TOPOLOGICAL GRAPH THEORY

AND THE HEAWOOD PROBLEM

A THESIS
submitted in partial fulfillment of the requirements
for the award of the dual degree of

Bachelor of Science - Master of Science

in

MATHEMATICS
by

SREEKANTH D
(13142)

DEPARTMENT OF MATHEMATICS
INDIAN INSTITUTE OF SCIENCE EDUCATION AND RESEARCH
BHOPAL
BHOPAL - 462066
April 2018
 i

CERTIFICATE

This is to certify that Sreekanth D, BS-MS (Dual Degree) student in Department

of Mathematics, has completed bonafide work on the dissertation entitled ‘Topo-
logical Graph Theory and The Heawood Problem’ under my supervision
and guidance.

April 2018 Dr. Kashyap Rajeevsarathy

IISER Bhopal

Committee Member Signature Date

Dr. Kashyap Rajeevsarathy

Dr. Nikita Agarwal

Dr. Prahlad Vaidyanathan

 ii

ACADEMIC INTEGRITY AND

COPYRIGHT DISCLAIMER

I hereby declare that this MS-Thesis is my own work and, to the best of my
knowledge, that it contains no material previously published or written by another
person, and no substantial proportions of material which have been accepted for
the award of any other degree or diploma at IISER Bhopal or any other educa-
tional institution, except where due acknowledgement is made in the document.

I certify that all copyrighted material incorporated into this document is in com-
pliance with the Indian Copyright (Amendment) Act (2012) and that I have
received written permission from the copyright owners for my use of their work,
which is beyond the scope of that law. I agree to indemnify and safeguard IISER
Bhopal from any claims that may arise from any copyright violation.

April 2018 Sreekanth D

IISER Bhopal
 iii

ACKNOWLEDGEMENT

Firstly I express my sincere gratitude toward my supervisor, Dr. Kashyap Ra-

jeevsarathy for his patience and guidance throughout my research. Next, I thank
all my teachers for inspiring me and giving me their full support. I also like to
thank the IISER Bhopal community for providing me with a friendly environment
for acquiring knowledge. Finally, I want to thank my family, friends and above
all, the almighty God for always being there for me.

Sreekanth D
 iv

ABSTRACT

Topological graph theory is a branch of graph theory that studies graphs as

topological spaces, their embeddings on surfaces and other properties alongside
the combinatorial and algebraic definition. The primary objective of topological
graph theory is to study graph embeddings on surfaces, which in layman’s terms,
pertains to understanding whether a given graph can be drawn on a surface
without crossings.
We will be focusing on graph embedding in closed orientable surfaces. We begin
by understanding planar embedding and the Kuratowski’s theorem, one of the
well-known results in topological graph theory [3] and its application. We will
then learn about graph embeddings on higher genus surfaces. Finally, we will
discuss the Heawood problem and its solution for orientable surfaces. In this
direction, we will derive the complete graph orientable embedding inequality
which gives a relation for the genus of a complete graph.
 CONTENTS

Certificate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

Academic Integrity and Copyright Disclaimer . . . . . . . . . . . . . ii

Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

1. PRELIMINARIES . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Walk, Path and Connectivity . . . . . . . . . . . . . . . . . . . 5
1.3 Graph Operations . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Surfaces and Simplicial Complexes . . . . . . . . . . . . . . . . 10
1.5 Graph Embedding . . . . . . . . . . . . . . . . . . . . . . . . . 11

2. PLANAR EMBEDDINGS OF GRAPHS . . . . . . . . . . . . . . 14

2.1 Planarity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Kuratowski’s Theorem . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Planarity Algorithms . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.1 Naive Planarity Algorithm . . . . . . . . . . . . . . . . . 22
2.3.2 Polynomial time Planarity Algorithm . . . . . . . . . . . 24
 Contents vi

3. NON-PLANAR EMBEDDINGS OF GRAPHS . . . . . . . . . . . 27

3.1 Band Decompositions . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Orientability . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3 Rotation System . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Edge-Deletion Surgery . . . . . . . . . . . . . . . . . . . . . . . 32
3.5 Orientable interpolation theorem . . . . . . . . . . . . . . . . . 32
3.6 Maximum Genus of a Graph . . . . . . . . . . . . . . . . . . . . 34
3.7 Heffter-Edmonds Algorithm . . . . . . . . . . . . . . . . . . . . 39

4. HEAWOOD PROBLEM . . . . . . . . . . . . . . . . . . . . . . . 40
4.1 Heawood problem . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Heawood inequality . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3 Complete Graph Embedding . . . . . . . . . . . . . . . . . . . . 44

Appendices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

A. COMPUTATIONAL COMPLEXITY BASICS . . . . . . . . . . . I

B. ADDITIONAL THEOREMS . . . . . . . . . . . . . . . . . . . . . II

Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
 1. PRELIMINARIES

In this introductory chapter, we provide the background to the material that we

present more formally in later chapters.

1.1 Graphs
Definition 1.1. A graph G is an ordered pair (V (G), E(G)) consisting of a set
V (G) of vertices and a set E(G) of edges, together with an incidence function
ψG that associates with each edge an unordered pair of distinct vertices called
the endpoints of that edge.

Fig. 1.1: Graph G

Example 1.2. Let G (see Figure 1.1) be a graph with vertex set V := {a, b, c, d, e},
edge set E := {e1 , e2 , e3 , e4 , e5 , e6 } and ψG (e1 ) = {a, b}, denoted simply by
e1 = ab.
 1. PRELIMINARIES 2

For a graph G, the cardinality of V (G) is called it’s order, denoted by #G.
Throughout this thesis, we will denote the cardinality of a set A by #A.

Definition 1.3. Let G be a graph, consider v1 , v2 ∈ V (G) and e1 , e2 ∈ E(G).

(i) The vertices v1 and v2 are said to be adjacent if there exists e ∈ E(G)
such that ψG (e) = {v1 , v2 }.

(ii) The edges e1 and e2 are said to be adjacent if e1 and e2 have a common
endpoint.

(iii) The neighbourhood of a vertex v1 is defined to be the set NG (v1 ) := {v ∈

V | v is adjacent to v1 }.

(iv) The valence of a vertex v1 is defined to be the cardinality of its neighbour-

hood, that is valence(v1 ) = #NG (v1 ).

Definition 1.4. Let G be a graph and v ∈ V (G).

(i) An edge e of form {v} is called a loop.

(ii) If E(G) is a multiset, then G is called a multigraph.

(iii) G is called simple if G is not a multigraph and contains no loop.

(iv) If #G < ∞ then G is called a finite graph.

(v) G is called k-regular if valence(v) = k for all v ∈ V (G).

In this thesis, a graph G is assumed to be undirected, finite and simple unless

stated otherwise.

Definition 1.5. Let G be a graph.

(i) A graph G0 is called a subgraph of G (denoted by G0 ⊆ G) if V (G0 ) ⊆

V (G), E(G0 ) ⊆ E(G) and ψG0 is a restriction of ψG on E(G0 ).
 1. PRELIMINARIES 3

(ii) If V (G0 ) ( V (G) and E(G0 ) ( E(G), then G0 is called a proper subgraph
of G (denoted by G0 ( G).

(iii) If V (G0 ) = V (G) then G0 is called a spanning subgraph of G.

(iv) The subgraph G0 is called an induced subgraph of G if E(G0 ) := {e ∈

E(G) | ψG (e) = {u, v} ∀ u, v ∈ V (G0 )}.

Definition 1.6. A graph Kn of n vertices is called a complete graph if for each

vertex v ∈ V (Kn ), NKn (v) = V (Kn ) (see Figure 1.2).

Fig. 1.2: Complete graph K5

F
Definition 1.7. A graph Km,n with vertex set V (Km,n ) = X Y where #X =
m and #Y = n, such that E(Km,n ) := {uv | u ∈ X, v ∈ Y } is called a
complete bipartite graph (see Figure 1.3).
 1. PRELIMINARIES 4

Fig. 1.3: Complete bipartite graph K3,3

Definition 1.8. A cycle Cn is a graph of n vertices with V (Cn ) := {v1 , v2 , ..vn }

such that E(Cn ) := {vi vi+1 | 1 ≤ i ≤ n − 1} ∪ {v1 vn }.

Definition 1.9. For n ≥ 4, a wheel Wn is defined to be a graph of n vertices

with V (Wn ) := V (Cn−1 ) ∪ {uv | u = V (K1 )} and v ∈ V (Cn−1 )} (See Figure
1.4).

Fig. 1.4: K4 or W4

Definition 1.10. Two graphs G and H are said to be isomorphic (written as

 1. PRELIMINARIES 5

G∼
= H), if there are bijections θ : V (G) −→ V (H) and φ : E(G) −→ E(H)
such that ψG (e) = uv if and only if ψH (φ(e)) = θ(u)θ(v) (see Figure 1.5).

Fig. 1.5: Isomorphic graphs

1.2 Walk, Path and Connectivity

Definition 1.11. Let G be a graph.

(i) A walk is a sequence of vertices v0 , v1 , ..., vi , ..., vn such that vi ∈ V (G) for
1 ≤ i ≤ n and vi vi+1 ∈ E(G) for 1 ≤ i ≤ n − 1. The vertices v0 and vn
are called the initial and terminal points, respectively.

(ii) A walk is said to be closed if v0 = vn .

(iii) A path is a walk v0 , ..., vn such that vi 6= vj for 0 < i, j < n.

The number of edges transversed in a path (or a walk) is called the length of the
path (or the walk).
 1. PRELIMINARIES 6

Definition 1.12. A tree T is a graph in which any two vertices of T are connected
by exactly one path.

Remark 1.13. let T be a tree such that #V (T ) = n (see [4]).

(i) #E(T ) = n − 1.

(ii) T is bipartite.

(iii) T has no simple cycles.

Definition 1.14. A graph G is called connected if there exists a path from u to

v, for all u, v ∈ V (G) (see Figure 1.6).

Definition 1.15. A vertex v ∈ V (G) of a graph G is called a cutpoint of G if

removing v along with its incident edges from G disconnects G (see Figure 1.6).

Definition 1.16. Let G be a graph and x, y ∈ V (G).

(i) The maximum number of pairwise internally disjoint paths from x to y in

G is called the local connectivity, which we denote by ρ(x, y).

(ii) G is called k-connected (see Figure 1.6) if ρ(x, y) ≥ k for all x, y ∈ V (G).
k is called the connectivity κ(G) of G.
 1. PRELIMINARIES 7

Fig. 1.6: Examples for connectedness.

1.3 Graph Operations

Let G be a graph, v ∈ V (G) and e = uv ∈ E(G).

Definition 1.17. The subgraph G − v with V (G − v) := V (G) \ v and E(G −

v) := E(G)\{uv ∈ E(G) | u ∈ NG (v)} is called the graph obtained by deleting
vertex v from G (see Figure 1.7).
 1. PRELIMINARIES 8

Fig. 1.7: Vertex deletion at v.

Definition 1.18. The subgraph G − e with V (G − e) := V (G) and E(G − e) :=

E(G) \ e is called the graph obtained by deleting edge e from G (see Figure 1.8).

Fig. 1.8: Edge deletion at e.

Definition 1.19. The contraction of an edge e of the graph G, denoted by G/e

(see Figure 1.9) is the graph obtained from G by the following steps:

(i) delete vertices u and v from G.

(ii) insert a new vertex u0 such that u0 v ∈ E(G/e), for all v ∈ NG (u) ∪ NG (v).
 1. PRELIMINARIES 9

Fig. 1.9: Contracing edge e.

Definition 1.20. An n-subdivision of an edge e of a graph G (see Figure 1.10) is

the graph H obtained by adding n vertices to edge e, that is, H = (V (H), E(H))
where V (H) := V (G) ∪ {v1 , ..., vn } and E(H) := (E(G) \ e) ∪ {vi vi+1 |1 ≤ i ≤
n − 1}. The reversal of subdivision is called smoothing.

Fig. 1.10: Subdividing edge e.

 1. PRELIMINARIES 10

1.4 Surfaces and Simplicial Complexes

Definition 1.21. A topological space M is called n-manifold if M is Haus-
dorff and each point of M has an open neighborhood homeomorphic to the
n-dimensional open ball or the n-dimensional half-ball.

Definition 1.22. Let M be an n-manifold.

(i) The boundary of M is the collection of all points in M that do not have a
neighbourhood homeomorphic to the n-dimensional open ball;

(ii) M is called closed if it is compact and its boundary is empty.

A surface is a 2-dimensional manifold. In this thesis we will be dealing with

closed, orientable surfaces of genus g ≥ 0.

Definition 1.23. A (geometric) k-simplex is the convex hull of k + 1 affinely

independent points or vertices in Euclidean n-space Rn , denoted by [v0 , v1 , ..., vk ].
The simplex determined by any subset of {v0 , ..., vk } is called a face of [v0 , v1 , ..., vk ].

Definition 1.24. A simplicial complex K (see Figure 1.11) is a finite collection

of simplexes in Rn satisfying the two conditions.

(i) Every face of every simplex in K is a simplex in K.

(ii) The intersection of any two simplexes in K is a simplex in K.

Definition 1.25. Let K be a simplicial complex.

(i) The point set |K| :=

S
s∈K S is called the carrier of K.

(ii) If m is the largest integer such that K contains an m-simplex, then K is

called a m-complex.
 1. PRELIMINARIES 11

Fig. 1.11: Example of simplicial complexes.

(iii) The collection of all k-simplexes of K for k ≤ r is called the r-skeleton of

K. Denoted by K (r) .

Definition 1.26 (Triangulation). A triangulation of a topological space X is a

homeomorphism h from the carrier of some simplical complex K to the space
X. The image of a simplex of K under h is called a simplex of triangulation.

1.5 Graph Embedding

Remark 1.27. Any graph G can be represented by a topological space in the
following sense:

(i) V (G) is represented by a collection of distinct point in R3 .

(ii) E(G) is represented by a collection of distinct, internally disjoint arcs, home-

omorphic to the closed interval [0, 1] such that boundary points of the arcs
represent the endpoints of the corresponding edge
 1. PRELIMINARIES 12

Definition 1.28. Let G be a graph (topological representation) and Sg , a surface

of genus g. A graph embedding is a continuous one-to-one function i : G −→ Sg
such that the function i0 : G −→ i(G) obtained by restricting the range of i is
a homeomorphism (see Figure 1.12).

Definition 1.29. Let G be a graph and Let i : G −→ S be an embedding on

G on surface S.

(i) The set F (G) = S \ i(G) is called the set of regions (or faces) of the em-
bedding i. Each element of F represents a maximal connected component
of S \ i(G) and is called a region (or face).

(ii) The graph embedding i is called a 2-cell embedding if each region is home-
omorphic to an open disk.

(iii) Two graphs are said to be homeomorphic if both can be obtained from the
same graph by a sequence of subdivisions of edges.

Definition 1.30. Consider the graph G and surfaces S, T . The two embeddings
i : G −→ S and j : G −→ T are called weakly equivalent if there exists a
homeomorphism h : S −→ T such that h(i(G)) = j(G).

Fig. 1.12: Graph embedding on S0 and corresponding regions.

 1. PRELIMINARIES 13

Definition 1.31. The genus γG of a graph G is defined to be the smallest

number g such that the graph G embeds in the orientable surface Sg .

Theorem 1.32 (Euler). The sum of the valences of the vertices of a graph G
equals twice the number of edges.

X
valence(v) = 2#E
v∈V (G)

valence(v). Observe that each e ∈ E(G) gets added

P
Proof. Let T = v∈V (G)
P
exactly twice in the sum T . Hence, T = e∈E(G) 2 = 2#E. This concludes the
proof.
 2. PLANAR EMBEDDINGS OF
GRAPHS

In this chapter we will study graph embeddings on genus 0 surfaces.

2.1 Planarity
Definition 2.1. A graph G is said to be planar if and only if it can be embedded
on a sphere S0 .

Remark 2.2. Let G be a graph. The following are equivalent.

(i) G is planar.

(ii) It can be embedded on a plane.

(iii) Genus γG = 0 (see Lemma 2.4).

The inequality in the following lemma is known as the edge-region inequality and
it gives a relation for the number of faces and the number of edges for any graph
embedding.

Lemma 2.3. Let i : G −→ S be an embedding of a connected simple graph

which is not a tree, with atleast three vertices into any surface S. Then

2#E ≥ 3#F
 2. PLANAR EMBEDDINGS OF GRAPHS 15

generally,
2#E ≥ girth(G)#F

where girth(G) is the length of the minimum cycle.

P
Proof. Observe that, f ∈F sf = 2#E, where sf denote the number of sides of
the region f (see Figure 2.1). Since G is simple, sf ≥ 3 for each f ∈ F and
hence the conclusion.

Fig. 2.1: An instance of edge-region inequality.

Lemma 2.4. Let i : G −→ S0 be an embedding of a connected graph G in

the sphere. Then #V (G) − #E(G) + #F (G) = 2.

Proof. This proof proceeds by induction on the number #F (G) of regions. First,
observe that if #F (G) = 1, then G must be a tree, since the Jordan curve
theorem implies that any cycle would separate the sphere. Thus, #E(G) =
#V (G) − 1 (by Remark 1.13) , from which it follows that #V (G) − #E(G) +
#F (G) = 2.
Now suppose that the Euler’s formula holds when the number of regions is at
most n, and suppose that #F (G) = n + 1. Then some edge e lies in the
boundary walk of two distinct regions. Since the two regions are distinct, the
subgraph G0 obtained by removing the edge e is connected. Then #F (G0 ) =
#F (G) − 1 = n, so by induction, #V (G0 ) − #E(G0 ) + #F (G0 ) = 2. Since
 2. PLANAR EMBEDDINGS OF GRAPHS 16

#V (G0 ) = #V (G), #E(G0 ) = #E(G) − 1, and #F (G0 ) = #F (G) − 1, it

follows that #V (G) − #E(G) + #F (G) = 2.

Proposition 2.5. If graph G is n-connected, n ≥ 2, then every set of n points

of G lie in a cycle.

Proof. By the definition of n-connected graph, G has no cutpoints and there

must be a maximum of atleast n number of pairwise internally disjoint paths
between any two vertices x, y in V (G). Thus, for any set P = {p1 , p2 , ..., pn } ⊂
V (G), we can find two internally disjoint paths between p0 and pn such that all
pi (i 6= 1, n) lie in either of the two path. This gives a cycle containing P .

2.2 Kuratowski’s Theorem

Kuratowski’s theorem give the criterion for a graph to be planar. The graphs
K3,3 and K5 forms the complete set of obstruction in planar embedding. These
graphs are called the Kuratowski’s graphs.

Lemma 2.6. K5 and K3,3 are non-planar.

Proof. For K5 , (see Figure 2.2)

Fig. 2.2: K5 , #E = 10 and #F = 7 = (2 + #E − #F )

 2. PLANAR EMBEDDINGS OF GRAPHS 17

The edge-region inequality yields, 2 . 10 (=20)  3 . 7 (=21). Hence, K5 is

non-planar.

Now for K3,3 , (see Figure 2.3)

Fig. 2.3: K3,3 , #E = 9 and #F = 5

Again by edge-region inequality, since K3,3 is bipartite girth(G) ≥ 4. Thus we

get, 2 . 9 (=18)  4 . 5 (=20). Hence, K3,3 is non-planar.

Theorem 2.7 (Kuratowski). A graph G is planar if and only if G has no subgraph

homeomorphic to K5 or K3,3 .

Proof. From lemma 2.6, K5 and K3,3 are non-planar. So any graph containing a
homeomorph of K5 or K3,3 are non-planar. Thus the converse is proved.

Now we want to prove that, a graph G is non-planar then G has a subgraph

homeomorphic to K5 or K3,3 . Consider the (edge) minimal counter example
H such that H is non-planar and does not contain a homeomorph of K3,3 or
K5 . The minimality ensures that removal of any edge makes H planar. We also
assume that H does not have any vertex of valence 2, since a valence 2 vertex
can be considered as the 1-subdivision of some edge and hence, we can smooth
out valence 2 vertices to obtain H.
 2. PLANAR EMBEDDINGS OF GRAPHS 18

First, we claim that H is at least 2-connected or in other words, H does not

have any cutpoints. To show this, suppose H has a cutpoint v, then removal
of v disconnects H. Since H is non-planar, some component of H must be
non-planar. This contradicts the minimality of H, and the claim follows.

Next, we claim that there exists edge e such that H − e has no cutpoints. To
show this, assume that H has a no such edge. Then, for any edge e0 ∈ E(H),
H − e0 has a cutpoint. This means that H − e is 1-connected for all e, which
in turn shows that H is 1-connected. This contradicts the first claim, and hence
the claim holds.

Now, choose an edge e = {u, v} of H whose removal does not affect con-
nectivity. Consider H − e = H 0 which is planar and 2-connected. Thus we can
find a planar embedding of H 0 with a cycle C (by Proposition 2.5) such that C
contains u and v and the number of regions enclosed by C is maximal among
other embeddings. Let C = v0 , v1 , ...vk = v, vk+1 , ..., vl , v0 and consider path P
(see Figure 2.4). From the maximality of C we can see that there is no path
connecting two vertices in the set {v0 , v1 , ...vk } that lies exterior to C and fur-
thermore, there is no path connecting two vertices in the set {vk+1 , ..., vl , v0 }
that lies exterior to C.

The non-planarity of H implies that there is some structure inside cycle C that
restricts the insertion of edge e between u and v. While taking all the possible
structures we finally arrive at the structures shown in Figure 2.4 such that any
obstructing structure is homeomorphic to one of the graphs in Figure 2.4.
 2. PLANAR EMBEDDINGS OF GRAPHS 19

Fig. 2.4: All possible structures of H.

In Figure 2.4, the bottom right graph is homeomorphic to K5 and the rest are
homeomorphic to K3,3 .

The above proof can be developed into the naive planarity algorithm which will
be discussed in the next section. A much more efficient planarity testing algo-
rithm can be obtained from the following proof of Kuratowski’s theorem due to
Thomassen [3]. The idea of Thomassen’s proof relies on the following result by
Tutte which we state without proof [4].

Proposition 2.8 (Tutte). A graph G is 3-connected then it is a wheel or can be

obtained from a wheel by a sequence of operations of the following two types:
 2. PLANAR EMBEDDINGS OF GRAPHS 20

(i) The addition of new edge

(ii) The replacement of a vertex v having valence (≥ 4) by two adjacent points

v 0 and v 00 such that each point formerly joined to v is joined to exactly
one of v 0 and v 00 so that in the resulting graph, valence(v 0 ) ≥ 3 and
valence(v 00 ) ≥ 3

The above proposition ensures the existence of wheel structure for the 3-connected
graph in the following proof of Kuratowski’s theorem.

Theorem 2.9. Let G be a 3-connected graph with five or more vertices. Then
there is some edge e of G such that the graph G/e is also 3-connected.

Proof. (Thomassen) Suppose for every edge e, the contracted graph G/e has a
set of two vertices that disconnects it. One of those two vertices must be the
vertex obtained by identifying the two endpoints of the edge e or else the same set
of two vertices would also disconnect G, thereby contradicting the 3-connectivity
of G. Thus for every edge e = uv together with some third vertex w disconnect
G. Accordingly, let us choose an edge e and a vertex w such that the largest
component H of the graph G − {u, v, w} is the largest for any disconnecting set
consisting of three vertices, two of which are adjacent.
Let x be a vertex adjacent to w such that x lies in a component of G − {u, v, w}
other than the maximum component H. Since vertices w and x are the endpoints
of an edge of G, it follows that G has a disconnecting set of the form {w, x, y}.
Now claim that some component of G − {w, x, y} is larger than H, a contra-
diction. To see this, let H 0 be the subgraph of G induced by the vertices of
H together with u and v. Since both u and v are adjacent to vertices of H
(otherwise G would not be 3-connected), the subgraph H 0 is connected. On one
hand, perhaps the vertex y is not in H 0 . Since w and x are not in H 0 either, it
 2. PLANAR EMBEDDINGS OF GRAPHS 21

follows that H 0 is contained in a component of G − {w, x, y}, contradicting the

maximality of H. On the other hand, perhaps y is in H’. If H 0 − y is connected,
then there is again a contradiction of the maximality of H, since H 0 − y has
one more vertex than H. If H 0 − y were not connected, then one component
of H 0 − y would contain both the vertices u and v, since u is adjacent to v and
hence all the other components of H 0 − y are connected to the rest of the graph
G through the vertices y and w. This would imply that {y, w} disconnects G,
contradicting the 3-connectivity of G. We conclude that for some edge e, the
contracted graph G/e is 3-connected.

Corollary 2.10. A graph G contains no homeomorph of K5 or K3,3 then G is

planar.

Proof. We prove by induction on the number of vertices. The statement is vacu-

ously true of all graphs with four or fewer vertices. We assume that the statement
is true for all graphs with fewer than n vertices, for n ≥ 5.
Consider the n + 1 case. By Theorem 2.9, we can choose an edge e = {u, v}
such that G/e with the identified vertex v 0 is still 3-connected. This means that
G-v 0 is 2-connected. Now consider the cycle C containing all the neighbours of
v 0 (see Figure 2.5).

Fig. 2.5: Wheel stucture with cycle C.

 2. PLANAR EMBEDDINGS OF GRAPHS 22

Fig. 2.6: Homeomorphs of K5 and K3,3 respectively.

Now expand v 0 back to u and v . By induction hypothesis, G will not contain

the graphs in Figure 2.6 since those graphs are homeomorphic to K5 and K3,3 as
noted. Hence, G is planar.

2.3 Planarity Algorithms

Planarity checking has applications in many fields such as in VLSI designing,
architecture etc. Thus, an efficient planarity algorithm is desirable. In this
section we will discuss two planarity algorithms.

2.3.1 Naive Planarity Algorithm

Naive planarity algorithm [3] (see Figure 2.7) is an exponential time algorithm
and hence very inefficient. But it is a much straight forward algorithm.

Algorithm Run Time = O(2#E )

(i) G = (V, E) ;Input graph.

(ii) For each subset of E whose removal leaves only one nontrivial component
H. O(2#E )
 2. PLANAR EMBEDDINGS OF GRAPHS 23

(i) Hsmooth = H - { vertices vt of valence 2 + the incendent edges } +

{new edges between vertices in NG (vt ) }

(ii) if Hsmooth ∼
= K5 or K3,3 .
G is non-planar.
else
G is planar.

Fig. 2.7: An instance of naive planarity algorithm.

 2. PLANAR EMBEDDINGS OF GRAPHS 24

2.3.2 Polynomial time Planarity Algorithm

The algorithm we discuss here has a worst case running time of polynomial
order. Thus it is much more efficient than the naive algorithm. The algorithm is
as follows (see Figure 2.8) :

(i) Let G = (V, E) a 3-connected graph and #VG = n.

(ii) Choose edge e such that G/e is 3-connected. Now take G/e = H.

(iii) Now repeat step 2 with each H obtained after simplical contraction of
another similar edge in H till a Graph H 0 of 4 vertices is obtained. Note
that, H 0 = K4 .

(iv) Now Reverse each contraction of H 0 . and check whether there exist a cycle
C such that

• 3 vertices in C adjacent to both u and v.

• There exists a path vi1 uj1 vi2 uj2 in C such that vi ∈ NG (v) and uj ∈
NG (u)

(v) If any of the above test returns true ⇒ the graph is non-planar.

Pseudocode

G = (V, E) a 3-connected graph with #V = n.

if (#E ≥ 3n − 6) then
;Graph is non-planar. Exit

e = ei ∈ E such that G/ei is 3-connected

 2. PLANAR EMBEDDINGS OF GRAPHS 25

;Defining any edge array for tracking the edge contraction

k=0
BK[k] = e
H =G ;initializing H

do while (#VH > 4) T1 ≈ O((n − 4)(3n − 6).n.n)

H = H/e ;contracting e in H
e = ei ∈ HG such that H/ei is 3-connected
BK[+ + k] = e

;Now we have H = K4 , since K4 is the only 3-connected four vertex graph.

;Reversing the contraction on H.

do while (k ≥ 0) T2 ≈ O(n4 )
H = restore(H, BK[k])
;The restore function restores the edge BK[k] on H.
C = cycle with all vertices in NG (BK[k])
test1 = check that 3 vertices in C adjacent to both u and v .
test2 = check ∃ path vi1 uj1 vi2 uj2 in C, vi ∈ nbhd(v) and uj ∈ nbhd(u)
if (test1 = true or test2 = true) then

;G is non-planar.
else
;G is planar.
k =k−1
 2. PLANAR EMBEDDINGS OF GRAPHS 26

Fig. 2.8: An instance of O(nk ) planarity algorithm.

 3. NON-PLANAR EMBEDDINGS OF
GRAPHS

3.1 Band Decompositions

Let G −→ S be a 2-cell embedding of a graph in a surface. We can surround
each vertex by a small disk in the surface S and each edge by a thin band such
that the union of all disks and bands is a neighbourhood of G in S whose shape
preserves that of the graph itself. The complement of this neighbourhood in
S gives a family of open disks, one just inside each region of the embedding.
Thus we define band decomposition of a surface which is the two-dimensional
version of the topological construction known as a handle decomposition of an
n-manifold.

Definition 3.1. For n=0, 1 and 2.

(i) A 1-band is a topological space b together with a homeomorphism h :

I × I → b, where I = [0, 1].
h(I × {j}) and h({j} × I) for j = 0, 1 are called the ends and the sides

of b respectively.

(ii) A 0-band or 2-band is simply a homeomorph of the unit disk (see Figure
3.1).
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 28

Fig. 3.1: Band Decomposition of K3,3 −→ S1

Definition 3.2. A band decomposition of a surface S is a collection B of 0-

bands, 1-bands and 2-bands satisfying these conditions:

(i) Different bands intersect only along arcs in their boundaries.

(ii) The union of all the bands is S.

(iii) Each end of each 1-band is contained in a 0-band.

(iv) Each side of each 1-band is contained in a 2-band.

(v) The 0-bands are pairwise disjoint and the 2-bands are pairwise disjoint.

Definition 3.3. Reduced Band decomposition of a graph embedding is the band

decomposition with the 2-bands removed (see Figure 3.2).
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 29

Fig. 3.2: Reduced band decomposition of graph imbedding in Klein bottle.

3.2 Orientability
Definition 3.4. Let B be a band decomposition. Then

(i) B is called locally oriented if each 0-band is assigned an orientation.

(ii) A 1-band is called orientation-preserving if the directions induced on its

ends by adjoining 0-bands are the same as those induced by one of the
two possible orientations of the 1-band, otherwise it is called orientation-
reversing (see Figure 3.3).
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 30

Fig. 3.3: Orientation of bands.

Definition 3.5. Consider the graph imbedding G → S with locally oriented

band decomposition and edge e.

(i) Type 0: if the corresponding 1-band of e is orientation preserving.

(ii) Type 1: if the corresponding 1-band of e is orientation reversing.

3.3 Rotation System

Definition 3.6. A rotation at a vertex v of a graph is an ordered list, unique up
to cyclic permutation, of the edges incident on that vertex.

Definition 3.7. A rotation system on a graph is an assignment of a rotation to

each vertex and a designation of orientation type for each edge.

Remark 3.8. Representing a rotation system as a diagram).

(i) Draw dots representing each vertex with spokes radiating from the dot
labelled in clockwise order according to the rotation at the vertex.

(ii) Draw curves joining spokes with the same label. Finally, all type-1 edges
are marked with a cross.
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 31

These kind of diagrams are called rotation projections (see Figure 3.4).

Fig. 3.4: Rotation projection of K4 (left) and its reduced band decomposition
(right).

Example 3.9. The rotation system for K4 shown in figure 3.4 can be given a
list format in edge form.
u.c1 ba
v.f ad
w.dbe
x.ef c1
By tracing along the boundary of the reduced band decomposition surface, we
can verify that that the embedding has two faces af ebcedf c and adb.

The following theorem which we state without proof tells the existence and
uniqueness of a rotation system for every locally oriented graph embedding (see
[3]).

Theorem 3.10. Every pure rotation system (all edges are of type 0) for a graph
G induces (up to orientation-preserving equivalence of embeddings) a unique
embedding of G into an oriented surface. Conversely, every embedding of a
graph G into an oriented surface induces a unique pure rotation system for G.
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 32

3.4 Edge-Deletion Surgery

Definition 3.11. Let G → S be a cellular embedding of a graph in a surface,
and e ∈ E(G). Consider the band decomposition B obtained by the following
operation on the band decomposition for G → S:

(i) delete the 2-bands that meet the e-band

(ii) delete the e-band

(iii) close the holes with one or two 2-bands as needed.

The operations performed to obtain B is called the edge-deletion surgery.

Remark 3.12. The effect of the surgery depends on three cases that arises when
the edge e is deleted.

(i) The two sides of e lie in different faces, f1 and f2 . Then deleting f1 − band,
f2 − band and e − band leaves one hole, which can be closed off by one
new 2-band.

(ii) Face f is pasted to itself along edge e without a twist. Deleting the f −band
and e − band leaves two distinct holes. These holes can be closed off with
two new 2-bands.

(iii) Face f us pasted to itself with a twist along edge e, so that the union of
the f − band and the e − band is a Mobius band. Deleting the f − band
and e − band leaves one hole, that can be closed with one new 2-band.

3.5 Orientable interpolation theorem

Definition 3.13. The genus range of a graph, denoted GR(G), is defined to be
the set of numbers g such that the graph G can be cellularly embedded in the
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 33

surface Sg .

Let γ(G) and γM (G) denote the minimum and maximum genus of G, respec-
tively.

Definition 3.14. The graph embeddings G → S and G → T are called adjacent

if there is an edge e in G such that the two embeddings (G − e) → S 0 and
(G − e) → T 0 are equivalent.

Remark 3.15. Adjacent embedding surfaces differ in genus by at most one. For
a graph G with two embeddings (G − e) → S 0 and (G − e) → T 0 equivalent,
then the most reasonable way to create embedding of G from the embedding of
G − e is to attach a handle and insert e through it or simply inserting e in the
same embedding if the resulting embedding remains cellular. This implies that
the respective adjacent embeddings of G differ by at most genus 1.

Theorem 3.16. Let G be a connected graph. Then the genus range GR(G) is
an unbroken interval of integers, that is, GR(G) = [γ(G), γM (G)].

Proof. Let G1 and Gn be the cellular embeddings with genus γ(G) and γM (G)
respectively. From Remark 3.11 we can find unique rotation systems R1 and Rn
for G1 and Gn respectively. Let Li denote the list format for rotation system
Ri . Then, there exists a permutation π that takes L1 to Ln . Then take the
decomposition of π into transpositions. For instance, let Li and Lj be list formats
in a transposition, then Lj may be obtained from Li by moving one edge symbol
at a time. These consecutive list formats represent adjacent embeddings. This
means that there exists a sequence of adjacent cellular embeddings of the G from
Sγ(G) (= G1 ) to SγM (G) (= Gn ). But we know that, the adjacent embedding
surfaces differ by at most one in genus and hence the conclusion.
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 34

3.6 Maximum Genus of a Graph

Objective of this section is to calculate the maximum genus γM (G) of graph G.
The most reasonable approach is to find if there exists a one-face embedding of
G on some surface Sg .

Definition 3.17. The Betti number β(G) of a connected graph G = (V, E), is
defined by the equation

β(G) = 1 − #V + #E

and it is equal to the number of edges in the complement of a spanning tree.

Lemma 3.18. Let d and e be adjacent edges in a connected graph G such that
G − d − e is a connected graph having an orientable one face embedding. Then
the graph G has a one face orientable embedding.

Proof. Let d = uv and e = vw be the two adjacent edges. Extend the one
face embedding (G − d − e) −→ S to a two-face embedding (G − e) −→ S by
placing the image of d across the single face. Observe that the vertex v lies in
both faces. Now, attach a handle from one face of (G − e) −→ S to the other
and place the image of e via the handle. This create a one-face embedding (see
Figure 3.5).
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 35

Fig. 3.5: An instance of the proof

Lemma 3.19. Let G = (V, E) be a connected graph such that every vertex has
valence atleast 3, and let G have a one-face orientable embedding G −→ S.
Then there exists adjacent edges d and e in G such that G − d − e has a one-face
orientable embedding.

Proof. Let d ∈ E whose two occurrences in the single boundary walk of the
embedding G −→ S are the closest together, among all other edges. Let the
boundary walk be dAdB where no edge appear twice in the sub walk. The edge
deletion surgery on d in the embedding G −→ S yields a two-face embedding
(G − d) −→ S 0 . The boundary walks of the two faces are A and B and the edge
e appears in both A and B. Thus the result of edge-deletion surgery on e in the
embedding (G − d) −→ S is a one-face embedding of G − d − e (see Figure
3.6).
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 36

Fig. 3.6: An instance of the proof

Definition 3.20. The deficiency ξ(G, T ) of a connected graph G with respect

to the spanning tree T is defined to be the number of components of G − T that
have an odd number of edges (see Figure 3.7).

Definition 3.21. The deficiency ξ(G, T ) of a connected graph G is defined by

ξ(G) := M in{ξ(G, T ) : T ∈ ST (G)}

where ST is the collection of all spanning trees of G.

 3. NON-PLANAR EMBEDDINGS OF GRAPHS 37

Fig. 3.7: Left: deficiency = 3 , Right: deficiency = 1

Lemma 3.22. Let T be a spanning tree for graph G, and let d and e be pair of
adjacent edges in G − T . Then ξ(G − d − e, T ) = 0 if and only if ξ(G, T ) = 0.

Proof. Every component of G − d − e − T that meets either of the edges d

or e has an even number of edges, since ξ(G − d − e, T ) = 0. It follows that
the number of edges in the components of G − T that contains the edges d
and e is even, and all other components of G − T has evenly many edges as in
G − d − e − T . This implies that ξ(G, T ) = 0. By a similar argument we can
prove the converse.

Theorem 3.23 (Xuong, 1979). Let G be a connected graph. Then G has a

one-face orientable embedding if and only if ξ(G) = 0.

Proof. We prove by induction. Assume that ξ(G) = 0 for any graph G with n or
fewer edges. Let G = (V, E) be a graph with n + 1 edges and valence(v) ≥ 3,
for all v ∈ V .
Suppose that G has one-face orientable embedding. Then by Lemma 3.19, there
exists adjacent edges d, e ∈ E such that G − d − e has one face embedding.
Then by the induction hypothesis ξ(G − d − e) = 0. This implies by Lemma 3.22
that there exists a spanning tree T in G − d − e such that ξ(G − d − e, T ) = 0.
This implies that ξ(G, T ) = ξ(G) = 0, since T spans G.
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 38

Conversely, suppose ξ(G) = 0. Then there exists a spanning tree T such that
ξ(G, T ) = 0. It follows from the above lemmas that there exists adjacent edges
d, e ∈ EG−T such that ξ(G − d − e, T ) = 0. This means that the graph G − d − e
has a one-face orientable embedding by induction hypothesis. It follow by Lemma
3.18 that G has a one-face orientable embedding.

Theorem 3.24 (Xuong, 1979). Let G be a connected graph. Then the minimum
number of faces in any orientable embedding of G is exactly ξ(G) + 1.
For any embedding of G,

#F ≥ ξ(G) + 1.

We can rephrase the above theorem as follows:

Theorem 3.25. The graph G has an orientable embedding with n + 1 or fewer

faces if and only if ξ(G) ≤ n.

Proof. We prove this by inducting on the number of faces n. It holds for n = 0.

Assume that the statement is true for all values of k less than or equal to n and
n > 0. We will now be using the arguments in Lemmas 3.18, 3.19, 3.22 and
Theorem 3.23 to prove the theorem.
Suppose G −→ S is an orientable embedding with #FG = n + 1. Perform
edge-deletion surgery on an edge e common to two faces of the embedding. The
resulting embedding G − e −→ S 0 has n faces. Then, by induction hypothesis,
ξ(G − e) ≤ n − 1 which implies that ξ(G) ≤ n.
Conversely, Suppose ξ(G) = n, then there exists a spanning tree T in G such
that ξ(G, T ) = n. Let H be a component of G − T with odd number of edges.
Now choose an edge e from H that does not disconnect H or such that one end
point of e has valence 1. It follows that ξ(G − e, T ) = n − 1. Thus, by the
 3. NON-PLANAR EMBEDDINGS OF GRAPHS 39

induction hypothesis, the graph G − e has an orientable embedding with at most

n faces. Hence, G has an orientable embedding with at most n + 1 faces.

Corollary 3.26 (Xoung, 1979). Let G be a connected graph. Then

1
γM (G) = (β(G) − ξ(G)).
2

Proof. Let g = γM (G). Then by Theorem 3.25, we have

2 − 2g = #V − #E + (ξ(G) + 1)

⇒ 2g = (1 + #E − #V ) − ξ(G)
1
⇒ g = (β(G) − ξ(G)).
2

3.7 Heffter-Edmonds Algorithm

An exponential time algorithm [3] for calculating the minimum genus γm of a
graph, based on rotation system enumeration.

Algorithm

(i) List all the pure rotation systems of a graph.

(ii) Compute the number of faces for each rotation system.

(iii) Choose the one having the most faces and calculate the genus.

Run time is O(2k ) because a regular (r + 1) valent graph with n vertices has
(r!)n pure rotation systems.
 4. HEAWOOD PROBLEM

In this chapter we will review the Heawood problem [3] and its solution with
respect to closed orientable surfaces Sg of genus g > 0. For a sphere S0 , the
problem is popularly known as the Four Color Theorem [4].

Definition 4.1. A graph G is said to be n-coloured with a set C of n distinct

elements (called colours) if there exists a surjective map ζn : V (G) −→ C such
that ζn (v1 ) 6= ζn (v2 ) for any adjacent vertices v1 , v2 ∈ V (G).

The n-colouring of a graph G can also be defined in terms of edge set E(G) or
by the face set F (G) of an embedding of G. We will be using the above (vertex)
definition throughout this thesis.

Definition 4.2. The chromatic number chr(G) of a graph G is defined to be

the smallest number n such that G has an n-colouring.

Definition 4.3. The chromatic number of a surface S is equal to the maximum

of the set of chromatic numbers of simplicial graphs that can be embedded in S.

Definition 4.4. A graph G is called chromatically critical if chr(G−e) < chr(G)

for any edge e ∈ E(G).

Remark 4.5. For complete graph Kn , the following holds.

(i) chr(Kn ) = n, since any two edges are adjacent in Kn .

(ii) Kn is chromatically critical. If we remove any edge e = uv from Kn then

we can colour u and v with the same colour, i.e. chr(Kn − e) < chr(Kn ).
 4. HEAWOOD PROBLEM 41

4.1 Heawood problem

For a surface S with Euler characteristic c ≤ 1, Percy John Heawood [3] showed
there is a finite maximum for chr(S), that is, there exists H(S) such that
chr(S) ≤ H(S). The finite maximum H(S) is called the Heawood number
of the surface S.

The determination of the chromatic numbers of the surfaces other than the
sphere is called the Heawood Problem. The solution of the Heawood problem
is that chr(S) = H(S), except for the Klein bottle. Due to the complexity
of the original solution of the Heawood problem [3], we will narrow down our
exploration of the Heawood problem to a basic overview of the implementation
of the Ringel-Young solution [3] for closed orientable surfaces. The initial step in
this pursuit is to derive the Heawood inequality. We will then use the Heawood
inequality to reduce the Heawood problem to finding the genus of the complete
graphs.

4.2 Heawood inequality

Lemma 4.6. Let S be a closed surface of Euler characteristic c, and let G be a
simplical graph embedded in S. Then

6c
average valence(G) ≤ 6 −
#V

Proof. From the Euler’s equation, we get

#V − #E + #F ≥ c
 4. HEAWOOD PROBLEM 42

and by the edge-region inequality we get #F ≤ 32 #E. Set #F to be 23 #E in

the above inequality. Then,

1
#V − #E ≥ c
3
#E 3c
⇒ + ≤3
#V #V
2#E 6c
⇒ + ≤6
#V #V
6c
⇒ average valence(G) ≤ 6 − ,
#V
P
since v∈V valence(v) = 2#E, By Theorem 1.32.

Lemma 4.7. Let S be a closed surface, and let G be a chromatically critical

graph such that chr(G) = chr(S). Then for every vertex v, chr(S) − 1 ≤
valence(v).

Proof. Suppose v is a vertex in G with valence(v) < chr(S) − 1. Since G is

chromatically critical, we get chr(G − v) ≤ chr(S) − 2. This means that we can
colour the neighbours of v in G−v with chr(S)−2 colours. This implies that we
can colour v and its neighbourhood with atleast chr(S) − 1 colours in G, that is,
chr(G) ≤ chr(S) − 1. This contradicts the assumption chr(G) = chr(S).

Theorem 4.8 (Heawood, 1890). Let S be a closed surface with Euler charac-
teristic c ≤ 1. Then
$ √ %
7+ 49 − 24c
chr(S) ≤ = H(S) (4.1)
2

Proof. If c = 1, H(S) = 6. S is the projective plane (see Figure 4.1) and

chr(S) = 6. The inequality holds.
 4. HEAWOOD PROBLEM 43

Fig. 4.1: Embedding of K6 in projective plane.

For c ≤ 0, let G be a graph embedded in S such that chr(G) = chr(S) and G

is chromatically critical. Then by Lemma 4.6 and 4.7, we get

6c
chr(S) − 1 ≤ average valence(G) ≤ 6 − ,
#V

which implies that

6c
chr(S) − 1 ≤ 6 −
#V
6c
⇒ chr(S) − 1 ≤ 6 −
chr(S)
⇒ chr2 (S) − 7chr(S) + 6c ≤ 0

Consider the inequality chr2 (S) − 7chr(S) + 6c ≤ 0, the quadratic polynomial

in the left side of the inequality yields
√ ! √ !
7− 49 − 24c 7 + 49 − 24c
chr(S) − chr(S) − ≤0
2 2

For c ≤ 0 ,the first factor is a non zero positive number and chr(S) is an integer.
Hence, we have the required result.
 4. HEAWOOD PROBLEM 44

4.3 Complete Graph Embedding

$ %
√
7+ 1+48g
For surface Sg of genus g, H(Sg ) = 2
. The Figure 4.2 shows the
Heawood number for surfaces of genus g where 1 ≤ g ≤ 23.

Fig. 4.2: Table showing H(Sg ) for corresponding genus g.

Theorem 4.9. & '

(n − 3)(n − 4)
γ(Kn ) ≥ (4.2)
12
Proof. By Euler equation, we have #V − #E + #F = 2 − 2γ(Kn ) and from
edge-region inequality, we have #F ≤ 32 #E. Setting #F = 32 #E, we get

1
#V − #E ≥ 2 − 2γ(Kn )
3
n2 − 7n + 12
⇒ γ(Kn ) ≥
12
(n − 3)(n − 4)
≥ .
12

Since γ(Kn ) is an integer, the conclusion follows.

Remark 4.10. It is possible to triangulate surfaces by complete graphs. By

complete graph orientable embedding inequality,
& '
(n − 3)(n − 4)
γ(Kn ) ≥
12

construct minimum embeddings of the complete graphs. The general form of

these embeddings seems to depend strongly on the residue class of n (mod 12).
 4. HEAWOOD PROBLEM 45

& ' !
(n−3)(n−4) (n−3)(n−4)
Let I(n) = 12
then I(n) − 12
measures how much an ori-
entable embedding Kn −→ SIn fails to! be a triangulation. If ((n − 3)(n − 4))
(mod 12) = 0 then I(n) − (n−3)(n−4)
12
= 0. This implies that Kn −→ SIn has
a triangulation by Kn (see [3]).
Appendices

46
 Appendix A

COMPUTATIONAL COMPLEXITY
BASICS

RAM (Random-Access Machine) Model

• No concurrent operations

• Common instructions (arithmetic, memory) take constant time.

• Data Type - Integer and Floating point.

Definition A.1 ( Run Time T (n) ). The number of primitive operations or steps
executed, for an algorithm for a particular input (size n).

Asymptotic Efficiency of algorithm: Growth of run time with respect to

increase in size of input.

Definition A.2 (Big-O Notation). f (n) is O(g(n)) if there exist positive number
c and N such that f (n) ≤ cg(n) ∀ n > N .

Run time T (n) = O(f (n)) ordered in decreasing order of asymptotic efficiency.

O(1) > O(log2 (n)) > O(n) > O(nlog2 (n)) > O(nc ) > O(2n ).

For more details refer [2].

 Appendix B

ADDITIONAL THEOREMS

Theorem B.1. Let G −→ Sg be a cellular embedding, for any g = 0, 1, 2, ...

Then χ(G −→ Sg ) = 2 − 2g.

Proof. Omitted, see [3] p.112.

Corollary B.2. Let i and j be distinct non-negative integers. Then Si and Sj

are not homeomorphic.

Proof. A homeomorphism f : Si −→ Sj would carry a cellular embedding

G −→ Si of relative Euler characteristic χ(G −→ Si ) = 2 − 2i to a cellular
embedding G −→ Sj of relative Euler characteristic χ(G −→ Si ) = 2 − 2i, in
violation of the invariance of Euler characteristic, which implies that χ(G −→
Sj ) = 2 − 2j.

Remark B.3. Let G −→ S be a cellular graph embedding, and let e be an edge

of the graph. Let F be the set of faces for G −→ S, and let F 0 be the set of
faces of the embedding obtained by edge-deletion surgery at e. Then in cases
(see Remark 3.11)

1. #F 0 = #F − 1, and the resulting surface is homeomorphic to S.

2. #F 0 = #F + 1.
 Appendix B. ADDITIONAL THEOREMS

3. #F 0 = #F.

Theorem B.4 (Classification of Surfaces). Every closed, connected, orientable

surface is homeomorphic to one of the standard surfaces Sg with g ≥ 0.

Proof. Omitted, see [3] p.128 .

 BIBLIOGRAPHY

[1] J. A. Bondy and U. S. R. Murty. Graph theory, volume 244 of Graduate

Texts in Mathematics. Springer, New York, 2008.

[2] Thomas H. Cormen, Charles E. Leiserson, Ronald L. Rivest, and Clifford

Stein. Introduction to Algorithms, 3rd Edition (MIT Press). The MIT Press,
3rd edition, 7 2009.

[3] Jonathan L. Gross and Thomas W. Tucker. Topological graph theory. Dover
Publications, Inc., Mineola, NY, 2001. Reprint of the 1987 original [Wiley,
New York; MR0898434 (88h:05034)] with a new preface and supplementary
bibliography.

[4] Frank Harary. Graph Theory. on Demand Printing Of 02787. Westview Press,
1969.

[5] James R. Munkres. Topology: a first course. Prentice-Hall, Inc., Englewood

Cliffs, N.J., 1975.