Mahima Ranjan Adhikari

Basic
Topology 3
Algebraic Topology and Topology
of Fiber Bundles
Basic Topology 3
Professor Mahima Ranjan Adhikari (1944–2021)
Mahima Ranjan Adhikari

Basic Topology 3
Algebraic Topology and Topology of Fiber
Bundles
Mahima Ranjan Adhikari
Institute for Mathematics, Bioinformatics,
Information Technology and Computer
Science (IMBIC)
Kolkata, West Bengal, India

ISBN 978-981-16-6549-3 ISBN 978-981-16-6550-9 (eBook)

https://doi.org/10.1007/978-981-16-6550-9

Mathematics Subject Classification: 55-XX, 22-XX, 14F35, 14Fxx, 18F15

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature
Singapore Pte Ltd. 2022
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
 Dedicated to
Prof. B. Eckmann and Prof. P. J. Hilton
for their fruitful interaction with them
during my academic visit to
ETH, Zurich, Switzerland
in 2003
Preface

Algebraic topology is an important branch of topology that utilizes algebraic tools to

study topological problems. The main aim of algebraic topology is to construct alge-
braic invariants on topological spaces to convert topological problems into algebraic
problems to have a better chance of a solution. For example, the algebraic invariants
stem from homotopy theory are the homotopy groups π n (X, x 0 ) of a pointed topolog-
ical space (X, x 0 ) for n ≥ 1. The other basic algebraic invariants are homology and
cohomology groups of a space. Their constructions are much more complicated than
those of homotopy groups. Fortunately, computations of homology and cohomology
groups are easier than those of homotopy groups. In algebraic topology, algebraic
invariants classify topological spaces up to homeomorphism. It is found that this clas-
sification is usually in most cases up to homotopy equivalence. It tries to measure
degrees of connectivity using homology, cohomology and homotopy groups.
Several approaches have been used in topology to associate a topological space
with a number of algebraic objects, for instance, groups, rings, etc. The basic idea
of algebraic topology is the correspondence (or functor) associating a collection
of certain algebraic objects to a collection of topological spaces and continuous
mappings of spaces to corresponding homomorphisms. This functorial approach
allows us to transform topological problem into the corresponding algebraic one. The
solvability of this ‘derived’ algebraic problem in many cases implies the solvability
of the initial topological problems.
The contents of Volume 3 are expanded into seven chapters and discuss geometric
topology and manifolds by using algebraic topology.
Chapter 1 provides a background on algebra, topology and Lie groups to facilitate
a smooth study of this volume.
Chapter 2 officially inaugurates a study of algebraic topology by conveying the
basic concepts of homotopy and fundamental groups born through the work of H.
Poincaré in his land-marking Analysis situs, Paris, 1895, and also discusses higher
homotopy groups constructed in 1935 by H. Hurewicz (1904–1956) in his paper
(Hurewicz, 1935), which are natural generalizations of fundamental groups. Homo-
topy theory studies those properties of topological spaces and continuous maps
which are invariants under homotopic maps, called homotopy invariants. Finally,

vii
viii Preface

this chapter presents some interesting applications of homotopy, fundamental and

higher homotopy groups in analysis, geometry, algebra, matrix theory, atmospheric
science, vector field and extension problems and some others.
Chapter 3 conveys the basic concepts of homology theory starting from its inven-
tion by Heny H. Poincaré (1854–1912) in 1895 to the approach formulating axiomati-
zation of homology, announced in 1952 by S. Eilenberg (1915–1998) and N. Steenrod
(1910–1971), now known as Eilenberg and Steenrod axioms. This approach simpli-
fies the proofs of many results by escaping avoidable difficulties to promote active
learning in homology and cohomology theories, which is the most important contri-
bution to algebraic topology after the invention homotopy and homology by Poincaré
in 1895. This functorial approach facilitates in variety of cases to solve topological
problems through the solvability of corresponding algebraic problems. The motiva-
tion of the study of algebraic topology comes from the study of geometric properties
of topological spaces from the algebraic viewpoint.
Chapter 4 discusses the topology of fiber bundles starting with general theory of
bundles and continues its study to Chap. 5. The topology of fiber bundles has created
general interest and promises for more work because it is involved of interesting
applications of topology to other areas such as algebraic topology, geometry, physics
and gauge groups. The theory of fiber bundles was first recognized during the period
1935–1940 through the work of H. Whitney (1907–1989), H. Hopf (1894–1971) and
E. Stiefel (1909–1978), J. Feldbau (1914–1945) and some others. A fiber bundle is
a bundle with an additional structure derived from the action of a topological group
on the fibers. A fiber bundle is a locally trivial fibration having covering homotopy
property.
Chapter 5 continues the study topology of fiber bundles from the viewpoint of
homotopy theory. Covering spaces provide tools to study the fundamental groups.
Fiber bundles provide likewise tools to study higher homotopy groups (which are
generalizations of fundamental groups). The notion of fiber spaces is the most fruitful
generalization of covering spaces. The importance of fiber spaces was realized during
1935–1950 to solve several problems relating to homotopy and homology.
Chapter 6 studies geometric topology primarily, which studies manifolds and
their embeddings in other manifolds. A particularly active area is low-dimensional
topology, which studies manifolds of four or fewer dimensions. This includes knot
theory, the study of mathematical knots. It proves more theorems and conveys further
applications of topological concepts and results discussed in earlier chapters with a
view to understand the beauty, power and scope of the subject topology. Moreover,
it provides alternative proofs of some results proved in the previous chapters such as
Brouwer–Poincaré theorem, Van Kampen theorem and Borsuk–Ulam theorem for
any finite dimension. It proves Ham Sandwich theorem and Lusternik–Schnirelmann
theorem.
Chapter 7 conveys the history of emergence of the concepts leading to the
development of algebraic topology as a subject with their motivations.
The list of chapters shows that the book covers a wide range of topics. Some are
more technical than others, but the reader without a great deal of technical knowl-
edge should still find most of the text accessible. Avoiding readymade proofs of
Preface ix

some theorems, their statements have appeared in this book in the form of problems
providing an opportunity for exploration of the topics of the book along with some-
what diverging from the basic thrust of the book, since solving problems plays a key
role in the study of mathematics.
The book is a clear exposition of the basic ideas of topology and conveys a
straightforward discussion of the basic topics of topology and avoids unnecessary
definitions and terminologies. Each chapter starts with highlighting the main results
of the chapter with motivation and is split into several sections which discuss related
topics with some degree of thoroughness and ends with exercises of varying degrees
of difficulties, which not only impart an additional information about the text covered
previously but also introduce a variety of ideas not treated in the earlier texts with
certain references to the interested readers for more study. All these constitute the
basic organizational units of the book.
The present book, together with the authors’ two other Springer books, Basic
Modern Algebra with Applications (M. R. Adhikari and Avishek Adhikari) and Basic
Algebraic Topology and its Applications (M. R. Adhikari), will form a unitary module
for the study of modern algebra, general and algebraic topology with applications in
several areas.
The author acknowledges Higher Education Department of the Government of
West Bengal for sanctioning the financial support to the Institute for Mathematics,
Bioinformatics and Computer Science (IMBIC), toward writing this book vide order
no. 432(Sanc)/EH/P/SE/SE/1G-17/07 dated August 29, 2017, and also to IMBIC,
the University of Calcutta, Presidency University, Kolkata, India, and Moulana Abul
Kalam Azad University of Technology, West Bengal, for providing the infrastructure
toward implementing the scheme.
The author is indebted to the authors of the books and research papers listed
in the Bibliography at the end of each chapter and are very thankful to Profes-
sors P. Stavrions (Greece), Constantine Udriste (Romania), Akira Asada (Japan) and
Avishek Adhikari (India) and also to the reviewers of the manuscript for their schol-
arly suggestions for improvement of the book. We are thankful to Md. Kutubuddin
Sardar for his cooperation towards the typesetting of the manuscript and to many UG
and PG students of Presidency University and Calcutta University, and many other
individuals who have helped in proofreading the book. Authors apologize to those
whose names have been inadvertently not entered. Finally, the author acknowledges,
with heartfelt thanks, the patience and sacrifice of the long-suffering family of the
author, especially Minati Adhikari, Dr. Shibopriya Mitra Adhikari, and the beloved
grand son Master Avipriyo Adhikari.

Kolkata, India Mahima Ranjan Adhikari

June 2021
A Note on Basic Topology—Volumes 1–3

The topic ‘topology’ has become one of the most exciting and influential fields of
study in modern mathematics, because of its beauty and scope. The aim of this subject
is to make a qualitative study of geometry in the sense that if one geometric object
is continuously deformed into another geometrical object, then these two geometric
objects are considered topologically equivalent, called homeomorphic. Topology
starts where sets have some cohesive properties, leading to define continuity of
functions.
The series of three books on Basic Topology is a project book funded by the
Government of West Bengal, which is designed to introduce many variants of a
basic course in topology through the study of point-set topology, topological groups,
topological vector spaces, manifolds, Lie groups, homotopy and homology theories
with an emphasis of their applications in modern analysis, geometry, algebra and
theory of numbers:
Topics in topology is vast. The range of its basic topics is distributed among
different topological subfields such as general topology, topological algebra, differ-
ential topology, combinatorial topology, algebraic topology and geometric topology.
Each volume of the present book is considered as a separate textbook that promotes
active learning of the subject highlighting elegance, beauty, scope and power of
topology.

Basic Topology—Volume 1: Metric Spaces and General

Topology

This volume majorly studies metric spaces and general topology. It considers the
general properties of topological spaces and their mappings. The special struc-
ture of a metric space induces a topology having many applications of topology
in modern analysis, geometry and algebra. The texts of Volume I are expanded into
eight chapters.

xi
xii A Note on Basic Topology—Volumes 1–3

Basic Topology—Volume 2: Topological Groups, Topology

of Manifolds and Lie Groups

This volume considers additional structures other than topological structures studied
in Volume 1 and links topological structure with other structures in a compatible
way to study topological groups, topological vector spaces, topological and smooth
manifolds, Lie groups and Lie algebra and also gives a complete classification of
closed surfaces without using the formal techniques of homology theory. Volume 2
contains five chapters.

Basic Topology—Volume 3: Algebraic Topology

and Topology of Fiber Bundles

This volume mainly discusses algebraic topology and topology of fiber bundles.
The main aim of topology is to classify topological spaces up to homeomorphism.
To achieve this goal, algebraic topology constructs algebraic invariants and studies
topological problems by using these algebraic invariants. Because of its beauty and
scope, algebraic topology has become an essential branch of topology. Algebraic
topology is an important branch of topology that utilizes algebraic tools to study
topological problems. Its basic aim is to construct algebraic invariants that classify
topological spaces up to homeomorphism. It is found that this classification, usually
in most cases, is up to homotopy equivalence.
This volume conveys a coherent introduction to algebraic topology formally inau-
gurated by H. Poincaré (1854–1912) in his land-marking Analysis situs, Paris, 1895,
through his invention of fundamental group and homology theory, which are topo-
logical invariants. It studies Euler characteristic, the Betti number and also certain
classic problems such as the Jordan curve theorem. It considers higher homotopy
groups and establishes links between homotopy and homology theories, axiomatic
approach to homology and cohomology inaugurated by Eilenberg and Steenrod. It
studies the problems of converting topological and geometrical problems to algebraic
one in a functorial way for better chance for solution.
This volume also studies geometric topology and manifolds by using algebraic
topology. The contents of Volume 3 are expanded into seven chapters.
Just after the concept of homeomorphisms is clearly defined, the subject of
topology begins to study those properties of geometric figures which are preserved by
homeomorphisms with an eye to classify topological spaces up to homeomorphism,
which stands the ultimate problem in topology, where a geometric figure is considered
to be a point set in the Euclidean space Rn . But this undertaking becomes hopeless,
when there exists no homeomorphism between two given topological spaces. The
concepts of topological properties and topological invariants play key tools in such
problems:
A Note on Basic Topology—Volumes 1–3 xiii

(a) The concepts of topological properties, such as, compactness and connect-
edness, introduced in general topology, solve this problem in very few cases
(studied in Basic Topology, Vol. 1).
(b) On the other hand, the subjects algebraic topology and differential topology
(studied in Volume 2) were born to solve the problems of impossibility in
many cases with a shift of the problem by associating invariant objects in the
sense that homeomorphic spaces have the same object (up to equivalence),
called topological invariants. Initially, these objects were integers and subse-
quent research reveals that more fruitful and interesting results can be obtained
from the algebraic invariant structures such as groups and rings. For example,
homology and homotopy groups are very important algebraic invariants which
provide strong tools to study the structure of topological spaces.
Contents

1 Prerequisite Concepts of Algebra, Topology, Manifold

and Category Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Some Basic Concepts on Algebraic Structures and their
Homomorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Groups and Fundamental Homomorphism Theorem . . . . 1
1.1.2 Fundamental Theorem of Algebra . . . . . . . . . . . . . . . . . . . . 4
1.1.3 Modules and Exact Sequences . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Some Basic Concepts of Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.1 Homeomorphic Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.2 Connectedness and Locally Connectedness . . . . . . . . . . . . 8
1.2.3 Compactness, Locally Compactness
and Paracompactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.4 Weak Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Partition of Unity and Lebesgue Lemma . . . . . . . . . . . . . . . . . . . . . . 10
1.3.1 Partition of Unity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.2 Lebesgue Lemma and Lebesgue Number . . . . . . . . . . . . . . 11
1.4 Separation Axioms, Urysohn Lemma and Tietze Extension
Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Function Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.6 Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.7 Topological Groups and Lie Groups . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7.1 Topological Groups: Definitions and Examples . . . . . . . . . 15
1.7.2 Actions of Topological groups and Orbit Spaces . . . . . . . . 16
1.7.3 Lie Groups and Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.8 Category, Functor and Natural Transformation . . . . . . . . . . . . . . . . . 19
1.8.1 Introductory Concept of Category . . . . . . . . . . . . . . . . . . . . 19
1.8.2 Introductory Concept of Functors . . . . . . . . . . . . . . . . . . . . 20
1.8.3 Introductory Concept of Natural Transformation . . . . . . . 23
1.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

xv
xvi Contents

2 Homotopy Theory: Fundamental Group and Higher Homotopy

Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.1 Motivation of the Study of Algebraic Topology . . . . . . . . . . . . . . . . 29
2.2 Homotopy: Introductory Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.2.1 Homotopy of Continuous Maps . . . . . . . . . . . . . . . . . . . . . . 31
2.2.2 Homotopy Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Homotopy Classes of Continuous Maps . . . . . . . . . . . . . . . . . . . . . . 36
2.3.1 Homotopy Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.2 Homotopy Classes of Maps to a Topological Group . . . . . 40
2.3.3 Homotopy Between Smooth Maps on Manifolds . . . . . . . 42
2.4 Retraction, Contraction and Deformation Retraction . . . . . . . . . . . . 45
2.4.1 Retraction and Deformation Retraction . . . . . . . . . . . . . . . . 45
2.4.2 Weak Retraction, and Weak and Strong
Deformation Retractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.4.3 Contractible Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.5 Homotopy Extension Property: Retraction and Deformation
Retraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.6 Paths and Homotopy of Paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.6.1 Paths in a Topological Space . . . . . . . . . . . . . . . . . . . . . . . . 55
2.6.2 Homotopy of Paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.6.3 Homotopy of Loops Based at a Point . . . . . . . . . . . . . . . . . 57
2.7 Euler Characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.7.1 Simplexes and Polyhedra . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.7.2 Homotopy and Euler Characteristic of a Finite
Graph . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2.7.3 Euler Characteristic of a Polyhedron . . . . . . . . . . . . . . . . . . 63
2.8 Exponential Maps and Homotopy Classification
of Complex-Valued Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.9 Fundamental Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.9.1 Basic Motivation of Fundamental Group . . . . . . . . . . . . . . 68
2.9.2 Construction of Fundamental Group . . . . . . . . . . . . . . . . . . 69
2.10 Homotopy Property of Connectedness . . . . . . . . . . . . . . . . . . . . . . . . 75
2.11 Functorial Property of π1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
2.12 Link between Fundamental Groups and Retractions . . . . . . . . . . . . 79
2.13 Simply Connectedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
2.13.1 A Characterization of Simply Connected Spaces . . . . . . . 82
2.13.2 Simply Connected Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.14 Fundamental Groups of Product Spaces . . . . . . . . . . . . . . . . . . . . . . . 83
2.15 Fundamental Groups of Hopf Spaces and Topological Groups . . . 84
2.15.1 H-spaces and their Fundamental Groups . . . . . . . . . . . . . . . 84
2.15.2 Fundamental Groups of Topological Groups . . . . . . . . . . . 87
2.16 Alternative Approach to Fundamental Group . . . . . . . . . . . . . . . . . . 88
2.17 Degree and Winding Number of a Loop on the Circle . . . . . . . . . . . 89
2.18 Vector Fields and their Applications . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.18.1 Basic Concepts of Vector Fields . . . . . . . . . . . . . . . . . . . . . . 93
Contents xvii

2.18.2 Vector Fields on S n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

2.18.3 Applications of Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . 97
2.18.4 Link Between Homotopy and Flow . . . . . . . . . . . . . . . . . . . 98
2.18.5 Topological Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
2.19 Computation of the Fundamental Groups . . . . . . . . . . . . . . . . . . . . . 101
2.19.1 Computation of the Fundamental Group
of the Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
2.19.2 Computation of the Fundamental Group
of the Punctured Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
2.19.3 Computation of Fundamental Group of Figure-Eight
F8 and a Wedge of Circles . . . . . . . . . . . . . . . . . . . . . . . . . . 105
2.20 Higher Homotopy Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
2.20.1 Introductory Concepts in Higher Homotopy Groups
πn (X, x0 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
2.20.2 Abelian Group Structure of πn (X, x0 ) for n > 1 . . . . . . . . 111
2.20.3 Functorial Property of πn . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
2.20.4 Role of Base Point x in Groups πn (X, x) for n ≥ 1 . . . . . 114
2.21 Freudenthal Suspension Theorem and Its Applications . . . . . . . . . . 115
2.21.1 Freudenthal Suspension Theorem . . . . . . . . . . . . . . . . . . . . 116
2.21.2 Hurewicz and Hopf Theorems on S n . . . . . . . . . . . . . . . . . . 117
2.21.3 Table of πi (S n ) for 1 ≤ i, n ≤ 8 . . . . . . . . . . . . . . . . . . . . . . 118
2.22 Degree of a Spherical Map on S n from Viewpoint
of Homotopy Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
2.23 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
2.23.1 Fundamental Theorem of Algebra and Algebraic
Completeness of the Field C . . . . . . . . . . . . . . . . . . . . . . . . . 120
2.23.2 Fixed Point and Homotopy: Brouwer Fixed-Point
Theorem for Dimension 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
2.23.3 Borsuk–Ulam Theorem for Dimension 2 . . . . . . . . . . . . . . 123
2.23.4 Lusternik–Schnirelmann Theorem for S 2 . . . . . . . . . . . . . . 125
2.23.5 Cauchy’s Integral Theorem of Complex Analysis . . . . . . . 125
2.23.6 Extension Problem in Homotopy Theory . . . . . . . . . . . . . . 126
2.24 Hopf Classification Theorem by Using Homotopy Theory . . . . . . . 127
2.25 Application of Degree of Spherical Maps on S n and Hopf
Classification Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
2.25.1 Brouwer Degree Theorem and Hopf Classification
Theorem by Using Homotopy Theory . . . . . . . . . . . . . . . . . 133
2.25.2 Brouwer-Poincaré Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 134
2.25.3 Separation of the Euclidean Plane and Jordan
Curve Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
2.25.4 Separation of the Euclidean Plane Problem: Tietze
Extension Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
2.25.5 Applications of the Euler Characteristic
in the Theory of Convex Polyhedra . . . . . . . . . . . . . . . . . . . 137
2.25.6 Interior and Boundary Point Problem of a Surface . . . . . . 138
xviii Contents

2.25.7 Homotopy Property of Infinite Dimensional Sphere

S∞ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
2.25.8 Homotopy Property of Infinite Symmetric Product
Space S P ∞ X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
2.25.9 Brouwer Fixed-Point Theorem for Dimension n . . . . . . . . 144
2.26 More Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
2.27 Invariance of Dimensions of Spheres and Euclidean Spaces . . . . . . 148
2.28 Exercises and Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . 150
2.28.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
2.28.2 Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . 158
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
3 Homology and Cohomology Theories: An Axiomatic Approach
with Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
3.1 Manifolds with Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
3.2 Simplicial Homology Theory and Euler-Poincaré Theorem . . . . . . 166
3.2.1 Simplicial Complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
3.2.2 Simplicial Homology Group . . . . . . . . . . . . . . . . . . . . . . . . . 169
3.2.3 Euler-Poincaré Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
3.2.4 Topological and Homotopy Invariance of Euler
Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
3.3 Topology of CW-Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
3.4 Singular Homology Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
3.4.1 Advantage of Singular Homology over Simplicial
Homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
3.4.2 The Singular Homology with Coefficient Group G . . . . . 180
3.5 Noether’s Algebraic Approach to Homology
and Cohomology Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
3.5.1 Homology Groups of Chain Complexes . . . . . . . . . . . . . . . 181
3.5.2 Cochain Complex and Its Cohomology Groups . . . . . . . . . 186
3.5.3 Cochain Complex and Its Cohomology Groups . . . . . . . . . 186
3.5.4 Singular Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
3.5.5 Functorial Approach to Singular Cohomology
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
3.6 Homology Theory: Eilenberg and Steenrod Axioms . . . . . . . . . . . . 190
3.6.1 Admissible Categories for Eilenberg and Steenrod
Axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
3.6.2 Eilenberg and Steenrod Axioms for Homology
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
3.7 The Uniqueness Theorem for Homology Theory . . . . . . . . . . . . . . . 194
3.8 Cohomology Theory: Eilenberg and Steenrod Axioms . . . . . . . . . . 197
3.9 The Reduced Homology and Cohomology Groups . . . . . . . . . . . . . 199
3.9.1 The Reduced Homology Groups . . . . . . . . . . . . . . . . . . . . . 200
3.9.2 The Reduced Cohomology Groups . . . . . . . . . . . . . . . . . . . 203
3.10 Invariance of Homology and Cohomology Groups . . . . . . . . . . . . . 203
Contents xix

3.10.1 Invariance of Homology Groups . . . . . . . . . . . . . . . . . . . . . 204

3.10.2 Invariance of Cohomology Groups . . . . . . . . . . . . . . . . . . . 205
3.11 Consequences of the Exactness and Excision Axioms
of Eilenberg and Steenrod . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
3.11.1 Consequence of the Exactness Axiom . . . . . . . . . . . . . . . . . 206
3.11.2 Consequence of Excision Axiom . . . . . . . . . . . . . . . . . . . . . 208
3.11.3 Additivity Property of Homology Theory . . . . . . . . . . . . . . 209
3.11.4 Mayer-Vietoris Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
3.12 Applications and Computations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
3.12.1 Computation of Homology Groups of S n . . . . . . . . . . . . . . 211
3.12.2 Suspension Space and Suspension Functor . . . . . . . . . . . . 211
3.13 Degrees of Spherical Maps from the Viewpoint of Homology
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
3.13.1 Brouwer Degree Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
3.13.2 Hopf Classification Theorem by Using Homology
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
3.13.3 More Properties of Degree Function . . . . . . . . . . . . . . . . . . 220
3.14 More Applications of Homology Theory . . . . . . . . . . . . . . . . . . . . . . 223
3.14.1 Brouwer Fixed-Point Theorem for Dimension n . . . . . . . . 223
3.14.2 Hurewicz Homomorphism: Relation Between
Homology and Homotopy Groups . . . . . . . . . . . . . . . . . . . . 225
3.15 The Lefschetz Number and Fixed-Point Theorems . . . . . . . . . . . . . 226
3.16 Exercises and Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . 228
3.16.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
3.16.2 Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . 232
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
4 Topology of Fiber Bundles: General Theory of Bundles . . . . . . . . . . . . 235
4.1 General Properties of Bundles in Topology . . . . . . . . . . . . . . . . . . . . 235
4.1.1 The Concept of Bundles in Topology . . . . . . . . . . . . . . . . . 236
4.1.2 General Properties of Cross Sections of Bundles . . . . . . . . 238
4.1.3 General Properties of Bundle Morphisms . . . . . . . . . . . . . . 240
4.2 Fiber Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
4.2.1 Introductory Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
4.2.2 Fiber Map and Mapping Fiber . . . . . . . . . . . . . . . . . . . . . . . 247
4.3 Vector Bundles over Topological Spaces and Manifolds . . . . . . . . . 248
4.3.1 Vector Bundles over Topological Spaces . . . . . . . . . . . . . . 248
4.3.2 Vector Bundles over Manifolds . . . . . . . . . . . . . . . . . . . . . . 254
4.4 Tangent and Normal Bundles over Manifolds . . . . . . . . . . . . . . . . . . 256
4.4.1 Tangent Bundle of a Smooth Manifold . . . . . . . . . . . . . . . . 256
4.4.2 Tangent Bundle over S n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
4.4.3 Normal Bundle over a Manifold . . . . . . . . . . . . . . . . . . . . . . 259
4.4.4 Normal Bundle over S n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
4.4.5 Orthonormal K-Frames over Spheres . . . . . . . . . . . . . . . . . 261
4.4.6 Canonical Vector Bundle γ n . . . . . . . . . . . . . . . . . . . . . . . . . 262
xx Contents

4.5 Covering Spaces and Covering Homomorphism . . . . . . . . . . . . . . . 262

4.5.1 Covering Spaces and Covering Projections . . . . . . . . . . . . 263
4.5.2 Automorphism Group of Covering Spaces
and Lifting Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
4.6 Construction of Fiber Bundles and Their Local Cross
Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
4.6.1 Construction of Fiber Bundle by Continuous Action . . . . 269
4.6.2 Local Cross Sections of Fiber Bundles . . . . . . . . . . . . . . . . 271
4.7 Hopf Fibering of Spheres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
4.7.1 Hopf Map p : S 3 → S 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
4.7.2 A Generalization of the Hopf Map p : S 3 → S 2 . . . . . . . . 275
4.8 G-bundles and Principal G-bundles . . . . . . . . . . . . . . . . . . . . . . . . . . 276
4.8.1 G-spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
4.8.2 G-Coverings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
4.8.3 G-bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 280
4.8.4 Principal G-bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
4.9 Charts and Transition Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
4.10 Principal G-bundles for Lie Groups G . . . . . . . . . . . . . . . . . . . . . . . . 289
4.11 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
4.11.1 Covering Spaces of R P n . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
4.11.2 Covering Space of Figure-Eight F8 . . . . . . . . . . . . . . . . . . . 293
4.12 More Geometrical Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293
4.13 Exercises and Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . 294
4.13.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295
4.13.2 Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . 300
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
5 Topology of Fiber Bundles: Homotopy Theory of Bundles . . . . . . . . . . 303
5.1 Homotopy Properties of Vector Bundles Over Manifolds . . . . . . . . 303
5.2 Fibrations and Cofibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
5.2.1 Homotopy Lifting Problems . . . . . . . . . . . . . . . . . . . . . . . . . 307
5.2.2 Fibration: Introductory Concepts . . . . . . . . . . . . . . . . . . . . . 309
5.2.3 Cofibration: Introductory Concepts . . . . . . . . . . . . . . . . . . . 313
5.3 Hurewicz Fiberings and Characterization of Fibrations . . . . . . . . . . 316
5.4 Homotopy and Path Lifting Properties of Covering Spaces . . . . . . 318
5.4.1 Introductory Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
5.4.2 Characterization of Path Liftings of Fibrations . . . . . . . . . 322
5.4.3 General Lifting Problems from Homotopy
Viewpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
5.4.4 Homotopy Classification of Covering Spaces . . . . . . . . . . 327
5.4.5 Isomorphism Theorem and Computation
of πn (S 1 ) = 0 for n ≥ 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
5.5 Galois Correspondence on Covering Spaces . . . . . . . . . . . . . . . . . . . 333
5.6 Homotopy Property of Universal Covering Spaces . . . . . . . . . . . . . 337
Contents xxi

5.7 Homotopy Properties of Vector Bundles and Their

Homotopy Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
5.7.1 Homotopy Properties of Vector Bundles . . . . . . . . . . . . . . . 339
5.7.2 Homotopy Classification of Vector Bundles . . . . . . . . . . . . 340
5.8 Homotopy Properties of Numerable Principal G-Bundles . . . . . . . 343
5.9 Classifying Spaces: The Milnor Construction . . . . . . . . . . . . . . . . . . 346
5.10 Applications and Computations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
5.10.1 Application of Galois Correspondence Theorem . . . . . . . . 349
5.10.2 Actions of Fundamental Groups on Fibers
of Covering Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
5.10.3 Fundamental Groups of Orbit Spaces . . . . . . . . . . . . . . . . . 352
5.10.4 Computing the Fundamental Group of R P n . . . . . . . . . . . 354
5.10.5 Computing the Fundamental Group of Klein’s Bottle . . . . 355
5.10.6 Computing the Fundamental Groups of Lens Spaces . . . . 356
5.10.7 Computing the Fundamental Group of Figure-Eight . . . . . 358
5.11 The Relative Homotopy Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
5.11.1 Standard Notations and Construction
of πn (X, A, x0 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
5.11.2 Boundary Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
5.11.3 The Induced Transformation . . . . . . . . . . . . . . . . . . . . . . . . . 363
5.11.4 Algebraic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 364
5.11.5 Functorial Property of the Relative Homotopy
Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
5.11.6 Role of Cells and Spheres in Computing Homotopy
Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
5.12 Isomorphism on πn (X, x0 ) Induced by a Curve and Role
of Base Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
5.13 Homotopy Sequence and Its Basic Properties . . . . . . . . . . . . . . . . . . 371
5.13.1 Homotopy Sequence of Triplets . . . . . . . . . . . . . . . . . . . . . . 371
5.13.2 Elementary Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
5.13.3 Exactness Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
5.13.4 Homomorphism and Isomorphism of Homotopy
Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
5.14 Homotopy Groups of Cells and Spheres . . . . . . . . . . . . . . . . . . . . . . 377
5.15 Stable Homotopy Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
5.15.1 Stable Homotopy Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
5.15.2 Some Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
5.16 Consequences of Homotopy Exact Sequence . . . . . . . . . . . . . . . . . . 382
5.16.1 Direct Sum Theorems in Higher Homotopy Groups . . . . . 382
5.16.2 Characterization of n-Connected Spaces . . . . . . . . . . . . . . 384
5.17 Homotopy Sequence of Fibering and Hopf Fibering
of Spheres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
5.17.1 Homotopy Sequence of Fibering . . . . . . . . . . . . . . . . . . . . . 385
5.17.2 Fiberings of Spheres Over Projective Spaces . . . . . . . . . . . 388
5.17.3 Fiberings of Spheres by Spheres . . . . . . . . . . . . . . . . . . . . . 389
xxii Contents

5.18 More Study on Hopf Map p : S 3 → S 2 . . . . . . . . . . . . . . . . . . . . . . . 390

5.18.1 Problems of Computing πm (S n ) . . . . . . . . . . . . . . . . . . . . . . 392
5.18.2 More Theorems on Hopf Maps . . . . . . . . . . . . . . . . . . . . . . 392
5.18.3 Hopf Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
5.18.4 Hurewicz Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394
5.18.5 Fiber Maps and Induced Fiber Spaces . . . . . . . . . . . . . . . . . 395
5.19 Homotopy Sequence of Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
5.19.1 Fundamental Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
5.19.2 Homotopy Sequence of a Bundle . . . . . . . . . . . . . . . . . . . . . 397
5.20 Bundle Space and Bundle Property . . . . . . . . . . . . . . . . . . . . . . . . . . 398
5.21 Table of πi (S n ) for 1 ≤ i, n ≤ 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
5.22 Action of π1 on πn and n-simplicity . . . . . . . . . . . . . . . . . . . . . . . . . . 400
5.22.1 Automorphism Induced by a Closed Curve . . . . . . . . . . . . 400
5.22.2 Isomorphism of πn Induced by a Curve . . . . . . . . . . . . . . . 400
5.22.3 n-simplicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
5.23 Further Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
5.24 Exercises and Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . 408
5.24.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 408
5.24.2 Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . 417
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
6 Geometric Topology and Further Applications of Algebraic
Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
6.1 Geometric Topology: Embedding Problem of the Circle
in R3 with Knot and Knot Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 421
6.2 Further Applications of Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 426
6.2.1 Isotopy and Its Applications . . . . . . . . . . . . . . . . . . . . . . . . . 427
6.2.2 Application of Topology to Theory of Numbers . . . . . . . . 427
6.3 Borsuk–Ulam Theorem with Applications . . . . . . . . . . . . . . . . . . . . 429
6.3.1 Borsuk–Ulam Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
6.3.2 Ham Sandwich Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
6.3.3 Lusternik–Schnirelmann Theorem for Higher
Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
6.3.4 Van Kampen Theorem: An Application of Graph
Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 432
6.3.5 Proof of Jordan Curve Theorem by Homology
Theory . . . . . . . . . . . .  ............................... 435
6.3.6 Homology Groups of Sin . . . . . . . . . . . . . . . . . . . . . . . . . 436
i∈ A
6.3.7 More Application of Euler Characteristic . . . . . . . . . . . . . . 437
6.3.8 Application of Algebraic Topology to Algebra . . . . . . . . . 437
6.3.9 Whitehead Theorem and Its Applications . . . . . . . . . . . . . . 439
6.3.10 Eilenberg–MacLane Spaces and Their Applications . . . . . 440
6.3.11 Adams Theorem on Vector Field Problem . . . . . . . . . . . . . 443
Contents xxiii

6.4 Exercises and Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . 444

6.4.1 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
6.4.2 Multiple Choice Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . 446
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 448
7 Brief History of Algebraic Topology: Motivation of the Subject
and Historical Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 449
7.1 Motivation of the Study of Topology . . . . . . . . . . . . . . . . . . . . . . . . . 451
7.2 Analysis Situs of Henri Poincaré . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 452
7.2.1 Brief History of Algebraic Topology . . . . . . . . . . . . . . . . . . 453
7.2.2 Motivation of Study of Algebraic Topology . . . . . . . . . . . . 453
7.3 Development of the Basic Topics in Algebraic Topology . . . . . . . . 454
7.3.1 Euler’s Polyhedral Formula . . . . . . . . . . . . . . . . . . . . . . . . . 454
7.3.2 Beginning of Algebraic Topology . . . . . . . . . . . . . . . . . . . . 455
7.4 Historical Note on Homology Theory . . . . . . . . . . . . . . . . . . . . . . . . 456
7.5 Historical Note on Homotopy Theory . . . . . . . . . . . . . . . . . . . . . . . . 457
7.5.1 Inauguration of Homotopy Theory in 1895
by Poincaré . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 457
7.5.2 Brouwer Fixed-Point Theorem and Degree
of Spherical Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
7.5.3 Brouwer–Poincaré Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 459
7.5.4 Freudenthal Suspension Theorem . . . . . . . . . . . . . . . . . . . . 459
7.5.5 Stable Homotopy Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
7.6 The Topology of Fiber Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
7.6.1 Historical Note on Category Theory . . . . . . . . . . . . . . . . . . 463
7.6.2 Homology and Cohomology Theories: Eilenberg
and Steenrod Axioms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 464
7.6.3 Euler–Poincaré Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 465
7.6.4 Topological and Homotopical Invariance of Euler
Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 466
7.6.5 Brief History of Combined Aspect of Combinatorial
and Set-Theoretic Topologies . . . . . . . . . . . . . . . . . . . . . . . . 467
7.6.6 Poincaré Conjecture and Its Solutions . . . . . . . . . . . . . . . . . 467
7.6.7 Interest of Poincaré in Various Scientific Work . . . . . . . . . 467
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468
About the Author

Mahima Ranjan Adhikari, Ph.D., M.Sc. (Gold Medalist), is the founder presi-
dent of the Institute for Mathematics, Bioinformatics, Information Technology and
Computer Science (IMBIC), Kolkata, India. He is a former professor at the Depart-
ment of Pure Mathematics, University of Calcutta, India. His research papers are
published in national and international journals of repute, including the Proceed-
ings of American Mathematical Society. He has authored nine textbooks and is the
editor of two, including: Basic Modern Algebra with Applications (Springer, 2014),
Basic Algebraic Topology and Applications (Springer, 2016), and Mathematical and
Statistical Applications in Life Sciences and Engineering (Springer, 2017).
Twelve students have been awarded Ph.D. degree under his guidance on various
topics such as algebra, algebraic topology, category theory, geometry, analysis, graph
theory, knot theory and history of mathematics. He has visited several universities and
research institutions in India, USA, UK, Japan, China, Greece, Sweden, Switzerland,
Italy, and many other counties on invitation. A member of the American Mathematical
Society, Prof. Adhikari is on the editorial board of several journals of repute. He was
elected as the president of the Mathematical Sciences Section (including Statistics)
of the 95th Indian Science Congress, 2008. He has successfully completed research
projects funded by the Government of India.

xxv
Chapter 1
Prerequisite Concepts of Algebra,
Topology, Manifold and Category Theory

This chapter conveys a few basic concepts of algebra, topology, manifold and cate-
gory theory for smooth study of Volume 3 of the present book series. For detailed
study of the concepts and results given in this chapter, [Adhikari and Adhikari, 2003,
2006, 2014], [Adhikari, 2016], [Dugundji, 1966], [Simmons, 1963], [Adhikari and
Adhikari, Volumes 1 and 2, 2022a, 2022b], [Alexandrov, 1979], [Borisovich et al.
1985], [Bredon, 1983], [Chatterjee et al. 2002], [MacLane, 1971], [Williard and
Stephen, 1970] and some other references are given in Bibliography.

1.1 Some Basic Concepts on Algebraic Structures and their

Homomorphisms

A group, ring, vector space, module or any algebraic system is defined as a nonempty
set endowed with special algebraic structures. A homomorphism (transformation) is
a function preserving the specific structures of the algebraic systems.

1.1.1 Groups and Fundamental Homomorphism Theorem

This subsection presents some basic concepts of group (abstract) theory which are
subsequently used.

Definition 1.1.1 Let G be a group. A subset C(G) of G defined by

C(G) = {g ∈ G : gx = xg for all x ∈ G}

forms a subgroup of G, called the center of G.

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 1
M. R. Adhikari, Basic Topology 3,
https://doi.org/10.1007/978-981-16-6550-9_1
 2 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

Definition 1.1.2 Let H be a subgroup of group G. Then for any x ∈ G,

(i) the set

xH = {xh : h ∈ H }

is called a left coset of H in G and

(ii) the set
Hx = {hx : h ∈ H }

is called a right coset of H in G.

Definition 1.1.3 A subgroup H of a group G is said to be a normal subgroup of

G, if
xH = Hx, ∀ x ∈ G.

Definition 1.1.4 (Quotient group) Let G be group, N be a normal subgroup of G and

G/N be the set of all cosets of N in G. Then G/N is a group under the composition

xN ◦ yN = xyN , ∀ xN and yN in G/N ,

called the factor group (or quotient group) of G by N , denoted by G/N .

Definition 1.1.5 Given an additive abelian group G, and a normal subgroup N of

G, the factor group G/N is defined under the group operation

(x + N ) + (y + N ) = (x + y)N .

This quotient group is sometimes called a difference group.

Definition 1.1.6 Let G and K be two groups. A map f : G → K is said to be a

homomorphism if
f (xy) = f (x)f (y), ∀ x, y ∈ G.

Remark 1.1.7 A homomorphism of groups maps identity element into the identity
element and inverse element into the inverse element. Special homomorphisms carry
special names.

Definition 1.1.8 A homomorphism f : G → K of groups is said to be

(i) a monomorphism if f is injective;

(ii) an embedding if f is a monomorphism;
(iii) an epimorphism if f is surjective;
(iv) an isomorphism if f is bijective;
(v) an isomorphism of G onto itself is called an automorphism of G and
(vi) a homomorphism of G into itself is called an endomorphism.
1.1 Some Basic Concepts on Algebraic Structures and their Homomorphisms 3

Remark 1.1.9 Two isomorphic groups are considered as replicas of each other,
because they have the identical algebraic properties. So two isomorphic groups are
identified and are considered the same group (up to isomorphism).
Definition 1.1.10 The kernel of a homomorphism f : G → K of groups, denoted
by kerf is defined by
kerf = {x ∈ G : f (x) = ek },

where ek is the identity element of K.

Theorem 1.1.11 Let f : G → K be a group homomorphism. Then kerf is a normal
subgroup of G. Conversely, if N is a normal subgroup of G, then the map

π : G → K/N , x → xN

is an epimorphism with N as its kernel.

Corollary 1.1.12 Let N be a subgroup of a given group G. Then N is a normal
subgroup of G if it is the kernel of some homomorphism.
Theorem 1.1.13 (First Isomorphism Theorem) Let f : G → K be a homomorphism
of groups. Then f induces an isomorphism

f˜ : G/kerf → Imf , x kerf → f (x).

Corollary 1.1.14 Let f : G → K be an epimorphism of groups. Then the groups

G/kerf and K are isomorphic.
Example 1.1.15 Let G = GL(n, R) be the general linear group over R, R∗ be the
multiplicative group of nonzero real numbers and

det : G → R∗ , M → det M

be the determinant function and N = {M ∈ GL(n, R) : detM = 1}. Then

(i) N is a normal subgroup of G and
(ii) the groups GL(n, R)/N and R∗ are isomorphic by Corollary 1.1.14.
Definition 1.1.16 (Commutator subgroup) Given a group G and a pair of elements
g, h ∈ G, the commutator of g and h denoted by [g, h] is the element

[g, h] = ghg −1 h−1 .

The subgroup K of G generated by the set S = {ghg −1 h−1 : g, h ∈ G} is called the

commutator subgroup of G, and it consists of all finite products of commutators
of G.
Theorem 1.1.17 Let G be a group and K be the commutator subgroup of K. Then
 4 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

(i) K is a normal subgroup of G;

(ii) the quotient group G/K is always commutative;
(iii) the group G is commutative if its commutator subgroup K is the trivial group.

1.1.2 Fundamental Theorem of Algebra

This subsection states the fundamental theorem of algebra and its proof is given in
Chap. 2 by using the tools of homotopy theory. The completeness of the field C of
complex numbers follows directly from the fundamental theorem of algebra.

Definition 1.1.18 A field F is said to be algebraically closed or complete if every

polynomial ring f (x) of degree n (n ≥ 1) over F, has a root in F.

Example 1.1.19 The field C is algebraically closed. It follows from fundamental

theorem of Algebra 2.23.1.

Theorem 1.1.20 (Fundamental Theorem of Algebra) Every nonconstant polyno-

mial with coefficients in the field C has a root in C.

1.1.3 Modules and Exact Sequences

Definition 1.1.21 A module M over a commutative ring R with nonzero identity

element e is an additive abelian group M together with an external law of composition

m : R × M → V,

the image of (α, x) under m abbreviated αx, such that the following conditions are
satisfied:

(i) ex = x;
(ii) α(x + y) = αx + αy;
(iii) (α + β)x = αx + βx;
(iv) (αβ)x = α(βx)
∀ x, y ∈ M , α, β ∈ R.

Example 1.1.22 (i) Every additive abelian group is a module over the ring (Z).
(ii) Every module over a field F is a vector space.
(iii) Every ideal of a ring R is a module over R.

Remark 1.1.23 The definitions of submodules and quotient modules are analogous
to the definitions corresponding to subgroups and quotient groups.
1.1 Some Basic Concepts on Algebraic Structures and their Homomorphisms 5

Definition 1.1.24 Let M and N be two modules over the same ring R. Then a group
homomorphism f is said to be an R-module homomorphism if

f (x + y) = f (x) + f (y), f (rx) = rf (x), for all x, y in M and r in R.

Definition 1.1.25 A sequence of modules or groups and their homomorphisms

fn fn+1 fn+2
· · · → Mn −−−→ Mn+1 −−−−→ Mn+2 −−−−→ Mn+3 → · · ·

(i) is said to be exact at Mn+1 if Im fn = ker fn+1 and

(ii) is said to be exact if it is exact at every Mn+1 in the sense that

Im fn = ker fn+1 , ∀ n.

Remark 1.1.26 There exist many results that compare groups or modules of an
exact sequence, but only some of them are given which are used in the book. As
every abelian group is a module over Z, every result valid for modules is also true
for abelian groups.

Proposition 1.1.27 Consider an exact sequence of four modules or abelian groups

f g h
{0} −−−→ M −−−→ N −−−→ {0},

where the end module or group {0} is the trivial module or group. Then g is an
isomorphism.

Proposition 1.1.28 Consider an exact sequence of five groups

f g
{0} → M −−−→ N −−−→ P → {0},

where group {0} is the trivial group. If h : P → N is a homomorphism such that g ◦ h

is the identity map on P and N is abelian, then the groups

N∼
= M ⊕ P.

1.2 Some Basic Concepts of Topology

This subsection communicates some basic concepts of topology. For their detailed
study, see Basic Topology, Volume 1.
6 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

1.2.1 Homeomorphic Spaces

This subsection communicates the concept of a homeomorphism in topology which

is a basic concept in topology. This concept is analogous to the concept of an isomor-
phism between algebraic objects such as groups or rings. Every homeomorphism is
a bijective map that preserves topological structure involved. So, every classification
problem in topology involves classification of topological spaces up to homeomor-
phism. Its precise definition is formulated in Definition 1.2.1.

Definition 1.2.1 Let X and Y be two topological spaces. A map

f :X →Y

is said to be continuous if
f −1 (U ) ⊂ X

is an open set in X for every open set U in Y .

Definition 1.2.2 A continuous map f : X → Y between topological spaces X and

Y is said to be a homeomorphism if f is bijective and

f −1 : Y → X

is also continuous.

Remark 1.2.3 If f : X → Y is a homeomorphism, then both the maps f and f −1

are continuous in the sense that
(i) the map f not only sends points of X to points of Y in a (1-1) manner,
(ii) but f also sends open sets of X to open sets of Y in a (1-1) manner.
This implies that X and Y are topologically the same in the sense that a topological
property enjoyed by X is also enjoyed by Y and conversely.

Example 1.2.4 If the map

f :X →Y

is a homeomorphism, then X is compact (or connected) if Y is compact (or con-

nected), because compactness and connectedness properties are topological in the
sense that they are shared by homeomorphic spaces.

Example 1.2.5 Let X and Y be topological spaces. A bijective map

f :X →Y
 1.2 Some Basic Concepts of Topology 7

is not necessarily continuous. For example, let R be the set of real numbers endowed
its usual topology σ and Rl be the same set endowed with the lower topology. Then
the identity map
f : R → Rl , x → x

is a bijection, but it is not continuous.

Example 1.2.6 (i) The open interval (0, 1) ⊂ R and the real line space R with
usual topology are homeomorphic spaces.
(ii) The open ball B = {x = (x1 , x2 ) ∈ R2 : x < 1} ⊂ R2 with Euclidean topology
is homeomorphic to the whole plane R2 .
(iii) The open square A = {(x, y) ∈ R2 : 0 < x, y < 1} ⊂ R2 with Euclidean topol-
ogy is homeomorphic to the open ball B defined in (ii).
(vi) The cone A = {(x, y, z) ∈ R3 : x2 + y2 = z 2 , z > 0} with Euclidean topology is
homeomorphic to the plane R2 .
(v) Consider the n-sphere S n = {x ∈ Rn+1 : x = 1, n ≥ 1}, endowed with
Euclidean topology having its north pole

N = (0, 0, . . . , 1) ∈ Rn+1

and its south pole

S = (0, 0, . . . , −1) ∈ Rn+1 .

(a) The stereographic projection

1
f : S n − N → Rn , x  → (x1 , x2 , . . . , xn ),
1 − xn+1

for every x = (x1 , x2 , . . . , xn+1 ) ∈ S n − N is a homeomorphism.

(b) The space S n − S is homeomorphic to S n − N .
(vi) A circle minus (deleted) any of its point is homeomorphic to a line segment, and
a closed arc is homeomorphic to a closed line segment.

Theorem 1.2.7 Consider the n-cube In = {(x1 , x2 , . . . , xn ) ∈ Rn : 0 ≤ xi ≤ 1}. Its

interior
Int In = {(x1 , x2 , . . . , xn ) ∈ Rn : 0 < xi < 1}

and its boundary ∂In = İn = In − Int In . Then

(i) In is homeomorphic to the n-ball Bn in Rn and
(ii) under this homeomorphism, ∂In = In − Int In corresponds to the (n − 1)-sphere
S n−1 .
 8 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

1.2.2 Connectedness and Locally Connectedness

This subsection conveys the concepts of connectedness and locally connectedness.

Definition 1.2.8 A topological space X is said to be connected if the only sets which
are both open and closed are ∅ and X .

Remark 1.2.9 Connectedness of topological spaces is an important topological

property and is characterized by the following theorem.

Theorem 1.2.10 A topological space X is connected if it is not the union of two

disjoint nonempty open sets.

Definition 1.2.11 Let X be a topological space and x be a point of X . Then X is

said to be locally connected at x, if for every open set V containing x there exists a
connected open set U with x ∈ U ⊂ V . The space X is said to be locally connected
if it is locally connected at x for all x ∈ X .

Example 1.2.12 The Euclidean space Rn is connected and locally connected for all
n ≥ 1.

Remark 1.2.13 The continuous image of a locally connected space may not be
locally connected.

Definition 1.2.14 A path in a topological space X is a continuous map f : I → X

from the closed unit interval I to X .

Definition 1.2.15 A topological space X is said to be path connected, if any two

points of X can be joined by a path.

Remark 1.2.16 A path-connected space is connected. A connected open subset of

a Euclidean space is path connected.

Example 1.2.17 For n > 0, the n-sphere S n is path-connected.

Definition 1.2.18 A topological space X is said to be locally path connected if for

each x ∈ X , and each nbd U of x, there is a path-connected nbd V of x which is
contained in U .

Example 1.2.19 The following spaces in real analysis are connected.

(i) The space R of real numbers;

(ii) Any interval in R;
(iii) Rn ;
(iv) Any ball or cube in Rn ;
(v) The continuous image of a connected space is connected.
1.2 Some Basic Concepts of Topology 9

1.2.3 Compactness, Locally Compactness and

Paracompactness

This subsection conveys the concept of compactness which is used throughout the
book and that of paracompactness which is specially used in the classification of
vector bundles.

Definition 1.2.20 An open covering of a topological space X is a family {Ui } of

open sets of X , whose union is the whole set X .

Definition 1.2.21 A topological space X is said to be compact if every open

covering of X has a finite subcovering.

Remark 1.2.22 This means that from any open covering {Ui } of a compact space X ,
n
we can choose finitely many indices ij , j = 1, 2, . . . , n such that Uij = X . If X
j=1
is a compact space, every sequence of points xn of X has a convergent subsequence,
which means, every subsequence xn1 , xn2 , . . . , xnt , . . . , converges to a point of X .
For metric spaces, this condition is equivalent to compactness.

Proposition 1.2.23 A compact subspace of a Hausdorff topological space X is

closed in X and every closed subspace of a compact space is compact.

Definition 1.2.24 A topological space X is said to be locally compact if each of

its points has a compact neighborhood (nbd).

Example 1.2.25 Any compact space, the space Rn , any discrete space, any closed
subset of a locally compact space are locally compact spaces. On the other hand, the
space Q of rational numbers is not locally compact.

Definition 1.2.26 A topological space X is said to be a Baire space if intersection

of each countable family of open dense sets in X is dense in X .

Example 1.2.27 Every locally compact Hausdorff space is a Baire space.

Definition 1.2.28 A topological space X is said to be compactly generated if X

is a Hausdorff space and each subset A of X satisfying the property that A ∩ C is
closed for every compact subset C of X is itself closed.

Remark 1.2.29 If X and Y are two topological spaces such that X is locally compact
and Y is compactly generated, then their Cartesian product is compactly generated.

Definition 1.2.30 A topological space X is said to be paracompact if every open

covering of X has a locally finite subcovering of X .

Example 1.2.31 (i) Rn is paracompact.

(ii) Every closed subspace of a paracompact space is paracompact; but a subspace
of a paracompact space is not necessarily paracompact.
 10 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

Definition 1.2.32 A topological space X is said to be countably compact if every

countable open covering of X has a finite subcovering.

Theorem 1.2.33 (Cantor’s intersection theorem) A topological space X is countably

compact if every descending chain of nonempty closed sets of X has a nonempty
intersection.

1.2.4 Weak Topology

This subsection communicates the concept of weak topology which is utilized to

construct some important topological spaces in this book.

Definition 1.2.34 Let

X1 ⊂ X2 ⊂ X3 ⊂ · · ·

be a chain of closed inclusions of topological spaces. Then its set union Xi ,
i≥1
which is the union of the sets Xi defines a topology by declaring a subset K ⊂ Xi
i≥1
to be closed if its intersection K ∩ Xi is closed in Xi for all i ≥ 1. This topology is
known as union topology. It is also called weak topology with respect to the
subspaces.
∞
Example 1.2.35 (i) The infinite sphere S ∞ = S n has the weak topology.
∞ n=0
(ii) The infinite projective space RP ∞ = RP n has the weak topology.
n=0 ∞
(iii) The infinite complex projective space CP ∞ = CP n has the weak topol-
n=0
ogy.

1.3 Partition of Unity and Lebesgue Lemma

This section conveys the concept of ‘partition of unity’ and states Lebesgue lemma
with Lebesgue number. A partition of unity subordinate to a given open covering of
a topological space is an important concept in mathematics.

1.3.1 Partition of Unity

Definition 1.3.1 Let U = {Uj : j ∈ A} be an open covering of a topological space

X . A partition of unity subordinate to U consists of a family of functions
1.3 Partition of Unity and Lebesgue Lemma 11

{fj : X → I : j ∈ A}

such that
(i) fj |(X −Uj ) = 0 for all j and each x ∈ X has a nbd V with the property fj |V = 0,
except
 for a finite number of indices j, and
(ii) fj (x) = 1 for all x ∈ X .
j

Remark 1.3.2 The sum fj (x) is always a finite sum.
j

Remark 1.3.3 Paracompactness of a topological space can be characterized with

the help of partition of unity.
Theorem 1.3.4 A topological space X is paracompact if every open covering U of
X admits a partition of unity subordinate to U.
Proof See [Dugundji, 1966]. ❑

1.3.2 Lebesgue Lemma and Lebesgue Number

Lebesgue lemma is used to prove many important results. This lemma is also called
Lebesgue covering lemma.
Lemma 1.3.5 (Lebesgue) Let X be a compact metric space. Given an open covering
{Uα : α ∈ A} of X , there exists a real number δ > 0, called Lebesgue number of
{Uα } having the property that every open ball of radius less than δ lies in some
element of {Uα }.

1.4 Separation Axioms, Urysohn Lemma and Tietze

Extension Theorem

This section imposes certain conditions on the topology to obtain some particular
classes of topological spaces initially used by P.S. Alexandroff (1896–1982) and H.
Hopf (1894–1971). Such spaces are important objects in algebraic topology. More-
over, this section presents Urysohn Lemma and Tietze Extension theorem which are
used in this book.
Definition 1.4.1 A topological space (X , τ ) is said to be a
(i) T1 -space (due to Frechet) if for every pair of distinct points p, q in X , there exist
two open sets U, V such that

p ∈ U, q ∈ V , p ∈
/ V , and q ∈
/ U.
 12 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

In other words, every pair of distinct points is weakly separated in (X , τ ):

equivalently, for every pair of distinct points p, q in X , there exist a neighborhood
of p which does not contain q and a neighborhood of q which does not contain
p.
(ii) Hausdorff space (due to Hausdorff) if any two distinct points are strongly sepa-
rated in (X , τ ):
equivalently, distinct points have disjoint neighborhoods.
(iii) Regular space (due to Vietoris) if any closed set F and any point p ∈ / F are
always strongly separated in (X , τ ).
(iv) Normal space (due to Tietze) if any two disjoint closed sets are strongly sep-
arated in (X , τ ), equivalently, each pair of disjoint closed sets have disjoint
neighborhoods.

Remark 1.4.2 It is not true that a nonconstant real-valued continuous function can
always be defined on a given space. But on normal spaces there always exist noncon-
stant real-valued continuous functions. Urysohn lemma characterizes normal spaces
by real-valued continuous functions.

Lemma 1.4.3 (Urysohn ) A topological space (X , τ ) is normal if and only if every

pair of disjoint closed sets P, Q in (X , τ ) are separated by a continuous real-valued
function f on (X , τ ), such that

0, for all x ∈ P
f (x) =
1, for all x ∈ Q

and
0 ≤ f (x) ≤ 1 for all x ∈ X .

1.5 Function Spaces

This section introduces the concept of function spaces topolozied by the compact
open topology. Function spaces play an important role in topology and geometry.

Definition 1.5.1 (Compact open topology) Let X and Y be topological spaces and
Y X (or F(X , Y )) be the set of all continuous functions f : X → Y . Then a topology,
called compact open topology, can be endowed on F(X , Y ) by taking a subspace for
the topology of all sets of the form

VK,U = {f ∈ F(X , Y ) : f (K) ⊂ U },

where K ⊂ X is compact and U ⊂ Y is open.

Let E : Y X × X → Y , (f , x) → f (x) be the evaluation map. Then given a function
h : Z → Y X , the composite
1.5 Function Spaces 13

h×1d E
ψ : Z × X −−−−−→ Y X × X −−−→ Y

i.e., ψ = E ◦ (h × 1d ) : Z × X → Y is a function.
Theorem 1.5.2 (Theorem of Exponential Correspondence) Let X be a locally com-
pact Hausdorff space and Y , Z be topological spaces. Then a function f : Z → Y X
is continuous if
E ◦ (f × 1d ) : Z × X → Y

is continuous.
Theorem 1.5.3 (Exponential Law) Let X be a locally compact Hausdorff space, Z
be a Hausdorff space, and Y be a topological space. Then the function

ψ : (Y X )Z → Y Z×X

defined by ψ(f ) = E ◦ (f × 1d ) is a homeomorphism.

Proposition 1.5.4 If X is a compact Hausdorff space and Y is metricized by a metric
d , then the space Y X is metricized by the metric d  defined by

d  (f , g) = sup{d (f (x), g(x)) : x ∈ X }.

1.6 Manifolds

This section defines manifolds which form an important class of geometrical objects
in topology. An n-manifold is a Hausdorff topological space which looks locally like
Euclidean n-space Rn , but not necessarily globally. A local Euclidean structure to
manifold by introducing the concept of a chart is utilized to use the conventional
calculus of several variables. Due to linear structure of vector spaces, for many
applications in mathematics and in other areas it needs generalization of metrizable
vector spaces, maintaining only the local structure of the latter. On the other hand,
every manifold can be considered as a (in general nonlinear) subspace of some vector
space. Both aspects are used to approach the theory of manifolds. Since dimension
of a vector space is a locally defined property, a manifold has a dimension.
Definition 1.6.1 An n-dimensional (topological) manifold or an n-manifold M is a
Hausdorff space with a countable basis such that each point of M has a neighborhood
homeomorphic to an open subset of Rn . An one-dimensional manifold is called a
curve, and a two-dimensional manifold is called a surface.
Example 1.6.2 S 2 , torus, RP 2 are examples of surfaces.
Remark 1.6.3 All manifolds M in this book are assumed to be paracompact to
ensure that M is a separable metric space.
14 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

Definition 1.6.4 An n-dimensional differentiable manifold or a smooth manifold

M is a Hausdorff topological space having a countable open covering {U1 , U2 , . . . }
such that

DM(1) for each Ui , there is a homeomorphism ψi : Ui → Vi , where Vi is an open

disk in Rn ;
DM(2) if Ui ∩ Uj = ∅, the homeomorphism ψji = ψj ◦ ψi−1 : ψi (Ui ∩ Uj ) → ψj
(Ui ∩ Uj ) is a differentiable map (or smooth maps) between open subsets
of Rn .

(Ui , ψi ) is called a local chart of M , and {(Ui , ψi )} is a set of local charts of M .

Definition 1.6.5 Let M and N be two smooth manifolds and f : M → N be a map.

A chart (ψ, U ) for M is said to be adapted to f by a chart (φ, V ) for N if f (U ) ⊂ V .
Then the map
φ ◦ f ◦ ψ −1 : ψ(U ) → φ(V )

is well-defined, and it is called the local representation of f at the point x ∈ U

in the given charts. The map f is said to be differentiable (or smooth ) at x if it has
a local representation at x which is differentiable (or smooth). This is well-defined
because a local representation is a map between open sets in Euclidean spaces.

Example 1.6.6 Rn , S n , RP n are n-dimensional differentiable manifolds.

Example 1.6.7 CP n is a 2n-dimensional differentiable manifold.

Definition 1.6.8 A Hausdorff space M is called an n-dimensional manifold with

boundary (n ≥ 1) if each point of M has a neighborhood homeomorphic to the open
set in the subpace of Rn .

Example 1.6.9 The n-dimensional disk Dn is an n-manifold with boundary.

Remark 1.6.10 Let S = {(Ui , ψi )} be a set of local charts of a differentiable man-

ifold M . Then S is said to be a differentiable structure on M . Every subset of S
which satisfies M = ∪ Ui , DM(1) and DM(2) is called a basis for the differential
structure S.

1.7 Topological Groups and Lie Groups

This section recalls the introductory concepts of topological groups and also Lie
groups (studied in Basic Topology: Volume 2) of the present book series. Lie
groups provide special topological groups which are also smooth manifolds satisfy-
ing compatibility conditions laid down in Definition 1.7.14. The detailed study of
topological groups and Lie groups is available in Basic Topology: Volume 2
 1.7 Topological Groups and Lie Groups 15

1.7.1 Topological Groups: Definitions and Examples

This subsection illustrates the concept of topological groups with examples. The
basic concept of a topological group is that it is an abstract group endowed with
a topology such that the multiplication and inverse operations are both continuous.
This concept was accepted by mathematicians in the early 1930s.

Definition 1.7.1 A nonempty set G is said to be a topological group if it satisfies

the following axioms:
TG(1): G is algebraically a group;
TG(2): G is topologically a Hausdorff space; i.e., every pair of distinct points of
G are strongly separated by disjoint open sets;
TG(3): group multiplication m : G × G → G, (x, y) → xy is continuous, where
the topology on G × G is endowed with product topology;
TG(4): group inversion v : G → G, x → x−1 is continuous.
Remark 1.7.2 Some authors do not assume ‘Hausdorff property’ for a topological
group. The conditions TG(3) & TG(4) in Definition 1.7.1 are equivalent to the single
condition
TG(5): the map ψ : G × G → G, (x, y) → xy−1 is continuous.
Example 1.7.3 Rn (under usual addition) and S 1 = {z ∈ C : |z| = 1} (under usual
multiplication of complex numbers) are important examples of topological groups.
Example 1.7.4 (i) The general real linear group GL (n, R) of all invertible n × n
matrices over R is an important topological group, which is neither compact nor
connected.
(ii) GL (n, C) is the set of all n × n nonsingular matrices with complex entries,
called the general complex linear group. GL (n, C) is a topological group with
dimC GL (n, C) = n2 . It is not compact.
(iii) The set U(n, C) = {M ∈ GL (n, C) : MM ∗ = I } forms a subgroup of GL (n, C),
where M ∗ is the transpose of the complex conjugate of M . This subgroup is com-
pact.
(iv) For the sympletic group SU (n, H) = {A ∈ GL (n, H) : AA∗ = I } which is an
quaternionic analogue, see Exercise 1.9 of Sect. 1.9.
Definition 1.7.5 Let G and K be two topological groups. A homomorphism f :
G → K is a continuous map such that f is a group homomorphism in abstract sense.
An isomorphism f : G → K between two topological groups is a homeomorphism
and is also a group homomorphism between G and K.
Example 1.7.6 The special orthogonal group SO(2, R) and the circle group S 1 are
both topological groups. These two topological groups are isomorphic because there
exists an isomorphism of topological groups
 
cos θ − sin θ
ψ : SO(2, R) → S 1 , → eiθ .
sin θ cos θ
16 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

1.7.2 Actions of Topological groups and Orbit Spaces

This subsection communicates the concept of topological group actions on topo-

logical spaces which provide many geometrical objects such as real and complex
projective spaces, torus, lens spaces, etc., as orbit spaces obtained by specifying
actions of topological groups.

Definition 1.7.7 Let G be a topological group with identity element e and X be a

topological space. If G × X has the product topology, then G is said to act on X
from the left if there is continuous map μ : G × X → X , the image μ(g, x), denoted
by g(x) or gx such that

(i) e(x) = x for all x ∈ X ;

(ii) (gk)(x) = g(k(x)), for all x ∈ X , and g, k ∈ G.

The group G is then called a topological transformation group of X relative

to the group action μ. It is sometimes denoted by the triple (G, X , μ). A topological
space X endowed with a left G-action on X is said to be a left G-space. A right
action and a right G-space are defined in an analogous way.

There is a one-to-one correspondence between the left and right G-structures on

X . So it is sufficient to study only one of them according to the situation.

Example 1.7.8 For the general linear group GL(n, R), the Euclidean n-space Rn is
a left GL(n, R)-space under usual multiplication of matrices.

Definition 1.7.9 Let X be a left G-space. Two given elements x, y in X are called
G-equivalent, if y = g(x) (i.e., y = gx) for some g ∈ G. The relation of being G-
equivalent is an equivalence relation on X and the corresponding quotient space
Xmod G endowed with quotient topology induced from X (i.e., the largest topology
such that the projection map p : X → Xmod G, x → G(x) is continuous, where
G(x) = {g(x) : g ∈ G} is called the orbit space of x ∈ G. For an element x ∈ X ,
G(x) is called the orbit of x and the subgroup G x of G defined by G x = {g ∈ G :
g(x) = x} is called the stabilizer or isotropy group at x of the corresponding group
action.

Example 1.7.10 (Geometrical) The actions of a given topological group on the

same topological space may be different. For example, let T be the torus in R3
obtained by rotating the circle C : (x − 3)2 + z 2 = 1 about the z-axis and Z2 =<
h >, generated by h. Consider the three actions

(i)
ψ1 : Z2 × T → T , (h, (x, y, z)) → (x, −y, −z),

which geometrically represents rotation of T through an angle 180◦ about the

x-axis;
 1.7 Topological Groups and Lie Groups 17

(ii)
ψ2 : Z2 × T → T , (h, (x, y, z)) → (−x, −y, z),

which geometrically represents rotation of T through an angle 180o about the

z-axis;
(iii)
ψ3 : Z2 × T → T , (h, (x, y, z)) → (−x, −y, −z)

which geometrically represents reflection of T about the origin.

Each of ψ1 , ψ2 and ψ3 is a homeomorphism of T of order 2 with the sphere,
torus and Klein bottle as the resulting orbit spaces, respectively.
Proposition 1.7.11 Let X be a left G-space and g ∈ G be an arbitrary point. Then
the map
ψg : X → X , x → g(x) (= gx)

is a homeomorphism.
Proof ψg is continuous for every g ∈ G, since the group action is continuous. So,
ψg and ψg−1 are two continuous maps such that ψg ◦ ψg−1 = IX = ψg−1 ◦ ψg , and
hence, it follows that ψg is a homeomorphism. ❑
Remark 1.7.12 Let X be a left G-space and homeo(X) be the set of homeomor-
phisms ψg : X → X for all g ∈ G. Then homeo(X) = {ψg : g ∈ G} is a group under
usual composition of mappings. Proposition 1.7.13 shows that this group is closely
related to the group action of G on X .
Proposition 1.7.13 Let X be a left G-space. Then the map

f : G → homeo(X), g → ψg

is a group homomorphism.
Proof Since for each g ∈ G, ψg is a homeomorphism, the map f is well defined.
Again, since for g, k ∈ G, f (gk) = ψgk = f (g)f (k) asserts that f is a group homo-
morphism. ❑

1.7.3 Lie Groups and Examples

This subsection communicates the concept of Lie groups which are continuous trans-
formation groups. Lie groups occupy a vast territory of topological groups carrying
a differentiable structure and play a key role in the study of topology, geometry and
physics. S. Lie developed his theory of continuous transformations with an eye to
investigate differential equations. The basic ideas of his theory appeared in his paper
(Lie 1880) published in ‘Math. Ann., Vol 16, 1880.’ A Lie group is a topological
group possessing the structure of a smooth manifold on which the group operations
are smooth functions. Its mathematical formulation is given in Definition 1.7.14.
 18 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

Definition 1.7.14 A topological group G with identity element e is said to be a real

Lie group if
(i) G is a real differentiable manifold;
(ii) the group multiplication

m : G × G → G, (x, y) → xy

and the group inversion

v : G → G, x → x−1

are both differentiable.

Definition 1.7.15 A topological group G is said to be a complex Lie group if

(i) G is a complex manifold;
(ii) the group multiplication

m : G × G → G, (x, y) → xy

and the group inversion

v : G → G, x → x−1

are both holomorphic.

Definition 1.7.16 The dimension of a Lie group is defined to be its dimension as

a manifold.

Example 1.7.17 Examples of Lie groups are plenty.

(i) The real line R is a Lie group under usual addition of real numbers.
(ii) R+ = {x ∈ R : x > 0} is a Lie group under usual multiplication of real numbers.
(iii) R2 is a Lie group under pointwise addition given by

(x1 , y1 ) + (x2 , y2 ) = (x1 + x2 , y1 + y2 ).

(iv) Every finite dimensional vector space V over R is a Lie group, since the map

f : V × V → V , (x, y ) → x + y

is linear and hence differentiable. Similarly, the map g : V → V , x → −x is

smooth. In particular, the additive group Rn with its standard structure as a
differentiable manifold is a Lie group.
 1.8 Category, Functor and Natural Transformation 19

1.8 Category, Functor and Natural Transformation

This section conveys the basic concepts of category theory, which provides a
convenient language to unify several mathematical results and unifies many basic
concepts of mathematics in an accessible way. This language born through the work
of S. Eilenberg (1913–1998) and S. MacLane (1900–2005) during 1942–1945 is
used throughout the present book. This section conveys the introductory concepts of
category, functor and natural transformation in the language of category theory.

1.8.1 Introductory Concept of Category

Definition 1.8.1 A category C consists of

(i) a certain family of objects X , Y , Z, . . . usually denoted by ob(C);
(ii) for every ordered pair of objects X , Y ∈ ob(C), a set of morphisms from X to
Y , denoted by [X , Y ] is specified;
(iii) for every ordered triple of objects X , Y , Z ∈ ob(C) and any pair of morphisms
f ∈ [X , Y ], g ∈ [Y , Z], their composition, denoted by g ◦ f ∈ [X , Z] is defined
with the following properties:
(iv) (associativity): if f ∈ [X , Y ], g ∈ [Y , Z], h ∈ [Z, K], then h ◦ (g ◦ f ) = (h ◦
g) ◦ f ∈ [X , K];
(v) (existence of identity morphism): for every object Y ∈ ob(C), there is a mor-
phism 1Y , called identity morphism with the property: for every f ∈ [X , Y ], g ∈
[Y , Z], both the equalities 1Y ◦ f = f and g ◦ 1Y = g hold.

Remark 1.8.2 The identity morphism in a category is unique for each of its object.

Example 1.8.3 (i) The family of all sets and their mappings form a category,
denoted by Set, where objects are all sets, morphisms are all possible mappings
between them and the composition is the usual composition of functions.
(ii) The family of all groups and their homomorphisms form a category, denoted by
Grp, where objects are all groups, morphisms are all possible homomorphisms
between them and the composition is the usual composition of mappings.
(iii) The family of all abelian groups and their homomorphisms form a category,
denoted by Ab, where objects are all abelian groups, morphisms are all possible
homomorphisms between them and the composition is the usual composition of
mappings.
(iv) The family of all vector spaces and their linear transformations form a category,
denoted by Vect.
(v) The family of all topological spaces and their continuous maps form a category,
denoted by T op.
(vi) The family of all pointed topological spaces and their base point preserving
continuous maps form a category, denoted by T op∗ .
 20 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

Remark 1.8.4 The concepts of bijective mappings of sets, isomorphism of groups,

rings or vector spaces, topological spaces and so on can be unified through the concept
of an equivalence in category theory.
Definition 1.8.5 A morphism f ∈ [X , Y ] in a category C is said to be an equivalence
if there is a morphism g ∈ [Y , X ] in the category C such that

f ◦ g = 1Y , g ◦ f = 1X .

If f is an equivalence in C, then g is also an equivalence in C and the objects X and

Y are said to be equivalent.
Example 1.8.6 The equivalences and equivalent objects in the following categories
are specified:
(i) In the category Set, equivalences are bijective mappings and equivalent objects
are precisely the sets having the same cardinality.
(ii) In the category Grp, equivalences are group isomorphisms and equivalent objects
are precisely isomorphic groups.
(iii) In the category T op, equivalences are homeomorphisms and equivalent objects
are precisely homeomorphic topological spaces.
(iv) In the category Ring, equivalences are ring isomorphisms and equivalent objects
are precisely isomorphic rings.

1.8.2 Introductory Concept of Functors

A functor is a natural mapping from one category to the other in the sense that
it preserves the identity morphism and composites of well-defined morphisms. It
plays a key role in converting a problem of one category to the problem of other
category to have a better chance for solution. Important examples of functors from
the topological viewpoint are available in Chaps. 2, 3 and 5.
Definition 1.8.7 Given two categories C1 and C2 , a covariant functor

F : C1 → C2 , X → F(X ), f → F(f )

from category C1 to the category C2 , consists of

(i) an object function which assigns to every object X ∈ C1 an object F(X ) ∈ C2
and
(ii) a morphism function which assigns to every morphism

f ∈ [X , Y ]

in the category C1 , a morphism

 1.8 Category, Functor and Natural Transformation 21

F(f ) ∈ [F(X ), F(Y )]

in the category C2 such that

(iii) F(1X ) = 1F(X ) for every identity morphism 1X ;
(iv) for morphisms f ∈ [X , Y ], g ∈ [Y , Z] in the category C1 , the equality

F(g ◦ f ) = F(g) ◦ F(f )

holds in the category C2 .

A contravariant functor is defined dually and is formulated in Definition 1.8.8.

Definition 1.8.8 Given two categories C1 and C2 , a contravariant functor

F : C1 → C2 , X → F(X ), f → F(f )

from category C1 to the category C2 , consists of

(i) an object function which assigns to every object X ∈ C1 an object F(X ) ∈ C2

and
(ii) a morphism function which assigns to every morphism f ∈ [X , Y ] in the cate-
gory C1 , a morphism
F(f ) ∈ [F(Y ), F(X )]

in the category C2 such that

(iii) F(1X ) = 1F(X ) ;
(iv) for morphisms f ∈ [X , Y ], g ∈ [Y , Z] in the category C1 , the equality

F(g ◦ f ) = F(f ) ◦ F(g)

holds in the category C2 .

Example 1.8.9 (i) There is a covariant functor F : Grp → Set whose object func-
tion assigns to every group G its underlying set |G| and to every group homo-
morphism f : G → K in the category Grp to its corresponding underlying set
function f : |G| → |K| in the category Set. This functor is known as forgetful
functor as it forgets the group structure.
(ii) Let V be an n-dimensional vector space over R and V d = L(V , R) be the
set of all linear transformations T : V → R. Then it is also an n-dimensional
vector space over R, called the dual space of V , denoted by V d . To each
linear transformation T : V1 → V2 , its dual map T d : V2d → V1d is defined
by T d (α)(x) = α(T (x)), ∀ x ∈ V1 , ∀ α ∈ V2d . If V n be the category of all n-
dimensional vector space over R and their linear transformations, then the above
results assert that there is a covariant functor

F : V n → V n , V → V d ∀ objects V ∈ V n , and T → T d ∀ morphisms T ∈ V n

22 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

i.e., the object function maps V to V d and morphism function maps T to T d in

the category V n to itself.
(iii) Given an object A ∈ C, there is a covariant functor

hC : C → C

whose object function assigns to every object X ∈ C to the object hC (X ) =

[C, X ] (the set of all morphisms from the object C to the object X in the category
C) and the morphism function assigns to every morphism f ∈ [X , Y ] in C, the
morphism
hC (f ) : hC (X ) → hC (Y ), g → f ◦ g

in the category C.
(iv) Given an object A ∈ C, there is a contravariant functor

hC : C → C

whose object function assigns to every object X ∈ C to the object hC (X ) =

[X , C] (the set of all morphisms from the object X to the object C in the category
C) and the morphism function assigns to every morphism f ∈ [X , Y ] in C, the
morphism
hC (f ) : hC (Y ) → hC (X ), g → g ◦ f

in the category C.

Proposition 1.8.10 Let C1 and C2 be two categories and F : C1 → C2 be a functor.

If X and Y are two objects equivalent in the category C1 , then the corresponding
objects F(X ) and F(Y ) are also equivalent objects in the category C2 .

Proof Let X and Y be two equivalent objects in the category C1 . Then there exist two
morphisms f ∈ [X , Y ], g ∈ [Y , X ] in the category C1 such that f ◦ g = 1Y and g ◦
f = IX . Then the result follows from the definition of a functor F (may be covariant
or contravariant). ❑

Corollary 1.8.11 Let X and Y be two objects in a category C1 and F be a functor

from the category C1 to the category C2 such that the objects F(X ) and F(Y ) are not
equivalent. Then the objects X and Y are also not equivalent.

Remark 1.8.12 The Corollary 1.8.11 is applied to show that the groups Z5 and Z9
are not isomorphic because the forgetful functor assigns to their underlying sets of
different cardinalities. But the cardinal number of the underlying sets of two groups
may be the same but the groups may not be isomorphic. For example, the groups
S3 and Z6 are not isomorphic but the cardinal number of their underlying sets is the
same.
1.8 Category, Functor and Natural Transformation 23

1.8.3 Introductory Concept of Natural Transformation

The subsection defines natural transformation between two functors. Natural trans-
formation is an important concept needed to compare functors with each other of the
same variance (i.e., when they are either both covariant or both contravariant).

Definition 1.8.13 Let C1 and C2 be two categories and F1 , F2 : C1 → C2 be two

functors of the same variance. A natural transformation

ψ : F1 → F2

is a function which assigns to every object X in the category C1 a morphism ψ(X )

such that for every morphism f ∈ [X , Y ] in C1 , the following equality holds:

ψ(Y ) ◦ F1 (f ) = F2 (f ) ◦ ψ(X )

(if both F1 and F2 are covariant functors)

or
ψ(X ) ◦ F1 (f ) = F2 (f ) ◦ ψ(Y )

(if both F1 and F2 are contravariant functors). In addition, if for each object X in the
category C1 , the natural transformation ψ(X ) is an equivalence in the category C2 ,
then ψ is said to be a natural equivalence.

Example 1.8.14 Given a commutative ring R, let ModR be the category of all R-
modules and their R-homomorphisms. Given a fixed R-module A, there is a covariant
functor πA from the category ModR to itself which assigns to an R-module M the
R-module [A, M ] (the module of all R-module homomorphisms from A to M ) and
for a morphism f : M → N in the category ModR the morphism πA (f ) is defined
by
πA (f ) : [A, M ] → [A, N ] : g → f ◦ g.

Then πA is a covariant functor. Similarly, for each fixed R-module A there is a con-
travariant functor π A defined from the category ModR to itself. Given a morphism
f ∈ [X , Y ] in the category ModR , a natural transformation f ∗ : πY → πX is defined
by
f ∗ (M ) : πY (M ) → πX (M ), h → h ◦ f , ∀ M in ModR .

Remark 1.8.15 For topological applications of category theory, see Chaps. 2, 3, 5

and 6.
24 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

1.9 Exercises

As solving exercises plays an essential role of learning mathematics, various types

of exercises are given in this section. They form an integral part of the book series.
1. Show that the circle group S 1 = {z ∈ C : |z| = 1} in the complex plane is a Lie
group (This group is written as U (1, C) or simply as U (1)).
2. Prove that the general linear group GL (n, H) over the quaternions H is a topo-
logical group but it is not compact.
[Hint : In absence of a determinant function in this case, use the result that
GL (n, H) is an open subset of an Euclidean space.]
3. Show that the special real linear group

SL(n, R) = {X ∈ GL (n, R) : det X = 1}

is a noncompact connected topological group and it is a real Lie group with

dimension n2 − 1.
[ Hint : Use the result that SL(n, R) is a subgroup of GL (n, R). It is a hypersur-
face of GL (n, R). ]
4. Prove that the special complex linear group

SL(n, C) = {X ∈ GL (n, C) : det X = 1}

is a noncompact connected topological group and is a complex Lie group of

dimension n2 − 1.
[Hint : SL(n, C) is a subgroup of GL (n, C). ]
5. Prove that the orthogonal group

O(n, R) = {A ∈ GL (n, R) : AAt = I = At A}

is a compact nonconnected topological group and it is a real Lie group with

dimension n(n−1)
2
.
6. Show that the special orthogonal group

SO(n, R) = O(n, R) ∩ SL(n, R)

is a real compact connected topological group and it is a real Lie group with
dimension n(n−1)
2
.
7. Prove that the general (complex) linear group GL(n, C) is a topological group
and it is a connected, noncompact complex Lie group with dimension n2 .
8. Show that the unitary group

U (n, C) = {A ∈ GL (n, C) : AA∗ = A∗ A = I }

1.9 Exercises 25

is a connected compact topological group, and it is a real Lie group with dimen-
sion n2 , where A∗ denotes the conjugate transpose of A (conjugate means reversal
of all the imaginary components).
[ Hint : It is a subgroup of GL (n, C). It is not a complex submanifold of
GL (n, C). It can be embedded as a subgroup of GL (2n, R) .]
9. Let SU (n, C) denote the special unitary group defined by

SL(n, C) = U (n, C) ∩ SL(n, C).

Show that the group

 
z w
SU (2, C) = {A = : z, w ∈ C and |z|2 + |w|2 = 1}
−w̄ z̄

is isomorphic to S 3 .
[Hint. Use the form of A.]
10. The quaternionic analog of orthogonal unitary groups is the sympletic group

SU (n, H) = {A ∈ GL (n, H) : AA∗ = I },

where A∗ denotes the quaternionic conjugate transpose of A. Prove that it is a

compact topological group.
11. Show that the 3-dimensional projective space RP 3 and SO(3, R) are homeo-
morphic.
12. Let G be a Lie group. Prove the following statements:

(i) the right translation R : G × G → G, (a, g) → Ra (g) = ga is a free and

transitive action and
(ii) left translation L : G × G → G, (a, g) → La (g) = ag is also a free and
transitive action.

13. Let G be a Lie group and M be a smooth manifold. Prove the following state-
ments:

(i) the isotropy group G x of any point x ∈ M is a Lie subgroup of G;

(ii) if G acts freely on M , then the isotropy group G x of any point x ∈ M is
trivial.
14. Show that the orthogonal group O (n + 1, R) acts on RP n transitively from left.
15. Show that orthogonal group O (n, R) acts transitively on the Grassmann manifold
(see Chap. 5) G n,r (r ≤ n).
16. Show that the special orthogonal group SO (n, R) acts transitively on the Stiefel
manifold (see Chap. 5) Vn,r = Vr (Rn ), (r ≤ n).
26 1 Prerequisite Concepts of Algebra, Topology, Manifold and Category Theory

References

Adhikari A, Adhikari MR. Basic topology, vol. 1: metric spaces and general topology. India:
Springer; 2022a.
Adhikari A, Adhikari MR, Basic topology, vol. 2: topological groups, topology of manifolds and
lie groups. India: Springer; 2022b.
Adhikari MR, Adhikari. A. Basic modern algebra with applications. New Delhi, New York, Hei-
delberg: Springer; 2014.
Adhikari MR. Basic algebraic topolgy and its applications. India: Springer; 2016.
Adhikari MR, Adhikari A, Groups, rings and modules with applications. Hyderabad: Universities
Press; 2003.
Adhikari MR, Adhikari A. Textbook of linear algebra: an introduction to modern algebra. New
Delhi: Allied Publishers; 2006.
Alexandrov PS, Introduction to set theory and general topology. Moscow; 1979.
Borisovich YU, Bliznyakov N, Izrailevich YA, Fomenko T. Introduction to topology. Moscow: Mir
Publishers; 1985.
Bredon GE. Topology and geometry. New York: Springer; 1983.
Chatterjee BC, Ganguly S, Adhikari MR. A textbook of topology. New Delhi: Asian Books; 2002.
Dugundji J. Topology. Newton, MA: Allyn & Bacon; 1966.
MacLane S. Categories for the working mathematician. Springer, New York (1971)
Simmons G.Introduction to topology and modern analysis. McGraw Hill York; 1963.
Williard S General topology. Addision- Wesley; 1970.
Chapter 2
Homotopy Theory: Fundamental Group
and Higher Homotopy Groups

This chapter officially inaugurates homotopy theory to begin a study of algebraic

topology by conveying the basic concepts of homotopy and fundamental groups
born through the work of H. Poincaré (1854–1912) in his land-marking ‘Analy-
sis Situs,’ Paris, 1895, and also discusses higher homotopy groups constructed in
1935 by H. Hurewicz (1904–1956) in his paper [Hurewicz, 1935], which are natural
generalizations of fundamental groups. Homotopy theory studies those properties of
topological spaces and continuous maps which are invariants under homotopic maps,
called homotopy invariants. Finally, this chapter presents some interesting applica-
tions of homotopy, fundamental and higher homotopy groups in analysis, geometry,
algebra, matrix theory, atmospheric science, vector field and extension problems and
some others.
Just after the concept of homeomorphisms is clearly defined, the subject of topol-
ogy begins to study those properties of geometric figures which are preserved by
homeomorphisms with an eye to classify topological spaces up to homeomorphism,
which stands as the ultimate problem in topology, where a geometric figure is con-
sidered to be a point set in the Euclidean space Rn . But this undertaking becomes
hopeless when there exists no homeomorphism between the two given topological
spaces. For example, the problem is whether the Euclidean plane R2 and the punc-
tured Euclidean plane R2 − {(0, 0)} are homeomorphic or not. It is difficult to solve
such a problem by the concepts of topological properties such as compactness and
connectedness, studied in Basic Topology, Volume 1 of the present book series.
So, it has become necessary to search for an alternative technique, which is created
in algebraic topology.
A basic problem in topology is to classify continuous maps between topological
spaces up to homeomorphism. This problem is known as classification problems of
topological spaces, and it aims to investigate whether given two topological spaces
are homeomorphic or not. To solve such a problem, either we have to find an explicit
expression of a homeomorphism between them or we have to show that no such
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 27
M. R. Adhikari, Basic Topology 3,
https://doi.org/10.1007/978-981-16-6550-9_2
28 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

homeomorphism exists. To solve the problem of impossibility (when there exists no

such homeomorphism), the problem is shifted in many cases to algebra by associating
invariant objects in the sense that homeomorphic spaces have the same algebraic
object (up to equivalence). These algebraic objects are well known as topological
invariants. They are also called algebraic invariants. Initially, these objects were
integers such as Euler characteristic of a polyhedron. But subsequent research
reveals that more fruitful and interesting results can be obtained from the algebraic
invariant structures such as groups and rings. For example, homology and homotopy
groups are very important algebraic invariants that provide strong tools to study the
structure of topological spaces.
On the other hand, it is a natural problem in homotopy theory to classify con-
tinuous maps between topological spaces up to homotopy: Two continuous maps
from one topological space to other are homotopic if one map can be continuously
deformed into the other map. This classification by an equivalence relation leads to
the concepts of the fundamental groups and the higher homotopy groups, which
are powerful topological invariants to solve many basic problems of topology.
The basic aim of algebraic topology is to devise ways to assign to every topolog-
ical space an algebraic object and to every continuous map from a topological space
to other a homomorphism between the corresponding algebraic objects in a functo-
rial way. This functorial approach facilitates in a variety of cases to solve topological
problems through the possible sovability of corresponding algebraic problems. This
technique defines topological invariants, which are also algebraic invariants. They
convert topological problems into algebraic ones to have a better chance for solu-
tion, which develop another branch of topology, known as algebraic topology. This
branch is one of the most important and powerful creations in mathematics which
uses algebraic tools to study topological spaces. The twentieth century witnessed
its greatest development. For example, in classical mechanics, a natural topology
(Euclidean topology) can be endowed on the configuration space and the phase
space of a system, which provides a qualitative study of the system. A nonvanish-
ing vector field on a nonempty subset X of the Euclidean space Rn and a flow on it
establish a close connection between topology and analysis. This chapter also studies
topological dynamics, which is the study of flows and gives an abstract form of
differential equations.
Historically, B. Riemann (1826–1866) made an extensive work generalizing the
concept of a surface to higher dimensions. He studied a special class of surfaces,
now called Riemann surfaces. While investigating the 3-dimensional and higher-
dimensional manifolds (topological) in 1895, Henri Poincaré in his ‘Analysis
Situs’ formally introduced the concepts of homotopy, fundamental group, homology
groups and Betti numbers. His monumental work embodied in his ‘Analysis Situs’
provided tools to solve problems on system of differential equations, and his research
establishes a surprising connection between analysis and topology (see Topological
Dynamics of Sect. 2.18.5).
Poincaré himself remarked in 1912: ‘Geometers usually distinguish two kinds
of geometry, the first of which they qualify as metric and the second is projective.
· · · But it is a third geometry from which quantity is completely excluded and
2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups 29

which is purely quantitative; this is analysis situs. In this discipline, two figures
are equivalent whenever one can pass from one to the other by a continuous
deformation, whatever else the law of this deformation may be, it must be con-
tinuous. Thus a circle is equivalent to an ellipse or even to an arbitrary closed
curve, but it not equivalent to a straight line segment since the segment is not
closed. A sphere is equivalent to a convex surface; it is not equivalent to a torus
since there is a hole in a torus and in a sphere there is not.’
Poincaré also posed a conjecture in 1904 known as ‘Poincaré conjecture.’ One
form his conjecture says: Is a compact n-manifold homotopically equivalent to S n
homeomorphic to S n ? For n = 3, G. Perelman (1966–) proved this conjecture in
2003 by using Ricci flow. For other values of n, it was solved by others before 1994
(see Chap. 7). Finally, this chapter communicates some interesting applications of
homotopy theory to algebra, matrix theory, atmospheric science, vector field and
extension problems. Algebraic topology also includes homology theory invented by
Poincaré in 1895 and cohomology theory (its dual theory) born thereafter, which are
mainly studied in Chapter 3.
For this chapter, the books [Adhikari, 2016], [Armstrong , 1993], [Hatcher, 2002],
[Hu, 1959], [Massey, 1967], [Patterson, 1959], [Rotman, 1988], [Spanier, 1966],
[Whitehead, 1978] and some others are referred in bibliography.

2.1 Motivation of the Study of Algebraic Topology

This section conveys the motivation of the study of combinatorial or algebraic topol-
ogy. This motivation comes from the study of geometric properties of topological
spaces from the algebraic viewpoint by assigning algebraic objects such as groups
modules and rings to topological spaces and assigning to continuous maps between
topological space homomorphisms between the corresponding algebraic objects in a
functorial way. The concepts of fundamental groups, homology and homotopy groups
were born through this study. More precisely, combinatorial or algebraic topology
was born in the 1890s (at the turn of the nineteenth century to the twentieth century)
through the remarkable work of Henri Poincaré in his ‘Analysis Situs’ dealing with
the theory of integral calculus in higher dimensions and dividing a topological space
into geometric elements corresponding to the vertices, edges and faces of polyhedra,
and their higher-dimensional analogues. His monumental work inaugurated homo-
topy and homology theories with the invention of fundamental group and homology
groups which are basic topological invariants including a generalization of the Euler
characteristic (see Sect. 2.7.3). The development of the methods of this branch of
topology concurred with the set-topology was facilitated by set-theoretic approach
of G. Cantor during 1874–1895 and F. Hausdorff during 1900–1910. Motivation of
construction of fundamental group is available in Sect. 2.9, that of higher homotopy
groups is available in Sect. 2.20, and that of homology groups is available in Chap. 3.
Historically, a systematic study of algebraic topology as a subject began with
precise formulations and correct proofs at the turn of the nineteenth to twentieth
30 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

century (1885–1904) through the work of Henri Poincare in his ‘Analysis Situs,’
Paris, 1895. But his deep insight did not attract mathematicians sufficiently until the
1920s, when the situation began to change with applications in many mathematical
theories. For example, the importance of homotopy invented by Poincaré was first
established by H. Hopf (1895–1971) in 1835 with his discovery of a new continuous
map, now known as Hopf map. The exponential growth of algebraic topology in both
theory and applications has been found since 1940. Algebraic topologists consider
H. Poincaré as founder and H. Hopf as co-founder of algebraic topology.
There is a natural question:
1. What is the main problem of topology?
2. Why we study algebraic topology?

1. Answer: The main problem of topology is to solve classification problems of

topological spaces by using topological properties such as compactness and con-
nectedness properties discussed in Basic Topology, Volume 1 of the present book
series. But the methods applied in general topology to solve the classification prob-
lems are workable at some very restrictive situations. For example, the Euclidean
line R and Euclidean space R3 are not homeomorphic, because deleting any point
from R keeps the remaining space disconnected; on the other hand, deleting any
point from R3 keeps the remaining space connected. But this technique fails to
examine whether higher-dimensional Euclidean spaces such as R100 and R1000
are homeomorphic or not. But algebraic topology provides powerful tools to solve
this problem in a quick way.
2. Answer: A characteristic of a topological space which is shared by homeomorphic
spaces is called a topological invariant in the sense that it is an invariant which
is preserved by a homeomorphism. The main objective in algebraic topology is to
create powerful tools to invade topological problems. They are called topological
invariants and are algebraic in nature. Fundamental groups, higher homotopy
groups, homology and cohomology groups (ring) are central topics of the study in
algebraic topology, and they are utilized in classification of topological spaces up
to homeomorphism. On the other hand, the motivation of combinatorial topology
(former name of algebraic topology) is to study a topological space by representing
it as a union of simple pieces with a specified arrangement, called combinatorial,
such that the properties of the original space depend on how the split pieces
are arranged. As algebraic topology solves many problems of mathematics and
beyond it by utilizing the tools of algebraic topology. This shows the beauty and
scope of the subject (see Chap. 7) of this volume.
Throughout this book, I = [0, 1] denotes the closed unit interval with the topology
induced on I by the natural topology of the real line space R.
2.2 Homotopy: Introductory Concepts 31

2.2 Homotopy: Introductory Concepts

This section starts studying algebraic topology by addressing introductory concepts

of homotopy theory, which form a key part of algebraic topology. The term homo-
topy coined by M. Dehn (1878–1952) and P. Heegaard (1871–1948) in 1907 is now
commonly used. The importance of homotopy lies on the fact that the most of the
topological (algebraic) invariants are also homotopy invariants. A homotopy is a
relation between continuous maps of topological spaces, but not between their sub-
spaces. In the mapping space C(X , Y ) (endowed with compact open topology) of
continuous maps from a topological space X to a topological space Y , a continuous
deformation of one mapping into the other is considered as a path Ht in C(X , Y ),
which starts from f and ends at g. A path in Y is generally studied by the particular
choice of X in C(X , Y ).
Historically, L. E. J. Brouwer (1881–1967) gave the precise definition of contin-
uous deformation by using the concept of homotopy of continuous maps. The aim
of homotopy is to make a qualitative study of geometry in the sense that if one geo-
metric object is continuously deformed into another geometrical object, then these
two geometric objects are considered topologically equivalent, called homeomor-
phic. For example, the geometric objects such as a circle, an ellipse and a square are
topologically equivalent (though their shapes are different). So, one of the aims of
topology is to determine whether two given geometric objects are homeomorphic or
not. To solve such a problem, it is necessary to find an expression of a homeomor-
phism between them or to show that no such homeomorphism exists, specially by
using topological invariants in algebraic topology.

2.2.1 Homotopy of Continuous Maps

This subsection formalizes the intuitive idea of continuous deformation of a contin-

uous map between two topological spaces by introducing the concept of homotopy
with illustrative examples. A homotopy is a family of continuous mappings param-
eterized by the unit interval. Like homeomorphism classes, homotopy classes are
generated by homotopy whose elements are continuous maps.

Definition 2.2.1 Two continuous maps α, β : X → Y from a topological space X

to a topological space Y are said to be homotopic, symbolized α  β, if there exists
a continuous map
H :X ×I→Y

such that H (x, 0) = α(x) and H (x, 1) = β(x), ∀ x ∈ X , where the topology on X ×
I is the product topology. The map H is said to be a homotopy between α and β,
abbreviated H : α  β. In particular, if β is a constant map, then the map α is said
to be nullhomotopic.
32 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Fig. 2.1 Straight-line f1 (x)

homotopy between f1 and f2 f1

f2
f2 (x)

Example 2.2.2 Given two continuous maps f1 , f2 : X → R2 , the map

H : X × I → R2 , (x, t) → (1 − t)f1 (x) + tf2 (x), ∀ x ∈ X , ∀ t ∈ I

is a homotopy from f1 to f2 , called a straight-line homotopy or linear homotopy,

depicted in Fig. 2.1.

Geometrically, the map H shifts the point f1 (x) to the point f2 (x) along the
straight-line segment joining f1 (x) and f2 (x) as shown in Fig. 2.1. So, the map H is
called a straight-line homotopy.

Example 2.2.3 Let f , g : R → R2 , x → (x, x3 ), (x, ex ) be two maps. Then, the

map
H : R × I → R2 , (x, t) → (x, tex + (1 − t)x3 )

is homotopy from f to g.

Example 2.2.4 Let IX , c : Rn → Rn be two maps such that IX : X → X , x → x is

the identity map on X and c : Rn → Rn , x → 0 = (0.0, . . . , 0) ∈ Rn is a constant
map. Then, the map
H : Rn × I → Rn , (x, t) → tx

is a continuous map such that H : c  IX . This shows that the identity map IX is
nullhomotopic and hence Rn is a contractible space with H as a contraction mapping
in the sense of Definition 2.4.9. Again, the maps F and G defined by

F : Rn × I → Rn , (x, t) → (1 − t 2 )x

and
G : Rn × I → Rn , (x, t) → (1 − t)x

are also different homotopies between the maps IX and c which imply that homotopy
between continuous maps is not unique.

Remark 2.2.5 (Geometrical Interpretation of Homotopy) Since Example 2.2.4

shows that homotopy between two continuous maps is not unique, it asserts that there
exist different ways of deforming a continuous map to another continuous map. Two
continuous maps f , g : X → Y are said to be homotopic if f can be continuously
2.2 Homotopy: Introductory Concepts 33

deformed into g by a continuous 1-parameter family of maps Ht : X → Y , given

by Ht (x) = H (x, t) with the property that H0 = f and H1 = g ∀ x ∈ X , ∀ t ∈ I .
Here, Ht forms a continuous family of maps in the sense that H is continuous with
respect to both x and t as a function from the product space X × I to the space Y .
If t is regarded as measuring time, then at t = 0, H0 = f and at t = 1, H1 = g.
Again as t increases from 0 to 1, the map f is continuously deformed into the map g.
In particular, if H is a contraction mapping of the topological space X to the point
x0 ∈ X , then H shrinks the whole space X continuously into the point x0 .

Pasting Lemma 2.2.6 is a key result in proving continuity of a certain class of

maps. For example, it is used to prove Theorem 2.2.7.

Lemma 2.2.6 (Pasting or Gluing Lemma) Let X be a topological space and A, B be

closed subsets in X such that X = A ∪ B. Given a topological space Y , if f1 : A → Y
and f2 : B → Y are continuous maps such that f1 (x) = f2 (x), ∀ x ∈ A ∩ B, then the
map 
f1 (x), if x ∈ A
f : X → Y , x →
f2 (x), if x ∈ B

is continuous.

Proof The map f : X → Y defined in this lemma is a well-defined unique map such
that f |A = f1 and f |B = f2 . We show that f is continuous. Let K be a closed set in Y .
Then, f −1 (K) = (A ∪ B) ∩ f −1 (K) = (A ∩ f −1 (K)) ∪ (B ∩ f −1 (K)) = f1−1 (K) ∪
f2−1 (K). Since each of the maps f1 and f2 is continuous, f1−1 (K) and f2−1 (K) are
both closed in X . This implies that f −1 (K) being the union of two closed sets is
closed in X . This proves that f is continuous. 

Theorem 2.2.7 proves that the relation of being homotopic ‘’ of continuous
maps f , g : X → Y is an equivalence relation on the set C(X , Y ) of all continuous
maps from X to Y .

Theorem 2.2.7 Let X and Y be two topological spaces and C(X , Y ) be the set of
all continuous maps from X to Y . Then, the relation of being homotopic ‘’ is an
equivalence relation on the set C(X , Y ).

Proof (i) ‘’ is reflexive.: A homotopy

F : X × I → Y , (x, t) → f (x) ∀ f ∈ C(X , Y )

shows f  f , ∀ f ∈ C(X , Y ), and hence, the relation ‘’ is reflexive.

(ii) ‘’ is symmetric : If f , g ∈ C(X , Y ) are maps such that f  g, then there exists
a homotopy F : f  g such that

F(x, 0) = f (x), and F(x, 1) = g(x), ∀ x ∈ X .

34 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Consider the map H : X × I → Y , (x, t) → F(x, 1 − t), which is continuous,

since it is the composite of continuous maps

X × I → X × I, (x, t) → (x, 1 − t)

and
X × I → Y , (x, t) → F(x, t).

Then, H : g  f . It shows that the relation ‘’ is symmetric.

(iii) ‘’ is transitive: Finally, if f , g, h ∈ C(X , Y ) are maps such that f  g, g  h,
then there exist homotopies F : f  g and H : g  h. Consider the map

F(x, 2t), 0 ≤ t ≤ 1/2
G : X × I → Y , (x, t) →
H (x, 2t − 1), 1/2 ≤ t ≤ 1.

By using pasting Lemma 2.2.6, it follows that G is continuous. Moreover,

G(x, 0) = F(x, 0) = f (x) and G(x, 1) = H (x, 1) = h(x), ∀ x ∈ X . It implies
that G : f  h and hence the relation ‘’ is transitive.
Consequently, the relation ‘’ is an equivalence relation on the set C(X , Y ). 

Definition 2.2.8 The quotient set C(X , Y )/ , abbreviated, [X , Y ], consists of all

homotopy classes [f ] of maps f ∈ C(X , Y ).

Remark 2.2.9 The set [X , Y ] plays a key role in algebraic topology, and it provides
some algebraic structures on this set by specifying the spaces X or Y . Fundamental
group defined in Sect. 2.16 and higher homotopy groups defined in Sect. 2.20 are
their outstanding examples.

The homotopy category Htp∗ of pointed topological spaces given in Definition

2.2.10 plays a key role in algebraic topology for representation of topological invari-
ants and many concepts in topology in the language of category theory in a unified
way. For example, see functorial property of π1 in Sect. 2.11 and functorial property
of πn in Sect. 2.20.

Definition 2.2.10 Topological spaces and their continuous maps form a category
denoted by T op, and the corresponding homotopy classes of continuous maps form
a category denoted by Htp. This category is called the homotopy category; i.e.,
Htp is the category whose objects are topological spaces and mor (X , Y ) = [X , Y ],
where the composition of maps is consistent with homotopies. In particular, pointed
topological spaces and their base point preserving continuous maps form a category
denoted by T op∗ and the corresponding homotopy classes of their base point pre-
serving continuous maps form a category, called the homotopy category of pointed
spaces, denoted by Htp∗ .
 2.2 Homotopy: Introductory Concepts 35

2.2.2 Homotopy Equivalence

This subsection communicates the concept of homotopy equivalence, which gener-

alizes the concept of homeomorphism (≈) and gives rise to homotopically equivalent
() spaces which are natural generalization of homeomorphic spaces.

Definition 2.2.11 Let X and Y be two topological spaces.

(i) Two continuous maps f : X → Y and g : Y → X are said to be homotopy equiv-
alences if f ◦ g  1Y and g ◦ f  1X , where 1X and 1Y are identity maps on X
and Y , respectively. Each of f and g is called a homotopy equivalence with
homotopy inverse of each other.
(ii) The topological spaces X and Y are said to be homotopy equivalent denoted
by X  Y , if there exists a homotopy equivalence f : X → Y . The relation 
on the set of topological spaces is an equivalence relation. If X  Y , then X and
Y are also called topological spaces in the same homotopy type.
Example 2.2.12 Let Rn be the Euclidean space with its origin 0 = (0, 0, . . . , 0).
Then, Rn  {0}.
Example 2.2.13 Let n be a nonnegative integer and 0 = (0, 0, . . . , 0) be the origin
of the Euclidean (n + 1)-space Rn+1 . Then, the inclusion map

i : S n → Rn+1 − {0}

is a homotopy equivalence by Proposition 2.27.5.

Example 2.2.14 An isomorphism in the homotopy category Htp is a homotopy
equivalence.
Example 2.2.15 (Homeomorphic spaces are homotopy equivalent but its converse
is not true). If two topological spaces X and Y are homotopy equivalent, it is a
natural question: whether they are homeomorphic or not. The answer is negative.
Two homeomorphic spaces are homotopy equivalent by Definition 2.2.11, but its
converse is not necessarily true. In support, consider the examples,
(i) The spaces Rn  {0} but they are not homeomorphic.
(ii) The spaces Dn  {0} but they are not homeomorphic.
(iii) X is the topological space S 1 together with the line segment I0 = {(x, 0) ∈ R2 :
0 ≤ x ≤ 1} obtained by joining the point (1, 0) ∈ S 1 with the point (2, 0), hav-
ing the subspace topology inherited from R2 . Then, X  S 1 but they are not
homeomorphic.
The homeomorphic spaces are of the same topological type. The homotopy equiv-
alent spaces are said to have the same homotopy type. Proposition 2.2.16 asserts that
the homeomorphic spaces have the same homotopy type, but the converse is not true
by Example 2.2.15.
36 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proposition 2.2.16 Let X and Y be two homeomorphic spaces. Then, they have the
same homotopy type.
Proof By hypothesis, X and Y are two homeomorphic spaces. Then, there exists a
homeomorphism f : X → Y . Let g = f −1 : Y → X be the inverse map of f . Then,
g is a continuous map such that

g ◦ f = f −1 ◦ f = 1X  1X and f ◦ g = f ◦ f −1 = 1Y  1Y .

This implies that f is a homotopy equivalence. It proves that X and Y are of the same
homotopy type. 

Proposition 2.2.17 proves that every continuous map homotopic to a homotopy

equivalence is also a homotopy equivalence.

Proposition 2.2.17 Let f : X → Y be a continuous map such that it is homotopic

to a continuous map g : X → Y . If f is a homotopy equivalence, then g is also a
homotopy equivalence.

Proof By hypothesis, f : X → Y is a continuous map such that f  g. Suppose that

f is a homotopy equivalence. Then, there exists a continuous map h : Y → X such
that

h ◦ f  1X and f ◦ h  1Y .

By hypothesis, f  g. Then, it follows that f ◦ h  g ◦ h and hence 1Y  g ◦ h ⇒

g ◦ h  1Y . Similarly, h ◦ g  1X . This proves that g is a homotopy equivalence. 

Definition 2.2.18 A continuous map f ∈ C(X , Y ) is said to be an inessential map

if f  c for some constant map c : X → Y . Otherwise, f is said to be an essential
map. An inessential map is also nullhomotopic.

Example 2.2.19 (i) The identity map f : I → I, t → t is homotopic to the constant

map f : I → I, t → 0. The homotopy

H : I × I → (t, s) → (1 − s)t

asserts that f is inessential.

(ii) The identity map 1S 1 : S 1 → S 1 is an essential map by Corollary 2.19.2.

2.3 Homotopy Classes of Continuous Maps

This section continues the study of the homotopy classes of maps formulated in
Definition 2.2.8 and describes the quotient set [X , Y ] = C(X , Y )/ , which precisely
consists of all homotopy classes [f ] of maps f ∈ C(X , Y ). If we keep the space X
2.3 Homotopy Classes of Continuous Maps 37

fixed and vary Y in the set [X , Y ], then this set is an invariant of the homotopy type
of Y in the sense that if Z is any topological space such that Y  Z, then the sets
[X , Y ] and [X , Z] are equivalent from the set-theoretic viewpoint; i.e., there exists a
bijective map ψ : [X , Y ] → [X , Z]. Analogous result holds for pairs of topological
spaces and also for pointed topological spaces. Many homotopy invariants are found
by specializing the set [X , Y ] by particular choice of the spaces X and Y . This section
works in the category Top and in the category Top∗ .
There are two problems that arise naturally in this section:

1. Corresponding to a given pointed topological space Y , does there exist a natural

product in the set [X , Y ] making the set [X , Y ] a group for every pointed topo-
logical spaces X ? A positive solution of this problem is available in Sect. 2.3.2.
2. Corresponding to a given pointed topological space X , does there exist a nat-
ural product in the set [X , Y ] making the set [X , Y ] a group for every pointed
topological space Y ?

To study these problems, this section works in the homotopy category Htp∗ of
pointed topological spaces and their base point preserving continuous maps. This
means that the set [X , Y ] is the set of morphisms from X to Y in the homotopy cat-
egory Htp ∗ of pointed topological spaces and their homotopy classes. This asserts
the set [X , Y ] depends on the homotopy types of the spaces X and Y .

2.3.1 Homotopy Sets

This subsection studies homotopy sets [X , Y ] and proves some special properties of
such sets.

Definition 2.3.1 (Induced maps) Given two continuous maps f : X → Y and g :

Y → Z, define maps

f ∗ : [Y , Z] → [X , Z], [g] → [g ◦ f ]

and

g∗ : [X , Y ] → [X , Z], [f ] → [g ◦ f ].

The maps f ∗ and g∗ are well defined, since the homotopy of g ◦ f depends precisely
on the homotopy classes of f and g. The maps f ∗ and g∗ are called maps induced by
f and g, respectively.

Theorem 2.3.2 proves properties of induced maps from the viewpoint of homo-
topy.
 38 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Theorem 2.3.2 Given three pointed topological spaces X , Y and Z, if f : Y → Z

is a base point preserving continuous map, then f induces a map

f∗ : [X , Y ] → [X , Z], [α] → [f ◦ α]

such that
(i) If f  h : Y → Z, then f∗ = h∗ ; i.e., homotopic maps induce the same map.
(ii) If 1Y : Y → Y is the identity map, then 1Y ∗ is the identity map; i.e., the identity
map induces the identity map.
(iii) If g : Z → W is another base point preserving continuous map, then (g ◦ f )∗ =
g∗ ◦ f∗ ; i.e., the induced map preserves the composites.

Proof The map

f∗ : [X , Y ] → [X , Z], [α] → [f ◦ α]

is well defined, because f∗ is independent of the choice of the representatives of the

classes in the sense that if α  β, then f ◦ α  f ◦ β.

(i) If f  h : Y → Z, then h∗ ([α]) = [h ◦ α] = [f ◦ α] = f∗ ([α]) ∀ [α] ∈ [X , Y ]

asserts that h∗ = f∗ .
(ii) 1Y ∗ : [X , Y ] → [X , Y ], [α]) → [1Y ◦ α] = [α], ∀ [α] ∈ [X , Y ] asserts that 1Y ∗
is the identity map.
(iii) (g ◦ f )∗ : [X , Y ] → [X , W ], [α] → [(g ◦ f ) ◦ α] = [g ◦ (f ◦ α)] = (g∗ ◦ f∗ )
[α], ∀ [α] ∈ [X , Y ] asserts that (g ◦ f )∗ = g∗ ◦ f∗ . 

Corollary 2.3.3 Let f : Y → Z be a homotopy equivalence. Then, its induced map

f∗ : [X , Y ] → [X , Z]

is bijective for any topological space X .

Proof By hypothesis, f : Y → Z is a homotopy equivalence. Then, there exists a

continuous map g : Z → Y such that g ◦ f  1Y and f ◦ g  1Z . Hence, the corol-
lary follows from Theorem 2.3.2. 

Corollary 2.3.4 For every pair of homotopy equivalent spaces Y and Z, there exists
a bijection
ψ : [X , Y ] → [X , Z]

for any topological space X .

Proof By hypothesis, Y  Z. Then, ∃ a homotopy equivalence f : Y → Z. Hence,

it follows that
f∗ = ψ : [X , Y ] → [X , Z]

is a bijection for any topological space X by Corollary 2.3.3. 

2.3 Homotopy Classes of Continuous Maps 39

Theorem 2.3.5 expresses the above results in the language of the category theory.
Theorem 2.3.5 Let Top∗ be the category of pointed topological spaces, Htp∗ be the
homotopy category of pointed topological spaces and Set be the category of sets and
their maps. Then for every object X ∈ Top∗ , there exists a covariant functor

πX : Htp∗ → Set ,

whose object function assigns to every object Y ∈ Htp∗ the set [X , Y ] ∈ Set and
morphism function assigns to every morphism [f ] ∈ Htp∗ , where f : Y → Z ∈ Top∗ ,
the morphism f∗ ∈ Set defined by

πX [f ] = f∗ : [X , Y ] → [X , Z], [α] → [f ◦ α].

Theorem 2.3.6 formulates the dual result of Theorem 2.3.2.

Theorem 2.3.6 Every base point preserving continuous map f : Y → Z induces a
map
f ∗ : [Z, X ] → [Y , X ], [α] → [α ◦ f ]

for any pointed space X such that

(i) If f  h, then f ∗ = h∗ .
(ii) If 1Y : Y → Y is the identity map, then 1Y ∗ is the identity map.
(iii) If g : Z → W is another base point preserving continuous map, then (g ◦ f )∗ =
f ∗ ◦ g∗.
Proof Similar to the proof of Theorem 2.3.2. 
Corollary 2.3.7 If f : Y → Z is a homotopy equivalence, then its induced map

f ∗ : [Z, X ] → [Y , X ]

is bijective for any pointed topological space X .

Proof By hypothesis, f : Y → Z is a homotopy equivalence. Then, there exists a
continuous map g : Z → Y such that g ◦ f  1Y and f ◦ g  1Z . Hence, the corol-
lary follows from Theorem 2.3.6. 
Corollary 2.3.8 For every pair of homotopy equivalent spaces Y and Z, there exists
a bijection

f ∗ : [Z, X ] → [Y , X ]

for any pointed topological space X .

Proof By hypothesis, Y and Z are homotopy equivalent spaces. Then, there exist
continuous maps f : X → Y and g : Y → X such that g ◦ f  1X , and f ◦ g  1Y .
Hence, the corollary follows from Theorem 2.3.6. 
40 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Theorem 2.3.9 formulates the dual result of Theorem 2.3.5.

Theorem 2.3.9 Let Top∗ be the category of pointed topological spaces, Htp∗ be the
homotopy category of pointed topological spaces and Set be the category of sets and
their maps. Then for every object X ∈ Top∗ , there exists a contravariant functor

π X : Htp∗ → Set .

Proof Proceed as in Theorem 2.3.5. 

Theorem 2.3.10 proves the converses of Corollaries 2.3.3 and 2.3.7.

Theorem 2.3.10 Let f : Y → Z is a base point preserving continuous map.

(i) If f∗ : [X , Y ] → [X , Z] is a bijective map for all pointed topological spaces X ,
then f is a homotopy equivalence.
(ii) If f ∗ : [Z, X ] → [Y , X ] is a bijective map for all pointed topological spaces X ,
then f is a homotopy equivalence.

Proof (i) By hypothesis, for the particular choice of X = Z, it follows that f∗ :

[Z, Y ] → [Z, Z] is a bijection. Hence, there exists a continuous map g : Z → Y
such that f∗ ([g]) = [1Z ]. This shows that f ◦ g  1Z . Similarly, it follows that
g ◦ f  1Y . This proves that f is a homotopy equivalence.
(ii) Proceed as in (i). 

2.3.2 Homotopy Classes of Maps to a Topological Group

This subsection is dedicated to the study of continuous maps from an arbitrary pointed
topological space to a topological group from the viewpoint of homotopy and proves
that given a pointed topological space X and a topological group G, the set [X , G]
admits a group structure. The essential feature which is retained in a topological
group G is a continuous multiplication with a unit. Every topological group G with
identity element e as base point is an object in the category Top∗ .

Definition 2.3.11 A topological group G is a Hausdorff topological space together

with group operations such that
TG(1) Group multiplication m : G × G → G, (x, y) → xy is continuous.
TG(2) Group inversion inv : G → G, x → x−1 is continuous.

Definition 2.3.12 A topological group is said to be commutative if its continuous

multiplication is commutative.

Theorem 2.3.13 Let G be a topological group with identity element e as base point.
Then, the set [X , G] admits a group structure for every topological space X ∈ Top∗ .
2.3 Homotopy Classes of Continuous Maps 41

Proof Let f , g : X → G be any two base point preserving continuous maps. Define
their pointwise multiplication

f · g : X → G, x → f (x)g(x),

where the right-side multiplication is the usual group multiplication μ of the topo-
logical group G. In the language of mapping, the product f · g is the composite map

 f ×g μ
f · g = μ ◦ (f × g) ◦  : X −−−→ X × X −−−−−→ G × G −−−→ G,

where  : X → X × X , x → (x, x) is the diagonal map. Carry this product f · g

over the homotopy classes by the rule [f ] ◦ [g] = [f · g]. This product on [X , G ] is
well defined by Exercise 15 of Sect. 2.28. Hence, the group structure of ([X , G], ◦)
follows from the corresponding properties of the topological group G. 

Corollary 2.3.14 Let G be a topological group. Then, every base point preserving
continuous map f : X → Y induces a group homomorphism f ∗ : [Y , G] → [X , G].

Proof By hypothesis, G is a topological group and f : X → Y is a base point pre-

serving continuous map. Then, [X , G] and [Y , G] are groups by Theorem 2.3.13.
Define
f ∗ : [Y , G] → [X , G], [β] → [β ◦ f ].

Then, f ∗ is well defined and it is a group homomorphism. 

Corollary 2.3.15 If G is a commutative topological group, then the set [X , G]

admits a commutative group structure for every pointed topological space X .

Proof Proceed as in Theorem 2.3.13. 

The above behavior of a topological group G with respect to the abstract group
[X , G] for every pointed topological space X is formulated in Theorem 2.3.16 in the
language of the category theory.

Theorem 2.3.16 Let Htp∗ be the homotopy category of pointed topological spaces
and Grp be the category of abstract groups and their homomorphisms. Then, every
topological group G determines a contravariant functor

π G : Htp∗ → Grp .

Proof The object function assigns to every object X in the category Htp∗ , the group
[X , G], which is an object in the category Grp . This assignment is well defined by
Theorem 2.3.13. The morphism function assigns to every morphism [f ] ∈ Htp∗ of
the homotopy class of the base point preserving continuous map f : X → Y , the
group homomorphism
42 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

f ∗ : [Y , G] → [X , G], [β] → [β ◦ f ],

which is a morphism in the category Grp . This assignment is well defined by Corol-
lary 2.3.14. Hence, by using Theorem 2.3.9, it follows that π G : Htp∗ → Grp is a
contravariant functor. 

Corollary 2.3.17 Let X be a pointed topological space such that X  G for some
topological group G. Then, there is a natural equivalence between the contravariant
functors π X and π G .

Proof By hypothesis, X  G and hence the corollary follows from Theorem 2.3.16.


Remark 2.3.18 Corollary 2.3.17 asserts that the functor π G can be considered as a
functor on the category of groups.

Example 2.3.19 Consider the commutative topological group S 1 = {z ∈ C : |z| =

1} under usual multiplication of complex numbers. Then, [X , S 1 ] is a commu-
tative group and if f : X → Y is a base point preserving continuous map, then
f ∗ : [Y , S 1 ] → [X , S 1 ] is a homomorphism of groups.

Example 2.3.20 Consider the noncommutative topological group S 3 consisting of

the set of unit quaternions

S 3 = {q = x + iy + jz + kw : x, y, z, w ∈ R and x2 + y2 + z 2 + w2 = 1}

under usual multiplication of quaternions of norm 1. Then, [X , S 3 ] is a noncommu-

tative group. Moreover, if f : X → Y is a base point preserving continuous map ,
then f ∗ : [Y , S 3 ] → [X , S 3 ] is group homomorphism.

2.3.3 Homotopy Between Smooth Maps on Manifolds

This section studies homotopy between smooth maps on manifolds defined in Chapter
3 of Basic Topology, Volume 2 of the present series of books. The motivation of
this study comes from the observation that in topology, there are certain properties
of a continuous map which are not changed if the map is deformed in a smooth
manner. Every smooth map is continuous, and the concept of smooth homotopy is
borrowed from the usual concept of homotopy between continuous maps. Definition
2.2.1 formulates the concept of homotopy of continuous maps between topological
spaces. In an analogous way, Definition 2.3.21 formulates the concept of homotopy
of smooth maps between smooth manifolds.
2.3 Homotopy Classes of Continuous Maps 43

Definition 2.3.21 Let M and N be two smooth manifolds and f , g : M → N be two

smooth maps. Then, the smooth maps f and g are said to be smoothly homotopic
if there exists a smooth map

H : M × R → N : H (x, 0) = f (x) and H (x, 1) = g(x), ∀ x ∈ M .

This gives rise to a family of smooth maps

Ht : M → N , x → H (x, t), ∀ t ∈ R.

The smooth map H is called a smooth homotopy or homotopy between smooth

maps f and g. In particular, if the map H is just continuous, the maps f and g are said
to be continuously homotopic or simply homotopic in the sense of in Definition
2.2.1. In general, {Ht } is said to be a smooth family of mappings for some indexing
set A ⊂ Rn if there exists a smooth map H such that

H : X × A → Y , (x, t) → Ht (x).

Corresponding to above H , define a smooth map

H̃ : M × R → N : H̃ (x, t) = f (x), ∀ t ≤ 0 and H̃ (x, t) = g(x), ∀ t ≥ 1.

The map H̃ is called the normalized homotopy corresponding to the smooth homo-
topy H .

Bump function formulated in Definition 2.3.22 is an important smooth function

in differential topology. For example, it is used to study smooth homotopy and
normalized homotopy (see Definition 2.3.21).
Definition 2.3.22 A smooth function B : R → R is said to be bump function if it
satisfies the conditions

0, for all t ≤ 0
B(t) =
1, for all t ≥ 1

and
0 < B(t) < 1 for all t such that 0 < t < 1.

Example 2.3.23 Given a smooth homotopy H formulated in Definition 2.3.21, the

smooth map H defined by

H̃ (x, t) = H (x, B(t))

44 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

is its corresponding normalized homotopy. Moreover, the map

F : M × R × R → N , (x, s, t) → H (x, (1 − t)s + tB(s))

defines a smoothly homotopy map between H and H̃ .

Remark 2.3.24 Geometrically, two smooth maps are smoothly homotopic if one
can be deformed to the other through smooth maps. For a continuous homotopy F
between two continuous maps f , g : M → N , the map F is taken as a continuous
map from M × I → N ; on the other hand, for a smooth homotopy F between smooth
maps f , g : M → N , the map F is taken as a smooth map from M × R → N . The
technical reason is that if the manifold M has a boundary, then the product space
M × I is not a smooth manifold.

Proposition 2.3.25 Let F(M , N ) be the set of all smooth maps f : M → N . Then,
the smooth homotopy H is an equivalence relation on F(M , N ).

Proof Let H be a smooth homotopy relation on F(M , N ).

(i) Then, the relation H on F(M , N ) is reflexive and symmetric by Definition 2.3.21.
(ii) To prove that the relation H on F(M , N ) is transitive, consider any three smooth
maps f , g, h ∈ F(M , N ) such that f is smooth homotopic to g and g is smooth
homotopic to h. Let F be a normalized smooth homotopy between f and g and
G be a normalized smooth homotopy between g and h. Define

F(x, 3t), if t ≤ 1/2
ψ : M × R → N , (x, t) →
G(x, 3t − 2), if t ≥ 1/2.

This implies that the map ψ is well defined and it is a smooth map, because F
and G are smooth maps and ψ(x, t) = g(x), ∀ t ∈ [1/3, 2/3]. This asserts that
ψ is a normalized homotopy between f , h ∈ F(M , N ) and hence it is proved
that the relation H is also transitive.
Consequently, the relation H is an equivalence relation on F(M , N ). 

Definition 2.3.26 Let M and N be two smooth manifolds and f , g : M → N be two

smooth maps. If d is a metric on N and δ : M → R is a positive-valued real function,
then g is said to be a δ-approximation to f , if

d (f (x), g(x)) < δ(x), ∀ x ∈ M .

Proposition 2.3.27 Continuously homotopic smooth maps are also smoothly homo-
topic.

Proof Let f , g : M → N be two smooth maps such that they are continuously homo-
topic and F : M × R → N be a normalized continuous homotopy between them.
If K = (−∞, 0] ∪ [1, ∞), then F is smooth on the closed set M × K, because,
2.3 Homotopy Classes of Continuous Maps 45

for F|M ×(−∞,0] = f and F|M ×[1,∞) = g, there exists a positive continuous function
δ : M × R → R by the smoothing theorem (see Basic Topology, Volume 2) such
that
(i) F can be δ-approximated by a smooth map H : M × R → N .
(ii) F = H |M ×K . 

2.4 Retraction, Contraction and Deformation Retraction

This section addresses the concepts of retraction, contraction, deformation retraction,

weak and strong deformation retractions and studies them by using homotopy theory.

2.4.1 Retraction and Deformation Retraction

This subsection introduces the concepts of retraction and deformation retraction

which are needed for our future study.
Definition 2.4.1 Let X be a topological space and A be a subspace of X , with
inclusion i : A → X .
(i) The subspace A is said to be a retract of X , if there exists a continuous map r :
X → A with the property that r ◦ i = 1A . If such a map r exists, then r : X → A
is said to be a retraction of X to A.
(ii) Additionally, if i ◦ r  1X , then r is called a deformation retraction of X to
A and A is called a deformation retract of X .
Remark 2.4.2 The retraction r formulated in Definition 2.4.1 can be expressed in
Fig. 2.2 making the triangle commutative in the sense that r ◦ i = 1A .
An interesting family of retractions is given in Proposition 2.4.3 and that of defor-
mation retraction is available in Proposition 2.27.6 for n ≥ 1.
Proposition 2.4.3 The n-disk Dn = {x ∈ Rn : ||x|| ≤ 1} is a retract of Rn .
Proof Let i : Dn → Rn be the inclusion map. Then, the continuous map

x, if ||x|| ≤ 1
r : R → D , x →
n n
,
x
||x||
, if ||x|| ≥ 1

Fig. 2.2 Diagram

representing retraction
r:X →A
46 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

is a retraction. This shows that Dn is a retract of Rn . 

Remark 2.4.4 (Geometrical interpretation of retraction): The map r defined in

Proposition 2.4.3 shows that

(i) r fixes every point x ∈ Dn .

(ii) r moves every point x in Rn − Dn along the straight line from the origin to x
onto the boundary ∂ Dn (≈ S n−1 ) of Dn .

Proposition 2.4.5 Let X = Rn − {0} be the punctured Euclidean space and S n−1
be (n − 1)-sphere. Then, S n−1 is a deformation retract of X .

Proof Let i : S n−1 → X be the inclusion map. Define a continuous map

x
r : X → S n−1 , x →
||x||

and another map

tx
H : X × I → X , (x, t) → (1 − t)x + .
||x||

Then, H is a continuous map such that

x
H (x, 0) = x; H (x, 1) = = (i ◦ r)(x), ∀ x ∈ X .
||x||

This shows that H : 1X  i ◦ r and hence i ◦ r  1X . Moreover, r ◦ i = 1S n−1 . Hence,

S n−1 is a deformation retract of X . 

Proposition 2.4.6 The retraction r : Rn → Dn formulated in Proposition 2.4.3 is

also a deformation retraction.

Proof To show this, define a map


(1 − t)x + tx/||x||, if ||x|| ≥ 1
F : R × I → R , (x, t) →
n n
x, if ||x|| < 1

Then, F is a continuous map such that F : 1Rn  i ◦ r. Since r is a retraction, it

proves that r is a deformation retraction.
Geometrically, F fixes every point strictly inside Dn and shifts other points of
n
R linearly from x to r(x) along the straight line joining the points x and origin
(0, 0, . . . , 0) ∈ Rn . 
 2.4 Retraction, Contraction and Deformation Retraction 47

2.4.2 Weak Retraction, and Weak and Strong Deformation

Retractions

This subsection introduces the concepts of weak retraction, and weak and strong
deformation retractions by imposing specific conditions on the inclusion map i :
A → X . These concepts are also used in our future study.

Definition 2.4.7 Let X be a topological space and A be a subspace of X . If i : A → X

is the inclusion map, then

(i) A is said to be a weak retract of X , if there exists a continuous map r : X → A

with the property that r ◦ i  1A . If such r exists, then r : X → A is said to be a
weak retraction of X to A.
(ii) The subspace A ⊂ X is said to be a weak deformation retract of X if the
inclusion map i is a homotopy equivalence.
(iii) The subspace A ⊂ X is said to be a strong deformation retract of X , if there
exists a retraction r : X → A with the property

1X  i ◦ r rel A.

If H : 1X  i ◦ r rel A, then the homotopy H is said to be a strong deformation

retraction of X to A.

Example 2.4.8 (i) Every retraction is a weak retraction, but its converse is not
true. For example, the comb space X ⊂ I2 is a weak retract of I2 but not a retract
of I2 . The comb space X = (I × 0) ∪ (0 × I) ∪ {1/n × I : n = 1, 2, . . . , } is a
subspace of the Euclidean space R2 (see Fig. 2.4.22). It is contractible in the
sense of Definition 2.4.9, but it is not a retract of R2 and the point (0, 1) is not a
strong deformation retract of X .
(ii) The n-sphere S n is strong deformation retract of the punctured Euclidean plane
Rn+1 − {0} with

tx
H : (Rn+1 − {0}) × I → Rn+1 − {0}, (x, t) → (1 − t)x +
||x||

a strong deformation retraction.

2.4.3 Contractible Spaces

This subsection studies contractible spaces which are precisely those topological
spaces that are homotopy equivalent to a one-point space, and hence, all contractible
spaces have the homotopy type of a topological space deformable to a single point.
48 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Definition 2.4.9 A topological space X is called contractible if the identity map

1X : X → X is nullhomotopic, i.e., if there exists a homotopy H : 1X  c for some
constant map c : X → X , x → x0 ∈ X . The homotopy H is then called a contrac-
tion mapping (or a contraction) of the topological space X to the point x0 .
Example 2.4.10 Every convex subspace X of the Euclidean space Rn is contractible.
To show it, consider the continuous map

H : X × I → X , (x, t) → (1 − t)x + tx0 , x, x0 ∈ X , t ∈ I.

If c : X → X , x → x0 is the constant map, then H : 1X  c and hence the space X

is contractible with H a contraction of X to the point x0 ∈ X . Since Rn , Dn , I are
convex subspaces of Rn , it follows by the first part that all of them are contractible
spaces.
Remark 2.4.11 (Geometrical interpretation of contraction) Let X be a contractible
space. Then, there exists a contraction H : 1X  c of the topological space X to the
point x0 ∈ X , which is geometrically interpreted as a continuous deformation of the
space X which shrinks ultimately the whole space X into the point x0 . This asserts
that the space X is contracted to a point of X by H .
Proposition 2.4.12 Let X be any contractible space. Then, it is path connected.
Proof By hypothesis, X is contractible to a point x0 ∈ X . If c : X → X , x → x0 is
a constant map, then there exists a homotopy H : 1X  c. Given any point p ∈ X ,
define a path g in X × I

g : I → X × I, t → (p, t).

Then, β = H ◦ g : I → X is a path in X from the point p to the point x0 . Since p

and x0 are arbitrary points of X , it follows that X is path connected. 
Corollary 2.4.13 Every contractible space is connected.
Proof Since every path-connected space is connected, the corollary follows from
Proposition 2.4.12. 
Proposition 2.4.14 characterizes contractibility of a topological space X in terms
of continuous maps from an arbitrary space to the space X .
Proposition 2.4.14 Let X be a topological space. It is contractible iff given a contin-
uous map f : Y → X from an arbitrary topological space Y , the map f is homotopic
to a constant map c : Y → X , y → x0 .
Proof First suppose that X is contractible. Then, 1X  c. Hence for an arbitrary
continuous map f : Y → X , the maps 1X ◦ f  c ◦ f . Since c ◦ f : Y → X , y → x0
is a constant map, it follows that f is homotopic to a constant map. Conversely,
suppose that given a continuous map f : Y → X from an arbitrary topological space
Y , the map f is homotopic to a constant map c : Y → X , y → x0 . Then taking in
particular, Y = X and f = 1X : X → X , it follows that X is contractible. 
2.4 Retraction, Contraction and Deformation Retraction 49

Corollary 2.4.15 Any two continuous maps from an arbitrary space to a contractible
space are homotopic.

Proof Let X be a contractible space, Y be an arbitrary space and f , g : Y → X

be any two continuous maps. If c : X → X : x → x0 , then by contractibility of X ,
it follows that 1X ◦ f  c ◦ f and 1X ◦ g  c ◦ g. This implies that f = 1X ◦ f 
c ◦ f = c ◦ g  1X ◦ g = g. This proves that f  g. 

Corollary 2.4.16 The identity map 1X : X → X of any contractible space X is

homotopic to any constant map of X to itself.

Proof Let X be a contractible space. Using Corollary 2.4.15, it is proved in particular

that the identity map 1X : X → X is homotopic to any constant map of X to itself.

Theorem 2.4.17 proves that contractible spaces are precisely the topological
spaces which are homotopy equivalent to a one-point space. In other words, this
proves that all contractible spaces have the homotopy type of any topological space
deformable to a single point.

Theorem 2.4.17 Let X be a topological space. Then, X is contractible iff it is of the

same homotopy type of a one-point space P = {p}.

Proof Let P = {p} be a one-point space. First suppose that the spaces X  P; i.e.,
they are of the same homotopy type. Then, there exist continuous maps f : X → P
and g : P → X such that

g ◦ f  1X and f ◦ g  1P .

Suppose that g(p) = x0 ∈ X , c : X → x0 and H : 1X  g ◦ f . Then, c is a constant

map such that g ◦ f = c and hence 1X  c. It proves that X is contractible. Next,
suppose that X is contractible. Then, the identity map 1X : X → X is homotopic
to some constant map c : X → X , x → x0 ∈ X under some homotopy H : 1X  c.
Hence, it follows that c ◦ i = 1P and H : 1X  i ◦ c. This proves that X  P. 

Corollary 2.4.18 proves that two contractible spaces are in the same homotopy
type, and any continuous map between contractible spaces is a homotopy equiva-
lence.
Corollary 2.4.18 Let X and Y be any two contractible spaces. Then, X and Y
are in the same homotopy type and any continuous map f : X → Y is a homotopy
equivalence.

Proof Let P is a one-point space. By hypothesis, X and Y are two contractible spaces.
Then, X  P and Y  P by Theorem 2.4.17 and hence it follows by symmetry and
transitivity of the relation  that X  Y . This proves the first part of the corollary.
For the second part, suppose that X  Y . Then, there exists a homotopy equivalence
f : X → Y . If g : X → Y is an arbitrary continuous map, then f  g by Corollary
2.4.15. 
50 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Definition 2.4.19 Let X be a topological space and c : X → X , x → p ∈ X be a

constant map. It is called contractible to the point p relative to the subset P = {p}
if there exists a homotopy

H : X × I → X such that H : 1X  c rel P.

Theorem 2.4.20 Let X be a topological space contractible to a point p ∈ X relative

to the one-point space P = {p}. Then for every nbd U of p in X , there is a nbd V ⊂ U
of the point p with the property that every point in V can be joined to the point p by
a path lying entirely in U.

Proof The theorem is proved by using the compactness property of I. If c : X →

X , x → p, then under the given condition, there exists a homotopy H : 1X  c rel P.
This implies that the line {p} × I is mapped by H to the point p ∈ X . If U is a nbd of
p in X , then from the continuity of the homotopy H , it follows that there exist nbd
Vt (p) of p ∈ X for each t ∈ I and a nbd Wt of t in I such that

H (Vt (p) × Wt ) ⊂ U.

By using the compactness of I, it follows that the open covering

{Wt : t ∈ I}

of I has a finite subcovering

Wt1 , Wt2 , . . . , Wtn

such that
H (Vti (p) × Wti ) ⊂ U, ∀ i = 1, 2, . . . , n.
n
This asserts that V (p) = Vt (p) is a nbd of p in X such that H (V (p) × I) ⊂
i=1 i
U . For x ∈ V (p), the image H (V (p) × I) ⊂ U, and hence it follows that the point x
can be joined to the point p by a path which lies entirely in U. 

Example 2.4.21 The comb space is a subspace X of the Euclidean plane R2 defined
by
X = (I × 0) ∪ (0 × I) ∪ {1/n × I : n = 1, 2, . . . , }.

Geometrically, the comb space X consists of horizontal line segment L joining the
point (0, 0) to the point (1, 0) and the vertical closed unit line segments standing on
the points (1/n, 0) for every n ∈ N together with the vertical line segment joining
the points (0, 0) with the point (0, 1) as shown in Fig. 2.3. It is an important example
of a contractible space by Proposition 2.4.22.

Proposition 2.4.22 Let X be the comb space. Then, it is contractible but it is not
contractible relative to the point {(0, 1)}.
2.4 Retraction, Contraction and Deformation Retraction 51

Fig. 2.3 Comb space X (0, 1)

L
(0, 0) (1/2, 0) (1, 0)

Proof X is contractible: Let L be the horizontal line segment joining the point
(0, 0) to the point (1, 0) with i : L → X be the inclusion map and p : X →
L, (x, y) → (x, 0) be the projection map. Then, p ◦ i = 1L (identity map on L).
Define the map

H : X × I → X , ((x, y), t) → (x, (1 − t)y).

Then, H : 1X  i ◦ p and p : X → L is a homotopy equivalence and hence X  L.

Since L ≈ I (homeomorphic), it follows that L  I. Since the space I is contractible
space, it is of the same homotopy type of a one-point space and hence the spaces I
and L are of the same homotopy type of one-point space. This implies that the space
X is of the same homotopy type of one-point space. This proves that the comb space
X is contractible by Theorem 2.4.17.
X is not contractible relative to the point {(0, 1)}: To show it, take any small
nbd U of the point (0, 1) ∈ X that contains an infinite number of path components
of X . Let D21 be the open disk with center (0, 1) and radius 21 . Consider the nbd
V = D21 ∩ X of the point (0, 1) in X . Then, V has no nbd of the type U such that
each of its points can be joined to the point (0, 1) by a path lying entirely in V .
This proves that the comb space X is not contractible relative to the point {(0, 1)} by
Theorem 2.4.20. 

Example 2.4.23 The concept of relative homotopy is stronger than that of homotopy
in the sense that if A be a subspace of X and f , g : X → Y are two continuous maps
such that f  g rel A. Then f  g. But its converse is not true. For example, consider
the comb space X , its identity map 1X : X → X , x → x and the constant map c :
X → X , (x, y) → (0, 1). Since X is contractible, the maps 1X  c by Corollary
2.4.16. But by Proposition 2.4.22, the comb space X is not contractible relative to
{(0, 1)}.

2.5 Homotopy Extension Property: Retraction

and Deformation Retraction

Homotopy extension property (HEP) is an important concept in topology to solve

many problems in homotopy theory. For example, it is used in Chapter 5 to study
cofibration of a continuous map. In this chapter, it is proved in Theorem 2.5.5 that
52 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Fig. 2.4 Homotopy

extension property of (X,A)
w.r.t Y

the concepts of weak retraction and retraction coincide under HEP. It is also proved
in Theorem 2.5.2 that HEP is a property in the homotopy category.

Definition 2.5.1 A pair (X , A) of topological spaces is said to have the homotopy

extension property (HEP) with respect to a topological space Y if for any two
continuous maps h : X → Y and H : A × I → Y such that H (x, 0) = h(x), ∀ x ∈ A,
there exists a continuous map

G : X × I → Y : G(x, 0) = h(x), x ∈ X and G|A×I = H .

In the mapping diagram, it is represented in Fig. 2.4, where h0 : X → X × I, x →

(x, 0) and the dotted arrow represents the map G.

Theorem 2.5.2 asserts that the homotopy extension problem is a problem in the
homotopy category.

Theorem 2.5.2 Let the pair (X , A) of spaces have the homotopy extension property
with respect to the space Y and f , g : A → Y be two homotopic maps. Then, f has
a continuous extension over X iff g has also a continuous extension over X .

Proof By hypothesis, f  g : A → Y . Then, there exists a homotopy

H : A × I → Y : H (x, 0) = f (x) and H (x, 1) = g(x), ∀ x ∈ A.

First suppose that f : A → Y has a continuous extension h : X → Y . Then,

H (x, 0) = f (x) = h(x), ∀ x ∈ A. As (X , A) has the HEP with respect to Y , there
exists a continuous map

G : X × I → Y : G|A×I = H

and hence the diagram in Fig. 2.4 is commutative. This asserts the existence of the
map G. Define a map
k : X → Y , x → G(x, 1).

Then, k is a continuous extension of g over X , because k(a) = G(a, 1) = H (a, 1) =

g(a), ∀ a ∈ A. Next suppose that g : A → Y has a continuous extension k : X → Y .
Then proceeding as above, it is also proved that f has also a continuous extension
h : X → Y. 
2.5 Homotopy Extension Property: Retraction and Deformation Retraction 53

Remark 2.5.3 Theorem 2.5.2 asserts that the homotopy extension property is a prop-
erty in the homotopy category in the sense that a continuous map f : A → Y can or
cannot be extended over X is a property depending on the homotopy class of that
map f .

Remark 2.5.4 Example 2.4.8 shows that the concepts of weak retraction and retrac-
tion are different. Does there exist conditions under which these two concepts coin-
cide? Theorem 2.5.5 gives its positive answer by providing suitable conditions under
which the concepts of weak retraction and retraction coincide.

Theorem 2.5.5 Let the pair (X , A) of spaces have the homotopy extension property
with respect to the subspace A. Then, A is a weak retract of X iff A is a retract of X .

Proof By hypothesis, the pair (X , A) of spaces have the homotopy extension property
with respect to the subspace A and i : A → X is the inclusion map. First suppose
that A ⊂ X is a retract of X and r : X → A is a retraction. Then, r ◦ i = 1A implies
that r ◦ i  1A . This proves that A is a weak retract of X . Conversely, suppose that
r : X → A is a weak retraction. Then, r ◦ i  1A asserts that there exists a homotopy

H : A × I → A : H (x, 0) = r(x) and H (x, 1) = 1A (x) = x.

Again as (X , A) has the HEP with respect to A, there exists a continuous map G : X ×
I → A extending H : A × I → A. Hence, G(x, 0) = r(x), ∀ x ∈ X and G|A×I = H .
Define a map
r̃ : X → A, x → G(x, 1).

Then, A is a retract of X with retraction r̃.



Definition 2.5.6 Let X be a topological space and A be a subspace of X .

(i) A homotopy D : A × I → X is said to be a deformation of A in X , if D(x, 0) =

x, ∀ x ∈ A.
(ii) Additionally, if D(x, 1) ∈ Y ⊂ A, ∀ x ∈ A, then D is said to be deformation of
A into Y and A is said to be deformable in X into Y .
(iii) The inclusion map i : A ⊂ X is said to have a right homotopy inverse f : X →
A, if there exists a continuous map

D : X × I → X such that D : 1X  i ◦ f .

Theorem 2.5.7 characterizes deformability of a topological space in its subspace

in terms of right homotopy inverse of the inclusion map.

Theorem 2.5.7 Let X be a topological space and A be a subspace of X . Then, X is

deformable into A iff the inclusion map i : A → X has a right homotopy inverse.
54 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proof By hypothesis, X is a topological space and A be a subspace of X with the

inclusion map i : A → X . First suppose that X is deformable into the subspace A of
X . Hence, there exists a continuous map

D : X × I → X : D(x, 0) = x and D(x, 1) ∈ A ⊂ X , ∀ x ∈ X .

Define
f : X → A : (i ◦ f )(x) = D(x, 1), ∀ x ∈ X .

This implies that D : 1X  i ◦ f and hence it is proved that i has a right homotopy
inverse.
Next suppose that i : A ⊂ X has a right homotopy inverse f : X → A. Then, 1X 
i ◦ f . Hence, there exists a continuous map

D : X × I → X such that D : 1X  i ◦ f .

This implies that

D(x, 0) = 1X (x) = x and D(x, 1) = (i ◦ f )(x) ∈ A ⊂ X , ∀ x ∈ X .

This proves that X is deformable into A. 

Remark 2.5.8 (Geometrical interpretation of retraction D) The map D : X × I →
X formulated in Definition 2.5.6 starts with the identity map 1X : X → X , then
moves every point of X continuously, including the points of A and finally, pushes
each point into a point of A. If, in particular, a topological space X is deformable
into a point x0 ∈ X , then X is contractible, and conversely, if X is contractible then
it is deformable into one of its points.
Proposition 2.5.9 The pair (Dn , S n−1 ) has the homotopy extension property.
Proof Let X be a given space and H : S n−1 × I → X and f : Dn → X be any two
continuous maps such that H (x, 0) = f (x), ∀ x ∈ S n−1 . Consider the map

f (x/(1 − t/2)), if ||x|| ≤ 1 − t/2
G : D × I → X : (x, t) →
n
H (x/||x||, 2(||x|| − 1 + t/2)), if ||x|| ≥ 1 − t/2.

Since X is an arbitrary topological space, it follows that (Dn , S n−1 ) has the homo-
topy extension property. 

2.6 Paths and Homotopy of Paths

This section studies the concepts of paths and homotopy of paths in arbitrary topo-
logical spaces.
2.6 Paths and Homotopy of Paths 55

2.6.1 Paths in a Topological Space

This subsection defines a path in a topological space with illustrative examples.

Definition 2.6.1 Given a topological space X , a continuous map α : I → X such

that α(0) = x0 ∈ X and α(1) = x1 ∈ X is said to be a path in X from the point x0
to the point x1 . The points x0 and x1 are called the initial and terminal points of
the path α in X , respectively.

Example 2.6.2 Given two points x0 , x1 ∈ R2 , the map α : I → R2 , t → (1 − t)x0 +

tx1 is a path in R2 from x0 to x1 . On the other hand, the map β : I → R2 , t →
tx0 + (1 − t)x1 is a path in R2 from x1 to x0 .

Example 2.6.3 Let (X , τ ) be a topological space and x0 ∈ X . Then, a constant map

c : I → X , x → x0 is said to be a constant path or null path in X at x0 .

2.6.2 Homotopy of Paths

This subsection considers homotopy of a special class of paths α : I → X in a topo-

logical space X .

Definition 2.6.4 Let X be a topological space and α, β : I → X be two paths in X .

Then, they are said to be path homotopic if they have the same initial point and
the same terminal point and there exists a continuous map H : I × I → X with the
conditions

H (t, 0) = α(t), H (t, 1) = β(t), ∀ t ∈ I

and
H (0, s) = x0 , H (1, s) = x1 , ∀ s ∈ I.

The map H is called a path homotopy between the paths α and β as shown in
Fig. 2.5, and it is represented by

H : α  β.
p

Remark 2.6.5 In Definition 2.6.4

(i) The first condition asserts that H is a homotopy between α and β.
(ii) The second condition asserts that for each t ∈ I, the path t → H (t, s) is a path
in X from x0 to x1 .
56 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Fig. 2.5 Path homotopy X

between α and β
x1

β
x0

Geometrically, the map

H :αβ
p

provides a continuous way of deforming the path α to the path β keeping the end
points of the paths remained fixed during the deformation as shown in Fig. 2.5.
Theorem 2.6.6 Let X be a topological space and P(X ) be the set of all paths in X
with the same initial point x0 and the same terminal point x1 . Then, the path homotopy
relation ‘’ on P(X ) is an equivalence relation.
p

Proof Let α1 , α2 , α3 ∈ P(X ) be paths in X such that α1 (0) = α2 (0) = α3 (0) = x0

and α1 (1) = α2 (1) = α3 (1) = x1 .

(i) Reflexivity: Let α ∈ P(X ) be an arbitrary path. Define a map

H : I × I → X , (t, s) → α(t).

Since H = α ◦ p1 is the composite of the projection map p1 : I × I → I, (t, s) →

t onto the first factor and the map α : I → X , which is continuous, it follows
that the map H is also continuous. Moreover,

H (t, 0) = α(t), H (t, 1) = α(t), ∀ t ∈ I, H (0, s) = x0 , H (1, s) = x1 , ∀ s ∈ I.

This implies that H : α  α, ∀ α ∈ P(X ). This shows that the path homotopy
p
relation is reflexive.
(ii) Symmetry: Let α1  α2 and H : α1  α2 . Define a map
p p

F : I × I → X , (t, s) → H (t, 1 − s).

Then, F is a continuous map such that

F(t, 0) = H (t, 1) = α2 (t), and F(t, 1) = H (t, 0) = α1 (t), ∀ t ∈ I

and
F(0, s) = H (0, 1 − s) = x0 , F(1, s) = H (1, 1 − s) = x1 .
2.6 Paths and Homotopy of Paths 57

Fig. 2.6 Path (linear) α(x)

homotopy H : α p β α

β
β(x)

Hence, F : α2  α1 . This asserts that this relation is symmetric.

p
(iii) Transitivity: Let α1  α2 , f2  α3 . Using the homotopies G : α1  α2 ,
p p p
and H : α2  α3 and pasting lemma, construct a homotopy F : α1  α3 (see
p p
Theorem 2.2.7 ).

This proves that the path homotopy relation ‘’ on P(X ) is an equivalence
p
relation. 

Definition 2.6.7 Given a topological space X , the quotient set obtained by the path
homotopy relation on P(X ) denoted by P(X )/ is called the set of path homotopy
p
classes of paths in X .

Example 2.6.8 Let α, β : X → R2 be two continuous maps. Define a linear (con-

tinuous ) map

H : X × I →, R2 , (x, t) → (1 − t)α(x) + tβ(x).

Then, H : α  β.
Geometrically, H shifts the point α(x) to the point β(x) along the straight-line
segment joining α(x) and β(x), as shown in Fig. 2.6. The map H is called a linear
homotopy.

2.6.3 Homotopy of Loops Based at a Point

This subsection studies a particular class of paths, called loops. More precisely,
given a topological space X , an arbitrary chosen point x0 ∈ X , called a base point,
the homotopy classes of loops in X based at x0 are studied in this subsection, which
leads to the concept of the fundamental group of X based at the point x0 .

Definition 2.6.9 A path α : I → X is called a loop in X based at x0 ∈ X if α(0) =

α(1) = x0 . If İ = {0, 1}, then it is a subspace of I. A loop α in X based at x0 is a
continuous map α : (I, İ) → (X , x0 ). In particular, the constant map δ : I → X , t →
x0 , ∀ t ∈ I, is called a constant loop or a null loop in X at x0 .
58 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Definition 2.6.10 Let α, β : (I, İ ) → (X , x0 ) be two loops in X based at the point

x0 . Then, α and β are said to be homotopic relative to the subspace İ = {0, 1} of
I denoted by α  β rel İ, if there exists ( ∃ ) a continuous map H : I × I → X such
that

H (t, 0) = α(t), ∀ t ∈ I, H (t, 1) = β(t), ∀ t ∈ I, and H (0, s) = H (1, s) = x0 , ∀ s ∈ I.

Remark 2.6.11 Geometrically, the continuous map H given in Definition 2.6.10
sends the square I × I into X so that the bottom of the square is mapped by α, the
top of the square is mapped by β and vertical sides of the square are mapped onto
the point x0 ∈ X . It shows intuitively that if t represents time, then the path α is
continuously deformed to β throughout the unit interval.
Theorem 2.6.12 Given a topological space (X , τ ), let L(X , x0 ) be the set of all
loops α : (I, İ ) → (X , x0 ) in X based at the point x0 ∈ X . Then, the relation α 
β rel İ on L(X , x0 ) is an equivalence relation.
Proof It follows from Theorem 2.6.6. 
Remark 2.6.13 The quotient set L(X , x0 )/  of homotopy classes of loops relative
to İ = {0, 1}, denoted by π1 (X , x0 ), admits a group structure, called the funda-
mental group of the pointed topological space (X , x0 ) (see Theorem 2.9.10). It is
a very important topological invariant and is studied in Sect. 2.9.

2.7 Euler Characteristic

This section considers geometric objects such as points, lines and surfaces in R3
through Euler characteristic, which is an important topological invariant (integral),
but it is different from the powerful topological properties such as compactness and
connectedness studied in Chap. 5, Basic Topology, Volume 1 of the present series
of books. Euler characteristic invented by L. Euler (1703–1783) in 1752, an integral
invariant, is the first topological invariant which distinguishes nonhomeomorphic
spaces. The search of other topological invariants led to invention of fundamen-
tal group (see Sect. 2.9) and homology group (see Chap. 3) by Henri Poincaré in
1895. If the X and Y are two homotopy equivalent spaces, then the characteristics
κ(X ) = κ(Y ). Moreover, κ(X ) ∈ Z, which is an algebraic object. Euler characteristic
establishes a relation between geometry and algebra through topology. A general-
ization of Euler characteristic through homology theory is available in Chap. 3.

2.7.1 Simplexes and Polyhedra

This subsection develops some machinery to compute Euler characteristic and funda-
mental group of a vast family of topological spaces by using the familiar concepts of
2.7 Euler Characteristic 59

simplexes and polyhedra. It is widely used in homology theory developed in Chapter

3. Historically, algebraic topology was born through the work of H. Poincaré based
on the idea of dividing a topological space into geometric elements corresponding to
the vertices, edges and faces of polyhedra, and their higher-dimensional analogues.
Such investigation presents many topological invariants including the Euler charac-
teristic.

Example 2.7.1 Some finite-dimensional simplexes. A zero-dimensional simplex

is a point; a one-dimensional simplex is a straight-line segment. A two-dimensional
simplex is a triangle (including the plane region which it bounds), and a three-
dimensional simplex is a tetrahedron. Finite-dimensional simplexes are used in the
development of homology theory (see Chap. 3).

Definition 2.7.2 Any (m + 1) of the (n + 1) vertices of a given n-dimensional sim-

plex sn determine an m-dimensional simplex, called an m-dimensional face of the
simplex sn for every m such that 0 ≤ m ≤ n.

Example 2.7.3 The 0-dimensional simplexes of a simplex are its vertices.

Definition 2.7.4 Let σp be a p-dimensional simplex. If f is a simplex such that the

vertices of f form a subset of the vertices of σp , then f is said to be a face of σp . A
face f of σp is said to be proper if f is neither ∅ nor the whole of σp .

Definition 2.7.5 A k-dimensional polyhedron P in Rn is a point set in Rn endowed

with the subspace topology inherited from the Euclidean topology on Rn , which can
be decomposed into simplexes of dimensions less than or equal to k, but there is at
least one k-simplex such that the simplexes satisfy the property
(i) Two simplexes of this decomposition have either a common face as their inter-
section
(ii) Or no point in common.
The collection of all of the simplexes which belong to simplicial decomposition of
a polyhedron is called a geometric complex. A polyhedron is a geometric object
such that the boundary of its two faces is an edge and its two edges meet at a vertex.

Example 2.7.6 A tetrahedron is polyhedron having 4 faces, 6 sides (edges) and 4

vertices.

Definition 2.7.7 Let X be a subspace of R3 such that it is homeomorphic to a

polyhedron P. The Euler characteristic of X , denoted by κ(X ), is given by

κ(X ) = V − E + F,

where
V denotes the number of vertices in P,
E denotes the number of edges in P and
F denotes the number of faces in P.
 60 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Example 2.7.8 (i) If X is one-point, then κ(X ) = 1.

(ii) If X is a line segment, then κ(X ) = 1.
(iii) If X a triangle (i.e., made of 3 edges of a triangle), then κ(X ) = 0.
(iv) If X is a triangular region, called triangular lamina, which includes the plane
region the triangle bounds (i.e., made of 3 vertices, 3 edges and one face of a
triangle), then κ(X ) = 1.
(v) If X is the circle S 1 , then κ(X ) = 0, since X is homeomorphic to the sides of a
triangle.

Example 2.7.9 (i) If X is the tetrahedron, then κ(X ) = 2.

(ii) If X is the cube, then κ(X ) = 2.
(iii) If X is the 2-sphere S 2 , then κ(X ) = 2, since X is homeomorphic to the surface
of a tetrahedron.

Example 2.7.10 (i) Let R be the real line space. If X is a one-point set and Y is a
line segment with topology induced from R, then κ(X ) = 1 = κ(Y ), but they
are not homeomorphic (though they are homotopy equivalent spaces).
(ii) The circle S 1 and the sphere S 2 are not homeomorphic, since κ(S 1 ) = 0 but
κ(S 2 ) = 2.

Remark 2.7.11 Two homeomorphic compact polyhedra have the same Euler char-
acteristic; i.e., if X and Y are two compact polyhedra, which are homeomorphic, then
their Euler characteristics κ(X ) = κ(Y ). Its proof follows from Euler characteristic
theorem for simplicial homology (see Chapter 3).

2.7.2 Homotopy and Euler Characteristic of a Finite Graph

This subsection studies finite graphs from the viewpoint of homotopy theory and
their Euler characteristics. The ‘Seven Bridge Problem of Königsberg’ posed by
Euler in 1752 initiated the concept of a new geometry, now called topology, without
the concept of distance, which may be considered as the starting point of graph the-
ory. Several problems of combinatorial nature may be solved by converting them in
the language of graph theory. The ‘Seven Bridge Problem of Königsberg’ is an out-
standing example. The diagrams commonly used by electrical engineers are practical
examples of graphs. Jordan curve Theorem 2.25.29 has wide applications in graph
theory (see Chap. 6).
The precise definition of a graph from the topological viewpoint is given in Def-
inition 2.7.12.

Definition 2.7.12 A (topological) graph G is a topological space having a collec-

tion of points, called vertices of G together with a collection of edges such that every
edge is either homeomorphic to the closed interval I = [0, 1] with subspace topol-
ogy inherited from the real line space R with usual topology and joins two distinct
vertices of G or homeomorphic to a circle S 1 and joins a given vertex of G to itself.
2.7 Euler Characteristic 61

Fig. 2.7 Graph G 1 is a tree

G1

Fig. 2.8 Graph G 2 is not a

tree

G2

Remark 2.7.13 If G is a geometric graph, then its underlying space |G| is a topologi-
cal space satisfying the conditions of Definition 2.7.12. On the other hand, Definition
2.7.12 has an alternative form in the language of simplicial complex given in Defi-
nition 2.7.14. Detailed description of a simplicial complex is available in Chapter 3.

Definition 2.7.14 A simplicial complex of dimension ≤ 2 is called a graph. A graph

which does not contain any loop is called a tree. An end of a graph is the vertex of
at most one-simplex.

Example 2.7.15 The graph G 1 shown in Fig. 2.7 is a tree. On the other hand, the
graph G 2 shown in Fig. 2.8 is not a tree.

Definition 2.7.16 A graph X is said to be contractible if its underlying space |X |

is contractible.

Theorem 2.7.17 Let X be any tree. Then, it is contractible.

Proof Case I: Let X be a finite tree. The theorem is proved by induction on the
number of vertices of X . If X has only one vertex or exactly one edge, then the the-
orem holds trivially. Assume that the theorem is true for all trees with n vertices. If
the tree X has n+1 vertices with vn+1 an end vertex, then there is a unique 1-simplex
s ∈ X with vn+1 as its vertex. If S = X − {vn+1 , (s)}, then S is a simplicial complex
such that its underlying space |S| = |X | − (s) ∪ vn+1 . Hence, S is also a tree, since
if s1 is 1-simplex in S such that |S| − (s1 ) is connected, then |X | − (s1 ) is to be
connected. Since S has only n-vertices, by induction hypothesis, S is contractible.
Consider the maps f : |X | → |S| such that f maps (s) ∪ vn+1 into the other vertex
of s and S onto itself. Let g : |S| → |X | be the inclusion map. Then, g ◦ f  1d and
f ◦ g  1d assert that |S| and |X | are of the same homotopy type. This implies that
|X | is contractible.

Case II : Let X be an arbitrary tree. Select a vertex v0 ∈ X . Now construct a homotopy

H : |X | × I → X , such that H (x, 0) = x, H (x, 1) = v0 and H (v0 , t) = v0 , ∀ x ∈ |X |, ∀ t ∈ I.

62 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

The construction of H is done in several successive steps. The homotopy H implies

any arbitrary tree X is also contractible. 

Definition 2.7.18 Let G be a finite graph with V vertices and E edges (number of
1-simplexes), then the Euler characteristic κ(G) is defined to be the integer

κ(G) = V − E.

The integer κ(G) of any graph G is invariant under subdivisions, since inserting
an additional vertex into G splits some 1-simplex into two 1-simplexes, and hence
this process enhances both the numbers V and E by one keeping the number κ(G) =
V − E, unchanged.

Proposition 2.7.19 For any tree G, its Euler characteristic κ(G) = 1.

Proof The proposition is proved by induction on the number n = V (the number of

vertices of G). For n = 1, the proposition is trivial. Assume that the proposition is
true for trees with n vertices. If the tree G has n+1 vertices with vn+1 an end vertex of
G, then there is a unique 1-simplex (s) with vn+1 as its vertex. If S = G − {vn+1 , (s)},
then S is a simplicial complex such that |S| = |G| − (s) ∪ vn+1 . S is also a tree, since
if s1 is an 1-simplex in S such that |S| − (s1 ) is connected, then |G| − (s1 ) is to be
connected. Since S has only n-vertices, by induction hypothesis, κ(S) = 1. But the
number of vertices of G = number of vertices of S plus 1 and also the number of
edges of G = number of edges of S plus 1. This implies that κ(G) = κ(S) = 1. 

Definition 2.7.20 A topological space X is said to be arcwise connected or pathwise

connected if any two points of X can be joined by a path in X .

Definition 2.7.21 Let G be an arcwise connected graph. If m is the maximum num-

ber of open 1-simplexes that can be deleted from G without disconnecting the space,
then the number m is said to be the number of basic circuits in G.

Theorem 2.7.22 Let G be an arcwise connected graph. If m is the number of basic

circuits in G, then m = 1 − κ(G).

Proof Case I : If G is a tree, then m = 0 and hence the theorem follows from Theorem
2.7.19.
Case II : If G is not a tree, let (s1 ) be an open 1-simplex such that X1 = |G| − (s1 )
is connected. If X1 is a tree, then the theorem follows from Case I. Again if X1 is not
a tree, let (s2 ) be an open 1-simplex such that X2 = |G| − (s1 ) ∪ (s2 ) is connected.
Continue the process, which will stop after only a finite number of steps, since in G,
there are only finite number of 1-simplexes. Hence, there exists some m ∈ N such
that

Xm = |G| − (s1 ) ∪ (s2 ) ∪ · · · ∪ (sm )

2.7 Euler Characteristic 63

is a tree. Hence, it follows that

κ(G) = κ(Xm ) − m = 1 − m =⇒ m = 1 − κ(G). 

Remark 2.7.23 For the fundamental group of a connected graph which is not a tree,
see Exercise 22 of Sect. 2.28.

2.7.3 Euler Characteristic of a Polyhedron

This subsection studies Euler characteristic of a polyhedron in R3 , which is a topo-

logical (integral) invariant. Euler’s theorem formulated in Corollary 2.7.26 is con-
sidered the first theorem on polyhedra which conveys the geometric properties of a
polyhedron without using the concept of distance. Euler characteristic is an integral
invariant. As Z is an algebraic object, the concept of Euler characteristic establishes
an interplay between geometry and algebra, giving the birth of combinatorial topol-
ogy and algebraic topology. Euler sent a letter to C. Goldbach (1690–1764) in 1750
giving his formula for a connected graph G on a 2-dimensional sphere S 2 :

V (number of vertices of G) - E (number of edges of G) + F (number of regions

of the sphere divided by the graph G) = 2, notationally, it is expressed as

V − E + F = 2.

Theorem 2.7.24 Let S be a closed surface divided into F regions, called faces with
the help of E arcs, joining in pairs, V vertices with the property that at least two arcs
meet at a vertex. Then, the integer

V−E+F

is independent of the choice of dividing up the surface S.

Proof Let T1 and T2 be two given divisions of the surface S into faces, arcs and
vertices. Construct a third division T3 containing all the regions, arcs and vertices
of both divisions T1 and T2 . This construction is possible, by adding new vertices at
points of intersection of the arcs of T1 and T2 . This process gives rise to new arcs
and new faces. To prove the theorem, it is sufficient to show that for any division
T obtained from a given division of S by adding new vertices, arcs and faces, the
integral value of V − E + F remains the same. To show it, let one new vertex be
inserted and this new vertex be joined to some of the existing vertices by m new arcs.
By this process, the number of faces is increased by m − 1, since one of the original
faces is replaced by m new faces. This asserts that the integral value of V − E + F
is unaltered by this process of construction. This value remains also unaltered if
new arcs are added without adding new vertices, since each new arc gives one new
64 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

face. This asserts that the integral value of V − E + F remains the same for any
division obtained from the existing division by adding new vertices, arcs and faces.
It concludes that V − E + F is the same for all divisions of the given surface S. 

Definition 2.7.25 The integral value of V − E + F for any closed surface S is well
defined and is called the Euler characteristic of S, abbreviated by κ(S).

Corollary 2.7.26 (Euler theorem on polyhedron) Let V be the number of vertices,

E the number of edges and F the number of faces of a polyhedron P homeomorphic
to the 2-sphere S 2 . Then its Euler characteristic

κ(P) = V − E + F = 2.

Proof It follows from Theorem 2.7.24 by using κ(S 2 ) = 2. 

Remark 2.7.27 Euler characteristic of a topological space X is independent of a

polyhedron P as long as X is homeomorphic to P by Poincaré–Alexander theorem;
see Exercise 2.28.1 of Sect. 2.28. Euler characteristic of a compact surface is a
topological invariant by Exercise 48 of Sect. 2.28. Thus, if X and Y are two compact
homeomorphic surfaces in R3 , then κ(X ) = κ(Y ). This implies that if κ(X ) = κ(Y ),
then X and Y cannot be homeomorphic. But its converse is not always true in the sense
that κ(X ) = κ(Y ) does not guarantee that the spaces X and Y are homeomorphic.
In support, see Example 2.7.30.

Example 2.7.28 (i) If S is the torus T , then κ(S) = 0, since for the torus S,

V = 1, E = 2, F = 1.

(ii) If S is the real projective plane RP 2 , then κ(S) = 1.

Proposition 2.7.29 S 1 is not homeomorphic to S 2 .

Proof Since κ(S 1 ) = 0 and κ(S 2 ) = 2, they cannot be homeomorphic, since Euler
characteristic of a compact surface is a topological invariant. 

Example 2.7.30 Two nonhomeomorphic spaces may have the same Euler charac-
teristic. For example, if X = [a, b] and Y = {x0 }, a one-point space in R, then
κ(X ) = 1 = κ({x0 }). They are not homeomorphic but they are homotopy equiva-
lent.

Remark 2.7.31 Example 2.7.30 motivates to study Euler characteristic from the
viewpoint of algebraic topology. More precisely, a study on Euler characteristic
such as Euler–Poincaré theorem, the topological and homotopy invariance of the
Euler characteristic of compact polyhedra by using simplicial homology groups of a
compact polyhedra is available in Chapter 3. But a study of Euler characteristic for
a surface is available in Exercises 44–48 of Sect. 2.28.
2.8 Exponential Maps and Homotopy Classification of Complex-Valued Functions 65

2.8 Exponential Maps and Homotopy Classification

of Complex-Valued Functions

This section studies exponential map and index function from the viewpoint of homo-
topy. More precisely, the homotopy classification of complex-valued continuous
functions from compact metric spaces is given in Theorem 2.8.4. In particular, the
homotopy of complex-valued continuous maps from S 1 is characterized with the help
of index numbers (see Exercise 53 of Sect. 2.28 ). For the study of continuous maps
from a topological space to the punctured complex plane C∗ = C − {0} from the
viewpoint of homotopy theory, the concept of exponential map is necessary, which
is given in Definition 2.8.1.

Definition 2.8.1 Given any topological space X , a continuous map f : X → C −

{0} is said to be exponential if the map f can be expressed as f = ek for some map
k : X → C.

Example 2.8.2 (i) Every continuous function f : X → R+ is exponential, since f

can be expressed as k = log f (usual logarithm function) for some map k : X →
C.
(ii) Let f : X → C − {0} be a continuous map such that f (X ) ∩ (−∞, 0) = ∅; i.e.,
f escapes the part of negative real axis. Then, f is exponential.

Proposition 2.8.3 Let G = C(X , C − {0}) be the group of all continuous maps from
X to C − {0} under pointwise multiplication

(f g)(x) = f (x)g(x)

and

H = {f ∈ G : f is exponential}.

Then, H is a subgroup of G.

Proof It follows from Definition 2.8.1. 

Theorem 2.8.4 characterizes homotopy classes of continuous maps from a com-

pact metric space to the space C∗ = C − {0} by exponential maps.

Theorem 2.8.4 Let X be a compact metric space and f , g : X → C − {0} be two

continuous maps. Then, f  g iff the map f /g is exponential.

Proof Let the map f /g be exponential. Then, there exists some map k : X → C
such that f /g = ek . Define a homotopy

H : f  g : X × I → C − {0}, (x, t) → g(x)e(1−t)k(x) .

66 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Conversely, let H : f  g : X × I → C − {0}. Then, H (x, 0) = f (x) and H (x, 1) =

g(x) for all x ∈ X . Since X is compact by hypothesis, and I is compact, their product
space X × I is also compact. Then, the continuous positive-valued function |H | :
X × I → R+ attains its minimum value m and hence

0 < m = inf {|H (x, t)| : x ∈ X , t ∈ I }.

Since H is uniformly continuous on X × I, there exists a real number δ > 0, such

that whenever |t − s| < δ, then

|H (x, t) − H (x, s)| < m, ∀ x ∈ X .

For an integer n > 1/δ, consider the maps

fm : X → C − {0}, x → H (x, m/n), ∀ x ∈ X , 1 ≤ m ≤ n.

Since each of the maps fm−1 /fm is exponential, H (x, 0) = f (x) = f0 (x), say, and
H (x, 1) = g(x) for all x ∈ X , it follows that

f /g = (f0 /f1 )(f1 /f2 )(f2 /f3 ) · · · (fn−1 /fn )

is exponential, because f0 (x) = H (x, 0) = f (x) and fn (x) = H (x, 1) = g(x) for all
x ∈ X. 

Corollary 2.8.5 Given a compact metric space X , a map f : X → C − {0} is expo-

nential iff f  c, where c : X → C − {0}, x → z0 ∈ C − {0} is a constant map.

Proof It follows from Theorem 2.8.4 as a particular case. 

Corollary 2.8.6 Given a compact contractible space X , every map f : X → C −

{0} is exponential.

Proof By hypothesis, X is contractible. Then, there exists a homotopy H : X × I →

X such that H : IX  c, where c : X → C − {0}, x → z0 ∈ C − {0} is a con-
stant map. Then, f ◦ H : f  c. Hence, it follows by Corollary 2.8.5 that f is
exponential. 

Example 2.8.7 leads to define the index number of a continuous map f : S 1 →

C − {0} formulated in Definition 2.8.8.

Example 2.8.7 Given a continuous map f : S 1 → C − {0}, define

ψf : [0, 2π ] → C − {0}, t → f (eit ).

Since [0, 2π ] is compact and contractible, it follows by Corollary 2.8.6 that there
exists an exponential map
2.8 Exponential Maps and Homotopy Classification of Complex-Valued Functions 67

g : [0, 2π ] → C − {0}

such that

f (eit ) = e g(t) , ∀ t ∈ [0, 2π ].

If h : [0, 2π ] → C − {0} is another map such that f (eit ) = eh(t) , ∀ t ∈ [0, 2π ].

Then, e g(t)−h(t) = 1, ∀ t ∈ [0, 2π ] implies that g − h assumes only values which are
integral multiples of 2π i. Again since the range of the map g − h is discrete, it
follows that g − h is constant. This shows that the number g(2π ) − g(0) is indepen-
dent of the choice of g satisfying the condition f (eit ) = e g(t) , ∀ t ∈ [0, 2π ]. Hence,
corresponding to each f , there exists a number, denoted by Ind f defined by

Ind f = [g(2π ) − g(0)]/2π i.

Definition 2.8.8 The number Ind f defined in Example 2.8.7 is called the index of
the continuous map f : S 1 → C − {0}, and it is denoted by Ind f .

Example 2.8.9 For the map f : S 1 → C − {0}, z → z m , its index Ind f is m.

Remark 2.8.10 For more study of exponential maps and index number of a map f :
S 1 → C − {0} and their relations with homotopy, see Exercises 49–54 of Sect. 2.28.

2.9 Fundamental Groups

This section starts with the concept of fundamental group and its motivation. The
fundamental group is an algebraic object which is assigned to a geometric space.
This group provides useful tools to measure the number of holes in the geometric
space. For example, the fundamental group of
(i) The Euclidean line is 0, which indicates that it has no hole.
(ii) The Euclidean plane is 0, which indicates that it has no hole.
(iii) The circle is Z, which indicates that it has one hole.
(iv) The torus is Z ⊕ Z which indicates that it has two holes.
On the other hand, in more general, the fundamental group π1 (X , x0 ) of an arbi-
trary pointed topological space (X , x0 ) is defined by using the homotopy classes of
loops in X based at the point x0 as its elements. For this construction, we start with
the set (X , x0 ) of all loops in the space X based at a point x0 . Two loops in X
based at x0 are said to be equivalent if one loop can be continuously deformed to
the other. This defines an equivalence relation ∼ on (X , x0 ) providing the quotient
set (X , x0 )/ ∼, denoted π1 (X , x0 ), which admits a group structure by Theorem
2.9.10, called the fundamental group of X based at x0 .
68 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

This group is a powerful topological invariant and characterizes the connectivity

properties of topological spaces related to properties of loops in these spaces. Its basic
motivation is to detect a hole in the plane by letting loops in the plane shrunk to a
point. This facilitates to attack some topological problems. One of the main problems
in topology is the classification of topological spaces up to homeomorphism. To
solve such a problem, either we have to find a homeomorphism between two given
topological spaces or we have to show that no such homeomorphism exists. In the
latter case, a special property or a characteristic is searched which is shared by
homeomorphic spaces. This search led to the invention of homotopy, fundamental
group and homology groups in algebraic topology.
Historically, Euler characteristic invented by L. Euler (1703–1783) in 1752,
an integral invariant, is the first invented topological invariant which distinguishes
nonhomeomorphic spaces. The search of other invariants establishes a connection
between topology and modern algebra in such a way that homeomorphic spaces have
isomorphic algebraic structures. More precisely, fundamental group and homology
groups invented by Poincaré in 1895 are the first powerful topological (algebraic)
invariants which came from such a search.
Fundamental group is the first of a sequence of functors πn designed in Sect. 2.20,
called homotopy group functors from the category of pointed topological spaces and
their continuous maps to the category of groups and their homomorphisms. Such
functors occupy a vast territory in algebraic topology and are still the subject of
intensive study. More precisely, given a pointed topological space (X , x0 ), the set
π1 (X , x0 ) is defined to be the set of homotopy classes of paths α : I → X that send
0 and 1 to x0 . Each such path is called a loop in X based at x0 . It is shown that
π1 (X , x0 ) admits a group structure. The group π1 (X , x0 ) depends on X as well as
on x0 ∈ X and is called the fundamental group or Poincare group of the space X
based at x0 . It is a homotopy-type invariant in the sense that homotopy equivalent
spaces (X , x0 ) and (Y , y0 ) have the isomorphic fundamental groups π1 (X , x0 ) and
π1 (Y , y0 ). The algebraic properties of the fundamental group reflect the topological
properties of some specified spaces. The study of this group is easier than the study
of the topological space X directly.
Higher homotopy groups πn (X , x0 ) of pointed topological spaces which are
natural generalization of π1 (X , x0 ) and the sequence of higher homotopy functors
πn in the homotopy category of pointed topological spaces are studied in Sect. 2.20.
On the other hand, the sequence of homology functors Hn given by Poincaré is
studied in Chap. 3. Further generalization of πn (X , x0 ) by defining πn (X , A, x0 ) of
any triplet (X , A, x0 ) of topological spaces and a detailed study of πn (X , A, x0 ) are
available in Chap. 5.

2.9.1 Basic Motivation of Fundamental Group

The concept of the fundamental group arose to distinguish two geometrical objects
such as a disk D∗ with a hole and a disk D without a hole in the Euclidean plane R2 as
2.9 Fundamental Groups 69

Fig. 2.9 A disk D∗ with a

hole

α
β

D∗

Fig. 2.10 A disk D without

a hole

displayed in Figs. 2.9 and 2.10. While considering homotopy, the essential property
of a circle is the existence of an inside hole. Topological spaces homeomorphic to a
circle have the same connectivity properties, and they are homotopy equivalent but
its converse is not true.
All loops in the punctured disk D∗ , as shown in Fig. 2.9, cannot be continuously
shrunk to a point, i.e., cannot be continuously deformed to a point in D∗ ; on the other
hand, all loops in the disk D, see Fig. 2.10, can be continuously shrunk to a point in
D. For example, the loop β in Fig. 2.9 cannot be continuously deformed to a point
as there is a hole in D∗ . Some loops in D∗ such as α may be continuously deformed
to a point but not all loops. This characterizes the difference between the spaces D
and D∗ and leads to the concept of fundamental group.

2.9.2 Construction of Fundamental Group

This subsection constructs the fundamental group π1 (X , x0 ) of a pointed topological

space (X , x0 ) by using the homotopy classes of loops in X based at the point x0 as
its elements. For this construction, we start with the set (X , x0 ) of all loops in the
space X based at a point x0 and define a group structure on the set of homotopy
equivalence classes of loops in (X , x0 ).

Definition 2.9.1 Let L(X , x0 ) be the set of all loops in the topological X based at
the point x0 ∈ X and the boundary points of I be İ = {0, 1} ⊂ R. Then, it follows
70 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

from Theorem 2.6.12 that the relation of homotopy α  β rel İ on L(X , x0 ) is an

equivalence relation. Hence, it generates the set of homotopy classes L(X , x0 )/ 
of loops relative to İ, denoted by π1 (X , x0 ).

A suitable composition on the set π1 (X , x0 ) is defined to make it a group.

Definition 2.9.2 Given α, β ∈ L(X , x0 ), their product α ∗ β : (I, İ ) → (X , x0 ) is

defined by


α(2t), 0 ≤ t ≤ 1/2
(α ∗ β)(t) = (2.1)
β(2t − 1), 1/2 ≤ t ≤ 1.

Then, α ∗ β is well defined and it is continuous by pasting lemma. It is a loop in

X based at x0 and hence α ∗ β ∈ L(X , x0 ).
Similarly, given three loops α, β, γ ∈L(X , x0 ), their product α ∗ β ∗ γ is defined by
⎧
⎪
⎨α(3t), 0 ≤ t ≤ 1/3
(α ∗ β ∗ γ )(t) = β(3t − 1), 1/3 ≤ t ≤ 2/3
⎪
⎩
γ (3t − 2), 2/3 ≤ t ≤ 1.

Then, α ∗ β ∗ γ ∈ L(X , x0 ).

Definition 2.9.3 If α ∈ L(X , x0 ), then its inverse α −1 : (I, İ ) → (X , x0 ) defined

by α −1 (t) = α(1 − t), ∀ t ∈ I asserts that α −1 ∈ L(X , x0 ).

Remark 2.9.4 The two paths α and α −1 have the same set of points of X , but they
are of opposite directions.

Proposition 2.9.5 If α1 , α2 , β1 , β2 ∈ L(X , x0 ) are loops such that α1  α2 , rel İ β1 

β2 rel İ, then α1 ∗ β1  α2 ∗ β2 rel İ.

Proof Let H : α1  α2 rel İ and K : β1  β2 rel İ. Then,

H (t, 0) = α1 (t), H (t, 1) = α2 (t), ∀ t ∈ I, H (0, s) = x0 = H (1, s), ∀ s ∈ I

and

K(t, 0) = β1 (t), K(t, 1) = β2 (t), ∀ t ∈ I, K(0, s) = x0 = K(1, s), ∀ s ∈ I.

Define a map F : I × I → X by

H (2t, s), 0 ≤ t ≤ 1/2
F(t, s) =
K(2t − 1, s), 1/2 ≤ t ≤ 1.
2.9 Fundamental Groups 71

Then, F is well defined. Moreover, it is continuous by pasting lemma. Again,


H (2t, 0), 0 ≤ t ≤ 1/2
F(t, 0) =
 K(2t − 1, 0), 1/2 ≤ t ≤ 1
α1 (2t), 0 ≤ t ≤ 1/2
=
β1 (2t − 1), 1/2 ≤ t ≤ 1
= (α1 ∗ β1 )(t), ∀ t ∈ I.

Similarly, F(t, 1) = (α2 ∗ β2 )(t), ∀ t ∈ I, F(0, s) = x0 = H (0, s), ∀ s ∈ I and

F(1, s) = x0 = K(1, s), ∀ s ∈ I. It asserts that F : α1 ∗ β1  α2 ∗ β2 rel İ. 

Proposition 2.9.6 If α, β ∈ L(X , x0 ) and α  β rel İ, then α −1  β −1 rel İ.

Proof Let H : α  β rel İ. Then,

H (t, 0) = α(t), H (t, 1) = β(t), ∀ t ∈ I and H (0, s) = x0 = H (1, s), ∀ s ∈ I.

Define
F : I × I → X , (t, s) → H (1 − t, s).

Then, F is a continuous function such that

F(t, 0) = H (1 − t, 0) = α(1 − t) = α −1 (t), ∀ t ∈ I ,
F(t, 1) = H (1 − t, 1) = β(1 − t) = β −1 (t), ∀ t ∈ I
and F(0, s) = H (1, s) = x0 , F(1, s) = H (0, s) = x0 .
It asserts that F : α −1  β −1 rel İ. 

Proposition 2.9.7 If α, β, γ ∈ L(X , x0 ), then α ∗ (β ∗ γ )  (α ∗ β) ∗ γ rel İ.

Proof Define α ∗ (β ∗ γ ) : (I, İ) → (X , x0 ) by the rule


α(2t), 0 ≤ t ≤ 1/2
(α ∗ (β ∗ γ ))(t) =
⎧ (β ∗ γ )(2t − 1), 1/2 ≤ t ≤ 1
⎨ α(2t), 0 ≤ t ≤ 1/2
= β(4t − 2), 1/2 ≤ t ≤ 3/4
⎩
γ (4t − 3), 3/4 ≤ t ≤ 1.
Then, α ∗ (β ∗ γ ) is well defined and it is continuous by pasting lemma. It is a
loop in X based at x0 and therefore α ∗ (β ∗ γ ) ∈ L(X , x0 ). On the other hand,

(α ∗ β)(2t), 0 ≤ t ≤ 1/2
((α ∗ β) ∗ γ )(t) =
⎧γ (2t − 1), 1/2 ≤ t ≤ 1
⎨ α(4t), 0 ≤ t ≤ 1/4
= β(4t − 1), 1/4 ≤ t ≤ 1/2
⎩
γ (2t − 1), 1/2 ≤ t ≤ 1.
It shows that (α ∗ β) ∗ γ ∈ L(X , x0 ).
Define a map F : I × I → X by the rule
72 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups
⎧
⎨ α(4t/(1 + s)), 0 ≤ t ≤ (1 + s)/4
F(t, s) = β(4t − 1 − s), (1 + s)/4 ≤ t ≤ (2 + s)/4
⎩
γ (1 − (4(1 − t)/(2 − s))), (2 + s)/4 ≤ t ≤ 1.
Then, F is well defined and continuous by pasting lemma. Clearly,
⎧
⎨ α(4t), 0 ≤ t ≤ 1/4
F(t, 0) = β(4t − 1), 1/4 ≤ t ≤ 1/2
⎩
γ (2t − 1), 1/2 ≤ t ≤ 1
= ((α ∗ β) ∗ γ )(t), ∀ t ∈ I,
⎧
⎨ α(2t), 0 ≤ t ≤ 1/2
F(t, 1) = β(4t − 2), 1/2 ≤ t ≤ 3/4
⎩
γ (4t − 3), 3/4 ≤ t ≤ 1
= (α ∗ (β ∗ γ ))(t), ∀ t ∈ I,
Hence, it follows that

F(0, s) = α(0) = x0 , and F(1, s) = γ (1) = x0 .

It proves that (α ∗ β) ∗ γ  α ∗ (β ∗ γ ) rel İ. 

Proposition 2.9.8 If α ∈ L(X , x0 ) and δ : I → X is the constant loop at x0 defined

by δ(t) = x0 , ∀ t ∈ I, then α ∗ δ  α rel İ and δ ∗ α  α rel İ.

Proof α ∗ δ : I → X is defined by the rule


α(2t), 0 ≤ t ≤ 1/2
(α ∗ δ)(t) =
 δ(2t − 1), 1/2 ≤ t ≤ 1
α(2t), 0 ≤ t ≤ 1/2
=
x0 , 1/2 ≤ t ≤ 1.

This asserts that α ∗ δ ∈ L(X , x0 ). Define a map F : I × I → X by the rule


α(2t/(1 + s)), 0 ≤ t ≤ (1 + s)/2
F(t, s) =
x0 , (1 + s)/2 ≤ t ≤ 1.

This asserts that F : α ∗ δ  α rel İ. Similarly, δ ∗ α  α rel İ. 

Proposition 2.9.9 If α ∈ L(X , x0 ), then α ∗ α −1  δ rel İ and α −1 ∗ α  δ rel İ.

Proof α ∗ α −1 : I → X is defined by the rule


α(2t), 0 ≤ t ≤ 1/2
(α ∗ α −1 )(t) =
α −1 (2t − 1), 1/2 ≤ t ≤ 1

α(2t), 0 ≤ t ≤ 1/2
=
α(1 − 2t − 1), 1/2 ≤ t ≤ 1

α(2t), 0 ≤ t ≤ 1/2
=
α(2 − 2t), 1/2 ≤ t ≤ 1.
2.9 Fundamental Groups 73

This asserts that α ∗ α −1 ∈ L(X , x0 ). Define H : I × I → X by


α(2t(1 − s)), 0 ≤ t ≤ 1/2
H (t, s) =
α((2 − 2t)(1 − s)), 1/2 ≤ t ≤ 1.

This shows that F : α ∗ α −1  δ rel İ. Similarly, α −1 ∗ α  c rel İ. 

Theorem 2.9.10 π1 (X , x0 ) is a group under the usual composition of homotopy

classes of loops in L(X , x0 ).

Proof Let [α], [β] ∈ π1 (X , x0 ). Then, α, β ∈ L(X , x0 ) and α ∗ β given in Definition

2.9.2 is in L(X , x0 ). This law of composition ‘∗’ is carried over π1 (X , x0 ) to define the
composition ‘◦’ by the rule [α] ◦ [β] = [α ∗ β]. The composition ‘◦’ is well defined
by Proposition 2.9.5, because it is independent of the choice of the representatives
of the classes. This composition is associative by Proposition 2.9.7 with [δ] as the
identity element by Proposition 2.9.8 and for any element [α] ∈ π1 (X , x0 ), it has an
inverse [α −1 ] ∈ π1 (X , x0 ) by Proposition 2.9.9. Consequently, π1 (X , x0 ) is a group
under the composition ‘◦’. This proves that π1 (X , x0 ) is a group under the usual
composition of homotopy classes of loops in L(X , x0 ). 

Remark 2.9.11 The inverse of an element in the group π1 (X , x0 ) is represented by

a loop traveling the same loop in the reverse direction. Thus, traveling a loop in two
opposite directions determines elements of the group π1 (X , x0 ) which are inverse to
each other.

Definition 2.9.12 The group π1 (X , x0 ) defined in Theorem 2.9.10 is said to be the

fundamental group or Poincaré group of the topological space X based at the point
x0 ∈ X .

Remark 2.9.13 For an equivalent definition of π1 (X , x0 ) and an alternative proof of

its group structure, see Sect. 2.16. The index ‘1’ in π1 (X , x0 ) used now appeared later
than the notation π(X , x0 ) used by Poincaré in 1895. It is also known as the first or
one-dimensional homotopy group. There is an infinite sequence of groups πn (X , x0 )
with n = 1, 2, 3, . . ., called higher-dimensional homotopy groups of the pointed
space (X , x0 ), and the first one is the fundamental group. The higher-dimensional
homotopy groups (see Sect. 2.20) were introduced by W. Hurewicz in 1935 in his
paper [Hurewicz, 1935]. For n = 0, π0 (X , x0 ) taken to be the set of path-connected
components of X is not necessarily a group.

Example 2.9.14 (i) The fundamental group of I = [0, 1] at the base point 1 ∈ I is
the trivial group, because if α : I → I is a loop, then H : I × I → I, (x, t) →
t + (1 − t)α(x) is a homotopy between α and the constant loop at the base point
1 ∈ I. This asserts that π1 (I, 1) consists of only element, denoted by 0.
(ii) For a contractible space X and a point x0 ∈ X , π1 (X , x0 ) = 0. It is also true that
π1 (X , x) = 0 for any x ∈ X . Hence for any tree X (topological) and a vertex v0
of X , the fundamental group π1 (X , vo ) = 0.
 74 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Fig. 2.11 Geometrical x1

description of ψα X
ᾱ
f
x0

(iii) For the n-dimensional Euclidean space Rn , its fundamental group π1 (Rn , x) = 0
for any x ∈ Rn .
(iv) For the n-dimensional Euclidean disk Dn , its fundamental group π1 (Dn , x) = 0
for any x ∈ Dn .
(v) For any convex set X ⊂ Rn , its fundamental group π1 (X , x0 ) = 0 for any x0 ∈ X .
Remark 2.9.15 It is a natural question : Does the fundamental group of a topological
space depend on its base points x0 , x1 ∈ X ? If it depends on base points, how are
the groups π1 (X , x0 ) and π1 (X , x1 ) related to arbitrary topological spaces? Theorem
2.9.16 gives its answer.
Theorem 2.9.16 Let X be a path-connected space and x0 , x1 ∈ X be any two distinct
points. Then, π1 (X , x0 ) and π1 (X , x1 ) are isomorphic groups.
Proof By hypothesis, X is a path-connected space and x0 , x1 ∈ X are two distinct
points. Then, there exists a path α : I → X , α(0) = x0 , α(1) = x1 , with its inverse
path ᾱ : I → X , t → α(1 − t), shown in Fig. 2.11. Define a map

ψα : π1 (X , x0 ) → π1 (X , x1 ), [f ] → [α ∗ f ∗ α].

It is geometrically represented in Fig. 2.11. Then, ψα is well defined and a group

isomorphism with its inverse ψᾱ : π1 (X , x1 ) → π1 (X , x0 ), [f ] → [α ∗ f ∗ α]. 
Remark 2.9.17 For any path-connected space X , the fundamental group π1 (X , x0 )
is independent of its base point x0 and hence it is unique up to isomorphism by
Theorem 2.9.16. This group is abbreviated to π1 (X ), without specifying the base
point of the path-connected space X .

Corollary 2.9.18 Let X be a topological space and x0 , x1 ∈ X . If α is a path in X

from x0 to x1 , then α induces an isomorphism

ψα : π1 (X , x0 ) → π1 (X , x1 ), [f ] → [α ∗ f ∗ α].

Proof It follows from the proof of Theorem 2.9.16. 

Proposition 2.9.19 relates path homotopic maps to isomorphism of fundamental
groups.
Proposition 2.9.19 If α and β are two path homotopic maps in X joining the points
x0 to x1 , then their induced isomorphisms ψα = ψβ : π1 (X , x0 ) → π1 (X , x1 ), i.e.,
2.9 Fundamental Groups 75

Fig. 2.12 Path homotopic X

maps x1
β

α
x0

(i) the isomorphism ψα : π1 (X , x0 ) → π1 (X , x1 ) and

(ii) the isomorphism ψβ : π1 (X , x0 ) → π1 (X , x1 )
defined in Corollary 2.9.18 are the same.

Proof If α and β are two homotopic paths as shown in Fig. 2.12, then ᾱ and β̄ are
also path homotopic. Therefore, it follows that for any loop f in X based at x0 , the
path ᾱ ∗ f ∗ α is a path homotopic to the path β̄ ∗ f ∗ β. Consequently, ψα ([f ]) =
ψβ ([f ]) ∀ [f ] ∈ π1 (X , x0 ). This asserts that ψα = ψβ : π1 (X , x0 ) → π1 (X , x1 ). 

2.10 Homotopy Property of Connectedness

This section establishes an interesting relation between homotopy and connected-

ness properties of topological spaces by showing that connectedness is a homotopy
property of the spaces in the sense that this property is preserved by every homo-
topy equivalence, which is in addition to its topological property proved in Chap. 5
of Basic Topology, Volume 1 of the present series of books.

Definition 2.10.1 Given a topological space X , let C(X ) denote the set of all con-
nected components or simply called components of X . If f : X → Y is an arbitrary
continuous map, then for each component C of X , f (C) ⊂ Y is connected and hence
it is contained in a component D of Y . This defines a map

ψ(f ) : C(X ) → C(Y ), C → D

which is well defined, since components of a topological space are mutually disjoint.

Proposition 2.10.2 If f , g : X → Y are homotopic maps, then ψ(f ) = ψ(g).

Proof Let f , g : X → Y be homotopic maps. Then, there exists a homotopy

F :X ×I→Y

such that F(x, 0) = f (x) and F(x, 1) = g(x) for all x ∈ X . Since I is connected, the
product space C × I of two connected sets is connected for every component C ∈
C(X ). Again since, F is continuous, it follows that F(C × I) ⊂ Y is also connected.
76 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

It shows that F(C × I) is contained in a component D in C(Y ). Again since both

f (C) and g(C) are contained in H (C × I), it follows that

[ψ(f )](C) = D = [ψ(g)](C)

for every C ∈ C(X ). This proves that ψ(f ) = ψ(g). 

Proposition 2.10.3 Given a homotopy equivalence f : X → Y , the map

ψ(f ) : C(X ) → C(Y )

is a bijection.
Proof As f : X → Y is a homotopy equivalence by hypothesis, there exists a con-
tinuous map g : Y → X such that f ◦ g  1Y and g ◦ f  1X , where 1X and 1Y are
identity maps on X and Y , respectively. Then by Proposition 2.10.2, it follows that

ψ(f ) ◦ ψ(g) = ψ(f ◦ g) = ψ(1Y )

and
ψ(g) ◦ ψ(f ) = ψ(g ◦ f ) = ψ(1X ),

where ψ(1X ) and ψ(1Y ) are identity maps on the sets C(X ) and C(Y ), respectively.
This asserts that ψ(f ) is a bijection with ψ(g) its inverse. 
Definition 2.10.4 A property of a topological space is said to be a homotopy prop-
erty if this property is preserved by every homotopy equivalence; i.e., a homotopy
property of a topological X is such a property that every topological space homotopy
equivalent to X has also the same property.
Corollary 2.10.5 Connectedness is a homotopy property.
Proof It follows from Proposition 2.10.3. 

2.11 Functorial Property of π1

Functors play a key role in algebraic topology. This section proves the functorial
property of π1 , where π1 (X , x0 ) denotes the fundamental group of the pointed space
(X , x0 ). In the language of category theory, π1 is a covariant functor from the homo-
topy category Htp∗ of pointed topological spaces and homotopy classes of continu-
ous maps to the category Grp of groups and homomorphisms. This functor transfers
topological problems into algebraic ones to have a better chance for solution. The
fundamental group of a pointed topological space is a topological (algebraic) invari-
ant as well as it is a homotopy invariant. The aim of this section is to study π1 through
the category theory and establish some important properties of fundamental group.
 2.11 Functorial Property of π1 77

Theorem 2.11.1 Let f : (X , x0 ) → (Y , y0 ) be a base point preserving continuous

map. Then, it induces a group homomorphism

f∗ : π1 (X , x0 ) → π1 (Y , y0 ), [α] → [f ◦ α]

such that
(i) For the identity map 1X : (X , x0 ) → (X , x0 ), its induced homomorphism 1X ∗ is
the identity automorphism on π1 (X , x0 ).
(ii) For the base point preserving continuous maps f : (X , x0 ) → (Y , y0 ) and g :
(Y , y0 ) → (Z, z0 ), their composite map g ◦ f induces homomorphism with the
property
(g ◦ f )∗ = g∗ ◦ f∗ : π1 (X , x0 ) → π1 (Z, z0 ).

(iii) For homotopic maps f  g : (X , x0 ) → (Y , y0 ) rel {x0 }, their induced homo-

morphisms
f∗ = g∗ : π1 (X , x0 ) → π1 (Y , y0 ).

(iv) For a homeomorphism f : (X , x0 ) → (Y , y0 ), its induced homomorphism

f∗ : π1 (X , x0 ) → π1 (Y , y0 )

is an isomorphism of groups.
(v) For a homotopy equivalence f : (X , x0 ) → (Y , y0 ) its induced homomorphism

f∗ : π1 (X , x0 ) → π1 (Y , y0 )

is an isomorphism of groups.

Proof Let α, β be two homotopic loops in X based at x0 and H : α  β rel İ be a

homotopy. Since f ◦ α is a loop in Y based at y0 , for every loop α in X based at x0 ,

f ◦ H : f ◦ α  f ◦ β rel İ.

It shows that f∗ is well defined, and moreover, it is a homomorphism.

(i) If 1X : (X , x0 ) → (X , x0 ) is the identity map, then 1X ◦ α = α for every loop α
in X based at x0 asserts that 1X ∗ ([α]) = [1X ◦ α] = [α], ∀ [α] ∈ π1 (X , x0 ). It
shows that 1X ∗ : π1 (X , x0 ) → π1 (X , x0 ) is the identity automorphism.
(ii) Given any element [α] ∈ π1 (X , x0 ), it follows that
(g ◦ f )∗ ([α]) = [(g ◦ f ) ◦ α] = [g ◦ (f ◦ α)] = g∗ ([f ◦ α]) = g∗ (f∗ ([α])) = (g∗ ◦ f∗ )[α].

It asserts that (g ◦ f )∗ = g∗ ◦ f∗ .
(iii) If f  g rel {x0 }, then ∀ [α] ∈ π1 (X , x0 ), f ◦ α  g ◦ α rel {y0 }. It asserts that

f∗ ([α]) = [f ◦ α] = [g ◦ u] = g∗ ([α]).
78 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Hence, it follows that f∗ = g∗ .

(iv) If f : (X , x0 ) → (Y , y0 ) is homeomorphism, there exists a continuous map g :
(Y , y0 ) → (X , x0 ) such that f ◦ g = 1Y , g ◦ f = 1X . Hence, it follows from (i)
and (ii) that f∗ : π1 (X , x0 ) → π1 (Y , y0 ) is an isomorphism with g∗ as its inverse.
(v) If f : (X , x0 ) → (Y , y0 ) is a homotopy equivalence, there exists a continuous
map g : (Y , y0 ) → (X , x0 ) such that f ◦ g  1Y , g ◦ f  1X . Hence, it from (iii)
that
f∗ : π1 (X , x0 ) → π1 (Y , y0 )

is an isomorphism with g∗ as its inverse. 

Corollary 2.11.2 (i) The fundamental group is a topological invariant.
(ii) The fundamental group is a homotopy invariant.
Proof (i) The fundamental group of a pointed topological space is invariant under
homeomorphisms by Theorem 2.11.1(iv) in the sense that homeomorphic
pointed topological spaces have isomorphic fundamental groups. This implies
that it is a topological invariant.
(ii) The fundamental group of a pointed topological space is invariant under homo-
topy equivalence by Theorem 2.11.1(v) in the sense that homotopy equivalent
pointed topological spaces have isomorphic fundamental groups. This implies
that it is a homotopy invariant. 
Theorem 2.11.3 expresses π1 in the language of category theory by using Theorem
2.11.1.

Theorem 2.11.3 Let T op∗ be the category of pointed topological spaces and their
base point preserving continuous maps, Grp be the category of groups and their
homomorphisms and Htp∗ be the homotopy category of pointed topological spaces
and their homotopy classes of maps. Then,
(i) π1 : T op∗ → Grp is a covariant functor from the category T op∗ to the category
Grp such that if f , g : (X , x0 ) → (Y , y0 ) are continuous and f  g rel {x0 }, then
π1 (f ) = f∗ = g∗ = π1 (g).
(ii) π1 : Htp∗ → Grp is also a covariant functor from the category Htp∗ to the
category Grp.
Proof (i) The assignment

π1 : T op∗ → Grp, (X , x0 ) → π1 (X , x0 )

defines an object function, and the assignment

π1 : T op∗ → Grp, f → π1 (f ) = f∗

defines a morphism function. Hence, (i) follows from Theorem 2.11.1.

(ii) It follows from (i) and Theorem 2.11.1. 
2.12 Link between Fundamental Groups and Retractions 79

2.12 Link between Fundamental Groups and Retractions

This section studies the concepts of retraction and strong deformation retractions
by using fundamental group. It is proved in Proposition 2.12.2 that every strong
deformation retract A of X has the isomorphic fundamental groups.
Proposition 2.12.1 Let X be a topological space and A be a subspace of X . Then,
(i) Every retraction r : X → A induces an epimorphism

r∗ : π1 (X , a) → π1 (A, a), ∀ a ∈ A.

(ii) The inclusion map i : A → X induces a monomorphism

i∗ : π1 (A, a) → π1 (X , a), ∀ a ∈ A.

Proof Let r : X → A be a retraction and i : A → X be the inclusion map. Then given

a point a ∈ A, the map r : (X , a) → (A, a) is a continuous map such that r(a) = a
and the induced maps
r∗ : π1 (X , a) → π1 (A, a)

and
i∗ : π1 (A, a) → π1 (X , a)

i r
are both homomorphisms. Since the composite map r ◦ i : (A, a) − → (X , a) −
→
(A, a) is the identity map on (A, a), the composite of induced homomorphisms

i∗ r∗
r∗ ◦ i∗ : π1 (A, a) −→ π1 (X , a) −→ π1 (A, a)

is the identity automorphism on π1 (A, a) by Theorem 2.11.1 by using the functorial

property of π1 . This asserts from group theory that r∗ is an epimorphism and i∗ is a
monomorphism for every a ∈ A. 
Proposition 2.12.2 Let X be a topological space and A ⊂ X be a strong deformation
retract of X . Then, the fundamental groups π1 (A, a) and π1 (X , a) are isomorphic
for each a ∈ A.
Proof Let A ⊂ X be a strong deformation retract of X . Hence, there exists a retrac-
tion r : (X , a) → (A, a) for every a ∈ A with the property that 1X  i ◦ r rel A.
This implies that the composite homomorphism i∗ ◦ r∗ is the identity automorphism.
Hence, i∗ is an epimorphism. Again, r ◦ i = id asserts that the composite homo-
morphism r∗ ◦ i∗ is an automorphism. Hence, i∗ is a monomorphism. This implies
that
i∗ : π1 (A, a) → π1 (X , a)

is an isomorphism for every a ∈ A. 

80 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Remark 2.12.3 Let X be a given topological space and x0 ∈ X . If Cx0 is the path
component of X containing x0 , then π1 (Cx0 , x0 ) = π1 (X , x0 ). Since all the loops
based at the point x0 and their homotopies in X relative to {x0 } lie entirely in the
subspace Cx0 , it asserts that the fundamental group π1 (X , x0 ) depends only on the
path component Cx0 . Unfortunately, this does not provide any information on the
complement set X - Cx0 .

2.13 Simply Connectedness

This section formalizes the intuitive concept of geometrical objects X having no

hole to prevent any closed path in X from shrinking to a point in X . For example,
every ordinary closed curve on the sphere can be deformed continuously into a point
without leaving the surface. On the other hand, there exist circles C and C  on the
surface of the torus as shown in Fig. 2.14 which cannot be deformed continuously
into a point without leaving its surface. The former example leads to the concept
of simply connected spaces, and the latter example leads to concept of nonsimply
connected spaces. Moreover, simply connected spaces are characterized in Theorem
2.13.9 with the help of homotopy theory.

Definition 2.13.1 A topological space X is said to be simply connected if it is

path connected and its fundamental group is trivial for some base point x0 ∈ X and
hence for every base point of X . In other words, a path-connected space X is simply
connected if π1 (X , x0 ) = 0 for some x0 ∈ X (hence for every base point x0 ∈ X ).

Remark 2.13.2 Geometrically, a path-connected topological space X is said to be

simply connected if there is no hole in X to prevent any closed path in X from shrink-
ing to a point in X . More precisely, a path-connected topological space is simply con-
nected if on the topological space every closed curve can be continuously deformed
into a point without leaving the space. For example, it appears geometrically that the
n-sphere S n is simply connected for n > 1 and hence, π1 (S n ) = 0, ∀ n > 1.

Example 2.13.3 The real line space R and the Euclidean n-space Rn are simply
connected spaces by Corollary 2.13.6. The circle S 1 is not simply connected, because
π1 (S 1 ) = 0 by Corollary 2.19.2, though it is path connected.

Theorem 2.13.4 Let α, β : I → X be two paths in a path-connected space X with

the same initial point and the same terminal point. Then, the paths α and β are
homotopic, X is simply connected.
2.13 Simply Connectedness 81

Fig. 2.13 Paths α, β in X X

x1
α

x0 β

Proof Let α, β : I → X be two paths in a path-connected space X from the point x0

to the point x1 , as shown in Fig. 2.13. Then, α ∗ β̄ is a loop in X based at x0 , where
β̄ is the inverse path of β in X .
By hypothesis, X is simply connected. Hence, it follows that α ∗ β̄  δx0 (constant
loop at x0 ). It shows that α ∗ β̄ ∗ β  δx0 ∗ β. Consequently, [(α ∗ β̄) ∗ β] = [δx0 ∗
β = [β]. Again, [(α ∗ β̄) ∗ β] = [α ∗ (β̄ ∗ β)] = [α ∗ δx1 ] = [α]. Hence, [α] = [β]
shows that α  β. 

Theorem 2.13.5 Let X be a contractible space. Then, it is simply connected.

Proof Since X is a contractible space by hypothesis, the identity map 1X  δ for

some constant map δ. Hence, there is a point x0 ∈ X and a homotopy F : X × I → X
such that F(x, 0) = x and F(x, 1) = x0 , ∀ x ∈ X . Define a path

αx = F(x, −) : I → X , t → αx (t) = F(x, t)

in X from αx (0) = F(x, 0) = x to αx (1) = F(x, 1) = x0 . Again, for any y ∈ X , αy is

a path from y to x0 and hence ᾱy (the inverse path of αy ) is a path in X from x0 to y.
Thus, any two points x and y can be joined by the path αx ∗ ᾱy in X . This asserts that
the space X is path-connected space. Since π1 (X , x0 ) = 0 (see Example 2.9.14), it
follows that the space X is simply connected. 

Corollary 2.13.6 The Euclidean n space Rn and any convex subspace of Rn are
simply connected.

Definition 2.13.7 A subspace X ⊂ Rn is called star convex if for some x0 ∈ X ,

(1 − t)x + tx0 ∈ X , ∀ t ∈ I

for any other point x ∈ X ; i.e., geometrically, it means that all the line segments
joining the point x0 to any other point x ∈ X completely lie in X .

Proposition 2.13.8 Every star convex subspace of Rn is simply connected.

Proof Let X ⊂ Rn be a star convex space and x0 ∈ X . Then, there exists a continuous
map
F : X × I → X , (x, t) → (1 − t)x + tx0 .

This shows that F : IX  cx0 (constant map at x0 ) and hence the space is contractible.
This asserts by Theorem 2.13.5 that X is simply connected. 
82 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

2.13.1 A Characterization of Simply Connected Spaces

Theorem 2.13.9 gives a characterization of simply connected spaces with the help
of homotopy of paths.

Theorem 2.13.9 Let X be a path-connected space. Then, it is simply connected

iff every pair of paths in X with the same initial point and same terminal point is
homotopic.
Proof First suppose that the path-connected space X is simply connected. Given
two points x0 , x1 ∈ X , let α and β be two paths in X from x0 to x1 . Then, it follows
by Theorem 2.13.4 that α  β. Next suppose that the space X is path connected
and [α] ∈ π1 (X , x0 ). Then by hypothesis, α  cx0 (constant path at x0 ). It shows that
[α] = [cx0 ] and hence π1 (X , x0 ) = 0. As X is path connected, it follows that the
space X is simply connected. 

2.13.2 Simply Connected Surfaces

This subsection continues the study of connected and disconnected spaces X , when
X is in particular a surface and also provides concrete examples of simply connected
surfaces. A connected surface S is said to be simply connected if every closed path
on S can be deformed continuously into a point without leaving the surface S.

Definition 2.13.10 A surface is said to be arc connected (or path connected) in

Rn if any two points on the surface can be joined by a continuous path.
Example 2.13.11 (i) The sphere S 2 is arc connected.
(ii) The hyperboloid of two sheets is not arc connected.
(iii) Every arc-connected surface is connected. A surface which is not connected
consists of two or more distinct surfaces.

Definition 2.13.12 A connected surface is said to be simply connected if every

closed curve on the surface can be continuously deformed to a point without leaving
the surface.
Example 2.13.13 (i) The sphere S 2 is a simply connected surface.
(ii) The torus T = S 1 × S 1 is not a simply connected surface, since every closed
curve on the surface of the torus T cannot be continuously deformed to a point
without leaving the surface of the torus T . For example, the closed curve C 
on T can be continuously deformed to a point without leaving the torus, but the
closed curves C and C  as shown in Fig. 2.14 cannot be continuously deformed
to a point.
2.14 Fundamental Groups of Product Spaces 83

Fig. 2.14 Circles C, C  , C 

on the surface of the torus
C 

C C

2.14 Fundamental Groups of Product Spaces

This section gives a relation between the fundamental group of a product space and
the fundamental group of its factor spaces.

From abstract group theory, it is well known that

(i) Given two groups G and H , their Cartesian product G × H admits a group
structure with pointwise multiplication given by

(g, h)(g  , h ) = (gg  , hh ), ∀ g, g  ∈ G, h, h ∈ H .

(ii) If f : K → G and s : K → H are any two homomorphisms from any group K,

then the map

ψ : K → G × H , k → (f (k), s(k))

is also a group homomorphism.

Moreover, from general topology, it is known that

(i) Given two canonical projections p1 : (X × Y ) → X , p2 : X × Y → Y and a

given a pair of continuous maps f : I → X , g : I → Y , there is a continuous
map
(f , g) : I → X × Y , t → (f (t), g(t)).

(ii) Conversely, any continuous map h : I → X × Y defines a pair of continuous

maps

p1 ◦ h : I → X and p2 ◦ h : I → Y .

Proof of Theorem 2.14.1 is based on the above algebraic and topological results.

Theorem 2.14.1 If X and Y are two pointed topological spaces having base points
x0 ∈ X and y0 ∈ Y , then the fundamental groups π1 (X , x0 ) ⊕ π1 (Y , y0 ) and π1 (X ×
Y , (x0 , y0 )) are isomorphic.
84 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proof If α : (I, İ) → (X × Y , (x0 , y0 )) is a loop in X × Y at (x0 , y0 ), then the canon-

ical projections p : X × Y → X and q : X × Y → Y are continuous maps. Hence,
they induce homomorphisms

(i) p∗ : π1 (X × Y , (x0 , y0 )) → π1 (X , x0 ), [α] → [p ◦ α] and

(ii) q∗ : π1 (X × Y , (x0 , y0 )) → π1 (Y , y0 ), [α] → [q ◦ α].

Define a map

f = (p∗ , q∗ ) : π1 (X × Y , (x0 , y0 )) → π1 (X , x0 ) ⊕ π1 (Y , y0 ), [γ ] → (p ∗ ([γ ]),

q∗ ([γ ])) = ([p ◦ γ ], [q ◦ γ ]).

Then, f = (p∗ , q∗ ) is a group isomorphism with its inverse isomorphism

g : π1 (X , x0 ) ⊕ π1 (Y , y0 ) → π1 (X × Y , (x0 , y0 )), ([α], [β]) → [(α, β)],

where (α, β) is defined by

(α, β) : I → X × Y , t → (α(t), β(t)). 

Corollary 2.14.2 If X = S 1 × R is a cylinder, then π1 (X ) ∼

= Z.

Proof It follows from Theorem 2.14.1, since both of S 1 and R are arcwise-connected
spaces with π1 (S 1 ) ∼
= Z and π1 (R) ∼
= 0. 

2.15 Fundamental Groups of Hopf Spaces and Topological

Groups

This section discusses Hopf spaces, named after H. Hopf (1894–1971), by general-
izing topological groups, and studies their fundamental groups. Fundamental groups
of arbitrary topological spaces may be abelian or nonabelian. For example, the fun-
damental group of the circle is abelian but that of figure-eight is nonabelian (see
Theorems 2.19.1 and 2.19.16). On the other hand, the fundamental group of every
H-space is always abelian and hence fundamental group of every topological group
is abelian, which is an interesting result proved in this section.

2.15.1 H-spaces and their Fundamental Groups

This subsection defines Hopf spaces and proves that their fundamental groups are
always abelian. Every topological group is an H -space, but its converse is not neces-
2.15 Fundamental Groups of Hopf Spaces and Topological Groups 85

sarily true (see Example 2.15.4 ). For the study of Hopf group formulated in Definition
2.15.3 from the viewpoint of homotopy theory, the book [Adhikari, 2016] is referred.

Definition 2.15.1 Let X be a pointed topological space with base point x0 ∈ X .

Then, X is said to be an Hopf space, in brief, H -space if there exists a continuous
multiplication
μ : X × X → X,

whose image is abbreviated, μ(x, y) = xy such that the constant map

c : X → X , x → x0

is a homotopy identity in the sense that the maps

α, β : X → X , x → x0 x, xx0

are homotopic.

Definition 2.15.2 An H -space X with associative continuous multiplication

μ : X × X → X , (x, y) → xy

and homotopy identity

c : X → X , x → x0

is said to be an Hopf group (H-group) if there exists a continuous map

ψ :X →X

such that each of the maps

X → X , x → ψ(x)x, xψ(x)

is homotopic to the map c rel {x0 }. The map ψ is said to be homotopy inverse for
X and μ.

Definition 2.15.3 An associative Hopf space with homotopy inverse is said to be an

Hopf group or H -group. It is sometimes called a generalized topological group.

Example 2.15.4 Every topological group is an Hopf space. Let (X , x0 ) be a topolog-

ical group with the identity element x0 . Then, the maps α, β : X → X , x → x0 x, xx0
are equal and hence they are homotopic. But its converse is not necessarily true. For
example,

(i) The infinite real projective space RP ∞ = n≥0 RP n endowed with weak topol-
ogy is an Hopf space but it is not a topological group.
86 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

(ii) The infinite complex projective space CP ∞ = n≥0 CP n endowed with weak
topology is an Hopf space but it is not a topological group.

Example 2.15.5 Every Lie group is an H -space. Hence, the general linear group
GL(n, R) and the orthogonal groups O(n, R) form an important family of H -spaces.

Remark 2.15.6 Example 2.15.4 asserts that every topological group X = G is an

Hopf space, where x0 = e is taken to be the identity element of the topological group.

Theorem 2.15.7 Let X be an H -space with continuous multiplication μ : X × X →

X and homotopy identity c : X → X , x → x0 . Then, π1 (X , x0 ) is abelian.

Proof The isomorphism f and g defined in Theorem 2.14.1 are now applied to prove
the theorem by taking, in particular, X = Y . Then,

f : π1 (X × X , (x0 , x0 )) → π1 (X , x0 ) ⊕ π1 (X , x0 )

is an isomorphism with its inverse isomorphism g : π1 (X , x0 ) × π1 (X , x0 ) → π1

(X ⊕ X , (x0 , x0 )) and 1X : X → X , x → x is the identity map. Now, for [α], [β] ∈
π1 (X , x0 ),

[β] = (μ ◦ (c, 1X ))∗ [β] = μ∗ ([(c, 1X ) ◦ β]) = μ∗ ([(c ◦ β, β)]) = (μ∗ ◦ g)([c ◦ β],
[β]) = (μ∗ ◦ g)(e, [β]),

where e is the identity element of π1 (X , x0 ). Similarly, [α] = (μ∗ ◦ g)([α], e).

Clearly,
μ∗ ◦ g : π1 (X , x0 ) ⊕ π1 (X , x0 ) → π1 (X , x0 )

is a homomorphism such that

(μ∗ ◦ g)(([α], [β])) = (μ∗ ◦ g)(([α], e),(e, [β])) = (μ∗ ◦ g)(([α], e)) · (μ∗ ◦ g)
(e, [β]) = [α] ◦ [β]

and also

(μ∗ ◦ g)(([α] · [β])) = (μ∗ ◦ g)((e, [β]), ([α], e)) = (μ∗ ◦ g)

(e, [β]) · (μ∗ ◦ g)([α], e) = [β] ◦ [α].

Hence,
[α] ◦ [β] = [β] ◦ [α], ∀ [α], [β] ∈ π1 (X , x0 )

asserts that the group π1 (X , x0 ) is abelian. 

 2.15 Fundamental Groups of Hopf Spaces and Topological Groups 87

2.15.2 Fundamental Groups of Topological Groups

This subsection proves that the fundamental group of every topological group is
abelian. As topological groups occupy a vast territory in topology, geometry and
algebra, it is necessary to study their fundamental groups.

Proposition 2.15.8 Let G be a topological group with identity element e. Then, the
fundamental π1 (G, e) is abelian.

Proof By hypothesis, G is a topological group with identity element e. Hence, it

is an H-space by Remark 2.15.6. Consequently, the proposition is proved by using
Theorem 2.15.7. 

Example 2.15.9 provides an extensive class of topological spaces having abelian

fundamental groups.

Example 2.15.9 (Some topological spaces having abelian fundamental groups)

There are plenty of such examples in matrix algebra. To show it, consider the
set M(n, R) of all square matrices of order n over the field R identified with the
2
Euclidean n2 -space Rn . The fundamental groups of GL(n, R), SL(n, R), O(n, R)
and SO(n, R) and their complex analogues GL(n, C), SL(n, C), U(n, C), SU(n, C),
and symplectic group Sp(n, H) based at identity matrix are all abelian by Proposition
2.15.8, where these classical topological groups of matrices are defined as follows.

(i) Let GL(n, R) be the set of all n × n nonsingular matrices over R. It forms a
topological group under usual multiplication of matrices, called general linear
group over R.
(ii) Let SL(n, R) be defined by SL(n, R) = {A ∈ GL(n, R) : det A = 1}. It is a sub-
group of GL(n, R), called special linear group.
(iii) Let O(n, R) be defined by O(n, R) = {A ∈ GL(n, R) : AAt = I } is a subgroup
of GL(n, R) called orthogonal group.
(iv) Let SO(n, R) be defined by SO(n, R) = {A ∈ O(n, R) : det A = 1}. It forms
topological group, called special orthogonal group.
(v) The topological groups GL(n, C), SL(n, C), U(n, C), SU(n, C) are defined in
analogous ways.
(vi) The quaternionic analogue of orthogonal analogue is the symplectic group
Sp(n, H) defined by

Sp(n, H) = {A ∈ GL(n, H) : AA∗ = I },

where A∗ is the quaternionic conjugate transpose of A, and conjugation is in

the sense of reversal of all three imaginary components. It is a subgroup of the
topological group GL(n, H).
88 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Remark 2.15.10 Given a pointed topological space (X , x0 ), its fundamental group

π1 (X , x0 ) may be abelian or nonabelian. But for a pointed topological space (X , x0 ),
if its fundamental group π1 (X , x0 ) is nonabelian, then there exists no multiplication
on X making it a topological group and even such a space cannot be equipped with
the structure of an H -space because of Theorem 2.15.7 and Proposition 2.15.8.

2.16 Alternative Approach to Fundamental Group

This section presents an alternative approach to the concept of fundamental groups

defined in Sect. 2.9, by considering a loop on a pointed topological space (X , x0 ) as a
continuous map from a circle into X , and a chosen point of the circle is being sent to
x0 by this map. The basic objective of the concept of fundamental groups is to classify
all loops in a topological space based at a point up to homotopy equivalence. This
leads to define an alternative definition of fundamental groups. This section gives
a convenient approach to define fundamental group initiated by Hurewicz (1904–
1956) in his paper [Hurewicz, 1935], which is equivalent to this concept inaugurated
by Poincaré in 1895, described in Theorem 2.9.10. For this alternative approach,
an element [f ] ∈ π1 (X , x0 ) is defined to be the homotopy class of the continuous
map f : (S 1 , 1) → (X , x0 ), where S 1 = {e2πit : 0 ≤ t ≤ 1} is the unit circle in the
complex plane C. The motivation of this approach was born through the observation
that every loop α : (I , İ ) → (X , x0 ) gives rise to a pointed continuous map

α̃ : (S 1 , 1) → (X , x0 ), e2πit → α(t),

and conversely, every continuous map β : (S 1 , 1) → (X , x0 ) gives a loop

β̃ : (I, İ) → (X , x0 ), t → β(e2πit ).

Proposition 2.16.1 For a pointed topological space (X , x0 ), the map

f : π1 (X , x0 ) → [(S 1 , 1), (X , x0 )], [α] → [α̃]

is bijective.

Proof Since every loop α : (I, İ) → (X , x0 ) determines a continuous map

α̃ : (S 1 , 1) → (X , x0 ), e2πit → α(t)

and every continuous map β : (S 1 , 1) → (X , x0 ) determines a loop

β̃ : (I, İ) → (X , x0 ), t → β(e2πit ),

the proposition follows, since the maps α̃ and β̃ are determined uniquely up to
homotopy. 
2.16 Alternative Approach to Fundamental Group 89

Let (X , x0 ) be the set of homotopy classes [(S 1 , 1), (X , x0 )] of the continuous

maps ψ : (S 1 , 1) → (X , x0 ). The identification map f described in Proposition 2.16.1
identifies (X , x0 ) with the group π1 (X , x0 ).

Theorem 2.16.2 (X , x0 ) is a group isomorphic to the fundamental group π1 (X , x0 ).

Proof Let α, β : (I, İ) → (X , x0 ) be two given loops. Then by Proposition 2.16.1,
there exist two continuous maps

α̃ : (S 1 , 1) → (X , x0 ), e2πit → α(t)

and
β̃ : (S 1 , 1) → (X , x0 ), e2πit → β(t).

Define a composition ‘◦’ on the (X , x0 ) = [(S 1 , 1) → (X , x0 )] by the rule

[α̃] ◦ [β̃] = [(α

∗ β)],

where α ∗ β is given in Definition 2.9.2. Then, (X , x0 ) is a group under the compo-

sition ‘◦’ with homotopy class of the constant map δ : S 1 → {x0 } as identity element
and the inverse of the homotopy class [α̃] is represented by the map

α̃  : (S 1 , 1) → (X , x0 ), eiθ → α̃(e−iθ ), 0 ≤ θ ≤ 2π.

Hence, the bijection

f : π1 (X , x0 ) → [(S 1 , 1), (X , x0 )], [α] → [α̃]

defines an isomorphism of groups. This establishes the equivalence of the two groups
π1 (X , x0 ) and (X , x0 ) defined in two different ways. 

Remark 2.16.3 At many situations, it is convenient to consider an element [α] ∈

π1 (X , x0 ) as the homotopy class of the map

α̃ : (S 1 , 1) → (X , x0 ), e2πit → α(t).

Again, the circle S 1 defined in the complex plane is a group under usual multipli-
cation of complex numbers. This multiplication rule can also be used to define the
composition law on π1 (S 1 , 1).

2.17 Degree and Winding Number of a Loop on the Circle

This section communicates the concept of degree of a loop f : (I, İ) → (S 1 , 1) on

the circle S 1 by using homotopy theory with an eye to compute the fundamental
90 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Fig. 2.15 Exponential map

p : R → S 1 , t → e2π it 2

1
p
R 0
1
−1
S1
−2

Fig. 2.16 Lifting f˜ of f in R

group of the circle S 1 . On the other hand, its generalization for the degree of spheri-
cal maps f : S n → S n (n ≥ 1) using homotopy theory is studied in Sect. 2.22 and by
using homology theory the same concept is available in Chapter 3. This section also
studies winding number of a closed curve from the viewpoint of homotopy theory.

The exponential map p : R → S 1 , t → e2πit is a continuous map onto S 1 , and

geometrically, it wraps the real line R rounding the circle infinite number of times
as represented in Fig. 2.15.

Definition 2.17.1 The exponential map p : R → S 1 , t → e2πit defines a group homo-

morphism p : (R, +) → (S 1 , ·) from the additive group of reals under usual addition
to the multiplicative group S 1 under usual multiplication of complex numbers with
its kernel ker p = Z.

Definition 2.17.2 Let f : (X , x0 ) → (S 1 , 1) be a continuous map and z0 ∈ ker p.

The map f is said to have a lifting if there is a continuous map f˜ : (X , x0 ) → (R, z0 )
making diagram in Fig. 2.16 commutative, i.e., satisfying the property p ◦ f˜ = f .
Then, f˜ is said to be lifting of f .

Remark 2.17.3 It is a natural question: Given a continuous map f : (X , x0 ) →

(S 1 , 1), if its lifting f˜ exists, is it unique? The answer is positive under suitable
situations given in Exercise 26 and Exercise 29 of Sect. 2.28.

Proposition 2.17.4 (Path Lifting Property) Given a loop α : (I, İ) → (S 1 , 1) on

S 1 , there exists a unique path α̃ : I → R such that p ◦ α̃ = α and α̃(0) = 0 as shown
in Fig. 2.17.

Proof Since the subspace I ⊂ R is convex and compact, corresponding to the given
α, there exists a unique lifting α̃ : I → R such that α(0) = 0 by using Exercise 29
of Sect. 2.28. 
2.17 Degree and Winding Number of a Loop on the Circle 91

Fig. 2.17 Path lifting α̃ of α

in R

Definition 2.17.5 Given a loop α : (I, İ) → (S 1 , 1), the winding number w(α) of
α is given in complex analysis by the integral
1 dz
w(α) = .
2π i α z

We now study the winding number from the viewpoint of homotopy theory.

Definition 2.17.6 The winding number of a closed curve α in the Euclidean plane
R2 around a given point a ∈ R2 is the total number of times that the curve α travels
anticlockwise around the point a and hence it is an integer and depends on orientation
of the curve. This number may be positive, 0 or negative. The winding number of a
loop α is the same as its degree by Definition 2.17.9.

Example 2.17.7 Given point a ∈ R2 , the space R2 − {a} is homotopy equivalent to

the circle S 1 . The set [f ] of homotopy classes of continuous maps f : S 1 → X based
at a point x0 of the topological space X is its fundamental group π1 (X , x0 ). The
fundamental group π1 (S 1 ) is isomorphic to Z (see Sect. 2.19.1), and the homotopy
class of a complex closed curve is precisely represented by its winding number. Given
a loop α : (I, İ) → (S 1 , 1), its winding number w(α) is given in complex analysis
by the integral
1 dz
w(α) = .
2π i α z

If p : R → S 1 , t → e2πit , then there exists a real-valued function α̃ such that

α = p ◦ α̃ as shown in Fig. 2.17. Taking z = α(t) = p ◦ α̃(t) = e2πiα̃(t) , it follows
that
dz = 2π ie2πiα̃(t) α̃  (t)dt = 2π iz α̃  (t)dt.

This shows that

1
1 dz
w(α) = = α̃  (t)dt = α̃(1) − α̃(0).
2π i α z 0

Example 2.17.8 If α : I → S 1 , t → e2πint , i.e., if α(t) = e2πint , then

α̃(t) = nt and w(α) = α̃(1) − α̃(0) = n.

92 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Geometrically, the map α : I → S 1 wraps the unit interval I around the circle S 1
anticlockwise n times if n ≥ 0 and clockwise |n| times if n < 0.

Definition 2.17.9 (Degree of a loop) Given a loop α : (I, İ) → (S 1 , 1) on S 1 , there

exists a unique path α̃ : I → R such that p ◦ α̃ = α and α̃(0) = 0. The winding
number w(α) defined by

w(α) = α̃(1) − α̃(0) = α̃(1)

is called the degree of α, denoted by deg α. It is an integer and is homotopy invariant

by Proposition 2.17.11.

Example 2.17.10 For the loop α : (I, İ) → (S 1 , 1), t → e2πnit , α̃(1) = n. Thus for
the map α : S 1 → S 1 , z → z n , deg α = n justifies term ‘ degree.’

Proposition 2.17.11 The degree of any loop α : (I, İ) → (S 1 , 1) is an integer, and
it is homotopy invariant.

Proof Let p : R → S 1 , t → e2πit be the exponential map. For the loop α : (I, İ) →
(S 1 , 1) if α̃ : I → R is its the unique lifting, then p ◦ α̃ = α. Hence, (p ◦ α̃)(1) =
α(1) = 1 implies that deg α = α̃(1) ∈ ker p = Z. Again homotopy lifting property
(HLP) asserts that if α, β : (I, İ) → (S 1 , 1) are two loops such that α  β rel İ,
then their corresponding unique liftings α̃, β̃ : I → R with α̃(0) = β̃(0) are such
that
α̃  β̃ rel İ and α̃(1) = β̃(1).

Since deg α = α̃(1), it follows by the above homotopy lifting property (see Exer-
cise 30 of Sect. 2.28) that the degree of any loop α : (I, İ) → (S 1 , 1) is homotopy
invariant. 

Corollary 2.17.12 If two differentiable functions

α, β : (I, İ) → (S 1 , 1)

are such that α  β rel İ, then they have the same winding numbers.

Proof It follows from Proposition 2.17.11. 

Proposition 2.17.13 characterizes homotopy of loops in terms of their winding

numbers.

Proposition 2.17.13 Let α, β : (I, İ) → (S 1 , 1) be any two loops. Then, the winding
numbers w(α) = w(β) iff α  β rel İ.

Proof Consider the exponential map

p : R → S 1 , t → e2πit .
2.17 Degree and Winding Number of a Loop on the Circle 93

If α  β rel İ, then it follows by Proposition 2.17.11 that w(α) = w(β). Conversely,
let w(α) = w(β) and α̃, β̃ : I → R be two liftings of α and β. Since I is convex, the
map

H̃ : I × I → R, (t, s) → (1 − s)α̃(t) + sβ̃(t)

is a well-defined and continuous map such that

H̃ : α̃  β̃ rel İ and hence H = p ◦ H̃ : α  β rel İ.


Corollary 2.17.14 Any loop (I, İ) → (S 1 , 1) is homotopic to the unique loop

αn : (I, İ) → (S 1 , 1), t → e2πint .

Proof Let α : (I, İ) → (S 1 , 1) be an arbitrary loop and w(α) = n be its winding
number around the origin (0, 0). Again, w(αn ) also moves n times around the origin
(0, 0). Hence, it follows from Proposition 2.17.13 that α  αn rel İ and for any
integer m = n, the loops α and αm cannot be homotopic rel İ. 

2.18 Vector Fields and their Applications

This section is devoted to the study of nonvanishing continuous vector fields on

spheres S n from the viewpoint of algebraic topology. On the other hand, Basic
Topology, Volume 2 studies vector fields on smooth manifolds from the view-
point of differential topology. Theorem 2.25.23 provides a necessary and sufficient
condition for existence of a nonvanishing vector field on S n . Vector fields establish
a close connection among geometry, analysis and topology.

2.18.1 Basic Concepts of Vector Fields

This subsection starts with basic concepts of vector fields on Euclidean space Rn .
Definition 2.18.1 A vector field on the Euclidean space Rn is a continuous vector-
valued function
v : Rn → Rm .

If x1 , x2 , . . . , xn are coordinates of any point x ∈ Rn , then the vector field v is

represented by m continuous real-valued functions

f : Rn → Rm , (x1 , x2 , . . . , xn ) → (f1 (x1 , x2 , . . . , xn ), . . . , fm (x1 , x2 , . . . , xn )),

94 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

in brief,
f : Rn → Rm , x → (f1 (x), . . . , fm (x)).

There are two natural problems:

(a) Does there exist a vector field tangent to a given surface?

(b) Has every continuous map f : X → X a fixed point for an arbitrary topological
space X ?

Proposition 2.18.2 Let A ⊂ R2 be a subset of the Euclidean plane R2 . A vector

field V (or VA ) on A is a continuous function which assigns to each point x ∈ A, a
vector v(x) in R2 with its initial point a ∈ A.

Remark 2.18.3 A vector field V on A ⊂ R2 may be intuitively thought as the velocity

of a particle on A under certain situations. As the essential components of a vector
v(x) are its length and direction, without loss of generality it is assumed that the
vector v(x) starts from the origin.

2.18.2 Vector Fields on Sn

This subsection conveys the concept of vector field on S n with its geometrical inter-
pretation.

Definition 2.18.4 A vector field on S n is a continuous map v : S n → Rn+1 (n ≥ 1)

such that the inner product < x, v(x) >= 0, ∀ x ∈ S n ; i.e., the vector v(x) is orthog-
onal to the vector v(x) in Rn+1 for every x ∈ S n . Moreover, if v(x) = 0 for all x ∈ S n ,
we say that the vector field is nonvanishing.

Remark 2.18.5 (Geometrical Interpretation) Definition 2.18.4 implies that a vector

field v on S n is a continuous map which assigns to every vector x of unit length in Rn+1 ,
a unit vector v(x) in Rn+1 such that x and v(x) are orthogonal, i.e., x ⊥ v(x), ∀ x ∈ S n .
If we consider the vector v(x) starting from the point x ∈ S n , then this vector v(x)
must be tangent to S n at each point x of S n . If x moves in S n , then end point of the
vector v(x) varies continuously in Rn+1 . Geometrically, Fig. 2.18 represents it for
the particular case, when n = 1.

Example 2.18.6 A continuous tangent vector field (or simply a vector field ) on
S 2 is an ordered pair (x, f (x)), where f : S 2 → R3 is a continuous map such that
f (x) is tangent to S 2 at the point x for every x ∈ S 2 and it is written simply by f .
This vector field f on S 2 is nonvanishing if f (x) = 0 for every x ∈ S 2 ; hence, a
nonvanishing (nonzero ) vector field f on S 2 is considered as continuous mapping
f : S 2 → R3 − {0}.
2.18 Vector Fields and their Applications 95

Fig. 2.18 Geometrical

interpretation of a vector
field on S 1
x
v(x)
y

v(y)

Definition 2.18.7 A continuous map An : S n → S n is said to be antipodal map if

An (x) = −x, ∀ x ∈ S n , where for any x = (x1 , x2 , · · · xn , xn+1 ) ∈ S n , its antipode −x
is given by
−x = (−x1 , −x2 , · · · , −xn , −xn+1 ) ∈ S n .

Theorem 2.18.8 provides a necessary and sufficient condition for existence of a

nonvanishing vector field on S n in terms of homotopy.

Theorem 2.18.8 Let An : S n → S n , x → −x be the antipodal map and 1S n : S n →

S n be the identity map. Then, there exists a continuous nonvanishing tangent vector
field on S n iff 1S n  An .

Proof First suppose that there exists a continuous nonvanishing tangent vector field

v : S n → Rn+1

on S n . Define a continuous map

F : S n × I → S n , (x, t) → (1 − 2t)x + 2 t − t 2 v(x)/||v(x)||.

Hence, it follows that F : 1S n  An . For the converse part, let a homotopy F : 1S n  A

can be approximated by a differentiable homotopy. This produces tangent curve
elements to S n . Since the tangent line to the curve βx : I → S n , t → F(x, t) at t = 0
contains a unit vector in the direction of increasing t ∈ I, it follows that this tangent
line is tangent to S n at x ∈ S n and it is nonvanishing. 

Proposition 2.18.9 There does not exist antipodal map f : S 2 → S 2 with the prop-
erty that x and f (x) are orthogonal for every x ∈ S 2 .

Proof Let f : S 2 → S 2 be a continuous map; for example, x and f (x) are orthogonal
vectors for every x ∈ S 2 . Suppose there is some point y ∈ S 2 for which f (y) = −y.
Then, the inner product < y, f (y) >= −1, and hence the vectors y and f (y) cannot
be orthogonal, which gives a contradiction, because by hypothesis < y, f (y) >= 0.
This contradiction proves that there does not exist antipodal map f : S 2 → S 2 with
the property that x and f (x) are orthogonal for every x ∈ S 2 . 
96 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proposition 2.18.10 For all odd integers n ≥ 1, there is a nonvanishing vector field
f : S n → S n.
Proof Let n be an odd integer. Define a continuous map

f : S n → S n , x = (x1 , x2 , . . . , xn+1 ) → (x2 , −x1 , x4 , −x3 , . . . , xn+1 , −xn ).

Then, the inner product

< x, f (x) >= (x1 x2 − x1 x2 ) + (x3 x4 − x3 x4 ) + · · · + (xn xn+1 − xn xn+1 ) = 0, ∀ x ∈ S n .

This shows that f : S n → S n is a continuous map such that the vectors x and f (x)
are orthogonal and hence it is a vector field on S n . 
Remark 2.18.11 For converse of Proposition 2.18.10, see Brouwer–Poincaré The-
orem 2.25.23.
Proposition 2.18.12 If v : S 2n+1 → R2n+2 is a nonvanishing tangent vector field on
S 2n+1 , then the map
v(x)
f : S 2n+1 → S 2n+1 , x →
||v(x)||

is homotopic to the identity map.

Proof Consider the map
πt v(x) πt
H : S 2n+1 × I → S 2n+1 , (x, t) → x cos + sin .
2 ||v(x)|| 2

Then, H is a homotopy between the identity maps on S 2n+1 and f .


Remark 2.18.13 Let fn : S n → S n be the antipodal map. If there exists a tangent
vector field on S n , then fn  1d , where 1d is the identity map on S n .
Proof Construct a continuous map

H : S n × I → S n , (x, t) → H (x, t) = α(t)x + β(t)fn (x) : ||H (x, t)||2 = 1,

where α and β are real-valued functions of t. This determines the equation

α(t)2 + β(t)2 fn (x), f (x) = 1.

Choose α(t)√ = 1 − 2t. Then, β(t) = 4(t − t )/||fn (x)|| , since fn (x) = 0,
2 2 2
√ gives
β(t) = 2 t − t /||fn (x)||.
2 Consequently, H (x, t) = (1 − 2t)x + 2 t − t 2
fn (x)/||fn (x)|| shows that H : 1d  fn , which implies that fn  1d . 
Remark 2.18.14 The following basic results on vector fields on n-sphere S n are also
proved in this chapter.
2.18 Vector Fields and their Applications 97

(i) Brouwer–Poincaré Theorem 2.25.23 asserts that the n-sphere S n admits a con-
tinuous nonvanishing vector field iff n is odd.
(ii) For all even integers n > 1, there is no vector field f : S n → S n proved in Corol-
lary 2.25.24.

2.18.3 Applications of Vector Fields

This subsection proves hairy ball and Brouwer fixed-point theorems by using the
concept of vector fields. A hair is said to be combed if it is kept flat, which means
mathematically that it is tangent to the sphere. To establish a criterion that a continu-
ous map from a topological space to itself must have a fixed point is one of the most
important applications of topology in mathematics and other sciences.
Historically, L. E. J. Brouwer (1881–1967) first proved during 1910–1912 that every
continuous map f : Dn → Dn , n ≥ 1 has a fixed point by using the concept of degree
of spherical maps f : S n → S n . This theorem is subsequently proved by using either
the homology or homotopy groups. But Brouwer used neither of them, which had
not been invented at that time. This theorem named after L. E. J. Brouwer is applied in
Corollary 2.23.8 to determine the nature of eigenvalues of a special class of matrices.

Remark 2.18.15 The absence of any nonvanishing vector field on S 2 presents a

popular result called hairy ball theorem 2.18.16 saying that the hair on S 2 cannot
be combed flat . So, this theorem is called ‘hairy ball theorem.’ It is proved by the
concept of nonzero vector field on S 2 .

Theorem 2.18.16 (Hairy Ball Theorem) There is no nonzero vector field on S 2 .

Proof Suppose that there is a nonzero vector field v : S 2 → R3 on S 2 . Then, v is a

continuous map such that v(x) = 0, ∀ x ∈ S 2 and the vector v(x) is tangent to S 2 at
x for every x ∈ S 2 , since by hypothesis v is a nonzero vector field. Consider the map

f : S 2 → S 2 ⊂ R3 , x → v(x)/||v(x)||.

Clearly, the map f is continuous. Since f (x) is tangent to S 2 at the point x for every
x ∈ S 2 , this produces a contradiction by Proposition 2.18.9. This contradiction proves
the theorem.
Geometrical interpretation of the proof : If one imagines that he has a hair
growing out from each point on the surface of a ball, then it is not possible to brush
them flat. Otherwise, the tangent vectors to the hairs would show that S 2 would admit
a continuous nonvanishing vector field. As it would contradict Proposition 2.18.9,
we conclude that a hairy ball cannot be combed flat. 

Definition 2.18.17 A continuous map v : D2 → R2 is said to be a vector field on

D2 . It is said to be nonvanishing if v(x) = 0, ∀ x ∈ D2 .
98 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proposition 2.18.18 Let v be a nonvanishing vector field on 2-dimensional disk D2 .

Then, there exist points x, y ∈ ∂ D2 such that

v(x) = rx, v(y) = −sy for some real numbers r, s > 0.

Proof Let v : D2 → R2 be a vector field such that there exists no point x ∈ ∂D2 with
v(x) = −sx, s > 0, i.e., v(x) points directly inward. Let f = v|∂ D2 be the restriction
map of v to ∂ D2 . Then, f has a continuous extension

f˜ : D2 → R2 − {0}.

Hence, it follows that f is nullhomotopic. Again, f is homotopic to the inclusion

map
i : S 1 = ∂ D2 → R2 − {0}

by a homotopy H : ∂D2 × I → R2 − {0} : (x, t) → tx + (1 − t)f (x). Clearly,

H (x, t) = 0 for t = 0, 1. If H (x, t) = 0 for some t ∈ (0, 1), then the equality
f (x) = − 1−t
t
x would hold. This would imply that the point f (x) points outward
at x, which contradicts our supposition. For the second part, consider −v for v. 

Proposition 2.18.18 is now used to prove Brouwer fixed-point theorem 2.18.19

for dimension 2. It is named after L. E. J. Brouwer (1881–1967). Its alternative proof
is given in Theorem 2.23.6 by showing that S 1 ⊂ D2 not a retract of D2 .

Theorem 2.18.19 (Brouwer Fixed-Point Theorem for Dimension 2) Every contin-

uous map f : D2 → D2 has a fixed point.

Proof If possible, f : D2 → D2 has no fixed point. Then, f (x) = x, ∀ x ∈ D2 . Con-

sider the map
v : D2 → R, x → f (x) − x.

Then, v defines a nonvanishing vector field on D2 . Clearly, v cannot point directly

outward at any point x ∈ ∂ D2 , because this would assert that there exists some
r > 0 such that f (x) − x = rx, and hence f (x) = (r + 1)x ∈
/ D2 . But this contradicts
Proposition 2.18.18. This contradiction proves the theorem. 

Remark 2.18.20 For more study on vector fields, and Brouwer fixed-point theorem,
see Chap. 3.

2.18.4 Link Between Homotopy and Flow

This subsection continues the study of vector fields by introducing the concept of
flow, which is a generalization of the concept of vector field. It establishes an interplay
between flow and homotopy. Let X be a topological space and ψt : X → X ( t ∈ R)
 2.18 Vector Fields and their Applications 99

be a family of continuous maps. By continuity of ψt , it is meant jointly continuous

in t and x.
Definition 2.18.21 A continuous family ψt : X → X (t ∈ R) of maps is said to be
a flow if
(i) ψ0 = 1, i.e., ψ0 (x) = x, ∀ x ∈ X .
(ii) ψt is a homeomorphism for all t ∈ R.
(iii) ψt ◦ ψs = ψt+s , i.e., ψt (ψs (x)) = ψt+s (x), ∀ x ∈ X and ∀ t, s ∈ R.
A flow is also known as a one parameter group of homeomorphisms.
Remark 2.18.22 The first condition asserts that the map ψ0 is the identity map. The
third condition defining property of a flow on X implies that the family {ψt } of flows
on X forms a group isomorphic to the additive group of real numbers. It asserts that
if a ball rolls for s seconds and then it rolls for t seconds, the result says that the
ball rolls for s + t seconds. The concept of flows plays a key role in topology and
differential equations [Hirsch et al., 1974].
A flow ψt on a topological space X is sometimes considered as a continuous map

ψ : X × R → X , (x, t) → ψt (x).

This motivates to link the concept of flow with homotopy in Proposition 2.18.23.

Proposition 2.18.23 Given a topological space X , every flow ψt : X → X is homo-

topic to IX .
Proof Define the map

F : X × I → X , (x, s) → ψ(x, (1 − s)t).

This map is continuous and is such that

F(x, 0) = ψ(x, t) = ψt (x)

and

F(x, 1) = ψ(x, 0) = ψ0 (x) = 1X (x), ∀ x ∈ X , ∀ t ∈ R.

This asserts that ψt is homotopic to 1X under F. 

2.18.5 Topological Dynamics

This subsection initiates a study of topological dynamics by using the concept of

a flow, which establishes a close connection between topology and analysis. The
100 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

concept of flow on a given topological space X is an extension of the flows in R2

arising from vector fields. The tools of vector fields and differential equations can be
used as tools in topology, facilitating topological ideas to penetrate in the premises
of differential equations. Topological dynamics is the study of flows and gives an
abstract form of differential equations. The study of vector field is closely related to
the study of differential equations

dx
= f (x, y)
dt

dy
= g(x, y).
dt
Remark 2.18.24 The three basic concepts in topological dynamics such as contin-
uous functions, vector fields and system of differential equations are closely inter-
linked, because vector fields and differential equations are used as tools in topology
enroading their entry into the premises of topology. For example, vector fields are
used to prove Brouwer fixed-point theorem, which is a classical result in topol-
ogy (see Theorem 2.18.19), and in more general, fixed-point theorem on compact
manifold (see Exercise 64 of Section 2.28) is also an important result of fixed-point
theory. Fixed-point theorems facilitate to study a given system of equations in several
variables.
There is a natural question: Whether a given system of equations of the form
f (x) = 0 has a solution or not, where f (x) = 0 is a polynomial equation or a com-
plicated equation of several variables. If g(x) = f (x) + x and g(x) has a fixed point,
then this fixed point will be a solution of the given equation f (x) = 0.
Definition 2.18.25 For a given p = (x, y) ∈ A ⊂ R2 , and a map v : A → R2 , its
image v(p) can be expressed as

v(x, y) = (U (x, y), V (x, y)),

where U and V are real-valued continuous functions of x and y.

A vector field v on a subset A ⊂ R2 is said to be continuous if the function

v : A → R2 , (x, y) → (U (x, y), V (x, y))

is continuous.
Example 2.18.26 implies that the study of vector fields v on a set A ⊂ R2 is
precisely the study of continuous transformations f : A → R2 .
Example 2.18.26 Let A ⊂ R2 and v : A → R2 be a map such that f : A → R2 and
f (x, y) is the end point of the vector v(x, y). Then,

f (x, y) = (x, y) + v(x, y) = (x + U (x, y), y + V (x, y)). (2.2)

 2.18 Vector Fields and their Applications 101

Hence, it follows that f is continuous, if v is a continuous vector field. Conversely,

given A ⊂ R2 and a transformation f : A → R2 , the vector field v is defined by taking
v(a) to be the vector from the point a ∈ A to the point f (a). Then, v is continuous if
f is continuous.

Example 2.18.27 Some important examples of vector fields.

(i) The force field arising from gravitation is a vector field.
(ii) Electromagnetism is a vector field.
(iii) Velocity vector of a moving fluid is a vector field.
(iv) Pressure gradient on a weather map is a vector field.

A vector field v(x, y) = (U (x, y), V (x, y)) gives a system of differential equa-
tions in two unknowns x and y representing the position of a moving particle in R2 ,
which depends on time t. This produces a system of differential equations

dx/dt = U (x, y), dy/dt = V (x, y). (2.3)

The system of Eq. (2.3) is said to be autonomous, if the functions U (x, y) and
V (x, y) do not depend on t. A solution of the system of Eq. (2.3) involves two
functions in x and y expressed in terms of t. These are the parametric equations of a
path in R2 . This idea may be applied to study the orbit space of a planet in Newtonian
mechanics or the orbit of a molecule or the path of a molecule of a fluid. The original
vector field v(x, y) conveys the tangent vector to the path of the particle at the point
(x, y) ∈ R2 .

Example 2.18.28 (i) The vector field v(x, y) = (1, y) has the exponential curves
given by y = cex as the solution curves, since here

dx/dt = U (x, y) = 1, dy/dt = V (x, y) = y =⇒ x = t + c1 , y = c2 et .

(ii) The vector field v(x, y) = (x, −y) has the rectangular hyperbolas given by xy = c
as the solution curves, since here dx/dt = U (x, y) = x, dy/dt = V (x, y) = −y
and hence x = c1 et , y = c2 e−t .
(iii) The vector field v(x, y) = (y, x) has the hyperbolas given by x2 − y2 = c as
the solution curves, since here dx/dt = U (x, y) = y, dy/dt = V (x, y) = x and
hence d2 x/dt 2 = dy/dt = x, d2 y/dt 2 = dx/dt = y. Consequently, x = c1 et +
c2 e−t , y = c1 et + c2 e−t .

2.19 Computation of the Fundamental Groups

This section computes the fundamental groups of the circle S 1 , punctured Euclidean
plane R2 − {(0, 0)} and figure-eight F8 which are used in our subsequent study.
102 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

2.19.1 Computation of the Fundamental Group of the Circle

This subsection computes the fundamental group π1 (S 1 , 1) of the circle by utilizing

the degree function. As the circle is a path-connected space, its fundamental group
does not depend on its base point. So without loss of generality, 1 ∈ S 1 is taken as
its base point. Its fundamental group π1 (S 1 ) is isomorphic to Z, the infinite cyclic
group by Corollary 2.19.2.

Theorem 2.19.1 The degree function ψ : π1 (S 1 , 1) → Z, [α] → deg α is a group

isomorphism.

Proof The map

ψ : π1 (S 1 , 1) → Z, [α] → deg α

is independent of the choice of the representative of the class [α] ∈ π1 (S 1 , 1) by

Proposition 2.17.11, and hence, the map ψ is well defined. Moreover, it clearly
follows that ψ is a homomorphism and also a monomorphism. Finally, we claim that
ψ is an epimorphism. To prove it, given an integer n, define a loop α : I → S 1 , t →
e2πint . Then, the path α̃ : I → R, t → nt is such that α̃(0) = 0 and lifts the path α.
Hence, the path α̃ is unique and α̃(1) = n = deg α. Consequently, it follows that the
map ψ is a group isomorphism. 

Corollary 2.19.2 π1 (S 1 ) ∼
= Z.

Proof Since the circle S 1 is path connected, the fundamental group π1 (S 1 ) does not
depend on the choice of the base point of S 1 . This implies by Theorem 2.19.1 that
π1 (S 1 ) = π1 (S 1 , 1) ∼
= Z. 

Corollary 2.19.3 The circle S 1 is not simply connected.

Proof It follows from Corollary 2.19.2. 

Remark 2.19.4 Geometrical interpretation of the result π1 (S 1 , 1) ∼ = Z: Let α :

I → S 1 be a loop based at the point 1 ∈ S 1 . Then, if t travels from 0 to 1, its image
point α(t) travels on the circle and at t =1, it returns to this base point. This shows that
the total number of times for its complete winding the circle is an integer. If the loops
α, β on S 1 based at the point 1 are such that α describes S 1 m times and β describes
it n times. With m > n, then the loops α, β cannot be homotopic; otherwise, α ∗ β
would be a path describing (m − n) times the circle S 1 . But such a path cannot be
a null path. This implies that assigning to each loop the number of times it winds
around S 1 establishes a bijective correspondence between the set of homotopy classes
of loops in S 1 with the set of integers Z, which is also an isomorphism between the
groups π1 (S 1 , 1) and Z.

Theorem 2.19.5 The circle S 1 ⊂ D2 is not a retract of D2 .

2.19 Computation of the Fundamental Groups 103

Proof Let 1S 1 : S 1 → S 1 be the identity map and i : S 1 → D2 be the inclusion map.

If S 1 is a retract of D2 , there is a retraction r : D2 → S 1 such that the equality
i∗
r ◦ i = 1S 1 holds. This asserts that the composite homomorphism π1 (S 1 , 1) −→
r∗
π1 (D2 , 1) −→ π1 (S 1 , 1) is the identity automorphism of π1 (S 1 , 1) = Z. This is
impossible, since D2 is contractible and hence π1 (D2 , 1) = 0. 
Corollary 2.19.6 There exists no retraction r : D2 → S 1 .
Proof If possible, there exists a retraction r : D2 → S 1 . Then, S 1 is a retract of D2 .
But it contradicts Theorem 2.19.5. 
Corollary 2.19.7 The identity map 1S 1 : S 1 → S 1 has no continuous extension over
D2 .
Proof If 1S 1 has a continuous extension r : D2 → S 1 , then its restriction to S 1 is the
identity map 1S 1 . This implies that r ◦ i = 1S 1 and hence S 1 is a retract of D2 . But
this contradicts Theorem 2.19.5 that asserts S 1 is not a retract of D2 . 
Remark 2.19.8 The homotopy class of a loop α : (I, İ) → (S 1 , 1) is completely
determined by deg α in Theorem 2.19.9.
Theorem 2.19.9 Let α, β : (I, İ) → (S 1 , 1) be two loops. Then, α  β rel İ if and
only if deg α = deg β.
Proof For the loops α  β rel İ, it follows that α̃  β̃ rel İ with α̃(1) = β̃(1). This
shows that deg α = deg β. Again, since the degree function ψ is injective, it follows
that if deg α = deg β, then [α] = [β], and hence α  β rel İ. 

2.19.2 Computation of the Fundamental Group of the

Punctured Plane

This subsection computes the fundamental group of the punctured Euclidean plane
by utilizing its special homotopic properties. It is an important geometrical object
for the study of topology.

Theorem 2.19.10 Let X = R2 − {0} be the punctured Euclidean plane, where 0

stands for (0, 0) ∈ R2 and x0 ∈ S 1 ⊂ X be an arbitrary point. Then, the homomor-
phism

i∗ : π1 (S 1 , x0 ) → π1 (X , x0 )

induced by the inclusion map

i : (S 1 , x0 ) → (X , x0 )

is an isomorphism.
104 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proof Define a continuous map

x
r : X → S 1 , x → .
||x||

Geometrically, the map r collapses each radial ray in X onto the point where the
ray intersects S 1 , and in particular, it maps every point x ∈ S 1 onto itself. As the
composite map r ◦ i
i r
(S 1 , x0 ) −
→ (X , x0 ) −
→ (S 1 , x0 )

is the identity map 1S 1 on S 1 , its induced homomorphism

(r ◦ i)∗ = r∗ ◦ i∗ = 1∗S 1 : π1 (S 1 , x0 ) → π1 (S 1 , x0 )

is the identity automorphism of π1 (S 1 , x0 ) by the functorial property of π1 and hence

i∗ is a monomorphism. Again, given a loop α in X based at the point x0 ,

α(t) α(t)
i ◦ r ◦ α = β : I → X , t → (i ◦ r)(α(t)) = i( )=
||α(t)|| ||α(t)||

is also a loop in X based at the point x0 . Define a map

α(t)
H : I × I → X , (t, s) → s + (1 − s)α(t).
||α(t)||

Clearly, H (t, s) = 0 for every (t, s) ∈ I × I and hence it is well defined. Moreover,
it is a continuous map such that H : α  β rel İ. This asserts that

(i∗ ◦ r∗ )([α]) = [i ◦ r ◦ α] = [β] = [α], ∀ [α] ∈ π1 (X , x0 ).

It implies that i∗ ◦ r∗ is the identity automorphism of π1 (X , x0 ) and hence i∗ is an

epimorphism. Consequently, i∗ is an isomorphism with r∗ as its inverse. 

Theorem 2.19.11 π1 (R2 − {0}) ∼

= Z.

Proof Let X = R2 − {0} be the punctured Euclidean plane and x0 ∈ S 1 ⊂ X be an

arbitrary point. Then, the inclusion map

i : (S 1 , x0 ) → (X , x0 )

induces an isomorphism

i∗ : π1 (S 1 , x0 ) → π1 (X , x0 )

by Theorem 2.19.10. Define a map

2.19 Computation of the Fundamental Groups 105

x
H : X × I → X , (x, t) → (1 − t)x + t .
||x||

Then, H (x, t) = 0 for every (x, t) ∈ X × I and hence it is well defined. Moreover,
H is a continuous map satisfying the properties:

x
H (x, 0) = x, ∀ x ∈ X , H (x, 1) = ∈ S 1 , ∀ x ∈ X , H (a, t) = a, ∀ a ∈ S 1 .
||x||

This asserts that S 1 is a strong deformation retract of R2 − {0} and hence by Propo-
sition 2.12.2 π1 (S 1 , x0 ) ∼
= π1 (R2 − {0}, x0 ) ∼
= Z. As R2 − {0} is path connected, it
follows that its fundamental group is independent of the choice of its base point x0
and hence it follows that π1 (R2 − {0}) ∼ = Z. 

Corollary 2.19.12 The punctured Euclidean plane R2 − {0} is not simply con-
nected.

Proof It follows from Theorem 2.19.11. 

Example 2.19.13 The circle S 1 is not simply connected by Corollary 2.19.3. On the
other hand, the n-sphere S n is simply connected for all n > 1 (see Theorem 2.26.7).

2.19.3 Computation of Fundamental Group of Figure-Eight

F8 and a Wedge of Circles

This subsection computes the fundamental group of figure-eight F8 , which is the

union of two circles in the plane with one point in common. Its fundamental group
is nonabelian, which is different from fundamental groups computed in previous
subsections, because they are abelian. Its computation by another method using the
concept of covering spaces is given in Chapter 5.

Remark 2.19.14 Theorem 2.19.15 proves that the fundamental group of figure-eight
F8 is not abelian and Corollary 2.19.18 asserts that this fundamental group is a free
group on two generators.

Definition 2.19.15 Let F8 ⊂ S 1 × S 1 be the subspace of the product space S 1 × S 1

defined by

F8 = {(u, v) ∈ S 1 × S 1 : u = (1, 0) or v = (1, 0)},

which is the union of two circles in the plane touching each of other at one point. F8
is known as figure-eight.
106 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Theorem 2.19.16 proves by using Van Kampen theorem that the fundamental
group of figure-eight is nonabelian. The computation of this group by two other
methods is also available in Chapter 5.

Theorem 2.19.16 The fundamental group of figure-eight is nonabelian.

Proof For the figure-eight F8 , let x0 be the common point of the two circles and
x1 , x2 be two points taking one point from each circle such that x1 = x0 = x2 . Then,
the spaces

(F8 − x1 , x0 )  (S 1 , s0 ) and (F8 − x2 , x0 )  (S 1 , s0 ),

for an arbitrary point s0 ∈ S 1 . Moreover,

((F8 − x1 , x0 ) ∩ (F8 − x2 , x0 ), x0 )  (s0 , s0 ).

Hence by using Van Kampen theorem (see Exercise 23 of Sect. 2.28 and see Chapter
6 for its detailed proof in another form by using graph theory), it follows that

π1 (F8 , x0 ) ∼
= Z ∗{1} Z,

which is the amalgamated product of the groups Z and Z over the trivial group {1}
(see Chapter 1). Hence, the group π1 (F8 , x0 ) is a free group having two generators.
If we take the closed paths α at x0 going once around the one circle C1 and another
closed path β based at x0 going once around the other circle C2 , then their homotopy
classes can be taken as generators. As α and β encircle different holes, their product
closed path α ∗ β ∗ α −1 cannot be deformed into β. This asserts that

[α] ∗ [β] ∗ [α −1 ] = [β] =⇒ [α] ∗ [β] = [β] ∗ [α].

This proves that the group π1 (F8 , x0 ) is not abelian. 

Theorem 2.19.17 Let X be a Hausdorff space such that X = X1 ∪ X2 , X1 ∩ X2 =

{x0 } and X1 and X2 are both homeomorphic to the circle S 1 . Then, π1 (X , x0 ) is a free
group on two generators and it is not abelian.

Proof Let x1 ∈ X1 and x2 ∈ X2 be two distinct points. Then, both π1 (X1 , x1 ) and
π1 (X2 , x2 ) are infinite cyclic groups. Hence by Theorem 2.19.16, it follows π1 (X , x0 )
is a free group on two generators by taking one generator the homotopy class of the
closed path α which goes once around the circle homeomorphic to X1 and the other
generator is the homotopy class of the closed path β which goes once around the
circle homeomorphic to X2 . 

Corollary 2.19.18 The fundamental group of the figure-eight is a free group on two
generators.
2.19 Computation of the Fundamental Groups 107

Proof It follows from Theorem 2.19.17 that the fundamental group of the figure-
eight F8 is a free group on two generators with one generator the homotopy class of
the closed path α which goes once around the one circle S11 and the other generator is
the homotopy class of the closed path β which goes once around the other circle S21 . 

Remark 2.19.19 Theorem 2.19.20 gives a generalization of Theorem 2.19.17 by

n
induction. It computes the fundamental group of one-point union of n circles Si1
i=1
n
(wedge ). On the other hand, the homology groups of Si1 (wedge) are computed
i=1
in Chapter 6.

Theorem 2.19.20 Let X be a Hausdorff space such that

X = X1 ∪ X2 ∪ · · · ∪ Xn : for i = j, Xi ∩ Xj = {x0 },

where each Xi with subspace topology is homeomorphic to the circle S 1 ( X is called

n n
the wedge of the n circles Xi denoted by Xi or Si1 ). Then, π1 (X , x0 ) is a free
i=1 i=1
group on n generators and it is not abelian.

Proof Use induction on n. For n = 2, it is proved in Corollary 2.19.18. Take a point

xi ∈ Xi such that xi = x0 . Then, π1 (X , x0 ) is a free group on n generators and it is not
abelian. Moreover, the homotopy class of each loop αi is a generator and the loops
{αi : i ∈ A} form a representative system of the free generators for the fundamental
group π1 (X , x0 ).

U = X − {xn } and V = X − {x1 , x2 , . . . , xn }

are both open sets such that

(i) X1 ∪ X2 ∪ · · · ∪ Xn−1 is a deformation retract of U.
(ii) Xn is a deformation retract of V .
(iii) U ∩ V is contractible.
Hence, the corollary follows from Theorem 2.19.16. 

Theorem 2.19.21 gives a further a generalization of Theorem 2.19.20 by taking X

as the wedge of the circles {Xi : i ∈ A} with Xi having the subspace topology of X .

Theorem 2.19.21 Let X be a Hausdorff space such that

(i)
X = Xi
i∈A
108 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

(ii) 
Xi = {x0 }
i∈A

where each Xi with subspace topology is homeomorphic to the circle S 1 (X is called

the wedge of the circles Xi ). Then, π1 (X , x0 ) is a free group. Moreover, if αi is a loop
in Xi , then its homotopy class is a generator of π1 (X , x0 ) and {αi : i ∈ A} forms a
representative system of free generators for the fundamental group π1 (X , x0 ).

Proof Proceed as in Theorem 2.19.17. 

Remark 2.19.22 If (X , x0 ) is a pointed topological space such that its fundamental

group π1 (X , x0 ) is nonabelian, then there exists no multiplication on X making the
space X a topological group and even such a space cannot be equipped with the
structure of an H -space structure or an H -group, because of Theorem 2.15.7 and
Proposition 2.15.8. This implies by Theorem 2.19.16 that there does not exist any
continuous multiplication on the topological space figure-eight F8 admitting it a
topological group structure or an H -space or an H -group structure.

2.20 Higher Homotopy Groups

The fundamental group of a topological space studies the properties of loops under
continuous deformation. It characterizes the topological properties of a space in
which the loops are defined. The study of these properties exhibited by loops in
dimension higher than two fails. This failure motivates to define higher-dimensional
analog of the loop (one-dimensional), leading to the concept of higher homotopy
groups. This section studies higher homotopy groups πn (X , x0 ), for integers n > 1,
of a pointed topological space (X , x0 ). The concept of higher homotopy groups of a
pointed space is a natural generalization of the concept of fundamental group of the
pointed space defined in Sect. 2.16. H. Hurewicz in his paper [Hurewicz, 1935] chose
spheres directly (in place of loops) to define higher homotopy groups formulated in
Definition 2.20.8 and classified continuous maps f from the n-sphere S n :

f : Sn → X

to an arbitrary topological space X up to homeomorphism. There is an alternative

approach to define higher homotopy groups starting with n-cubes In . It is formu-
lated in Proposition 2.20.11. In topology, homotopy groups are applied to classify
topological spaces; on the other hand, they are used in physics to classify continu-
ous maps. The relative homotopy groups πn (X , A, x0 ) are natural generalization of
πn (X , x0 ) (see Chap. 5). Higher homotopy groups carry special importance in the
study of bundles (see Chap. 5).
2.20 Higher Homotopy Groups 109

2.20.1 Introductory Concepts in Higher Homotopy Groups

πn (X, x0 )

This subsection introduces the basic concepts in higher homotopy groups πn (X , x0 ),

for integers n > 1, needed for our further study of algebraic topology.

Definition 2.20.1 Let X be a topological space and A be a subspace of X . Then, the

pair (X , A) is said to be a topological pair of spaces. If A = ∅, the pair (X , ∅) is
identified with the space X . If A = {x0 }, the pair (X , x0 ) is identified with the pointed
topological space X having base point x0 .

Example 2.20.2 If s0 ∈ S n , then (S n , s0 ) is a pointed topological space. On the other

hand, (Dn+1 , S n ) is a pair of spaces, where the n-sphere S n is the boundary of the
(n + 1)-disk Dn+1 in Rn+1 .

Definition 2.20.3 Let (X , A) and (Y , B) be two pairs of topological spaces. A map

f : (X , A) → (Y , B) is said to be continuous if f : X → Y is continuous and
f (A) ⊂ B.

Remark 2.20.4 The homotopy of continuous maps between two topological spaces
is now extended to pairs of topological spaces by using the usual convention of
writing (X × I, A × I) to represent the product space (X , A) × I.

Definition 2.20.5 Let (X , A) and (Y , B) be two pairs of topological spaces and f , g :

(X , A) → (Y , B) be two continuous maps such that for a subset X  ⊂ X , f (x ) =
g(x ), ∀ x ∈ X  . Then, they are homotopic relative to X  , symbolized, f  g rel X  ,
if ∃ a continuous map
H : (X × I, A × I) → (Y , B)

such that H (x, 0) = f (x) and H (x, 1) = g(x), ∀ x ∈ X and H (x , t) = f (x ) = g(x )
for every x ∈ X  and for every t ∈ I. It is expressed by the symbol H : f  g rel X  .
In particular, if X  = ∅, then the term relative to X  is not used.

Definition 2.20.6 Given a pointed topological space (X , x0 ), the topological space

F n (X , x0 ) is defined to be the space of all continuous maps f : (In , ∂ In ) → (X , x0 )
endowed with compact open topology.

Remark 2.20.7 Motivation of Hurewicz original Definition 2.20.8 of the absolute

group πn (X , x0 ).
This definition was given by Hurewicz (1904–1956) in 1935. He studied the groups
πn (S n , s0 ) during 1935–36 by considering elements of πn (S n , s0 ) as the homotopy
classes of continuous maps f : (S n , s0 ) → (X , x0 ), where S n is a fixed n-sphere. Let
S n be oriented by a choice of a generator τn of the homology group Hn (S n ) (see
Chap. 3) and take a continuous map g : (In , ∂ In ) → (S n , s0 ) such that g∗ (un ) = τn ,
110 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

where un is a generator of πn (X , x0 ) and g maps In − ∂ In topologically onto S n − s0 .

Then, the composite map

f ◦ g : (In , ∂ In ) → (X , x0 ) ∈ F n (X , x0 ).

If h is the second map having the properties of g, then g∗ (un ) = h∗ (un ). It implies that
g  h, since they represent the same generator of πn (X , x0 ). This implies that the
homotopy class of the composite map f ◦ g depends only on f and hence f represents
a unique element c(f ) in πn (X , x0 ). Conversely, given k ∈ F n (X , x0 ) the continuous
map f = k ◦ g −1 is single-valued and hence any element of πn (X , x0 ) is a c(f ). This
assignment sets up a 1-1 correspondence between the elements of πn (X , x0 ) and the
homotopy classes of continuous map f : (S n , s0 ) → (X , x0 ). It is pointed out that
c(f ) depends on orientation of S n : Its other orientation reverses the sign of c(f ). Its
generalization for higher homotopy groups is given in Chapter 5.

Definition 2.20.8 Let (X , x0 ) be a pointed space and (S n , s0 ) be the n- sphere with

base point s0 . Then, the set homotopy classes of continuous maps f : (S n , s0 ) →
(X , x0 ), abbreviated by the set πn (X , x0 ) = [(S n , s0 ); (X , x0 )], admit an abelian
group structure for all n > 1, called the n-th homotopy group of the pointed space
(X , x0 ), denoted by πn (X , x0 ).

The group structure of πn (X , x0 ) can be equally well defined in a way analogous

to that of the fundamental group. Let In be the unit n-cube and ∂ In = I˙n be its
boundary defined as follows

In = {t = (t1 , t2 , . . . , tn ) : 0 ≤ ti ≤ 1, i = 1, 2, . . . , n};

I˙n = ∂ In = {(t1 , t2 , . . . , tn ) ∈ In : ti = 0 or ti = 1 for atleast one i}.

Consider the set n (X , x0 ) of all continuous maps f : (In , ∂ In ) → (X , x0 )

endowed with the compact open topology. Then, n (X , x0 ) consists of all continuous
maps
f : In → X : f (t) = x0 , ∀ t ∈ ∂ In .

f (In ) is called an n-loop in X based at the point x0 ∈ X . In view of identification of

the points on the boundary of In , these n-loops become topologically equivalent to
n-spheres S n .
Definition 2.20.9 Two continuous maps f , g ∈ n (X , x0 ) are said to be homotopic
relative to the subspace ∂ In , written as f  g rel ∂ In if there exists a continuous map

G : In × I → X : G(t, 0) = f (t), G(t, 1) = g(t), ∀ t ∈ In and G(t, s) = x0 , ∀ t ∈ ∂ In , ∀ s ∈ I.

This homotopy is an equivalence relation ∼ between n-loops on n (X , x0 ), the

class corresponding to the n-loop f is as usual denoted by [f ], and the corresponding
2.20 Higher Homotopy Groups 111

quotient set n (X , x0 )/ ∼ is denoted by πn (X , x0 ). It can be endowed with a group

structure by defining a product f ∗ g of f and g connecting along a common part of
boundaries as formulated in Definition 2.20.10.

Definition 2.20.10 A composition ∗ is defined on the set n (X , x0 ) by the rule:

 
f (2t1 , t2 , . . . , tn ), 0 ≤ t1 ≤ 1/2
(f ∗ g)(t) = ∀f ,g ∈ n (X , x0 ) , t = (t1 , t2 , . . . , tn ) ∈ In
g(2t1 − 1, t2 , . . . , tn ), 1/2 ≤ t1 ≤ 1
(2.4)

f ∗ g is continuous and hence f ∗ g ∈ n (X , x0 ). Consequently, the composition

∗ is well defined.

Proposition 2.20.11 πn (X , x0 ) is a group under the composition ◦ defined by [f ] ◦

[g] = [f ∗ g], called the higher homotopy group of the pointed space (X , x0 ) for
n > 1.

Proof The composition ◦ defined by [f ] ◦ [g] = [f ∗ g] is clearly well defined. The

unit element e is defined by the homotopy class of the constant map

e(t) = e(t1 , t2 , . . . , tn ) = x0 , ∀ t ∈ In .

The inverse element of [f ] is defined by the homotopy class of

f −1 (t1 , t2 , . . . , tn ) = f (1 − t1 , t2 , . . . , tn ).

Proceed as in the case of π1 (X , x0 ) to complete the proof of the group structure

of πn (X , x0 ) for n > 1. 

Remark 2.20.12 The higher homotopy group πn (X , x0 ) of all pointed spaces (X , x0 )

is abelian for all n > 1, because there exists a rotation of S n interchanging the two
hemispheres of S n and keeping its base point s0 unchanged. Its analytical proof is
available in Theorem 2.20.14. On the other hand, the fundamental group π1 (X , x0 )
is not abelian for all pointed spaces (X , x0 ). For example, the fundamental group of
figure-eight F8 is not abelian by Theorem 2.19.16.

2.20.2 Abelian Group Structure of πn (X, x0 ) for n > 1

This subsection proves some properties of πn (X , x0 ). For example, Theorem 2.20.14

proves that πn (X , x0 ) is abelian for all n > 1, and for all pointed topological spaces
(X , x0 ) which is not true for fundamental groups for all (X , x0 ). This provides a
difference between πn (X , x0 ) for all n ≥ 2 and fundamental group π1 (X , x0 ).
 112 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Definition 2.20.13 Given a pointed topological space (X , x0 ), its loop space denoted
by (X , x0 ) is defined to be the space of all loops in X based at the point x0 endowed
with compact open topology. It is a pointed topological space with the constant map
c : I → x0 its base point.
Theorem 2.20.14 Let (X , x0 ) be a pointed topological space. Then, the higher
homotopy group πn (X , x0 ) is abelian for all n > 1.
Proof Since the loop space (X , x0 ) endowed with compact open topology is an
H -space, with the constant loop c as its homotopy identity, it follows that the group
π2 (X , x0 ) = π1 ( (X , x0 ), c) is abelian by Theorem 2.15.7. Using this result, the
theorem is proved by induction on n. 
Proposition 2.20.15 Let X = {x0 } be a topological space consisting of a single
element. Then, πn ({x0 }) = 0, ∀ n ≥ 0.
Proof By the given condition, there exists only one map f : S n → {x0 }, which is the
constant map and hence the proposition follows. 

2.20.3 Functorial Property of πn

This subsection proves that πn is a covariant functor from the homotopy category
T op∗ of pointed topological spaces to the category Ab of abelian groups and homo-
morphisms for every n ≥ 2. For the case, when n = 1, the functorial property of π1
has been proved in Theorem 2.11.3, which says that

π1 : Htp∗ → Grp

is a covariant functor from the category Htp∗ to the category Grp.

Theorem 2.20.16 Let f : (X , x0 ) → (Y , y0 ) be a base point preserving continuous

map. Then, f induces a homomorphism

f∗ : πn (X , x0 ) → πn (Y , y0 ), [σ ] → [f ◦ σ ], ∀ n ≥ 1.

such that
(i) For the identity map 1X : (X , x0 ) → (X , x0 ), its induced homomorphism 1X ∗ is
the identity automorphism on πn (X , x0 ).
(iii) For the base point preserving continuous maps f : (X , x0 ) → (Y , y0 ) and g :
(Y , y0 ) → (Z, z0 ), their composite map g ◦ f : (X , x0 ) → (Z, z0 ) induces a
homomorphism (g ◦ f )∗ with the property

(g ◦ f )∗ = g∗ ◦ f∗ : πn (X , x0 ) → πn (Z, z0 ).
 2.20 Higher Homotopy Groups 113

(iii) For homotopic maps f  g : (X , x0 ) → (Y , y0 ) rel {x0 }, their induced homo-

morphisms
f∗ = g∗ : πn (X , x0 ) → πn (Y , y0 ).

(iv) For a homeomorphism f : (X , x0 ) → (Y , y0 ), its induced homomorphism

f∗ : πn (X , x0 ) → πn (Y , y0 )

is an isomorphism.
(v) For a homotopy equivalence f : (X , x0 ) → (Y , y0 ), its induced homomorphism

f∗ : πn (X , x0 ) → πn (Y , y0 )

is an isomorphism.

Proof Case n = 1 : For fundamental group, it has been proved in Theorem 2.11.3.

For Case n > 1, take any element σ ∈ n (X , x0 ) and consider the assignment
σ → f ◦ σ. This assignment induces a transformation

f∗ : πn (X , x0 ) → πn (Y , y0 ), [σ ] → [f ◦ σ ], ∀ n > 1.

Then, f∗ carries the zero element of πn (X , x0 ) to the zero element of πn (Y , y0 ) and

for any two elements σ, τ ∈ n (X , x0 ),

f (σ ∗ τ ) = (f ◦ σ ) ∗ (f ◦ τ )

asserts that the induced transformation

f∗ : πn (X , x0 ) → πn (Y , y0 ), [σ ∗ τ ] → f∗ [σ ] ◦ f∗ [τ ], ∀ n > 1

is a homomorphism of groups, called the induced homomorphism of f . The prop-

erties (i) − (v) follow from the definition of induced homomorphism, which are
similar to the corresponding properties for fundamental groups proved in Theorem
2.11.3.


Corollary 2.20.17 (Homotopy invariance) Let (X , x0 ) and (Y , y0 ) be two homotopy

equivalent spaces. Then, the groups πn (X , x0 ) and πn (Y , y0 ) are isomorphic for every
n ≥ 1.

Proof By hypothesis, (X , x0 ) and (Y , y0 ) are homotopy equivalent spaces. Then,

there exists a homotopy equivalence

f : (X , x0 ) → (Y , y0 ).
114 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Hence, it follows by Theorem 2.20.16(v) that

f∗ : πn (X , x0 ) → πn (Y , y0 )

is an isomorphism for every n ≥ 1. 

In Theorem 2.20.18 and thereafter, the following notations are used:

(i) T op∗ stands for the category of pointed topological spaces and their base point
preserving continuous maps.
(ii) Ab stands for the category of abelian groups and their homomorphisms.
(iii) Htp∗ stands for the homotopy category of pointed topological spaces and their
homotopy classes of maps.

Theorem 2.20.18 (Functorial property of πn ) For every n ≥ 2,

(i) πn : T op∗ → Ab is a covariant functor such that if f , g : (X , x0 ) → (Y , y0 ) are
continuous and f  g rel {∂ In }, then πn (f ) = f∗ = g∗ = πn (g).
(ii) πn : Htp∗ → Ab is also a covariant functor.

Proof (i) The assignment

πn : T op∗ → Ab, (X , x0 ) → πn (X , x0 )

gives an object function, and the assignment

πn : T op∗ → Ab, f → πn (f ) = f∗

is well defined and gives a morphism function. This proves the first part of the
theorem by using Theorem 2.20.16.
(ii) The assignment
πn : Htp∗ → Ab, (X , x0 ) → πn (X , x0 )

defines an object function, and the assignment

πn : Htp∗ → Ab, [f ] → πn (f ) = f∗

is well defined and gives a morphism function. This proves the last part of the
theorem by using Theorem 2.20.16. 

2.20.4 Role of Base Point x in Groups πn (X, x) for n ≥ 1

This subsection studies the role of base point of a homotopy group and proves
Theorem 2.20.19 saying that for a path-connected space X , the groups πn (X , x)
do not depend on the choice of base points in the sense that all the groups πn (X , x)
2.21 Freudenthal Suspension Theorem and Its Applications 115

are isomorphic with various base points x ∈ X for every n ≥ 1. Its detailed study is
available in Chapter 5.
Theorem 2.20.19 Let X be a path-connected space and x0 , x1 ∈ X be two arbitrary
points. Then there exists an isomorphism

αn : πn (X , x1 ) → πn (X , x0 ), ∀ n ≥ 1.

Proof Proof I (Geometric proof): By hypothesis, X is a path-connected space.

Hence there exists a path α : I → X from the point x0 to the point x1 . Pull the image
of the boundary ∂ In of In along the path α back to the point x0 with the image of
In being pulled in an arbitrary way. The map thus obtained represents the homotopy
class of an element θ ∈ πn (X , x0 ). It depends only on σ and the homotopy class of
α. This correspondence σ → θ constructs the map αn

αn : πn (X , x1 ) → πn (X , x0 ), σ → θ, ∀ n ≥ 1.

This yields the required isomorphism αn of Theorem 2.20.19.

Proof II (Analytical proof): An analytical proof formulating the geometric con-
struction of αn described in Proof I, is available in Chap. 5. 

Corollary 2.20.20 Let X and Y be two homotopy equivalent path-connected spaces.

Then the groups πn (X ) and πn (Y ) are isomorphic and independent of base points
for every n ≥ 1.

Proof By hypothesis, X and Y are two homotopy equivalent path-connected spaces.

Let x0 ∈ X and y0 ∈ Y . Then there exists a homotopy equivalence

f : X → Y : f (x0 ) = y0 .

This f induces an isomorphism

f∗ : πn (X , x0 ) → πn (Y , y0 ), ∀ n ≥ 1.

Hence the corollary follows from Theorem 2.20.19. 

2.21 Freudenthal Suspension Theorem and Its Applications

This section defines Freudenthal suspension homomorphism, which is an isomor-

phism under some specified condition prescribed in Theorem 2.21.3. Hans Freuden-
thal proved this theorem in 1937 showing the stable range for homotopy groups. This
theorem is known as Freudenthal suspension theorem named after Hans Freudenthal.
The most interesting feature of the theorem is that the suspension homomorphism
lowers the dimension of homotopy groups by one and it facilitates the study of the
116 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

problem of computing the homotopy groups πm+n (S n ), one of the deepest problem
in homotopy theory. The other aspect of this theorem is to introduce the concept of
stabilization of homotopy groups leading to stable homotopy theory (see Chap. 5).

2.21.1 Freudenthal Suspension Theorem

This subsection describes a geometric construction of Freudenthal suspension homo-

morphism
σn : πn (S n ) → πn+1 (S n+1 ).

H. Freudenthal proved in 1937 a basic theorem in homotopy theory saying that

σn : πn (S n ) → πm+1 (S n+1 ), [α] → [α̃]

is an isomorphism for m < 2n − 1 and is surjective for m ≤ 2n − 1, where α̃ :

S n+1 → S n+1 is an continuous extension of the continuous map α : S n → S n deter-
mined uniquely up to homotopy. This theorem is known as Freudenthal Suspension
Theorem.

Definition 2.21.1 (Geometric Construction) To define Freudenthal suspension

homomorphism geometrically, consider

πm (S n ) = {[α] such that α : (S m , 1) → (S n , 1) is a continuous map }

and S n as the equator of S n+1 defined as the subspace of S n+1 consisting of all points
of S n+1 with last coordinate 0. Let N and S be the north and south points of S n+1
defined by

N = (0, 0, . . . , 1) ∈ S n+1 , and S = (0, 0, . . . , −1) ∈ S n+1 .

Let [α] ∈ πm (S n ). Then α : S m → S n is a continuous map. Extend α to a continu-

ous function α̃ : S n+1 → S n+1 as follows: α̃|S m = α and it maps the equator of S m+1
to the equator of S n+1 . The map is then extended radially as shown in Fig. 2.19.

Fig. 2.19 Radially extended

map
t α(t)

x Sm S n α(x)

S m+1 S n+1
2.21 Freudenthal Suspension Theorem and Its Applications 117

The arc from the north pole to a point x ∈ S m is mapped linearly onto the arc from
the north pole of S n+1 to α(x). This defines the map α̃ on the northern hemisphere.
For the southern hemisphere it is similarly defined. The extended map α̃ is called the
suspension of α.
We are now in a position to define Freudenthal suspension homomorphism σn .
Definition 2.21.2 (Freudenthal suspension homomorphism) The natural map

σn : πn (S n ) → πn+1 (S n+1 ), [α] → [α̃]

is a homomorphism, called the Freudenthal suspension homomorphism.

Theorem 2.21.3 (Freudenthal suspension theorem ) Freudenthal suspension homo-
morphism

σn : πm (S n ) → πm+1 (S n+1 ), [α] → [α̃]

is an isomorphism for m < 2n − 1 and is surjective for m ≤ 2n − 1, where α̃ :

S n+1 → S n+1 is an continuous extension of the continuous map α : S n → S n deter-
mined uniquely up to homotopy.
Proof The paper [Freudenthal, 1937] is referred. 

2.21.2 Hurewicz and Hopf Theorems on Sn

This subsection proves Hurewicz and Hopf Theorems on S n , which are two basic
results in homotopy theory. The proof is based on Freudenthal Suspension Theorem
2.21.3.
Theorem 2.21.4 (Hurewicz) πm (S n ) = 0, ∀ m, n satsfying 0 < m < n.
Proof Let k be any positive integer such that k < m. Then k + m + 1 < 2n =⇒
m − k < 2(n − k) − 1 and hence by Freudenthal suspension theorem 2.21.3, it fol-
lows that

πm (S n ) ∼
= πm−1 (S n−1 ) ∼
= πm−2 (S n−2 ) ∼
= πm−1 (S n−1 ) ∼
= ··· ∼
= π1 (S n−m+1 ) = 0

by Theorem 2.26.7, since n − m + 1 > 1. 

Theorem 2.21.5 (Hopf) πn (S n ) ∼
= Z for every integer n ≥ 1.
Proof Since π1 (S 1 ) ∼
= Z, it follows from the above discussion that Z ∼
= π1 (S 1 ) ∼
=
π2 (S ). For n > 1, n < 2n − 1 asserts by Freudenthal suspension theorem that
2

π2 (S 2 ) ∼
= π3 (S 3 ) ∼
= π4 (S 4 ) ∼
= ··· ∼
= πn (S n ) ∼
= Z. 
118 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Table 2.1 Table of πi (S n ) for 1 ≤ i ≤ 8, 1 ≤ n ≤ 8

i=1 2 3 4 5 6 7 8
n=1 Z 0 0 0 0 0 0 0
2 0 Z Z Z2 Z2 Z12 Z2 Z2
3 0 0 Z Z2 Z2 Z12 Z2 Z2
4 0 0 0 Z Z2 Z2 Z × Z12 Z2 × Z2
5 0 0 0 0 Z Z2 Z2 Z24
6 0 0 0 0 0 Z Z2 Z2
7 0 0 0 0 0 0 Z Z2
8 0 0 0 0 0 0 0 Z

Remark 2.21.6 Theorem 2.21.5 proved by H. Hopf is called Hopf degree theorem.
The justification towards naming it degree theorem is given in Sect. 2.22.

2.21.3 Table of πi (Sn ) for 1 ≤ i, n ≤ 8

This subsection gives the Table 2.1 displaying a small sample of the values of the
groups πi (S n ), the results are extracted from the paper [Toda, 1962]. It is an immediate
consequence of Freudenthal Suspension Theorem 2.21.3. An analogue table is also
given in Chapter 5 for readers interested in topology of fiber bundles.

2.22 Degree of a Spherical Map on Sn from Viewpoint

of Homotopy Theory

This section gives the concept of degree of a spherical map by using homo-
topy theory in Definition 2.22.2 by establishing a close relation between the group
πn (S n ) ∼
= Z and degree of a continuous spherical map f : S n → S n . On the other
hand, degree of a spherical map by using homology theory is equally well-defined
in Chapter 3. The concept of degree of such maps play a key role in analysis and
topology and it generalizes the concept of degree of a loop on the circle, studied in
Sect. 2.17.

Proposition 2.22.1 Let f : S n → S n be a continuous map and αn be a generator of

the group πn (S n ) ∼
= Z. If f∗ : πn (S n ) → πn (S n ) is the homomorphism induced by f ,
then f∗ (αn ) determines an integer d such that f∗ (αn ) = d αn .
(i) The integer d is independent of the choice of the generator αn of πn (S n ).
2.22 Degree of a Spherical Map on S n from Viewpoint of Homotopy Theory 119

(ii) If f , g : S n → S n are two continuous maps such that f  g. Then the homomor-
phisms f∗ = g∗ : πn (S n ) → πn (S n ), ∀ n ≥ 1.
Proof (i) Let x ∈ πn (S n ) be an arbitrary element. There is some integer p such that
x = pαn . Now,

f∗ (x) = f∗ (pαn ) = pf∗ (α) = pd αn = d (pαn ) = dx, ∀ x ∈ πn (S n ).

It also follows in particular that

f∗ (−αn ) = d (−αn ).

Consequently, it asserts that the degree d corresponding to f : S n → S n does not

depend on a particular choice of the generator of πn (S n ).
(ii) It follows from the functorial property of πn given in Corollary 2.20.17. 
Definition 2.22.2 Let f : S n → S n be a continuous map and αn be a generator of
the group πn (S n ) = Z. If

f∗ : πn (S n ) → πn (S n ), [β] → [f ◦ β]

is the homomorphism induced by f , then f∗ (αn ) determines an integer d such that

f∗ (αn ) = d αn . The integer d is independent of the choice of the generator and hence it
is well-defined and is called the degree of the spherical map f : S n → S n , denoted
by deg f = d .
The above discussion is summarized in a basic result embodied in Proposition
2.22.3.
Proposition 2.22.3 Let f : S n → S n be a continuous map. Then f has degree d iff
its induced homomorphism

f∗ : πn (S n ) → πn (S n )

has the property

f∗ (x) = dx, ∀ x ∈ πn (S n ).

Example 2.22.4 The continuous map f : S 1 → S 1 , z → z m has degree m. Geomet-

rically, f rounds the circle S 1 completely m times.
Remark 2.22.5 Given a continuous spherical map f : S n → S n , its degree deg f = d
is the number of the homotopy classes [f ] ∈ πn (S n ). Because, for the continuous map
f : S n → S n,

f∗ (αn ) = f∗ ([1d ]) = [f ].

Since αn is the homotopy class of the identity map 1d : S n → S n , it follows that d αn

is the homotopy class of f .
120 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

2.23 Applications

This section presents some interesting applications of homotopy and fundamental

groups to algebra, matrix theory, atmospheric science, vector field and extension
problems.

2.23.1 Fundamental Theorem of Algebra and Algebraic

Completeness of the Field C

This subsection proves the fundamental theorem of algebra through homotopy the-
ory. The proof of this theorem in an algebraic way is complicated, but it is less
complicated by using complex analysis. One role of algebraic topology is mainly
‘algebra serving topology,’ but in this subsection the role is reversed to prove ‘Funda-
mental Theorem of Algebra’ by using the concept of homotopy followed by proving
the algebraic completeness of the field C of complex numbers.

Theorem 2.23.1 (Fundamental Theorem of Algebra) Let C be the field of complex

numbers. Every nonconstant polynomial over C has a root in C.

Proof Proof I : Let f (z) = a0 + a1 z + · · · + an−1 z n−1 + an z n , an = 0, n ≥ 1 be an

arbitrary polynomial over C. If the theorem is not true, the restriction of the map

ψ : C → C − {0}, z → f (z)

to different circles Cr : |z| = r, r ≥ 0 are loops in C − {0}. Define the two maps

f (re2πit )/f (r)

F : I × [0, ∞) → S 1 , (t, r) →
|f (re2πit )/|f (r)|

and

F(t, s/(1 − s)), 0 ≤ t ≤ 1, 0 ≤ s < 1
H : I × I → S ⊂ C, (t, s) →
1
e2πint , 0 ≤ t ≤ 1, s = 1.

The continuity of F asserts that

lim H (t, s) = lim F(t, s/(1 − s)) = lim F(t, r) = e2πint .

s→1 s→1 r→∞

This proves the continuity of H . If H (t, 0) = g0 (t) and H (t, 1) = g1 (t), then

H : g0  g1 rel İ .
2.23 Applications 121

This shows that deg g0 = deg g1 by Theorem 2.24.1. Again, by Definition 2.17.9, it
follows that deg g0 = 0 and deg g1 = n ≥ 1. This contradiction implies that f (z) has
a root in C.
Proof II : Let f (z) = a0 + a1 z + · · · + an−1 z n−1 + an z n , an = 0, n ≥ 1 be an
arbitrary polynomial over C. If the theorem is not true, then f (z) = 0, ∀ z ∈ C.
Define a map
f (tz)
Ht : S 1 → S 1 , z →
|f (tz)|

for every nonnegative real number t. Then H0 is a constant map at a0 . Taking t large
enough, it follows that Ht is homotopic to the map

g : S 1 → S 1 , z → zn .

Since any two of the maps Ht are homotopic, H0 and g are homotopic. Again two
closed paths in S 1 are homotopic iff they have the same degree. This implies that
deg H0 = deg g, which is not possible, since deg H0 = 0 and deg g = n. 
Definition 2.23.2 A field F is called algebraically complete (or closed) if every
polynomial over F of degree ≥ 1, has a root in F.
Corollary 2.23.3 The complex field C is algebraically complete.
Proof Theorem 2.23.1 asserts that the field C is algebraically complete. 
Example 2.23.4 The field C of complex numbers is algebraically complete but the
real field R is not complete. Because there are some polynomials over R of degree
≥ 1 having no root in R. For example, x2 + 2 has no root in R. But it has a root in C.
Corollary 2.23.5 The real field R is embedded in the algebraically complete field C.
Proof The corollary follows from Corollary 2.23.3. 

2.23.2 Fixed Point and Homotopy: Brouwer Fixed-Point

Theorem for Dimension 2

The aim of this subsection is to study links between fixed-point theory and homotopy.
For example, this subsection proves Brouwer fixed-point theorem (BFT) for dimen-
sion 2 as an application of Theorem 2.19.5 asserting that S 1 ⊂ D2 not a retract of
D2 . Moreover, Proposition 2.23.11 establishes a close relation between fixed point
theorem and homotopy. There is a natural question: if X = D2 − {0} is the punc-
tured disk, does there exist a continuous map f : X → X without fixed points? For
existence of such an example see Exercise 43 of Sect. 2.28.
Theorem 2.23.6 proves Brouwer fixed-point theorem for dimension 2. Its gener-
alization for an arbitrary finite dimension n is proved in Theorem 2.25.58.
122 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Theorem 2.23.6 (Brouwer Fixed-Point Theorem for Dimension 2) Every continu-

ous map h : D2 → D2 has a fixed point.

Proof Suppose h has a fixed point. Then there exists a point x ∈ D2 such that h(x) =
x. If no such point x ∈ D2 exists, then h(x) = x, ∀ x ∈ D2 . Define a map

r : D2 → S 1 , x → r(x) ∈ S 1 ,

where r(x) is the point of intersection of the circle S 1 with the ray in R2 starting from
the point h(x) and passing through the point x. Since this ray intersects S 1 at exactly
one point, the map r is well-defined. This implies that

x = (1 − t)h(x) + tr(x)

for some t > 0. Hence the point r(x) is determined by the rule

(x − (1 − t)h(x))
r(x) = .
t

This shows that r is continuous. Moreover, r(x) = x, ∀ x ∈ S 1 shows that r : D2 →

S 1 is a retraction which implies that S 1 is a retract of D2 . But it contradicts Theorem
2.19.5. 

Remark 2.23.7 For an alternative proof of Theorem 2.23.6 by using the concept of
vector field is already given in Theorem 2.18.19. Brouwer fixed-point theorem for
dimension n = 1 is also proved in Theorem 2.26.6. Its generalization for an arbitrary
finite dimension n is given in Theorem 2.25.58.

Corollary 2.23.8 Every real 3 × 3 matrix with positive real entries has a positive
real eigenvalue.

Proof Let A be a real 3 × 3 matrix with positive real entries and associated linear
transformation L : R3 → R3 with respect to the standard basis of the real vector
space R3 and X1 = {(x, y, z) ∈ R3 : x, y, z ≥ 0} be the first octant of R3 . If X =
S 2 ∩ X1 ⊂ R3 , then the subspace X is homeomorphic to the disk D2 . This implies
that Brouwer fixed-point theorem is valid for every continuous map f : X → X . For
every w = (x, y, z) ∈ X , all the components x, y, z of the point w must be nonnegative
and at least one of them is positive. Since by hypothesis, all the entries of the matrix
A are positive by hypothesis, the vector L(x) is a vector having its all components
positive. Hence the map
L(w)
f : X → X , w →
||L(w)||
L(x0 )
is continuous. This asserts that f has a fixed point x0 (say). Therefore, x0 = ||L(x0 )||
shows that L(x0 ) = ||L(x0 )||x0 . This implies that the linear transformation L and
2.23 Applications 123

hence the matrix A has the positive real eigenvalue ||L(x0 )||. It proves that the matrix
A has a positive real eigenvalue. 
Remark 2.23.9 A generalization of the Corollary 2.23.8 is given in the Perron–
Frobenius theorem in Rn . It asserts that any square matrix with positive entries
has a unique eigenvector with positive entries (up to a multiplication by a positive
constant) and the corresponding eigenvalue has multiplicity one and is strictly greater
than the absolute value of any other eigenvalue.
Definition 2.23.10 For a point x ∈ S n , its antipode is the point −x ∈ S n and the
pair {x, −x} of points is said to be a pair of antipode points. A continuous map
f : S n → S m is called antipodal if f (−x) = −f (x), ∀ x ∈ S n .
Proposition 2.23.11 establishes a close relation between fixed point theorem and
homotopy.
Proposition 2.23.11 Let the continuous map h : S 1 → S 1 be nullhomotopic, i.e., h
is homotopic to a constant map. Then
(i) h has a fixed point.
(ii) There exists some point x0 ∈ S 1 such that h(x0 ) = −x0 (i.e., h maps the point x0
to its antipode −x0 ).
Proof (i) Let h : S 1 → S 1 be nullhomotopic. Then the map h has a continuous
extension
h̃ : D2 → S 1 ⊂ D2

(see Theorem 2.23.22). Hence by Brouwer fixed-point theorem, there is a point

x0 ∈ D2 such that h̃(x0 ) = x0 . But h̃(x0 ) ∈ S 1 , since Im h̃ ⊂ S 1 . Hence h̃(x0 ) =
x0 shows that x0 ∈ S 1 . This concludes that x0 is a fixed point of h.
(ii) Consider the map g : S 1 → S 1 , x → −x. Then the composite map g ◦ h : S 1 →
S 1 is also nullhomotopic. This shows that g ◦ h has a fixed point x0 . Thus x0 =
(g ◦ h)(x0 ) = g(h(x0 )) = −h(x0 ) shows that h(x0 ) = −x0 .

Corollary 2.23.12 There is no antipodal continuous map h : S 2 → S 1 .

2.23.3 Borsuk–Ulam Theorem for Dimension 2

This subsection proves the Borsuk–Ulam theorem for dimension 2 and applies this
theorem to solve the following two problems.

Problem 1: Can the standard sphere S 2 be put in the Euclidean plane R2 ? Its
negative answer is available in Remark 2.23.17.
Problem 2 : Does there exist a pair of antipode points {x, −x} on the earth S 2
(assumed spherical) having the same temperature and same barometric pressure at
any point of time? Its affirmative answer is available in Remark 2.23.17.
124 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Remark 2.23.13 Borsuk–Ulam Theorem 2.23.14 for dimension 2 solves the prob-
lems posed as above. A generalization of Borsuk–Ulam Theorem 2.23.14 for higher
dimension asserts that given any continuous map f : S n → Rn (n ≥ 2), there exists a
pair of antipodal points {x, −x} on S n with the property that f (x) = f (−x). Its proof
is available in Chap. 6.

Theorem 2.23.14 (Borsuk–Ulam Theorem for dimension 2) Given any continuous

map f : S 2 → R2 , there exists a pair of antipodal points {x, −x} on S 2 with the
property that f (x) = f (−x).

Proof If possible, f (x) = f (−x) for all x ∈ S 2 . Then there exists a continuous map
f (x) − f (−x)
h : S 2 → S 1 , x → .
||f (x) − f (−x)||

This asserts that h(−x) = −h(x), ∀ x ∈ S 2 and hence it contradicts the Corollary
2.23.12. 

Remark 2.23.15 A generalization of Theorem 2.23.14 asserting that given two inte-
gers m, n with m > n ≥ 0, there does not exist a continuous map f : S m → S n pre-
serving the antipodal points is proved by using homology theory in Chapter 6.

Corollary 2.23.16 The sphere S 2 cannot be embedded in the Euclidean plane R2 .

Proof Suppose f : S 2 → R2 is an embedding. Then it is continuous and injective. By

applying Borsuk–Ulam Theorem 2.23.14, it follows that there exists a pair {x, −x} of
antipode points on S 2 with the property that f (x) = f (−x). This gives a contradiction,
since the map f is injective. 

Remark 2.23.17 Corollary 2.23.16 asserts there exists no homeomorphism from

the sphere S 2 on the plane R2 and hence 2-sphere S 2 cannot be put in the plane R2 .
This implies that no map of the earth can be drawn (up to homeomorphism) on
a page of an atlas. This gives a negative answer of the Problem 1.

Remark 2.23.18 Atmospherically, Borsuk–Ulam Theorem 2.23.14 asserts that

there exist two different places on the earth represented by a pair of points x and
−x having the same temperature and same barometric pressure at any point of time.
To show it, consider the earth as 2-sphere S 2 . Let T , P be functions on the earth
defining temperature and barometric pressure at any point of time and at a place on
the earth. These are continuous functions and define a map

f : S 2 → R2 , x → (T (x), P(x)).

Then by Borsuk–Ulam Theorem 2.23.14, at any point of time, there exists a pair of
points x and −x on the earth S 2 such that the temperature and barometric pressure
both are identical at x and −x. This gives an affirmative answer of the Problem 2.
2.23 Applications 125

2.23.4 Lusternik–Schnirelmann Theorem for S2

This subsection proves Lusternik–Schnirelmann theorem by applying Borsuk–Ulam

Theorem 2.23.14.

Theorem 2.23.19 (Lusternik–Schnirelmann theorem for S 2 ) If S 2 = G 1 ∪ G 2 ∪

G 3 , where each G i is a closed subset of S 2 , (i.e., if S 2 is covered by its three closed
subsets G 1 , G 2 and G 3 ), then one of the sets contains a pair of antipodal points.

Proof Define a function

f : S 2 → R2 , x → (d (x, G 1 ), d (x, G 2 )).

where d (x, G i ) denotes the Euclidean distance of the point x from G i . Since f is
continuous, it identifies a pair of opposite points by Theorem 2.23.14. This asserts
that there is a point x0 ∈ S 2 such that

d (x0 , G i ) = d (−x0 , G i ) for 1 ≤ i ≤ 2.

If d (x0 , G i ) > 0, for 1 ≤ i ≤ 2, then x0 , − x0 ∈ G 3 , since S 2 = G 1 ∪ G 2 ∪ G 3 .

Again, if d (x, G i ) = 0 for some i, then both x0 and −x0 lie in G i , since each G i is a
closed subset of S 2 . 

Remark 2.23.20 A generalization of Lusternik–Schnirelmann Theorem 2.23.19 for

S n asserts that if S n = G 1 ∪ G 2 ∪ · · · ∪ G n , where each G i is a closed subset of S n ,
then one of the sets contains a pair of antipodal points. Its proof is similar to that of
Theorem 2.23.19. Its detailed proof is available in Chapter 6.

2.23.5 Cauchy’s Integral Theorem of Complex Analysis

Cauchy’s integral theorem is one of the basic theorems of complex analysis. This
subsection proves this theorem from the viewpoint of homotopy by utilizing the
concept of winding number of a differentiable loop in the complex plane and the
exponential map p : R → S 1 , t → e2πit .

Theorem 2.23.21 (Cauchy’s Integral Theorem) Let X ⊂ C be an open subset and

f : X → C be an analytic function. If β is a simple closed piecewise differentiable
loop in X with the property that β is nullhomotopic, then β f (z)dz = 0.

Proof As β is nullhomotopic, there exists a constant curve α in X such that β  α.

If H : β  α, define a map

g : I → Z, t → w(βt ; z0 ).
126 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

That is, g(t) = w(βt ; z0 ), where βt (s) = H (s, t) for 0 ≤ s, t ≤ 1 and z0 is a fixed
point in C − X and w(β; z0 ) denotes the winding number of β about the point z0 .
Clearly, the map g is continuous on I. Again since g is an integral valued function and
g(0) = 0, it follows that g(t) ≡ 0. Consequently, the winding number w(β; z0 ) = 0
for all z0 ∈ C − X . This proves the theorem by using Exercise 31 of Sect. 2.28. 

2.23.6 Extension Problem in Homotopy Theory

Extension problem in topology is a basic problem. This subsection solves an exten-

sion problem by using homotopy of maps but it is different from Tietze extension
theorem of real-valued continuous functions on normal spaces (see Chap. 6).

Extension problem formulated in Theorem 2.23.22 is solved by homotopy theory.

Theorem 2.23.22 (Extension Problem) Let S n ⊂ Dn+1 be the n-sphere in the

Euclidean (n + 1)-space Rn+1 and X be an arbitrary topological space. Then a con-
tinuous map f : S n → X has a continuous extension over the (n + 1)-disk Dn+1 iff
f is homotopic to some constant map c : S n → X (i.e., iff f is nullhomotopic).

Proof Suppose c : S n → X , y → x0 ∈ X is a constant map such that f  c. Then

there exists a homotopy

H : S n × I → X , (x, 0) → f (x), and (x, 1) → x0

such that H : f  c. Define a map


x0 , 0 ≤ ||y|| ≤ 1/2
f˜ : Dn+1 → X , y → y
H ( ||y|| , 2 − 2||y||), 1/2 ≤ ||y|| ≤ 1.

Let A = {y ∈ Dn+1 : 0 ≤ ||y|| ≤ 1/2} and B = {y ∈ Dn+1 : 1/2 ≤ ||y|| ≤ 1}. Then f˜
y
agrees on their intersection A ∩ B, since at ||y|| = 21 , H ( ||y|| , 1) = x0 . Again A, B ⊂
Dn+1 are closed sets such that Dn+1 = A ∪ B. Since the map f˜ is well defined and
is continuous on each of closed sets A and B, by Pasting Lemma, f˜ is continuous.
Again, since ∀ y ∈ S n , ||y|| = 1 and f˜ (y) = H (y, 0) = f (x), it follows that f has
a continuous extension f˜ over Dn+1 . Conversely, if f˜ : Dn+1 → X is a continuous
extension of f : S n → X , then f˜ (y) = f (y) ∀ y ∈ S n . Let s0 ∈ S n and f (s0 ) = x0 ∈
X . Define a map

H : S n × I → X , (y, t) → f˜ ((1 − t)y + ts0 ).

Again, ((1 − t)y + ts0 ) ∈ Dn+1 ∀ y ∈ S n , since Dn+1 is a convex set. Clearly, H is a
continuous map such that H : f  c. 
 2.23 Applications 127

Another extension problem is solved in Corollary 2.23.23 by the value of c(f )

formulated in Remark 2.20.7.

Corollary 2.23.23 c(f ) formulated in Remark 2.20.7 vanishes iff f has a continuous
extension over Dn+1 .

Proof It follows from definition of c(f ) by using Theorem 2.23.22. 

Remark 2.23.24 For a study of extension problem by using the degree function, see
Theorem 2.25.1.

2.24 Hopf Classification Theorem by Using Homotopy

Theory

Hopf classification theorem 2.24.2 characterizes the homotopy of continuous spher-

ical maps by their degrees formulated in Definition 2.22.2 by homotopy theory,
which are integers and hence it gives a complete classification of spherical maps by
integers. On the other hand, the same homotopy classification theorem of continuous
spherical maps by their degrees formulated by homology theory is available in
Chapter 3.

The main tool used here is the concept degree of spherical maps given in Definition
2.22.2 and their basic properties are given in Theorem 2.24.1.

Theorem 2.24.1 (i) If 1d : S n → S n is the identity map, then its degree is 1.

(ii) If f  c : S n → S n , where c is a constant map, then its deg f is 0.
(iii) If f , g : S n → S n are continuous maps, then deg (f ◦ g) = deg f deg g.
(iv) If f  g : S n → S n , then deg f = deg g.
(v) If f , g : S n → S n are continuous maps such that deg f = deg g, then f  g.

Proof It follows from the definition of degree of a continuous spherical map f :

S n → S n. 

Theorem 2.24.2 (Hopf classification theorem) Two continuous maps f , g : S n →

S n are homotopic iff deg f = deg g.

Proof It follows from Theorem 2.24.1. An alternative proof is given in Theorem

2.25.20. 
128 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

2.25 Application of Degree of Spherical Maps on Sn

and Hopf Classification Theorem

This section applies the degree of continuous map to solve some classical problems
such as extension problem given in Theorem 2.25.1. Moreover, Hopf classification
theorem 2.25.20 completely classifies continuous spherical maps f : S n → S n by
degree function deg : f → deg f .
Given a continuous spherical map f : S n → S n , its degree deg f = d is the number
of the homotopy classes [f ] ∈ πn (S n ). Because, for the continuous map f : S n →
S n , f∗ (αn ) = f∗ [1d ] = [f ], since αn is the homotopy class of the identity map 1d :
S n → S n and hence d α is the homotopy class of f . For any continuous map f : S n →
S n , the deg f is equally well defined by homology theory and is studied in Chap. 3.
Let f : S n → Rn+1 − {0} be a continuous map. It is a natural question: whether f has
a continuous extension f˜ over the closed (n + 1)- disk Dn+1 ? Since the punctured
Euclidean space Rn+1 − {0} is homotopy equivalent to S n by a homotopy equivalence
g, it follows that the groups πn (S n ) and πn (Rn+1 − {0}) are isomorphic by Corollary
2.20.17 by the isomorphism g∗ . Hence the deg(g ◦ f ) of the map g ◦ f : S n → S n
leads to define the degree of a map f : S n → Rn+1 − {0}, called the characteristic
of the vector field f , usually denoted by κS n (f ). It is used to solve an extension
problem in Theorem 2.25.1.
Theorem 2.25.1 A continuous map f : S n → Rn+1 − {0} has a continuous exten-
sion f˜ over the closed (n + 1)-disk Dn+1 iff κS n (f ) = 0.
Proof If f is homotopic to a constant map c, i.e., if f is nullhomotopic, then
κS n (f ) = 0. Moreover, the continuous extension f˜ over the closed (n + 1)- disk Dn+1
determines a homotopy

H : S n × I, (x, t) → f˜ (tx).

H is a continuous map such that H : f  c. Hence the theorem follows from Theorem
2.23.22. 
Corollary 2.25.2 Let a continuous map f : S n → Rn+1 − {0} has a continuous
extension f˜ over the closed (n + 1)-disk Dn+1 . If κS n (f ) = 0, then f˜ has a fixed
point.
Remark 2.25.3 Given a continuous extension f˜ : Dn+1 → Rn+1 of the continuous
map f : S n → Rn+1 − {0}, Corollary 2.25.2 is applied to solve the existence problem
of a solution to the equation f˜ (x) = 0.
Theorem 2.25.4 Let f : S n → S n be a continuous map of degree d . Then its sus-
pension map

f = fn+1 : S n+1 → S n+1 , [x, t] → [f (x), t]

has also the same degree d .

 2.25 Application of Degree of Spherical Maps on S n … 129

Fig. 2.20 Freudenthal

suspension isomorphism σn
diagram

Proof To prove the theorem, consider the commutative diagram in Fig. 2.20 of
groups and homomorphisms where σn is the Freudenthal suspension isomorphism
and
fn∗ : πn (S n ) → πn (S n )

is the homomorphism of a continuous map fn : S n → S n of degree d . Since σn is an

isomorphism and the diagram in Fig 2.20 is commutative, it follows that

σn ◦ fn∗ ◦ σn−1 = f(n+1)

∗
.

Hence it follows that

f(n+1)∗ (x) = σn ◦ fn∗ ◦ σn−1 (x) = d σn [σn−1 (x)] = dx, ∀ x ∈ πn+1 (S n+1 ).

This proves the theorem by Proposition 2.22.3. 

Corollary 2.25.5 Given any positive integer n, there is a continuous map f : S n →

S n such that deg f = d.

Proof For n = 1, it follows from Theorem 2.19.1. Suppose it is true for some n ≥ 1
and f : S n → S n is a continuous map such that deg f = d . Let fn : S n → S n be a
continuous map of degree d . Then its suspension map

f = fn+1 : S n+1 → S n+1

has also the same degree d by Theorem 2.25.4. By induction on n the proof of the
corollary is completed. 

Theorem 2.25.6 Let f , g : S n → S n be any two continuous maps. Then

(i) deg (f ◦ g) = deg f deg g.
(ii) If f = 1S n is identity map, then deg f = 1.
(iii) If f is a homeomorphism, then deg f = 1 or -1.
(iv) If f  g, then deg f = deg g.
(v) If f  c : S n → S n , where c is a constant map, then its deg f is 0.

Proof (i) (f ◦ g)∗ (x) = f∗ (g∗ (x)) = f∗ (deg gx) = deg f deg g, ∀ x ∈ S n =⇒
deg(f ◦ g) = deg f deg g;
(ii) f = 1S n , f∗ (x) = x = 1x, ∀ x ∈ S n =⇒ deg f = 1.
130 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

(iii) For a homeomorphism f with its inverse f −1 , their composite map

f ◦ f −1 = 1S n =⇒ deg f deg f −1 = 1 =⇒ deg f = 1 or − 1,

since deg f and deg f −1 are both integers.

(iv) It follows from Proposition 2.22.1.
(v) It follows trivially. 
Definition 2.25.7 The map

rm : S n → S n , x = (x1 , x2 , . . . , xm−1 , xm , . . . , xn+1 ) → (x1 , x2 , . . . , xm−1 − xm , . . . , xn+1 )

is called the reflection map of S n about the xm axis for m = 1, 2, . . . , n + 1 and the
map

A : S n → S n , x = (x1 , x2 , . . . , xm−1 , xm , . . . , xn+1 )  → −x = (−x1 , −x2 , . . . , −xm−1 − xm , . . . , −xn+1 ),

i.e., A(x) = − x, ∀ x ∈ S n , is called the antipodal map.

Proposition 2.25.8 The degree of the refection map
rm : S n → S n , x = (x1 , x2 , . . . , xm−1 , xm , . . . , xn+1 ) → (x1 , x2 , . . . , xm−1 , −xm , . . . , xn+1 ),

is −1, i.e., deg rm = −1 for all m = 1, 2, . . . n + 1

Proof For each reflection map
rm : S n → S n , x = (x1 , x2 , . . . , xm−1 , xm , . . . , xn+1 ) → (x1 , x2 , . . . , xm−1 , −xm , . . . , xn+1 ),

deg rm = −1, because rm = 1S n but rm ◦ rm = 1S n and hence deg rm deg rm = 1

and deg rm = 1. This proves that the only possibility is that deg rm = −1. This is
true for every m = 1, 2, . . . , n + 1. 
Proposition 2.25.9 The degree of the antipodal map A : S n → S n , x → −x is
(−1)n+1 for all n ≥ 1.
Proof The map A is the composite map

A = r1 ◦ r2 ◦ · · · ◦ rn+1 .

Hence it implies that

deg A = (−1)n+1 ,

since deg rm = −1, ∀ m = 1, 2, . . . , n + 1. 

Theorem 2.25.10 relates the nonexistence of fixed points of spherical maps to
their degrees.
Theorem 2.25.10 Let f : S n → S n be a continuous map such that f has no fixed
point for all integers n > 0, then deg f = (−1)n+1 .
2.25 Application of Degree of Spherical Maps on S n … 131

Proof By hypothesis f has no fixed point. It asserts that the line segment (1 −
t)f (x) − tx = 0, ∀ x ∈ S n , ∀ t ∈ I. Otherwise, if (1 − t)f (x) − tx = 0 for some x ∈
S n and any t ∈ I, then

1 − t = ||(1 − t)f (x)|| = ||tx|| =⇒ 1 − t = t,

since ||f (x)|| = 1 = ||x||, ∀ x ∈ S n and hence from above it follows that f (x) = x,
which is not possible, since f (x) = x by hypothesis. Hence the map H : S n × I → S n
defined by
(1 − t)f (x) − tx
H (x, t) =
||(1 − t)f (x) − tx||

is well-defined and continuous. Then H (x, 0) = f (x) and H (x, 1) = −x = A(x),

∀ x ∈ S n show that H : f  A and hence it follows that deg f = (−1)n+1 from Propo-
sition 2.25.9. 

Theorem 2.25.10 is now applied to solve the following fixed-point problems by

homotopy.

Corollary 2.25.11 Let f : S 2n → S 2n be a continuous map such that f  1S n . Then

f has a fixed point.

Proof If f  1S n , then deg f = 1. Suppose f has no fixed point. Then by deg f =

(−1)2n+1 = −1 by using Theorem 2.25.10, a contradiction. This contradiction shows
that f has a fixed point. 

Corollary 2.25.12 Let f : S n → S n be a continuous map such that

deg f = (−1)n+1 .

Then f has a fixed point.

Proof By hypothesis,
deg f = (−1)n+1 .

If possible, f has no fixed point. Then it follows by Theorem 2.25.10 that

deg f = (−1)n+1 .

This contradiction proves the corollary. 

Proposition 2.25.13 If f , g : S n → S n be two continuous maps such that f (x) =

g(x), ∀ x ∈ S n . Then
deg f + (−1)n deg g = 0.
 132 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proof Proceed as in Theorem 2.25.10 by showing first that

(1 − t)f (x) − tg(x) = 0, ∀ x ∈ S n , ∀ t ∈ I. 

Definition 2.25.14 A topological space X is said to have the fixed-point property
if every continuous map
f :X →X

has a fixed point.

Example 2.25.15 Examples of topological spaces having fixed-point property.
(i) Let X = [a, b] be a subspace of the real line space R. Then every continuous
map f : X → X has a fixed point by Brouwer fixed-point theorem 2.26.6 of
dimension 1. This implies that [a, b] has the fixed-point property.
(ii) Let X = D2 be the unit disk in R2 . Then every continuous map f : X → X has a
fixed point by Brouwer fixed-point theorem 2.23.6 of dimension 2. This implies
that D2 has the fixed-point property.
(iii) Let X = Dn be the unit disk in Rn . Then every continuous map f : X → X has a
fixed point by Brouwer fixed-point Theorem 2.25.58 of a finite dimension n ≥ 0.
This implies that Dn has the fixed-point property.
(iv) On the other hand, the sphere S n has no fixed-point property by Proposition
2.25.16.
All topological spaces do not enjoy the fixed-point property. For example, by
Proposition 2.25.16, the sphere S n has no fixed-point property.
Proposition 2.25.16 For any positive integer n, the sphere S n has no fixed-point
property.
Proof Let f : S n → S n be an arbitrary continuous map. Since the antipodal map
A : S n → S n , x → −x has no fixed point, it follows that S n has no fixed-point
property. 
Proposition 2.25.17 proves the nonexistence of continuous nonvanishing vector
field on S n for every even integer n.
Proposition 2.25.17 If n is an even integer, there is no continuous nonvanishing
vector field f : S n → S n .
Proof Let v be a continuous nonvanishing vector field on S n . The map

v(x)
f : S n → S n , x →
||v(x)||

is a continuous map homotopic to the identity map on S n . As by hypothesis, n is an

even integer, it follows from Corollary 2.25.11 that f has a fixed point x0 such that
< x0 , f (x0 ) >= 1. This implies that the vector field must vanish at some point of
S n. 
2.25 Application of Degree of Spherical Maps on S n … 133

Remark 2.25.18 An alternative proof of Proposition 2.25.17 is given in Corollary

2.25.24.

2.25.1 Brouwer Degree Theorem and Hopf Classification

Theorem by Using Homotopy Theory

Hopf classification theorem 2.24.2 characterizes the homotopy of continuous spher-

ical maps by their degrees given in Definition 2.22.2 by homotopy theory, which are
integers and hence it gives a complete classification of spherical maps by integers.
The same homotopy classification theorem of continuous spherical maps by their
degrees defined by homology theory is available in Chapter 3.

Theorem 2.25.19 (Brouwer degree theorem) Let f , g : S n → S n be two continuous

maps such that f  g, then deg f = deg g.

Proof Let f  g. Then their induced homomorphisms f∗ and g∗ are such that

f∗ = g∗ : πn (S n ) → πn (S n ) for all integers n ≥ 1.

This asserts that deg f = deg g. 

Theorem 2.25.20 (Hopf Classification Theorem) Let f , g : S n → S n be two con-

tinuous maps. Then f  g, iff deg f = deg g.

Proof First. let f  g. Then it follows by Theorem 2.25.19 that deg f = deg g.
Conversely, let deg f = deg g. If f , g : S 1 → S 1 are continuous maps such that
deg f = deg g, then by Theorem 2.19.9, it follows that f  g. This shows that the
theorem is true for n = 1. Use induction on n and apply Theorem 2.25.4 to complete
the proof. 

Corollary 2.25.21 Let Id : S n → S n be the identity map and n be an even integer.

Then Id is not homotopic to any continuous map f : S n → S n free from fixed points.

Proof Let n be an even integer and f : S n → S n be a continuous map having no fixed

point. Then by Theorem 2.25.10, the map

deg f = (−1)n+1 = −1.

Since deg 1d = 1 and deg f = −1, the maps f and 1d cannot be homotopic by Hopf
classification theorem 2.25.20. 

Remark 2.25.22 Hopf classification theorem 2.25.20 completely classifies con-

tinuous spherical maps f : S n → S n by degree function deg : f → deg f . The same
classification is also given in Chapter 3 by degree function through homology theory.
134 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

2.25.2 Brouwer-Poincaré Theorem

Brouwer- Poincaré Theorem 2.25.23 asserts that there is a continuous nonvanishing

vector field f : S n → S n (n ≥ 1), iff n is odd. On the other hand, Corollary 2.25.24
shows that for all even integers n ≥ 1, there is no vector field f : S n → S n .

Theorem 2.25.23 (Brouwer-Poincaré) The n-sphere S n admits a continuous non-

vanishing vector field iff n is odd.

Proof Suppose the integer n ≥ 1 is odd. Then it follows by Proposition 2.18.10

that there is a nonvanishing vector field f : S n → S n . For the converse, let v be a
nonvanishing vector field on S n . Define

H : S n × I → S n , (x, t) → x cos(π t) + v(x) sin(π t),

then H is a homotopy between the identity map on S n and the antipodal map A : S n →
S n . Hence the antipodal map A has degree 1. But deg A = (−1)n+1 by Proposition
2.25.9. It proves that n is odd. 

Corollary 2.25.24 The n-sphere S n admits no continuous nonvanishing vector field

if n is even.

Proof It follows by method of contradiction. If v is a nonvanishing vector field, then

the map
v(x)
H : S 2n × I → S 2n , (x, t) → x cos π t + sin π t
||v(x)||

defines a homotopy between the identity map and the antipodal map. This asserts
a contradiction that the degree of the identity map on S 2n is (−1)2n+1 = −1. This
contradiction proves the corollary. 

Remark 2.25.25 If n is odd, the difficult problem of determining the maximum

number of linearly independent nowhere vanishing vector fields on S n was solved
by J.F Adams in 1962 by using K-theory [Adams, 1962].

2.25.3 Separation of the Euclidean Plane and Jordan Curve

Theorem

This subsection studies separation of Euclidean plane R2 by proving some theorems

including Jordan Curve Theorem. Many questions naturally arise while studying
the topology of R2 as a continuation of study of analysis. Jordan curve theorem is
one them. An alternative proof of Theorem 2.25.31 with a homological proof of the
theorem is given in Chapter 6.
2.25 Application of Degree of Spherical Maps on S n … 135

Fig. 2.21 Jordan curve

theorem

Recall that a surface S is said to be simply connected, if every closed curve on S

can be continuously deformed into a point without leaving the surface S.

Example 2.25.26 (i) The sphere S 2 is simply connected.

(ii) The torus T is not simply connected.

Definition 2.25.27 A subset X ⊂ R2 is said to separate the Euclidean space R2

if R2 − X has more than one component.

Definition 2.25.28 A subspace J ⊂ R2 homeomorphic to the circle S 1 is said to be

a Jordan curve or a simply closed curve in R2 . It is geometrically represented in
Fig. 2.21.

Theorem 2.25.29 (Jordan Curve Theorem) If J is a Jordan curve in R2 , then R2 − J

has exactly two components, one is bounded and the other one is unbounded. The
curve J is their common boundary.

Remark 2.25.30 Jordan Curve Theorem 2.25.29 appears to be trivial geomet-

rically but it is difficult to prove. It was a conjecture posed by Jordan in 1892 and
its first correct proof was given by Oswald Veblen (1880–1960) in 1905. Its gen-
eral statement known as known as Jordan–Brouwer theorem (see Chap. 6) was later
proved by Brouwer. Original form of Jordan curve theorem 2.25.31 is proved first,
from which its another form is given in Corollary 2.25.32. This proof is based on
homotopy theory. The proof of Jordan curve theorem by homology theory is available
in Chapter 6.

Theorem 2.25.31 (Jordan Curve Theorem) Let J be a subspace of the Euclidean

plane R2 , which is homeomorphic to the circle S 1 . Then J separates the Euclidean
plane R2 .
136 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proof Let p : S 2 − {N } → R2 be the stereographic projection with f : R2 → S 2 −

{N } its inverse homeomorphism, where N = (0, 0, 1) is the north pole of S 2 . Given a
point q ∈ f (J ), let g : R2 → S 2 − {q} be a homeomorphism. Let X ⊂ R2 be defined
by

X = g −1 (f (J ) − {q}).

Then X is homeomorphic to the real line R. So, X may be considered as a line

extended up to infinity on both sides in R2 . Hence the spaces R2 − J , S 2 − f (J ) and
R2 − X have the same number of components. To prove the theorem, it is sufficient
to prove that R2 − X is not connected. If possible, let R2 − X be connected. By
construction of X , it follows that R2 − X is path connected, since every connected
3 3
open subset is path connected. Let R+ and R− be the upper and lower open half-
spaces of R given by z > 0 and z < 0. Define the sets
3

U = R+
3
∪ {(x, y, z) : (x, y) ∈ R2 − X , −1 < z ≤ 0}

and
V = R−
3
∪ {(x, y, z) : (x, y) ∈ R2 − X , 0 ≤ z < 1}.

Hence it follows that U ∪ V = R3 − X , where U ∩ V is homeomorphic to the prod-

uct space (R2 − X ) × (−1, 1), which is path connected. Again since U and V are
simply connected, because, any loop on them can be moved vertically until it stays
3
in either R+ or R−3
followed by deforming it to a point and hence the space R3 − X
is simply connected by using Exercise 60 of Section 2.28 and hence its fundamen-
tal group π1 (R3 − X ) ∼ = {0}. Moreover, there is a homeomorphism h : R3 → R3
with the image h(X ) of X is the z-axis by using Exercise 63 of Section 2.28.
Then it follows that the spaces R3 − X and R3 − (z axis) are homeomorphic. But
the space R3 − (z axis) is homotopy equivalent to space R2 − {0}. This asserts
that the fundamental groups π1 (R3 − X ) and π1 (R2 − {0}) are isomorphic. But
π1 (R2 − {0}) ∼= π1 (S 1 ) ∼
= Z. On the other hand, π1 (R3 − X ) ∼
= {0}. This contra-
diction implies that the space R2 − X is not connected. This proves that the space J
separates R2 . 
To prove Jordan curve theorem it is sufficient to prove its equivalent statement in
Corollary 2.25.32.
Corollary 2.25.32 (An Alternative Form of Jordan Curve Theorem) If J is Jordan
curve in R2 , then R2 − J has exactly two components, one bounded and the other
one is unbounded. The curve J is their common boundary as shown in Fig. 2.21.
Proof The corollary follows from Theorem 2.25.31. 
Remark 2.25.33 The Jordan curve theorem 2.25.29 also asserts that every Jordan
curve divides the surface of the sphere into two disjoint regions, for which the given
Jordan curve is a common boundary. Because, S 2 − J is locally path connected, its
path components and connected components coincide.
2.25 Application of Degree of Spherical Maps on S n … 137

2.25.4 Separation of the Euclidean Plane Problem: Tietze

Extension Theorem

This subsection proves that analogous of Jordan Curve theorem 2.25.31 fails if S 1 is
replaced by I = [0, 1] in Definition 2.25.28. The proof is based on Tietze extension
theorem, which asserts that if A is a closed subspace of a normal space X , then every
continuous map g : A → R has a continuous extension over X (see Chap. 6, Basic
Topology, Volume 1 of the present series of books).

Theorem 2.25.34 Let X ⊂ R2 be a subspace which is homeomorphic to the closed

interval [0, 1]. Then X does not separate the Euclidean plane R2 .

Proof If possible, let R2 − X have more than one component. By hypothesis, X is

homeomorphic to the closed set I = [0, 1]. Then X is compact and hence it is bounded
and R2 − X has a unique unbounded component. Let B be a bounded component of
R2 − X . Construct a disk D2 with center at the origin having sufficiently large radius
so that X ∪ B is in its interior. For b ∈ B, let r : D2 − {b} → S 1 be the retraction
along the straight lines joining the point b to the points of the boundary S 1 of D2 .
Consider its restriction map

f = r|D2 −B : D2 − B → S 1 .

By hypothesis, X is homeomorphic to [0, 1] hence it follows that the map g =

r|X : X → S 1 has a lifting g̃ : X → R such p ◦ g̃ = g, where p : R → S 1 , t → e2πit .
Hence by Tietze extension theorem g̃ extends to a map h̃ : X ∪ B → R. If h =
p ◦ h̃ : X ∪ B → S 1 , then the map f ∪ h : D2 → S 1 obtained by gluing the maps f
and h together is continuous by the pasting lemma. But (f ∪ h)(x) = f (x), ∀ x ∈
S 1 . This implies that the map f ∪ h is a retraction, which contradicts the Corollary
2.19.6 saying that there exists no retraction of D2 onto its boundary ∂ D2 = S 1 . This
contradiction proves that X does not separate the Euclidean plane R2 . 

2.25.5 Applications of the Euler Characteristic in the Theory

of Convex Polyhedra

This subsection proves a classical theorem asserting that there are only five different
types of platonic solids in Theorem 2.25.38. It is an interesting application of the Euler
characteristic in the theory of convex polyhedra. Its proof is based on considering the
surface of a convex polyhedron as glued together a finite number of convex polygons
with respect to identity map on edges glued.
 138 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Definition 2.25.35 For a given polyhedron P, if n edges meet at each vertex and
each face is a convex m-gon, then the polyhedron P is said to be of type [m, n]. In
particular, P is said to regular if every m-gon is regular.
Remark 2.25.36 If the type [m, n] of a polygon P is known, then the number V of
the vertices, the numbers E of the edges and the number F of the faces of P can be
calculated.
Definition 2.25.37 A platonic solid is a polyhedron such that its faces are congruent
regular polygons and each vertex lies in the same number of edges. It is also called
a regular simple polyhedron.
Theorem 2.25.38 There are only five platonic solids which are precisely of types:
[3, 3], [4, 3], [3, 4], [5, 3], and [3, 5].
Proof Let P be a platonic solid with V number of vertices, E number of edges, and
F number of faces. If n is the number of edges meeting in each vertex and m is the
number of edges in each face, then from the geometric point of view, it is assumed
that n, m, V ≥ 3 and it follows that

nV = 2E and m F = 2E,

since n edges meet in each vertex and each edge joins two vertices and two faces.
This asserts by Euler formula V − E + F = 2 for P such that

V E F V−E+F 2 4mn
= −1 = −1 = −1 = −1 = .
n−1 2 m n − 2−1 + m−1 n − 2−1 + m−1 2m − mn + 2n

Then the values V, E and F which are all positive integers, are determined. Hence it
follows that for m, n ∈ N, there are only five solutions of [m, n] which are precisely,
[3, 3], [4, 3], [3, 4], [5, 3], and [3, 5]. They represent geometrically precisely the
five regular polyhedra which are tetrahedron ([3, 3]), cube ([ 4,3]), octahedron ([3,
4]) , dodecahedron ([5, 3]), and icosahedron ([3, 5]). 
Example 2.25.39 Possible values of V, E, F are given for
(i) tetrahedron ([3, 3]): V = 4, E = 6, F = 4.
(ii) cube ([ 4,3]): V = 8, E = 12, F = 6.
(iii) octahedron ([3, 4]): V = 6, E = 12, F = 8.
(iv) dodecahedron ([5, 3]): V = 12, E = 30, F = 20.
(v) icosahedron ([3, 5]): V = 20, E = 30, F = 12.

2.25.6 Interior and Boundary Point Problem of a Surface

This subsection applies homotopy and fundamental group to prove in Theorem

2.25.40 that a point cannot lie both in the interior and the boundary of a surface.
2.25 Application of Degree of Spherical Maps on S n … 139

A surface is Hausdorff space in which every point has a nbd homeomorphic either
to R2 or to the closed half space R+ 2
= {(x, y) ∈ R2 : y ≥ 0}. The interior points of
S, denoted by Int S, are precisely, the points of S, each having a nbd homeomorphic
to R2 . On the other hand, the set of points x of S for each of which, there is a nbd
Ux and a homeomorphism f : R+ @
→ Ux such that f (x) = 0, form the boundary of
S, denoted by ∂S.

Theorem 2.25.40 Given a surface S, let Int S be its interior and ∂S be its boundary.
Then are disjoint.

Proof Suppose there is a point x ∈ Int S ∩ ∂S. Then there exist nbds U1 , U2 of the
point x in S and homeomorphisms

h1 : R+
2
→ U1 and h2 : R2 → U2

with the property that h1 (0) = x = h2 (0). Select a half disk D+ ⊂ R+

2
with center at
origin and radius small enough such that h1 (D+ ) ⊂ U2 . Consider the map

h = h−1
2 ◦ h1 : D+ → R .
2

Since, the maps h1 , h2 are homeomorphisms, the subset h(D+ ) ⊂ R2 is a nbd of the
point 0 in R2 . Take a disk D in R2 with center at the origin and radius sufficiently
small so that D ⊂ h(D+ ). Let ∂D denote the boundary circle of radius ρ of D and
x
p : R2 − {0} → ∂D, x → ρ( )
||x||

denote the radial projection. Then the restriction map

r|h(D+ −{0}) : h(D+ − {0}) → ∂D

is a retraction. Hence this map induces an epimorphism

r∗ : π1 (h(D+ − {0}) → π1 (∂(D)),

which is not possible since π1 (h(D+ − {0})) = 0 and π1 (∂(D)) = Z. This contra-
diction proves that the sets Int S and ∂S are disjoint. 

Theorem 2.25.41 Given two homeomorphic surfaces S1 and S2 , every homeomor-

phism f : S1 → S2
(i) sends the interior Int S1 of S1 to the interior Int S2 of S2 ; and
(ii) sends the boundary ∂S1 of S1 to the boundary ∂S2 of S2 .

Proof (i) Suppose x ∈ Int S1 . Then there exists a nbd Ux of x in the surface S1 and a
homeomorphism h : R2 → Ux . As by hypothesis, f is a homeomorphism, f (Ux )
140 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

is a nbd of f (x) in the surface S2 and f ◦ h : R2 → f (Ux ) is a homeomorphism.

This shows that the point f (x) is in Int S2 and hence the first part is proved.
(iii) Again since, f −1 : S2 → S1 is also a homeomorphism, it follows by the first part
that f −1 sends the interior Int S2 of S2 to the interior Int S1 of S1 . Thus f maps
Int S1 of S1 onto Int S2 of S2 . Then the second part follows under the given
homeomorphism f , since Int S and ∂S for any surface S are disjoint by Theorem
2.25.40. 

Corollary 2.25.42 Homeomorphic surfaces contain homeomorphic boundaries.

Proof It follows from Theorem 2.25.41. 

2.25.7 Homotopy Property of Infinite Dimensional Sphere

S∞

This subsection studies homotopy property of the infinite dimensional sphere S ∞

and proves that this space is contractible.

For construction of S ∞ , we need the concepts of infinite dimensional Euclidean

space R∞ and infinite dimensional unitary space C∞ .

Definition 2.25.43 (i) (Infinite dimensional Euclidean space R∞ ) It consists

∞

of all sequences x = (x1 , x2 , . . . , xn , . . .) of real numbers such that |xn |2
1
converges, i.e.,
∞

R∞ = {x = (x1 , x2 , . . . , xn , . . .) : xn ∈ R and |xn |2 < ∞}.
1

R∞ is a vector space over R under pointwise addition and scalar multiplication.

R∞ endowed with a norm function
∞

x = ( |xn |2 )1/2 ,
1

is a real Banach space. The space R∞ is called infinite dimensional Euclidean

space.
(ii) (Infinite dimensional unitary space C∞ ) It is defined in a way analogous to
infinite dimensional Euclidean space R∞ . The space C∞ is a complex Banach
space and as a topological space Cn is homeomorphic to R2n and C∞ is home-
omorphic to R∞ .

The concept of S ∞ is formulated in Definition 2.25.44.

2.25 Application of Degree of Spherical Maps on S n … 141

Fig. 2.22 Commutative

diagram involving S 2n+1
and Cn+1

Definition 2.25.44 (Infinite dimensional sphere S ∞ ) It is the subspace of R∞ con-

sisting of all real sequences x = (x1 , x2 , x3 , · · · ) such that x12 + x22 + x32 + · · · = 1,
i.e.,
S ∞ = {x = (x1 , x2 , x3 , · · · ) ∈ R∞ : x12 + x22 + x32 + · · · = 1}

and endowed with weak topology in the sense that a subset A ⊂ S ∞ is closed iff
A ∩ S n is closed for each n ≥ 1, where the chain of inclusions are represented in the
commutative diagram as shown in Fig.2.22.
S ∞ as the subspace of C∞ consists of the sequences z = (z1 , z2 , . . . ) over C such
that |z1 |2 + |z2 |2 + · · · = 1.
Proposition 2.25.45 proves the contractibility of the topological space S ∞ , though
n
S is not contractible for any finite value of n.
Proposition 2.25.45 The infinite dimensional sphere S ∞ is contractible.
Proof To prove the proposition, define the map
H : S ∞ × I → S ∞ , (x1 , x2 , x3 , . . . , t) → ((1 − t)x1 , tx1 + (1 − t)x2 , tx2 + (1 − t)x3 , . . . )/Nt ,

where the denominator Nt = [((1 − t)x1 )2 + (tx1 + (1 − t)x2 )2

+ (tx2 + (1 − t)x3 ) + . . . ] , is the norm of the nonzero vector of the numerator.
2 1/2

Parameterize H as

Ht (x1 , x2 , x3 , . . . ) = H (x1 , x2 , x3 , . . . , t), ∀ t ∈ I.

Then H0 (x1 , x2 , x3 , . . . ) = (x1 , x2 , x3 , . . . ), since N0 = 1 and H1 (x1 , x2 , x3 , . . . ) =

(0, x1 , x2 , x3 , . . . ), since N1 = 1. This asserts that H0 is the identity map 1d : S ∞ →
S ∞ , the image of H1 is given by X = {x ∈ S ∞ : x1 = 0} and H : H0  H1 .
Define another homotopy

G : X × I → S ∞ : G(x1 = 0, x2 , x3 , . . . , t) → (t, (1 − t)x2 , (1 − t)x3 , . . . )/Nt ,

where Nt = [t 2 + ((1 − t)x2 )2 + ((1 − t)x3 )2 + · · · ]1/2 .

Let i : X → S ∞ be the inclusion map and c : X → S ∞ be a constant map. Then
G : i  c.
Define the map

∞ H (x, 2t), 0 ≤ t ≤ 1/2
G ∗ H : X × I → S , (x, t) →
G(x, 2t − 1), 1/2 ≤ t ≤ 1.
142 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

The map G ∗ H is clearly a contraction. Hence it is proved that the space S ∞ is

contractible. 
Corollary 2.25.46 The inclusion map i : S n−1 → S n is nullhomotopic in the sense
that it is homotopic to a constant map.
Remark 2.25.47 The infinite dimensional sphere S ∞ is contractible. On the other
hand, the n-sphere S n is not contractible for any finite integer n ≥ 0, because πn (S n ) ∼
=
Z = {0} by Hopf Theorem 2.21.5 (see also Remark 2.25.56).

2.25.8 Homotopy Property of Infinite Symmetric Product

Space SP ∞ X

This subsection studies homotopy property of the infinite symmetric product space
SP ∞ X for any pointed topological space X and proves that this space is contractible.
Construction of SP ∞ X : Given a pointed topological space X with base point x0 , let
X n = X × × · · · × X denote its nth Cartesian product for every n ≥ 1. Let Sn denote
the symmetric groups on the set {1, 2, . . . , n} of n elements. Consider a right action

σ : X n × S n → X n , (x1 , x2 , . . . , xn ) · σ = (xσ (1) , xσ (2) , . . . , xσ (n) ).

Then its orbit space denoted by X n mod Sn is called the nth symmetric product
of X and the equivalence class of (x1 , x2 , . . . , xn ) is denoted by [x1 , x2 , . . . , xn ].
Consider the inclusion maps

i : SP n X → SP n+1 X , [x1 , x2 , . . . , xn ] → [x0 , x1 , . . . , xn ], ∀ n ≥ 1

and construct the union SP ∞ X = SP n X endowed with the weak topology in

n
the sense that a subset A ⊂ SP ∞ X is closed iff A ∩ SP n X is closed for each n ≥ 1,
where the chain of inclusions are represented in the diagram as shown in Fig.2.23.
The topological space SP ∞ X is called the infinite symmetric product space of
X . The elements of SP n X are thus unordered n-tuples [x1 , x2 , . . . , xn ], ∀ n ≥ 1 and
SP ∞ X is a pointed topological space with the base point 0 = [x0 ]. Clearly, there is a
natural inclusion i : X → SP ∞ X , where X = SP 1 X is taken. Every continuous map
f : X → Y in Top ∗ induces a sequence of continuous maps

f n : X n → Y n , (x1 , x2 , . . . , xn ) → (f (x1 ), f (x2 ), . . . , f (xn )).

Since these maps are compatible with the action, they induce another sequence of
continuous maps SP n (f ) : SP n X → SP n Y and hence they induce a continuous map

SP ∞ (f ) = f∗ : SP ∞ X → SP ∞ Y

for every continuous map f : X → Y in Top ∗ .

 2.25 Application of Degree of Spherical Maps on S n … 143

Fig. 2.23 Commutative

diagram involving infinite
symmetric product spaces
and their maps

Remark 2.25.48 Every base point preserving continuous map f : X → Y induces

a map f∗(n) : SP n X → SP n Y , compatible with the action of the group Sn making the
diagram in Fig. 2.23 commutative.
This induces a continuous map f∗ : SP ∞ X → SP ∞ Y in Top ∗ . Moreover,
(i) Proposition 2.25.49 proves the functorial properties of SP ∞ in category Top ∗ .
(ii) Proposition 2.25.50 proves that SP ∞ : Top ∗ → Top ∗ is a covariant functor.
(iii) Theorem 2.25.54 proves that SP ∞ : Htp ∗ → Htp ∗ is a covariant functor.

Proposition 2.25.49 Let Top ∗ be the category of pointed topological spaces.

(i) If f = 1X : X → X is the identity map on X , then its induced map

f∗ = 1SP∞ X : SP ∞ X → SP ∞ X is the identity map on SP ∞ X .

(ii) If f : X → Y and g : Y → Z are in Top ∗ , then (g ◦ f )∗ = g∗ ◦ f∗ : SP ∞ X →
SP ∞ Z.

Proof It follows that from the definition of the induced map f∗ . 

Proposition 2.25.50 SP ∞ : Top ∗ → Top ∗ is a covariant functor.

Proof The objective function assigns to every topological X in Top ∗ , the topological
space SP ∞ X in Top ∗ and the morphism function assigns to every continuous map
f : X → Y in Top ∗ , the continuous map SP ∞ (f ) = f∗ : SP ∞ X → SP ∞ Y in Top ∗ .
By using Proposition 2.25.49, it follows that SP ∞ : Top ∗ → Top ∗ is a covariant
functor. 

Theorem 2.25.51 proves that for any map f : X → Y in Top ∗ , the map

SP ∞ (f ) : SP ∞ X → SP ∞ Y

is homotopy preserving.

Theorem 2.25.51 If f  g : X → Y in Top ∗ , then SP ∞ (f )  SP ∞ (g).

Proof By hypothesis, f  g : X → Y in Top ∗ . Then there exists a homotopy H :

X × I → Y such that

H (x, 0) = f (x) and H (x, 1) = g(x), ∀ x ∈ X .

Define a sequence of maps

H (n) : X n × I → Y n , (x1 , x2 , . . . , xn , t) → (H (x1 , t), H (x2 , t), . . . , H (xn , t)), ∀ n ≥ 1.

144 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Then every map H (n) is continuous, since its projection onto each coordinate function
is continuous. Again, since the symmetric group Sn acts on X n × I by permuting the
coordinate of X n and fixing I, it follows that every continuous map H (n) respects
this action. Hence H (n) induces maps SP n (H ) : SP n X → SP n Y . Now passing to the
limit, it induces a continuous map

SP ∞ (H ) : SP ∞ X × I → SP ∞ Y .

Parametrize H as Ht : X → Y , x → H (x, t), ∀ t ∈ I and consider the map

SP ∞ (H ) : SP ∞ X × I → SP ∞ Y , (x, t) → SP ∞ (Ht )(x).

This implies that

SP ∞ (H ) : SP ∞ (f )  SP ∞ (g). 

Corollary 2.25.52 Let X and Y be two topological spaces in Top ∗ such that X  Y .
Then the corresponding infinite symmetric spaces satisfy the same relation in the
sense that SP ∞ X  SP ∞ Y .

Proof By hypothesis, X  Y . Then the exists a homotopy equivalence f : X → Y

with homotopy inverse g : Y → X . This implies that SP ∞ (g) is a homotopy inverse
of SP ∞ (f ). This proves that SP ∞ X  SP ∞ Y . 

Corollary 2.25.53 For every contractible space X ∈ Top ∗ , the space SP ∞ X is also
contractible.

Proof By hypothesis, the space X is contractible. Then X  {∗}. This asserts that
SP ∞ X  SP ∞ {∗} = {∗}. It proves that the space SP ∞ X is also contractible. 

Proposition 2.25.50 says that SP ∞ : Top ∗ → Top ∗ is a covariant functor on the

category Top ∗ . It is also a covariant functor on the category Htp ∗ as proved in Theorem
2.25.54.

Theorem 2.25.54 SP ∞ : Htp ∗ → Htp ∗ is a covariant functor.

Proof Using Theorem 2.25.51 and Proposition 2.25.50, the theorem is readily
proved. 

2.25.9 Brouwer Fixed-Point Theorem for Dimension n

L. E. J. Brouwer (1881–1967) took the first step toward connecting homotopy

and homology by demonstrating in 1912 that two continuous mappings of a two-
dimensional sphere into itself can be continuously deformed into each other if and
only if they have the same degree (that is, if they are equivalent from the point of view
of homology theory). The papers of H. Poincaré (1854–1912) during 1895-1904 can
2.26 More Applications 145

be considered as blue prints for theorems to come. The results of Brouwer during
1910–1912 may be considered the first one of the proofs in algebraic topology. He
proved the celebrated theorem ‘Brouwer fixed-point theorem 2.25.58’ by using the
concept of degree of a continuous spherical map defined by Brouwer himself. This
section proves Brouwer fixed-point theorem for an arbitrary dimension n ≥ 0. In par-
ticular, for n = 2, i.e., Brouwer fixed-point theorem for dimension 2 is also proved
in Theorem 2.23.6 and that of for dimension n = 1 is also proved in Theorem 2.26.6.

Proposition 2.25.55 The n-sphere S n is not contractible for every finite n ≥ 0.

Proof If possible, suppose S n is contractible. Then the identity map 1S n : S n → S n
is homotopic to some constant map. But the identity map 1S n : S n → S n has degree
1 for n ≥ 1, and any constant map f : S n → S n has degree 0. This contradicts Hopf
classification theorem 2.25.20 and hence S n is not contractible for n ≥ 1. For n =
0, S 0 = {−1, 1} is a discrete space. Hence S 0 cannot be contractible. 
Remark 2.25.56 The finite dimensional n-sphere S n is not contractible. On the other
hand, the infinite dimensional sphere S ∞ is contractible (see also Remark 2.25.47 )
Theorem 2.25.57 (Brouwer no retraction theorem) There does not exist any con-
tinuous onto map f : Dn+1 → S n which leaves each point of S n fixed for every finite
n ≥ 0.
Proof Suppose for every integer n ≥ 0, there exists a continuous map f : Dn+1 → S n
such that f (x) = x for all x ∈ S n . Define a homotopy

H : S n × I → S n , (x, t) → f ((1 − t)x).

This shows that S n is contractible. But this contradicts Proposition 2.25.55. 

Theorem 2.25.58 proves Brouwer fixed-point theorem for dimension n ≥ 0. Sep-
arate proofs of Brouwer fixed-point theorem for dimension n = 1 and n = 2 are
given in Theorem 2.26.6 and Theorem 2.23.6 respectively. This theorem provides
a sufficient condition such that every continuous map from a particular topological
space to itself must have a fixed point. This result is one of the most important results
having applications of topology in mathematics and other sciences.

Theorem 2.25.58 (Brouwer fixed-point theorem for dimension n) Every continuous

map f : Dn+1 → Dn+1 has a fixed point for every finite n ≥ 0.
Proof If possible, f has no fixed point. This implies that f (x) = x for each x ∈ Dn .
For n = 0, as it an immediate contradiction, it is well assumed from now that n ≥ 1.
By assumption, for each x ∈ Dn , the points x and f (x) are distinct. For any x ∈ Dn
we now consider the half-line in the direction from f (x) to x. Let h(x) denote the
point of intersection of this ray with S n . Then the map h : Dn+1 → S n is continuous.
Moreover, h(x) = x for every x ∈ S n . This contradicts the Brouwer no retraction
theorem 2.25.57. This proves that f (x) has a fixed point. 
146 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

2.26 More Applications

This section conveys more applications of homotopy and fundamental groups.

Theorem 2.26.1 Let f : S 2 → S 2 be a continuous map.
(i) If f has no fixed point, then f is homotopic to the
antipodal map g : S 2 → S 2 , x → −x.
(ii) If f is nullhomotopic, then it has a fixed point.
Proof (i) Let f : S 2 → S 2 have no fixed point. Consider the map
(1 − t)g(x) + tf (x)
H : S 2 × I → S 2 , : (x, t) → ∈ S 2,
||(1 − t)g(x) + tf (x)||

provided (1 − t)g(x) + tf (x) = 0. But (1 − t)g(x) + tf (x) cannot be 0, otherwise

we have a contradiction. Because if (1 − t)g(x) + tf (x) = 0, then f (x) = −(1 −
t)/t)g(x) implies that ||f (x)|| = 1 = || − (1 − t)/t)g(x)|| and hence (1-t) /t = 1
would imply that f (x) = −g(x) = x. This shows that f has a fixed point, which
is a contradiction.
(ii) Let f : S 2 → S 2 be nullhomotopic. Suppose the continuous map f : S 2 →
S has no fixed point. Then by (i), f is homotopic to the antipodal map g : S 2 →
2

S 2 , x → −x and hence deg f = deg g. But deg f = 0, since f is nullhomotopic

and deg g = 0. Hence it produces a contradiction, which asserts that f has a fixed
point. 
Proposition 2.26.2 If v : S 2n+1 → R2n+2 is a nowhere vanishing tangent vector field
on S 2n+1 , then the map

v(x)
f : S 2n+1 → S 2n+1 , x →
||v(x)||

is homotopic to the identity map.

Proof Consider the map
πt v(x) πt
H : S 2n+1 × I → S 2n+1 , (x, t) → x cos + sin .
2 ||v(x)|| 2

Then H is a homotopy between the identity map on S 2n+1 and f .


Torus (2-torus) is an important geometrical object, which is homeomorphic to the
product space S 1 × S 1 . It can be studied through homotopy theory.

Theorem 2.26.3 For the 2-torus T 2 = S 1 × S 1 , its fundamental group π1 (S 1 ×

S 1) ∼
= Z ⊕ Z.
2.26 More Applications 147

Proof Using the Theorem 2.14.1, it follows that the fundamental groups π1 (S 1 , 1) ×
π1 (S 1 , 1) and π1 (S 1 × S 1 , (1, 1))) are isomorphic. Again π1 (S 1 , 1) ∼
= Z by Theorem
2.19.1. Since both the spaces S 1 and S 1 × S 1 are path connected, hence it follows
that π1 (S 1 × S 1 ) ∼
= Z×Z∼ = Z ⊕ Z.

n
  
Corollary 2.26.4 For the n-torus T = S 1 × S 1 × · · · × S 1 (n components),
n

n
  
π1 (Tn ) ∼
= Z ⊕ Z ⊕ · · · ⊕ Z.

Proposition 2.26.5 Let X = [a, b] be a subspace of the real line space R and A =
{a, b} be the two-point subspace of X . Then there is no retraction r : X → A.

Proof If possible, there is a retraction r : X → A. Then its restricted map r|A = 1A

(identity map on A) and hence r ◦ i = 1A asserts that the map r is surjective. X =
[a, b] is a connected space, but A is not connected, because, the open sets {a} and {b}
of the discrete space A constitute a nontrivial separation of the space A. This gives a
contradiction. This proves that there is no retraction r : X → A. 

Theorem 2.26.6 (Brouwer fixed-point theorem of dimension 1) Let X = [a, b] be

a subspace of the real line space R. Then every continuous map f : X → X has a
fixed point.

Proof Without loss of generality, assume that X = [−1, 1]. If possible, f : X → X

has no fixed point. Then for every point x ∈ X , f (x) = x. Define a map
x − f (x)
r : X → X , x → .
|x − f (x)|

This map r is well defined and continuous, because, by assumption, for every point
x ∈ X , f (x) = x and f is continuous. If A = {−1, 1}, then r : X → A is a retraction.
This contradicts Proposition 2.26.5. 

Theorem 2.26.7 For n > 1, the n- sphere S n is simply connected.

Proof If N = (0, 0, . . . , 0, 1) ∈ Rn+1 is the north pole of S n and S = (0, 0, . . . , 0,

−1) ∈ Rn+1 is the south pole of S n , then U1 = S n − {N } and U2 = S n − {S} are both
open sets homeomorphic to Rn , which is a simply connected space. Consequently,
U1 and U2 are both simply connected. Again, since U1 ∩ U2 = S n − {N } − {S} ≈
Rn − {0} and the space Rn − {0} is path connected, by using Exercise 25 of Section
2.28. it follows that for n > 1, the n-sphere S n is simply connected. 

Corollary 2.26.8 The Euclidean plane R2 cannot be homeomorphic to the Euclidean

space Rn for n > 2.
148 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Proof Consider the punctures n-dimensional Euclidean space Rn − {0} for n > 2
and the punctured plane R2 − {0}. The former space Rn − {0} is homotopy equivalent
to S n−1 ; on the other hand R2 − {0} is homotopy equivalent to S 1 . Since S n−1 is simply
connected for n > 2 by Theorem 2.26.7 but S 1 is not so by Corollary 2.19.12, hence
the corollary follows. 

2.27 Invariance of Dimensions of Spheres and Euclidean

Spaces

This section solves a homeomorphism problem and a homotopy equivalence

problems on spheres and Euclidean spaces of different dimensions by proving the
topological and homotopy invariance of dimensions of spheres and Euclidean spaces
using homotopy theory. It is proved by using homotopy theory in this section that
the dimension of the spheres and Euclidean spaces are both topological invariants.
The same properties by another method using the homology theory is also proved in
Chapter 5.

Proposition 2.27.1 Let m and n be two distinct nonnegative integers. Then the
spheres S m and S n cannot be homotopically equivalent.

Proof Without loss of generality, assume that 0 ≤ m < n. Then it follows from
Theorem 2.21.4 and Theorem 2.21.5 that

πn (S n ) ∼
= Z, and πn (S m ) = {0},

(though the usual convention in algebraic topology to use 0 in place of the trivial group
{0}). This asserts that the spheres S m and S n cannot be homotopically equivalent.


Corollary 2.27.2 Let m and n be two distinct nonnegative integers. Then the spheres
S m and S n cannot be homeomorphic.

Proof It follows from Proposition 2.27.1. 

Corollary 2.27.3 The dimension of the sphere is a topological invariant.

Proof It follows from Corollary 2.27.2. 

Remark 2.27.4 (i) Proposition 2.27.1 solves homotopy equivalence problem on

spheres of different dimensions.
(ii) Corollary 2.27.2 solves the homeomorphism problem on spheres of different
dimensions.

Proposition 2.27.5 Let n be a nonnegative integer and 0 = (0, 0, . . . , 0) be the

origin of the Euclidean (n + 1)-space Rn+1 . Then the inclusion map
2.27 Invariance of Dimensions of Spheres and Euclidean Spaces 149

i : S n → Rn+1 − {0}

is a homotopy equivalence.

Proof Let X = Rn+1 − {0} and i : S n → X be the inclusion map. Define maps
x
r : X → S n , x →
||x||

and
tx
H : X × I → X , (x, t) → (1 − t)x + .
||x||

Then H and r are continuous maps such that

x
H (x, 0) = x and H (x, 1) = = (i ◦ r)(x), ∀ x ∈ X .
||x||

This shows that H : 1X  i ◦ r and hence i ◦ r  1X . Moreover, r ◦ i = 1S n . This

asserts that i is a homotopy equivalence and hence S n  Rn+1 − {0}. 

Corollary 2.27.6 Let X = Rn+1 − {0} be the punctured Euclidean (n + 1)-space

and S n be the n-sphere in Rn+1 . Then S n is a deformation retract of X .

Proof It follows from Proposition 2.27.5 that S n is a deformation retract of X with

r as a deformation retraction. 

Theorem 2.27.7 Let m and n be two distinct positive integers. Then the Euclidean
m-space Rm and the Euclidean n-space Rn cannot be homotopically equivalent.

Proof Suppose that Rm and Rn are homeomorphic. Then there exists a homeomor-
phism
f : Rm → Rn .

If f (0) = y ∈ Rn , and if

t : Rn → Rn , x → x − y

is the translation of Rn , then the map

h = t ◦ f : Rm → Rn

is a homeomorphism such that h sends the origin 0 ∈ Rm to the origin 0 ∈ Rn . This

determines a homeomorphism

h : h|Rm −{0} : Rm − {0} → Rn − {0}.

150 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

Since S m−1 is homotopy equivalent to Rm − {0} and S n−1 is homotopy equivalent

to Rn − {0} by Proposition 2.27.5, it follows that S m−1 and S n−1 are homotopically
equivalent. By hypothesis, m and n be two distinct positive integers and hence m − 1
and n − 1 are two distinct nonnegative integers such that S m−1 and S n−1 are homo-
topically equivalent. But it contradicts the Proposition 2.27.2. 

Corollary 2.27.8 Let m and n be two distinct positive integers. Then the Euclidean
m-space Rm and the Euclidean n-space Rn cannot be homeomorphic.

Proof It follows from Theorem 2.27.7. 

Corollary 2.27.9 The dimension of the Euclidean space is a topological invariant.

Proof It follows from Corollary 2.27.8. 

Remark 2.27.10 (i) Theorem 2.27.7 solves a homotopy equivalence problem on

Euclidean spaces of different dimensions.
(ii) Corollary 2.27.8 solves a homeomorphism problem on Euclidean spaces of dif-
ferent dimensions.

2.28 Exercises and Multiple Choice Exercises

As solving exercises plays an essential role of learning mathematics, various types

of exercises and multiple choice exercises are given in this section. They form an
integral part of the book series.

2.28.1 Exercises

1. Let X , Y , Z be topological spaces and f1 , g1 ∈ C(X , Y ) and f2 , g2 ∈ C(Y , Z) be

maps such that f1  g1 and f2  g2 . Show that f2 ◦ f1  g2 ◦ g1 : X → Z. Hence
prove that the composites of homotopic maps are homotopic.
2. Given a matrix M in the topological general linear group GL(2, R), show that
there is a path α : I = [0, 1] → GL(2, R) such that
(i) if det(M ) > 0, then
α(0) = M

and  
10
α(1) = .
01

(ii) if det(M ) < 0, then

α(0) = M
2.28 Exercises and Multiple Choice Exercises 151

and  
1 0
α(1) = .
0 −1

(iii) Hence, show that GL(2, R) has two connected components.

[ Hint: Construct a family

 of  paths, each taking the matrix
 M to  a sim-
11 1 1
pler form such as M = , when det M > 0 and M = , when
01 0 −1
det M < 0. Then

for the case (i), the path

 
1 1−t
α : I = [0, 1] → GL(2, R), t → ,
0 1

and

for the case (ii), the path

 
1 1−t
α : I = [0, 1] → GL(2, R), t →
0 −1

solve the problem.]

3. Show that

(i) Every contractible space is path connected.

(ii) If X is a contractible space and Y is a path-connected space, then the set of
homotopy classes [X , Y ] of continuous maps from X to Y consists of single
element.

4. Show that the pair (X , A) of topological spaces with A a closed subset of X has
the homotopy extension property iff the space (X × 0) ∪ (A × I) is a retract of
X × I.
5. (Homotopy invariance of mapping cones and mapping cylinders) Let f , g :
X → Y be two homotopic maps. Show that the corresponding

(i) Mapping cones Cf and Cg are homotopy equivalent.

(ii) Mapping cylinders Mf and Mg are homotopy equivalent.

6. Let X be a topological space and A be a closed subspace of X . If the pair

(X × I, X × {0} ∪ (A × I) ∪ (X × {1}) of spaces has the homotopy extension
property (HEP) with respect to X, show that A is a deformation retract of X iff
A is a strong deformation retract of X .
152 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

7. Show that a topological space is contractible iff it is deformable into one of its
points.
8. Let X and Y be two topological spaces having the same homotopy type. Show
that X is path connected iff Y is also so.
9. Show that
(i) The cylinder has the same homotopy type of the circle.
(ii) The Möbius strip has the same homotopy type of the circle.
10. Given a path-connected space X and two points x0 , x1 ∈ X , show that the fun-
damental group π1 (X , x0 ) is abelian if and only if for each pair of paths α, β in
X from x0 to x1 , the homomorphisms

ψα : π1 (X , x0 ) → π1 (X , x1 ), [f ] → [ α ∗ f ∗ α].

and
ψβ : π1 (X , x0 ) → π1 (X , x1 ), [f ] → [ β ∗ f ∗ β].

are the same, where α and β are the inverse paths of α and β, respectively.
11. Given a loop α : (I, İ) → (S 1 , 1) on S 1 , show that

deg(α n ) = n deg α.

12. Show that the map f : S 1 → S 1 , z → z n induces a monomorphism

f∗ : π1 (S 1 , 1) → π1 (S 1 , 1), [α] → [f ◦ α].

13. Show that S 1 and S n (n ≥ 2) are not of the same homotopy type.

[ Hint: π1 (S 1 ) ∼
= Z. On the other hand, π1 (S n ) ∼
= 0, ∀ n ≥ 2. ]
14. Show that given any topological space X , the homotopy set [X , I] consists of a
single element.
15. Let Xβ , Yβ (β ∈ A) be pointed topological spaces and fβ  gβ : Xβ → Yβ : β ∈
A be base point preserving homotopic maps. Show that β∈A fβ  β∈A gβ
relative to the base point.

[ Hint : Let Hβ : Xβ × I → Yβ be a homotopy between fβ and gβ . Define

H : β∈A Xβ × I → ×Yβ , (xβ , t) → (Hβ (xβ , t))

is continuous and is the required homotopy relative to base point.]

16. Given a pointed topological space A, let [A, X ] denote the set of homotopy
classes of base point preserving continuous maps from A to X .
Prove that every base point preserving continuous map f : X → Y induces a
function
2.28 Exercises and Multiple Choice Exercises 153

f∗ : [A, X ] → [A, Y ]

such that
(i) If f  g, then f∗ = g∗ .
(ii) If 1X : X → X is the identity map, then 1X ∗ : [A, X ] → [A, X ] is the identity
function.
(iii) If g : Y → Z is another continuous map, then (g ◦ f )∗ = g∗ ◦ f∗ .
Hence show that if X  Y , then there exists a bijective correspondence between
the sets [A, X ] and [A, Y ].

What are the corresponding results for the sets [X , A] for a fixed pointed topo-
logical space A ?

[ Hint Use Theorems 2.3.2 and 2.3.6 and their corollaries. ]

17. Consider the topological group S 1 = {z ∈ C : |z| = 1} under usual multiplica-
tion of complex numbers. Show that

(i) Given any pointed topological space X , the pointwise multiplication makes
the set of base point preserving continuous maps X → S 1 an abelian group.
(ii) The set [X , S 1 ] admits the structure of an abelian group.
(iii) If f : Y → X is a base point preserving continuous, then its induced map

f ∗ : [X , S 1 ] → [Y , S 1 ]

is a group homomorphism.

18. Given any topological space X , show that the cardinal number card (I, X ]) of
the homotopy set [I , X ] is the number path-connected components of X .
19. Show that a path-connected space is simply connected iff for any two paths in
the space with the same initial point and the same terminal point are homotopic.
20. m Let X be a topological space and U, V be two simply connected open set in
X such

(i) X = U ∪ V .
(ii) U ∩ V is path connected.
Show that X is also simply connected.

21. Let X be a topological group with identity element x0 and (X , x0 ) be the set
of all loops in X based at x0 . For any two loops f , g ∈ (X , x0 ), define a loop
f #g by
(f #g)(t) = f (t)g(t).

Show that

(i) The set (X , x0 ) endowed with the binary operation  # forms a group.
154 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

(ii) The operation  # on (X , x0 ) induces a group  # on π1 (X , x0 ).

(iii) The two group operations ◦ (usual) and  # on π1 (X , x0 ) are identical.
(iv) The group π1 (X , x0 ) is abelian.
[ Hint: Compute (f ◦ x0 )#(x0 ◦ g) and use (i), (ii), (iii). ]

22. Let X be a connected graph which is not a tree. Show that π1 (X ) is a nontrivial
free group.
23. (Van Kampen theorem) Let (X , τ ) be a topological space and U, V be two
open sets in (X , τ ) such that
(i) X = U ∪ V .
(ii) U, V and U ∩ V are nonempty path connected open subsets in (X , τ ).
Then for any point x0 ∈ U ∩ V , show that

π1 (X , x0 ) ∼
= π1 (U, x0 ) ∗π1 (U ∩V ,x0 ) π1 (V , x0 )

where the right hand side denotes the amalgamated product of the groups
π1 (U, x0 ) and π1 (V , x0 ) over the group π1 (U ∩ V , x0 ) through the group homo-
morphisms i∗ and j∗ defined by the inclusion maps

i : (U, x0 ) → (X , x0 ) and

j : (V , x0 ) → (X , x0 ).

Hence show that

(i) The fundamental group of the wedge of n circles is a free group having n
generators.
(ii) If U ∩ V is simply connected, then π1 (X , x0 ) is the free product of the groups
π1 (U, x0 ) and π1 (V , x0 ).
(iii) π1 (S n , s) = 0, ∀ n > 1 and ∀ s ∈ S n .

[ Hint: See Chapter 6 for the proof of Van Kampen theorem in an alternative
form. For the other part, consider the open sets U and V in S n defined by

U = {(x1 , x2 , . . . , xn , xn+1 ) ∈ S n : xn+1 < 1}

and
V = {(x1 , x2 , . . . , xn , xn+1 ) ∈ S n : xn+1 > −1},

which are homeomorphic to Rn . This implies that π1 (U ) = 0 = π1 (V ).

Then use Van Kampen theorem. ]

24. A subspace Y ⊂ Rn is called star-shaped if there exists a point y ∈ Y such that

for any point x ∈ Y , the line segment [y, x] joining the points y and x entirely
2.28 Exercises and Multiple Choice Exercises 155

lies in Y . Show that any continuous map from an arbitrary topological space X
to a star-shaped subspace Y of Rn is nullhomotopic.
25. If {Ua : a ∈ A} be an open covering of a space X , where each Ua is simply
connected, then prove that X is itself simply connected if
(i) ∩ Ua = ∅.
(ii) For a = b ∈ A, Ua ∩ Ub is path connected.
26. Let X be a connected space and f˜ , g̃ be two liftings of f : (X , x0 ) → (S 1 , 1). If
f˜ , g̃ : (X , x0 ) → (R, z0 ) be two lifting of f such that f˜ (x0 ) = g̃(x0 ), then show
that f˜ = g̃.
27. Show that if n is odd, then the antipodal map A : S n → S n , x → −x is homotopic
to the identity map 1d : S n → S n .
28. Let f : S n → S n be a continuous map such that f (x) = x, ∀ x ∈ X . Show that f
is homotopic to the map g :: S n → S n , x → −x.
29. Let X ⊂ Rn be compact and convex. If f : (X , x0 ) → (S 1 , 1) is continuous and
z0 ∈ Z, show that its lifting

f˜ : (X , x0 ) → (R, z0 )

with p ◦ f˜ = f is unique, where p : R → S 1 , t → e2πit is the exponential map.

30. (Homotopy Lifting Property) .Let α, β : (I, İ) → (S 1 , 1) be two loops such
that α  β rel İ, and p ◦ α̃ = α, p ◦ β̃ = β and α̃(0) = β̃(0) = 0, where p :
R → S 1 , t → e2πit is the exponential map. Show that

α̃  β̃ rel İ and α̃(1) = β̃(1).

31. If f is a piecewise differentiable loop in the complex plane C and z0 is a point in

C but not in Im f , show that the winding number w(f ; z0 ) is given by the integral
1 dz
w(f ; z0 ) = .
2π i z − z0
f

32. Show that the circle S 1 is path connected but not simply connected.
33. Given a topological space X , let s0 ∈ S n ⊂ Dn+1 be an arbitrary point and
f : S n → X be a continuous map. Show the following three statements are equiv-
alent.

(i) The map f is nullhomotopic (i.e., f is homotopic to a constant map).

(ii) The map f has a continuous extension over Dn+1 .
(iii) The map f is nullhomotopic relative to {s0 }.

34. Show that the map f : S 1 → R2 − {(0, 0)}, z → z n is not nullhomotopic.

35. Let f : S 1 → S 1 be an antipode preserving map. Show that f is not nullhomo-
topic.
36. Show that there is no antipode preserving continuous map f : S 2 → S 2 .
156 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

37. Show that there exists a continuous map ψ : Dn → S n−1 with the property that
for the inclusion map i : S n−1 → Dn , the composite

ψ ◦ i = 1d : S n−1 → S n−1

is the identity map iff the identity map 1d is nullhomotopic.

38. Let f : S 1 → R2 be a continuous map. Show that there exists a point x0 ∈ S 1
such that f (x0 ) = f (−x0 ).
39. Let f : S 2 → R2 be a continuous map such that f (x) = −f (x), ∀ x ∈ S 2 . Show
that exists a point x0 ∈ S 2 such that f (x0 ) = 0.
40. Let f : S 2 → S 2 be a continuous map. Show that
(i) Either f has a fixed point or
(ii) There is a point x0 ∈ S 2 such that f (x0 ) = −x0 .
41. (Hopf) Show that a continuous map f : S n → S n of degree 0 is homotopic to a
constant map.
42. Show that there does not exist any homeomorphic copy of the ordinary sphere
S 2 in the Euclidean plane R2 .

[ Hint: Use Borsuk–Ulam Theorem 2.23.14].

43. Let X be the disk with g holes. Construct (if possible) a continuous map
f : X → X without fixed points.

[ Hint: Let 0 be an arbitrary point inside one hole of the disk. Let f : X → X be
a continuous map mapping each point X to the point where the ray OX meets
the boundary circle of the disk. Then under f only the points of its boundary
point are fixed. If you turn the circle by a nonzero angle, then a continuous map
is obtained without fixed points.]
44. Let S1 and S2 be two surfaces with boundaries ∂S1 and ∂S2 , which are homeo-
morphic to S 1 . Then they can be glued boundarywise by the homeomorphism
f : ∂S1 → ∂S2 . If S1 ∪f S2 is the corresponding factor space, show that the Euler
characteristic satisfies the relation

κ(S1 ∪f S2 ) = κ(S1 ) + κ(S2 ).

45. Show that the Euler characteristic of

(i) The sphere S 2 is 2.

(ii) The disk is 1.
(iii) The 2-torus T is 0.
(iv) The Möbius strip is 0.
(v) The sphere Sn2 with n holes is 2 -n.
2.28 Exercises and Multiple Choice Exercises 157

Hence show that the sphere S 2 and the torus 2- torus T are homotopically dis-
tinct in the sense that one cannot be deformed into the other by a continuous
deformation without tearing or cutting.
46. Show that the cylinder and the Möbius strip are not homeomorphic.

[ Hint: Use Theorem Th 10.21.39. ]

47. (Poincaré - Alexander theorem) Let X be any topological space homeomorphic
to a polyhedron P. Show that its Euler characteristic κ(X ) is independent of the
polyhedron P provided P is homeomorphic to X .
48. Let S1 and S2 be two compact surfaces. Show that they are homeomorphic iff
(i) κ(S1 ) = κ(S2 ).
(ii) S1 and S2 are both orientable or both nonorientable.

49. Let f , g : X → C be two continuous maps such that

|f (x) − g(x)| < |f (x) + |g(x)|, ∀ x ∈ X .

Show that
(i) Both the maps f /g and g/f are exponential.
(ii) f is exponential if g is also exponential.
50. Let f , g : S 1 → C − {0} be two continuous maps. Let Ind f denotes the index of
f given in Definition 2.8.8.

Show that

(i) Ind (f g) = Ind f + Ind g, where f g : S 1 → C − {0}, x → f (x) · g(x).

(ii) Ind f =0, iff f is an exponential map.
(iii) Ind f = Ind ( |ff | ).

51. Let f : S 1 → S 1 be a continuous map such that deg f = 1. Show that f  1S 1 .

[Hint: Consider the exponential map p : R → S 1 and the lift of the composite
map f ◦ p to a map R → R. ]
52. If f : S 1 → S 1 is a continuous map such that f (1) = 1. Show that Ind f and the
index of the loop α : (I, İ) → (S 1 , 1), t → f (e2πit ) are the same.
53. Show that for the maps f , g : S 1 → C − {0}, the following statements are equiv-
alent:

(i) f  g.
(ii) Ind f = Ind g.
(iii) The map f /g is exponential.
158 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

54. If f : S 1 → C − {0} is a continuous map such that its index Ind f = n, show
that f is homotopic to the exactly one map defined by

g : S 1 → C − {0}, z → z n .

55. Let f , g : (In , ∂ In ) → (X , x0 ) be two homotopic maps relative to ∂ In and ψ, φ :

I → X are two homotopic paths relative to end points. If Ft , G t : In → X : t ∈ I,
are homotopies of f along the path ψ and that of g along the path φ respectively,
show that F1 and G 1 are homotopic relative to ∂ In .
56. Let the set M (n, R) of all n × n real matrices be identified with the Euclidean
2
Rn -space and M ∈ M (n, R). Then its associated linear transformation TM :
R → Rn has a continuous extension f : S n → S n . Show that deg f = det M.
n

57. Show that every rotation of S 2n has a fixed point.

58. Show that there exists a continuous tangent vector field on S n for every odd
integer n.
59. Show that for every even integer n > 0, there exists no continuous tangent vector
field on S n .
60. Let X be a topological space such that it can be expressed as the union of two
simply connected open sets X1 and X2 with their path-connected intersection
X1 ∩ X2 . Show that X is simply connected.
61. Show that there exists no continuous map f : D2 → S 1 such that its restriction
map f |S 1 : S 1 → S 1 is the identity map.
[ Use the results that π(S 1 , 1) = Z and π1 (D2 , 1) = {0}. ]
62. If X and Y are two arcwise connected topological spaces, show that

πn (X × Y ) ∼
= πn (X ) ⊕ πn (Y ), ∀ n ≥ 1.

63. Let X be the closed subset of R3 defined in Theorem 2.25.31, which is homeo-
morphic to R. Show that there exists a homeomorphism ψ : R3 → R3 such that
its image ψ(X ) is the z-axis.
64. Let M be compact manifold of dimension n and v be a nonzero vector field in
M . Show that there is a continuous map f : M → M without a fixed point such
that f is homotopic to the identity map 1M : M → M .
65. Let M be a compact differentiable manifold and m0 ∈ M . Show that its funda-
mental group π1 (M , m0 ) is finitely generated.
66. Given any two connected n -manifolds M and N (n > 2), show that the fun-
damental group π1 (M #N ) of their connected sum, is the free product of the
fundamental groups π1 (M ) and π1 (N ).

2.28.2 Multiple Choice Exercises

Identify the correct alternative (s) (there may be none or more than one ) from the
following list of exercises:
2.28 Exercises and Multiple Choice Exercises 159

1. Let X be a subspace of R3 and κ(X ) be its Euler characteristic.

(i) If X is the tetrahedron, then κ(X ) = 2.

(ii) If X is the cube, then κ(X ) = 3.
(iii) If X is the 2-sphere S 2 , then κ(X ) = 2.

2. Let X be a subspace of R3 homeomorphic to a polyhedron and κ(X ) be its Euler

characteristic.

(i) κ(X ) is an integer.

(ii) κ(X ) is a topological invariant.
(iii) κ(X ) is a homotopy invariant.

3. Let π1 (X , x0 ) be the fundamental group of a topological space X based at the

point x0 ∈ X .

(i) If X is any topological group with identity element x0 , then π1 (X , x0 ) is

abelian.
(ii) If X is any Hopf group with homotopy identity element x0 , then π1 (X , x0 )
is abelian.
(iii) If X is the circle S 1 and x ∈ X is an arbitrary point, then π1 (X , x) ∼
= Z for
all x ∈ X .

4. Let C be the field of complex numbers.

(i) Every polynomial with coefficients in C may not have a root in C.
(ii) The field C is algebraically closed.
(iii) The field R of real numbers is algebraically embedded in the field C.

5. Let Rn be the n-dimensional Euclidean space.

(i) Every Jordan curve J separates the plane R2 .
(ii) The punctured Euclidean 5-space R5 − {0} is homotopy equivalent to the 4
-sphere S 4 .
(iii) The Euclidean plane R2 can be continuously deformed into the Euclidean
space R3 .
6. Let Dn = {x ∈ Rn : ||x|| ≤ 1} be the unit disk in the Euclidean n space Rn .
(i) A continuous map f : D2 → D2 may not have a fixed point.
(ii) Every continuous map f : D2 − {0} from the punctured disk : D2 − {0} to
itself has a fixed point.
(iii) Every continuous map S 2 → S 2 has a continuous extension over D3 if f is
homotopic to a constant map, where S 2 is the boundary of D3 .

7. Let S n be the n-sphere in Rn+1 .

(i) The odd dimensional sphere S 2n+1 admits a nowhere nonvanishing tangent
vector field for every n ≥ 0.
160 2 Homotopy Theory: Fundamental Group and Higher Homotopy Groups

(ii) Every vector field on an even dimensional sphere S 2n vanishes.

(iii) If v : S 2n+1 → R2n+2 is a nowhere vanishing tangent vector field on S 2n+1 ,
then the map
v(x)
f : S 2n+1 → S 2n+1 , x →
||v(x)||

is homotopic to the identity map.

References

Adams JF. Algebraic topology: a student’s guide. Cambridge: Cambridge University Press; 1972.
Adams JF. Vector fields on spheres. Ann of Math. 1962;75:603–32.
Adhikari MR, Adhikari A. Groups. Rings and modules with applications. Hyderabad: Universities
Press; 2003.
Adhikari MR, Adhikari A. Textbook of linear algebra: An introduction to modern algebra. New
Delhi: Allied Publishers; 2006.
Adhikari A, Adhikari MR. Basic topology, vol. 1: metric spaces and general topology. India:
Springer; 2022a.
Adhikari A, Adhikari MR, Basic topology, vol. 2: topological groups, topology of manifolds and
lie groups. India: Springer; 2022b.
Adhikari MR. Basic algebraic topology and its applications. India: Springer; 2016.
Adhikari MR, Adhikari, A. Basic modern algebra with applications. India, New York, Heidelberg,
Springer; 2014.
Martin A. Introduction to homotopy theory. New York: Springer; 2011.
Armstrong MA. Basic topology. New York: Springer; 1993.
Bredon GE. Topology and geometry. New York: Springer; 1983.
Brown R. Topology: a geometric account of general topology, homotopy types, and the fundamental
groupoid. New York: Wiley; 1988.
Chatterjee BC, Ganguly S, Adhikari MR. Introduction to topology. New Delhi: Asian Books; 2002.
Freudenthal H. Über die Klassen von Sphärenabbildungen. Compositio Mathematica. 1937;5:299–
314.
Gray B. Homotopy theory, an introduction to algebraic topology. New York: Academic Press; 1975.
Hatcher A. Algebraic topology. Cambridge University Press; 2002.
Henle M. A combinatorial introduction to topology. New York: Dover Publications; 1979.
Hirsch M, Smale S. Differential equations, dynamical systems and linear algebra. New York: Aca-
demic Press; 1974.
Hu ST. Homotopy Theory. New York: Academic Press; 1959.
Hurewicz W. Beitrage der Topologie deformation. Proc K Akad Wet Ser. 1935;38:112–119, 521-
528.
Lefschetz S. Introduction to topology. Princeton; 1949.
Lie S. Theorie der Transformations gruppen. Math Annalen. 1880;16:441–528.
Massey WS. A basic course in algebraic topology. New York: Springer; 1991.
Maunder CRF. Algebraic topology. London: Van Nostrand Reinhhold; 1970.
Geometry Nakahara M. Topology and physics. Bristol: Institute of Physics Publishing, Taylor and
Francis; 2003.
Patterson EM. Topology. Oliver and Boyd; 1959.
Prasolov VV. Intuitive topology. India: Universities Press; 1995.
Rotman JJ. An introduction to algebraic topology. New York: Springer; 1988.
2.28 Exercises and Multiple Choice Exercises 161

Singer IM, Thorpe JA. Lecture notes on elementary topology and geometry. New York: Springer;
1967.
Spanier E. Algebraic topology. New York: McGraw-Hill Book Company; 1966.
Steenrod N. The topology of fibre bundles. Prentice: Prentice University Press; 1951.
Toda H. Composition methods in homotopy groups of spheres. Ann Math Stud. 49:19621
Whitehead GW. Elements of homotopy theory. New York: Springer; 1978.
Chapter 3
Homology and Cohomology Theories: An
Axiomatic Approach with Consequences

As the classical constructions of homology and cohomology groups are available in

almost every textbook of algebraic topology, this chapter develops homology and
cohomology theories with an emphasis on the axiomatic approach of Eilenberg and
Steenrod with various interesting applications for accessible presentation. It starts
with a brief discussion on simplicial homology theory with Euler-Poincaré theo-
rem, and then it defines singular homology theory which is a natural generalization
of simplicial homology theory. Their dual theories are also discussed. This chapter
also gives a presentation of an approach formulating axiomatization of homology
and cohomology theories which makes the subject algebraic topology elegant and
provides a quick access to further study. These axioms, now called Eilenberg and
Steenrod axioms for homology and cohomology theories, were announced by S.
Eilenberg (1915–1998) and N. Steenrod (1910–1971) in 1945, but they first appeared
in their celebrated book ‘The Foundations of Algebraic Topology’ in 1952. This
approach classifies and unifies different homology (cohomology) groups. Chapter 2
studies fundamental and higher homotopy groups, where the base point plays a key
role. There are several advantages of homology groups over them.
Advantages of homology groups over fundamental and higher homotopy
groups:
(i) All homology groups and higher homotopy groups (of dimension more than 1)
are abelian but all fundamental groups are not abelian.
(ii) Every homotopy group needs a base point, but a homology group needs no base
point.
(iii) From the intuitive viewpoint, homology groups are less intuitive than the fun-
damental group and the higher homotopy groups. Computation of fundamental
group and the higher homotopy groups is more difficult than that of homology
groups.
Historically, homology invented by Henri Poincaré in 1895 was studied by him
during 1895–1904. This homology, now called the simplicial homology, is one of the
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 163
M. R. Adhikari, Basic Topology 3,
https://doi.org/10.1007/978-981-16-6550-9_3
164 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

most fundamental powerful inventions in mathematics. Poincaré in 1899 remarked

on his homology theory which runs as fo1lows:
‘ Assume that one can find in V a manifold of p + 1 dimension whose boundary
consists of n manifolds of p dimension V1 , V2 , . . . , Vn I will express this fact with
the relation V1 + V2 + · · · + Vn ∼ 0, that I will call it homology.’ Henri Poincaré,
1899.
For the study of geometric properties of a topological space, Henri Poincaré pre-
scribed a method of constructing homology groups of the topological space in 1895
and contributed to the fundamental idea of converting the topological problems to
algebraic ones for the first time in the history of topology. He started with a geometric
object, which is a topological space and gave a combinatorial data providing a sim-
plicial complex. These data were used to construct homology groups by utilizing the
tools of linear algebra and boundary relations. The classical simplicial homology
theory is involved of tedious discussion on the concepts of triangulability of the
topological spaces, orientations of simplexes, incidence numbers, subdivisions, sim-
plicial approximation and also the topological invariance of the simplicial homology
groups. Emmy Noether (1882–1935) formulated an algebraic approach correspond-
ing to the geometric approach of homology theory invented by Poincaré S. Lefschetz
(1884–1972) extended simplicial homology theory to singular homology theory in
1931 in his paper [ Lefschetz, 1933] by considering continuous maps from the stan-
dard simplexes to a topological space X , which are called singular simplexes in X
and by using the algebraic properties of singular chain complexes. There are other
important homology groups (topological invariants) such as homology groups for
compact metric spaces constructed by L. Vietoris (1891-2002) in 1927 and homol-
ogy groups for compact Hausdorff spaces constructed by E. Čech (1893–1960) in
1932.
Lefschetz remarked
‘Others (topological invariants) were discovered by Poincaré. They are all tied
up with his homology theory which is perhaps the most profound and far reach-
ing creation in all topology.’
Historically, algebraic topologists began around 1940 to compare different def-
initions of homology and cohomology formulated in the earlier years. Eilenberg
and Steenrod formulated a new approach in 1945 by considering a small number
of their properties (without focusing the tools used for construction of homology
and cohomology groups) as axioms to characterize a theory of homology and coho-
mology. This approach is the most surprising contribution to algebraic topology
since the invention of the homology groups by Poincaré in 1895. Now, it is called the
axiomatic approach formulated by a set of seven axioms announced by S. Eilenberg
and N. Steenrod in 1945 and published in their book in 1952. This approach gives the
subject algebraic topology conceptual coherence and elegance and provides quick
techniques for computing homology and cohomology groups. It unifies different
homology groups (modules) on the category of compact triangulable spaces.
Their dual theories are called cohomology theories. Homology and cohomology
theories are basic topics of study of algebraic topology and are used in different
3.1 Manifolds with Motivation 165

disciplines. Initially, no important result connecting these theories was found but the
twentieth century witnessed their greatest development.
For this chapter, the books [Adhikari 2016], [Bredon 1993], [Eilenberg and Steen-
rod 1952], [Gray 1975], [Hatcher 2002], [Hilton and Wylie 1960], [Hu 1966],
[Poincaré 2010] , [Rotman 1988], [Switzer 1975], [Adhikari and Adhikari 2003]
and [Adhikari and Adhikari 2006] some others are referred in the bibliography.

3.1 Manifolds with Motivation

Before starting simplicial homology theory in Sect. 3.2, one needs an idea of a mani-
fold. This section provides the concept of manifolds with its motivation. A manifold
is a generalization of the familiar geometric objects like curves and surfaces which
are considered locally homeomorphic to R and R2 , respectively. An n-dimensional
real manifold looks locally like Rn but not necessarily globally. A local Euclidean
structure to manifold by introducing the concept of a chart is utilized to use the
conventional calculus of several variables. Due to linear structure of vector spaces,
for their applications in mathematics and in other areas, it needs generalization of
metrizable vector spaces, maintaining only the local structure of the latter. On the
other hand, every manifold can be considered as a (in general, nonlinear) subspace of
some vector space. Both aspects are used to approach the theory of manifolds. Since
dimension of a manifold is a locally defined property, a manifold has a dimension.
Our main interest in this chapter is on finite dimensional topological manifold.

Remark 3.1.1 A topological manifold is a Hausdorff space M such that every point
of M has a nbd homeomorphic to the Euclidean n-space Rn , i.e., every point of
M has a nbd homeomorphic to an open subset of Rn . This means that every topo-
logical manifold is locally Euclidean. In particular, the n-dimensional Euclidean
space Rn is a topological space as well as an n-dimensional real vector space. If
x = (x1 , x2 , . . . , xn ) ∈ Rn , then the real numbers x1 , x2 , . . . , xn are called the coor-
dinates of the point x. A manifold is a geometric object in topology which looks
locally like a small piece of Euclidean space, where a local coordinate system can
be introduced. As calculus is developed on the local geometry of Euclidean space,
tools of calculus can be used to develop the structure and properties of manifolds. For
example, every point on the sphere lies on small curved disk which can be flattened
into a disk in the Euclidean plane R2 . On the other hand, the vertex of the cone has
no nbd which looks like a small piece of R2 . Curves are manifolds of dimension one,
and surfaces are manifolds of dimension two.

Definition 3.1.2 formalizes the intuitive idea of an n-dimensional manifold M

such that each point of M is in a nbd that looks like an open ball in Rn .

Definition 3.1.2 Let M be a Hausdorff space. Then M is said to be an n-dimensional

manifold or simply n-manifold, if there is an open covering C = {Vi : i ∈ A} of M
such that for every i ∈ A, there is a map ψi : Vi → Rn mapping Vi homeomorphically
 166 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

onto an open subset Ui of Rn homeomorphically. Each ordered pair (ψi , Vi ) is called

a local chart (or simply a chart or a coordinate system on Vi ), and the family
ψ = {(ψi , Vi ) : i ∈ A} is called an atlas of the manifold M .

Example 3.1.3 (i) The Euclidean n-space Rn is a manifold of dimension n. Here

for each x ∈ Rn , the map ψx is taken to be the identity map on Rn .
(ii) The n-sphere Sn is a manifold of dimension n. Here for the points x, y ∈ Sn , x =
y, the map ψx is taken to be the stereographic projection ψx : S n − {y} → Rn ,
which is a homeomorphism.
(iii) The n-dimensional projective space RP n (the space of all lines through the origin
0 ∈ Rn+1 ) is a manifold of dimension n.
(iv) Let M (n, R) be the set of all n × n matrices over R, identified with the Euclidean
2
Rn -space and GL(n, R) = {M ∈ M (n, R) : detM = 0}. The set GL(n, R) of
nonsingular matrices is a manifold of dimension n2 .
(v) The 2-torus S 1 × S 1 with product topology is a compact and connected manifold,
which is a closed manifold of dimension 2.

Remark 3.1.4 A detailed study of manifolds including differentiable manifolds is

available in Chap. 3, Basic Topology, Volume 2 of the present series of books.

3.2 Simplicial Homology Theory and Euler-Poincaré

Theorem

This section begins with the homology theory born through the work of Henri
Poincaré (1854–1912) published in his land-marking paper ‘Analysis Situs’ of
1895 [Poincaré 1895].
Poincaré himself defined homology in 1895 in the following way:
Let us consider a manifold V of p dimensions; now let W be a manifold of q dimen-
sions ( q ≤ p) which is part of V . Assume the complete boundary of W consists of λ
manifolds of q − 1 dimensions V1 , V2 , . . . , Vλ We express this fact with the notation
V1 + V2 + · · · + Vλ ∼ 0. More generally, the notation k1 V1 + k2 V2 − k3 V3 − k4 V4 ∼ 0,
where the k  s are integers and V  s are manifolds of q − 1 dimensions, will denote that
there exists a manifold W of q dimension in V such that the complete boundary of
W consists of k1 manifolds similar to V1 , k2 manifolds similar to V2 , k3 manifolds
similar to V3 but oppositely oriented, and k4 manifolds similar to V4 , but oppositely
oriented. Relations of this form are called homologies. Homologies can be considered
like ordinary equations”. [Poincaré 1895]

The motivation leading to the concept of simplicial homology:

While studying two-dimensional manifolds, one can distinguish intuitively non-
homeomorphic manifolds. But for a study of higher-dimensional manifolds, geo-
metric intuition is less effective. So an alternative approach came through the work
of Henri Poincaré: Let M n and N n be two n-dimensional manifolds, and X ⊂ M n
and Y ⊂ N n be two compact submanifolds (subspaces which are C 0 -manifolds).
3.2 Simplicial Homology Theory and Euler-Poincaré Theorem 167

If any m-dimensional submanifold in M n (m ≤ n) is the boundary of an (m + 1)-

submanifold of M n , there is an m-dimensional submanifold in N n , and then the
manifolds M n and N n are nonhomeomorphic. For example, any one-dimensional
compact submanifold of the 2-sphere S 2 is a boundary, but there is one-dimensional
manifold such as the circle S 1 on the torus T = S 1 × S 1 which is not the boundary
of any two-dimensional submanifold of T . On the other hand, if there are subman-
ifolds, which are not boundaries both in M n and N n , we can find certain quantities
associated with these manifolds.

3.2.1 Simplicial Complex

This subsection formally defines simplicial complex based on which simplicial

homology groups are formulated. It is assumed that the readers are familiar with
the concept of standard simplexes in Euclidean space Rn .

Definition 3.2.1 A finite simplicial complex K is a finite collection of simplexes

in some Euclidean space Rn such that
(i) If σp is a simplex in K, then all of its faces are also in K.
(ii) If σp and σq are simplexes in K, then σp ∩ σq is either empty or is a common
face of σp and σq in K.
The dimension of K denoted by dim K is defined to be −1 if K = ∅ and to be m ≥ 0
if m is the greatest integer such that K has an m-simplex.

Remark 3.2.2 A simplicial complex K is said to be finite if it consists of a finite

number of simplexes. Let K be any simplicial complex and |K| ⊂ Rn be the set-
theoretic union of all simplexes in K endowed with a topology such that a subset
Y ⊂ |K| is closed if Y ∩ σpi is closed in σpi for every σpi ∈ K. If K is finite, then this
topology coincides with the subspace topology on |K| inherited from the Euclidean
topology on Rn .

Definition 3.2.3 Let K be a finite dimensional simplicial complex. A topological

space X is said to be a polyhedron if X is homeomorphic to |K| and then K is said
to be a triangulation of X.

Remark 3.2.4 A triangulation of a topological space X consists of a simplicial com-

plex K together with a homeomorphism h : |K| → X . Since a finite simplicial com-
plex is made up of a finite number of simplexes lying in a Euclidean space, its
polyhedron K is compact, hence, every triangulable space is compact, and it is also
a metric space.

Example 3.2.5 Every closed surface is triangulable.

168 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Definition 3.2.6 (Orientation of a simplicial complex) An oriented p-simplex σp for

p ≥ 1 is obtained from a p-simplex σp =< v0 , v1 , . . . , vp > by selecting an ordering
of its vertices. An oriented simplicial complex is obtained from a simplicial complex
by choosing an orientation of each of its simplexes.

Proposition 3.2.7 follows from the above discussion.

Proposition 3.2.7 Let K be a finite-dimensional-oriented simplicial complex. Then
(i) Every polyhedron is a normal Hausdorff space.
(ii) |K| is compact.
(iii) K is finite iff |K| is compact.
Let K be an oriented simplicial complex of dimension n and (K) be the set of the
linear combinations of oriented p-simplexes σp over the field Q given by

(K) = q1 σ1 + q2 σ2 + · · · + qn σn : each qp ∈ Q and each σp is an oriented
p-simplex of Kfor p = 0, 1, 2, . . . , n} .

Every element of (K) is called a rational p-chains of K. Then (K) forms a

vector space over Q under usual compositions. Let V (K) be the subspace of (K),
generated by the elements of the form σ + ρ, where σ and ρ are the same p-simplex of
K with opposite orientations. The quotient space (K)/V (K) denoted by Cp (K; Q)
is called the vector space of rational p-chains of K. Its dimension is the number of
p-simplexes in K. Consider a rational p-chain
β
cp = q1 σp1 + q2 σp2 + · · · + qn σpβi = i=1
i
qi σpi ,

where −σp is denoted by (−1)σp . The boundary of a p-simplex σpi (p > 0) in K is

the rational (p − 1)-chain
j j
∂σp = j [σpi , σp−1 ] σp−1 ,

j
where [σpi , σp−1 ], called incidence number, is an integer defined as follows:
j j
(i) If σp−1 is not a face of σpi , then [σpi , σp−1 ] is taken to be 0.
j j
(ii) If σp−1 is a face of σpi , then [σpi , σp−1 ] is taken to be 1 or −1 according as the
j
orientation of σp−1 induced by the orientation of σpi agrees or not.

Proposition 3.2.8 Let K be an oriented simplicial complex and p > 1. Then

∂p−1 ◦ ∂p : Cp (K) → Cp−2 (K)

is the trivial homomorphism, i.e., ∂p−1 ◦ ∂p = 0 for all p > 1.

Proof It follows from the definition of the boundary homomorphism ∂. 

3.2 Simplicial Homology Theory and Euler-Poincaré Theorem 169

3.2.2 Simplicial Homology Group

This subsection prescribes a method of construction of homology groups, called sim-

plicial homology groups invented by H. Poincaré in 1895 before giving formalization
of the axioms of homology and cohomology theories. Homology group is a basic
topological invariant different from the fundamental group and higher homotopy
groups studied in Chap. 2. Instead of the closed loop at a base point in a topological
space X , for defining the fundamental group, here the sum of paths σ : I → X in
a formal sense called 1-chains and in general p-chains are considered. For the path
σ : I → X , its boundary ∂σ = σ (1) − σ (0) is a formal sum of signed points. Hence
for a chain c = σi , its boundary ∂c is defined by

∂c = ∂σi ,

which is to be closed if ∂c = 0.

Definition 3.2.9 Let K be an oriented simplicial complex and p ≥ 0 be an integer.

Then the image of
∂p+1 : Cp+1 (K) → Cp (K)

is a subgroup of Cp (K), denoted by Bp (K), which is called the p-dimensional bound-

ary group of K, and each element bp ∈ Bp (K) is called a p-dimensional boundary
of K.

Definition 3.2.10 Let K be an oriented simplicial complex and p > 0 be an integer.

Then the kernel of
∂p : Cp (K) → Cp−1 (K)

is a subgroup of Cp (K) denoted by Zp (K), which is called the p-dimensional cycle

group of K, and each element zp ∈ Zp (K) is called a p-cycle of K.

Proposition 3.2.11 Let K be an oriented n-dimensional simplicial complex. Then

Bp (K) is a subgroup of Zp (K) for every integer p with 0 ≤ p ≤ n.

Proof It follows from Proposition 3.2.8. 

Definition 3.2.12 Let K be an oriented simplicial complex and p ≥ 0 be an integer.

Then the quotient group

Hp (K; Z) = Zp (K; Z)/Bp (K; Z)

is called the p-dimensional simplicial homology group of K in coefficient ring Z.

Example 3.2.13 K be the 2-simplex <v0 , v1 , v2 > having orientation induced by the
ordering v0 < v1 < v2 .
(i) 0-simplexes of K are <v0 >, <v1 > and <v2 >.
 170 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

(ii) Positively oriented 1- simplexes of K are <v0 , v1 >, <v1 , v2 > and <v0 , v2 >.
(iii) Positively oriented 2-simplex of K is <v0 , v1 , v2 >.
Then
(i) H0 (K; Z) ∼
= Z.
(ii) H1 (K; Z) ∼
= {0}.
(iii) H2 (K; Z) ∼
= {0}.

Definition 3.2.14 The quotient group Hp (K; Q) = Zp (K; Q)/Bp (K; Q) is called the
pth simplicial homology group of K with coefficients in Q. It is a vector space over
Q. Instead, the quotient group Hp (K; Z) = Zp (K; Z)/Bp (K; Z), the pth simplicial
homology group of K with coefficients in Z is a module over Z.

There are several definitions of simplicial maps, but we use the Definition 3.2.15.
Definition 3.2.15 Let K and L be two simplicial complexes. A simplicial map
between them is a continuous map f : |K| → |L| between their corresponding poly-
hedra which takes simplexes of K linearly onto simplexes of L.

Definition 3.2.16 Let f : |K| → |L| be a simplicial map between oriented simplicial
complexes K and L. Then it induces a group homomorphism

f∗ : Hn (K; Z) → Hn (L; Z), [zn ] → [f (zn )].

f∗ is called the homomorphism induced by f on homology groups.

Proposition 3.2.17 Let Sim be the category of oriented simplicial complexes and
their simplicial maps and Ab be the category of abelian groups and their homomor-
phisms. Then Hn : Sim → Ab is a covariant functor for every integer n ≥ 0.

Proof Here the object function is defined by

Hn : Sim → Ab, K → Hn (K; Z)

and for every simplicial map f : |K| → |L| in the category Sim, the morphism func-
tion is defined by

f → f∗ : Hn (K; Z) → Hn (L; Z), [zn ] → [f (zn )].

Corollary 3.2.18 Let X and Y be two homeomorphic compact polyhedral spaces.

Then
Hn (X ; Z) ∼
= Hn (Y ; Z), ∀ n ≥ 0.

Proof It follows from the functorial property of Hn . 

3.2 Simplicial Homology Theory and Euler-Poincaré Theorem 171

Corollary 3.2.19 Let X and Y be two homotopy equivalent compact polyhedral

spaces. Then
Hn (X ; Z) ∼
= Hn (Y ; Z), ∀ n ≥ 0.

Proof It follows from the functorial property of Hn . 

Definition 3.2.20 A simplicial pair (K, L) is a pair of the simplicial complex K and
its subcomplex L, and a simplicial map f : (K, L) → (X , A) of simplicial pairs is a
simplicial map f : K → X such that f (L) ⊂ A.

3.2.3 Euler-Poincaré Theorem

Euler-Poincaré Theorem 3.2.24 establishes a close link among geometry, topology

and algebra with the help of the Euler characteristic of compact polyhedra.

Definition 3.2.21 Let K be a simplicial complex of dimension n and βp denotes the

number of p-simplexes in K, for p = 0, 1, 2, ..., n. Then the alternative sum

κ(K) = p=0
n
(−1)p βp (3.1)

is called the Euler characteristic of K with Betti numbers βp .

Remark 3.2.22 Betti numbers βp in Equation (3.1) coined by Poincaré (named after
E. Betti) play an important role in algebraic topology to classify topological spaces
based on the connectivity of a p-dimensional simplicial complex K. The number βp is
the same as the rank of the pth homology group Hp (K; Q) with rational coefficients.

Example 3.2.23 By Theorem 3.12.9, the homology group of S n with coefficient

group Z is 
Z, if p = 0, or n
Hp (S n ; Z) ∼
=
{0}, otherwise.

It shows that the Betti numbers β0 = 1, βn = 1, and all other Betti numbers are 0.

Theorem 3.2.24 (Euler-Poincaré theorem) The Euler characteristic of an oriented

simplicial complex K of dimension n is given by

κ(K) = p=0
n
(−1)p βp = p=0
n
(−1)p rank(Hp (K; Q)). (3.2)

Proof Consider the chain complex

∂n+1 ∂n ∂p+1 ∂1 ∂0
C∗ : {0} −−−−−−→ Cn −−−−→ Cn−1 → · · · → Cp+1 −−−−−−→ Cp → · · · −−−−→ C0 −−−−→ {0},
172 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

where Cp stands for the vector space Cp (K; Q), βp = dimQ Cp (K, Q) is the same as
the number of p-simplexes in K, for p = 0, 1, 2, . . . , n and {0} stands for the trivial
group (though the usual convention in algebraic topology is to use 0 in place of the
trivial group {0}). Then

Bp = ∂p+1 (Cp+1 ) ∼
= Cp+1 /ker ∂p+1 = Cp+1 /Zp+1 , ∀ p ≥ 0,

where Bp stands for Bp (K; Q) and Zp stands for Zp (K; Q). This asserts that

dim Bp (K; Q) = dim Cp+1 (K; Q) − dim Zp+1 (K; Q), ∀ p ≥ 0.

Again, Hp (K; Q) ∼
= Zp (K; Q)/Bp (K, Q) implies dim Hp (K; Q) = dim Zp (K; Q) −
¯
dim Bp (K; Q), ∀ p ≥ 0. These two relations assert that

dim Cp+1 (K; Q) = dim Zp (K; Q) − dim Hp (K; Q) + dim Zp+1 (K; Q), ∀ p ≥ −1.

By taking the alternative sum for p = −1, 0, 1, 2, . . . , n, it follows that the Euler
characteristic is given by the formula

κ(K) = p=0
n
(−1)p βp = p=0
n
(−1)p rankHp (K; Q),

where dim Cp (K; Q) = βp and dim Hp (K; Q) = rank Hp (K; Q). 

Corollary 3.2.25 The Euler characteristic of an oriented simplicial complex K of

dimension n is also given by the formula

κ(K) = p=0
n
(−1)p βp = p=0
n
(−1)p rankHp (K; Z).

Proof It follows from Theorem 3.2.24. 

Corollary 3.2.26 Let X be a compact polyhedron. Then its Euler characteristic is

given by the formula

κ(X ) = p=0
n
(−1)p βp = p=0
n
(−1)p rankHp (X ; Z).

Proof It follows from Theorem 3.2.24 by taking K as a triangulation of X . 

The above discussion is summarized in a basic result in Corollary 3.2.27.

Corollary 3.2.27 The rank of the free abelian part of Hp (K; Q) of a finite oriented
complex K is the Betti number βp of K.
3.2 Simplicial Homology Theory and Euler-Poincaré Theorem 173

3.2.4 Topological and Homotopy Invariance of Euler

Characteristics

Euler-Poincaré theorem is a powerful result in topology. For example, this subsec-

tion applies this theorem to prove topological invariance of Euler characteristics in
Theorem 3.2.28 in the sense that two homeomorphic compact polyhedra have the
same Euler characteristic and homotopy invariance of compact polyhedra in Theo-
rem 3.2.29 in the sense that two homotopy equivalent compact polyhedra have the
same Euler characteristics.

Theorem 3.2.28 Two homeomorphic compact polyhedra have the same Euler char-
acteristics.

Proof Euler-Poincaré theorem of simplicial homology asserts that if X is compact

polyhedron, then its Euler characteristic

κ(X ) = p=0
n
(−1)p βp = p=0
n
(−1)p rankHp (X ; Z), (3.3)

which follows from Eq. (3.2). Homology group is a topological invariant in the sense
that if X and Y are two homeomorphic compact polyhedra, then

Hp (X ; Z) ∼
= Hp (Y ; Z), ∀ p ≥ 0.

Consequently, it follows by using Eq. (3.3) that κ(X ) = κ(Y ). 

Theorem 3.2.29 Two homotopy equivalent compact polyhedra have the same Euler
characteristics.

Proof Since homology group is a homotopy invariant in the sense that, if X and
Y are two homotopy equivalent compact polyhedra, Hp (X ; Z) ∼ = Hp (Y ; Z) for all
p ≥ 0. Consequently, it asserts by Eq. (3.3) that κ(X ) = κ(Y ). 

Remark 3.2.30 Theorem 3.2.28 establishes a close link among geometry, topology
and algebra by Euler characteristics of compact polyhedra, which are integers. For a
generalization of theorem 3.2.29 asserting that if the compact polyhedra X and Y are
two homotopy equivalent spaces, then κ(X ) = κ(Y ). Moreover, κ(X ) ∈ Z, which
is an algebraic object.

3.3 Topology of CW-Complexes

This section introduces the concept of CW-complexes by generalizing the concept

of simplicial complexes. The process of this generalization is made by building up
topological spaces by attaching cells successively, starting from a discrete set of
points.
 174 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Definition 3.3.1 Let X be a Hausdorff space. It is said to be a CW -complex if X

is the union of disjoint subspaces ea (a ∈ A), called cells such that it satisfies the
axioms:
(i) Axiom (i): Corresponding to each cell ea , there is an integer n ≥ 0, called its
dimension. Then the cell ea is symbolized by ean .
Let X n denote the union of all cells eak : k ≤ n, which is called the n-skeleton
of X .
(ii) Axiom (ii): There is a map κa : (Dn , S n−1 ) → (X , X n−1 ), called a characteristic
map with the property that its restriction κa | ( Dn − S n−1 ) is a homeomorphism
from Dn − S n−1 onto ean .
(iii) Axiom (C): X is closure finite in the sense that if for each cell ean , the intersection
K(ean ) of all subcomplexes containing ean is a finite subcomplex. (A subset A ⊂ X
is said to be a subcomplex of X , if A is a union of cells ea and ea ⊂ A for ea ⊂ A
in the sense of Definition 3.3.2).
(iv) Axiom (W) : X has the weak topology in the sense that for every subset B ⊂ X ,
the subset B is closed iff B ∩ eak is compact for every cell eak .

X satisfying the above conditions (i) and (ii) is called a cell complex. A cell
complex satisfying the above additional conditions (iii) and (iv) is called a CW -
complex, where C stands for closure finite and W stands for weak topology in
the sense of Axioms (C) and (W).

Definition 3.3.2 Let X be a CW -complex and A ⊂ X . Then A is said to be a sub-

complex of X , if A is a union of cells ea and ea ⊂ A, if ea ⊂ A.

Example 3.3.3 (i) If K is a simplicial complex, then |K| is a CW -complex. A

polyhedron is usually representable as CW -complex with less number of cells
than the original number of simplexes.
(ii) S n is a cell complex having two cells such as e0 = {(1, 0, . . . , 0} and en = S n −
e0
(see Proposition 3.3.4).
(iii) The real line space R is a CW -complex with 0-cells e0 the integers and 1-cells
e1 the intervals [n, n + 1] for all integers n.
(iv) The 2-sphere S 2 considered as a cell complex with each point a 0-cell, is not a
CW -complex, because, it is closure finite but it does not have the weak topology,
and hence, it fails Axiom (W) of Definition 3.3.1.
∞ ∞
(v) (Wedge of the circles Si1 ) Let X = Si1 be an infinite 1-point union of
i=1 i=1
circles. Then X is a CW -complex. In particular, the figure-eight defined in
Chap. 2 is the 1-point union of two circles and is a CW -complex.
(vi) D3 with one 0-cell for every point of S 2 and the cell e3 = D3 − S 2 , is not a
CW -complex, because it has the weak topology but it is not closure finite, and
hence, it fails the Axiom (C) of Definition 3.3.1.
 3.3 Topology of CW-Complexes 175


(vii) RP ∞ = ∞ n
n=1 RP is a closure finite cell complex, and it has the weak topology
given by
RP n ⊂ RP n+1
∞ as an subcomplex (see Exercise 3 of Sect. 3.16).
∞
(viii) CP = n=1 CP is a closure finite cell complex, and it has the weak topology
given by
CP n ⊂ CP n+1
∞ as an subcomplex (see Exercise 3 of Sect. 3.16).
∞
(ix) HP = n=1 HP is a closure finite cell complex, and it has the weak topology
given by
HP n ⊂ HP n+1 as a subcomplex (see Exercise 3 of Sect. 3.16).

Proposition 3.3.4 S n is a CW -complex.

Proof Let e0 = s0 = {(−1, 0, . . . , 0)} ∈ S n and en = S n − e0 .

κn : (Dn , S n−1 ) → (S n , s0 )

be the standard characteristic map. Then κn |en is a homeomorphism onto κn (en ).

Consider the characteristic map ψ : D0 → s0 . Hence, it follows that S n is a CW -
complex with one 0-cell and one n-cell with characteristic maps κn and ψ. 

Definition 3.3.5 Let X and Y be two cell complexes. A map f : X → Y is said to

be cellular if
f (X k ) ⊂ Y k , ∀ X k ,

where X k denotes a k-skeleton of X and Y k denotes a k-skeleton of Y .

Proposition 3.3.6 (i) CW -complexes and their cellular maps forms a category,
denoted CW.
(ii) Pointed CW -complexes and their cellular maps form a category, denoted CW ∗ .

Proof Take all CW -complexes as objects and their cellular maps as morphisms for
the category CW and take all pointed CW -complexes X , where ∗ ∈ X is a 0-cell as
objects and their cellular maps as morphisms for the category CW ∗ . 

Theorem 3.3.7 proves a close link between a CW -complex and its path component.

Theorem 3.3.7 (i) Each path component of a CW -complex is also a CW -complex.

(ii) Each path component of a CW -complex is both open and closed.
(iii) A CW -complex is connected iff it is path connected.

Proof Let X be an arbitrary CW -complex.

(i) Since X is a disjoint union of cells and each cell is path connected, it follows that
every path component P of X is a union of cells, and hence, P is a CW -complex.
(ii) Let P be a path component of X . Then P is closed in X . Let U be the union of
its other path components in X . Then U is a CW -subcomplex of the complex
X , and it is also closed. P is also open in X , since it is the complement of the
closed set U in X .
176 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

(iii) Let P be a path component of X . Since P is both open and closed in X , it

follows that it is connected. Then (iii) follows, since path components of X are
the components of X . 

Definition 3.3.8 A CW -complex pair (X , A) is a pair consisting of a topological

space X and a closed subspace A of X together with a sequence of closed subspaces
(X , A)n , called the n-skeleton of X relative to A, for all n ≥ 0 such that
(i) (X , A)0 is constructed from A by adjoining 0-cells.
(ii) For every n ≥ 1, the pair (X , A)n is constructed from (X , A)n−1 by adjoining
n-cells.

(iii) X = (X , A)n .
(iv) X has the weak topology with respect to {(X , A)n }n .
For A = ∅, the CW -complex pair (X , ∅) coincides with the CW -complex X .

Example 3.3.9 For the simplicial pair (K, L), the pair (|K|, |L|) is a CW -complex
pair such that (|K|, |L|)n = |K n ∪ Ln |.

Remark 3.3.10 For more study of CW -complexes, see Exercises 1–4 of Section
3.16.

3.4 Singular Homology Theory

This section gives a brief discussion on Singular Homology Theory introduced by

S. Lefschetz in his paper [ Lefschetz, 1933]. He extended simplicial homology theory
to singular homology theory by considering continuous maps from the standard
simplexes to a topological space X , which are called singular simplexes in X and by
using the algebraic properties of singular chain complexes.

Definition 3.4.1 Let e0 , e1 , e2 , . . . , en be the standard basis of Rn+1 . Then the stan-
dard n-simplex n is defined by

n = {i=0
n
λi ei : λi = 1, where 0 ≤ λi ≤ 1} ⊂ Rn+1 .

It is also defined by

n = {(λ0 , λ1 , . . . , λn ) ∈ Rn+1 : 0 ≤ λi ≤ 1 and λi = 1} ⊂ Rn+1 .

Definition 3.4.2 A singular n-simplex σn in a topological space X is a continuous

map
σn : n → X

where n is the standard n-simplex.

3.4 Singular Homology Theory 177

Example 3.4.3 (i) A singular 0-simplex σ0 is identified to a point in X , since, 0

is a one-point set.
(ii) A singular 1-simplex σ1 is a path in X , since 1 is the closed interval I .

Definition 3.4.4 Given a set of points {v0 , v1 , . . . , vn : vi ∈ Rm }, its affine combi-

nation is a point

x = λ0 v0 + λ1 v1 + · · · + λn vn : λi ∈ R and i=0
n
λi = 1.

A convex combination is an affine combination for which 0 ≤ λi ≤ 1, ∀ i.

Example 3.4.5 A convex combination of x, y ∈ Rn , is of the form

λx + (1 − λ)y, ∀ λ ∈ I.

Definition 3.4.6 Given a set of points {v0 , v1 , . . . , vn : vi ∈ Rm }, the set [v0 , v1 , . . .

, vi , . . . , vn ] consisting of all convex combinations of v0 , v1 , . . . vi , . . . , vn i.e., denotes
the convex set spanned by the points v0 , v1 , . . . , vi , . . . , vn .

Definition 3.4.7 An ordered set of points {v0 , v1 , . . . , vn } ⊂ Rm , is said to be affine

independent if

{v1 − v0 , v2 − v0 , . . . , vn − v0 }

is a linearly independent subset of the real vector space Rn .

Definition 3.4.8 If [v0 , v1 , . . . , vn ] denotes an n-simplex n, then the map

Tn : n → Rm , i=0
n
λi ei → i=0
n
λi vi

is called an affine singular n-simplex.

Remark 3.4.9 The image of Tn is a convex subspace of Rm , called the convex space
spanned by vi s which may not be independent.

Definition 3.4.10 Let [v0 , v1 , . . . v̂i , . . . , vn ] denote the affine (n − 1)- simplex n−1
obtained from the affine n-simplex [v0 , v1 , . . . vi , . . . , vn ] by deleting the i- th vertex
vi (symolized by v̂i ) counted from 0. Then the map

[v0 , v1 , . . . v̂i , . . . , vn ] : n−1 → n

is called the nth face map, denoted by fi n .

Definition 3.4.11 Given a singular n-simplex in a topological space X , its ith face
is defined
σni = σn ◦ fi n ,
178 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

and its boundary is defined by

∂σn = i=0
n
(−1)i σni , ∀ n ≥ 1.

For an n-chain c = nσ σ, define

∂n c = ∂ nσ σ = nσ ∂n σ ∀ n ≥ 1, and if n = 0 , ∂0 (c) = 0,

which is extended to n (X ), the free abelian group based on the singular n-simplexes,
to obtain a homomorphism

∂n : n (X ) → n−1 (X ).

Proposition 3.4.12 The composite homomorphism ∂n ◦ ∂n+1 = 0, ∀ n ≥ 0.

Proof It follows from definition of ∂n . 

The concepts of chains, cycles and boundaries defined for simplicial complex are
extended in a natural way to define singular complex associated with a topological
space X .

Definition 3.4.13 For every integer n ≥ 0, the singular n-chain group Cn (X ; Z)

is defined as the free abelian group with basis on all singular n-simplexes in X and
C−1 (X ; Z) = {0} is taken.
(i) An element of Cn (X ; Z) can be represented as a formal linear combination of
singular n-simplices with integer coefficients.
(ii) The elements of the group Cn (X ; Z) are called singular n-chains in X with
coefficient group Z.
The group of singular n-cycles in X , denoted by Zn (X ; Z), is the ker ∂n and the
group of singular n-boundaries, denoted by Bn (X ; Z) is the Im ∂n+1 . Hence, it
follows that
(i) Bn (X ; Z) and Zn (X ; Z) are both subgroups of Cn (X ; Z) ∀ n ≥ 0.
(ii) Moreover, the condition ∂n ◦ ∂n+1 = 0 asserts that Bn (X ; Z) is a subgroup of
Zn (X ; Z).

Definition 3.4.14 Given a topological space X , for every integer n ≥ 0, the quotient
group
Hn (X ; Z) = Zn (X ; Z)/Bn (X ; Z) = ker ∂n /Im ∂n+1

is called the nth singular homology group of X with coefficient group Z. For
an n-cycle zn ∈ Zn (X ; Z), the coset zn + Bn (X ; Z) is called the homology class of
zn ,denoted by [zn ]. The notion of the nth singular homology group Hn (X ; G) of X
with coefficient group G (abelian) is analogous.
3.4 Singular Homology Theory 179

Definition 3.4.15 Let f : X → Y be a continuous map of topological spaces. Then

it induces a group homomorphism

f∗ : Hn (X ; Z) → Hn (Y ; Z), [zn ] → [f (zn )].

Proposition 3.4.16 Hn : Top → Ab is a covariant functor from the category of

topological spaces and their continuous maps to the category of abelian groups
and their homomorphisms, for every integer n ≥ 0.

Proof Here, the object function is defined by

Hn : Top → Ab, X → Hn (X ; Z)

and for every continuous map f : X → Y in the category Top , the morphism function
is defined by
f → f∗ : Hn (X ; Z) → Hn (Y ; Z), [zn ] → [f (zn )].

Corollary 3.4.17 Let X and Y be two homeomorphic spaces. Then

Hn (X ; Z) ∼
= Hn (Y ; Z), ∀ n ≥ 0.

Proof It follows from the functorial property of Hn . 

Proposition 3.4.18 Let X be a topological space and {Xj : j ∈ A} be the set of path
components of X . Then every inclusion map ij : Xj → X induces an 1-1 homomor-
phism

ij∗ : Hn (Xj ; Z) → Hn (X ; Z), ∀ n ≥ 0

of groups such that for the natural homomorphism

ψ : ⊕j∈A Hn (Xj ; Z) → Hn (X ; Z), ∀ n ≥ 0,

ψ|Hn (Xj ; Z) = ij∗ , ∀ j ∈ A, ∀ n ≥ 0.

Proof Using the fact that the image of a singular simplex under a continuous map
completely lies in one path component, it follows that every inclusion map ij : Xj →
X induces a homomorphism

ij+ : Cn (Xj ; Z) → Cn (X ; Z).

Hence passing to the corresponding homology groups, the proposition is proved. 

180 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

3.4.1 Advantage of Singular Homology over Simplicial

Homology

There are several advantages of singular homology theory over simplicial homol-
ogy theory as a generalization of simplicial homology theory. They include mainly
as stated below:
(i) Singular homology theory is defined on all topological spaces, not on just poly-
hedra.
(ii) The induced homomorphism of a continuous map is easier to define in singular
homology than in simplicial homology.

Remark 3.4.19 Because of this advantage, singular homology theory has devel-
oped extensively. Instead of Z, called coefficient group of the singular homology
theory, any abelian group G may be taken equally well as its coefficient group.
For detailed study of the singular homology theory, the book [Adhikari 2016] is
referred.

3.4.2 The Singular Homology with Coefficient Group G

This subsection extends homology theory with integral coefficient formulated in Def-
inition 3.4.14 to homology theory with an arbitrary abelian group G as its coefficient
group.

Definition 3.4.20 Let X be an arbitrary topological space and G be an abelian group.

Then Z0 (X ; G) of 0-cycles on X is an abelian group, and B0 (X ; G) of 0-dimensional
boundaries is a subgroup of Z0 (X ; G). Its quotient group

Z0 (X ; G) ∼
H0 (X ; G) = = F,
B0 (X ; G)

where F is the free abelian group on the set of path-connected components of the
topological space X . The group H0 (X ; G) is called its 0-dimensional homology
group of X with G as its coefficient group. In a analogous way, its 1-dimensional
homology group is the quotient group

Z1 (X ; G)
H1 (X ; G) =
B1 (X ; G)

is called the first homology group of X with coefficient in G, where C1 (X ; G) is the

is the abelian group of 1-chains, Z1 (X ; G) is the subgroup of 1-cycles of C1 (X ; G) on
X and B1 (X ; G) of 1-dimensional boundaries is a subgroup of Z1 (X ; G). In general,
3.4 Singular Homology Theory 181

the n-dimensional homology group of X with coefficient group G, is the quotient

group
Zn (X ; G)
Hn (X ; G) =
Bn (X ; G)

3.5 Noether’s Algebraic Approach to Homology and

Cohomology Theories

This section gives an algebraic approach formulated by Emmy Noether (1882–

1935) corresponding to the geometric approach of homology theory invented
by Poincaré and described in Sect. 3.6.1. Inspired by her algebraic approach, P.
Alexandroff (1896–1982) and H. Hopf (1894–1971) jointly published a detailed
study of homology theory in 1935 for the first time. Its dual algebraic approach to
cohomology theories is also given in Sect. 3.5.2.

3.5.1 Homology Groups of Chain Complexes

The geometric approach of homology theory invented by Poincaré is shifted to alge-

braic approach by Emmy Noether. In her algebraic approach, homology groups are
defined for any topological space. This approach has created homological algebra,
a new branch of mathematics. The classical method of construction consists of con-
struction of homology groups of chain complexes associated with topological spaces.
Definition 3.5.1 A sequence C∗ of abelian groups {Cn } and their homomorphisms
{∂n }
∂n+1 ∂n ∂n−1 ∂1 ∂0
C∗ : · · · −−−−−→ Cn −−−→ Cn−1 −−−−−→ · · · −−−→ C0 −−−→ {0}

is said be a chain complex if

∂n−1 ◦ ∂n = 0, ∀ n ≥ 1.

The groups Cn are called chain groups and the homomorphisms ∂n : Cn → Cn−1 are
called boundary homomorphisms of the chain complex C∗ . This chain complex is
sometimes written as C∗ = {Cn , ∂n }.
Definition 3.5.2 Given a chain complex C∗ = {Cn , ∂n }, the condition

∂n ◦ ∂n+1 = 0, ∀ n ≥ 1

asserts that Im ∂n+1 is a subgroup of ker ∂n . The quotient group ker ∂n / Im ∂n+1 is
called the n th homology group of the chain complex C∗ denoted by Hn (C∗ ). On the
182 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Fig. 3.1 Chain map

f : C∗ → C∗

other hand, ker ∂n is called the group of n-dimensional cycles, denoted by Zn and
Im ∂n+1 is called the n-dimensional boundaries, denoted by Bn .
(i) An element of Zn = ker∂n is called an n-dimensional cycle.
(ii) An element of Bn = Im∂n+1 is called an n-dimensional boundary.
(iii) Elements of Hn (C∗ ) = Zn /Bn are called homology classes, denoted by [z] for
every z ∈ Zn .
Definition 3.5.3 Let M n be an n-dimensional manifold and {Cim } be the set of all m-
dimensional cycles (which are m-dimensional submanifolds with boundary) of M n
and Bm+1 be a submanifold of M n with a boundary consisting of connected manifolds
C1m , C2m , . . . , Cim , which are in {Cim }. Then the cycle C1m + C2m + · · · + Cim is said to
be homologous to 0.
Definition 3.5.4 Two cycles in an n-dimensional manifold M n are said to be homol-
ogous or equivalent if they differ by a cycle homologous to 0. The set of equivalence
classes of m-dimensional cycles is called the m-dimensional homology group of
M n.
Definition 3.5.5 Given two chain complexes C∗ = {Cn , ∂n } and C∗ = {Cn , ∂n }, a
sequence f = {fn : Cn → Cn } of homomorphisms is said to be a chain map from C∗
to C∗ , if these homomorphisms commute with the boundary homomorphisms in the
sense

fn ◦ ∂n+1 = ∂n+1 ◦ fn+1 , ∀ n ≥ 0,

i.e., if every square in Fig. 3.1 is commutative.

Proposition 3.5.6 Given two chain complexes of abelian groups C∗ = {Cn , ∂n } and
C∗ = {Cn , ∂n }, for a chain map f = {fn : Cn → Cn },
(i) its images fn (Zn ) ⊂ Zn for every n, i.e., fn sends n-cycles of C∗ into n-cycles of
C∗ . and
(ii) fn (Bn ) ⊂ Bn for every n, i.e., fn sends n-boundaries of C∗ into n-boundaries of
C∗ .
Proof Using the commutativity of each square in Fig. 3.1, the proposition follows.

Theorem 3.5.7 Given two chain complexes C∗ = {Cn , ∂n } and C∗ = {Cn , ∂n }, every
chain map f : C∗ → C∗ induces a homomorphism

Hn (fn ) = fn∗ : Hn (C∗ ) → Hn (C∗ ), [z] → [fn (z)].

3.5 Noether’s Algebraic Approach to Homology and Cohomology Theories 183

Proof It follows from Proposition 3.5.6 

Definition 3.5.8 The homomorphism Hn (fn ) = fn∗ : Hn (C∗ ) → Hn (C∗ ) defined in

Theorem 3.5.7 is said to be the homomorphism induced by fn in homology groups
for every integer n.

Proposition 3.5.9 (a) Given two chain maps f : C∗ → C∗ and g : C∗ → C∗ , their
composite
g ◦ f : C∗ → C∗

is also a chain map such that

(g ◦ f )∗ = g∗ ◦ f∗ : Hn (C∗ ) → Hn (C∗ ).

(b) If 1d : C∗ → C∗ is the identity chain map, then its induced map

(1d )∗ : Hn (C∗ ) → Hn (C∗ )

is the identity automorphism.

Proof It follows from Definition 3.5.8 of induced homomorphism in the homology

groups. 

Definition 3.5.10 Given two chain complexes C∗ = {Cn , ∂n } and C∗ = {Cn , ∂n },
f , g : C∗ → C∗ their two chain maps f , g : C∗ → C∗ are said to be chain homo-
topic, denoted by f  g, if there is a sequence

{Hn : Cn → Cn+1 }

of homomorphisms such that


∂n+1 Hn + Hn−1 ∂n = fn − gn : Cn → Cn , ∀ n ≥ 0.

On the other hand, a chain map

f : C∗ → C∗

is said to be a chain homotopy equivalence if there exists a chain map

h : C∗ → C∗

such that
h ◦ f  1d and f ◦ h  1d .

Proposition 3.5.11 Let S(C∗ , C∗ ) be the set of all chain maps from C∗ to C∗ . Then
the chain homotopy relation on S(C∗ , C∗ ) is an equivalence relation.
184 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Proof It follows from Definition 3.5.10. 

Theorem 3.5.12 Any two chain homotopic maps induce the same homomorphisms
in the homology.

Proof Let f , g : C∗ → C∗ be two chain maps such that f  g in the sense of Def-

inition 3.5.10. Then there is a chain homotopy {Hn : Cn → Cn+1 } such that for
[z] ∈ Hn (C∗ ), the image ∂n ([z]) = 0. This implies that

fn ([z]) − gn ([z]) = ∂n+1 Hn ([z])

is a boundary. Consequently,

[fn [z]] = [g, [z]], ∀ [z] ∈ Hn (C∗ ) =⇒ fn∗ ([z]) = gn∗ ([z]), ∀ [z] ∈ Hn (C∗ ) =⇒ fn∗ = gn∗ .

This proves that

f∗ = g∗ : Hn (C∗ ) → Hn (C∗ ), ∀ n ≥ 0.

Theorem 3.5.13 (a) Chain complexes and chain maps form a category, called
chain complex category, denoted by Comp .
(b) Hn is a covariant functor from the category Comp of chain complexes and chain
maps to the category A b of abelian groups and homomorphisms for every n ≥ 0,
i.e., Hn : Comp → Ab is a covariant functor for every integer n ≥ 0.
(c) Hn : Top → Ab is also a covariant functor for every integer n, where Top is the
category of topological spaces and their continuous maps.

Proof (a) Here , chain complexes are taken as objects and chain maps are taken as
morphisms. Then under the composition of chain maps coordinatewise : {gn } ◦
{fn } = {gn ◦ fn }, they form the category Comp .
(b) Here, the object function assigns to every chain complex the sequence of its
homology groups, and morphism function assigns to every chain map f between
chain complexes the induced map f∗ between their homology group. This asserts
that , Hn : Comp → Ab is a covariant functor for every integer n, by Proposition
3.5.9.
(c) Here, the object function assigns to every topological space X ∈ Top , its homol-
ogy groups Hn (X ; G). To define the morphism function, consider a continuous
map f : X → Y ∈ Top . If σ : n → X is an n-simplex in the space X , then

f ◦σ : n →Y

is an n-simplex in the space Y . This gives rise to a homomorphism

f# : Cn (X ; Z) → Cn (Y ; Z), kσ σ → kσ (f ◦ σ )

3.5 Noether’s Algebraic Approach to Homology and Cohomology Theories 185

(extending by linearity), which is independent of n. Since an element of

Hn (X ; Z) = Zn (X ; Z)/Bn (X ; Z), is a coset αn + Bn (X ; Z), where αn ∈ Zn (X ;
Z). Then f induces the map

Hn (f ) = f∗ : Hn (X ; Z) → Hn (Y ; Z), αn + Bn (X ; Z) → f∗ (αn ) + Bn (Y ; Z).

This is well defined and is a homomorphism. Moreover, it follows from Definition

of f∗ that
(i) Hn (1X ) : Hn (X ; Z) → Hn (X ; Z) is the identity homomorphism and
(ii) For any two maps f : X → Y and g : Y → Z in Top

Hn (g ◦ f ) = Hn (g) ◦ Hn (f ) : Hn (X ; Z) → Hn (Z; Z).

Hence it is proved that

Hn : Top → Ab

is a covariant functor for each integer n ≥ 0. 

Corollary 3.5.14 Homeomorphic spaces induce isomorphic homology groups.
Proof To prove the corollary, it is sufficient to show that if X , Y ∈ Top are homeo-
morphic, then the corresponding homology groups

Hn (X ; Z) ∼
= Hn (Y ; Z), ∀ n ≥ 0.

It follows from the functorial property of

Hn : Top → Ab

for each n ≥ 0 proved in Theorem 3.5.13.


Remark 3.5.15 For any X ∈ Top ,
(i) Every homology group Hn (X ; Z) is a topological invariant of X in the sense
that its homeomorphic spaces have isomorphic homology group for every integer
n ≥ 0.
(ii) rank Hn (X ; Z) is an invariant of X for every integer n ≥ 0 in the sense that if
the topological spaces X and Y are homeomorphic, then

rank Hn (X ; Z) = rank Hn (Y ; Z), ∀ n ≥ 0.

This rank is called the nth Betti number of X .

Definition 3.5.16 The sequence {Hn : Top → Ab} of covariant functors Hn given
in Theorem 3.5.13 is called the homology sequence from the category Top to the
category Ab .
186 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

3.5.2 Cochain Complex and Its Cohomology Groups

Cohomology groups are dual to homology groups defined for any topological space.
The classical method of their construction consists of construction of cohomology
groups associated with topological spaces. The basic difference between these two
theories is that
(i) Homology theory
Hn : Top → Ab

is a covariant functor from the category Top of topological spaces and their
continuous maps to the category of abelian groups Ab and their homomorphisms,
for every integer n ≥ 0.
(ii) On the other hand, cohomology theory

H n : Top → Ab

is a contravariant functor from the category Top of topological spaces and their
continuous maps to the category of abelian groups Ab and their homomorphisms,
for every integer n ≥ 0.

3.5.3 Cochain Complex and Its Cohomology Groups

Let Ab denote the category of abelian groups and their homomorphisms and Top
denote the category of topological spaces and their continuous maps. The motivation
of cohomology theory comes from homological algebra.
Definition 3.5.17 For any X ∈ Top and any abelian group G ∈ Ab, the singular
cochain group C n (X ; G) with coefficient group G, is defined by

C n (X ; G) = Hom(Cn (X , G), G),

where the right-hand side is the group of all homomorphisms from the singular chain
group (Cn (X ; G) to the abelian group G.
Definition 3.5.18 Let G be an abelian group and β ∈ (C n (X ; G) be an arbitrary
element. Then
β : Cn (X ; G) → G

is a group homomorphism. Define a map

δ : Cn+1 (X ; G) → G, β → β ◦ ∂

Then δ is the composite homomorphism

3.5 Noether’s Algebraic Approach to Homology and Cohomology Theories 187

∂ β
Cn+1 (X ; G) −→ Cn (X ; G) −→ G.

This asserts that

δβ( σ ) = i (−1)i (σ |[v0 ,v1 ,...,vˆi ,...,vn+1 ])

for all singular (n + 1) simplexes σ, where the notation hat over vi , i.e., v̂i indicates
that the vertex vi is omitted from the vertex set

v0 , v1 , . . . , vi , . . . , vn .

Then it follows that δ ◦ δ = 0.

For simplicity, the notation C ∗ = {C n , δ n } stands for notation C ∗ (X ; G) =

{C (X ; G), δ n }
n

Definition 3.5.19 A sequence C ∗ = {C n , δ n }, n ∈ Z of abelian groups C n and their

homomorphisms δ n : C n−1 → C n is said to be a cochain complex if

δ n+1 ◦ δ n = 0 ∀ n ≥ 0.

This asserts that the sequence (3.4) of abelian groups {C n } and their homomorphisms
δ n : C n−1 → C n is a cochain complex if δ n+1 ◦ δ n = 0 ∀ n ≥ 0 :

δn δ n+1
C∗ : · · · −
→ C n−1 −−−→ C n −−−−−→ C n+1 −
→ ··· (3.4)

Definition 3.5.20 In Definition 3.5.19,

(i) δ n is called a coboundary homomorphism.
(ii) The elements of Z n = ker δ n+1 are called n-cocycles.
(iii) The elements of Bn = Im δ n are called n-coboundaries
of the cochain complex C ∗ given in (3.4).

Proposition 3.5.21 Z n and Bn form groups for all n for the cochain complex C ∗
given in (3.4).

Proof It follows from respective definitions. 

Proposition 3.5.22 Bn is a subgroup of Z n for all n for the cochain complex C ∗

given in (3.4).

Proof The cochain complex C ∗ given in (3.4) implies that δ n+1 ◦ δ n = 0. This proves
that Bn is a subgroup of Z n for all n. 
 188 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Fig. 3.2 Cochain map

f = {f n : C n → C n }

Definition 3.5.23 (Cohomology group of C ∗ ) Given a cochain complex C ∗ given

in (3.4), the quotient group
Z n (C ∗ )/Bn (C ∗ ),

denoted by H n (C ∗ ) (or simply H n ), is said to be the n-dimensional cohomology

group of the cochain complex C ∗ .

3.5.4 Singular Cohomology

Definition 3.5.23 leads to singular cohomology group of an arbitrary topological

space formulated in Definition 3.5.24.
Definition 3.5.24 (singular cohomology group) Given an abelian group G and
a topological space X , let C n (X ; G) denote the group Hom(Cn (X ), G)), ∀ n ≥ 0,
where Cn (X ) denotes the group of singular n-chains in X with coefficient group G
(see Definition 3.4.13).
(i) An element of C n (X ; G) is called an n-dimensional singular cochain with
coefficients in G (in brief n-cochain).
(ii) An element of Z n (X ; G) = ker δ n+1 is called an n-cocycle with coefficient in G.
(iii) An element of Bn (X ; G) = Im δ n is called an n-coboundary with coefficient in
G.
(iv) The quotient group Z n (X ; G)/Bn (X ; G) denoted by the quotient group H n (X ; G)
is called the n-dimensional singular cohomology group of X with coefficients
in G.
(v) An element of H n (X ; G) is a coset α + Bn (X ; G), where α is a cocycle. This is
called the cohomology class of α written as [α] ∈ H n (X ; G).
Definition 3.5.25 For n ≥ 0, given two cochain complexes C ∗ = {C n , δ n } and
C ∗ = {C n , δ n }, a sequence

f = {f n : C n → C n }

is called a cochain map from C ∗ to C ∗ if the diagram in Fig. 3.2

commutes,i.e., if
f n+1 ◦ δ n = δ n ◦ f n

holds for all n ≥ 0.

3.5 Noether’s Algebraic Approach to Homology and Cohomology Theories 189

3.5.5 Functorial Approach to Singular Cohomology Theory

This subsection presents a functorial approach to singular homology theory. Let Ab

denote the category of abelian groups and their homomorphisms and Top denote
the category of topological spaces and their continuous maps. The motivation of
cohomology theory comes from homological algebra.

Proposition 3.5.26 Given an abelian group G, there is a contravariant functor

Hom ( , G) : Ab → Ab .

Proof The object function is defined by assigning to every group A ∈ Ab , the group
Hom (A, G) of homomorphisms from A to G and morphism function is defined by
assigning to a group homomorphism f : A → B ∈ Ab, the homomorphism

f ∗ : Hom (B, G) → Hom (A, G), ψ → ψ ◦ f .

This implies the proposition.



Theorem 3.5.27 Given an integer n ≥ 0, for every abelian group G, the cohomol-
ogy

H n ( ; G) : Top → Ab

is a contravariant functor.

Proof Here, the object function is defined by assigning to every topological space
X ∈ Top , the cohomology group H n (X ; G) ∈ Ab, and morphism function is defined
by assigning to a continuous map f : X → Y ∈ Top , the homomorphism

f ∗ : H n (Y ; G) → H n (X ; G), α + Bn (Y ; G) → f # (α) + Bn (X ; G) = α ◦ f + Bn (X ; G)

where f # is defined by

f # : Hom(Cn (Y ; G), G) → Hom(Cn (X ; G), G), k → k ◦ f#

190 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

and f# is defined in Theorem 3.5.13

f# : Cn (X ; G) → Cn (Y ; G), kσ σ → kσ (f ◦ σ ).

Clearly, f ∗ is well defined, and it is a homomorphism. Moreover, it follows from

Definition of f ∗ that
(i) H n (1X ) : H n (X ; Z) → H n (X ; Z) is the identity homomorphism and
(ii) For any two maps f : X → Y and g : Y → Z in Top

H n (g ◦ f ) = H n (f ) ◦ H n (g) : Hn (Z; G) → H n (X ; G).

Hence it is proved that H n : Top → Ab is a contravariant functor for each n ≥ 0. 

Remark 3.5.28 For any X ∈ Top ,
(i) Every cohomology group H n (X ; Z) is an invariant of X in the sense that its
homeomorphic spaces have isomorphic cohomology group for every integer
n ≥ 0;
(ii) Rank H n (X ; Z) is an invariant of X for every integer n ≥ 0.
Definition 3.5.29 The sequence {H n : Top → Ab} of contravariant functors H n
given in Theorem 3.5.27 is called the cohomology sequence from the category
Top to the category Ab .
Remark 3.5.30 Dual concepts and results of homology mentioned in Section 3.5.1
hold in cohomology.

3.6 Homology Theory: Eilenberg and Steenrod Axioms

Eilenberg and Steenrod announced an outstanding new approach to homology and

cohomology theories in 1945 by taking a small number of properties of the existing
homology and cohomology theories (not focusing the tools used for construction
of homology and cohomology groups) as axioms to characterize both homology
and cohomology. This axiomatic approach was published in 1952 [Eilenberg and
Steenrod 1952]. The most surprising result is the proof that on the category of all
topological pairs having homotopy type of finite CW-complex pairs all homology
(resp. cohomology) theories satisfying these axioms have isomorphic groups. This
result proves that there is only one concept of homology on that category [Eilenberg
and Steenrod 1952]. The Eilenberg and Steenrod axioms (in brief, E-S axioms) for
homology or cohomology functors provide an elegant and quick access to the further
study of algebraic topology, and Dimension Axiom locates the coefficient group at
the right dimension.
This section presents Eilenberg and Steenrod axioms on the category T op2 of
topological pairs and their continuous maps, which is the largest admissible category
 3.6 Homology Theory: Eilenberg and Steenrod Axioms 191

in the sense that it contains other admissible categories as subcategories. The earliest
definition of homology group given by Poincaré in 1895 stands matched with the
Eilenberg and Steenrod axioms on the admissible category Simp defined in Exam-
ple 3.6.5. However, for modern study of homology and cohomology theories (with
generalized homology and cohomology theories), the books [Adams 1972; Adhikari
2016] are referred.

3.6.1 Admissible Categories for Eilenberg and Steenrod

Axioms

This subsection specifies admissible category C in Definition 3.6.4 to formulate

Eilenberg and Steenrod axioms for homology and cohomology theories.

Proposition 3.6.1 (i) The pairs of topological spaces and their continuous maps
form a category, called the category of topological pairs, denoted by T op2 .
(ii) The pairs of topological spaces and the homotopy classes of their continuous
maps relative to a subspace form a category, called the homotopy category of
topological pairs, denoted by Htp2 .

Proof It follows from the corresponding definitions. 

Definition 3.6.2 (Topological triple) A triple (X : A, B) is said to be a topological

triple, if X is a topological space such that A and B are subspaces of X with B ⊂ A.

Definition 3.6.3 (Topological triad) A triple (X : A, B) is said to be a topological

triad, if X is a topological space such that (A, B) is an ordered pair of subspaces A
and B of X .

The concept of topological triad is used in Sect. 3.11.4 on Mayer-Vietoris theorem.

Definition 3.6.4 A category C is said to be an admissible category for the Eilenberg

and Steenrod axioms if C satisfies the following conditions:
(i) C consists of topological pairs (X , A), as objects, i.e., if (X , A) ∈ Ob C, then
the pairs (X , X ), (A, A), X ≡ (X , ∅), A ≡ (A, ∅), ∅ ≡ (∅, ∅) ∈ Ob C.
(ii) C consists of morphisms f : (X , A) → (Y , B), for every pair of objects (X , A)
and (Y , B) ∈ Ob (C) with all their possible inclusions, i.e., if f : (X , A) →
(Y , B) is a morphism in C, then all the inclusion maps induced by f on the
corresponding subpairs are also morphisms in C.
(iii) For any object (X , A) ∈ Ob C, the object (X × I, A × I) ∈ Ob C and the maps

Ht : X → X × I, x → (x, t)

are morphisms in C.
 192 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

(iv) There is one-point spaces X0 ∈ Ob C, with the property that the constant map

c : X → X0 , x → X0

is a morphism in C.

Example 3.6.5 Some examples of admissible category for Eilenberg and Steen-
rod axioms for homology and cohomology theories:
(i) All topological pairs (X , A) and their continuous maps form an admissible cate-
gory for Eilenberg and Steenrod axioms for homology and cohomology theories,
denoted by T op2 .
(ii) All the pairs (X , A) of spaces where the spaces X and A have triangulations
together with their simplicial maps form an admissible category for Eilenberg
and Steenrod axioms for homology and cohomology theories, denoted by Simp.
(iii) All compact pairs (X , A) in the sense that X is compact and A is closed in X and
all continuous maps of such pairs form an admissible category for Eilenberg and
Steenrod axioms for homology and cohomology theories, denoted by Cc .

3.6.2 Eilenberg and Steenrod Axioms for Homology Theory

This subsection presents Eilenberg and Steenrod axioms for homology theory on the
category C = T op2 of all topological pairs (X , A) of spaces and their continuous
maps.
A homology theory H on the category T op2 consists of three functions H =
{H , ∗, ∂} which satisfy the following axioms for every integer n ≥ 0. :

(i) The first function H assigns to every topological pair (X , A) of spaces in T op2
and every integer n ≥ 0, an abelian group Hn (X , A), called the n-dimensional
homology group of the topological pair (X , A) in the homology theory H. For
A = ∅, it is called n-dimensional (absolute) homology group of the space X .
(ii) The second function ∗ assigns to every continuous map f : (X , A) → (Y , B) in
T op2 and every integer n ≥ 0, a homomorphism

f∗ = fn∗ : Hn (X , A) → Hn (Y , B),

called the homomorphism induced by the map f in the homology theory H.

(iii) The third function ∂ assigns to each topological pair (X , A) in T op2 and every
integer n ≥ 1, a homomorphism

∂ = ∂n : Hn (X , A) → Hn−1 (A),

called the boundary operator on the group Hn (X , A) in the homology theory

H.
3.6 Homology Theory: Eilenberg and Steenrod Axioms 193

Fig. 3.3 Diagram

connecting boundary
operator ∂ and induced
homomorphisms in H

Moreover, these functions satisfy the following seven axioms, called the Eilenberg–
Steenrod axioms for homology theory H on T op2 ;
Axiom H(1) (Identity Axiom). If 1X : (X , A) → (X , A) is the identity map on a
topological pair (X , A) in T op2 , then its induced homomorphism

1X ∗ : Hn (X , A) → Hn (X , A)

is the identity automorphism of the homology group Hn (X , A) for every integer

n ≥ 0.
Axiom H(2) (Composition Axiom). If f : (X , A) → (Y , B) and g : (Y , B) →
(Z, C) are two continuous maps in T op2 , then

(g ◦ f )∗n = g∗n ◦ f∗n : Hn (X , A) → Hn (Z, C)

for every integer n ≥ 0.

Remark 3.6.6 In the language of category theory, the above axioms H(1) and
H(2) assert that for every fixed integer n ≥ 0, the functions Hn form a covariant
functor from the category T op2 to the category Ab. Each

Hn : T op2 → Ab

is called the homology functor in the homology theory H on the category T op2 .
We use the notation Hn (f ) = f∗n (Fig. 3.3).
Axiom H(3) (Commutativity Axiom). If f : (X , A) → (Y , B) is a continuous map
in category T op and if g : A → B is a continuous map in T op, defined by g(x) =
f (x) for all x ∈ A, then the diagram in Fig. 3.3 is commutative in the sense that
g∗ ◦ ∂ = ∂ ◦ f∗ for every integer n ≥ 0.
Remark 3.6.7 This axiom connects the homology functor Hn with boundary oper-
ator ∂ and induced homomorphisms in the homology theory H.
Axiom H(4) (Exactness Axiom). If (X , A) is a topological pair in T op2 and i : A →
X , j : X → (X , A) are the inclusion maps, then the beginningless sequence

i∗ j∗ ∂
· · · → Hn (A) −−−→ Hn (X ) −−−→ Hn (X , A) −−−→ Hn−1 (A) → · · ·

of groups and homomorphisms is exact, called the homology exact sequence of

(X , A).
194 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Remark 3.6.8 The axioms formulated in H(1)-H(4) are all algebraic axioms.

Axiom H(5)(Homotopy Axiom) If two continuous maps f , g : (X , A) → (Y , B) in

T op2 are homotopic in T op2 , then

f∗n = g∗n , ∀ n ≥ 0.

AxiomH(6)(Excision Axiom). Let U be an open set of a topological space X , where

its closure U is contained in the interior Int A= Å of a subspace A of X (i.e., U ⊂ Å).
If the inclusion map
i : (X − U, A − U ) → (X , A)

is in T op2 , then the induced homomorphism

i∗ : Hn (X − U, A − U ) → Hn (X , A)

is an isomorphism for every integer n ≥ 0. The map i is called the excision of the
open set U , and i∗ is called its n-dimensional excision isomorphism.
Axiom H(7)(Dimension Axiom). The n-dimensional homology group Hn (X ) of a
one-point space X = {point} in the homology theory H consists of a single element
for every integer n = 0, in symbol, Hn (point) = {0}, for n = 0.

Definition 3.6.9 If the homology theory H satisfies only the first six axioms H(1)-
H(6), then H is called a generalized homology theory on the category T op2 . Such
homology theories form the basic topics of modern algebraic topology.

Definition 3.6.10 For a homology theory H on an admissible category C, the 0-

dimensional homology group
G = H0 (point)

is called the coefficient group of the homology theory H.

Example 3.6.11 The simplicial homology theory applies to the category Simp of
pairs (X , A) of spaces, where X and A have triangulations K and L, respectively, for
which L is a subcomplex of K. On the other hand, the singular homology theory
applies to all pairs of spaces (X , A), where X is a topological space and A is a
subspace of X .

3.7 The Uniqueness Theorem for Homology Theory

This section proves in Theorem 3.7.3 the uniqueness of homology groups in

axiomatic homology theories satisfying E-S axioms on the category T op2 with iso-
morphic coefficient groups. A surprising result is that on the category T op2 of all
topological pairs having homotopy types of finite CW-complex pairs (see Sect. 3.3),
3.7 The Uniqueness Theorem for Homology Theory 195

Fig. 3.4 Natural

equivalence of homology
functors Hn and Hn

Fig. 3.5 Commutative

diagram involving hn and
induced homomorphisms f∗
and f


all homology theories satisfying the E-S axioms have isomorphic groups [Eilenberg
and Steenrod 1952].
This uniqueness theorem is a key result in algebraic topology. For example, the
simplicial and singular homology (cohomology) theories on the category of compact
polyhedra coincide in the sense that for any compact polyhedra |K|, the simplicial
homology groups HnS (|K|; G) and the singular homology group Hn (|K|; G) with
coefficient group G are isomorphic for all n ≥ 0.
To prove Theorem 3.7.3, let H = {H , ∗, ∂} and H = {H  , 
, ∂  } be two arbitrary
homology theories on the category T op . Suppose G = H0 (point), G  = H0 (point)
2

are their coefficient groups.

Definition 3.7.1 Let H and H be two homology theories on T op2 . An isomor-

phism (natural) ψ : H → H is a sequence of natural equivalences

ψn : Hn → Hn , ∀ n ≥ 0

such that the rectangular diagram in Fig. 3.4 is commutative in the sense that ∂ ◦
ψn−1 = ψn ◦ ∂  for all pairs (X , A) in T op2 and for all n ≥ 0.

Theorem 3.7.2 proves that at some situation described in this theorem, there exists
a unique homomorphism

hn : Hn (X , A) → Hn (X , A).

Theorem 3.7.2 Let G and G  be abelian groups and h : G → G  be a homomor-

phism. Then for every pair (X , A) in T op2 and every integer n ≥ 0, there exists a
unique homomorphism
hn : Hn (X , A) → Hn (X , A)

such that
(i) h0 = h on G = H0 (point).
(ii) For every map f : (X , A) → (Y , B) in T op2 and every integer n ≥ 0, the dia-
gram in Fig. 3.5 is commutative, i.e., hn ◦ f∗ = f
 ◦ hn .
196 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Fig. 3.6 Diagram

connecting boundary
homomorphisms with
homomorphism hn

(iii) For every pair of spaces (X , A) in T op2 and every integer n ≥ 0, the diagram
in Fig. 3.6 is commutative, i.e., hn−1 ◦ ∂ = ∂  ◦ hn .

Proof For proof, see [Eilenberg and Steenrod 1952]. 

Theorem 3.7.3 proves that there is only one concept of homology on that category.
This uniqueness theorem is very important in the development of algebraic topology.

Theorem 3.7.3 (The uniqueness theorem for homology theory) Let G and G  be
two abelian groups and h : G → G  be an isomorphism of groups. Then the unique
homomorphism
hn : Hn (X , A) → Hn (X , A)

defined in Theorem 3.7.2 is also an isomorphism for every pair of spaces (X , A) in

T op2 and every integer n ≥ 0.

Proof By hypothesis, G and G  are two abelian groups, and h : G → G  is an iso-

morphism of groups. Let k : G  → G be the isomorphism of groups defined by
k = h−1 . Then by Theorem 3.7.2, there exists a unique homomorphism

kn : Hn (X , A) → Hn (X , A)

satisfying the conditions (i)-(iii) of Theorem 3.7.2 for every pair of topological spaces
(X , A) in T op2 and every integer n ≥ 0. This shows that
(i) kn ◦ hn = identity automorphism of the groups Hn (X , A) for every integer n ≥ 0.
(ii) hn ◦ kn = identity automorphism of the groups Hn (X , A) for every integer n ≥ 0.
.
This proves that
hn : Hn (X , A) → Hn (X , A)

is an isomorphism of groups. 

Corollary 3.7.4 Given a coefficient group G, there exists only one homology theory
in the category T op2 .

Proof It follows from Theorem 3.7.2. 

Remark 3.7.5 Corollary 3.7.4 justifies the name the uniqueness theorem.
3.8 Cohomology Theory: Eilenberg and Steenrod Axioms 197

3.8 Cohomology Theory: Eilenberg and Steenrod Axioms

This section presents an axiomatic approach to cohomology theory given by Eilen-

berg and Steenrod, which is dual (parallel) to the homology theory described in
Section 3.6. In fact these two theories differ in only one point: Homology functors
are covariant; on the other hand, cohomology functors are contravariant, and hence,
most arrows in cohomology theory change their directions compared to the directions
with homology theories. Therefore, one can expect a dual theorem in cohomology
theory for every theorem established in homology theory. The Eilenberg–Steenrod
axioms for cohomology functors make the subject algebraic topology quickly acces-
sible.
The most surprising result is the proof that on the category of all topological
pairs having homotopy type of finite CW-complex pairs, all cohomology theories
satisfying these axioms have isomorphic groups. This basic theorem asserts that
there is only one concept of cohomology on that category.
A cohomology theory K on the category T op2 consists of three functions K =
{H , ∗, δ} satisfying the following axioms for each integer n ≥ 0:
(i): The first function H assigns to each topological pair (X , A) in the category
T op2 and every integer n ≥ 0, an abelian group H n (X , A), called the n-dimensional
cohomology group of the topological pair (X , A) in the cohomology theory K. For
A = ∅, it is called the n-dimensional (absolute) cohomology group of the space X .
(ii): The second function ∗ assigns to every f : (X , A) → (Y , B) in T op2 and
every integer n ≥ 0, a homomorphism

f ∗ = fn∗ : H n (Y , B) → H n (X , A),

called the homomorphism induced by the map f in the cohomology theory K.

(iii): The third function δ assigns to each topological pair (X , A) in T op2 , a
homomorphism
δ = δ(X , A, n) : H n−1 (A) → H n (X , A),

called the coboundary operator on the group H n−1 (A) in the cohomology theory
K.
Moreover, these three functions satisfy the following axioms C(1)–C(7), called
the Eilenberg–Steenrod axioms, in brief, E-S axioms for cohomology theory on
T op2 :
Axiom C(1)(Identity Axiom). If 1X : (X , A) → (X , A) is the identity map on a
topological pair (X , A) in T op2 , then the induced homomorphisms

1∗X : H n (X , A) → H n (X , A)

are the identity automorphism of the cohomology group H n (X , A) for every integer
n ≥ 0. Axiom C(2)(Composition Axiom). If f : (X , A) → (Y , B) and g : (Y , B) →
(Z, C) are maps in T op2 , then
198 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Fig. 3.7 Diagram

connecting coboundary
operator δ with induced
homomorphisms in K

(g ◦ f )∗n = fn∗ ◦ gn∗ : H n (Z, C) → H n (X , A)

for every integer n ≥ 0.

Remark 3.8.1 The above axioms C(1) and C(2) assert that for every fixed integer
n, the function H n forms a contravariant functor from the category T op2 to the
category Ab. The functor H n is called the n-dimensional cohomology functor in
the cohomology theory K.

We use the notation H n (f ) = fn∗ .

Axiom C(3)(Commutativity Axiom). If f : (X , A) → (Y , B) is a map in T op2
and if g : A → B is the map in T op defined by g(x) = f (x) for all x ∈ A, then the
diagram in Fig. 3.7 is commutative, i.e., δ ◦ g ∗ = f ∗ ◦ δ for every integer n ≥ 0.

Remark 3.8.2 The Axiom C(3) connects the cohomology functors in the cohomol-
ogy theory K with the coboundary operator δ and induced homomorphisms.

Axiom C(4)(Exactness Axiom). If (X , A) is a topological pair of spaces in T op2

and
i : A → X , and j : X → (X , A)

are inclusion maps, then the cohomology sequence

δ j∗ i∗
· · · → H n−1 (A) −−−→ H n (X , A) −−−→ H n (X ) −−−→ H n (A) → · · ·

of the topological pair (X , A) is exact.

Remark 3.8.3 The above four axioms C(1)-C(4) are algebraic axioms.
Axiom C(5)(Homotopy Axiom). If two maps f , g : (X , A) → (Y , B) in T op2 are
homotopic in T op2 , then their induced transformations are such that

fn∗ = gn∗

for every integer n ≥ 0.

Axiom C(6)(Excision Axiom). Let U be an open set of a topological space X whose
closure U is contained in the interior Int A= Å of a subspace A of X (i.e., U ⊂ Å.)
If the inclusion map i : (X − U, A − U ) → (X , A) is in T op2 , then the induced
homomorphism
i∗ : H n (X , A) → H n (X − U, A − U )
3.8 Cohomology Theory: Eilenberg and Steenrod Axioms 199

is an isomorphism for every integer n. The map i is called the excision of the open
set U , and i∗ is called its n-dimensional excision isomorphism.
Axiom C(7)(Dimension Axiom). The n-dimensional cohomology group H n (X ) of
a one-point space X = {point} consists of a single element for every integer n = 0,
in symbol,
H n (point) = {0}, for n = 0.

Definition 3.8.4 If K satisfies only the first six axioms C(1)-C(6), then K is called
a generalized cohomology theory on the categoryT op2 .

Definition 3.8.5 For a cohomology theory K, the 0-dimensional cohomology group

G = H 0 (point) is called the coefficient group of the cohomology theory K.

Theorem 3.8.6 (the uniqueness theorem for cohomology theory) Let G and G  be
two abelian groups and h : G → G  be an isomorphism of groups. Then h induces
an isomorphism
hn : Hn (X , A) → Hn (X , A)

for every pair of spaces (X , A) in T op2 and for every integer n ≥ 0.

Proof It follows from Theorem 3.7.2 by duality principle (see [Eilenberg and Steen-
rod 1952]). 

Remark 3.8.7 Theorem 3.8.6 justifies the name the uniqueness theorem for coho-
mology theory.

3.9 The Reduced Homology and Cohomology Groups

This section studies reduced homology and cohomology theories on the admissible
category T op2 . Let (X , A) be a topological pair in T op2 and i : A → X , j : X →
(X , A) be the inclusion maps. If f : (X , A) → (Y , B) is a continuous map in T op2 ,
then it defines two continuous maps

g : A → B, and h : X → Y

such that their induced homomorphisms in homology theory H = {H , ∗, ∂} on the

category T op2 satisfy the following relations:
(i)
i∗ ◦ g∗ = h∗ ◦ i∗ ;

(ii)
j∗ ◦ h∗ = f∗ ◦ j∗ ,
 200 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

and
(iii)
∂ ◦ f∗ = g∗ ◦ ∂.

3.9.1 The Reduced Homology Groups

This subsection studies the reduced homology groups and their relations with homol-
ogy groups in a homology theory H with coefficient group G.

Definition 3.9.1 Let (X , A) be a topological pair in T op2 and i : A → X , j : X →

(X , A) be the inclusion maps. If

f : (X , A) → (Y , B)

is a continuous map in T op2 , then it defines two continuous maps

g : A → B, and h : X → Y

such that
(i) For the kernel Kn (X , A) of

f∗ : Hn (X , A) → Hn (Y , B)

ker f∗ = Kn (X , A) ⊂ Hn (X , A) for every integer n ≥ 0.

(ii) For the kernel Kn (A) of

g∗ : Hn (A) → Hn (B)

ker g∗ = Kn (A) ⊂ Hn (A) for every integer n ≥ 0.

(iii) For the kernel Kn (X ) of

h∗ : Hn (X ) → Hn (Y )

ker h∗ = Kn (X ) ⊂ Hn (X ) for every integer n ≥ 0.

Proposition 3.9.2 The homomorphisms i∗ , j∗ and ∂ have the following properties:

 3.9 The Reduced Homology and Cohomology Groups 201

(i) i∗ (Kn (A)) ⊂ Hn (X )∀n ≥ 0;

(ii) j∗ (Kn (X )) ⊂ Hn (X , A)∀n ≥ 0;
(iii) ∂(Kn (X , A)) ⊂ Hn−1 (A)∀n ≥ 0.

Proof It follows from the respective definitions. 

Definition 3.9.3 The beginningless infinite sequence (3.5) (obtained by using Propo-
sition 3.9.2)

i∗ j∗ ∂
· · · → Kn (A) −−−→ Kn (X ) −−−→ Kn (X , A) −−−→ Kn−1 (A) → · · · (3.5)

of groups and homomorphisms is called the kernel sequence of the continuous map

f : (X , A) → (Y , B)

in the homology theory H on the admissible category T op2 .

Let (X , A) = (∅, ∅), (Y , B) = (0, 0) ∈ T op2 be two topological pairs in the homol-
ogy theory H on category T op2 , where 0 is the distinguished one-point space in this
category. Then there is a unique map

ψ : (X , A) → (0, 0).

Definition 3.9.4 The kernel sequence of the map ψ : (X , A) → (0, 0) represented

by

j∗ ∂
··· → H n (X ) −−−→ H
n (A) −−−→ H i∗
n (X , A) −−−→ H
n−1 (A) → · · · (3.6)

is called the reduced homology sequence of (X , A) in the homology theory H.

Consider the homology sequence of the pair (0, 0) in the homology theory H with
coefficient group G = {0}

i∗ j∗ ∂ i∗
· · · → Hn (0) −−−−→ Hn (0) −−−−→ Hn (0, 0) −−−→ Hn−1 (0) → · · · → H0 (0) −−−−→ H0 (0) → · · · .

All the groups of the above sequence are trivial except for H0 (0) = G. The
groups H 0 (A) are called the reduced homology 0-dimensional homol-
0 (X ) and H
ogy groups the spaces X and A, respectively, in the reduced homology sequence of
(X , A) in the homology theory H.
The above discussion is summarized in the basic and important result.

Theorem 3.9.5 Let H be a homology theory with coefficient group G = {0} on the
category T op2 and X = ∅. Then
 202 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

(i) n (X , A) = Hn (X , A), for every integer n ≥ 0.

H
(ii) n (A) = Hn (A) for every integer n = 0.
H
(iii) n (X ) = Hn (X ) for every integer n = 0.
H
(iv) H0 (A) = H0 (A) ⊕ G.
(v) H0 (X ) = H0 (X ) ⊕ G.
(vi) The identity map 1X : X → X induces the identity automorphism

n (X ) → H
1X ∗ : H n (X ).

(vii) For the continuous maps f : X → Y and g : Y → Z that are in the above cate-
gory,

n (X ) → H
(g ◦ f )∗ = g∗ ◦ f∗ : H n (Z),

for every integer n.

(viii) For the homotopic maps f , g : X → Y in the above category,

n (X ) → H
f∗ = g∗ : H n (Y ),

for every integer n.

(ix) For the homotopically equivalent spaces X and Y , in the above category, reduced
homology groups

n (X ) ∼
H =Hn (Y ),

for every integer n.

(x) For the homeomorphic topological spaces X and Y , in the above category,
reduced homology groups

n (X ) ∼
H =Hn (Y ),

for every integer n.

(xi) The reduced homology group H n (X ) of a topological space X is both a homotopy
and a topological invariant of X .
(xii) The reduced homology sequence of (X , A)
j∗ ∂
n (A) −−−→ H
··· → H
i∗
n (X ) −−−→ H
n (X , A) −−−→ H
n−1 (A) → · · ·

is exact.

Proof Left as an exercise. 

Proposition 3.9.6 Let X = ∅ be a contractible space. Then H n (X ) = G for every

integer n in the homology theory H with coefficient group G.
3.9 The Reduced Homology and Cohomology Groups 203

Proof By hypothesis, X is contractible. Then there is a unique map f : X →

P for some one-point space P, which is a homotopy equivalence. Then it induces
an isomorphism

f∗ : Hn (X ; G) → Hn (P; G)

for every integer n. This asserts that

n (X ) = ker f∗ = G
H

for every integer n. 

Corollary 3.9.7 Let X = ∅ be a contractible space. Then Hn (X ) = {0} for every

integer n = 0 and H0 (X ) = G.

Corollary 3.9.8 Let (X , A) be a topological pair.

(i) If X = ∅ is contractible, then

n (X , A; G) → H
∂ : Hn (X , A; G) = H n−1 (A; G)

is an isomorphism for every integer n.

(ii) If A = ∅ is contractible, then

n (X ; G) → H
j∗ : H n (X , A : G) = Hn (X , A; G)

is an isomorphism for every integer n.

Proof It follows from exactness of the reduced homology sequence for the pair
(X , A). 

3.9.2 The Reduced Cohomology Groups

The reduced homology groups and their relations with homology groups in a homol-
ogy theory H with coefficient group G. are studied in in Sect. 3.9.1.
Remark 3.9.9 The assertions in Sect. 3.9.1 have duals in a cohomology theory H
with coefficient group G, and they are left as exercises.

3.10 Invariance of Homology and Cohomology Groups

This section considers homology theory H = {H , ∗, ∂} with an abelian group G as its

coefficient group on the category T op2 of topological pairs and utilizes the Eilenberg
204 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

and Steenrod axioms for homology and cohomology theories to prove invariance
of homology groups in Sect. 3.10.1 and invariance of cohomology groups in Sect.
3.10.2.

3.10.1 Invariance of Homology Groups

This subsection proves invariance of homology groups in the sense that homology
group is a topological invariant. Moreover, it is proved that it is also a homotopy
invariant.

Theorem 3.10.1 Every homotopy equivalence f : (X , A) → (Y , B) in the category

T op2 induces an isomorphism

f∗ : Hn (X , A) → Hn (Y , B), for every integer n ≥ 0.

Proof Let f : (X , A) → (Y , B) be a homotopy equivalence with its homotopy

inverse g : (Y , B) → (X , A). Then

g ◦ f  1(X ,A) =⇒ (g ◦ f )∗ = g∗ ◦ f∗ = 1d ,

by using homotopy and identity axioms of Eilenberg and Steenrod. Similarly,

f ◦ g  1(Y ,B) =⇒ f∗ ◦ g∗ = 1d .

This asserts that f∗ is an isomorphism of groups with its inverse g∗ . 

Corollary 3.10.2 Every homeomorphism f : (X , A) → (Y , B) in the category T op2

induces an isomorphism

f∗ : Hn (X , A) → Hn (Y , B), for every integer n ≥ 0.

Proof Let f : (X , A) → (Y , B) be a homeomorphism with its inverse g : (Y , B) →

(X , A). Then g ◦ f = 1(X ,A) and f ◦ g = 1(Y ,B) . Hence the corollary follows by using
Theorem 3.10.1. 

Corollary 3.10.3 (i) Every homology group is a topological invariant.

(ii) Every homology group is a homotopy invariant.

Proof (i) Hn (X , A) is a topological invariant by Corollary 3.10.2 in the sense that if

f : (X , A) → (Y , B) is a homeomorphism in the category T op2 , then it induces
an isomorphism

f∗ : Hn (X , A) → Hn (Y , B) for every integer n ≥ 0.

3.10 Invariance of Homology and Cohomology Groups 205

(ii) Hn (X , A) also a homotopy invariant by Theorem 3.10.1 in the sense that if

f : (X , A) → (Y , B) is a homotopy equivalence in the category T op2 , then it
induces an isomorphism

f∗ : Hn (X , A) → Hn (Y , B) for every integer n ≥ 0.

Corollary 3.10.4 If two topological spaces X and Y are homotopically equivalent,

then the corresponding homology groups

Hn (X ) ∼
= Hn (Y ) ∀n ≥ 0.

Proof It follows from Theorem 3.10.1 by taking A = ∅ and B = ∅, in particular. 

Corollary 3.10.5 If two topological spaces X and Y are homeomorphic, then the
corresponding homology groups

Hn (X ) ∼
= Hn (Y ) ∀n ≥ 0.

Proof It follows from Corollary 3.10.2 by taking A = ∅ and B = ∅, in particular. 

Corollary 3.10.6 If X is a contractible space, then the homology group of X with

coefficient group G is 
∼ G, if p = 0
Hp (X ; G) =
{0}, otherwise.

Proof By hypothesis X is contractible. Hence X is homotopically equivalent to the

singleton space {∗}. This proves the corollary by using Corollary 3.10.4. 

3.10.2 Invariance of Cohomology Groups

This subsection considers cohomology theory H = {H , ∗, δ} with an abelian group

G as its coefficient group on the category T op2 of topological pairs proves invariance
of cohomology groups in the sense that homeomorphic pairs of topological spaces
have isomorphic cohomology groups.

Theorem 3.10.7 A homotopy equivalence f : (X , A) → (Y , B) in the category

T op2 induces an isomorphism

f∗ : Hn (Y , B) → Hn (X , A), for every integer n ≥ 0.

Proof The proof is similar to that of Theorem 3.10.1 

206 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Corollary 3.10.8 A homeomorphism f : (X , A) → (Y , B) in the category T op2

induces isomorphisms

f ∗ : H n (Y , B) → H n (X , A), for every integer n ≥ 0.

Proof The proof is similar to that of Corollary 3.10.2. 

Corollary 3.10.9 (i) Cohomology group is a topological invariant.

(ii) Cohomology group is a homotopy invariant.

Proof (i) H n (X , A) is a topological invariant by Corollary 3.10.8 .

(ii) H n (X , A) also a homotopy invariant by Theorem 3.10.7.


3.11 Consequences of the Exactness and Excision Axioms

of Eilenberg and Steenrod

This section considers homology theory H = {H , ∗, ∂} on the category T op2 of

topological pairs and establishes some immediate consequences of Eilenberg and
Steenrod axioms by using Exactness Axiom H(4) and Excision Axiom H(6) on the
homology theory H.

3.11.1 Consequence of the Exactness Axiom

This section studies the effect of the Exactness Axiom: H(4) on homology groups in
the homology theory H with coefficient group G.

Proposition 3.11.1 Let X be a topological space, A be a subspace of X and the

inclusion map i : A → X be a homotopy equivalence. Then Hn (X , A) = {0} for
every integer n ≥ 0.

Proof By hypothesis, i : A → X is a homotopy equivalence. Hence by Theorem

3.10.1 it follows that its induced homomorphism

i∗ : Hn (A) → Hn (X )

is an isomorphism for every integer n ≥ 0. Consider the exact homology sequence

(3.7) of the pair of topological spaces (X , A) on the category T op2 .

i∗ j∗ ∂ i∗
· · · → Hn (A) −−−→ Hn (X ) −−−→ Hn (X , A) −−−→ Hn−1 −−−→ Hn−1 (X ) → · · · (3.7)
3.11 Consequences of the Exactness and Excision Axioms of Eilenberg and Steenrod 207

Two isomorphisms i∗ in the exact sequence (3.7) assert that Hn (X , A) consists of

exactly a singleton element for every integer n ≥ 0 and hence Hn (X , A) = {0} for
every integer n ≥ 0. 
Corollary 3.11.2 For any topological space X , the homology group

Hn (X , X ) = {0}

for every integer n ≥ 0.

Proof Taking in particular, A = X in Proposition 3.11.1, the corollary follows. 
Remark 3.11.3 By specifying the topological spaces X and A in the pair (X , A) of
topological spaces, some interesting relations among the homology groups of (X , A),
X and A are established.
Proposition 3.11.4 Let X be a topological space and A be a retract of X . Then
(i) The inclusion map i : A → X induces a monomorphism

i∗ : Hn (A) → Hn (X ) for each integer n ≥ 0.

(ii) The inclusion map j : X → (X , A) induces an epimorphism

j∗ : Hn (X ) → Hn (X , A) for each integer n ≥ 0.

(iii) The boundary operator

∂ : Hn (X , A) → Hn−1 (A) is a trivial homomorphism for each integer n ≥ 0.

(iv) Hn (X ) ∼
= Hn (A) ⊕ Hn (X , A) for each integer n ≥ 0.
Proof Since A is a retract of X , there exists a retraction r : X → A such that r ◦ i :
A → A is the identity map 1A on A. Hence the axioms H(1) and H(2) assert that the
composite of the homomorphisms

i∗ r∗
Hn (A) −−−→ Hn (X ) −−−→ Hn (A) is the identity automorphism of the group Hn (A)
(3.8)
This asserts that i∗ is a monomorphism and r∗ is an epimorphism, and the abelian
group Hn (X ) decomposes into the direct sum

Hn (X ) = Im i∗ ⊕ ker r∗ for each integer n.

Consider the exact homology sequence (3.7) of the pair (X , A):

i∗ j∗ ∂ i∗
· · · → Hn (A) −−−→ Hn (X ) −−−→ Hn (X , A) −−−→ Hn−1 (A) −−−→ Hn−1 (X ) → · · ·
208 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Since i∗ : Hn−1 (A) → Hn−1 (X ) is a monomorphism, the above exact sequence

(3.7) asserts that ∂ is a trivial homomorphism and j∗ is an epimorphism.
Finally, since i∗ : Hn (A) → Hn (X ) is a monomorphism, Im i∗ ∼ = Hn (A) for each
integer n ≥ 0. Moreover the exactness of the above sequence asserts that ker j∗ =
Im i∗ . Since Hn (X ) = Im i∗ ⊕ ker r∗ and j∗ is an epimorphism, it follows that

ker r∗ ∼
= Hn (X )/Im i∗ = Hn (X )/ker j∗

by Isomorphism Theorem for groups. This proves that

Hn (X ) = Hn (A) ⊕ Hn (X , A) for each integer n ≥ 0.

Corollary 3.11.5 Let X be a topological space and x0 is a point in X . Then in the

homology theory H with coefficient group G,

H0 (X ) ∼
= G ⊕ H0 (X , x0 ),

Hn (X ) ∼
= Hn (X , x0 ) for n = 0.

Proof As {x0 } is a singleton subspace of the topological space X , it is a retract of X .

Hence the corollary is proved by using Proposition 3.11.4 and Corollary 3.10.6. 

3.11.2 Consequence of Excision Axiom

This subsection proves an Isomorphism Theorem 3.11.6 by using Excision Axiom

H(6) for homology theory.

Theorem 3.11.6 Let X be a topological space and U be an open set of X such that
U is contained in a subspace A of X . Then in a homology theory H on Top2 with
coefficient group G, the excision

e : (X − U, A − U ) → (X , A)

induces an isomorphism

e∗ : Hn (X − U, A − U ; G) → Hn (X , A; G) for each integer n ≥ 0,

if there exists an open set V in X such that the closure V of V is contained in U and
the inclusion map

i : (X − U, A − U ) → (X − V , A − V )is a homotopy equivalence.

3.11 Consequences of the Exactness and Excision Axioms of Eilenberg and Steenrod 209

Proof Let Å denote the interior of A in X . Since V ⊂ U ⊂ A by hypothesis, it follows

that V ⊂ Å. Then the excision

ẽ : (X − V , A − V ) → (X , A)

induces an isomorphism for each integer n ≥ 0

ẽ∗ : Hn (X − V , A − V ; G) → Hn (X , A; G)

by Excision Axiom H(6). Moreover, since

i : (X − U, A − U ) → (X − V , A − V )

is a homotopy equivalence, it induces an isomorphism

i∗ : Hn (X − U, A − U ; G) → Hn (X − V , A − V ; G) for each integer n ≥ 0.

Again, composite of maps

ẽ
e = ẽ ◦ i : (X − U, A − U ) → (X − V , A − V ) −−−→ (X , A)

induces an isomorphism e∗ by Composition Axiom H(2),

e∗ = ẽ∗ ◦ i∗ : Hn (X − U, A − U ; G) → Hn (X , A; G) for each integer n ≥ 0.

3.11.3 Additivity Property of Homology Theory

This subsection proves additivity property of homology theory in Proposition

3.11.7 by using Excision Axiom H(6).

Proposition 3.11.7 Let X and Y be topological spaces with X + Y be their topo-

logical sum. If
iX : X → X + Y , and iY : Y → X + Y

are the inclusion maps, then

iX∗ ⊕ iY∗ : Hn (X ; G) ⊕ Hn (Y ; G) → Hn (X + Y ; G)

is an isomorphism for any homology theory H satisfying E-S axioms with coefficient
group G.
210 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Proof Consider the exact homology sequence

iX ∗ j∗ ∂
· · · → Hp (X ; G) −−−−−→ Hp (X + Y ; G) −−−−→ Hp (X + Y , X ; G) −−−→ Hp−1 (X ; G) → · · ·
(3.9)
of groups and homomorphisms for the pair (X + Y , X ; G). The inclusion map iY :
(Y , ∅) → (X + Y , X ) is an excision map by taking U = X in Excision Axiom, and
hence by this axiom, there is an isomorphism

k∗ : Hp (Y ; G) → Hp (X + Y ; G)∀p ≥ 0,

where k = j ◦ 1Y . This asserts that 1∗Y ◦ k∗−1 is a splitting of the above long exact
sequence (3.9). 

Remark 3.11.8 The property of homology theory H proved in Proposition 3.11.7 is

taken as an axiom, called the additivity axiom. For example, the additivity axiom
added by J.W. Milnor (1931-) in 1962 with Eilenberg–Steenrod axioms to establish
the uniqueness of homology is not required for the category T op, but it contributes
for the category of all polyhedra (compact or not) and hence for the category of all
CW-complexes (not discussed in this book) [Milnor 1962].

3.11.4 Mayer-Vietoris Theorem

This subsection proves Mayer-Vietoris theorem by using ‘Excision Axiom’ in homol-

ogy theory H = {H , ∗, ∂}, which provides a technique to compute homology groups.

Definition 3.11.9 A topological triad (X ; A, B) is said to be a proper triad with

respect to a homology theory H if the inclusion maps

i : (A, A ∩ B) → (A ∪ B, B)
j : (B, A ∩ B) → (A ∪ B, A)

induce isomorphisms

i∗ : Hn (A, A ∩ B) → Hn (A ∪ B, B)
j∗ : Hn (B, A ∩ B) → Hn (A ∪ B, A)

in the homology theory H for every integer n.

Theorem 3.11.10 (Mayer-Vietoris Theorem) Let X , X1 , X2 and A be topological

spaces such that X = X1 ∪ X2 , A = X1 ∩ X2 . If the inclusion (X1 , A) → (X , X2 ) is
an excision, then there is a long exact sequence in homology, called Mayer-Vietoris
sequence of the proper topological triad (X ; X1 , X2 ) :
3.11 Consequences of the Exactness and Excision Axioms of Eilenberg and Steenrod 211

Fig. 3.8 Diagram for Mayer-Vietoris theorem

α β 
··· −
→ Hn (A) −−−→ Hn (X1 ) ⊕ Hn (X2 ) −−−→ Hn (X ) −−−→ Hn−1 (A) −
→ ···

Proof Consider the commutative diagram with two long exact homology sequences
as shown in Fig. 3.8 provided by axiom H(4), where by assumption α : Hn (X1 , A) →
Hn (X , X2 ) is an isomorphism by Exicison Axiom H(6). Then use four lemma to the
above diagram to complete the proof. 

3.12 Applications and Computations

This section presents some applications derived as further consequences of Eilenberg

and Steenrod axioms. Moreover, this section computes the ordinary homology groups
of S n with coefficients in an arbitrary abelian group G. Let C0 be the full subcategory
of C, whose objects are topological spaces with base points.

3.12.1 Computation of Homology Groups of Sn

Homology groups of n-sphere S n in a homology theory H with a coefficient G play

a key role in the development of topology. For example, degree of a spherical map is
a useful concept in topology. It is defined in Sect. 3.13 by using homology theory H
with the coefficient group G as an infinite cyclic group with the help of Hn (S n ; G)
This subsection computes the homology groups of S n in a homology theory H with
coefficient group G in Theorem 3.12.9, which determines the degree of a spherical
map f : S n → S n through homology theory.

3.12.2 Suspension Space and Suspension Functor

This subsection recalls the concepts of suspension space and suspension functor,
which are used in some computations.
212 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Definition 3.12.1 Given a pointed topological space X with base point x0 . its sus-
pension space, denoted by X , is defined to be the quotient space
X ×I
X =
(X × 0) ∪ (x0 × I) ∪ (X × 1)

endowed with quotient topology. For the point (x, t) ∈ X × I, the symbol [x, t]
denotes its corresponding point in X under the quotient map

p : X × I → X , (x, t) → [x, t].

This implies that

[x0 , 0] = [x0 , t] = [x, 1], ∀ x ∈ X and ∀ t ∈ I.

The point [x0 , 0] ∈ X is also denoted by x0 . Hence the space X is a pointed space
with base point x0 .

Example 3.12.2 For any integer n ≥ 0, the suspension space of the n-sphere is the
(n + 1)-sphere S n+1 .

Proposition 3.12.3 The suspension  : T op∗ → T op∗ is a covariant functor.

Proof The object function is defined by assigning to every object X ∈ T op∗ , its
suspension space X ∈ T op∗ , i.e., the object function is the assignment

X → X in the category T op∗

in the category T op∗ . The morphism function is defined by assigning to every

morphism f : X → Y ∈ T op∗ ,(which is a base point preserving continuous map
f : X → Y ) the morphism is the assignment

f : X → Y , [x, t] → [f (x), t] in the category T op∗ .

Then  satisfies the following properties:

(i) If f = 1X : X → X is the identity map in the category T op∗ , then (1X ) is also
an identity map in the category T op∗ , and
(ii) If f : X → Y and g : Y → Z are morphisms in the category T op∗ , then

(g ◦ f ) = (g) ◦ (f ).

This proves that  : T op∗ → T op∗ is a covariant functor. The functor  is

known as suspension functor.

3.12 Applications and Computations 213

Proposition 3.12.4 proves homotopy invariance property of the suspension functor

 in the sense that suspension functor sends homotopic maps to homotopic maps in
the category T op∗ .
Proposition 3.12.4 (Homotopy invariance of ) In the category T op∗ , if

f  g : X → Y,

then
f  g : X → Y .

Proof It follows from Definition of f . 

Theorem 3.12.5 proves a basic theorem in reduced homology theory, known as
suspension isomorphism theorem.
Theorem 3.12.5 (suspension isomorphism) Let X be a nonempty topological
space. Then there exists is an isomorphism

n (X ; G) → H
σn : H
 n+1 (X ; G)

in the reduced homology theory H with coefficient group G, where  is the suspension
functor.
Proof Let X be a nonempty topological space with a base point x0 ∈ X . Then it can
be embedded in its suspension space X by an embedding

i : X → X , x → p(x, 1/2),

where the standard projection map

p : X × I → X , (x, t) → [x, t] : [x, 0] = [x0 , t] = [x , 1], ∀ x, x ∈ X and ∀ t ∈ I.

Let U and V be two subspaces of X defined by

1 1
U = {p(x, t) : x ∈ X and t ∈ [0, ], i.e., 0 ≤ t ≤ }
2 2
and
1 1
V = {p(x, t) : x ∈ X and t ∈ [ , 1], i.e., ≤ t ≤ 1}.
2 2
Then U and V are both contractible spaces such that

U ∪ V = X and U ∩ V = X .

Since V is contractible, it follows by using the reduced homology exact sequence of

the pair (V , X ) of topological spaces that
214 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

n (X ; G)
∂ : Hn+1 (V , X ; G) → H

is an isomorphism for every integer n. Again, since U is contractible, it follows

by using the reduced homology exact sequence of the pair (X , U ) of topological
spaces that
n+1 (X ; G) → Hn+1 (X , U ; G)
j∗ : H

is an isomorphism for every integer n. Consider the inclusion map

i : (V , X ) → (X , U )

The inclusion map i is the excision of the open set Y = Int U = X − V from the
topological pair (X , U ). Define the open set

1
W = {p(x, t) : x ∈ X and 0 ≤ t < }
3

of X . Clearly, the closure W is contained in Y , and the inclusion map

i1 : (V , X ) → (X − W, U − X )

is a homotopy equivalence. This asserts by using the Excision Axiom that the induced
homomorphism
i∗ : Hn+1 (V , X ; G) → Hn+1 (X , U ; G)

is an isomorphism for every integer n. Consequently, the composite isomorphism

n (X ; G) → H
σn = j∗−1 ◦ ı∗ ◦ ∂ −1 : H
 n+1 (X ; G)

is an isomorphism for every integer n.



Definition 3.12.6 The isomorphism

n+1 (X ; G)
σn : Hn (X ; G) → H


for every integer n defined in Theorem 3.12.5 is called the suspension isomorphism
on the reduced homology group H n (X ; G).

Proposition 3.12.7 Let m, n be two integers with m ≥ 0. Then in a homology theory

H with coefficient group G,

Hn (S m , {∗}; G) ∼
= Hn+1 (S m+1 , {∗}; G) ∼
= Hn−m (S 0 , {∗}; G) ∼
= Hn−m ({∗}; G).
3.12 Applications and Computations 215

Proof The first two isomorphisms follow from suspensions S m+1 = S m =  m+1 S 0 .
The other isomorphism follows from the triad (S 0 , {−1}, {1}) and inclusion ({−1}, ∅)
→ (S 0 , {+1}). 
Theorem 3.12.8 computes the reduced homology groups of some common spaces,
from which the homology groups of spheres follow in Theorem 3.12.9.
Theorem 3.12.8 Let Dn+ be the closed upper half of the sphere S n . Then for every
integer n ≥ 0, the homology groups with coefficient group G are given by

p (S n ; G) ∼ G, if p = n
(i) H =
{0}, otherwise

(ii) Hp (D+ , S ; G) =
n n−1 ∼ G, if p = n
{0}, otherwise

G, if p = n
(iii) Hp (S n , Dn+ ; G) ∼
=
{0}, otherwise

Proof Denote the statements (i), (ii) and (iii) by 

(Sn ), (Dn ) and (Sn ), respectively.
They are proved by recursive method. The statement (S0 ) follows by using the Exci-
sion and Dimension Axioms asserting that

Hp (S 0 , D0+ ; G) ∼
= Hp (point) ∼
=G

for p = 0 and it is {0}, otherwise.

The equivalence of the statements 
(Sn ) and (Sn ) from the exact homology sequence
of the inclusion map Dn+ → S n

i∗
p (Dn+ ; G) −−−−→ H
j∗ ∂
p (S n ; G) −−−−→ Hp (S n , Dn+ ; G) −−−→ Hp−1 (Dn+ ; G) = {0}.
{0} = H
(3.10)

The equivalence of the statements (Dn ) and (Sn ) follows similarly asserting that

Hp (S n , Dn+ ; G) ∼
= Hp (S n − V , Dn+ − V ; G) ∼
= Hp (Dn+ , S n−1 ; G),

where V is some small nbd of the north pole of the sphere S n . Use the Excision
Axiom for the left-hand isomorphism and use the Homotopy Axiom for the right-
hand isomorphism.
Consider the exact sequence of the pair (Dn , S n−1 ) of spaces in reduced homology
theory

{0} = H p−1 (S n−1 ; G) → H

p (Dn ; G) → Hp (Dn , S n−1 ; G) → H p−1 (Dn ; G) = {0}.

since Dn is contractible. Hence it follows that

Hp (Dn , S n−1 ; G) ∼
=Hp−1 (S n−1 ; G)
216 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

The validity of the statement (S0 ) together with the equivalence of the statements
(D0 ). (S0 ) and (
S0 ) assert that

(D1 ) =⇒ (S1 ) =⇒ (
S1 ) =⇒ (D2 ) =⇒ · · ·


The above discussion is summarized in the important Theorem 3.12.9.
Theorem 3.12.9 (Homology Groups of Spheres) Let H be a homology theory with
coefficient group G. Then
⎧
⎪
⎨G ⊕ G, if n = m = 0
Hn (S m ; G) ∼
= G, if n = m = 0 or n = 0, m = 0
⎪
⎩
{0}, otherwise.

Theorem 3.12.10 is a basic result in homology theory, known as suspension iso-

morphism theorem.
Theorem 3.12.10 (suspension isomorphism) Let X be a topological space and
x0 ∈ X . Then there is an isomorphism

n : Hn (X , {x0 }; G) → Hn+1 (X , {∗}; G)

σn = 

in the homology theory H with coefficient group G, where  is the suspension functor.
Proof Let CX be the cone over X with vertex x0 . Then CX /X ≈ X . This asserts
that
Hn (CX /X ; G) ∼
= Hn (X ; G)

and
Hn+1 (CX , X ; G) ∼
= Hn (X , {x0 }; G).

Now, use the projection map

(CX , X ) → (CX /X , {∗})

to prove the theorem. 

Corollary 3.12.11 There exists an isomorphism

σ : Hn (S n : G) → Hn+1 (S n ; G) = Hn+1 (S n+1 ; G),

called suspension isomorphism on Hn (S n ; G) in the homology theory H with coef-

ficient group G.
Proof It follows from Theorem 3.12.10 by taking X = S n , in particular. 
3.13 Degrees of Spherical Maps from the Viewpoint of Homology Theory 217

3.13 Degrees of Spherical Maps from the Viewpoint

of Homology Theory

This section is devoted to the study of degrees of spherical maps by using homology
theory H with the coefficient group G as an infinite cyclic group. A study of degrees
of spherical maps by using homotopy theory has already been made in Chap. 2.
Historically, the concept of degrees of spherical maps was defined and studied by
L. E. J. Brouwer (1881–1967) during the period 1910–1912. It provides an important
tool in algebraic topology. This concept establishes a key link between homotopy
and homology theories with various applications. Some of them are studied in this
chapter, specially in its Sects. 3.13– 3.15. For example, Hopf classification theorem
3.13.7 characterizes homotopy of spherical maps with the help of their degrees, which
are integers. Definition 3.13.1 formulates the degree of a spherical map through
homology theory (which was defined in Chap. 2 by homotopy theory).

Definition 3.13.1 Let f : S n → S n be a continuous map for n ≥ 1. Then it induces

a homomorphism in the corresponding homology groups in a homology theory H
with coefficient group G

f∗ : Hn (S n ; G) → Hn (S n ; G) with Hn (S n ; G) ∼
= G.

By hypothesis, G is an infinite cyclic group. If α is a generator of Hn (S n ; G), then

there is some integer m such that

f∗ (α) = mα.

Given an arbitrary element x ∈ Hn (S n ; G), there is some integer p such that x =

pα. Consequently,

f∗ (x) = f∗ (pα) = pf∗ (α) = pmα = mpα = mx.

Moreover, f∗ (−α) = m(−α) and the integer m is independent of the choice of the
generator α of G. The integer m is called the degree of f , denoted by

deg f = m.

It follows from Definition 3.13.1 that a continuous f : S n → S n is of degree m iff the

induced homomorphism

f∗ : Hn (S n ; G) → Hn (S n ; G)

has the property that

f∗ (x) = mx, ∀ x ∈ Hn (S n ; G).

218 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

This asserts that if deg f = m, then the homomorphism f∗ is well defined by

f∗ : Hn (S n ; G) → Hn (S n ; G), x → mx.

Example 3.13.2 Consider the map

fn : S 1 → S 1 , z  → z n .

Then f−1 : S 1 → S 1 , z → z −1 sends every point z ∈ S 1 to its reciprocal z −1 . This

means that f−1 is the reflection of S 1 with respect to real axis. This asserts that
deg f−1 = −1. In general, deg fn = n.

3.13.1 Brouwer Degree Theorem

This subsection proves Brouwer degree theorem by using degree functions d : f →

deg f of spherical maps f : S n → S n through homology theory H instead of homo-
topy theory.

Theorem 3.13.3 (Brouwer degree theorem) Let f , g : S n → S n be two continuous

maps. If f  g, then deg f = deg g.

Proof Let f  g. Then by Homotopy Axiom for homology theory

f∗ = g∗ : Hn (S n ; G) → Hn (S n ; G) for all integers n.

This asserts that deg f = deg g. 

Proposition 3.13.4 Let 1S n : S n → S n , x → x be the identity map on S n . Then

deg 1S n = 1.

Proof Since 1S n : S n → S n , x → x induces the identity homomorphis on the homol-

ogy groups
1S n ∗ : Hn (S n ; G) → Hn (S n ; G) for all integers n.

This proves that deg 1S n = 1.

Remark 3.13.5 Theorem 3.13.3 raises the problem: Is the converse of Brouwer
degree theorem true? An affirmative answer is proved in Hopf classification Theorem
3.13.7.
3.13 Degrees of Spherical Maps from the Viewpoint of Homology Theory 219

Fig. 3.9 Diagram

connecting the suspension
operator σ and induced
homomorphisms in H

3.13.2 Hopf Classification Theorem by Using Homology

Theory

This subsection solves the problem raised in Remark 3.13.5 by proving that the
converse of Theorem 3.13.3 is also true by Hopf classification theorem 3.13.7. This
theorem provides a complete homotopy classification of spherical maps by using
their degrees defined by homology theory, which are integers.
Brouwer proved that if f and g are continuous maps on the 2-sphere which have
the same degree, then f and g are homotopic. H. Hopf (1894–1971) generalized
Brouwer theorem to an arbitrary dimension n by proving in 1927 that the converse
of Brouwer degree theorem is also true for arbitrary dimension n. These two results
taken together are known as Hopf classification theorem obtained by degree func-
tion. It is defined through homology theory and is proved in Theorem 3.13.7. On
the other hand, the same classification is also given in Chap. 2 by degree function
defined through homotopy theory.

Theorem 3.13.6 If a map f : S n → S n is of degree m, then its suspension map

h = f : S n+1 → S n+1 , [x, t] → [f (x), t]

is also of degree m.

Proof To prove the theorem, consider the commutative diagram of groups and
homomorphisms, where σ : Hn (S n ) → Hn+1 (S n+1 ) is the suspension isomorphism.
Hence it follows that h∗ = σ ◦ f∗ ◦ σ −1 . We claim that deg h = m. By hypotheses,
deg f = m. Then for any x ∈ Hn+1 (S n+1 ),

h∗ (x) = (σ ◦ f∗ ◦ σ −1 ) (x) = σ (f∗ ◦ (σ −1 ) (x)) = σ (mσ −1 ) (x)) = mx.

This implies that

h∗ : Hn+1 (S n+1 ) → Hn+1 (S n+1 ), x → mx.

This asserts that deg h = m (Fig. 3.9). 

Hopf classification theorem 3.13.7 characterizes homotopy of spherical maps with
the help of their degrees, which are integers.
220 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Theorem 3.13.7 (Hopf classification theorem by using homology theory) Let

f , g : S n → S n be two continuous maps. Then f  g, iff deg f = deg g.

Proof First let f  g. Then it follows by Brouwer degree theorem 3.13.3 that deg f =
deg g. Conversely, let deg f = deg g. If f , g : S 1 → S 1 are continuous maps such
that deg f = deg g, then it follows that f  g. This shows that the theorem is true
for n = 1. To use induction on n, suppose that the theorem is true for n − 1, i.e.,
suppose for continuous maps f , g : S n−1 → S n−1 , deg f = deg g. Then by using
Theorem 3.13.6, it follows that deg f = deg f and deg g = deg g. This asserts by
induction hypothesis that f and g are homotopic. 

3.13.3 More Properties of Degree Function

This subsection proves more properties of degree functions of spherical maps and
relates to the fixed-point property of spherical maps. Their proofs are similar to the
proofs given in Chap. 2.

Proposition 3.13.8 For every integer k ∈ In = {0, 1, 2, 3, . . . , n}, the reflection rk

of S n is about the xk − axis.

rk : S n → S n , x → (x0 , x1 , x2 , . . . , xk−1 , −xk , xk+1 , . . . , xn )

has degree −1.

Proof For n = 0, S 0 = {x, y} consists of two points x = 1 and y = −1 on the real

axis. Then f : S 0 → S 0 interchanges x and y. Now use additive property 3.11.7 of
homology theory. Clearly, deg r0 = −1, because r0 =  n−1 (f−1 ), where fn : S 1 →
S 1 , z → z n . Since for each k = 1, 2, 3, . . . , n, every map

rk : S n → S n , x → (x0 , x1 , x2 , . . . , xk−1 , −xk , xk+1 , . . . , xn )

is homotopic to r0 by rotation of S n , it follows that

deg r0 = −1 = deg r1 = deg r2 = · · · = deg rn .

This proves that every reflection rk of S n about the xk - axis has degree −1. 

Proposition 3.13.9 Let g, h : S n → S n be two continuous maps. Then

deg (g ◦ h) = deg g deg h.

Proof Let ψ = g ◦ h : S n → S n . Then its induced homomorphism

ψ∗ : Hn (S n ) → Hn (S n )
3.13 Degrees of Spherical Maps from the Viewpoint of Homology Theory 221

is such that

ψ∗ (x) = g∗ (h∗ (x)) = (deg g deg f )(x), ∀ x ∈ Hn (S n ).

It proves that deg ψ = deg g deg h. 

Definition 3.13.10 The continuous map

f : S n → S n , x = (x0 , x1 , x2 , . . . , xn ) → (−x0 , −x1 , −x2 , . . . , −xn ) = −x

is said to be antipodal.

Proposition 3.13.11 The antipodal map

f : S n → S n , x = (x0 , x1 , x2 , . . . , xn ) → (−x0 , −x1 , −x2 , . . . , −xn ) = −x

has the degree deg f = (−1)n+1 .

Proof The antipodal map f is the product of (n + 1) reflections

f = r0 ◦ r1 ◦ r2 ◦ · · · ◦ rn .

This proves that by Proposition 3.13.8 that

deg f = (−1)n+1 .

Remark 3.13.12 Proposition 3.13.13 is proved by using the concept of antipodal

map If x is a point of S n , then its antipode is the point −x ∈ S n . If f : S n → S n is
an antipodal map, then f (x) = −x, ∀ x ∈ S n , and if f is antipode preserving, then
f (−x) = −f (x), ∀ x ∈ S n .

Proposition 3.13.13 For any positive integer n, the sphere S n has no fixed-point
property.

Proof Let f : S n → S n be an arbitrary continuous map. Since the antipodal map f :

S n → S n , x → −x has no fixed point, it follows that S n has no fixed-point property.


Proposition 3.13.14 For any positive integer n, if two continuous maps f , g : S n →

S n are such that f (x) = g(x), ∀ x ∈ S n , then

deg f + (−1)n deg g = 0.

Proof By the given condition,

222 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

(1 − t)f (x) − tg(x) = 0, ∀ x ∈ S n and ∀ t ∈ I.

Because, if (1 − t)f (x) − tg(x) = 0, for some x ∈ S n and for some t ∈ I , then

1 − t = ||(1 − t)f (x)|| = t||g(x)||, since f (x), g(x) ∈ S n =⇒ 1 − t = t =⇒ f (x) = g(x) for some x ∈ S n ,

which contradicts our hypothesis. This asserts that the map

(1 − t)f (x) − tg(x)
H : S n × I → S n , (x, t) →
||(1 − t)f (x) − g(x)||

is well defined. H is also a continuous map such that

H : f  k, where k = r ◦ g and r : S n → S n is the antipodal map.

Hence

deg f = deg (r ◦ g) = deg r deg g = (−1)n+1 deg g =⇒ deg f + (−1)n deg g = 0.


Proposition 3.13.15 Let f : S → S be a continuous map such that f has no fixed
n n

point. Then
deg f = (−1)n+1 , ∀ n ≥ 1.

Proof By the given condition, it follows that

(1 − t)f (x) − tx = 0. ∀ t ∈ I

otherwise, 1 − t = t, ∀ t ∈ I would imply f (x) = x, contradicting the hypothesis

that f has no fixed point. Then the map
(1 − t)f (x) − tx
H : S n × I → S n , (x, t) →
||(1 − t)f (x) − tx||

is well defined and continuous and is such that f is homotopic to the antipodal
map
A : S n → S n , x → −x

under H and hence deg f = deg A = (−1)n+1 .

The proposition also follows from Proposition 3.13.14 by taking g as identity map
on S n . 
Corollary 3.13.16 Let f : S n → S n be a continuous map such that

deg f = (−1)n+1 .
3.13 Degrees of Spherical Maps from the Viewpoint of Homology Theory 223

Then f has a fixed point.

Proof It follows from Proposition 3.13.15 by way of contradiction. 
Corollary 3.13.17 Let f : S → S be continuous map such that f is homotopic
2n 2n

to the identity map. Then f has a fixed point.

Proof As f is homotopic to the identity map, deg f = +1. If possible, f has no
fixed point, then by Proposition 3.13.15 deg f = (−1)2n+1 = −1. This contradiction
shows that f has a fixed point. 
Corollary 3.13.18 Let Id : S n → S n be the identity map and n be an even integer.
Then Id is not homotopic to a map free from fixed points.
Proof It follows from Corollary 3.13.17. 

3.14 More Applications of Homology Theory

This section is devoted to prove some classical results : Brouwer fixed-point theorem
for any degree, invariance of dimensions of spheres and Euclidean planes. It also
solves some extension problems by using homology theory.

3.14.1 Brouwer Fixed-Point Theorem for Dimension n

This subsection proves Brouwer fixed-point theorem and its immediate consequences
by using homology theory. Historically, L.E.J. Brouwer (1881–1967) took the first
step toward connecting homotopy and homology by demonstrating in 1912 that two
continuous mappings of a two-dimensional sphere into itself can be continuously
deformed into each other if and only if they have the same degree (i.e., if they are
equivalent from the point of view of homology theory).
The papers of H. Poincaré (1854–1912) during 1895–1904 can be considered
as blue prints for theorems to come. The results of Brouwer during 1910–1912
may be considered the first one of the proofs in algebraic topology. He proved the
celebrated theorem ‘Brouwer fixed-point theorem 3.14.3’ by using the concept of
degree of a continuous spherical map defined by Brouwer himself. This subsection
proves Brouwer fixed-point theorem for finite dimension n ≥ 0 by using the tools of
homology theory. On the other hand, the same theorem is proved in Chap. 2 using
the tools of homotopy theory.
Proposition 3.14.1 Let Dn be the n-disk in Rn and S n−1 be its boundary. The identity
map
1S n−1 : S n−1 → S n−1

has no continuous extension over Dn for every n ≥ 1.

224 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Proof Clearly, S n−1 ⊂ Dn . Suppose that there exists some continuous extension

f : Dn → S n−1

of the identity map 1S n−1 for every n ≥ 1. Consider the two possible cases:
Case I For n = 1, the sphere S n−1 = S 0 = {±1} is a disconnected space. On
the other hand, Dn is connected for every n ≥ 1. In this particular case, Dn = D1
is connected, and hence, the image f (D1 ) is also connected. Since by assumption,
f : Dn → S n−1 is a continuous extension of the identity map 1S n−1 , it follows that
f (Dn ) = S n−1 . But in this case, since S 0 is not connected, we have a contradiction.
Case II For every n > 1, take k = n − 1 > 0, and consider the homology theory
H with coefficient group Z. Then it follows that

Hk (Dn ; Z) = Hn−1 (Dn ; Z) = {0} and Hk (S n−1 ; Z) = Hn−1 (S n−1 ; Z) ∼

= Z.

Since by hypothesis, 1S n−1 : S n−1 → S n−1 is the identity map, it follows that its
induced homomorphism

1S n−1 ∗ : Hk (S n−1 ) → Hk (S n−1 )

is the identity automorphism of the group Hk (S n−1 ). Since Hk (S n−1 ) ∼

= Z, it follows
that 1S n−1 ∗ is not the trivial homomorphism. Consider the inclusion map i : S n−1 →
Rn . Then, the map
f ◦ i : S n−1 → S n−1

is the composite of the continuous maps

i f
S n−1 −−−→ Dn −−−→ S n−1

and is the identity map 1S n−1 . Hence it follows that

Z∼
= Hn−1 (S n−1 ; Z) −−−→ Hn−1 (Dn ; Z) −−−→ Hn−1 (S n−1 Z) ∼
i∗ f∗
=Z

is the identity automorphism on the group Z in the homology theory H with coeffi-
cient group Z. But this is not possible, since

Hn−1 (S n−1 ; Z) ∼
= Z = {0} and Hn−1 (Dn ; Z) = {0}.

This produces a contradiction. This contradiction proves the proposition 

Remark 3.14.2 Theorem 3.14.3 proves Brouwer fixed-point theorem for any
finite dimension n by using the tools of homology theory used in Proposition 3.14.1.
An alternative proof of this theorem is also given by using the functorial property of
3.14 More Applications of Homology Theory 225

homology functor in a homology theory H. In Chap. 2, the same theorem is proved

by homotopy theory.

Theorem 3.14.3 (Brouwer fixed-point theorem for dimension n) Every continuous

map f : Dn+1 → Dn+1 has a fixed point for every n ≥ 0.

Proof If possible, f has no fixed point for any n ≥ 0. This implies that f (x) = x for
each x ∈ Dn . For n = 0, as it an immediate contradiction, it is well assumed from
now that n ≥ 1. By assumption, for each x ∈ Dn , the points x and f (x) are distinct.
For any x ∈ Dn , we now consider the half line in the direction from f (x) to x. Let h(x)
denote the point of intersection of this ray with S n . Then the map h : Dn+1 → S n is
continuous. Moreover, h(x) = x for every x ∈ S n . This implies that h : Dn+1 → S n
is an extension of the identity 1S n : S n → S n . But this contradicts Proposition 3.14.1.
This contradiction proves that f has a fixed point for every n ≥ 0.
Altenative proof: Using the functorial property of homology functor in a homol-
ogy theory H with a nonzero coefficient group G, it follows from the above discussion
that
h∗ ◦ i∗ : Hn (S n : G) → Hn (S n ; G)

is the identity automorphism on the group G in the homology theory H. But this is
not possible, since

Hn (S n ; G) ∼
= G = {0} and Hn (Dn+1 ; G) = {0}.

Remark 3.14.4 For more consequences of Eilenberg and Steenrod axioms for
homology and cohomology theories, see Exercises of Section 3.16 .

3.14.2 Hurewicz Homomorphism: Relation Between

Homology and Homotopy Groups

Hurewicz homomorphism defined by W. Hurewicz (1904–1956) in 1935 [Hurewicz

1935] establishes a close relation between homology and homotopy groups of a
specified class of topological spaces. His original definition is now simplified as
follows.

Definition 3.14.5 Let (X , x0 ) be a pointed topological space and gn be the standard

generator of the homology group Hn (S n ; Z) for n = 1, 2, 3, . . . . If f represents an
element α ∈ πn (X , x0 ), then
(i) The induced homomorphism

f∗ : Hn (S n ; Z) → Hn (X ; Z)
226 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

defines an element f∗ (gn ) ∈ Hn (X ; Z) and

(ii) f∗ defines a homomorphism

h : πn (X , x0 ) → Hn (X ; Z), α → f∗ (gn ), ∀ n = 1, 2, 3, . . . .

The homomorphism h is well defined, because, it is independent of the choice of the

representative f . This is called the Hurewicz homomorphism.
For n = 1, the fundamental group π1 (X , x0 ) is not abelian in general, but its abeli-
azation is isomophic to H1 (X ; Z), i.e., H1 (X ; Z)∼
=π1 (X , x0 )/[π1 (X , x0 ), π1 (X , x0 )],
where [π1 (X , x0 ), π1 (X , x0 )] is the commutator subgroup of π1 (X , x0 ), X is path
connected.
Hurewicz isomorphism 3.14.6 provides a sufficient condition under which
Hurewicz homomorphism h : πn (X , x0 ) → Hn (X ; Z) between the homotopy and
homology groups are isomorphic.
Theorem 3.14.6 (Hurewicz isomorphism)
(a) If X is a simply connected pointed topological space, then the following state-
ments are equivalent:
(i) πk (X ) = 0, ∀ k < n;
(ii) H̃k (X ) = 0, ∀ k < n.
(b) Moreover, (a) implies that the Hurewicz homomorphism h

h : πn (X , x0 ) → Hn (X ; Z), α → f∗ (gn )

formulated in Definition 3.14.5 is an 1 (n + 1) isomorphism .

Proof See [Gray 1975; Spanier 1966] 
Corollary 3.14.7 Let X be pointed simply connected space. Then X is n-connected
iff
Hm (X ) = 0, ∀ m = 2, 3, . . . n.

Remark 3.14.8 For special properties of Hurewicz homomorphism, see Exercises

21–23 of Sect. 3.16.

3.15 The Lefschetz Number and Fixed-Point Theorems

This section introduces the concept of Lefschetz number defined by S.Lefschetz

(1884–1972) in 1923 corresponding to each continuous map f : |K| → |K| from a
polyhedron into itself. The number is denoted by f . It is an integral-valued topo-
logical invariant and generalizes the Euler characteristic. It is closely linked with
the degree of a spherical map by Theorems 3.15.5. More properties of the Lefschetz
number are given in Exercise 24 of Section 3.16.
3.15 The Lefschetz Number and Fixed-Point Theorems 227

Definition 3.15.1 (Lefschetz number) Let X be a compact triangulable space with

a given triangulation K of dimension n and f : X → X be a continuous map. Then
there exists a homeomorphism ψ : |K| → X . The homology groups Hq (K; Q) with
rational coefficients are all vector spaces over Q and the homomorphisms
ψ
fp∗ = (ψ −1 ◦ f ◦ ψ)∗ : Hp (K; Q) → Hp (K; Q)

are linear transformations. Then

n
ψ
f = (−1)p trace fp∗
p=0

is defined to be the Lefschetz number of f , which is well defined, since the trace
of the corresponding matrices is independent of choice of the basis and the number
does not dependent on the triangulation of X .
Theorem 3.15.2 Let X be a compact triangulable space. If the identity map 1X :
X → X is homotopic to a fixed-point free map, then its Euler characteristic κ(X ) =
0.
Proof By hypothesis, X is a compact triangulable space. Then the Lefschetz number
1X of the identity map 1X : X → X is the characteristic κ(X ) of X . Since the
homotopic maps have the same Lefschetz number, it follows that κ(X ) = 0. 
Theorem 3.15.3 (Lefschetz fixed-point theorem) If X is a compact triangulable
space and f : X → X is a continuous map such that its Lefschetz number f = 0,
then f has a fixed point.
Proof Consider X as the polyhedron of a finite simplicial complex K and f : |K| →
|K|1 as a simplicial map. If f has no fixed point, then f (σ ) = σ for a simplex σ of
K. Orient each p-simplex of K in some way to obtain a basis of the vector space
Cp (K; Q) over Q. Then the linear map

fp : Cp (K; Q) → Cp (K; Q)

with respect to the above basis has the matrix representation Mfp with its diagonal
elements zeros, and hence, its trace is tracefp = 0. Since the Lefschetz number of f
at the chain level and at homology level is the same, it follows that
n n
(−1)p trace fp∗ = f = (−1)p trace fp . (3.11)
p=0 p=0

Then it follows from the relation (11.8) that f = 0, since trace fp = 0 for each p.
But it contradicts our hypothesis that f = 0. This contradiction asserts that f has a
fixed point.

228 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Remark 3.15.4 Theorem 3.15.5 asserts that the degree of a spherical map f : S n →
S n is closely related to its Lefschetz number f .

Theorem 3.15.5 For any spherical map f : S n → S n , its Lefschetz number f =

1 + (−1)n deg f .

Proof Given f : S n → S n , the only nonzero groups Hn (S n ; Q) are Q in dimension

n and 0, where in dimension n, the homomorphism f∗ induced by f

f∗ : Hn (S n ; Q) → Hn (S n ; Q) ∼
=Q

is obtained just multiplication by deg f . This proves the theorem. 

Lefschetz number of any spherical map is characterized in Corollary 3.15.6 by its

homotopic maps.

Corollary 3.15.6 For two spherical maps f , g : S n → S n , their Lefschetz numbers

f and g are the same iff f is homotopic to g.

Proof It follows from Theorem 3.15.5 by using Hopf classification theorem 3.13.7.


Corollary 3.15.7 If the degree of a spherical map f : S n → S n is neither 1 nor -1,

then f has a fixed point.

Proof It follows from Theorems 3.15.5 and 3.15.3. 

3.16 Exercises and Multiple Choice Exercises

As solving exercises plays an essential role of learning mathematics, various types

of exercises and multiple choice exercises are given in this section. They form an
integral part of the book series.

3.16.1 Exercises

This section considers exercises in a given homology theory H with coefficient group
G on the category T op2 , unless stated otherwise.
1. Prove that every CW -complex is a normal space.
2. Show that
(i) The torus T is a CW -complex with one 0-cell, two 1-cells and one 2-cell.
(ii) The real projective space RP n is a CW -complex with 1-cell of each dimen-
sion 0, 1, . . . , n.
3.16 Exercises and Multiple Choice Exercises 229

(iii) The complex projective space CP n is a CW -complex with one cell of dimen-
sion 2k for each k satisfying 0 ≤ k ≤ n, i.e., (n + 1) cells

e2k : 0 ≤ k ≤ n, symbolized , CP n = e0 ∪ e2 ∪ e4 ∪ · · · ∪ c2n .

(iv) The quaternionic projective space HP n is a CW -complex with one cell of

dimension 4k for each k satisfying 0 ≤ k ≤ n, i.e., (n + 1) cells

e4k : 0 ≤ k ≤ n, symbolized HP n = e0 ∪ e4 ∪ e8 ∪ · · · ∪ c4n .

3. {Xn } be a sequence of

CW -complexes such that Xn is a subcomplex of Xn+1 . Show
that the space X = Xn endowed with the weak topology is a CW -complex
and every Xn is a subcomplex of X . Hence prove that

(i) RP ∞ = ∞ n
n=1 RP endowed with the weak topology is a CW -complex.
∞ ∞
(ii) CP = n=1 CP endowed with the weak topology is a CW -complex.
n

(iii) HP ∞ = ∞ n
n=1 HP endowed with the weak topology is a CW -complex.
[ Hint: Consider RP n ⊂ RP n+1 as a subcomplex. Then RP ∞ endowed with
weak topology is a closure finite cell complex and hence RP ∞ is a CW -
complex. ]
4. Prove that the space of an infinite one-point union of circles is a CW -complex.
In particular, the figure-eight space F8 is a CW -complex.
[ Hint: Let X = ∨∞ ∞ 1
i Sn endowed with the weak topology as a subset of i Sn
1

having the product topology. X is a closure finite cell complex. Consider F8 =

S 1 ∨ S 1. ]
5. Using Dimension Axiom, show that the homology groups of any singleton space
X in the homology theory H with coefficient group G, are

G if p = 0
Hp (X ; G) ∼
=
{0}, otherwise.

6. Show that the homology groups of a discrete topological space X = {x1 , x2 , . . . ,

xn } in the homology theory H with coefficient group G, are
⎧ n
⎨  
Hp (X ; G) ∼
= G ⊕ G ⊕ · · · ⊕ G if p = 0
⎩
{0}, otherwise.

7. In the homology theory H with coefficient group G, show that for every integer
n ≥ 0,
(i) n (S n ) ∼
H = G.
(ii) Hm (S n ) ∼
 = {0}, if m = n.
230 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

8. In the homology theory H with coefficient group G, show that for every integer
n ≥ 0,
(i) Hn (S n ; G) ∼
= G, if n = 0;
(ii) H0 (S 0 ; G) ∼
= G ⊕ G;
(iii) Hm (S n ; G) ∼
= {0}, if m = n = 0.
9. For the pair (Rn , S n−1 ) of spaces, show that the homology groups of (Rn , S n−1 )
in the homology theory H with coefficient group G are

G, if p = n
Hp (Rn , S n−1 ; G) ∼
=
{0}, otherwise.

10. Show that the homology groups of an n-dimensional manifold M n (n ≥ 1) in

the homology theory H with coefficient group G are

Hp (M , {x}; G) =
n ∼ G, if p = n
{0}, otherwise.

for every point x ∈ M n .

11. Given an integer k, show that there is a continuous map f : S n → S n such that
deg f = k.
12. Let f : S n → S n be a continuous map for n > 0. Show that

(i) If deg f = m and  f : S n+1 → S n+1 is its suspension map, then

deg f = deg (f ) = m, ∀ n > 0.

(ii) The suspension map  f : S n+1 → S n+1 of f has degree m iff

deg f = m.

13. Show that the antipodal map on S n is not homotopic to the identity map on S n
for every even integer n > 1.
14. If n is an even integer, by using homology theory H, show that S n admits no
nonzero continuous tangent vector field.
[ Hint: To prove it by contradiction method, use Exercise 13. ]
15. Let X be a topological space and X1 , X2 be two subsets of X such that
(i) X1 is closed and
(ii) X = IntX1 ∪ IntX2 .
Show that the inclusion map

i : (X1 , X1 ∪ X2 ) → (X , X2 )
3.16 Exercises and Multiple Choice Exercises 231

induces an isomorphism

i∗ : Hn (X1 , X1 ∪ X2 ) → Hn (X , X2 )

is an isomorphism for every integer n.

[ Hint: Apply Excision Axiom H(6).]
16. Construct a statement of Mayer-Vietoris theorem 3.11.10 in the cohomology
theory and prove it.
17. For any topological pair (X , A) with X = ∅ and A = ∅, show that the reduced
homology sequence of (X , A)
j∗ ∂
··· → H p (X ) −−−→ H
p (A) −−−→ H i∗
p (X , A) −−−→ H
p−1 (A) → · · ·

of groups and homomorphisms is exact in the homology theory H.

18. Let f : (X , A) → (Y , B) be a map of pair of spaces. If both f : X → Y and
f |A : A → B are homotopy equivalences, show that f∗ : Hn (X , A) → Hn (Y , B)
is an isomorphism for all n.
[Hint. Consider the commutative diagram of two rows of exact sequences as
shown in Fig. 3.10, and use five lemma result
19. Show that the kernel sequence ( 3.5) of a map f : (X , A) → (Y , B) is exact if
there exists a map

g : (Y , B) → (X , A) such that f ◦ ghomotopic to identity map on (Y , B).

20. Show that for every pair of topological spaces (X , X )

Hn (X , X ) = {0}, ∀ n ≥ 0.

[ Hint: Apply ‘Exactness Axiom’ for the pair (X , X ).]

21. (Hurewicz) Given a pointed topological space (X , x0 ), if

π0 (X , x0 ) = {0}, π1 (X , x0 ) = {0}, π2 (X , x0 ) = {0}, . . . , πn−1 (X , x0 ) = {0}, for all n ≥ 2,

show that
(i)

H1 (X ; Z) = {0}, H2 (X ; Z) = {0}, . . . , Hn−1 (X ; Z) = {0}, for all n ≥ 2;

(ii) Hurewicz homomorphism

h : πn (X , x0 ) → Hn (X ; Z), α → f∗ (gn ),

given in Definition 3.14.5, is an isomorphism for all n ≥ 2.

232 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

Fig. 3.10 Commutative diagram of exact sequences

Hint : See [ Adhikari, 2016]

22. Given a path-connected topological X , show that under usual notation, there
exists an epimorphism
h : π1 (X ) → H1 (X ; Z),

which determines an isomorphism

h∗ : π1 (X )/ker h → H1 (X ; Z)

with ker h the commutator subgroup of π1 (X ). Hence prove that h is an isomor-

phism iff the fundamental group π1 (X ) is abelian.
23. A pointed topological space (X , x0 ) is said to be n-connected, if πk (X , x0 ) =
0, ∀ k ≤ n. Given an n-connected topological space (X , x0 ), show that
(i) Hk (X˜ ; Z) = {0}, ∀ k ≤ n.
n+1 (X ; Z) is an isomor-
(ii) The Hurewicz homomorphism h : πn+1 (X , x0 ) → H
phism for all n ≥ 1.
24. Let X be a compact triangulable space and f : X → X be a continuous map with
its Lefschitz number f . Show that
(i) f = 0 implies f has a fixed point.
(ii) Every contractible compact triangulable space has the fixed-point property
(Brouwer fixed-point theorem).
(iii) If the identity map 1X : X → X is homotopic to a fixed-point free map
f : X → X , the Euler characteristic κ(X ) of X is zero.

3.16.2 Multiple Choice Exercises

Identify the correct alternative (s) (there may be none or more than one) from the
following list of exercises:

1. (i) The spheres S 10 and S 11 are homeomorphic.

(ii) The spheres S 100 and S 111 are homotopically equivalent.
(iii) The spheres S 11 and S 12 are neither homeomorphic nor homotopically equiv-
alent.
3.16 Exercises and Multiple Choice Exercises 233

2. (i) The Euclidean spaces R10 and R11 are homeomorphic.

(ii) The Euclidean spaces R10 and R11 are homotopically equivalent.
(iii) The Euclidean spaces R10 and R11 are neither homeomorphic nor homo-
topically equivalent.
3. (i) Let R5 − {0} be punctured five-dimensional Euclidean space and S 4 be the
four-dimensional sphere. Then the inclusion map

i : S 4 → R5 − {0}

is a homotopy equivalence.
(ii) The homology group of Hn (R5 − {0}) of the punctured Euclidean space
R5 − {0} and the homology group of Hn (S 4 ) of the sphere S 4 are isomorphic
for all n > 1.
(iii) The cohomology group of H n (R5 − {0}) of the punctured Euclidean space
R5 − {0} and the homology group of H n (S 4 ) of the sphere S 4 are isomorphic
for all n > 1.
4. Let H be a homology theory with coefficient group G = {0}. Then for the sub-
space S 4 of the punctured Euclidean space R5 − {0} ,
n (S 4 ) = Hn (S 4 ) for every integer n = 0.
(i) H
0 (S 4 ) ⊕ G.
(ii) H0 (S 4 ) = H
(iii) For X = R5 − {0}, the identity map 1X : X → X induces the identity auto-
morphism of reduced homology groups

n (X ; G) → H
1X ∗ : H n (X ; G), ∀ n > 1.

5. (i) Homology group of a topological space is a topological invariant.

(ii) Cohomology group of a topological space is a topological invariant.
(iii) If X and Y are two homotopically equivalent topological spaces, then their
homology groups Hn (X ; G) and Hn (Y ; G) are isomorphic.
6. (i) The 101-dimensional sphere S 101 admits a nowhere nonvanishing tangent
vector field.
(ii) Every tangent vector field on the 100-dimensional sphere S 100 vanishes.
(iii) If v : S 51 → R52 is a nowhere vanishing tangent vector field on the 51-
dimensional sphere S 51 , then the map

v(x)
f : S 51 → S 51 , x →
||v(x)||

is homotopic to the identity map.

234 3 Homology and Cohomology Theories: An Axiomatic Approach with Consequences

References

Adams JF. Algebraic topology: a student’s guide. Cambridge: Cambridge University Press; 1972.
Adhikari, M.R., Adhikari, A.: Groups. Rings and Modules with Applications. Universities Press,
Hyderabad (2003)
Adhikari, M.R., Adhikari, A.: Textbook of Linear Algebra: An Introduction to Modern Algebra.
Allied Publishers, New Delhi (2006)
Adhikari A, Adhikari MR. Basic topology, vol. 1: metric spaces and general topology. India:
Springer; 2022a.
Adhikari A, Adhikari MR, Basic topology, vol. 2: topological groups, topology of manifolds and
lie groups. India: Springer; 2022b.
Adhikari MR, Adhikari A. Basic modern algebra with applications. New Delhi, New York, Heidel-
berg: Springer; 2014.
Adhikari MR. Basic algebraic topolgy and its applications. India: Springer; 2016.
Bredon GE. Topology and geometry. New York: Springer; 1993.
Dieudonné J. A history of algebraic and differential topology, 1900–1960. Modern Birkhäuser;
1989.
Eilenberg S, Steenrod N. Foundations of algebraic topology. Princeton: Princeton University Press;
1952.
Gray B. Homotopy theory. An introduction to algebraic topology. New York: Academic Press;
1975.
Hatcher A. Algebraic topology. Cambridge University Press; 2002.
Hilton PJ, Wylie S. Homology theory. Cambridge: Cambridge University Press; 1960.
Hu ST. Homology theory. Oakland CA: Holden Day; 1966.
Hurewicz W. Beitrage de Topologie der Deformationen. Proc. K. Akad. Wet. Ser. A 1935;38:112–
119, 521–528.
Milnor JW. On axiomatic homology theory . Pacific J. Math 1962;12:337-341.
Poincaré H. Analysis situs. J. l’Ecole Polyt. 1895;1:1–121
Poincaré H. Papers on topology: analysis situs and its five supplements. Amer. Math. Soc. 2010;37.
(Translated by Stillwell J, History of Mathematics)
Rotman JJ. An introduction to algebraic topology. New York: Springer; 1988.
Spanier E. Algebraic topology. McGraw-Hill; 1966.
Switzer RM. Algebraic topology-homotopy and homology. Berlin, Heidelberg, New York: Springer;
1975.
Chapter 4
Topology of Fiber Bundles: General
Theory of Bundles

A bundle in topology is a union of fibers parametrized by its base space and glued
together by the topology of the total space. On the other hand, a fiber bundle is a
bundle with an additional structure derived from the action of a topological group
on the fibers. A fiber bundle is a locally trivial fibration having covering homotopy
property. The theory of fiber bundles, in particular, vector bundles, establishes a
very strong link between algebraic topology and differential topology. Topology of
fiber bundles is studied in Chaps. 4 and 5. This chapter studies General Theory of
Bundles. On the other hand, Chap. 5 develops the theory of bundles based on their
homotopy properties. The topology of fiber bundles has created general interest and
promises for more work, because it is involved of interesting applications of topology
to other areas such as algebraic topology, geometry, physics and gauge groups.
Historically, the theory of fiber bundles was first recognized during the period 1935–
1940 through the work of H. Whitney (1907–1989), H. Hopf (1894–1971), E. Stiefel
(1909–1978), J. Feldbau (1914–1945) and some others.
For this chapter, the books (Adhikari, 2022a), (Bredon, 1993), (Husemoller, 1966),
(Steenrod, 1951), (Adhikari, 2022b), (Adhikari, 2014), (Armstrong, 1983), (Hu,
1959), (Mayer, 1972), (Massey, 1991), (Maunder, 1970), (Munkers, 1984), (Rot-
man, 1988), (Spanier, 1966).

4.1 General Properties of Bundles in Topology

This section introduces the concept of bundles in topology and studies bundles.
Bundles in topology play a key role in the theory of both fiber bundles and vector
bundles. So we start with the study of preliminaries of the concept of bundles. Since
a bundle in topology is a union of fibers parametrized by its base space and glued
together by the topology of its total space, it provides a basic underlying structure

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 235
M. R. Adhikari, Basic Topology 3,
https://doi.org/10.1007/978-981-16-6550-9_4
236 4 Topology of Fiber Bundles: General Theory of Bundles

for both the fiber bundles and vector bundles. They are special families of bundles
with additional structures and are closely related to topology. Their deep study uses
homotopy theory (see Chap. 5). The concept of fiber spaces is the most fruitful
generalization of covering spaces, and its importance was first realized during 1935–
1950 which facilitated to solve several problems involving homotopy and homology
theories. On the other hand, a vector bundle is a bundle admitting an additional vector
space structure on each of its fibers.

4.1.1 The Concept of Bundles in Topology

A bundle is a triple consisting of two topological spaces, one is called total space and
the other is called base space connected by a continuous map from the total space to
the base space, called the projection of the bundle space. This concept is formalized
in Definition 4.1.1.

Definition 4.1.1 A bundle ξ = (X , p, B) is an ordered triple which consists of a

topological space X , called the total space of ξ, a topological space B, called the
base space of ξ, and a continuous onto map p : X → B, called the projection of the
bundle ξ. The space Fb = p−1 (b) ⊂ X with subspace topology inherited from the
topology of X is called the fiber of ξ over b for each point b ∈ B. It is also sometimes
abbreviated as Xb .
 
Remark 4.1.2 The total space X = Xb = p−1 (b) and any two fibers Xb and
b∈B b∈B
Xa are disjoint if b = a. This implies that every point of X lies in exactly one fiber.
The notation X (ξ) is sometimes used to denote the total space and B(ξ) the base
space of the bundle ξ to avoid any confusion.

Example 4.1.3 The fibers of a bundle may be of different types. For example, con-
sider the bundle ξ = (X , p, B) displayed in Fig. 4.1. The total space X of the bundle
ξ = (X , p, B) is decomposed into four different types of fibers such as a line segment,
two line segments, a point together with a line segment and a point.

Fig. 4.1 Bundle ξ having

four types of fibers
4.1 General Properties of Bundles in Topology 237

Definition 4.1.4 A bundle η = (Y , q, B) is called a subbundle of a bundle ξ =

(X , p, B) if
(i) Y is a subspace of X and
(ii) q = p|Y : Y → B is its projection.

Moreover, if A is a subspace of B, then the bundle (Y , q, A) is a particular subbundle

of ξ.

Example 4.1.5 Given a bundle ξ = (X , p, B), if η = (Y , q, A) is a bundle such that

there are inclusion maps
(i) i : Y → X and
(ii) j : A → B with q = p|Y is the restriction of p over Y ,
then η is a subbundle of the bundle ξ.

Definition 4.1.6 (Induced bundle) Given a bundle ξ = (X , p, B) and a continuous

map f : A → B from a space A, the induced bundle f ∗ (ξ) = (Y , q, A) of ξ over A
under f is the bundle (Y , q, A), where the total space Y is defined by

Y = {(a, x) ∈ A × X : f (a) = p(x)}

and the projection q is defined by

q : Y → A, (a, x) → a.

Proposition 4.1.7 Given a bundle ξ = (X , p, B) and a continuous map f : A → B

from a space A, let f ∗ (ξ) = (Y , q, A) be the induced bundle of ξ over A under f . If

p:X →B

is an open map, then

q:Y →A

is also an open map.

Proof It follows from the Definition 4.1.6 of the induced bundle. 

Definition 4.1.8 (Product bundle) The bundle (B × F, p, B), where p : B × F →

B, (b, x) → b is the projection from the product space B × F on the first factor, is
called the product bundle over B with fiber F.

Definition 4.1.9 A trivial bundle ξ = (X , p, B) with fiber F is a bundle such that

(i) Its total space X is homeomorphic to the product space B × F.
(ii) p : B × F → B, (b, x) → b is the projection from the product space B × F on
the first factor.
238 4 Topology of Fiber Bundles: General Theory of Bundles

Definition 4.1.10 Let ξ = (X , p, B) be a bundle with fiber F. An open covering

{Ui : i ∈ K} of its base space B with a family of homeomorphisms

ψi : p−1 (Ui ) → Ui × F

is said to be a local trivialization if

p ◦ ψi−1 (b, x) = b, ∀ b ∈ Ui , ∀ x ∈ F and ∀ i ∈ K.

Example 4.1.11 (M öbius band) It is obtained from the product space L × F by

identifying the two ends with a twist described in Basic Topology, Volume 1, Chap.
3, of the present series of books, where L is a line segment. Consider projection

p : L × F → L, (x, y) → x.

Then the Möbius band is a bundle whose base space is the circle B, obtained from
the line segment L by identifying its end points. Its fiber F is a line segment. From
the viewpoint of manifold, the Möbius band is the quotient manifold R2 mod Z, as
described in Basic Topology, Volume 2 of the present series of books.

Example 4.1.12 (Klein bottle) It is a bundle obtained from the product space L × F
by identifying the two ends of the cylinder L × F described in Basic Topology,
Volume 1, Chap. 3, of the present series of books, where L is a line segment. Its base
space is the circle B, obtained from a line segment L by identifying its end points.
Its fiber F is a circle. Klein bottle cannot be topologically embedded in R3 . From
the viewpoint of manifold, Klein bottle is the quotient manifold R2 mod (Z × Z)
described in Basic Topology, Volume 2 of the present series of books.

4.1.2 General Properties of Cross Sections of Bundles

This subsection studies the cross sections of a bundle, some of them are identified
with familiar geometric objects.

Definition 4.1.13 A cross section s of a bundle ξ = (X , p, B) is a continuous map

s:B→X

such that the composite

p ◦ s = 1B ,

where 1B : B → B is the identity map on B.

Proposition 4.1.14 For a bundle ξ = (X , p, B), every cross section

4.1 General Properties of Bundles in Topology 239

s:B→X

is injective and s(b) ∈ p−1 (b), ∀ b ∈ B.

Proof Let ξ = (X , p, B) be a bundle and s : B → X be a cross section. Then for

every
b ∈ B, (p ◦ s)(b) = 1B (b) = b

implies that s(b) ∈ p−1 (b). It asserts that s(b) ∈ p−1 (b) for every b of the base space
B. Again the condition p ◦ s = 1B proves that the map s : B → X is injective. 

Proposition 4.1.15 Let ξ = (B × F, p, B) be a product bundle. Then its cross sec-

tions s are precisely of the form

s : B → B × F, b → (b, fs (b)),

such that fs : B → F is a continuous map which is uniquely determined by the cross

section s.

Proof Every cross section s : B → B × F takes the form

s : B → B × F, b → (gs (b), fs (b)) : p ◦ s = 1B ,

where the maps fs : B → F and gs : B → B are uniquely determined by s. Thus

p(s(b)) = 1B (b), ∀ b ∈ B and hence

p(gs (b), fs (b)) = b asserts that gs (b) = b, ∀ b ∈ B,

which implies that s(b) = (b, fs (b)), ∀ b ∈ B.

Conversely, given a continuous map s : B → B × F, suppose that

s(b) = (b, fs (b)), ∀ b ∈ B.

Then
(p ◦ s)(b) = b, ∀ b ∈ B asserts that p ◦ s = 1B ,

which proves that s is a cross section of the product bundle ξ = (B × F, p, B). 

Remark 4.1.16 The map fs : B → F defined in Proposition 4.1.15 has the property
given in Corollary 4.1.17.

Corollary 4.1.17 The cardinality of the set S(ξ) of all cross sections of a product
bundle ξ = (B × F, p, B) and the set C(ξ) of all continuous maps B → F are same.

Proof Define a map φ using the map fs : B → F formulated in Proposition 4.1.15

240 4 Topology of Fiber Bundles: General Theory of Bundles

Fig. 4.2 Rectangle

representing morphism of
bundles

φ : S(ξ) → C(ξ), s → fs .

Since the map φ : S(ξ) → C(ξ), s → fs is a bijection, the corollary follows. 

Proposition 4.1.18 characterizes the cross sections of a subbundle of a bundle in
terms of cross sections of its mother bundle, which follows from the above discussion.
Proposition 4.1.18 Given a subbundle ξ = (X , p , B) of a bundle ξ = (X , p, B),
a continuous map s : B → X is a cross section of ξ iff s(b) ∈ X for every b ∈ B.

4.1.3 General Properties of Bundle Morphisms

This subsection conveys the concept of bundle morphisms needed for the study of
fiber and vector bundles specially, for comparing fiber bundles or vector bundles over
the same or different base spaces likewise group homomorphisms (for example, see
Proposition 4.1.21).
Definition 4.1.19 Given two bundles ξ = (X , p, B) and η = (Y , q, A), a bundle
morphism or a fiber map consists of a pair of continuous maps

(f , g) : ξ → η : f : X → Y and g : B → A

such that q ◦ f = g ◦ p, which means that in mapping diagram Fig. 4.2, the rectangle
is commutative.

Remark 4.1.20 Proposition 4.1.21 proves that the continuous map f : X → Y in

Fig. 4.2 is fiber preserving in the sense that f (p−1 (b)) ⊂ q−1 (g(b)), ∀ b ∈ B.
Proposition 4.1.21 The map f : X → Y in Fig. 4.2 is fiber preserving
Proof For each b ∈ B, the map f sends fibers of ξ over b into the fibers of η over f (b).
Because, for each x ∈ X , the equality (q ◦ f )(x) = (g ◦ p)(x) holds. This shows that
the pair of maps (f , g) sends the pair (x, p(x)) into the pair (f (x), g(p(x))) by (f , g).
Hence, it follows that for every b ∈ B

f (p−1 (b)) ⊂ q−1 (g(b)).


4.1 General Properties of Bundles in Topology 241

Fig. 4.3 Triangle

representing B-morphism of
bundle f : ξ → η

Fig. 4.4 Cross section of

bundle

Definition 4.1.22 formalizes the interesting particular concept of B-morphism,

when ξ and η are both bundles over the same base space B.

Definition 4.1.22 Let ξ = (X , p, B) and η = (Y , q, B) be two bundles over the

same base space B. Then a continuous map f : X → Y is said to be a bundle
B-morphism (in brief, B-morphism) written as f : ξ → η if the triangle in the Fig. 4.3
is commutative in the sense that p = q ◦ f .

Proposition 4.1.23 The map f : X → Y given in Definition 4.1.22 is fiber preserv-

ing.

Proof By hypothesis q ◦ f = p and hence f (p−1 (b)) ⊂ q−1 (b) for every b ∈ B. 

Proposition 4.1.24 shows that a cross section of a bundle satisfies the general
property of a bundle morphism.

Proposition 4.1.24 Let ξ = (X , p, B) be a bundle over B. Then its cross sections

are precisely the B-morphisms s : (B, 1B , B) → (X , p, B).

Proof For any cross section s of ξ, since p ◦ s = 1B , it follows that the triangle in
Fig. 4.4 is commutative. This shows that s is a B-morphism. For its converse, let
s : (B, 1B , B) → (X , p, B) be a B-morphism. Then it follows that s is a cross section
of ξ. 

Example 4.1.25 For a subbundle (Y , q, A) of a bundle (X , p, B) with inclusion maps

i : Y → X and j : A → B, the pair (i, j) : (Y , q, A) → (X , p, B) is a bundle mor-
phism.

Definition 4.1.26 Given two bundles ξ = (X , p, B) and η = (Y , q, A), a bundle map

(f , g) : ξ → η is said to be a bundle isomorphism if both the maps

f : X → Y and g : B → A

are homeomorphisms.
242 4 Topology of Fiber Bundles: General Theory of Bundles

Example 4.1.27 Let ξ = (X , p, B) and η = (Y , q, A) be two bundles and (f , g) :

ξ → η be a bundle morphism. If (f , g) is a bundle isomorphism, then the pair
(f −1 , g −1 ) : η → ξ is also a bundle isomorphism with both the pairs of maps

(f ◦ f −1 , g ◦ g −1 ) and (f −1 ◦ f , g −1 ◦ g)

are identity bundle morphisms.

Definition 4.1.28 Given a bundle ξ = (X , p, B) and a nonempty subset A of B,
the restricted bundle of ξ to A, written as ξ|A, is the bundle (Y , q, A), such that
Y = p−1 (A) and q = p|Y .
Example 4.1.29 Let ξ = (X , p, B) be a bundle. If X is a nonempty subspace of X
and p = p|X : X → B, then (X , p , B) is a restricted bundle of ξ.
Definition 4.1.30 Given two bundles ξ = (X , p, B) and η = (Y , q, B) over the same
base space B, if there exists a homeomorphism f : X → Y , then f is called a B-
isomorphism. If for every point b ∈ B, there is an open nbd Ub of b and an Ub -
isomorphism between the restricted bundles ξ|Ub and η|Ub , then the given bundles
are said to be locally isomorphic.
Proposition 4.1.31 A bundle which is locally isomorphic to a locally trivial bundle
is also locally trivial.
Proof Let B be the set of all bundles over B and ∼ be the equivalence relation on the
set B defined by two bundles ξ ∼ η in B iff they are locally isomorphic. This asserts
that if ξ is locally isomorphic to a locally trivial bundle, then ξ is locally trivial. 
Definition 4.1.32 (Canonical morphism) Given a bundle ξ = (X , p, B) and a con-
tinuous map f : A → B, if the bundle f ∗ (ξ) = (Y , q, A) is the induced bundle
under f in the sense of Definition 4.1.6, then the pair of maps (g, f ), where
g : Y → X , (a, x) → x forms a bundle morphism

(g, f ) : f ∗ (ξ) → ξ

called the canonical morphism of the induced bundle.

Remark 4.1.33 Proposition 4.1.34 establishes an interesting result asserting that
if f ∗ (ξ) = (Y , q, A) is the induced bundle of ξ = (X , p, B) by a continuous map
f : A → B, then the fibers q−1 (a) and p−1 (f (a)) are homeomorphic for every a ∈ A.
Proposition 4.1.34 Let ξ = (X , p, B) be a bundle and f : A → B be a continuous
map. If f ∗ (ξ) = (Y , q, A) is the induced bundle under f and (g, f ) : f ∗ (ξ) → ξ is
the canonical bundle morphism, then for every a ∈ A, the restricted map

g|q−1 (a) : q−1 (a) → p−1 (f (a))

is a homeomorphism.
4.1 General Properties of Bundles in Topology 243

Fig. 4.5 Construction of

cross section of the bundle
(Y , q, A) by using (f , g)

Fig. 4.6 Construction of

induced cross section of
(Y , q, A) by using s and f

Proof It follows from the property of the induced bundle. 

Proposition 4.1.35 Given a bundle ξ over B and two continuous maps g : A → B
and h : C → A,
(i) The induced bundle 1∗B (ξ) and the bundle ξ are B-isomorphic.
(ii) The induced bundles h∗ (g ∗ (ξ)) and (g ◦ h)∗ (ξ) are C-isomorphic.
Proof It follows from definition of the induced bundle. 
Proposition 4.1.36 Given a bundle morphism (f , g) : (X , p, B) → (Y , q, A) and a
cross section s : B → X of (X , p, B), if the map g : B → A is a homeomorphism,
then
s∗ = f ◦ s ◦ g −1 : A → Y

is a cross section of the bundle (Y , q, A) (Fig. 4.5).

Proof It follows from the construction of s∗ . 

Corollary 4.1.37 Given two bundles (X , p, B) and (Y , q, B) over the same base
space B and a bundle B-morphism f : (X , p, B) → (Y , q, B), for every cross section
s of (X , p, B), there exists an induced cross section s∗ of (Y , q, B) satisfying the
property s∗ = f ◦ s.

Proof It follows from Proposition 4.1.36 by taking g = 1B in diagram in Fig .4.6.

Definition 4.1.38 (Trivial bundle) A topological space F is said to be the fiber
space of a bundle ξ = (X , p, B) if for every b ∈ B, the fiber p−1 (b) with subspace
topology inherited from the topology of X , is homeomorphic to F. The bundle ξ
is called a trivial bundle with fiber F if ξ is B-isomorphic to the product bundle
(B × F, q, B) with
q : B × F → B, (b, f ) → b

the projection. It is said to be locally trivial if it is locally a product.

Example 4.1.39 The normal bundle N (S n ) over S n is an 1-dimensional real trivial
bundle for every integer n ≥ 1 (see Definition 4.4.9).
 244 4 Topology of Fiber Bundles: General Theory of Bundles

4.2 Fiber Bundles

This section discusses basic concepts of fiber bundles which form a special class
of topological spaces widely used in different areas. A fiber bundle is a topological
space which looks locally a direct product of two topological spaces. The concept of
fiber bundles arose through the study of some problems in topology and geometry
of manifolds around 1930. It is locally the product of its base space and a discrete
space. The motivation of fiber bundles came through the study of the covering space

p : R → S 1 , t → e2πit

over S 1 . This study leads to introduce the concept of fiber bundles in this section.
For its general theory, see (Steenrod, 1951).

4.2.1 Introductory Concepts

This subsection introduces the concept of a fiber bundle and illustrates it by several
examples.

Definition 4.2.1 An ordered quadruple ξ = (X , p, B, F) is said to be a fiber bundle

if
(i)X is a topological space, called the total space of ξ.
(ii)B is a topological space, called the base space of ξ.
(iii)F is a topological space, called the fiber of ξ.
(iv) p : X → B is a continuous surjective map, called the fiber bundle projection
or simply projection of ξ.
(v) the inverse image p−1 (b) = Fb is homeomorphic to F for each point b ∈ B. The
subspace Fb is called the fiber over b.
(vi) B has an open covering {Ua }a∈A such that for each a ∈ A, there is a homeomor-
phism
ψa : Ua × F → p−1 (Ua )

with the composite map p ◦ ψa is the projection

pUa : Ua × F → Ua

to the first factor in the sense that p ◦ ψa = pUa , which means that

(p ◦ ψa )(x, y) = x, ∀ x ∈ Ua , ∀ y ∈ F.

It is sometimes symbolized as
4.2 Fiber Bundles 245

p
F → X −−−→ B

and it is called a covering of B if F is a discrete space. In that case, the space X is

called a covering space over B with p a covering projection. Then p−1 (b0 ) is discrete
for b0 ∈ B. The space F is homeomorphic to p−1 (b) for each b ∈ B.
Definition 4.2.2 A trivial fiber bundle is a fiber bundle ξ = (X , p, B, F) such that
its total space X is homeomorphic to the product space B × F, and the projection is
given by
p : B × F → B, (b, x) → b.

Remark 4.2.3 Since a fiber F of a fiber bundle ξ = (X , p, B, F) is a topological

space, the set Homeo(F) of all homeomorphisms of F forms a group G under
usual composition of functions. This is called the structure group G for the fiber
bundle ξ = (X , p, B, F), and it leads to the concept of G-bundles (see Sect. 4.8). In
particular, if B is paracompact, then the projection

p:X →B

is a fibration (see Chap. 5).

Proposition 4.2.4 For the exponential map

p : R → S 1 , t → e2πit ,

the ordered quadruple ξ = (R, p, S 1 , Z) is a fiber bundle.

Proof Consider the unit circle S 1 = {z ∈ C : |z| = 1}. To prove the proposition,
consider the open sets U1 = S 1 − {1} and U2 = S 1 − {−1}. Then p−1 (U1 ) = R − Z.
Consider the map

ψ1 : U1 × Z → p−1 (U1 ), (z, n) → n + (1/2πi) log z

where log z is the principal value of the logarithm function on C − {t ∈ R : t ≥ 0}.

Then its inverse map exists and is given by

ψ1−1 : p−1 (U1 ) → U1 × Z, t → (e2πit , [t]),

where [t] denotes the greatest integer contained in t for t ∈ p−1 (U1 ) = R − Z. Hence,
it follows that ψ1 is a homeomorphism such that the triangle in Fig. 4.7 is commu-
tative. Consider the map

ψ2 : U2 × Z → p−1 (U2 ), (z, n) = n + (1/2πi) log z,

where log z is the principal value of the logarithm function on C − {t ∈ R : t ≤ 0}

with its inverse map
 246 4 Topology of Fiber Bundles: General Theory of Bundles

Fig. 4.7 Homeomorphism

ψ1 commuting the triangle

Fig. 4.8 Homeomorphism

ψ2 commuting the triangle

1
ψ2−1 : p−1 (U2 ) → U2 × Z, t → (e2πit , [t + ]).
2
Proceeding as before it follows that ψ2 is a homeomorphism such that the triangle in
Fig. 4.8 is commutative. This proves that the ordered quadruple ξ = (R, p, S 1 , Z) is
a fiber bundle. 

Example 4.2.5 (More examples of fiber bundles)

(i) Given a connected space B and a point b0 ∈ B, if p : X → B is a covering pro-
jection, then
(X , p, B, p−1 (b0 ))

is a fiber bundle.
(ii) The n-dimensional torus
n
  
T = S1 × S1 × · · · × S1
n

and the map

p : Rn → T n , (t1 , t2 , . . . , tn ) → (e2πit1 , e2πit2 , . . . , e2πitn )

form a fiber bundle (Rn , p, T n , F) with fiber F which is the set of integer lattice
points in Rn .

(iii) ξ = (SO(n, R), p, S n−1 ) is a fiber bundle over S n−1 with fiber SO((n − 1), R).
To show it, consider SO((n − 1), R) as a subgroup of SO(n, R), which con-
sists of matrices A ∈ SO(n, R) that keep the vector en = (0, 0, . . . , 1) fixed. The
continuous map
p : SO(n, R) → S n−1 , A → Aen

is the projection map of ξ. Then p(AB) = A, ∀ B ∈ SO(n − 1, R) shows that

p−1 (x) homeomorphic to the group SO(n − 1, R) for every x ∈ S n−1 .
4.2 Fiber Bundles 247

4.2.2 Fiber Map and Mapping Fiber

This subsection studies fiber maps, induced maps, trivial fibering and mapping fiber
by imposing certain conditions on bundles studied in Sect. 4.1. A fiber bundle ξ =
(X , p, B) is sometime written simply as p : X → B. Given two fiberings p : X → B
and q : Y → A, a continuous map f : X → Y is said to be a fiber map if f sends
fibers into fibers. This concept is formulated in Definition 4.2.6.

Definition 4.2.6 Let p : X → B and q : Y → A be two fiberings. A continuous map

f :X →Y

is said to be a fiber map if for every point b ∈ B, there exists a point a ∈ A, such
that
f (p−1 (b)) ⊂ q−1 (a).

Given two fiberings, Definition 4.2.7 formulates the concept of an induced map on
their base spaces.

Definition 4.2.7 (Induced map) Let p : X → B and q : Y → A be two fiberings,

and f : X → Y be a fiber map. Then f induces a map

f∗ : B → A, b → (q ◦ f )(p−1 (b)).

For an arbitrary subset V ⊂ A,

f∗−1 (V ) = (p ◦ f −1 )(q−1 (V ))

asserts that f∗ is continuous if the map p is either open or closed. The map f∗ is called
the induced map of the given fiber map f .

Remark 4.2.8 Let p : X → B be a bundle. Then p is open, and hence, the induced
map f∗ = g : B → A is a continuous map by Definition 4.2.7 such that the diagram
in Fig. 4.9 is commutative in the sense that q ◦ f = g ◦ p .

Example 4.2.9 (i) (Trivial fibering) The fibering q : Y → A is said to be trivial

over the base space B, with the projection p : X → B as a fiber map, where
A = B = Y , q = 1B (identity map on B) and the induced map g = 1B .

Fig. 4.9 Rectangle

involving fiber map
248 4 Topology of Fiber Bundles: General Theory of Bundles

(ii) Let p : X → B be the trivial fibering over the base space B. Then every map
f : X → Y is a fiber map, where B = X and p = 1X .
Definition 4.2.10 (Mapping Fiber) Let f : (X , x0 ) → (Y , y0 ) be a base point pre-
serving continuous map. Its mapping fiber Ff is the pointed topological space

Ff = {(x, α) ∈ X × Y I : α(0) = y0 and α(1) = f (x)} ⊂ X × Y I

with product topology and base point (x0 , α0 ), where α0 is the constant path at y0 .
Remark 4.2.11 Definition 4.2.10 implies that the elements of the mapping fiber Ff
are precisely the ordered pairs (x, α) such that α is a path in Y from the point y0
to the point f (x). In particular, the loop space (Y , y0 ) based at the point y0 ∈ Y
consists of exactly of all ordered pairs of the form (x0 , α). The map

h : (Y , y0 ) → Ff : α → (x0 , α)

is clearly injective. The map

p : Ff → X , (x, α) → x

is the projection map.

4.3 Vector Bundles over Topological Spaces and Manifolds

A vector bundle is a special type of a fiber bundle for which every fiber admits a
vector space structure compatible on its neighboring fibers and the structure group
is a group of linear automorphism of the corresponding vector space. The concept
of vector bundles was born through the study of tangent vector fields to smooth
geometric objects such as manifolds, viz., spheres, projective spaces, etc. Roughly
speaking, a vector bundle over a topological B is a family of vector spaces contin-
uously parameterized by the base space B. This section studies vector bundles over
topological spaces in Subsect. 4.3.1 as well as over manifolds in Sect. 4.3.2 as base
spaces. The theory of vector bundles over a manifold provides a convenient language
to study many problems on the manifold.
For vector bundles, F denotes one of the fields R, C or the division ring H of
quaternionic numbers and Fn denotes the n-dimensional vector space over F.

4.3.1 Vector Bundles over Topological Spaces

This subsection communicates the concept of vector bundles over topological spaces
and studies it. In particular, vector bundles over manifolds are studied in Sect. 4.3.2.
4.3 Vector Bundles over Topological Spaces and Manifolds 249

Definition 4.3.1 (Vector bundle) A fiber bundle ξ = (X , p, B, Fn ) together with the

structure of an n-dimensional vector space Fn over F on every fiber Xb = p−1 (b) for
b ∈ B is said to be an n-dimensional F-vector bundle over a space B if
(i) There is an open covering {Ua : a ∈ A} of B.
(ii) For each a ∈ A, there is a homeomorphism ψa such that

ψa : Ui × Fn → p−1 (Ua ) : p ◦ ψa = pUa

and
(ψa |{b} × Fn ) : {b} × Fn → p−1 (b)

is an isomorphism of vector spaces over F for every b ∈ Ua .

Each ψa is called a coordinate transformation. An F-vector bundle ξ is said to be
a real, complex or quaternionic vector bundle if F = R, C or H. The vector bundle
ξ = (X , p, B) is also called an n-dimensional F-vector bundle over B.

Example 4.3.2 (i) If the fiber of a vector bundle ξ is Rn , then ξ is said to be finite
dimensional with dim ξ =n.
(ii) If the fiber of ξ is an infinite-dimensional Banach space and the structure group
is the group of invertible bounded operators of the Banach space, the bundle ξ
is said to be infinite dimensional.

Example 4.3.3 (Vector bundle γrn = (X , p, G r (Fn ), Fn ) ) Let G r (Fn ) be the Grass-
mann manifold of r-dimensional subspaces of Fn . Then the bundle γrn =
(X , p, G r (Fn ), Fn ), where

X = {(V , y) ∈ G r (Fn ) × Fn }

and
p : X → Fn , (V , y) → y is the orthogonal projection of y into V ,

is an n-dimensional F vector bundle.

Remark 4.3.4 Proposition 4.3.5 proves that the set of all cross sections of an F-
vector bundle ξ = (X , p, B) form a module MF over the ring R of F-vector-valued
continuous maps on B.

Proposition 4.3.5 Given an n-dimensional F-vector bundle ξ = (X , p, B) over B,

the set of all cross sections of ξ forms a module MF over the ring R of F-vector-valued
continuous maps on B.

Proof Given cross sections s, s of ξ and a continuous map f : B → F, the map

s + s : B → X , b → s(b) + s (b)
250 4 Topology of Fiber Bundles: General Theory of Bundles

is a cross section of ξ and the map

fs : B → X , b → f (b)s(b)

is also a cross section of ξ for all continuous maps f : B → F. Again, the map

θ : B → X , b → 0 ∈ p−1 (b)

is a cross section (zero cross section). Let ψ : U × Fn → p−1 (U ) be a local coordi-

nate of ξ over U . If
(i) ψ −1 (s(b)) = (b, g(b)), ∀ b ∈ B and
(ii) ψ −1 (s (b)) = (b, g (b)), ∀ b ∈ B
for some continuous maps g : U → Fn and g : U → Fn , then

(i)
ψ −1 ((s + s )(b)) = (b, g(b) + g (b)), ∀ b ∈ B;

(ii)
ψ −1 ((fs)(b)) = (b, f (b)g(b)), ∀ b ∈ B

and
(iii)
ψ −1 (0)(b) = (b, 0) ∀ b ∈ B.

Hence, it follows that s + s , fs and 0 are continuous maps such that they are all cross
sections. This proves that MF is a module over the ring R. 

Definition 4.3.6 Given two vector bundles ξ = (X , p, B) and η = (Y , q, A), a vector

bundle morphism
(f , g) : ξ → η

is a pair of continuous maps

f : X → Y and g : B → A

such that the diagram in the Fig. 4.10 is commutative in the sense that q ◦ f = g ◦ p
and
f |p−1 (b) : p−1 (b) → q−1 (g(b))

is a linear map (transformation) for every b ∈ B. In particular for B = A, the given

vector bundles become ξ = (X , p, B) and η = (Y , q, B). Then a morphism of the
form (f , 1B ) : ξ → η as shown in diagram Fig. 4.11 is called a B-morphism of
vector bundles f : ξ → η.
4.3 Vector Bundles over Topological Spaces and Manifolds 251

Fig. 4.10 Morphism of

vector bundles

Fig. 4.11 B-morphism

f : ξ → η of vector bundles

Remark 4.3.7 Let f : ξ → η is a B-morphism of vector bundles. Then it follows

from Definition 4.3.6 that
(i) q ◦ f = p and
(ii) f |p−1 (b) : p−1 (b) → q−1 (b) is a linear map for every b ∈ B.

Definition 4.3.8 (Isomorphism of vector bundles) Given two n-dimensional F-

vector bundles ξ = (X , p, B, Fn ) and η = (Y , q, A, Fn ), a vector bundle morphism

(f , g) : ξ → η

is said to be an isomorphism or an equivalence if there exists a homeomorphism

f : X → Y such that
f |p−1 (b) : p−1 (b) → q−1 (g(b))

is a linear isomorphism for every b ∈ B. In particular, if B = A and g = 1B , then the

given vector bundles become ξ = (X , p, B, Fn ) and η = (Y , q, B, Fn ) and

f :ξ→η

is called an equivalence over B, and in this case, ξ and η are said to be equivalent
over the base space B, denoted by ξ ∼
= η.
Remark 4.3.9 Any two equivalent vector bundles have the same dimension.
Proposition 4.3.10 Let VF be the set of all F-vector bundles over B. Equivalent
relation ∼
= of F-vector bundles over the same base space B is an equivalence relation
on VF .
Proof The reflexivity and transitivity of the given relation follow from Definition
4.3.8. To prove the symmetry of the relation, take any two bundles ξ = (X , p, B, Fn )
and η = (Y , q, B, Fn ) with f : ξ → η an equivalence. Then f : X → Y is an injective
continuous map. To show that f is open, it suffices to prove that f |p−1 (Uα ) is open,
for an open covering {Uα } of B. If it is expressed in terms of local coordinates
as (x, v) → (x, Ax v), where Ax is a nonsingular linear transformation depending
continuously on x then map has a continuous inverse, because matrix inversion is
continuous. This shows that f |p−1 (Uα ) is a homeomorphism. This implies that the f is
open. 
252 4 Topology of Fiber Bundles: General Theory of Bundles

Definition 4.3.11 (Trivial vector bundle) Let ξ = (X , p, B, Fn ) be a vector bundle.

If ξ is isomorphic to the product bundle B × Fn → B, then it is said to be a trivial
vector bundle.

Proposition 4.3.12 Given an n-dimensional product bundle ξ = (B × Fn , p, B, Fn )

and an m-dimensional product bundle η = (B × Fm , p, B, Fm ), the B-morphisms
f : ξ → η are expressed in the form

f (b, x) = (b, g(b, x)), i.e., f : B × Fn → B × Fm , (b, x) → (b, g(b, x)),

where the map

g : B × Fn → Fm , (b, x) → g(b, x) is linear in x.

Proof Let L(Fn , Fm ) be the vector space of all linear transformations T : Fn → Fm .

Then the two vector spaces L(Fn , Fm ) and Fmn are isomorphic and the map

g : B × Fn → Fm

is continuous iff the map

α : B → L(Fn , Fm ), b → g(b, −)

is continuous. Hence the proposition follows. 

Theorem 4.3.13 (i) Given a vector bundle ξ = (X , p, B) over B, the vector bundle
1∗B induced by the identity map 1B : B → B and the vector bundle ξ are B-
isomorphic.
(ii) Again, given a pair of continuous maps f and g

f g
A −−−→ B1 −−−→ B,

their two induced vector bundles f ∗ (g ∗ (ξ))and (g ◦ f )∗ (ξ) are A-isomorphic.

Proof By hypothesis, f : A → B1 and g : B1 → B are two continuous maps, and

g ◦ f is their composite map

f g
A −−−→ B1 −−−→ B,

which is also continuous. From the definition of the induced bundles, it follows that
if g ∗ (ξ) = (X1 , p1 , B1 ), then

X1 = {(b1 , x) ∈ B1 × X : g(b1 ) = p(x)}

4.3 Vector Bundles over Topological Spaces and Manifolds 253

and
p1 : X1 → B1 , (b1 , x) → b1 .

In particular, if g = 1B : B → B, then 1∗B (ξ) and ξ are B-isomorphic. Consider the

A-vector bundle induced by the map g ◦ f : A → B with its total space X2 = E((g ◦
f )∗ (ξ)) defined by

E((g ◦ f )∗ (ξ)) = {(a, x) ∈ A × X : (g ◦ f )(a) = p(x)}

and the total space of E(f ∗ (g ∗ (ξ))) is defined by

E(f ∗ (g ∗ (ξ))) = {(a, y) ∈ A × X1 : f (a) = p1 (y)}
= {(a, (b1 , x)) : g(b1 ) = p(x) and f (a) = p1 (y) = b1 }
= {(a, (f (a), x)) : (g ◦ f )(a) = p(x)}.
Hence, it follows that the two induced vector bundles f ∗ (g ∗ (ξ)) and (g ◦ f )∗ (ξ)
are A-isomorphic, because the map

α : (g ◦ f )∗ (ξ) → f ∗ (g ∗ (ξ)), (a, x) → (a, (f (a), x))

is a vector bundle isomorphism from (g ◦ f )∗ (ξ) to f ∗ (g ∗ (ξ)). 

The above discussion can be expressed in the language of category theory formu-
lated in Theorem 4.3.14.
Theorem 4.3.14 Let T op be the category of topological spaces and their continuous
maps and Set be the category of sets and set functions. If Vectn (B) denotes the set
of isomorphism classes of n-dimensional vector bundles over B, then

Vectn : T op → Set

is a contravariant functor.
Proof (i) The object function is defined as follows: Vectn assigns to every object
B ∈ T op, the vector bundle Vectn (B), which is the set of isomorphism classes
of n-dimensional vector bundles over B.
(ii) The morphism function is defined as follows: Let {ξ} denotes the isomorphism
class of a vector bundle ξ over B. Then for every morphism f : B1 → B in the
category T op, the function

f ∗ : Vectn (B) → Vectn (B1 ), {ξ} → {f ∗ (ξ)}

defines the morphism function.

(iii) Finally, using Theorem 4.3.13, it follows that

Vectn : T op → Set

is a contravariant functor.

 254 4 Topology of Fiber Bundles: General Theory of Bundles

4.3.2 Vector Bundles over Manifolds

The theory of vector bundles over a manifold M provides a convenient language to

describe many ideas in manifold. Given a manifold M of dimension n, and an open
covering {Ua : a ∈ A} of M , a vector bundle over M is constructed by gluing together
a family of product bundles {Ua × Rn } by an action of the group of GL(n, R) on Rn .

Definition 4.3.15 Let M be a smooth manifold of dimension n and N be a smooth

manifold of dimension m. Then a smooth map f : M → N
(i) is said to be an immersion at a point x ∈ M , if n ≤ m and rank of f at x is n. It
is said to be an immersion if it is an immersion at every point x ∈ M .
(ii) is said to be a submersion at a point x ∈ M , if n ≥ m and rank of f at x is m. It
is said to be a submersion if it is an submersion at every point x ∈ M .
(iii) is an embedding if it is an immersion and a homeomorphism onto its image
f (M ).
(iv) is a diffeomorphism, if n = m and f is a surjective embedding.
Example 4.3.16 The natural inclusion map i : S 2 → R3 is an immersion, since at
each point of S 2 , the map i has rank 2, the dimension of S 2 .
Definition 4.3.17 An n-dimensional R-vector bundle over a manifold B is an ordered
triple ξ = (X , p, B) consisting of a pair of smooth manifolds X and B connected by
a smooth surjective map p : X → B such that the following conditions are satisfied:
(a) (VB1 ) For every b ∈ B, the inverse image Xb = p−1 (b) is an n-dimensional
vector space over R.
(b) (VB2 ) For every b ∈ B, there exists an open nbd U of b and a diffeomorphism
ψ : U × Rn → p−1 (U ) with the properties that
(i) The triangle in Figure 4.12 is commutative, where

p1 : U × Rn → U, (y, v) → y

is the projection map onto the first factor.

(ii) For every y ∈ B, there exists a linear isomorphism

ψy : Rn → p−1 (y), v → ψ(y, v)

The vector bundle ξ = (X , p, B) is also denoted by p : X → B.

(i) The manifold X is called the total space and the manifold B is called the base
space and the map p is called the projection of the vector bundle ξ.
(ii) The inverse image Xb = p−1 (b) is called its fiber over b.
(iii) The defining condition VB2 is called the local triviality, the pair (U, ψ) is
called a vector bundle chart, U is called a trivializing open set and the family
{(U, ψ)} is called a vector bundle atlas of the vector bundle ξ.
4.3 Vector Bundles over Topological Spaces and Manifolds 255

Fig. 4.12 Commutative

diagram of vector bundles
involving ψ and projection
maps

(iv) The dimension of ξ is the dimension of its fiber.

Remark 4.3.18 Definition 4.3.17 asserts that a vector bundle ξ = (X , p, B) of

dimension n over a manifold B is a product bundle B × Rn obtained by gluing together
a family of product bundles {Ua × Rn }, where {Ua : a ∈ A} is an open covering of
B and is obtained by an action of the linear group GL(n, R) on Rn . The dimension
of a vector bundle ξ = (X , p, B, Rn ) is actually the dimension of its fibers (instead
of dimension of X or B). It is well defined, because, for any fiber Xb over b ∈ B, the
map

ψ : B → R, b → dim Xb

is locally constant, and it is constant on each component of B. If ψ is constant on the

entire B, then dim Xb is constant for all b ∈ B and this value is the dimension of ξ.

Example 4.3.19 Let V be an n-dimensional vector space over R. If B is a manifold,

then the projection p : B × V → B, (b, v) → b defines a vector bundle, called a
product bundle.

Proposition 4.3.20 The projection map

p : p−1 (U ) → U

given in Definition 4.3.17 is an submersion.

Proof It follows from commutativity of triangle in Figure 4.12 that p = p1 ◦ ψ −1

holds locally. Since ψ −1 is a diffeomorphism and p1 is an submersion and hence p
is a submersion (see Basic Topology, Volume 2). 

Example 4.3.21 Let ξ = (X , p, B) be a vector bundle over a manifold B and S be a

submanifold of B. Then the restricted bundle ξ|S is vector bundle p : p−1 (S) → S.

Example 4.3.22 (Whitney sum of vector bundles) Let ξ1 = (X1 , p1 , B) and ξ2 =

(X2 , p2 , B) be two vector bundles over B. Then their Whitney sum, denoted by ξ1 ⊕
ξ2 , is the vector bundle ξ = (X1 ⊕ X2 , p, B), where the total space is given by

X1 ⊕ X2 = {(x1 , x2 ) ∈ X1 × X2 : p1 (x1 ) = p2 (x2 )}

and the projection p is defined by

p : X1 ⊕ X2 → B, (x1 , x2 ) → p1 (x1 ) = p2 (x2 )

 256 4 Topology of Fiber Bundles: General Theory of Bundles

and its fiber over b ∈ B, is the direct sum p1−1 (b) ⊕ p2−1 (b).

There is a natural question: does there exist any similarity between the concepts
of a vector bundle and a manifold? Remark 4.3.23 gives its answer.

Remark 4.3.23 Similarity between a vector bundle and a manifold.

(i) A vector bundle is analogous to a manifold in the sense that both of them are built
up from elementary objects by gluing together by specified maps. For example,
the elementary objects for vector bundle are trivial bundles U × Rn , and the
gluing maps are morphisms of the form

ψ : U × Rn → U × Rn , (x, y) → (x, g(x)) for some g : U → GL(n, R).

(ii) On the other hand, the elementary objects for manifolds are open subsets of Rn ,
and the gluing maps are homeomorphisms.
(iii) More generally, a vector bundle over a topological space B consists of a family
{Xb }b∈B of disjoint vector spaces parametrized by the space B. The union X =

b∈B Xb of these vector spaces is a space X and the map

p : X → B, Xb → b

is continuous, and it is locally trivial in the sense that X looks locally like the
product U × Rn .
(iv) On the other hand, given a manifold M of dimension n, and an open covering
{Ui : i ∈ A}, a vector bundle over M is constructed by gluing together a family
of product bundles {Ui × Rn }i∈A by an action of the group of GL(n, R) on Rn .

4.4 Tangent and Normal Bundles over Manifolds

This section studies tangent and normal bundles over manifolds specially over S n .

4.4.1 Tangent Bundle of a Smooth Manifold

This subsection introduces the concept of tangent bundle over a smooth manifold and
studies it. If M is an n-dimensional
 smooth manifold and Tx (M ) is the tangent space
at x ∈ M , then T (M ) = {Tx (M )} (disjoint union) together with a projection p :
x∈M
T (M ) → M forms a tangent bundle of the manifold M . It is formalized in Definition
3.5.26. The tangent bundle T (M ) is a smooth manifold of dimension 2n by Theorem
4.4.3.
4.4 Tangent and Normal Bundles over Manifolds 257

Definition 4.4.1 (The tangent bundle of a manifold) The tangent bundle of a smooth
n manifold M is the bundle (T (M ), p, M ), where
(i) The total space T (M ) is the disjoint union of all tangent spaces Tx (M ) as x runs
over M . This is the set of all ordered pairs (x, v) such that x ∈ M and v ∈ Tx (M ).
(ii) The map
p : T (M ) → M , (x, v) → x

is called the projection map of the tangent bundle. Tx (M ) is called the tangent
space at x, and v ∈ Tx (M ) is called the tangent vector with initial point x.
More precisely, let M be an n-dimensional smooth manifold, Tx (M ) bethe vector
space of all tangent vectors to M at any point x ∈ M and T (M ) = {Tx (M )}
x∈M
(disjoint union). Then the set T (M ) is the disjoint union set of all tangent spaces
Tx (M ), and there is projection map

p : T (M ) → M , (x, v) → x.

The subspace Fx = p−1 (x) = Tx (M ) with topology inherited from M is called

the fiber over x. If M is a smooth manifold of dimension n, then T (M ) is a 2n-
dimensional manifold (see Theorem 4.4.3). A cross section of the tangent bundle
(T (M ), p, M ) is a smooth map s : M → T (M ) such that p ◦ s = 1M and hence
(p ◦ s)(x) = x, ∀ x ∈ M .

Remark 4.4.2 If M is a smooth manifold of dimension n, then T (M ) is the set of

all tangent vectors (x, v) at all points x ∈ M . If p assigns to every vector (x, v) its
initial point x, then Tx (M ) = p−1 (x) = Fx is the tangent plane at the point x, which
is a linear space. A cross section of the tangent bundle T (M ) of M is a vector field
over M .

Theorem 4.4.3 gives a unique differential structure on T (M ) and proves if

dim M = n, then dim T (M ) = 2n.

Theorem 4.4.3 Let M be a smooth manifold of dimension n, then its tangent bundle
T (M ) is a smooth manifold of dimension 2n.

Proof Let p : T (M ) → M , (x, v) → x be the projection map of the tangent bundle

T (M ) and (ψ, U ) be a chart of M . Then it determines a map

Tψ : p−1 (U ) → ψ(U ) × Rn ⊂ Rn × Rn , (x, v) → (ψ(x), d ψx (v))

Clearly, Tψ is a bijective map with its inverse

Tψ−1 : ψ(U ) × Rn → p−1 (U ), (a, w) → (b.d ψx−1 (w) : b = ψ −1 (a).

If (ψ, U ) and (φ, V ) be two compatible charts of M , then the map

258 4 Topology of Fiber Bundles: General Theory of Bundles

Tφ ◦ Tψ−1 : ψ(U ∩ V ) × Rn → φ(U ∩ V ) × Rn , (a, w)  → Tφ (b, dψ−1 (W )) = (φ(b), d φb ◦ d ψb−1 (w))

= (φ ◦ ψ −1 (a, d φb ◦ d ψb−1 (w)).

It asserts that Tφ ◦ Tψ−1 is a homeomorphism. Hence, T (M ) has a unique topology

such that
(i) Each Tψ is a homeomorphism.
(ii) T (M ) is second countable and Hausdorff.
(iii) The projection map p : T (M ) → M , (x, v) → x is continuous.
Again, since Tφ ◦ Tψ−1 is a diffeomorphism, the family of charts {(p−1 (U ), Tψ )} forms
a smooth atlas on M. Hence, it follows that T (M ) is a smooth manifold of dimension
2n.


4.4.2 Tangent Bundle over Sn

This section continues the study of tangent bundle over a very particular smooth
manifold S n .

Definition 4.4.4 (Tangent bundle over S n ) The tangent bundle over the n-sphere
S n in Rn+1 for n ≥ 1, is a fiber bundle

ξ = T (S n ) = (X , p, S n , Rn ),

where X = {(x, y) ∈ Rn+1 × Rn+1 : ||x|| = 1 and x, y = 0} and the projection map

p : X → S n , (x, y) → x.

Then ξ is an n-dimensional real vector bundle. To show the local triviality condition
VB2 , take the open covering {Ui } of S n defined by

Ui = {x = (x1 , x2 , . . . , xi , . . . , xn+1 ) ∈ Rn+1 : ||x|| = 1, xi = 0, 1 ≤ i ≤ n + 1}.

Then Ui ⊂ S n is the open set Ui = {x ∈ Rn+1 : ||x|| = 1, xi = 0, 1 ≤ i ≤ n + 1} and

it is not connected, because it has two components corresponding to xi > 0 and
xi < 0. Define the map

ψi : Ui × Rn → p−1 (Ui ), (x, y) → (x, fi (y) − x, fi (y)x),

 4.4 Tangent and Normal Bundles over Manifolds 259

where

fi : Rn → Rn+1 , y = (y1 , y2 , . . . , yn ) → (y1 , y2 , . . . , yi−1 , 0, yi , . . . , yn ).

This asserts that p ◦ ψ = p1 and every ψi is a linear map on each fiber. The point
x = (x1 , x2 , . . . , xi , . . . , xn+1 ) ∈ Rn+1 lies outside Im fi , since xi = 0 and hence if
y = 0, then
fi (y) − x, fi (y)x = 0.

This implies that ψi is injective and it is an isomorphism on each fiber. It follows that
ψi is also a diffeomorphism such that

p ◦ ψi = pUi .

Remark 4.4.5 The tangent bundle over the n-sphere S n in the Euclidean (n + 1)-
space Rn+1 is the subbundle ξ = (T (S n ), p, S n ) of the product bundle (S n ×
Rn+1 , p, S n ), whose total space T (S n ) is defined by

T (S n ) = {(b, x) ∈ S n × Rn+1 : b, x = 0},

and the projection map p is defined by

p : T (S n ) → S n , (b, x) → b.

(i) An element of the total space T (S n ) is said to be a tangent vector to S n at the

point b ∈ S n .
(ii) The fiber p−1 (b) ⊂ T (S n ) is a vector space of dimension n.
(iii) A cross section of the tangent bundle ξ over S n is said to be a tangent vector field
(or simply vector field) over S n .

Remark 4.4.6 Consider the three-sphere S3 = {(z, w) ∈ C2 : |z|2 + |w|2 = 1} as

the particular case of S n , when n = 3. The two-sphere S 2 may be considered as
the extended complex plane. Then the map

p : S 3 → S 2 , (z, w) → z/w

is well defined. It has a fiber space structure (see Remark 4.4.5). This asserts that its
fiber space structure is related to the properties of complex numbers.

4.4.3 Normal Bundle over a Manifold

This subsection studies normal bundles over manifolds

260 4 Topology of Fiber Bundles: General Theory of Bundles

Definition 4.4.7 Given an n-dimensional submanifold M of Rm , the normal space

Nx (M ) of M , at x ∈ M , defined by

Nx (M ) = {(x, v) ∈ M × Rm : v ⊥ Tx (M )}.

Its normal bundle (N (M ), p, M ) is the vector bundle,(N (M ), p, M ), where the

total space 
N (M ) = {Nx (M )}
x∈M

(disjoint union) and the projection

p : N (M ) → M , (x, v) → x.

Remark 4.4.8 Given an n-dimensional submanifold M of Rm , its normal bundle

(N (M ), p, M ) is a vector bundle with fiber dimension m − n. Moreover,
(i) N (M ) is a manifold of dimension m.
(ii) The projection
p : N (M ) → M , (x, v) → x

is a submersion.

4.4.4 Normal Bundle over Sn

This section continues the study of normal bundle over a smooth manifold M by
considering in particular, M = S n .
Definition 4.4.9 (Normal bundle over S n ) The normal bundle N (S n ) over S n for
every integer n ≥ 1 is the fiber bundle N ξ = (X , q, S n , R1 ), where

X = {(x, y) ∈ S n × Rn+1 : y = rx, for some r ∈ R}

and
q : X → S n , (x, y) → x.

The maps
ψ : S n × R1 → X , (x, r) → (x, rx)

and
φ : X → S n × R1 , (x, y) → (x, x, y)

are homeomorphisms such that ψ is a homeomorphism with its inverse φ. This asserts
that ξ = N (S n ) is an 1-dimensional real trivial bundle.
4.4 Tangent and Normal Bundles over Manifolds 261

Remark 4.4.10 The normal over S n is the subbundle (N (S n ), q, S n ) of the product

bundle (S n × Rn+1 , p, S n ) whose total space N (S n ) is defined by

N (S n ) = {(b, x) ∈ S n × Rn+1 : x = tb for some t ∈ R}

and the projection q is defined by

q : N (S n ) → S n , (b, x) → b.

(i) An element of N (S n ) is called a normal vector to S n at the point b ∈ S n .

(ii) The fiber q−1 (b) ⊂ T (S n ) is a vector space of dimension 1. A cross section of
the normal bundle ξN over S n is called a normal vector field on S n .

4.4.5 Orthonormal K-Frames over Spheres

This subsection describes orthonormal k-frames over spheres as a subbundle (X , p, B)

of the product bundle (S n × (S n )k , p, S n ).

Example 4.4.11 (Orthonormal k-frames over spheres) The bundle ξok of orthonor-
mal k-frames over S n for k ≤ n is the subbundle (X , p, B) of the product bundle
(S n × (S n )k , p, S n ), where total space X is given by

X = {(b, x1 , x2 , . . . , xk ) ∈ S n × (S n )k : b, xi  = 0 and xi , xj  = δij , 1 ≤ i, j ≤ k},

where the Kronecker δij is defined by


1, if i = j
δij =
0, otherwise

and the projection p is given by

p : X → S n , (b, x1 , x2 , . . . , xk ) → b.

(i) An element (b, x1 , x2 , . . . , xk ) of the total space X of an orthonormal k-frames

over S n for k ≤ n is an orthonormal system of k-tangent vectors to S n at b ∈ S n .
(ii) A cross section of ξok over S n is said to be a field of k-frames.
262 4 Topology of Fiber Bundles: General Theory of Bundles

4.4.6 Canonical Vector Bundle γ n

This subsection describes real, complex and quaternionic canonical vector bundle
γ n over real, complex and quaternionic Grassman manifolds G r (Rn ), G r (Cn ) and
G r (Hn ).

Example 4.4.12 (Canonical vector bundle γ n ) The real Grassman manifold G r (Rn )
is the set of all r-dimensional vector subspaces (or r-planes through the origin)
of Rn . Then G 1 (Rn ) = RP n−1 . The canonical r-dimensional vector bundle γrn =
(X , p, G r (Rn )) over the Grassmann manifold G r (Rn ) of r-frames in Rn (r ≤ n) is
the subbundle of the product bundle (G r (Rn ) × Rn , p, G r (Rn )) with the total space
X consisting of the subspace of pairs (V , x) ∈ G r (Rn ) × Rn with x ∈ Rn and the
orthogonal complement vector bundle of γrn , denoted by γr∗ is the subbundle of
n

∗n
(G r (R ) × R , p, G r (R )) defined by γr = (Y , p, G r (R )), where Y = {(V , x) ∈
n n n n

G r (Rn ) × Rn : V , x = 0 ( i.e., x ⊥ V )}. Then for r = 1, in particular, γ1n on RP n−1

= G 1 (Rn ), is one dimensional and is called the canonical line bundle.

By natural inclusion

G r (Rn ) ⊂ G r (Rn+1 ) ⊂ G r (Rn+2 ) ⊂ · · · .


The topological space G r (R∞ ) is defined by G r (R∞ ) = G r (Rn ) endowed with
r≤n
weak topology.
Similarly, G r (Cn ), G r (Hn ) and G r (C∞ ) and G r (H∞ ) are defined, where H is the
division ring of quaternions. In general, for F = R, C or H, the canonical vector
bundle over G r (F∞ ) is written as γr . Since G r (Rn ) ⊂ G r (Rn+t ) for integers t ≥ 1,
the vector bundle γrn can be viewed as

γrn = γrn+1 (G r (Rn ).

Analogously, the complex and quaternionic canonical vector bundles over G r (Cn )
(or G r (Hn )) are defined.

4.5 Covering Spaces and Covering Homomorphism

This section introduces the concepts of covering spaces and covering homomor-
phisms which are basic concepts in topology of fiber bundles. For example, they
are used to compute fundamental groups of some important topological spaces (see
Chap. 5), and the algebraic features of these groups are expressed in the geometric
language of the corresponding covering spaces. All topological spaces in this section
are assumed to be Hausdorff.
4.5 Covering Spaces and Covering Homomorphism 263

4.5.1 Covering Spaces and Covering Projections

This subsection presents covering spaces, which form another class of important
bundles. The covering projections and covering spaces play a key role in homotopy
theory.

Definition 4.5.1 A topological space X is called locally path connected if for every
point x ∈ X and every nbd U of x, there is an open set V such that
(i) x ∈ V ⊂ U .
(ii) Every pair of elements in V can be joined by a path in U.
On the other hand, the space X is called semi locally path connected if for every
point x ∈ X , there is an open nbd U of x such that every closed path in U based at
the point at x is null homotopic in the sense that it is homotopic to a constant map in
X.

Example 4.5.2 Every path-connected space is not locally path connected. For exam-
ple, consider the topological space X ⊂ R2 , which is the union of the graph
{(x, sin 1/x) : x ∈ (0, 1]} and a path connecting the points (1, 0) and (0, 1) with
subspace topology of R2 . Then X is path connected, but it is not locally path con-
nected.

Proposition 4.5.3 characterizes locally path connectedness in terms of path com-

ponents.

Proposition 4.5.3 Let X be a topological space. Then it is locally path connected

iff every path component of any open subset of X is open.

Proof It follows from Definition 4.5.1. 

Definition 4.5.4 Let X and B be two path connected, locally path-connected spaces
and the map p : X → B be continuous. Then p is said to be a covering projection
(map) if
(i) p is a surjection;
(ii) for each point b ∈ B, there is an open set U in the space B containing the point b
such that p−1 (U ) is a disjoint union of open sets in X , each of which is mapped
homeomorphically by p onto U called sheets. Then the ordered pair (X , p) is
called a covering space of B. Every open set in the disjoint union of open sets
in X is sometimes called a sheet or an admissible open set for the covering
space (X , p) and U is called evenly covered by p.

Remark 4.5.5 Definition 4.5.4 says that given a continuous surjective map p : X →
B, an open set U of B is evenly covered by p if p−1 (U ) is a union of disjoint open
sets Vi , called sheets such that p|Vi : Vi → U is a homeomorphism for each i and the
open set U is an admissible open set in B.
264 4 Topology of Fiber Bundles: General Theory of Bundles

Definition 4.5.6 (Trivial covering projection) A covering projection p : X → B is

called trivial, if it is isomorphic to the projection p : B × F → B from the product
space B × F onto B, where F is any set endowed with discrete topology, i.e.,all of
its points are closed.
Example 4.5.7 Let S 1 be the unit circle in the complex plane. Then the map

p : R → S 1 , x → e2πix

is a covering projection and (R, p) is a covering space. Consider the open sets U1 =
S 1 − {1} and U2 = S 1 − {−1}. They are evenly covered by p. Because, p−1 (U2 ) =
 1 1
(n − , n + ). This asserts that the sheets are open intervals. Similarly, U1 is
n∈Z
2 2
also evenly covered by p.
Example 4.5.8 For every nonzero real number t, the map

p : R → S 1 , x → eitx

is a covering projection and (R, p) is a covering space of S 1 .

Example 4.5.9 For every n ∈ N, the map

pn : S 1 → S 1 , z → z n

is a covering map with (S 1 , pn ) a covering space of S 1 .

Example 4.5.10 There are many noncovering spaces. For example, let B = S 1 and
X be a finite open spiral over S 1 as shown in Fig. 4.13. Let p : X → B be the map ,
which projects every point x ∈ X to the point on the circle S 1 directly below it. Then
(X , p) is not a covering space, because the condition (ii) of Definition 4.5.4 is not
satisfied in this case.

Proposition 4.5.11 Let X̃ be a locally path-connected space and p : X̃ → X be a

projection map. Then p is an open mapping
Proof Let V be an open set in X̃ and x ∈ p(V ) and x̃ ∈ p−1 (x) and U be an admis-
sible nbd of x. This shows that x̃ is a point of V such that p(x̃) = x. Let W be the
component of p−1 (U ) which contains the point x̃. Since by hypothesis, X̃ is locally
path connected, it follows that W is open in X̃ by Proposition 4.5.3 . Clearly, p maps
the open set W ∩ V to the open subset p(W ∩ V ) in X , because p maps W home-
omorphically onto U. Then x ∈ p(W ∩ V ) ⊂ p(V ). Since x is an arbitrary point of
p(V ), it follows that p(V ) is a union of open sets, which is an open set and hence p
is an open map. 
Theorem 4.5.12 is an interesting result without any assumption of path connectedness
or local path connectedness property of Y .
4.5 Covering Spaces and Covering Homomorphism 265

Fig. 4.13 Finite spiral X

with projection p : X → S 1

Fig. 4.14 Triangular

diagram involving f , g and p

Theorem 4.5.12 Given a covering space (X̃ , p) of X and a topological space Y , if

f , g : Y → X̃ are two continuous maps such that p ◦ f = p ◦ g, as shown in Fig. 4.14,
then the set A = {y ∈ Y : f (y) = g(y)} is both open and closed in Y .

Proof We first claim that A is open in Y . Given any point y ∈ A and an evenly covered
nbd U of (p ◦ f )(y), the path component V of p−1 (U ) in which f (y) lies is an open set
in X̃ . This shows that f −1 (V ) and g −1 (V ) are both open in Y . Since by hypothesis,
f (y) ∈ V and f (y) = g(y), then y ∈ f −1 (V ) ∩ g −1 (V ). Let z ∈ f −1 (V ) ∩ g −1 (V ).
Then f (z), g(z) ∈ V and (p ◦ f )(z) = (p ◦ g)(z). Since p maps V homeomorphi-
cally onto U , it follows that f (z) = g(z) and hence z ∈ A. Thus, it follows that A is
an open set.

We also claim that A is closed in Y . If possible suppose A is not closed in Y .

Let z be a limit point of A which is not in A. Then there is an elementary nbd U
of (p ◦ f )(z) = (p ◦ g)(z) such that the points f (z) and g(z) must lie in different
path components V1 and V2 , say, of p−1 (U ). The open set f −1 (V1 ) ∩ g −1 (V2 ) in Y
contains a point y ∈ A. But this is not possible, because V1 ∩ V2 = ∅ and f (y) =
g(y) ∈ V1 ∩ V2 , This asserts that all limit points of A are in A and hence A is closed.


Corollary 4.5.13 and Corollary 4.5.14 prove the uniqueness of the lifting of a
continuous map. Its independent proof is given in Proposition 4.5.25.
266 4 Topology of Fiber Bundles: General Theory of Bundles

Corollary 4.5.13 Given a covering space (X̃ , p) of X , a connected space Y and

two continuous maps f , g : Y → X̃ such that p ◦ f = p ◦ g, if f (y) = g(y) at some
point y ∈ Y , then f = g.
Proof Since by hypothesis, Y is a connected space, the only sets that are both open
and closed in Y are Y and ∅. Hence, it follows by using Theorem 4.5.12 that either
A = Y or A = ∅. This shows that either f (y) = g(y) at every y ∈ Y or f (y) = g(y) at
every y ∈ Y . Since by hypothesis f (y) = g(y) at some y ∈ Y , it follows that A = ∅
and hence A = Y . This implies that f (y) = g(y), ∀ y ∈ Y . It proves that f = g. 
Corollary 4.5.14 proves the uniqueness of lifting. Its alternative form is given in
Proposition 4.5.25.
Corollary 4.5.14 (Uniqueness of lifting) Let p : X̃ → X be a covering space, Ỹ
be a path-connected space and f : Ỹ → X be a continuous map having two liftings
f˜1 , f˜2 : Ỹ → X̃ . If f˜1 and f˜2 agree at some point of Ỹ , then f˜1 = f˜2 .
Proof By hypothesis, p : X̃ → X is a covering space, Ỹ be a connected space and
f : Ỹ → X is a continuous map having two liftings f˜1 , f˜2 : Ỹ → X̃ . Then p ◦ f˜1 =
p ◦ f˜2 . Again, since, f˜1 (y) = f˜2 (y) at some point y ∈ Ỹ , by hypothesis, it follows from
Corollary 4.5.13 that f˜1 = f˜2 .. 
Example 4.5.15 (i) Given a covering map p : X → B, if B is any subspace of B,
then its restriction X = p−1 (B ) → B is a covering map.
(ii) If p : X → B is a covering map of an open subset B in the Euclidean plane R2 ,
then X can be endowed with the structure of a differentiable surface with p a
local diffeomorphism. For a connected open set U in B, the components C of
p−1 U over which the covering is trivial along with the homeomorphisms of U
onto C obtained by the inverse of p determine charts covering X .
(iii) Every covering of a manifold has a natural manifold structure.
(iv) If B is connected, all the fibers of the covering space p : X → B have the same
cardinality.
Definition 4.5.16 Let p : X → B be covering map such that every fiber p−1 (b) has a
finite cardinality n. Then the covering is said to be n-sheeted. One fails to distinguish
its different n-sheets unless the covering is trivial.
Proposition 4.5.17 (Product of covering spaces) Given two covering spaces (X̃1 , p1 )
and (X̃2 , p2 ), the map

p1 × p2 : X̃1 × X̃1 , → X1 × X2 , (x̃1 , x̃2 ) → (p1 (x̃1 ), p2 (x̃2 ))

is a covering map with (X̃1 × X̃2 , p1 × p2 ) a covering space of X1 × X2 .

Proof Let (x1 , x2 ) ∈ X1 × X2 and U1 be an open nbd of x1 evenly covered by p1 and
U2 be an open nbd of x2 evenly covered by p2 . Then U1 × U2 is a nbd of (x1 , x2 ) in
X1 × X2 evenly covered by p1 × p2 . Hence it follows that p1 × p2 is a covering map
with (X̃1 × X̃2 , p1 × p2 ) the corresponding covering space of X1 × X2 . 
4.5 Covering Spaces and Covering Homomorphism 267

Example 4.5.18 (i) The map

p : R2 → S 1 × S 1 , (x1 , x2 ) → (e2πix1 , e2πix2 )

from the Euclidean plane to the torus is a covering projection with (R2 , p) a
covering space of S 1 × S 1 . Because, given a point (z1 , z2 ) ∈ S 1 × S 1 , a small
rectangle U constructed by the product of two open arcs in S 1 containing z1 and
z2 , respectively, is an admissible nbd having its inverse image consisting of a
countably infinite family of open rectangles in the Euclidean plane R2 .
(ii) The exponential map
p : R → S 1 , x → e2πix

defines the projection map

(p, p) : R × R → S 1 × S 1

in the natural way. For every positive integer n, by induction on n, the product
map
(p, p, . . . , p) = pn : Rn → T n

is a covering projection with the corresponding covering space (Rn , pn ) over

n
the n-dimensional torus T n =  S 1 .
1

4.5.2 Automorphism Group of Covering Spaces and Lifting

Problem

This subsection starts with the concepts of covering homomorphisms and deck trans-
formations.
Definition 4.5.19 (Covering homomorphism) Given two covering spaces (X̃ , p)
and (Ỹ , q) over the same base space X , a covering homomorphism h from (X̃ , p) to
(Ỹ , q) is a continuous map h : X̃ → Ỹ such that the diagram in Fig. 4.15 is commuta-
tive in the sense that q ◦ h = p. The homomorphism h is said to be an isomorphism
if h is a homeomorphism. If there exists an isomorphism between the covering spaces
(X̃ , p) to (Ỹ , q), then they are called isomorphic or equivalent covering spaces. If
there exists no isomorphism between them, they are called distinct covering spaces.
In particular, an isomorphism of a covering space onto itself is called an automor-
phism or a deck transformation.

Example 4.5.20 Every homomorphism of path connected covering spaces is a cov-

ering projection. Because, if h : X̃ → Ỹ is a homomorphism of covering spaces, then
(X̃ , h) is a covering space of Ỹ .
268 4 Topology of Fiber Bundles: General Theory of Bundles

Fig. 4.15 Covering

homomorphism

Proposition 4.5.21 The collection of all covering spaces of a base space X and
their homomorphisms form a category, denoted by Cov.

Proof Take the collection of all covering spaces of a base space X as the class of
objects and their homomorphisms as the class of morphisms. Let (X̃ , p) be a covering
space of X . Then 1X̃ : X̃ → X̃ is a covering homomorphism. For the covering spaces
(X̃ , p), (Ỹ , q) and (Z̃, r) of X and covering homomorphisms h : X̃ → Ỹ , g : Ỹ →
Z̃, their composite g ◦ h : X → Z is also a covering homomorphism from (X̃ , p) to
(Z̃, r). 

Example 4.5.22 Isomorphisms in this category Cov are precisely the isomorphisms
of covering spaces as defined above.

Proposition 4.5.23 discusses the algebraic structure of the set Aut (X̃ /X ) of all
automorphisms of the covering space X .

Proposition 4.5.23 Let Aut (X̃ /X ) be the set of all automorphisms of the covering
spaces of X . Then (Aut (X̃ /X ), ◦) admits a group structure under usual composition
of maps.

Proof The identity map 1X̃ : X̃ → X̃ is itself an automorphism and the inverse of an
automorphism is again an automorphism. Hence under usual composition of maps,
Aut (X̃ /X ) is a group. 

Definition 4.5.24 The group (Aut (X̃ /X ), ◦) defined in Proposition 4.5.23 is called
the automorphism group of covering spaces of X , and their elements are called
the deck or covering transformations.

Proposition 4.5.25 proves the uniqueness of liftings. Its alternative form is given
in Corollary 4.5.14.

Proposition 4.5.25 (Uniqueness of liftings) Let (X̃ , p) and (Ỹ , q) be two covering
spaces of the same base space X , where X̃ is connected. If f , g : X̃ → Ỹ are two
covering homomorphisms such that f (x̃0 ) = g(x̃0 ) for some x̃0 ∈ X̃ , then f = g.

Proof Let (X̃ , p) and (Ỹ , q) be two covering spaces of the same base space X .
Suppose that f , g : X̃ → Ỹ are two covering homomorphisms such that f (x̃0 ) =
g(x̃0 ) for some x̃0 ∈ X̃ . Then each of f and g is considered as liftings of the map
p : X̃ → X with respect to the covering projection q : Ỹ → X . Since by hypothesis,
X̃ is connected and f , g both agree at the same point x̃0 ∈ X̃ , g = h. This proves the
uniqueness of the lifting. 
4.6 Construction of Fiber Bundles and Their Local Cross Sections 269

Fig. 4.16 Commutativity of

the triangle for the covering
space (X , p)

Theorem 4.5.26 Let (X , p) be a covering space of B.

(i) If h ∈ Cov(X /B) = Aut (X /B) and h = 1X , then h has no fixed point.
(ii) If h, g ∈ Aut (X /B) and ∃ x ∈ X with h(x) = g(x), then h = g.

Proof (i) Let x ∈ X be a point such that h(x) = x. Then x is a fixed point of h.
For b = p(x), consider the commutative triangle shown in Fig. 4.16 Since both
h and 1X complete the diagram in Fig. 4.16, it follows that h = 1X . This implies
a contradiction. This contradiction proves that h has no fixed point.
(ii) The map h−1 g ∈ Aut (X /B) has a fixed point, namely x and hence by (i) it follows
that
h−1 g = 1X ,

which proves that h = g.



4.6 Construction of Fiber Bundles and Their Local Cross

Sections

This section describes construction of fiber bundle by continuous action by a discrete

topological group.

4.6.1 Construction of Fiber Bundle by Continuous Action

This subsection constructs a fiber bundle by using a continuous action

X × G → X , (x, g) → xg

of a discrete topological group G on a topological space X .

Let G be a topological group and X be a topological space. Then G is said to
act on X (from the right) properly discontinuously if for any point x ∈ X and any
sequence {gn } of distinct points in G, the sequence {xgn } of points in X does not
converge to any point of X in the sense that each orbit xG is a closed discrete subset
of X . This is formulated in Definition 4.6.1.
270 4 Topology of Fiber Bundles: General Theory of Bundles

Definition 4.6.1 Let G be a discrete topological group with the identity element
e and X be a topological space. Then G is said to act (from the right) properly
discontinuously on X if
PD(i) For every point x ∈ X , there is a nbd Ux of x in X with the property:
Ux g ∩ Ux = ∅ implies g = e.
PD(ii) For any two elements x, y ∈ X , y ∈
/ xG, there are nbds Vx and Vy of x and
y, respectively, such that

Vx g ∩ Vy = ∅, ∀ g ∈ G.

The topological space X /G = X mod G, topologized by the identification map

p : X → X mod G, x → xG

is the orbit space obtained by the action of G on X .

Remark 4.6.2 If an action of topological group G on a topological space X is prop-

erly discontinuous, then it is evenly covered by p in the sense that any point in X has
a nbd U such that U g ∩ U k = ∅ for every pair of distinct points g, k ∈ G.

Example 4.6.3 Every free action of a finite group on a Hausdorff space is properly
discontinuous.

Proposition 4.6.4 The projection p : X → X mod G, x → xG is a covering map,

where X is a Hausdorff space.

Proof By hypothesis, X is assumed to be Hausdorff. Hence the Hausdorff property

of X mod G follows from the condition PD(ii) that distinct points x, y ∈ X , which
are not equivalent, have nbds U and V of x and y respectively such that

U g ∩ Vk = ∅, ∀ g, k ∈ G,

since, U g ∩ Vk = (k −1 g)U ∩ V . Hence it follows that the projection

p : X → X mod G, x → xG

is a covering map. 

Proposition 4.6.5 prescribes an important method in which fiber bundles arise in

a natural way.

Proposition 4.6.5 Let G be a discrete topological group acting properly discontin-

uously on a topological space X . Then (X , p, X mod G, G) forms a fiber bundle
over X mod G.

Proof For an element y ∈ X mod G,

(i) There exists an element x ∈ X such that p(x) = y.
4.6 Construction of Fiber Bundles and Their Local Cross Sections 271

(ii) There exists a nbd Ux of x in X by the condition PD(i).

If Vy = p(Ux ), then

p−1 (Vy ) = Ux g
g∈G

asserts that Vy is open and y ∈ Vy . Again, pUx : Ux → Vy is a homeomorphism. Then

the map
ψy : Vy × G → p−1 (Vy ), (z, g) → ((p|Ux )−1 (z))g,

is a homeomorphism with its inverse

ψy−1 : p−1 (Vy ) → Vy × G, (z g) → (p(z ), g).

This implies that ψy is a homeomorphism such that p ◦ ψy = pVy . This asserts that the
open sets {Vy } form a covering of the space X mod G. It proves that (X , p, X mod
G, G) forms a fiber bundle over X mod G. 

For convenience of future study, Definition 4.6.1 is reformulated in Defini-

tion 4.6.6.
Definition 4.6.6 Let G be the group of homeomorphisms of a topological space X .
Then G is said
(i) To act (from the right) properly discontinuously on X if for any x ∈ X , there
is a nbd, Ux of x in X such that for g, g ∈ G if Ux g intersects Ux g , then g = g .
(ii) To act without fixed points if the only element of G having fixed point is its
identity element 1X : X → X .

4.6.2 Local Cross Sections of Fiber Bundles

This subsection studies the local cross sections of fiber bundles ξ.

Definition 4.6.7 Given a bundle ξ = (X , p, B), the map p : X → B is said to have

a local cross section s at a point b ∈ B if there is a nbd Ub in B and a continuous
map
s : Ub → X

is such that p ◦ s = 1Ub .

Example 4.6.8 There are bundles ξ having no cross section. For example, consider
the bundle ξ = (D2 , p, S 1 ), where p : D2 → S 1 is the map letting p(x) to be the
point of S 1 , where the ray in R2 starting from origin and passing through trough the
point x. If ξ has a cross section s : S 1 → D2 , then p ◦ s = 1S 1 would imply that the
272 4 Topology of Fiber Bundles: General Theory of Bundles

map s : S 1 → D2 , is injective. Again since s : S 1 → D2 is continuous and injective,

it contradicts the Borsuk-Ulam Theorem (see Chap. 2) which asserts that every
continuous map f : S 1 → D2 maps at least one pair of antipodal points of S 1 to the
same point.

Example 4.6.9 The bundle defined in Example 4.6.8 is not a fiber bundle, because
every fiber bundle admits a cross section.

Proposition 4.6.10 gives a sufficient condition for the existence of local cross
section of a bundle at a point.

Proposition 4.6.10 Given a topological group G and a closed subgroup H of G,

the projection map
p : G → G/H , g → gH

admits a local cross section at every point of G/H .

Proof Consider the quotient space G/H topolozied by quotient topology. Define an
action σ of G on the space G/H , formulated by

σ : G × G/H → G/H , (g, g H ) → (gg )H .

For the proof of the given proposition, it is sufficient to show that the continuous
map
p : G → G/H , g → gH

admits a local cross section at the coset H . Because, if (U, σ) is a local cross section
for p at the point H , then for any other point gH of G/H , the set gU is a nbd of gH
and the map
σg : gU → G, g H → g(σ(g −1 g H ))

is continuous such that

p ◦ σg = 1gU .

This proves that σg is a local cross section. 

Theorem 4.6.11 Given a topological group G and a closed subgroup K of G, if the

map
p : G → G/K, g → gK

admits a local cross section at K, then for every closed subgroup A ⊂ K, the natural
projection
q : G/A → G/K, gA → gK,

is a fiber bundle having its fiber K/A.

4.7 Hopf Fibering of Spheres 273

Proof By hypothesis, G is a topological group and K is a closed subgroup of G.

Then by using Proposition 4.6.10, it follows that the projection

p : G → G/K, g → gK

has a local cross section at every point of G/K. Take any point x ∈ G/K and a local
cross section (U, α) of p at x, define a map

φ : U × K/A → G/A, (y, hA) → α(y)hA.

Then φ is a continuous map having the property that for all y ∈ U and for all k ∈ K,

(q ◦ φ)(y, hA) = q(α(y)hA) = α(y)hA = α(y)K = p(α(y)) = y = 1U (y).

Again, the map

ψ : q−1 (U ) → U × KA, gA → (gK, α(gK)−1 gA), ∀ gA ∈ q−1 (U )

is continuous. This implies that

ψ ◦ φ = 1d and φ ◦ ψ = 1d ,

where 1d is the identity maps on U × K/A and q−1 (U ) in the respective cases. Hence
the theorem follows. 

Example 4.6.12 Fiber bundles obtained from the decomposition of compact Lie
groups modulo their closed subgroups provide a rich supply of fiber bundles, and it
serves as a valuable source of examples of fiber bundles, see (Samelson, 1952) and
(Borel, 1955).

4.7 Hopf Fibering of Spheres

This section studies the three fiberings of spheres:

p : S 2n−1 → S n , for n = 2, 4, 8,

the early examples of bundles spaces, of which, the map p : S 3 → S 2 of the 3-sphere
on the 2-sphere defined by Hopf in 1935 is known as a Hopf map, is the simplest.
274 4 Topology of Fiber Bundles: General Theory of Bundles

4.7.1 Hopf Map p : S3 → S2

This subsection is devoted to study the Hopf map p : S 3 → S 2 .

Theorem 4.7.1 The Hopf fibering (S 3 , p, S 2 ) has fibers a family of great circles.

Proof To construct the Hopf fibering (S 3 , p, S 2 ), consider S 3 and S 2 defined as

follows:
S 3 = {(z1 , z2 ) ∈ C2 : z1 z 1 + z2 z 2 = |z1 |2 + |z2 |2 = 1}

and the 2-sphere S 2 represented as the complex projective line consisting of pairs
[z1 , z2 ] of complex numbers, not both zero, given by the equivalence relation

[z1 , z2 ] ∼ [λz1 , λz2 ] for λ = 0.

Consider the map p : S 3 → S 2 defined by

p : S 3 → S 2 , (z1 , z2 ) → [z1 , z2 ]/(|z1 |2 + |z2 |2 )1/2 ).

Since every pair [z1 , z2 ] is normalized on dividing by z1 z 1 + z2 z 2 , it follows that p is

well defined and continuous. Again, consider the circle S 1 = {z ∈ C : |z| = 1} and
two points α = [1, 0] and β = [0, 1] of S 2 . Then U1 = S 2 − {α} and U2 = S 2 − {β}
are two open sets which form an open covering of S 2 . Since every point of U1 can
be expressed by a pair [u, 1], the map
zu z
ψU1 : U1 × S 1 → S 2 , ([u, 1], z) → ( , )
(z1 z 1 + z2 z 2 ) 1/2 (z1 z 1 + z2 z 2 )1/2

is well defined and it maps U1 × S 1 homeomorphically onto p−1 (U1 ) with the prop-
erty
pψU1 (u, z) = u, ∀ u ∈ U1 , z ∈ S 1 .

This constructs ψU1 . Similarly, construct ψU2 . This completes the construction of
the Hopf fibering (S 3 , p, S 2 ). For (z1 , z2 ) ∈ S 3 , the fiber p−1 [z1 , z2 ] = {(λz1 , λz2 ) :
λ ∈ S 1 }. This implies that the fibers of the Hopf fibering (S 3 , p, S 2 ) are precisely the
great circles of S 3 . Hence, it follows that the Hopf fibering (S 3 , p, S 2 ) has fibers a
family of great circles, because the inverse image of p of a point of S 2 is just a great
circle of S 3 and this fibering has fibers a family of great circles. 

Corollary 4.7.2 The 3-sphere S 3 is decomposed into a family of great circles with
the 2-sphere S 2 as a quotient space of the Hopf fibering (S 3 , p, S 2 ).

Proof Consider the Hopf fibering (S 3 , p, S 2 ). It follows from the proof of Theorem
4.7.1 that the 3-sphere is decomposed into a family of great circles, which are called
fibers, having the 2-sphere as a quotient space (or decomposition space). 
4.7 Hopf Fibering of Spheres 275

Remark 4.7.3 The decompositions of compact Lie groups modulo their closed sub-
groups produce fiber bundles, which serve as a valuable source of examples. We rec-
ommend the survey articles by [H. Samelson, Topology of Lie groups, Bull. Amer.
Math. Soc.52 (1952), 2–37] and [A. Borel, Topology of Lie groups and characteristic
classes, Bull. Amer. Math. Soc.61 (1955), 397–432].

4.7.2 A Generalization of the Hopf Map p : S3 → S2

This subsection generalizes the Hopf map p : S 3 → S 2 through the study of some
spaces that arise in projective geometry.

For F = R, C or H, the right vector space Fn consists of elements, which are

ordered sets of n elements of F. If x = (x1 , x2 , . . . , xn ) ∈ Fn and β ∈ F, then xβ =
n
(x1 β, . . . , xn β). Using the usual inner product x and y in Fn by x, y = xi yi ,
1
where xi is the conjugate of xi , it follows that

y, x = x, y, xβ, y = βx, y and x, (yβ) = x, yβ.

Two nonzero elements x, y ∈ Fn are said to be orthogonal denoted by x ⊥ y iff

x, y = 0. The orthogonality relation is symmetric, because, x, y = 0 iff y, x = 0.
Let S be the unit sphere in Fn defined by the locus x, x = 1 and G n be the orthogonal,
unitary or symplectic group according as F = R, C or H, then each G n is a compact
Lie group. Let FP n be the projective space associated with F and it is topolozied by
considering it as a quotient space of Fn+1 − {0}. Geometrically, it may be considered
n+1
  
as the set of all lines through the origin in F n+1
= F ⊕ F ⊕ · · · ⊕ F, since, every
point of Fn+1 − {0} determines a line through the origin 0 and if x and y are nonzero
elements of Fn+1 , then x ∼ y iff there is an element β(= 0) ∈ F such that y = xβ. As
it is an equivalence relation, it defines FP n as the quotient set of equivalence classes
endowed with the quotient topology. In particular,

(i) RP n is called the n-dimensional real projective space.

(ii) CP n is called the n-dimensional complex projective space.
(iii) HP n is called the n-dimensional quaternionic projective space.
The natural projection map Fn+1 − {0} → FP n , w → [w] is continuous and defines
maps on restriction to the unit sphere in Fn which are S n ⊂ Rn , S 2n+1 ⊂ Cn and
S 4n+3 ⊂ Hn . These maps are

pn : S n → RP n ,

qn : S 2n+1 → CP n ,
 276 4 Topology of Fiber Bundles: General Theory of Bundles

and
rn : S 4n+3 → HP n .

Remark 4.7.4 Usually, the common notation p is used instead of pn , qn or rn , unless

there is any confusion. Example 4.7.5 plays a key role in computing the homotopy
groups of sphere (results are only partly known) (see Chap. 5) and hence it reflects
the importance of bundle theory.

Example 4.7.5 (Real, complex and Quaternionic Hopf bundles)

(i) (Real Hopf bundle) ξ = (S n , p, RP n , Z2 ) is a locally trivial fiber bundle with
fiber Z2 .
(ii) (Complex Hopf bundle) η = (S 2n+1 , p, CP n , S 1 ) is a trivial fiber bundle with
fiber S 1 .
(iii) (Quaternionic Hopf bundle) γ = (S 4n+3 , p, HP n , S 3 ) is a locally trivial fiber
bundle with fiber S 3 .

4.8 G-bundles and Principal G-bundles

This section introduces the concepts of G-bundles and principal G-bundles obtained
by an action of a topological group G on a topological space and studies these bun-
dles. A G-bundle is a bundle with an additional structure obtained from an action
of the topological group G on a topological space. Transformation groups are also
derived from an action of topological groups on topological spaces.

The principal G-bundles for a Lie group G named after Sophus Lie (1842–1899)
are studied in Sect. 4.10. The main result of this section is the Corollary 4.8.21 which
asserts that if X is a simply connected space and G is a properly discontinuous group
of homeomorphisms of X , then the fundamental group π1 (X mod G) of the orbit
space X mod G is isomorphic to G. The homotopy property of numerable principal
G-bundles and Milnor construction are studied in Chap. 5. In this section, G denotes
a topological group and X denotes a topological space.

4.8.1 G-spaces

This subsection introduces the concept of G-spaces obtained by an action of a topo-

logical group on a topological space.

Definition 4.8.1 A right action σ of a topological group G with identity element e

on a topological space X is a continuous map

σ : X × G → X,
 4.8 G-bundles and Principal G-bundles 277

the image σ(x, g) denoted by x · g or simply xg such that

(i) x · (gh) = (x · g) · h, ∀ g, h ∈ Gand x ∈ X and
(ii) x · e = x, ∀ x ∈ X .
Then the pair (X , G) endowed with an action σ is called a right G-space or a
topological transformation group. Similarly, a left action of G on X is defined,
and it gives a left G-space. One can convert a right G-space into a left G-space
by defining
g · x = x · g −1

So, it is sufficient to consider either right or a left G-space. Any such space is
called simply a G-space or a topological transformation group.

Definition 4.8.2 Let G be a topological group with identity e and X be topological

group. An action
σ : X × G → X , (x, g) → x · g

of G on X is said to be
(i) free if for all g(= e) ∈ G, x · g = x for every x ∈ X and
(ii) effective or trivial if for all g(= e) ∈ G, there exists an element x ∈ X such
that x · g = x.
(iii) transitive, if given any two x, y ∈ X , there exists an element g ∈ G, such that
x · g = y.

Remark 4.8.3 Let σ : X × G → X be an action of a topological group G on a

topological space X . Then the action
(i) σ is free if the isotropy group G x = {g ∈ G : g(x) = x · g = x} = {e} for every
x ∈ X , which means that g(x) = x for some x ∈ X asserts that g = e.
(ii) σ is effective or trivial, if g(x) = x for every x ∈ X , then g = e, i.e., if the
homomorphism
f : G → homeo(X ), g → ψg ,

is a monomorphism, where homeo(X ) is the group of homeomorphisms of X .

(iii) is transitive, if only one orbit is generated by this action σ.

Example 4.8.4 The condition PD(i) of Definition 4.6.1 implies that the topological
group G acts on X freely. Because, if x ∈ X and x · g = x for some g ∈ G, then, for
any nbd U of x, the point x ∈ U g ∩ U and so U g ∩ U = ∅, which is possible only
when g = e.

Definition 4.8.5 Given a right G-space X , the set X mod G = {xG : x ∈ X } of all
orbits of X under an action σ of G on X , with the quotient topology, which is the
largest topology such that the projection map

p : X → X mod G, x → xG
278 4 Topology of Fiber Bundles: General Theory of Bundles

is continuous. The quotient space X mod G is called the orbit space of X modulo
G.
Remark 4.8.6 Let X be G-space. Then x, y ∈ X are said to be G-equivalent,
denoted by x ∼ y iff y = x · g for some g ∈ G. This is an equivalence relation on X
and the equivalent class determined by x is an orbit of x, denoted by xG.
Proposition 4.8.7 Given a right G-space X , for every point g ∈ G, the map

ψg : X → X , x  → x · g

is a homeomorphism and the projection

p : X → X mod G, x → xG

is an open map.
Proof The map ψg is a homeomorphism with its inverse ψg−1 for every g ∈ G. To

prove the second part, let U be an open subset of X . Then p−1 (p(U )) = U g is a
 g∈G

union of open sets in X and hence U g is an open set in X mod G. This asserts
g∈G
that p(U ) is an open set of X mod G for every open set U of X . 
Remark 4.8.8 The quotient maps X → Xmod G, x → G(x) are identified with the
covering maps. We now use the symbol gx or g(x) for the symbol g · x.
Theorem 4.8.9 is an important theorem in topology. Its geometrical applications
are available in Section 4.12.
Theorem 4.8.9 Let G be a compact topological group and X be a Hausdorff space.
If G acts (from left) on X , and G x is the isotropy group at x ∈ X with orbit orb(x) =
G(x), the orbit of x, then the map

ψ : G/G x → G(x), gG x → g(x) = gx

is a homeomorphism.
Proof Let {gG x } be the set of all left cosets of G x in G. Clearly, the map ψ is
continuous by quotient topology on G/G x . It is onto, because for any y ∈ G(x) it
can be expressed as y = gy (x) for some gy ∈ G and hence ψ(gy G x ) = gy (x) = y.
Again, for g, h ∈ G if g(x) = h(y), then g −1 h ∈ G x , and hence gG x = hG x implies
that ψ is injective. Consequently, ψ is continuous one-one and onto map from a
compact space to a Hausdorff space. This proves that ψ is a homeomorphism. 
Definition 4.8.10 (G-morphism) Given two right G-spaces X and Y , a continuous
map f : X → Y is said to be a G-morphism if f (xg) = f (x)g holds for all x ∈ X and
for all g ∈ G.
4.8 G-bundles and Principal G-bundles 279

Remark 4.8.11 Given two two right G-spaces X and Y , a G-morphism map f :
X → Y sends xG into f (x)G such that

f (xG) ⊂ f (x)G, ∀ x ∈ X .

4.8.2 G-Coverings

This subsection studies G-coverings obtained by the action of a topological group

as orbit spaces.

Definition 4.8.12 A covering p : X → B is said to be a G-coverings, if it is obtained

as a properly discontinuous action of a topological group G on X . It is said to be a
trivial covering of B, if this covering is the product

G × X → X.

Definition 4.8.13 Let p : X → B and q : Y → B be two G-coverings. An isomor-

phism between them is a homeomorphism

ψ:X →Y

such that

(i) q ◦ ψ = p and hence the diagram in Fig. 4.17 is commutative and

(ii) ψ(gx) = gψ(x), ∀ g ∈ G and ∀ x ∈ X .

Proposition 4.8.14 Let ξ be an arbitrary G-covering . Then it is locally trivial as a

G-covering.

Proof ξ : p : X → B be a G-covering. Then every point in B has a nbd U such that the
G-covering qU : p−1 (U ) → U is isomorphic to the trivial G-covering U × G → U.
Let U = p(V ). Then a local trivialization is defined by

p−1 (U ) → U × G, gv → (p(v), g).

This proves that ξ is locally trivial as a G-covering. 

Fig. 4.17 Isomorphism f of

G-coverings
280 4 Topology of Fiber Bundles: General Theory of Bundles

4.8.3 G-bundles

This subsection is devoted to the study of G-bundles which admits an additional

structure obtained from the action of a topological group G on a topological space.

Definition 4.8.15 Let (X , p, B) be a bundle. Then it is said to be a G-bundle if

there exists some G-space structure on X such that the two the bundles (X , p, B) and
(X , pX , X mod G) are isomorphic in the sense that there exists a homeomorphism
f : X mod G → B with the property that the pair of maps

(1X , f ) : (X , pX , X mod G) → (X , p, B)

makes the diagram in Fig. 4.18 commutative (see Definition 4.1.26).

Proposition 4.8.16 (Existence of a bundle morphism) Let X and Y be two G-spaces

and f : X → Y be a G-morphism. Then there exists a bundle morphism

(f , f˜ ) : (X , pX , X mod G) → (Y , pY , Y mod G).

Proof Corresponding to G-spaces X and Y , let ξ(X ) = (X , pX , X mod G) and

ξ(Y ) = (Y , pY , Y mod G) be the G-bundles. Then the G morphism f : X → Y pro-
duces a quotient map

f˜ : X mod G → Y mod G, xG → f (x)G.

Fig. 4.18 Rectangular

commutative diagram for
G-bundle (X , p, B)

Fig. 4.19 G-bundle

morphism
4.8 G-bundles and Principal G-bundles 281

Fig. 4.20 Covering

transformation ψ

The commutativity of the rectangle in the Fig. 4.19 proves that the pair of maps (f , f˜ )
is a bundle morphism. 

Definition 4.8.17 Let p : X → B be a covering space. Then a covering transfor-

mation is a homeomorphism
ψ : X → X : p ◦ h = p,

i.e., ψ is a homeomorphism such that the triangular diagram in Fig. 4.20 is commu-
tative.
If Cov(X /B) denotes the set of all covering transformations of ψ : X → X , then it
is group under usual composition of maps, called the group of covering transfor-
mations of the covering space p : X → B.

Definition 4.8.18 A covering space p : X → B is said to be regular if the image

p∗ π1 (X , x0 ) of induced homomorphism

p∗ : π1 (X , x0 ) → π1 (B, b0 )

is a normal subgroup of π1 (B, b0 ).

Remark 4.8.19 Let G = Cov(X /B) be the group of covering transformations of a

covering space p : X → B. This group resembles π1 (B, b0 ). Then the action of G on
X is properly discontinuous. It is a natural question: is its converse true? Its answer
is available in Theorem 4.8.20, which considers action of properly discontinuous
group G of homeomorphisms of a space X in the sense of Definition 4.6.6.

Theorem 4.8.20 answers the problem raised in Remark 4.8.19.

Theorem 4.8.20 Let G be a properly discontinuous group of homeomorphisms of

a topological space X and X mod G be its orbit space.
(i) Then the projection p : X → X mod G, x → xG is a covering projection.
(ii) If X is connected, then this covering projection is regular and G is its group of
covering transformations.

Proof i) follows from Proposition 4.6.4. The proof the theorem is also given as
follows: by hypothesis, p : X → X mod G, x → xG is the usual projection, then p
is continuous and open by Proposition 4.8.7. If U is an open subset of X satisfying
the condition that whenever U g meets U g , then g = g , then p(U ) is evenly covered
by p. By hypothesis on U, it follows that {U g}g∈G is a disjoint family of open sets
whose union is p−1 (p(U )). To prove the theorem, it suffices to show that

p|U g : U g → p(U )
282 4 Topology of Fiber Bundles: General Theory of Bundles

is a bijection. For x ∈ U ,

p(xg) = p(x) =⇒ p(U g) = p(U ).

Again, if p(xg) = p(x g) for x, x ∈ U, then there exists some s in G such that xg =
xsg. This implies that U g intersects x sg and g = sg. Hence s = e and xg = x g and
p : U g → p(U ). Since by hypothesis, G is properly discontinuous, the sets p(U ) are
evenly covered by p and form an open covering of X mod G. Since p(xg) = p(x),
it follows that G is contained in the group of covering transformations of p. Hence,
the group G and the group of covering transformations are same. Since the group
of covering transformations is transitive on each fiber, it is proved that the covering
projection p is regular. It proves (ii). 

Corollary 4.8.21 identifies properly discontinuous group G of homeomorphisms

of a simply connected space X with the fundamental group of the orbit space
X mod G.

Corollary 4.8.21 Given a simply connected space X and a properly discontinuous

group G of homeomorphisms of X , the fundamental group π1 (X mod G) and the
group G are isomorphic.

Proof Using Theorem 4.8.20, it follows that G is the group of covering transforma-
tions of the regular covering projection p → X mod G, x → xG. Hence, the Corol-
lary follows from Theorem 4.8.20. 

4.8.4 Principal G-bundles

This subsection continues the study of G-bundles by introducing the concept of

principal G-bundles. In particular, principal G-bundles for Lie groups are studied in
Sect. 4.10.

Definition 4.8.22 (Principal G-bundle) Given a topological group G and a topo-

logical space B, a principal G-bundle over B consists of
(a) a fiber bundle p : X → B and
(b) an action X × G → X , (x, g) → xg
such that
PG(i) the shearing map

T : X × G → X × X , (x, g) → (x, xg)

maps the product space X × G homeomorphically to its image space

T (X × G);
4.9 Charts and Transition Functions 283

PG(ii) for the space B = X mod G, the projection p : X → X mod G is the quo-
tient map;
PG(iii) for every point b ∈ B, there exists an open nbd U of b such that the bundle

p : p−1 (U ) → U

is G-bundle isomorphic to the trivial bundle

q:U ×G →U

in the sense that there exists a homeomorphism

ψ : p−1 (U ) → U × G

with q ◦ ψ = p, where the action is given by (x, g )g = (x, gg ).

Remark 4.8.23 (i) The shearing map T formulated in PG(i) is injective iff the
action of G on X is free. This implies by PG(i) that the action of G on the total
space X of a principal bundle is always free. If G and X are compact, then a free
action satisfies PG(i).
(ii) Every free action produces a translation function α : Y → G, where Y = {(x, x ·
g) ∈ X × X } is the image of the shearing map T . Condition PG(i) is equivalent
to a free action with a continuous translation function.

Definition 4.8.24 (Locally trivial principal G-bundle) A fiber bundle ξ = (X , p, B, G)

with a continuous action σ : X × G → X is said to be a (locally trivial) principal
G-bundle, if given an open covering {Ui } of B and for every i, there is a homeomor-
phism
ψi : Ui × G → p−1 (Ui )

such that for all b ∈ Ui and g ∈ G.

(i) (p ◦ ψi )(b, y) = b and
(ii) ψi (b, g) = ψi (b, e) · g.

Example 4.8.25 (Product principal G -bundle) The product G-space B × G is prin-

cipal under the action of G given by (b, t)s = (b, ts).

Remark 4.8.26 For more study of principal G-bundles, see Chap. 5.

4.9 Charts and Transition Functions

This section continues the study of principal G-bundles corresponding to a topologi-

cal group G by introducing the concepts of charts and transition functions of principal
284 4 Topology of Fiber Bundles: General Theory of Bundles

G-bundles and proves a bijective correspondence in Theorem 4.9.8 between the sets
of the equivalence classes of principal G-bundles ξ over a fixed base space and the
equivalence classes of sets of transition functions determined by an atlas of the G-
bundles ξ. This result together with Theorem 4.9.11 establishes a close connection
between n-dimensional F-vector bundles over B and principal G-bundles over B,
where F = R, C, H and G = GL(n, F).

Definition 4.9.1 Given a topological group G with identity e and a principal G-

bundle ξ = (X , p, B, G), a chart (ψ, U ) of ξ is a pair, which consists of
(a) an open set U ⊂ B and
(b) a homeomorphism ψ : U × G → p−1 (U )
such that
(i)
p ◦ ψ = pU

and
(ii)
ψ(b, g) = ψ(b, e)g, ∀ b ∈ U and ∀ g ∈ G

Given an open covering {Uα : α ∈ K} of B, the family of charts {(ψα , Uα ) : α ∈ A}

of B is said to be an atlas of ξ if each homeomorphism

ψα : Uα × G → p−1 (Uα )

satisfies the properties

(p ◦ ψα ) = pUα and ψα (b, g) = ψα (b, e)g, ∀ b ∈ Uα , ∀ g ∈ G.

If an atlas of ξ includes all its charts, then it is said to be a complete atlas.

Example 4.9.2 Let ξ be an arbitrary principal G-bundle. Then it has at least one
atlas.

Definition 4.9.3 Given a topological group G with identity e and a principal G-

bundle ξ = (X , p, B, G), a collection of transition functions T for ξ consists of

(a) an open covering {Uα : α ∈ K} of B and

(b) a family of continuous functions {gαβ : Uα ∩ Uβ → G, ∀ α, β ∈ K} for Uα ∩
Uβ = ∅ such that

gαβ (b)gβγ (b) = gαγ (b), ∀ b ∈ Uα ∩ Uβ ∩ Uγ (= ∅).

Each function gαβ ∈ T is called a transition function on Uα ∩ Uβ and it is some-

times written by the pair (Uα ∩ Uβ , gαβ ).
4.9 Charts and Transition Functions 285

Proposition 4.9.4 The family of transition functions {gαβ : α, β ∈ K} formulated

in Definition 4.9.3 have the following properties.
(i) gαα (b) = gαα (b)gαα (b) ∈ G, ∀ b ∈ B, α ∈ K.
(ii) gαα (b) = e, ∀ b ∈ B, α ∈ K.
(iii) gαβ (b) = [gβα (b)]−1 ∀ b ∈ Uα ∩ Uβ , α, β ∈ K.

Proof Consider the relation gαβ (b)gβγ (b) = gαγ (b), ∀ b ∈ Uα ∩ Uβ ∩ Uγ (= ∅).
(i) In particular, take α = β = γ in the relation gαβ (b)gβγ (b) = gαγ (b). Then it
follows that gαα (b) = gαα (b)gαα (b) ∈ G, ∀ b ∈ B, α ∈ K.
(ii) (i) asserts that gαα (b) = e, ∀ b ∈ B, α ∈ K, since gαα (b) ∈ G and G is a group.
(iii) In particular, take γ = α in the relation gαβ (b)gβγ (b) = gαγ (b). Then using (i),
it follows that

gαβ (b) = [gβα (b)]−1 ∀ b ∈ Uα ∩ Uβ , α, β ∈ K.

Definition 4.9.5 Let {(Uα ∩ Uβ , gαβ ) : α, β ∈ K} and {(Uα ∩ Uβ , gα β ) : α , β ∈

A } be two sets of transition functions of two principal G-bundles ξ = (X , p, B, G)
and ξ = (X , p , B, G) over the same base space B. Then they are said to be equiv-
alent if there exist a family of continuous maps

fα α : Uα ∩ Uα → G with Uα ∩ Uα = ∅

such that

gα β (b) = fα α (b)gαβ (b)[fβ β (b)]−1 , ∀ b ∈ Uα ∩ Uβ ∩ Uα ∩ Uβ : α, β ∈ K and α‘, β ∈ K‘

Definition 4.9.6 Given two n-dimensional F-vector bundles ξ = (X , p, B, Fn ) and

ξ = (X , p , B , Fn ), a morphism

ψ:ξ→ξ

consists of a pair of continuous maps ψ : X → X and ψ̃ : B → B such that

(i) p ◦ ψ = ψ̃ ◦ p and
(ii) ψ|p−1 (b) : p−1 (b) → p −1 (ψ̃(b)) is a linear map for every b ∈ B.
In particular. the identity morphism 1d : ξ → ξ consists of the pair of the identity
map 1X : X → X and the identity map 1B : B → B.

Theorem 4.9.7 establishes a close relation between vector bundles and principal
G-bundles.
Theorem 4.9.7 (i) For every principal G-bundle ξ = (X , p, B, G) over the same
base space B, there exists a unique set of transition functions T = {(Uα ∩
Uβ , gαβ ) : α, β ∈ K} with the property
286 4 Topology of Fiber Bundles: General Theory of Bundles

ψβ : Uβ × G → p−1 (Uβ ) : (b, g)  → ψα (b, gαβ (b)g), ∀ α, β ∈ K, ∀ b ∈ Uα ∩ Uβ , g ∈ G.

(ii) For two G-bundles ξ = (X , p, B, G) and the ξ = (X , p , B , G) and the corre-

sponding sets ξ˜ and ξ˜ of transition functions if ψ : ξ → ξ is a bundle morphism,
then it induces a unique morphism of sets of transition functions

f : ξ˜ → ξ˜

such that
f˜ = ψ̃ : B → B

and

ψ ◦ ψα (b, g) = ψα (ψ̃(b), fα α (b)g), ∀ b ∈ Uα ∩ ψ̃ −1 (Uα ), ∀ g ∈ G.

Proof (i) For α, β ∈ K, the map

ψαβ = ψα −1 ◦ (φβ |(Uα ∩ Uβ ) × G) : (Uα × Uβ ) × G → (Uα × Uβ ) × G : pUα ∩Uβ ◦ ψαβ = pUα ∩Uβ

Writing ψαβ in the form ψαβ (b, fαβ (b, g)) for some fαβ : (Uα ∩ Uβ ) × G → G.
Hence
ψβ (b, g) = ψα (b, fαβ (b, g)), ∀ b ∈ Uα ∩ Uβ , g ∈ G.

This asserts that

ψα (b, fαβ (b, g)) = ψβ (b, g) = ψβ (b, e)g = ψα (b, fαβ (b, e))g = ψα (b, fαβ (b, e)g). ∀ b ∈ Uα ∩ Uβ , g ∈ G.

This implies that

fαβ (b, g) = fαβ (b, e)g, ∀ b ∈ Uα ∩ Uβ , g ∈ G.

Now, taking, gαβ (b) = fαβ (b, e), ∀ b ∈ Uα ∩ Uβ and ∀ α, β ∈ K. g ∈ G, it fol-

lows that φβ has the requisite property. Moreover, for any b ∈ Uα ∩ Uβ ∩ Uα .

ψα (b, gα (b)g) = ψα (b, g) = ψβ (b, gβα (b)g) = ψα (b, gαβ (b)gβα (b)g).

This asserts that gαα (b) = gαβ (b)gβα (b). This proves that ξ˜ = {(Uα , gαβ ) :
α, β ∈ A} is a set of transition functions.
(ii) For any α, β, α , β ∈ K, consider the map

ψα α = ψα−1 ◦ ψ(ψα |(Uα ∩ ψ̃ −1 (Uα × G)) : Uα ∩ ψ̃ −1 (Uα ∩ ψ̃ −1 (Uα )) × G → Uα × G

4.9 Charts and Transition Functions 287

satisfies the relation

pUα ◦ θα β = ψ̃ ◦ pUα ∩ ψ̃ −1 (Uα ),

where θα β is given by

θα β (b, g) = (ψ̃(b), hα β (b, g))

for some hα β : (Uβ ∩ φ̃−1 (Uα )) × G → G. This shows that

ψ ◦ ψα (b, g) = ψα (ψ̃(b), hα α (b, g)), ∀ b ∈ Uα ∩ ψ̃ −1 (Uα ), ∀ g ∈ G.

Hence it follows that

ψα (ψ̃(b), hα α (b, g)) = (ψ ◦ ψj (b, e))g = ψα (ψ̃(b), hαj (b, e)g) = ψα (ψ̃(b), hα α (b, e)g).

This shows that

hα‘α (b, g) = hα α (b, e)g.

Proceed as in (i) to prove (ii) except its uniqueness. To show its uniqueness,
suppose ξ˜ = {(Uα ∩ Uβ , gαβ ) : α.β ∈ K} and ξ˜ = {(Uα ∩ Uβ , gα β ) : α , β ∈
A }. Then

ψα (ψ̃(b), fα α (b)gαβ (b)g)=ψ ◦ ψj (b, gβγ (b)g) = ψ ◦ ψγ (b, g) = ψβ‘ (ψ̃(b), fβ β (b))

= ψα (ψ̃(b), gα β (ψ̃(b))fβ β (b)g).

Hence it follows that

fα α (b)gαβ (b) = gα β (ψ̃(b))fβ β (b).

This proves that {fβ β } is a morphism of sets of transition functions.



Theorem 4.9.8 relates the set of the equivalence classes of principal G-bundles ξ
over a fixed base space B and the equivalence classes of sets of transition functions
associated with an atlas of ξ.

Theorem 4.9.8 Let G be a topological group and ξ be any principal G-bundles over
a fixed base space B. Then there exists a (1–1)-correspondence between the set SP
of the equivalence classes of principal G-bundles ξ and the set ST of equivalence
classes of sets of transition functions associated with an atlas of ξ.
288 4 Topology of Fiber Bundles: General Theory of Bundles

Proof Let ξ and ξ be any two equivalent principal G-bundles over the same base
space B with
ψ:ξ→ξ

an equivalence of G-bundles. Then there exists a morphism f (ψ) : ξ˜ → ξ˜ by Propo-

sition 4.9.7(b). Since ξ and ξ are G-bundles over the same base space, the map
ψ̃ : B → B coincides with 1B . This defines a correspondence

θ : SP → ST , ψ → f (ψ).

To show that θ is surjective, take any set of transition functions ξ˜ ∈ ST . Then by

Ex.4.13.1 of Sect. 4.13 there exist a principal G-bundle ξ and an atlas {(Uj : ψj ) :
j ∈ J} of ξ such that ξ˜ is the corresponding set of transition functions. This proves
that θ is surjective. To show that θ is injective, take two principal G-bundles ξ and ξ
such that the corresponding sets ξ˜ and ξ˜ of transition functions are equivalent, by an
equivalence f : ξ˜ → ξ˜ . Then there exists a morphism φ : ξ → ξ by Exercise 4.13.1
of Sect. 4.1.3 inducing f . As before, φ̃ : B → B is 1B . Define the morphism

f −1 : ξ˜ → ξ˜

by the rule
f −1 = {faj−1 : α ∈ A, j ∈ K}.

Then morphism φ : ξ → ξ of G-bundles is the inverse of ψ. Because, ψ ◦ φ and

φ ◦ ψ are both identity maps. This implies that the principal G-bundles ξ and ξ are
equivalent. This proves the map θ is injective. Hence, it follows that the map θ is
bijection. 
Consider the topological group GL (n, F) = G of the group of all nonsingular
n × n matrices over F. There is a natural problem:
(i) does there exist a (1–1)-correspondence between the set of the equivalence
classes of n-dimensional F-vector bundles over a topological space B and
the set of the equivalence classes of sets of transition functions for B and
G = GL (n, F)?
(ii) does there exist a (1–1)-correspondence between the equivalence classes of sets
of transition functions for B with principal GL (n, F)-bundles ?
We now define chart and atlas of a vector bundle in a way analogous to Definition
4.9.1.
Definition 4.9.9 A chart (U, ψ) of an n-dimensional vector bundle ξ = (X , p, B, Fn )
is a pair consisting of
(a) an open set U ⊂ B and
(b) a homeomorphism ψ : U × Fn → p−1 (U )
such that
4.10 Principal G-bundles for Lie Groups G 289

(i) p ◦ φ = pU and
(ii) φ is linear on all fibers p−1 (b) for b ∈ B.
On the other hand, an atlas is a family {(Uk , ψk ) : k ∈ K} of charts such that {Uk :
k ∈ K} is an open covering of B.

Construction 4.9.10 Let ξ = (X , p, B, Fn ) be a given n-dimensional F-vector bun-

dle with an atlas {(Uk , ψk ) : k ∈ K}. Construction of a set of transition functions
{(Ui , gik ) : i, k ∈ K} for B and the group GL (n, F) is now described :
The maps

ψik = φi −1 ◦ (φk |(Ui ∩ Uk ) × Fn ) : (Ui ∩ Uk ) × Fn → (Ui ∩ Uk ) × Fn , ∀ i, k ∈ K

are well defined and take the form

ψik (b, u) = (b, fik (b, u)) for some fik : (Ui ∩ Uk ) × Fn → Fn .

Hence it follows that for a fixed b ∈ Ui ∩ Uk , the map fik (b, −) : Fn → Fn is a linear
isomorphism and hence fik (b, −) is in GL (n, F). If gik (b) = fik (b, −), then

fik (b, b) = gik (b)u.

It proves that

φk (b, u) = φi (b, gik (b)u), ∀ b ∈ Ui ∩ Uk , u ∈ Fn .

Summarizing the above construction and discussion a basic result interlinking

n-dimensional F-vector bundles over B with principal G-bundles over B, where
G = GL(n, F) is formulated in Theorem 4.9.11.

Theorem 4.9.11 There is a (1–1) correspondence between the set of equivalence

classes of n-dimensional F-vector bundles over a fixed base space B and the set of
equivalence classes of the set of transition functions for B and the general linear
group GL (n, F).

4.10 Principal G-bundles for Lie Groups G

This section continues the study of principal G-bundles over differentiable manifolds
when G is a Lie group. Throughout this section G denotes an arbitrary Lie group.

Definition 4.10.1 A principal (differentiable) G-bundle is a triple (E, p, M ) such

that p : E → M is a differentiable mapping of differentiable manifolds. Further-
more, E is given a differentiable right G-action E × G → E such that the following
conditions hold:
 290 4 Topology of Fiber Bundles: General Theory of Bundles

Fig. 4.21 Local

trivialization

(i) Ex = p−1 (x), x ∈ M are the orbits for the G-action.

(ii) (Local trivialization) Every point in M has an open neighbourhood U and a
diffeomorphism ψ : p−1 (U ) → U × G such that the diagram in Fig. 4.21 com-
mutes, i.e., ψx = ψ|Ex maps Ex to {x} × G; and ψ is equivariant, i.e.,

ψ(xg) = ψ(x)g, ∀ x ∈ p−1 (U ), g ∈ G,

where G acts on U × G by

σ : (U × G) × G → U × G, ((x, g ), g) → (x, g g).

E is called the total space, M the base space and Ex = p−1 (x) the fiber at x ∈ M .
Sometimes we use the notation E to denote the G-bundle (E, p, M ).

Remark 4.10.2 (i) Let (E, p, M ) be a principal G-bundle. Then p is surjective and
open.
(ii) The orbit space E mod G is homeomorphic to M .
(iii) The G-action is free, i.e., x · g = x ⇒ g = e, ∀ x ∈ E, g ∈ G.
(iv) For each x ∈ E, the mapping G → Ex given by g → x · g, is a diffeomorphism.
(v) If N ⊂ M is a submanifold (e.g., if N is an open subset), then the restriction to
N E|N = (p−1 (N ), p, N ) is again a principal G-bundle with base space N .

Example 4.10.3 (i) For an n-dimensional real vector bundle (V , p, M ) the associ-
ated frame bundle (F(V ), p̃, M ) is a principal G = GL (n, R)-bundle.
(ii) For an n-dimensional real vector bundle V equipped with Riemannian metric,
(F0 (V ), p̃, M ) is a principal O(n, R)-bundle.
(iii) Let G be any Lie group and M be a differential manifold. Then (M × G, p, M )
with p the projection onto the first factor, is a principal G-bundle called the
product bundle.

Definition 4.10.4 Let G be a Lie group and (E, p, B) and (F, q, B) be two principal
G-bundles over the same base space B. An isomorphism ψ : E → F is a diffeomor-
phism of the total spaces such that
(i) The diagram in Fig. 4.22 commutes, i.e., ψb = ψ|Eb maps Eb = p−1 (b) to Fb =
q−1 (b) and
(ii) ψ is equivariant, i.e., ψ(xg) = ψ(x)g, ∀ x ∈ E, ∀ g ∈ G.

Remark 4.10.5 The map ψb : Eb → Fb is also a diffeomorphism for each b ∈ B.

4.11 Applications 291

Fig. 4.22 Isomorphism of

principal G-bundles

4.11 Applications

This section studies covering spaces of real projective spaces RP n and figure-eight
F8 . Chap. 5 computes their fundamental groups.

4.11.1 Covering Spaces of RP n

This subsection constructs covering spaces of real projective spaces RP n We start

with the real projective plane RP 2 obtained as a quotient space of the 2-sphere S 2 by
identifying every pair of antipodal points with p : S 2 → RP 2 the natural projection.
It is topologized by defining U to be open in RP 2 iff p−1 (U ) is open in S 2 . By using
this covering space the fundamental group π1 (RP n ) is computed in Chap. 5.

Theorem 4.11.1 (S 2 , p) is a covering space of RP 2 and RP 2 is a surface.

Proof Consider the natural projection map p : S 2 → RP 2 . Let y ∈ RP 2 , x ∈ p−1 (y)

and
A : S 2 → S 2 , z → −z

be the antipodal map. Given an  < 1, choose an  nbd U of x in S 2 such that U

contains no pair {z, A(z)} of antipodal points of S 2 , this choice is possible, since
d (z, A(z)) = 2, where d is the Euclidean metric of R3 . This implies that p : U →
p(U ) is a bijective map and

A : S 2 → S 2 , z → −z

is a homeomorphism. Then A(U ) is an open set in S 2 . Moreover, p(U ) is open in

RP 2 , because, p−1 (p(U )) = U ∪ A(U ) is open in S 2 . This shows that p is an open
map. Consequently, it follows that

p : U → p(U )

is a homeomorphism, since it is bijective, continuous and open. It follows similarly

that

p : A(U ) → p(A(U )) = p(U )

292 4 Topology of Fiber Bundles: General Theory of Bundles

is a homeomorphism.

p(U ) is a nbd of p(x) = y, which is evenly covered by p, since p−1 (p(U )) is the
union of two open sets U and A(U ) each of them is mapped homeomorphically by
p onto p(U ). This proves that (S 2 , p) is a covering space of RP 2 .

To prove the second part, consider a countable basis {Un } for S 2 . This implies
that {p(U )} is a countable basis of RP 2 , Consider the Hausdorff space RP 2 . Given
two distinct points x1 , x2 ∈ RP 2 , the set p−1 (x1 ) ∪ p−1 (x2 ) consists of four points.
If 2 is the minimum distance between them and U1 is the -nbd of one of the points
p−1 (x1 ) and U2 is the -nbd of one of the points p−1 (x2 ), then the sets U1 ∪ A(U1 )
and U2 ∪ A(U2 ) are disjoint. This shows that p(U1 ) and p(U2 ) are disjoint ends of x1
and x2 respectively in RP 2 . The space RP 2 is also a surface, because, S 2 is a surface
and every point of RP 2 has a nbd homeomorphic to an open subset of S 2 . 

Theorem 4.11.2 gives a generalization of Theorem 4.11.1 for n > 1.

Theorem 4.11.2 (S n , p) is a covering space of RP n , where p is the map identifying

antipodal points of S n for n > 1.

Proof To show that (S n , p) is a covering space of RP n , consider open sets

Ui+ = {(x1 , x2 , . . . , xn+1 ) ∈ S n : xi > 0}

and
Ui− = {(x1 , x2 , . . . , xn+1 ) ∈ S n : xi < 0},

which cover S n . For Ui = p(Ui+ ) = p(Ui− ), clearly p−1 (Ui ) = Ui+ ∪ Ui− and the
open sets Ui+ and Ui− are disjoint and homeomorphic to Ui , since the map p|Ui+ is
1–1, continuous and open. This proves that (S n , p) is a covering space of RP n , for
every integer n > 1. 

Definition 4.11.3 Given a covering space (X , p) of B, the multiplicity of the covering

space is defined to be the cardinal number of a fiber. If its multiplicity is n, covering
space (X , p) is said to be an n-sheeted covering space of B or that (X , p) is an n
-fold cover of B.

Example 4.11.4 (i) (S 2 , p) is a double covering of RP 2 , Because, p identifies pairs

of antipodal points of S 2 , and hence, the number of sheets of this covering is 2.
(ii) Consider the covering space (R, p) of S 1 defined in Example 4.5.7. Its number of
sheets is countably infinite. Because, the covering projection p : R → S 1 sends
every integer to the point 1 ∈ S 1 . This asserts that p−1 (1) = Z. This proves that
the number of sheets of the covering space (R, p) of S 1 is countably infinite.
4.12 More Geometrical Applications 293

Fig. 4.23 Covering space of

figure-eight

4.11.2 Covering Space of Figure-Eight F8

This section constructs a covering space of figure-eight F8 and computes its funda-
mental group in Chap. 5 by using this covering space.

Example 4.11.5 (Covering of Figure-eight F8 ) the figure eight F8 is the union of

two circles C1 and C2 with a point x0 in common. We construct a covering space
(X , p) of F8 . Let X be the subspace of the Euclidean plane R2 , which consists of the
x-axis and the y-axis, along with the small circles tangent to these axes, one circle
tangent to the x-axis at each nonzero integer point and one circle tangent to the y-axis
at each nonzero integer point as shown in Fig. 4.23.
Geometrical construction of the projection map p : X → F8 . The map p wraps
the x-axis around the circle C1 , and it wraps the y-axis around the other circle C2
such that the integer points are mapped by p in each case into the base point x0 of
the figure-eight F8 . Hence, every circle tangent to an integer point on the x-axis is
mapped by p onto X homeomorphically, and every circle tangent to an integer point
on the y-axis is mapped by p homeomorphically onto C1 , In either of these cases the
point of tangency is mapped onto the point x0 . Then p is a covering map.

This shows that (X , p) is a covering space of the figure-eight F8 .

4.12 More Geometrical Applications

This section communicates geometrical applications of Theorem 4.8.9.

Theorem 4.12.1 For every integer n ≥ 2, the sphere S n−1 and the factor space
O(n, R)/O(n − 1, R) are homeomorphic.

Proof Proof I: Consider the orthogonal (real) topological space G = O(n, R) for
n ≥ 2. It is a compact topological group and X = S n−1 ⊂ Rn is Hausdorff space. A
matrix M ∈ O(n, R) is a transformation of the Euclidean space Rn and it preserves
lengths of the vectors and hence it is a map of S n−1 to itself. Again, O(n − 1, R) is
regarded as the subgroup of G = O(n, R) obtained by keeping the last coordinate
294 4 Topology of Fiber Bundles: General Theory of Bundles

fixed. Let s0 = (0, 0. . . . , 1). This point is kept fixed by O(n − 1, R). This gives a
mapping from O(n, R) into X = S n−1 . More precisely, define the map

ψ : M → M (0, 0. . . . , 1)t .

If A ∈ O(n − 1, R), then ψ(MA) = ψ(M ) asserts that the map ψ factors via the
left coset space O(n, R)/O(n − 1, R). The topological group O(n, R) acts on S (n−1)
transitively and the induced map

ψ ∗ : O(n, R)/O(n − 1, R) → S n−1

a continuous bijective map from a compact space to a Hausdorff space. Hence ψ ∗ is

a homeomorphism. This proves the theorem.

Proof II: Every orthogonal matrix M ∈ O(n, R) represents a transformation of

Rn that preserves length of vectors. This implies that it is a map from S n−1 to itself.
Consider O((n − 1), R) as a subgroup of the orthogonal group O(n, R), where the
last coordinate is fixed. The isotropy group at the point (0, 0, . . . .0, 1) is O((n −
1), R). Hence, the corollary follows from Theorem 4.8.9 

Theorem 4.12.2 For every integer n ≥ 2, the sphere S 2n−1 and the factor space
U (n, C)/U (n − 1, C) are homeomorphic.

Proof Since S 2n−1 is the set of all unit vectors in Cn , the topological group U (n, C)
acts on S (2n−1) transitively. Because, there is a unitary matrix sending any vector of
length 1 sending any other vector of length 1. Now proceed as in Theorem 4.8.9 to
prove the theorem.. 

Theorem 4.12.3 For every integer n ≥ 2, the sphere S 4n−1 and the factor space
Sp(n, H) / Sp(n − 1, H) are homeomorphic.

Proof Since S 4n−1 is the set of all unit vectors in H, the topological group Sp(n, H)
acts on S (4n−1) transitively. Because, there is a unitary matrix sending any vector of
length 1 sending any other vector of length 1. Now proceed as in Theorem 4.8.9 to
prove the theorem. 

4.13 Exercises and Multiple Choice Exercises

As solving exercises plays an essential role of learning mathematics, various types

of exercises and multiple choice exercises are given in this section. They form an
integral part of the book series.
4.13 Exercises and Multiple Choice Exercises 295

4.13.1 Exercises

In this section, F denotes F = R, C or H.

1. Let p : X → B be covering space. Show that the cardinality of any fiber p−1 (b)
does not depend on the choice of the point b ∈ B.
2. Let p : X → B be a covering map, with B a locally connected space (i.e., every
nbd of a point of B, contains a connected nbd of the point ). Show that

(i) B is a union of connected open sets U such that every component of p−1 (U )
is mapped homeomorphically onto U by p.
(ii) Given a component X of X , the image B = p(X ) is also a component of B.
(iii) The restriction p|X : X → p(X ) is also a covering map.

3. Let p : X → B and p : X → B be two covering spaces and ψ : X → X be a

continuous map such that p ◦ ψ = p. If X , X and B are connected and B is
locally connected, show that ψ is a covering map.
4. Let S be a compact surface and p : X → S be an n-sheeted covering space. Show
that

(i) X is also a compact surface.

(ii) The Euler characteristic κ(X ) = n× Euler characteristic κ(S).
(iii) If S is a sphere with g handles and X is a sphere with k handles, then

k = ng − n + 1.

5. Consider the action σ : Z × R, (n, t) → n + t (translation). Show that the quo-

tient map p : R → R mod Z, t → orb(t) is identified with the covering map
p : R → S 1 , t → eit .
6. Let G = {−1, +1} ∼ = Z2 be a group of two elements. Consider the action

σ : G × S n , (−1, x) → −x, (1, x) → x.

Show that

(i) The quotient group S n mod G is the real projective space RP n .

(ii) The quotient map p : S n → RP n is two-sheeted.

7. Let G n be the group of the nth roots of unity. Show that the action

(i) σ of G n on C

σ : G n × C → C, (g, z) → gz (usual multiplication of complex numbers)

is not properly discontinuous;

296 4 Topology of Fiber Bundles: General Theory of Bundles

(ii) On the other hand, the action ψ of G n on the open subset

ψ : G n × C − {0} → C − {0}, (g, z)  → gz (usual multiplication of complex numbers)

is properly discontinuous.
8. Show the action

σ : Zn × Rn → Rn , ((n1 , n2 , . . . , nn ), (x1 , x2 , . . . , xn ))  → (n1 + x1 , n2 + x2 , . . . , nn + xn )

is properly is continuous. Identify its orbit space Rn mod Zn with the n-torus T n .
9. Let X be the quotient space obtained from the disk D2 by identifying its antipodal
points. Show that the space X is homeomorphic the real projective space RP 2 .
10. For the group G = Z2 , show that any 2-sheeted covering has the unique G-
covering structure.
11. Let p : X → B = Xmod G be a G-covering and f , g : X → X be two G-
isomorphisms. If B is connected and f (x0 ) = g(x0 ), for some x0 ∈ X , that
f = g.
12. (Klein bottle as an orbit space) Let G be the group of homeomorphisms of R2
and H be the subgroup of G generated by the translation

T : R2 → R2 , (x, y) → (x + 1, y + 1)

and by the mapping

f : R2 → R2 , (x, y) → (−x, y + 1).

Show that
(i) this action of H on R2 is properly discontinuous;
(i) the orbit space R2 mod H is homeomorphic to the Klein bottle.
13. (Möbius band as an orbit space) Let G be the group of homeomorphisms of
R2 and H be the subgroup of G generated by the translation

T : R2 → R2 , (x, y) → (x + 1, −y).

(i) this action H on R2 is properly discontinuous;

(ii) the orbit space R2 mod H is homeomorphic to the Möbius band.
14. Let G be a connected Lie group and p : X → G be a covering space. Show that
X is also a Lie group.
15. Let t = 0, 1 or − 1 be a real number and X = Rn − {0}. Consider the action

σ : Z × X → X , (m, x) → t m x.

Show that
4.13 Exercises and Multiple Choice Exercises 297

(i) The action σ is properly discontinuous.

(ii) The orbit space X mod Z is homeomorphic to the product space S 1 × S n−1 .
16. Let G be the group of diffeomorphism of a differentiable manifold X and G acts
on X properly discontinuously. Endow the orbit space X mod G a differentiable
manifold structure such that the projection

p : X → X mod G

is a local diffeomorphism.
17. (Spin group) Let S 3 be the topological group of unit quaternions in H

S 3 = {q = a + bi + cj + dk : a, b, c, d ∈ R and a2 + b2 + c2 + d 2 = 1}.

This group S 3 in H is called the spin group, denoted by spin(3). Identifying the set
{bi + cj + dk} with R3 and taking q ∈ S 3 , consider an orthogonal transformation
of R3 having determinant 1, which is an element of SO(3, R) defined by the
mapping v → v · q · q-1 . Show that
(i) The quotient map

p : S 3 → SO(3, R)

is an epimprphism of groups having ker p = {−1, +1}.

(ii) p : S 3 → SO(3, R) is a 2-sheeted covering.
18. Let G be a Lie group with identity element e and G e denote the connected
component of G, which contains e. Show that
(i) G e is a normal subgroup of G and it is itself a Lie group.
(ii) The quotient group G/G e is discrete.
[Hint: G e is closed under both the operations of multiplication and inversion
inherited from the Lie group G. Moreover, G e is sent to G e by conjugation,
which is continuous is used to prove the normality of G e .]
19. Let ξ = (X , p, B, Fn ) be an n-dimensional vector bundle over the base space B.
Prove the following statements:

(i) If θ = (Y , q, B, Fn ) is also an n-dimensional vector bundle over the same

base space B of ξ, then a B-morphism f : ξ → θ is a vector bundle isomor-
phism iff the map
f : p−1 (b) → q−1 (b)

is a linear isomorphism for every b ∈ B.

(ii) If f : D → B is a continuous map, then its induced bundle f ∗ (ξ) = (Y , q, B,
Fn ) is a vector bundle over D such that
298 4 Topology of Fiber Bundles: General Theory of Bundles

(fξ , f ) : f ∗ (ξ) → ξ

is a morphism of vector bundles, with the map fξ defined by

fξ : Y → X , (d , x) → x.

(iii) If the base space B of ξ is such that B = B1 ∪ B2 with B1 = A × [a, c]

and B2 = A × [c, b], where a, b, c ∈ R : a < c < b and the restricted bun-
dles ξ|B1 = (X1 , p1 , B1 ) and ξ|B2 = (X2 , p2 , B2 ) are trivial bundles, then the
bundle ξ is also trivial.

20. Prove the following statements

(i) If η̃ = {(Uk , gik ) : i, k ∈ K} is a set of transition functions for a given topo-
logical space B and a topological group G, there exists a principal G-bundle
η = (X , p, B, G) and an atlas {(Uk , ψk ) : k ∈ K} for η such that η̃ is the set
of transition functions for the atlas.
(ii) Let η = (X , p, B, G) and η = (X , p , B , G) be two principal G-bundles
with atlases {(Uk , ψk ) : k ∈ K} and {(Ua , ψa ) : a ∈ A} and the correspond-
ing sets of transition functions of η and η . If f : η̃ → η̃ is a morphism of
sets of transition function, then there is a morphism ψ : η → η of principal
G-bundles inducing f .
21. Let G be a topological group (with identity e) acting on a topological space X
(from left). An open set U in X , is said to be proper if

gU ∩ U = ∅, ∀ g ∈ G − {e}.

If every point x ∈ X has a proper open nbd, then it is said that G acts properly
on X . Let ξ : (X , p, B) be a covering space. Then ξ is said to be a regular
covering if the subgroup p∗ π1 (B, b0 ) of π1 (B, b0 ) is normal. Let X be a locally
path-connected space and G act properly on X . For the natural projection

p : X → mod G, x → xG,

prove the following statements:

(i) (X , p, Xmod G) is a regular covering space.

(ii) If X is semi locallyy and 1-connected, then the two groups Cov(X /Xmod G)
and G are isomorphic, i.e., Cov(X /Xmod G) ∼ = G.
(iii) If X is a simply connected space, then the fundamental π1 (Xmod G, ∗) is
isomorphic to the group G.

22. Let X and B be path-connected spaces. Prove the following statements:

4.13 Exercises and Multiple Choice Exercises 299

(i) If (X , p) is regular covering space of B, then p induces the same group

p∗ π1 (X , x0 ) = p∗ π1 (X , x1 ) for every pair of points x0 , x1 lying in the same
fiber.
(ii) For every simply connected space X , any covering space (X , p) of B is
regular.
(iii) For any covering space (X , p) of B, if the fundamental group π1 (B, b0 ) is
abelian, then the covering space (X , p) of B is regular.
(iv) A covering space (X , p) of a connected and locally path-connected space B
with a base point b0 ∈ B is regular if and only if the group Aut (X /B) acts
transitively on the fiber p−1 (b0 ) over the point b0 .

23. Let (X , p) be a covering space of B and b0 ∈ B. Prove the following statements

(i) If X is locally path connected, then given any two points x0 , x1 ∈ Y =

p−1 (b0 ), then there exists an f ∈ Cov(X /B) with f (x0 ) = x1 iff there exists
an h ∈ Aut (Y ) with h(x0 ) = x1 .
(ii) If B is locally path connected and the fiber p−1 (b0 ) = Y is considered as a
G = π1 (B, b0 )-set, then the map

φ : Cov(X /B) → Aut (Y ), h → h|Y

is an isomorphism.
(iii) Let a group G act transitively on the fiber p−1 (b0 ) = Y and let y0 ∈ Y . For
the normalizer NG (G 0 ) of the isotropy group G 0 of y0 ,

Aut (Y ) ∼
= NG (G 0 )/G 0 .

(iv) If B is locally path connected and x0 ∈ p−1 (b0 ), then

Aut (X /B) ∼
= NG (p∗ π(X , x0 ))/p∗ π1 (X , x0 )

and in particular, the fundamental group

π1 (S 1 , 1) ∼
= Z.

(v) If the base space B is an H-space, then every covering space of B is regular.
[ Hint : By hypothesis, B is an H-space. It implies that π1 (B, b0 ) is abelian
for b0 ∈ B and hence every covering space of B is regular. ]

24. If (X , p) and (Y , q) are covering spaces of the same path-connected and locally
path-connected base space B and if b0 ∈ B, x0 ∈ X and y0 ∈ Y be points such
that
p(x0 ) = b0 = q(y0 ) and q∗ π1 (Y , y0 ) ⊂ p∗ π1 (X , x0 ),

then show that

300 4 Topology of Fiber Bundles: General Theory of Bundles

(i) There is a unique continuous map f : (Y , y0 ) → (X , x0 ) with p ◦ f = q.

(ii) X is a quotient space of Y in the sense that (Y , f ) forms a covering space of
X.
25. Given three path connected and locally path-connected spaces X , Y and B and
a covering space (X , p) of B with x0 ∈ X , y0 ∈ Y and b0 ∈ B and p(x0 ) = b0 , if
f : (Y , y0 ) → (B, b0 ) is a continuous map such that f∗ π1 (Y , y0 ) ⊂ p∗ π1 (B, b0 ),
show that there is a continuous map

f˜ : (Y , y0 ) → (X , x0 ) : p ◦ f˜ = f .

26. If ξ = (X , p, B, Fn ) and η = (Y , q, B, Fn ) are two vector bundles and f : X → Y

is a continuous map such that

f |p−1 (b) : p−1 (b) → q−1 (b)

is a linear isomorphism for each b ∈ B, show that f is an isomorphism of vector

bundles.
27. Prove the following statements
(i) If A is a closed subgroup of a Lie group G, show that every normal subgroup
N ⊂ A determines a fiber bundle with bundle map

p : G/N → G/A, gN → gA.

(ii) Let G = {+1, −1} be the two-element group and the n-sphere S n be the
G-space with action given by the relation x(+1) = x, x(−1) = −x. Then
the principal Z2 -space produces a principal Z2 -bundle with the real n-
dimensional projective space RP n as its base space.
(iii) Every covering projection p : X → B is a principal G-bundle, with G the
group of covering transformations with the discrete topology.
(iv) Every fiber bundle p : X → B is an open map.
(v) If X is a G-space, then automorphisms of the trivial G-bundle p2 : X × B →
B are in (1–1)-correspondence with continuous functions f : B → G

4.13.2 Multiple Choice Exercises

Identify the correct alternative (s) (there may be more than one) from the following
list of exercises:

1. Consider the quadruple ξ formulated below:

(i) The quadruple ξ = (R, p, S 1 , Z) with the map p : R → S 1 , t → e2πit is a
fiber bundle with fiber Z.
4.13 Exercises and Multiple Choice Exercises 301

(ii) The quadruple ξ = (Rn , p, T n , F), where T n is the n-dimensional torus

T n = S 1 × S 1 × · · · × S 1,

with the map

p : Rn → T n , (t1 , t2 , . . . , tn ) → (e2πit1 , e2πit2 , . . . , e2πitn )

form a fiber bundle with fiber F which is the set of integer lattice points in
Rn .
(iii) The quadruple ξ = (SO(n, R), p, S n−1 ) is a fiber bundle over S n−1 with fiber
SO((n − 1), R), where SO((n − 1), R) is a subgroup of SO(n, R), which
consists of matrices A ∈ SO(n, R) that keep the vector en = (0, 0, . . . , 1)
fixed and the map

p : SO(n, R) → S n−1 , A → Aen

is the projection map of ξ.

2. Let M be a smooth manifold of dimension n with its tangent bundle T (M )
and the projection map p : T (M ) → M , (x, v) → x. Given a chart (ψ, U ) of M
consider the map

Tψ : p−1 (U ) → ψ(U ) × Rn ⊂ Rn × Rn , (x, v) → (ψ(x), d ψx (v))

(i) T (M ) is a smooth manifold of dimension 2n.

(ii) each Tψ is a homeomorphism.
(iii) T (M ) is second countable and Hausdorff.

3. The tangent bundl over the n-sphere S n in the Euclidean (n + 1) − space Rn+1
is the subbundle ξ = (T (S n ), p, S n ) of the product bundle (S n × Rn+1 , p, S n ),
whose total space T (S n ) is defined by

T (S n ) = {(b, x) ∈ S n × Rn+1 : b, x = 0},

and the projection map p is defined by

p : T (S n ) → S n , (b, x) → b.

(i) An element of the total space T (S n ) is a tangent vector to S n at the point

b ∈ S n.
(ii) The fiber p−1 (b) ⊂ T (S n ) is a vector space of dimension n.
(iii) A cross section of the tangent bundle ξ over S n is a vector field over S n .

4. The normal bundle over S n in the Euclidean (n + 1) − space Rn+1 is the sub-
bundle ξ = (N (S n ), q, S n ) of the product bundle (S n × Rn+1 , p, S n ) whose total
302 4 Topology of Fiber Bundles: General Theory of Bundles

space N (S n ) is defined by

N (S n ) = {(b, x) ∈ S n × Rn+1 : x = tb for some t ∈ R}

and the projection q is defined by

q : N (S n ) → S n , q(b, x) = b.

(i) An element of N (S n ) is a normal vector to S n at the point b ∈ S n .

(ii) The fiber q−1 (b) ⊂ T (S n ) over the point b ∈ S n is a vector space of dimen-
sion 1.
(iii) A cross section of the normal bundle ξ over S n is a normal vector field on
S n.

5. Consider the Hopf fibering H = (S 3 , p, S 2 ).

(i) The fibering H has fibers a family of great circles.
(ii) The 3-sphere S 3 is decomposed into a family of great circles with the 2-
sphere S 2 as a quotient space of the Hopf fibering H.
(iii) The Hopf map p : S 3 → S 2 for the Hopf fibering H = (S 3 , p, S 2 ) has its
generalization through some spaces that arise in projective geometry.

References

Adhikari A, Adhikari MR. Basic topology, vol. 1: metric spaces and general topology. India:
Springer; 2022a.
Adhikari A, Adhikari MR, Basic topology, vol. 2: topological groups, topology of manifolds and
lie groups. India: Springer; 2022b.
Adhikari MR, Adhikari A. Groups. Rings and Modules with Applications. Universities Press,
Hyderabad 2003
Adhikari MR, Adhikari A. Textbook of Linear Algebra: An Introduction to Modern Algebra. Allied
Publishers, New Delhi 2006
Adhikari MR, Adhikari A. Basic modern algebra with applications. New Delhi, New York, Heidel-
berg: Springer; 2014.
Armstrong A. Basic Topology. New York: Springer-Verlag; 1983.
Bredon GE. Topology and Geometry. New York: Springer; 1993.
Hu ST. Homotopy Theory. New York: Academic Press; 1959.
Mayer J. Algebraic Topology. New Jersy: Prentice-Hall; 1972.
Massey WS. A Basic course in algebraic topology. New York, Berlin, Heidelberg: Springer; 1991.
Maunder CRF. Algebraic topology. London: Van Nostrand Reinhhold; 1970.
Munkers JR. Elements of algebraic topology. Addition-Wesley-Publishing Company; 1984.
Rotman JJ. An introduction to algebraic topology. New York: Springer; 1988.
Spanier E. Algebraic topology. New York: McGraw-Hill Book Company; 1966.
Steenrod N. The topology of fibre bundles. Prentice: Prentice University Press; 1951.
Chapter 5
Topology of Fiber Bundles: Homotopy
Theory of Bundles

This chapter continues the study topology of fiber bundles, which has created general
interest as it involves interesting applications of topology to other areas such as
algebraic topology, geometry, physics, gauge groups and addresses the homotopy
theory of bundles. Covering spaces provide tools to study the fundamental groups.
On the other hand, fiber bundles provide similar tools to discuss higher homotopy
groups, which are natural generalizations of fundamental groups. The concept of fiber
spaces is the most important generalization of the concept of covering spaces. The
importance of fiber spaces was greatly realized during 1935–1950 to solve several
problems involving homotopy and homology.
The motivation of the study of fiber bundles and vector bundles was born through
the distribution of signs of the derivatives of the plane curves at every point. Histor-
ically, the recognition of the theory of fiber bundles as a discipline of mathematics
came through the work of H. Whitney (1907–1989), H. Hopf (1894–1971), E. Stiefel
(1909–1978), J. Feldbau (1914–1945) and some others.
For this chapter the books [Adams, 1974], [Adhikari et al. 2022a, 2022b, 2003,
2006, 2014, 2016], [Arkowitz, 2011], [Bredon, 1993], [Dugundji, 1966], [Hopf,
1935], [Hu, 1959], [Mukherjee, 2015], [Steenrod, 1951] and some others are referred
in the Bibliography.

5.1 Homotopy Properties of Vector Bundles Over

Manifolds

This section extends the concept of homotopy for continuous maps to smooth maps
and studies the homotopy properties of vector bundles. Intuitively, two smooth maps
are said to be smoothly homotopic if one can be deformed to the other through
smooth maps. The main result of this section is Theorem 5.1.7 asserting the homotopy
invariance of smooth homotopic maps, in the sense that smooth homotopic maps

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 303
M. R. Adhikari, Basic Topology 3,
https://doi.org/10.1007/978-981-16-6550-9_5
304 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

induce the same vector bundle over smooth manifolds upto equivalence. Some results
on smooth homotopy are also available in Exercises 5.24.1 of Sect. 5.24.

Proposition 5.1.1 Let M and N be two smooth manifolds and ξ = (X, p, M) be a

vector bundle over M. If f : N → M is a smooth map, then there exists
(i) a unique vector bundle η = (Y, q, N ) (up to equivalence) and
(ii) a bundle morphism
(g, f ) : η → ξ

such that for every b ∈ N , the fiber preserving map

gb = g| p−1 (b) : p −1 (b) → p −1 ( f (b))

is a linear isomorphism.

Proof Let M and N be two smooth manifolds and ξ = (X, p, M) be a vector bundle
over M and f : N → M be a smooth map. Construct the bundle η = (Y, q, N ) with
the total space Y defined by

Y = {(b, x) ∈ N × X : f (b) = p(x).

and the projection map q is defined by

q : Y → N , (b, x) → b.

Define
g : Y → X, (b, x) → x.

Consider the commutative rectangular diagram as shown in Fig. 5.1

For local triviality of η, take a vector bundle chart for ξ

ψ : U × Rn → p −1 (U ).

Define

ψ1 : f −1 (U ) × Rn → q −1 ( f (U )), (b, v) → (b, ψ( f (b), v)).

Then ψ1 is a vector bundle chart for η, with its inverse formulated by

ψ1−1 (b, x) = (b, p2 ◦ ψ −1 (x)),

where b ∈ f −1 (U ), x ∈ p −1 (U ) so that f (x) = p(x) and p2 is the projection map

onto the second factor. This proves that g is a diffeomorphism and also it an isomor-
phism on every fiber. 
 5.1 Homotopy Properties of Vector Bundles Over Manifolds 305

Fig. 5.1 Morphism of vector

bundles over manifolds

Definition 5.1.2 The bundle η formulated in Proposition 5.1.1 is called the induced
bundle. It is induced from the bundle ξ by f , and it is denoted by f ∗ ξ or by f ∗ (ξ).
The induced bundle f ∗ ξ is also called the pull-back of ξ by f. The pair of maps
(g, f ) formulated in the same definition is called the canonical bundle map of the
induced map f ∗ ξ.

Definition 5.1.3 Let ξ = (X, p, M) be a vector bundle over a manifold M. Then

η = (Y, q, M) is said to be a subbundle of ξ if Y is a submanifold of X and q = p|Y .
On the other hand, if N a submanifold of M, the restricted bundle of ξ to N , denoted
by ξ| N , is bundle (Y, q, N ), where Y = p −1 (N ) and q = p|Y .

Proposition 5.1.4 Let ξ = (X, p, M) be a vector bundle over a manifold M.

(i) If M1 be a submanifold of M and f : M1 → M is the inclusion map, then the
restricted vector bundle ξ| M1 and the induced vector bundle f ∗ ξ are isomorphic.
(ii) If f 1 : M1 → M and f 2 : M2 → M1 are smooth maps, then the induced bundles
( f 1 ◦ f 2 )∗ (ξ) and f 2∗ ( f 1∗ (ξ)) are isomorphic.
(iii) If 1 M : M → M is the identity map on M, then its induced bundle (1 M )∗ (ξ) is
isomorphic to ξ.

Proof It follows from the definition of induced vector bundle. 

Proposition 5.1.5 Let ξ1 = (X 1 , p1 , M1 ) and ξ2 = (X 2 , p2 , M2 ) be two vector bun-

dles. Then every vector bundle morphism

(g, f ) : ξ1 → ξ2

can be expressed as a product

(g, f ) = (h, f ) ◦ (k, 1 M1 ) : g = h ◦ k,

where (k, 1 M1 ) is a homomorphism, and (h, f ) is the canonical bundle map of the
induced bundle f ∗ (ξ2 ).

Proof Let f ∗ (ξ2 ) = (X, p, M1 ). Consider the maps

h : X → X 2 , (b, x) → x and k : X 1 → X : x → ( p1 (x), g(x)).

Then I m K ⊂ X, because, f ◦ p1 = p2 ◦ g. This implies that k is linear on every

fiber and g = h ◦ k. 
306 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Remark 5.1.6 Theorem 5.1.7 proves that smooth homotopic maps induce the same
vector bundle up to equivalence. Its immediate consequence proves Corollary 5.1.8
saying that every vector bundle over a contractible manifold is trivial.

Theorem 5.1.7 (Homotopy invariance) Let f, g : M → N be two smooth homo-

topic maps and ξ be a vector bundle over M. Then their induced vector bundles over
the manifold M are isomorphic, in notation,

f ∗ (ξ) ∼
= g ∗ (ξ).

Proof By hypothesis, f, g : M → N be two smooth homotopic maps. Let H : M ×

R → N be a homotopy between f and g and p : M × R → M be the projection
map. Then there is an isomorphism between vector bundles H ∗ (ξ) and p ∗ Ht∗ (ξ) over
the closed subinterval M × {t}, because, F = Ht ◦ p on M × {t}, which is closed
in the product topology. Hence, by Exercise 5.24.1 of Sect. 5.24, there exists an
isomorphism

ψ : H ∗ (ξ) → p ∗ Ht∗ (ξ)

over some vertical strip N × (t − δ, t + δ). This implies that the isomorphism class
of Ht∗ (ξ) is a locally constant function of t and it is constant, because the real line
space R is connected. This proves that f ∗ (ξ) ∼= g ∗ (ξ). 

Corollary 5.1.8 is an immediate consequence of Theorem 5.1.7.

Corollary 5.1.8 Every vector bundle over a contractible manifold is trivial.

Proof Let ξ = (X, p, M) be a vector bundle over a contractible manifold M and

b0 ∈ M. Then the identity map 1 M : M → M is homotopic to the constant map
c : M → b0 . Hence it follows by Theorem 5.1.7 that f ∗ (ξ) ∼ = g ∗ (ξ). But the vector
∗
bundle g (ξ) is trivial, because, it is a vector bundle over a point. This proves that
the vector bundle ξ = (X, p, M) over a contractible manifold M is trivial. 

5.2 Fibrations and Cofibrations

From topological viewpoint, the fibers over different points of the base space of a
fiber bundle are homeomorphic. So, it is natural to study these fibers from viewpoint
of homotopy theory. For this study, we investigate a structure in which they are
precisely homotopy equivalent. This investigation generalizes the concept of fiber
bundle to the concept of a fibration.
This section is devoted to the study of the concepts of fibrations and cofibrations
and establishes a close connection between a fibration and a covering projection.
They are dual concepts and well connected with the concepts of the homotopy lifting
property (HLP) and the homotopy extension property (HEP). Duality principle in
5.2 Fibrations and Cofibrations 307

Fig. 5.2 Lifting f˜ of f

homotopy theory was born through these concepts. They play a key role in the study of
homotopy theory by providing strong mathematical tools to invade many problems.
It is interesting that any continuous map can be expressed as a fibration and also as a
cofibration up to homotopy. Historically, the concept of fibration born in geometry
and topology was implicitly found in 1937 in the work of K. Borsuk (1905–1982)
but explicitly first appeared in the work of J. H. C Whiteney (1904–1960) during the
period 1935–1940, while investigating sphere bundles.

5.2.1 Homotopy Lifting Problems

This subsection starts with recalling the homotopy lifting problem of a continuous
map. What is this problem? Given a continuous map p : X → B, a topological space
Y and a continuous map f : Y → B, the lifting problem for f is to examine whether
there is a continuous map f˜ : Y → X (represented by an dotted arrow in the Fig. 5.2)
such that this diagram is commutative in the sense that f = p ◦ f˜. The concept of
homotopy lifting property (HLP) and that of the homotopy extension property (HEP)
provide key tools in the study of homotopy theory. The first one leads to the concept
of fibration, the second one leads to the concept of cofibration, and they are dual
concepts.
The concept of homotopy lifting property (HLP) is very important in homotopy
theory, and it leads to the concept of fibration. On the other hand, the homotopy
extension property (HEP), the dual concept of the homotopy lifting property (HLP),
leads to the concept of cofibration, a dual concept of fibration.

Definition 5.2.1 Let p : X → B and f : Y → X be two continuous maps and H :

Y × I → B be also continuous map such that

H (y, 0) = p f (y), ∀ y ∈ Y.

Then the map p : X → B is said to have the homotopy lifting property (HLP) with
respect to Y , if there exists continuous map

H̃ : Y × I → X : H̃ (y, 0) = f (y), ∀ y ∈ Y and H = p ◦ H̃ .

Remark 5.2.2 Definition 5.2.1 asserts that the existence of H̃ is equivalent to

the existence of a continuous map H̃ : Y × I → X indicated by the dotted arrow
in the diagram as shown in Fig. 5.3 such that this diagram is commutative. If
308 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.3 Homotopy lifting

H̃ of p

Fig. 5.4 Homotopy lifting

problem for p

Fig. 5.5 Homotopy lifting

problem

h 0 (y) = (y, 0) ∀ y ∈ Y, then a homotopy lifting problem can be represented by the

commutative diagram in Fig. 5.4.
The two continuous maps f : Y → X and H : Y × I → B in the above diagram
are said to form the data for the given lifting problem. H̃ lifts the homotopy of H
of p ◦ f to a homotopy of f, because, the map H is a homotopy of the composite
p ◦ f and a solution to the problem is to determine a homotopy

H̃ : Y × I → X o f f : p ◦ H̃ = H.

Remark 5.2.3 Proposition 5.2.4 proves that the lifting problem for continuous maps
f : Y → B to X is a problem of homotopy category. Because, given two continuous
maps p : X → B and f : Y → B the map f can or cannot be lifted to X depends
on the homotopy class of f. This asserts that the lifting problem for continuous maps
f : Y → B to X is a problem of homotopy category.

Proposition 5.2.4 If a continuous map p : X → B has the HLP with respect to a

space Y, then one of the homotopic maps f g : Y → B can be lifted to X if the
other one can also be lifted to X.

Proof By hypothesis the continuous map p : X → B has the HLP with respect to a
space Y . If f is defined by

f : Y → B, y → H (y, 0), then f˜ : Y → X

is a lifting of f ; if g is defined by
5.2 Fibrations and Cofibrations 309

Fig. 5.6 Homotopy lifting

problem

g : Y → B, y → H (y, 1),

then H : f g : Y →B. and H̃ : f˜ g̃ : Y →X, where g̃ : Y → X : y → H̃ (t, 1)

is a lifting of g to X. Consequently, it follows that if H : f g : Y → B and if f
has a lifting f˜, then the homotopy H can be lifted to H̃ : Y × I → X : f˜ g̃. This
implies if f g : Y → B and if the map f has a lifting f˜ : Y → X, the homotopy
H can be lifted to H̃ as shown in diagram in Fig. 5.6 and hence g has a lifting
g̃ : Y → X with H̃ : f˜ g̃. By symmetry, it follows that if f g : Y → B and if
the map g has a lifting g̃, then f has also a lifting f˜ such that f˜ g̃. 

5.2.2 Fibration: Introductory Concepts

This subsection conveys the concept of a fibration which was explicitly first found
in the work of Whitney during 1935–1940 on sphere bundles. This concept is a
natural generalization of fiber bundles in the sense that in a fiber bundle, the fibers
over different points of its base space are homeomorphic; on the other hand in a
fibration, the fibers over different points of its base space are homotopy equivalent.
Fibrations (or Hurewicz fiber spaces) [Hurewicz, 1955] provide powerful tools to
study homotopy theory. For example, a covering map is a fibration, and a continuous
map p : X → B has the HLP with respect to a space Y if every problem represented
by the commutative diagram in Fig. 5.5 has a solution (Fig. 5.6).

Definition 5.2.5 Let p : X → B be a base point preserving continuous onto map.

Then it is said to be a fibration (or fiber map or Hurewicz fiber space) if p has the
HLP with respect to each space Y in the sense of Definition 5.2.1.
(i) The space X is called the total space, and the space B is called the base space
of the fibration.
(ii) For every point b ∈ B, the space p −1 (b) = F ⊂ X is called the fiber over b.
(iii) A Serre fibration is a continuous surjective map p : X → B such that it has
the HLP with respect to disk Dn ⊂ Rn for every n ≥ 1. It is also called a weak
fibration.
310 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.7 Diagram involving

classifying space K and
classifying map g for p

Notation: The fibration p : X → B with the fiber space F over some point of B
and the inclusion i : F → X is denoted by
p
F → X −
→ B.

Example 5.2.6 The trivial projection

p : B × F → B, (b, f ) → b

is an important example of fibration.

Example 5.2.7 Every covering projection is a fibration by Theorem 5.2.15.

Definition 5.2.8 Let g : B → K be a continuous map and K I be the space of all

continuous maps α : I → K topolozied by the compact open topology. The path
space Pg of g is defined by

Pg = {(b, α) : α(0) = k0 ∈ K and α(1) = g(b)} ⊂ B × K I .

Definition 5.2.9 uses the notation of the diagram in Fig. 5.3.

Definition 5.2.9 (Principal fibration) Let p : X → B be a fibration. It is said to be a

principal fibration if there exist a topological space K and a continuous map g : B →
K such that the diagram in Fig. 5.7 is commutative with p1 : Pg → B, (b, α) → b a
projection. The space K is called the classifying space , and g is called the classifying
map for the principal fibration p : X → B with X Pg .

Theorem 5.2.10 solves a lifting problem associated with a principal fibration with
the help of a classifying map.

Theorem 5.2.10 Let p : X → B be a principal fibration and f : Y → B be a con-

tinuous map. Then f has a lifting f˜ : Y → X iff g ◦ f is homotopic to a constant
map, where g : B → K is the classifying map.

Proof By hypothesis p : X → B is a principal fibration with a classifying map

g : B → K and f : Y → B is a continuous map. Then X Pg with projection p1 :
Pg → B, (b, α) → b. This implies that there exists two continuous maps u : X →
Pg and v : Pg → X such that v ◦ u 1 X and u ◦ v 1Pg . Again, p1 ◦ u = p and
p ◦ v = p1 . First suppose that f : Y → B has a lifting f˜ : Y → X. Then by the given
5.2 Fibrations and Cofibrations 311

Fig. 5.8 Existence of lifting

of f for the principal
fibration p

Fig. 5.9 Diagram extending

the classifying map g

conditions and chasing the diagram as shown in Fig. 5.8 , there exists a homotopy
H :g◦ f c, where c : Y → K , y → k0 is the constant map. Conversely, let H :
g◦ f c. Then

H : Y × I → K : H (y, 0) = (g ◦ f )(y) = g( f (y)) and H (y, 1) = c(y) = k0 , ∀ y ∈ Y.

Define two maps

Hy : I → K , t → H (y, t)

and

f˜ : Y → Pg → X, y → ( f (y), Hy ) → v( f (y), Hy ).

This implies that f has a lifting f˜ : Y → X, because,

( p ◦ f˜)(y) = ( pov)( f (y), Hy ) = p1 ( f (y), Hy ) = f (y), ∀ y ∈ Y

and hence p ◦ f˜ = f. 

Theorem 5.2.11 solves a lifting problem of a continuous map f associated with

a principal fibration with the help of classifying map from the mapping cone of f.

Theorem 5.2.11 Let p : X → B be a a principal fibration and f : Y → B be a

continuous map. Then for this fibration, f has a lifting f˜ iff there exists a continuous
map g̃ : C f → K extending the classifying map g in the diagram in Fig. 5.9, where
C f is the mapping cone of f.

Proof Consider the mapping cone C f of f : Y → B constructed from the mapping

cylinder M f by identifying Y × {0} ∪ {∗} × I with ∗ in B in the category Top ∗ of
pointed topological spaces. First suppose that there is a homotopy
312 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

H :c g ◦ f : Y → K,

where c : Y → k0 ∈ K is a constant map. Construct the map


(y, t) → H (y, t),
g̃ : C f → K ,
b → g(b) for b ∈/ f (Y ).

Then
g̃(y, 0) = y0 and g̃(y, 1) = g f (y) = g ◦ f, ∀ y ∈ Y

proves that g̃ is the required extension of g. Next suppose that g̃ is an extension

of g. Then there exists a homotopy

H : Y × I → K , (y, t)  → g̃(y, t) : H (y, 0) = g̃(y, 0) = k0 and H (y, 1) = g̃(y, 1) = (g ◦ f )(y) ∀ y ∈ Y.

This proves that g ◦ f c. 

Proposition 5.2.12 If p : X → B is a fibration and β : I → B is any path in B such
that β(0) ∈ p(X ), then the path β can be lifted to a path β̃ : I → X in X .
Proof Under the given condition, the path β can be regarded as a homotopy β :
{y0 } × I → B, where {y0 } is a one-point space. If x0 is a point in X such that
p(x0 ) = β(0), then there is a map f : {y0 } → X such that p f (y0 ) = β(y0 , 0). This
implies by the HLP of p that there exists a path

β̃ : I → X : β̃(0) = x0 and p ◦ β̃ = β.


Example 5.2.13 Given a topological space X , let P(X ) be the space of all paths
β : I → X in X topolozied by the compact open topology. Consider the map

p : P(X ) → X × X, β → (β(0), β(1))

and two other maps p0 , p1

p0 , p1 : P(X ) → X, β → β(0), β(1).

Then all these three maps p, p0 and p1 are fibrations.

Example 5.2.14 Given two fibrations p : X → Y and q : Y → B, then their com-
posite q ◦ p : X → B is also a fibration.
Theorem 5.2.15 Let p : X → B be an arbitrary covering projection. Then it is a
fibration.
5.2 Fibrations and Cofibrations 313

Fig. 5.10 Homotopy lifting

problem

Fig. 5.11 Commutative

triangle representing a
cofibration f

Proof By hypothesis, p : X → B is a covering projection. Suppose the diagram

in Fig. 5.10 represents the corresponding homotopy lifting problem. Then for every
point y ∈ Y , there exists a unique path β y : I → X in X such that β y (0) = f (y) and
pβ y (t) = H (y, t). This defines a continuous map

H̃ : Y × I → X, (y, t) → β y (t).

Hence it follows that p is a fibration. 

5.2.3 Cofibration: Introductory Concepts

This subsection communicates the concept of cofibration, a dual concept of a fibra-

tion. This subsection works in the category Top∗ of pointed topological spaces and
base point preserving continuous maps. The duality principle between fibration and
cofibration is based on the idea that while defining a fibration as a map satisfying
homotopy lifting property (HLP), if the directions of all arrows are reversed, a dual
concept is obtained, which is called a cofibration. This implies that a continuous map
f : X → Y is a cofibration if it satisfies the condition: given a map g̃ : Y → Z in
Top ∗ and a homotopy F t : Y → Z such that there is a continuous map Ft : X → Z
making the triangle in Fig. 5.11 commutative, in the sense that F t ◦ f = Ft .

Definition 5.2.16 Let f : X → Y be a continuous map in the category Top ∗ . Then

it is said to be a cofibration if it satisfies the following conditions: given
(i) any topological space Z ∈ Top ∗ ,
(ii) a continuous map g : Y → Z and
(iii) a homotopy H : X × I → Z starting from g ◦ f ,
there exists a homotopy K : Y × I → Z , starting from g such that H = K ◦ ( f ×
1d ) making the three triangles shown in Fig. 5.12 commutative where j0 : Y → Y ×
I, y → (y, 0) and j0 : X → X × I, x → (x, 0).

Remark 5.2.17 For any subspace A ⊂ X ∈ Top ∗

314 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.12 Diagram

representing cofibration of f

(i) the inclusion i : A → X is a cofibration if the pair (X, A) of spaces has the
absolute homotopy extension property with respect to any space Y in the sense
that if given continuous maps h : X → Y and H : A × I → Y such that h(x) =
H (x, 0), ∀ x ∈ A, there is a continuous map G : X × I → Y with the property

G(x, 0) = h(x), ∀ x ∈ X and G| A×I = H.

(ii) On the other hand, its converse is not necessarily true. Because Definition 5.2.16
of a cofibration works in the category Top ∗ but the absolute homotopy extension
property deals with maps and homotopies that are not necessarily in the category
Top ∗ .
Theorem 5.2.18 Let f : X → Y be an arbitrary continuous map in the category
Top ∗ . Then it is the composite of a cofibration and a homotopy equivalence.
Proof By hypothesis, f : X → Y is an arbitrary continuous map in the category
Top ∗ . Consider the mapping cylinder M f in the category Top ∗ constructed from Y
and (X × I)/(x0 × I) by identifying the points (x, 1) and f (x) for all x ∈ X. This
defines an inclusion map

g : X → M f : x → [(x, 0)].

Define another map h : M f → Y by sending every element y ∈ Y to y itself and

every other elements [(x, t)] to f (x). Then f = h ◦ g. To prove the theorem it is
sufficient to show that g is a cofibration and h is a homotopy equivalence. To show
that g is a cofibration, take any continuous map u : M f → Z in Top ∗ and a homotopy

H : X × I → Z : H (x, 0) = (u ◦ g)(x), ∀ x ∈ X,

which means that the homotopy starts from u ◦ g.

Define two continuous maps FX and FY

u(x, (2t − s)/(2 − s)), 0 ≤ s ≤ 2t
FX : (X × I) × I → Z , (x, t, s) →
H (x, s − 2t), 2t ≤ s ≤ 1
and
FY : Y × I → Z , (y, s) → u(y), ∀ s ∈ I.

Then FX (x, 1, s) = u(x, 1) = (u ◦ f )(x) = FY ( f (x), s), and FX , FY give rise

together a homotopy
5.2 Fibrations and Cofibrations 315

F : Mf × I → Z

such that it starts from u and satisfies the relation

F ◦ (g × 1d )(x, s) = F(x, 0, s) = H (x, s).

This implies that G ◦ (g × 1d ) = H and hence it proves that g is a cofibration.

To show that h is a homotopy equivalence, consider the continuous map j : Y →
M f defined to be (the restriction) of the identification map onto M f . Then

h ◦ j = 1Y and j ◦ h : M f → M f , y → y and (x, t) → f (x).

Define a homotopy

G : M f × I → M f , (y, t, s) → y and (x, t, s) → (x, t + s(1 − t)).

This asserts that G : 1 M f j ◦ h. This proves that h is a homotopy equivalence. 

Remark 5.2.19 Since each continuous map f : X → Y in Top ∗ is also the composite
of a homotopy equivalence and a fiber map, it implies that the dual of the Theorem
5.2.18 is also valid in Top ∗ .

Theorem 5.2.20 and Corollary 5.2.21 both characterize cofibration in terms of

retraction.

Theorem 5.2.20 Let X be topological space and K be a closed subset of X. Then

the inclusion
i : K → X

is a cofibration if the subspace

X × {0} ∪ K × I ⊂ X × I

is a retract of X × I.

Proof By hypothesis, K is a closed subset of a topological space X and i : K → X

is the inclusion map.
First suppose that
i : K → X

is a cofibration. Then corresponding to a given pair of continuous maps

f : X → X × {0} ∪ K × I, x → (x, 0)

and
H : K × I → X × {0} ∪ K × I, (k, t) → (k, t),
316 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

there exists a retraction

r : X × I → X × {0} ∪ K × I.

Since r is a retraction, the first part is proved. Next suppose that there exists a
retraction
r : X × I → X × {0} ∪ K × I.

Corresponding to a given topological space Y , a continuous map f : X → Y , and a

homotopy H : K × I → Y such that H (k, 0) = f (i(k)), ∀ k ∈ K , define a contin-
uous map

( f ◦ p X ◦ r )(x, t), if (x, t) ∈ r −1 (X × {0})
F : X × I → Y, (x, t) →
(H ◦ r )(x, t), if (x, t) ∈ r −1 (K × I).

Then F is continuous, because the projection p X is continuous, X × {0} and K × I

are closed sets in the product space X × I. This proves that i is a cofibration. 
Corollary 5.2.21 Let X be topological space and K be a closed subset of X. Then
the inclusion
i : K → X

is a cofibration iff the inclusion

j : X × {0} ∪ K × I → X × I

is a retraction.
Proof It follows from Theorem 5.2.20. 

5.3 Hurewicz Fiberings and Characterization of Fibrations

This section continues the study of fibrations, characterizes path liftings of fibrations
with the help of their fibers and proves Hurewicz theorem. This theorem is due
to W. Hurewicz (1904–1956). It gives a sufficient condition for a map p : X →
B to be a fibration [Hurewicz, 1955]. The concept of fibering plays a key role in
the homotopy theory of bundles specially in the application of homotopy theory
to geometric problems and provide rich supply of tools to compute the homotopy
groups of different spaces.
Definition 5.3.1 communicates the concept of Hurewicz fibering, named after W.
Hurewicz and studies it.
Definition 5.3.1 A continuous map p : X → B is said to be a Hurewicz fibering
if it has the homotopy lifting property (HLP) with respect to any space Y .
5.3 Hurewicz Fiberings and Characterization of Fibrations 317

Proposition 5.3.2 Given topological spaces Y and Z , and a continuous map

f : Y → Z , consider the space X f = {(y, w) ∈ Y × Z I : w(0) = f (x)} ⊂ Y × Z I
with subspace topology induced from the product topology on Y × Z I , where
I = [0, 1] has the subspace topology inherited from the real line space R. It has
the following properties.
(i) The spaces X f and Y are homotopy equivalent.
(ii) The map
β : X f → Z , (y, w) → w(1)

is a Hurewicz fibering having fiber

X f = {(y, w) ∈ Y × Z I : w(0) = f (x), w(1) = z ∗ ∈ Z }.

Proof Define the three maps

α : X f → Z , (y, w) → w(1),

β : X f → Y, (y, w) → y

and
γ : Y → X f , y → (y, c f (y) ),

where c f (y) denotes the constant path at f (y) in Z . Then α ◦ γ = f and β ◦ γ = 1Y ,

and the maps make the diagram in Fig. 5.13 commutative.
(i) The continuous map

F : X f × I → X f , (y, w, t) → (y, wt ) : wt (s) = w(st)

is a homotopy such that γ ◦ β 1 X f . Moreover, β ◦ γ = 1Y implies that β ◦ γ

1Y . Hence it is proved that Y Xf.
(ii) Let K ⊂ Y × Z I and h : K ×{0} → X f : h(k, 0) = (h 1 (k), h 2 (k)) ∈ X f , ∀ k ∈
K ⊂ Y × Z I . Using the commutativity of rectangular diagram in Fig. 5.14,
define the map

H : K × I → X f ⊂ Y × Z I , (k, t) → (H1 (k, t), H2 (k, t)),

where H1 and H2 are defined by

H1 (k, t) = h 1 (k)

and 
h 2 (k)(s(1 + t)), 0 ≤ s ≤ 1/(1 + t)
H2 (k, t)(s) =
F(k, (1 + t)s − 1), 1/(1 + t) ≤ s ≤ 1.
318 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.13 Triangle for

Hurewicz fibering involving
α, β, γ

Fig. 5.14 Commutative

diagram for Hurewicz
fibering

Since the equality ( f ◦ H1 )(k, t) = H2 (k, t)(0), ∀ k ∈ K , ∀ t ∈ I holds, the

map H is continuous, because the maps H1 and H2 are both continuous.
Moreover, H (k, 0)(s) = h(s) and (α ◦ H )(k, t) = F(k, t). It asserts that X f is
its fiber. 

Theorem 5.3.3 (Hurewicz ) Given a paracompact space B and a covering map p :

X → B, there is an open covering {Ua : a ∈ A} of B such that for every Ui ∈ {Ua }

p| p−1 (Ui ) : p −1 (Ui ) → Ui

is a fibration. Then p is also a fibration.

Proof The proof is long. See [Dugundji, 1966, p. 400]. 

Corollary 5.3.4 Given a paracompact space B and a fiber bundle (X, p, B, F), the
projection p → B is a fibration.

Proof The corollary follows directly from Hurewicz Theorem 5.3.3. 

5.4 Homotopy and Path Lifting Properties of Covering

Spaces

This section continues the study of covering spaces initiated in Chap. 4 and gives
homotopy classification of covering spaces and proves Monodromy Theorem 5.4.7,
which characterizes equivalence of two lifted paths in a covering space in terms of its
projection map. It also studies General Lifting Problems from homotopy viewpoint.
Covering spaces of a pointed topological space (X, x0 ) carry a close relation with the
higher homotopy groups πn (X, x0 ) for n ≥ 2. For example, Theorem 5.4.26 asserts
that every covering space induces an isomorphism between the higher homotopy
group of its total space and that of its base space at every dimension . Finally, it is
proved in Theorem 5.4.27 that πn (S 1 ) = 0 for every n ≥ 2.
5.4 Homotopy and Path Lifting Properties of Covering Spaces 319

5.4.1 Introductory Concepts

Definition 5.4.1 Let X be a topological space.

(i) It is called locally path connected if for every point x of X and every nbd U of
x, there exists an open set V with the property x ∈ V ⊂ U such that every pair
of points of V can be joined by a path in U.
(ii) It is called semilocally simply connected if every point x ∈ X has a nbd U with
the property that the homomorphism

i ∗ : π1 (U, x) → π1 (X, x)

induced by the inclusion

i : U → X

is trivial in the sense that every closed path in U at x is homotopic to a constant

map in X.
Clearly, X is locally path connected iff every path component of each open subset
of X is open.

Definition 5.4.2 (i) Given a covering space ( X̃ , p) of X and a path f : I → X, a

path f˜ : I → X̃ is said to be a lifting or a covering path of f if p ◦ f˜ = f,
diagrammatically, if they make the triangle given in Fig. 5.15 commutative.
(ii) On the other hand, if H : I × I → X is a homotopy, then a homotopy H̃ : I ×
I → X̃ such that p ◦ H̃ = H , is called a lifting or covering homotopy of H.

Theorem 5.4.3 (The Path Lifting Property) Given a covering space ( X̃ , p) of X

and a path f : I → X starting at a point x0 ∈ X with x̃0 ∈ p −1 (x0 ), there exists a
unique covering path f˜ : I → X̃ given in Fig. 5.16 of f starting at the point x̃0 with
the property p ◦ f˜ = f.

Fig. 5.15 Lifting of f in X̃

Fig. 5.16 Path lifting

property (PLP)
320 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Proof Existence of f˜: Let [a, b] ⊂ I be a subinterval of I such that f ([a, b]) ⊂ U
for an admissible nbd U of x = f (a) in X. Then for any x̃ ∈ p −1 (x), the point x̃
will be an element of a unique sheet V , say. This defines a map

g̃ : ([a, b], a) → ( X̃ , x̃) : g̃ = ( p|V )−1 ◦ ( f |[a,b] )

such that
p ◦ g̃ = f |[a,b] .

For any t ∈ I, let Ut be an admissible nbd of f (t). Then the family C = { f −1 (Ut ), t ∈
I} forms an open cover of the compact metric space I, and hence C has a Lebesgue
number λ. This asserts that for 0 < δ < λ and a subset I1 with diameter less than δ, the
subinterval I1 ⊂ f −1 (Ut ) for some t ∈ I. Hence f (I1 ) ⊂ Ut partitions the interval
I by the points t1 = 0, t2 , . . . , tk = 1, where ti+1 − ti < δ for 1 ≤ i ≤ k − 1. This
shows that there exists a continuous map g̃1 : [0, t2 ] → X̃ such that p ◦ g̃1 = f |[0,t2 ]
and g̃1 (o) = x0 . Analogously, there is a continuous map g̃2 : [t2 , t3 ] → X̃ such that
p ◦ g̃2 = f |[t2 ,t3 ] and g̃2 (t2 ) = g̃1 (t2 ) . Proceeding in this way, for 1 ≤ i ≤ k − 2,
there is a continuous map
g̃i+1 : [ti+1 , ti+2 ] → X̃

such that
p ◦ g̃i+1 = f |[ti+1 ,ti+2 ] and g̃i+1 (ti+1 ) = g̃i (ti+1 ).

Using gluing lemma for continuity of functions, and assembling the functions gi ,
there exists a continuous function

f˜ : I → X̃ : f˜i (t) = g̃i (t), ∀ t ∈ [ti , ti+1 ].

The uniqueness of f˜: Since I is connected, and by assumption any two lifts of f
agree at the point 0 ∈ I, the uniqueness of f˜ follows from Exercise 5.24.1 of Sect.
5.24. 
Corollary 5.4.4 (Homotopy Lifting Property (HLP)) Given a covering space ( X̃ , p)
of X and a homotopy H : I × I →X such that H (0, 0) = x0 and a point x̃0 ∈
p −1 (x0 ), there exists a unique homotopy H̃ : I × I → X̃ with the property that
H̃ (0, 0) = x̃0 .
Proof Subdividing I × I into rectangles and proceed as in Theorem 5.4.3 (here
subdivide I × I in place of I). 
Theorem 5.4.5 gives a generalization of Corollary 5.4.4.
Theorem 5.4.5 (The Generalized Homotopy Lifting Property) Given a covering
space ( X̃ , p) of X and a compact space A, if f : A → X̃ is continuous and H :
A × I → X is a homotopy starting from p ◦ f , then there exists a homotopy H̃ :
A × I → X̃ starting from f lifts H. Moreover, if H is a homotopy relative to a
subset A of A, then H̃ has also the same property.
5.4 Homotopy and Path Lifting Properties of Covering Spaces 321

Proof Proceed as above. 

Theorem 5.4.6 Given a covering space ( X̃ , p) of X, if x̃0 ∈ p −1 (x0 ), then for any
path

f : I → X : f (0) = x0 ,

there exists a unique path

f˜ : I → X̃ : f˜(0) = x̃0 and p ◦ f˜ = f.

Proof Let B = {b} be a singleton space and f : B → X, b → x0 be a given map.

Consider the homotopy

H : B × I → X, (b, t) → f (b).

Then by the Homotopy Lifting Property, there is a continuous map

H̃ : B × I → X̃ : H̃ (b, 0) = x̃0 and p ◦ H̃ = H.

Then,
f˜ : I → X̃ , t → H̃ (b, t)

is a path in X̃ starting from x̃0 such that

( p ◦ f˜)(t) = ( p ◦ H̃ (b, t)) = H (b, t) = f (t), ∀ t ∈ I ⇒ p ◦ f˜ = f.

This shows that the path f˜ : I → X̃ is unique. 

Theorem 5.4.7, known as ‘Monodromy theorem’, characterizes equivalence of
two lifted paths in a covering space with the help of its projection map.
Theorem 5.4.7 (The Monodromy Theorem) Given a covering space ( X̃ , p) of X
and two paths f˜ and g̃ with their common initial point x̃0 , they are equivalent in the
sense f˜ g̃ rel I iff p ◦ f˜ and p ◦ g̃ are equivalent paths in X.
Proof First suppose that f˜, g̃ : I → X̃ be two equivalent paths in X̃ . Then there is a
homotopy H : f˜ g̃ rel İ, and hence p ◦ H : I × I → X is a continuous map such
that
p ◦ H : p ◦ f˜ p ◦ g̃ rel İ.

This proves that p ◦ f˜ and p ◦ g̃ are equivalent paths in X. Conversely, suppose

that p ◦ f˜ and p ◦ g̃ are equivalent paths in X . Then there is a continuous map
F : I × I → X such that
F : p ◦ f˜ p ◦ g̃ rel İ.

Hence by using Homotopy Lifting Property, it follows that there is a unique homotopy
322 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

F̃ : I × I → X : F̃(0, 0) = x̃0 and p ◦ F̃ = F.

Then the restricted map F|(I × {0}) defines a path

f˜ : I → X̃ : t → F̃(t, 0) : f˜(0) = x̃0 ,

Using the uniqueness property of the covering paths, it follows F̃(t, 0) = f˜(t),
∀ t ∈ I. The restricted map F|(I × {0}) defines a path

f˜ : I → X̃ : t → F̃(t, 0) : f˜(0) = x̃0 .

Similarly, it follows that F̃(t, 1) = g̃(t). Again, the restricted map F|({0} × I)
defines a path

g̃ : I → X̃ : s → F̃(0, s) : g̃(0) = x̃0 .

It projects by p to the constant path at x0 = p(x0 ). A constant path s → x0 in X also

projects under p to the constant path s → x0 in X . Hence by uniqueness theorem
s → F̃(0, s) is a constant path based at x0 . Similarly, the path s → F̃(s, t) is a
constant path based at some point x̃1 ∈ p −1 (x0 ). This shows that H̃ : f˜ g̃ rel İ
and hence f˜ and g̃ are equivalent paths in X̃ . 
Corollary 5.4.8 Given a covering space ( X̃ , p) of X and x0 ∈ X, x̃0 ∈ p −1 (x0 ),
the induced homomorphism

p∗ : π1 ( X̃ , x̃0 ) → π1 (X, x0 )

is a monomorphism.
Proof Suppose that [ f˜], [g̃] ∈ π1 (X, x0 ) and [ f˜] = [g̃]. This implies that p∗ ([ f˜]) =
[ p ◦ f˜] and p∗ ([g̃]) = [ p ◦ g̃]. Hence it follows that

p ◦ f˜ p ◦ g̃ rel İ ⇔ f˜ g̃ rel İ

On the other hand

f˜  g̃ rel I˙ ⇔ p ◦ f˜  p ◦ g̃ rel İ,

because, otherwise, a contradiction would be arrived by Theorem 5.4.7. This proves

that p∗ is well defined and it is a monomorphism. 

5.4.2 Characterization of Path Liftings of Fibrations

This subsection studies liftings of fibrations and their fibers. For example, Theorem
5.4.9 characterizes path liftings of fibrations with the help of their fibers.
5.4 Homotopy and Path Lifting Properties of Covering Spaces 323

Theorem 5.4.9 A fibration p : X → B has the unique path lifting property (PLP)
iff its every fiber has no nonconstant paths.
Proof First suppose that p : X → B is a fibration with unique path lifting. Let b ∈ B
and f be a path in the fiber p −1 (b) and g be the constant path in p −1 (b) such that
g(0) = f (0). Then
p ◦ f = p ◦ g ⇒ f = g.

It shows that f is a constant path. Conversely, let p : X → B be a fibration such

that every fiber has no nontrivial path. If f and g are two paths in X such that
p ◦ f = p ◦ g and f (0) = g(0), then for every t ∈ I define a path h t in X by

f ((1 − 2t )t), 0 ≤ t ≤ 1/2
h t : I → X, t →
g((2t − 1)t), 1/2 ≤ t ≤ 1.

It implies that
(i) h t : I → X is a path in X from f (t) to g(t) for every t ∈ I and
(ii) p ◦ h t is a closed path in B, which is homotopic to the constant path at ( p ◦
f )(t) rel İ.
Since p has the HLP, there is a map

H : I × I → X : H (t , 0) = h t (t )

and H sends the subspace

{0} × I ∪ I × {1} ∪ {1} × I ⊂ I × I

to the fiber p −1 ( p f (t)). Since p −1 ( p f (t)) has no nonconstant paths, F maps 0 ×

I, I × 1 and 1 × I to a single point. Hence it follows that H (0, 0) = H (1, 0). This
proves that h t (0) = h t (1) and f (t) = g(t). 
Proposition 5.4.10 Given a pointed topological space X with base point x0 , let
P(X ) be the space of all paths in X starting at x0 and endowed with compact open
topology. Then the map
p : P(X ) → X, β → β(1)

is fibration with fiber (X ).

Proof Given an arbitrary topological space Y and a pair of continuous maps

f : Y → P(X ) and H : Y × I → X with H0 = H ( , 0) = p ◦ f : Y → X,

define

( f (y))(s(t + 1)), 0 ≤ s ≤ t+1
1
G : Y × I × I → X, (y, t, s) →
H (y, s(t + 1) − 1), t+1
1
≤ s ≤ 1.
324 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Then G is a continuous map such that it defines a map

F : Y × I → X I : F(y, t)(0) = f (y)(0) = x0 , ∀ y ∈ Y, t ∈ I.

Hence it follows that

F ∈ P(X ) and F(y, 0)(s) = f (y)(s), ∀ y ∈ Y, ∀ s ∈ I.

This asserts that F0 = F( , 0) = f and p ◦ F = H . This proves that

(i) F is the required lifting of H and
(ii)
p −1 (x0 ) = {β ∈ P(x) : β(1) = x0 } = (X ). 

5.4.3 General Lifting Problems from Homotopy Viewpoint

It is a natural problem: given a covering space ( X̃ , p) of X and a continuous map

f : A → X̃ , does there exist a continuous map f˜ : A → X̃ such that p ◦ f˜ = f ?
The answer is negative. In support consider the exponential map

p : R → S 1 , t → e2πit

studied in Example 5.4.11. This subsection studies lifting problems of arbitrary

continuous maps with the help of fundamental groups and solves a lifting problem
in Theorem 5.4.12.

Example 5.4.11 The exponential map

p : R → S 1 , t → e2πit

is a covering projection. The identity map 1 S 1 : S 1 → S 1 cannot be lifted to a con-

tinuous map ψ = 1˜S 1 : S 1 → R such that p ◦ ψ = 1 S 1 , which makes the triangle
in Fig. 5.17 commutative. Because, p ◦ ψ = 1 S 1 would imply that ψ is an embed-
ding of S 1 into R. Since S 1 is compact, ψ(S 1 ) must be a compact connected subset
X of R such that it would be homeomorphic to S 1 under present situation by the
embedding ψ. But it is not possible, because the fundamental groups π1 (X ) = 0,
π1 (S 1 ) = Z and π1 is a topological invariant. This example asserts that the identity
map 1 S 1 : S 1 → S 1 cannot be lifted to a continuous map ψ = 1˜S 1 : S 1 → R.

Theorem 5.4.12 provides a necessary and sufficient condition under which an

arbitrary continuous map f : A → X can be lifted, which is proved by using the
tools of homotopy theory.
5.4 Homotopy and Path Lifting Properties of Covering Spaces 325

Fig. 5.17 Covering

projection for exponential
map p

Fig. 5.18 Lifting of f to f˜

Theorem 5.4.12 (Lifting Theorem) Given a covering space ( X̃ , p) of X and a

connected and locally path-connected space A, if f : A → X is any continuous map,
then for arbitrary three points a0 ∈ A, x0 ∈ X and x̃0 ∈ X such that f (a0 ) = x0 and
p(x̃0 ) = x0 , there exists a unique continuous map

f˜ : A → X̃

such that
(i) f˜(a0 ) = x̃0 and
(ii) p ◦ f˜ = f if and only if f ∗ (π1 (A, a0 )) ⊂ p∗ (π1 ( X̃ , x̃0 )).

Proof Let there exist a continuous map f˜ : A → X̃ such that the given conditions
hold. Then the triangle in Fig. 5.18 is commutative, and hence by using the functorial
property of the fundamental group functor π1 , it follows that the triangle in Fig. 5.19
is also commutative. This asserts that

f ∗ (π1 (A, a0 )) = p∗ ( f˜∗ (π1 (A, a0 ))) ⊂ p∗ (π1 ( X̃ , x̃0 )).

Conversely, let the algebraic property f ∗ (π1 (A, a0 )) ⊂ p∗ (π1 ( X̃ , x̃0 )) hold. Since
by hypothesis, A is connected, it has only one component. Again, since A is locally
path connected, this component is a path component, and hence it follows that A is
path connected. For any point a ∈ A, construct a path

α : I → A : α(0) = a0 and α(1) = a.

Hence it follows that f ◦ α : I → X is a path such that

( f ◦ α)(0) = f (α(0)) = f (a0 ) = x0 .

By path Lifting Property 5.4.3, there is a unique path α̃ : I → X̃ which lifts f ◦ α

in X̃ with α̃(0) = x̃0 ( see Fig. 5.16). Define a map
326 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.19 Lifting problem

for f and f˜ converted in
homotopy theory

Fig. 5.20 Diagram

corresponding to lifting
Theorem 5.4.12

f˜ : A → X, a → α̃(1).

Clearly, f˜ is well-defined. To prove the continuity, take any point a ∈ A and an open
nbd U of f˜(a). We now find an open nbd Wa of a with f˜(Wa ) ⊂ U. We take an
open admissible nbd W of p f˜(a) = f (a) with the property that W ⊂ p(U ). Let C
be the path component of p −1 (W ) which contains the point f˜(a), and let W be an
open admissible nbd of f (a) such that

W ⊂ p(U ∩ C).

Then the path component of p −1 (W ) containing the point f˜(a) is contained in U.

Since f is continuous and the path-connected set A is locally connected, there is a
path connected nbd Wa of a such that f (Wa ) ⊂ V . This implies that f˜(Wa ) ⊂ U
(Fig. 5.20). 

Corollary 5.4.13 Suppose that A is simply connected and locally path connected
and f : (A, a0 ) → (X, x0 ) is continuous. If (X, p) is a covering space of X and if
x̃0 ∈ p −1 (x0 ), then f has the unique lifting

f˜ : (A, a0 ) → ( X̃ , x̃0 ).

Proof By hypothesis A is simply connected. Hence π1 (A, a0 )=0 and p∗ π1 (A, a0 ) =

{0} ⊂ p∗ π1 ( X̃ , x̃0 ). This implies that there is a unique lifting f˜ : (A, a0 ) → ( X̃ , x̃0 )
of f . 

Corollary 5.4.14 follows immediately from above discussion.

Corollary 5.4.14 Let X be a connected and locally path-connected space, and

( X̃ , p) and (Y, q) be two covering spaces of the same base space X. Let x0 ∈ X
and x̃0 ∈ X̃ , y0 ∈ Y be base points such that

p(x̃0 ) = x0 = q(y0 ).
5.4 Homotopy and Path Lifting Properties of Covering Spaces 327

If p∗ π1 ( X̃ , x̃0 ) = q∗ π1 (Y, y0 ), then there exists a unique continuous map

f : (Y, y0 ) → ( X̃ , x̃0 ) : p ◦ f = q,

i.e., f satisfies the property: p ◦ f = q.

Example 5.4.15 (S n , p) is a covering space of the real projective space R P n of

multiplicity 2. Because, the n-sphere S n is simply connected for n ≥ 2, and hence it
follows that if x̃0 ∈ p −1 (x0 ), then x0 ∈ R P n . Hence it follows that for any continuous
map
f : (S n , s0 ) → (R P n , x0 ),

there exists the unique lifting

f˜ : (S n , s0 ) → (S n , x̃0 ).

5.4.4 Homotopy Classification of Covering Spaces

This subsection completely classifies covering spaces of a fixed base space with the
help of group theory and characterizes such covering spaces in terms of conjugate
subgroups of the fundamental groups of base spaces.
We use in this subsection the following results of group theory .
(i) Two subgroups A and B of a group G are conjugate subgroups iff A = g −1 Bg
for some g ∈ G.
(ii) For the subgroups A and B of a group G, the G-sets G/A and G/B are G-
isomorphic iff A and B are conjugate subgroups of G.

Theorem 5.4.16 Given a path-connected space X, let ( X̃ , p) be a covering space of

the base space X, where X̃ is also path connected. If x0 ∈ X, then the family of groups
p∗ π1 ( X̃ , z), as z runs over F = p −1 (x0 ), forms a conjugacy class of subgroups of
π1 (X, x0 ).

Proof It is proved in two steps:

(i) Step I For any two points z 0 , z 1 ∈ F, the subgroups p∗ π1 ( X̃ , z 0 ) and p∗ π1 ( X̃ , z 1 )
of π1 (X, x0 ) are conjugate.
(ii) Step II For any subgroup H of π1 (X, x0 ) conjugate to the subgroup p∗ π1
( X̃ , z 0 ), there exists some z ∈ F such that

H = p∗ π1 ( X̃ , z).

(i) Given a path w : I → X̃ , from z 0 to z 1 , define a map

βw : π1 ( X̃ , z 0 ) → π1 ( X̃ , z 1 ), [ f ] → [w̄ ∗ f ∗ w].
328 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

It is an isomorphism of groups (see Chap. 2). Hence it follows that

( p∗ ◦ βw )π1 ( X̃ , z 0 ) = [ p ◦ w]−1 p∗ π1 ( X̃ , z 0 ) [ p ◦ w].

which asserts that p∗ π1 ( X̃ , z 1 ) and p∗ π1 ( X̃ , z 0 ) are conjugate subgroups of

π1 (X, x0 ),
(ii) By hypothesis, H is a subgroup of π1 (X, x0 ) conjugate to the subgroup
p∗ π1 ( X̃ , z 0 ), Then there exists some [k] ∈ π1 (X, x0 ) such that

H = [k]−1 p∗ π1 ( X̃ , z 0 )[k].

Suppose that the unique lifting k̃ of k in X̃ starting at z 0 ends at the point

k̃(1) = z ∈ X̃ . Now proceeding as in (i), it follows that

p∗ π1 ( X̃ , z) = [ p ◦ k̃]−1 p∗ π1 ( X̃ , z 0 )[ p ◦ k̃] = [k]−1 p∗ π1 ( X̃ , z 0 )[k] = H.

This proves that the family { p∗ π1 ( X̃ , z) : z ∈ F} constitutes a complete conju-

gate class of subgroups of the group π1 (X, x0 ).

Definition 5.4.17 The family { p∗ π1 ( X̃ , z) : z ∈ F} constituting conjugate class of
subgroups of the group π1 (X, x0 ) given in Theorem 5.4.16 is known as the conjugate
class determined by the covering space ( X̃ , p) of X.
Theorem 5.4.18 characterizes covering spaces of a base space X in terms of
conjugacy classes of subgroups of the fundamental group π1 (X ), and hence every
conjugacy class of a subgroup of π1 (X, x) determines uniquely the covering spaces
of X up to isomorphism (Fig. 5.22).
Theorem 5.4.18 Given a path connected and locally path-connected space X, let
( X̃ , p) and (Ỹ , q) be two path connected covering spaces of X such that x̃0 ∈ p −1 (x0 )
and ỹ0 ∈ q −1 (x0 ), where x0 ∈ X. Then the covering spaces ( X̃ , p) and (Ỹ , q) are
isomorphic iff they determine the same conjugacy class of subgroups of π1 (X, x0 )),
i.e., iff
p∗ π1 ( X̃ , x0 ) and q∗ π1 (Ỹ , y0 )

are conjugate subgroups of π1 (X, x0 ).

Proof First suppose that the covering spaces ( X̃ , p) and (Ỹ , q) are isomorphic with
a homeomorphism h : Ỹ → X̃ such that the triangle in Fig. 5.21 is commutative.
Hence p ◦ h = q.
If h( ỹ0 ) = x̃1 , then h induces an isomorphism

h ∗ : π1 (Ỹ , ỹ0 ) → π1 ( X̃ , x̃1 ).

Consequently,
5.4 Homotopy and Path Lifting Properties of Covering Spaces 329

Fig. 5.21 Isomorphisms

between covering spaces

h ∗ (π1 (Ỹ , ỹ0 )) = π1 ( X̃ , x̃1 ).

This implies that

( p∗ ◦ h ∗ )(π1 (Ỹ , ỹ0 )) = p∗ (π1 ( X̃ , x̃1 )).

This asserts that

q∗ (π1 (Ỹ , ỹ0 )) = p∗ π1 ( X̃ , x̃1 ).

Hence p∗ π1 ( X̃ , x̃1 ) is a subgroup of π1 (X, x0 ), and it is conjugate to the subgroup

p∗ π1 ( X̃ , x̃0 ) by Theorem 5.4.16. This shows that p∗ π1 ( X̃ , x̃0 ) and q∗ π1 (Ỹ , ỹ0 ) are
conjugate subgroups of π1 (X, x0 ). Conversely, let the two subgroups of π1 ( X̃ , x̃0 ) be
conjugate. By Theorem 5.4.16 we can choose a different base point ỹ0 in Ỹ such that
the two groups are equal. We now consider the lifting of p in the diagram in Fig. 5.22,
where q is a covering map. Since by hypothesis, the space X is path connected; it is
also locally path connected. Now, p∗ π1 ( X̃ , x̃0 ) ⊂ q∗ π1 (Ỹ , ỹ0 ) and these two groups
subgroups of π1 (X, x0 ) are equal. Using Theorem 5.4.16 the map p can be lifted to
p̃ : X → Ỹ such that p̃(x̃0 ) = ỹ0 . This implies yhat q ◦ p̃ = p.
If we now reverse the role of X̃ and Ỹ , then the map q : Ỹ → X can also be lifted
to q̃ : Ỹ → X̃ such that q̃( ỹ0 ) = x̃0 as shown in Fig. 5.23.
To show that p̃ and q̃ in Fig. 5.24 are inverses of each other, consider the diagram
in Fig. 5.25.
Consider the two liftings of p described below:
(i) q̃ ◦ p̃ : X̃ → X̃ is a lifting of the map p : X̃ → X such that (q̃ ◦ p̃)(x̃0 ) = x̃0
and
(ii) The identity map 1 X̃ : X̃ → X̃ is another such lifting of p.
This asserts by uniqueness property of lifting that q̃ ◦ p̃ = 1 X̃ . Proceeding in a
similar way, it follows that p̃ ◦ q̃ = 1Ỹ . Hence it is proved that
(i) p̃ : X̃ → Ỹ is a homeomorphism with its inverse q̃ : Ỹ → X̃ and
(ii) the covering spaces ( X̃ , p) and (Ỹ , q) are isomorphic. 

Example 5.4.19 (Complete classification of covering spaces of S 1 ) Consider the

covering spaces of S 1 . Since the fundamental group π1 (S 1 , 1) ∼ = Z is abelian, its
two subgroups are conjugate iff they are equal. Hence, two covering spaces of S 1
are isomorphic iff their corresponding subgroups of π1 (S 1 ) are same. The subgroups
in Z are the subgroups < n >, consisting of precisely all the multiples of n, for
330 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.22 Lifting of p

Fig. 5.23 Lifting of q

Fig. 5.24 Liftings of p

and q

Fig. 5.25 Lifting of p

involving p̃ and q̃

n = 0, 1, 2, . . . . Since R is simply connected, the subgroup corresponding to the

usual covering space (R, p) of S 1 is the trivial subgroup of Z. On the other hand, if
p; S 1 → S 1 , z → z n , then the subgroup corresponding to the covering space (S 1 , p)
of S 1 is the subgroup < n > of Z. This asserts that every path-connected covering
space of S 1 is isomorphic to one of these coverings. This asserts that any covering
space of S 1 is isomorphic either
(i) to (R, p) or
(ii) to one of the coverings (S 1 , qn ), where qn : S 1 → S 1 : z → z n , which geomet-
rically wraps S 1 around itself n times.

Example 5.4.20 The double covering (S 2 , p) of R P 2 determines the conjugacy

class contains only the trivial subgroup. Because, S 2 is simply connected, and hence
π1 (S 2 , s) = 0.

Example 5.4.21 The conjugacy class determined by the covering space (R2 , r ) of
the torus contains only the trivial subgroup. Because, the plane R2 is simply con-
nected.

Example 5.4.22 For the infinite spiral X , the projection map

5.4 Homotopy and Path Lifting Properties of Covering Spaces 331

Fig. 5.26 Homotopy

diagram corresponding to f˜0

q : X → S1

obtained by projecting each point on X to the point on the circle directly below it is
covering space of S 1 . Since X is contractible, π1 (X ) = 0 and hence (X, q) determines
the conjugacy class of π1 (S 1 ) consisting of only the trivial subgroup. Again, for
the exponential map p : R → S 1 , the covering space (R, p) also determines the
conjugacy class of π1 (S 1 ) consisting of only the trivial subgroup. This asserts that
the covering spaces (X, q) and (R, p) of S 1 are isomorphic by Theorem 5.4.18.

Proposition 5.4.23 Given a covering map p : ( X̃ , x̃0 ) → (X, x0 ),

(i) its induced homomorphism

p∗ : π1 ( X̃ , x̃0 ) → π1 (X, x0 )

is a monomorphism and
(ii) the subgroup p∗ (π1 ( X̃ , x̃0 )) ⊂ π1 (X, x0 ) has elements consisting of homotopy
classes of loops in X based at x0 which lift to X̃ starting at the point x̃0 ∈ X̃ are
loops.

Proof Given an element α ∈ ker p∗ represented by a loop f˜0 : I → X̃ with a homo-

topy Ht : I → X of f 0 = p ◦ f˜0 to the trivial loop f 1 , there exists a lifted homotopy
of loop Ht : I → X̃ ( see Fig. 5.26 ) started at f˜0 and ending at a constant loop, since
the lifted homotopy H t is a homotopy of paths fixing the end points and t varies each
point of Ht gives a path lifting a constant path, which is clearly a constant path. This
implies that [ f˜0 ] = 0 ∈ π1 ( X̃ , x̃0 ) and hence p∗ is a monomorphism. 

Remark 5.4.24 If p : X̃ → X is a covering map, then p is also onto. But its induced
homomorphism
p∗ : π1 ( X̃ , x̃0 ) → π1 (X, x0 )

needs not be an epimorphism. However, p∗ is a monomorphism.

Proposition 5.4.25 Given a path-connected space X̃ and a covering projection

p : ( X̃ , x̃0 ) → (X, x0 )

(i) there is a surjective map ψ : π1 (X, x0 ) → p −1 (x0 ).

(ii) Moreover, if X̃ is simply connected, then ψ is bijective.
332 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Proof It can be proved by using the technique for computation of π1 (S 1 , 1) described

in Chap. 3. 

5.4.5 Isomorphism Theorem and Computation of πn (S1 ) = 0

for n ≥ 2

This subsection proves that πn (S 1 ) = 0 for n ≥ 2 by applying Theorem 5.4.26 say-

ing that every covering space induces an isomorphism between the higher homotopy
group of its total space and that of its base space at every dimension n ≥ 2.

Theorem 5.4.26 If ( X̃ , p) is a covering space of X such that

(i) x0 ∈ X and
(ii) x̃0 ∈ p −1 (x0 )
then p induces an isomorphism

p∗ : πn ( X̃ , x̃0 ) → πn (X, x0 ), ∀ n ≥ 2.

Proof By hypothesis, p : X̃ → X is a covering map with p(x̃0 ) = x0 . Let σ̃ ∈

πn ( X̃ , x̃0 ) be an arbitrary element and let it is represented by a continuous map
f˜ : (In , ∂In ) → ( X̃ , x̃0 ). Then

σ̃ = [ f˜] ∈ πn ( X̃ , x̃0 ) = [S n , X̃ ].

Consequently, p induces a homomorphism

p∗ : πn ( X̃ , x̃0 ) → πn (X, x0 ), [ f˜] → [ p ◦ f˜].

p∗ is a monomorphism : Let σ̃ and τ̃ be two elements of πn ( X̃ , x̃0 ) such that

p∗ (σ̃) = p∗ (τ̃ ). If they are represented by f˜ and g̃, respectively, where f˜, g̃ : S n →
X̃ are pointed maps, then [ p ◦ f˜] = [ p ◦ g̃]. This implies that p ◦ f˜ p ◦ g̃. Hence
it follows by covering homotopy theorem that f˜ g̃. This implies [ f˜] = [g̃] and
hence it is proved that p∗ is a monomorphism.
p∗ is an epimorphism: Let σ = [ f ] ∈ πn (X, x0 ) be an arbitrary element. Consider
the triangle as shown in diagram in Fig. 5.27.
Then there exists a unique lifting f˜ : S n → X̃ of f : S n → X such that p ◦ f˜ =
f , because S n is simply connected for every integer n ≥ 2. This proves that there
exists an element σ̃ ∈ πn ( X̃ , x̃0 ) such σ̃ = [ f˜]. This implies that p∗ [ f˜] = [ f ] = σ
and hence it is proved that p∗ is an epimorphism. 

As an immediate application of Theorem 5.4.26, it is proved in Theorem 5.4.27 that

πn (S 1 ) = 0 for every integer n ≥ 2.
5.4 Homotopy and Path Lifting Properties of Covering Spaces 333

Fig. 5.27 Lifting of f to f˜

Theorem 5.4.27 πn (S 1 ) = {0} for every integer n ≥ 2.

Proof Consider the well-known covering map

p : R → S 1 , t → e2πit

and apply Theorem 5.4.26. Then p induces an isomorphism

p∗ : πn (R) → πn (S 1 ), ∀ n ≥ 2.

Since R is a contractible space, it follows that

πn (R) = {0}, ∀ n ≥ 2.

This proves that πn (S 1 ) = {0} for every n ≥ 2. 

5.5 Galois Correspondence on Covering Spaces

This section continues to study the classification problem of covering spaces over a
fixed base space X. The main tool of this classification is provided by the Galois cor-
respondence between a special family of covering spaces of X and subgroups of the
fundamental group of the base space π1 (X ). The Galois correspondence ψ is derived
from the map that assigns the subgroup p∗ (π1 ( X̃ , x̃0 )) of π1 (X, x0 ) corresponding
to every covering space p : ( X̃ , x̃0 ) → (X, x0 ) under certain conditions prescribed
in Theorem 5.5.6. The Galois correspondence ψ is both injective and surjective.

Theorem 5.5.1 Given a path connected, locally path connected and semilocally
path space X and a subgroup H of π1 (X, x0 ) there is a covering space p : X H → X
with
p∗ (π1 (X H , x̃0 )) = H

for some choice of the base point x̃0 ∈ X H .

Proof By hypothesis X is a semilocally path-connected space. Hence for every point

x ∈ X,
(i) there is a nbd Wx of x such that each closed path in Wx based at x is nullhomotopic
in X by its semilocality property and
334 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(ii) there is an open connected nbd Ux of x with

x ∈ Ux ⊂ Wx

by locally path connectedness property of X.

Then every closed path in Ux at x is null- homotopic in X , and Ux is evenly
covered by p.
Construction of X H : Let P(X, x0 ) denote the family of all paths f in X with
f (0) = x0 and topologized by the compact open topology. Define an equivalence
relation ‘∼’on P(X, x0 ) by the rule

f ∼ g mod H iff f (1) = g(1) and [ f ∗ g −1 ] ∈ H.

Let X H be the set of all such equivalence classes [f], topologized by the quotient
topology. For the constant loop c ∈ P(X, x0 ), take x̃0 = [c] ∈ X H and define

p : X H → X, [ f ] → f (1).

Then p(x̃0 ) = x0 . Since any two paths in the basic nbds U[ f ] and U[g] are identified
in X H , the whole nbds are identified. This asserts that the natural projection

p : XH → X

is a covering space with p(x̃0 ) = x0 . Since any loop β in X based at x0 has its lifting
to X H starting at x̃0 ends at [β] and the image of this lifted path in X H is a loop iff
[β] ∼ [c0 ] (equivalently, [β] ∈ H ) the image of

p∗ : π1 (X H , x̃0 ) → π1 (X, x0 )

coincides with H . This concludes that p∗ (π1 (X H , x̃0 )) = H for some choice of the
base point x̃0 ∈ X H . 
Corollary 5.5.2 Under the condition of Theorem 5.5.1, every subgroup G of the
fundamental group π1 (X, x0 , ) can be realized as the fundamental group of the topo-
logical space X G .
Proof Let G be a subgroup of the fundamental group π1 (X, x0 ). Consider the topo-
logical X G constructed in the proof of Theorem 5.5.1. Then the corollary follows
from Theorem 5.5.1. 
Corollary 5.5.3 For any connected, locally path connected, semilocally simply con-
nected space X, every covering space q : Y → X is isomorphic (equivalent) to a
covering spaces of the form p : X G → X for some group G.
Proof Let x0 ∈ X be a base point of X and y0 ∈ Y lie in the fiber q −1 (x0 ) over x0 . If
G = q∗ π1 (Y, y0 ), then p∗ π1 (X G , x0 ) = G. Hence it follows Theorem 5.5.1 shows
that the covering spaces p : X G → X and q : Y → X are isomorphic. 
5.5 Galois Correspondence on Covering Spaces 335

Fig. 5.28 Triangular

diagram for two isomorphic
coverings of X

Corollary 5.5.4 Given a connected, locally path connected, semilically simply con-
nected space X, if p : X̃ → X is a covering space of X , then every open contractible
set U in X is evenly covered by p.
Corollary 5.5.5 Given a connected, locally path connected, semilically simply con-
nected space X, it has a universal covering space X̃ (i.e., X̃ is simply connected ) iff
X̃ is semilocally simply connected.
Theorem 5.5.6 characterizes specified isomorphic coverings in terms of funda-
mental groups.
Theorem 5.5.6 (Classification theorem) Given a path connected and locally path-
connected space X, the two coverings p : X̃ → X and q : Ỹ → B are isomorphic
through a homeomorphism f : X̃ → Ỹ sending a base point x̃0 ∈ p −1 (x0 ) to a base
point ỹ0 ∈ q −1 (x0 ) iff
p∗ (π1 ( X̃ , x̃0 )) = q∗ (π1 (Ỹ , ỹ0 )).

Proof First suppose that there exists a homeomorphism f : ( X̃ , x̃0 ) → (Ỹ , ỹ0 ) mak-
ing the triangular diagram as shown in Fig. 5.28 commutative. Then p = q ◦ f and
q = p ◦ f −1 assert that p∗ (π1 ( X̃ , x̃0 )) = q∗ (π1 (Ỹ , ỹ0 )).
Conversely, let p∗ (π1 ( X̃ , x̃0 )) = q∗ (π1 (Ỹ , ỹ0 )). Then by the lifting property,
(i) the map p is lifted to p̃ : ( X̃ , x̃0 ) → (Ỹ , ỹ0 ) such that

q ◦ p̃ = p;

(ii) the map q is also lifted to q̃ : (Ỹ , ỹ0 ) → ( X̃ , x̃0 ) such that

p ◦ q̃ = q.

Then by unique lifting property, it follows that p̃ ◦ q̃ = 1d and q̃ ◦ p̃ = 1d ,

because each of these composite lifts fix the base points. This asserts that the induced
homomorphism p∗ is an isomorphism with the induced homomorphism q∗ its inverse
isomorphism. 
Theorem 5.5.7 gives a generalization of the classification Theorem 5.5.6.
Theorem 5.5.7 (Generalized classification theorem) Given a path connected,
locally path connected and semilocally simply connected space X, there is a bijection
between the set of base point preserving isomorphism classes of path-connected cov-
ering spaces p : ( X̃ , x̃0 ) → (X, x0 ) and the set of subgroups of π1 ( X̃ , x˜0 ), obtained
336 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

by assigning the subgroups p∗ (π1 ( X̃ , x̃0 )) to the corresponding covering spaces of

(X, x0 ). If the base points are ignored, this correspondence gives a bijection between
isomorphism classes of path connected covering and locally path-connected space
X, the covering spaces p : X̃ → X and conjugacy classes of subgroups of π1 (X, x0 ).

Proof By using Theorem 5.5.6 the first part follows. To prove the second part, let
x̃0 , x̃1 ∈ p −1 (x0 ) be two base points of X̃ and

f˜ : I → X̃ : f˜(0) = x̃0 , f˜(1) = x̃1 .

Then p ◦ α̃ determines a loop f in X , and hence it represents an element, say g ∈

π1 (X, x0 ). For the pair of subgroups H1 = p∗ (π1 ( X̃ , x1 )) and H0 = p∗ (π1 ( X̃ , x0 ))
of the group G = π1 (X, x0 ), there is an inclusion

g −1 H0 g ⊂ H1 ,

f −1 ∗ f˜ ∗ 
because, given a loop f˜ at x̃0 , the path  f is a loop at x̃1 . It follows similarly
that

g H1 g −1 ⊂ H0 .

Using conjugation by g −1 it follows that

H1 ⊂ g −1 H0 g.

Hence it follows that

g −1 H0 g = H1 .

The above discussion asserts that a change of the base point from x̃0 to x̃1 produces a
change from H0 to the conjugate subgroup H1 = g −1 H0 g. Conversely, correspond-
ing to a change H0 to a conjugate subgroup H1 = g −1 H0 g in G, take a loop β
representing g, which lifts to a path β̃ starting at x̃0 and ending at x̃1 = β̃(1). Hence
it follows as above that
H1 = g −1 H0 g. 

Theorem 5.5.8 (Galois correspondence) Given be path connected and locally path-
connected space X, the Galois correspondence ψ arising from the function that
assigns to each covering space p : ( X̃ , x̃0 ) → (X, x0 ) the subgroup p∗ (π1 ( X̃ , x˜0 ))
of π1 (X, x0 ) is a bijection.

Proof Let X be be a path connected and locally path-connected topological space and
C( X̃ , x̃0 ) be the set of all covering spaces ξ p = p : ( X̃ , x̃0 ) → (X, x0 ) and SG be the
set of all subgroups of π1 (X, x0 ) of the form p∗ (π1 ( X̃ , x˜0 )). Define a correspondence

ψ : C( X̃ , x̃0 ) → SG, ξ p → p∗ (π1 ( X̃ , x˜0 )).

5.6 Homotopy Property of Universal Covering Spaces 337

Then ψ is a well-defined set function. By using Proposition 5.4.23 it follows that

ψ is injective. Again, since to each subgroup G of π1 (X, x0 ), there is a covering
space p : ( X̃ , x˜0 ) → (X, x0 ) with p∗ π1 ( X̃ , x̃0 ) = G, it follows by using classifica-
tion Theorem 5.5.6 that ψ is surjective. This shows that the correspondence ψ is a
bijection. 
Definition 5.5.9 ψ defined in Theorem 5.5.8 is called a Galois correspondence.

5.6 Homotopy Property of Universal Covering Spaces

This section is devoted to the study universal covering spaces ( X̃ , p) of the base
space (X, x0 ). This study is based on homotopy theory and proves that all universal
covering spaces of the same base space are isomorphic and establishes their relations
with base spaces X by considering the conjugacy class of the trivial subgroup {0} of
π1 (X, x0 ).

Definition 5.6.1 Given a topological space X, a covering space ( X̃ , p) of X, where

X is simply connected, ( if it exists in the sense that the space X̃ is path connected
and π1 ( X̃ , x̃0 ) = {0} for every x̃0 ∈ X̃ ) ) is called a universal covering space of X.
Example 5.6.2 (i) For p : R → S 1 , t → e2πt , the covering space (R, p) is a uni-
versal covering space of S 1 , because the space R is simply connected.
(ii) If pn : S n → R P n is the map which identifies the antipodal points of S n , then
the covering space
(S n , pn )

is universal covering space of R P n .

Theorem 5.6.3 asserts that all universal covering spaces of the same base space
are isomorphic proving its uniqueness up to isomorphism of the same base space,
and hence it is called the ‘universal covering space’. Consequently, henceforth a
universal covering space of a base space is called the universal covering space.
Theorem 5.6.3 (i) All universal covering spaces of the same base space are iso-
morphic.
(ii) Given the universal covering space ( X̃ , p) of X and a covering space (Ỹ , q) of
X , there exists a continuous map

p̃ : X̃ → Ỹ

with the property that ( X̃ , p̃) is a covering space of Ỹ .

Proof (i) All universal covering spaces of the same base space X are isomorphic,
because any universal covering space of X, determines the conjugacy class of
the trivial subgroup of π1 (X, x0 ) by Theorem 5.4.18.
338 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.29 Lifting p̃ of p

(ii) To prove it, consider the commutative triangle in Fig. 5.29, and take arbitrary
base points x̃0 ∈ X̃ , ỹ0 ∈ Ỹ and x0 ∈ X with the property

p(x̃0 ) = q( ỹ0 ) = x0 .

Clearly,
p∗ π1 ( X̃ , x̃0 ) ⊂ q∗ π1 (Ỹ , ỹ0 ),

because π1 (( X̃ , x̃0 )) = {0}. The existence of a continuous map p̃ : ( X̃ , x̃0 ) →

(Ỹ , ỹ0 ) with the property
q ◦ p̃ = p

is proved by using the Lifting Theorem 5.4.12. This shows that p̃ is a covering
projection and hence ( X̃ , p̃) is a covering space of Ỹ . 

5.7 Homotopy Properties of Vector Bundles and Their

Homotopy Classification

This section continues the study of vector bundles initiated in Chap. 4. The present
study is devoted in homotopy properties of vector bundles and proves two theorems
on the homotopy classification of vector bundles formulated in Theorem 5.7.8 and
Corollary 5.7.10.
The strong structure of vector bundles facilitates to prove that there exists a natural
bijective correspondence between the set of isomorphism classes of n-dimensional
vector bundles over a paracompact space B and the set of homotopy classes of
maps from B into a Grassmann manifold of n-dimensional subspaces in an infinite-
dimensional space. This result leads to define topological K-theory by Grothendick in
1961. This theory closely relates algebraic topology with several other branches such
as algebraic geometry, analysis, ring theory and number theory. Interested readers
are referred to the books [Adhikari, 2016] and [Husemoller, 1966].
5.7 Homotopy Properties of Vector Bundles and Their Homotopy Classification 339

Fig. 5.30 Commutative

diagram for the induced
bundle (Y, q, A)

5.7.1 Homotopy Properties of Vector Bundles

Definition 5.7.1 (Induced bundle) Let ξ = (X, p, B) be a vector bundle and f :

A → B be a a continuous map from a topological space A. Then the induced vector
bundle f ∗ (ξ) = (Y, q, A) of ξ over A under f is the vector bundle (Y, q, A), where
the total space Y is defined by

Y = {(a, x) ∈ A × X : f (a) = p(x)}

and the projection q is defined by

q : Y → A, (a, x) → a.

If g : Y → X, (a, x) → x, then the pair of maps (g, f ) : (Y, q, A) → (X, p, B)

make the diagram in Fig. 5.30 commutative. The pair of maps

g : Y → X, (a, x) → x and f : A → B

form a bundle morphism (g, f ) : f ∗ (ξ) → ξ called the canonical morphism of

the induced bundle. The map q is a fiber preserving map.

Theorem 5.7.2 proves that for any vector bundle ξ over a paracompact base space
A, any pair of homotopic maps f, g : B → A induce B-isomorphic vector bundles,
which is an important homotopy property of vector bundles.

Theorem 5.7.2 Given a paracompact space B, a vector bundle ξ over A, and a

pair of homotopic maps f, g : B → A, there exists a B-isomorphism between the
induced bundles f ∗ (ξ) and g ∗ (ξ) over the base B.

Proof By hypothesis, B is paracompact, ξ is a vector bundle over A and f, g : B →

A is a pair of homotopic maps. Then there exists a homotopy

G : B × I → A : G(x, 0) = f (x), G(x, 1) = g(x), ∀ x ∈ B.

Hence it follows from Exercise 5.24.1 of Sect. 5.24 that

(i) the induced vector bundles f ∗ (ξ) and G ∗ (ξ)|(B × {0}) are B-isomorphic and
(ii) the induced bundles g ∗ (ξ) and G ∗ (ξ)|(B × {1}) are B-isomorphic.
Finally, since there exists a B-isomorphism
340 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(α, β) : G ∗ (ξ) : (B × {0}) → G ∗ (ξ)|(B × {1})

by using the same exercise, it is proved that the induced vector bundles f ∗ (ξ) and
g ∗ (ξ) are B-isomorphic. 

Corollary 5.7.3 Let B be a contractible paracompact space. Then every vector

bundle over B is trivial.

Proof By hypothesis, B is a contractible paracompact space. By using the con-

tractibility property of B, it follows that the identity map 1 B : B → B is homotopic
to a constant map c : B → B. If ξ is an n-dimensional vector bundle over B, then it
follows that
(i) 1∗B (ξ) and ξ are B-isomorphic and
(ii) f ∗ (ξ) is B-isomorphic to the product bundle (B × Fn , p, B).
Finally, since the maps 1 B c, by using Theorem 5.7.2, it follows that the vector
bundle ξ is B-isomorphic to the product bundle (B × F n , p, B) which is a trivial
bundle. 

5.7.2 Homotopy Classification of Vector Bundles

This subsection solves the homotopy classification problems of vector bundles by

proving two basic results formulated in Theorem 5.7.8 and in Corollary 5.7.10. One
result says that there is a bijective correspondence between isomorphism classes of
n-dimensional vector bundles over a paracompact space B and the homotopy classes
of maps from B to Grassmann manifold G n (F ∞ ) and the other one says that every
n-dimensional F-vector bundle over a paracompact space B is isomorphic to the
vector bundle induced by a continuous map from the base space B to the Grassmann
manifold G n (F∞ ).
This section uses the following notations
(i) F stands for R, C or H.
(ii) G n (Fm ) stands for Grassman manifold of n-dimensional subspaces of the vector
space Fn .
(iii) H stands for the category of paracompact spaces and their homotopy classes.
(iv) Set stands for the category of sets and their functions.
(v) Vectn (B) stands for the set of B-isomorphic classes of n-dimensional vector
bundles over B.
(vi) For an n-dimensional vector bundle ξ over B, the family {ξ} stands for the B-
isomorphism classes in Vectn (B) of ξ, and [ f ] stands for the homotopy class of
f : A → B between two paracompact spaces A and B.
Theorem 5.7.4 proves the functorial property of Vect n (−).
5.7 Homotopy Properties of Vector Bundles and Their Homotopy Classification 341

Theorem 5.7.4 Vectn : H → Set is a contravariant functor from the category of

paracompact spaces and their homotopy classes to the category of sets and set
functions .

Proof The assignment B → Vectn (B) for every object B in H defines the object
function

Vectn : H → Set, B → Vectn (B)

and the assignment {ξ} → { f ∗ (ξ)} for every morphism [ f ] in H, where A, B are
paracompact spaces and f : A → B is a continuous map, defines the morphism
function

Vectn ([ f ]) : Vectn (B) → Vectn (A) : {ξ} → { f ∗ (ξ)}.

The above functions are well-defined by Theorem 5.7.2. Moreover, they satisfy
the properties
(i) For the identity map 1 B : B → B, the vector bundles 1∗B (ξ) and ξ are B-
isomorphic.
(ii) If [g] denotes the homotopy class of g : C → A between the paracompact spaces
C and A, then the induced vector bundles g ∗ ( f ∗ (ξ)) and ( f ◦ g)∗ (ξ) are such
that they are C-isomorphic.
These functorial properties of Vectn prove that

Vectn : H → Set

is a contravariant functor. 

Definition 5.7.5 (Grassmann manifold G n (F∞ )) The natural inclusion

G n (Fm ) ⊂ G n (Fm+1 ) ⊂ G n (Fm+2 ) ⊂ · · ·


defines a topological space G n (F∞ ) = G n (Fm ) with weak topology, called
n≤m
Grassmann manifold G n (F∞ ).

Remark 5.7.6 If γn∞ denotes the n-dimensional vector bundle over Grassmann man-
ifold G n (F∞ ), then φn = [−, G n (F ∞ )] : H → Set is a contravariant functor.

Theorem 5.7.7 proves the equivalence of two contravariant functors [−, G n (F∞ )]
and Vectn from the category H of paracompact spaces and their homotopy classes
to the the category Set of sets and set functions. This equivalence plays a key role
to solve a homotopy classification problem of vector bundles in Theorem 5.7.8.

Theorem 5.7.7 There exists a natural equivalence

342 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.31 Rectangular

diagram involving natural
equivalence

ψ : [−, G n (F ∞ )] → Vectn

between the two contravariant functors.

Proof For every object B ∈ H, define

ψ(B) : [B, G n (F∞ )] → Vectn (B), [ f ] → { f ∗ (γn∞ )}.

Then ψ(B) is a well-defined morphism in the category Set.

To show that ψ is a natural transformation, consider the homotopy class [ f ] of the
continuous map f : A → B between paracompact spaces A and B and also consider
the rectangular diagram in Fig. 5.31
The diagram in Fig. 5.31 is commutative, because, for any [h] ∈ [B, G n (F∞ )],

(Vect n ([ f ]) ◦ ψ(B))([h]) = Vect n ([ f ]){h ∗ (γn∞ )} = {( f ∗ h ∗ (γn ))}.

Consequently, ψ is a natural transformation. To prove that ψ is a natural equiv-

alence, it is sufficient to prove that for every object B ∈ H, the map ψ(B) is a
bijection.
Consider

ψ(A)φ([ f ][h]) = ψ(A)([h ◦ f ]) = {(h ◦ f )∗ (γn∞ )}.

Clearly, for every B ∈ H, the map ψ(B) is injective by Exercise 5.24.1 of Sect. 5.24
and is also surjective by Exercise Ex:5.13(4) 5.24.1 of Sect. 5.24, Hence it implies
that for each B ∈ H, the map ψ(B) is a bijection. It proves that ψ is an equivalence.


Theorem 5.7.8. solves a homotopy classification problem of vector bundles over

a paracompact space B in Theorem 5.7.8 by using the homotopy classes of maps
from B to Grassmann manifold G n (F ∞ ).

Theorem 5.7.8 (Homotopy classification of vector bundles) There exists a bijective

correspondence between the set of isomorphism classes of n-dimensional F-vector
bundles on a paracompact space B and the set of homotopy classes of free continuous
maps from B to Grassmann manifold G n (F∞ ).

Proof For every object B ∈ H, define the map

5.7 Homotopy Properties of Vector Bundles and Their Homotopy Classification 343

ψ(B) : [B, G n (F∞ )] → Vectn (B), [ f ]) → { f ∗ (γn∞ )}.

Since ψ(B) is a bijection by Theorem 5.7.7, the theorem follows. 

Definition 5.7.9 (Representation of the functor Vectn ) The natural equivalence ψ :
[−, G n (F∞ )] → V ectn , proved in Theorem 5.7.7, is called a representation of the
contravariant functor Vectn .
Corollary 5.7.10 (Classification of vector bundles) Let ξ = (X, p, B, Fn ) be an
arbitrary n-dimensional F-vector bundle over a paracompact space B. Then it is
isomorphic to the vector bundle induced by a continuous map from the base space
B to the Grassmann manifold G n (F∞ ).
Proof By hypothesis, ξ = (X, p, B, Fn ) is an n-dimensional F-vector bundle over
a paracompact space B. Since ψ(B) is a bijection as proved in Theorem 5.7.7, it
follows that exists a continuous map f : B → G n (F∞ ) such that f ∗ ({γn∞ }) = {ξ}.
This proves that vector bundles f ∗ (γn∞ ) and ξ are B-isomorphic. 

5.8 Homotopy Properties of Numerable Principal

G-Bundles

This section is devoted to the study of numerable principal G-bundles over B asso-
ciated with a given topological group G. This study is based on the viewpoint of
homotopy theory and provides a contravariant functor

K G : Ht p → Set,

which plays an key role in the study of homotopy theory.

Theorem 5.8.1 Let η be a numerable principal G-bundle over B × I.
(i) Then the three bundles η, (η|(B × {1})) × I and (η|(B × {0})) × I are G-
isomorphic.
(ii) If the map h t : B → B × I, b → (b, t), then the principal G-bundles h ∗0 (η) and
h ∗1 (η) are B-isomorphic.
Proof Consider the map f : B × I → B × I, (b, t) = (b, 1) Then by Exercise
5.24.1 of Sect. 5.24, the bundles η and f ∗ (η) are isomorphic principal G-bundles
over B × I. Since the bundles f ∗ (η) and (η|(B × {1}) × I) are isomorphic principal
G-bundles over B × I, it follows that η and (η|(B × {1}) × I) are also isomorphic
principal G-bundles. Analogously, the bundles η and (η|(B × {0}) × I) are also
isomorphic principal G-bundles. This proves that the bundles η, (η|(B × {1})) × I
and (η|(B × {0})) × I are G-isomorphic. This gives part (i). For part (ii), since
f ◦ h 0 = h 1 and the bundles f ∗ (η)|(B × {0}) and η|(B × {0}) are G-isomorphic,
it is proved that the induced bundles h ∗1 (η) = h ∗0 ◦ f ∗ (η) and h ∗0 (η) are isomorphic
principal G-bundles. 
344 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Corollary 5.8.2 Let η = (X, p, B) be a numerable principal G-bundle and f, g :

A → B be two homotopic maps. Then their induced bundles f ∗ (η) and g ∗ (η) are
isomorphic principal G-bundles over A.

Proof By hypothesis, f g : A → B, consider the homotopy Ht : f g:A→

B and the map h t : A → A × I, a → (a, t) : t = 0, 1. Then H0 ◦ h 0 = f and H1 ◦
h 1 = g assert by Theorem 5.8.1 that the bundles f ∗ (η) and g ∗ (η) are isomorphic
principal G-bundles over A. 

Theorem 5.8.3 Given any topological space B, let K G (B) denote the set of iso-
morphism classes of numerable principal G-bundles over B. Then there exists a
contravariant functor
K G : Ht p → Set

from the homotopy category Ht p of topological spaces and their homotopy classes
of maps to the category Set of sets and their functions.

Proof To prove the theorem the object function and morphism functions are defined
as follows:
(i) For every object B ∈ H t p, the object K G (B) ∈ Set is defined to be the set
of isomorphism classes of numerable principal G-bundle over B. Hence the
assignment
K G : Ht p → Set, B → K G (B)

defines the object function and

(ii) for the homotopy class [ f ] of every continuous map f : A → B, the function

K G ([ f ]) : K G (B) → K G (A), [ f ] → { f ∗ (ξ)}

is well-defined by Corollary 5.8.2, and it defines the morphism function.

Finally, given any two continuous maps f : A → B and g : B → C and a numerable
principal G-bundle η over C, the induced bundle (g ◦ f )∗ (η) and f ∗ (g ∗ (η)) are
isomorphic over C. This implies that

K G ([g] ◦ [ f ]) = K G ([ f ]) ◦ K G ([g]).

Similarly, the function K G ([1C ]) is the same as the identity function on K G ([C]),
since η and 1C ∗ (η) are isomorphic. This proves that K G is a contravariant functor. 

Corollary 5.8.4 For any numerable principal G-bundle ξ = (X, p, B, G), every
homotopy equivalence f : A → B induces a bijection

K G ([ f ]) : K G (B) → K G (A).

Proof If f : A → B is a homotopy equivalence, then there exists a continuous map

5.8 Homotopy Properties of Numerable Principal G-Bundles 345

g: B → A:g◦ f 1 A and f ◦ g 1B .

Since K G is a contravariant functor by Theorem 5.8.3, it is proved that

K G ([g ◦ f ]) = K G ([ f ]) ◦ K G ([g]) : K G (A) → K G (A)

is the identity map. This implies that K G ([ f ]) is a surjective map. Similarly, it follows
that
K G ([ f ◦ g]) = K G ([g]) ◦ K G ([ f ]) : K G (B) → K G (B)

is the identity map. This implies the map K G ([ f ]) is injective. Consequently the map
K G ([ f ]) is a bijection. 

Corollary 5.8.5 Every numerable principal G-bundle ξ = (X, p, B, G) over a con-

tractible space B is trivial.

Proof By hypothesis, the base space B of ξ is contractible. Hence the space B is

homotopy equivalent to a point {∗}. Since the set K G ({∗}) consists of only one point,
which is precisely, the isomorphism class of the trivial bundle, corollary is proved
by using Corollary 5.8.4. 

Theorem 5.8.6 Let ξ0 = (X, p0 , B0 , G) be a fixed numerable principal G-bundle.

Then
ψξ0 : Ht p → Sets, [−, B0 ] → K G

is a covariant functor.

Proof By hypothesis, ξ0 = (X, p0 , B0 , G) is a fixed numerable principal G-bundle.

Let ξ = (X, p, B, G) be an arbitrary numerable principal G-bundle. Keeping B0
fixed and varying B, define the object function

Ht p → Sets, B → K G (B) = [B, B0 ]

and the morphism function

ψξ0 (B)([ f ]) : Ht p → Sets, [ f ] → [ f ∗ (ξ0 )], ∀ [ f ] ∈ [B, B0 ].

Both the functions are well-defined by Theorem 5.8.3 and its Corollary 5.6.3.
This proves that
ψξ0 : [−, B0 ] → K G

is a covariant functors from the homotopy category Ht p to the category Set. 

Definition 5.8.7 (Universal G-bundle) A principal G-bundle ξ0 = (X, p0 , B0 , G)

is said to be universal if
(i) ξ0 is numerable and
346 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(ii) ψξ0 : [−, B0 ] → K G defined in Theorem is a natural equivalence.

Theorem 5.8.8 characterizes universal G-bundles in terms of homotopic maps

from one base space to the other base space.

Theorem 5.8.8 Let ξ0 = (X 0 , p0 , B0 , G) be a numerable principal G-bundle. It is

universal iff
(i) for every numerable principal G-bundle ξ = (X, p, B, G), there exists a con-
tinuous map h : B → B0 with the property that ξ and h ∗ (ξ0 ) are B-isomorphic;
and
(ii) if f, g : B → B0 are two continuous maps with the property that f ∗ (ξ0 ) and
g ∗ (ξ0 ) are isomorphic, then f g.

Proof Under the condition (i) it follows that the map

ψξ0 (B) : [B, B0 ] → K G (B)

is surjective. Again, under condition (ii), it follows that

ψξ0 (B) : [B, B0 ] → K G (B)

is injective. Hence the theorem follows. 

5.9 Classifying Spaces: The Milnor Construction

This section describes Milnor method of construction of classifying spaces and also
universal principal fiber spaces. He started with topological G and take an join X G
of G. He defines an action σ : X G × G → X G and a topology on X G to make it
a G-space. The orbit space X G modG thus obtained is denoted by BG , and then
the projection map p : X G → BG is continuous and onto. The resulting bundle
ωG = (X G , p, BG ) is a numerable (universal) principal G− bundle. The space is
called the classifying space of K G . This method of construction is known as Milnor
construction.
The construction of the bundle ωG = (X G , p, BG ) described in Theorem 5.9.1 is
known as Milnor construction. The bundle ωG = (X G , p, BG ) provides an important
family of numerable principal G-bundle.

Theorem 5.9.1 (Milnor construction) Given a topological group G, the universal

fiber space X G is defined as an infinite join

XG = G ∗ G ∗ · · · ∗ G

and an element x, t ∈ X G is denoted by

 5.9 Classifying Spaces: The Milnor Construction 347

x, t = (t0 x0 , t1 x1 , . . . , tr xr , . . . ),

 each xi ∈ G and ti ∈ [0, 1] = I such that only a finite number ti = 0 and

where
ti = 1.
ti ≥0
Two elements x, t = x , t  ∈ X G are said to be equal iff
(i) ti = ti for every i and
(ii) xi = xi for every i with ti = ti > 0.
Define an action σ of G on X G from the right

σ : X G × G → X G , x, tg = xg, tg or (t0 x0 , t1 x1 , . . . )g = (t0 x0 g, t1 x1 g1 , . . . )

and topologize the set X G in such a way that X G admits a G-space structure. Consider
two families of maps
f i : X G → I, ∀ i ≥ 0,

which assigns to the element (t0 x0 , t1 x1 , . . . ) ∈ X G the component ti ∈ [0, 1] and

gi : f i−1 (0, 1] → G, ∀ i ≥ 0,

which assigns to the element (t0 x0 , t1 x1 , . . . ) the component xi ∈ G. Then xi cannot

be uniquely defined outside f i−1 (0, 1] in a natural way. Because, if f ∈ X G and
g ∈ G, then there exist the following relations between the action of G and the
maps f i and gi : gi ( f g) = gi (α)g and f i ( f g) = f i (α). Endow the smallest topology
on the set X G such that every map f i : X G → [0, 1] and gi : f −1 (0, 1] → G is
continuous, with f i−1 (0, 1] the subspace topology. From the definition of the action
sigma, it follows that X G is a G-space with the G-set structure map

σ : XG × G → XG

continuous. Denote
(i) the orbit space by X G mod G by BG ,
(ii) the quotient map by p : X G → BG and
(iii) the resulting bundle ωG = (X G , p, BG ).
This bundle ωG = (X G , p, BG ) constructed by Milnor is a numerable principal
G-bundle.

Example 5.9.2 illustrates Milnor construction at some concrete situation to

obtain a numerable principal G-bundle ωG , where the topological groups G are
Z2 , S 1 and S 3 .

Example 5.9.2 (Numerable principal G-bundle)

 348 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(i) Consider the n-sphere S n in Rn+1 for n ≥ 1 and the antipodal map

f : S n → S n , x → −x.

Then f 2 = f ◦ f is identity map 1 S n and the group G = {1 S n , f } ∼

= Z2 , which
is a subgroup of the group of homeomorphisms of S n . Here the space X G = S n .
Define the action σ of G on X G = S n

x if g = 1 S n
σ : S n × G → S n , (x, g) →
−x if g = f.

This asserts that the quotient space X G mod G = BG = R P n and the resulting
bundle ωG = (S n , p, R P n ) is a numerable principal G-bundle of dimensions
≤ n − 1.
(ii) Let G = S 1 be the unit circle in the complex plane C and S 2n+1 be the (2n + 1)-
sphere in R2n+2 ( identified with Cn+1 ). This topological group G = S 1 is
regarded as a subgroup S 1 ⊂ C − {0} with usual multiplication of nonzero com-
plex numbers. Consider the right action σ of S 1 on S 2n+1

σ : S 2n+1 × S 1 → S 2n+1 , ((z 0 , z 1 , . . . , z n ), eiθ ) → (eiθ z 0 , eiθ z 1 , . . . , eiθ z n ).

Let X G = G ∗ G ∗ · · · ∗ G be the infinite join. Then X G mod G = BG = CP n

is the complex n-dimensional projective space. This implies that wG =(S 2n+1 , p,
CP n ), is a principal numerable G-bundle of dimensions ≤ 2n.
(iii) Let G = S 3 . This topological group is regarded as a subgroup S 3 ⊂ H − {0}
with usual multiplication of nonzero quaternionic numbers. Let X G = G ∗ G ∗
· · · ∗ G be the infinite join. Then proceeding as before, it follows that wG =
(S ∞ , p, HP ∞ ), where S ∞ denotes the infinite dimensional sphere in infinite
dimensional quaternionic space H∞ with weak topology, and HP ∞ denotes
infinite dimensional quaternionic projective space.

The concept of classifying spaces introduced by Milnor in 1956 is formulated in

Definition 5.9.3.

Definition 5.9.3 (Classifying space of K G ) Given a topological group G, the func-

tor
K G : Ht p → Set

is said to be a representation of the functor K G , if there exist a space BG , called the

classifying space of K G and a bundle ξG = (X G , pG , BG ), called universal bundle
in K G (BG ), such that there is a natural equivalence

ψ : K G → [−, BG ]

of functors defined from the category Ht p to the category Set.

5.9 Classifying Spaces: The Milnor Construction 349

Remark 5.9.4 It follows from Definition 5.9.3 that

(i) for any topological group G, the map

ψ(B) : [B, BG ] → K G (B), [ f ] → [ f ∗ (ξG )])

is a bijection and
(ii) for any topological space which is homotopy equivalent to BG is also a classifying
space for K G .
Theorem 5.9.5 (Milnor) Let G be a topological group. Then the G-bundle ωg =
(X G , p, BG ) is a numerable principal G-bundle, and this bundle is a universal G-
bundle.
Proof It follows from Theorem 5.9.1 by using Milnor method of construction of
ωg = (X G , p, BG ). 

5.10 Applications and Computations

This section presents an application of Galois correspondence arising from the

function that assigns to each covering space p : (X, x0 ) → (B, b0 ) the subgroup
p∗ (π1 (X, x0 )) of π1 (B, b0 ) and communicates applications of covering to compute
fundamental groups of some important spaces such as real projective spaces, Klein
bottle, lens space and figure-eight.

5.10.1 Application of Galois Correspondence Theorem

This subsection communicates an application of Galois correspondence ψ formulated

in Theorem 5.5.8 by using the action σ of fundamental group of the base space of a
covering space on its fiber formulated in Theorem 5.10.2.
Theorem 5.10.1 Let (X, p) be a covering space over a connected, locally path
connected and semilocally simply connected base space B. Then
(i) the components of the total space X are in bijective correspondence with orbits
obtained by the action of the fundamental group π1 (B, b0 ) on the fiber p −1 (b0 )
over b0 ∈ B and
(ii) the subgroup assigned to the component of X containing a given lifting b̃0 = x0
of the point b0 by the Galois correspondence ψ is the stabilizer group G x0 of
x0 ∈ X, where the action σ on the fiber keeps x0 fixed.
Proof By hypothesis, (X, p) is a covering space over a connected, locally path-
connected and semilocally simply connected base space B. Let x0 , x1 ∈ p −1 (b0 ) be
two arbitrary points.
350 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(i) Suppose that the points x0 and x1 lie in the same component of X. Then there
exists a path
β : I → X : β(0) = x0 and β(1) = x1 .

Then [ p ◦ β] is an element of π1 (B, b0 ) such that

σ : p −1 (b0 ) × π1 (B, b0 ), (x0 , [ p ◦ β]) → x1 .

This asserts that the action σ is transitive. On the other hand, if x0 , x1 ∈ p −1 (b0 )
are points lying in different components of X, then there exists no path con-
necting them. Consequently, there exists a bijective correspondence between
the components of the total space X and the orbits obtained by the action σ of
the fundamental group π1 (B, b0 ) on the fiber p −1 (b0 ), because the set of ele-
ments of p −1 (b0 ) in a given component forms an orbit. Hence it follows that this
correspondence is a bijection.
(ii) Suppose that the point x0 ∈ p −1 (b0 ) is an arbitrary points and x0 lies in some
component C x0 of X. Then by Galois correspondence ψ, the subgroup of
π1 (B, b0 ) corresponding to the component C x0 of X is the image p∗ (G) of the
group G = π1 (C x0 , x0 ). Since every loop β ∈ p∗ (G) lifts back to a loop in C x0
by the unique lifting property, it follows that β carries the x0 to itself and hence
it is an element of the stabilizer group G x0 of x0 under the action σ. Conversely,
suppose f ∈ π1 (B, b0 ) is in the stabilizer group G x0 of x0 under the action σ.
Then the lift f˜ of α and hence f˜ is a loop in X based at the x0 . α̃ ∈ G, which
implies α ∈ p∗ (G). This proves that p∗ (G) is the stabilizer group G x0 of x0 . 

5.10.2 Actions of Fundamental Groups on Fibers

of Covering Spaces

This subsection considers action of the fundamental group of the base space of a
covering space on a fiber. This action plays an important role in the study of the
covering space.
Theorem 5.10.2 proves some important results obtained by action of the funda-
mental group of the base space of a covering space on its fiber.
Theorem 5.10.2 Let (X, p) be a covering space of B, and Fb0 = Y = p −1 (b0 ) be
the fiber over b0 ∈ B. If X is path connected, then the map

σ : Y × π1 (B, b0 ) → Y, (x, [ f ]) → f˜(1),

(where f˜ is the unique lifting of f : (I, 0) → (B, b0 ) such that f˜(0) = x )

satisfies the following properties:

(i) the group π1 (B, b0 ) acts transitively on Y ;
 5.10 Applications and Computations 351

(ii) If y0 ∈ Y ⊂ X , then the isotropy group G y0 = p∗ π1 (X, y0 ); and

(iii) the cardinality |Y | = [π1 (B, b0 ) : p∗ π1 (X, y0 )].
Proof We first show that

σ : Y × π1 (B, b0 ) → Y, (x, [ f ]) → f˜(1)

is a right action. Clearly, σ is well-defined, because

(i) f˜ is the unique lifting of

f : (I, 0) → (B, b0 ) : f˜(0) = x

and
(iii) σ is independent of the choice of the representative of the class [ f ] by the
Monodromy Theorem 5.4.7.
Next, we show that it satisfies the conditions of an action. If f is a constant path at
b0 , then f˜ is also a constant path at x ∈ Y such that

σ(x, [ f ]) = x · [ f ] = f˜(1) = x.

If [ f ], [g] ∈ π1 (B, b0 ), and f˜ is the lifting of f with f˜(0) = x and g̃ is the lifting
of g with g̃(0) = f˜(1), then f˜ ∗ g̃ is a lifting of f ∗ g that begins at x and ends
at g̃(1). This asserts that x · [ f ∗ g] = (x · [ f ])[g]. Hence it follows that σ : Y ×
π1 (B, b0 ) → Y is an action.
(i) To show that σ is transitive, take a fixed point y0 ∈ Y and any point x ∈ Y . By
hypothesis, X is path connected. Hence there is a path α̃ in X from y0 to x such
that p ◦ α̃ is a closed path in B at b0 having lifting with initial point y0 is α̃.
Hence it follows that

[ p ◦ α̃] ∈ π1 (B, b0 ) and y0 · [ p ◦ α̃] = α̃(1) = x.

This asserts that π1 (B, b0 ) acts transitively on Y .

(ii) To prove (ii), take any closed path f based at the point b0 ∈ B and G = π1 (B, b0 ).
If f˜ is the lifting of f with f˜(0) = y0 , then the corresponding isotropy group
of σ at y0 is given by

G y0 = {[ f ] ∈ π1 (B, b0 ) : y0 · [ f ] = y0 }.

This implies that if [ f ] ∈ G y0 , then y0 · [ f ] = y0 and f˜(1) = y0 = f˜(0). Hence

it follows that f˜ ∈ π1 (X, y0 ). This shows that G y0 ⊂ p∗ π1 (X, y0 ). On the other
hand, if [ f ] = [ p ◦ g̃] for some [g̃] ∈ π1 (X, y0 ), then f˜ = g̃, because both are
liftings of f having the same initial point y0 . This implies that

f˜(1) = g̃(1) ⇒ y0 · [ f ] = f˜(1) = y0 ⇒ [ f ] ∈ G y0

352 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

and hence p∗ π1 (X, y0 ) ⊂ G y0 . This proves that the isotropy group G y0 =

p∗ π1 (X, y0 ).
(iii) Since the group π1 (B, b0 ) acts transitively on Y , with isotropy group G y0 =
p∗ π1 (X, y0 ) at y0 ∈ Y, it follows that the cardinality |Y | of Y is given by

|Y | = [π1 (B, b0 ) : p∗ π1 (X, y0 )]. 

Corollary 5.10.3 Let X be a path-connected space and (X,p) be a universal covering

space of B, and b0 ∈ B. Then the cardinality of the fiber Y = p −1 (b0 ) is given by

|Y | = |π1 (B, b0 )|.

Proof Since by the given condition, π1 (X, x0 ) = 0, it follows by using from Theorem
5.10.2(iii)that
|Y | = |π1 (B, b0 )|. 

Corollary 5.10.4 Let X be a path-connected space. Given a covering space (X, p)

of b0 ∈ B, x0 ∈ p −1 (b0 ), if the induced homomorphism

p∗ : π1 (X, x0 ) → π1 (B, b0 )

is onto, then the map p : X → B induces an isomorphism of groups

p∗ : π1 (X, x0 ) → π1 (B, b0 ).

Proof p∗ is a monomorohism by Corollary 5.4.8, and it is also an epimorphism ,

since by hypothesis, p∗ is onto. This proves that p∗ is an isomorphism. 

5.10.3 Fundamental Groups of Orbit Spaces

This subsection computes the fundamental groups of some important spaces which
are obtained as orbit spaces. For example, we compute the fundamental groups of
projective spaces, lens spaces, figure-eight and Klein’s bottles by representing them
as orbit spaces.

Definition 5.10.5 A topological group G with identity e acting on a topological

space X is said to satisfy the condition (A) :
if for every point x ∈ X, there exists a nbd Ux of x such that

g (Ux ) ∩ Ux = ∅

implies g = e, where
g : X → X, x → gx
5.10 Applications and Computations 353

is a homeomorphism.
The special group action of the group G of homeomorphisms of X satisfying the
condition (A) reformulated in Definition 5.10.6 is said to be properly discontinuous.
Definition 5.10.6 An action σ of a topological group G on a topological space X

σ:G×X →X

is said to be properly discontinuous, if X has a nbd U such that the nbds gU and
kU are disjoint for every pair of distinct elements g, k ∈ G.
For the particular case, when G is a discrete topological group is also interesting.
For example, consider Example 5.10.7.
Example 5.10.7 Let G n be the finite group of n elements which are precisely the n
n-th roots of unity and C∗ = C − {0}. The action

σ : G n × C∗ , (g, z) → gz ( usual multiplication of complex numbers )

is properly discontinuous.
Example 5.10.8 The automorphism group Aut (X/B) of the covering spaces (X, p)
of a fixed base space B satisfies the condition (A) of Definition 5.10.5.
Definition 5.10.9 (Regular covering space) Let (X, p) be a covering space of B.
Then it is said to be regular if p∗ π1 (X, x0 ) is a normal subgroup of π1 (B, b0 ).
Example 5.10.10 Let B be a connected, locally path-connected space and G sat-
isfies the condition (A) on X then (X, p) is a regular covering space of X mod
G, where p : X → modG, x → Gx is the natural projection.
Theorem 5.10.11 Let σ : G × X → X be an action of a topological group G on X
satisfying the condition (A) of Definition 5.10.5.
(i) For any path-connected space X, the group G is the group of automorphisms of
the covering space
p : X → X mod G, x → Gx

(ii) For any is path connected and locally path-connected space X , the group G is
isomorphic to the quotient group π1 (X mod G)/ p∗ π1 (X ).
(iii) For any simply connected space X , the groups π1 (X mod G) and G are isomor-
phic.
Proof (i) By hypothesis, X is path connected. The automorphism group contains
G as a subgroup and equals this group, because, if f is any covering (deck)
transformation, then given any point x ∈ X, x and f (x) are in the same orbit
and hence there is some g ∈ G having the property g(x) = f (x). This implies
that f = g, since covering transformations of a connected covering space are
uniquely determined at this situation.
354 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(ii) By hypothesis , X is a path connected and locally path-connected space. Hence

this part follows.
(iii) By X is a simply connected space. Then for any point y ∈ X mod G, the funda-
mental group π1 (X, x0 ) = {e}, ∀ x0 ∈ p −1 (y). This implies that p∗ π1 (X, x0 ) =
{e}. Hence part (5.10.11) follows from part (5.10.11). 

5.10.4 Computing the Fundamental Group of R P n

This subsection computes by the fundamental groups of R P n for n ≥ 2. Theorem

5.10.15 computes it by using group action and shows that π1 (R R n ) ∼
= Z2 , ∀ n ≥ 2.
For the particular case, when n = 2, it is computed in an alternative method in
Theorem 5.10.18. The n-dimensional real projective n-space R P n is obtained from
the n-sphere S n = {x ∈ Rn+1 : ||x|| = 1} by identifying its antipodal points. It is
formulated in Definition 5.10.12.

Definition 5.10.12 Let A : S n → S n , x → −x be the antipodal map. Then A ◦ A =

1 S n and hence A generates an action of the two element group G = {+1, −1} defined
by the relation (+1)x = x and (−1)x = −x for all x ∈ S n . This action on S n has the
orbit space S n mod G, denoted by R P n , called the real n-dimensional projective
space.

Recall the following two theorems for their next applications.

Theorem 5.10.13 (S 2 , p) is a covering space of R P 2 , and R P 2 is a surface.

Proof See Chap. 4. 

Theorem 5.10.14 (S n , p) is a covering space of R P n , where p is the map identifying

antipodal points of S n for n > 1.

Proof See Chap. 4. 

Theorem 5.10.15 π1 (R R n ) ∼
= Z2 for n ≥ 2.

Proof Consider the action of the G = {+1, −1} on S n given in Definition 5.10.12
and the covering space (S n , R P n ) of multiplicity 2. Since S n is simply connected for
n ≥ 2 ( see Chap. 2), it follows by the Theorem 5.10.11 that the fundamental group
of orbit space is G. Since S n mod G = R P n , it follows that

π1 (S n mod G) = π1 (R P n ) = G ∼
= Z2 ∀ n ≥ 2. 

Remark 5.10.16 (i) Geometrically, a generator of the group π1 (R P n ) is any loop

obtained by projecting a continuous path on the sphere S n , which connects two
antipodal points.
5.10 Applications and Computations 355

(ii) The above action is free in the sense that gx = x ⇒ g = e. Does there exist
any other finite group G acting freely on S n and defining covering space S n →
S n mod G? The answer is Z 2 is the only nontrivial group that can act freely on
S n if n is even.

Theorem 5.10.17 provides an alternative approach for computing π1 (R R n ) for

n ≥ 2.

Theorem 5.10.17 π1 (R R n ) ∼
= Z2 for n ≥ 2.

Proof Proof I As S n is simply connected for n ≥ 2, so from the covering space

p : S n → R P n it follows by the Theorem 5.10.11 that the fundamental group of
orbit space is G. Thus π1 (S n mod G) = G ⇒ π(Rn ) = G ∼ = Z2 for n ≥ 2.
Proof II Consider the universal covering space (S n , q) of R P n where q identifies
the antipodal points of S n . Then by using Exercise 5.24.1 of Sect. 5.24, it follows
that
|π1 (R P n )| = 2

and hence it is proved that π1 (R P n ) ∼

= Z2 . 

Theorem 5.10.18 computes π1 (R P 2 , y) in a little different method.

Theorem 5.10.18 π1 (R P 2 , y) ∼
= Z2 .

Proof Consider the projection map

p : S2 → R P 2.

It is a covering map by Theorem 5.10.13. Since S 2 is simply connected, we apply The-

orem 5.4.25 to obtain a bijective map ψ between π1 (R P 2 , y) and the set p −1 (y). This
proves that π1 (R P 2 , y) ∼
= Z2 , because p −1 (y) is a two-element set and π1 (R P 2 , y)
is a group of order 2 and any group of order 2 is isomorphic to Z2 . 

5.10.5 Computing the Fundamental Group of Klein’s Bottle

This subsection computes the fundamental group of Klein’s bottle, which is a well-
known topological space constructed in Volume I of the present book series of Basic
Topology. Let G be the group of transformations of the Euclidean plane R2 generated
by α and β. Consider the action σ of G on R2

σ : G × R2 → R2 , (α, (x, y)), (β, (x, y)) → (x + 1, y), (1 − x, y + 1).

356 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

This implies that

α−1 (x, y) = (x − 1, y) and β −1 (x, y) = (1 − x, y − 1).

Since R2 is simply connected and the action σ satisfies condition (A) of Definition
5.10.5, then by Theorem 5.10.11, it follows that

π1 (R2 mod G) G.

The quotient space R2 mod G is the Klein’s bottle obtained by the action of σ on
R2 and gives a representation of Klein’s bottle as an orbit space whose fundamental
group is generated by α and β, because

β −1 αβ(x, y) = β −1 α(1 − x, y + 1)
= β −1 (2 − x, y + 1) = (1 − 2 + x, y)
= (x − 1, y) = α−1 (x, y), ∀ (x, y) ∈ R2

asserts that β −1 αβ = α−1 . Hence it follows that the fundamental group of the Klein
bottle is generated by α and β.

5.10.6 Computing the Fundamental Groups of Lens Spaces

This subsection computes the fundamental group of lens spaces (L(m, p)) con-
structed by H. Tietze (1888–1971) in 1908. They are three-dimensional manifolds
and form an important class of topological spaces for the study of algebraic topology.
Construction of lens space (L(m, p)): Consider S 3 = {(z 1 , z 2 ) ∈ C2 : |z 1 |2 +
|z 2 |2 = 1} ⊂ C2 .
Let m > 1 be an integer space and p be an integer relatively prime to m and
2πi
S = {(z 1 , z 2 ) ∈ C2 : |z 1 |2 + |z 2 |2 = 1} ⊂ C2 = R4 . Let ρ = e m be a primitive m-
3

th root of unity.
Define a continuous map
2πi 2πi p
f : S 3 → S 3 , (z 1 , z 2 ) → (ρz 1 , ρ p z 2 ) = (e m z1, e m z 2 ).

This implies that f : S 3 → S 3 is a homeomorphism of period m in the sense that

f = 1 S 3 . Consider an action σ of Zm on S 3 induced by the homeomorphism f
m

σ : Zm × S 3 → S 3 , (k, (z 1 , z 2 )) → f k (z 1 , z 2 ).
2πi
Geometrically, this action σ is generated by the rotation z → e m z of the unit sphere
S 3 , and hence it has no fixed point. Analytically, it is proved that the action σ has no
5.10 Applications and Computations 357

2πir
fixed point , because given an integer r : 0 < r < m, the equation z = e m z has a
solution z = 0 but z = 0 is not a point of S 3 .
The orbit spaces S 3 mod Zm thus obtained is called a lens space denoted by
L(m, p). Thus the lens space is the quotient space S 3 /∼, where

(z 1 , z 2 ) ∼ (z 1 , z 2 ) ⇔ (z 1 , z 2 ) = f k (z 1 , z 2 ) for some k ∈ Z m .

Construction of generalized lens space L(m, p1 , . . . , pn−1 ): Consider

S 2n−1 = {(z 1 , z 2 , . . . , z n ) ∈ Cn : |z 1 |2 + |z 2 |2 + · · · + |z n |2 = 1} ⊂ Cn .

Given an integer m > 1 and integers p1 , p2 , . . . , pn−1 relatively prime to m, let

2πi
ρ = e m be a primitive m-th root of unity. Define a continuous map
2πi 2πi p1 2πi pn−1
f : S 2n−1 →S 2n−1 , (z 1 , z 2 , . . . , z n ) → (ρz 1 , ρ p1 z 2 , . . . , ρ pn−1 z n ) = (e m z1 , e m z2 , . . . , e m zn )

This implies that f : S 2n−1 → S 2n−1 is a homeomorphism and the homeomorphism

f induces an action σ of Zm on S 2n−1

σ : Zm × S 2n−1 → S 2n−1 , (k, (z 1 , z 2 , . . . , z n )) → f k (z 1 , z 2 , . . . , z n ).

Theorem 5.10.19 π1 (L(m, p)) ∼

= Zm .

Proof Since the group Zm is finite and the space S 3 is Hausdorff, the action σ of Zm
on S 3 satisfies condition (A) of Definition 5.10.5. Hence Zm π1 (S 3 mod Zm ) =
π1 (L(m, p)). 

Theorem 5.10.20 π1 (L(m, p1 , . . . , pn−1 )) ∼

= Zm of the generalized lens space
L(m, p1 , . . . , pn−1 ).

Proof Since by construction the lens space (L(m, p1 , . . . , pn−1 ) is obtained as the
orbit spaces S 2n−1 mod Z m , proceed as in Theorem 5.10.19 to prove that

π1 (L(m, p1 , . . . , pn−1 )) ∼
= Zm . 

Corollary 5.10.21 π1 (R P 2 )) ∼
= Z2 .

Proof Consider the particular the group Z2 = {1 S 2 , A}, where 1 S 2 : S 2 → S 2 is the

identity map and A : S 2 → S 2 , x → −x the antipodal map. Hence it follows that
the orbit space S 2 mod Z2 = L(2, 1) obtained by the action of Z2 on S 2 is the real
projective plane R P 2 . This proves that π1 (R P 2 )) ∼
= Z2 . 
358 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

5.10.7 Computing the Fundamental Group of Figure-Eight

This subsection computes the fundamental group of figure-eight in two methods.

This group is not abelian. The computation of this group by van Kampen theorem is
available in Chap. 2.

Theorem 5.10.22 The fundamental group of the figure-eight F8 is not abelian.

Proof Proof I: Geometric method. Consider the universal covering space (X, p) of
figure-eight and the path f˜ : I → X : t → (t, 0) which goes along the x-axis from
the origin (0, 0) to the point (1, 0). Let g̃ : I → X, t → (0, t) which goes along the
y-axis from the origin (0, 0) to the point (0, 1). If f = p ◦ f˜ and g = p ◦ g̃, then
f and g are loops on the figure-eight F8 based at x0 , going around the circles C1
and C2 , respectively. To prove that f ∗ g and g ∗ f are not path homotopic, lift each
of them to a path in X beginning at the origin. Then the path f ∗ g lifts to a path
such that it goes along the x-axis from the origin to (1, 0) and goes thereafter once
around the circle tangent to the x-axis at (1, 0). On the other hand, the path g ∗ f
lifts to a path in X that goes along the y-axis from the origin to (0, 1) and then goes
once around the circle tangent to the y-axis at (0, 1). Consequently, it follows that
[ f ∗ g] = [g ∗ f ], because the lifted paths do not end at the same point. This implies
that [ f ] ◦ [g] = [g] ◦ [ f ].
Proof II: Graph theoretic method. Let G be a free group on two letters α
and β and X = Graph (G, α, β) be the graph constructed as follows: The vertices
of X are precisely the elements of G. This implies that the vertices of X are the
reduced words α and β. The edges of X are of the two types: (g, gα) : g ∈ G and
(g, gβ) : 1g ∈ G. Again (g, gα), (g, gβ), (gα−1 , g) and (g, gβ −1 ) are the only four
edges corresponding to the vertex g. Define a map
⎧
⎪
⎨hg, for the vertex x = g ∈ G
σ : G × X → X, (h, x) → (hg, hgα), for the edge x = (g, gα)
⎪
⎩
(hg, hgβ), for the edge x = (g, gβ)

Denote σ(h, x) by hx. The definition of σ asserts an action of G on X. Because,

for the identity element 1G ∈ G,

1G · g = g, 1G · (g, gα) = (g, gα) and 1G · (g, gβ) = (g, gβ),

and since h 2 g ∈ X is a vertex, then

(h 1 h 2 ) · g = h 1 h 2 · g, h 1 · (h 2 · g) = h 1 · (h 2 · g) = h 1 h 2 · g, ∀ h 1 , h 2 ∈ G

and finally, for all g, h 1 , h 2 ∈ G.

5.10 Applications and Computations 359

h 1 · (h 2 · (g, gα)) = h 1 · (h 2 g, h 2 gα) = h 1 · (h 2 g, h 1 (h 2 gα)) = (h 1 h 2 ) · (g, gα)

= h 1 · (h 2 · (g, gβ)) = h 1 · (h 2 g, h 2 gβ)
= h 1 · (h 2 g, h 1 (h 2 gβ)) = h 1 h 2 · (g, gβ).

The orbit space X mod G thus obtained by the action is the figure-eight space. Its two
loops are the images of the edges (g, gα) and (g, gβ). Since X is simply connected,
it follows from Theorem 5.10.11 that π1 (X mod G, ∗) ∼ = G, where G is the free
group on two generators. 

5.11 The Relative Homotopy Groups

This section introduces the concept of the relative homotopy groups πn (X, A, x0 )
by generalizing both the concepts of the fundamental group π1 (X, x0 ) and absolute
homotopy groups πn (X, x0 ) studied in Chap. 2. Homotopy groups play an important
role in the study of fiber bundles. The concept of higher homotopy groups was born to
study the problem to classify homotopically the continuous maps of an n-sphere into
a given space. For n = 1, this group is known as the fundamental group π1 (X, x0 )
of a pointed topological space (X, x0 ). The same technique works in defining the
absolute homotopy groups πn (X, x0 ) for n > 1. Both the fundamental groups and
absolute homotopy groups are already studied in Chap. 2. The close connection
between homotopy and homology groups and the existence of relative homology
groups Hn (X, A), studied in Chap. 3, stimulate to define the relative homotopy
groups πn (X, A, x0 ) for any triplet of topological spaces (X, A, x0 ) and provides an
important system, highly related to homology theory except for certain properties:
(i) π0 (X, x0 ) and π0 (X, A, x0 ) fail to be ordinarily groups.
(ii) the groups π1 (X, x0 ) and π2 (X, A, x0 ) are not ordinarily abelian.
(iii) the excision property for homology (see Chap. 3) fails to hold for homotopy.

5.11.1 Standard Notations and Construction of πn (X, A, x0 )

The main objective of this subsection is to generalize the concepts of absolute homo-
topy groups πn (X, x0 ) studied in Chap. 2 by defining the relative homotopy groups
πn (X, A, x0 ). For this purpose, this subsection first explains the standard notations
to be used in construction of πn (X, A, x0 ).

Definition 5.11.1 A triplet (X, A, x0 ) of three topological spaces consists of a space

X , a nonempty subspace A of X and a point x0 ∈ A. In particular, if x0 is the only
point of A, then the triplet is simply denoted by the pair (X, x0 ).

Example 5.11.2 Let In = {t = (t1 , t2 , . . . , tn ) : ti ∈ I, i = 1, 2, . . . , n} ⊂ Rn be the

topological product of the closed unit interval I = [0, 1] for n ≥ 1. The space In is
360 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

called the n-cube in the Euclidean space Rn . Its initial (n − 1)-face given by tn = 0
is identified with the space In−1 . Let Jn−1 denote the union of all remaining (n − 1)-
faces of In . Consequently, ∂ In = I˙n = In−1 ∪ Jn−1 and ∂ In−1 = In−1 ∩ Jn−1 . The
ordered triple (In , ∂In , Jn−1 ) of three spaces constitutes a triplet, whose form is more
general than the triplet (X, A, x0 ).

Definition 5.11.3 Let X be a topological space and A be a subspace of X with

x0 ∈ A. By a continuous map

f : (In , In−1 , Jn−1 ) → (X, A, x0 ),

we mean a continuous map f : In → X such that

(i) f (In−1 ) ⊂ A, i.e., f maps In−1 into A, and
(ii) f (Jn−1 ) = x0 , i.e., f maps Jn−1 into x0 .
In particular, f sends ∂ In into A and ∂ In−1 on the point x0 .
Let F n = F n (X, A, x0 ) denote the set of all such maps f topolozied by the
compact open topology.

Example 5.11.4 If f : ∂In → A is a continuous map such that f (∂ In−1 ) = x0 ∈ A,

then f : (∂In , ∂In−1 ) → (A, x0 ) is a continuous map and F n (X, x0 ) denote the set
of all such maps f topolozied by the compact open topology.

Definition 5.11.5 Let (X, A, x0 ) be a given triplet. Then for any pair of maps f, g ∈
F n (X, A, x0 ) their sum f ∗ g is defined by

f (2t1 , t2 , . . . , ti , . . . , tn ), if 0 ≤ t1 ≤ 1/2
( f ∗ g)(t) =
g(2t1 − 1, t2 , . . . , ti , . . . , tn ), if 1/2 ≤ t1 ≤ 1

∀ t = (t1 , t2 , . . . , tn ) ∈ In . Clearly, f ∗ g is continuous, because, for n ≥ 2 and

t1 = 1/2, the above both lines reduce to x0 and hence f ∗ g ∈ F n (X, A, x0 ) for
n ≥ 2. For n = 1, this is also true providing A = {x0 }.

Remark 5.11.6 Given any two elements f, g ∈ F n (X, A, x0 ), f ∗ g can be equally

well defined by

f (t1 , t2 , . . . , 2ti , . . . , tn ), if 0 ≤ ti ≤ 1/2
( f ∗ g)(t) =
g(t1 , t2 , . . . , 2ti − 1, . . . , tn ), if 1/2 ≤ ti ≤ 1

This implies that this definition of f ∗ g does not depend on the particular coordinate
ti we use.

Definition 5.11.7 Let f, g ∈ F n (X, A, x0 ) be two continuous maps. They are said
to be homotopic relative to the system {In−1 , A; Jn−1 , x0 } if there exists a continuous
map
 5.11 The Relative Homotopy Groups 361

Ht : In → X, ∀ t ∈ I

such that
(i) H0 = f,
(ii) H1 = g and
(iii) Ht ∈ F n (X, A, x0 ).
It is symbolized as f g rel {In−1 , A; Jn−1 , x1 }. Geometrically, it means that
f and g can be joined by a continuous curve in the space F n (X, A, x0 ).

Construction of πn (X, A, x0 ): The homotopy relation on F n (X, A, x0 ) is an

equivalence relation, and hence it divides F n (X, A, x0 ) into disjoint equivalence
classes, called homotopy classes. The set of homotopy classes of all these maps
relative to the system {In−1 , A; Jn−1 , x0 }) is denoted by πn (X, A, x0 ). If [ f ] denotes
the homotopy class of the map f ∈ F n (X, A, x0 ) and 0 denotes the homotopy class
of the constant map c(I n ) = x0 , then for n = 0, the set π0 (X, A, x0 ) denotes the set
of all path components of the space F n (X, A, x0 ). In F n (X, A, x0 ), if f i gi (i =
1, 2) rel {In−1 , A; Jn−1 , x1 }, then the two homotopies can be combined by pasting
lemma to provide a homotopy

f1 ∗ f2 g1 ∗ g2 rel {In−1 , A; Jn−1 , x1 },

This property of f ∗ g in F n (X, A, x0 ) asserts that if [ f ], [g] ∈ πn (X, A, x0 ), all

products f ∗ g for f ∈ [ f ] and g ∈ [g] will lie in a single homotopy class [k] ∈
πn (X, A, x0 ). This justifies to define a composition, called addition in πn (X, A, x0 )
by the rule
[ f ] + [g] = [ f ∗ g], ∀ [ f ], [g] ∈ πn (X, A, x0 )

in the sense that this composition does not depend on the particular choice of the
representatives of the homotopy classes.

Theorem 5.11.8 πn (X, A, x0 ) is an abelian group under the composition ‘ +’ for

for every n > 1.

Proof Associativity law: To prove this law, it is sufficient to show that for any three
elements f, g, h ∈ F n (X, A, x0 ) the property that ( f ∗ g) ∗ h f ∗ (g ∗ h) holds.
For this purpose, take a homotopy of the t1 - axis which stretches the closed interval
[0, 1/4] into the closed interval [0, 1/2], translates the closed interval [1/4, 1/2]
into the closed interval [1/2, 3/4] and contracts the closed interval [1/2, 1] into the
closed interval [3/4, 1].

Zero element: It is the homotopy class [c] of the constant map defined by c(In ) =
x0 , because the homotopy c ∗ f f for every f ∈ F n (X, A, x0 ) is obtained by
deforming the t1 interval to shrink the closed interval [0, 1/2] to the point 0 and
expands the closed interval [1/2, 1] into the closed interval [0, 1]. Additionally, if a
map f ∈ F n (X, A, x0 ) sends In into A represents the zero element. Because, if H is
362 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

a homotopy of In onto itself which contracts In into its face tn = 1, then a homotopy
is defined by

H (t, s) = (t1 , t2 , . . . , tn−1 , (1 − s)tn + s).

This implies that F(t, s) = f (H (t, s)) is a homotopy of f into the constant map c.

Existence of negative element: For any f, its inverse element

f (t) = f (1 − t1 , t2 , . . . , tn )

is also an element of F n (X, A, x0 ) such that both f ∗ f and f ∗ f are homotopic to

the constant map c.
Hence it follows that πn (X, A, x0 ) forms a group under the above additive opera-
tion +. This is called the n-th relative homotopy group of|, X modulo A at x0 . The
class [c] = 0 is the group-theoretic zero element of πn (X, A, x0 ), and the inverse
element of [ f ] is the element f ◦ θ, where

θ : In → In , t = (t1 , t2 , . . . , tn ) → (1 − t1 , t2 , . . . , tn ).

The group πn (X, A, x0 ) is abelian for n > 2 : If Jn−1 is pinched to a point s0 ,

then the triplet (In , In−1 , Jn−1 ) would admit a configuration, which is equivalent to
the triplet (Dn , S n−1 , s0 ) of spaces consisting of the unit Euclidean n-cell Dn , its
boundary ∂ D n = S n−1 , and a base point s0 ∈ S n−1 . This implies that an element of
πn (X, A, x0 ) can be equally well defined as a homotopy class (relative to the system
{S n−1 , A; s0 , x0 } ) of the maps f : (Dn , S n−1 , s0 ) → (X, A, x0 ). Since for every n >
2, there exists a rotation of Dn which keeps the point s0 fixed and interchanges two
halves of Dn , it follows that the group πn (X, A, x0 ) is abelian. This completely proves
that (πn (X, A, x0 ), + is an abelian group for every integer n > 1. 

Proposition 5.11.9 gives a sufficient condition for existence of a nonconstant

function f ∈ Fn (X, A, x0 ) representing the zero element of the group πn (X, A, x0 ).
The result is useful for our study.

Proposition 5.11.9 Let f ∈ Fn (X, A, x0 ) be a function such that f (In ) ⊂ A, then

[f] = 0

Proof By hypothesis, f ∈ Fn (X, A, x0 ) and f (In ) ⊂ A. Define a homotopy

Ht (t1 , t2 , . . . , tn−1 , tn ) = f (t1 , . . . , tn−1 , t + tn − ttn ).

Then Ht ∈ Fn (X, A, x0 ), ∀ t ∈ I and is such that H0 = f , and H1 (In ) = x0 . This

proves that [ f ] = 0. 
5.11 The Relative Homotopy Groups 363

5.11.2 Boundary Operator

This subsection defines the boundary operator

∂ : πn (X, A, x0 ) → πn−1 (A, x0 ), ∀ n ≥ 1.

It is an important concept in homotopy theory having interesting algebraic properties.

Definition 5.11.10 Let a given element α ∈ πn (X, A, x0 ) be represented by a con-

tinuous map
f : (In , In−1 , Jn−1 ) → (X, A, x0 ).

Then α = [ f ]. We now consider two possible cases.

Case I: If n = 1, then f sends In−1 to a point of A, which gives a path component
β of π0 (X, A, x0 ).
Case II: If n > 1, consider the restriction g = f |In−1 . Then g is a continuous map

g : (In−1 , ∂In−1 ) → (A, x0 ).

This implies that g represents an element [g] = β ∈ πn−1 (A, x0 ), which is inde-
pendent of the choice of the map f representing the given element α. Consider the
assignment α → β by setting ∂([ f ]) = [g]. This assignment induces a transforma-
tion

∂ : πn (X, A, x0 ) → πn−1 (A, x0 ), α → β, ∀ n ≥ 1,

which is called the boundary operator. Since

∂(α + β) = ∂α + ∂β, ∀ α, β ∈ πn (X, A, x0 )

for all n > 1, it follows that ∂ is a homomorphism, called the boundary homomor-
phism all n > 1.

5.11.3 The Induced Transformation

This subsection gives the concept of a transformation

f ∗ : πn (X, A, x0 ) → πn ((Y, B, y0 )

induced by a continuous map f : (X, A, x0 ) → (Y, B, x0 ) and relates this transfor-

mation with the boundary operator ∂. Moreover, Definition 5.11.11 formulates the
induced transformation
 364 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

f ∗ : πn (X, A, x0 ) → πn ((Y, B, y0 )

for n = 0, n ≥ 2 in general and for n = 1 at some particular situation.

Definition 5.11.11 Let f : (X, A, x0 ) → (Y, B, y0 ) be an arbitrary continuous map.
Then f : X → Y is a continuous map such that f (A) ⊂ B and f (x0 ) = y0 .
Case I for n = 0 : Since f carries the path components of X into the path compo-
nents of Y, the assignment f induces a transformation f ∗ : π0 (X, x0 ) → π0 (Y, y0 ),
which carries the zero element of π0 (X, x0 ) into the zero element of π0 (Y, y0 ).
Case II for n ≥ 2: Take any map g ∈ F n (X, A, x0 ). Then the composite map
f ◦ g ∈ F n (Y, B, y0 ). Hence the assignment f → f ◦ g defines a continuous map

f  : F n (X, A, x0 ) → F n (Y, B.y0 ).

Since the map f  is continuous, it carries the path components of F n (X, A, x0 )

to the path components of F n (Y, B, y0 ). Moreover, if g h in F n (X, A, x0 ), then
f ◦g f ◦ h in F n (Y, B, y0 ). Consequently, f induces a map

f ∗ : πn (X, A, x0 ) → πn (Y, B, y0 )

of homotopy classes. Since f (g ∗ k) = ( f ◦ g) ∗ ( f ◦ g), ∀ g, h ∈ F n (X, A, x0 ), it

follows that f ∗ is a homomorphism, called the homomorphism induced by f.
Case III for n = 1 In this case, A = {x0 }, B = {y0 }. Hence πn (X, A, x0 ) and
πn (Y, B, y0 ) become the fundamental groups π1 (X, x0 ) and π1 (Y, y0 ), respectively,
with f ∗ a homomorphism between them.
The above discussion is summarized in a basic result in homotopy theory formu-
lated in Theorem 5.11.12.
Theorem 5.11.12 Every continuous map f : (X, A, x0 ) → (Y, B, y0 ) induces a
transformation
f ∗ : πn (X, A, x0 ) → πn (Y, B, y0 )

of the corresponding homotopy classes such that

(i) the transformation f ∗ : π0 (X, x0 )→π0 (Y, y0 ) sends the zero element of π0 (X, x0 )
into the zero element of π0 (Y, y0 );
(ii) f ∗ is a homomorphism of groups for every n ≥ 2;
(iii) if A = {x0 }, B = {y0 } and n = 1, then f ∗ : π1 (X, x0 ) → π1 (Y, y0 ) is a homo-
morphism of fundamental groups.

5.11.4 Algebraic Properties

For any triplet (X, A, x0 ), the homotopy groups, the boundary operator ∂ and the
induced transformations have seven fundamental properties of which three algebraic
properties are proved in this section and others in subsequent sections.
5.11 The Relative Homotopy Groups 365

Proposition 5.11.13 Let 1d : (X, A, x0 ) → (X, A, x0 ) be the identity map. Then it

induces the identity transformation

1d∗ : πn (X, A, x0 ) → πn (X, A, x0 ), ∀ n ≥ 0.

Proof It follows from the definition of the induced transformation

1d∗ : πn (X, A, x0 ) → πn (X, A, x0 ), [ f ] → [1d ◦ f ] = [ f ], ∀ [ f ] ∈ πn (X, A, x0 ).

Proposition 5.11.14 Let f : (X, A, x0 )→(Y, B, y0 ) and g : (Y, B, y0 )→(Z , C, z 0 )

be two continuous maps of triples. Then their induced transformations have the
property:

(g ◦ f )∗ = g∗ ◦ f ∗ : πn (X, A, x0 ) → πn (Z , C, z 0 ), ∀ n ≥ 0.

Proof By the given condition, f and g induce transformations of homotopy classes

f ∗ : πn (X, A, x0 ) → πn (Y, B, y0 )

and

g∗ : πn (Y, B, y0 ) → πn (Z , C, z 0 ),

respectively. Hence the proposition follows the definitions of the induced transfor-
mations f ∗ and g∗ . 

Proposition 5.11.15 relates the boundary operator with the induced transforma-
tion.

Proposition 5.11.15 Let f : (X, A, x0 ) → (Y, B, y0 ) be a continuous function and

g = f | A = (A, x0 ) → (B, y0 ) be the restriction of f to its subspace A. Then

∂ ◦ f ∗ = g∗ ◦ ∂ : πn (X, A, x0 ) → πn−1 (B, y0 ), ∀ n ≥ 1.

Proof To prove the proposition it is sufficient to prove the commutativity of the

rectangular diagram as shown in Fig. 5.32 for every n ≥ 1. The commutativity of the
diagram in Fig. 5.32 follows from definitions of ∂, the induced transformations f ∗
and g∗ for every n ≥ 1. 

Remark 5.11.16 More algebraic properties are available in Sects. 5.11.5 and 5.13.3.
366 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.32 Commutativity of

the rectangle involving ∂ and
induced transformations

5.11.5 Functorial Property of the Relative Homotopy Groups

This section continues the study homotopy theory by proving the functorial property
of the relative homotopy groups πn (X, A, x0 ), ∀ n ≥ 2 and homotopy properties
of maps f ∈ F n (X, A, x0 ) and also considers the homotopy equivalence of a map
f ∈ F n (X, A, x0 ).

Proposition 5.11.17 For any two homotopic maps f, g : (X, A, x0 ) → (Y, B, b0 ),

their induced transformations

f ∗ : πn (X, A, x0 ) → πn (Y, B, b0 )

and

g∗ : πn (X, A, x0 ) → πn (Y, B, b0 )

are the same for every n.

Proof By hypothesis, the maps f, g : (X, A, x0 ) → (Y, B, b0 , ) are homotopic.

Then there exists a homotopy
Ht : f g

in F n (X, A, x0 ). To prove the proposition, it is sufficient to show that f ∗ (β) =

g∗ (β), ∀ β ∈ πn (X, A, x0 ). We now consider two possible cases.
Case I: Let β ∈ πn (X, A, x0 ) be an arbitrary element. Then for n = 0, A = {x0 }
and B = {y0 }, the element β gives a path component of X . If x ∈ β, then f ∗ (β) and
g∗ (β) are path components of Y containing the points f (x) and g(x), respectively.
Define a path
β : I → Y, t → Ht (x).

Then the path β joins the point f (x) to the point g(x) and hence f ∗ (β) = g∗ (β), ∀ β ∈
πn (X, A, x0 ). This proves that f ∗ = g∗ .
Case II: Let β ∈ πn (X, A, x0 ) be an arbitrary element. For n > 0, take a map

h ∈ Fn (X, A, x0 ) : [h] = β.

Then f ∗ (β) and g∗ (β) are represented by the maps f ◦ h and g ◦ h, respectively.
Consider the homotopy
Ht ◦ h : f ◦ h g ◦ h.
5.11 The Relative Homotopy Groups 367

It implies that f ∗ ([h]) = g∗ ([h]) and hence f ∗ (β) = g∗ (β), ∀ β ∈ πn (X, A, x0 )

asserts f ∗ = g∗ . 

Definition 5.11.18 Let f : (X, A, x0 ) → (Y, B, y0 ) be a continuous map. It is said

to be a homotopy equivalence if there exists a continuous map g : (Y, B, y0 ) →
(X, A, x0 ) such that
(i) g ◦ f is homotopic to the identity map on (X, A, x0 ) and
(ii) f ◦ g is homotopic to the identity map on (Y, B, y0 ).

The above discussion is summarized in a basic result formulated in Proposition

5.11.19.

Proposition 5.11.19 (Homotopy invariance) Every homotopy equivalence f :

(X, A, x0 ) → (Y, B, b0 ) induces an isomorphism

f ∗ : πn (X, A, x0 ) → πn (Y, B, b0 ), ∀ n ≥ 2.

Corollary 5.11.20 Let X and Y be two path-connected topological spaces. If they

are homotopy equivalent, then the groups

πn (X, A, x0 ) ∼
= πn (Y, B, y0 ), ∀ n ≥ 2.

Remark 5.11.21 If the topological spaces X and Y are path connected and homotopy
equivalent, then the groups πn (X, A, x0 ) and πn (Y, B, y0 ) are isomorphic for every
n > 1.

Proof It follows from Proposition 5.11.19. 

Corollary 5.11.22 For every n ≥ 2, the groups πn (X, A, x0 ) depend on the homo-
topy type of the triplet (X, A, x0 ) of topological spaces.

Corollary 5.11.23 If A is a strong deformation retract of X , then the inclusion map

i : A → X induces isomorphisms

i ∗ : πn (A, x0 ) → πn (X, x0 ), ∀ n ≥ 1.

Proof It follows from Corollary 5.11.22. 

Theorem 5.11.24 If Ht p 2 denotes the homotopy category of triplets and their con-
tinuous maps, and Ab denotes the category of abelian groups and homomorphisms,
then
πn : Ht p 2 → Ab

is a covariant functor for every integer n > 2.

Proof For every integer n > 2, we utilize the assignments (X, A, x0 )→πn (X, A, x0 )
and f → f ∗ to define
368 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(i) the object function : (X, A, x0 ) → πn (X, A, x0 ) and

(ii) the morphism function : f → f ∗ .
Hence it follows that for every integer n > 2,

πn : Ht p 2 → Ab

is a covariant functor. 

5.11.6 Role of Cells and Spheres in Computing Homotopy

Groups

There are some standard homotopy tricks to compute homotopy groups. Certain
bundles are used in such computations. This subsection conveys such a technique.
Definition 5.11.25 A pair of an n-cell and its boundary, denoted by (E n , S n−1 ),
consists of a topological space E n and a subspace homeomorphic to (In , ∂ In ).
Example 5.11.26 The Euclidean n-cell Dn defined by Dn = {x ∈ Rn : ||x|| ≤ 1}
and its boundary S n−1 defined by S n−1 = {x ∈ Rn : ||x|| = 1} is a pair of an n-cell
and its boundary. This pair is considered as the prototype.
Definition 5.11.27 An n-cell (E, S) is said to be oriented in the sense that there is
an orientation of (E, S) obtained by a choice of a generator τn of Hn (E, S).
Given a base point s0 ∈ S, let f : (E, S, s0 ) → (X, A, x0 ) be a continuous map
of triplets. We take a continuous map

g : (In , In−1 , Jn−1 ) → (E, S, s0 )

such that g∗ (u n ) = τn . We may suppose that g maps In − @ I n topologically onto

(E − Smaps In − @ I n topologically onto (E − S and maps In−1 − @ I n−1 ) topo-
logically onto (S, s0 ), because, if Jn−1 is pinched to a point, then the quotient space
of (In , In−1 , Jn−1 ) is homeomorphic to (E, S, s0 ). This implies that the composite
map f ◦ g ∈ F n (X, A, x0 ). If

h : (In , In−1 , Jn−1 ) → (E, S, s0 )

is a second map satisfying the property of g, then g∗ (u n ) = h ∗ (u n ), and hence it

follows that both g and h represent the generator of πn (E, S, s0 ). This proves that
g h in F n (E, S, s0 ). This asserts that the homotopy class of f ◦ g depends only
on f and it determines a unique element c( f ) ∈ πn (X, A, x0 ). The element c( f )
is called the element represented by the map f. Since a homotopy of f sending
S to A and s0 to x0 gives rise to a homotopy of f ◦ g, it follows that the element
c( f ) depends only on the homotopy class of f. Given a map k ∈ F n (X, A, x0 ), let
5.11 The Relative Homotopy Groups 369

f = k ◦ g −1 . The map f is single valued and is continuous ( g −1 may not be single

valued). This implies that k is of the form f ◦ g. Hence it follows that every element
σ ∈ πn (X, A, x0 ) is a c( f ) for some continuous map f : (E, S, s0 ) → (X, A, x0 ).

Summarizing the above discussion a basic result essentially due to Hurewicz is

given in Theorem 5.11.28.
Theorem 5.11.28 Let (E, S) be an oriented n-cell with s0 ∈ S. Then the homotopy
classes of continuous maps f : (E, S, s0 ) → (X, A, x0 ) completely determine the
homotopy group πn (X, A, x0 ) of (X, A, x0 ).
Corollary 5.11.29 Let S n be oriented by a choice of a generator τn of Hn (S n ), where
S n is fixed n-sphere. Then the homotopy classes of continuous maps f : (S n , s0 ) →
(X, x0 ) completely determine the absolute homotopy group πn (X, x0 ) of (X, x0 ).
Remark 5.11.30 Original definition of the absolute homotopy group πn (X, x0 ) given
by Hurewicz in 1935 is embedded in Corollary 5.11.29. Its elements are homotopy
classes of continuous maps f : (S n , s0 ) → (X, x0 ) completely determine the abso-
lute homotopy group πn (X, x0 ) of (X, x0 ).

5.12 Isomorphism on πn (X, x0 ) Induced by a Curve and

Role of Base Point

This section studies the role of base points of the homotopy groups πn (X, x0 ) and the
properties of the homomorphisms induced by a curve on X. Its generalization in the
context of isomorphisms of πn (X, A, x0 ) induced by continuous curve σ : I → A
and the role played by the base point x0 is available in 5.22.2.
Proposition 5.12.1 Let X be any topological space and ψ : I → X be a path con-
necting two given points x0 , x1 ∈ X. Then ψ induces the identity map

ψ0 : π0 (X, x1 ) → π0 (X, x0 ).

Proof By hypothesis, ψ(0) = x0 and ψ(1) = x1 . Since π0 (X, x1 ) denotes the set of
all path components of X containing the point x1 and similarly π0 (X, x0 ) denotes
the set of all path components of X containing the point x0 and the points x0 , x1 lie
in the same path component, the neutral elements of both π0 (X, x1 ) and π0 (X, x0 )
are the same. Denote by

ψ0 : π0 (X, x1 ) → π0 (X, x0 )

the identity map on π0 (X, x1 ) = π0 (X, x0 ). 

Theorem 5.12.2 Let X be any topological space and given any two points x0 , x1 ∈
X, let ψ : I → X be a path from x0 to x1 ∈ X. Then ψ induces an isomorphism
370 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Fig. 5.33 Commutativity of

the rectangle involving ∂ and
induced transformations

ψ∗ = ψn : πn (X, x1 ) → πn (X, x0 ), ∀ n ≥ 1,

which depends only on the homotopy class of the path ψ relative to its end points.
Moreover,

(i) if ψ : X → x0 is the degenerate path, then

ψ∗ = ψn : πn (X, x0 ) → πn (X, x0 ), ∀ n ≥ 1

is the identity automorphism;

(ii) if φ : I → X is another path such that φ(0) = ψ(1), then

(ψ ∗ φ)∗ = (ψ ∗ φ)n = ψn ◦ φn ,

where ψ ∗ φ denotes of product path in X from x0 to x1 .

(iii) For any path ψ : I → X and every continuous map f : X → Y, the rectangular
diagram of groups and homomorphisms as shown in Fig. 5.33 is commutative
for every n ≥ 1, where f (ψ) = φ, f (x0 ) = y0 and f (x1 ) = y1 .

Proof By hypothesis, ψ(0) = x0 and ψ(1) = x1 . Given an arbitrary element σ ∈

πn (X, x1 ), let σ = [ f ] for some representative

f : (In , ∂ In ) → (X, x1 )

of σ. Define a partial homotopy of f

Ht : ∂ In → X, ∀ t ∈ I

by taking
Ht (∂ In ) = ψ(1 − t), ∀ t ∈ I.

Then Ht has a continuous extension by homotopy extension property of ∂ In in In .

H̃t : In → X, ∀ t ∈ I : H̃0 = f.

The homotopy H̃t is called a homotopy of f along the curve ψ. Since H̃1 sends
(∂ In ) into the point ψ(0) = x0 , it represents an element α ∈ π(X, x0 ). This element
depends only on σ. Consider the transformation
5.12 Isomorphism on πn (X, x0 ) Induced by a Curve and Role of Base Point 371

ψn : πn (X, x1 ) → πn (X, x0 ), ∀ n ≥ 1, σ → α,

where the element α ∈ π(X, x0 ) is defined as above. It depends only on the homotopy
class of the path ψ on X by using exercise 5.24.1 of Sect. 5.24. We claim that ψn is
an isomorphism.
ψn is a homomorphism. Let β, γ ∈ πn (X, x1 ) be two arbitrary elements be rep-
resented by two maps
f, h : (In , ∂ In ) → (X, x1 )

and
Ft , Ht : In → X, ∀ t ∈ I

be homotopies of f and h, respectively, along the curve ψ. Then F1 represents ψn (β)

and H1 represents ψn (γ). 

5.13 Homotopy Sequence and Its Basic Properties

This section conveys the concept of homotopy sequence of a triplet and proves
its exactness property. The homotopy groups πn and their homomorphisms such
as ∂ and f ∗ have basic properties in homotopy theory analogous to those enjoyed
by homology groups and their homomorphisms in homology theory (see Chap. 3).
Such properties provide powerful tools for the study of both homotopy and homology
theories, specially for computing homotopy and homology groups of some important
spaces. This section proves also some immediate consequences of the homotopy exact
sequences.

5.13.1 Homotopy Sequence of Triplets

This subsection defines homotopy sequence of the triplet (X, A, x0 ). Consider the
inclusion maps
i : (A, x0 ) → (X, x0 )

and
j : (X, x0 ) → (X, A, x0 ).

They induce transformations

i ∗ : πn (A, x0 ) → πn (X, x0 ),

and
 372 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

j∗ : πn (X, x0 ) → πn (X, A, x0 ).

Then i ∗ , j∗ and the boundary operator ∂ constitute together a beginningless sequence

j∗ ∂ i∗ j∗
· · · → πn+1 (X, x0 ) −−−−→ πn+1 (X, A, x0 ) −−−→ πn (A, x0 ) −−−→ πn (X, x0 ) −−−−→ πn (X, A, x0 )
∂ ∂ i∗
−−−→ · · · π1 (X, A, x0 ) −−−→ π0 (A, x0 ) −−−→ π0 (X, x0 )
(5.1)

This sequence denoted by π(X, A, x0 ) is known as the homotopy sequence of the

triplet (X, A, x0 ).

Every set in the sequence (5.1) has its zero element, and every transformation in
the sequence (5.1) (5.1) carries the zero element into the zero element.

Definition 5.13.1 (Exactness property homotopy sequence) The homotopy sequence

(5.1) of any triplet (X, A, x0 ) is said to be exact if at every term of the sequence (5.1)
except the last one, the kernel of every homomorphism on the right coincides with
the image of the preceding homomorphism on the left. In other words, the homotopy
sequence (5.1) is said to be exact if for n ≥ 0,
(i) Im j∗ = ker ∂ in πn+1 (X, A, x0 );
(ii) Im ∂ = ker i ∗ in πn (A, x0 );
(iii) Im i ∗ = ker j∗ in πn (X, x0 ).

Remark 5.13.2 The exactness property of the homotopy sequence (5.2) of any triplet
(X, A, x0 ) is proved in Theorem 5.13.8.

5.13.2 Elementary Properties

This subsection proves some basic properties formulated in Propositions 5.13.3–

5.13.7.

Proposition 5.13.3 If f : (X, A, x0 ) → (X, A, x0 ) is the identity map, then

f ∗ : πn (X, A, x0 ) → πn (X, A, x0 )

is the identity transformation.

Proof It follows from Proposition 5.11.13. 

Proposition 5.13.4 For any two continuous maps f : (X, A, x0 ) → (Y, B, y0 ) and
g : (Y, B, y0 ) → (Z , C, z 0 ), their induced transformations

f ∗ : πn (X, A, x0 ) → πn (Y, B, b0 )
5.13 Homotopy Sequence and Its Basic Properties 373

and
g∗ : πn (Y, B, y0 ) → πn (Z , C, z 0 )

are such that they satisfy the property

(g ◦ f )∗ = g∗ ◦ f ∗ : πn (X, A, x0 ) → πn (Z , C, z 0 )

for each dimension n.

Proof It follows from Proposition 5.11.14. 

Proposition 5.13.5 For any continuous map f : (X, A, x0 ) → (Y, B, y0 ), if k is its

restriction, i.e., if
k = f | A : (A, x0 ) → (B, y0 ),

then
∂ ◦ f ∗ = k∗ ◦ ∂.

Proof It follows from Proposition 5.11.15. 

Proposition 5.13.6 For any two continuous maps f : (X, A, x0 ) → (Y, B, y0 ) and
g : (Y, B, y0 ) → (Y, B, y0 ), if they are connected by a homotopy Ht which maps
A × I into B and carries {x0 } × I into y0 , then f ∗ = g∗ for every n.

Proof It follows from Proposition 5.11.17. 

Proposition 5.13.7 If X = {x0 } consists of single point, then πn (X, x0 ) consists of

only the zero element at each dimension.

Proof It follows from Definition of πn (X, x0 ). 

5.13.3 Exactness Property

This subsection proves the exactness property of the homotopy sequence π(X, A, x0 )
of the triplet (X, A, x0 ).

Theorem 5.13.8 (Exactness property) The sequence

j∗ ∂ i∗ j∗
· · · → πn+1 (X, x0 ) −−−−→ πn+1 (X, A, x0 ) −−−→ πn (A, x0 ) −−−−→ πn (X, x0 ) −−−−→ πn (X, A, x0 )
∂ ∂ i∗
−−−→ · · · π1 (X, A, x0 ) −−−→ π0 (A, x0 ) −−−−→ π0 (X, x0 )
(5.2)

is exact.

Proof To prove the exactness of the given homotopy sequence, it is sufficient to

show that for every n ≥ 1,
374 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(i) Im j∗ = ker ∂ in πn+1 (X, A, x0 ), which proves the exactness at the term
πn+1 (X, A, x0 );
(ii) Im ∂ = ker i ∗ , which proves the exactness at the term πn (A, x0 );
(iii) Im i ∗ = ker j∗ , which proves the exactness at the term πn (X, x0 ).

(i) Proof of (i) To show that Im j∗ = ker ∂ in πn+1 (X, A, x0 ), take an element
α ∈ πn+1 (X, x0 ). Suppose f ∈ F n+1 (X, x0 ) represents the element α. Then f
is an element in F n+1 (X, x0 ) such that it represents the element j ∗ (α). By
definition of F n+1 (X, x0 ), it follows that f carries In into the point x0 . This
implies that the restricted map ∂ f = f |In sends In into the point x0 and hence
∂ f represents the zero element of πn (A, x0 ). This asserts that (∂ ◦ j∗ )(α) = 0
showing I m j∗ ⊂ ker ∂. For the reverse inclusion, suppose f ∈ F n+1 (X, A, x0 )
represents an element α∈πn+1 (X, A, x0 ) such that ∂(α) = 0. Then the map ∂ f
is homotopic to the constant map. Hence there exists a homotopy

Ht : In → A : H0 = f |In , H1 (In ) = x0 and Ht (∂In ) = x0 , ∀ t ∈ I.

Construct a map

Ht (s), if s ∈ In
Ft : ∂In+1 → A, s →
x0 , if s ∈ Jn

for all t ∈ I. Then F0 = f |∂In+1 . Hence by homotopy extension property (HEP),

the homotopy Ft has a extension F t : In+1 → X such that F0 = f . Again since
 
F1 (∂I ) = F1 (∂I ) = x0 , F1 represents an element [g] ∈ πn+1 (X, x0 ).
n+1 n+1

Since F t ∈ F n+1 (X, A, x0 ), it follows that j∗ ([g]) = α. If n = 1, [ f ] is rep-

resented by a path f : I → X such that f (0) ∈ A and f (1) = x0 . The given
condition ∂(α) = 0 shows that f (0) is contained in the same path compo-
nent of A as x0 . Hence there exists a homotopy Ft : I → X such that F0 =
f, Ft (0) ∈ A, Ft (1) = x0 and F1 (0) = x0 . Consequently, F1 represents an ele-
ment [g] ∈ π1 (X, x0 ), and hence the homotopy Ft shows that j∗ ([g]) = α. Thus.
ker ∂ ⊂ Im j∗ . Combining the above inclusions, it follows that

Im j∗ ⊂ ker ∂ ⊂ Im j∗ .

This implies that I m j∗ = ker ∂ and hence it completes the proof of exactness of
the given homotopy sequence at the term πn+1 (X, A, x0 ).
(ii) Proof of (ii) To show that Im ∂ = ker i ∗ , take an element α ∈ πn (X, A, x0 ) and
suppose f ∈ F n (X, A, x0 ) represents the element α and hence α = [ f ]. Then
the element (i ∗ ◦ ∂)(α) is represented by g = f |In−1 . Construct a homotopy

G t : In−1 → X, (t1 , . . . , tn−1 ) → f (t1 , t2 , . . . , tn−1 , t), ∀ t ∈ I

5.13 Homotopy Sequence and Its Basic Properties 375

Then G 0 = g, G 1 (In−1 ) = x0 and G t ∈ F n−1 (X, x0 ) for n > 1. Hence (i ∗ ◦

∂)(α) = 0 implies i ∗ ◦ ∂ = 0. This shows that Im ∂ ⊂ ker i ∗ .

For the reverse inclusion, first let n > 1 and α ∈ πn (A, x) be represented by f ∈
F n−1 (A, x0 ) such that i ∗ (α) = 0. Then there exists a homotopy Ft : In−1 → X
such that F0 = f, F1 (In−1 ) = x0 and Ft (∂In−1 ) = x0 . Define a map

g : I n → X, (t1 , t2 , . . . , tn−1 , tn ) → Ftn (t1 , t2 , . . . , tn−1 ).

Then g ∈ F n (X, A, x0 ) and represents an element [g] ∈ πn (X, A, x0 ). Since

g|In1 = f, it follows that ∂([g]) = α.
For the case, when n = 1, the element α is a path component of A and i ∗ (α) = 0
implies that α is contained in the path component of X that contains the point
x0 . Take any point x ∈ X and define a path

f : I → X : f (0) = x and f (1) = x0 .

Then this path f represents an element β ∈ π1 (X, A, x0 ). Since f (0) is a point

of α, it follows that ∂(β) = α. Combining the above inclusions, it shows that

Im ∂ ⊂ ker i ∗ ⊂ Im ∂.

This implies that I m ∂ = ker ı∗ and hence it completes the proof of exactness
of the given homotopy sequence at the term πn (A, x0 ).
.
(iii) Proof of (iii) Im i ∗ = ker j∗ : First we show that Im i ∗ ⊂ ker j∗ . For n ≥ 1, take
any element α ∈ πn (A, x0 ) and choose a representative f ∈ F n (A, x0 ) of the
element α. Then the element ( j∗ ◦ i ∗ )(α) ∈ πn (X, A, x0 ) is represented by the
map j ◦ i ◦ f ∈ F n (X, A, x0 ). Since ( j ◦ i ◦ f )(In ) ⊂ A, it follows by Propo-
sition 5.11.9 that ( j∗ ◦ i ∗ )(α) = 0, since α ∈ πn (A, x0 ) is an arbitrary element.
This shows that j∗ ◦ i ∗ = 0. Hence Im i ∗ ⊂ ker j∗ .

For the reverse inclusion, take an element α ∈ πn (X, x0 ) be such that j∗ (α) = 0
and choose a function f ∈ Fn (X, x0 ) to represent the element α. Then j∗ (α) = 0
shows that there is a homotopy Ft : In → X such that F0 = f, F1 (In ) = x0 , and
Ft ∈ F n (X, A, x0 ). Define a homotopy

F2tn (t1 , t2 , . . . , tn−1 , 0), if 0 ≤ 2tn ≤ t
G t : In → X, (t1 , t2 , . . . , tn−1 , tn ) → n −t
Ft (t1 , t2 , . . . , tn−1 , 2t2−t ), if t ≤ 2tn ≤ 1

Then G 0 = f, G 1 (In ) ⊂ A and G t (∂In ) = x0 for all t ∈ I . Hence G 1 represents

an element [g] ∈ πn (A, x0 ) and the homotopy G t shows that i ∗ ([g]) = α. Hence
ker j∗ ⊂ Im i ∗ .
Combining the above inclusions, it follows that
376 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Im i ∗ ⊂ ker j∗ ⊂ Im i ∗ .

This implies that Im i ∗ = ker j∗ and hence it completes the proof of the exactness
of the given homotopy sequence at the term πn (X, x0 ). 

Definition 5.13.9 gives a very useful extension of the concept of homotopy

sequence of (X, A, x0 ) for the triple (X, A, B), where X, A and B are three topo-
logical spaces such that X ⊃ A ⊃ B and x0 ∈ B is a base point.

Definition 5.13.9 (Generalized homotopy sequence of (X, A, B)) This is the

sequence

j∗ ∂ i∗ j∗
· · · → πn+1 (X, B) −−−−→ πn+1 (X, A) −−−→ πn (A, B) −−−→ πn (X, B) −−−−→ πn (X, A) −
→
···
(5.3)
where i : A → X and j : B → A are inclusion maps. The operator ∂ is the com-
posite

∂ k∗
πn+1 (X, A) −−−→ πn (A) −−−→ πn (A, B)
(5.4)

where k is the inclusion map. The sequence (5.3) ends at the term π2 (X, A) and
reduces to the sequence of a pair when B = {x0 }. This sequence is called generalized
homotopy sequence of (X, A, B).

Theorem 5.13.10 (Exactness property of generalized homotopy sequence) The

sequence (5.3) is exact.

Proof Proceed as in Theorem 5.13.8. 

5.13.4 Homomorphism and Isomorphism of Homotopy

Sequences

This subsection studies homotopy sequences induced by a continuous map of triplets.

Definition 5.13.11 Let f : (X, A, x0 ) → (Y, B, y0 ) be a continuous map. If f 1 :

(X, x0 ) → (Y, y0 ) and f 2 : (A, x0 ) → (B, y0 ) are continuous maps induced by f,
then the maps f, f 1 and f 2 induce homomorphisms of their corresponding homotopy
groups as shown in the diagram in Fig. 5.34.

Proposition 5.13.12 The diagram in Fig. 5.34 is commutative.

5.13 Homotopy Sequence and Its Basic Properties 377

Fig. 5.34 Homomorphism of homotopy sequences

Proof Propositions 5.13.4 and 5.13.5 together prove the commutativity of each rect-
angle in the diagram. 
Remark 5.13.13 The commutative diagram in Fig. 5.34 is called the homomor-
phism of the homotopy sequence of (X, A, x0 ) into the homotopy sequence of
(Y, B, y0 ). This two sequences are connected by the homomorphisms by the induced
homomorphisms f ∗, f 1∗ and f 2∗ making each rectangle in the diagram commuta-
tive. In particular, if f is a homeomorphism, then f induces an isomorphism of the
homotopy of (X, A, x0 ) into the homotopy sequence of (Y, B, y0 ).

5.14 Homotopy Groups of Cells and Spheres

This section studies isomorphism induced by a homotopy equivalence of one homo-

topy sequence into the homotopy sequence of the other.
Definition 5.14.1 Let f : (X, A, x0 ) → (Y, B, y0 ) be a continuous map. It is said to
be a homotopy equivalence if there is a continuous map g : (Y, B, y0 ) → (X, A, x0 )
such that
(i) g ◦ f is homotopic to the identity map of (X, A, x0 ); and
(ii) f ◦ g is homotopic to the identity map of (Y, B, y0 ). The homotopies shift the
subspaces A and B on themselves and keep the points x0 and y0 fixed.
Proposition 5.14.2 Every homotopy equivalence f : (X, A, x0 ) → (Y, B, y0 )
induces an isomorphism of the homotopy sequence of (X, A, x0 ) onto the homo-
topy sequence of (Y, B, y0 ).
Proof By hypothesis, f : (X, A, x0 )→(Y, B, y0 ) is a homotopy equivalence. Hence
it follows from Proposition 5.13.6 that f induces an isomorphism of the homotopy
sequence of (X, A, x0 ) onto the homotopy sequence of (Y, B, y0 ). 
Proposition 5.14.3 Let f : (X, A, x0 ) → (X, A, x0 ) be the identity map. If f is
homotopic to the constant map c : (X, A, x0 ) → x0 , then every homotopy group of
(X, A, x0 ) contains only the zero element.
Proof By hypothesis, the identity map f : (X, A, x0 ) → (X, A, x0 ) is homotopic to
the constant map c : (X, A, x0 ) → x0 . Hence it follows by using Proposition 5.13.7
that every homotopy group of (X, A, x0 ) contains only the zero element. 
378 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Proposition 5.14.4 Let E be an m-cell (open or closed) and x0 ∈ E be any point.

Then πn (E, x0 ) = 0 at every dimension n.
Proof Since every cell E is contractible to any of its points, the identity map of E is
homotopic to constant map. Hence the proposition follows from Proposition 5.14.3.

Let (E, S) be a closed m-cell in the sense that it is homeomorphic to (Im , ∂In )
and let x0 ∈ S be any point. Consider the part (5.5) of the homotopy sequence of
(E, S, x0 )
j∗ ∂ i∗
πn+1 (E, x0 ) −−−→ πn+1 (E, S, x0 ) −−−→ πn (S, x0 ) −−−→ πn (E, x0 )
··· (5.5)

Since πn+1 (E, x0 ) = 0, the image I m j∗ = 0 and hence ker ∂ = 0 by exactness

property of the sequence (5.5 ). Again, since πn (E, x0 ) = 0, the kernel ker i ∗ =
πn (S, x0 ). Hence by the exactness property of the sequence, it follows that I m ∂ =
πn (S, x0 ). This implies that

∂ : πn+1 (E, S, x0 ) ∼
= πn (S, x0 ).

The above discussion is summarized in a basic more general result given in Propo-
sition 5.14.5.
Proposition 5.14.5 If every third term of an exact sequence of groups and homo-
morphisms is zero, then the remaining adjacent pairs of groups are isomorphic.
Remark 5.14.6 Two basic theorems of homotopy theory embodied in Theorems
5.14.7 and 5.14.8 have been proved in Chap. 2 by applying Freudenthal suspension
theorem saying that

σn : πn (S n ) → πn+1 (S n+1 ), [α] → [α̃]

is an isomorphism for m < 2n − 1 and is surjective for m ≤ 2n − 1, where α̃ :

S n+1 → S n+1 is an continuous extension of the continuous map α : S n → S n deter-
mined uniquely upto homotopy. On the other hand, the proof of the first theorem
by simplicial approximation theorem and the second theorem by using Hurewicz
natural homomorphism

ψ : πn (X, A, x0 ) → Hn (X, A)

are available in [Steenrod, 1951]. Hurewicz homomorphism and relation between

homology and homotopy groups have already been studied in Chap. 3. The impor-
tance of the natural homomorphism between homotopy and homology theory is
realized from the statement of Theorem 5.14.9.
Theorem 5.14.7 The homotopy groups πm (S n ) = 0 for m < n.
5.14 Homotopy Groups of Cells and Spheres 379

Proof See Chap. 2 or see proof of Theorem 5.18.10. 

Theorem 5.14.8 The first nontrivial homotopy groups of the n-sphere S n is the
πn (S n ), and it is the infinite cyclic group.

Proof See Chap. 2. For an alternative proof see [Steenrod, 1951, p. 78]. 

Theorem 5.14.9 gives a sufficient condition for equality of the groups πn (X, A, x0 )
and Hn (X, A) upto an isomorphism.

Theorem 5.14.9 (Isomorphism theorem of Hurewicz ) Let (X, A, x0 ) be a triplet

such that
(i) the subspace A is simply connected;
(ii) X is also simply connected.
(iii) Moreover, if πm (X, A, x0 ) = 0 for 2 ≤ m < n, then

ψ : πn (X, A, x0 ) → Hn (X, A)

is an isomorphism.

Proof The proof is very long (see [Hu, 1933] ). 

Corollary 5.14.10 If the triplet (X, A, x0 ) satisfies the condition of Theorem 5.14.9,
then Hm (X, A) = 0 for 1 ≤ m < n.

Proof By hypothesis, πn (X, A, x0 ) = 0 for 2 ≤ m < n. Hence the Hurewicz homo-

morphism ψ maps isomorphically onto Hn (X, A). Conversely, if (X, A, x) satis-
fies the first two conditions and if Hm (X, A) = 0 for 2 ≤ m < n, then by itera-
tion, it can be proved that πn (X, A, x0 ) = 0 for n = 2 followed by m = 2, 3, . . . , n.
Hence it follows that the condition (iii) can be replaced by the condition saying that
Hm (X, A) = 0 for 2 ≤ m < n. Under the conditions (i), (ii) and (iii’), the result still
holds. 

Corollary 5.14.11 The first nontrivial homology group and the first nontrivial
homotopy group have the same dimension and are isomorphic under the natural
isomorphism.

Corollary 5.14.12 πn (S n ) ∼
= Hn (S n ) ∼
= Z.

Corollary 5.14.13 If (E n , S n−1 ) is an n-cell with boundary S n−1 and x0 ∈ S n−1 ,

then
πn (E n , S n−1 , x0 ) ∼
= Hn (E n , S n−1 ) ∼
= Z.
380 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

5.15 Stable Homotopy Groups

This section discusses stable homotopy groups based on the Freudenthal suspension
theorem saying that suspension homomorphism

E : πm (S n ) → πm+1 (S n+1 )

is an isomorphism for m < 2n − 1 and is onto for m ≤ 2n − 1. [Freudenthal, 1938],

which has been discussed in Chap. 2. It is a natural generalization of Freudenthal
suspension theorem. The term ‘stable’ is used in topology if a phenomenon occurs
essentially in the same way which is independent of dimension provided that the
dimension is sufficiently large. Stable homotopy theory is an important theory in
algebraic topology, and it witnessed its greatest development in the late 1950s. It
is applied to develop spectral homology and cohomology theory and specially K -
theory by Atiyah and Hirzebruch. R. Thom (1923–2002) used this theory to reduce
the problem of classifying manifolds up to cobordism to a solvable problem in stable
homotopy theory.
For stable homotopy theory with applications in homology and cohomology the-
ories, the paper [Adams, 1974], the books [Adhikari, 1916] and [Gray, 1975] are
referred.

5.15.1 Stable Homotopy Groups

Consider an n-connected C W -complex and the suspension map

 : πr (X ) → πr +1 (X )

is an isomorphis m for r < 2n + 1. In particular, for r ≤ n the suspension map  is

an isomorphism and  X is an (n + 1)-connected CW-complex. Now, consider the
sequence ( 5.6 ) of groups and homomorphisms

πr (X ) −
→ πr +1 ( X ) −
→ ··· −
→ πr +m ( m X ) −
→ ··· (5.6)

The homomorphism
πr +m ( m X ) → πr +m+1 ( m+1 X )

is an isomorphism for r + m < 2(n + m) + 1, i.e., for m > r − 2n + 1, because,

 m X is (n + m)-connected. This implies that for fixed n and r and sufficiently large
enough m, all the groups in the sequence of groups and homomorphisms (5.6) are
isomorphic. This leads to the concept of the stable homotopy group denoted by πrs (X )
of X, for which the sequence (5.6 ) is stabilized.
5.15 Stable Homotopy Groups 381

Definition 5.15.1 (Stable homotopy group) Given an (n − 1)-connected space X,

for r ≥ 0, the r -th stable homotopy group of X, denoted by πrs (X ), is defined to be
the group

πn+m ( X ), ∀ m > r − 2n + 1.

This is well-defined, since, adding any finite number of groups and homomor-
phisms to the beginning of (5.6) does not affect the resulting stable homotopy group.

5.15.2 Some Examples

Example 5.15.2 Consider the homotopy exact sequence (5.8) of the fibration: p :
S3 → S2 :
p∗
· · · → π3 (S 1 , s0 ) −
→ π3 (S 3 , s0 ) −−−−→ π3 (S 2 , s0 ) −
→ π2 (S 1 , s0 ) → · · · (5.7)

The exactness property of the above sequence (5.7) of groups and homomorphisms
asserts that the homomorphism p∗ induced by p

p∗ : π3 (S 3 , s0 ) → π3 (S 2 , s0 )

is an isomorphism, because π3 (S 1 , s0 ) = ∼ π2 (S 1 , s0 ) = 0. This implies that

π3 (S , s0 ) ∼
2
= Z. It provided the first example given by H. Hopf, where πm (S n , s0 ) = 0
for m > n. The map p is known the Hopf map, studied earlier in Chap. 2.
Example 5.15.3 For each q, consider

   
→∼
π2q+2 (S q+2 , s0 ) − →∼
=π2q+3 (S q+3 , s0 ) − →∼
=··· − →∼
=πq+n (S n , s0 ) − =···

We denote the common group πn+q (S n , s0 ), by πqS . It is called the kth stable homo-
topy group for the fibration: p : S 3 → S 2 .
Theorem 5.15.4 (Hopf) π3 (S 2 ) ∼
= π2 (S 2 ) ∼
= Z.
Proof Consider the Hopf fibration

S 1 → S 3 → S 2

and its corresponding homotopy exact sequence (5.8):

p∗
· · · → π3 (S 1 , s0 ) −
→ π3 (S 3 , s0 ) −−−−→ π3 (S 2 , s0 ) −
→ π2 (S 1 , s0 ) → · · · (5.8)

Proceed as Example 5.15.2 to prove the theorem using the result that π2 (S 2 ) ∼
= Z.

382 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

5.16 Consequences of Homotopy Exact Sequence

This section proves some results in homotopy theory as consequences of the exactness
property of the homotopy sequence (5.2).

5.16.1 Direct Sum Theorems in Higher Homotopy Groups

This section proves direct sum theorems as immediate consequences of the exact-
ness property of the homotopy sequence (5.2).

Theorem 5.16.1 Let (X, A, x0 ) be a triplet and r : X → A be a retraction of X. If

x0 ∈ A and i : A → X and i : (X, A) → (X, A, x0 ) are inclusion maps, then

πn (X, x0 ) ∼
= πn (A, x0 ) ⊕ πn (X, A, x0 )

for any n ≥ 2 and the inclusion map i : A → X induces a monomorphism

i ∗ : πn (A, x0 ) → πn (X, x0 ), ∀ n ≥ 1.

Proof By hypothesis, r ◦ i = 1 A . Hence the induced homomorphism

r∗ ◦ i ∗ : πn (A, x0 ) → πn (A, x0 )

is the identity automorphism on πn (A, x0 ) for every integer n ≥ 1. This asserts that
i ∗ : πn (A, x0 ) → πn (X, x0 ) is a monomorphism and r∗ : πn (X, x0 ) → πn (A, x0 )
is an epimorphism for every n ≥ 1. Since the group πn (X, x0 ) is abelian. Again
for n ≥ 2, the relation r∗ ◦ i ∗ = 1d implies that the group πn (X, x0 ) decomposes
into the direct sum πn (X, x0 ) = G ⊕ K , where G = Im i ∗ and K = ker r∗ . Clearly.
G∼ = πn (A, x0 ), because, i ∗ is a monomorphism. The exactness of the homotopy
sequence (5.2) of the triplet (X, A, x0 ) asserts that j∗ : πn (X, x0 ) → πn (X, A, x0 )
is an epimorphism for every integer n ≥ 2 and ker j∗ = Im i ∗ = G and j∗ maps K
isomorphically onto πn (X, A, x0 ) and hence K ∼ = πn (X, A, x0 ). 

Corollary 5.16.2 Let (X, A, x0 ) be a triplet, r : X → A be a retraction of X. If

x0 ∈ A and i : A → X and j : (X, A) → (X, A, x0 ) are inclusion maps, then the
group πn (X, x0 ) decomposes into direct sum

πn (X, x0 ) ∼
= G ⊕ K , ∀ n ≥ 2,

where G is the image of i ∗ in πn (X, x0 ), and K is the kernel of the homomorphism

r∗ : πn (X, x0 ) → πn (A, x0 ).

Proof It follows from the proof of direct sum theorem 5.16.1. 

5.16 Consequences of Homotopy Exact Sequence 383

Remark 5.16.3 For a given triplet (X, A, x0 ), if A is a retract of X , then the group
π2 (X, A, x0 ) is abelian.

Definition 5.16.4 Let (X, A, x0 ) be a triplet. Then the topological space X is said
to be deformable into its subspace A relative to a point x0 ∈ A, if there exists a
homotopy Ht : X → X satisfying the following conditions

H0 (x) = x, H1 (x) ∈ A and Ht (x0 ) = x0 , ∀ x ∈ X, t ∈ I.

Theorem 5.16.5 Let (X, A, x0 ) be a triplet and the topological space X be

deformable into its subspace A relative to a point x0 ∈ A. If i : A → X be the
inclusion map, then πn (A, x0 ) decomposes as a direct sum

πn (A, x0 ) ∼
= πn (X, x0 ) ⊕ πn+1 (X, A, x0 ), ∀ n ≥ 2

and
i ∗ : πn (A, x0 ) → πn (X, x0 ) is an epimorphism, ∀ n ≥ 1.

Proof By the given conditions, there exists a homotopy

Ht : X → X : H0 (x) = x, H1 (x) ∈ A and Ht (x0 ) = x0 , ∀ x ∈ X, t ∈ I.

Then the continuous map

f : (X, x0 ) → (A, x0 ), x → H1 (x)

is such that i ◦ f = H1 H0 rel x0 . This implies that i ∗ ◦ f ∗ : πn (X, x0 )→πn (X, x0 )

is the identity automorphism on πn (X, x0 ). Consequently, f ∗ is a monomorphism,
and i ∗ is an epimorphism for every integer n ≥ 1. Since for every integer n ≥ 2,
the group πn (A, x0 ) is abelian, it follows from i ∗ ◦ f ∗ = 1d , the group πn (A, x0 )
decomposes as direct sum

πn (A, x0 ) = Im f ∗ ⊕ ker i ∗ ,

where i ∗ is an epimorphism and f ∗ is a monomorphism for any n ≥ 1. This implies

that Im f ∗ ∼
= πn (X, x0 ). Finally, the exactness of the homotopy sequence (5.2) of the
triplet (X, A, x0 ) asserts that

∂ : πn+1 (X, A, x0 ) → πn (A, x0 )

is a monomorphism and hence ker i ∗ = Im ∂ ∼

= πn+1 (X, A, x0 ). 

Corollary 5.16.6 Let (X, A, x0 ) be a triplet and i : A → X be the inclusion map.

If the continuous map f : (X, x0 ) → (A, x0 ) is such that the maps i ◦ f 1X :
(X, x0 ) → (X, x0 ), then the group πn (A, x0 ) decomposes as a direct sum
384 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

πn (A, x0 ) ∼
= πn (X, x0 ) ⊕ πn+1 (X, A, x0 ) = G ⊕ K , ∀ n ≥ 2,

where G = I m f ∗ and K = ker i ∗ are subgroups of πn (A, x0 ).

Proof It follows from the proof of Theorem 5.16.5. 

5.16.2 Characterization of n-Connected Spaces

This subsection characterizes n-connected spaces in terms of homomorphisms of

homotopy groups induced by inclusion maps by using homotopy exact sequence.

Definition 5.16.7 A pair (X, A) of topological spaces is said to be

(i) 0-connected if every path component of X intersects A;
(ii) n-connected if (X, A) is 0-connected and πk (X, a) = 0 for 1 ≤ k ≤ n and ∀ a ∈
A.

Proposition 5.16.8 gives a characterization of n-connected spaces by homotopy

exact sequence.

Proposition 5.16.8 A pair (X, A) of topological spaces with inclusion map i :

(A, x0 ) → (X, x0 ) is n-connected (n ≥ 0) iff the induced map i ∗ : πk (A, x0 ) →
πk (X, x0 )
(i) is a bijection for k < n; and
(ii) it is a surjection for k = n and ∀ x0 ∈ A.

Proof It follows from the homotopy sequence (5.2) of the triplet (X, A, x0 ). 

5.17 Homotopy Sequence of Fibering and Hopf Fibering

of Spheres

This section is devoted to the study of the homotopy sequence of fibering. Moreover,
it describes Hopf fibering:

p : S 2n−1 → S n f or n = 2, 4, 8,

which are the early examples of bundles spaces. They are used in computing higher
homotopy groups of certain topological spaces. H. Hopf (1894–1975) described
various fiberings of spheres by spheres in his papers [Hopf, 1931, 1935]. The map
p : S 3 → S 2 of the 3-sphere on the 2-sphere defined by Hopf in 1935, known as a
Hopf map, is the simplest.
5.17 Homotopy Sequence of Fibering and Hopf Fibering of Spheres 385

5.17.1 Homotopy Sequence of Fibering

Consider the projection p : X → B. Let b0 ∈ B be the base point of B and F =

p −1 (b0 ) = ∅ be the fiber space of p. For x0 ∈ F, this subsection studies the homo-
topy sequence of the triplet (X, F, x0 ).

Define a continuous map

q : (X, F, x0 ) → (B, b0 ) : p = q ◦ j,

where j : (X, x0 ) → (X, F, x0 ) is the inclusion map. Then q induces a bijection

q∗ : πn (X, F, x0 ) → πn (B, b0 )

for every n ≥ 1 by Exercise 5.24.1 of Sect. 5.24. Hence q∗−1 is an isomorphism.

Consider the homomorphism

k∗ : ∂ ◦ q∗−1 : πn (B, b0 ) → πn−1 (F, x0 )

for ever integer n ≥ 1. This produces a beginninless sequence

p∗ k∗ i∗ p∗
· · · −−−−→ πn+1 (B, b0 ) −−−→ πn (F, x0 ) −−−→ πn (X, x0 ) −−−−→ πn (B, x0 )
k∗ p∗ k∗ i∗
−−−→ · · · −−−−→ π1 (B, b0 ) −−−→ π0 (F, x0 ) −−−→ π0 (X, x0 )
(5.9)

Theorem 5.17.1 proves the exactness of the sequence (5.9), which is known as
the homotopy sequence of the fibering p : X → B based at the point x0 .
Theorem 5.17.1 (Homotopy sequence of the fibering) The sequence (5.9) is exact.
Proof The exactness of the sequence (5.9) follows from the exactness of the sequence
(5.2) proved in Theorem 5.13.8. 
Proposition 5.17.2 Let the fiber space F in the homotopy sequence (5.9) of fibering
is totally disconnected. Then the induced homomorphism

p∗ : πn (X, x0 ) → πn (B, b0 )

(i) is an isomorphism for n ≥ 2; and

(ii) p∗ is a monomorphism for n = 1.
Proof By the given condition, the groups πn (F, x0 ) = 0, ∀ n ≥ 1. Using exactness
property of the homotopy sequence (5.9), the proposition is proved. 
Theorem 5.17.3 If p : X → B be a fibering with a cross section s : B → X, such
that for every b0 ∈ B the element x0 = s(b0 ) ∈ F = p −1 (b0 ), then
386 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

∼ πn (B, b0 ) ⊕ πn (F, x0 ), ∀ n ≥ 2, and

(i) πn (X, x0 ) =
(ii) p∗ : πn (X, x0 ) → πn (B, b0 ) is an epimorphism for every n ≥ 1.
Proof By hypothesis s : B → X is a cross section. Hence p ◦ s = 1 B induces the
identity automorphism

p∗ ◦ s∗ : πn (B, b0 ) → πn (B, b0 ).

This implies that for n ≥ 1,

(i) s∗ : πn (B, b0 ) → πn (X, x0 ) is a monomorphism; and
(ii) p∗ : πn (X, x0 ) → πn (B, b0 ) is an epimorphism.
For the particular case, when n ≥ 2, the group πn (X, x0 ) is abelian. Since p∗ ◦
s∗ = 1d , it follows that πn (X, x0 ) is the direct sum of the groups

πn (X, x0 ) = Im s∗ ⊕ ker p∗ .

Then Im s∗ = ∼ πn (B, b0 ), because s∗ is a monomorphism. Finally, it is proved by using

the exactness property of the sequence (5.9) that i ∗ is a monomorphism for every
n ≥ 1, because, p∗ is an epimorphism for every n ≥ 1. This implies that ker p∗ =
Im i ∗ ∼
= πn (F, x0 ). 
Theorem 5.17.4 If (X, A, x0 ) is a triplet such that the subspace A is contractible
in X relative to a point x0 ∈ A, then
∼ πn (X, x0 ) ⊕ πn−1 (A, x0 ), ∀ n ≥ 3;
(i) πn (X, A, x0 ) =
(ii) the inclusion map i : A → X induces homomorphisms i ∗ mapping πn (A, x0 )
into the zero element of πn (X, x0 ) for every n ≥ 1.
Proof By hypothesis, A is contractible in X relative to a point x0 ∈ A. Hence there
exists a homotopy

Ft : A → X : F0 = i, F1 (A) = x0 , and Ft (x0 ) = x0 .

This implies that i ∗ maps πn (A, x0 ) into the zero element of πn (X, x0 ) for every
n ≥ 1. This proves (ii). Next we consider the case for n ≥ 2. Since i ∗ = 0, by using
the exactness property of the homotopy sequence (5.2) of (X, A, x0 ), it follows that
∂ is an epimorphism and j∗ is a monomorphism. Consider πn (X, A, x0 ) as an exten-
sion of πn (X, x0 ) by πn−1 (X, x0 ).
Let β ∈ πn−1 (A, x0 ) be represented by a continuous map g : (In−1 , ∂In−1 ) →
(A, x0 ). Define a continuous map

k : (In , In−1 , Jn−1 ) → (X, A, x0 ), (t1 , t2 , . . . , tn ) → (Ftn ◦ g)(t1 , t2 , . . . , tn−1 )

and a homomorphism

f ∗ : πn−1 (A, x0 ) → πn (X, A, x0 ), β → α = [k], ∀ n ≥ 2.

5.17 Homotopy Sequence of Fibering and Hopf Fibering of Spheres 387

Then the homomorphism ∂ ◦ f ∗ =1d is the identity automorphism of πn−1 (A, x0 ),

because, k|In−1 = g. This implies that f ∗ is a monomorphism for each n ≥ 2. If n ≥ 3,
then the group πn (X, A, x0 ) is abelian and hence ∂ ◦ f ∗ = 1d asserts by group theory
that the group πn (X, A, x0 ) decomposes into the direct sum

πn (X, A, x0 ) = Im f ∗ ⊕ ker ∂
∼
= πn−1 (A, x0 ) ⊕ πn (X, x0 ),

because f ∗ is a monomorphism implies I m f ∗ ∼

= πn−1 (A, x0 ) and j∗ is a monomor-
phism implies ker ∂ = Im j∗ ∼
= πn (X, x0 ). 

Corollary 5.17.5 Let X be a fiber space over a base space B with projection p :
X → B. If x0 ∈ B is a base point such that the fiber F = p −1 (b0 ) is contractible in
X, then the group πn (B, x0 ) decomposes into the direct sum

πn (B, x0 ) ∼
= πn (X, x0 ) ⊕ πn−1 (F, x0 ).

Proof Consider the homotopy sequence of the fibering p : X → B. Then the corol-
lary follows by using Theorem 5.17.4. 

Theorem 5.17.6 If F → X → B is a fiber bundle such that the inclusion map

i : F → X is homotopic to a constant map. Then the long exact homotopy sequence
(5.9) of the fibering p : X → B based at x0 is split into short exact sequences giving
isomorphisms

πn (B) ∼
= πn (X ) ⊕ πn−1 (F).

Proof By hypothesis the inclusion map i : F → X is homotopic to a constant map.

Hence its induced maps i ∗ : πn (F, x0 ) → πn (X, x0 ) appearing in the long exact
sequence (5.9) have the property i ∗ = 0. This produces a short exact sequence for
every n ≥ 1

p∗ k∗
→ πn (X, x0 ) −−−−→ πn (B, b0 ) −−−→ πn−1 (F, x0 ) −
0− → 0.

Finally, since p : X → B has the homotopy lifting property (HLP) with respect
to all Euclidean disks, there exists a splitting map ψ : πn (B, b0 ) → πn (X, x0 ) in the
sense that p∗ ◦ ψ is the identity map on πn (B, b0 ). Hence the map splits for every
n ≥ 1 the short exact sequence

p∗ k∗
→ πn (X, x0 ) −−−−→ πn (B, b0 ) −−−→ πn−1 (F, x0 ) −
0− → 0.

This asserts by group theory that the group πn (B, x0 ) decomposes into the direct
sum
πn (B, x0 ) ∼
= πn (X, x0 ) ⊕ πn−1 (F, x0 ). 
 388 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Corollary 5.17.7 For the Hopf fiberings

p : S 2n−1 → S n : n = 2, 4, 8

the groups πm (S n ) decomposes into the direct sum

(i)
πm (S n ) ∼
= πm (S 2n−1 ) ⊕ πm−1 (S n−1 ) : n = 2, 4, 8 and for all m ≥ 2

and in particular
(ii)

πm (S 4 ) ∼
= πm (S 7 ) ⊕ πm−1 (S 3 ) and πm (S 8 ) ∼
= πm (S 15 ) ⊕ πm−1 (S 7 ), ∀ m ≥ 2.

Proof Let F be the fiber in each of the Hopf fiberings p : S 2n−1 → S n for n = 2, 4, 8.
Since the fiber F is an (n − 1)-sphere S n−1 and is contractible in (2n − 1)-sphere
S 2n−1 , the corollary follows by using Corollary 5.17.5. 

Corollary 5.17.8 The Hopf bundles p : S 3 → S 2 , S 7 → S 4 and S 15 → S 8 produce

isomorphisms
(i) πm (S 2 ) ∼
= πm (S 3 ), ∀ m ≥ 3.
4 ∼
(ii) πm (S ) = πm−1 (S 3 ), if 2 ≤ m ≤ 6.
(iii) πm (S 8 ) ∼
= πm−1 (S 7 ), if 2 ≤ m ≤ 14.
4 ∼
(iv) π7 (S ) = Z ⊕ π6 (S 3 ).
(v) π15 (S 8 ) ∼
= Z ⊕ π14 (S 7 ).

Proof It follows from Corollary 5.17.7. 

Corollary 5.17.9 The groups π7 (S 4 ) and π15 (S 8 ) contain Z summands.

Proof Using the results given in Corollary 5.17.8 that π7 (S 4 ) ∼

= Z ⊕ π6 (S 3 ) and
8 ∼
π15 (S ) = Z ⊕ π14 (S ), the corollary follows.
7


5.17.2 Fiberings of Spheres Over Projective Spaces

This subsection considers the fiberings of spheres over projective spaces.

Theorem 5.17.10 (Fiberings of spheres over projective spaces) Let CP n denote the
n-dimensional complex space. Then
(i) π1 (CP n ) = 0,
(ii) π2 (CP n ) ∼
= Z,
(iii) πm (CP n ) ∼
= πm (S 2n−1 ) for m > 2.
5.17 Homotopy Sequence of Fibering and Hopf Fibering of Spheres 389

Proof Let S denote the unit sphere in the complex n-space Cn and CP n be the n-
dimensional complex space and q : Cn → CP n be the natural projection. Consider
the fibering p = q| S : S → CP n . Then S is an (2n − 1) sphere, and the fiber F is
a 1-sphere. For n = 1, CP n consists of a single point, and for n > 1, the exactness
property of the homotopy sequence of the fibering p : S → CP n asserts that
(i) π1 (CP n ) = 0;
(ii) π2 (CP n ) ∼
= Z;
(iii) πm (CP n ) ∼
= πm (S 2n−1 ) for m > 2. 
Theorem 5.17.11 Let S q → S m → S n is a fiber bundle. Then

q = n − 1 and m = 2n − 1.

Proof Suppose n ≤ m and q ≤ m and q + n = m. If q = m, then n = 0, and S 0 is not

connected. Since S m → S n is a surjection, this gives a contradiction and hence q <
m, and S q → S m is homotopic to a constant map. Consequently, πi (S n ) ∼ = πi (S m ) ⊕
πi−1 (S ), ∀ i > 0. This asserts that q > 0 and m > n. Then for i = 1, 2, . . . , n, it
q

follows that πi (S q ) = 0 if i < n − 1 and πn−1 (S q ) ∼

= Z. This implies that q = n − 1
and hence it follows that m = 2n − 1. 

5.17.3 Fiberings of Spheres by Spheres

This subsection is devoted to study Hopf fibering: p : S 2n−1 → S n , for n = 2, 4, 8.

There was a long-standing problem: whether a continuous map p : S k → S n for
k > n > 1 is necessarily nullhomotopic? H. Hopf (1894–1971) solved this problem
in 1935 through his discovery of famous map p : S 3 → S 2 , named after him. For the
fibering: p : S 2n−1 → S n , for n = 2, 4, 8, its fiber space F is S n−1 for n = 2, 4, 8.
The simplest of Hopf fiberings is the one with the Hopf map p : S 3 → S 2 , where

S 3 = {(z, w) ∈ C2 : zz + ww = 1},

and S 2 represents the complex projective line consisting of pairs [z, w] of complex
numbers, not both zero, determined by the equivalence relation

[z, w] ∼ [λz, λw] : λ = 0

and the projection map

p : S 3 → S 2 , (z, w) → [z, w].

The map is onto, because, every pair [z, w] can be normalized on division by
1
λ = (zz + ww) 2 .
390 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

The map p is well-defined, because for |λ| = 1, if (z, w) ∈ S 3 , then (λz, λw) ∈ S 3
and their image points are the same, i.e., p(z, w) = p(λz, λw). Conversely, if
p(z, w) = p(z , w ), then (z , w ) = (λz, λw) for some λ such that |λ| = 1. Con-
sequently, the inverse image of a point of S 2 under the map is given by the inverse
image of any point S 2 on multiplication by eiθ : 0 ≤ θ ≤ 2π. Geometrically, this
inverse image represents a great circle of S 3 , and hence S 3 is decomposed into a
family of great circles having S 2 as a decomposition space. Clearly, p is continuous.
Definition 5.17.12 (Hopf map) The continuous onto map

p : S 3 → S 2 , (z, w) → [z, w]

is called the Hopf map.

5.18 More Study on Hopf Map p : S3 → S2

This section generalizes the Hopf map p : S 3 → S 2 through the study of some spaces
that arise in projective geometry. The constructions of Hopf fiberings p : S 7 → S 4
from the quaternions and p : S 15 → S 8 from Calyley numbers are similar to that
of p : S 3 → S 2 , and they are described below. Recall from Chap. 4 that for F =
R, C or H, the right vector space Fn consists of elements, which are ordered sets of
n elements of F. If x = (x1 , x2 , . . . , xn ) ∈ Fn and α ∈ F, then xα = (x1 α, . . . , xn α).
n
Using the usual inner product x and y in F by x, y =
n
x i yi , where x i is the
1
conjugate of xi , it is proved that
(i) y, x = x, y;
(ii) xα, y = αx, y;
(iii) x, (yα) = x, yα.
Remark 5.18.1 The orthogonality relation is symmetric, because, x, y = 0 iff
y, x = 0. Let S be the unit sphere in Fn defined by the locus x, x = 1 and G n be
the orthogonal, unitary or sympletic group according as F = R, C or H, then each
G n is a compact Lie group. Let FP n be the projective space associated with F, and it
is topolozized by considering it as a quotient space of Fn+1 − {0}. It can be thought
n+1

of the set of all lines through the origin in F = F ⊕ F ⊕ · · · ⊕ F, since, every
n+1

point of Fn+1 − {0} determines a line through the origin 0 and if x and y are nonzero
elements of Fn+1 , then x ∼ y iff there is an element λ(= 0) ∈ F such that y = xλ.
This is an equivalence relation and defines FP n as the quotient set of equivalence
classes endowed with the quotient topology. In particular, R P n , CP n and HP n are
called n-dimensional real, complex and quaternionic projective spaces. The natural
projection map Fn+1 − {0} → FP n , w → [w] is continuous, and defines maps on
restriction to the unit sphere of Fn
 5.18 More Study on Hopf Map p : S 3 → S 2 391

pn : S n → R P n ,

qn : S 2n+1 → CP n ,

and
rn : S 4n+3 → HP n .

Usually, the common notation p is used instead of pn , qn or rn , unless there is any

confusion.

Remark 5.18.2 (i) p : S 7 → S 4 is a Hopf fibering with fibers 3-spheres and

(ii) p : S 15 → S 8 is a Hopf fibering with fibers 7-spheres.

Example 5.18.3 plays a key role in computing the homotopy groups of sphere
(results are only partly known), and hence it reflects the importance of bundle theory.

Example 5.18.3 (Real, complex and quaternionic Hopf bundles)

(i) (Real Hopf bundle) ξ = (S n , p, R P n , Z2 ) is a locally trivial fiber bundle with
fiber Z2 .
(ii) (Complex Hopf bundle) η = (S 2n+1 , p, CP n , S 1 ) is a trivial fiber bundle with
fiber S 1 .
(iii) (Quaternionic Hopf bundle) γ = (S 4n+3 , p, HP n , S 3 ) is a locally trivial fiber
bundle with fiber S 3 .

Definition 5.18.4 A continuous map f : X → Y is said be inessential if there exists

a homotopy Ht : X → Y such that H0 = f and H1 (X ) = y0 ∈ Y, otherwise, the map
is said to be essential.

Proposition 5.18.5 Let B be a topological space with more than one point. If S n is
a fiber space over the base space B, then the projection p : S n → B is an essential
map in the sense of Definition 5.18.4.

Proof It is proved by the method of contradiction. Suppose that p : S n → B is an

inessential map. Then there exists a homotopy Ht : S n → B for all t ∈ I such that
H0 = p and H1 (S n ) = b0 ∈ B. If 1 S n : S n → S n is the identity map, then p ◦ 1 S n =
p. Since p : S n → B has the bundle property, it has the covering homotopy property
by Theorem 5.20.2. Hence there exists a homotopy Ft : S n → S n such that F0 = 1 S n
and p ◦ Ft = Ht , ∀ t ∈ I. By hypothesis, since B has more than one point, F1 sends
S n into the fiber p −1 (b0 ), which is a proper subset of S n . This implies that the map
F : S n × I → S n such that F(x, t) = Ft (x), ∀ x ∈ X, ∀ t ∈ I is inessential. This
asserts that the identity map 1nS is homotopic to a constant map. But it is not possible
by degree theorem (see Chap. 3). This proves that p : S n → B is an essential map.

392 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

5.18.1 Problems of Computing πm (Sn )

This subsection communicates the problems of computing the homotopy groups

πm (S n ), the simplest noncontractable spaces. The homotopy groups πm (S n ) are not
completely known. So, computing the homotopy groups completely is one of the
major unsolved problems in homotopy theory. The homotopy group πm (S n ) for
m ≤ n are known. It has been proved that
(i) πm (S n ) = 0 = {0}, for m < n by Theorem 5.18.10.
(ii) πm (S 1 ) = 0 = {0}, for m > 1, by Theorem 5.4.27.
(iii) π1 (S 1 ) ∼
= Z (see Chap. 2 ).
(iv) πn (S n ) ∼
= Z (see Chap. 2).

Remark 5.18.6 There is a natural question: is πm (S n ) = {0} for every integer m > n.
H. Hopf first solved this problem in 1931 by showing that π3 (S 2 ) is not trivial by
Hopf Theorem 5.18.7. The sample Table 5.21 shows that there are other examples
for πm (S n ) (for m > n) are known for particular pair of integers m and n but not
known in all possible cases.

5.18.2 More Theorems on Hopf Maps

This subsection continues the study of Hopf maps started in Chap. 2. In general,
continuous maps p : S 2n−1 → S n for n = 2, 4, 8 introduced by H. Hopf in 1935
[Hopf, 1935] are now called Hopf maps, which are early examples of bundle spaces
with the property that p is not homotopic to a constant map. The aim of his study was
to investigate certain homotopy groups of spheres. Historically, it was not known until
1930 whether a given continuous map p : S m → S n for m > n > 1 is not homotopic
to a constant map. Hopf presented in 1930 the first example of a continuous map
p : S 3 → S 2 which is not homotopic to a constant map by proving that π3 (S 2 ) = 0
( see Theorem 5.18.7).

5.18.3 Hopf Theorem

Theorem 5.18.7, known as Hopf theorem and proved by Hopf in 1930 exhibits first
example of a continuous map p : S 3 → S 2 which is not homotopic to a constant map
by proving that π3 (S 2 ) = {0} (see Theorem 5.18.7).

Theorem 5.18.7 π3 (S 2 ) = {0}.

Proof Consider the 3-sphere S 3 = {(z, ω) ∈ C × C : |z|2 + |ω|2 = 1} and the Hopf
map p : S 3 → S 2 . Define an equivalence relation ∼ on S 3 :
5.18 More Study on Hopf Map p : S 3 → S 2 393

Fig. 5.35 Commutative

diagram involving Hopf map
p : S3 → S2

(z, ω) ∼ (z , ω ) ⇔ (z, ω) = (αz , αω ) : α ∈ C and |α| = 1.

(i) If X = S 3 / ∼ is the quotient space topologized by the quotient topology and

(ii) p : S 3 → X, (z, ω) → [(z, ω)] is the projection map,
then p −1 [(z, ω)] called the fiber over [(z, ω)], is a great circle of S 3 . Since X is
homeomorphic to S 2 , by replacing X by S 2 , we have the Hopf map p : S 3 → S 2 . This
implies that S 3 is decomposed into a family of great circles with S 2 as a quotient space.
To prove the theorem, it is sufficient to show that p is not homotopic to a constant
map. We prove it by method of contradiction. Suppose that there exists a homotopy
F : S 3 × I → S 2 between p and a constant map c. Then it defines a homotopy
 : S 3 × I → S 3 such that the triangle in diagram Fig. 5.35 is commutative.
F
The map F  is continuous and gives a homotopy between the identity map 1 S 3 on
S and a constant map c. This implies that S 3 is contractible, but it is not true. This
3

contradiction proves that π3 (S 2 ) = {0}. 

Theorem 5.18.8 proves that the Hopf map p : S 7 → S 4 is not homotopic to any
constant map.

Theorem 5.18.8 π7 (S 4 ) = {0}.

Proof Consider the 7-sphere S 7 as the topological space

S 7 = {(z, ω) ∈ H × H : ||z||2 = 1},

where H denotes the division ring of quaternions. Let Y denote the unit disk in H
given by the topological space

Y = {z ∈ H : ||z|| ≤ 1}.

Let X be the quotient space obtained by identifying the boundary ∂Y of Y to a single

point. Then X is homeomorphic to S 4 , since the real dimension of Y is 4. Proceed
as in Theorem 5.18.8 to prove that π7 (S 4 ) = {0}. 

Theorem 5.18.9 π15 (S 8 ) = {0}.

Proof Consider the Hopf map

p : S 15 → S 8 .

To prove the theorem proceed performing likewise construction in R16 as in The-

orem 5.18.7 and in Theorem 5.18.8. 
394 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

5.18.4 Hurewicz Theorem

Theorem 5.18.10 known as Hurewicz and proved by Hurewicz in 1935 gives a suffi-
cient condition imposed on m and n such that πm (S n ) = {0}. This theorem is proved
by simplicial approximation theorem saying that if K and L are finite simplicial com-
plexes and f : |K | → |L| be a continuous map between their polyhedra, and then for
a chosen sufficiently large r, there exists a simplicial approximation g : |K r | → |L|
to f : |K r | → |L| and f g [Adhikari, 2016].

Theorem 5.18.10 (Hurewicz) πm (S n ) = {0}, ∀ m, n satsfying 0 < m < n.

Proof Let [ f ] ∈ πm (S n ) be an arbitrary element. Suppose that [ f ] is represented by

a continuous map
f : (S m , 1) → (S n , 1).

Since S m can be represented as the boundary complex of simplexes of dimension

m + 1 and similarly, S n as the boundary complex of simplexes of dimension n + 1,
the map
f : (S m , 1) → (S n , 1)

has a simplicial approximation g that cannot map a simplex onto a simplex of higher
dimension. This implies that the map g is continuous that cannot be onto. Then there
exists a point s0 ∈ S n such that s0 ∈ / I m g and g is a map such that I m g = g(S m ) is
contained in a contractible space. Hence it follows that g is homotopic to a constant
map c : S m → S n . This implies that [ f ] = [g] = [c] = 0. Since [ f ] is an arbitrary
element of πm (S n ) , it is proved that πm (S n ) = {0}, ∀ m, n satisfying 0 < m < n.


Theorem 5.18.11 For any r > 1, πr (S 1 , s0 ) = 0.

Proof Consider the exponential map

p : R → S 1 , t → e2πit .

Then p : R → S 1 is a covering map. Since S r is simply connected, every contin-

uous map f : S r → S 1 has a lifting to a continuous map f˜ : S r → R by the lifting
property such that f˜ is homotopic to a constant map c̃. Hence there exists a homotopy

H̃t : f˜ c̃.

Project this homotopy H̃t on S 1 to obtain a homotopy

Ht : f c.

This proves that πr (S 1 , s0 ) = 0, ∀ r > 1. 

5.18 More Study on Hopf Map p : S 3 → S 2 395

Fig. 5.36 Rectangle

involving fiber map and its
induced map

5.18.5 Fiber Maps and Induced Fiber Spaces

Given two fiberings p : X → B and q : Y → A, a continuous map f : X → Y is

said to be a fiber map if f sends fibers into fibers. This concept is formulated in
Definition 5.18.12.
Definition 5.18.12 Let p : X → B and q : Y → A be two fiberings. A continuous
map f : X → Y is said to be a fiber map if for every point b ∈ B, there exists a
point a ∈ A, such that
f ( p −1 (b)) ⊂ q −1 (a).

Every fiber map induces a map on base spaces.

Definition 5.18.13 Let p : X → B and q : Y → A be two fiberings and f : X → Y
be a fiber map. Then f induces a map

f ∗ : B → A, b → (q ◦ f )( p −1 (b)).

Then for an arbitrary subset V ⊂ A

f ∗ (V ) = ( p ◦ f −1 )(q −1 (V ))

asserts that f ∗ is continuous.

Remark 5.18.14 For bundles p : X → B and q : Y → A, let f : X → Y be a fiber
map. Then p is open and hence the induced map f ∗ = g : B → A is a continuous
map such that the diagram in Fig. 5.36 is commutative in the sense that q ◦ f = g ◦ p.

Example 5.18.15 (i) (Trivial fibering) The fibering q : Y → A is said to be trivial

over the base space B, with the projection p : X → B as a fiber map, where
A = B = Y, q = 1 B (identity map on B) and the induced map g = 1 B .
(ii) Let p : X → B be the trivial fibering over the base space B. Then every map
f : X → Y is a fiber map, where B = X and p = 1 X .

5.19 Homotopy Sequence of Bundles

This section proves Theorem 5.19.1, which is basic result in bundle theory and defines
homotopy sequence of a bundle in Sect. 5.19.2. For its exactness property and direct
sum theorems related to bundles, see Exercises 5.24.1 and 5.24.1 of Sect. 5.24. Direct
396 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

sum theorems related to bundles are analogous to Theorem 5.17.3 plays a key role
providing a very strong necessary condition for the existence of a cross section for
homotopy groups of dimensions ≥ 2. An exceptional behavior occurs for n = 1.
For example, consider the Klein bottle. It is bundle with base space is the circle B,
obtained from a line segment L by identifying its end points. Its fiber F is also a
circle, and it admits a cross section. If the fundamental group π1 (X, x0 ) of the Klein
bottle were a direct sum of πn (B, b0 ) ⊕ πn (F, x0 ), then the fundamental group would
be an abelian group, because both the groups π1 (B, b0 ) and π1 (F, x0 ) are infinite
cyclic. But this is not true in this case, because the fundamental group of the Klein
bottle is a group on two generators α and β having the only one relation αβ = β −1 α.
Möbius band is a bundle obtained from the product space L × F by identifying
the two ends of the cylinder L × F

5.19.1 Fundamental Property

Theorem 5.19.1 proves a fundamental property of p∗ in homotopy theory which is

not enjoyed in homology theory.

Theorem 5.19.1 (Fundamental property) Let ξ = (X, p, Y ) be a bundle over Y

and B ⊂ Y be a subspace. If A = p −1 (B), x0 ∈ A and y0 = p(x0 ), then the map
p : (X, A, x0 ) → (Y, B, y0 ) induces an isomorphism

p∗ : πn (X, A, x0 ) → πn (Y, B, y0 ), ∀ n ≥ 2.

Proof The map p : (X, A, x0 ) → (Y, B, y0 ) induces a homomorphism by Theorem

5.11.12
p∗ : πn (X, A, x0 ) → πn (Y, B, y0 ), ∀ n ≥ 2.

p∗ is a monomorphism: It is proved by showing that kernal of p∗ is zero. Let

α ∈ πn (X, A, x0 ) be an element such that p∗ (α) = 0. If α is represented by f ∈
F n (X, A, x0 ), then there exists a homotopy

Ht ∈ F n (Y, B, y0 ) : H0 = p ◦ f and H1 = c,

where c is a constant map. Then there a covering homotopy Ht of f such that

Ht (Jn−1 ) = x0 , because Ht (Jn−1 ) = y0 . Again, Ht (In−1 ) ⊂ A, because, Ht (In−1 ) ⊂
B. This implies that Ht is a homotopy in F n (X, A, x0 ) of f into a map

f : (In , In−1 , Jn−1 ) → (X 0 , X 0 , x0 ),

5.19 Homotopy Sequence of Bundles 397

where X 0 is the fiber over y0 . Define a continuous map

Ft : In → In , (t1 , t2 , . . . , tn ) → (t1 , t2 , . . . , tn−1 , (1 − t)tn + t), ∀ t ∈ I.

Then Ft is a homotopy of In over itself into its face tn = 1. This face stays in the
subspace Jn−1 , which is also deformed into itself. Define

Ft : In → In , (t1 , t2 , . . . , tn ) → f (Ft (t1 , t2 , . . . , tn−1 , tn )) ∀ t ∈ I.

This implies that Ft is a homotopy in F n (X, A, x0 ) of f into a constant map. This

proves that α = 0 and hence kernal of p∗ is zero.
p∗ is an epimorphism: Take an arbitrary element β ∈ (Y, B, y0 ) and let f ∈
F n (Y, B, y0 ) represent the element α. Define

G t : In → In , (t1 , t2 , . . . , tn ) → f (Ft (t1 , t2 , . . . , tn−1 , tn )) ∀ t ∈ I.

This implies that G t is a homotopy of f into a constant map. Let g be the map,
which carries In into x0 . Then there exists an element α ∈ πn (X, A, x0 ) such that
p∗ (α) = β.
This completes the proof of the theorem. 

Corollary 5.19.2 Let ξ = (X, p, Y ) be a bundle over Y and y0 ∈ Y be a base point.

If F = p −1 (y0 ), x0 ∈ F and y0 = p(x0 ), then the map p : (X, F, x0 ) → (Y, y0 )
induces an isomorphism

p∗ : πn (X, F, x0 ) → πn (Y, y0 ), ∀ n ≥ 2.

5.19.2 Homotopy Sequence of a Bundle

This subsection defines homotopy sequence of a bundle. For its exactness property,
see Exercise 5.24.1 of Sect. 5.24.
Let ξ = (X, p, Y ) be a bundle over Y and y0 ∈ Y be a base point. Let F =
p −1 (y0 ), x0 ∈ F and y0 = p(x0 ). Suppose

i : (F, x0 ) → (X, x0 ), x → x

and

j : (X, x0 , x0 ) → (X, F, x0 ), x → x

are the inclusion maps. Then the homotopy sequence of the triplet (X, F, x0 ) is
i∗ j∗ ∂
· · · −−→ πn (F, x0 ) −−−→ πn (X, x0 ) −−−−→ πn (X, F, x0 ) −−−→ πn−1 (F, x0 ) −
→ ···
398 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Let q denote p considered as a map q : (X, F, x0 ) → (Y, y0 , y0 ). Then q ◦ j :

(X, x0 ) → (Y, y0 ) is the same as the map p : (X, x0 ) → (Y, y0 ). Use Corollary
5.19.2 which asserts that the map p : (X, F, x0 ) → (Y, y0 ) induces an isomorphism

q∗ : πn (X, F, x0 ) → πn (Y, y0 ), ∀ n ≥ 2

to define
 = ∂ ◦ q∗−1 : πn (Y, y0 ) → πn (F, x0 ), ∀ n ≥ 2.

Then the sequence of groups and homomorphisms

i∗ p∗ 
· · · −−→ πn (F, x0 ) −−−→ πn (X, x0 ) −−−−→ πn (Y, y0 ) −−−→ πn−1 (F, x0 ) −
→ ···

 i∗ p∗
· · · −−→ π2 (Y, y0 ) −−−→ π1 (F, x0 ) −−−−→ π1 (X, x0 ) −−−−→ π1 (Y, y0 ) −
→ ···

is called the homotopy sequence of the bundle ξ based at x0 .

5.20 Bundle Space and Bundle Property

Definition 5.20.1 A continuous p : X → B is said have the bundle property if

there exists a topological space D such that for every b ∈ B, there exists an open
nbd U of the point b and a homeomorphism ψU such that

ψU : U × D → p −1 (U ) : pψU (y, z) = y, ∀ y ∈ U, z ∈ D.

Then X is called the bundle space, B is called the base space, and D is called
the director space relative to the projection p : X → B. The nbd U is called a
decomposing nbd (space), and the homeomorphism ψU is called the decomposing
function.
Theorem 5.20.2 (Covering homotopy theorem) If a map p : X → B has the bun-
dle property, then it has the covering homotopy property for every paracompact
Hausdorff space.
Proof See [p. 50, Steenrod, 1951]. 
Corollary 5.20.3 Let X be a bundle space over the base space B relative to projec-
tion p : X → B. Then X is fiber space over the space B relative to projection p.
Proof It follows from Theorem 5.20.2 
Theorem 5.20.4 The Hopf map

p : S 3 → S 2 , (z, w) → [z, w]

is a 1-sphere bundle.
5.21 Table of πi (S n ) for 1 ≤ i, n ≤ 8 399

Proof Represent 1-sphere as S 1 = {z ∈ C : |z| = 1}. For the points z 1 = [1, 0] and
z 2 = [0, 1] of S 2 , consider the open sets

U = S 2 − {z 1 } and V = S 2 − {z 2 }.

Hence it follows that U and V forms an open covering of S 2 and each point in U
is represented by a pair [z, 1]. Consider the map
wz w
ψU : U × S 1 → S 2 , ([z, 1], w) → ( 1 , 1 ).
(zz + 1) 2 (zz + 1) 2

Then ψU maps U × S 1 homeomorphically onto p −1 (U ) such that pψU (u, d) =

u, ∀ u ∈ U and d ∈ D some topological space D (called director space). 

Corollary 5.20.5 The 3-sphere is a bundle space over the 2-sphere S 2 relative to
the Hopf map p : S 3 → S 2 .

Proof It follows from the proof of Theorem 5.20.4. 

The above discussion is summarized in a result formulated in Theorem 5.20.6.

Theorem 5.20.6 The 3-sphere S 3 is decomposed into a family of great circles, called
fibers of the Hopf map f : S 3 → S 2 with the 2-sphere S 2 as a decomposition space.

5.21 Table of π i (Sn ) for 1 ≤ i, n ≤ 8

This section exhibits Table 5.1 displaying a small sample of the values of the groups
πi (S n ) extracted from the paper [Toda, 1962]. This table is a consequence of Freuden-
thal suspension theorem formulated by Freudenthal in 1937 (see Sect. 5.15). An
analogue table is also given in Chap. 2.

Table 5.1 Sample table of πi (S n ) for 1 ≤ i, n ≤ 8

i=1 2 3 4 5 6 7 8
n=1 Z 0 0 0 0 0 0 0
2 0 Z Z Z2 Z2 Z12 Z2 Z2
3 0 0 Z Z2 Z2 Z12 Z2 Z2
4 0 0 0 Z Z2 Z2 Z × Z12 Z2 × Z2
5 0 0 0 0 Z Z2 Z2 Z24
6 0 0 0 0 0 Z Z2 Z2
7 0 0 0 0 0 0 Z Z2
8 0 0 0 0 0 0 0 Z
400 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

5.22 Action of π1 on πn and n-simplicity

This section studies action of π1 on πn and conveys an important action. The fun-
damental group π1 (X, x0 ) acts on πn (X, x0 ) as a group of automorphisms for every
n ≥ 1.

5.22.1 Automorphism Induced by a Closed Curve

For the study of this subsection, we use Theorem 5.12.2, which asserts that for any
path-connected space X, the group πn (X, x0 ) is independent of the choice of the base
point x0 ∈ X in the sense that any path α in X from x0 to x1 induces an isomorphism

α∗ : πn (X, x1 ) → πn (X, x0 ).

This isomorphism α∗ satisfies the following properties for any two points x0 ,
x1 ∈ X
(i) if α : I → X is a path from x0 to x1 and β : I → X is a path from x1 to x2 , then
their product path α ∗ β : I → X induces an isomorphism (α ∗ β)∗ such that

(α ∗ β)∗ = α∗ ◦ β∗ .

(ii) if α γ : I → X are two paths from x0 to x1 , (keeping their end points fixed),
then they induce the same isomorphism, i.e., α∗ = γ∗ .
The above discussion is summarized in a basic result formulated in Theorem
5.22.1.

Theorem 5.22.1 Let X be a pointed topological space with base point x0 ∈ X. Then
for every n ≥ 1, the fundamental group π1 (X, x0 ) acts on πn (X, x0 ) as a group of
automorphisms.

Remark 5.22.2 For n = 1, the action ψ of π1 on itself is by inverse automorphism.

For n > 1, the action endows the abelian group πn (X, x0 ) a module structure over
the group ring Z[π1 (X, x0 )]. It is said that πn is a π1 -module instead of Z[π1 (X, x0 )]-
module.

5.22.2 Isomorphism of πn Induced by a Curve

This subsection shows that if the space A of the triple (X, A, x0 ) is path-connected,
then the groups πn (X, A, x0 ) are independent of the choice of the base point x0 ∈
A, ∀ n ≥ 2. Theorem 5.22.3 gives a generalization of Theorem 5.12.2.
5.22 Action of π1 on πn and n-simplicity 401

Theorem 5.22.3 Let (X, A, x0 ) be a triple of spaces and A be path-connected. Then

πn (X, A, x0 ) is independent of the choice of base point x0 ∈ A.

Proof Let ψ : I → X be a path in A from the point x0 to the point x1 in A. To prove

the theorem we have to prove that ψ induces an isomorphism

ψ∗ = ψn : πn (X, A, x1 ) → πn (X, A, x0 ), ∀ n ≥ 2

such that
(i) if θ is a path from x0 to x1 in A and η is a path from x1 to x2 in A then

(θ ∗ η)∗ = (θ ∗ η)n = θn ◦ ηn = (σ ∗ φ)n ,

where θ ∗ η denotes product path in A from x0 to x1 .

(ii) if ψ and φ are two homotopic paths in A from x0 to x1 relative to end points,
then
ψn = φn , ∀ n ≥ 2.

Geometric proof Given an element σ ∈ πn (X, A, x0 ), let f represent the element

σ, construct a homotopy Ft of f such that
(i) Ft moves Jn−1 along the inverse path ψ −1 into x0 keeping the image of J n−1 a
point at every step and
(ii) Ft deforms In−1 over A.
Hence, it follows that the composite map represents an element β ∈ πn (X, A, x0 ).
The assignment σ → β defines an isomorphism

ψn : πn (X, A, x1 ) → πn (X, A, x0 ), σ → β, ∀ n ≥ 2.

Analytical proof Proceed as in Theorem 5.12.2 for an analytical proof or see

[Steenrod, 1951]. 

Corollary 5.22.4 If X is any topological space and ψ : I → X is any path connect-

ing two given points x0 , x1 ∈ X. Then ψ induces an isomorphism

ψn : πn (X, x1 ) → πn (X, x0 ), ∀ n ≥ 1,

which depends only on the homotopy class of the path ψ relative to its end points.

5.22.3 n-simplicity

Definition 5.22.5 A group G is said to act simply on a group H, if g · h = h, ∀ g ∈

G and ∀ h ∈ H. This action is also called a trivial action of G on H.
402 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

Definition 5.22.6 A topological space X is said to be n-simple, if for every base

point x0 ∈ X, its fundamental group π1 (X, x0 ) acts simply on the group πn (X, x0 ).

Proposition 5.22.7 Let X be a path-connected space. Then it is n-simple, iff there

exists a point x0 ∈ X such that the fundamental group π1 (X, x0 ) acts simply on
πn (X, x0 ).

Proof It follows by using Definition 5.22.6. 

Example 5.22.8 Every simply connected space is n-simple for any n ≥ 1.

Example 5.22.9 Every path-connected topological group is n-simple for any n ≥ 1

by Proposition 5.22.11.

Theorem 5.22.10 characterizes n-simplicity of a topological space by specified

pairs of homotopic maps.

Theorem 5.22.10 Let X be a topological space and S n be the n-sphere with a

base point s0 ∈ S n . Then X is n-simple iff for any point x0 ∈ X and for any pair of
continuous maps f and g such that

f, g : S n → X : f (s0 ) = g(s0 ) = x0 ,

the homotopy relation f g implies f g r el s0 .

Proof First suppose that the space X is n-simple and f g. We claim that f
g r el s0 . Then the fundamental group π1 (X, x0 ) acts simply on the group πn (X, x0 )
by Proposition 5.22.7. Since by hypothesis, f g, there exists a homotopy

Ht : S n → X : H0 = f and H1 = g.

Let [ f ] = α ∈ πn (X, x0 ) and [g] = β ∈ πn (X, x0 ). Then α, β ∈ πn (X, x0 ) are the

elements represented by f, g, respectively. Define a path in X

γ : I → X, t → Ht (s0 ).

Then
γ(0) = H0 (s0 ) = f (s0 ) = x0 = γ(1)

implies γ represents an element u ∈ π1 (X, x0 ) such that α = u · β. Since by hypoth-

esis, π1 (X, x0 ) acts simply on the group πn (X, x0 ), it follows that u · β = β and
hence it follows that α = β. This asserts that f g r el s0 . To prove its converse,
suppose that f g implies f g r el s0 . Let α ∈ πn (X, x0 ) be an arbitrary ele-
ment represented by a continuous map f : S n → X such that f (s0 ) = x0 . Hence the
element uα ∈ πn (X, x0 ) is represented by a continuous map g : S n → X such that
g(s0 ) = x0 . Then f g and hence by hypothesis, f g r el s0 . This asserts that
u · α = α and hence it is proved that X is n-simple. 
5.22 Action of π1 on πn and n-simplicity 403

Proposition 5.22.11 Let G be a path-connected topological group. Then it is n-

simple for every n ≥ 1.

Proof Let S n be the n-sphere with a base point s0 ∈ S n . By hypothesis, G is a path-

connected topological group with its identity element g0 = e. Let f, g : S n → G be
two maps such that f g and f (s0 ) = g0 = g(s0 ). Then there exists a homotopy

Ht : S n → G : H0 = f and H1 = g.

Define a continuous map

Ft : S n → G, s → (Ht (s0 ))−1 Ht (s).

Then F0 = f, F1 = g and Ft (s0 ) = g0 , ∀ t ∈ I implies that Ft : f g r el s0 .

Hence it follows by Theorem 5.22.10 that G is n-simple for every n ≥ 1. 

Theorem 5.22.12 Let G be a Lie group and H be a closed connected subgroup of

G. Then G/H is n-simple for every integer n ≥ 1.

Proof Consider the natural projection

p : G → G/H, g → g H.

If p(H ) = x0 ∈ G/H, then to prove the theorem, it is sufficient to show that the
group π1 (G/H, x0 ) acts trivially on πn (G/H, x0 ), since G acts transitively on G/H.
Define

F n (G/H, x0 ) = { f : (In , ∂ I n ) → (G/H, x0 ) such that f is continuous}.

Let α : I → X/A be a closed curve based at the point x0 . If α(t) is regarded as a

homotopy of x0 , then a covering homotopy determines a curve α (t) such that

α (0) = e and pα(t) = α (t).

This implies that α (t) ∈ H. Adjoin a curve in H from the point α (1) to the point e
to obtain a closed curve β (t) such that p ◦ β = β α. Then

H : In × I → G/H, (t, s) → β (1 − s) f (t)

is a homotopy of f around the curve β −1 back into f. Since f ∈ F n (G/H, x0 ) is an

arbitrary map, the theorem follows. 

Remark 5.22.13 The action of π1 on πn is used in algebraic topology such as to

prove the homological version of Whitehead theorem which asserts that if X and
Y are both simply connected C W -complexes and if a continuous map f : X → Y
induces isomorphisms
404 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

f ∗ : Hn (X ) → Hn (Y )

for all homology groups, then f is a homotopy equivalence [Adhikari, 2016, p. 526].

5.23 Further Applications

Theorem 5.23.1 studies infinite earring initiated in Chap. 4 from the homotopy view-
point and commutes its fundamental group.

Theorem 5.23.1 Let X be the union of a countably infinite family of circles with
subspace topology in the plane R2 , called the ‘infinite earring or shrinking wedge
of circles’ in the plane R2 . Then
(i) π1 (X ) = {0};
(ii) X has no universal covering.

Proof Let Cn be the circle of radius 1/n in R2 with center at (1/n, 0), for each n ≥ 1
and X be the subspace of R2 which is the union of these circles as shown in Fig. 5.37.
Hence X is the union of a countably infinite family of circles, which is called the
infinite earring or ‘shrinking wedge of circles’ in the plane R2 .
(i) Let U be a nbd of the origin x0 in X . Consider the inclusion map

i : U → X.

Then the homomorphism

i ∗ : π(U, x0 ) → π1 (X, x0 )

between the corresponding fundamental groups induced by the inclusion

i : U → X

is not trivial. Because, for any integer n, there is a retraction r : X̃ → Cn which

sends each circle Ci for i = n to the point x0 . Take n sufficiently large such
that Cn ⊂ U. Then the inclusion maps j : U → X and inclusion k : U → U
induce monomorphism j∗ and k∗ in the triangle of groups and homomorphisms
in Fig. 5.38 commutative. This implies that the homomorphism j∗ can not be
trivial. This proves the part (i) of the theorem.
(ii) The part (ii) follows from part (i). 

Theorem 5.23.2 is a basic theorem in homotopy theory. It is now proved by using

homotopy sequence of a fibration.
 5.23 Further Applications 405

Fig. 5.37 Infinite earring

Fig. 5.38 Commutative

diagram induced by
inclusion maps

Theorem 5.23.2 Let X be a path-connected space and x0 ∈ X be an arbitrary point.

If  X x0 be the space of all loops in X based at x0 endowed with compact open topology.
Then
(i) πn ( X x0 ) ∼
= πn+1 (X );
(ii) The group π1 ( X x0 ) is abelian;
(iii) The groups πn (X ) are all abelian for n ≥ 2.

Proof E be the space of all paths α in X starting at the point x0 endowed with
compact open topology and ending at different pints x ∈ X. Then α : I → X is a
continuous map such that α(0) = x0 and α(1) = x. If

p : E → X, α → α(1) = x,

then ξ = (E, p, X ) is a fibration represented as

 X x0 −
→E−
→X

with fiber  X x0 , total space E and base space X. Since each α is contractible on itself
to the point x0 , it follows that E is also contractible and hence πn (E) = 0. Consider
the homotopy sequence of the fibration ξ = (E, p, X )

 X x0 −
→E−
→X
406 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

· · · −−→ πn+1 (E) −−→ πn+1 (X ) −−→ πn ( X x0 ) −−→ πn (E) · · ·

Since the above sequence is exact and πn+1 (E) = 0, it follows that πn ( X x0 ) ∼
=
πn+1 (X ). Hence, the theorem follows. 

Proposition 5.23.3 If (X, A) be a pair of topological spaces such that both X and
A are path connected, then
(i) the set π1 (X, A, x0 ) and
(ii) the set of cosets a H of the subgroup H of π1 (X, x0 ) represented by loops in A
at x0
are equivalent as sets.

Proof By the given condition, π1 (X, A, x0 ) represents the set of homotopy classes
of paths in X starting from an arbitrary point in A. Construct a map

φ : π1 (X, x0 ) → π1 (X, A, x0 )

by taking a loop at x0 as an element of π1 (X, A, x0 ). We claim that φ is a bijection.

φ is surjective By hypothesis, A is path connected, and hence every element
of π1 (X, A, x0 ) is homotopic to a loop based at the point x0 . This implies that φ is
surjective.
φ is injective Since any two loops α, β ∈ π1 (X, x0 ) are homotopic rel A iff
[α−1 ∗ β] is represented by a loop in A. This implies that φ is injective. 

Proposition 5.23.4 For a given triplet (X, A, x0 ), if A is a strong deformation

retract of X and the inclusion, i : (A, x0 ) → (X, x0 ) is the inclusion, them its induced
homomorphism
i ∗ : πn (A, x0 ) → πn (X, x0 )

is an isomorphism for every n > 0.

Proof By hypothesis, A ⊂ X is a strong deformation retract of X . Then there is a

retraction r : X → A such that i ◦ r 1 X rel A. Then the map i a homotopy equiv-
alence and consequently, its induced homomorphism i ∗ : πn (A, x0 ) → πn (X, x0 ) is
an isomorphism for all n > 1. 

Corollary 5.23.5 For a given triplet, (X, A, x0 ) if A is a strong deformation retract

of X , then πn (X, A, x0 ) = 0 for every integer n > 0.

Proof Consider the inclusion map

i : (A, x0 ) → (X, x0 ).

Then its induced homomorphism

i ∗ : πn (A, x0 ) → πn (X, x0 )
5.23 Further Applications 407

in the corresponding homotopy groups is an isomorphism for every n > 0 by Propo-

sition 5.23.4. Hence the corollary follows from the exactness property of the sequence
(5.2) of the triplet (X, A, x0 ). 

Definition 5.23.6 A continuous map p : X → B is said to have the polyhedra cov-

ering homotopy property (PCHP) if it has the covering homotopy property for
every triangulable space Y . If p has PCHP, then p is said to be a fibering.

Proposition 5.23.7 Given a continuous map f : (X, A, x0 ) → (Y, B, y0 ) of triplets,

if f : X → Y is a fibering and A = f −1 (B), then the induced homomorphism

f ∗ : πn (X, A, x0 ) → πn (Y, B, y0 )

is an isomorphism for each n > 1.

Proof f ∗ is an epimorphism: Let β ∈ πn (Y, B, y0 ) be an arbitrary element rep-

resented by a map g : (In , In−1 , Jn−1 ) → (Y, B, y0 ). Since the subspace Jn−1 is a
strong deformation retract of the space In , there is a continuous map

k : In → X : f ◦ h = g and k(Jn−1 ) = x0 .

This defines a continuous map of triplets

k : (In , In−1 , Jn−1 ) → (X, A, x0 ),

because A = f −1 (B), f ◦ k = g, k(In−1 ) ⊂ A. Then f ∗ ([k]) = β asserts that the

homomorphism f ∗ is an epimorphism. f ∗ is a monomorphism : Let β, γ ∈
πn (X, A, x0 ) be two elements such that f ∗ (α) = f ∗ (β) represented by g, k ∈ πn
(X, A, x0 ). Then maps f ◦ g and f ◦ k represent the same element of πn (Y, B, y0 ).
Hence , there exists a map

F : (In × I, In−1 × I, Jn−1 × I)→(Y, B, y0 ) : F(ω, 0)=( f ◦ g)(ω) and H (ω, 1) = ( f ◦ k)(ω), ∀ ω ∈ In .

If C = (In × {0}) ∪ (Jn−1 × I) ∪ (In × {1}), then C is a closed subspace of In × I.

Construct a continuous map
⎧
⎪
⎨g(ω), if ω ∈ In , t = 0
H : C → X, (ω, t) → x0 , if ω ∈ Jn−1 , t ∈ I
⎪
⎩
k(ω), if ω ∈ In , t = 1.

Then f ◦ H = F|C . Clearly, H has an extension

 : In × I → X : f ◦ H
H  = F,
408 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

(In−1 × I) ⊂
because, C is a strong deformation retract of In × I. This implies that H
A. Hence the map

 : (In × I, Jn−1 × I, Jn−1 × I) → (X, A, x0 )

H

satisfies the properties

(ω, 0) = g(ω), ∀ ω ∈ In and
(i) H
(ω, 1) = k(ω), ∀ ω ∈ In .
(ii) H
This shows that the maps g and h represent the same element of πn (X, A, x0 ).
This proves that if f ∗ (α) = f ∗ (β), then α = β. 
Proposition 5.23.8 Let X = {x0 } be a one-point space consisting of a single point
x0 . Then πn (X, x0 ) = 0 for every n ≥ 0.
Proof By hypothesis, X = {x0 }. then for each n. Since the map f : In → X is the
only map of In onto X , which is a constant map for every n ≥ 0, it follows that
πn (X, x0 ) = 0 for every n ≥ 0. 

5.24 Exercises and Multiple Choice Exercises

As solving exercises plays an essential role of learning mathematics, various types

of exercises and multiple choice exercises are given in this section. They form an
integral part of the book series.

5.24.1 Exercises

1. Let p : (X, x0 ) → (B, b0 ) be a fibration having fiber F = p −1 (b0 ) ⊂ X and the

mapping fiber F p = {(x, α) ∈ X × B I : α(0) = b0 and α(1) = p(x)} with
product topology. Show that the fiber F and the mapping fiber F p are homotopy
equivalent spaces.
2. Given any fibration p : X → B and a continuous map f : A → B, let H1 =
{[h : A → X : p ◦ h = f ]} and H2 = {[h̃ : A → X : p ◦ h̃ f ]} be two sets
of homotopy classes. Show that there exists a bijection

ψ : H1 → H2 .

3. Given a normal space X and its subspace A, show that the inclusion

i : A → X

is a cofibration iff the inclusion

5.24 Exercises and Multiple Choice Exercises 409

j : A → V

is a cofibration for some open neighborhood V of A in X .

4. Given a a continuous map : A → X with mapping cylinder M f , and the inclusion

i : A → M f , x → [x, 0],

show that the inclusion i is a cofibration.

5. Let B be a path-connected space and p : X → B be a fibration having fiber
F = p −1 (b0 ). If
i : F → X

is the inclusion, show that for any topological space Z , the sequence of homotopy
sets
i∗ p∗
[Z , F] −−−→ [Z , X ] −−−−→ [Z , B]

is exact.
6. Let A be be a subspace of the topological space X and q : X → X/A the natural
projection map. If i : A → X is a cofibration, with cofiber X/A, show that for
any topological space Z , the 3 terms sequence of homotopy sets

q∗ i∗
[X/A, Z ] −−−−→ [X, Z ] −−−→ [A, Z ]

is exact.
7. Let ξ and η be two principal G-bundles over the same space B and ψ = ( f, f˜) :
ξ → η be a morphism of principal G-bundles. If f˜ : B → B is the identity map
1 B , show that ψ is an equivalence of principal G-bundles.
8. Let ξ = (X, p, B, Fn ) be an n-dimensional vector bundle. A Gauss map of
ξ in Fm (n ≤ m ≤ ∞) is a continuous map f : X → Fm such that f | p −1 (b) :
p −1 (b) → Fn is a linear monomorphism. Prove the following statements:
(i) corresponding to every n-dimensional vector bundle ξ over a paracompact
space B, there exists Gauss map f : X → Fm for ξ,
(ii) corresponding to an open covering {Ui : 1 ≤ i ≤ m} of B, such that ξ|Ui is
trivial, then ξ has a Gauss map f : X → Fmn .
9. Show that
(i) given a covering space ( X̃ , p) of X and x0 ∈ X, x̃0 ∈ p −1 (x0 ), the induced
homomorphism
p∗ : π1 ( X̃ , x̃0 ) → π1 (X, x0 )

is a monomorphism.
[ Hint: Suppose that [ f˜], [g̃] ∈ π1 (X, x0 ) and [ f˜] = [g̃]. This implies that
p∗ ([ f˜]) = [ p ◦ f˜] and p∗ ([g̃]) = [ p ◦ g̃]. Hence it follows that
410 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

p ◦ f˜ p ◦ g̃ rel I˙ ⇔ f˜ g̃ rel I˙.

On the other hand

f˜  g̃ rel I˙ ⇔ p ◦ f˜  p ◦ g̃ rel I˙

because, otherwise, a contradiction would be arrived by Theorem 5.4.7. This

proves that p∗ is well defined and it is a monomorphism. ]
(ii) homotopic closed path in the punctured Euclidean plane R2 − {0} has the
same winding number.
(iii) every covering of a rectangle ( may be open or closed) is trivial.
[ Hint: Use homotopy lifting theorem.]
10. Let X be a connected covering space of a path-connected space B with covering
projection p : (X, x0 ) → (B, b0 ) such that p(x0 ) = b0 . Show that
(i) the induced homomorphism p∗ : πn (X, x0 ) → πn (B, b0 ) is an isomorphism
for any n ≥ 2.
(ii) πn (S 1 ) = {0} for any n ≥ 2.

[ Hint: p∗ : πn (R) → π1 (S 1 ) is an isomorphism for any n ≥ 2 by (a). Since all the

groups of the contractible space R are

X,

(ii) follows.]
11. Prove the following statements:
(i) Given a covering space (X, p) of B and two points b0 , b1 ∈ B, if F0 is the
fiber over b0 and F1 is the fiber over b1 , the fibers F0 and F1 are the home-
omorphic.

[Hint: Since each fiber is discrete and any two fibers have the same cardinal
numbers by Theorem 5.10.2(iii), they are homeomorphic.]
(ii) Given a path-connected topological group G and a discrete normal subgroup
N of G, the natural homomorphism,

p : G → G/N , g → g N ,

forms a covering space of G/N .

(iii) The map
p : S 1 → S 1 , z → z 2

is a covering map having its generalization to the map

p : S 1 → S 2 , z → z n .
5.24 Exercises and Multiple Choice Exercises 411

(iv) The antipode preserving continuous

f : S1 → S1

is not nullhomotopic.
12. Given any manifold M, let N be submanifold of M such that N is a closed subset
of M. If ξ = (X, p, M) is a vector bundle over M, show that every smooth section
s of the restricted bundle ξ|N can be extended to a smooth section of ξ.
[ Hint: Consider the smooth section s as a map with values in a vector space and
an open nbd U of for every point p ∈ N and a section s̃ of ξ|U such that s̃ = s
on U ∩ N . Use Smoothing theorem (ssee Chap. 3, Volume II) and partition of
unity subordinate to an open covering of M consisting of such open nbds {Ui }
in N together with the open set M − N . ]
13. Let ξ = (X, p, M) and ξ = (X , p , M) be two vector bundles of same dimen-
sion over a manifold M and N be a closed submanifold of M. Show that every
isomorphism

ψ : ξ|N → ξ |N

has an extension to an isomorphism

ψ̃ : ξ|U → ξ |U

over an open nbd U of N .

14. Show that the 4-manifold S 2 × S 2 is simply connected, but it is not homeomor-
phic to S 4 .
[Hint: Use the results : π2 (S 2 × S 2 ) ∼
= Z ⊕ Z and π2 (S 4 ) = 0.]
15. Let (X, x0 ) and (B, b0 ) be pointed topological spaces. Consider the homotopy
exact sequence (5.9) of the fibering p : X → B. Prove the following statements:
(a) If F is a retract of B, then πn (X, x0 ) ∼
= πn (B, b0 ) ⊕ πn (F, x0 ) for every
n ≥ 2 and p∗ is an epimorphism for every n ≥ 1.
(b) If X is deformable into F, then πn (F, x0 ) ∼= πn (X, x0 ) ⊕ πn+1 (B, b0 ) for
every n ≥ 2 and p∗ = 0 for every n ≥ 1.
(c) If F is contractible in X , then πn (B, b0 ) ∼
= πn (X, x0 ) ⊕ πn−1 (F, x0 ) for
every n ≥ 2 and p∗ is a monomorphism for every n ≥ 1.
[ Hint : As n ≥ 2, use Propositions 5.16.1 & 5.16.5 and exactness property of
the homotopy sequence of a fibering p : X → B.]
16. (a) Show that for any triplet (X, A, x0 ) the formula a + b − a = (∂a)b holds
for all a, b ∈ π2 (X, A, x0 ), where ∂ : π2 (X, A, x0 ) → π1 (A, x0 ) is the usual
boundary operator, and (∂a)b denotes the action of ∂a on b.
(b) Deduce from (a) that the image of the map j∗ : π2 (X, x0 ) → π2 (X, A, x0 )
lies in the entire of π2 (X, A, x0 ).
412 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

17. Show that a continuous map f : (Dn , S n−1 , s0 ) → (X, A, x0 ) defines the zero
element in πn (X, A, x0 ) iff f g rel S n−1 for some g : (Dn , S n−1 , s0 ) → (X, A,
x0 ) such that g(D ) ⊂ A.
n

18. Let p : X → B be a weak fibration with p(x0 ) = b0 If b0 ∈ A ⊂ B, x0 ∈

p −1 (b0 ), Y = p −1 (A), show that the induced transformation p∗ : πn (X, Y, y0 ) →
πn (B, A, b0 ) is a bijection for every n ≥ 1.
[Hint. Use mathematical induction on n starting from n = 1.]
19. Let p : E → B be a locally trivial fiber bundle and b0 ∈ B. If F = p −1 (b0 )
and f 0 ∈ F, show that for every n > 1, p∗ : πn (E, F, f 0 ) → πn (B, b0 , b0 ) is an
isomorphism.
20. Let p : X → B be a covering of X with discrete fiber F. Suppose b0 ∈ B and
x0 ∈ p −1 (b0 ), show that
(i) p∗ : πn (X, x0 ) → πn (B, b0 ) is an isomorphism for all n > 1 and a monomor-
phism for n = 1;
(ii) if X is 0-connected, then the points of F are in 1-1 correspondence with the
cosets of p∗ (πn (X, x0 )) in π1 (B, b0 ).
[Hint. Since F is discrete, πn (F, x0 ) = πn ({x0 }, x0 ) = 0 for all n ≥ 1.]
21. If B is locally path connected and (X, p) is regular, then for x0 ∈ p −1 (b0 ),

Aut (X/B) ∼
= π1 (B, b0 )/ p∗ π1 (X, x0 )

[ Hint: Use Monodromy theorem of the regular covering space].

22. Let O (n, R) be the topological group of real orthogonal n × n matrices and
SO (n, R) be the subspace of O (n, R) of real orthogonal matrices of deter-
minant 1. Show that the inclusion map i : SO (n, R) → O (n, R) induces an
isomorphism

i ∗ : πn (SO (n, R), 1) → πn (O (n, R), 1) for n ≥ 1.

[Hint. Consider the exact homotopy sequence

i∗
· · · → πn+1 (Z2 , 1) → πn (SO (n, R), 1) −−−→ πn (O (n, R), 1) → πn (Z2 , 1),

where πn (Z2 , 1) = 0 for n ≥ 1.]

23. Suppose there exist fiber bundles S n−1 → S 2n−1 → S n , for all n. Show that the
groups πi (S n ) would be finitely generated free abelian groups computable by
induction, and nonzero for i ≥ n ≥ 2.
24. Let p : S 3 → S 2 be the Hopf bundle and q : T 3 → S 3 be the quotient map
collapsing the complement of a ball in the 3-dimensional torus T 3 = S 1 × S 1 ×
S 1 to a point. Show that p ◦ q : T 3 → S 2 induces the trivial map ( p ◦ q)∗ :
πn (T 3 ) → πn (S 2 ), but not homotopic to a constant map.
25. Let X be a path-connected space with a base point x0 ∈ X and f : S n → X
be a continuous map such that f (s0 ) = x0 , where s0 is a base point of S n . If
5.24 Exercises and Multiple Choice Exercises 413


Y =X D n+1 , and i : X → Y is inclusion, show that induced homomorphism
f
i ∗ ; πm (X, x0 ) → πm (Y, y0 )
(i) is an isomorphism if m < n;
(ii) is an epimorphism if m = n and
(iii) ker i ∗ is generated by α−1 [ f ]α ∈ πn (X, x0 ), where α ∈ π1 (X, x0 ).
26. Let ξ = (X, p, A, G) be a principal G-bundle. Show that ξ has an H-structure
iff there exists a map f˜ : A → B H such that Bα ◦ f˜ f ξ , where f ξ : A → BG
is the unique map(upto homotopy) such that f ξ∗ (ξG ) ∼
= ξ.
27. Let ξ be a numberable principal G-bundle over B × I and f be the map

f : B × I → B × I, (b, t) → (b, 1)

Show that
(i) there exists a G-morphism (g, f ) : ξ → ξ, and
(ii) the principal G-bundles ξ and f ∗ (ξ) over B × I are isomorphic.
28. Given a covering space ( X̃ , p) of X and a connected space Y, if f, g : Y → X̃
are two continuous maps such that p ◦ f = p ◦ g, and if they agree at some
point y0 ∈ Y, then show that f = g.
29. Prove the following statements :
(i) If B is a paracompact space and p : X → B is the projection of a fiber
bundle, then p is a fibration.
(ii) Given a vector bundle ξ over B × I, there exists an open covering {Ui }i∈A
of B such that ξ|(Ui × I) is trivial.
(iii) Given a paracompact space B, a continuous map

f : B × I → B × I, (b, t) → (b, 1)

and a vector bundle

ξ = (X, p, B × I, Fn ),

there exists a continuous map g : X → X with a morphism of vector bundles

(g, f ) : ξ → ξ and a linear isomorphism g on each fiber
(iv) Given a vector bundles ξ over B × I, the vector bundles f ∗ (ξ|(B × {1}) and
ξ are isomorphic.
(v) The two vector bundles ξ and ξ|((B × {1}) × I) over the same base space
B × I are isomorphic.
(vi) There exists an isomorphism of vector bundles

( f, g) : ξ|(B × {0}) → ξ|(B × {1})

414 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

30. Let γn∞ represent the n-dimensional vector bundle over the Grassmann manifold
G n (Fm ). Prove the following statements:
(i) if B is a paracompact space, then each n-dimensional vector bundle over B
is B-isomorphic to the induced vector bundle f ∗ (γn∞ ) for some continuous
map f : B → G n (F∞ );
(ii) For any two continuous maps f, g : B → G n (Fm ) such that their induced
vector bundles f ∗ (γn m ) and g ∗ (γn m ) are B-isomorphic, if

i : G n (Fm ) → G n (F2m ) for 1 ≤ m ≤ ∞

is the natural inclusion, then

i◦ f i ◦ g.

(iii) for any paracompact space B, the fiber bundle p : X → B is a fibration.

31. Show that for every n-dimensional vector bundle ξ = (X, p, B, Fn ) over a para-
compact space B, there exists a continuous function f : B → G n (F∞ ) such that
the vector bundles ξ and f ∗ ({γn∞ }) are B-isomorphic.
[ Hint: See Chap. 4. ]
32. Let f, g : (In , ∂ In ) → (X, x0 ) be two homotopic maps relative to ∂ In and ψ, φ :
I → X are two homotopic paths relative to end points. If Ft , G t : In → X : t ∈ I
are homotopies of f along the path ψ and that of g along the path φ, respectively,
show that F1 and G 1 are homotopic relative to ∂ In .
33. Let p : X → B be fibration and b, b ∈ B are two points such that they have the
same path component. Show that

p −1 (b) p −1 (b ).

34. Let
F−
→X−
→B

be a fibration sequence with fiber F, total space X and contractible base space
B. Show that
(i) X F × B and
(ii) the inclusion F → X is a homotopy equivalence.
35. Prove the following statements:
(i) An inclusion A → X of topological spaces is a cofibration iff the inclusion

(A × I) ∩ (X × 0) → X × I
5.24 Exercises and Multiple Choice Exercises 415

is a retraction.
(ii) Let A and X be two topological spaces such that A is contractible and closed
in X. If the inclusion A → X is a cofibration, then the quotient map

p : X → X/A

is a homotopy equivalence.
36. Let ξ = (X, p, B) be a covering space. Then ξ is said to be regular if the
subgroup p∗ π1 (B, b0 ) of π1 (B, b0 ) is normal. Let G be the group of covering
transformations of ξ = (X, p, B). Prove the following statements:
(i) If G is a properly discontinuous group of homeomorphisms of a space X ,
then the projection p : X → X mod G is a covering projection.
(ii) If X is connected, then this covering space ξ = (X, p, X, modG) is regular,
and G is its group of covering transformations
(iii) If a topological space X is simply connected and G is a properly discontin-
uous group of homeomorphisms of X , then the groups G is isomorphic to
the fundamental group π1 (X mod G).
37. Let η be a numberable principal G-bundle over B × I and f : B × I → B ×
I, (b, t) = (b, 1) be a map. Show that
(i) there exists a G-morphism (g, f ) : η → η and
(ii) the principal G-bundles η and f ∗ (η) are isomorphic over B × I.
38. Prove the following statements:
(i) Let (X, p) be a universal covering space of B, where B is locally path
connected. Then for any point b0 ∈ B, the groups Aut (X/B) ∼ = π1 (B, b0 ).
(ii) A principal G-bundle ξ = (X, p, B, G) is a universal G-bundle iff its total
space X is contractible.
(iii) Let A be a contractible space such that it is a closed subspace of a topological
space X. If i : A → X is cofibration, then the natural quotient map

p : X → X/A

is a homotopy equivalence.
(iv) Let B and B be two base spaces of two universal G-bundles. Then the
spaces B and B are homotopy equivalent.
(v) p : X → B be a fibration such that the points b, b ∈ B lie in the same path
component of B. Then their fibers p −1 (b) and p −1 (b ) in X are homotopy
equivalent.
39. If a map f : X → Y satisfies the covering homotopy property, show that the
inverse images of the points of Y are homotopy equivalent.
40. Given a continuous map p : (X, A, x0 ) → (Y, B, b0 ), let p : X → Y be a fiber-
ing and A = p −1 (B) be a subspace of X. Prove that p induces a bijection
416 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

p∗ : πn (X, A, x0 ) → πn (Y, B, y0 ), ∀ n > 0.

41. Let p : X → Y be a fibration and x0 ∈ X, y0 ∈ Y be two points such that p(x0 ) =

y0 . If y0 ∈ B ⊂ Y and A = p −1 (B) ⊂ X, show that p induces an isomorphism

p∗ : πn (X, A, x0 ) → πn (Y, B, y0 ), ∀ n > 0.

42. Let ξ = (X, p, B) be a principal G-bundle over B and f, g : A → B are homo-

topic maps. Show that the induced bundles f ∗ (ξ) and g ∗ (ξ) are isomorphic
G-bundles over A.
43. Show that the homotopy sequence of any bundle based at a point defined in Sect.
5.19.2 is exact,
[Hint : Using notations of Sect. 5.13.3, proceed as in exactness property of homo-
topy sequence of any triplet given in Theorem 5.13.8 for the triplet (X, F, x0 )
in this case].
44. ( Direct sum theorems for bundles ) Let ξ = (X, p, B) be a bundle over B and
b0 ∈ B be a base point. If F = p −1 (b0 ), x0 ∈ F and ξ admits a cross section,
show that
(i) πn (X, x0 ) ∼
= πn (B, b0 ) ⊕ πn (F, x0 ), ∀ n ≥ 2;
(ii) for n = 1, the fundamental group π1 (X, x0 ) contains two subgroups G
and H such that G is invariant and isomorphic to the fundamental group
π1 (F, x0 ) and p∗ maps H isomorphically onto the fundamental group
π1 (B, b0 ) and every element of π1 (B, b0 ) is uniquely determined as the
product of an element of G with an element of H ;
(iii) p∗ : πn (X, x0 ) → πn (B, b0 ) is an epimorphism for every n ≥ 1.
Hence prove that

πn (B × F, (b0 , x0 )) ∼
= πn (B, b0 ) ⊕ πn (F, x0 ), ∀ n ≥ 1.

[ Hint: For the first part, use the result that the product space is a bundle and it
admits a cross section and hence proceed as in Theorem 5.17.3 for n ≥ 2. For
n = 1, use (ii). ]
45. (Long exact sequence for a fibration) Let p : X → B be a fibration and
b0 ∈ B be a base point. If F = p −1 (b0 ), x0 ∈ F, show that p induces an a
homomorphism p∗ such that
(i)
p∗ : πn (X, F, x0 ) → πn (B, b0 , b0 )

is an isomorphism for all n > 0 and

(ii) for any path-connected base space B, there is a long exact sequence

i∗ p∗ ∂
· · · −−→ πn (F, x0 ) −−−→ πn (X, x0 ) −−−−→ πn (B, b0 ) −−−→ πn−1 (F, x0 ) −
→ ···
5.24 Exercises and Multiple Choice Exercises 417

46. Show that a covering projection p : X → B is a principal G-bundle for the group
G of covering transformations with the discrete topology.

47. Let X be a G-space X . Show that the set of all automorphisms of the trivial
G-bundle with projection
p2 : X × B → B

are in bijective correspondence with the set of all continuous maps

f : B → G.
p
48. Show that a fiber bundle ξ : X −−−→ B is trivial if the base B is contractible.
49. Prove that a fibration p : X → B is a Serre fibration iff its the base B is para-
compact.
50. Let G n,k = G k (Rn ) denote the Grassmann manifold of k-planes through the
origin in Rn . Show that
(i) O(n, R) acts transitively on G n,k ;
(ii) G n,k O(n, R)/O(k, R) × O(n − k, R).
See Bredon p 464
51. Let (X, p) be the universal covering space of B and Aut (X/B) be the group of
all automorphisms of (X,B).
Prove that
(i) the automorphism group Aut (X/B) is isomorphic to the fundamental group
π1 (B) of B and
(ii) if |π1 (B)| is the order of the group π1 (B), then |π1 (B)|=number of sheets
of the universal covering space.

5.24.2 Multiple Choice Exercises

Identify the correct alternative (s) ( there may be more than one ) from the following
list of exercises:
1. Let (X, p, B) be a universal covering space of a connected space B. If b0 ∈ B,
x0 ∈ p −1 (b0 ), then the induced homomorphism

p∗ : π1 (X, x0 ) → πn (B, b0 )

(i) is an isomorphism for n = 1;

(ii) is an isomorphism for all n > 1;
(iii) is an isomorphism for n > 5.
418 5 Topology of Fiber Bundles: Homotopy Theory of Bundles

2. Let R P m be the real projective space and Sm be the unit sphere in Rm+1 . Then

πn (R P m ) ∼
= πn (S m )

(i) if n = 1;
(ii) only if n > 1;
(iii) only if n > 6.
3. Let ξ = (X, p, B) be a principal G-bundle over B and f, g : Y → B are homo-
topic maps.
(i) The induced bundles f ∗ (ξ) and g ∗ (ξ) are isomorphic G-bundles over Y .
(ii) the induced bundles f ∗ (ξ) and g ∗ (ξ) are isomorphic G-bundles over B.
(iii) The induced bundles f ∗ (ξ) and g ∗ (ξ) are neither isomorphic G-bundles over
B nor Y.
4. Consider a fibration sequence

F−
→X−
→B

with fiber F, total space X and contractible base space B. Then

(i) the spaces X and the product space F × B are homotopy equivalent;
(ii) the inclusion i : F → X is a homotopy equivalence;
(iii) the inclusion i : F → X is a homeomorphism.
5. Let f : (X, A, x0 ) → (Y, B, y0 ) be a continuous map of triplets. If f : X → Y
is a fibering and A = f −1 (B), then the induced homomorphism

f ∗ : πn (X, A, x0 ) → πn (Y, B, y0 )

(i) is a monomorphism but it is not an epimorphism;

(ii) is an epimorphism but it is not an monomorphism;
(iii) is an isomorphism.
6. Let Ht p 2 denote the homotopy category of triplets and their continuous maps,
and Ab denote the category of abelian groups and homomorphisms. Then

πn : Ht p 2 → Ab

(i) is a covariant functor for every integer n > 2.

(ii) is a covariant functor for every integer n ≥ 2.
(iii) is a contravariant functor for every integer n > 2.
5.25 Additional Reading 419

References

Adams JF. Stable homotopy and generalized homology. University of Chicago Press; 1974.
Adhikari A, Adhikari MR. Basic topology, vol 1: metric spaces and general topology. India:
Springer; 2022a.
Adhikari A, Adhikari MR. Basic topology, vol 2: topological groups, topology of manifolds and lie
groups. India: Springer; 2022b.
Adhikari MR, Adhikari A, Groups, rings and modules with applications. Hyderabad: Universities
Press; 2003.
Adhikari MR, Adhikari A, Textbook of linear algebra: an introduction to modern algebra. New
Delhi: Allied Publishers; 2006.
Adhikari MR, Adhikari A. Basic modern algebra with applications. New Delhi, New York, Heidel-
berg: Springer; 2014.
Adhikari MR. Basic algebraic topology and its applications. India: Springer; 2016.
Arkowitz M. Introduction to homotropy theory. New York: Springer; 2011.
Bredon GE. Topology and geometry. New York: Springer; 1993.
Dugundji J. Topology. Newtown, MA: Allyn & Bacon; 1966.
Hopf H. Üeber die Abbildungen von Sphären niedriger Dimension. Fund Math. 1935;25:427–40.
Hu ST. Homotropy theory. New York: Academic Press; 1959.
Mukherjee A. Differential topology. New Delhi: Hindustan Book Agency; 2015.
Steenrod N. The topology of fibre bundles. Prentice: Prentice University Press; 1951.
Chapter 6
Geometric Topology and Further
Applications of Algebraic Topology

Geometric topology primarily studies manifolds and their embeddings in other man-
ifolds. A particularly active area is low-dimensional topology, which studies mani-
folds of four or fewer dimensions. This includes knot theory, which makes a study
of mathematical knots. This chapter gives a brief study of geometric topology by
communicating the concepts of knots and knot groups. It also gives further appli-
cations of topological concepts and results discussed in earlier chapters with a view
to understand the beauty, power and scope of the subject topology. Moreover, it
provides alternative proofs of some results proved in the previous chapters such as
Brouwer–Poincaré theorem, Van Kampen theorem, Borsuk–Ulam theorem for any
finite dimension. It proves Ham Sandwich theorem and Lusternik–Schnirelmann
theorem.

For this chapter the books [Adams, 1958, 1960, 1972], [Adhikari and Adhikari,
2014, 2016, 2022a, 2022b], [Aguilar et al. 2002], [Arkowitz and Martin, 2011],
[Armstrong, 1983], [Bredon, 1993], [Basak, 2017], [Dold and Thom, 1958], [Hu,
1966], [Steenrod, 1951] and some others are referred in the Bibliography.

6.1 Geometric Topology: Embedding Problem of the Circle

in R3 with Knot and Knot Groups

The main aim of geometric topology is to study manifolds and their maps includ-
ing embeddings of one manifold into another. This section studies the embedding
problems of geometric topology such as various embeddings of the circle S 1 in
the Euclidean 3-space R3 or in the 3-sphere S3 . Such problems arise in geometry
leading to the concept of knots. These problems make a return to geometry and are

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 421
M. R. Adhikari, Basic Topology 3,
https://doi.org/10.1007/978-981-16-6550-9_6
 422 6 Geometric Topology and Further Applications …

mainly studied in low-dimensional topology. For possible solution of such prob-

lems, knot groups are studied through the concepts of fundamental group and the
one-point compactification of R3 . This study displays an interplay among geome-
try, topology and algebra. Physicists and bio-scientists use knot theory in their study.
Historically, H. Tietze (1880–1964) laid the foundations of knot theory.

Definition 6.1.1 A knot K is a homeomorphic image of an embedding f : S 1 →

R3 , and the fundamental group
π1 (R3 − K )

of the complement of the knot K in R3 is called the knot group of K .

Example 6.1.2 As a knot K is a subspace of R3 , which is homeomorphic to the

circle, it is represented by its projection in the plane of the paper. For example, trivial
knot, figure-eight knot and square knots and some others are interesting. For their
geometrical representation see Remark 6.1.4.
(i) The trivial knot or unknot is defined by the standard embedding

i : S 1 → R3

It is called the circle knot, which is the simplest knot and consists of the unit
circle in the x y-plane with its knot group Z.
(ii) For trefoil knot see Fig. 6.1.
(iii) For figure-eight knot see Fig. 6.2.
(iv) For square knot see Fig. 6.3.

Remark 6.1.3 Since a knot K is represented by its projection in the plane of the
paper, ‘trivial knot’ or ‘unknot’ consists of the unit circle in the x y-plane.

Remark 6.1.4 For representing a knot geometrically and working with it relatively
comfortable, the standard practice is to project it into the plane in such a way that its
projection only crosses itself at a finite number of points, at most two pieces of the
knot meet at such crossing, and it does so at ‘right angles.’ For example, the knots
represented in Figs. 6.1, 6.2 and 6.3 known as ‘trefoil knot,’ ‘figure-eight knot’ and
‘square knot,’ respectively.

Definition 6.1.5 Let K 1 and K 2 be two knots. They are said to equivalent as knots
if there exists a homeomorphism

f : R3 → R3 : f (K 1 ) = K 2 .

In other words, two knots are said to be equivalent or same if there exists a
homeomorphism
6.1 Geometric Topology: Embedding Problem of the Circle… 423

Fig. 6.1 Trefoil knot

Fig. 6.2 Figure-eight knot

Fig. 6.3 Square knot

f : R3 → R3

that sends one knot onto the other knot.

Remark 6.1.6 Geometrically, two knots are equivalent if one knot can be deformed
into the other knot by a continuous deformation.
Definition 6.1.7 Let S m be the m-dimensional sphere in Rm+1 and S n be the n-
dimensional sphere in Rn+1 . If m < n and

f : Sm → Sn

is an embedding, then f (S m ) is called the m-dimensional knot in S n . It is also

called an higher-dimensional knot.
Remark 6.1.8 The higher-dimensional knots given in Definition 6.1.7 are studied
through the generalized Jordan curve theorem given in Exercise 8 of Sect. 6.4.1 by
using homology and cohomology theories [Basak, 2017].
Remark 6.1.9 The 3-sphere is considered as the one-point compactification of the
Euclidean space R3 , the latter space R3 is not compact (see Basic Topology, Volume
1). This result is utilized in Theorem 6.1.10.
424 6 Geometric Topology and Further Applications …

Theorem 6.1.10 Given a knot K , the inclusion map

i : (R3 − K ) → (S 3 − K )

induces an isomorphism between the corresponding fundamental groups

i ∗ : π1 (R3 − K ) → π1 (S 3 − K ).

Proof By definition, K is a compact subset of R3 . Again, the subspace S 3 − K is

the union of the open set R3 − K and an open ball D obtained by including the
compactification point and the complement of a large closed ball in R3 containing
the knot K . Since both the spaces D and D ∩ (R3 − K ) are simply connected and
D ∩ (R3 − K ) is homeomorphic to S 2 × R, it follows that the inclusion map

i : (R3 − K ) → (S 3 − K )

induces an isomorphism.

i ∗ : π1 (R3 − K ) → π1 (S 3 − K ).

Definition 6.1.11 (Torus knot) Given two relatively prime positive integers ( p, q),
the torus knot K p,q
K = K p,q ⊂ R3

is the image of the embedding

h : S 1 → S 1 × S 1 ⊂ R3 , z → (z p , z q ),

with the torus S 1 × S 1 having the natural embedding in R3 .

Remark 6.1.12 Geometrically, the torus knot of the form K p,q ⊂ R3 is the image
in the torus of the line with the equation px = qy in R3 is a knot that winds p times
around the torus one way and it winds q times around the other way. More precisely,
the torus knot K = K p,q winds the torus a total of p times in the longitudinal direction
and q times in the meridian direction.

Definition 6.1.13 Let f : R3 → R3 be a homeomorphism. Then f is called isotopic

to the identity map if there is a homotopy

F : R3 × I → R3

having the property

Ft : R3 → R3 , x → F(x, t)
6.1 Geometric Topology: Embedding Problem of the Circle… 425

is a homeomorphism such that F0 = 1R3 ( identity map on R3 ) and F1 = f.

Remark 6.1.14 Given two knots K 1 and K 2 , let f : R3 → R3 be a homeomorphism

isotopic to the identity map such that f (K 1 ) = K 2 .
(i) Then Ht (K 1 ) provides a family of continuous maps which move gradually from
the knot K 1 to the knot K 2 as t increases from 0 to 1.
(ii) Since S 3 is the one-point compactification of R3 , a homeomorphism f : R3 →
R3 has a unique extension to a homeomorphism f˜ : S 3 → S 3 ( see Chap. 5).

Definition 6.1.15 Let f : R3 → R3 be a homeomorphism. It is called orientation

preserving (or orientation reversing) if its extension homeomorphism

f˜ : S 3 → S 3

preserves (or reverses) the orientation of S 3 .

Example 6.1.16 Every homeomorphism f : R3 → R3 , which is isotopic to the

identity map 1R3 on R3 is orientation preserving, since every homeomorphism

Ft : R3 → R3

has a continuous extension which is the homeomorphism

t : S 3 → S 3
F

because homotopic maps have the same degree by Hopf’s classification theorem (see
Chaps. 2 and 3).

Example 6.1.17 Reflection in a plane which is a homeomorphism of R3 , transforms

every knot to its mirror image. Since it is orientation reversing, it is not isotopic to
the identity. On the other hand every orientation-preserving homeomorphism of R3
is isotopic to the identity.

Proposition 6.1.18 Equivalent knots have homeomorphic complements in R3 .

Proof Suppose that K 1 and K 2 are two equivalent knots. Then there exists a homeo-
morphism f : R3 → R3 such that f (K 1 ) = K 2 . Hence, its restriction map f |(R3 −K 1 )
determines a homeomorphism

f : R3 − K 1 → R3 − K 2 .

This asserts that the knots K 1 and K 1 have homeomorphic complements in

R3 . 

Remark 6.1.19 As the knot group π1 (R3 − K ) of K is the complement of K in R3 ,

it is utilized to classify various knots.
426 6 Geometric Topology and Further Applications …

Definition 6.1.20 Let K be a given knot. It is said to be untied if there is an isotopy

of R3 that sends K to the standard circle S 1 ⊂ R3 .
Remark 6.1.21 The circle knot is a trivial knot having its knot group isomorphic to
Z, which is the infinite cyclic group. This asserts that if a given knot K has the non
abelian knot group, then the knot K cannot be a trivial knot, and hence, this knot K
cannot be untied.

Remark 6.1.22 If there is a homeomorphism h : R3 → R3 which is isotopic to the

identity such that h(K 1 ) = K 2 for two knots K 1 and K 2 , then the knots Ht (K 1 ) give
a continuous family of maps which move gradually from K 1 to K 2 as t increases
from 0 to 1. Since S 3 is the one-point compactification of R3 , a homeomorphism
h : R3 → R3 has a unique extension to a homeomorphism h̃ : S 3 → S 3 .

Definition 6.1.23 A homeomorphism h : R3 → R3 is said to be orientation preserv-

ing (or orientation reversing) if its extension homeomorphism h̃ : S 3 → S 3 preserves
(or reverses) the orientation of S 3 .

Example 6.1.24 A homeomorphism which is isotopic to the identity is orienta-

tion preserving, because we can extend each homeomorphism Ht : R3 → R3 to the
homeomorphism H t : S 3 → S 3 , since homotopic maps have the same degree.

Example 6.1.25 Reflection in a plane is a homeomorphism of R3 and transforms

a knot to its mirror image. It is orientation reversing and cannot be isotopic to the
identity.

Remark 6.1.26 Any orientation-preserving homeomorphism of R3 is isotopic to the

identity.

Definition 6.1.27 A knot K is said to be untied if there is an isotopy of R3 that

would take K to the standard circle S 1 ⊂ R3 .

Remark 6.1.28 Circle knot is a trivial knot. If a knot K is trivial, then the funda-
mental group of its complement (which is homeomorphic to the solid torus) is the
infinite cyclic group. Hence the knot group of K is abelian. This shows that if the
knot group of a knot K is not abelian, then K cannot be a trivial knot which means
that K cannot be untied.
Remark 6.1.29 For some sort of reasonable presentation for a knot group in terms
of generators and relations the book Armstrong [Armstrong, 1983] is referred.

6.2 Further Applications of Topology

This section gives further applications of topological concepts and results discussed
in earlier chapters with a view to understand the beauty, power and scope of the
subject topology.
6.2 Further Applications of Topology 427

6.2.1 Isotopy and Its Applications

This subsection studies the concept of isotopy, which is closely related to the concept
of embedding.

Definition 6.2.1 Let X and Y be two topological spaces and H : X × I → Y be a

continuous map. Then the family of continuous maps

{Ht : X → Y, x → H (x, t)}

is said to be an isotopy, if every Ht : X → Y is an embedding.

Definition 6.2.2 Let f, g : X → Y be two embeddings. They are said to be isotopic

denoted by f ∼= g if there exists an isotopy

{Ht : X → Y } such that H0 = f, and H1 = g.

Definition 6.2.3 Let f : X → Y be an embedding. It is said to be an isotopic equiv-

alence if there exists an embedding g : Y → X such that

= 1Y : Y → Y and g ◦ f ∼
f ◦g∼ = 1 X : X → X.

Example 6.2.4 Every homeomorphism is an isotopic equivalence, but its converse

is not. For example, consider the closed interval I and its open interval X = {t :
0 < t < 1} ⊂ I. Then the inclusion map i : X → I is an isotopic equivalence but
it is not a homeomorphism. The family of homeomorphisms given in Exercise 1 of
Sect. 6.4.1 forms of a family of isotopic equivalences.

6.2.2 Application of Topology to Theory of Numbers

It is well known that the set of prime integers is infinite. This result is proved in this
section by using topological tools.

Proposition 6.2.5 Let  be a family of subsets of N consisting of the empty set ∅

and all those subsets of N which are expressible in the form

X n = {n, n + 1, n + 2, n + 3, . . . : n ∈ N}.

Then  forms a topology on N.

Proof X 1 = N shows that the whole set N is in . Again since,  is totally ordered
by set inclusion, it follows that intersection of any two sets in  is also in . Let
S be a subfamily of  − {∅, N} in the sense that S = {X n : n ∈ J ⊂ N}. As J is a
subset of positive integers, it has a smallest positive integer p. Hence
 428 6 Geometric Topology and Further Applications …

{X n : n ∈ J } = { p, p + 1, p + 2, p + 3, . . .} = X p ∈ .

Hence it follows that  forms a topology on N. 

Proposition 6.2.6 (i) Given a nonempty subset X ⊂ N, let there exist a positive
integer n X such that X contains no arithmetic progressions (AP) of length greater
than n X . Then subsets of N having the property together with ∅ and the set N
form a collection of closed set for some topology on N.
[ Use Van der Waerden’s theorem which asserts that given an integer n ∈ N,
there is an integer n 0 such that for any subset X ⊂ {1, 2, . . . , n 0 } = Y, either X
or Y − X contains an AP.
Let A and B be two subsets of N which contain no AP of length at least the given
n. If A ∪ B contains a sufficiently long AP, then A or B contains an AP of length
more than n implies a contradiction.]
(ii) The collection of all infinite APs in N forms a base for some topology on N.
[ Use the result that the finite intersection of APs in N is also an AP.]
(iii) Using this topology on N show that the set of prime integers is infinite.
[ Hint: The sets A(k, d) = {k, k + d, k + 2d, . . . : k = 1, 2, . . . , d} are open,
pairwise disjoint, and form a covering of N. Hence, it follows that each of them
is closed. As a particular situation, for each prime integer p, the sets of the form
{ p, 2 p, 3 p, . . . , } forms a covering of N − {1}. This shows that the set of prime
integers cannot be finite; otherwise, if the set were finite, then the set {1} would be
open. This shows that it is not a union of arithmetic progressions. This concludes
that the set of prime integers cannot be finite and hence it is infinite.]

Proposition 6.2.7 There are infinitely many primes in Z.

Proof First a topology is defined on Z by using doubly infinite APs. Given k, d ∈

Z, d = 0, define a set A(k, d) = {k + nd : n ∈ Z}. Call a nonempty subset U ⊂ Z
to be open if it is a union of sets of the form A(k, d) or U = ∅. Then Z is an open
set and an arbitrary union of sets of the form A(k, d) is also of the same form. Let U
and V are of the same form and a ∈ U ∩ V be an arbitrary element. Let a ∈ A(k, d1 )
and a ∈ A(k, d2 ) be two APs containing k such that

a ∈ A(k, d1 ) ⊂ U

and
a ∈ A(k, d2 ) ⊂ V.

Then a ∈ A(k, d1 d1 ) ⊂ A(k, d1 ) ∩ A(k, d2 ) ⊂ U ∩ V. This asserts that U ∩ V is an

open set and every nonempty open set is infinite. A(k, d) is also a closed set. A(k, d)
can be expressed as

d−1
A(k, d) = Z − A(k + r, d).
r =1
6.3 Borsuk–Ulam Theorem with Applications 429

As any integer m ∈ Z − {−1, 1} has at least one prime divisor p, the integer m
is contained in A(0, p). Hence it follows that

Z − {−1, 1} = A(0, p).
p is prime

If possible, suppose there are finitely many primes in Z. Then the finite union of
closed sets 
A(0, p)
p is prime

being a closed set in Z, the above equality implies that the subset {−1, 1} is open in
Z, which is not true, as it is finite. This implies a contradiction. 

6.3 Borsuk–Ulam Theorem with Applications

The aim of this section is to prove Borsuk–Ulam theorem 6.3.1 in a general form
by using homology theory for all finite dimensions by generalizing this theorem for
two-dimensional case proved in Chap. 2 by homotopy theory. Finally, this section
provides some of its applications such as Ham Sandwich theorem and Lusternik–
Schnirelmann theorem obtained as direct consequences of Borsuk–Ulam theorem
6.3.1. Historically, S. Ulam (1909–1984) formulated the Borsuk–Ulam theorem with-
out any correct proof but K. Borsuk (1905–1982) proved it first in 1933. Since then,
different alternative proofs are found in literature was a conjecture posed by Borsuk.

6.3.1 Borsuk–Ulam Theorem

This subsection proves Borsuk–Ulam theorem by using homology theory.

Theorem 6.3.1 (Borsuk–Ulam Theorem ) Given any pair of integers m, n with

m > n ≥ 0, there does not exist any continuous map

h : Sm → Sn ,

which preserves antipodal points in the sense h(x) = h(−x), ∀ x ∈ S m .

Proof This theorem is proved by method of contradiction. Suppose there exists a

continuous map
h : S m → S n : h(x) = h(−x), ∀ x ∈ S m

and S n is obtained from S m by taking the last m − n coordinates equal to zero. Let
430 6 Geometric Topology and Further Applications …

i : S n → S m

be the usual inclusion map. Consider the composite map

i ◦ h : Sm → Sm .

This map preserves the antipodal points. This implies that deg (i ◦ h) must an odd
integer. Again consider the induced composite homomorphism

h∗ i∗
(i ◦ h)∗ = i ∗ ◦ h ∗ : Hm (S m ) −−−−→ Hm (S n ) −−−→ Hm (S m )

is the trivial homomorphism in the homology theory because it is a sequence covariant

functors and the homology group Hm (S n ) = 0, f or m > n. This proves that deg (i ◦
h) = 0, which produces a contradiction. 
Remark 6.3.2 As a direct consequence of Borsuk–Ulam theorem 6.3.1, it follows
that geometrically, every continuous map h : S n → Rn identifies a pair of antipodal
points of S n .
Corollary 6.3.3 Let h : S n → Rn be continuous map such that it preserves the
antipodal points of S n . Then there exists a point x ∈ S n such that h(x) = 0.
Proof Suppose there exists no point x ∈ S n such that h(x) = 0. Then

h(x) = 0, ∀ x ∈ S n .

Define a map
h(x)
f : S n → S n−1 , x → .
||h(x)||

Then f is well-defined and continuous. By hypothesis, h preserves antipodal points.

This implies that the continuous map f also preserves antipodal points. This contra-
dicts the Borsuk–Ulam theorem 6.3.1. 
Corollary 6.3.4 The n-sphere S n is not embeddable in Rn .
Proof Since the sphere S n is not homeomorphic to a subspace of Rn by Remark
6.3.2, it is proved that the sphere S n cannot be embedded in Rn . 

6.3.2 Ham Sandwich Theorem

This subsection proves Ham Sandwich theorem 6.3.5 as a direct consequence of

Borsuk–Ulam theorem 6.3.1.
Theorem 6.3.5 (Ham Sandwich Theorem) Given n bounded convex subsets X 1 ,
X 2 , . . . , X n of Rn , there exists a hyperplane bisecting all of the X i ’s simultaneously.
6.3 Borsuk–Ulam Theorem with Applications 431

Proof Case I: First suppose that n = 3 and consider the continuous map

f : S 3 → R3 , ( f 1 (x), f 2 (x), f 3 (x)),

where f i : S 3 → R is a continuous map such that for every x ∈ S 3 , the numerical

value f i (x) denotes the volume of the part of X i lying on the same side of the hyper-
plane Hx at the point x and passing through the point (0, 0, 0, 1/2) for i = 1, 2, 3.
Then there exists a point s0 ∈ S 3 such that f (s0 ) = f (−s0 ) by Borsuk–Ulam theorem
6.3.1. This asserts that f 1 (s0 ) = f 1 (−s0 ), f 2 (s0) = f 2 (−s0) and f 3 (s0 ) = f 3 (−s0 ).

Case II: For n > 3, proceed as in Case I. 

6.3.3 Lusternik–Schnirelmann Theorem for Higher

Dimension

Borsuk–Ulam theorem 6.3.1 is applied to prove Lusternik–Schnirelmann theorem

6.3.6 for S n , which generalizes Lusternik–Schnirelmann for dimension 2, proved in
Chap. 2.

Theorem 6.3.6 ( Lusternik–Schnirelmann theorem for S n ) If S n is covered by n + 1

closed sets X 1 , X 2 , . . . , X n+1 of S n , then one of them contains a pair of antipodal
points.


n+1
Proof By the given condition, X i = S n ⊂ Rn+1 . Consider Rn as the Euclidean
i=1
space with usual metric d. Let d(x, X i ) be the distance of x from the closed set X i .
Construct a continuous map

f : S n → Rn , x → (d(x, X 1 ), . . . , d(x, X n )).

Then Borsuk–Ulam theorem 6.3.1 implies that f must identify a pair of antipodal
points of S n . This implies that there exists a point s0 ∈ S n such that d(s0 , X i ) =
d(−s0 , X i ) for 0 ≤ i ≤ n. Consider the two possible cases.

Case I: If d(s0 , X i ) = 0 for some i, then the pair of antipodal points s0 , −s0 ∈ X i .
Because every X i being a closed set by hypothesis, both the points s0 , −s0 ∈ X i .

Case II: If d(s0 , X i ) > 0 for all i = 1, 2, . . . , n, then the pair of antipodal points
s0 , −s0 ∈ X n+1 because X i ’s form a cover of S n by hypothesis. 
432 6 Geometric Topology and Further Applications …

6.3.4 Van Kampen Theorem: An Application of Graph

Theory

This subsection proves Van Kampen theorem 6.3.16 by using graph-theoretic results
together with algebraic concept of free product of two groups. This theorem provides
a technique for computing the fundamental groups of topological spaces which are
decomposed into simpler spaces having their fundamental groups already known and
expressing the edge group as a set of generators and relations. This theorem is also
known as Seifert–Van Kampen theorem.

Definition 6.3.7 A tree T is a 1-dimensional subcomplex of a complex having its

polyhedron both path connected and simply connected.

Definition 6.3.8 A tree MT is said to be maximal if for any tree T ∗ containing MT ,

the trees T ∗ = MT .

Definition 6.3.9 The r -skeleton K r of a simplicial complex K is the subcomplex

of K which consists of all n-faces of simplexes of K having n ≤ r.

Definition 6.3.10 Given a map f : |K | → |L| between polyhedra and a point x in

|K |, the point f (x) ∈ |L| belongs to the interior of a unique simplex of L, called the
carrier of f (x).

Definition 6.3.11 Let f : |K | → |L| be a continuous map between polyhedra. Then

a simplicial map s : K → L is said to be simplicial approximation of f, if s(x) is
a point of the carrier of f (x) for every point x in K .

Proposition 6.3.12 A maximal tree of a complex contains all its vertices.

Proof Let MT be a maximal tree of a complex K which does not contain all its
vertices. There exists at least one vertex v of K such that v is in K − MT . Since K
is path connected for any vertex w of MT , the two vertices w and v can be joined
by a path in |K |. Then this path can be replaced by an edge path wv1 v2 · · · vm v by
simplicial approximation theorem. Let vk be the last vertex of this edge lying in MT .
Now, a new subcomplex S of K is constructed by including the vertex vk+1 and the
edge generated by vk vk+1 · · · . Then |S| is a deformation retract of |MT |, and hence,
S is also tree. But it contradicts the assumption that MT is a maximal tree. This
contradiction asserts that the maximal tree MT does contain all the vertices of K . 

Definition 6.3.13 Let K be a complex and S be a subcomplex of K with |S| be

simply connected. List the vertices of K as v = v0 , v1 , v2 , . . . , vm and denote the
generators gi j for every pair of vertices vi and v j of K as follows:

(i) gi j = 1, if vi and v j span a simplex of S,

(ii) gi j g jk = gik if vi , v j and vk span a simplex of K ,
 6.3 Borsuk–Ulam Theorem with Applications 433

(iii) gii = 1,
(iv) gi−1
j = g ji .

The group generated by gi j is denoted by G(K , S).

Construct another group E(K , S) called the edge group of K based at a vertex of
K in Definition 6.3.14.

Definition 6.3.14 Define another group E(K , S) called the edge group of K based
at a vertex of K as follows:
An edge path in K is a sequence v0 , v1 , . . . , vn of vertices of K such that every
consecutive pairs vi and vi+1 spans a simplex of K with the possibility vi = vi+1 .
Define a simplicial version of homotopy by defining two edge paths of K to be
equivalent if one can be obtained from the other by a finite number of operations of
the forms

(i) if v j−1 = v j , then replace · · · v j−1 v j · · · by · · · v j · · · and conversely

(ii) or, if {v j−1 , v j , v j+1 } spans a simplex of K , which may not be a 2-simplex,
replace · · · v j−1 v j v j+1 · · · by · · · v j−1 v j+1 · · · and conversely (like two sides of
a triangle can be replaced by the third side and vice versa).

This is an equivalence relation between the edge paths. The class corresponding
to the edge path vo v2 · · · v j is denoted by [vo v2 · · · v j ]. The set of the equivalence
classes of edge loops at the vertex v of K , denoted by v forms a group under the
binary operation (juxtaposition)

[vv1 v2 · · · vn v] · [vw1 w2 · · · wn v] = [vv1 v2 · · · vn vvw1 w2 · · · wn v]

with the identity element [v] and the inverse of its element [vv1 v2 · · · vn v] is the
element [vvn · · · v2 v1 v].
This is called the edge group of K based at its vertex v and is denoted by
E(K , v).

Theorem 6.3.15 proves that the groups E(K , S) and E(K , v) are isomorphic.

Theorem 6.3.15 Let S be a subcomplex of K such that |S| be simply connected.

Then the groups G(K , S) and E(K , v) are isomorphic.

Proof To prove the theorem first define a homomorphism

α : G(K , S) → E(K , v), gi j → [ei v j e−1

j ].

Define another homomorphism

β : E(K , v) → G(K , S), [vvk vt vm . . . vn v] → g0k gkt gtm gn0 .

434 6 Geometric Topology and Further Applications …

Then
α ◦ β = 1 E(K ,v) and β ◦ α = 1G(K ,S)

assert that α is an isomorphism with β as its inverse isomorphism. 

Given two simplicial complexes S, K in the same Euclidean space, if they inter-
sect in a common subcomplex and |S|, |K |, |S ∩ K | are all path-connected spaces
with their known fundamental groups, then the fundamental group π1 (|S ∪ K |) is
calculated by Van Kampen theorem 6.3.16.

Consider the two possibilities:

(i) If the simplicial complexes S and K intersect in a single vertex, then any edge
loop in S ∪ K based at this vertex is a product of loops, each of which is in either
S or K . This facilitates to calculate the free product π1 (|S|) ∗ π1 (|K |) for the
fundamental group of |S ∪ K |.
(ii) For the general situation, an analogous arguments holds, except that the free
product π1 (|S|) ∗ π1 (|K |) effectively counts the homotopy classes of these loops
lying in |S ∩ K | twice (one in each of π1 (|S|), π1 (|K |)). So, in this case, some
extra relations are required as given in Van Kampen theorem 6.3.16.

Theorem 6.3.16 (Van Kampen Theorem) Let S, K be two simplicial complexes in

the same Euclidean space with |S|, |K | and |S ∩ K | are path-connected spaces.
Suppose
i : |S ∩ K | → |S|,

j : |S ∩ K | → |K |

are inclusion maps and v is a vertex of S ∩ K , which is taken as a base point of S ∩ K .

Then the fundamental group π1 (|S ∪ K |, v) is the free product π1 (|S|, v) ∗ π1 (|K |, v)
with the relations i ∗ (x) = j∗ (x) for every x ∈ π1 (|S ∩ K |, v).

Proof Let MT be a maximal tree in S ∩ K . Extend it to a maximal tree MT1 in S and a

maximal tree MT2 in K . Then MT1 ∪ MT2 is a maximal tree in S ∪ K . By proposition
6.3.12 and theorem 6.3.15, it follows that the group π1 (|S ∪ K |) is generated by
elements gi j corresponding to edges of S ∪ K − MT1 ∪ MT2 , with relations gi j g jk =
gik provided by the triangles of S ∪ K . But this is precisely the group obtained by
taking a generator bi j for each edge of S − MT1 , a generator ci j for each edge of
K − MT2 , with relation of the form bi j b jk = bik , ci j c jk = cik corresponding to the
triangles of S, K with additional relations bi j = ci j , whenever bi j and ci j correspond
to the same edge of S ∩ K . Since the edges of S ∩ K − TM , considered as edges of
S give a set of generators for i ∗ (π1 (|S ∩ K |)). Similarly the same edges, considered
also as edges of K , give a set of generators for j∗ (π1 (|S ∩ M|)). 
6.3 Borsuk–Ulam Theorem with Applications 435

6.3.5 Proof of Jordan Curve Theorem by Homology Theory

This subsection continues the study of Jordan curve theorem initiated in Chap. 2 and
studies it and its generalization from the viewpoint of homology theory. Recall that
a homeomorphic image of a circle in the plane R2 is a Jordan curve J. This theorem
says that the complement in the plane R2 of a Jordan curve J consists of two open
components, each of which as J as its boundary. It is one of the most classical the-
orems in topology, and it is also one of the oldest problems of a purely topological
nature. It is related to connectedness and continuum theory. Intuitively, this theorem
is simple asserting that a Jordan curve J ( which is a subspace of R2 homeomorphic to
S 1 ) separates R2 into two complementary components. But its proof is not obvious.
The first correct proof was given by O. Veblem in 1905 [Veblem, 1905]. This sub-
section proves Jordan curve theorem and its generalization by using homology theory.

Lemma 6.3.17 (i) If X is subspace of S n homeomorphic to Dk for some k ≥ 0,

then the reduced homology groups

i (S n − X ) = 0, for every i;
H

(ii) If A is a subspace of S n homeomorphic to S k for some k with 0 ≤ k < n, then


i (S − A) ∼
n Z, if i = n − k − 1
H =
0, otherwise.

Proof (i) Apply induction on k. For k = 0, the proof is trivial, because the
space S n − X is homeomorphic to Rn . Next, if f : Ik → X is a homeomor-
phism, then the open sets D = S n − f (Ik−1 × [0, 1/2]) and S = S n − f (Ik−1 ×
[1/2, 1]), then D ∩ S = S n − X and D ∪ S = S n − f (Ik−1 × {1/2}). By induc-
tion 
Hi (D ∪ S) = 0 for all i. This asserts by Mayer–Vietoris sequence that there
are isomorphisms

ψ:H i (D) ⊕ H
i (S n − X ) → H i (S) for every i.

Since the two components of ψ are induced by the inclusions S n − X → D and

S n − X → S, it follows that there exists an i-dimensional cycle β in S n − X that
is not a boundary in S n − X. This implies that β is also not a boundary in at least
one of D and S. For, i = 0, ‘cyclic’ is considered augmented chain complexes,
which are studied in reduced homology. In an analogous way, subdivide the
last I factor of Ik into quarters, eights, . . . to get a nested sequence of closed
subintervals I1 ⊃ I2 ⊃ · · · with intersection one point p ∈ I, such that β is not a
boundary in S n − f (Ik−1 × Im ) for any m. By induction on k, β is the boundary of
a chain α in S n − f (Ik−1 × { p}). This implies that α is a finite linear combination
of singular simplices with compact image in S n − f (Ik−1 × { p}). This asserts by
436 6 Geometric Topology and Further Applications …

compactness α is a chain in S n − f (Ik−1 × Im ) for some m. This contradiction

implies that β is a boundary in S n − X . Hence (i) follows by induction on k.
(ii) Apply induction on k. For k = 0 the case is trivial because in this case S n − A ≈
S n−1 × R. Represent the space A as a union of two subspaces X 1 ∪ X 2 , where X 1
and X 2 are homeomorphic to D k and X 1 ∩ X 2 is homeomorphic to S k−1 . Apply
Mayer–Vietoris sequence for C = S n − X 1 and B = S n − X 2 , both of which
have trivial reduced homology groups by (i). Hence there exist isomorphisms

i (S n − A) ∼
H =Hi+1 (S n − (X 1 ∩ X 2 )), ∀ i.

This implies (ii). 

Theorem 6.3.18 (Jordan curve ) The complement of a Jordan curve J in the plane
R2 consists of two open components, each of which as J has its boundary.

Proof Use (ii) of Lemma 6.3.17 to prove that a subspace of S 2 homeomorphic

to S 1 separates S 2 into two complementary open complements, since open sub-
sets of S n are locally path connected. Finally, use R2 in place of S 2 to complete
the proof because deleting a point from an open set in S 2 does not change its
connectedness, 

Theorem 6.3.19 gives a generalization of the Jordan curve theorem 6.3.18 and its
proof is analogous.

Theorem 6.3.19 (Generalized Jordan curve theorem) Every subspace of S n home-

omorphic to S n−1 separates it into two components, and these components have the
same homology group same as the homology group of a point. Moreover, both the
complementary regions are homeomorphic to open balls.

Proof Proceed as in proof of Theorem 6.3.18. 


6.3.6 Homology Groups of Sin
i∈A


This subsection computes the homology groups of the wedge product Sin of any
i∈A
family {Sin : i ∈ A} of the n-spheres.

Theorem 6.3.20 Let A be an indexing set, and Sin be a copy of the n-sphere for each
i ∈ A. Then the reduced homology groups

 Z(i), if m = n
m (
H Sin ) ≡
i∈A
0, otherwise,
6.3 Borsuk–Ulam Theorem with Applications 437

where Z(i) is a free abelian group with generators i ∈ A.
i∈A
  
Proof Consider the topological spaces ( Sin ) and  Sin = Sin+1 . Since
i∈A i∈A i∈A
they are homotopy equivalent, the theorem follows. 

6.3.7 More Application of Euler Characteristic

Euler characteristic has various applications. For example, it is proved using Euler
characteristic that there are only five regular simple polyhedra (see Chap. 2). This
subsection studies Euler characteristic of a finite C W -complex.
Theorem 6.3.21 Let X be a finite CW-complex such that χ(X ) = 0. If ψt : X → X
is a flow, then there exists a fixed point of ψt for every t ∈ R.
Proof By hypothesis, ψt : X → X is a flow. If ∧ψt and ∧1 X denote the Lefschetz
numbers of the map ψt and the identity map 1 X on X respectively, then

∧ψt = ∧1 X = χ(X ) = 0.

Hence by Lefschetz fixed-point theorem, it follows that there exists a fixed point
x0 (t) of ψt . Construct a sequence of subspaces

X n = {x ∈ X : ψ1/2n (x) = x}

for n = 1, 2, . . . , . This implies that X n ⊃ X n+1 , and every space X n is a nonempty

closed set such that
X∞ = X n = ∅.
n

This implies that X ∞ is a set of points fixed under all rational numbers of dyadic
form k/2n . Since the set of rational numbers are dense is the real number space R,
every element in X ∞ is a fixed point of ψt for any t ∈ R. 

6.3.8 Application of Algebraic Topology to Algebra

In general, algebraic topology involves algebraic techniques to obtain topological

information but some algebraic results can be proved conveniently in a way where
the direction is reversed. For example, Theorem 6.3.22 in algebra is proved by using
tools of algebraic topology given in Exercise 4 of Sect. 6.4.1.
Theorem 6.3.22 Any subgroup G of a free group F is free.
438 6 Geometric Topology and Further Applications …

Proof By hypothesis, G is an arbitrary subgroup of a free group F. Let B be a basis

of F. Then corresponding to the basis B, there exists a graph G B by taking G B
the wedge of circles such that π1 (G B ) ∼
= F. Hence there exists a covering space
p : E → G B by Exercise 4 of Sect. 6.4.1 such that
(i) the induced homomorphism

p∗ : π1 (E) → π1 (G B )

is a monomorphism and
(ii) p∗ (π1 (E)) = G.
Since p∗ is a monomorphism, π1 (E) ∼ = G. Finally, it follows that the group G ∼
=
π1 (E) is free, because E is a graph by using Exersise 4 of Sect. 6.4.1. 
Remark 6.3.23 It has been proved in an earlier chapter that the antipodal map

A : S n → S n , x → −x

generates an action of Z2 on S n with the projective space R P n the corresponding

orbit space S n mod Z2 . It is a compact connected manifold of dimension n, and this
action is free. Hence a natural question arises: does there exist any other finite group
which acts freely on S n ? Theorem 6.3.25 gives its partial answer which asserts that
Z2 is the only nontrivial group that can act freely on S n for an even integer n. To
prove this result, we first prove Proposition 6.3.24.
Proposition 6.3.24 If a topological group G acts on a topological space X, then
every g ∈ G, defines a homeomorphism

ψg : X → X, x → gx.

Proof For every g ∈ G, the map

ψg : X → X, x → gx

is continuous and ψg is such that

(i) ψg ◦ ψg−1 = 1 X and
(ii) ψg ◦ ψg−1 = 1 X .
Hence it follows that ψg is a homeomorphism. 
Theorem 6.3.25 If n is an even integer, Z2 is the only nontrivial group that can act
freely on S n .
Proof Consider an action of an arbitrary topological G on S n

ψ : G × S n , (g, x) → gx.
6.3 Borsuk–Ulam Theorem with Applications 439

Then for every g ∈ G, the map

ψg : S n → S n , x → gx

is a homeomorphism by Proposition 6.3.24. Since the degree of a homeomorphism

is ±1, the action of G on S n determines a degree function

d : G → {±1}, g → deg ψg .

The degree function d is a homomorphism. If the action is free, then d maps every
nontrivial element of G to (−1)n+1 , since ψg has no fixed point. Consequently, if n
is even, d has trivial kernel and hence G ⊂ Z2 . This proves the theorem. 
Remark 6.3.26 Sect. 6.4.1 provides for more results of algebra that can be proved
by using algebraic topology.

6.3.9 Whitehead Theorem and Its Applications

Whitehead Theorem 6.3.29 is a basic result in algebraic topology proved by J. H. C.

Whitehead in [Whitehead, 1949], where he introduced the concept of C W complex.
Definition 6.3.27 Let T op∗ be the category of pointed topological spaces and their
continuous maps. Then a map f : (X, x0 ) → (Y, y0 ) ∈ T op∗ with y0 = f (x0 ) is
said to be a weak homotopy equivalence if its induced map
(i) f ∗ : π0 (X, x0 ) → π0 (Y, y0 ) is a bijection and
(ii) f ∗ : πn (X, x0 ) → πn (Y, y0 ) is an isomorphism of groups for all n ≥ 1.
On the other hand, f is said to be an m equivalence for some integer m ≥ 1 if
(i) f ∗ : πm (X, x0 ) → πm (Y, y0 ) is an isomorphism of groups for all 0 < m < n and
(ii) f ∗ : πm (X, x0 ) → πm (Y, y0 ) is an epimorphism for m = n.
A weak homotopy equivalence f : (X, x0 ) → (Y, y0 ) ∈ T op∗ with y0 = f (x0 ) is
sometimes written in brief as f : X → Y.
Proposition 6.3.28 Let f : X → Y be a weak homotopy equivalence between C W
complexes and K be a C W complex with base point k0 a 0-cell. Then the induced
map
f ∗ : [K , X ] → [K , Y ], [α] → [ f ◦ α]

is a bijection.
Proof Let M f be the mapping cylinder of the given map f : X → Y ∈ T op∗ with
inclusion i : X → M f and g : M f → Y be a homotopy equivalence. Then f =
g ◦ i. This implies that i is also a weak homotopy equivalence. This asserts that
440 6 Geometric Topology and Further Applications …

i ∗ : [K , X ] → [K , Y ]

is a bijection. This proves that f ∗ = g∗ ◦ i ∗ is a bijection, since g∗ is a bijection. 

Every homotopy equivalence is a weak homotopy equivalence by definition, but

its converse is true under certain conditions prescribed in Whitehead theorem 6.3.29.
Theorem 6.3.34 provides a sufficient condition under which the concepts a weak
homotopy equivalence and homotopy equivalence coincide.

Theorem 6.3.29 (Whitehead) Let f : X → Y be a weak homotopy equivalence

between C W complexes X and Y. Then it is also a homotopy equivalence.

Proof By hypothesis, f : X → Y is a weak homotopy equivalence between C W

complexes. Then by Proposition 6.3.28, it follows that

f ∗ : [Y, X ] → [Y, Y ], [α] → [ f ◦ α]

is a bijection. Then there exists a continuous map g : Y → X such that f ◦ g  1Y .

This implies that g is also a weak homotopy equivalence. Similarly, it can be proved
that g ◦ f  1 X . Consequently, f is a homotopy equivalence with g its homotopy
inverse. 

Corollary 6.3.30 (Another form of Whitehead theorem) Let f : X → Y be a weak

homotopy equivalence between C W complexes X and Y such that its induced homo-
morphism

f ∗ : πn (X ) → πn (Y )

is an isomorphism for every integer n ≥ 1.

6.3.10 Eilenberg–MacLane Spaces and Their Applications

Eilenberg–MacLane spaces form an important family of topological spaces having

only one nontrivial homotopy groups. Such spaces introduced by S. Eilenberg (1915–
1998) and S. MacLane (1909–2005) in 1945 are named after them. The importance
of Eilenberg–MacLane spaces is twofold. Because Eilenberg–MacLane spaces
(i) develop homotopy theory and
(ii) they closely link the study of cohomology theory with homotopy theory.

Definition 6.3.31 Let G be an arbitrary group. Given a positive integer n, an

Eilenberg–MacLane space of type K (G, n) is a pointed C W -complex X such that
(i) X has only one nontrivial homotopy group;
 6.3 Borsuk–Ulam Theorem with Applications 441

(ii) πn (X ) = G and all other homotopy groups of X (i.e., in all dimensions except
n) vanish;
(iii) the group G is to be abelian for all n > 1.
The concept of an Eilenberg–MacLane space K (G, n) is well-defined for all
n ≥ 1, because there is only one space of type K (G, n) upto homotopy equivalence.

Definition 6.3.32 An Eilenberg–MacLane space is of the form

(i) K (G, 1) is a path-connected space having fundamental group isomorphic to a

given group G and a contractible universal covering space.
(ii) K (G, 0) is defined to be the group G with the discrete topology.

Example 6.3.33 Consider the homotopy groups of the infinite dimensional real pro-
jective space RP∞ , of the infinite dimensional complex projective space CP∞ and
of unit circle S 1 in C :
Z2 , if i = 1,
πi (RP∞ ) =
0, if i = 1.

Z, if i = 2,
πi (CP∞ ) =
0, if i = 2.

Z, if i = 1,
πi (S 1 ) =
0, if i = 1.

Hence it follows that

(i) K (Z2 , 1) = RP∞ (infinite dimensional real projective space).

(ii) K (Z, 2) = CP ∞ (infinite dimensional complex projective space).
(iii) K (Z, 1) = S 1 (unit circle in C), but S 2 is not an Eilenberg–MacLane space of
type K (Z, 2).

Theorem 6.3.34 provides a sufficient condition under which the concepts of a weak
homotopy equivalence and homotopy equivalence coincide, which are in general
different.

Theorem 6.3.34 (Whitehead) For every abelian group G, there exists a weak homo-
topy equivalence

ψn : K (G, n) → K (G, n + 1), ∀ n ≥ 1,

which is also a homotopy equivalence.

Proof For every integer n ≥ 1, using the earlier results such as

(i) πn (K (G, n + 1)) ∼

= πn+1 (K (G, n + 1)) ∼
= G and
(ii) πn (K (G, n + 1)) ∼
= πn (K (G, n)) ∼
= G.
442 6 Geometric Topology and Further Applications …

Hence it follows that there exists a continuous map

ψn : K (G, n) → K (G, n + 1)

for every n ≥ 1 and its induced homomorphism

ψn ∗ : πn (K (G, n)) → πn (K (G, n + 1))

is an isomorphism for every n ≥ 1. This implies that

ψn : K (G, n) → K (G, n + 1)

is a weak homotopy equivalence for every n ≥ 1, since all other homotopy groups
are trivial. Finally, since the loop space K (G, n + 1) is in the homotopy type of a
C W -complex, it follows that ψn is a homotopy equivalence. 

Theorem 6.3.35 Let X be a C W -complex and K (G, n) be an Eilenberg- MacLane

space. Then the set [X, K (G, n)] of homotopy classes of maps f : X → K (G, n)
admits a group structure for every n ≥ 1.

Proof By Whitehead theorem 6.3.34 the Eilenberg–MacLane space K (G, n) is

homotopy equivalent to the H -space K (G, n + 1), ∀ n ≥ 1. This proves that the
set [X, K (G, n)] admits a group structure ( see Chap. 2). 

Dold–Thom theorem 6.3.36 establishes a close relation between the groups

πi (S P ∞ (X )) and Hi (X ; Z) for every C W -complex X in T op∗ .

Theorem 6.3.36 (Dold Theorem) The functor

F1 : T op∗ → Gr p, X → πi (S P ∞ (X ))

and the functor

F2 : T op∗ → Gr p, X → Hi (X ; Z)

coincide for every i ≥ 1.

Proof See [Dold and Thom, 1958] 

Corollary 6.3.37 For a connected C W -complex X , there is a natural isomorphism

ψ : πn (S P ∞ (X )) → H n (X ; Z),

for every n ≥ 1.

Proof It follows from Dold–Thom theorem 6.3.36. 

Corollary 6.3.38 S P ∞ (S n ) is a K (Z, n).

6.3 Borsuk–Ulam Theorem with Applications 443

Proof Take in particular, X = S n in Dold–Thom theorem 6.3.36 to prove the corol-

lary. 

Example 6.3.39 (i) The infinite symmetric product space S P ∞ (S n ) of S n is an

Eilenberg–MacLane space K (Z, n) for every integer n ≥ 1 because it is a CW-
complex having only one nontrivial homotopy group such as πn (K (Z, n)) ∼= Z.
(ii) The n-symmetric product space S P n (S 2 ) ≈ CP n .
(iii) The infinite symmetric product space S P ∞ (S 2 ) ≈ CP ∞ .

Remark 6.3.40 For the inclusion map

X = S P 1 (X ) → S P ∞ X,

its induced homomorphism

πn (X ) → πn (S P ∞ (X )) = Hn (X ; Z)

is the Hurewicz homomorphism (see Exercise 7 of Sect. 6.4.1). In particular for

X = S 1 , the map
S P n (S n ) → S P ∞ (S n )

induces on π1 by Hurewicz theorem, an isomorphism

σ1 : Z → Z.

6.3.11 Adams Theorem on Vector Field Problem

Adams theorem 6.3.43 solved a long a standing problem on vector field in 1960
saying that there exists a continuous map

f : S 2n−1 → S n

having Hopf invariant one only when n = 2, 4 and 8. This theorem is named after J.
F. Adams who solved this problem in his papers [Adams, 1958, 1960].

Definition 6.3.41 Let f : S 2n−1 → S n be a continuous map for n ≥ 1. Then there

exists a unique integer depending only on the homotopy class of f. This integer
denoted by H( f ) is called the Hopf invariant of the map f.

Remark 6.3.42 The assignment

ψ : π2n−1 (S n ) → Z, f :→ H( f )
444 6 Geometric Topology and Further Applications …

is a homomorphism such that H( f ) = 1 only when n = 2, 4 and 8 by Theorem

6.3.43.

Adams theorem 6.3.43 proves an algebraic result saying that R, C and H are the
only nontrivial real division algebra. This proof is beyond the scope of the book and
only referred.

Theorem 6.3.43 (Adams) there exists a continuous map

f : S 2n−1 → S n

having Hopf invariant one only when n = 2, 4 and 8.

Proof See Adams [(Adams, 1958, 1960)]. 

6.4 Exercises and Multiple Choice Exercises

As solving exercises plays an essential role in learning mathematics, various types

of exercises and multiple choice exercises are given in this section. They form an
integral part of the book series.

6.4.1 Exercises

1. Let Rn be the n-dimensional Euclidean space. Prove the following statements:

(i) every disk Dr = {x ∈ Rn : ||x|| = r } of an arbitrary radius r is homeomor-

phic to Dn ;
(ii) the unit dis Dn = {x ∈ Rn : ||x|| = 1} in Rn is homeomorphic to Rn ;
(iii) every disk Dr = {x ∈ Rn : ||x|| = r } of an arbitrary radius r is also home-
omorphic to Rn .
(iv) the above family of homeomorphisms forms of a family of isotopic equiva-
lences.

2. Let a connected graph be embedded in a sphere such that any face has exactly
three edges. Show that for this graph

(i) E = 3V − 6;
(ii) F = 2V − 4

where V, E, and F denotes the number of vertices, edges and faces of the graph,
respectively.
3. Let K be a simplicial complex in Rn . Show that
6.4 Exercises and Multiple Choice Exercises 445

(i)|K | is a closed bounded subset of Rn ;

(ii)|K | is a closed compact space;
(iii)every point of |K | belongs to the interior of exactly one simplex of K ;
(iv) |K | can be obtained by taking the simplexes of K and giving their union the
identification topology;
(v) if |K | is connected, then it is also path connected.

4. Prove the following statements:

(i) If G is a connected graph with maximal tree T, then the fundamental group
π1 (G) of G is a free group having a basis [ek ] corresponding to the edges
ek of G - T.
(ii) If G is any graph, then every covering space X of G having vertices and
edges are the corresponding lifts of vertices and edges of the graph G.
(iii) If B is a path connected, locally path connected and also is a semilocally
simply connected space, then corresponding to any subgroup H of the group
G = π1 (B), there exists a covering space p : E → B and a point e0 ∈ E
such that H = p∗ π1 (E, e0 ).
5. (Simplicial approximation theorem) Let f : |K | → |L| be a continuous map
between polyhedra. Show that for sufficiently large r, there is a simplicial approx-
imation s : |K r | → |L| of f.
6. Let X = S 2 S 1 be the one-point union of the 2-sphere S 2 and the circle S 1 ,
and Y = S 2 ∪ l, where l denotes the line segment joining the north and south
poles of l. Show that their fundamental groups π(X ) and π(Y ) are isomorphic.
[ Hint: The spaces X an Y are homotopically equivalent.]
7. ( Hurewicz theorem) Let (X, x0 ) and (Y, y0 ) be two path-connected pointed
spaces and f : (X, x0 ) → (Y, y0 ) be a base point preserving continuous map
such that its induced homomorphism in the homotopy theory

f ∗ : πm (X, x0 ) → πm (Y, y0 )

is an isomorphism for every integer m ≤ n − 1 and an epimorphism for m = n.

Show that its induced homomorphism in the homology theory H with integral
coefficients also such that

f ∗ : Hm (X, x0 ) → Hm (Y, y0 )

is an isomorphism for every integer m ≤ n − 1 and an epimorphism for m = n.

[Hint: See Chaps. 2 and 3. ]
8. (Generalized Jordan curve theorem) If X ⊂ S n is homeomorphic to the
sphere S m ( m < n), show that the homology groups of the complement S n − X
with coefficient group Z are
446 6 Geometric Topology and Further Applications …
⎧
⎪
⎨Z ⊕ Z, if m = n − 1 or i = 0
Hi (S − X ; Z) = Z,
n
if m < n − 1 and i = 0 or i = n − m − 1
⎪
⎩
0, otherwise.

9. Let S 3 denote the topological group of quaternions of unit modulus. Consider

the 2-sphere S 2 = {(a, b, c, d) ∈ S 3 : a = 0} as a subspace of S 3 . Show that
(i) for every y ∈ S 3 , the map

ψ y : S 3 → S 3 , x → yx y −1

is a continuous map such that it maps S 2 into S 2 ;

(ii) there is a map f : S 3 → S O(3, R) which induces a homeomorphism

f ∗ : R P 3 → S O(3, R);

(iii) the fundamental group π1 (S O(n, R)) ∼

= Z2 ∀ n ≥ 3;
(iv) the two-dimensional homotopy group π1 (S O(n, R)) = {0}, ∀ n ≥ 3.
10. Prove the following statements:
(i) If X is a path connected, commutative, associative H -space X with a
strict identity element, then it is in a weak homotopy type of a product
of Eilenberg–MacLane spaces.
(ii) If C0 is the category of pointed topological spaces having homotopy type of
C W complexes, then functor S P ∞ defines Eilenberg–MacLane spaces on
C0 .
[Hint: Use Exercise 10(i) because if X is a CW-complex,
 then S P ∞ (X ) is
path connected and has the weak homotopy type of n K (Hn (X ), n).]
(iii) The Eilenberg–MacLane K (Zm , 1) is an infinite dimensional lens space
l ∞ (m) = S ∞ mod Zm , that cannot be replaced by any finite dimensional
C W -complex.
(iv) If X = ∅ be a closed connected subspace of S 3 , then the complement S 3 − X
is an Eilenberg–MacLane space K (Z, 1).
(v) If X is a torus knot, then S 3 − X is an Eilenberg–MacLane space K (Z, 1).

6.4.2 Multiple Choice Exercises

Identify the correct alternative(s) (there may be more than one) from the following
list of exercises:

1. Consider the following topological spaces.

(i) The 2-sphere S 2 has the fixed-point property.
6.4 Exercises and Multiple Choice Exercises 447

(ii) The one-point union S 1 ∨ S 1 of two circles has the fixed-point property.
(iii) The torus S 1 × S 1 has the fixed-point property.
2. (i) The 3-sphere S 3 has no fixed-point property.
(ii) There is a continuous unit tangent vector field over 3-sphere S 3 .
(iii) Every rotation of 4-sphere S 4 has a fixed point.
3. (i) For every point x of the 5-dimensional complex projective space CP 5 , its
inverse image p −1 of the identification map

p : S 11 → CP 5 , z → [z]

is a great circle of the 11-dimensional sphere S 11 .

(ii) Let H be the division ring of quaternions. Then every polynomial over H
has a root in H.
(iii) The lens space L(2, 1) is homeomorphic to the real projective space R3 .
4. (i) The antipodal map f : S 11 → S 11 , x → −x is of degree −1.
(ii) The antipodal map f : S 16 → S 16 , x → −x is of degree +1.
(iii) If a continuous map f : S 15 → S 15 has continuous extension over the 16-
dimensional Euclidean space R16 , where S 15 is the boundary of R16 , then
f is of degree 0.
5. (i) The action of O(n, R) on the n-dimensional Euclidean plane Rn

ψ : O(n, R) × Rn , (A, x) → Ax

is not transitive.
(ii) This action ψ is transitive.
(iii) In particular, this action on S n−1 is transitive having orbit space of the point
x is the spheres of radius ||x||.
6. Let X (x, y) = −y ∂/∂x + x ∂/∂ y be a vector field in the Euclidean plane R2 .
Then, the action

ψ : R × R2 → R2 , (t, (x, y)) → (x cos t − y sin t, x sin t + y cos t)

is a flow generated by the vector field X satisfying the following properties:

(i) the flow through (x, y) is the circle having the center at the origin;
(ii) if ψt ≡ ψ(t, −) then ψt = ψ2nπ+t
(iii) one-parameter group {ψt = ψ(t, −)} is isomorphic to S O(2, R) or to the
circle group S 1 ∼
= U (1, R).
448 6 Geometric Topology and Further Applications …

References

Adams JF. On the nonexistence of elements of Hopf invariant one. Bull Amer Math Soc.
1958;64:279–82.
Adams JF. On the nonexistence of elements of Hopf invariant one. Ann Math. 1960;72:20–104.
Adams JF. Algebraic topology: A student’s guide. Cambridge: Cambridge University Press; 1972.
Adhikari A, Adhikari MR. Basic topology, vol. 1: metric spaces and general topology. India:
Springer; 2022a.
Adhikari A, Adhikari MR, Basic topology, vol. 2: topological groups, topology of manifolds and
lie groups. India: Springer; 2022b.
Adhikari MR. Basic Algebraic topology and its applications. India: Springer; 2016.
Adhikari MR, Adhikari. Basic modern algebra with applications. Springer, New Delhi, New York,
Heidelberg; 2014.
Aguilar, Gitler, S, Prieto, C. Algebraic topology from a homotopical view point. Springer, New
York, 2002.
Arkowitz M. Introduction to homotopy theory. New York: Springer; 2011.
Armstrong A. Basic topology. New York: Springer; 1983.
Basak S. A study of some aspects of knots and links, Ph.D Thesis, University of Calcutta, India;
2017.
Bredon G. Topology and Geometry, Springer, GTM 139; 1993.
Dold A, Thom R. Quasifaserungen und unendliche symmetrische Produkte. Ann Math.
1958;67(2):239–81.
Steenrod N, The topology of fibre bundles. Princeton: Princeton University Press; 1951.
Whitehead JHC. Combinatorial homotopy I. Bull. Am. Math. Soc. 1949;55(5):213–245.
Chapter 7
Brief History of Algebraic Topology:
Motivation of the Subject and Historical
Development

This chapter conveys the history of emergence of the concepts leading to the devel-
opment of algebraic topology as a subject with their motivations. Just after the concept
of homeomorphisms is clearly defined, the subject of topology begins to study those
properties of geometric figures which are preserved by homeomorphisms with an
eye to classify topological spaces up to homeomorphism, which stands the ultimate
problem in topology, where a geometric figure is considered to be a point set in the
Euclidean space Rn . But this undertaking becomes hopeless, when there exists no
homeomorphism between the two given topological spaces.

(i) The concept of topological property such as compactness and connectedness

introduced in general topology solves this problem in a very few cases which is
studied in Basic Topology Volume 1. A study of the subspaces of the Euclidean
plane R2 gives an obvious example.
(ii) On the other hand, the subject algebraic topology was born to solve the prob-
lems of impossibility in many cases with a shift of the problem by associat-
ing invariant objects in the sense that homeomorphic spaces have the same
object (up to equivalence). Initially, these objects were integers, and subsequent
research reveals that more fruitful and interesting results can be obtained from
the algebraic invariant structures such as groups and rings. For example, homol-
ogy and homotopy groups are very important algebraic invariants (they are
also called topological invariants) which provide strong tools to study the
structure of topological spaces. These algebraic objects are assigned to topolog-
ical spaces in such a way that natural operations on the latter correspond to
natural operations on the former in the sense continuous maps correspond to
group homomorphisms and homeomorphisms to isomorphisms, etc., (its con-
verse is not necessarily true). This approach of assignment in the language of
category theory is called functorial. In this way, it is often possible to distinguish
between different topological spaces by demonstrating that certain assigned alge-

© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2022 449
M. R. Adhikari, Basic Topology 3,
https://doi.org/10.1007/978-981-16-6550-9_7
450 7 Brief History of Algebraic Topology …

braic objects are not isomorphic. Algebraic topology is now used to invade many
problems of contemporary mathematics.
(iii) The homology groups are algebraic invariants that stem from homology theory
inaugurated by Heny Poincaré (1854–1912) in 1895. Development of homol-
ogy theory discussed in Chap. 3 starts from its invention by Heny Poincaré in
1895 to the approach formulating axiomatization of homology, announced in
1952 by S. Eilenberg (1913–1998) and N. Steenrod (1910–1971), now known
as Eilenberg and Steenrod axioms. This approach simplifies the proofs of many
results by escaping avoidable difficulties to promote active learning in homology
and cohomology theories, which is the most important contribution to algebraic
topology after the invention of homotopy and homology by Poincare’ in 1895.
This functorial approach facilitates in variety of cases to solve topological prob-
lems through the sovability of corresponding algebraic problems. The motivation
of the study of algebraic topology comes from the study of geometric properties
of topological spaces from the algebraic viewpoint.
(iv) The homotopy groups are also important algebraic invariants studied in Chap.
2 and they stem from homotopy theory. The early development of homotopy
theory was found through the work H. Poincaré, L. E. J. Brouwer (1881–1966),
H. Hopf (1894–1971), W. Hurewicz (1904–1956), H. Freudenthal (1905–1990)
and some others.
(v) The concepts born in the development of homology and homotopy theories to
solve topological problems have found outstanding applications to other areas
of mathematics leading to the starting points of many theories such as category
theory, homological algebra, K-theory, to mention a few (see Adhikari, 2016).
This is a remarkable feature in the history of topology.

The systematic study of algebraic topology as a subject began with precise for-
mulations and correct proofs at the turn of the nineteenth to twentieth century (1895–
1904) through the work of Henri Poincaré (1854–1912) in his land-marking ‘Analysis
situs,’ Paris, 1895. Unfortunately, his deep insight did not invite sufficiently attraction
of mathematicians until the 1920s, when the situation began to change with appli-
cations in many mathematical theories. For example, the importance of homotopy
invented by Poincaré was first established by H. Hopf (1895–1971) in 1835 with his
discovery of a new continuous map, now known as Hopf map. The fibration

p : S3 → S2

is also known as Hopf fibration plays a central role in both geometry and algebraic
topology. He proved a surprising result that π3 (S 2 ) = 0. Geometrically, it means that
there is a ‘three-dimensional hole’ in S 2 that cannot be filled. Many mathematicians
consider the discovery of Hopf map as the starting point of modern homotopy
theory. The development of the ideas of homotopy and fiber bundles after 1935
closely links to the study of differential topology.
The exponential growth of algebraic topology both in theory and applications has
been found in 1940. Algebraic topologists consider H. Poincaré founder and H.
7.1 Motivation of the Study of Topology 451

Hopf as co-founder of algebraic topology. As many fundamental ideas of algebraic

topology, specially in homotopy theory, were born through the work of W. Hurewicz
(1904–1956) in 1935–1936, he is also considered co-founder of algebraic topology,
at least for homotopy theory.
A characteristic of a topological space which is shared by homeomorphic spaces
is called a topological invariant in the sense that it is an invariant which is pre-
served by a homeomorphism. The main objective in algebraic topology is to create
and study topological invariants, which are algebraic in nature and they are also
algebraic invariants. Fundamental groups, higher homotopy groups, homology and
cohomology groups (ring) are central topics of study in algebraic topology. The con-
cept of topological invariant is utilized in classification of topological spaces up to
homeomorphism. On the other hand, the motivation of combinatorial topology is
to study a topological space by representing it as a union of simple pieces with a
specified arrangement, called combinatorial, such that the properties of the original
space depend on how the spitted pieces are arranged.

7.1 Motivation of the Study of Topology

This section begins with the motivation of the study of topology. Two natural ques-
tions arise:
(1) What is the subject topology?
(2) Why we study this subject?
(1) There are many different answers of (1). One may call the subject topology as
a qualitative study of geometry without reference to distance in the sense that
if one geometric object is obtained from another geometric object by a continuous
deformation, then these two geometric objects are considered to be topologically
same, called homeomorphic. So it is also called a rubber sheet geometry. Accord-
ingly, the geometric objects such as a circle, ellipse and a square are topologically
the same, though they are geometrically different.
(2) There are also many different answers of (2). The simplest answer is topology
is both highly elegant and useful which come from beauty, scope and power of the
subject. Its beauty comes from both its various interesting geometric constructions. Its
usefulness comes from the basic properties of continuous functions and geometric
objects with their applications in mathematics and also beyond mathematics. For
example, it facilitates a study of practically all branches of mathematics, including
algebra, real analysis, complex analysis, functional analysis, graph theory, number
theory, dynamical systems, and differential equations and many more.
What is the main problem of study in topology? The main problem in topology
is the classification problem of topological spaces up to homeomorphism. To solve
this classification problem, given two topological spaces, either we have to find an
452 7 Brief History of Algebraic Topology …

explicit expression of a homeomorphism between these two spaces or we have to

show that it is not possible to construct such a homeomorphism. Algebraic and
differential topology were born to prove this impossibility. The usual technique
is to assign ‘invariant’ objects which are shared by homeomorphic spaces (i.e., same
for homeomorphic spaces). The earliest invariant objects were Euler characteristics
which are integers. Subsequently, integral invariants are generalized by inventing
algebraic invariants such as groups, rings and modules which offer more information
about the structure of the concerned topological spaces. For example, fundamental
group, homotopy and homology groups provide deep insight into the structure
of the topological spaces. Homology underwent developed first since its invention
by H. Poincaré in 1895; on the other hand, homotopy did not develop until 1930.
Since then, there has been an explosive development of homotopy theory and its
connection with homology theory has become a central theme of topology. Many
concepts initially introduced in homotopy and homology theories such as K -theory,
Brouwer fixed-point theorem and so on have found surprising applications to other
areas of mathematics and also beyond mathematics.

7.2 Analysis Situs of Henri Poincaré

Topology is now one of the most exciting and powerful fields of research in modern
mathematics because of its importance. Its origins may be traced back several hundred
years. Historically, topology emerged as a distinct field of mathematics with the
publication famous memoirs Analysis Situs and its Five Supplements of Henri
Poincaré (1854–1912 ) from 1895 onward (1895–1904). Historically, before 1895,
several topological ideas were found in mathematics during the previous century
and a half. Mathematical terminology changes with time. For example, the Latin
terminology Analysis Situs is essentially due to G. Leibniz (1646–1716) based on
the concept limit and continuity and the German terminology Topologie is due to
J. B. Listing (1802–1882) in 1847. The Latin terminology Analysis Situs means
‘analysis of position.’ Poincaré preferred the terminology Analysis Situs of his above
memoirs and O. Veblen (1880–1960) used this terminology in his famous Colloquium
volume of 1922. In English language, the term topology was first used in 1883 in
mathematical sense. But confusion arose because of using the same term in botanical
sense also since 1659. However, the other alternative older terms of topology were
superseded by the term topology, perhaps, because of simplicity of its derived terms
such as topological and topologist, etc.
A new geometry was born through Analysis Situs of Poincaré. He remarked in
1912.
“Geometers usually distinguish two kinds of geometry, the first of which they
qualify as metric and the second as projective. . . .. But it is a third. . . .; this is
analysis situs. In this discipline, two figures are equivalent whenever one can
pass from one to the other by a continuous deformation; whatever else the law
of this deformation may be, it must be continuous. Thus a circle is equivalent to
7.2 Analysis Situs of Henri Poincaré 453

an ellipse or even to an arbitrary closed curve, but it not equivalent to a straight

line segment since this segment is not closed. A sphere is equivalent to a convex
surface; it is not equivalent to a torus since there is a hole in a torus and in a
sphere there is not.”

7.2.1 Brief History of Algebraic Topology

This section highlights the emergence of the ideas leading to development of alge-
braic topology and communicates the contributions of some mathematicians who
inaugurated new concepts and new theories or proved basic results of fundamental
importance in algebraic topology starting from the creation of fundamental group
and homology group by H. Poincaré in 1895, which are the first fundamental and
powerful inventions in algebraic topology. This section also communicates the moti-
vation of the study of algebraic topology with historical development of the subject
starting from invention of Euler characteristic in 1752, which is a numerical topo-
logical invariant followed by invention of other topological invariants which are
algebraically groups such as fundamental groups (may be nonabelian) and higher
homotopy groups (abelian) associated with pointed topological spaces together with
different types of homology and cohomology groups (always abelian) for arbitrary
topological spaces.

7.2.2 Motivation of Study of Algebraic Topology

The main aim of algebraic topology is to devise methods to construct topological

invariants. This subsection conveys the concepts of several topological invariants with
their invention. The main areas of algebraic topology include a study of homotopy
groups, homology groups and cohomology groups (rings). The first homotopy group,
known as the fundamental group, provides information about loops in a topological
spare which facilitates intuitively to know the basic shape, or holes, of a topological
space. The concepts of higher homotopy groups are the generalization of the funda-
mental group which provide a sequence of topological invariants. Homology groups
assigned to a topological space provide a sequence of abelian groups defined on a
chain complex. Cohomology groups are dual concepts of topological groups defined
on co-chain complex associated with a topological space also provide a sequence of
abelian groups which are topological invariants. The role of topological invariants is
to reformulate statements about topological spaces and continuous maps into state-
ments about groups and homomorphism to have a better chance for solution. This
transformation of topological problems into algebraic ones is done through homo-
topy, homology and cohomology theories by using topological invariants such as
fundamental groups, higher homotopy groups, homology and cohomology groups.
It is easier to define homotopy groups than homology groups but it is difficult to com-
454 7 Brief History of Algebraic Topology …

pute homotopy groups, ever for simple spaces. Moreover, the fundamental group of
many topological spaces are nonabelian. An older name for the algebraic topology
was combinatorial topology in the sense that the investigating topological spaces
were constructed from simpler topological spaces by some technique. Topologists
were successful in investigating during 1920s and 1939s to convert topological prob-
lems to algebraic problems , which led to rename algebraic topology of combinatorial
topology.

7.3 Development of the Basic Topics in Algebraic Topology

The following subsections convey the motivation and historical development of the
basic topics discussed in this book titled Basic Topology: Volume 3, starting from
Euler’s polyhedral formula.

7.3.1 Euler’s Polyhedral Formula

Euler’s land-marking polyhedral formula

V−E+F=2

in combinatorial qualities of polyhedra is considered the first important topological

invariant. The Euler characteristic invented by L. Euler (1703–1783) in 1752 is
an integral invariant, which distinguishes nonhomeomorphic spaces. For topological
and homotopical invariance of Euler characteristics, see Sect. 7.6.4. The search of
other invariants has established connections between topology and modern algebra
in such a way that homeomorphic spaces have isomorphic algebraic structures.

Remark 7.3.1 Theorem 7.3.5 saying that there are only five different types of pla-
tonic solids is a classical result proved in Chap. 2 as an interesting application of the
Euler characteristic in the theory of convex polyhedra. Its proof is based on consid-
ering the surface of a convex polyhedron as glued together a finite number of convex
polygons with respect to identity map on edges glued.

Definition 7.3.2 For a given polyhedron P, if n edges meet at each vertex and
each face is a convex m-gon, then the polyhedron P is said to be of type [m, n]. In
particular, P is said to regular if every m-gon is regular.

Remark 7.3.3 If the type [m, n] of a polygon P is known, then the number V of
the vertices, the numbers E of the edges and the number F of the faces of P can be
calculated.
 7.3 Development of the Basic Topics in Algebraic Topology 455

Definition 7.3.4 A platonic solid is a polyhedron such that its faces are congruent
regular polygons and each vertex lies in the same number of edges, it is also called
a regular simple polyhedron.

Theorem 7.3.5 There are only five platonic solids which are precisely of types:

[3, 3], [4, 3], [3, 4], [5, 3], and [3, 5].

Remark 7.3.6 The platonic solids

(i)[3, 3] represents geometrically tetrahedron , where V = 4, E = 6, F = 4;
(ii)[4, 3] represents geometrically cube , where V = 8, E = 12, F = 6;
(iii)[3, 4] represents geometrically octahedron, where V = 6, E = 12, F = 8;
(iv) [5, 3] represents geometrically dodecahedron, where V = 12, E = 30, F =
20; and
(v) [3, 5] represents geometrically icosahedron, where V = 20, E = 30, F = 12.
These are the only five different types of platonic solids.

7.3.2 Beginning of Algebraic Topology

This subsection highlights the emergences of the ideas leading to algebraic topology
and communicates the contributions of some mathematicians who inaugurated new
concepts and new theories or proved basic results of fundamental importance in
algebraic topology starting from the creation of fundamental group and homology
group by H. Poincaré in 1895, which are the first fundamental and powerful inventions
in algebraic topology. Actually, Algebraic Topology was born as a subject through
the work of H. Poincaré based on the idea of dividing a topological space into
geometric elements corresponding to the vertices, edges, and faces of polyhedra, and
their higher-dimensional analogues. Such investigation presents many topological
invariants including the Euler characteristic.
Historically, fundamental group and homology groups are the first important
topological invariants of homotopy and homology theories which came from such
a search embodded in the work of H. Poincaré (1854–1912) in his land-marking
‘Analysis situs,’ Paris, 1895. He invented homology theory, now called, simplicial
homology in 1895 with an aim to study geometric properties of a topological space
by converting topological problems to algebraic ones for the first time in the history of
topology. The term ‘Homotopy’ was first used by M. Dehn and P. Heegaard in 1907.
L. E. J. Brouwer (1881–1967) gave the precise definition of continuous deformation
by using the concept of homotopy of continuous maps. The Jordan Curve Theorem
stated by Jordan in 1892 is a classical theorem. Its first rigorous proof given by Oswald
Veblen (1880–1960) in 1905 is one of the remarkable development of algebraic
topology. W. Hurewicz made significant contributions to algebraic topology. The
invention of the higher homotopy groups πn by W. Hurewicz in 1935–1936 is a
456 7 Brief History of Algebraic Topology …

natural generalization of the fundamental group to higher-dimensional analogue of

the fundamental group. More precisely, πn is a sequence of covariant functors defined
by Hurewicz from topology to algebra by extending the concept of fundamental
group formulated by πn (X ) = [S n , X ]. Lens spaces defined by H. Tietze (1880–
1964) in 1908 form an important class of 3-manifolds in the study of their homotopy
classification.

7.4 Historical Note on Homology Theory

Before formal of invention homology by Henri Poincaré in 1895, this concept was
found in the work of B. Riemann (1826–1866) during 1850–1860 on the notion of the
connectivity order of geometrical objects (now called Riemann surfaces). A surface
X is called (r + 1)—connected if there is a family of pairwise (r + 1) closed paths,
known as cobounding curves, on X such that all of them taken together bound a region
in X but no proper subfamily of this family can do it. Such family of cobounding
curves is called maximal. For example the connectivity order of the sphere is 1. The
connectivity order is independent of the choice of the maximal cobounding curves,
which is proved by Riemann. For example, the connectivity order of the torus T is
2, because a single circle on its surface can not separate it but any pair of disjoint
closed curves can do it. E. Betti (1823–1892) developed the concept of connectivity
in 1870s for submanifolds in the Euclidean spaces Rn . The concept of connectivity
order based on the cofounding curves on a region of a manifold leads to the concept of
homology theory invented by H Poincaré (1854–1912) in 1895. He assigned to the n-
cells of an oriented simplicial complex K a matrix M(n, Z), called the n-th incidence
matrix having entries 0,-1 or + 1 depending on the nature of the orientations on its
n − 1-cells induced by the given n-cells. He reduced the incidence matrix to normal
form to obtain the Riemann’s connectivity order, called Betti number of the given
oriented surface. He also utilized tosion coefficients the nature of the closed curves
in nonorientable surfaces not bounding a region but their multiples can bound it.
To define homology groups, Poincaré started with a geometric object (a topological
space) which is given by combinatorial data (a simplicial complex) and then he
constructed homology groups by using the linear algebra and boundary relations by
these data. His homology theory is called the classical simplicial homology theory.
It involves of tedious discussion on the concepts of triangulablity of the topolog-
ical spaces, orientations of simplexes, incidence numbers, subdivisions, simplicial
approximation and also the topological invariance of the simplicial homology groups.
Emmy Noether (1882–1935) formulated an algebraic approach corresponding to
the geometric approach of homology theory invented by Poincaré.
7.5 Historical Note on Homotopy Theory 457

7.5 Historical Note on Homotopy Theory

The early development of homotopy theory was found through the work H. Poincaré,
L. E. J. Brouwer (1881–1967), H. Hopf (1894–1971), W. Hurewicz (1904–1956), H.
Freudenthal (1905–1990) and some others. But historically, a systematic study of
homotopy theory in the premises of algebraic topology with precise formulations and
correct proofs began at the turn of the nineteenth to twentieth century (1885–1904)
through the work of Henri Poincare in his ‘Analysis situs,’ Paris, 1895. But his deep
insight did not attract mathematicians sufficiently until the 1920s, when the situation
began to change with applications in many mathematical theories. L. E. J. Brouwer
(1881–1967) gave the precise definition of continuous deformation by using the
concept of homotopy of continuous maps. Homotopy theory excepting fundamental
group was first applied as a secondary tool of homology theory. This section addresses
the contribution of Brouwer, Hopf, Hurewicz and Freudenthal toward development
of homotopy theory since its inauguration in 1895 by H. Poincaré.

7.5.1 Inauguration of Homotopy Theory in 1895 by Poincaré

The intuitive idea of continuous deformation which led to the concept of homotopy
is found in Lagrange’s method in calculus of variation. This idea was also available
in the mathematical work of many many mathematicians of the nineteenth century.
But it was Brouwer who first formally defined the concepts of homotopy between
two continuous maps in 1911, though Poincaré introduced the idea of fundamental
group π(X, x0 ) of a pointed topological space in his paper of 1895 consisting of its
elements: each element is a class of a loop that can be considered as a continuous
map
α : [0, 1] → X : α(0) = α(1) = x0 .

The study of Poincaré on the fundamental group (also called Poincaré’s group) is
considered the beginning of homotopy theory. The concepts of fundamental groups
and homotopy were born through the work of Henri Poincaré in his land-marking
‘Analysis situs,’ Paris, 1895. This work officially inaugurated homotopy theory along
with homology theory. These two theories form the basic parts of algebraic topology.
For this reason, Poincaré is called the founder of algebraic topology. His study to solve
classification problems in mathematical analysis and Euclidean geometry by creating
topological invariants, which are also algeraicgical invariants. He devised ways to
assign to a topological space a group or module in such a way that homeomorphic
spaces have isomorphic algebraic objects.
458 7 Brief History of Algebraic Topology …

7.5.2 Brouwer Fixed-Point Theorem and Degree of Spherical

Map

L. E. J. Brouwer (1881–1967) gave the precise definition of continuous deformation

by using the concept of homotopy of continuous maps. He initiated significant work
in 1912 connecting homology and homotopy groups of certain spaces and proved
that two continuous maps of a two-dimensional sphere into itself can be continuously
deformed into each other if and only if they have the same degree (i.e., if they are
equivalent from the view point of homology theory). His definition of the degree of
a spherical map is more intuitive than its definition from the viewpoint of homology
theory. He defined deg f , the degree of f as the number of times of the domain sphere
wraps around the range sphere and proves its homotopy invariance. He showed that
for self maps of S n , the homotopy class of a continuous map is characterized by its
degree. His definition shows that if f : S 1 → S 1 , z → z n , then deg f = n. The most
basic results of Brouwer proved in this volume are

(i) Brouwer no retraction theorem saying that there exists no continuous onto
map f : Dn → S n−1 which leaves every point of S n−1 fixed for every integer
n ≥ 1.
(ii) Brouwer degree theorem asserting that if f, g : S n → S n are two continuous
maps such that f  g, then deg f = deg g.
(iii) the topological invariance of dimension of Rn saying that if m and n are
two distinct positive integers, then the Euclidean m-space Rm and the Euclidean
n-space Rn cannot be homeomorphic.
(iv) Brouwer fixed-point theorem asserts that every continuous map f : Dn → Dn
has a fixed point for every integer n ≥ 0.

These results of Brouwer has made a key role in laying the foundations of algebraic
topology.
Brouwer fixed-point theorem (named after his name) for Dn in the n-dimensional
Euclidean space Rn was first proved and studied by L. E. J. Brouwer during 1910–
2012. It is a celebrated theorem in topology. This theorem is also proved by using the
homology or homotopy groups. But Brouwer used neither of them. Instead, he used
the notion of degree of spherical maps f : S n → S n . An important application of
Brouwer fixed-point theorem for Dn to algebra is the Perron–Frobenius theorem in
Rn . It asserts that any square matrix with positive entries has a unique eigenvector with
positive entries (up to a multiplication by a positive constant), and the corresponding
eigenvalue has multiplicity one and is strictly greater than the absolute value of any
other eigenvalue.
Historically, Brouwer first established a close link between homotopy and homology
by showing in 1912 that two continuous mappings of a two-dimensional sphere S 2
into itself can be continuously deformed into each other if and only if they have
the same degree, the concept defined by Brouwer himself from the viewpoint of
homology, i.e., if they are equivalent from the viewpoint of homology theory invented
7.5 Historical Note on Homotopy Theory 459

by Poincaré in 1895. The papers of H. Poincaré during 1895–1904 can be considered

as blueprints for theorems to come. The results of Brouwer during 1910–1912 may be
considered the first one of the proofs in algebraic topology. He proved the celebrated
theorem Brouwer fixed-point theorem 7.5.1 by using the concept of degree of a
continuous spherical map defined by Brouwer himself.
Theorem 7.5.1 (Brouwer fixed-point theorem for dimension n) Every continuous
map f : Dn+1 → Dn+1 has a fixed point for every finite n ≥ 0.

7.5.3 Brouwer–Poincaré Theorem

Brouwer-Poincaré Theorem 7.5.4 asserts that there is a continuous nonvanishing

vector field f : S n → S n (n ≥ 1), iff n is odd. On the other hand, Corollary 7.5.5
shows that for all even integers n ≥ 1, there is no vector field f : S n → S n .

Definition 7.5.2 A vector field on S n is a continuous map v : S n → Rn+1 (n ≥

1) such that the inner product < x, v(x) >= 0, ∀ x ∈ S n , i.e., the vector v(x) is
orthogonal to the vector x in Rn+1 for every x ∈ S n . Moreover, if v(x) = 0 for all
x ∈ S n , we say that the vector field is nonvanishing.

Remark 7.5.3 (Geometrical Interpretation) Definition 7.5.2 implies that a vector

field v on S n is a continuous map which assigns to every vector x of unit length
in Rn+1 , a unit vector v(x) in Rn+1 such that x and v(x) are orthogonal, i.e., x ⊥
v(x), ∀ x ∈ S n . If we consider the vector v(x) starting from the point x ∈ S n , then
this vector v(x) must be tangent to S n at each point x of S n . If x moves in S n , then
endpoint point of the vector v(x) varies continuously in Rn+1 .

Theorem 7.5.4 (Brouwer–Poincaré) The n-sphere S n admits a continuous nonva-

nishing vector field iff n is odd.

Corollary 7.5.5 The n-sphere S n admits no continuous nonvanishing vector field if

n is even.

Remark 7.5.6 If n is odd, the difficult problem of determining the maximum number
of linearly independent nowhere vanishing vector fields on S n was solved by J.F
Adams in 1962 by using K -theory Adams (1962).

7.5.4 Freudenthal Suspension Theorem

Freudenthal Suspension Theorem 7.5.8 proved in 1938 by H. Freudenthal (1905–

1990) is a basic theorem in algebraic topology. For each pair of positive inte-
gers, m and n, there is a natural homomorphism σm : πm (S n ) → πm+1 (S n+1 ). This
460 7 Brief History of Algebraic Topology …

homomorphism is called the Freudenthal suspension homomorphism defined by H.

Freudenthal in 1937. He observed that the suspension operation on topological spaces
shifts by one their low-dimensional homotopy groups. This observation was impor-
tant in understanding the special behavior of homotopy groups of spheres, because
every sphere can be formed topologically as a suspension of a lower-dimensional
sphere and this subsequently forms the basis of stable homotopy theory. So, it said
that the founding result of stable homotopy theory is the Freudenthal suspension
theorem proving that homotopy groups become stable in the sense of isomorphic
after performing sufficiently many iterated suspensions. This result leads to the con-
cept of stable homotopy groups of a topological space. Its direct consequences are
enormous. For example, Corollary 7.5.9 proved by Hurewicz says that the homotopy
groups πm (S n ) = 0 for 0 < m < n.
Definition 7.5.7 (Suspension homomorphism) The natural homomorphism

σ : πn (S n ) → πn+1 (S n+1 ), [α] → [α̃]

is a homomorphism, called the suspension homomorphism.

H. Freudenthal proved the following suspension theorem in 1938. In his honor, this
suspension theorem is known as the Freudenthal suspension theorem.

Theorem 7.5.8 (The Freudenthal suspension theorem) The suspension homorphism

σ : πm (S n ) → πm+1 (S n+1 )

is an isomorphism for m < 2n − 1 and is onto for m ≤ 2n − 1.

An immediate consequence of the Theorem 7.5.8 is the following corollaries:

Corollary 7.5.9 (Hurewicz) The homotopy groups πm (S n ) = 0 for 0 < m < n.

Corollary 7.5.10 (Hopf) For every integer n ≥ 1, πn (S n ) ∼

= Z. (This result is known
as Hopf degree theorem).

7.5.5 Stable Homotopy Groups

This section studies stable homotopy groups based on the Freudenthal suspension
theorem saying that suspension homomorphism

E : πm (S n ) → πm+1 (S n+1 )

is an isomorphism for m < 2n − 1 and is onto for m ≤ 2n − 1. Freudenthal (1938)

has been discussed in Chap. 2. It is a natural generalization of Freudenthal suspen-
sion theorem. In algebraic topology we use the word ‘stable’ when a phenomenon
7.5 Historical Note on Homotopy Theory 461

occurs essentially in the same way independent of dimension provided perhaps that
the dimension is sufficiently large. The importance of stable homotopy theory was
reinforced by two related developments in the late 1950s. One is the introducing of
spectral homology and cohomology theory and specially K -theory by Atiyah and
Hirzebruch. The other one is the work of Thom which reduces the problem of classi-
fying manifolds up to cobordism to a problem, a solvable problem in stable homotopy
theory. Moreover, this section studies homotopy groups of a spectrum. Historically,
stable phenomena found implicitly before 1937 in the study of reduced homology
and cohomology functors that are invariant under suspension without limitations
on dimension. Stable homotopy theory appeared as an important topic of algebraic
homotopy with Adam’s introduction of his spectral and conceptual use of the concept
of stable phenomena in his solution to the Hopf invariant problems. Every homology
and cohomology theory can be constructed by using homotopy-theoretical techniques
only. This establishes a close relation of homology and cohomology theories with
stable objects in homotopy theory.
Construction of the stable homotopy group: Consider an n-connected C W -
complex and the suspension map

 : πr (X ) → πr +1 (X )

is an isomorphism for r < 2n + 1. In particular, for r ≤ n the suspension map  is

an isomorphism and  X is an (n + 1)-connected CW-complex. Now, consider the
sequence (7.1 ) of groups and homomorphisms

πr (X ) −
→ πr +1 ( X ) −
→ ··· −
→ πr +m ( m X ) −
→ ··· (7.1)

The homomorphism
πr +m ( m X ) → πr +m+1 ( m+1 X )

is an isomorphism for r + m < 2(n + m) + 1, i.e., for m > r − 2n + 1, because,

 m X is (n + m)-connected. This implies that for fixed n and r and sufficiently large
enough m, all the groups in the sequence of groups and homomorphisms (7.1) are
isomorphic. This leads to the concept of the stable homotopy group for which the
sequence (7.1 ) is stabilized.

Definition 7.5.11 (Stable homotopy group) Given an (n − 1) -connected space X,

for r ≥ 0, the r -th stable homotopy group of X, denoted by πr (X ) is defined to be
the group
πn+m ( X ), ∀ m > r − 2n + 1.

This is well defined, since, adding any finite number of groups and homomor-
phisms to the beginning of (7.1) does not affect the resulting stable homotopy group.

Example 7.5.12 Consider the homotopy exact sequence (7.2) of the fibration: p :
S3 → S2 :
462 7 Brief History of Algebraic Topology …

p∗
· · · → π3 (S 1 , s0 ) −
→ π3 (S 3 , s0 ) −−−−→ π3 (S 2 , s0 ) −
→ π2 (S 1 , s0 ) → · · · (7.2)

The exactness property of the sequence (7.2) of groups and hommorphisms asserts
that the homomorphism p∗ induced by p

p∗ : π3 (S 3 , s0 ) → π3 (S 2 , s0 )

∼ π2 (S 1 , s0 ) = 0. This implies that π3 (S 2 , s0 )

is an isomorphism, because π3 (S 1 , s0 ) =
∼
= Z. It provided the first example given by H.Hopf, where πm (S n , s0 ) = 0 for m > n.
The map p is known the Hopf map, studied earlier in Chap. 2.

Example 7.5.13 For each q, consider

   
→∼
π2q+2 (S q+2 , s0 ) − →∼
=π2q+3 (S q+3 , s0 ) − →∼
=··· − →∼
=πq+n (S n , s0 ) − =···

We denote the common group πn+q (S n , s0 ), by πqS . It is called the kth stable homo-
topy group for the fibration : p : S 3 → S 2 :.

7.6 The Topology of Fiber Bundles

(i) The theory of fiber bundles, in particular, vector bundles, establishes a very
strong link between algebraic topology and differential topology. The topol-
ogy of fiber bundles has created general interest and promises for more work,
because it is involved of interesting applications of topology to other areas
such as algebraic topology, geometry, physics and gauge groups.
(ii) Whitney gave the first general definition of fibration in 1935 and his own
idea on fibration was subsequently developed by him. He studied a special
type of fibration, now, called locally trivial fiber space, which is a quadruple
ξ = (X, p, B, F) consisting of the total space X, base space B, fiber space
F and the projection map p : X → B.
(iii) H. Poincaré considered in 1883 a special type of fiber spaces, which were
the covering spaces over the open subsets of C with discrete fibers. Their
generalizations are found during the period 1913-1934.
(iv) Mathematicians realized around 1940 that the covering homotopy property
(CHP) of a fiber space ξ = (X, p, B, F) asserts that the projection

p : X → B : p(x0 ) = b0

which is a continuous surjective map satisfying the condition that for each
point b ∈ B, there exist nbd U and a homeomorphism

ψ : U × F → p −1 U
7.6 The Topology of Fiber Bundles 463

such that
p(ψ(x, t)) = x, ∀ x ∈ U, t ∈ F

which induces an isomorphism

p∗ : πn (X, F, x0 ) → πn (B, b0 ), ∀ n ≥ 1.

This gives rise beginningless an exact homotopy sequence (7.3)

j∗ p∗ ∂ j∗
· · · → πn+1 (F, x0 ) −−−−→ πn+1 (X, x0 ) −−−−→ πn+1 (B, b0 ) −−−→ πn (F, x0 ) −−−−→ πn (X, x0 )
∂ j∗ p∗ ∂
−−−→ · · · π1 (F, x0 ) −−−−→ π0 (X, x0 ) −−−−→ π0 (B, b0 ) −−−→ 0
(7.3)

which establishes a close link of homotopy groups of the three spaces such
as X, B and F of ξ = (X, p, B, F).
(v) A real (resp. complex) vector field is a fiber space ξ = (X, p, B, F) with F
is a vector space Rn (resp. Cn ).
(vi) The theory of fiber bundles was first recognized during the period 1935–1940
through the work of H. Whitney (1907–1989 ), H. Hopf (1894–1971) and E.
Stiefel (1909–1978 ), J. Feldbau (1914–1945) and some others.
(vii) A fiber bundle is a bundle with an additional structure derived from the
action of a topological group on the fibers. A fiber bundle is a locally trivial
fibration having covering homotopy property. Historically, the first formal
definition of fiber bundles was given by Whitney. Chapter 4 of this book
studies general theory of bundles and its Chap. 5 is developed based on
their homotopy properties. The study topology of fiber bundles has created
general interest as it is involved of interesting applications of topology to
other areas such as algebraic topology, geometry, physics and gauge groups
and addresses the homotopy theory of bundles. Covering spaces provide
tools to study the fundamental groups. Fiber bundles provide likewise tools
to study higher homotopy groups (which are generalizations of fundamental
groups).
(viii) The notion of fiber spaces is the most fruitful generalization of covering
spaces. The importance of fiber spaces was realized during 1935–1950 to
solve several problems relating to homotopy and homology. The motivation
of the study of fiber bundles and vector bundles came from the distribution
of signs of the derivatives of the plane curves at each point.

7.6.1 Historical Note on Category Theory

Category theory plays an important role in the modern study of algebraic topol-
ogy, because, homotopy, homology and cohomology theories can be conveniently
expressed in the language of category theory. This theory is also very important in
464 7 Brief History of Algebraic Topology …

mathematics to unify different concepts in mathematics and specially in the study of

homotopy, homology and cohomology theories and their development in a unified
language. The basic objective of category theory is to unify many basic concepts
and results of mathematics in an accessible way. Historically, category theory was
born through the work of Eilenberg (1913–1998) and S MacLane (1909–2005) dur-
ing 1942–1945. A category is a certain collection of mathematical objects (possibly
with an additional structure) and morphisms which are like mappings agreeing with
this structure. The objects may be a set, group, ring, vector space, module, a sequence
of abelian groups, topological space, etc. and morphisms are collections of mapping
preserving this structure. On the other hand, a functor is a natural mapping from
one category to the other in the sense that it preserves the identity morphism and
composites of well-defined morphisms. It plays a key role in converting a problem of
one category to the problem of other category to have a better chance for solution. A
natural transformation is a certain function from one functor to other one satisfy-
ing some specific properties. These concepts together with their dual concepts form
the foundation of category theory, which provides a convenient language to unify
several mathematical results. This language is used throughout the present book.

7.6.2 Homology and Cohomology Theories: Eilenberg and

Steenrod Axioms

Homology and cohomology theories are basic theories in algebraic topology. While
investigating the 3-dimensional and higher dimensional manifolds in 1895, Henri
Poincaré in his ‘Analysis Situs’ formally introduced the concepts of homotopy,
fundamental group, homology groups and Betti numbers . His monumental work
embodied in his ‘Analysis Situs’ aimed at solving problems on system of differential
equations and his research establishes a surprising connection between analysis
and topology. Historically, homology invented by Henri Poincaré in 1895 was stud-
ied by him during 1895–1904. This homology, called simplicial homology is one of
the most fundamental powerful inventions in mathematics. He started with a geo-
metric object (a topological space) which is given by combinatorial data (a simplicial
complex), then the linear algebra and boundary relations by these data were used to
construct homology groups. The shifting of geometric approach of Henri Poincaré
to the algebraic approach of Emmy Nother (1882–1935) to homology theory highly
motivated P. Alexandroff (1896–1982) and H. Hopf to study homology theory jointly
from the algebraic viewpoint in 1935. The classical simplicial homology theory
invented by Poincaré is involved of tedious discussion on the concepts of trian-
gulalbity of the topological spaces, orientations of simplexes, incidence numbers,
subdivisions, simplicial approximation, and also the topological invariance of the
simplicial homology groups. Poincaré remarked in 1899 saying that
“ Assume that one can find in V a manifold of p + 1 dimension whose boundary
consists of n manifolds of p dimension v1 , v2 , . . . , vn ; I will express this fact
with the relation v1 + v2 + · · · + vn ∼ 0, that I will call it homology.”
7.6 The Topology of Fiber Bundles 465

There are other homology theories constructed by different mathematicians

such as L. Vietories (1891–2002) in 1927, by E.Čech (1893–1960) in 1932 and S.
Lefschitz (1884–1972) in 1933 and some others. For example, Vietories constructed
homology groups for compact metric spaces; E.Čech constructed homology groups
for compact Hausdorff spaces and Lefschitz constructed singular homology groups
by generalizing singular homology groups invented by Poincaré. These theories are
conveniently used depending on the nature of the topological problems. On the other
hand, the axiomatic approach formulated by seven axioms, now called Eilenberg
and Steenrod axioms for homology and cohomology theories , were announced by
S. Eilenberg (1913–1998) and N. Steenrod (1910–1971) in 1945 but first appeared
in their celebrated book ‘The Foundations of Algebraic Topology’ in 1952. This
approach classifies and unifies different homology (cohomology groups) and is the
most important contribution to algebraic topology since invention of the homology
groups by Poincaré in 1895. Eilenberg and Steenrod proved that on the category of
compact triangulable spaces all the homology and cohomology theories satisfying
their axioms have isomorphic groups. This asserts that there exists only one homology
theory and only one cohomology theory on this category.

7.6.3 Euler–Poincaré Theorem

Euler–Poincaré theorem 7.6.2 is a powerful result in topology. It establishes a close

link among geometry, topology and algebra with the help of Euler characteristics of
compact polyhedra.

Euler- Poincaré characteristic is a generalization of Euler characteristic

formula.

Definition 7.6.1 Let K be a simplicial complex of dimension n and β p denotes the

number of p simplexes in K, for p = 0, 1, 2, ..., n. Then the alternative sum

κ(K ) =  np=0 (−1) p β p (7.4)

is called the Euler characteristic of K with Betti numbers β p .

Betti numbers β p in Eq. (7.4) coined by Poincaré (named after E. Betti ), play an
important role in algebraic topology to classify topological spaces based on the
connectivity of a p-dimensional simplicial complex K. The number β p is the same
as the rank of the p-th homology group H p(K ; Q) with rational coefficients.

Theorem 7.6.2 (Euler–Poincaré theorem) The Euler characteristic of an oriented

simplicial complex K of dimension n is given by

κ(K ) =  np=0 (−1) p β p =  np=0 (−1) p rank H p (K ; Q). (7.5)

466 7 Brief History of Algebraic Topology …

Corollary 7.6.3 The Euler characteristic of an oriented simplicial complex K of

dimension n is also given by the formula

κ(K ) =  np=0 (−1) p β p =  np=0 (−1) p rank H p (K ; Z).

Corollary 7.6.4 Let X be a compact polyhedron. Then its Euler characteristic is

given by the formula

κ(X ) =  np=0 (−1) p β p =  np=0 (−1) p rank H p (X ; Z).

Corollary 7.6.5 The rank of the free abelian part of H p (K ; Q) of a finite oriented
complex K is the Betti number β p of K .

Remark 7.6.6 Since homology group is a homotopy as well as a topological invari-

ant, Euler characteristic is also so (see Sect. 7.6.4).

7.6.4 Topological and Homotopical Invariance of Euler

Characteristics

Euler–Poincaré theorem 7.6.2 is applied to prove a powerful results in topology.

For example, this theorem applies to prove topological invariance of Euler character-
istics in Theorem 7.6.7 in the sense that two homeomorphic compact polyhedra have
the same Euler characteristics and homotopical invariance of compact polyhedra in
Theorem 7.6.8 in the sense that two homotopy equivalent compact polyhedra have
the same Euler characteristics (see Chap. 3).

Theorem 7.6.7 Two homeomorphic compact polyhedra have the same characteris-
tics.

Theorem 7.6.8 Two homotopy equivalent compact polyhedra have the same Euler
characteristics.

Remark 7.6.9 Theorem 7.6.7 establishes a close link among geometry, topology
and algebra by Euler characteristics of compact polyhedra, which are integers. For a
generalization of theorem 7.6.8 asserting that if the compact polyhedra X and Y are
two homotopy equivalent spaces, then κ(X ) = κ(Y ). Moreover, κ(X ) ∈ Z, which
is an algebraic object.
7.6 The Topology of Fiber Bundles 467

7.6.5 Brief History of Combined Aspect of Combinatorial

and Set-Theoretic Topologies

A union of combinatorial and set-theoretic aspects of topology was achieved

first by L. E. J. Brouwer through his investigation of the concept of dimension
during 1908–1912. The unified theory was laid on a solid foundation in the period
1915–1930 by J. W. Alexander, P. S. Alexandrov (1896–1982), S. Lefschetz (1884–
1972) and others. Until 1930 topology was called ‘analysis situs (position analysis).’
Analysis situs conveyed the qualitative properties of geometric figures both in
the ordinary space as well as in the space of more than 3-dimensions. It was
Lefschetz who first used and popularized the name topology by publishing a book
with this title in 1930.

7.6.6 Poincaré Conjecture and Its Solutions

Poincaré stated a conjecture known as Poincaré conjecture in 1894 or 1904 by

saying that a compact smooth n-dimensional manifold, which is homotopy equivalent
to the n-sphere S n is homeomorphic to S n . Its equivalent statement says: is a compact
n-manifold homotopically equivalent to S n homeomorphic to S n ? For n = 3, G.
Perelman (1966-) proved this conjecture in 2003 by using Ricci flow. For other
values of n, it was solved by others before 1994:
(i) For n = 4, M. Freedman (1951-) proved that the conjecture is true and wins
Fields medal for this proof.
(ii) For n = 5, C. Zeeman (1925–2016 ) demonstrated the conjecture in 1961.
(iii) For n = 6, J. R. Stallings (1935–2008) proved in 1961 that the conjecture is
true.
(iv) For n ≥ 7, S. Smale (1930-) proved that the conjecture is true and also extended
his proof for all n ≥ 5. He wins the Fields medal in 1966 for this work.

7.6.7 Interest of Poincaré in Various Scientific Work

H. Poincaré, founder of algebraic topology has contributed fundamental work in

pure and applied mathematics, celestial mechanics, dynamical system, mathemat-
ical physics. He published 30 books and over 500 papers contributing significant
work in a variety of areas such as algebra, analysis, differential equation, complex
analysis, algebraic geometry, celestial mechanics, relativity theory, mathematical
physics, philosophy of mathematics and even in popular science. Like combinatorial
or algebraic topology, his original work on chaotic deterministic system has fonded
the modern chaos theory. In 1905, he first showed the requirement the gravitational
468 7 Brief History of Algebraic Topology …

waves emanating from a body and propagating at the speed of light in the Lorentz
transformations. His work on perfect invariance of all of Maxwell’s equations has
lead to the formulation of the theory of special relativity. The Poincaré group intro-
duced by Poincaré used in topology and physics are named after him.
For additional reading the readers may refer to Adams (1972), Adhikari (2016),
Adhikari and Adhikari (2003, 2006, 2014, 2022a, 2022b), Barrat (1955), Bredon
(1993), Dieudonné (2016), Eilenberg and Steenrod (1952), Freedman (1982), Gray
(1975), Hatcher and Allen (2002), Hilton and Wylie (1960), Hu (1966), Hurewicz
(1935), Milnor (1962), Poincaré (1895, 2010), Rotman (1988), Spanier (1966),
Switzer (1975).

References

Adams JF. Algebraic topology: a student’s guide. Cambridge: Cambridge University Press; 1972.
Adhikari A, Adhikari MR. Basic topology, vol. 1: metric spaces and general topology. India:
Springer; 2022a.
Adhikari A, Adhikari MR, Basic topology, vol. 2: topological groups, topology of manifolds and
lie groups. India: Springer; 2022b.
Adhikari MR, Adhikari A, Groups, rings and modules with applications. Hyderabad: Universities
Press; 2003.
Adhikari MR, Adhikari A, Textbook of linear algebra: an introduction to modern algebra. New
Delhi: Allied Publishers; 2006.
Adhikari MR, Adhikari A, Basic modern algebra with applications. New Delhi, New York: Springer;
2014.
Adhikari MR, Basic algebraic topology and its applications. India: Springer; 2016.
Barrat MG. Track groups I. Proc Lond Math Soc. 1955;5(3):71–106.
Bredon GE. Topology and geometry. New York: Springer; 1993.
Dieudonné J, A history of algebraic and differential topology, 1900–1960, Modern Birkhäuser;
1989.
Eilenberg S, Steenrod N. Foundations of algebraic topology. Princeton: Princeton University Press;
1952.
Freedman M. The topology of four-dimensional manifolds. J Diff Geom. 1982;17:357–453.
Gray B. Homotopy theory. An introduction to algebraic topology. New York: Academic Press;
1975.
Hatcher A. Algebraic topology. Cambridge: Cambridge University Press, 2002.
Hilton PJ, Wylie S. Homology theory. Cambridge: Cambridge University Press; 1960.
Hu ST. Homology theory. Oakland CA: Holden Day; 1966.
Hurewicz W. Beitrage de Topologie der Deformationen. Proc K Akad Wet Ser A 1935; 38:112–119,
521–528.
Milnor JW. On axiomatic homology theory. Pacific J Math. 1962; 12:337–341.
Poincaré H. Analysis situs. J de l’Ecole, Polyt. 1895; 1:121
Poincaré H. Papers on topology: analysis situs and its five supplements,Translated by Stillwell, J.,
History of Mathematics, 2010; 37 Amer Math Soc.
Rotman JJ. An introduction to algebraic topology. New York: Springer; 1988.
Spanier E. Algebraic topology. McGraw-Hill; 1966.
Switzer RM. Algebraic topology-homotopy and homology. Berlin, Heidelberg, New York: Springer;
1975.