve cycles.
Graph Theory 149

Two graphs G and H are said to be 2-isomorphic if there exists a one-one correspondence
between their edge sets such that the edges of a cycle in G correspond to the edges of a
cycle in G2 and vice versa.

Example Figure 6.17(a) shows 2-isomorphic graphs.

1-isomorphism Graphs which become isomorphic after the splitting of all of their cut
vertices are said to be 1-isomorphic.

Example Figure 6.17(b) shows 1-isomorphic graphs.

Fig. 6.17

Consider the two non-isomorphic graphs G1 and G2 shown in Fig. 6.17(a) and (b). They
both have a single cut vertex, namely v. If this cut vertex is split into two vertices in each of
G1 and G2 , we obtain the edge disjoint graphs G01 and G02 as shown in (c) and (d). Clearly,
G01 and G02 are isomorphic. Thus G1 and G2 are 1-isomorphic.

The next result gives a correspondence between the edges of a graph with the edges of
its dual.

Theorem 6.16 Edges in a plane graph G form a cycle in G if and only if the correspond-
ing dual edges form a bond in G∗ .
150 Planarity

Proof Let D ⊆ E(G). It is sufficient to prove that D contains a cycle if and only if the
set D∗ of dual edges contains a bond of G∗ . Now, let D contain a cycle C. Then by the
Jordan Curve theorem, some region of G lies inside C and some region lies outside C.
These regions correspond to vertices v∗ , w∗ in G∗ , one drawn inside C and one outside C.
A v∗ − w∗ path in G∗ crosses C and hence uses an edge of G∗ , that is dual to an edge of C.
Thus D∗ disconnects v∗ from w∗ and hence D∗ contains a bond (Fig. 6.18(a)).
Conversely, let D∗ contain a bond. We show that D contains a cycle. Assume that D does
not contain a cycle. Then D encloses no region. It remains possible to reach each region of
G from every other without crossing D. Hence G∗ − D∗ is connected, and so D∗ contains no
bond. This is a contradiction. Hence D contains a cycle. q

Fig. 6.18(a)

Theorem 6.17 All duals of a planar graph G are 2-isomorphic and any graph H
2-isomorphic to a dual G∗ of G is itself a dual of G.

Proof Let G∗ (V ∗ , E ∗ ) be a dual of G. Then there is an edge bijection φ : E → E ∗ such

that cycles in G∗ correspond to bonds in G. Let H(W, F) be any other dual of G. Then there
is an edge bijection ψ : E → E ∗ such that bonds in G∗ correspond to cycles in H . Therefore
φ ψ : E ∗ → F is an edge bijection which preserves cycles between G∗ and H . Hence G∗ and
H are 2-isomorphic.
Conversely, let H be a graph which is 2-isomorphic to a dual G∗ of G. Then there is an
edge bijection χ : E ∗ → F which preserves cycles between G∗ and H . Also, there is an edge
bijection φ : E → E ∗ such that cycles in G∗ correspond to bonds in G. Thus we have an edge
bijection χφ : F → E such that cycles in F correspond to bonds in G. Hence H is a dual of
G. q

The following result provides the condition for two given graphs to be duals of each other.

Theorem 6.18 A necessary and sufficient condition for two planar graphs G1 and G2 to
be the duals of each other is that there should be a one-one correspondence of the edges in
G1 with the edges in G2 such that the subset of edges in G1 forms a cycle if and only if the
corresponding set of edges in G2 forms a bond.

Proof
Necessity Consider a plane representation of a planar graph G1 and let G∗1 be the dual of
G1 . Let C be an arbitrary cycle in G1 .
Graph Theory 151

Clearly, C will form some closed simple curve in the plane representation of G1 , dividing
the plane into two areas. Thus the vertices of G∗1 are partitioned into two non-empty,
mutually exclusive subsets, one inside C and the other outside. In other words, the set of
edges C∗ in G∗1 corresponding to the set C in G1 is a bond in G∗1 .
Similarly, corresponding to a bond S∗ in G∗1 , there is a unique cycle consisting of the
corresponding edge set S in G1 such that S is a cycle.
Sufficiency Let G be a planar graph, and let G0 be a graph for which there is a one-one
correspondence between the bonds of G and the cycles of G0 and vice versa. Let G∗ be the
dual graph of G. Therefore there is a one-one correspondence between the bonds of G and
cycles of G∗ . So there is a one-one correspondence between the cycles of G0 and G∗ . Thus
G0 and G∗ are 2-isomorphic. Hence G0 is the dual of G. q

Let G be a plane graph and G∗ be the dual of G. So to every vertex of G we have a

region of G∗ . Let the region φ of the plane graph G∗ corresponding to the vertex v of G has
e∗1 , e∗2 , . . ., e∗k as its boundary edges. Then each of these edges e∗i crosses the corresponding
edge ei of G, and these edges are all incident at the vertex v. It follows that φ contains the
vertex v. As G∗ is a plane graph, we can construct the dual of G∗ , called the double dual of
G and is denoted by G∗∗ .

We now have the following observation.

Theorem 6.19 Let G be a plane connected graph. Then G is isomorphic to its double
dual G∗∗ .

Proof Let G be a plane connected graph and G∗ be the dual of G. Now, any region φ of
dual G∗ contains exactly one vertex of G, namely its corresponding vertex v of G. This is
because the number of the regions of G∗ is same as the number of vertices of G.
Thus in the construction of G∗∗ , we take the vertex v to be the vertex in G∗∗ corresponding
to the region φ of G∗ . This choice gives the required isomorphism. q

Definition: A connected graph G is called self-dual if it is isomorphic to its dual G∗ . For

example, K4 is self-dual.

The next result is a characterisation of bipartite graphs in terms of duality.

Theorem 6.20 A connected plane graph G is bipartite if and only if its dual graph G∗ is
Eulerian.

Proof Let G be a bipartite graph. Then G does not contain odd cycles. Since G is a plane,
all the regions of G are of even length. So degree of every region in G is even. Let G∗ be
the dual graph of G. Since the vertex degrees of G∗ are same as the corresponding degrees
of the regions of G, degree of every vertex in G∗ is even. Hence G∗ is Eulerian.
Conversely, assume G∗ is Eulerian. Then degree of every vertex in G∗ is even. Since
the degree of a vertex in G∗ corresponds to the degree of the region in G, degree of every
region in G is even. This shows that G contains only even cycles. Hence G is bipartite. q
152 Planarity

Dual of a subgraph: Let G be a plane graph and let G∗ be its dual. Let e be an edge in
G and let e∗ be the corresponding edge in G∗ . We delete the edge e and find dual of G − e.
If edge e is on the boundary of two regions, then removal of e merges these two regions
into one.
Thus the dual (G − e)∗ is obtained from G∗ by deleting the corresponding edge e∗ and
then fusing the two end vertices of e∗ in G∗ − e∗ . On the other hand, if edge e is not on the
boundary, then e∗ forms a self-loop. In that case, G∗ − e∗ is the same as (G − e)∗ .
Thus, if a graph G has a dual G∗ , then the dual of any subgraph of G can be obtained by
successive application of this procedure.
Dual of a homeomorphic graph: Let G be a plane graph and G∗ be its dual. Let e be
an edge in G and let e∗ be the corresponding edge in G∗ . We create an additional vertex
in G by introducing a vertex of degree two in edge e (e now becomes two edges in series).
This simply adds an edge parallel to e∗ in G∗ . Likewise, the reverse process of merging two
edges in series will simply eliminate one of the corresponding parallel edges in G∗ . Thus,
by this procedure, if the graph G has dual G∗ , the dual of any graph homeomorphic to G
can be obtained from G∗ . These are illustrated in Figure 6.18(b).

The following important result is due to Whitney [267].

Theorem 6.21 (Whitney) A graph has a dual if and only if it is planar.
Proof We have only to prove that a nonplanar graph does not have a dual. Let G be
a nonplanar graph. Then according to Kuratowski’s theorem, G contains K5 or K3, 3 , or a
sub-division of K5 or K3, 3 .
We know that a graph G has a dual only if every subgraph H of G and every subdivision
of G has a dual. Thus to prove the result, we show that neither K5 nor K3, 3 has a dual.

Fig. 6.18(b)
Graph Theory 153

a. Suppose that K3, 3 has a dual D. Then the bonds in K3, 3 correspond to cycles in D and
vice-versa. Since K3, 3 has no bond consisting of two edges, D has no cycle consisting
of two edges. That is, D contains no pair of parallel edges. Since every cycle in K3, 3
is of length four or six, D has no bond with less than four edges. Therefore the degree
of every vertex in D is at least four. As D has no parallel edges and the degree of every
vertex is at least four, D has at least five vertices each of degree four or more (≥ 4).
That is, D has at least 5×4
2 = 10 edges. This is a contradiction to the fact that K3, 3 has
nine edges. Hence K3, 3 has no dual.
b. Suppose K5 has a dual H . We note that K5 has
i. 10 edges,
ii. no pair of parallel edges,
iii. no bond with two edges, and
iv. bonds with only four or six edges.
Therefore graph H has

i. 10 edges,
ii. no vertex with degree less than three,
iii. no bond with two edges, and
iv. cycles of length four or six only.

Now graph H contains a hexagon (a cycle of length six) and no more than three edges
can be added to a hexagon without creating a cycle of length three, or a pair of parallel
edges. Since both of these are not present in H , and H has 10 edges, there must be at least
seven vertices in H . The degree of each of these vertices is at least three. This implies that
H has at least 11 edges, which is a contradiction. Hence K5 has no dual. q

Definition: Two plane graphs G and G∗ are said to be duals (or combinatorial duals) of
each other if there is a one-one correspondence between the edges of G and G∗ such that if
H is any subgraph of G, and H ∗ is the corresponding subgraphs of G∗ , then rank (G∗ −H ∗ ) =
rank G∗ − nullity H .

6.6 Polyhedron
A polyhedron is a solid bounded by surfaces, called faces, each of which is a plane. That
is, a solid bounded by plane surfaces is called a polyhedron. For example, brick, window
frame, tetrahedron.
A polyhedron is said to be convex if any two of its interior points can be joined by
a straight line lying entirely within the interior. For example, a brick is convex, but a
window frame is not convex.
154 Planarity

The vertices and edges of a polyhedron, which form a skeleton of the solid, give a simple
graph in three dimensional space. It can be shown that for a convex polyhedron this graph
is planar.
To see this, imagine the faces of the convex polyhedron to be made of rubber, with
one face, say at the base, missing. Then taking hold of the edges at the missing face,
we are able to stretch out the rubber to form one plane sheet. The vertices and edges
in this transformation now form a plane graph. Moreover, each face of this plane graph
corresponds to a face of a solid, with the exterior face of the graph corresponding to the
missing face we used in stretching process.
Clearly, the plane graph is also connected, the degree of every vertex is at least 3 and
the degree of every face is also at least 3. Also, the graph is simple.
A simple connected plane graph G is called polyhedral if d(v) ≥ 3, for each vertex v of
G, and d(φ ) ≥ 3, for every face φ of G.
The following is a simple but useful property of polyhedral graphs.

Theorem 6.22 Let P be a convex polyhedron and G be its polyhedral graph. For each
n ≥ 3, let ni denote the number of vertices of G of degree i and let f i denote the number of
faces of G of degree i. Then

a. ∑ ini = ∑ i f i = 2m , where m is the number of edges of G.

i≥3 i≥3

b. The polyhedron P, and so the graph G, has at least one face bounded by a cycle of
length n for either n = 3, 4 or 5.

Proof
a. The expression ∑ ini is simply ∑ d(v), since d(v) ≥ 3 for each vertex v.
i≥3 v∈V (G)

We know, ∑ d(v) = 2m, so that ∑ ini = 2m .

v∈V (G) i≥3

For each n ≥ 3, the expression i f i is obtained by going round the boundary of each
face having a cycle of length i as its boundary, counting up the edges as we go. If we
do this over all possible i, then we count each edge twice, and so ∑ i f i = 2m .
i≥3

b. Assume to the contrary that P has no faces bounded by cycles of length 3, 4 or 5.

Then f 3 = f 4 = f 5 = 0.
So by (a), 2m = ∑ i f i ≥ ∑ 6 f i = 6 ∑ f i = 6 f , where f is the total number of faces
i≥6 i≥6 i≥6
of P. Thus, f ≤ 31 m.
Also by (a), 2m = ∑ ini ≥ ∑ 3ni = 3 ∑ ni = 3n, where n is the total number of
i≥3 i≥3 i≥3
vertices of P. Thus, n ≤ 23 m .
Now, by Euler’s formula, m = n + f − 2 and we have
Graph Theory 155

2 1
m ≤ m + m − 2 = m − 2 , and this implies that 2 ≤ 0,
3 3

which is impossible. Hence P must have a face of length 3, 4 or 5. q

Regular polyhedra: A polyhedron is called regular if it is convex and its faces are con-
gruent regular polygons (so that the polyhedral angles are all equal). The regular polyhedra
are also called Platonic bodies or Platonic solids.
A fact that has been known for at least 2000 years is that there are only five regular
polyhedra, which is proved next by using graph theoretic argument.

Theorem 6.23 The only regular polyhedra are the tetrahedron, the cube, the octahedron,
the dodecahedron and the icosahedron.

Proof Let P be a regular polyhedron and let G be its corresponding polyhedral graph.
Let n, m and f denote the number of vertices, edges and faces of G. Since the faces of
G are congruent to each other, each one is bounded by the same number of edges. Thus,
d(φ ) = k, k ≥ 3 for each face φ of G. Also, we know that there is at least one cycle of length
3, 4 or 5, therefore, 3 ≤ k ≤ 5. Similarly, since the polyhedral angles are all equal to each
other, the graph G is regular, say r-regular, where r ≥ 3. Thus we have

rn = 2m and k f = 2m, so that

2m = rn = k f . (6.23.1)

Using this in Euler’s formula n − m + f = 2, we have

8 = 4n − 4m + 4 f = 4n − 2m + 4 f − 2m = 4n − rn + 4 f − k f .

Therefore, 8 = (4 − r)n + (4 − k) f . (6.23.2)

Since n and f are both positive and also 3 ≤ k ≤ 5 and r = 3, there are only five possible
cases (Fig. 6.19).
Case 1 Let r = 3 and k = 3. Then 8 = (4 − 3)n + (4 − 3) f , so that 8 = n + f .
From (6.23.1.), n = f . So, n + f = 8 gives n = 4, f = 4.
Thus, n = 4, m = 6, f = 4, which is a tetrahedron.

Case 2 Let r = 3, k = 4. Then from (6.23.2), 8 = (4 − 3)n + (4 − 4) f .

So, n = 8. Then, (6.23.1) gives 3n = 4 f , so that f = 6. Also, m = 12.
So, n = 8, m = 12, f = 6, and thus is a cube.
156 Planarity

Fig. 6.19
Graph Theory 157

Case 3 Let r = 3, k = 5. Then (6.23.2) gives 8 = (4 − 3)n + (4 − 5) f , so that 8 = n − f .

From (6.23.1), 3n = 5 f . Solving 8 = n − f and 3n = 5 f , we get n = 20, f = 12.
Thus, n = 20, m = 30, f = 12, which is a dodecahedron.

Case 4 Let r = 4, k = 3. Then we get n = 6, f = 8.

So, n = 6, m = 12, f = 8, which is an octahedron.

Case 5 Let r = 5, k = 3. Then we have n = 12, f = 20.

Thus, n = 12, m = 30, f = 20, which is an icosahedron. q

The following is an elegant necessary condition for a plane graph to be Hamiltonian and
is due to Grinberg [86].
n
Theorem 6.24 If G is a loopless plane graph having a Hamiltonian cycle C, then ∑ (i −
i=2
2)(φi0 − φi00 ) = 0, where φi0 and φi00 are the number of regions of G of degree i contained in
interior C and exterior C respectively.

Proof Let E 0 and E 00 denote the sets of edges of G contained in int Cand ext C, respec-
tively, and let |E 0 | = m0 and |E 00 | = m00 . Then int C contains exactly m0 + 1 regions (Fig. 6.20)
and so
n
∑ φi0 = m0 + 1, (6.24.1)
i=2

Since G is loopless, φ10 = φ100 = 0.

Fig. 6.20
158 Planarity

Also, each edge in int C is on the boundary of exactly two regions in int C and each edge
of C is on the boundary of exactly one region in int C. Thus, counting the edges of all the
regions in int C, we obtain
n
∑ iφi00 = 2m0 + n. (6.24.2)
i=2

Eliminating m0 between (6.24.1) and (6.24.2), we get

n
∑ (i − 2)φi00 = n − 2. (6.24.3)
i=2

n
Similarly, ∑ (i − 2)φi00 = n − 2. (6.24.4)
i=2

The required result follows from equation (6.24.3) and (6.24.4). q

Example Consider the Herschel graph G shown in Figure 6.21. Clearly, G has 9 regions
and all the regions are of degree 4. Thus, if G were Hamiltonian, then 2(φ40 − φ400 ) = 0. This
implies φ40 = φ400 , which is impossible, since φ40 + φ400 = number of regions of degree 4 in
G = 9 is odd. Hence G is non-Hamiltonian.

Fig. 6.21

Konig observed that every graph is embeddable on some orientable surface. This is seen
by drawing an arbitrary graph G on the plane, possibly with edges that cross over each
other and then attaching a handle to the plane at each crossing and allowing one edge to
go over the handle and the other under it. Konig further showed that any embedding of
a graph on an orientable surface with a minimum number of handles has all its regions
simply connected.
Obviously, planar graphs can be embedded on a sphere. A toroidal graph is a graph
embedded on a torus, for instance, K5 , K3, 3 , K7 and K4, 4 .
Graph Theory 159

Definition: The genus γ (G)of a graph G is the minimum number of handles which are
added to the sphere so that G can be embedded on the resulting surface. Obviously γ (G) = 0
if and only if G is planar. The homeomorphic graphs have the same genus.
The genus of a polyhedron is the number of handles required on the sphere for a surface
to contain the polyhedron.

Definition: The least number of planar subgraphs whose union is a given graph G is
called the thickness of G. Clearly the thickness of any planar graph is 1. Also, K5 and K3, 3
have thickness 2, while the thickness of K9 is 3.
The crossing number of a graph G is the minimum number of pairwise intersections of
its edges when G is drawn in the plane. Clearly, crossing number of a graph is zero if and
only if G is planar.
The following result is due Euler and the proof can be found in Courant and Robbins [61].

Theorem 6.25 For a polyhedron of genus γ with n vertices, m edges and f regions

n − m + f = 2 − 2γ .

We have the following observations.

If G is a connected graph of genus γ , then

 3(n − 2 + 2γ ), i f every region in G is a triangle,
m=
2(n − 2 + 2γ ), i f every region in G is a quadrilateral.


From this, we have the following.

If G is a connected graph of genus γ , then γ ≥ 61 m − 12 (n − 2), and if G has no triangles,
then γ ≥ 41 m − 12 (n − 2).

The proof of the following result is due to Ringel and Youngs [223] and some details
can be found in Harary [104].

Theorem 6.26 For n ≥ 3, the genus of the complete graph is

(n − 3)(n − 4)
γ (Kn ) = .
12

6.7 Decomposition of Some Planar Graphs

Schnyder [232] proved that each triangulated planar graph G can be decomposed into three
edge disjoint trees. If instead triangles all faces (or regions) of G are quadrilaterals, and G
is without loops and multiple edges, Petrovic [183] proved that two trees are sufficient.
160 Planarity

The following two lemmas are used in proving Theorem 6.27.

Lemma 6.1 If G is a planar graph on n(n ≥ 4) vertices whose all regions are quadrilater-
als, then |E(G)| = 2n − 4.

Lemma 6.2 If G is a planar graph on n(n ≥ 4) vertices whose all regions are quadrilater-
als, then G contains at least three vertices of degree ≤ 3.

The following result is due to Petrovic [183].

Theorem 6.27 (Petrovic) Let G be a planar graph on n(n ≥ 4) vertices whose all faces
are quadrilaterals, and vr and vb any two non-adjacent vertices of G (we assume vr is
coloured red and vb blue). Then the edges of G can be partitioned into red and blue ones so
that red ones form a spanning tree Tr of G − vb , and blue ones a spanning tree Tb of G − vr .

Proof We call Tr the red, and Tb the blue tree. A vertex of G is red (respectively blue)
if it belongs to Tr (respectively Tb ). A vertex is red-blue if it belongs to both Tr and Tb .
Thus vr is the only red, and vb is only blue vertex in G. We proceed by induction on n. For
n = 4, the assertion is obvious. Assume that it holds for each graph on less than n(n > 4)
vertices, and consider a graph G on n vertices. Let u (u 6= vr , vb ) be a vertex of G of degree
≤ 3 existing by Lemma 6.2.
a. d(u) = 2. Let uv1 , uv2 be the edges, and uv1 xv2 u, uv1 yv2 u the faces incident with u.
Since G 6= C4 , each of vertices v1 and v2 has degree ≥ 3. It implies that all regions of
G − u are quadrilaterals. By induction hypothesis G − u can be partitioned into trees
Tr0 and Tb0 . Now, independent of the kind of vertices v1 and v2 , (red, blue or red-blue)
we can always colour the edges uv1 and uv2 so that the obtained trees Tr and Tb are as
desired.
b. d(u) = 3. Let uv1 , uv2 , uv3 be the edges and uv1 xv2 u, uv2 yv3 u, uv3 zv1 u be the faces
incident with u.
First assume that all vertices x, y, z are distinct. Since no triangle can be partitioned
into disjoint quadrilaterals whose vertices are vertices of the triangle and some its inner
points, xy, yz, zx ∈/ E(G). Suppose one of vertices x, y, z, is blue and there is no edge
connecting it with the opposite vertex of the hexagon v1 xv2 yv3 z. Without loss of generality,
we may assume it is x. Then xv3 ∈ / E(G). Denote by G00 the graph obtained from G by
deleting the edges xv1 and xv2 and identifying vertices u and x. Since each face of G00 is
a quadrilateral it can be partitioned into two trees Tr00 and Tb00 , where V (Tr00 )= V (G00 − vb )
and V (Tb00 ) = V (G00 − vr ). If edges uv1 and uv2 are coloured differently in G00 0 , say uv1 red
and uv2 blue, we colour xv1 red and xv2 blue obtaining the required trees Tr = Tr0 + xv1 and
Tb = Tb00 + xv2 in G. Therefore assume that uv1 and uv2 are monocoloured, say red. Then one
of vertices v1 and v2 , say v1 , is red−blue. As u is also a red-blue vertex, there is the unique
blue u − v1 path P in G00 . If P contains uv3 , we colour xv1 and xv2 blue and red, respectively.
If not, we colour both edges xv1 and xv2 red, and recolour uv1 blue. It is routine to check
that obtained trees Tr = Tr00 + xv1 , Tb = Tb00 + xv2 in the first case, and Tr = Tr00 − uv1 + xv1 + xv2 ,
Tb = Tb00 + uv1 in the second, are as required.
Graph Theory 161

Now, assume none of the vertices x, y, z satisfy the condition above. Since at most one of
edges xv3 , yv1 , zv2 can exist, we may assume without loss of generality that x is red-blue, y
is the red, z is the blue vertex, xv3 ∈ E(G) and yv1 ∈/ E(G). Then each region of the graph G000
obtained from G by deleting edges yv2 and yv3 and identifying y and u is a quadrilateral. By
induction ypothesis, G000 can be composed into trees Tr000 and Tb000 , where V (T 000 ) = V (G000 − z)
and V (Tb000 ) = V (G000 − y). Colouring all edges incident with y red, and recolouring the edge
uv3 blue, we obtain the decomposition of G into trees Tr and Tb , where V (Tr ) = V (G − z) and
V (Tb ) = V (G − y).
Suppose two of vertices x, y, z, say y and z, coincide. Then d(v3 ) = 2. If v3 is red-
blue, we finish the proof as in case (a). If x is red-blue we proceed as above. So we
may assume that {v3 , x} = {vr , vb ). Let v3 = vr , x = vb . By induction hypothesis the graph
GIV = G − v3 can be decomposed into trees TrIV and TbIV , where V (TrIV ) = V (GIV − x) and
V (TbIV ) = V (GIV − u). Then edges uv1 , uv2 are coloured red and edges xv1 , xv2 , blue. It
implies that edges yv1 , yv2 are coloured differently, yv1 , red and yv2 , blue. Now, colouring
v3 u and v3 y red and recolouring uv1 blue, we get from TrIV and TbIV the desired trees Tr and
Tb , where V (Tr ) = V (Tr ) = V (G − x) and V (Tb ) = V (G − v3 ).
If x = y = z, then v(G) = {u, x, v1, v2 , v3 }, E(G) = {uv1 , uv2 , uv3, xv1 , xv2 , xv3 } and the state-
ment holds trivially. q

6.8 Exercises
1. If G is a connected planar graph of order less than 12, prove that δ (G) = 4.
2. If G is a planar graph of order 24 and is regular of degree 3, then what is the number
of regions in a planar representation of G?
3. Prove that Euler’s formula fails for disconnected graphs.
4. Show that every graph with at most three cycles is planar.
5. Find a simple graph G with degree sequence [4, 4, 3, 3, 3, 3] such that
a. G is planar,
b. G is nonplanar.
6. Show that every simple bipartite cubic planar graph contains a C4 .
7. Prove that a simple planar graph has at least 4 vertices of degree 5 at most.
8. Let G be a planar, triangle free graph of order n. Prove that G has no more than 2n − 4
edges.
9. Let G be of order n = 11. Show that at least one of G or G is non planar.
10. Show that the average degree of a planar graph is less than 6.
11. Use Kuratowski’s theorem to prove that the Peterson graph is nonplanar.
162 Planarity

12. Show that the edges forming a spanning tree in a planar graph G correspond to the
edges forming a set of chords in the dual G∗ .
13. Prove that the complete graph K4 is self-dual.