Mahima Ranjan Adhikari - Basic Algebraic Topology and Its Applications-Springer (2016)

Mahima
Ranjan Adhikari
Basic Algebraic
Topology and
its Applications
Basic Algebraic Topology and its Applications
Mahima Ranjan Adhikari
Basic Algebraic Topology

and its Applications
123
Mahima Ranjan Adhikari
Institute for Mathematics, Bioinformatics,
Information Technology and Computer
Science (IMBIC)
Kolkata
India
ISBN 978-81-322-2841-7 ISBN 978-81-322-2843-1 (eBook)

DOI 10.1007/978-81-322-2843-1
Library of Congress Control Number: 2016944483
© Springer India 2016

This work is subject to copyright. All rights are reserved 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, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer (India) Pvt. Ltd.
To
my parents Naba Kumar and
Snehalata Adhikari
who created my interest in mathematics
at my very early childhood
Preface
Algebraic topology is one of the most important creations in mathematics which

uses algebraic tools to study topological spaces. The basic goal is to find algebraic
invariants that classify topological spaces up to homeomorphism (though usually
classify up to homotopy equivalence). The most important of these invariants are
homotopy groups, homology groups, and cohomology groups (rings). The main
purpose of this book is to give an accessible presentation to the readers of the basic
materials of algebraic topology through a study of homotopy, homology, and
cohomology theories. Moreover, it covers a lot of topics for advanced students who
are interested in some applications of the materials they have been taught. Several
basic concepts of algebraic topology, and many of their successful applications in
other areas of mathematics and also beyond mathematics with surprising results
have been given. The essence of this method is a transformation of the geometric
problem to an algebraic one which offers a better chance for solution by using
standard algebraic methods.
The monumental work of Poincaré in “Analysis situs”, Paris, 1895, organized
the subject for the first time. This work explained the difference between curves
deformable to one another and curves bounding a larger space. The first one led to
the concepts of homotopy and fundamental group; the second one led to the concept
of homology. Poincaré is the first mathematician who systemically attacked the
problems of assigning algebraic invariants to topological spaces. His vision of the
key role of topology in all mathematical theories began to materialize from 1920.
This subject is an interplay between topology and algebra and studies algebraic
invariants provided by homotopy, homology, and cohomology theories. The
twentieth century witnessed its greatest development.
The literature on algebraic topology is very vast. Based on the author’s teaching
experience of 50 years, academic interaction with Prof. B. Eckmann and Prof.
P.J. Hilton at E.T.H., Zurich, Switzerland, in 2003, and lectures at different insti-
tutions in India, USA, France, Switzerland, Greece, UK, Italy, Sweden, Japan, and
many other countries, this book is designed to serve as a basic text of modern
algebraic topology at the undergraduate level. A basic course in algebraic topology
vii
viii Preface
presents a variety of phenomena typical of the subject. This book conveys the basic
language of modern algebraic topology through a study of homotopy, homology,
and cohomology theories with some fruitful applications which display the great
beauty of the subject. For this study, the book displays a variety of topological
spaces: spheres, projective spaces, classical groups and their quotient spaces,
function spaces, polyhedra, topological groups, Lie groups, CW-complexes,
Eilenberg–MacLane spaces, infinite symmetric product spaces, and some other
spaces. As well as, the book studies a variety of maps, which are continuous
functions between topological spaces.
Characteristics which are shared by homeomorphic spaces are called topological
invariants; on the other hand, characteristics which are shared by homotopy
equivalent spaces are called homotopy invariants. The Euler characteristic is an
integral invariant, which distinguishes non-homeomorphic spaces. The search of
other invariants has established connections between topology and modern algebra
in such a way that homeomorphic spaces have isomorphic algebraic structures.
Historically, the concepts of fundamental groups, higher homotopy groups, and
homology and cohomology groups came from such a search. The natural emphasis
is: to solve a geometrical problem of global nature, one first reduces it to a
homotopy-theoretic problem; this is then transformed to an algebraic problem
which provides a better chance for solution. This technique has been the most
fruitful one in algebraic topology. The notions initially introduced in homology and
homotopy theories for applications to problems of topology have found fruitful
applications in other parts of mathematics. Homological algebra and K-theory are
their outstanding examples.
The materials discussed here have appeared elsewhere. Each chapter opens with
a short introduction which summarizes the information that sets out its central
theme and closes with a list of sources for the use of readers to expand their
knowledge. This does not mean that other sources are not good. Our contribution is
the selection of the materials and their presentation. Each chapter is split into
several sections (and subsections) depending on the nature of the materials which
constitute the organizational units of the text. Each chapter provides exercises with
further applications and extension of the theory. Some exercises carry hints which
should not be taken as ideal ones. Many of them can be solved in a better way. The
title of the book suggests the scope and power of algebraic topology and its text is
expanded over 18 chapters and two appendices displayed below.
Chapter 1 assembles together some basic concepts of set theory, algebra,
analysis, set topology, Euclidean spaces, manifolds with some standard notations
for smooth reading of the book.
Chapter 2 is devoted to the study of basic elementary concepts of homotopy theory
with illustrative examples. A homotopy formalizes the intuitive idea of continuous
deformation of a continuous map between two topological spaces. It displays a
variety of phenomena and related problems such as homotopy classification of
continuous maps up to homotopy equivalence introduced by Hurewicz (1904–1956)
in 1935, contractible spaces, H-groups (Hopf groups) and H-cogroups through their
homotopy properties. Finally, this chapter presents interesting immediate
Preface ix
applications of homotopy in dealing with some extension problems, lifting problems,

and proving “Fundamental theorem of algebra” by using homotopic concepts.
Chapter 3 continues the study of homotopy theory through the concept of
fundamental groups invented by H. Poincaré in 1895 which conveys the first
transition from topology to algebra by assigning an algebraic structure to the set of
relative homotopy classes of loops in a functorial way. The group structure of these
homotopy classes of loops is proved in Sects. 3.1 and 3.2 in two different ways.
This group earlier called Poincaré group is now known as fundamental group. It is
the first influential invariant of homotopy theory and is also the first of a series of
algebraic invariants πn , called homotopy groups studied in Chap. 7. This chapter
computes fundamental group of the circle by using the degree of a continuous map
f : S1 ! S1 , and studies Brouwer fixed point theorem for dimension 2, fundamental
theorem of algebra, vector field problems on D2 and knot groups by using the tools
of fundamental groups.
Chapter 4 continues the study of the fundamental groups and presents a thorough
discussion of covering spaces which are deeply connected with fundamental groups.
Algebraic features of the fundamental groups are expressed by the geometric
language of covering spaces. This chapter is designed to utilize the power of the
fundamental groups and also to establish an exact correspondence between the
various connected covering spaces of a given topological space B and subgroups of
its fundamental group π1 ðBÞ, like Galois theory, with its correspondence between
field extensions and subgroups of Galois groups, which is an amazing result. This
chapter also studies the concepts of fibrations and cofibrations with their applications
born in geometry and topology.
Chapter 5 continues the study of homotopy theory through fiber bundles, vector
bundles, and K-theory. Covering spaces provide tools to study the fundamental
groups. Fiber bundles provide likewise tools to study higher homotopy groups
(which are generalization of fundamental groups and described in Chap. 7). 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. This chapter also discusses homotopy classification of vector
bundles, Milnor’s construction of a universal fiber bundle for a topological group G
with homotopy classification of principal G-bundles and presents the introductory
concept of K-theory born in connecting the rich structure of vector bundles over a
paracompact space B with the set of homotopy classes of maps from B into the
Grassmann manifold of n-dimensional subspaces in infinite-dimensional space. This
theory plays a vital role in applications of algebraic topology to analysis, algebraic
geometry, topology, ring theory, and number theory.
Chapter 6 builds up interesting topological spaces called polyhedra from
simplexes followed by a study of their homotopy properties and develops some tools
for computing the fundamental groups of a large class of topological spaces. The
geometrical objects such as points, edges, triangles, and tetrahedra are examples of
low-dimensional simplexes. Simplicial complexes provide a convenient way to
x Preface
study manifolds. This chapter considers how simplexes may be fitted together to
construct simplicial complexes which play an important role to construct interesting
topological spaces such as polyhedra for the study of algebraic topology. They form
building blocks of homology theory which begins in Chap. 10. The concept of
triangulation is utilized to solve extension problems and that of edge-path to show
that edge-group EðK; vÞ is isomorphic to the fundamental group π1 ðjKj; vÞ for any
simplicial complex K. Finally, van Kampen theorem is proved by using
graph-theoretic results. This chapter also proves simplicial approximation theorem
given by L.E.J. Brouwer (1881–1967) and J.W. Alexander (1888–1971) around
1920 by utilizing a certain good feature of simplicial complexes introduced by
Alexander. This theorem plays a key role in the study of homotopy and homology
theories.
Chapter 7 continues to study homotopy theory displaying the construction of a
sequence of functors πn given by W. Hurewicz (1904–1956) in 1935 from topology
to algebra by extending the concept of fundamental group invented by H. Poincaré in
1895. The basic idea of homotopy groups is to classify all continuous maps from Sn
to pointed topological space X up to homotopy equivalence. To study topological
spaces X of low dimension, the fundamental group π1 ðXÞ is very useful. But it needs
refined tools for the study of higher dimensional spaces. For example, fundamental
group cannot distinguish spheres Sn with n 2. Such a limitation of low dimension
can be removed by considering the natural higher dimensional analogues of π1 ðXÞ.
More precisely, this chapter defines the nth (absolute) homotopy group and
generalizes it to a (relative) homotopy group of a triplet and studies algebraic,
functorial and fibering properties with the exactness of homotopy sequence of
fibering, Hopf maps introduced by H. Hopf (1894–1971) in 1935 for the investi-
gation of certain homotopy groups of Sn , action of π1 on πn , Freudenthal suspension
theorem given by H. Freudenthal (1905–1990) in 1937 for the investigation of the
homotopy groups πm ðSn Þ for 0\m\n, weak fibration which plays a key role in the
study of higher homotopy groups, and the nth cohomotopy set πn ðX; AÞ on which K.
Borsuk (1905–1982) endowed in 1936 with an abelian group structure under certain
conditions on ðX; AÞ. This chapter also discusses some interesting applications of
higher homotopy groups.
Chapter 8 continues to study homotopy theory through a suitable special class of
topological spaces, called CW-complexes introduced by J.H.C. Whitehead (1904–
1960) in 1949 to meet the need for further development of homotopy theory. This
class of spaces is broader and has some better categorical properties than simplicial
complexes, but still retains a combinatorial nature that allows for computation (often
with a much smaller complex). The concept of CW-complexes is introduced as a
natural generalization of the concept of polyhedra by relaxing all “linearity condi-
tions” in simplicial complexes, instead cells are attached by arbitrary continuous
maps starting with a discrete set, whose each point is regarded as a 0-cell.
A CW-complex is built up by successive adjunctions of cells of dimensions
1; 2; 3; . . .: There is an analogy between what can be done topologically with a space,
and what can be done algebraically with its chain groups. In the class of
CW-complexes this analogy attains its highest strength. The category of
Preface xi
CW-complexes is a suitable category for a systematic study of algebraic topology.

Algebraic topologists now feel that a study of CW-complex should be included in
the basic course of algebraic topology, and this study should move to the theorem
that every continuous map between CW-complexes is homotopic to a cellular map.
This chapter studies the basic aspects of CW-complexes and relative CW-complexes
with their homotopy properties and proves Whitehead theorem, Freudenthal
suspension theorem (general form), and cellular approximation theorem with their
applications.
Chapter 9 continues to study homotopy theory through the different products in
homotopy groups such as the Whitehead product introduced by J.H.C. Whitehead
in 1941, mixed product introduced by McCarty in 1964, and Samelson product.
Whitehead product provides a technique at least in some cases for constructing
nonzero elements of the homotopy group πp þ q1 ðXÞ of a pointed topological space
X. Moreover, this chapter finds a generalization of Whitehead product, establishes
certain relation between Whitehead and Samelson products, and studies mixed
products corresponding to a fiber space and a topological transformation group
acting on it.
Chapter 10 begins to study homology and cohomology theories. Homotopy
groups are very difficult to compute. There is an alternative approach of construction
of a different topological invariant, the so-called homology group, which historically
came earlier than homotopy groups. Homology (cohomology) theory is a covariant
(contravariant) functor from the category of topological spaces to the category of
abelian groups. Homology (simplicial) invented by H. Poincaré in 1895 is one of the
most fundamental influential inventions in mathematics. The basic idea of homology
is that it starts with a geometric object (a space) which is given by combinatorial data
(a complex). Then the linear algebra and boundary relations determined by this data
are used to construct homology groups. The simplicial devices in simplicial
homology theory are gradually generalized to singular homology by using the
algebraic properties of the singular complex. ^
There exist different homology
theories: simplicial, singular, cellular, and Cech homology theories which are
studied in this chapter. The most important homology theory in algebraic topology is
the singular homology. Simplicial homology is the primitive version of singular
homology. Cohomology theory is also discussed. In some sense, homology theory
and cohomology theory are dual to each other. More precisely, this chapter begins
with a study of the concepts of chain complex, boundary, cycle introduced by
W. Mayer (1887–1947) in 1929 from a purely algebraic viewpoint. This chapter
presents a construction of the homology groups of a simplicial complex in two steps:
first by assigning to each simplicial complex a certain complex, called chain complex
followed by assigning to the chain complex its homology group. This construction
assigns a group structure to cycles that are not boundaries with an extension to the
concept of relative simplicial homology groups and generalizes simplicial homology
theory to singular homology theory. These two theories are related by the basic result
that the singular homology of a polyhedron is isomorphic to the simplicial homology
of any of its triangulated simplicial complexes. This chapter examines the relations
xii Preface
between absolute homology groups of simplicial chain complexes and the relative
homology groups of relative simplicial chain complexes by using the language of
exact sequences and shows that the relative homology groups Hp ðK; LÞ for any pair
ðK; LÞ of simplicial complexes fit into a long exact sequence. This chapter also
discusses homology groups Hn ðX; GÞ with an arbitrary coefficient group G (abelian),
Mayer–Vietoris sequences in singular and simplicial homology theories, cup
product, and gives the Künneth formula and Eilenberg–Zilber theorem which are
used for computing homology or cohomolgy of product spaces, and Euler
characteristic & Jordan curve theorem from the viewpoint of homology theory.
Chapter 11 studies a special class of CW-complexes having only one nonzero
homotopy groups, called Eilenberg–MacLane spaces which were introduced by
S. Eilenberg (1915–1998) and S. MacLane (1909–2005) in 1945. Such spaces form
a very important class of CW-complexes in algebraic topology. Their importance is
twofold: they develop both homotopy and homology theories. They are closely
linked with the study of cohomology operations. This chapter presents Eilenberg–
MacLane spaces with their construction and studies their homotopy properties. The
construction process of Eilenberg–MacLane spaces KðG; nÞ for all possible ðG; nÞ
depends on a very natural class of spaces, called Moore spaces of type ðG; nÞ,
denoted by MðG; nÞ: This chapter also studies Postnikov towers to meet the need
for construction of Eilenberg–MacLane spaces.
Chapter 12 presents an approach formulating axiomatizaton of ordinary
homology and cohomology theories. These axioms, now called Eilenberg and
Steenrod axioms were announced by S. Eilenberg (1915–1998) and N.E. Steenrod
(1910–1971) in 1945, but first appeared in their celebrated book Foundations of
Algebraic Topology in 1952. This approach came from the problem of comparing
the various definitions of homology and cohomology given in the previous years.
Eilenberg and Steenrod initiated a new approach by taking a small number of their
properties (not focussing on machinery used for construction of homology and
cohomology groups) as axioms to characterize a theory of homology and
cohomology. This axiomatic approach simplifies the proofs of many lengthy and
complicated theorems and escapes the avoidable difficulty to motivate the students
who are learning homology and cohomology theories for the first time as their
systematic study. This axiomatic approach classifies and unifies different homology
groups on the category of compact triangulable spaces and inaugurates its dual
theory called cohomology theory. This approach is the most important contribution
to algebraic topology since the invention of the homology groups by Poincaré in
1895.
Chapter 13 continues the study of homology and cohomology theories by
presenting some of their interesting properties which directly follow from the
Eilenberg and Steenrod axioms for homology and cohomology theories such as
homotopy equivalence in these theories, relations between cofibrations and
homology theory, and finally computes the ordinary homology groups of Sn with
coefficients in an arbitrary abelian group G.
Preface xiii
Chapter 14 presents further interesting applications of the homotopy, homology,

and cohomology theories. The notions initially introduced in these theories to solve
problems of topology that have fruitful applications, and proves many interesting
theorems such as Hopf’s classification theorem, hairy ball theorem, ham sandwich
theorem, Borsuk–Ulam theorem, Lusternik–Schnirelmann theorem, Lefschetz fixed
point theorem, and Jordan curve theorem. It also proves some results related to
graph theory, fixed point theory of continuous maps, vector fields, and applications
to algebra. Moreover, this chapter indicates some applications of algebraic topology
in physics, chemistry, economics, biology, and medical science with specific
references.
Chapter 15 conveys the concept of a spectrum originated by F.L. Lima (1929–)
in 1958 and constructs its associated spectral homology and cohomology theories,
and generalized homology and cohomology theories (which have been proved to be
very useful theories) to distinguish them from ordinary homology and cohomology
theories. Their properties and relations to homotopy theory are also discussed. For
example, the ordinary homology group of certain topological spaces X can be
thought of as an approximation to πn ðXÞ. Moreover, this chapter constructs a new
Ω-spectrum A, generalizing the Eilenberg–MacLane spectrum KðG; nÞ and also
constructs its associated cohomology theory h ð; AÞ which generalizes the ordinary
cohomology theory of Eilenberg and Steenrod. This chapter conveys K-theory as a
generalized cohomology theory and also studies the Brown representability
theorem, stable homotopy groups, the cohomology operations, and Poincaré duality
theorem.
Chapter 16 studies a theory known as “obstruction theory” by utilizing the tools
of cohomology theory to encounter two basic problems in algebraic topology such
as extension and lifting problems. Obvious examples are the homotopy extension
and homotopy lifting problems. The homotopy classifications of continuous maps,
together with the study of extension and lifting problems, play a central role in
algebraic topology. The term “obstruction theory” refers to a technique for defining
a sequence of cohomology classes that are obstructions to finding solution to the
extension, lifting or relative lifting problems. Obstruction theory leads to make an
attempt to find a general solution. This theory originated in the classical work of
Hopf, Eilenberg, Steenrod, and Postinikov in around 1940. Certain sets of
cohomology elements, called obstructions, are associated with both a single map in
the case of extension and with a pair of maps in the case of homotopies. These are
invariants depending only on the topological spaces and their continuous mappings.
In polyhedra these are the characteristics for the existence or non-existence of the
desired extensions and homotopies. The underlying idea of associating cohomology
elements with continuous mappings was implicitly used by H. Whitney (1907–
1989) and first explicitly formulated by N.E. Steenrod (1910–1971). This chapter
uses cohomology theory to yield algebraic indicators for obstacles to extension and
lifting problems of continuous maps and proves Eilenberg extension theorem. It
presents some applications of obstruction theory to prove a homological version of
Whitehead theorem, stepwise extension of cross-section and obstruction for
homotopy between relative lifts.
xiv Preface
Chapter 17 presents some similarities and interesting relations among homotopy,

homology, and cohomology. In earlier chapters, some relations between these
theories have been discussed. This chapter continues to convey more relations
through Hurewicz homomorphism, Eilenberg–MacLane spaces, Dold–Thom
theorem, Brown’s representation theorem, Hopf invariant and Adams classical
theorem on Hopf invariant. Historically, L.E.J. Brouwer first connected homology
and homotopy in 1912 by proving 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 and only if they are equivalent from the view point of
homology theory). Hopf’s classification theorem generalizes Brouwer’s result to an
arbitrary dimension. The homotopy groups resemble the homology groups in
many respects under suitable situations proved by Hurewicz in his celebrated
“equivalence theorem”. There is also a lack of similarities between these two
theories essentially due to the absence in higher homotopy groups the excision
property for homology and also due to the absence in higher homotopy groups a
theorem analogous to van Kampen theorem for fundamental group.
Chapter 18 focuses a brief history of algebraic topology highlighting the
emergences of the ideas leading to new areas of study in algebraic topology and
conveys the contributions of some mathematicians who introduced new concepts or
proved theorems of fundamental importance or inaugurated new theories in
algebraic topology starting from the creation of fundamental group and homology
group by H. Poincaré in 1895, which are the first basic and influential inventions in
algebraic topology. The literature on algebraic topology is very vast. Some concepts
studied now in algebraic topology had been found in the work of B. Riemann
(1826–1866), C. Felix Klein (1849–1925), and H. Poincaré (1854–1912). But the
foundation of algebraic (combinatorial topology) was laid in the decade beginning
1895 by H. Poincaré through the publication of his famous series of memoirs
“Analysis Situs” from 1895 onwards. J.W. Alexander (1888–1971) used the word
“topological” in the titles of his research papers in the 1920s. This chapter also
conveys more names with their contributions in algebraic topology. The early
development of homotopy theory was essentially due to H. Poincaré, L.E.F.
Brouwer, H. Hopf, W. Hurewicz, H. Freudenthal, and many others. W. Hurewicz
first established a connection between homology and homotopy groups for ðn 1Þ-
connceted spaces, when n 2. H. Hopf pioneered a study of maps into spheres
during 1926–1935 and inaugurated the homotopy theory with the discovery of the
Hopf map followed by the research of W. Hurewicz, and Freudenthal. Since then
homotopy theory has made a rapid progress and now plays an important role in
mathematics. Homology, invented by Henry Poincaré during 1895–1901, is one
of the most fundamental influential inventions in mathematics. He started with a
geometric object (a space) which is given by combinatorial data (a simplicial
complex), then the linear algebra and boundary relations by these data are used to
construct homology groups. There are other homology theories:
Preface xv
(i) Homology groups for compact metric spaces introduced by L. Vietoris

(1891–2002) in 1927; ^
(ii) Homology groups for compact Hausdorff spaces introduced by E.Cech
(1893–1960) in 1932;
(iii) Singular homology groups are first defined by S. Lefschitz (1884–1972) in
1933.
All these homology theories lived in isolation. Algebraic topologists in around
1940 started comparing various definitions of homology and cohomology given in
the previous years. Eilenberg and Steenrod initiated a new approach in 1945 by taking
a small number of their properties (not focusing on machinery used for construction
of homology and cohomology groups) as axioms to characterize a theory of
homology and cohomology. This approach is the most important contribution to
algebraic topology since the invention of the homology groups by Poincaré and is
called the axiomatic approach given by a set of seven axioms announced by
S. Eilenberg and N. Steenrod in 1945 and published in their book in 1952. This
approach classifies and unifies different homology groups on the category of compact
triangulated spaces and inaugurated its dual theory for cohomology theories. This
chapter also conveys the contributions of more mathematicians, S. MacLane,
J.H. Whitehead, Serre, Brown, Milnor, and Grothendieck, to name a few.
Apppendix A studies classical topological groups and Lie groups that occupy
a vast territory in topology and geometry. Lie groups are special topological
groups and also manifolds carrying a differential structure. For example,
GLðn; RÞ; GLðn; CÞ; GLðn; HÞ; SLðn; RÞ; SLðn; CÞ; Oðn; RÞ; Uðn; CÞ; SLðn; HÞ are
some important classical Lie Groups. Historically, S. Lie (1842–1899) investigated
group of transformations. He developed his theory of transformation groups to
solve his integration problems. Such groups are now called Lie groups after his
name. The Fifth Problem of Hilbert announced at the ICM 1900, Paris, is linked to
Sophus Lie theory of transformation groups which asserts that Lie groups act as
groups of transformations on manifolds.
Appendix B discusses category theory through the study of categories, functors,
and natural transformations with an eye to study algebraic topology which consists
of the construction and use of functors from some category of topological spaces
into an algebraic category. This theory plays an important role for the study of
homotopy, homology, and cohomology theories, which constitute the basic text of
this book in addition to adjoint functor, representable functor, abelianization
functor, Brown functor, and infinite symmetric product functor. All constructions in
algebraic topology are in general functorial. Fundamental groups, higher homotopy
groups, and homology and cohomology groups are not only invariants of the
underlying topological space, in the sense that two topological spaces which are
homeomorphic have the isomorphic associated groups (or modules) but their
associated morphisms also correspond to a continuous mapping of topological
spaces an induced group (or module) homomorphism on the associated groups
(modules), and these homomorphisms can be used to show non-existence (or, much
xvi Preface
more deeply, existence) of mappings. So the readers of algebraic topology cannot

escape learning the concepts of categories, functors and natural transformations.
The author acknowledges the Science and Engineering Research Board (SERB)
under the Department of Science & Technology of the Government of India for
sanctioning of the financial support towards writing this book under the Utilization
of Scientific Expertise of Retired Scientists (USERS) scheme (G.O. No HR/
UR/17/2010 dated December 27, 2011) and also to West Bengal University of
Technology (now renamed as Moulana Abul Kalam Azad University of
Technology) for providing infrastructure towards implementing the scheme. The
author expresses his sincere thanks to Springer for publishing this book. The author
is very thankful to the reviewers of the manuscript and also to Dr. Avishek Adhikari
of the University of Calcutta for their scholarly suggestions for improvement of the
book. Thanks are also due to K. Sardar for his cooperation in dealing with the
typing of one version after another of the manuscript and also to many other
individuals who have helped in proofreading the book. Author’s thanks are due to
the institute IMBIC, Kolkata, for providing the author with the library and other
facilities towards the manuscript development work of this book. Finally, the author
acknowledges, with heartfelt thanks, the patience and sacrifice of long-suffering
family of the author, specially author’s wife Minati, son Avishek, daughter-in-law
Shibopriya, and grandson little Avipriyo.
Kolkata, India Mahima Ranjan Adhikari

March 2016
Contents
1 Prerequisite Concepts and Notations . . . . . . . . . . . . . . . . . . . . . . . 1

1.1 Set Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Groups and Fundamental Homomorphism Theorem . . . . . . . . 4
1.3 Group Representations, Free Groups, and Relations . . . . . . . . 7
1.3.1 Linear Representation of a Group . . . . . . . . . . . . . . 7
1.3.2 Free Groups and Relations. . . . . . . . . . . . . . . . . . . 8
1.3.3 Betti Number and Structure Theorem for Finite
Abelian Group . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Exact Sequence of Groups. . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.5 Free Product and Tensor Product of Groups . . . . . . . . . . . . . 13
1.5.1 Free Product of Groups . . . . . . . . . . . . . . . . . . . . . 13
1.5.2 Tensor Product of Groups . . . . . . . . . . . . . . . . . . . 14
1.6 Torsion Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 Actions of Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.8 Modules and Vector Spaces. . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8.1 Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8.2 Direct Sum of Modules . . . . . . . . . . . . . . . . . . . . . 17
1.8.3 Tensor Product of Modules . . . . . . . . . . . . . . . . . . 17
1.8.4 Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.9 Euclidean Spaces and Some Standard Notations . . . . . . . . . . . 22
1.10 Set Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
1.10.1 Topological Spaces: Introductory Concepts . . . . . . . 23
1.10.2 Homeomorphic Spaces . . . . . . . . . . . . . . . . . . . . . 25
1.10.3 Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.10.4 Connectedness and Locally Connectedness . . . . . . . 28
1.10.5 Compactness and Paracompactness . . . . . . . . . . . . . 29
1.10.6 Weak Topology . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.11 Partition of Unity and Lebesgue Lemma . . . . . . . . . . . . . . . . 30
1.11.1 Lebesgue Lemma and Lebesgue Number . . . . . . . . . 31
xvii
xviii Contents
1.12 Separation Axioms, Urysohn Lemma, and Tietze

Extension Theorem. . . . . . . . . . . . . . . . . . . . . . . . . ...... 31
1.13 Identification Maps, Quotient Spaces, and Geometrical
Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
1.14 Function Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
1.15 Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.16 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
1.17 Additional Reading. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2 Homotopy Theory: Elementary Basic Concepts . . . . . ..... . . . . . 45
2.1 Homotopy: Introductory Concepts and Examples ..... . . . . . 47
2.1.1 Concept of Homotopy . . . . . . . . . . . . ..... . . . . . 47
2.1.2 Functorial Representation. . . . . . . . . . ..... . . . . . 57
2.2 Homotopy Equivalence . . . . . . . . . . . . . . . . . . ..... . . . . . 58
2.3 Homotopy Classes of Maps . . . . . . . . . . . . . . . ..... . . . . . 62
2.4 H-Groups and H-Cogroups . . . . . . . . . . . . . . . ..... . . . . . 64
2.4.1 H-Groups and Loop Spaces . . . . . . . . ..... . . . . . 65
2.4.2 H-Cogroups and Suspension Spaces . . ..... . . . . . 75
2.5 Adjoint Functors . . . . . . . . . . . . . . . . . . . . . . ..... . . . . . 79
2.6 Contractible Spaces . . . . . . . . . . . . . . . . . . . . ..... . . . . . 83
2.6.1 Introductory Concepts . . . . . . . . . . . . ..... . . . . . 83
2.6.2 Infinite-Dimensional Sphere and Comb Space . . . . . 85
2.7 Retraction and Deformation . . . . . . . . . . . . . . . ..... . . . . . 88
2.8 NDR and DR Pairs. . . . . . . . . . . . . . . . . . . . . ..... . . . . . 95
2.9 Homotopy Properties of Infinite Symmetric
Product Spaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
2.10 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
2.10.1 Extension Problems . . . . . . . . . . . . . . . . . . . . . . . 97
2.10.2 Fundamental Theorem of Algebra . . . . . . . . . . . . . . 99
2.11 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
3 The Fundamental Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
3.1 Fundamental Groups: Introductory Concepts . . . . . . . . . . . . . 108
3.1.1 Basic Motivation . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.1.2 Introductory Concepts . . . . . . . . . . . . . . . . . . . . . . 109
3.1.3 Functorial Property of π1 . . . . . . . . . . . . . . . . . . . . 117
3.1.4 Some Other Properties of π1 . . . . . . . . . . . . . . . . . 119
3.2 Alternative Definition of Fundamental Groups . . . . . . . . . . . . 124
3.3 Degree Function and the Fundamental Group
of the Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.4 The Fundamental Group of the Punctured Plane. . . . . . . . . . . 130
3.5 Fundamental Groups of the Torus . . . . . . . . . . . . . . . . . . . . 131
Contents xix
3.6 Vector Fields and Fixed Points . . . . . . . . . . . . . . . . . . . . . . 132

3.7 Knot and Knot Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.8 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
3.8.1 Fundamental Theorem of Algebra . . . . . . . . . . . . . . 136
3.8.2 An Alternative Proof of Brouwer Fixed Point
Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
3.8.3 Borsuk–Ulam Theorem . . . . . . . . . . . . . . . . . . . . . 139
3.8.4 Cauchy’s Integral Theorem of Complex
Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
3.9 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4 Covering Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
4.1 Covering Spaces: Introductory Concepts and Examples . . . . . . 148
4.1.1 Introductory Concepts . . . . . . . . . . . . . . . . . . . . . . 148
4.1.2 Some Interesting Properties of Covering Spaces . . . . 151
4.1.3 Covering Spaces of RPn . . . . . . . . . . . . . . . . . . . . 152
4.2 Computing Fundamental Groups of Figure-Eight
and Double Torus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
4.3 Path Lifting and Homotopy Lifting Properties . . . . . . . . . . . . 155
4.4 Lifting Problems of Arbitrary Continuous Maps . . . . . . . . . . . 158
4.5 Covering Homomorphisms: Their Classifications
and Galois Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.5.1 Covering Homomorphisms and Deck
Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.5.2 Classification of Covering Spaces
by Using Group Theory. . . . . . . . . . . . . . . . . . . . . 163
4.5.3 Classification of Covering Spaces and Galois
Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . 167
4.6 Universal Covering Spaces and Computing π1 (RPn) . . . . . . . . 170
4.6.1 Universal Covering Spaces . . . . . . . . . . . . . . . . . . 170
4.6.2 Computing π1 (RPn) . . . . . . . . . . . . . . . . . . . . . . . 172
4.7 Fibrations and Cofibrations . . . . . . . . . . . . . . . . . . . . . . . . . 174
4.7.1 Homotopy Lifting Problems . . . . . . . . . . . . . . . . . . 174
4.7.2 Fibration: Introductory Concepts. . . . . . . . . . . . . . . 176
4.7.3 Cofibration: Introductory Concepts . . . . . . . . . . . . . 179
4.8 Hurewicz Theorem for Fibration and Characterization
of Fibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
4.9 Homotopy Liftings and Monodromy Theorem . . . . . . . . . . . . 184
4.9.1 Path Liftings and Homotopy Liftings . . . . . . . . . . . 185
4.9.2 Monodromy Theorem . . . . . . . . . . . . . . . . . . . . . . 185
4.10 Applications and Computations . . . . . . . . . . . . . . . . . . . . . . 186
4.10.1 Actions of Fundamental Groups . . . . . . . . . . . . . . . 186
4.10.2 Fundamental Groups of Orbit Spaces . . . . . . . . . . . 187
xx Contents
4.10.3 Fundamental Group of the Real Projective

Space RPn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
4.10.4 The Fundamental Group of Klein’s Bottle . . . . . . . . 189
4.10.5 The Fundamental Groups of Lens Spaces . . . . . . . . 190
4.10.6 Computing Fundamental Group of Figure-Eight
by Graph-theoretic Method . . . . . . . . . . . . . . . . . . 191
4.10.7 Application of Galois Correspondence. . . . . . . . . . . 191
4.11 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
5 Fiber Bundles, Vector Bundles and K-Theory . . . . . . . . . . . . . . . . 197
5.1 Bundles, Cross Sections, and Examples. . . . . . . . . . . . . . . . . 198
5.1.1 Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
5.1.2 Cross Sections . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
5.1.3 Morphisms of Bundles . . . . . . . . . . . . . . . . . . . . . 201
5.1.4 Examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
5.2 Fiber Bundles: Introductory Concepts . . . . . . . . . . . . . . . . . . 206
5.3 Hopf and Hurewicz Fiberings . . . . . . . . . . . . . . . . . . . . . . . 210
5.3.1 Hopf Fibering of Spheres. . . . . . . . . . . . . . . . . . . . 210
5.3.2 Hurewicz Fibering . . . . . . . . . . . . . . . . . . . . . . . . 212
5.4 G-Bundles and Principal G-Bundles . . . . . . . . . . . . . . . . . . . 213
5.5 Homotopy Properties of Numerable Principal G-Bundles. . . . . 218
5.6 Classifying Spaces: The Milnor Construction . . . . . . . . . . . . . 220
5.7 Vector Bundles: Introductory Concepts . . . . . . . . . . . . . . . . . 223
5.8 Charts and Transition Functions of Bundles. . . . . . . . . . . . . . 229
5.9 Homotopy Classification of Vector Bundles. . . . . . . . . . . . . . 233
5.10 K-Theory: Introductory Concepts . . . . . . . . . . . . . . . . . . . . . 236
5.11 Principal G-Bundles for Lie Groups G . . . . . . . . . . . . . . . . . 240
5.12 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
5.13 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 241
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
6 Geometry of Simplicial Complexes and Fundamental
Groups of Polyhedra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
6.1 Geometry of Finite Simplicial Complexes . . . . . . . . . . . . . . . 250
6.2 Triangulations and Polyhedra . . . . . . . . . . . . . . . . . . . . . . . . 253
6.3 Simplicial Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
6.4 Barycentric Subdivisions . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
6.5 Simplicial Approximation . . . . . . . . . . . . . . . . . . . . . . . . . . 261
6.6 Computing Fundamental Groups of Polyhedra . . . . . . . . . . . . 264
6.7 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Contents xxi
6.7.1 Application to Extension Problem. . . . . . . . . . . . . . 265

6.7.2 Application to Graph Theory . . . . . . . . . . . . . . . . . 266
6.7.3 van Kampen Theorem . . . . . . . . . . . . . . . . . . . . . . 267
6.8 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
7 Higher Homotopy Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
7.1 Absolute Homotopy Groups: Introductory Concept . . . . . . . . . 274
7.2 Absolute Homotopy Groups Defined by Hurewicz . . . . . . . . . 277
7.3 Functorial Properties of Absolute Homotopy Groups. . . . . . . . 278
7.4 The Relative Homotopy Groups: Introductory Concepts . . . . . 281
7.5 The Boundary Operator and Induced Transformation . . . . . . . 282
7.5.1 Boundary Operator . . . . . . . . . . . . . . . . . . . . . . . . 282
7.5.2 Induced Transformations . . . . . . . . . . . . . . . . . . . . 283
7.6 Functorial Property of the Relative Homotopy Groups . . . . . . 284
7.7 Homotopy Sequence and Its Exactness . . . . . . . . . . . . . . . . . 285
7.7.1 Homotopy Sequence and Its Exactness . . . . . . . . . . 285
7.7.2 Some Consequences of the Exactness
of the Homotopy Sequence . . . . . . . . . . . . . . . . . . 287
7.8 Homotopy Sequence of Fibering and Hopf Fibering . . . . . . . . 289
7.8.1 Homotopy Sequence of Fibering. . . . . . . . . . . . . . . 289
7.8.2 Hopf Fiberings of Spheres . . . . . . . . . . . . . . . . . . . 290
7.8.3 Problems of Computing πm ðSn Þ . . . . . . . . . . . . . . . 290
7.9 More on Hopf Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
7.10 Freudenthal Suspension Theorem and Table of πi ðSn Þ
for 1 I; n 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
7.10.1 Freudenthal Suspension Theorem . . . . . . . . . . . . . . 293
7.10.2 Table of πi ðSn Þ for 1 I; n 8 . . . . . . . . . . . . . . . . 294
7.11 Action of π1 on πn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
7.12 The nth Cohomotopy Sets and Groups . . . . . . . . . . . . . . . . . 295
7.13 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 297
7.14 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 300
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
8 CW-Complexes and Homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 305
8.1 Cell-Complexes and CW-Complexes: Introductory
Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306
8.1.1 Cell-Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . 307
8.1.2 CW-Complexes. . . . . . . . . . . . . . . . . . . . . . . . . . . 308
8.1.3 Examples of Spaces Which Are Neither
CW-Complexes Nor Homotopy Equivalent
to a CW-Complex . . . . . . . . . . . . . . . . . . . . . . . . . 312
8.2 Cellular Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
xxii Contents
8.3 Subcomplexes of CW-Complexes . . . . . . . . . . . . . . . . . . . . . 313

8.4 Relative CW-Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
8.5 Homotopy Properties of CW-Complexes, Whitehead
Theorem and Cellular Approximation Theorem . . . . . . . . . . . 315
8.5.1 Homotopy Properties of CW-Complexes . . . . . . . . . 315
8.5.2 Whitehead Theorem . . . . . . . . . . . . . . . . . . . . . . . 317
8.5.3 Cellular Approximation Theorem . . . . . . . . . . . . . . 318
8.6 More on Homotopy Properties of CW-Complexes . . . . . . . . . 319
8.7 Blakers–Massey Theorem and a Generalization of
Freudenthal Suspension Theorem . . . . . . . . . . . . . . . . . . . . . 320
8.8 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
8.9 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
9 Products in Homotopy Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
9.1 Whitehead Product Between Homotopy Groups
of CW-Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
9.2 Whitehead Products Between Homotopy Groups
of H-Spaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
9.3 A Generalization of Whitehead Product. . . . . . . . . . . . . . . . . 335
9.4 Mixed Products in Homotopy Groups . . . . . . . . . . . . . . . . . . 336
9.4.1 Mixed Product in the Homotopy Category
of Pointed Topological Spaces . . . . . . . . . . . . . . . . 336
9.4.2 Mixed Product Associated with Fibrations . . . . . . . . 337
9.5 Samelson Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
9.5.1 The Samelson Product. . . . . . . . . . . . . . . . . . . . . . 338
9.5.2 The Iterated Samleson Product . . . . . . . . . . . . . . . . 339
9.6 Some Relations Between Whitehead and Samelson
Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
9.7 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
9.7.1 Adams Theorem Using Whitehead Product . . . . . . . 341
9.7.2 Homotopical Nilpotence of the Seven Sphere S7 . . . . 342
9.8 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
10 Homology and Cohomology Theories . . . . . . . . . . . . . . . . . . . . . . 347
10.1 Chain Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
10.2 Simplicial Homology Theory . . . . . . . . . . . . . . . . . . . . . . . . 352
10.2.1 Construction of Homology Groups of a Simplicial
Complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
10.2.2 Induced Homomorphism and Functorial
Properties of Simplicial Homology . . . . . . . . . . . . . 360
10.2.3 Computing Homology Groups of Polyhedra . . . . . . . 361
Contents xxiii
10.3 Relative Simplicial Homology Groups . . . . . . . . . . . . . . . . . 362

10.4 Exactness of Simplicial Homology Sequences . . . . . . . . . . . . 364
10.5 Simplicial Cohomology Theory: Introductory Concepts . . . . . . 365
10.6 Simplicial Cohomology Ring . . . . . . . . . . . . . . . . . . . . . . . . 367
10.7 Singular Homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
10.7.1 Singular Homology Groups . . . . . . . . . . . . . . . . . . 369
10.7.2 Reduced Singular Homology Groups. . . . . . . . . . . . 372
10.7.3 Relative Singular Homology Groups . . . . . . . . . . . . 373
10.8 Eilenberg–Zilber Theorem and Künneth Formula . . . . . . . . . . 374
10.8.1 Eilenberg–Zilber Theorem . . . . . . . . . . . . . . . . . . . 374
10.8.2 Künneth Formula . . . . . . . . . . . . . . . . . . . . . . . . . 374
10.9 Singular Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
10.10 Relative Cohomology Groups . . . . . . . . . . . . . . . . . . . . . . . 376
10.11 Hurewicz Homomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . 377
10.12 Mayer–Vietoris Sequences. . . . . . . . . . . . . . . . . . . . . . . . . . 379
10.12.1 Mayer–Vietoris Sequences in Singular
Homology Theory . . . . . . . . . . . . . . . . . . . . . . . . 379
10.12.2 Mayer–Vietoris Sequences in Simplicial
Homology Theory . . . . . . . . . . . . . . . . . . . . . . . . 380
10.13 Computing Homology Groups . . . . . . . . . . . . . . . . . . . . . . . 381
10.13.1 Homology Groups of a One-Point Space . . . . . . . . . 381
10.13.2 Homology Groups of CW-complexes . . . . . . . . . . . 382
10.14 Cellular Homology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
10.15 Čech Homology and Cohomology Groups. . . . . . . . . . . . . . . 384
10.16 Universal Coefficient Theorem for Homology and
Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
10.16.1 Homology with Arbitrary Coefficient Group. . . . . . . 385
10.16.2 Universal Cohomology Theorem
for Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . 387
10.17 Betti Number and Euler Characteristic . . . . . . . . . . . . . . . . . 388
10.17.1 Euler Characteristics of Polyhedra. . . . . . . . . . . . . . 388
10.17.2 Euler Characteristic of Finite Graphs . . . . . . . . . . . . 390
10.17.3 Euler Characteristic of Graded Vector Spaces. . . . . . 390
10.17.4 Euler–Poincaré Theorem for Finite
CW-complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
10.18 Cup and Cap Products in Cohomology Theory . . . . . . . . . . . 392
10.18.1 Cup Product. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393
10.18.2 Cap Product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
10.19 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
10.19.1 Jordan Curve Theorem_. . . . . . . . . . . . . . . . . . . . . 396
10.19.2 Homology Groups of Snα . . . . . . . . . . . . . . . . . .
α2a 397
xxiv Contents
10.20 Invariance of Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . 398

10.21 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406
11 Eilenberg–MacLane Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 407
11.1 Eilenberg–MacLane Spaces: Introductory Concept . . . . . . . . . 407
11.2 Construction of Eilenberg–MacLane Spaces KðG; nÞ. . . . . . . . 409
11.2.1 A Construction of KðG; 1Þ . . . . . . . . . . . . . . . . . . . 409
11.2.2 A Construction of K(G, n) for n [ 1 . . . . . . . . . . . . 409
11.2.3 Moore Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 410
11.2.4 Killing Homotopy Groups . . . . . . . . . . . . . . . . . . . 410
11.2.5 Postnikov Tower: Its Existence and Construction . . . 411
11.2.6 Existence Theorem . . . . . . . . . . . . . . . . . . . . . . . . 413
11.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
11.4 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
12 Eilenberg–Steenrod Axioms for Homology and Cohomology
Theories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 419
12.1 Eilenberg–Steenrod Axioms for Homology Theory . . . . . . . . . 420
12.2 The Uniqueness Theorem for Homology Theory . . . . . . . . . . 422
12.3 Eilenberg–Steenrod Axioms for Cohomology Theory . . . . . . . 424
12.4 The Reduced 0-dimensional Homology
and Cohomology Groups. . . . . . . . . . . . . . . . . . . . . . . . . . . 427
12.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
12.5.1 Invariance of Homology Groups . . . . . . . . . . . . . . . 428
12.5.2 Invariance of Cohomology Groups . . . . . . . . . . . . . 428
12.5.3 Mayer–Vietoris Theorem . . . . . . . . . . . . . . . . . . . . 428
12.6 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 429
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
13 Consequences of the Eilenberg–Steenrod Axioms. . . . . . . . . . . . . . 433
13.1 Immediate Consequences. . . . . . . . . . . . . . . . . . . . . . . . . . . 433
13.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440
13.2.1 Cofibration and Homology. . . . . . . . . . . . . . . . . . . 440
13.2.2 Computing Ordinary Homology Groups of Sn . . . . . 441
13.3 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
Contents xxv
14 Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 445
14.1 Degrees of Spherical Maps and Their Applications . . . . . . . . . 445
14.1.1 Degree of a Spherical Map . . . . . . . . . . . . . . . . . . 446
14.1.2 Hopf Classification Theorem . . . . . . . . . . . . . . . . . 449
14.1.3 The Brouwer Fixed Point Theorem . . . . . . . . . . . . . 450
14.2 Continuous Vector Fields . . . . . . . . . . . . . . . . . . . . . . . . . . 451
14.3 Borsuk–Ulam Theorem with Applications . . . . . . . . . . . . . . . 453
14.3.1 Borsuk–Ulam Theorem . . . . . . . . . . . . . . . . . . . . . 453
14.3.2 Ham Sandwich Theorem . . . . . . . . . . . . . . . . . . . . 454
14.3.3 Lusternik–Schnirelmann Theorem . . . . . . . . . . . . . . 455
14.4 The Lefschetz Number and Fixed Point Theorems . . . . . . . . . 455
14.5 Application of Euler Characteristic . . . . . . . . . . . . . . . . . . . . 458
14.6 Application of Mayer–Vietoris Sequence . . . . . . . . . . . . . . . . 460
14.7 Application of van Kampen Theorem . . . . . . . . . . . . . . . . . . 461
14.8 Applications to Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . 462
14.9 Application of Brown Functor . . . . . . . . . . . . . . . . . . . . . . . 463
14.10 Applications Beyond Mathematics . . . . . . . . . . . . . . . . . . . . 464
14.10.1 Application to Physics. . . . . . . . . . . . . . . . . . . . . . 464
14.10.2 Application to Sensor Network. . . . . . . . . . . . . . . . 465
14.10.3 Application to Chemistry . . . . . . . . . . . . . . . . . . . . 465
14.10.4 Application to Biology, Medical Science
and Biomedical Engineering. . . . . . . . . . . . . . . . . . 466
14.10.5 Application to Economics . . . . . . . . . . . . . . . . . . . 466
14.10.6 Application to Computer Science . . . . . . . . . . . . . . 467
14.11 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 472
15 Spectral Homology and Cohomology Theories. . . . . . . . . . . . . . . . 475
15.1 Spectrum of Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 476
15.2 Spectral Reduced Homology Theory. . . . . . . . . . . . . . . . . . . 477
15.3 Spectral Reduced Cohomology Theory . . . . . . . . . . . . . . . . . 480
15.4 Generalized Homology and Cohomology Theories . . . . . . . . . 481
15.5 The Brown Representability Theorem . . . . . . . . . . . . . . . . . . 481
15.6 A Generalization of Eilenberg–MacLane Spectrum
and Construction of Its Associated Generalized
Cohomology Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
15.6.1 Construction of a New Ω-Spectrum A . . . . . . . . . . . 484
15.6.2 Construction of the Cohomology Theory
Associated with A. . . . . . . . . . . . . . . . . . . . . . . . . 485
15.7 K-Theory as a Generalized Cohomology Theory . . . . . . . . . . 487
15.8 Spectral Unreduced Homology and Cohomology Theories. . . . 488
15.9 Cohomology Operations . . . . . . . . . . . . . . . . . . . . . . . . . . . 489
15.9.1 Cohomology Operations of Type ðG; n; T; mÞ
and Eilenberg–MacLane Spaces . . . . . . . . . . . . . . . 490
xxvi Contents
15.9.2 Cohomology Operation Associated

with a Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . 491
15.9.3 Stable Cohomology Operations. . . . . . . . . . . . . . . . 492
k
15.9.4 A Characterization of the Group fOm g. . . . . . . . . . 493
15.10 Stable Homotopy Theory and Homotopy Groups
Associated with a Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . 496
15.10.1 Stable Homotopy Groups. . . . . . . . . . . . . . . . . . . . 496
15.10.2 Homology Groups Associated with a Spectrum . . . . 498
15.10.3 Homotopy Groups Associated with a Spectrum . . . . 499
15.11 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
15.11.1 Poincaré Duality Theorem . . . . . . . . . . . . . . . . . . . 500
15.11.2 Homotopy Type of the Eilenberg–MacLane
Space KðG; nÞ . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
15.11.3 Application of Representability Theorem
of Brown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
15.11.4 More Applications of Spectra . . . . . . . . . . . . . . . . . 504
15.11.5 Homotopical Description of Singular
Cohomology Theory . . . . . . . . . . . . . . . . . . . . . . . 505
15.12 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
16 Obstruction Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 511
16.1 Basic Aim of Obstruction Theory . . . . . . . . . . . . . . . . . . . . . 512
16.1.1 The Extension Problem . . . . . . . . . . . . . . . . . . . . . 513
16.1.2 The Lifting Problem . . . . . . . . . . . . . . . . . . . . . . . 514
16.1.3 Relative Lifting Problem . . . . . . . . . . . . . . . . . . . . 514
16.1.4 Cross Section Problem . . . . . . . . . . . . . . . . . . . . . 515
16.2 Notations and Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . 515
16.3 The Obstruction Theory: Basic Concepts. . . . . . . . . . . . . . . . 516
16.3.1 The Obstruction Cochains and Cocycles . . . . . . . . . 516
16.3.2 The Deformation and Difference Cochains . . . . . . . . 519
16.3.3 The Eilenberg Extension Theorem . . . . . . . . . . . . . 520
16.3.4 The Obstruction Set for Extension . . . . . . . . . . . . . 521
16.3.5 The Homotopy Index . . . . . . . . . . . . . . . . . . . . . . 522
16.4 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
16.4.1 A Link between Cohomolgy and Homotopy
with Hopf Theorem . . . . . . . . . . . . . . . . . . . . . . . 523
16.4.2 Stepwise Extension of A Cross Section . . . . . . . . . . 524
16.4.3 Homological Version of Whitehead Theorem . . . . . . 526
16.4.4 Obstruction for Homotopy Between
Relative Lifts . . . . . . . . . . . . . . . . . . . . . . . . . . . . 526
16.5 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 530
Contents xxvii
17 More Relations Between Homology and Homotopy . . . . . . . . . . . . 533

17.1 Some Similarities and Key Links . . . . . . . . . . . . . . . . . . . . . 534
17.1.1 Some Similarities . . . . . . . . . . . . . . . . . . . . . . . . . 534
17.1.2 Hurewicz Homomorphism Theorem:
A Key Link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
17.2 Relative Version of Hurewicz Homomorphism Theorem . . . . . 536
17.3 Alternative Proof of Homological Version
of Whitehead Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
17.4 Dold–Thom Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 538
17.5 The Hopf Invariant and Adams Theorem. . . . . . . . . . . . . . . . 539
17.5.1 Hopf Invariant . . . . . . . . . . . . . . . . . . . . . . . . . . . 539
17.5.2 Vector Field Problem and Adams Theorem . . . . . . . 541
17.6 Exercises. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545
18 A Brief History of Algebraic Topology . . . . . . . . . . . . . . . . . . . . . 547
18.1 Poincaré and his Conjecture. . . . . . . . . . . . . . . . . . . . . . . . . 548
18.2 Early Development of Homotopy Theory . . . . . . . . . . . . . . . 550
18.3 Category Theory and CW-Complexes . . . . . . . . . . . . . . . . . . 554
18.4 Early Development of Homology Theory . . . . . . . . . . . . . . . 555
18.5 Hopf Invariant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557
18.6 Eilenberg and Steenrod Axioms . . . . . . . . . . . . . . . . . . . . . . 558
18.7 Fiber Bundle, Vector Bundle, and K-Theory . . . . . . . . . . . . . 559
18.8 Eilenberg–MacLane Spaces and Cohomology Operations . . . . 561
18.9 Generalized Homology and Cohomology Theories . . . . . . . . . 562
18.10 Ω-Spectrum and Associated Cohomology Theories. . . . . . . . . 563
18.11 Brown Representability Theorem . . . . . . . . . . . . . . . . . . . . . 564
18.12 Obstruction Theory. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 567
Appendix A: Topological Groups and Lie Groups . . . . . . . . . . . . . . . . 569
Appendix B: Categories, Functors and Natural Transformations. . . . . . 581
List of Symbols. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609

About the Author
Mahima Ranjan Adhikari, Ph.D. is the former Professor of Pure Mathematics at

the University of Calcutta. His main interest lies in algebra and topology. He has
published a number of papers in several international journals including the
Proceedings of American Mathematical Society and five textbooks. Under his
supervision, 11 students have already been awarded the Ph.D. degree. He is a
member of the American Mathematical Society and serves on the editorial board of
several journals and research monographs. He was the President of the mathe-
matical science section of the 95th Indian Science Congress, 2008.
He has visited several institutions in India, USA, UK, Japan, France, Greece,
Sweden, Switzerland, Italy, and in many other countries on invitation. While vis-
iting E.T.H., Zurich, Switzerland, in 2003, he made an academic interaction with
Prof. B. Eckmann and Prof. P.J. Hilton. He is currently the President of the Institute
for Mathematics, Bioinformatics, Information Technology and Computer Science
(IMBIC). He is also the principal investigator of an ongoing project funded by the
Government of India.
This book is written based on author’s teaching experience of 50 years. He is the
joint author (with Avishek Adhikari) of another Springer book “Basic Modern
Algebra with Applications”, 2014.
xxix
Chapter 1
Prerequisite Concepts and Notations
This chapter assembles together some basic concepts and results of set theory,
algebra, analysis, set topology, Euclidean spaces, manifolds with standard notations
for smooth reading of the book. It is assumed that the readers are familiar with these
basic concepts. However, for their detailed study, the books Adhikari and Adhikari
(2014), Dugundji (1966), Herstein (1964), Maunder (1970), Spanier (1966), and
some other books are referred in Bibliography.
1.1 Set Theory
This section conveys some basic concepts of set theory (naive) initiated around 1870
by the German mathematician Georg Cantor (1845–1918) which are used throughout
the book. Set theory occupies an important position in mathematics. Many concrete
concepts and examples are based on it. It is assumed that the readers are familiar
with the sets
N (set of natural numbers/positive integers)

Z (set of integers)
Q (set of rational numbers)
R (set of real numbers)
C (set of complex numbers)
For precise description of many concepts of mathematics and also for mathemat-
ical reasoning the concepts of relations(functions) and cardinality of sets are very
important, which are discussed first.
A binary relation ρ on a nonempty set X is a subset of X × X , which is said to be
an equivalence relation if ρ is reflexive, i.e., (x, x) ∈ ρ for each x ∈ X ; symmetric,
i.e., (x, y) ∈ ρ implies (y, x) ∈ ρ and transitive i.e., (x, y) ∈ ρ and (y, z) ∈ ρ imply
(x, z) ∈ ρ for x, y, z ∈ X .
© Springer India 2016 1

M.R. Adhikari, Basic Algebraic Topology and its Applications,
DOI 10.1007/978-81-322-2843-1_1
2 1 Prerequisite Concepts and Notations
Definition 1.1.1 Let X be a nonempty set and ρ be an equivalence relation on X .

The disjoint classes [x] into which the set X is partitioned by ρ constitute a set, called
the quotient set of X by ρ, denoted by X/ρ, where [x] denotes the class (determined
by ρ) containing the element x of X . Each element x of the class [x] is called a
representative of [x].
Example 1.1.2 Given a positive integer n, the quotient set Zn consists of all n distinct
classes [0], [1], . . . , [n − 1]. The set Zn is called the residue classes of Z modulo n.
Remark 1.1.3 The set Zn provides very strong different algebraic structures (depend-
ing on n). The visual description of Z12 is a 12-h clock.
Definition 1.1.4 Given a nonempty set I , if there exists a set X i for each i ∈ I , then
the collection of the sets {X i : i ∈ I } is called a family of sets and I is called an
indexing set of the family.
For i = j, X i may be equal to X j . The collection of sets {X i : i ∈ I } is finite or

infinite according as I is a finite set or an infinite set.
Definition 1.1.5 A relation ρ on X is said to be antisymmetric if (x, y) ∈ ρ and

(y, x) ∈ ρ imply x = y for x, y ∈ X . A reflexive, antisymmetric and transitive rela-
tion on X is called a partial order relation.
Definition 1.1.6 If ρ is a partial order relation on a set X , then the pair (X, ρ) is
called a partially ordered set or a poset. A partially ordered set in which every pair of
elements is comparable, is called an ordered set and the set is called totally ordered.
Zorn’s Lemma Let (X, ≤) be a nonempty partially ordered set. If every subset
A ⊆ X , which is totally ordered by ≤, has an upper bound in X , then X has at least
one maximum element.
Remark 1.1.7 Zorn’s lemma is indispensible to prove many results of mathematics.
Definition 1.1.8 A map f : X → Y is said to be

(i) injective (or an injection) if different elements of X have different images in
Y , i.e., x = x implies f (x) = f (x ) for x, x in X ;
(ii) surjective (or onto or a surjection) if every element of Y is the image of some
element of X , i.e., for every element y in Y , there is some x in X such that
y = f (x);
(iii) bijective (or a bijection) if f is both injective and surjective.
Definition 1.1.9 Let f : X → Y and g : Y → Z be two maps. Their composite map

denoted by g ◦ f is the map g ◦ f : X → Z , x → g( f (x)), x ∈ X.
1.1 Set Theory 3
Proposition 1.1.10 Let f : X → Y and g : Y → Z be two maps. Then g ◦ f has

the properties:
(i) If f and g are injective, then g ◦ f injective;
(ii) If f and g are surjective, then g ◦ f is surjective;
(iii) If g ◦ f is surjective, then g is surjective;
(iv) If g ◦ f is injective, then f is injective;
(v) If f and g are bijective, then g ◦ f is bijective;
(vi) If g ◦ f is bijective, then f is injective and g is surjective. If in particular, if
g ◦ f = 1 X , the identity map of X , then f is injective and g is surjective.
The bijections of sets define equivalent sets.
Definition 1.1.11 Two sets X, Y are said to be equivalent denoted by X ∼ Y if there

exists a bijection from the set X to the set Y.
Example 1.1.12 Let S 2 = {(x, y, z) ∈ R3 : x 2 + y 2 + z 2 = 1} be the unit sphere in

the Euclidean space R3 and C∞ = C ∪ {∞} be the extended complex plane. Then
the stereographic projection f : S 2 → C∞ is a bijection. Thus C∞ is represented as
the sphere S 2 called Riemann sphere.
Example 1.1.13 A permutation of a set X is a bijection from X onto itself.
Definition 1.1.14 A set X is said to be finite if either X is empty or X ∼ Zn for

some integer n > 1. Two nonempty finite sets are equivalent if they have the same
number of elements.
Proposition 1.1.15 A finite set cannot be equivalent to a proper subset of the set.
Definition 1.1.16 A nonempty set which is not finite is said to be an infinite set.
Example 1.1.17 Z, N, R, C are infinite sets but Zn is a finite set.
Definition 1.1.18 A set X is said to be countable if either X is finite or there exists

a bijection N → X (in the latter case X is said to be infinitely countable and its
elements can be arranged as a sequence {xn }, n ∈ N).
Proposition 1.1.19 Every infinite set contains a countable set.
Example 1.1.20 The set Q is countable but the set R is not countable.
The notion of counting is extended by assigning to every set X (finite or infinite)

an object |X |, called the cardinal number or cardinality, defined in such a way that
|X | = |Y | iff there exists a bijection between the sets X and Y . If X is a finite set of
n elements, then |X | = n. The cardinal number d or c of an infinite set X asserts that
the set is countable or not. Hence |N| = |Q| = d but |R| = c.
Definition 1.1.21 Let α and β be the cardinal numbers of two disjoint sets X and Y ,
then α + β, αβ and β α are defined by α + β = |X ∪ Y |, αβ = |X × Y | and β α =
|Y X |, where Y X denotes the sets of all maps f : X → Y .
Proposition 1.1.22 For any cardinal numbers α, β and γ,
(i) (α + β) + γ = α + (β + γ) (associativity for addition);
(ii) (αβ)γ = α(βγ) (associativity for multiplication);
(iii) α + β = β + α (commutativity for addition);
(iv) αβ = βα (commutativity for multiplication);
(v) α(β + γ) = αβ + αγ (distributive property);
(vi) (αβ)γ = αγ β γ ;
(vii) α < 2α (Cantor’s Theorem);
(viii) α ≤ β and β ≤ α imply that α = β.
1.2 Groups and Fundamental Homomorphism Theorem
This section conveys some basic results of group theory which are used throughout
the book. Originally, a group was defined as the set of permutations (i.e., bijections)
on a nonempty set X with the property that combination (called composition) of two
permutations is also a permutation on X . Earlier definition of a group is generalized
to the present concept of an abstract group by a set of axioms.
Definition 1.2.1 A group G is a nonempty set G together with a binary operation
(called composition), that is, a rule that assigns to each ordered pair (a, b) in G × G,
an element of G, denoted by ab (or a ·b called a multiplication) such that
G(1) ab(c) = a(bc) for all a, b, c in G (associative law);
G(2) there exists an element e in G such that ae = ea = a for all a in G (existence
of identity);
G(3) for each a in G, there is an element a in G such that aa = a a = e (existence
of inverse).
Remark 1.2.2 In a group G, e is unique and for each a in G, a is also unique. The
element a denoted by a −1 , is called the inverse of a for each a ∈ G. In additive
notation, ab is written as a + b; e is as 0 (zero) and a −1 as −a.
A group G is said to be commutative (or abelian) if ab = ba for all a, b in G.
We usually use the term ‘abelian group’ when the composition law is in additive
notation. A group G is said to be finite if its underlying set G is finite; otherwise, it
is said to be infinite.
Example 1.2.3 Given a nonempty set X , let P(X ) denote the set of all permutations
(bijective mappings) on X . Then under usual composition of mappings P(X ) is a
group, called permutation group on X . In particular, if X contains only n elements,
then P(X ) is called the symmetric group on n elements, denoted by Sn .
1.2 Groups and Fundamental Homomorphism Theorem 5
Example 1.2.4 (General Linear Groups) GL (n, R) (GL (n, C)) is the group of all
invertible n × n real (complex) matrices under usual multiplication of matrices and
is called general linear group of order n over R(C).
Example 1.2.5 (Circle group) The set S 1 = {z ∈ C : |z| = 1} forms a group under
usual multiplication of complex numbers, called the circle group in C.
An arbitrary subset of a group forming a group, called a subgroup contained in a

larger group, sometimes creates interest and plays an important role in group theory
and algebraic topology.
Proposition 1.2.6 A nonempty subset H of a group G is a subgroup of G if and only

if ab−1 ∈ H for all a, b ∈ H .
Example 1.2.7 (i) The additive group of integers is a subgroup of the additive
group of real numbers. The additive group Z of integers is an example of
infinite group. On the other hand the group Zn is a finite group.
(ii) Given a group G, the center Z (G) of G, defined by Z (G) = {g ∈ G : gx =
xg for all x ∈ G} is subgroup of G.
Definition 1.2.8 For any subgroup H of an arbitrary group G, the set a H = {ah ∈
G : h ∈ H } is said to be a left coset of H in G for every a ∈ G. On the other hand,
for every a ∈ G, the right coset H a is defined by H a = {ha ∈ G : h ∈ H }.
Using the concept of cosets, Lagrange theorem establishes a relation between the
order of a finite group and orders of its subgroups.
Theorem 1.2.9 (Lagrange Theorem) The order of a subgroup of a finite group

divides the order of the group.
Remark 1.2.10 The converse of Lagrange’s theorem claims that if a positive integer
m divides the order n of a finite group G, then G contains a subgroup of order m.
But it is not true in general. For example, the alternating group A4 of order 12 has
no subgroup of order 6. However, under certain particular situations, converse of
Lagrange’s theorem is partially true. In its support, consider the following results.
Theorem 1.2.11 Let G be a finite group of order n and m be a positive divisor of n.

(i) If G is an abelian group, then corresponding to every positive divisor m of n,
G contains a subgroup of order m.
(ii) If m = p, is a prime integer, then G has a subgroup of order p.
(iii) If m is a power of a prime p, then G has a subgroup of order m (Sylow first
theorem.).
We now define a special map between groups, called a homomorphism which

preserves compositions of the groups.
Definition 1.2.12 Let G and H be groups. Then a map f : G → H is said to be a

homomorphism if f (x y) = f (x) f (y) for all x, y in G.
Particular homomorphisms carry special names having interesting properties.
Definition 1.2.13 Let f : G → H be a homomorphism of groups. Then

(a) f is said to be
(i) an epimorphism if f is surjective;
(ii) a monomorphism if f is injective;
(iii) an isomorphism if f is bijective;
(iv) an endomorphism if G = H ;
(v) an automorphism if G = H and f is an isomorphism.
(b) (i) the kernel of f , defined by ker f = {x ∈ G : f (x) = e H } is a subgroup
of G.
(ii) the image of f , defined by Im f = {y ∈ H : y = f (x) for some x ∈ G} is
a subgroup of H .
Remark 1.2.14 A homomorphism between groups preserves the identity element

and inverse elements.
Proposition 1.2.15 Let f : G → H be a group homomorphism. Then f is a

monomorphism if and only if ker f = {eG }.
We are now interested to present isomorphic replicas of an abstract group.
Definition 1.2.16 A subgroup H of a group is said to be normal if for all g ∈

G, g H g−1 ⊆ H, where g H g −1 = {ghg −1 : h ∈ H } (i.e., g H = H g for all g ∈ G).
Remark 1.2.17 Normal subgroups form an important class of subgroups in group

theory, which was first recognized by the French mathematician Evariste Galois
(1811–1832).
Example 1.2.18 If f : G → H is a group homomorphism, then ker f is a normal

subgroup of G.
We now construct a quotient group (or factor group) by a normal subgroup N of

a group using the relation a N = N a for each element a of the group.
Definition 1.2.19 Let N be a normal subgroup of a group G and G/N be the set of
all cosets of N in G. Then the set G/N is a group under the composition a H ·bH =
(ab)H . The group G/N is called quotient group or factor group of G by N .
Remark 1.2.20 If |G| denotes the order of a finite group G and N is a normal
subgroup of the group G, then |G/N | = |G|/|N |.
Using the fact that the kernel of a homomorphism is a normal subgroup, the
following theorem is proved.
1.2 Groups and Fundamental Homomorphism Theorem 7
Theorem 1.2.21 (Fundamental Homomorphism Theorem or First Isomorphism

Theorem) Let f : G → G be a group homomorphism. Then f induces an isomor-
phism f˜ : G/ker f → Im f, g ker f → f (g). In particular, if f is an epimorphism,
then f˜ is an isomorphism.
Finitely generated groups are very important in algebraic topology. Let G be a

group and S be a nonempty subset of G. Then the intersection of all subgroups of G
containing S is also a subgroup of G, denoted by S. It is the smallest subgroup of
G which contains S.
Definition 1.2.22 If S is a nonempty subset of a group G, then S is called the

subgroup generated by S. If S = G, then G is said to be generated by S. If S is
a finite set and the group G = S, then G is said to be finitely generated with S a
set of generators. In particular, if S = {x} and G = S, then G is said to be a cyclic
group and x is said to be a generator of G.
Theorem 1.2.23 Let G be a cyclic group. Then

(i) G is isomorphic to the group Z if and only if G is infinite.
(ii) G is isomorphic to the group Zn if and only if G is finite and |G| = n.
1.3 Group Representations, Free Groups, and Relations
This section conveys the concepts of group representations, free groups, and relations
which are used in subsequent chapters.
1.3.1 Linear Representation of a Group
A group representation describes an abstract group in terms of linear operators of

vector spaces (see Sect. 1.8). This subsection introduces the concept of group repre-
sentation.
Definition 1.3.1 Let G be a group and V be a vector space over a field F. If GL(V )
is the general linear group on V , then a representation of G on V is a group homo-
morphism
ψ : G → GL(V )
such that ψ(g1 g2 ) = ψ(g1 ) ◦ ψ(g2 ) for all g1 , g2 ∈ G. The vector space V is called the
representation space and dimension of V is called the dimension of the representation.
The homomorphism ψ is sometimes called a linear representation of the group G.
1.3.2 Free Groups and Relations
This subsection introduces the concepts of free groups and relations, which are used
in computation of fundamental groups and some other groups. The free groups used
in multiplication notation here are not necessarily abelian.
Definition 1.3.2 A subset X = {x j } of a group G with identity e is called a free set
of generators of G if every element g ∈ G − {e} is uniquely expressable as
g = x1i1 x2i2 . . . x nin (1.1)
where n is a positive integer and i k ∈ Z. We assume that x j = x j+1 for any j (i.e.,
no adjacent x j are equal). If i j = 1 for some i j , we write x 1j as x j . Again, if i j = 0
for some i j , the term x 0j is dropped from the expression of g.
Example 1.3.3 The expression g = a 5 b−7 c b8 is acceptable but the expression h =

a 5 a −7 b0 is not acceptable.
Definition 1.3.4 If a group G has a free set of generators, it is called a free group.
Given a set X , there exists a free group G such that X is a free set of generators of
G. Each element of X is called ‘letter’.The cardinality of a free set of generators of
G is called the rank of G.
Definition 1.3.5 The product
w = x1i1 x2i2 . . . x nin (1.2)
is called a word, where x j ∈ X and i j ∈ Z. If i j = 0 and x j = x j+1 , the word is

called reduced word.
Remark 1.3.6 It is always possible to reduce a word by finite steps.
Definition 1.3.7 A word with no letter is called an empty word denoted by 1.
Definition 1.3.8 The set of all reduced words forms a free group, called the free
group generated by X , denoted by F(X ), and under the multiplication is the juxta-
position of two words followed by reduction, the unit element is the empty word and
the inverse of w = x1i1 x2i2 . . . x nin is w −1 = xn−in . . . x 2−i2 x1−i1 .
Example 1.3.9 Let X = {x}. Then F(X ) ∼ = Z. An arbitrary group G is in general
defined by the generators and certain conditions (constraints) R on them. If {x j } is
the set of generators, then the conditions (constraints) commonly written as
R : x ij11 x ij22 . . . x ijnn = 1 (1.3)
are called relations.

Example 1.3.10 The cyclic group of order n generated by {x} satisfies the relation
R : x n = 1.
1.3 Group Representations, Free Groups, and Relations 9
Definition 1.3.11 A group G is defined by generators X = {x j } and relations R =

{Rk : Rk = 1} if G ∼
= F/N , where F is free on X and N is the normal subgroup of
F, generated by {Rk }. The ordered pair (X ; R) is called a presentation of the group
G, denoted by X ; R.
Example 1.3.12 (i) Zn = x : x n and Z = x : ∅.

(ii) Z ⊕ Z = x m y n : m, n ∈ Z is free abelian group generated by X = {x, y}. Then
x y = yx. Since x yx −1 y −1 = 1, there is a relation R = x yx −1 y −1 . Hence the
presentation of Z ⊕ Z is x, y : x yx −1 y −1 .
Definition 1.3.13 A group G is said to be finitely presented with a set S as its

generating set S and a set R of relations describing the group G if both the sets S
and R are finite and then G has the presentation as G = S : R.
Example 1.3.14 The cyclic group Zn has the presentation x : x n .
Example 1.3.15 The dihedral group Dn is completely determined by two generators

x, y and the defining relations x n = e, y 2 = e, (x y)2 = e and x k = e for 0 < k < n.
Hence Dn has the presentation x, y : x n , y 2 , (x y)2 .
Example 1.3.16 The quaternion group has the presentation x, y : x 4 , x 2 = y 2 ,

yx = x 3 y.
The concepts of direct product and direct sum of groups are frequently used in
algebraic topology.
Definition 1.3.17 Let G and K be two groups. The set P = G × K = {(g, k) : g ∈

G, k ∈ K } is a group under pointwise multiplication defined by (g, k)·(g , k ) =
(gg , kk ). This group is called the direct product of the groups G and H .
We generalize this definition as follows:
Definition 1.3.18 Let {G i : i ∈ I } be a family of groups. Then an element f ∈

i∈I
G i = G, is a map f : I → G i such that f (i) ∈ G i for i ∈ I . The set G = G i
i∈I
i∈I
forms a group under the composition defined by ( f g)(i) = f (i)·g(i) for all i ∈ I ,
where the right-hand multiplication is the usual multiplication in the group G i for
all i ∈ I . This group is called the direct product of the groups {G i : i ∈ I }. Let ei be
the identity element of G i , for all i ∈ I .

We now consider the subset G i of the product group G i , defined by
i∈I
i∈I
⊕G i = {(gi ) ∈ G i : gi = ei for all i except for a finite number of indices}.
i∈I
Definition 1.3.19 ⊕G i forms a subgroup of G i , called the direct sum of the given
i∈I
family {G i : i ∈ I } of subgroups. In particular, if I is a finite set, then the concepts
of direct product and direct sum coincide.
Free abelian groups in additive notation are used in algebraic topology.
Definition 1.3.20 Let G be an additive abelian group. G is said to be a free group

with a basis B if
b ∈ B, the cyclic group b is infinite; and

(i) for each
(ii) G = b (direct sum).
b∈B
Theorem 1.3.21 If G is a free abelian group with a basis B = {b1 , b2 , . . . , bn }, then

n is uniquely determined by G, and G is then said to be free abelian group of rank n.
Definition 1.3.22 An abelian group G is said to be a finitely generated abelian group

if G has a finite basis.
Remark 1.3.23 Basis of a finitely generated abelian group is not unique.
For example, {(1, 0), (0, 1)} and {(−1, 0), (0, −1)} are two different bases of the
group Z ⊕ Z.
Theorem 1.3.24 Any two bases of a free abelian group have the same cardinality.
This theorem leads to the following definition.
Definition 1.3.25 Let F be a free abelian group with a basis B. The cardinality of
B is called the rank of F. In particular, if F is finitely generated, then the number of
elements in a basis of F is the rank of F.
Example 1.3.26 (i) Z is a free abelian group of rank 1, finitely generated by 1

(or −1).
(ii) Z ⊕ Z is a free abelian group of rank 2.
(iii) Z2 is finitely generated by 1 but not free, because 1 + 1 = 0 shows that 1 is
not linearly independent.
Definition 1.3.27 An abelian group G has rank r (possibly infinite) if there exists a
free abelian subgroup F of G such that
(a) rank of F is r ; and
(b) the quotient group G/F is of finite order.
Theorem 1.3.28 Given a family of abelian groups {G i }i∈I , there exists an abelian
group G and a family of monomorphisms fi : G i → G such that G = ⊕ f i (G i ).
1.3 Group Representations, Free Groups, and Relations 11
1.3.3 Betti Number and Structure Theorem for Finite

Abelian Group
This subsection states some basic concepts and theorems such as fundamental the-
orem of finitely generated abelian group, Betti number, and structure theorem for
finite abelian group which are very key algebraic results used in algebraic topology
Theorem 1.3.29 (Fundamental theorem of finitely generated abelian groups) Every

finitely generated abelian group G (not necessarily free) can be expressed uniquely
as
r summands

∼
G = Z ⊕ Z ⊕ · · · ⊕ Z ⊕Zn ⊕ Zn ⊕ · · · ⊕ Zn ,
1 2 t
for some integers r, n 1 , n 2 , . . . , n t such that

(i) r ≥ 0 and n j ≥ 2 for all j; and
(ii) n i |n i+1 , for 1 ≤ i ≤ t − 1.
Definition 1.3.30 The integer r in Theorem 1.3.29 is called the free rank or Betti
number of the group G given by E. Betti (1823–1892) and the integers n 1 , n 2 , . . . , n t
are called invariant factors of G.
r summands

Remark 1.3.31 Z ⊕ Z ⊕ · · · ⊕ Z is a free abelian group of rank r .
Theorem 1.3.32 (Structure Theorem for finite abelian groups) Any nonzero finite
abelian group G can be expressed uniquely as G ∼
= Zn 1 ⊕ Zn 2 ⊕ · · · ⊕ Zn t such that
n i |n i+1 , for 1 ≤ i ≤ t − 1.
Theorem 1.3.33 Two finite abelian groups are isomorphic if and only if they have
the same invariant factors.
1.4 Exact Sequence of Groups
This section conveys some results of exact sequences of groups and their homomor-
phisms which are frequently applied in algebraic topology. For this section the book
Adhikari and Adhikari (2014) is referred.
Definition 1.4.1 A sequence of groups and their homomorphisms
f n+1 fn
··· −
→ G n+1 −−−−−→ G n −−−−→ G n−1 −
→ ···
is said to be exact if ker f n = Im f n+1 for all n. Clearly, f n ◦ f n+1 = 0 for an exact
sequence.
f
Proposition 1.4.2 (i) In the short exact sequence 0 −
→ G −−−→ K , f is a
monomorphism.
f
(ii) In the short exact sequence G −−−→ K −
→ 0, f is an epimorphism.
f
(iii) The sequence 0 − → G −−−→ K − → 0 is exact if and only if f is an isomor-
phism;
(iv) If G is a normal subgroup of K and i : G → K is the inclusion map (i.e.,
i(x) = x for all x ∈ G), then the sequence
i p
0−
→ G −−−→ K −−−→ K /G −
→0
is an exact sequence, where 0 denotes the trivial group and p is the natural
homomorphism defined by p(x) = x + G for all x ∈ K .
Proposition 1.4.3 Given exact sequences of groups and homomorphisms
fi gi
→ G i −−−→ K i −−−→ Hi −
0− →0
for each element i ∈ I , the sequence

⊕ fi ⊕gi
0−
→ G i −−−−→ K i −−−−→ Hi −
→0
i∈I i∈I i∈I
is also exact
Theorem 1.4.4 (The Five Lemma) Let the diagram in Fig. 1.1 of groups and homo-
morphisms be commutative with two exact rows. If α is an epimorphism, λ is a
monomorphism and β, δ are isomorphisms, then γ is an isomorphism.
Proposition 1.4.5 Given an exact sequence of abelian groups and homomorphisms
f g
→ G −−−→ K −−−→ H −
0− →0
if h : H → K is a homomorphism such that g ◦ h : H → H is the identity automor-

phism of H , then K ∼
= G ⊕ H.
Fig. 1.1 Five lemma f g h k

A /B /C /D /E
diagram
α β γ δ λ
f g h
k

A / B / C / D / E
1.4 Exact Sequence of Groups 13
Remark 1.4.6 It is not true that all sequences of the form
0−
→G−
→K −
→H−
→0
are always exact. But if H is a free abelian group, then this is true.
Proposition 1.4.7 (a) Given abelian groups G, H, K and homomorphisms f, g

f g
with G −−−→ K ←−−− H , where f is an epimorphism and H is a free abelian
group, there exists a homomorphism h : H → G such that f ◦ h = g.
(b) Given an exact sequence of abelian groups and homomorphisms
f g
→ G −−−→ K −−−→ H −
0− →0
where H is a free abelian group, the sequence splits and K ∼

= G ⊕ H.
Remark 1.4.8 For the three lemma and four lemma of groups, see the Sect. 1.8.
1.5 Free Product and Tensor Product of Groups
This section presents the concepts of two important products such as free product
and tensor product of groups which are frequently used in algebraic topology.
1.5.1 Free Product of Groups
This subsection conveys the concept of free product of groups.
Definition 1.5.1 Let G and H be groups (not necessarily abelian). Their free product
denoted by G ∗ H is a group satisfying the following condition:
if there are homomorphisms i and j such that given a pair of homomorphisms
f : G → K and g : H → K for any group K , there exists a unique homomorphism
h : G ∗ H → K making the diagram in Fig. 1.2 commutative.
For example, Z ∗ Z is a free group (of rank 2).
Fig. 1.2 Free product G ∗ H G; ∗ HcG

ww GGG j
www
i GG
ww GG
ww G

G GG h
w
H
GG w
GG ww
G
f GG wwwg
# {ww
K
Remark 1.5.2 An alternative description of free product G ∗ H may be given with

the help of presentations of groups G and H .
Definition 1.5.3 Let G = X : R and H = Y : S be presentations of the groups
G and H in which the sets X and Y are generators (and thus the relations R and S)
are disjoint. Then a presentation of G ∗ H is given by
G ∗ H = X ∪ Y : R ∪ S.
1.5.2 Tensor Product of Groups
This subsection conveys the concept of tensor products of groups.

Definition 1.5.4 Let G and H be two abelian groups. Their tensor product denoted
by G ⊗ H is the group defined as the abelian group generated by all pairs of the form
(g, h) with g ∈ G, h ∈ H satisfying the bilinearity relations (g + g , h) = (g, h) +
(g , h) and (g, h + h ) = (g, h) + (g, h ).
For example, Zm ⊗ Zn = Z(m,n) , where (m, n) is the gcd of m and n: on the other
hand, this tensor product is 0, if m and n are relatively prime.
1.6 Torsion Group
This section introduces the concept of torsion group. Let G be an abelian group,
and p be a given prime integer. If f : G → G be the homomorphism defined by
f (x) = px, then there exists an exact sequence
i f h
→ ker f −−−→ G −−→ G −−→ G ⊗ Z p −
0− → 0, (1.4)
where h(x) = x ⊗ 1.
Definition 1.6.1 The sequence (1.4) is called a short free resolution of G and ker f
is called torsion group written Tor (G, Z p ).
Definition 1.6.2 For each abelian group G there exists an exact sequence
i
→ R −−−→ F −
0− →G−
→0 (1.5)
with F free abelian. For any abelian group H , the torsion product of G and H is
defined by Tor (G, H ) = ker (i ⊗ I B ).
Definition 1.6.3 (Torsion subgroup) For any abelian group G,
T (G) = {x ∈ G : x has finite order}

1.6 Torsion Group 15
is a subgroup of G, called its torsion subgroup.

Remark 1.6.4 If for every homomorphism f : G → G of abelian groups, T ( f ) is
defined by
T ( f ) = f |T (G) ,
then T defines a functor from the category of abelian groups and their homomor-
phisms into itself.
1.7 Actions of Groups
This section presents an action of a group G on a nonempty set X which assigns

to each element of G a permutation on the set X and unifies the historical concept
of the group of transformations and the modern axiomatic concept of a group. The
concept of group actions is useful in the study of algebraic topology.
Definition 1.7.1 A group G with identity e, is said to act on a nonempty set X from
the left (or X is said to be a left G-set), if there is a map σ : G × X → X, written as
σ(g, x) = g·x (or gx) such that for all x ∈ X and for all g1 , g2 ∈ G,
(i) e·x = x;
(ii) (g1 g2 )·x = g1 ·(g2 ·x).
Then σ is said to be a left action of G on X and X is said to be a left G-set. Similarly,
a right G-set and a right action of G on X are defined.
Remark 1.7.2 If X is a left G-set, then x ·g = g −1 ·x defines a right G-set structure
on X . As there is a bijective correspondence between left and right G-set structures
it is sufficient to study only one of them.
There may exist different actions of G on X .
Example 1.7.3 Let H be a subgroup of a group G and h ∈ H, x ∈ G. Then
(i) h ·x = hx (left translation);
(ii) h ·x = hxh −1 (conjugation by h)
are both actions of H on G.
Theorem 1.7.4 Let X be a left G-set. Then for any g in G, the map X → X, x → g·x
is a permutation on X .
Remark 1.7.5 This theorem identifies the notion of a G-set with the notion of a
representation of G by permutations on X .
Definition 1.7.6 Given a left G-set X the quotient set X/ρ by an equivalence relation
ρ on X defined by (x, y) ∈ ρ iff g·x = y for some g ∈ G, is called an orbit set and
[x] is called an orbit of x ∈ X , denoted by or b(x) or Gx. The action is said to be
transitive if or b(x) = X .
Definition 1.7.7 Let X be a left G-set. For each x ∈ X , the set G x = {g ∈ G : g·x =
x} is a subgroup of G, called the isotropy group or the stabilizer group of x.
Remark 1.7.8 The isotropy group G x is a subgroup of G for every x ∈ X . For any
two elements in the same isotropy group are conjugate. If G x = {e} for all x ∈ X ,
then the action of G on X is said to be free.
Theorem 1.7.9 Let X be a left G-set. Then |or b(x)| is the index [G : G x ] for every
x ∈ X . In particular, if G acts on X transitively, then |X | = [G : G x ].
Theorem 1.7.10 Let G be a group. Then

(i) for each g ∈ G, conjugation σg : G → G, x → gxg −1 for all x ∈ G is an auto-
morphism of G.
Definition 1.7.11 σg defined in Theorem 1.7.10 is called an inner automorphism of

G induced by g.
1.8 Modules and Vector Spaces
This section conveys some basic concepts of modules and vector spaces needed for
subsequent chapters. Modules and vector spaces play an important role in algebraic
topology, specially in the study of homology and cohomology theories. The concept
of modules over a ring is a generalization of the concept of abelian groups (which
are modules over the ring Z of integers) and vector spaces are modules over a field
(or a division ring).
1.8.1 Modules
This subsection begins with basic concepts of module theory. We may define a module
as an action of a ring on an additive abelian group as follows.
Definition 1.8.1 Let R be a ring. A(left) R-module M is an additive abelian group M

together with an action (called scalar multiplication) μ : R × M → M, (r, x) → r x
such that for all r, s ∈ R and x, y ∈ M,
M(i) r (x + y) = r x + r y;
M(ii) (r + s)x = r x + sx;
M(iii) r (sx) = (r s)x.
If R has an identity element 1, then an R-module M is said to be an unitary R-module
if
M(iv) 1x = x for all x ∈ M.
1.8 Modules and Vector Spaces 17
If R is a field F (or a division ring), then the F-module M is called an F-vector

space or a vector space over F.
A ring R itself may be considered as an R-module by taking scalar multiplication
to be the usual multiplication of the ring R.
1.8.2 Direct Sum of Modules
Given R-modules M1 , M2 , . . . , Mt , forget for the time being the scalar multiplication
and form the direct sum of abelian groups M1 , M2 , . . . , Mt . Then M1 ⊕ M2 ⊕ · · · ⊕
Mt is an R-module, called the direct sum under the scalar multiplication defined by
r (m 1 , m 2 , . . . , m t ) = (r m 1 , r m 2 , . . . , r m t ). In particular, the direct sum of t-copies
of R, denoted by R (t) , is called a free R-module. Let 1 ∈ R. If ei ∈ R (t) is the t-
copies having 1 in the ith place and 0 elsewhere, then every element x ∈ R (t) can be
expressed uniquely as
x= ri ei , ri ∈ R.
The concepts of direct sum and direct product of a finite number of R-modules
coincide.
1.8.3 Tensor Product of Modules
Tensor product of abelian groups and R-modules are frequently used in algebraic
topology.
Let M, N , T be given R-modules. A map f : M × N → T is said to be R-bilinear
(or bilinear) if
f (r1 m 1 + r2 m 2 , n) = r1 f (m 1 , n) + r2 f (m 2 , n)
and
f (m, r3 n 1 + r4 n 2 ) = r3 f (m, n 1 ) + r4 f (m, n 2 )
for all ri ∈ R, for all m, m i ∈ M and for all n, n i ∈ N .

Let R be a commutative ring with 1 and M, N be R-modules. Consider the free
R-module F on the set M × N , i.e., the elements
of F can be uniquely expressed as
a finite linear combinations of the form: ri j (xi , y j ), where xi ∈ M, y j ∈ N and
i, j
ri j ∈ R. Let S be the submodule of F generated by all elements of F of the form:
(x + x , y) − (x, y) − (x , y);
(x, y + y ) − (x, y) − (x, y );
(r x, y) − r (x, y) and
(x, r y) − r (x, y).
Then the quotient module F/S is called the tensor product of M and N , denoted by
M ⊗ N.
If π : F → F/S is the canonical projection and i : S → F, then ψ = π ◦ i : S →
F/S satisfies the relations
ψ((x + x , y) − (x, y) − (x , y)) = 0;
ψ((x, y + y ) − (x, y) − (x, y )) = 0;
ψ((r x, y) − r (x, y)) = 0;
ψ((x, r y) − r (x, y)) = 0.
If we write ψ(x, y) = x ⊗ y, then the elements x ⊗ y of M ⊗ N satisfy the follow-
ing identities:
(x + x ) ⊗ y = x ⊗ y + x ⊗ y;
x ⊗ (y + y ) = x ⊗ y + x ⊗ y ;
(r x) ⊗ y = r (x ⊗ y) = x ⊗ r y.
Clearly, the map ψ : M × N → M ⊗ N , (x, y) → x ⊗ y is R-bilinear, and M ⊗ N
satisfies the following universal properties:
Proposition 1.8.2 Let M and N be two R-modules. Given an R-module T and

an R-bilinear map f : M × N → T, there exists a unique R-homomorphism g :
M × N → T such that the diagram in Fig. 1.3 is commutative, i.e., g ◦ ψ = f.
Proposition 1.8.3 Let M, N and T be R-modules. Then there exist isomorphisms

such that
(a) M⊗N ∼ = N ⊗ M;
(b) (M ⊗ N ) ⊗ T ∼
= M ⊗ (N ⊗ T );
(c) R⊗M ∼ = M;
(d) (M ⊕ N ) ⊗ T ∼
= M ⊗ T ⊕ N ⊗ T.
Theorem 1.8.4 (Structure theorem for finitely generated modules over a principal
ideal domain) Let R be a principal ideal domain and M be a finitely generated
R-module. Then
r copies

M∼
= R ⊕ R ⊕ · · · ⊕ R ⊕R/ q1 ⊕ R/ q2 ⊕ · · · ⊕ R/ qt
for some integer r ≥ 0 and nonzero nonunit elements q1 , q2 , . . . , qt of R are such

that q1 |q2 | . . . |qt (i.e., q1 divides q2 , q2 divides q3 , etc.)
Fig. 1.3 Commutativity of ψ

M ×N /M ⊗N
H HH v
the triangle for uniqueness of HH v
H vg
g f HHH v
# {v
T
The elements q1 , q2 , . . . , qt uniquely determined up to units are called invariant

factors of M and the integer r is called the free rank or Betti number of M.
Corollary 1.8.5 (Fundamental theorem of finitely generated abelian groups) Let G

be a finitely generated abelian group. Then
r copies

G∼
= Z ⊕ Z ⊕ ··· ⊕ Z.
Theorem 1.8.6 (The three lemma) Let the diagram in Fig. 1.4 of R-modules and
their homomorphisms be commutative with two rows of exact sequences.
(i) If α, γ and f are R-monomorphisms, then β is also an R-monomorphism.
(ii) If α, β and g are R-epimorphisms, then β is also an R-epimorphism.
(iv) If α, γ are R-isomorphisms, f is an R-monomorphism and g is an R-
epimorphism, then β is an R-isomorphism.
Theorem 1.8.7 (The four lemma) Let the diagram in Fig. 1.5 of R-modules and
their homomorphisms be commutative with two rows of exact sequences.
(i) If α, γ are R-epimorphisms and δ is an R-monomorphism, then β is an R-
epimorphism.
(ii) If α is an R-epimorphism and β, δ are R-monomorphism, then γ is an R-
monomorphism.
Theorem 1.8.8 (The five lemma) Let the diagram in Fig. 1.6 of R-modules and their
homomorphisms be commutative with exact rows of exact sequences. If α, β, δ, λ are
R-isomorphisms, then γ is also an R-isomorphism.
Fig. 1.4 Three lemma f

/B g
/C
A
diagram
α β γ
f g
A / B / C
Fig. 1.5 Four lemma f

/B g
/C h /D
A
diagram
α β γ δ
f g h

A / B / C / D
Fig. 1.6 Five lemma f g h k

A /B /C /D /E
diagram
α β γ δ λ
f g h
k

A / B / C / D / E
1.8.4 Vector Spaces
This subsection conveys some results of vector spaces. A vector space is a combi-
nation of an additive abelian group and a field (division ring) and interlinked by an
external law of composition and is a module over a field.
Definition 1.8.9 A vector space or a liner space over a field F is an additive abelian
group V together with an external law of composition (called scalar multiplication)
μ : F × V → V , the image of (α, v) under μ is denoted by αv such that for all
α, β ∈ F and u, v ∈ V
V(1) 1v = v, where 1 is the multiplicative identity element in F;
V(2) (αβ)v = α(βv);
V(3) (α + β)v = αv + βv;
V(4) α(u + v) = αu + αv.
Definition 1.8.10 A vector space V is said to be a direct sum if its subspaces
V1 , V2 , . . . , Vn denoted by V = V1 ⊕ V2 ⊕ · · · ⊕ Vn if every element v of V can
be expressed uniquely as v = v1 + v2 + · · · + vn , vi ∈ Vi , i = 1, 2, . . . , n.
Definition 1.8.11 Let V be a vector space over a field F and U be a subspace of V .
Then (U, +) is a subgroup of the abelian group (V, +). The abelian group (V /U, +)
is a vector space over F, under the scalar multiplication
μ : F × V /U → V /U, (α, v + U ) → αv + U.
Definition 1.8.12 The vector space V /U is called the quotient space of V by U and
the map p : V → V /U, v → v + U is called the canonical homomorphism.
Theorem 1.8.13 (Existence Theorem) Every vector space V over F has a basis.
Theorem 1.8.14 Let V be a nonzero vector space over F.
(i) Let S be an arbitrary linearly independent subset of V . If S is not a basis of
V , then S can be extended to a basis of V .
(ii) If V has a finite basis consisting of n elements, then any other basis of V is
also finite consisting of n elements.
(iii) Cardinality of every basis of V is the same.
This theorem leads to define the dimension of a vector space.
Definition 1.8.15 Let V be a nonzero vector space over F. The cardinality of every
basis of V is the same and this common value is called the dimension of V denoted
by dimV. If V has finite basis, then V is said to be finite dimensional; otherwise it
is said to be infinite dimensional. If V = {0}, then V is said to be 0-dimensional.
Definition 1.8.16 Let V and W be vector spaces over the same field F. A linear
transformation T : V → W is a mapping such that
L(1) T (x + y) = T (x) + T (y), ∀ x, y ∈ V (additive law);

L(2) T (αx) = αT (x), ∀ x ∈ V and ∀ α ∈ F (homogeneity law).
Conditions L(1) and L(2) can be combined together to obtain an equivalent

condition
L(3) T (αx + β y) = αT (x) + βT (y), ∀ α, β ∈ F and ∀ x, y ∈ V .
A linear transformation T : V → W is said to be an isomorphism if T is a bijec-
tion.
Proposition 1.8.17 Given a vector space V of dimension n, the set L(V, V ) of
all linear transformations (linear operators) T : V → V is also a vector space of
dimension n 2 .
Definition 1.8.18 The group GL(V) of all nonsingular linear operators T : V → V
is called a general linear group over V .
Remark 1.8.19 The group GL(V) is used to define a linear representation of a group
(see Sect. 1.3).
Theorem 1.8.20 (First Isomorphism Theorem) Let T : U → V be a linear trans-
formation. Then the vector spaces U/ker T and ImT are isomorphic.
Linear transformations and matrices are closely related.
Proposition 1.8.21 Given an m × n matrix M over R, the map TM : Rn → Rm ,
X → M X , is a linear transformation where X is viewed as a column vector of Rm .
Conversely, if B = {E 1 , E 2 , . . . , E n } is a set of unit column vectors of Rn , then there
is a linear transformation T : Rn → Rm with T (E j ) = M j ; where M j is a column
vector in Rn such that M is the matrix whose column vectors are M1 , M2 , . . . , Mn .
Definition 1.8.22 Let V be an n-dimensional vector space over F and T : V → V
be a linear transformation. An element λ in F is called an eigenvalue of T if there
exists a nonzero vector v ∈ V such that T (v) = λv. This vector v (if it exists) is
called an eigenvector of T corresponding to the eigenvalue λ.
For a square matrix over F we have an analog of this definition.
Definition 1.8.23 A real division algebra is a finite dimensional real vector space
with a bilinear multiplication having a both-sided identity element and satisfying the
condition that each nonzero element has a both-sided multiplicative inverse.
Examples of real division algebra:
Example 1.8.24 (i) The real numbers R form a real division algebra.
(ii) The complex numbers C form a real division algebra.
(iii) The real quaternions H form a real division algebra.
(iv) The Cayley numbers form a real division algebra (an eight-dimensional non-
associative algebra).
Remark 1.8.25 J.F.Adams proved that there are no other examples of real division
algebra (see Chap. 17).
1.9 Euclidean Spaces and Some Standard Notations
In mathematical problems, subspaces of an n-dimensional Euclidean space arise

frequently. Such spaces are used both in theory and application of topology. Some
standard notations used throughout the book are given.
∅ : empty set
Z : ring of integers (or set of integers)
Zn : ring of integers modulo n
R : field of real numbers
C : field of complex numbers
Q : field of rational numbers
H : division ring of quaternions

n 2 n
Rn : Euclidean n-space, with x = xi and x, y = xi yi
i=1 i=1
for x = (x 1 , x2 , . . . , xn ) and y = (y1 , y2 , . . . , yn ) ∈ Rn .
n
C : complex n-space
I : [0, 1]
I˙ : {0, 1} ⊂ I
In : n-cube = {x ∈ Rn : 0 ≤ xi ≤ 1 for 1 ≤ i ≤ n} for x = (x1 , x2 , . . . , x n )
D n
: n-disk or n-ball = {x ∈ Rn : x ≤ 1}
S n
: n-sphere = {x ∈ Rn+1 : x = 1} = ∂ D n+1 (the boundary of the
(n + 1)-disk D n+1 )
R P : real projective space = quotient space of S n with x and −x identified for
n
all x ∈ S n
CP : complex projective space = space of all complex lines through the origin
n
in the complex space Cn+1

: disjoint union of sets or spaces
×, : product of sets, groups, modules, or spaces
∼
= : isomorphism
≈ : homeomorphism
iff : if and only if
X ⊂ Y or Y ⊂ X : set-theoretic containment (not necessarily proper)
1.10 Set Topology
It is assumed that the readers are familiar with the basic concepts of set topology
used throughout the book. However, some of them are given in brief.
1.10 Set Topology 23
1.10.1 Topological Spaces: Introductory Concepts
Definition 1.10.1 Let X be a set. It is called a topological space if there is a family

τ of subsets of X , called open sets such that
O(1) the union of any number of open sets is an open set;
O(2) the intersection of a finite number of open sets is an open set;
O(3) empty set ∅ is an open set;
O(4) the whole X is an open set.
Sometimes we write the topological space as (X, τ ) to avoid any confusion regard-
ing the topology τ .
Example 1.10.2 (i) (Trivial or indiscrete topology) The two subsets ∅ and the whole
set X constitute a topology of X , called trivial topology.
(ii) (Discrete topology) The family of all subsets of X constitutes a topology of X ,
called the discrete topology of X . This topology is different from trivial topology,
if X has more than one element.
Definition 1.10.3 Let (X, τ ) be a given topological space. A family of open sets
is said to form an open base (basis) of the topology τ if every open set (relative to
the topology τ ) is expressible as the union of some sets belonging to .
Remark 1.10.4 The unions of all sub-collections of an open base of a topology

τ constitute the topology τ . Thus the topology τ is completely determined by an
open base . But there may exists different open bases for a particular topology.
For example, any topology τ always forms an open base of itself. On the other hand
for the discrete topology, the subsets consisting of one point only also form an open
base.
Example 1.10.5 (Natural topology) The null set ∅ and all open intervals (a, b), where
a and b are rational numbers that form a base of a topology of the set of real numbers
R. This topology is called the natural topology or usual topology of R and the set R
endowed with this topology is called the real number space.
Definition 1.10.6 The direct product (or product) X × Y of topological spaces has
a natural topology, called the product topology: a subset U ⊂ X × Y is open if U is
the union of the sets of the form U1 × U2 , where U1 is open in X and U2 is open in
Y.
Remark 1.10.7 Let X, Y , and Z be topological spaces. Then a function f : X ×

Y → Z may be considered as a function f (x, y) of two variables, with values in Z
and f is continuous if it is continuous jointly in both variables x and y.
Definition 1.10.8 Let (X, τ ) be a topological space. A subset V of X is said to be

a neighborhood (abbreviated nbd) of a point p ∈ X if exists an open set U ∈ τ such
that p ∈ U ⊂ V .
Definition 1.10.9 (X, τ ) be a topological space and A be a subset of X . A point

p ∈ X is said to be limit point of the set A if every nbd of p intersects A in at least
one point other than p. The set formed by all the limit points of A is called the derived
set of A, denoted by A .
Definition 1.10.10 Let A be a subset of a topological space (X, τ ). The union of

all open sets of X contained in A is called the interior of A, denoted by Int(A). It is
sometime denoted by Å or by simply Int A.
Remark 1.10.11 The interior of A is the set Int(A) consisting of all points a ∈ A for
which A is a nbd of a.
Definition 1.10.12 A subset A of a topological space X is said to be dense in X if

for any point x ∈ X , any nbd of x contains at least one point from A (i.e., A has
nonempty intersection with every nonempty open subset of X ). In other words, A is
said to be dense in X if the only closed subset of X containing A is X itself.
Remark 1.10.13 The real number space R with the natural topology has the rational
numbers Q as a countable dense subset. This implies that the cardinality of a dense
subset of a topological space may be strictly smaller than the cardinality of the space
itself.
It is sometimes convenient to define a topology on a set by the axioms of closed

sets.
Definition 1.10.14 A subset F ⊂ X of a topological space is said to be closed if

X \ F (= X − F) is open.
Remark 1.10.15 A topological space can also be defined in terms of closed sets.
Definition 1.10.16 Let X be a set. It is called a topological space if there is a family

of subsets of X , called closed sets such that
C(1) the union of finitely many closed sets is a closed set;
C(2) the intersection of any number of closed sets is a closed set;
C(3) X is a closed set;
C(4) empty set ∅ is a closed set.
Definition 1.10.17 Let (X, τ ) be a topological space and A be a subset of X . The

intersection of all closed sets of X containing A is called the closure of A, denoted
by Ā. If Ā = X , then A is a dense set.
Remark 1.10.18 Let (X, τ ) be a topological space. Let A ⊆ X . Then the closure Ā
is given by Ā = A ∪ A . where A is the derived set of A.
Example 1.10.19 (i) The set of all rational numbers Q is dense in R with usual
topology.
(ii) The set of all irrational numbers is also dense in R with usual topology.
Remark 1.10.20 Let (X, τ ) be a topological space and A be a subset of X . Then

Int(A) = X − (X − A).
Every subset Y of a topological space X is itself a topological space in a natural

way.
Definition 1.10.21 Let X be a topological space and Y be a subset of X . A set U ⊂ Y

is defined to be open in Y if there exists an open set V ⊂ X such that V ∩ Y = U .
This topology on Y is called the relative topology induced by the topology of X on Y .
1.10.2 Homeomorphic Spaces
This subsection presents the concept of a homeomorphism in topology which is a

bijective function that preserves topological structure involved and is analogous to
the concept of an isomorphism between algebraic objects such as groups or rings,
which is a bijective map that preserves the algebraic structures involved. Its precise
definition is now given.
Definition 1.10.22 Let X and Y be topological spaces. A function 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.
A continuous function f : X → Y is said to be a homeomorphism if f is bijective
and f −1 : Y → X is also continuous.
Remark 1.10.23 For a homeomorphism f : X → Y , both f and f −1 are continuous

means that f not only sends points of X to points of Y in a (1-1) manner, but f also
sends open sets of X to open sets of Y in a (1-1) manner. This asserts 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. For example, if f : X → Y is a homeomorphism,
then X is compact (or connected) iff Y is compact (or connected).
Remark 1.10.24 Let X and Y be topological spaces. A bijective function f : X → Y

may not be continuous. For example, let R be the set of real numbers in its usual
topology and Rl be the same set in the lower topology. Then the identity function
f : R → Rl , x → x is a bijection but not continuous.
Example 1.10.25 (i) The open interval (0, 1) and the real line R with usual topol-
ogy are homeomorphic.
(ii) A homeomorphism f : (0, 1) → R cannot be extended to I = [0, 1].
(iii) The open ball B = {x = (x1 , x2 ) ∈ R2 : x < 1} is homeomorphic to the
whole plane R2 .
(iv) The open square A = {(x, y) ∈ R2 : 0 < x, y < 1} is homeomorphic to the
open ball B defined in (iii).
(v) The cone A = {(x, y, z) ∈ R3 : x 2 + y 2 = z 2 , z > 0} is homeomorphic to the
plane R 2 .
(vi) Let S n be the n-sphere defined by S n = {x ∈ R n+1 : x = 1, n ≥ 1}, N =

(0, 0, . . . , 1) ∈ Rn+1 be the north pole of S n and S = (0, 0 . . . , −1) ∈ Rn+1
be the south pole of S n . Then
(a) S n − S is homeomorphic to S n − N ;
(b) The stereographic projection f : S n − N → Rn , defined by f (x) =
1
(x , x , . . . , xn ), for every x = (x1 , x2 , . . . , xn+1 ) ∈ S n − N is a
1−xn+1 1 2
homeomorphism.
(vii) A circle minus a point is homeomorphic to a line segment, and a closed arc
is homeomorphic to a closed line segment.
[Hint:
(i) The function f : (0, 1) → R, x → log 1−x x
is a homeomorphism.
(ii) Use the fact that any continuous function defined on [0, 1] must be bounded.
(iii) The function f : B → R2 , x → (1−x) x
is a homeomorphism with its inverse
g : R → B, x → (1+x) .
2 x
(iv) Using (i) there exists a homeomorphism f : (0, 1) → R. Consider the func-
tions g : A → R2 , (x, y) → ( f (x), f (y)) and h : R2 → A, (x, y) →
( f −1 (x), f −1 (y). Then g is a homeomorphism. Now use the result that
A ≈ R2 and R2 ≈ B.
(v) Consider the functions f : A → R2 , (x, y, z) → (x, y) and g : R2 → A,
(x, y) → (x, y, x + y ). Then f is a homeomorphism with g as its inverse.
2 2
(vi) (a) Let r : S n − S → S n − N be the reflection map defined by r (x1 , x2 , . . . ,

xn+1 ) = (x1 , x2 , . . . , −x n+1 ). Then r is a homeomorphism.
(b) Let g : Rn → S n − N be the function defined by g(y1 , y2 , . . . , yn ) =
(t y1 , t y2 , . . . , t yn , 1 − t), where t = (1+y 2 +y 2 2+···+y 2 )1/2 ∈ R. Then f is
1 2 n
a homeomorphism with g as its inverse.]
Theorem 1.10.26 Let I n = {(x1 , x2 , . . . , xn ) ∈ Rn : 0 ≤ xi ≤ 1} be the n-cube and

Int I n = {(x1 , x2 , . . . , xn ) : 0 < xi < 1} be the interior of I n and ∂ I n = I˙n = I n −
Int I n be the boundary of I n . Then I n is homeomorphic to the n-ball B n in Rn and
under this homeomorphism ∂ I n = I n − Int I n corresponds to S n−1 .
Proof Let Iˆn = {(x1 , x2 , . . . , xn ) : −1 ≤ xi ≤ 1}. Then there exists a homeomor-

phism f : Iˆn → I n . Under this homeomorphism ∂ I n corresponds to
∂ Iˆn = {(x1 , x2 , . . . , xn ) : −1 ≤ xi ≤ 1 and xi = ±1 for some i}.
Define f 1 : Iˆn → B n and g1 : B n → Iˆn by
i |)
f 1 (x1 , x2 , . . . , xn ) = max(|x
x
(x1 , x 2 , . . . , x n ), if x = 0
= 0, if x = 0
||x||
and g1 (x1 , x2 , . . . , xn ) = max(|x i |)
(x1 , x2 , . . . , xn ), if x = 0
= 0, if x = 0.
√
The continuity at 0 follows from the fact that max(|xi |) ≤ ||x|| ≤ n max (|xi |).
Clearly, f 1 and g1 are inverses to each other. ❑
1.10.3 Metric Spaces
This subsection conveys the concept of metric spaces with special reference to Banach
spaces.
Definition 1.10.27 A nonempty set X is said to have a metric or a distance function
f : X × X → R if for every pair of elements x, y in X
(i) d(x, y) ≥ 0, equality holds iff x = y;
(ii) d(x, y) = d(y, x);
(iii) d(x, y) + d(y, z) ≥ d(x, z) for all z ∈ X .
d(x, y) is called the distance between x and y and the pair (X, d) is called a metric
space or X is said to be metrized by d.
A metric space X can be made into a topological space in a natural way by defining
as open sets all unions of the open balls β (x) = {y ∈ X : d(x, y) < }, for x ∈ X
and > 0.
We define neighborhoods and limit points in the usual way.
Definition 1.10.28 Let (X, d) be a metric space. A Cauchy sequence in X is a
function f : N → X such that for every positive real numbers , there exists a positive
integer m such that d( f (i), f ( j)) < for all i > m and j > m.
Definition 1.10.29 A complete metric space is a metric space in which every
sequence is convergent.
Example 1.10.30 [0, 1] is a complete metric space but (0, 1] is not so.
Definition 1.10.31 A normed linear space is a vector space X on which a real-valued
function || || : X → R is called a norm function is defined such that
N(1) ||x|| ≥ 0 and ||x|| = 0 iff x = 0;
N(2) ||x + y|| ≤ ||x|| + ||y||;
N(3) ||αx|| = |α|| |x|| for x, y ∈ X and α ∈ R or C.
Remark 1.10.32 A normed linear space is a metric space with respect to the metric
induced by the metric defined by d(x, y) = ||x − y||.
Definition 1.10.33 A Banach space X is a normed linear space which is complete
as a metric space, i.e., every Cauchy sequence in X is convergent.
Definition 1.10.34 A Hilbert space is a complete Banach space X in which function
, : X × X → C is defined satisfying the following conditions:
H(1) αx + β y, z = α x, z + β y, z;
H(2) x, y = y, x;
H(3) x, x = ||x||2 .
Remark 1.10.35 A Hilbert space is a complex Banach space whose norm is defined
by an inner product.
1.10.4 Connectedness and Locally Connectedness
This subsection conveys the concepts of connectedness and locally connectedness.
Definition 1.10.36 A topological space X is said to be connected if the only sets

which are both open and closed are ∅ and X .
Remark 1.10.37 Connectedness of topological spaces is an important topological

property and is characterized by the following theorem.
Theorem 1.10.38 A topological space X is connected iff it is not the union of two
disjoint nonempty sets.
Definition 1.10.39 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 1.10.40 The Euclidean space Rn is connected and locally connected for
all n ≥ 1.
Remark 1.10.41 The continuous image of a locally connected space may not be
locally connected.
Definition 1.10.42 A path in a topological space X is a continuous map f : I → X

from the closed unit interval I to X .
Definition 1.10.43 A topological space X is said to be path-connected, if any two

points of X can be joined by a path.
Remark 1.10.44 A path-connected space is connected. A connected open subset of

a Euclidean space is path-connected.
Example 1.10.45 For n > 0, the n-sphere S n is path-connected.
Definition 1.10.46 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 1.10.47 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.
1.10.5 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 1.10.48 An open covering of a topological space X is a family of {Ui }

of open sets of X , whose union is the whole set X .
Definition 1.10.49 A topological space X is said to be compact if every open cov-

ering of X has a finite subcovering.
Remark 1.10.50 This means that from any open covering {Ui } of a compact space X ,

n
we can choose finitely many indices i j , j = 1, 2, . . . , n such that Ui j = X . If X
j=1
is a compact space, every sequence of points x n of X has a convergent subsequence,
which means, every subsequence x n 1 , xn 2 , . . . , xn t , . . . , converges to a point of X .
For metric spaces, this condition is equivalent to compactness.
Proposition 1.10.51 A compact subspace of a topological space X is closed in X

and every closed subspace of a compact space is compact.
Definition 1.10.52 A topological space X is said to be locally compact if each of

its points has a compact neighborhood.
Example 1.10.53 Any compact 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 1.10.54 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.
Example 1.10.55 Every locally compact space is a Baire space.
Definition 1.10.56 A topological space X is said 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 1.10.57 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 1.10.58 A topological space X is said to be paracompact if every open

covering of X has a locally finite subcovering of X .
Example 1.10.59 (i) R n is paracompact.

(ii) Every closed subspace of a paracompact space but a subspace of a paracompact
space is not necessarily paracompact.
Fig. 1.7 Universal

property
of the union X i with Xi ; → Xi
;; i≥1
i≥1 ;; f }
; }
weak topology fi ;; }
; ~}
Y
Definition 1.10.60 A topological space X is said to be countably compact if every

countable open covering of X has a finite subcovering.
Theorem 1.10.61 (Cantor’s intersection theorem) A topological space X is count-

ably compact iff every descending chain of nonempty closed nonempty sets of X has
a nonempty intersection.
1.10.6 Weak Topology
This subsection presents the concept of weak topology which is used in construction
of some important topological spaces. Let X1 ⊂ X 2 ⊂ X 3 ⊂ · · · be a chain of closed
inclusions of topological spaces. Its union X i as the union of the sets X i defines
i≥1
a topology by declaring a subset A ⊂ X i to be closed iff its intersection A ∩ X i
i≥1
is closed in X i for all i ≥ 1. This topology is called union topology. It is also called
∞
weak topology with respect to the subspaces. For example, S ∞ = Sn , R P ∞ =
n=0
∞
∞

R P n , CP ∞ = CP n have the weak topology.
n=0 n=0

Theorem 1.10.62 The union X i has the following universal property:
i≥1
If a family { f i : X i → Y : i ≥ 1} of continuous
maps is such that f i+1 | X i = f i :
X i → Y , then there exists a union map f : X i → Y satisfying the property that
f | X i = f i : X i → Y , represented by the commutative diagram as in the Fig. 1.7.
1.11 Partition of Unity and Lebesgue Lemma
This section discusses the concept of ‘partition of unity’ and states Lebesgue lemma
with Lebesgue number. A partition of unity subordinate to a given open covering is
an important concept in mathematics.
1.11 Partition of Unity and Lebesgue Lemma 31
Definition 1.11.1 Let U = {U j : j ∈ J } be an open covering of a topological space

X . A partition of unity subordinate to U consists of a family of functions { f j : X →
I }, j ∈ J such that f j |(X −U j ) = 0 for all j and each x ∈ X has a neighborhood V

with the property f j |V = 0, except for a finite number of indices j, and f j (x) = 1
j
for all x ∈ X .

Remark 1.11.2 The sum f j (x) is always a finite sum.
j
Example 1.11.3 If {g j : U j → R} is a family of continuous functions, where {U j :

j ∈ J } is an open
covering of a topological space X , then the function g : X → R
such that g(x) = f j (x)g j (x) is well defined and continuous.
j
Paracompactness of a topological space can be characterized with the help of
partition of unity.
Theorem 1.11.4 A topological space X is paracompact iff every open covering U

of X admits a partition of unity subordinate to U.
Proof See (Dugundji 1966). ❑
1.11.1 Lebesgue Lemma and Lebesgue Number
Lebesgue Lemma is used to prove many important results. This lemma is also called
Lebesgue Covering Lemma.
Lemma 1.11.5 (Lebesgue) Let X be a compact metric space. Given an open cover-
ing {Uα : α ∈ A} of X , there exists a real number δ > 0 (called Lebesgue number of
{Uα }), such that every open ball of radius less than δ lies in some element of {Uα }.
1.12 Separation Axioms, Urysohn Lemma, and Tietze

Extension Theorem
This section imposes certain conditions on the topology to obtain some particu-
lar 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 1.12.1 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.
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 neigh-
borhood 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
separated 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
separated in (X, τ ), equivalently, each pair of disjoint closed sets have disjoint
neighborhoods.
Remark 1.12.2 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 1.12.3 (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.
Theorem 1.12.4 (Tietze extension theorem) If E is a closed subspace of a normal

space X , then every continuous map g : E → I has a continuous extension over X .
1.13 Identification Maps, Quotient Spaces,

and Geometrical Construction
This section presents geometrical construction of some quotient spaces. The concept
of quotient spaces is very important in topology and geometry to formalize the
intuitive idea of ‘gluing’ or ‘identifying’ or ‘pasting’ mathematically.
1.13 Identification Maps, Quotient Spaces, and Geometrical Construction 33
Definition 1.13.1 Let X be a topological space and ρ be an equivalence relation

on X . If p : X → X/ρ, x → [x] is the natural surjective map, then the collection
of all subsets U ⊂ X/ρ such that p−1 (U ) is an open set of X , forms the largest
topology on X/ρ such that p is continuous. The set X/ρ is called a quotient space of
X , with the quotient topology and p is called an identification map. The process
of identification is sometimes called ‘gluing’ or ‘pasting’.
Example 1.13.2 Let I be the closed unit interval and ρ be an equivalence relation on
I such that [0] = [1] = {0, 1} and [x] = {x} for 0 < x < 1. Then I /ρ is the quotient
space homeomorphic to the circle S 1 . In other words, S 1 is obtained from I by
identifying the end points 0 and 1 of I .
Example 1.13.3 If we identify all the points of the circumference of a disk D 2 , then
the resulting quotient space is homeomorphic to the sphere S 2 .
Example 1.13.4 Let X be a topological space. If we define an equivalence relation ρ

on X × I by (x, t)ρ(y, s) iff t = s = 1, then the resulting quotient space, is called the
cone over X , denoted by C X . Thus C X = X × I /X × {1} and this identified point
[x, 1] is its vertex. Thus if we consider I as a pointed space with base point 0, then
the reduced cone C X over X is defined to be the quotient space obtained from X × I
by collapsing X × 0 ∪ x 0 × I to a point, i.e., C X = X × I /(X × 0) ∪ x0 × I . We
use [x, t] to denote the point of C X corresponding to the point (x, t) ∈ X × I under
the identification map
p : X × I → C X.
The space X is embedded as a closed subspace of C X by the map
x → [x, 1].
Example 1.13.5 For topological spaces X and Y , every continuous map f : X ×

I → Y satisfying the condition f (x, 1) = y0 ∈ Y for all x ∈ X , induces a continuous
map f˜ : C X → Y, [x, t] → f (x, t). In particular, if X = S n and Y = D n+1 , f :
S n × I → D n+1 , (x, t) → (1 − t)x, then f˜ : C S n → D n+1 , [x, t] → (1 − t)x is a
homeomorphism. Thus D n+1 is the cone over S n with vertex f (x, 1) = y0 .
Example 1.13.6 Let I × I be the unit square.

(a) If we identify the point (0, t) with (1, t) of the square I × I for all t ∈ I , then
the quotient space obtained by such identification is the Euclidean cylinder with
two disjoint circles as their boundaries. If the top and bottom circles of a cylinder
are glued together, the resulting quotient space is called is called the ‘torus’ as
shown in Fig. 1.8.
Thus the torus is the quotient space obtained from unit square I × I by identi-
fying (t, 0) with (t, 1) and also (0, t) with (1, t).
Fig. 1.8 Torus as the quotient space of unit square
Fig. 1.9 Möbius band as the quotient space of unit square
(b) If we identify the point (0, t) with (1, 1 − t) of the square I × I for all t ∈ I ,
then the resulting quotient space is homeomorphic to the space in R3 , called
Möbius band (or strip) named after A.F. Möbius (1790–1868) who invented it
1858. It is described as shown in Fig. 1.9.
Möbius band is the quotient space M = (0, 1) × [0, 1]/(x, 0) ∼ (x, 1). It can
be embedded in R3 , because M is homeomorphic to a subspace of R3 . It is
a non-orientable surface with only one side and only one boundary. It can be
realized as a ruled surface and is used as conveyor belts.
(c) If we identify the point (0, t) with (1, t) and the point (t, 0) with (1, 1 − t) of
the square I × I for all t ∈ I , then the resulting quotient space is called the
Klein bottle named after F. Klein (1849–1925) who invented it in 1882. It is
described as shown in Fig. 1.10.
Klein Bottle is also defined as the quotient space K = (S 1 × I )/(x, 0) ∼
(x −1 , 1).
As Klein Bottle is the continuous image of I × I , it is a compact and connected
space but it cannot be embedded in R3 , because it is not homeomorphic to any
subspace of R3 , but it can be embedded in R4 . Klein bottle is a closed manifold
in the sense that it is a compact manifold without boundary.
(d) The quotient space obtained from I × I by identifying (t, 0) with (1 − t, 1) and
also (0, t) with (1, 1 − t) as shown in Fig. 1.11 is called the real projective plane
R P 2 . It cannot be embedded in R3 .
(e) The quotient space obtained from square by identifying its boundary to a point
is the 2-sphere S 2 .
Fig. 1.10 Klein bottle as the quotient space of the unit square
Fig. 1.11 Real projective

plane R P 2
Remark 1.13.7 (i) The surface of a cylinder can be considered as a disjoint union
of a family of line segments parametrized continuously by points of a circle. The
Möbious can be considered in a similar way.
(ii) The two-dimensional torus embedded in R3 can be considered as a union of
circle parametrized by points of another circle.
Example 1.13.8 (Mapping Cylinder) Let X and Y be topological spaces and f :

X → Y be continuous. Let (X × I ) Y denote the disjoint union of topological
spaces X × I and Y . Then both X × I and Y are open sets of (X × I ) Y . If we
define an equivalence relation ρ on (X × I ) Y by (x, t)ρy iff y = f (x) and t = 1,
then the quotient space M f = ((X × I ) Y )/ρ is called the mapping cylinder of f .
Thus M f is the space obtained from Y and (X × I )/(x0 × I ) by identifying for each
x ∈ X , the points (x, 1) and f (x) as shown in Fig. 1.12, in which the thick line is
supposed to be identified to a point (the base point of M f ). We denote (x, t)ρ in M f
by [x, t] and yρ in M f by [y]. Then [x] = [x, 1] = [ f (x)], ∀ x ∈ X . The space Y
is embedded in M f under the map y → [y]. In particular, if Y is a one-point space,
then f : X → Y is a constant map and M f is C X , the cone over X .
Example 1.13.9 Let X and Y be topological spaces with base points x0 and y0 ,
respectively. Given a continuous map f : X → Y , the mapping cone as shown in
Fig. 1.13.
C f is the quotient space obtained from Y and C X by identifying the point [x, 1]
of C X with the point f (x) of Y for all x ∈ X . The base point of C f is the point to
which [x 0 , t] and y0 are identical for all t ∈ I .
Fig. 1.12 Mapping cylinder

M f of f x0 x X
y0 f (x) f (X)
Fig. 1.13 Mapping cone C f 1

of f
y0 = f (x0 )
f (X) Y
Example 1.13.10 (Wedge) Let (X, x0 ) and (Y, y0 ) be two pointed topological spaces.
Their wedge (or one-point union) X ∨ Y is the quotient space of their disjoint union
X Y in which the base points are identified. In general, if X i is a collection of
disjoint spaces, with base point xi ∈ X i , then their wedge (or one-point union) Xi
i∈I
is the quotient space X/X 0 , where X = X i and X 0 is the subspace of X consisting

i∈I
of all base points xi ; the base point of X i is the point corresponding to X 0 . In
i∈I
other words, X i is the space obtained from X by identifying together the base
i∈I
point xi .
Example 1.13.11 (Smash product) Let X and Y be two pointed spaces with base
points x0 and y0 respectively. Then their smash product (or reduced product) X ∧ Y
is defined to be the quotient space X × Y/(X ∨ Y ). We may think X ∧ Y as a reduced
version of X × Y obtained by collapsing X ∨ Y to a point.
Example 1.13.12 (Reduced suspension space) Let X be a pointed topological space

with base point x0 . Then the suspension of X , denoted by X , is defined to be the
quotient space of X × I in which (X × 0) ∪ (x 0 × I ) ∪ (X × 1) is identified to a
single point. It is sometimes called the reduced suspension space. If (x, t) ∈ X × I ,
we use [x, t] to denote the corresponding point of X under the identification map
f : X × I → X such that [x, 0] = [x0 , t] = [x , 1] for all x, x ∈ X and for all
t ∈ I . The point [x 0 , 0] ∈ X is also denoted by x0 . Thus X is a pointed space with
base point x 0 and X = X ∧ S 1 . In particular, S n ≈ S n+1 for n ≥ 0. Moreover,
if f : X → Y is a base point preserving continuous map, then f : X → Y is
defined by f ([x, t]) = [ f (x), t].
Proposition 1.13.13 Let Y be a quotient space of X with quotient map p : X → Y

and f : Y → Z be a map from Y to some space Z . Then f is continuous iff the map
f ◦ p : X → Z is continuous.
1.14 Function Spaces
This section introduces the concept of function spaces topologized by the compact
open topology. Function spaces play an important role in topology and geometry.
Definition 1.14.1 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 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 be the evaluation map defined by E( f, x) = f (x). Then
given a function h : Z → Y X , the composite
h×1d E
ψ : Z × X −−−−−→ Y X × X −−−→ Y
i.e., ψ = E ◦ (h × 1d ) : Z × X → Y is a function.
Theorem 1.14.2 (Theorem of Exponential Correspondence) If X is a locally com-

pact Hausdorff space and Y, Z are topological spaces, then a function f : Z → Y X
is continuous if and only if E ◦ ( f × 1d ) : Z × X → Y is continuous.
Theorem 1.14.3 (Exponential Law) If X is a locally compact Hausdorff space, Z

is a Hausdorff space, and Y is a topological space, then the function ψ : (Y X ) Z →
Y Z ×X defined by ψ( f ) = E ◦ ( f × 1d ) is a homeomorphism.
Proposition 1.14.4 If X is a compact Hausdorff space and Y is metrized by a metric

d, then the space Y X is metrized by the metric d defined by
d ( f, g) = sup{d( f (x), g(x)) : x ∈ X }.

1.15 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 space, 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. Our study
is confined to finite dimensional manifolds (although there are infinite dimensional
manifolds).
We are familiar with curves and surfaces. Manifolds are generalizations of curves
and surfaces to arbitrary dimensional objects. The concept of manifolds can be traced
to the work of B. Riemann (1826–1866) on differential and multivalued functions. A
curve in R3 is parametrized locally by a single number t as (x(t), y(z), z(t)), on the
other hand two numbers u and v parametrize a surface as (x(u, v), y(u, v), z(u, v)).
A curve is considered locally homeomorphic to R1 (real line space) and a surface
to R 2 . A manifold in general, is a topological space which is locally homeomorphic
to Rn for some n. More precisely, if every point of a topological space M has a
neighborhood homeomorphic to an open subset of Rn , we call M an n-dimensional
topological manifold.
Definition 1.15.1 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
that is homeomorphic to an open subset of Rn . An one-dimensional manifold is called
a curve and a two-dimensional manifold is called a surface.
For example, S 2 , torus, R P 2 are surfaces. All manifolds M in this book are
assumed to be paracompact to ensure that M is a separable metric space.
Definition 1.15.2 An n-dimensional differentiable manifold M is a Hausdorff topo-

logical space having a countable open covering {U1 , U2 , . . .} such that
DM(1) for each Ui , there is homeomorphism ψi : Ui → Vi , where Vi is an open
disk in Rn ;
DM(2) if Ui ∩ U j = ∅, the homeomorphism ψ ji : ψ j ◦ ψi−1 : ψi (Ui ∩ U j ) →
ψ j (Ui ∩ U j ) is a differentiable map.
(Ui , ψi ) is called a local chart of M and {(Ui , ψi )} is a set of local charts of M.
Example 1.15.3 Rn , S n , R P n are n-dimensional differentiable manifolds.
Example 1.15.4 CP n is a 2n-dimensional differentiable manifold.

1.15 Manifolds 39
Definition 1.15.5 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 subspace x1 ≥ 0 of Rn .
Example 1.15.6 The n-dimensional disk D n is an n-manifold with boundary.
Remark 1.15.7 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.
Example 1.15.8 (Stiefel manifold) Any ordered set of r (r ≤ n) independent vectors

in the Euclidean n-space Rn is called an r -frame. Let Vr (Rn ) be the set of (orthonor-
mal) r-frames in Rn . Then Vr (Rn ) is the subspace of (v1 , v2 , . . . , vr ) ∈ (S n−1 )r such
that vi , v j = δi j (Kronecker delta). Since Vr (Rn ) is a closed subset of a compact
space, it is also compact. Corresponding to each r-frame (v1 , v2 , . . . , vr ) ∈ (S n−1 )r ,
there exists an associated r-dimensional subspace v1 , v2 , . . . , vr with a basis
{v1 , v2 , . . . , vr }. Each r-dimensional subspace of Rn is of the form v1 , v2 , . . . , vr .
The manifold Vr (Rn ) is called the Stiefel manifold of (orthonormal) r-frames in Rn .
It may be considered as the manifold of all orthonormal (r − 1) frames tangent to
S n−1 . In particular, for r = 2, V2 (Rn ) is the manifold of unit vectors tangent to S n−1 .
The orthogonal (real) group O(n, R) acts transitively on Vr (Rn ) with isotropy group
O(n − r, R) and hence O(n, R)/O(n − r, R) ≈ Vr (Rn ).
Example 1.15.9 (Grassmann manifold) An r-dimensional (r ≤ n) linear subspace

of Rn is called an r-plane. Let G r (Rn ) be the set of r-planes of Rn through the origin,
with the quotient topology defined by, the identification map
Vr (Rn ) → G r (Rn ), (v1 , v2 , . . . , vr ) → v1 , v2 , . . . , vr .
Then G r (Rn ) is a compact space. Clearly, V1 (Rn ) = S n−1 andG 1 (Rn ) = R P n−1 .
∞
The natural inclusion G r (R ) ⊂ G r (R ) gives G r (R ) =
n n+1
G r (Rn ) with the
r ≤n
induced topology. For g ∈ O(n, R) and V ∈ Gr (Rn ), g(V ) is another r -plane in Rn .
Any r -frame is taken into any other by some g ∈ O(n, R), so the same is also true
of r -plane. Hence O(n, R) acts transitively on Gr (Rn ). The isotropy group of the
standard Rr ⊂ Rn is O(r, R) × O(n − r, R). Hence
Gr (Rn ) ≈ O(n, R)/O(r, R) × O(n − r, R).
Hence we get the following homeomorphisms:
O(n, R)/O(1, R) × O(n − 1, R) ≈ R P n−1
U (n, C)/U (1, C) × U (n − 1, C) ≈ CP n−1

and
S p (n, H)/S p (1, H) × S p (n − 1, H) ≈ HP n−1
where O(n, R), U(n, C) and Sn (n, H) represent the orthogonal group over R, the
unitary group over C and sympletic group over quaternions H respectively (see
Appendix A).
Example 1.15.10 (The Mbius strip or Mbius band) It is a surface with only one side
and only one boundary. The Mbius strip has the mathematical property of being
non-orientable.
Remark 1.15.11 Throughout the book the terms ‘continuous function’ and ‘contin-
uous map’ (or ‘map’) are interchangeable in the context of topological spaces and
the term ‘space’ means a topological space unless otherwise stated.
1.16 Exercises
1. If f : I → Rn is a one-one continuous function, show that f (I ) and I are home-

omorphic spaces.
[Hint: I is compact and Rn is a metric space and hence Hausdorff. The func-
tion f : I → f (I ) ⊂ Rn is onto and by hypothesis one-one continuous. Hence
f −1 : f (I ) → I exists. Let A be a closed subspace of I . Then A is also compact
and ( f −1 )−1 (A) = f (A) is a closed subset of f (I ), which is also Hausdorff and
hence f (A) is closed.]
2. Show that the family of open disks form an open base of the topology of the
Euclidean plane.
3. Show that
(a) the spaces R and Rn (n > 1) are not homeomorphic.
[Hint: If possible, there exists a homeomorphism f : Rn → R. If x ∈ Rn ,
then the spaces Rn − {x} and R − { f (x)} must be homeomorphic. But this
is not true, because Rn − {x} is connected but R − { f (x)} is not so.]
(b) the spaces I and I n (n > 1) are not homeomorphic but there exists a bijection
between them.
4. Show that a space X is connected iff all continuous functions f : X → Y =
{0, 1} are constant functions f (x) = 0 or f (x) = 1.
[Hint: Use the result that the continuous image of a connected space is con-
nected.]
5. Show that
(i) the continuous image of a path-connected space is path-connected;
(ii) a space which is connected and locally path-connected is path-connected.
6. Let I n be the n-cube and ∂ I n be its boundary. Show that the spaces Rn and
I n − ∂ I n are homeomorphic.
1.16 Exercises 41
[Hint: By Example 1.10.25(i), I − ∂ I = (0, 1) is homeomorphic to R1 . Use

n-products

this result to show that I n − ∂ I n and (I − ∂ I ) × · · · × (I − ∂ I ) are homeo-
morphic.]
7. (Topologist’s sine curve) Let Z = {(x, y) ∈ R2 : y = sin x1 , 0 < x ≤ π}. The
map f : (0, π) → Z , x → (x, sin 1x ) is continuous. Since (0, π) is connected
and Z is the continuous image of f, Z is connected. This space T = Z ∪ {(0, y) :
|y| ≤ 1} is known as ‘Topologist’s sine curve.’
Show that
(a) if a subset B of T is such that Z ⊆ B ⊆ T, then B is connected;
(b) in particular, if B is obtained by adjoining an additional point (0, 1/2)(say),
or a part of the y-axis lying between the points (0, −1) and (0, 1) to Z , then
B is connected.
8. Let X be a topological space and A be a subset of X . Show that A is connected
iff its closure Ā is the same as A.
9. Show that every topological space with a connected dense subset is itself con-
nected.
10. Show that
(a) Rn is connected;
(b) any n-ball D n or n-cube I n in Rn is connected;
(c) the n-sphere S n in Rn+1 is connected for n > 0.
[Hint (c): Rn+1 − {0} is connected. Consider the map
f : Rn+1 − {0} → S n , (x1 , x2 , . . . , xn+1 )

⎛ ⎞
x x x
→ ⎝ , , . . . , ⎠ .
1 2 n+1
2 2
xi xi xi2
Then f is continuous and surjective.]

11. Show that the product of two compact spaces is compact.
12. Show that
2
(i) G L(n, R) is a subspace of Rn with the relative topology;
(ii) G L(n, R) is not connected.
13. Let X be a metric space. Show that the following three conditions on X are
equivalent:
(i) X is compact (Heine–Borel property);
(ii) X is countably compact, i.e., every infinite subset of X has a limit point
(Bolzano–Weierstress property);
(iii) X is sequentially compact, i.e., every sequence in X has a convergence
subsequence.
14. Show that the connectedness and compactness of topological spaces are topo-
logical properties in the sense that if f : X → Y is a homeomorphism, then
(i) X is connected iff Y is connected;
(ii) X is compact iff Y is compact.
15. Let X and Y be topological spaces and f : X → Y be a continuous and surjec-
tive. Show that if X is compact, then Y is also so.
[Hint: Let {V j } j∈J be an open covering of Y . Then { f −1 (V j ) : j ∈ J } is an
open covering of X . Since X is compact, there exists a finite subcovering
{ f −1 (V1 ), . . . , f −1 (Vr )}. Again since f is surjective, f ( f −1 (Vk )) = Vk , for
r
r
k = 1, 2, . . . , r and Vk = Y , because f −1 (Vk ) = X.]
k=1 k=1
16. Let R+ be the space of positive real numbers topologized as a subset of R1 . The
space R+ × S n is in the product topology. Show that the map
f : Rn+1 − {0} → R+ × S n , (x 1 , x2 , . . . , xn+1 )

n+1 1/2
x1 xn+1
→ x, ,..., , where ||x|| = xi2 ,
||x|| ||x|| i=1
is a homeomorphism.
17. Let X be a regular space and A be a proper subspace of X . Show that
(i) the quotient space X/A is a Hausdorff space;
(ii) if X is a normal space and A is closed, then X/A is normal.
18. Let G be a group and H be (not necessarily abelian) subgroup of G and G/H
be the set of all left cosets of H in G. Now G acts on G/H by the action
G × G/H → G/H, (g, g H ) → gg H.
Show that this action is transitive and H is the stabilizer of the coset H .
19. Let a group G act on the setX transitively and H be the stabilizer group of a
point. Show that X is G-isomorphic to G/H , the set of all left cosets of H in G
on which G acts by left translation.
[Hint Let H = G x0 be the stabilizer group of a point x0 ∈ X . As the action is
transitive, for each x ∈ X , ∃ an element gx ∈ G such that gx x0 = x. Define a
map
ψ : X → G/H, x → gx H.
Then ψ is a G-isomorphism.]
20. If H and K are subgroups of a group G, show that the G-sets G/H and G/K
are G-isomorphic iff H and K are conjugate subgroups in G.
[Hint: If H and K are subgroups of the group G, then they are conju-
gate subgroups iff H = g −1 K g for some g ∈ G. Let ψ : G/H → G/K be a
1.16 Exercises 43
G-isomorphism. Then ∃g ∈ G such that ψ(H ) = gK . If h ∈ H , then gK =

ψ(H ) = ψ(h H ) = hψ(H ) = hgK shows that g −1 hg ∈ K and g −1 H g ⊂ K .
Again ψ(g −1 H ) = g −1 ψ(H ) = g −1 gK = K shows that ψ −1 (K ) = g −1 H .
Consequently, g −1 H g ⊂ H . Hence g −1 H g = K shows that H and K are con-
jugate subgroups in G.
Conversely, let g −1 H g = K for some g ∈ G. Show that the map
ψ : G/H → G/K , a H → agK
is a G-isomorphism.]
21. Let (X, A) be a normal pair (i.e., X is normal and A is closed in X ) and
f : A → S n be an open map. Show that there exists an open set U of X con-
taining A and an open extension f˜ : U → S n of f .
[Hint: Consider f as a map of A into Rn+1 . Then by Tietze’s extension Theo-
rem 1.12.4, there exists an extension g : X → Rn+1 of f . Let U = X − g−1 (0),
where 0 is the origin of Rn+1 . Define
g(x)
f˜ : U → S n , x → .
||g(x)||
]
22. Let (X, A) be a normal pair such that X × I is normal. Show that every map
f : X × {0} ∪ A × I → S n admits a continuous extension f˜ : X × I → S n .
[Hint: Use Ex. 21 of Sect. 1.16, for the pair (X × I, B), where B = X × {0} ∪
A × I .]
1.17 Additional Reading
[1] Artin, M., Algebra, Prentice-Hall, Englewood Cliffs, 1991.

[2] Chatterjee, B.C., Ganguly, S., and Adhikari, M.R., A Textbook of Topology, Asian
Books Pvt.Ltd., New Delhi, 2002.
[3] Lang, S., Algebra, Addition-Wesley, Reading, 1965.
[4] Munkres, J.R., Topology, A First Course, Prentice-Hall, New Jersey, 1975.
[5] Switzer, R.M., Algebraic Topology-Homotopy and Homology, Springer-Verlag,
Berlin, Heidelberg, New York, 1975.
References
Adhikari, M.R., Adhikari, A.: Basic Modern Algebra with Applications. Springer, New York (2014)
Artin, M.: Algebra. Prentice-Hall, Englewood Cliffs (1991)
Chatterjee, B.C., Ganguly, S., Adhikari, M.R.: A Textbook of Topology. Asian Books Pvt. Ltd.,
New Delhi (2002)
Dugundji, J.: Topology. Allyn & Bacon, Newtown (1966)

Herstein, I.: Topics in Algebra. Blerisdell, New York (1964)
Lang, S.: Algebra. Addition-Wesley, Reading (1965)
Maunder, C.R.F.: Algebraic Topology. Van Nostrand Reinhhold, London (1970)
Munkres, J.R.: Topology, A First Course. Prentice-Hall, New Jersey (1975)
Spanier, E.: Algebraic Topology. McGraw-Hill Book Company, New York (1966)
Switzer, R.M.: Algebraic Topology-Homotopy and Homology. Springer, Berlin (1975)
Chapter 2
Homotopy Theory: Elementary Basic
Concepts
This chapter opens with a study of homotopy theory by introducing its elementary
basic concepts such as homotopy of continuous maps, homotopy equivalence,
H -group, H -cogroup, contractible space, retraction, deformation with illustrative
geometrical examples and applications. The study of homotopy theory continues
explicitly up to Chap. 9 of the present book. Its many key concepts are also applied
to other chapters. The basic aim of homotopy theory is to investigate ‘algebraic prin-
ciples’ latent in homotopy equivalent spaces. Such principles are also important in
the study of topology and geometry as well as in many other subjects such as algebra,
algebraic geometry, number theory, theoretical physics, chemistry, computer science,
economics, bioscience, medical science, and some other subjects.
Algebraic topology flows mainly through two channels: one is the homotopy the-
ory and other one is the homology theory. The concept of homotopy is a mathematical
formulation of the intuitive idea of a continuous deformation from one geometrical
configuration to other in the sense that this concept formalizes the naive idea of
continuous deformation of a continuous map. On the other hand, the concept of
homology is a mathematical precision to the intuitive idea of a curve bounding an
area or a surface bounding a volume. Cohomology theory which is a dual concept
of homology theory is also closely related to homotopy theory. The idea guiding
the development of mathematical theory of homotopy, homology, and cohomology
is described nowadays in the language of category theory by constructing certain
functors.
Algebraic topology is one of the most important creations in mathematics which
uses algebraic tools to study topological spaces. The basic goal is to find algebraic
invariants that classify topological spaces up to homeomorphism (though usually
classify up to homotopy equivalence). The most important of these invariants are
homotopy, homology, and cohomology groups. This subject is an interplay between
topology and algebra and studies algebraic invariants provided by homotopy and
homology theories. The twentieth century witnessed its greatest development.

DOI 10.1007/978-81-322-2843-1_2
46 2 Homotopy Theory: Elementary Basic Concepts
A basic problem in homotopy theory is to classify continuous maps 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. On the other hand,
the basic problem in algebraic topology is to devise ways to assign various algebraic
objects such as groups, rings, modules to topological spaces and homomorphisms
to the corresponding algebraic structures in a functorial way. More precisely, although
the ultimate aim of topology is to classify topological spaces up to homeomorphism,
the main problem of algebraic topology is the ‘classification’ of topological spaces
up to homotopy equivalence, the concept introduced by W. Hurewicz (1904–1956)
in 1935. So in algebraic topology a homotopy equivalence plays a more influential
role than a homeomorphism, because the basic tools of algebraic topology such as
homotopy groups, and homology & cohomology groups are invariants with respect
to homotopy equivalence.
Homotopy theory constitutes a basic part of algebraic topology and studies topo-
logical spaces up to homotopy equivalence which is a weaker relation than topological
equivalence in the sense that homotopy classes of spaces are larger than homeomor-
phic classes. The concept of the homotopy equivalence gives rise to the classification
of topological spaces according to their homotopy properties. The basic idea of this
classification is to assign to each topological space ‘invariants’, which may be inte-
gers, or algebraic objects in such a way that homotopy equivalent spaces have the
same invariants (up to isomorphism), called homotopy invariants, which characterize
homotopy equivalent spaces completely. The main numerical invariants of homotopy
equivalent spaces are dimensions and degrees of connectedness.
Historically, the idea of homotopy for the continuous maps of unit interval was
originated by C. Jordan (1838–1922) in 1866 and that of for loops was introduced
by H. Poincaré (1854–1912) in 1895 to define an algebraic invariant called the fun-
damental group, which is studied, in Chap. 3. The monumental work of Poincaré
in ‘Analysis situs’, Paris, 1895, organized the subject for the first time. This work
explained the difference between curves deformable to one another and curves bound-
ing a larger space. The first one led to the concepts of homotopy and fundamental
group; the second one led to the concept of homology. Poincaré is the first mathe-
matician who systemically attacked the problems of assigning algebraic (topological)
invariants to topological spaces. His vision of the key role of topology in all math-
ematical theories began to materialize from 1920. Of course, many of the ideas he
developed had their origins prior to him, with L. Euler (1707–1783), and B. Riemann
(1826–1866) above all. H. Hopf (1894–1971) introduced the concept of H -spaces
and H -groups from the viewpoint of homotopy theory. Some of his amazing results
have made a strong foundation of algebraic topology. Many topologists regard H.
Poincaré as the founder and regard H. Hopf and W. Hurewicz as cofounders of many
key concepts in algebraic topology.
Throughout this book a space means a topological space and a map means a
continuous function between topological spaces; the terms: map (or continuous map)
and continuous function will be used interchangeably in the context of topological
spaces, unless specified otherwise.
2 Homotopy Theory: Elementary Basic Concepts 47
For this chapter the books Eilenberg and Steenrod (1952), Hatcher (2002),
Maunder (1970), Spanier (1966) and some others are referred in Bibliography
2.1 Homotopy: Introductory Concepts and Examples
This section is devoted to the study of the concept of homotopy formalizing the
intuitive idea of continuous deformation of a continuous map between two topolog-
ical spaces and presents introductory basic concepts of homotopy with illustrative
examples. Homeomorphism generates equivalence classes whose members are topo-
logical spaces. On the other hand, homotopy generates equivalence classes whose
members are continuous maps. The term homotopy was first given by Max Dehn
(1878–1952) and Poul Heegaard (1871–1948) in 1907. It is sometimes replaced by
a complicated function between two topological spaces by another simpler function
sharing some important properties of the original function. An allied concept is the
notion of deformation. This leads to the concept of homotopy of functions.
The relation between topological spaces of being homeomorphic is an equivalence
relation. So it divides any set of topological spaces into disjoint classes. The main
problem of topology is the classification of topological spaces. Given two topological
spaces X and Y , are they homeomorphic? This is a very difficult problem. Algebraic
topology transforms such topological problems into algebraic problems which may
have a better chance for solution. The algebraic techniques are usually not delicate
enough to classify topological spaces up to homeomorphism. The notion of homo-
topy of continuous functions defines somewhat coarser classification. This leads to
the concept of a continuous deformation. The relation of homotopy of continuous
functions generalizes path connectedness of a point, which is a fundamental concept
of homotopy theory.
2.1.1 Concept of Homotopy
The intuitive concept of a continuous deformation is now explained with the concept
of homotopy. Moreover the concept of ‘flow’ which is also known as one parameter
group of homeomorphisms is conveyed through homotopy. Let I = [0, 1] be the
closed unit interval with topology induced by the natural topology on the real line R
(sometimes written as R1 ).
Definition 2.1.1 Let X and Y be topological spaces. Two continuous maps f, g :

X → Y are said to be homotopic (or f is said to be homotopic to g), if there
exists a continuous map F : X × I → Y such that F(x, 0) = f (x) and F(x, 1) =
g(x), ∀ x ∈ X . The map F is said to be a homotopy between f and g, written
F : f g.
Remark 2.1.2 Geometrically, two continuous maps f, g : X → Y are said to be

homotopic if f can be continuously deformed into g by a continuous family of
maps Ft : X → Y defined by Ft (x) = F(x, t) such that F0 = f and F1 = g, ∀ x ∈
X, ∀ t ∈ I . By saying that the maps Ft form a continuous family, we mean that F is
continuous with respect to both x and t as a function from the product space X × I
to Y . Clearly, a homotopy F between two continuous maps f, g : X → Y can be
considered as a special case of extension:
consider in the topological space X × I the subspace
A = (X × {0}) ∪ (X × {1}) ⊂ X × I
and consider the continuous map
G : A → Y, (x, 0) → f (x), (x, 1) → g(x);
then a homotopy F from f to g is an extension of G from A to X × I.
Definition 2.1.3 Let X be a topological space. A continuous map f : I → X such

that f (0) = x0 and f (1) = x1 , is called a path in X from x0 to x1 . The point x0 is
called the initial point and the point x 1 is called the final or terminal point of the
path f .
Definition 2.1.4 Two paths f, g : I → X are said to be homotopic if they have the
same initial point x0 , the same final point x1 and there exists a continuous map
F : I × I → X such that
F(t, 0) = f (t), F(t, 1) = g(t), ∀ t ∈ I (2.1)

and F(0, s) = x0 and F(1, s) = x1 , ∀ s ∈ I (2.2)
We call F a path homotopy between f and g as shown in Fig. 2.1 and is written
F : f g.
p
Remark 2.1.5 The condition (2.1) says that F is a homotopy between f and g and
the condition (2.2) says that for each t ∈ I, the path t → F(t, s) is a path in X from
x 0 to x1 . In other words, (2.1) shows that F represents a continuous way of deforming
Fig. 2.1 Path homotopy X
x1
f
g
x0
2.1 Homotopy: Introductory Concepts and Examples 49
the path f to the path g and (2.2) shows that the end points of the path remain fixed
during the deformation.
We now prove the following two lemmas of point set topology which will be used
throughout the book.
Lemma 2.1.6 (Pasting or Gluing lemma) Let X be a topological space and A, B

be closed subsets of X such that X = A ∪ B. Let Y be a topological space and
f : A → Y and g : B → Y be continuous maps. If f (x) = g(x), ∀ x ∈ A ∩ B,
then the function h : X → Y defined by

f (x), ∀ x ∈ A
h(x) =
g(x), ∀ x ∈ B
is continuous.
Proof h defined in the lemma is the unique well-defined function X → Y such that
h| A = f and h| B = g. We now show that h is continuous. Let C be a closed set
in Y . Then h −1 (C) = X ∩ h −1 (C) = (A ∪ B) ∩ h −1 (C) = (A ∩ h −1 (C)) ∪ (B ∩
h −1 (C)) = (A ∩ f −1 (C)) ∪ (B ∩ g −1 (C)) = f −1 (C) ∪ g −1 (C). Since each of f
and g is continuous, f −1 (C) and g −1 (C) are both closed in X . Hence h −1 (C) is
closed in X . Consequently, h is continuous. ❑
This lemma can be generalized as follows:
Lemma 2.1.7 (Generalized Pasting or Gluing lemma) Let a topological space

n
X be a finite union of closed subsets X i : X = X i . If for some topological
i=1
space Y , there are continuous maps f i : X i → Y that agree on overlaps (i.e.,
f i | X i ∩X j = f j | X i ∩X j , ∀ i, j), then ∃ a unique continuous function f : X → Y with
f | X i = f i , ∀ i.
Proof The proof is similar to proof of Lemma 2.1.6. ❑
Theorem 2.1.8 Let P(X ) denote the set of all paths in a space X having the same
initial point x0 and the same final point x 1 . Then the path homotopy relation ‘’ is
p
an equivalence relation on P(X ).
Proof Let f, g, h ∈ P(X ). Then f (0) = g(0) = h(0) = x0 and f (1) = g(1) =
h(1) = x1 . Let a map F : I × I → X be defined by F(t, s) = f (t), ∀ t, s ∈ I .
Then F is continuous, because it is the composite of the projection map onto the
first factor and the continuous map f . Hence F is a continuous map such that
F(t, 0) = f (t), F(t, 1) = f (t), ∀ t ∈ I and F(0, s) = x0 , F(1, s) = x 1 , ∀ s ∈ I .
Thus F : f f, ∀ f ∈ P(X ). Next, let f g and F : f g. Then F : I × I → X
p p p
is a continuous map such that F(t, 0) = f (t), F(t, 1) = g(t), ∀ t ∈ I and F(0, s) =
x 0 , F(1, s) = x 1 , ∀ s ∈ I . Let G : I × I → X be the map defined by G(t, s) =
F(t, 1 − s). Since the maps I → I, t → t and s → 1 − s are both continuous, G is

continuous. Now G(t, 0) = F(t, 1) = g(t), G(t, 1) = F(t, 0) = f (t), ∀ t ∈ I and
G(0, s) = F(0, 1 − s) = x 0 , G(1, s) = F(1, 1 − s) = x1 . Hence G : g f .
p
Finally, let f g and g h. Then ∃ continuous maps F, G : I × I → X such
p p
that F : f g and G : g h. Consequently, for all t, s ∈ I , F(t, 0) = f (t),
p p
F(t, 1) = g(t), F(0, s) = x0 , F(1, s) = x1 , G(t, 0) = g(t), G(t, 1) = h(t),
G(0, s) = x0 and G(1, s) = x1 . We now define a map H : I × I → X by the
equations
F(t, 2s), 0 ≤ s ≤ 1/2
H (t, s) =
G(t, 2s − 1), 1/2 ≤ s ≤ 1
At s = 21 , F(t, 2s) = F(t, 1) = g(t) and G(t, 2s − 1) = G(t, 0) = g(t), ∀ t ∈ I

show that F and G agree at t × 21 . Moreover, F is continuous on I × [0, 12 ] and
G is continuous on I × [ 12 , 1]. Hence by Pasting lemma, H is continuous. Now,
H (t, 0) = F(t, 0) = f (t), ∀ t ∈ I , H (t, 1) = G(t, 1) = h(t), ∀ t ∈ I ,

F(0, 2s), 0 ≤ s ≤ 1/2
H (0, s) =
G(0, 2s − 1), 1/2 ≤ s ≤ 1
= x0 , ∀ s ∈ I
and
F(1, 2s), 0 ≤ s ≤ 1/2
H (1, s) =
G(1, 2s − 1), 1/2 ≤ s ≤ 1
= x1 , ∀ s ∈ I
Hence H : f h. Consequently, ‘’ is an equivalence relation on P(X ). ❑

p p
Definition 2.1.9 The quotient set P(X )/ is called the set of path homotopy classes
p
of paths in X .
Example 2.1.10 Let f, g : X → R2 be two continuous maps. Define F : X × I →R2

by the rule F(x, t) = (1 − t) f (x) + tg(x), ∀ x ∈ X, ∀ t ∈ I . Then F : f g. In
this example, F shifts the point f (x) to the point g(x) along the straight line seg-
ment joining f (x) and g(x), as shown in Fig. 2.2. The map F is called a straight line
homotopy.
Fig. 2.2 Straight line f (x)

homotopy f
g
g(x)
Example 2.1.11 If X = Y = Rn and f (x) = x and g(x) = 0 ≡ (0, . . . , 0) ∈ Rn ,

∀ x ∈ Rn , i.e., if f = 1 X (identity map on X ) and g is the constant map at 0,
then F : X × I → X , defined by F(x, t) = (1 − t)x is a homotopy from f to g,
i.e., F : f g. Again G : X × I → X , defined by G(x, t) = (1 − t 2 )x is also a
homotopy from f to g. These examples show that homotopy between two maps is
not unique.
Remark 2.1.12 As there are many homotopies between two maps, we can deform a
map f into a given map g in different ways.
Example 2.1.13 Let X denote the punctured plane X = R2 − {0}. Then the paths
f (t) = (cos πt, sin πt), g(t) = (cos πt, 2 sin πt) are path homotopic; the straight
line homotopy between them is an acceptable path homotopy.
On the other hand, the straight line homotopy between the paths f (t) = (cos πt,
sin πt) and h(t) = (cos πt, − sin πt) is not acceptable, because it passes through 0
and hence it does not entirely lie in the space X = R2 − {0}, as shown in the Fig. 2.3.
There does not exist any path homotopy in X between the paths f and h, because
one cannot deform f into g continuously passing through the hole at 0.
Example 2.1.14 Let Dn = {x ∈ R n : ||x|| ≤ 1} be the n-disk. If f (x) = x and g(x) =

0, ∀ x ∈ Dn , then g f . Define F : Dn × I → Dn by F(x, t) = t x, ∀ x ∈ Dn , ∀ t ∈
I . Now ||t x|| = |t| · ||x|| ≤ 1 ⇒ t x ∈ Dn , ∀ t ∈ I and ∀ x ∈ Dn ⇒ F is well
defined. Clearly, F is continuous and F : g f . Similarly, G : Dn × I → Dn
defined by G(x, t) = (1 − t)x is a continuous map such that G : f g.
Example 2.1.15 Let f, g : I → I be defined by f (t) = t and g(t) = 0, ∀ t ∈ I .

Then F : I × I → I defined by F(t, s) = (1 − s)t is a continuous map such that
F : f g.
Example 2.1.16 Let Y be a subspace of Rn and f, g : X → Y be two continuous

maps such that for every x ∈ X, f (x) and g(x) can be joined by a straight line segment
in Y , then F : f g, where F :X×I →Y is defined by
Fig. 2.3 Path homotopy (0, 2)
(0, 1)
g
(−1, 0) (1, 0)
(0, 0)
(0, −1)
F(x, t) = (1 − t) f (x) + tg(x). Since f (x) and g(x) can be joined by a line segment
in Y by hypothesis, F is well defined. To prove the continuity of F, we take x, u ∈
X and t, s ∈ I . Then F(u, s) = (1 − s) f (u) + sg(u). Now F(u, s) − F(x, t) =
(s − t)(g(u) − f (u)) + (1 − t)( f (u) − f (x)) + t (g(u) − g(x)). Let > 0 be an
arbitrary small positive number. Then
||F(u, s) − F(x, t)|| ≤ |(s − t)|||g(u) − f (u)|| + |(1 − t)||| f (u) − f (x)||
+ |t|||g(u) − g(x)|| (2.3)
Again, f and g being continuous, ∃ open neighborhoods U1 and U2 of x in X such that

for u ∈ U1 ∩ U2 , || f (u) − f (x)|| < /3, ||g(u) − g(x)|| < /3. Then for u ∈ U1 ∩
U2 , ||g(u) − f (u)|| ≤ ||g(u) − g(x)|| + ||g(x) − f (x)|| + || f (x) − f (u)|| < c,
where c is the positive constant ||g(x) − f (x)|| + 2/3. Thus if |s − t| < /3c, then
from (2.3) it follows that
||F(u, s) − F(x, t)|| < (2.4)
Since the set (U1 ∩ U2 ) × (t − /3c, t + /3c) is open in X × I , this shows that F is
continuous. Finally, F(x, 0) = f (x) and F(x, 1) = g(x), ∀ x ∈ X . Consequently,
F : f g.
Remark 2.1.17 Geometrically, the above homotopy F deforms f into g along the
straight line segment in Y joining the points f (x) and g(x) for every x ∈ X . The
function F is called a straight line homotopy.
Example 2.1.18 Let X be a topological space and S n be the n-sphere in Rn+1 . If f, g :

X → S n are two continuous maps such that f (x) = −g(x) for any x ∈ X , then f
(1−t) f (x)+tg(x)
g. To show this define the map F : X × I → S n by F(x, t) = ||(1−t) f (x)+tg(x)||
. For
each x ∈ S n and t ∈ I, (1 − t) f (x) + tg(x) ∈ Rn+1 . The given condition f (x) =
−g(x) for any x ∈ X shows that the line segment joining f (x) and g(x) cannot pass
through the origin 0 = (0, 0, . . . , 0) ∈ Rn+1 . In other words, (1 − t) f (x) + tg(x) =
0 for any t ∈ I and any x ∈ X . Hence F(x, t) ∈ S n , ∀ (x, t) ∈ X × I and F is
well defined. We now consider f, g as f, g : X → Rn+1 − {0}. Then by Example
2.1.16, ∃ a straight line homotopy G : X × I → Rn+1 − {0} defined by G(x, t) =
(1 − t) f (x) + tg(x), i.e., G : f g. Again consider the map h : Rn+1 − {0} → S n
defined by h(x) = ||x||
x
. Then F = h ◦ G is the composite of two continuous maps
f (x)
h and G and hence F is continuous. Finally, F(x, 0) = || f (x)||
= f (x), because
g(x)
f (x) ∈ S , ∀ x ∈ X and F(x, 1) =
n
||g(x)||
= g(x), because g(x) ∈ S n , ∀ x ∈ X .
Consequently, F : f g.
Example 2.1.19 Let X be a topological space and f : X → S n be a continuous non-

surjective map. Then f is homotopic to a constant map c : X → S n . By hypothesis
f (X ) S n ⇒ ∃ a point s0 ∈ S n such that s0 ∈
/ f (X ). Define a constant map c : X →
S n by c(x) = −s0 , ∀ x ∈ X. Then f (x) = −c(x) for any x ∈ X . Hence f c by
Example 2.1.18.
Example 2.1.20 Let S 1 = {z ∈ C : |z| = 1} = {eiθ : 0 ≤ θ ≤ 2π} be the unit circle

in C. Then the maps f, g : S 1 → S 1 defined by f (z) = z and g(z) = −z are homo-
topic. Consider the map F : S 1 × I → S 1 defined by F(eiθ , t) = ei(θ+tπ) . Clearly,
F is the composite of the maps
S 1 × I → S 1 × S 1 → S 1 , (eiθ , t) → (eiθ , eitθ ) → ei(θ+tπ) ,
where the second map is the usual multiplication of complex numbers. Consequently,
F is a continuous map such that F : f g.
We now extend Theorem 2.1.8 to the set C(X, Y ) of all continuous maps from
X to Y by extending the concept of path homotopy (obtained by replacing I by any
topological space X ).
Theorem 2.1.21 Given topological spaces X and Y , the relation ‘’ (of being
homotopic) is an equivalence relation on the set C(X, Y ).
Proof Each f ∈ C(X, Y ) is homotopic to itself by a homotopy H : X × I → Y
defined by H (x, t) = f (x). Thus f f, ∀ f ∈ C(X, Y ). Next suppose H : f
g, f, g ∈ C(X, Y ). Define G : X × I → Y by G(x, t) = H (x, 1 − t). Then G is
continuous, because G is the composite of continuous maps
X × I → X × I → Y, (x, t) → (x, 1 − t) → H (x, 1 − t),
where the first map is continuous, because the projection maps (x, t) → x and
(x, t) → (1 − t) are continuous and the second map is H . Then G(x, 0) = H (x, 1) =
g(x) and G(x, 1) = H (x, 0) = f (x), ∀ x ∈ X . Thus G : g f . Finally, let
f, g, h ∈ C(X, Y ) be such that F : f g and G : g h. Define a map H : X ×
I → Y by
F(x, 2t), 0 ≤ t ≤ 1/2
H (x, t) =
G(x, 2t − 1), 1/2 ≤ t ≤ 1
Then H is continuous by Pasting lemma. Finally, H (x, 0) = F(x, 0) = f (x) and

H (x, 1) = G(x, 1) = h(x), ∀ x ∈ X ⇒ H : f h. Consequently, ‘’ is an equiv-
alence relation on C(X, Y ). ❑
Definition 2.1.22 The quotient set C(X, Y )/ is called the set of all homotopy
classes of maps f ∈ C(X, Y ), denoted by [X, Y ] and for f ∈ C(X, Y ), [ f ] ∈ [X, Y ]
is called the homotopy class of f .
Remark 2.1.23 The set [X, Y ] was first systemically studied by M.G. Barratt in 1955
in his paper (Barratt 1955). This set plays the central role in algebraic topology and
is used throughout the book. Some of its properties are displayed in Sect. 2.3.
We now show that composites of homotopic maps are homotopic.
Theorem 2.1.24 Let f 1 , g1 ∈ C(X, Y ) and f 2 , g2 ∈ C(Y, Z ) be maps such that
f 1 g1 and f 2 g2 . Then the composite maps f 2 ◦ f 1 and g2 ◦ g1 : X → Z are
homotopic.
Proof Let F : f 1 g1 and G : f 2 g2 . Then f 2 ◦ F : X × I → Z is a continu-

ous map such that ( f 2 ◦ F)(x, 0) = f 2 (F(x, 0)) = f 2 ( f 1 (x)) = ( f 2 ◦ f 1 )(x), ∀ x ∈
X and ( f 2 ◦ F)(x, 1) = f 2 (F(x, 1)) = f 2 (g1 (x)) = ( f 2 ◦ g1 )(x), ∀ x ∈ X . Conse-
quently,
f 2 ◦ F : f 2 ◦ f 1 f 2 ◦ g1 (2.5)
Again we define H : X × I → Z by H (x, t) = G(g1 (x), t). Thus H is the compos-

ite
g1 ×1d G
X × I −−−−−−→ Y × I −−−→ Z ,
(x, t) → (g1 (x), t) → G(g1 (x), t).
Then H is a continuous map such that H (x, 0) = G(g1 (x), 0) = f 2 (g1 (x)) =
( f 2 ◦ g1 )(x), ∀ x ∈ X and H (x, 1) = G(g1 (x), 1) = g2 (g1 (x)) = (g2 ◦ g1 )(x),
∀ x ∈ X. Hence
H : f 2 ◦ g1 g2 ◦ g1 (2.6)
Consequently, by transitive property of homotopy relation, it follows from (2.5) and

(2.6) that f 2 ◦ f 1 g2 ◦ g1 . ❑
Remark 2.1.25 Theorem 2.1.24 asserts in the language of category theory that topo-
logical spaces and homotopy classes of continuous maps form a category denoted
by Htp called homotopy category of topological spaces (see Appendix B). Thus
Htp is the category whose objects are topological spaces and mor (X, Y ) consists of
homotopy classes of continuous maps from X to Y , where the composition of maps
is consistent with homotopies (see Theorem 2.1.24).
Given a topological space X the concept of a flow ψt : X → X (t ∈ R) is closely

related to homotopy.
Definition 2.1.26 A continuous family ψt : X → X (t ∈ R) of maps is called a

flow if
(i) ψ0 = 1d ;
(ii) ψt is a homeomorphism for all t ∈ R;
(iii) ψt+s = ψt ◦ ψs .
Remark 2.1.27 It is sometimes convenient to consider a flow ψt as a continuous map
ψ : X × R → R, (x, t) → ψt (x).
A flow is also known as one parameter group of homeomorphisms.
Proposition 2.1.28 ψt : X → X is homotopic to I X .

Proof Consider the map
F : X × I → X, (x, s) → ψ(x, (1 − s)t).
This shows that every ψt is homotopic to 1 X . ❑
We now extend the Definition 2.1.1 of homotopy of continuous maps for pairs of
topological spaces.
Definition 2.1.29 A topological pair (X, A) consists of a topological space X and a

subspace A of X . If A = ∅, the empty set, we shall not distinguish between the pair
(X, ∅) and the space X . A subpair (X , A ) of (X, A) is a pair such that X ⊂ X and
A ⊂ A.
Definition 2.1.30 A continuous map f : (X, A) → (Y, B) is a continuous function

f : X → Y such that f (A) ⊂ B.
Given a topological pair (X, A), (X, A) × I represents the pair (X × I, A × I ).
Definition 2.1.31 Given pairs of topological spaces (X, A) and (Y, B), two continu-
ous maps f, g : (X, A) → (Y, B) are said to be homotopic if ∃ a continuous map F :
(X × I, A × I ) → (Y, B) such that F(x, 0) = f (x) and F(x, 1) = g(x), ∀ x ∈ X.
Then the map F is called a homotopy from f to g and written F : f g.
We now consider a more restricted type of homotopy of continuous maps between

pairs of topological spaces, which extends the concept of path homotopy obtained
by replacing I by any topological space X and {0, 1} by a subspace of X under
consideration.
Definition 2.1.32 Let f, g : (X, A) → (Y, B) be two continuous maps of pairs of

topological spaces and X ⊂ X be such that f | X = g| X (i.e., f (x ) = g(x ), ∀ x ∈
X , which implies that f and g agree at x , ∀ x ∈ X ). Then f and g are said to
be homotopic relative to X if there exists a continuous map F : (X × I, A × I ) →
(Y, B) such that F(x, 0) = f (x), F(x, 1) = g(x), ∀ x ∈ X and F(x , t) = f (x ) =
g(x ), ∀ x ∈ X and ∀ t ∈ I, and written F : f g rel X .
If X = ∅, we omit the phrase relative to X .
Remark 2.1.33 f g rel X ⇒ f g rel X for any subspace X ⊂ X .
Geometrical Interpretation: For t ∈ I , if we define h t : (X, A) → (X × I, A × I )

by h t (x) = (x, t), then h 0 (x) = (x, 0) and h 1 (x) = (x, 1). Thus F : f g rel X ⇒
F ◦ h 0 = f, F ◦ h 1 = g, and F ◦ h t | X = f | X = g| X , ∀ t ∈ I ⇒ the collection
{F ◦ h t }t∈I is a continuous one parameter family of maps from (X, A) to (Y, B) agree-
ing on X and satisfying the relations f = F ◦ h 0 and g = F ◦ h 1 . Thus f g rel X
represents geometrically a continuous deformation deforming f into g by maps all
of which agree on X . For example, f g rel {0} in Example 2.1.14.
Example 2.1.34 Consider D 2 = {z ∈ C : z = r eiθ , 0 ≤ r ≤ 1} and S 1 = {z ∈ C :

z = eiθ , 0 ≤ θ ≤ 2π}. Then S 1 ⊂ D 2 . Let f : (D 2 , S 1 ) → (D 2 , S 1 ) be the identity
map and g : (D 2 , S 1 ) → (D 2 , S 1 ) be the reflection in the origin, i.e., g(r eiθ ) =
r ei(θ+π) . Then f g rel {0} under the homotopy F : (D 2 , S 1 ) × I → (D 2 , S 1 )
defined by F(r eiθ , t) = r ei(θ+tπ) . Moreover, G : (D 2 , S 1 ) × I → (D 2 , S 1 ) defined
by G(r eiθ , t) = ei(θ−tπ) is also a homotopy G : f g rel {0} (compare Example
2.1.14).
Remark 2.1.35 There may exist different homotopies from f to g relative to a sub-
space and thus homotopy from f to g is not unique.
Example 2.1.36 Let X be a topological space, A ⊂ X and Y be any convex subspace

of Rn . If f, g : X → Y are two continuous maps such that f | A = g| A , then f
g rel A by a homotopy G : X × I → Y defined by G(x, t) = (1 − t) f (x) + tg(x).
We now generalize Theorem 2.1.21.
Theorem 2.1.37 The relation between continuous maps from (X, A) to (Y, B) of
being homotopic relative to a subspace X ⊂ X is an equivalence relation.
Proof Reflexivity Let f : (X, A) → (Y, B) be a continuous map. Define F : (X ×

I, A × I ) → (Y, B) by the rule F(x, t) = f (x), ∀ x ∈ X, ∀ t ∈ I . Then F : f
f rel X .
Symmetry Let F : f g rel X . Define G : (X × I, A × I ) → (Y, B) by the rule
G(x, t) = F(x, 1 − t), ∀ x ∈ X, ∀ t ∈ I . For continuity of G see Theorem 2.1.21.
Then G : g f rel X .
Transitivity Let f, g, h : (X, A) → (Y, B) be three continuous maps such that F :
f g rel X and G : g h rel X . Define H : (X × I, A × I ) → (Y, B) by the rule

F(x, 2t), 0 ≤ t ≤ 1/2
H (x, t) =
G(x, 2t − 1), 1/2 ≤ t ≤ 1
Then H is continuous by Pasting Lemma 2.1.6. Moreover, H : f h rel X ❑
Remark 2.1.38 It follows from Theorem 2.1.37 that the set of continuous maps
from (X, A) to (Y, B) is partitioned into disjoint equivalence classes by the relation
of homotopy relative to X denoted by [X, A; Y, B]. This set is very important in the
study of algebraic topology. Given a continuous map f : (X, A) → (Y, B), [ f | X ]
represents the homotopy class in [X, A; Y, B] determined by f .
We now generalize Theorem 2.1.24 for homotopies relative to a subspace.
Theorem 2.1.39 Let f 0 , f 1 : (X, A) → (Y, B) be homotopies relative to X ⊂ X

and g0 , g1 : (Y, B) → (Z , C) be homotopies relative to Y , where f 1 (X ) ⊂ Y ⊂ Y .
Then the composites g0 ◦ f 0 , g1 ◦ f 1 : (X, A) → (Z , C) are homotopic relative to
X , i.e., composites of homotopic maps are homotopic.
Proof Let F : f 0 f 1 rel X and G : g0 g1 rel Y . Then the composite mapping
F g0
(X × I, A × I ) −−−→ (Y, B) −−−→ (Z , C)
is a homotopy relative to X from g0 ◦ f 0 to g0 ◦ f 1 , i.e.,
g0 ◦ F : g0 ◦ f 0 g0 ◦ f 1 rel X (2.7)
Again the composite mapping
f 1 ×1d G
(X × I, A × I ) −−−−−−→ (Y × I, B × I ) −−−→ (Z , C)
is a homotopy relative to
f 1−1 (Y ) from g0 ◦ f 1 to g1 ◦ f 1 (2.8)
Since X ⊂ f 1−1 (Y ), (2.7) and (2.8) show that g0 ◦ f 0 g0 ◦ f 1 rel X and g0 ◦ f 1

g1 ◦ f 1 rel X and hence g0 ◦ f 0 g1 ◦ f 1 rel X by transitivity of the relation .❑
2.1.2 Functorial Representation
This subsection summarizes the earlier discussion in the basic result from the view-
point of category theory which gives important examples of categories, functors and
natural transformations, the concepts defined in Appendix B.
Theorem 2.1.39 shows that there is a category, called the homotopy category
of pairs of spaces whose objects are topological pairs and whose morphisms are
homotopy classes relative to a subspace. This category contains as full subcategories
the homotopy category Htp of topological spaces and also the homotopy category
Htp∗ of pointed topological spaces.
Theorem 2.1.40 There is a covariant functor from the category of pairs of topologi-
cal spaces and their continuous maps to the homotopy category whose object function
is the identity function and whose morphism function sends a continuous map f to its
homotopy class [ f ]. Moreover, for any pair (P, Q) of topological spaces there is a
covariant functor π(P,Q) from the homotopy category of pairs to the category of sets
and functions defined by π(P,Q) (X, A) = [P, Q; X, A] and if f : (X, A) → (Y, B)
is continuous, then f ∗ = π(P,Q) ([ f ]) : [P, Q; X, A] → [P, Q; Y, B] is defined by
f ∗ ([g]) = [ f ◦ g] for g : (P, Q) → (X, A).
If α : (P, Q) → (P , Q ), then there is a natural transformation α∗ : π(P ,Q ) →
π(P,Q) . Similarly, we can define a contravariant functor π (P,Q) for a given (P, Q) of

pair of topological spaces and a natural transformation α∗ : π (P,Q) → π (P ,Q ) .
2.2 Homotopy Equivalence
This section studies the concept of homotopy equivalence introduced by W. Hurewicz

(1935) to establish a connection between homotopy and homology groups of a certain
class of topological spaces. The problem of classification of continuous maps from
one topological space to other is closely related to the problem of classification of
topological spaces according to their homotopy properties. This problem led to the
concept of homotopy equivalence which is not only a generalization of the concept
of homeomorphism but also gives a new foundation for the development of the
combinatorial invariants of topological spaces and manifolds. The higher homotopy
groups and the homology groups are invariants of the homotopy equivalent class of
a topological space.
Classification of topological spaces up to homotopy equivalences is the main prob-
lem of algebraic topology. This is a weaker relation than a topological equivalence in
the sense that homotopy classes of continuous maps of topological spaces are larger
than their homeomorphism classes. Although the main aim of topology is to classify
topological spaces up to homeomorphism; in algebraic topology, a homotopy equiv-
alence plays a more important role than a homeomorphism. Because the basic tools
of algebraic topology such as homotopy and homology groups are invariants with
respect to homotopy equivalences.
Definition 2.2.1 A continuous map f : (X, A) → (Y, B) is called a homotopy

equivalence if [ f ] is an equivalence in the homotopy category of pairs. In particular, a
map f ∈ C(X, Y ) is said to be a homotopy equivalence if ∃ a map g ∈ C(Y, X ) such
that g ◦ f 1 X (existence of left homotopy inverse of f ) and f ◦ g 1Y (existence
of right homotopy inverse of f ). In such a situation g is unique and the map g is
called a homotopy inverse of f .
Remark 2.2.2 Let f be a homotopy equivalence with g as its homotopy inverse. Then
[g] = [ f ]−1 in the homotopy category Htp which has the same objects in the category
Top of topological spaces and their continuous maps but the morphism in Htp are
the homotopy classes of continuous maps, so that their morphisms morHtp (X, Y ) =
[X, Y ]. The isomorphisms in the category Top are homeomorphisms and in the
category Htp are homotopy equivalences.
Example 2.2.3 Let Y be the (n − 1)-sphere S n−1 ⊂ Rn ⊂ Rn+q and X be the subset
of R n+q of points not lying on the plane x1 = · · · = xn = 0. Then the inclusion map
i : Y → X is a homotopy equivalence.
Define f : X → Y by f (x 1 , x2 , . . . , xn+q ) = (r x1 , . . . , r x n , 0, . . . , 0), where r =
(x 12 + x22 + · · · + xn2 )−1/2 . Then f ◦ i = 1Y . Again define
H : X × I → X, (x1 , x 2 , . . . , x n+q , t) → (r 1−t x1 , . . . , r 1−t xn , t xn+1 , . . . , t xn+q )
is a homotopy from i ◦ f to 1 X . Consequently, i is a homotopy equivalence.

2.2 Homotopy Equivalence 59
Example 2.2.4 Let D 2 be the unit disk in R2 and p ∈ D 2 . Let i : P = { p} → D 2

be the inclusion map and c : D 2 → P be the constant map. Then c ◦ i = 1 P . Again
the map H : D 2 × I → D 2 defined by H (x, t) = (1 − t)x + t p, being a homotopy
from 1 D2 to i ◦ c, i.e., i ◦ c 1 D2 . Consequently, i is a homotopy equivalence.
Definition 2.2.5 If f ∈ C(X, Y ) is a homotopy equivalence, then X and Y are said

to be homotopy equivalent spaces, denoted by X Y .
We now extend Definition 2.2.5 for pairs of topological spaces.
Definition 2.2.6 Two pairs of topological spaces (X, A) and (Y, B) are said to be
homotopy equivalent, written (X, A) (Y, B), if ∃ continuous maps f : (X, A) →
(Y, B) and g : (Y, B) → (X, A) such that g ◦ f 1 X and f ◦ g 1Y , the homotopy
being the homotopy of pairs.
Remark 2.2.7 Homeomorphic spaces are homotopy equivalent but its converse is
not true in general.
Consider the following example:
Example 2.2.8 Let X be the unit circle S 1 in R2 and Y be the topological space
S 1 , together with the line segment I1 joining the points (1, 0) and (2, 0) in R2 , i.e.,
I1 = {(r, 0) ∈ R2 : 1 ≤ r ≤ 2}. Then X and Y are of the same homotopy type but
they are not homeomorphic (Fig. 2.4).
X and Y cannot be homeomorphic, because removal of the point (1, 0) from Y
makes Y disconnected. On the other hand, removal of any point from X leaves X
connected. We claim that X Y . We take f : X → Y to be the inclusion map and
g : Y → X defined by
y, if y ∈ X
g(y) =
(1, 0), if y ∈ I1
The continuity of g follows from Pasting lemma. Then f ◦ g, 1Y : Y → Y are two

continuous maps such that ( f ◦ g)(y) = f (g(y)) = g(y), ∀ y ∈ Y and

y, if y ∈ X
1Y (y) =
(r, 0), if y = (r, 0) ∈ I1 .
Since for every y ∈ Y, ( f ◦ g)(y) and 1Y (y) can be joined by a straight line segment in
Y , it follows that f ◦ g 1Y (see Example 2.1.16). Again g ◦ f = 1 X 1 X . Thus f
Fig. 2.4 Homotopy S1 S1

equivalent but
non-homeomorphic spaces I1
(2, 0)
(1, 0)
X = S1 Y = S 1 ∪ I1
and g are two continuous maps such that f ◦ g 1Y and g ◦ f 1 X . Consequently,

X Y.
This example shows that X Y does not imply X ≈ Y .
Example 2.2.9 The topological spaces X and Y in Example 2.2.8 are homotopy
equivalent. The unit disk D 2 is homotopy equivalent to a one-point topological space
{ p} ⊂ D 2 (see Example 2.2.4).
As the name suggests, the relation of being homotopy equivalent is an equivalence

relation on the set of topological spaces (or pairs of topological spaces).
Theorem 2.2.10 The relation ‘ ’ between topological spaces (or pairs of topolog-
ical spaces) of being homotopy equivalent is an equivalence relation.
Proof Reflexivity: 1 X : X → X is a homotopy equivalence ⇒ X X for all X .

Symmetry: Let X Y . Then ∃ continuous maps f : X → Y and g : Y → X such
that g ◦ f 1 X and f ◦ g 1Y ⇒ g : Y → X is a homotopy equivalence with
homotopy inverse f : X → Y . Consequently, Y X . Thus X Y ⇒ Y X .
Transitivity: Let X Y and Y Z . Then ∃ continuous maps f : X → Y, g :
Y → X, h : Y → Z and k : Z → Y such that g ◦ f 1 X , f ◦ g 1Y , h ◦ k 1 Z
and k ◦ h 1Y . Now h ◦ f : X → Z and g ◦ k : Z → X are continuous maps
such that (h ◦ f ) ◦ (g ◦ k) = h ◦ ( f ◦ g) ◦ k h ◦ 1Y ◦ k = h ◦ k 1 Z and (g ◦
k) ◦ (h ◦ f ) = g ◦ (k ◦ h) ◦ f g ◦ 1Y ◦ f = g ◦ f 1 X . Consequently, h ◦ f is
a homotopy equivalence with g ◦ k homotopy inverse ⇒ X Z . The equivalence
relation divides the set of spaces up to homotopy equivalent classes. ❑
Definition 2.2.11 The homotopy equivalent class containing X is called the homo-
topy type of X .
Remark 2.2.12 Two topological spaces X and Y are homotopy equivalent or of

the same homotopy type if there exists a homotopy equivalence f ∈ C(X, Y ). For
example, D2 and { p} in Example 2.2.9 are homotopy equivalent spaces. The home-
omorphic spaces are said to have the same topological type. On the other hand, the
homotopy equivalent spaces are said to have the same homotopy type.
Proposition 2.2.13 Two homeomorphic spaces have the same homotopy type.
Proof Let X and Y be two homeomorphic spaces and f : X → Y be a homeomor-

phism. Then its inverse g = f −1 : Y → X is continuous and satisfies the conditions:
g ◦ f = f −1 ◦ f = 1 X 1 X and f ◦ g = f ◦ f −1 = 1Y 1Y by reflexivity of the
relation . Consequently, f is a homotopy equivalence. Hence X and Y are of the
same homotopy type. ❑
Remark 2.2.14 The converse of Proposition 2.2.13 is not true. For example, the
disk D n is of the same homotopy type of a single point { p} ⊂ D n but D n is not
homeomorphic to { p}.
2.2 Homotopy Equivalence 61
Proposition 2.2.15 Any continuous map homotopic to a homotopy equivalence is a

homotopy equivalence.
Proof Let C(X, Y ) denote the set of all continuous maps from X to Y and f ∈
C(X, Y ) be a homotopy equivalence. Suppose g ∈ C(X, Y ) is such that f g. Now
f is a homotopy equivalence ⇒ ∃ h ∈ C(Y, X ) such that h ◦ f 1 X and f ◦ h
1Y . Again f g ⇒ f ◦ h g ◦ h ⇒ 1Y g ◦ h ⇒ g ◦ h 1Y . Similarly, h ◦ g
1 X . Consequently, g is a homotopy equivalence. ❑
Definition 2.2.16 A continuous map f : X → S n is called inessential if f is homo-
topic to a continuous map of X into a single point of S n (i.e., if f is homotopic to
a constant map). Otherwise f is called essential. In general, a map f ∈ C(X, Y ) is
said to be nullhomotopic or inessential if it is homotopic to some constant map.
Example 2.2.17 Let X = Y = I . Define f, g : I → I by f (t) = t and g(t) = 0, ∀ t ∈
I . Then f is the identity map and g is a constant map. Define F : I × I → I by
F(t, s) = (1 − s)t. Then F : f g ⇒ f is nullhomotopic.
Remark 2.2.18 Two nullhomotopic maps may not be homotopic.
Example 2.2.19 Let X be a connected space and Y be not a connected space. Let y0
and y1 be points in distinct components of Y . Let f 0 (x) = y0 and f 1 (x) = y1 , ∀ x ∈
X be two constant maps from X to Y . If possible, let f 0 f 1 . Then ∃ a continuous
map F : X × I → Y such that F : f 0 f 1 . Since X × I is connected and F is
continuous, F(X × I ) must be connected, which contradicts the fact that Y is not
connected.
Proposition 2.2.20 Let f, g : (X, A) → (Y, B) be pairs of continuous maps such
that f g as maps of pairs. Then the induced maps f˜, g̃ : X/A → Y/B (corre-
sponding quotient spaces) are also homotopic.
Proof Let H : (X × I, A × I ) → (Y, B) be a homotopy between f and g. Then

H induces a function H̃ : (X/A) × I → Y/B such that the diagram in Fig. 2.5 is
commutative, where p and q are the identification maps. Since H̃ ◦ ( p × 1d ) =
q ◦ H is continuous, I is locally compact and Hausdorff, p × 1d is an identification
map, it follows that H̃ is a (based) homotopy between f˜ and g̃, where base points of
X/A and Y/B are respectively the points to which A and B are identified. ❑
Corollary 2.2.21 Let f : (X, A) → (Y, B) be a homotopy equivalence of pairs.
Then f˜ : X/A → Y/B is a (based) homotopy equivalence.
Proof As f : (X, A) → (Y, B) is a homotopy equivalence, there exists a map g :
(Y, B) → (X, A) such that g ◦ f 1 X and f ◦ g 1Y . Hence the corollary follows
from the Proposition 2.2.20. ❑
Fig. 2.5 Diagram for X ×I

H /Y
identification map
p×1d q

X/A × I / Y /B
H̃
2.3 Homotopy Classes of Maps
This section continues the study of homotopy classes of continuous maps given in
Sect. 2.2. These classes play an important role in the study of algebraic topology as
depicted throughout the book. Homotopy theory studies those properties of topolog-
ical spaces and continuous maps which are invariants under homotopic maps, called
homotopy invariants.
Let [X, Y ] be the set of homotopy classes of continuous maps from X to Y : by
keeping X fixed and varying Y , this set is an invariant of the homotopy type of
Y , in the sense that if Y Z , then there exists a bijective correspondence between
the sets [X, Y ] and [X, Z ]. Similar result holds for pairs of topological spaces and
hence for pointed topological spaces. Many homotopy invariants can be obtained
from the sets [X, Y ] on which some short of algebraic structure is often given.for
particular choice of X and Y . Most of the classical invariants of algebraic topology
are homotopy invariants. Many homotopy invariants can be obtained by specializing
the sets [X, Y ].
The following two natural problems are posed in this section but solved in Sect. 2.4.
(i) Given a pointed topological space Y , does there exist a natural product defined
in [X, Y ] admitting the set [X, Y ] a group structure for all pointed topological
spaces X ?
(ii) Given a pointed topological space X , does there exist a natural product defined
in [X, Y ] admitting the set [X, Y ] a group structure for all pointed topological
space Y ?
In this section we work in the homotopy category Htp∗ of pointed topological
spaces and their base point preserving continuous maps. Thus [X, Y ] is the set of
morphisms from X to Y in the homotopy category Htp ∗ of pointed topological
spaces. This set [X, Y ] only depends on homotopy types of X and Y . Given two
continuous maps f : X → Y and g : Y → Z , we can compose them and obtain g ◦
f : X → Z . The homotopy class of g ◦ f depends only on the homotopy classes of
f and g. So the composition with g gives a function
g∗ : [X, Y ] → [X, Z ]
and the composition with f gives a function
f ∗ : [Y, Z ] → [X, Z ].
Theorem 2.3.1 Let X, Y, Z be pointed topological spaces and f : Y → Z be a base

point preserving continuous map. Then f induces a function. f ∗ : [X, Y ] → [X, Z ]
satisfying the following properties:
(a) If f h : Y → Z , then f ∗ = h ∗ ;
(b) If 1Y : Y → Y is the identity map, then 1Y ∗ is the identity function;
2.3 Homotopy Classes of Maps 63
(c) If g : Z → W is another base point preserving continuous map, then (g ◦ f )∗ =

g∗ ◦ f ∗ .
Proof Define f ∗ : [X, Y ] → [X, Z ] by the rule f∗ ([α]) = [ f ◦ α], ∀ [α] ∈ [X, Y ].
Since α β ⇒ f ◦ α f ◦ β ⇒ f ∗ ([α]) = f ∗ [β] ⇒ f ∗ is independent of the
choice of the representatives of the classes. Hence f ∗ is well defined.
(a) Consider the functions f ∗ , h ∗ : [X, Y ] → [X, Z ]. Then h ∗ ([α]) = [h ◦ α] =
[ f ◦ α] = f ∗ ([α]), since f h ⇒ f ◦ α h ◦ α = f ∗ ([α]), ∀ [α] ∈ [X, Y ].
Hence h ∗ = f ∗ .
(b) 1Y ∗ : [X, Y ] → [X, Y ] is given by 1Y ∗ ([α]) = [1Y ◦ α] = [α], ∀ [α] ∈ [X, Y ].
Hence 1Y ∗ is the identity function.
(c) (g ◦ f )∗ : [X, Y ] → [X, W ] is given by (g ◦ f ) ∗ ([α]) = [(g ◦ f ) ◦ α] = [g ◦
( f ◦ α)] = (g∗ ◦ f ∗ )[α], ∀ [α] ∈ [X, Y ]. Hence (g ◦ f )∗ = g∗ ◦ f ∗ .
❑
Corollary 2.3.2 If f : Y → Z is a homotopy equivalence, then f ∗ : [X, Y ] →

[X, Z ] is a bijection for every topological space X .
Proof If f ∈ C(Y, Z ) is a homotopy equivalence, then ∃ g ∈ C(Z , Y ) such that

g ◦ f 1Y and f ◦ g 1 Z . Hence (g ◦ f )∗ = g∗ ◦ f ∗ is the identity function and
( f ◦ g)∗ = f ∗ ◦ g∗ is also identity function ⇒ f ∗ is a bijection with g∗ as its
inverse. ❑
Corollary 2.3.3 If Y Z , then there exists a bijection ψ : [X, Y ] → [X, Z ] for

every topological space X .
Proof If Y Z , then ∃ a homotopy equivalence f ∈ C(Y, Z ). Hence f∗ = ψ :

[X, Y ] → [X, Z ] is a bijection for every X by Corollary 2.3.2. ❑
Corollary 2.3.4 Given a pointed topological space X , there exists a covariant

functor π X from the homotopy category of pointed topological spaces to the cat-
egory of sets and functions defined by π X (Y ) = [X, Y ] and if f : Y → Z is a base
point preserving continuous map, then π X ( f ) = f ∗ : [X, Y ] → [X, Z ] is defined by
f ∗ ([α]) = [ f ◦ α].
Proof The Corollary follows from Theorem 2.3.1. ❑
We obtain the corresponding dual results.
Theorem 2.3.5 A base point preserving continuous map f : Y → Z induces a func-

tion f ∗ : [Z , X ] → [Y, X ] for every pointed space X , satisfying the following prop-
erties:
(a) f h : Y → Z ⇒ f ∗ = h ∗ ;
(b) 1Y : Y → Y is the identity map ⇒ 1Y ∗ is the identity function;
(c) If g : Z → W is another base point preserving continuous map, then (g ◦ f )∗ =
f ∗ ◦ g∗ .
Proof Similar to the proof of Theorem 2.3.1. ❑
Corollary 2.3.6 If f : Y → Z is a homotopy equivalence, then f ∗ : [Z , X ] →

[Y, X ] is a bijection for every pointed topological space X .
Corollary 2.3.7 If Y Z , then ∃ a bijection ψ : [Z , X ] → [Y, X ] for every pointed

topological space X .
Corollary 2.3.8 Given a pointed topological space X , there exists a contravariant

functor π X from the homotopy category of pointed topological spaces to the category
of sets and functions.
Converses of Corollaries 2.3.2 and 2.3.6 are also true.
Theorem 2.3.9 If f : Y → Z is a base point preserving continuous map such that

(a) f ∗ : [X, Y ] → [X, Z ] is a bijection for all pointed topological spaces X , then
f is a homotopy equivalence.
(b) f ∗ : [Z , X ] → [Y, X ] is a bijection for all pointed topological spaces X , then
Proof (a) In particular, f ∗ : [Z , Y ] → [Z , Z ] is a bijection (by hypothesis) ⇒ ∃ a

continuous map g : Z → Y such that f ∗ ([g]) = [1 Z ] ⇒ f ◦ g 1 Z . Similarly,
g∗ ([ f ]) = [1Y ] ⇒ g ◦ f 1Y . Consequently, f is a homotopy equivalence.
(b) Similar to (a).
❑
2.4 H-Groups and H-Cogroups
This section conveys the concept of a grouplike space, called an H -group and its
dual concept, called an H -cogroup as a continuation of the study of the set [X, Y ] by
considering the problem: when is the set [X, Y ] a group for every pointed topological
space X (or for every pointed topological space Y )? The concepts of H -groups
and H -cogroups arose through the study of such problems. These concepts develop
homotopy theory. The loop spaces and suspension spaces of pointed topological
spaces play an important role in the study of homotopy theory. Loop spaces of pointed
spaces provide an extensive class of H -groups. On the other hand suspension spaces
of pointed topological spaces form an extensive class of H -cogroups, a dual concept
of H -group.
We consider topological spaces Y such that [X, Y ] admits a group structure for
all X . There is a close relation between the natural group structures on [X, Y ] for
all X and ‘grouplike’ structure on Y . Before systematic study of the homotopy sets
[X, Y ] or [(X, x0 ), (Y, y0 )] by M.G. Barratt (1955) in his paper, the concept of an
H -space introduced by H. Hopf in 1933 arose as a generalization of a topological
group which is used to solve the above problem.
2.4 H -Groups and H -Cogroups 65
2.4.1 H-Groups and Loop Spaces
This subsection continues to study H -groups by specializing the the sets [X, Y ] and
presents loop spaces which form an important class of H -groups. Given pointed
topological spaces X and Y , we often give the set [X, Y ] some sort of algebraic
structure. With this objective this subsection studies a grouplike space which is a
group up to homotopy, called an H -group. More precisely, this subsection introduces
the concepts of H -groups to obtain algebraic structures on the set of certain homotopy
classes of continuous maps and introduces the concept of an H -group with loop
spaces as illustrative examples. An H -group is a generalization of a topological
group. Such groups were first introduced by H. Hopf in 1941 and they are named in
his honor. Loop spaces of pointed topological spaces constitute an extensive class of
H -groups.
The motivation of this study is to describe an additional structure needed on
a pointed space P so that π P (X ) = [X, P] is a group and for f : X → Y, f ∗ =
π P ( f ) : [Y, P] → [X, P] is a group homomorphism. If f : X → Y and g : X →
Z are continuous maps, we define ( f, g) : X → Y × Z to be the map ( f, g)(x) =
( f (x), g(x)), ∀ x ∈ X.
If Y is a topological group, then [X, Y ] admits a group structure by
Theorem 2.4.1.
Now the following two natural questions arise:
(i) Given a pointed topological space Y , does there exist a natural product defined
in [X, Y ] for all pointed topological spaces X ?
(ii) Given a pointed topological space X , does there exist a natural product defined
in [X, Y ] for all pointed topological spaces Y ?
We start with a topological group P (see Appendix A) followed by H -groups and
H -cogroups. The essential feature which is retained in an H space is a continuous
multiplication with a unit. There is a significant class of topological spaces which
are H -spaces but not topological groups.
Theorem 2.4.1 Let X be any pointed topological space and P be a topological

group with identity element as base point. Then [X, P] can be given the structure of
a group.
Proof Given two base point preserving continuous maps f, g : X → P, let their
product f ·g be defined by pointwise multiplication, i.e., f ·g : X → P is defined by
( f ·g)(x) = f (x)g(x), where the right side is the group multiplication μ in P. Thus
f ·g = μ ◦ ( f × g) ◦ , where (x) = (x, x) is the diagonal map, i.e., f ·g is the
composite
f ×g μ
X −−−→ X × X −−−−−→ P × P −−−→ P
Then f ·g is another continuous map from X to P. Moreover, given further maps

f , g : X → P such that f f and g g then f ·g f ·g (by Ex. 2 of Sect. 2.11)
⇒ [ f ]·[g] = [ f ·g] ⇒ the law of composition f ·g carries over to give an operation

‘◦’ on [X, P]. Then the group structure on [X, P] follows from the corresponding
properties of the topological group. Consequently, ([X, P], ◦) is a group. ❑
Corollary 2.4.2 If P is a topological group and f : X → Y is a base point preserv-

ing continuous map, then f induces a group homomorphism f ∗ : [Y, P] → [X, P]
defined by f ∗ ([α]) = [α ◦ f ], ∀ [α] ∈ [Y, P].
Theorem 2.4.3 Given a topological group P, there exists a contravariant functor

π P from the homotopy category of pointed topological spaces to the category of
groups and homomorphisms.
Proof It follows from Theorem 2.4.1 and Corollaries 2.4.2 and 2.3.8. ❑
Remark 2.4.4 Given a topological group P, the group structure on [X, P] is endowed
from the group structure on the set of base point preserving continuous maps from
X to P. We come across some situations in which [X, P] admits a natural group
structure, but the set of base point preserving continuous maps from X to P has no
group structure. If P is a pointed topological space having the same homotopy type

as some topological group P , then π P is naturally equivalent to π P . Hence π P can
be regarded as a functor to the category of groups.
Example 2.4.5 S 1 = {z ∈ C : |z| = 1} is an abelian topological group under usual

multiplication of complex numbers. Then [X, S 1 ] is an abelian group and if f : X →
Y , then f ∗ : [Y, S 1 ] → [X, S 1 ] is a homomorphism of groups.
Example 2.4.6 S 3 is a topological group (the multiplicative group of quaternions of

norm 1). Then [X, S 3 ] is a group and if f : X → Y , then f ∗ : [Y, S 3 ] → [X, S 3 ] is
group homomorphism.
Remark 2.4.7 If Y is a topological group, then a product f ·g : X → Y is given

by ( f ·g)(x) = f (x)·g(x) ∀ x ∈ X . To solve the problems (i) and (ii) we search for
certain other classes of pointed topological spaces, called Hopf spaces (H -spaces)
and Hopf groups (H -groups).
Definition 2.4.8 A pointed topological space P with a base point p0 , is called an

H -space if there exists a continuous multiplication μ : P × P → P, ( p, p ) → pp
for which the (unique) constant map c : P → p0 ∈ P is a homotopy identity, i.e.,
each composite
(c,1 P ) μ (1 P ,c) μ
P −−−−−−→ P × P −−−→ P and P −−−−−−→ P × P −−−→ P
is homotopic to 1 P , i.e., if each of the triangles in Fig. 2.6 is homotopy commutative;

sometimes it is written as an ordered pair (P, μ).
Fig. 2.6 H -space (c,1 ) (1 ,c)

P DD
P
/P ×P o P
P
DD z
DD zz
DD zzz
DD z
D μ zz
1P DD zz
DD zzz 1P
DD zz
D! z
}z
P
Remark 2.4.9 The above homotopy commutativity means that the maps P →
P, p → p0 p, pp0 are homotopic to 1 P rel { p0 }. In other words, ∃ homotopies
L and R from P × I → P such that L( p, 0) = p0 p, L( p, 1) = p, L( p0 , t) =
p0 , R( p, 0) = pp0 , R( p, 1) = p and R( p0 , t) = p0 , ∀ p ∈ P and ∀ t ∈ I .
Definition 2.4.10 An H -space (P, μ) is said to be homotopy associative if the square

in Fig. 2.7 is homotopy commutative, i.e., μ ◦ (μ × 1 P ) μ ◦ (1 P × μ), i.e., the two
maps
P × P × P → P, ( p1 , p2 , p3 ) → ( p1 p2 ) p3 , p1 ( p2 p3 ) are homotopic rel { p0 }.
Definition 2.4.11 A continuous map φ : P → P is said to be homotopy inverse for

(1 P ,φ) μ (φ,1 P )
P and μ if each of the composites P −−−−−−→ P × P −−−→ P and P −−−−−−→
μ
P × P −−−→ P is homotopic to the constant map c : P → P, p → p0 ∈ P, i.e.,
each of the maps P → P, p → pφ( p), φ( p) p is homotopic to c rel { p0 }.
Definition 2.4.12 An associative H -space P with an inverse is called an H -group
or a generalized topological group. The point p0 ∈ P is called the homotopy unit of
(P, μ).
Definition 2.4.13 A multiplicative μ on an H -space P is said to be homotopy abelian
if the triangle in Fig. 2.8 is homotopy commutative, where T ( p, p ) = ( p , p), i.e.,
the two maps P × P → P, ( p, p ) → pp , p p are homotopic rel { p0 }.
Example 2.4.14 (i) Every topological group is an H -space with homotopy inverse.
But its converse is not true in general (see (ii)). In particular, Lie groups (for
Fig. 2.7 Homotopy μ×1P

P ×P ×P /P ×P
associative H -space
1P ×μ μ
μ
P ×P /P
Fig. 2.8 Abelian H -space T /P ×P

P × P GG w
GG
GG www
w
μ GG ww μ
G# {w
w
P
Fig. 2.9 Homotopy μ

P ×P /P
homomorphism
α×α α

P × P / P
μ
example, the general linear group, G L(n, R) or the orthogonal group O(n, R),
see Appendix A) are H -spaces.
(ii) The infinite real projective space R P ∞ = R P n and infinite complex projec-
n≥0
tive space CP ∞ = CP n are H -spaces but not topological groups.
n≥0
Definition 2.4.15 Let (P, μ) and (P , μ ) be two H -spaces. Then a continuous map
α : P → P is called a homotopy homomorphism if the square in Fig. 2.9 is homo-
topy commutative.
Clearly, H -groups (H -spaces) and homotopy homomorphisms form a category.
Theorem 2.4.16 A pointed topological space having the same homotopy type of an
H -space (or an H -group) is itself an H -space (or H -group) in such a way that the
homotopy equivalence is a homotopy homomorphism.
Proof Let (P, μ) be an H -space and P be a pointed topological space having the
homotopy type of the space P. Then there exist continuous maps f : P → P and
g : P → P such that g ◦ f 1 P and f ◦ g 1 P . Define μ : P × P → P to be
the composite
g×g μ f
P × P −−−−−→ P × P −−−→ P −−−→ P i.e., μ = f ◦ μ ◦ (g × g).
Then μ is a continuous multiplication in P . Moreover, the composites
(1,c ) μ
P −−−−−→ P × P −−−→ P (2.9)
g (1,c) μ f
and P −−−→ P −−−−−→ P × P −−−→ P −−−→ P (2.10)
are equal. As P is an H -space, the composite in (2.10) is homotopic to the compos-

ite f ◦ g, because f ◦ μ ◦ (1, c) ◦ g f ◦ 1 P ◦ g f ◦ g. Again, f ◦ g 1 P ⇒
μ ◦ (1, c ) 1 P by (2.9) and (2.10). Similarly, μ ◦ (c , 1) 1 P . Consequently, P
is an H -space with continuous multiplication μ . Since the diagram in Fig. 2.10 is
homotopy commutative, g is a homotopy homomorphism and so is f . If μ is homo-
topy associative or homotopy abelian, and if φ : P → P is a homotopy inverse for
P, then the composite map f ◦ φ ◦ g : P → P is a homotopy inverse for P . ❑
Generalizing the Theorem 2.4.1 we have the following theorem:


P × P / P
commutative square
g×g g
μ
P ×P /P
Theorem 2.4.17 If X is any pointed topological space and P is an H -group, then

[X, P] can be given the structure of a group.
Proof Similar to the proof of Theorem 2.4.1. ❑
Theorem 2.4.18 Let P be a pointed space with base point p0 and p1 , p2 : P × P →
P be the projections from the first and the second factors respectively. If i 1 , i 2 : P →
P × P are inclusions defined by i 1 ( p) = ( p, p0 ), i 2 ( p) = ( p0 , p) for all p ∈ P,
then the pointed space P is an H -space iff there exists a map μ : P × P → P such
that μ ◦ i 1 μ ◦ i 2 . Moreover, this map μ satisfies the condition [μ] = [ p1 ]·[ p2 ] and
if f 1 , f 2 : X → P are maps then [ f 1 ]·[ f 2 ] is the homotopy class of the composite
f1 × f2 μ
X −−−→ X × X −−−−−−
→ P × P −−−→ P
Proof It follows from hypothesis that p1 ◦ i 1 = p2 ◦ i 2 = 1 P and p1 ◦ i 2 = p2 ◦

i 1 = c, where c : P → p0 is the constant map. If μ = P × P → P is a map such
that μ ◦ i 1 1 P μ ◦ i 2 , then this μ is a multiplication admitting P the structure of
an H -space.
Conversely suppose P is a pointed space and μ : P × P → P is a map such that
μ ◦ i 1 1 P μ ◦ i 2 . Then given maps f 1 , f 2 : X → P define f 1 · f 2 = μ ◦ ( f 1 ×
f 2 ) ◦ . This composition is compatible with homotopy, and induces a natural prod-
uct in [X, P]. Hence P is an H -space and p1 · p2 = μ ◦ ( p1 × p2 ) ◦ = μ ◦ 1 P =
μ. Consequently, P is an H -space. ❑
Remark 2.4.19 For an arbitrary H -space P, the multiplication defined in [X, P] may
not be associative. If f 1 , f 2 , f 3 : X → P are maps, then ( f 1 · f 2 )· f 3 and f 1 ·( f 2 · f 3 )
are by definition the homotopy classes of the composites
3 f1 × f2 × f3 μ×1 P μ
X −−−−→ X × X × X −−−−−−−−→ P × P × P −−−−−→ P × P −−−→ P,
3 f1 × f2 × f3 1 P ×μ μ
X −−−−→ X × X × X −−−−−−−−→ P × P × P −−−−−→ P × P −−−→ P,
where 3 = ( × 1 X ) ◦ = (1 X × ) ◦ : X → X × X × X is the diagonal

map. Hence the conditionμ ◦ (μ × 1 P ) μ ◦ (1 P × μP) is sufficient for associa-
tivity. It is also necessary that if X = P × P × P and f 1 , f 2 , f 3 are projections of
P × P × P into P, then ( f 1 × f 2 × f 3 ) ◦ 3 is the identity map.
The above discussion can be stated in the form of the following interesting result:
Theorem 2.4.20 The set [X, P] admits a monoid structure natural with respect to
X iff P is a homotopy associative H -space.
Theorem 2.4.21 Let (P, p0 ) be an H -group with multiplication μ and homotopy

inverse φ. Then for every pointed topological space (X, x0 ), the set [(X, x0 ), (P, p0 )]
denoted [X, P] can be given the structure of a group if we define the product [ f ]·[g]
to be the homotopy class of the composite map
f ×g μ
X −−−→ X × X −−−−−→ P × P −−−→ P,
where is the diagonal map given by (x) = (x, x). The identity element of the
group is the class [c] of the constant map c : X → p0 and the inverse of [ f ] is given
by [ f ]−1 = [φ ◦ f ]. If μ is homotopy commutative, then [X, P] is abelian.
Proof Define the product [ f ]·[g] = [μ ◦ ( f × g) ◦ ]. We claim that [ f ]·[g] is well
defined. To show this, let H : X × I → P be a homotopy between f and f and
G : X × I → P a homotopy between g and g . Define a homotopy M : X × I → P
by Mt (x) = M(x, t) = μ(H (x, t), G(x, t)). Then M0 = μ ◦ ( f × g) ◦ and M1 =
μ ◦ ( f × g ) ◦ ⇒ μ ◦ ( f × g) ◦ μ ◦ ( f × g ) ◦ ⇒ [ f ]·
[g] = [ f ]·[g ] ⇒ the multiplication is independent of the choice of representatives
of the classes. ⇒ the multiplication is well defined.
We now prove the associativity of the multiplication. Let h : (X, x0 ) → (P, p0 )
be a third map. Then
[ f ]·([g]·[h]) = [μ ◦ ( f × {μ ◦ (g × h) ◦ }) ◦ ]
= [μ ◦ (1 × μ) ◦ ( f × g × h) ◦ (1 × ) ◦ ]
= [μ ◦ (μ × 1) ◦ ( f × g × h)
◦ ( × 1) ◦ ] by homotopy associativity of μ
= [μ ◦ ({μ ◦ ( f × g) ◦ } × h) ◦ ]
= ([ f ]·[g])·[h].
Again [ f ]·[c] = [μ ◦ ( f × c) ◦ ] = [μ ◦ (1, c) ◦ f ] = [1 P ◦ f ] = [ f ] and

similarly, [c]·[ f ] = [ f ], ∀ [ f ] ∈ [X, P]] ⇒ [c] is an identity element for [X, P].
Finally, [φ ◦ f ]·[ f ] = [μ ◦ ((φ ◦ f ) × f ) ◦ ] = [μ ◦ (φ, 1) ◦ f ] = [c] and [ f ]·
[φ ◦ f ] = [c] ⇒ [ f ]−1 = [φ ◦ f ]. Consequently, if (P, p0 ) is an H -group, then
[X, P] is a group. If μ is homotopy commutative, the last part follows immediately
❑
The converse of the Theorem 2.4.21 is also true.
Theorem 2.4.22 The set [X, P] admits a group structure natural with respect to X
iff P is an H -group.
Proof It follows from the Theorems 2.4.20 and 2.4.21. ❑
Remark 2.4.23 The set [X, P] can be endowed with a monoid structure natural with
respect to X iff P is a homotopy associative H -space.
Theorem 2.4.24 If g : X → Y is a base point preserving continuous map and P is

an H -group, then the induced function g ∗ : [Y, P] → [X, P] defined by g ∗ ([α]) =
[α ◦ g] is a group homomorphism. In particular, if g is a homotopy equivalence, then
g ∗ is an isomorphism.
Proof Given continuous maps f 1 , f 2 : Y → P, we have ( f 1 · f 2 ) ◦ g = μ ◦

( f 1 × f 2 ) ◦ ◦ g = μ ◦ ( f 1 × f 2 ) ◦ (g × g) ◦ = μ ◦ ( f 1 ◦ g × f 2 ◦ g) ◦ =
( f 1 ◦ g)·( f 2 ·g). Then g ∗ ([ f 1 ]·[ f 2 ]) = g ∗ [ f 1 ]·g ∗ [ f 2 ] ⇒ g ∗ is a homomorphism. If g
is a homotopy equivalence, then ∃ a continuous map f : Y → X such that f ◦ g 1 X
and g ◦ f = 1Y . Then ( f ◦ g)∗ = g ∗ ◦ f ∗ = 1d and (g ◦ f )∗ = f ∗ ◦ g ∗ = 1d ⇒ g ∗
is an isomorphism. ❑
Theorem 2.4.25 If P is an H -group, π P is a contravariant functor from the homo-

topy category of pointed topological spaces to the category of groups and homomor-
phisms. If P is an abelian H -group, then the functor π P takes values in the category
of abelian groups.
Proof Define the object function by π P (X ) = [X, P], which is a group for every
pointed space X by Theorem 2.4.22. If g : X → Y is a base point preserving map,
define the morphism function by π P (g) = g ∗ by Theorem 2.4.24. Then the theorem
follows. ❑
The converse of the Theorem 2.4.25 is also true.

Theorem 2.4.26 If P is a pointed topological space such that π P takes values in the
category of groups, then P is an H -group (abelian if π P takes values in the category
of abelian groups). Moreover, for any pointed space X , the group structures on
π P (X ) and on [X, P] given in the Theorem 2.4.21 coincide.
Proof Let p1 : P × P → P and p2 : P × P → P be the projections on the first and

second factor respectively. Let μ : P × P → P be a map such that [μ] = [ p1 ]·[ p2 ],
where · is the product in the group [P × P, P]. For any continuous maps f, g : X →
P, the induced map ( f, g)∗ : [P × P, P] → [X, P] is a homomorphism and
[μ ◦ ( f, g)] = ( f, g)∗ [μ] = ( f, g)∗ ([ p1 ]·[ p2 ]) = ( f, g)∗ [ p1 ]·( f, g)∗ [ p2 ] = [ f ]·[g]
implies that the multiplication in [X, P] is induced by the multiplication map μ.

Let X be a one-point space. The unique map X → P represents the identity ele-
ment of the group [X, P]. Since the unique map P → X induces a homomorphism
[X, P] → [P, P], it follows that the composite P → X → P, which is the constant
map c : P → P represents the identity element of [P, P]. Hence it follows that
μ ◦ (1 P , c) 1 P and μ ◦ (c, 1 P ) I P . Consequently, P is an H -space. To prove
that μ is homotopy associative, let q1 , q2 , q3 : P × P × P → P be the projections.
Then [μ ◦ (1 P × μ)] = (1 P × μ)∗ [μ] = (1 P × μ)∗ [ p1 ]·(1 P × μ)∗ [ p2 ] = [q1 ]·[μ ◦
(q2 , q3 )] = [q1 ]·([q2 ]·[q3 ]). Similarly, [μ ◦ (μ × 1 P )] = ([q1 ] ◦ [q2 ]) ◦
[q3 ]. Since [P × P × P, P] has an associative multiplication, it follows that
μ ◦ (1 P × μ) μ ◦ (μ × 1 P ).
Finally, we show that P has a homotopy inverse. Let φ : P → P be the map such
that [1 P ]·[φ] = [c]. Then μ ◦ (1 P , φ) c. Similarly, μ ◦ (φ, 1 P ) c. Hence, φ is a
homotopy inverse for P and μ.
Consequently, P is an H -group. Moreover, if [P × P, P] is an abelian group, a
similar argument shows that P is an abelian H -group. ❑
Given two H -groups P and P , we now compare between the contravariant func-

tors π P and π P .
Theorem 2.4.27 Let α : P → P , be a continuous map between H -groups. Then

α induces a natural transformation α∗ : π P → π P in the category of H -groups iff
α is a homomorphism.

Proof For each pointed topological space X , define α∗ (X ) : π P (X ) → π P (X ) by
the rule α∗ (X )[h] = [α ◦ h], ∀ [h] ∈ π P (X ). Then diagram in Fig. 2.11 is commu-

tative, for every f : Y → X , because, (π P ( f ) ◦ α∗ (X ))[h] = π P ( f )([α ◦ h]) =
[(α ◦ h) ◦ f ] and (α∗ (Y ) ◦ π P ( f ))[h] = α∗ (Y )[h ◦ f ] = [α ◦ (h ◦ f )], which are
equal. Hence α∗ is a natural transformation.
The converse part is left as an exercise. ❑
We now investigate the question of existence of homotopy inverses for a homotopy

associative H -space.
Theorem 2.4.28 If P is a homotopy associative H -space, then P is an H -group

if and only if the shear map ψ : P × P → P × P, given by ψ(x, y) = (x, x y) is a
Proof Case I. First we consider the particular case when P is a topological group.
Then the map ψ is a homeomorphism with inverse ψ −1 : P × P → P × P defined
by ψ −1 (u, v) = (u, u −1 v). Let j = p2 ◦ ψ −1 ◦ i 1 , where i 1 : P → P × P is the
inclusion, defined by i 1 (y) = (y, y0 ), where y0 is the base point of P and p1 , p2 :
P × P → P be the projections on the first and the second factor respectively. Then
[ j] is the inverse of the homotopy class of the identity map 1 P ∈ [P, P], so that the
composites
j×1 P μ
P −−−→ P × P −−−−−→ P × P −−−→ P,
1P × j μ
P −−−→ P × P −−−−−→ P × P −−−→ P
are each nullhomotopic.
Fig. 2.11 Natural α∗ (X)

/ π P (X)
π P (X)
transformation α∗

π P (f )=f ∗ f ∗ =πP (f )
α∗ (Y )

π P (Y ) /π P
(Y )
Case II. We now consider the general case. Let ψ be a homotopy equivalence with
homotopy inverse φ. Define j ∈ [P, P] by j = p2 ◦ φ ◦ i 1 . Then
( p1 ◦ ψ)(x, y) = p1 (ψ(x, y)) = p1 (x, x y) = x

= p1 (x, y), ∀ (x, y) ∈ P × P ⇒ p1 ◦ ψ = p1 .
Again, ( p2 ◦ ψ)(x, y) = p2 (x, x y) = x y = μ(x, y), ∀ (x, y) ∈ P × P ⇒ p2 ◦

ψ = μ.
Hence
1d

p1 p1 ◦ ψ ◦ φ = ( p1 ◦ ψ) ◦ φ = p1 ◦ φ,
1d

p2 p2 ◦ ψ ◦ φ = ( p2 ◦ ψ) ◦ φ = μ ◦ φ.
In particular, p1 ◦ φ ◦ i 1 p1 ◦ i 1 = 1 P , since ( p1 ◦ i 1 )(y) = p1 (y, y0 ) = y =

1 P (y), ∀ y ∈ P.
Hence
μ ◦ (1 P × j) ◦ = μ ◦ ( p1 ◦ φ ◦ i 1 × p2 ◦ φ ◦ i 1 ) ◦
= μ ◦ ( p1 × p2 ) ◦ (φ ◦ i 1 × φ ◦ i 1 ) ◦
= μ ◦ ( p1 × p2 ) ◦ ◦ φ ◦ i 1
1d

= μ ◦ φ ◦ i 1 = p2 ◦ ψ ◦ φ ◦i 1 p2 ◦ i 1 c,
where c : P → p0 ∈ P is the constant map.

Hence j is a right inverse of the identity map.
It follows from the above argument that every element of [X, P] has a left
inverse, and hence [X, P] is a group. Conversely, if P is an H -group, then the
map φ : (u, v) → (u, j (u)v) is a homotopy inverse of the shear map ψ, because
(ψ ◦ φ)(u, v) = ψ(φ(u, v)) = ψ(u, j (u)v) = (u, u j (u)v) = (u, p0 v) and since p0
is a homotopy unit, it follows that ψ ◦ φ 1d . Similarly, φ ◦ ψ 1d . Consequently,
ψ is a homotopy equivalence. ❑
Remark 2.4.29 Some of the techniques which apply to topological groups can be
applied to H -spaces, but not all. From the viewpoint of homotopy theory, it is not
the existence of a continuous inverse which is the important distinguishing feature,
but rather the associativity of multiplication. If we consider S 1 , S 3 and S 7 as the
complex, quaternionic and Cayley numbers of unit norm, these spaces have contin-
uous multiplication. The multiplication in the first two cases are associative but not
associative in the last case. S ! and S 3 are topological groups. The spheres S 1 , S 3
and S 7 are the only spheres that are H -spaces proved by J.F. Adams (1930–1989) in
(1962).
Remark 2.4.30 Every topological group is an H -group.

We now describe another important example of an H -group. Loop spaces form an

important class of grouplike spaces, called H -groups.
Definition 2.4.31 (Loop Space) Let Y be a pointed topological space with base point
y0 . The loop space of Y (based at y0 ) denoted Y (or (Y, y0 )), is defined to be the
space of continuous functions α : (I, I˙) → (Y, y0 ), topologized by the compact open
topology. Then (Y, y0 ) is considered as a pointed space with base point α0 equals
to the constant map c : I → y0 .
The elements of Y are called loops in Y .

Theorem 2.4.32 (Y, y0 ) is an H -group.
Proof Define a map μ : Y × Y → Y by

α(2t), 0 ≤ t ≤ 1/2
μ(α, β)(t) =
β(2t − 1), 1/2 ≤ t ≤ 1.
To show that μ is continuous, consider the evaluation map E : Y × I → Y

defined by E(α, t) = α(t). Since I is locally compact, by Theorem of exponential
correspondence (see Theorem 1.14.2 of Chap. 1) it is sufficient to show that the
composite map
μ×1d E
Y × Y × I −−−−−→ Y × I −−−→ Y
is continuous.
Then the theorem of exponential correspondence and the Pasting lemma show the
continuity of μ, since the above composite is continuous on each of the closed sets
Y × Y × [0, 21 ] and Y × Y × [ 21 , 1].
μ is associative: To show this define G : Y × Y × Y × I → Y by the rule

⎧
⎪
⎨ α( 1+s ),
4t
0 ≤ t ≤ (1 + s)/4
G(α, β, γ, s)(t) = β(4t − 1 − s), (1 + s)/4 ≤ t ≤ (2 + s)/4
⎪
⎩ γ( 4t−2−s ),
2−s
(2 + s)/4 ≤ t ≤ 1.
The continuity of G follows from the Pasting lemma. Clearly, G : μ ◦ (μ × 1d )

μ ◦ (1d × μ)
Existence of homotopy unit: If c : Y → Y is the constant map whose value is
the constant loop c : I → Y , c(t) = y0 , then μ ◦ (β, c) β and μ ◦ (c, β) β for
every loop β.
The first homotopy is given by F : Y × I → Y, where

β( 1+s
2t
), 0 ≤ t ≤ (1 + s)/2
F(β, s)(t) =
y0 , (1 + s)/2 ≤ t ≤ 1.
The continuity of F follows from Pasting lemma. The second homotopy is defined
in an analogous manner.
Existence of homotopy inverse: Let φ : Y → Y be a map such that φ(α)(t) =
α(1 − t). Then φ determines homotopy inverses. The homotopy H : Y × I →
Y , where
α(2(1 − s)t), 0 ≤ t ≤ 1/2
H (α, s)(t) =
α(2(1 − s)(1 − t)), 1/2 ≤ t ≤ 1,
begins at μ ◦ (α, φ(α)) and ends at c. The second homotopy is given in an analogous
manner.
Consequently, Y is an H -group. ❑
Definition 2.4.33 Given a pointed space X , iterated loop spaces n X are defined
inductively: n X = (n−1 X ) for ≥ 1 and 0 X is taken to be X .
Corollary 2.4.34 For n ≥ 1, n X is an H -group for every pointed space X .
Theorem 2.4.35 is a covariant functor from the category of pointed topological

spaces and continuous maps to the category of H -groups (Hopf groups) and their
continuous homomorphisms.
Proof If f : X → Y is base point preserving continuous maps, then f : (X ) →

(Y ) defined by (f )(α)(t) = f (α(t)) is a homomorphisms of H -groups. The
object function is given by X → (X ) and the morphism function is given by
f : (X ) → (X ). Then is a covariant functor. ❑
Theorem 2.4.36 For every pointed topological space Y , Y is an H -group and for
every pointed space X, [X, Y ] is a group. If f : X → X is a base point preserving
continuous map, then f ∗ : [X , Y ] → [X, Y ] is a group homomorphism.
Proof The theorem follows from Theorems 2.4.32 and 2.4.24. ❑
2.4.2 H-Cogroups and Suspension Spaces
This subsection conveys the dual concepts of H -groups, called H -cogroups intro-
duced by Beno Eckmann (1917–2008) and Peter John Hilton (1923–2010) in 1958. It
involves wedge products of pointed topological spaces. Suspension spaces of pointed
topological spaces form an extensive class of H -cogroups.
Recall that the wedge X ∨ X is viewed as the subspace X × {x 0 } ∪ {x 0 } × X of the
product space X × X . If pi : X × X → X, for i = 1, 2 are the usual projections onto
the first or second coordinate respectively, then define ‘projections’ qi : X ∨ X → X,
for i = 1, 2 by qi = pi | X ∨X ; each qi sends the appropriate copy of x ∈ X , namely,
(x, x 0 ) or (x0 , x) into itself.
Definition 2.4.37 A pointed topological space (X, x0 ) is called an H -cogroup

if there exists a base point preserving continuous map μ : X → X ∨ X, called comul-
tiplication, such that q1 ◦ μ 1 X q2 ◦ μ, (1 X ∨ μ) ◦ μ (μ ∨ 1 X ) ◦ μ
(co-associativity) and there exists a base point preserving continuous map h : X → X
such that (1 X , h) ◦ μ c (h, 1 X ) ◦ μ, (h is called an inverse), where c : X → X
is the constant map at x0 .
Remark 2.4.38 In an H -cogroup given maps f : X → Z and g : Y → Z in Top∗ ,

the map ( f, g) : X ∨ Y → Z is defined by the characteristic property:
( f, g)| X = f and ( f, g)|Y = g.
H -cogroup is now defined more explicitly keeping similarity with the definition
of H -group.
Definition 2.4.39 A pointed topological space X with base point x 0 is called an

H -cogroup if there exists a base point preserving continuous map
μ : X → X ∨ X,
called H -comultiplication such that the following conditions hold:

Existence of homotopy identity. If c : X → X is the (unique) constant map at x0 ,
then each composite map
μ (c,1 X ) μ (1 X ,c)
X −−−→ X ∨ X −−−−−−→ X and X −−−→ X ∨ X −−−−−−→ X
is homotopic to 1 X .
Homotopy associativity. The diagram in Fig. 2.12 is commutative up to homotopy,
i.e.,
(1 X ∨ μ) ◦ μ (μ ∨ 1 X ) ◦ μ.
Existence of homotopy inverse. There exists a map h : X → X such that each

composite map
μ (1 X ,h) μ (h,1 X )
X −−−→ X ∨ X −−−−−−→ X and X −−−→ X ∨ X −−−−−−→ X
is homotopic to c : X → X .

X /X ∨X
associativity of H -cogroups
μ (1X ∨μ)
(μ∨1X )
X ∨X /X ∨X ∨X
Fig. 2.13 Diagram for X

888
abelian H -cogroup 88
μ
88 μ
88
88
88
8
T /X ∨X
X ∨X

X /X ∨X
homomorphism of
H -cogroups
α α∨α
γ
Y /Y ∨Y
Definition 2.4.40 An H -cogroup X is said to be abelian if the triangle in Fig. 2.13

is homotopy commutative, where T (x, x ) = (x , x), ∀ x, x ∈ X.
Definition 2.4.41 Let X and Y be H -cogroups with comultiplications μ and γ

respectively. Then the map α : X → Y is said to be a homomorphism of H -cogroups
if diagram in Fig. 2.14 is homotopy commutative, i.e., α ∨ α ◦ μ γ ◦ α.
Remark 2.4.42 The definition of an H -cogroup closely resembles to that of an H -

group. We merely turn all the maps round and use the one-point union instead of the
product.
Theorem 2.4.43 If X is an H -cogroup and Y is any pointed space, then [X, Y ] can
be given the structure of a group. Moreover, if g : Y → Z is a base point preserving
continuous map, then the induced function g∗ : [X, Y ] → [X, Z ] is a homomorphism
in general and it is an isomorphism if g is a homotopy equivalence.
Proof Given f 1 , f 2 : X → Y , define a product in [X, Y ] by the rule f 1 · f 2 =
◦ ( f ∨ g) ◦ μ, where : X ∨ X → X is the folding map, defined by (x0 , x) =
(x, x 0 ) = x. Proceed as in proofs of Theorems 2.4.22 and 2.4.24. ❑
Dualizing the Theorems 2.4.16, 2.4.25–2.4.27, following theorems are proved.
Theorem 2.4.44 A pointed space having the same homotopy type of an H -cogroup is
itself an H -cogroup in such a way that the homotopy equivalence is a homomorphism.
Theorem 2.4.45 If Q is a H -cogroup, then π Q is a covariant functor from the
homotopy category of pointed spaces with values in the category of groups and
homomorphisms. If Q is an abelian H -cogroup, this functor takes values in the
category of abelian groups.
Theorem 2.4.46 If Q is a pointed topological space such that π Q takes values in
the category of groups, then Q is an H -cogroup (abelian if π Q takes values in the
category of abelian groups). Furthermore, for a pointed topological space X the
group structure on π Q (X ) is identical with that determined by the H -cogroup Q as
in Theorem 2.4.43.
Theorem 2.4.47 If α : Q → Q is a continuous map between H -cogroups, then

there is a natural transformation from π Q to π Q in the category of groups if and only
if α is a homomorphism.
We now describe suspension spaces which are dual to loop spaces. Suspension
spaces give an extensive class of H -cogroups which are dual to H -groups. The
impact of suspension operator is realized from a classical theorem of H. Freudenthal
(1905–1990) known as Freudenthal suspension theorem (see Chap. 7).
Example 2.4.48 (Suspension Space) Let X be a pointed topological space with base
point x 0 . The suspension space of X , denoted by X , is defined to be the quotient
space of X × I in which (X × 0) ∪ (x0 × I ) ∪ (X × 1) has been identified to a single
point. This is sometimes called the reduced suspension. If (x, t) ∈ X × I, we use
[x, t] to denote the corresponding point of X under the quotient map X × I → X .
Then [x0 , 0] = [x 0 , t] = [x , 1], ∀ x, x ∈ X and ∀ t ∈ I . The point [x0 , 0] ∈ X
is also denoted by x0 and X is a pointed space with base point x0 . Moreover, if
f : X → Y is a base point preserving continuous map, then f : X → Y is
defined by
f ([x, t]) = [ f (x), t].
Consequently, is a covariant functor from the category Top ∗ of pointed spaces and
continuous maps to itself.
Remark 2.4.49 If f g : X → Y , then f g : X → Y.
We now show that is also a covariant functor from the category Top ∗ to
the category of H -cogroups and homomorphisms. We define a comultiplication
γ : X → X ∨ X by the formula as shown in Fig. 2.15

([x, 2t], x0 ), 0 ≤ t ≤ 1/2
γ([x, t]) =
(x 0 , [x, 2t − 1]), 1/2 ≤ t ≤ 1.
Clearly, γ is continuous and makes X an H -cogroup.
Theorem 2.4.50 For any pointed topological space X , X is an H -cogroup. More-

over, if f : X → Y is a base point preserving continuous map, then f : X →
Y , [x, t] → [ f (x), t] is a homomorphism of H -cogroups.
Proof As X and Y are both H -cogroups, the proof follows from the definition
of f . ❑
Theorem 2.4.51 For any pair of pointed spaces X and Y , [ X → Y ] is a group.
Proof As X is an H -cogroup, the theorem follows from Theorem 2.4.43. ❑

2.5 Adjoint Functors 79
Fig. 2.15 Comultiplication

γ
x0
γ
x0
ΣX x0
ΣX ∨ ΣX
2.5 Adjoint Functors
This section provides an example of a special pair of functors, called adjoint functors
in the language of category theory. This categorical notion of adjoint functors was
introduced by Daniel Kan (1927–2013) in 1958. There is a close relation between
the loop functor and the suspension functor in the category Top ∗ .
Proposition 2.5.1 The functors and defined from the category Top ∗ of pointed
spaces and continuous maps to itself form a pair of adjoint functors in the sense
that for pointed topological spaces X and Y in Top ∗ there is an equivalence
mor ( X, Y ) ≈ mor (X, Y ), where both sides are the set of morphisms in the cat-
egory Top ∗ .
Proof If g : X → Y is in Top ∗ , then the corresponding morphism g : X →

Y is defined by g [x, t] = g(x)(t), ∀ x ∈ X and ∀ t ∈ I . Thus if h : Y → Y ,
then (h ◦ g) = h ◦ g : X → Y , and if f : X → X , then (g ◦ f ) = g ◦ f :
X → Y . Then the correspondence g ↔ g gives a natural equivalence from the
functor (−, −) to the functor (−, −) on the category Top ∗ . ❑
This natural equivalence plays an important role in the homotopy category Htp ∗
of pointed topological spaces.
Theorem 2.5.2 There exists a natural equivalence from the functor [−, −] to the
functor [−, −] on the category Htp ∗ .
Proof For pointed spaces X and Y , a homotopy G : X × I → Y maps x0 × I into

y0 . Therefore it defines a map F : X × I /x 0 × I → Y . Since (X × I /x 0 × I ) can
be identified with X × I /x0 × I by the homeomorphism [(x, t), t ] ↔ ([x, t ], t),
∀ x ∈ X, t, t ∈ I, it follows that homotopies F : X × I /x0 × I → Y correspond
bijectively to homotopies F : X × I /x0 × I → Y . Consequently, the equivalence
defined in Proposition 2.5.1 gives rise to an equivalence [ X, Y ] ≈ [X, Y ] such

that if g : X → Y and g : X → Y are related by g [x, t] = g(x)(t), then [g ]
corresponds to [g]. Hence there is a natural equivalence from the functor [−, −]
to the functor [−, −]. ❑
Definition 2.5.3 In the language of category theory, the functors and are called
adjoint functors in the sense of Theorem 2.5.2.
The above results are summarized in the basic Theorem 2.5.4.
Theorem 2.5.4 The suspension functor is a covariant functor from the category
Top ∗ of pointed topological spaces and continuous maps to the category of H -
cogroups and continuous homomorphisms. Moreover, the functor preserves homo-
topies, i.e., if f 0 , f1 : X → Y0 are homotopic by the homotopy Ft , then f 0 , f 1 are
homotopic by the homotopy Ft , which is a continuous homomorphism for each
t ∈ I.
Corollary 2.5.5 The suspension functor is a covariant functor from the homotopy
category of pointed topological spaces and homotopy classes of continuous maps to
the category of H -cogroups and continuous homomorphisms.
We now show that for n ≥ 1, the sphere S n admits an extensive family of H -

cogroups.
Proposition 2.5.6 For n ≥ 0, S n+1 is an H -cogroup.
Proof To show this it is sufficient to prove that S n ≈ S n+1 . Let p0 = (1, 0, . . . , 0)

be the base point of S n . We consider Rn+1 as embedded in Rn+2 as the set of points
in Rn+2 whose (n + 2)nd coordinate is 0. Then S n is embedded as an equator in
S n+1 . Again S n = {x ∈ Rn+2 : ||x|| = 1 and xn+2 = 0} and D n+1 is also embedded
in D n+2 , where D n+1 = {x ∈ Rn+2 : ||x|| ≤ 1 and xn+2 = 0}. Let H+ and H− be
two closed hemispheres of S n+1 defined by the equator S n . Then H+ = {x ∈ S n+1 :
x n+2 ≥ 0}, called upper hemisphere, H− = {x ∈ S n+1 : xn+2 ≤ 0}, called lower
hemisphere are such that S n+1 = H+ ∪ H− and S n = H+ ∩ H− .
The maps
p+ : (D n+1 , S n ) → (H+n+1 , S n ), (x 1 , x2 , . . . , xn+1 , 0)

⎛ ⎞

n+1
→ ⎝x1 , x2 , . . . , xn+1 , 1 − xi2 ⎠
i=1
and p− : (D n+1
,S )→
n
(H−n+1 ,S ), (x1 , x 2 , . . . , xn+1 , 0)
n
⎛ ⎞

n+1
→ ⎝x1 , x 2 , . . . , x n+1 , −1 − xi2 ⎠
i=1
are homeomorphisms. Again for t ∈ I, x ∈ S n , the point t x + (1 − t) p0 ∈ D n+1 .

2.5 Adjoint Functors 81
Clearly, the map f : (S n ) → S n+1 defined by

−1
p− (2t x + (1 − 2t) p0 ), 0 ≤ t ≤ 1/2
f ([x, t]) = −1
p+ ((2 − 2t)x + (2t − 1) p0 ), 1/2 ≤ t ≤ 1
is well defined and bijective. It is a homeomorphism, since S 1 ∧ S n and S n+1 are

both compact.
Hence f : (S n ) ≈ S n+1 ⇒ S n+1 is an H -cogroup, since (S n ) is an H -
cogroup. ❑
Definition 2.5.7 Given a pointed topological space X , its iterated suspension spaces
are defined inductively:
n X = ( n−1 X ) for n ≥ 1, and 0 X is taken to be X.
Remark 2.5.8 The groups [ X, Y ], [X, 2 Y ] and [ 2 X, Y ] are isomorphic for

any pointed space Y .
Corollary 2.5.9 For every integer n ≥ 0, any pointed topological space X , n X is

homeomorphic to S n ∧ X.
Proof Since S 0 = {−1, 1}, it follows that S 0 ∧ X ≈ X ≈ 0 X . Suppose n X =

S n ∧ X. Then
n+1 X = ( n X ) ≈ (S n ∧ X ) = S 1 ∧ (S n ∧ X ) ≈ (S 1 ∧ S n ) ∧ X ≈ S n+1 ∧ X.
Hence the corollary follows by induction on n. ❑
Corollary 2.5.10 For every integer n ≥ 0, the (n + 1)-space S n+1 is an H -cogroup.
Remark 2.5.11 The Corollary 2.5.10 shows that for n ≥ 1, the space S n admits an
extensive family of H -cogroups.
Definition 2.5.12 (Adjoint functors) In the language of category theory the equiva-
lence between [ X, Y ] and [ X, Y ] is expressed by saying that in the homotopy
subcategory of pointed Hausdorff spaces of the homotopy category Htp ∗ of pointed
topological spaces, the functors and are adjoint.
Remark 2.5.13 Recall that given a pointed topological space X , we have formed
iterated loop spaces n X inductively: n X = (n−1 X ) for ≥1 and 0 X is taken
to be X . and we have similarly formed iterated suspension spaces inductively:
n X = ( n−1 X ) for n ≥ 1, and 0 X is taken to be X.
Then the groups [ X, Y ], [X, 2 Y ] and [ 2 X, Y ] are isomorphic for any pointed
topological space Y .
Theorem 2.5.14 (i) n X is an abelian H -group for n ≥ 2 for all pointed

topological spaces X.
(ii) n X is an abelian H -cogroup for n ≥ 2 for all pointed topological spaces X.
(iii) For any pair of pointed Hausdorff spaces X and Y , the adjoint functors and
give an isomorphism
ψ : [ X, Y ] → [X, Y ]
of groups. For n ≥ 2, the isomorphisms
ψ : [ X, n−1 Y ] → [X, n Y ]
are of abelian groups;
Proof (i) Let X be a pointed topological space with base point x 0 . Then for any
pointed topological space Y ,
[X, n Y ] ∼
= [ X, n−1 Y ] = [ X, (n−2 Y )]
is an abelian group. Hence if [ f ], [g] ∈ [X, n Y ], then [ f ]·[g] = [g]·[ f ]. This

shows that
μ ◦ ( f × g) ◦ μ ◦ (g × f ) ◦ .
In particular, if X = n Y × n Y, f = p1 , the projection on the first factor and

g = p2 , the projection on the second factor, then we have
( f × g) ◦ (x, y) = ( p1 × p2 )((x, y), (x, y)) = (x, y).
This implies that ( f × g) ◦ = 1d . On the other hand,
(g × f ) ◦ (x, y) = ( p2 × p1 )((x, y), (x, y)) = (y, x) = T (x, y).
This implies that (g × f ) ◦ = T . Hence μ μ ◦ T shows that μ is homotopy

commutative. Consequently, (i) follows from Theorem 2.4.36.
(ii) Similarly, n X is homotopy commutative. Hence (ii) follows from
Theorem 2.4.50.
(iii) It follows from Ex. 29 of Sect. 2.11. ❑
Remark 2.5.15 If X is an H -cogroup and Y is an H -group, the products available

in [X, Y ] determine isomorphic groups which are abelian.
2.6 Contractible Spaces 83
2.6 Contractible Spaces
This section studies a special class of topological spaces, called contractible spaces,
for each of which there exists a homotopy that starts with the identity map and
ends with some constant map. This introduces the concept of contractible spaces.
The concept of contractible spaces is very important. Contractible spaces are in a
natural sense, the trivial objects from the view point of homotopy theory, because
all contractible spaces have the homotopy type of a space reduced to a single point.
Such spaces are connected topological objects having no ‘holes’ or ‘cycles’ and
have nice intrinsic properties. The simplest nonempty space is one-point space. We
characterize the homotopy type of such spaces.
2.6.1 Introductory Concepts
This subsection opens with introductory concepts of contractible spaces.
Definition 2.6.1 A topological space X is said to be contractible if the identity

map 1 X : X → X is homotopic to some constant map of X to itself. If c : X → X
defined by c(x) = x0 ∈ X is such that 1 X c, then a homotopy F : 1 X c is called
a contraction of the space X to the point x0 .
Example 2.6.2 Any convex subspace X of Rn is contractible. Because, any contin-

uous map H : X × I → X defined by H (x, t) = (1 − t)x + t x 0 , x, x 0 ∈ X, t ∈ I
is such that H (x, 0) = x = 1 X (x), ∀ x ∈ X and H (x, 1) = x 0 = c(x), ∀ x ∈ X .
Hence H : 1 X c ⇒ X is contractible and H is a contraction of X to the point
x0 ∈ X . In particular, Rn , D n , I are contractible spaces.
Geometrical meaning: A contraction H : 1 X c can be interpreted geometrically

as a continuous deformation of the space X which ultimately shrinks the whole space
X into the point x0 ∈ X and hence X can be contracted to a point of X .
Can we contract a topological space X to an arbitrary point x 0 ∈ X ?
To answer this question we need the following Proposition:
Proposition 2.6.3 A topological space X is contractible if and only if an arbitrary

continuous map f : Y → X from any topological space Y to X is homotopic to a
constant map.
Proof Let X be contractible. Then the identity map 1 X : X → X is homotopic to

some constant map c : X → X, x → x 0 (say). Let f : Y → X be any continuous
map. Now 1 X c ⇒ 1 X ◦ f c ◦ f . But c ◦ f : Y → X, y → x0 is a constant
map. Thus f is homotopic to a constant map. For the converse, we take Y = X and
f = 1 X : X → X . Then by hypothesis, 1 X : X → X is a constant map. Hence X is
contractible. ❑
Corollary 2.6.4 Any two continuous maps from an arbitrary space to a contractible
space are homotopic.
Proof Let X be a contractible space and f, g : Y → X be two continuous maps

from an arbitrary space Y to the space X . Now 1 X c, where c : X → X is defined
by x → x0 ∈ X ⇒ 1 X ◦ f c ◦ f and 1 X ◦ g c ◦ g ⇒ f = 1 X ◦ f c ◦ f =
c ◦ g 1 X ◦ g = g ⇒ f g. ❑
Corollary 2.6.5 If X is contractible, then the identity map 1 X : X → X is homotopic

to any constant map of X to itself.
Proof If X is contractible, then by Corollary 2.6.4 it follows in particular that 1 X is

homotopic to any constant map of X to itself. ❑
Remark 2.6.6 In absence of the condition of contractibility of X , the Corollary 2.6.4

fails.
Example 2.6.7 Let X be a connected topological space and Y = {y0 , y1 }, (y0 = y1 )

with discrete topology, i.e., Y is a discrete space consisting of two distinct elements.
Consider the constant maps f, g : X → Y defined by constant f (x) = y0 and g(x) =
y1 , ∀ x ∈ X. Then f and g are not homotopic (see Example 2.2.19)
We now characterize contractible spaces.
Theorem 2.6.8 A topological space X is contractible if and only if X is of the same

homotopy type of a one-point space P = { p}.
Proof Suppose X is contractible. Then 1 X : X → X is homotopic to some constant

map co : X → X, x → x0 ∈ X . Let H : 1 X c0 . Define maps i : P → X and c :
X → P by i(P) = x0 and c(x) = p, ∀ x ∈ X . Then c ◦ i = 1 P . Moreover, H :
1 X i ◦ c, because H (x, 0) = x and H (x, 1) = c0 . Hence X P. Conversely, let
X P. then ∃ continuous maps f : X → P and g : P → X such that g ◦ f
1 X and f ◦ g 1 P . Leth g( p) = x0 ∈ X and H : 1 X g ◦ f . Since (g ◦ f )(x) =
g( f (x)) = g( p) = x0 , ∀ x ∈ X, g ◦ f : X → X, x → x0 , is the constant map c0 .
Thus 1 X c0 ⇒ X is contractible. ❑
Corollary 2.6.9 Two contractible spaces have the same homotopy type, and any
continuous map between contractible spaces is a homotopy equivalence.
Proof Let X and Y be two contractible spaces and P be a one-point space.

Then X P and Y P ⇒ X Y by symmetry and transitivity of the relation
. Hence ∃ a homotopy equivalence f ∈ C(X, Y ). Let g : X → Y be an arbitrary
continuous map. Then f g by Corollary 2.6.4 ⇒ g is a homotopy equivalence by
Proposition 2.2.15. ❑
Remark 2.6.10 Contractible spaces are precisely those spaces which are homotopy
equivalent to a point space. Thus all contractible spaces have the homotopy type of
a space reduced to a single point.
Definition 2.6.11 A topological space X is said to be contractible to a point a ∈ X

relative to the subset A = {a} if ∃ a homotopy H : X × I → X such that H : 1 X
c rel A, where c : X → X, x → a is a constant map.
Theorem 2.6.12 If a topological space X is contractible to a point a ∈ X relative
to the subset A = {a}, then for each neighborhood U of a in X, ∃ a neighborhood
V of a contained in U such that any point of V can be joined to a by a path lying
entirely inside U .
Proof Let the space X be contractible to a point a ∈ X relative to the subset A =
{a}. Then there exists a continuous map F such that F : 1 X c rel A ⇒ the line
{a} × I is mapped by F to the point a ∈ X . We now take a neighborhood U of
a. Then the continuity of F ⇒ for each t ∈ I , neighborhoods Vt (a) of a in X and
W (t) of t in I are such that F(Vt (a) × W (t)) ⊂ U . Since I is compact, the open
covering {W (t) : t ∈ I } of I has a finite subcovering W (t1 ), W (t2 ), . . . , W (tn ) (say)
n
such that F(Vti (a) × W (ti )) ⊂ U , for i = 1, 2, . . . , n. Thus V (a) = Vti (a) is
i=1
a neighborhood of a in X such that F(V (a) × I ) ⊂ U . Now, if x ∈ V (a), then
considering the image F(V (a) × I ) in U , it follows that the point x can be joined
to the point a by a path which lies inside U . ❑
Proposition 2.6.13 Every contractible space is path-connected.
Proof Let X be contractible to a point x0 ∈ X and H : 1 X c, where c : X →
X, x → x0 ∈ X is a constant map. Now H (x, 0) = 1 X (x) = x and H (x, 1) =
c(x) = x 0 , ∀ x ∈ X . Given a ∈ X define a path f : I → X × I by f (t) = (a, t).
Then α = H ◦ f : I → X is a continuous map such that α(0) = H ( f (0)) =
H (a, 0) = a and α(1) = H ( f (1)) = H (a, 1) = x0 ⇒ α is a path from a to x0 . In
other words, X is path-connected. ❑
2.6.2 Infinite-Dimensional Sphere and Comb Space
We now examine the contractibility of the infinite-dimensional sphere S ∞ . We also

study comb space which is contractible in absolute sense but not contractible in
relative sense. First we describe R∞ , C∞ and S ∞ .
Definition 2.6.14 The set of all sequences x = (x 1 , x2 , . . . , x n , . . .) of real numbers
∞

such that |xn |2 converges, is denoted by R∞ . Under coordinatewise addition and
1
scalar multiplication, R∞ is a vector space over R. Moreover, R∞ endowed with a
∞
norm function defined by x = ( |xn |2 )1/2 is called a real Banach space. The
1
space R∞ is called infinite-dimensional Euclidean space. Similarly, the infinite-
dimensional unitary space C∞ is defined.
Remark 2.6.15 The space C∞ is a complex Banach space. Clearly, as a topological

space Cn is homeomorphic to R2n and C∞ is homeomorphic to R∞ . The space S ∞
is now defined.
Definition 2.6.16 The infinite-dimensional sphere S ∞ is the subspace of R∞ (under

weak topology) consisting of all real sequences (x 1 , x2 , x 3 , . . .) such that x12 + x22 +
x32 + · · · = 1 (i.e., S ∞ = {(x1 , x2 , x 3 , . . .) ∈ R∞ : x 12 + x 22 + x 32 + · · · = 1}).
As the diagram in Fig. 2.16 is commutative, we may consider S ∞ as the subspace of

∞
C consisting of the sequences (z 1 , z 2 , . . .) over C such that |z 1 |2 + |z 2 |2 + · · · = 1.
We are now in a position to prove the contractibility of S ∞ .
Proposition 2.6.17 The infinite-dimensional sphere S ∞ is contractible.
Proof Consider the map
F : S ∞ × I → S ∞ , (x 1 , x2 , x 3 , . . . , t)
→ ((1 − t)x1 , t x1 + (1 − t)x 2 , t x2 + (1 − t)x 3 , . . .)/Nt ,
where Nt = [((1 − t)x1 )2 + (t x 1 + (1 − t)x2 )2 + (t x 2 + (1 − t)x3 )2 + · · · ]1/2 ,

which is the norm of the nonzero vector of the numerator. We may parametrize
F as Ft (x1 , x2 , x 3 , . . .) = F(x1 , x2 , x 3 , . . . , t).
Then F0 (x1 , x2 , x3 , . . .) = (x1 , x2 , x 3 , . . .), since N0 = 1 and F1 (x1 , x2 , x 3 , . . .) =
(0, x1 , x2 , x3 , . . .), since N1 = 1. Consequently, F0 is the identity map 1d : S ∞ →
S ∞ , the image of F1 is the set X = {x ∈ S ∞ : x1 = 0} and F : F0 F1 .
Consider another homotopy
H : X × I → S ∞ , H (x 1 = 0, x2 , x3 , . . . , t) → (t, (1 − t)x2 , (1 − t)x3 , . . .)/Nt ,
where Nt = [t 2 + ((1 − t)x2 )2 + ((1 − t)x 3 )2 + · · · ]1/2 .

If i : X → S ∞ is the inclusion map, then H : i c, where c is a constant map.
Let H ∗ F : S ∞ × I → S ∞ be defined by

F(x, 2t), 0 ≤ t ≤ 1/2
(H ∗ F)(t) =
H (x, 2t − 1), 1/2 ≤ t ≤ 1.
where x = (x 1 , x2 , x 3 , . . .) ∈ S ∞ .
Then H ∗ F is a contraction. Consequently, S ∞ is a contractible space. ❑
Fig. 2.16 Commutative / Cn

diagram involving Cn and S 2n−1
_ _
Cn+1
/ Cn+1
S 2n+1
Remark 2.6.18 The infinite-dimensional sphere S ∞ is contractible. On the other

hand, the n-sphere S n is not contractible for any integer n ≥ 0 (see Proposition
14.1.13 of Chap. 14).
Corollary 2.6.19 The inclusion map i : S n−1 → S n is nullhomotopic.
Remark 2.6.20 We now search for a topological space which is contractible in
absolute sense but not contractible in relative sense.
Example 2.6.21 (Comb Space) The subspace Y of the plane R2 defined by

1
Y = (x, y) ∈ R : 0 ≤ y ≤ 1, x = 0, (n ∈ N) or y = 0, 0 ≤ x ≤ 1
2
n
is called the comb space, i.e., Y consists of the horizontal line segment L joining
(0, 0) to (1, 0) and vertical unit closed line segments standing on points (1/n, 0) for
each n ∈ N, together with the line segment joining (0, 0) with (0, 1) as shown in
Fig. 2.17. It is an important example of a contractible space.
Proposition 2.6.22 Comb space Y is contractible but not contractible relative to

{(0, 1)}.
Proof First part: First we show that L Y . Let p : Y → L , (x, y) → (x, 0) be the
projection map and i : L → Y be the inclusion map. Then ( p ◦ i)(x, 0) = p(x, 0) =
(x, 0), ∀ (x, 0) ∈ L ⇒ p ◦ i = 1 L (identity map on L). Define F : Y × I → Y
by the rule F((x, y), t) = (x, (1 − t)y). Then F((x, y), 0) = (x, y) = 1Y (hx, y),
∀ (x, y) ∈ Y and F((x, y), 1) = (x, 0) = (i ◦ p)(x, y), ∀ (x, y) ∈ Y show that
F : 1Y i ◦ p ⇒ p ∈ C(Y, L) is a homotopy equivalence ⇒ Y L. Again L ≈
I ⇒ L I . Moreover, I being a contractible space, I is of the same homotopy type
of a one-point space and hence L is of the same homotopy type of one-point space.
Consequently, Y is of the same homotopy type of one-point space. In other words,
Y is contractible by Theorem 2.6.8.
Second part: Any small neighborhood V of (0, 1) has infinite number of path com-
ponents. Let D be the open disk around (0, 1) of radius 12 . Then the neighborhood
U = D ∩ Y of (0, 1) in Y cannot have any neighborhood V each of whose points
can be joined to (0, 1) by a path entirely lying in U ⇒ Y is not contractible relative
to {(0, 1)} by Theorem 2.6.12, otherwise we would reach a contradiction by the same
theorem. ❑
Fig. 2.17 Comb space (0, 1)
L
(0, 0) (1/2, 0) (1, 0)
Remark 2.6.23 The concept of relative homotopy is stronger than the concept of
homotopy.
Let A be a subspace of X and f, g : X → Y be two continuous maps such that
f g rel A. Then f g. But its converse is not true.
Example 2.6.24 Let Y be the comb space, 1Y : Y → Y be the identity map and
c : Y → Y be the constant map defined by c(x, y) = (0, 1), ∀ (x, y) ∈ Y . Then IY
and c agree on {(0, 1)} and hence 1Y c by Corollary 2.6.5, since Y is contractible.
But the comb space Y is not contractible relative to {(0, 1)} (see Proposition 2.6.22).
2.7 Retraction and Deformation
This section mainly studies inclusion maps from the viewpoint of homotopy theory.
We consider whether an inclusion map i : A → X has a left inverse or a right inverse
or a left homotopy inverse or a right homotopy inverse or two sided inverse or
two-sided homotopy inverse. More precisely, the concepts of retraction and weak
retraction are introduced and it is proved that these two concepts coincide under
suitable homotopy extension property (HEP).
Let A be a subspace of a topological space X and i : A → X be the inclusion
map. Then a continuous map f : A → Y from A to a subspace Y is said to have
a continuous extension over X if ∃ a continuous map F : X → Y such that the
diagram in Fig. 2.18 is commutative, i.e., F ◦ i = f . Thus F is said to be a continuous
extension of f over X if F| A = f .
Definition 2.7.1 A subspace A of X is called a retract of X if there exists a continuous

map r : X → A such that r ◦ i = 1 A , i.e., if i has a left inverse in the category Top
of topological spaces and continuous maps, i.e., if r (x) = x, ∀ x ∈ A. Such a map
r is called a retraction of X to A. On the other hand, if i ◦ r 1 X , A is called a
deformation retract of X and r is called a deformation retraction.
Thus A is a retract of X if ∃ a continuous map r : X → A making the diagram in
Fig. 2.19 is commutative.
Fig. 2.18 Continuous i /X

extension of f A@
@@ ~
@@ ~
f @@ ~~
~ F
Y
Fig. 2.19 Retraction and i /X

retract A@
@@ ~
@@ ~
1A @@ ~~
~ r
A
2.7 Retraction and Deformation 89
Remark 2.7.2 The main property of a retract A of X is that any continuous map
f : A → Y has at least one continuous extension f˜ : X → Y , namely, f˜ = f ◦ r ,
where r : X → A is a retraction.
Example 2.7.3 The circle A as shown in Fig. 2.20 is a retract of the annulus X . The
arrows indicate the action of the retraction. The whole of X is mapped onto A keeping
points in A fixed.
Example 2.7.4 Consider the inclusion map i : D n → Rn . Define a map r : Rn →

D n by the rule
x
, if ||x|| > 1
r (x) = ||x||
x, if ||x|| ≤ 1.
Then r is a retraction and hence D n is a retract of Rn . Geometrically, the map r fixes

points in D n and shifts points x outside of D n along a straight line from the origin
to x onto the boundary S n−1 of D n .
Definition 2.7.5 A subspace A of a topological space X is called a weak retract of

X if there exists a continuous map r : X → A such that r ◦ i 1 A , i.e., if i has a
left homotopy inverse, i.e., if i has a left inverse in the homotopy category Htp of
topological spaces and continuous maps.
Thus A is a weak retract of X if ∃ a continuous map r : X → A making the diagram
in Fig. 2.21 homotopy commutative. Such a map r is called a weak retraction of X
to A.
Remark 2.7.6 A is retract of X ⇒ A is a weak retract of X , because r ◦ i = 1 A 1 A .

But its converse is not true.
Example 2.7.7 Consider X = I 2 and A =comb space (see Example 2.6.21). Then
A X . As A and X are both contractible spaces, the inclusion map i : A → X is
Fig. 2.20 The circle A (deep

black) is a retract of the
annulus X
A
Fig. 2.21 Weak retraction i /X

and weak retract A@
@@ ~
@@ ~
1A @@ ~~
~ r
A
a homotopy equivalence by Corollary 2.6.9. Hence ∃ a continuous map r : X → A

such that r ◦ i 1 A . This shows that r is a weak homotopy equivalence. Clearly
r ◦ i = 1 A . Consequently, A is a weak retract of X but not a retract of X .
We now search conditions under which the concepts of retraction and weak retrac-
tion coincide. For this purpose we introduce the concept of Homotopy Extension
Property for the pair of spaces (X, A).
Definition 2.7.8 Let (X, A) be pair of topological spaces and Y be an arbitrary

topological space. Then the pair (X, A) is said to have the Homotopy Extension
Property (HEP) with respect to the space Y if given continuous maps g : X → Y
and G : A × I → Y such that g(x) = G(x, 0), ∀ x ∈ A, there is a continuous map
F : X × I → Y with the property F(x, 0) = g(x), ∀ x ∈ X and F| A×I = G.
If h 0 (x) = (x, 0), ∀ x ∈ X , the existence of F is equivalent to the existence

of a continuous map represented by the dotted arrow which makes the diagram in
Thus (X, A) has the HEP with respect to Y if ∃ a continuous map F : X × I → Y
such that the square and the two triangles in the diagram in Fig. 2.22 are commutative.
Proposition 2.7.9 If (X, A) has the HEP with respect to Y and if f 0 , f 1 : A → Y

are homotopic, then f 0 has a continuous extension over X iff f 1 has also a continuous
extension over X .
Proof Let f 0 , f 1 : A → Y be two continuous maps such that f 0 f 1 . Then ∃ a

homotopy G : A × I → Y such that G(x, 0) = f 0 (x) and G(x, 1) = f 1 (x), ∀ x ∈
A. Let f 0 : A → Y have a continuous extension g0 : X → Y . Then G(x, 0) =
f 0 (x) = g0 (x), ∀ x ∈ A. As (X, A) has the HEP with respect to Y, ∃ a map
F : X × I → Y extending G : A × I → Y and therefore the diagram in Fig. 2.23 is
commutative. The existence of F follows from the HEP of (X, A) with respect to
Y . Define a map g1 : X × Y by g1 (x) = F(x, 1), ∀ x ∈ X . Then g1 is an extension
Fig. 2.22 Homotopy h0

A /A×I
~ Oo x
extension property (HEP) i ~~~ G xxx
~~ x
~~ xxx
g {x
X UUUUUU /Y i×1d
UUUU cF
F F
UUUU F
UUUU F
h0 UUUU F
*
X ×I
Fig. 2.23 Homotopy h0

A /A×I
o
extension property of (X, A) i ~~~ O G xxx
x
~~ xx
w.r.t. Y ~~ x
~
g {xx
X UUUUUU /Y i×1d
UUUU cF
F
UUUU FF
UUUU F
h0 UUUU F
*
X ×I
of f 1 over X , because, g1 (a) = F(a, 1) = G(a, 1) = f 1 (a), ∀ a ∈ A ⇒ g1 | A = f 1 .

Moreover g1 is continuous and hence f 1 has a continuous extension g1 over X . ❑
Remark 2.7.10 A continuous map f : A → Y can or cannot be extended over X is
a property of the homotopy class of that map. Thus the homotopy extension property
implies that the extension problem for continuous maps A → Y is a problem in the
homotopy category.
The map r : Rn → D n defined in Example 2.7.4 is a deformation retraction. To show
this define a homotopy F : Rn × I → Rn by the rule

(1 − t)x + t x/||x||, if ||x|| ≥ 1
F(x, t) =
x, if ||x|| < 1
Then F : 1d i ◦ r shows that r is a deformation retraction.

Geometrically, F fixes points in D n and shifts points x outside of D n linearly from
x to r (x) along the straight line determined by x and the origin 0 = (0, 0, . . . , 0).
Theorem 2.7.11 If (X, A) has the HEP with respect to A, then A is a weak retract
of X iff A is a retract of X .
Proof Let A ⊂ X be a retract of X and r : X → A be a retraction. Then r ◦ i =
1 A ⇒ r ◦ i 1 A ⇒ A is a weak retract of X . Conversely, let r : X → A be a weak
retraction. Then r ◦ i 1 A , where i : A → X is the inclusion map. Then ∃ a homo-
topy G : A × I → Y such that G(x, 0) = r (x), G(x, 1) = 1 A (x) = x, ∀ x ∈ A. As
(X, A) has the HEP with respect ot A, ∃ a continuous map F : X × I → A extending
G : A × I → A. Hence F(x, 0) = r (x), ∀ x ∈ X and F| A×I = G. Define a map
r : X → A by the rule r (x) = F(x, 1), ∀ x ∈ X . Now, for all a ∈ A, (r ◦ i)(a) =
r (i(a)) = F(a, 1) = G(a, 1) = a ⇒ r ◦ i = 1 A ⇒ A is a retract of X and r is a
retraction of X into A. ❑
We can as well consider inclusion maps with right homotopy inverses.
Definition 2.7.12 Given a subspace X ⊂ X , a deformation D of X in X is
a homotopy D : X × I → X such that D(x , 0) = x , ∀ x ∈ X . If moreover,
D(X × 1) ⊂ A ⊂ X , D is said to a deformation of X into A and X is said to
be deformable in X into A. If X = X , then a space X is said to be deformable into
a subspace A of X if it is deformable in itself into A.
Theorem 2.7.13 A topological space X is deformable into a subspace A of X iff
the inclusion map i : A → X has a right homotopy inverse.
Proof Let X be deformable into a subspace A of X . Then ∃ a continuous map D : X ×
I → X such that D(x, 0) = x and D(x, 1) ∈ A ⊂ X, ∀ x ∈ X . Let f : X → A be
defined by the equation (i ◦ f )(x) = D(x, 1), ∀ x ∈ X . Then D : 1 X i ◦ f ⇒
i has a right homotopy inverse. Conversely, let i : A ⊂ X has a right homotopy
inverse f : X → A. Then 1 X i ◦ f. Let F : X × I → X be such that F : 1 X
i ◦ f . Then F(x, 0) = 1 X (x) = x, ∀ x ∈ X and F(X × 1) = (i ◦ f )(X ) ⊂ A(i.e.,
F(x, 1) = i( f (x)) = f (x), ∀ x ∈ X ) ⇒ X is deformable into A. ❑
Remark 2.7.14 The homotopy D which starts with identity map 1 X : X → X ,

simply moves each point of X continuously, including the points of A and finally,
pushes every point into a point of A. In particular, if X is deformable into a point
x 0 ∈ X , then X is contractible and vice verse (see Ex. 4 of Sect. 2.11).
Remark 2.7.15 An inclusion map i : A → X has never a right inverse in the category
of topological spaces and continuous maps in the trivial case when A = X .
We now consider inclusion maps which are homotopy equivalences.
Definition 2.7.16 A subspace A of a topological space X is called a weak deforma-

tion retract of X if the inclusion map i : A → X is a homotopy equivalence.
Theorem 2.7.17 A subspace A of a topological space X is a weak deformation

retract of X iff A is a weak retract of X and X is deformable into A.
Proof Let A be a weak deformation retract of X . Then i : A → X is a homotopy

equivalence. Hence ∃ a continuous map r : X → A such that r ◦ i 1 A and i ◦
r 1 X . Thus i has a right homotopy inverse and also a left homotopy inverse.
Consequently, A is a weak retract of X and X is deformable into A.
Conversely, let A be a weak retract of X and X be deformable into A. Then
i : A → X has a left homotopy inverse f (say) and a right homotopy inverse g(say).
Now f ◦ i 1 A and i ◦ g 1 X ⇒ f = f ◦ 1 X f ◦ (i ◦ g) = ( f ◦ i) ◦ g 1 A ◦
g = g ⇒ f g. Hence f ◦ i 1 A and i ◦ f 1 X ⇒ i is a homotopy equivalence
⇒ A is a weak deformation retract of X . ❑
We now consider a deformation D which deforms X into A, but the points of A

do not move at all.
This led to the concept of strong deformation retract introduced by Borsuk in
1933.
Definition 2.7.18 A subspace A of a topological space X is called a strong defor-

mation retract of X if there exists a retraction r : X → A such that 1 X i ◦ r rel A.
If F : 1 X i ◦ r rel A, then F is called a strong deformation retract of X to A.
There is an intermediate concept between the concepts of weak deformation retrac-
tion and strong deformation retraction.
Definition 2.7.19 A subspace A of a topological space X is called a deformation

retract of X if ∃ a retraction r : X → A such that 1 X i ◦ r . If F : 1 X i ◦ r , then
F is called a deformation retraction of X to A.
Remark 2.7.20 A homotopy F : X × I → X is a deformation retraction of X to

A iff F(x, 0) = 1 X (x) = x, ∀ x ∈ X, F(x, 1) = x, ∀ x ∈ A and F(X × 1) ⊂ A.
A map F is called a strong deformation retraction iff it also satisfies the condition
F(x, t) = x, ∀ x ∈ A and ∀ t ∈ I . Thus F is a strong deformation retraction of X
to A iff F satisfies the conditions:
(i) F(x, 0) = x, ∀ x ∈ X ;
(ii) F(x, 1) = x, ∀ x ∈ A;
(iii) F(X × 1) ⊂ A ⊂ X ;
(iv) F(x, t) = x, ∀ x ∈ A and ∀ t ∈ I.
Example 2.7.21 Let X = Rn+1 − {0} and S n be the n-sphere (n ≥ 1). Then S n ⊂ X
is a strong deformation retract of X . Let i : S n → X be the inclusion. Define a
retraction r : X → S n by r (x) = ||x||
x
. Geometrically, this map shifts points x ∈ X
n
to the boundary S along a straight line from the origin. Define a continuous map
F : X × I → X by F(x, t) = (1 − t)x + ||x|| tx
, ∀ x ∈ X, ∀ t ∈ I . Then
(i) F(x, 0) = x, ∀ x ∈ X ;
(ii) F(x, 1) = x, ∀ x ∈ S n ;
(iii) F(X × 1) ⊂ S n ⊂ X and
(iv) F(x, t) = x, ∀ x ∈ S n and ∀ t ∈ I.
Clearly, r ◦ i is the identity map on S n and i ◦ r is homotopic to the identity map

on X . Hence i is a homotopy equivalence. Geometrically, this homotopy F moves
linearly along the straight path defined above from x to r (x). Hence F is a strong
deformation retraction of X to S n . Therefore S n is a strong deformation retract of
X = Rn+1 − {0}.
Remark 2.7.22 We now explain Example 2.7.21 geometrically for n = 1 as shown

in Fig. 2.24.
Let l be an arbitrary half-line starting from the origin 0 = (0, 0). Then it intersects
the circle S 1 at exactly one point l P (say). Since 0 = (0, 0) is not a point of R2 − {0},
the lines l − {0} are disjoint and their union is R2 − {0}. Define a map r : R2 − {0} →
S 1 by r (x) = l P , ∀ x ∈ l. Then r is a retraction and S 1 is a retract of R2 − {0}. Define
a deformation D : (R2 − {0}) × I → R2 − {0} by D(x, t) = (1 − t)x + t ||x|| x
. Then
as before D is a strong deformation retraction of R − {0} relative to S into S 1 .
2 1
Example 2.7.23 Consider the product space X = D n × I and the subspace A =

(S n−1 × I ) ∪ (D n × {0}). If P is the point (0, 2) in Rn × R, a retraction r : X → A
is defined by taking r (x) to be the point where the line joining P and x meets A.
Consequently, the map
F : X × I → X, (x, t) → F(x, t) = t r (x) + (1 − t)x
is such that F(x, 0) = x and F(x, 1) = r (x). This shows that F is a strong defor-
mation retraction.
Fig. 2.24 Half line starting

l
from the origin and
intersecting the circle lP
0
Fig. 2.25 Construction of X

X
Example 2.7.24 Let X be the topological space given by I together with a family of
segments approaching it as shown in Fig. 2.25.
Then I is a deformation retract of X but not a strong deformation retract.
Proposition 2.7.25 A topological space X is deformable into a retract A ⊂ X if

and only if A is a deformation retract of X .
Proof Let X be deformable into a retract A ⊂ X and i : A → X be the inclusion

map. Then there exists a retraction r : X → A such that r ◦ i = 1 A . Hence r is a left
inverse of i and thus r is a left homotopy inverse of i. Again as X is deformable into
A, i has a right homotopy inverse by Theorem 2.7.13, which is also r , i.e., 1 X i ◦ r .
Consequently, A is a deformation retract of X . Conversely, let A be a deformation
retract of X . Then ∃ a retraction r : X → A such that 1 X i ◦ r and r ◦ i = 1 A .
Consequently, X is deformable into a retract A. ❑
We now show that if (X, A) has the HEP with respect to A, then the concepts of
weak deformation retraction and deformation retraction coincide.
Theorem 2.7.26 If (X, A) has the HEP with respect to A, then A is a weak defor-
mation retract of X if and only if A is a deformation retract of X .
Proof Suppose (X, A) has the HEP with respect to A. Then A is a weak retract of
X and X is deformable into A if and only if A is a retract of X and X is deformable
into A by Theorem 2.7.11. Then the theorem follows from Proposition 2.7.25. ❑
We now show that under suitable HEP the concepts of strong deformation retrac-
tion and deformation retraction coincide.
Theorem 2.7.27 If A ⊂ X and (X × I, (X × {0}) ∪ (A × I ) ∪ (X × 1)) has the

HEP with respect to X and A is closed in X , then A is a deformation retract of X if
and only if A is a strong deformation retract of X .
Proof Let X = X × I and A = (X × {0}) ∪ (A × I ) ∪ (X × 1). Then A ⊂ X .

Let (X , A ) has the HEP with respect to X and A be closed in X . Suppose i : A → X
and A is a deformation retract of X . We claim that A is also a strong deformation
retract of X . Since A is a deformation retract of X, ∃ a retraction r : X → A such
that 1 X i ◦ r . Let F : X × I → X be a continuous map such that F : 1 X i ◦ r .
Then F(x, 0) = 1 X (x) = x, ∀ x ∈ X and F(x, 1) = (i ◦ r )(x) = r (x), ∀ x ∈ X .
We now define a map G : A × I → X by the equations
Fig. 2.26 Homotopy

h0
/ A × I
A n 99
extension property of i}}} N 99
} 99
(X , A ) w.r.t. X }}} G
99
}~
WW F / 99i×1d
X WWWWW X eK 99
WWWWW K
WWWWW KG 999
WWWWW K
h0 WWWWW K99
+
X ×I
G((x, 0), t ) = x, ∀ x ∈ X, ∀ t ∈ I (2.11)

G((x, t), t ) = F(x, (1 − t )t), ∀ x ∈ A, ∀ t, t ∈ I (2.12)
G((x, t), t ) = F(r (x), 1 − t ), ∀ x ∈ X, ∀ t, t ∈ I (2.13)
Then G is well defined, because for x ∈ A, G((x, 0), t ) = x = F(x, 0) by the

first two Eqs. (2.11) and (2.12) and G((x, 1), t ) = F(x, 1 − t ) = F(r (x), 1 − t )
by the last two Eqs. (2.12) and (2.13). Again G is continuous, because its restriction
to each of the closed sets (X × {0}) × I, (A × I ) × I and (X × 1) × I is continuous.
For (x, t) ∈ A , G((x, t), 0) = F(x, t), because F(x, 0) = x and since r : X → A
is a retraction.
F(r (x), 1) = (i ◦ r )(r (x)) = r (x) = F(x, 1). Therefore G restricted to A × {0}
can be extended to (X × I ) × {0}. Then by HEP of (X , A ) w.r.t X in the hypothesis,
∃ a homotopy G : X × I → X extending G : A × I → X (see Fig. 2.26). Define
H : X × I → X by H (x, t) = G ((x, t), 1). Then we have the equations
H (x, 0) = G ((x, 0), 1) = G((x, 0), 1) = x, ∀ x ∈ X ;
H (x, 1) = G ((x, 1), 1) = F(r (x), 0) = r (x), ∀ x ∈ X ;
and H (x, t) = G ((x, t), 1) = G((x, t), 1) = F(x, 0) = x, ∀ x ∈ A, ∀ t ∈ I .
Therefore H : 1 X i ◦ r rel A. Hence A is a strong deformation retract of X .
Conversely, if A is a strong deformation retract of X , then A is automatically a
deformation retract of X . ❑
2.8 NDR and DR Pairs
This section defines the concepts of NDR-pair and DR-pair which are closely related
to the concepts of retraction and homotopy extension property for compactly gener-
ated spaces (see Sect. B.4 of Appendix B). N. Steenrod (1910–1971) proved in 1967
the equivalence between the NDR condition and the homotopy extension property
(Steenrod 1967).
We now use the concept of compactly generated space defined in Appendix B.
Definition 2.8.1 Let X be a compactly generated topological space and A ⊂ X be
a subspace. Then (X, A) is said to be an NDR-pair (NDR stands for ‘neighborhood
deformation retract’) if there exist continuous maps u : X → I and h : X × I → X
such that
NDR(i) A = u −1 (0);
NDR(ii) h(x, 0) = x, ∀ x ∈ X ;
NDR(iii) h(a, t) = a, ∀ t ∈ I, a ∈ A;
NDR(iv) h(x, 1) ∈ A for all x ∈ X such that u(x) < 1.
In particular, A is a retract of its neighborhood U = {x ∈ X : u(x) < 1}, and hence
is a neighborhood retract of X .
Definition 2.8.2 A pair (X, A) is called a DR-pair (DR stands for “deformation
retract”) if in addition to NDR(i)–NDR(iii), another condition NDR(v): h(x, 1) ∈ A
(instead of NDR(iv)) holds for all x ∈ X .
Remark 2.8.3 The concepts of DR-pair and NDR-pair are closely related to the
concepts of retraction and HEP (see Ex. 32 of Sect. 2.11).
2.9 Homotopy Properties of Infinite Symmetric Product

Spaces
This section conveys homotopy properties of infinite symmetric product spaces

defined for spaces in Top ∗ (see Sect. B.2.5 of Appendix B). These spaces link homo-
topy theory with homology theory via Elienberg–MacLane spaces (see Chaps. 11
and 17) and form an important class of topological spaces in the study of algebraic
topology. So it has become necessary to study such spaces from homotopy viewpoint.
We have constructed in Sect. B.2.5 of Appendix B the finite symmetric product
S P n X and infinite symmetric product S P ∞ X of a pointed topological space X .
Both S P n and S P ∞ are functors from the category Top ∗ to itself (see Sect. B.2.5
of Appendix B). A continuous map f : X → Y in Top ∗ induces maps f n : X n →
Y n , (x1 , x2 , . . . , xn ) → ( f (x 1 ), f (x2 ), . . . , f (xn )). These maps are compatible with
the action of the symmetric group Sn of the set {1, 2, . . . , n} and hence induce maps
S P n ( f ) : S P n X → S P n Y between the corresponding orbit spaces and also induce
maps S P ∞ ( f ) = f ∗ : S P ∞ X → S P ∞ Y (see Sect. B.2.5 of Appendix B).
Theorem 2.9.1 If f, g : X → Y are in Top ∗ and f g, then S P ∞ ( f ) S P ∞ (g).
Proof Let F : X × I → Y be a map such that F : f g. For all n ≥ 1, define F n :
X n × I → Y n , (x 1 , x2 , . . . , x n , t) → (F(x, t), F(x2 , t), . . . , F(xn , t)). Then F n is
continuous, because its projection onto each coordinate is continuous. Since Sn acts
on X n × I by permuting the coordinate of X n and fixing I , and F n respects this action,
F n induces maps S P n (F) : S P n X → S P n Y , which passing to the limit induces a
map S P ∞ (F) : S P ∞ X × I → S P ∞ Y . Define
h t : X → Y, x → F(x, t) and S P ∞ (F) : S P ∞ X

× I → S P ∞ Y, (x, t) → S P ∞ (h t )(x).
Hence S P ∞ (F) : S P ∞ ( f ) S P ∞ (g). ❑

2.9 Homotopy Properties of Infinite Symmetric Product Spaces 97
Corollary 2.9.2 If spaces X and Y in Top ∗ are homotopy equivalent, then the spaces
S P ∞ X and S P ∞ Y are also homotopy equivalent. In particular, if X is contractible,
then S P ∞ X is contractible.
Proof Let f : X → Y be a homotopy equivalence with homotopy inverse g : Y →
X . Then S P ∞ (g) is a homotopy inverse of S P ∞ ( f ). Consequently, the spaces
S P ∞ X and S P ∞ Y are homotopy equivalent. Again S P ∞ {∗} = {∗} proves the sec-
ond part. ❑
Theorem 2.9.3 S P ∞ : Htp ∗ → Htp ∗ is a covariant functor.
Proof It follows from Theorem 2.9.1 and Proposition B.2.18. ❑
2.10 Applications
This section presents some interesting immediate applications of homotopy. It deals

with some extension problems and proves ‘Fundamental Theorem of Algebra’ by
using homotopic concepts.
2.10.1 Extension Problems
This subsection solves some extensions problems with the help of homotopy.
Theorem 2.10.1 A continuous map f : S n → Y from S n to any space Y can be
continuously extended over D n+1 if and only if f is nullhomotopic, i.e., iff f is
homotopic to a constant map.
Proof Let c : S n → Y be a constant map defined by c(S n ) = y0 ∈ Y such that f c.
Then exists a homotopy H : S n × I → Y such that H (x, 0) = f (x) and H (x, 1) =
c(x) = y0 , ∀ x ∈ S n . We now construct a map F : D n+1 → Y by the rule

y0 , 0 ≤ ||x|| ≤ 1/2
F(x) =
H x
||x||
,2 − 2||x|| , 1/2 ≤ ||x|| ≤ 1.
Since at ||x|| = 21 , H ( ||x||

x
, 1) = y0 , F is well defined. Again, since its retraction to
each of the closed sets C 1 = {x ∈ D n+1 : 0 ≤ ||x|| ≤ 1/2} and C2 = {x ∈ D n+1 :
1/2 ≤ ||x|| ≤ 1} is continuous, F agrees on C1 ∩ C2 and D n+1 = C1 ∪ C2 , F is
continuous by Pasting lemma. Moreover, ∀ x ∈ S n , ||x|| = 1 and hence F(x) =
H (x, 0) = f (x) ⇒ F is a continuous extension of f over D n+1 . Thus f c ⇒ f
has a continuous extension over D n+1 . Conversely, let F : D n+1 → Y be a con-
tinuous extension of f : S n → Y . Then ∀ x ∈ S n , F(x) = f (x). Suppose p0 ∈ S n
and f ( p0 ) = y0 ∈ Y. We now define a mapping H : S n × I → Y by H (x, t) =
F((1 − t)x + t p0 ). H is well defined, because D n+1 is a convex set. Moreover, H

is continuous and H : f c. ❑
Theorem 2.10.2 Any continuous map from S n to a contractible space has a contin-
uous extension over D n+1 .
Proof Let c : S n → Y be a constant map from S n to a contractible space Y and

f : S n → Y be an arbitrary continuous map. Then f c by Corollary 2.6.4. Hence
it follows by Theorem 2.10.1 that f has a continuous extension over D n+1 . ❑
Theorem 2.10.3 Let p0 be an arbitrary point of S n and let f : S n → Y be contin-

uous. Then the following statements are equivalent.
(a) f is nullhomotopic.
(b) f can be continuously extended over D n+1 .
(c) f is nullhomotopic relative to { p0 }.
Proof (a) ⇒ (b) follows from Theorem 2.10.1.

(b) ⇒ (c) Let F : D n+1 → Y be a continuous extension of f over D n+1 . Suppose
f ( p0 ) = y0 ∈ Y . Define a map H : S n × I → Y by H (x, t) = F((1 − t)x +
t p0 ) as shown in Fig. 2.27.
Then ∀ x ∈ S n , H (x, 0) = F(x) = f (x), H (x, 1) = F( p0 ) = f ( p0 ) = y0 =
c(x) and H ( p0 , t) = F( p0 ) = f ( p0 ) = y0 = c( p0 ), ∀ t ∈ I. Hence H : f
c rel { p0 }.
(c) ⇒ (a) It follows trivially.
❑
Proposition 2.10.4 There exists a continuous map f : D n → S n−1 with f ◦ i = 1d

iff the identity map 1d : S n−1 → S n−1 is nullhomotopic.
Proof Suppose there exists such a map f : D n → S n−1 . Define a homotopy
H : S n−1 × I → S n−1 , (x, t) → f (t x).
Then H (x, 1) = x, ∀ x ∈ S n−1 and H (x, 0) = f (0), ∀ x ∈ S n−1 , i.e., H (x, 0) is

independent of x. Hence 1d is homotopic to a constant map. Conversely, let there
exist H : S n−1 × I → S n−1 such that H (x, 0) = c and H (x, 1) = x. Define
Fig. 2.27 Construction of H p0
x
2.10 Applications 99

H x
||x||
, ||x|| , if x = 0
f :D →S
n n−1
, x →
c, if x = 0
Since S n−1 is compact, H is uniformly continuous. Hence for every > 0, e a δ > 0
(depends on but not on x) such that ||H (x, t) − c|| < δ if t < . This shows that f
is continuous at 0 ∈ D n . ❑
Proposition 2.10.5 Let (X, A) be a normal pair (i.e., X is normal and A is closed
in X ) such that X × I is normal and f : X → S n be a continuous function. Then
every homotopy of f | A can be extended to a homotopy of f .
Proof Let H : A × I → S n be a homotopy of f | A . Then H can be extended to
a continuous map H̃ : X × {0} ∪ A × I be setting H (x, 0) = f (x) for all x ∈ X .
Then by using Ex. 22 of Sect. 1.16 of Chap. 1 it follows that there exists an extension
F : X × I → S n , which is required homotopy. ❑
Proposition 2.10.6 If a continuous map f : X → S n is essential, then f (X ) = S n
(i.e., f is a surjection).
Proof If f (X ) = S n , then there exists an element y ∈ S n − f (X ). Since S n − y is
contractible to a point and f (X ) ⊂ S n − y, it follows that f is inessential. This is a
contradiction. ❑
Proposition 2.10.7 Let (X, A) be a normal pair such that X × I is normal. Then
every inessential map f : A → S n admits an inessential extension f˜ : X → S n .
Proof Let g : X → S n be a map of X into a single point of S n . Then g| A is homotopic
to f . Hence the existence of f˜ follows from Proposition 2.10.5. ❑
Definition 2.10.8 A topological space X is said to be aspherical if every continuous
map f : S n → X extends to a continuous map f˜ : D n+1 → X .
Example 2.10.9 Every convex subspace of Euclidean space and every contractible
space are aspherical.
2.10.2 Fundamental Theorem of Algebra
This subsection applies the tools of homotopy to prove the celebrated fundamental
theorem of algebra which shows that the field of complex numbers is algebraically
closed. There are several methods to prove the fundamental theorem of algebra. We
now present a proof by homotopy. For an alternative proof see Theorem 3.8.1 of
Chap. 3.
Theorem 2.10.10 Let C denote the field of complex numbers, and C ρ ⊂ C ≈ R2
denote the circle at the origin and of radius ρ. Let f ρn : Cρ → C − {0} be the restric-
tion to Cρ to the map z → z n . If none of the maps f ρn is nullhomotopic (n ≥ 1 and
ρ > 0), then every nonconstant polynomial over C has a root in C.
Proof of Fundamental Theorem of Algebra by Homotopy:
Proof Without loss of generality we consider the polynomial g(z) = a0 + a1 z +

· · · + an−1 z n−1 + z n , ai ∈ C. We choose

n−1
ρ > max 1, |ai | (2.14)
i=0

n−1
We define a map F : Cρ × I → C by F(z, t) = z n + (1 − t)ai z i . Then F(z, t) =
i=0
0 for any (z, t) ∈ C ρ × I . Otherwise, F(z, t) = 0 for some z ∈ Cρ and t ∈ I would
n−1
n−1
n−1
imply z = −
n
(1 − t)ai z . This implies ρ ≤
i n
(1 − t)|ai |ρ ≤
i
|ai |ρi ≤
i=0 i=0 i=0

n−1
|ai |ρn−1 for ρ > 1, because ρi ≤ ρn−1 for ρ > 1.
i=0

n−1
Hence ρ ≤ |ai |, by canceling ρn−1
i=0
⇒ a contradiction to the relation (2.14).
In other words, F(z, t) = 0 for any z with |z| = 1 and for any t ∈ I . We now
assume that g has a root in C. We define G : Cρ × I → C − {0} by G(z, t) = g((1 −
t)z). Since g has no root in C, G(z, t) = 0 and hence the values of G must lie in
C − {0}. Now G : g|Cρ k, where k is the constant map z → g(0) = a0 at a0 . Hence
g|Cρ is nullhomotopic. Again g|Cρ f ρn . Thus f ρn is nullhomotopic by symmetric and
transitive properties of the relation . This contradicts the hypothesis. Consequently,
g has a complex root. ❑
2.11 Exercises
1. For all n ≥ 0, show that the topological spaces S 1 ∧ S n and S n+1 are homeo-
morphic.
[Hint: Let S n+1 be the (n + 1)-sphere in Rn+2 , S n be equator, D n+1 be the n + 1-
disk embedded in D n+2 , H+n+1 be upper hemisphere, H−n+1 be lower hemisphere
and s0 = (1, 0, 0, . . . , 0) be base point. Now proceed as in Proposition 2.5.6.]
2. Given a collection of pointed topological spaces X α , Yα (α ∈ A), and maps f α
gα : X α → Yα , show that × f α ×gα .
[Hint: Let Fα : X α × I → Yα be a homotopy between f α and gα . Then F :
(×X α ) × I → ×Yα , defined by F((x α ), t) = (Fα (xα , t)), ∀ t ∈ I is continuous
and a homotopy between × f α and ×gα (relative to base point).]
2.11 Exercises 101
3. Consider the homotopy set [A, X ], where A is a fixed space. Show that a contin-
uous map f : X → Y induces a function f ∗ : [A, X ] → [A, Y ] satisfying the
following properties:
(i) If f g, then f ∗ = g∗ ;
(ii) If 1 X : X → X is the identity map, then 1 X ∗ : [A, X ] → [A, X ] is the iden-
tity function;
(iii) If g : Y → Z is another continuous map, then (g ◦ f )∗ = g∗ ◦ f ∗ .
Deduce that if X Y , then ∃ a bijection between the sets [A, X ] and [A, Y ].
What are the corresponding results for the sets [X, A] for a fixed space A?
[See Theorems 2.3.1 and 2.3.5 and their corollaries.]
4. Show that
(a) S 1 = {z ∈ C : |z| = 1} is a topological group under usual multiplication of
complex numbers.
(b) For any space X , pointwise multiplication endows the set of continuous
maps X → S 1 with the structure of an abelian group. It is compatible with
homotopy and then the set [X, S 1 ] acquires the structure of a group.
(c) If f : Y → X is continuous then f ∗ : [X, S 1 ] → [Y, S 1 ] is a homomor-
phism.
5. Show that a space X is contractible iff it is deformable into one of its points.
6. Show that if A is a deformation retract of X , then A and X have the same
homotopy type.
7. Show that any one-point subset of a convex subspace Y of Rn is a strong defor-
mation retract of Y .
8. Let X be the closed unit square and A be the comb space. Show that A is weak
deformation retract of X but not a deformation retract of X .
9. Show that the point (0, 1) of the comb space X is a deformation retract of X but
not a strong deformation retract of X .
10. Let X be a Hausdorff space and A ⊂ X be a retract of X . Prove that A is closed
in X . Hence show that an open interval (0, 1) cannot be a retract of any closed
subset of the real line R 1 .
11. Show that
(a) A continuous map f : X → Y is nullhomotopic iff it has a continuous exten-
sion over the cone C X = (X × I )/X × {1}.
(b) Given a continuous map f : X → Y , its mapping cylinder M f = (X × I ) ∪
Y = (X × I ) ∪ Y/∼, where for all x ∈ X, ∼ identifies (x, 1) with f (x).
(c) S 1 = {z ∈ C − {0} : |z| = 1} is a strong deformation retract of C − {0}.
(d) S 1 and C − {0} have the same homotopy type.
(e) For all f : X → Y , the space Y is a deformation retract of the its mapping
cylinder M f .
(f) Any continuous map from a closed subset of Rn into a sphere is extendable
over the whole of Rm iff f is essential.
(g) Two constant maps ki : X → Y, x → yi , i = 0, 1 are homotopic iff ∃ a con-

tinuous curve γ : I → Y from y0 to y1 .
12. Let [X, Y ] denote the set of homotopy classes of maps f : X → Y . Show that
(i) for any space X, [X, I ] has a single element;
(ii) if X is path-connected, then [I, X ] has a single element;
(iii) a contractible space is path-connected;
(iv) if X is contractible, then for any space Y, [Y, X ] has a single element;
(v) if X is contractible, and Y is path-connected, then [X, Y ] has a single ele-
ment.
13. Show that a retract of a contractible space is contractible.
14. Show that R n+1 − {0} is homotopy equivalent to S n .
15. Show that the space X = {(x, y, z) ∈ R3 : y 2 > x z} is homotopy equivalent to a
circle. Interpret this result by considering the roots of the equation ax 2 + 2hx y +
by 2 = 0.
16. Let X = {( p, q) ∈ S n × S n : p = −q}. Show that the map f : S n → X defined
by f ( p) = ( p, p) is a homotopy equivalence.
17. In R2 , define A1 = {(x1 , x2 ) ∈ R2 : (x1 − 1)2 + x 22 = 1}, A2 = {(x1 , x2 ) ∈ R2 :
(x1 + 1)2 + x 22 = 1}. Suppose Y = A1 ∪ A2 , X = Y \ {(2, 0), (−2, 0)}, A =
0 = {(0, 0)}. Show that A is a strong deformation retract of X .
18. (a) Let X and Y be pointed topological spaces. Show that there exists a bijection
ψ : [ X, Y ] → [X, Y ] such that it is natural in X and in Y in the sense
that if f : X → X and g : Y → Y are base point preserving continuous
maps, then the diagrams in the Fig. 2.28 and in Fig. 2.29 are commutative,
where the horizontal arrows represents the corresponding isomorphism.
(b) Show that for n ≥ 2 and any pointed Hausdorff space X the iterated loop
1
spaces n X (= (n−1 X ) = (n−1 X ) S ) are homotopy commutative H -
groups. Hence prove that for n ≥ 2 and pointed spaces X, Y , the groups
[X, n Y ] are abelian.
[Hint: See Theorem 2.5.14.]
19. Let (X, A) have the absolute homotopy extension property (AHEP) (in the sense
that A has HEP in X with respect to every space Y ) and A be contractible. Show
that the identification map p : X → X/A is a homotopy equivalence.
Fig. 2.28 Naturality of ψ in ∼

= / [X, ΩY ]
[ΣX, Y ]
X
(Σf )∗ f∗
∼
= / [X , ΩY ]

[ΣX , Y ]∗
Fig. 2.29 Naturality of ψ in ∼

= / [X, ΩY ]
[ΣX, Y ]
Y
g∗ (Ωg)∗
∼
= / [X, ΩY ]
[ΣX, Y ]∗
2.11 Exercises 103
20. (a) Let G be a fixed H -group with base point e with continuous multiplication
μ : G × G → G and homotopy inverse φ : G → G. Show that there exists
a contravariant functor π G : Htp ∗ → Grp .
(b) For each homotopy associative H -space K , show that π K is a contravariant
from Htp ∗ to the category of monoids and their homomorphisms.
(c) Show that π G is homotopy type invariant for each H -group G.
(d) Let G be a pointed topological space such that π G assumes values in Grp .
Show that G is an H -group. Moreover, for any pointed space X , show that
the group structure on π G (X ) and [X, G] coincide.
(e) Let α : G → H be a homomorphism of H -groups. Show that α induces
a natural transformation N (α) : π G → π H , where N (α)(X ) : [X, G] →
[X, H ] is defined by N (α)(X )([ f ]) = [α ◦ f ], ∀ [ f ] ∈ [X, G].
21. Given a closed curve C in the plane R2 × {0}, show that there exists a continuous
deformation deforming C into a spherical closed curve C̃ and conversely given
a spherical curve C̃, show that there exists a continuous deformation deforming
C̃ into a closed curve C in the plane R2 × {0} such that total normal twists of C
and C̃ remain the same.
22. (M. Fuchs) Prove that two topological spaces X and Y have the same homotopy
type iff they are homeomorphic to a strong deformation retract of a space Z .
23. Using the notation of Theorem 2.4.18, show that a pointed space P is an H -space
iff there is a continuous map μ : P × P → P such that μ ◦ i 1 = μ ◦ i 2 = c. The
map μ satisfies the condition [μ] = [ p1 ]·[ p2 ] and if f, g : X → P are base point
preserving continuous maps, then [ f ]·[g] is the homotopy class of the composite
f ×g μ
X −−−→ X × X −−−−−→ P × P −−−→ P.
24. Let X and Y be topological spaces and f : X → Y be a continuous map. Show

that Y is a strong deformation retract of its mapping cylinder M f .
25. Show that a continuous map f : X → Y has a left homotopy inverse iff X is a
retract of its mapping cylinder M f .
26. Show that a continuous map f : X → Y has a right homotopy inverse iff the
mapping cylinder M f deforms into X .
27. Show that
(a) A continuous map f : X → Y is a homotopy equivalence iff X is a defor-
mation retract of the mapping cylinder M f ;
(b) If D is such a deformation retraction, then D|Y ×{1} is a homotopy inverse to
f and for any homotopy inverse g, there is a deformation retract of M f into
X which gives g.
(c) Let X be a normal space. If A ⊂ X is the set of zeros of a continuous map
f : X → I , and if A is a strong deformation retract of a neighborhood U of
A in X , then (X × {0} ∪ A × I ) is a strong deformation retract of X × I .
(d) S 1 is a deformation retract of R2 \ {0}(= R2 − {0}).
(e) Möbius strip is homotopy equivalent to S 1 .
28. Show that the suspension

: Htp∗ → Htp ∗
is an endofunctor (i.e., a functor from Htp ∗ to itself).

29. Let X and Y be pointed Hausdorff spaces. Show that
(i) Both [ X, Y ] and [X, Y ] are groups.

(ii) The groups [ X, Y ] and [X, Y ] are isomorphic.
(iii) If X is an H -cogroup and Y is an H -group, then the products available on
[X, Y ] determine isomorphic groups which are abelian;
30. Let A be a closed (or open) subspace of X in Top ∗ . Then the inclusion i : A → X
induces closed (or open) inclusions S P ∞ (i) : S P ∞ A → S P ∞ X .
31. Let (X, A) be a pair of topological spaces such that X is a compact Hausdorff
space, A is closed in X and A is a strong deformation retract of X . Let p : X →
X/A be the identification map and p(A) = y ∈ X/A. Show that {y} is a strong
deformation retract of X/A.
32. (Steenrod) Let the space X be compactly generated and A be closed in X . Show
that the following statements are equivalent:
(i) (X, A) is an NDR-pair;
(ii) (X × I, X × {0} ∪ A × I ) is a DR-pair;
(iii) X × {0} ∪ A × I is a retract of X × I ;
(iv) (X, A) has the homotopy extension property (HEP) with respect to arbitrary
topological spaces.
33. (i) Let f, g : (X, A) → (Y, B) be two continuous maps of pair of spaces such
that f g. Show that that their induced maps f ,
g : X/A → Y/B are also
homotopic.
(ii) Let f : (X, A) → (Y, B) be a homotopy equivalence. Show that the induced
map f : X/A → Y/B is a (based) homotopy equivalence.
34. Let (X, A) be a pair of topological spaces such that A is closed in X and X × I
is a normal space. If there is a neighborhood U such that U is a retract of (X ×
{0} ∪ (A × I )), show that any continuous map G : (X × {0}) ∪ (A × I ) → Y
has a continuous extension over X × I .
[1] Adams, J.F., Algebraic Topology: A student’s Guide, Cambridge University

Press, Cambridge, 1972.
[2] Adhikari, M.R., and Adhikari, Avishek, Basic Modern Algebra with Applica-
tions, Springer, New Delhi, New York, Heidelberg, 2014.
2.12 Additional Reading 105
[3] Adhikari, Avishek and Rana, P.K., A Study of Functors Associated with
Topological Groups, Studia Universiatis, Babes-Bolyai Mathematica, XLVI(4),
3–14, 2001
[4] Arkowitz, Martin, Introduction to Homotopy Theory, Springer, New York, 2011.
[5] Armstrong, M.A., Basic Topology, Springer-Verlag, New York, 1983.
[6] Aguilar, Gitler, S., Prieto, C., Algebraic Topology from a Homotopical View
Point, Springer-Verlag, New York, 2002.
[7] Chatterjee, B.C., Ganguly, S., and Adhikari, M.R., A Textbook of Topology,
Asian Books Pvt.Ltd., New Delhi, 2002.
[8] Croom, F.H., Basic Concepts of Algebraic Topology, Springer-Verlag, New
York, Heidelberg, Berlin, 1978.
[9] Dugundji, J., Topology, Allyn & Bacon, Newtown, MA, 1966.
[10] Dieudonné, J., A History of Algebraic and Differential Topology, 1900–1960,
Modern Birkhäuser, 1989.
[11] Gray, B., Homotopy Theory, An Introduction to Algebraic Topology, Academic
Press, New York, 1975.
[12] Hilton, P.J., An introduction to Homotopy Theory, Cambridge University Press,
Cambridge, 1983.
[13] Hu, S.T., Homotopy Theory, Academic Press, New York, 1959.
[14] Massey, W.S., A Basic Course in Algebraic Topology, Springer-Verlag, New
York, Berlin, Heidelberg, 1991.
[16] Rotman, J.J., An Introduction to Algebraic Topology, Springer-Verlag, New
York, 1988.
[17] Satya, Deo Algebraic Topology: A primer, Hindustan Book Agency, New Delhi,
2003.
[19] Whitehead, G.W., Elements of Homotopy Theory, Springer-Verlag, New York,
Heidelberg, Berlin, 1978.
References
Adams, J.F.: Vector fields on spheres. Ann. Math. 75, 603–632 (1962)
Adams, J.F.: Algebraic Topology: A Student’s Guide. Cambridge University Press, Cambridge
(1972)
Adhikari, A., Rana, P.K.: A study of functors associated with topological groups, Studia Universiatis,
Babes-Bolyai Mathematica, vol. XLVI(4), pp. 3–14 (2001)
Adhikari, M.R., Adhikari, A.: Basic Modern Algebra with Applications. Springer, New York (2014)
Arkowitz, M.: Introduction to Homotopy Theory. Springer, New York (2011)
Armstrong, M.A.: Basic Topology. Springer, New York (1983)
Aguilar, M., Gitler, S., Prieto, C.: Algebraic Topology from a Homotopical View Point. Springer,
New York (2002)
Barratt, M.G.: Track groups I. Proc. Lond. Math. Soc. 5(3), 71–106 (1955)
New Delhi (2002)
Croom, F.H.: Basic Concepts of Algebraic Topology. Springer, New York (1978)
Dieudonné, J.: A History of Algebraic and Differential Topology, 1990-1960. Modern Birkhäuser,
Basel (1989)
Eilenberg, S., Steenrod, N.: Foundations of Algebraic Topology. Princeton University Press, Prince-
ton (1952)
Gray, B.: Homotopy Theory, An Introduction to Algebraic Topology. Acamedic Press, New York
(1975)
Hatcher, A.: Algebraic Topology. Cambridge University Press, Cambridge (2002)
Hilton, P.J.: An Introduction to Homotopy Theory. Cambridge University Press, Cambridge (1983)
Hu, S.T.: Homotopy Theory. Academic Press, New York (1959)
Hurewicz, W.: Beitrage der Topologie der Deformationen. Proc. K. Akad. Wet. Ser. A 38, 112–119,
521–528 (1935)
Massey, W.S.: A Basic Course in Algebraic Topology. Springer, New York (1991)
Rotman, J.J.: An Introduction to Algebraic Topology. Springer, New York (1988)
Satya, D.: Algebraic Topology: A Primer. Hindustan Book Agency, New Delhi (2003)
Steenrod, N.: A convenient category of topological spaces. Mich. Math J. 14, 133–152 (1967)
Whitehead, G.W.: Elements of Homotopy Theory. Springer, New York (1978)
Chapter 3
The Fundamental Groups
This chapter continues the study of homotopy theory though the concept of
fundamental groups invented by H. Poincaré (1854–1912) in 1895, which conveys
the first transition from topology to algebra by assigning a group structure on the
set of relative homotopy classes of loops in a functorial way. Its group structure is
proved in Sects. 3.1 and 3.2 in two different ways. This group earlier called Poincaré
group, is now known as fundamental group. It plays an influential role in the study
of algebraic topology.
Properties and characteristics which are shared by homeomorphic spaces are
called topological properties and topological invariants; on the other hand those by
homotopy equivalent spaces are called homotopy properties and homotopy invari-
ants. The Euler characteristic invented by L. Euler (1703–1783) in 1752 is an integral
invariant, which distinguishes non-homeomorphic spaces. The search of other invari-
ants has established connections between topology and modern algebra in such a way
that homeomorphic spaces have isomorphic algebraic structures. Historically, the
concept of fundamental group introduced by Poincaré in 1895 is the first important
invariant of homotopy theory which came from such a search. His work explained
the difference between curves deformable to one another and curves bounding a
larger space. The first one led to the concepts of homotopy and fundamental group.
Fundamental group is one of the basic homotopy invariants. It is a very powerful
invariant in algebraic topology and is the first of a series of algebraic invariants πn
associated with a topological space with a base point.
The classification of topological spaces up to homeomorphism is the main prob-
lem of topology. Given two topological spaces, either we have to find an explicit
expression for a homeomorphism between them or we have to show that no such
homeomorphism exists. In the latter case, it does not suffice to consider any special
mapping, and it is impossible to consider all the mappings. So for proving nonex-
istence of a homeomorphism we use indirect arguments. In particular, we find a
property or a characteristic shared by homeomophic spaces. This is the basic moti-
vation of invention of homotopy and homology groups in algebraic topology.
DOI 10.1007/978-81-322-2843-1_3
108 3 The Fundamental Groups
Using the tools of the fundamental groups, this chapter introduces the concept of
degree function of a continuous map f : (I, I˙) → (S 1 , 1) and develops the neces-
sary tools to compute and study the fundamental group of the circle. It also studies
Brouwer fixed point theorem for dimension 2, fundamental theorem of algebra, vec-
tor field problems on D 2 , and knot groups, and finally computes fundamental groups
of some important spaces.
For this chapter the books Armstrong (1983), Bredon (1993), Croom (1978),
Massey (1991), Maunder (1970), Munkres (1975), Rotman (1988), Switzer (1975),
Whitehead (1978) and some others are referred in the Bibliography.
3.1 Fundamental Groups: Introductory Concepts
This section introduces the concept of fundamental groups and starts studying the
basic elementary properties of fundamental groups with an eye to apply them as tools
for the study of subsequent chapters. It is one of the several key homotopy invariants
which exist associated with topological spaces.
3.1.1 Basic Motivation
The basic motivation of the concept of the fundamental group is given by a geometric
approach. Consider the disk X with a hole and another disk Y without a hole as shown
in Figs. 3.1 and 3.2.
Any loop in Fig. 3.1 cannot be continuously shrunk to a point; on the other hand
any loop in Fig. 3.2 can be continuously shrunk to a point.
For example, the loop α in Fig. 3.1 cannot be continuously shrunk to a point due to
existence of a hole in X , but some loops in X such as β may be continuously shrunk
to a point but not all loops. This characterizes the difference between the spaces
X and Y . This difference leads to the concept of fundamental group of a pointed
topological space.
Fig. 3.1 A disk X with a

hole
β
α
X
3.1 Fundamental Groups: Introductory Concepts 109
Fig. 3.2 A disk Y without a

hole
Fundamental group is the first of a sequence of functors πn (see Chap. 7), called
homotopy group functors from the category of pointed topological spaces to the
category of groups. 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, x 0 ), the set π1 (X, x0 ) is defined to be the set of homotopy classes of paths
f : I → X that send 0 and 1 to x0 . Each such path is called a loop in X based at x 0 . It is
shown that π1 (X, x0 ) admits a group structure.The group π1 (X, x 0 ) depends on X as
well as on x 0 ∈ X and is called the fundamental group or Poincaré group of the space
X based at x 0 . 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, x 0 ) and
π1 (Y, y0 ).
Definition 3.1.1 Let X be a topological space and u : I → X be a continuous map.

Then u is said to be a path in X , u(0) is called the initial point and u(1) is called the
terminal point of the path u.
If u and v are two paths in X such that u(1) = v(0), then we can define a new
path, called the product of u and v denoted by u ∗ v as follows:

u(2t), 0 ≤ t ≤ 1/2
(u ∗ v)(t) = (3.1)
v(2t − 1), 1/2 ≤ t ≤ 1
u ∗ v : I → X is continuous by Pasting Lemma. The initial point of u ∗ v is the

initial point of u and the terminal point of u ∗ v is the terminal point of v.
If w is a third path in X such that v(1) = w(0), then the paths u ∗ (v ∗ w) and
(u ∗ v) ∗ w are defined by
⎧
⎪
⎨u(2t), 0 ≤ t ≤ 1/2
(u ∗ (v ∗ w))(t) = v(4t − 2), 1/2 ≤ t ≤ 3/4
⎪
⎩
w(4t − 3), 3/4 ≤ t ≤ 1
and ⎧
⎪
⎨u(4t), 0 ≤ t ≤ 1/4
((u ∗ v) ∗ w)(t) = v(4t − 1), 1/4 ≤ t ≤ 1/2
⎪
⎩
w(2t − 1), 1/2 ≤ t ≤ 1
These two paths in X are not necessarily the same paths, because at t = 12 , images
of these two paths may not be the same, since u(1) and v(1) may not be equal. This
shows that the product of paths is in general not an associative operation. Even for a
fixed x0 ∈ X , the product of loops in X based at x0 need not be associative, because
their respective images at t = 14 may not be equal. To overcome this difficulty we
consider an equivalence relation on the set (X, x0 ) of all loops in X based at x0 ∈ X .
Definition 3.1.2 A path u : I → X is called a loop in X based at x0 ∈ X if

u(0) = u(1) = x 0 . If I˙ = {0, 1}, then a loop f in X based at x0 is a continuous map
u : (I, I˙) → (X, x0 ). In particular, the constant map c : I → X, t → x 0 , ∀ t ∈ I ,
is called a constant path or a null loop in X at x0 .
Definition 3.1.3 Let u, v : (I, I˙) → (X, x 0 ) be two loops in X based at x 0 . Then u
and v are said to be homotopic relative to the subspace I˙ = {0, 1} of I denoted by
u v rel I˙, if ∃ a continuous map
F : I × I → X such that
F(t, 0) = u(t), ∀ t ∈ I ,
F(t, 1) = v(t), ∀ t ∈ I ,
and F(0, s) = F(1, s) = x 0 , ∀ s ∈ I .
Let (X, x 0 ) be the set of all loops in X based at x 0 . Then it follows from Theorem
2.1.37 of Chap. 2 that ‘’ is an equivalence relation on (X, x 0 ). This gives the set
of homotopy classes of loops relative to I˙ = {0, 1}, denoted by π1 (X, x0 ). Thus
π1 (X, x0 ) is the quotient set (X, x 0 )/.
We want to define a composition on π1 (X, x 0 ) to make it a group. First, we define
composition ∗ on (X, x 0 ) and then we carry it to (X, x 0 )/ = π1 (X, x0 ).
Definition 3.1.4 Given u, v ∈ (X, x0 ) their product u ∗ v : (I, I˙) → (X, x0 ) is

defined by

u(2t), 0 ≤ t ≤ 1/2
(u ∗ v)(t) = (3.2)
v(2t − 1), 1/2 ≤ t ≤ 1
Then at t = 21 , u(2t) = u(1) = x 0 = v(0) = v(2t − 1) shows that u ∗ v is well

defined and continuous by Pasting Lemma. Moreover (u ∗ v)(0) = u(0) = x0 =
v(0) = (u ∗ v)(1) =⇒ u ∗ v is a loop in X based at x0 =⇒ u ∗ v ∈ (X, x0 )
We now extend this definition for the product of three loops. Given loops u, v, w :
(I, I˙) → (X, x 0 ), their product u ∗ v ∗ w : I → X is defined by
⎧
⎪
⎨u(3t), 0 ≤ t ≤ 1/3
(u ∗ v ∗ w)(t) = v(3t − 1), 1/3 ≤ t ≤ 2/3
⎪
⎩
w(3t − 2), 2/3 ≤ t ≤ 1
Then as before, u ∗ v ∗ w ∈ (X, x0 ).
Definition 3.1.5 If u ∈ (X, x0 ), then its inverse u −1 : (I, I˙) → (X, x0 ) is defined
by u −1 (t) = u(1 − t), ∀ t ∈ I .
Clearly, u −1 ∈ (X, x0 ). Thus u ∈ (X, x0 ) =⇒ u −1 ∈ (X, x 0 ).
Remark 3.1.6 u and u −1 give the same set of points of X but their directions are
opposite.
Proposition 3.1.7 If u 1 , u 2 , v1 , v2 ∈ (X, x0 ) and u 1 u 2 rel I˙, v1 v2 rel I˙,

then u 1 ∗ v1 u 2 ∗ v2 rel I˙.
Proof Let F : u 1 u 2 rel I˙ and G : v1 v2 rel I˙. Then F(t, 0) = u 1 (t), F(t, 1) =
u 2 (t), ∀ t ∈ I, F(0, s) = x 0 = F(1, s), ∀ s ∈ I and G(t, 0) = v1 (t), G(t, 1) =
v2 (t), ∀ t ∈ I, G(0, s) = x0 = G(1, s), ∀ s ∈ I .
Define a map H : I × I → X by

F(2t, s), 0 ≤ t ≤ 1/2
H (t, s) =
G(2t − 1, s), 1/2 ≤ t ≤ 1
Then H is well defined. Moreover, it is continuous by Pasting Lemma, since its

restrictions to I × [0, 12 ] and I × [ 21 , 1] are continuous, and both functions agree on
{ 21 } × I and their restrictions to [0, 12 ] × I and [ 21 , 1] × I are continuous. Again,

F(2t, 0), 0 ≤ t ≤ 1/2
H (t, 0) =
G(2t − 1, 0), 1/2 ≤t ≤1
u 1 (2t), 0 ≤ t ≤ 1/2
=
v1 (2t − 1), 1/2 ≤ t ≤ 1
= (u 1 ∗ v1 )(t), ∀ t ∈ I.
Similarly, H (t, 1) = (u 2 ∗ v2 )(t), ∀ t ∈ I, H (0, s) = x 0 = F(0, s), ∀ s ∈ I and

H (1, s) = x0 = G(1, s), ∀ s ∈ I . Consequently, H : u 1 ∗ v1 u 2 ∗ v2 rel I˙. ❑
Proposition 3.1.8 If u, v ∈ (X, x0 ) and u v rel I˙, then u −1 v −1 rel I˙.
Proof Let F : u v rel I˙. Then F(t, 0) = u(t), F(t, 1) = v(t), ∀ t ∈ I and
F(0, s) = x 0 = F1 (1, s).
Define G : I × I → X by G(t, s) = F(1 − t, s).
Then G is a continuous function such that
G(t, 0) = F(1 − t, 0) = u(1 − t) = u −1 (t), ∀ t ∈ I,

G(t, 1) = F(1 − t, 1) = v(1 − t) = v −1 (t), ∀ t ∈ I
and G(0, s) = F(1, s) = x0 , G(1, s) = F(0, s) = x0 .
Consequently, G : u −1 v −1 rel I˙. ❑
Proposition 3.1.9 If u, v, w ∈ (X, x0 ), then u ∗ (v ∗ w) (u ∗ v) ∗ w rel I˙
Proof u ∗ (v ∗ w) : (I, I˙) → (X, x0 ) is defined by

u(2t), 0 ≤ t ≤ 1/2
(u ∗ (v ∗ w))(t) =
⎧(v ∗ w)(2t − 1), 1/2 ≤t ≤1
⎨ u(2t), 0 ≤ t ≤ 1/2
= v(4t − 2), 1/2 ≤ t ≤ 3/4
⎩
w(4t − 3), 3/4 ≤ t ≤ 1
Then u ∗ (v ∗ w) is well defined, continuous by Pasting Lemma and a loop in X

based at x 0 and therefore u ∗ (v ∗ w) ∈ (X, x0 ). On the other hand,

(u ∗ v)(2t), 0 ≤ t ≤ 1/2
((u ∗ v) ∗ w)(t) =
⎧w(2t − 1), 1/2 ≤ t ≤ 1
⎨ u(4t), 0 ≤ t ≤ 1/4
= v(4t − 1), 1/4 ≤ t ≤ 1/2
⎩
w(2t − 1), 1/2 ≤ t ≤ 1
As before, (u ∗ v) ∗ w ∈ (X, x 0 ).
Define a map H : I × I → X by the rule
⎧
⎨ u(4t/(1 + s)), 0 ≤ t ≤ (1 + s)/4
H (t, s) = v(4t − 1 − s), (1 + s)/4 ≤ t ≤ (2 + s)/4
⎩
w(1 − (4(1 − t)/(2 − s))), (2 + s)/4 ≤ t ≤ 1
Then H is well defined. Moreover it is continuous by Pasting Lemma. Now

⎧
⎨ u(4t), 0 ≤ t ≤ 1/4
H (t, 0) = v(4t − 1), 1/4 ≤ t ≤ 1/2
⎩
w(2t − 1), 1/2 ≤ t ≤ 1
= ((u ∗ v) ∗ w)(t), ∀ t ∈ I,
Fig. 3.3 Construction of (0,1) (1/2,1)(3/4,1) (1,1)

homotopy H
A B
R1 R2 R3
s
(0,0) (1/4,0)(1/2,0) t (1,0)
⎧
⎨ u(2t), 0 ≤ t ≤ 1/2
H (t, 1) = v(4t − 2), 1/2 ≤ t ≤ 3/4
⎩
w(4t − 3), 3/4 ≤ t ≤ 1
= (u ∗ (v ∗ w))(t), ∀ t ∈ I,
H (0, s) = u(0) = x0 , and H (1, s) = w(1) = x0 .

Hence (u ∗ v) ∗ w u ∗ (v ∗ w) rel I˙. ❑
The motivation for writing H comes from the diagram in Fig. 3.3.
Divide the square I × I into the three regions R1 , R2 , and R3 given by
R1 : 0 ≤ t ≤ (s + 1)/4, 0 ≤ s ≤ 1;
R2 : (s + 1)/4 ≤ t ≤ (s + 2)/4, 0 ≤ s ≤ 1;
R3 : (s + 2)/4 ≤ t ≤ 1, 0 ≤ s ≤ 1.
The two slanted lines are given by the equations: s = 4t − 1 and s = 4t − 2.

For a fixed s ∈ I , the horizontal line AB has three pieces. When s moves from
0 to 1, these pieces also change their positions. For s = 0, we obtain a partition
defining (u ∗ v) ∗ w and for s = 1, we obtain a partition defining u ∗ (v ∗ w). The
map H defined by u on R1 , v on R2 and w on R3 , each of which is continuous.
On their common boundary, each pair of maps agree. Then by Pasting Lemma H is
continuous and yields the required homotopy.
Proposition 3.1.10 If u ∈ (X, x0 ) and c : I → X is the constant loop at x 0

defined by c(t) = x0 , ∀ t ∈ I , then u ∗ c u rel I˙ and c ∗ u u rel I˙
Proof u ∗ c : I → X is defined by

u(2t), 0 ≤ t ≤ 1/2
(u ∗ c)(t) =
c(2t − 1), 1/2 ≤ t ≤ 1
u(2t), 0 ≤ t ≤ 1/2
=
x0 , 1/2 ≤ t ≤ 1
Then u ∗ c ∈ (X, x0 ).
Define a map H : I × I → X by

u(2t/(1 + s)), 0 ≤ t ≤ (1 + s)/2
H (t, s) =
x0 , (1 + s)/2 ≤ t ≤ 1
Then H : u ∗ c u rel I˙. Similarly, c ∗ c u r el I˙. ❑
Proposition 3.1.11 If u ∈ (X, x0 ), then u ∗ u −1 c rel I˙ and u −1 ∗ u c rel I˙
Proof u ∗ u −1 : I → X is given by

u(2t), 0 ≤ t ≤ 1/2
(u ∗ u −1 )(t) = −1
u (2t − 1), 1/2 ≤t ≤1
u(2t), 0 ≤ t ≤ 1/2
=
u(1 − 2t − 1), 1/2 ≤ t ≤ 1
u(2t), 0 ≤ t ≤ 1/2
=
u(2 − 2t), 1/2 ≤ t ≤ 1
Then u ∗ u −1 ∈ (X, x0 ). Define H : I × I → X by

u(2t (1 − s)), 0 ≤ t ≤ 1/2
H (t, s) =
u((2 − 2t)(1 − s)), 1/2 ≤ t ≤ 1
Then H : u ∗ u −1 c rel I˙.

Similarly, u −1 ∗ u c rel I˙. ❑
Theorem 3.1.12 π1 (X, x0 ) is a group.
Proof Let [u], [v] ∈ π1 (X, x 0 ). Then u, v ∈ (X, x 0 ) and u ∗ v(∈ (X, x0 )) is
defined by (3.1). This law of composition ‘∗’ is carried over to π1 (X, x0 ) to give the
composition ‘◦’ by the rule [u] ◦ [v] = [u ∗ v]. The composition ‘◦’ is well defined
by Proposition 3.1.7, because it is independent of the choice of the representatives of
the classes. This composition is associative by Proposition 3.1.9, [c] is the identity
element by Proposition 3.1.10 and any element [u] ∈ π1 (X, x 0 ) has an inverse
[u −1 ] ∈ π1 (X, x 0 ) by Proposition 3.1.11. Consequently, π1 (X, x0 ) is a group under
the composition ‘◦.’ ❑
Definition 3.1.13 π1 (X, x0 ) is called the Fundamental group or Poincaré group of

X based at x0 .
Remark 3.1.14 For an equivalent definition of π1 (X, x 0 ), and an alternative proof of

its group structure see Sect. 3.2.
Remark 3.1.15 The index ‘1’ in the notation π1 (X, x 0 ) appeared later than the nota-
tion π(X, x0 ) used by Poincaré in 1895. It is sometimes called the first or one-
dimensional homotopy group. There is an infinite sequence of groups πn (X, x 0 )
with n = 1, 2, 3, . . . , the first of them is the fundamental group. The higher dimen-
sional homotopy groups (see Chap. 7) were introduced by W. Hurewicz in 1935. For
n = 0, π0 (X, x 0 ), which is the set of path-connected components of X , is not a group
as a rule.
Example 3.1.16 (a) If X is a contractible space and x 0 ∈ X , then π1 (X, x0 ) = 0.
(b) π1 (Rn , x) = 0 for any x ∈ Rn .
(c) π1 (D n , d0 ) = 0 for any d0 ∈ D n
[Hint: (a) Let X be a contractible space. Then any continuous map f : (I, I˙) →
(X, x 0 ) is homotopic to the constant map c at x0 relative to I˙ =⇒ f c rel I˙ =⇒
[ f ] = [c] =⇒ π1 (X, x 0 ) = [c] = 0. (b) and (c) follow from (a).]
Example 3.1.17 If X is any convex set in Rn , then π1 (X, x 0 ) = 0.
We shall compute the fundamental group of the circle a little later.
It is natural to ask: does π1 (X, x 0 ) depend on the choice of the base point x0 ? How
are π1 (X, x0 ) and π1 (X, x 1 ) related for two different points x0 , x1 ∈ X ? If X is an
arbitrary topological space, then a loop in X at x0 being itself path-connected, lies
completely in the path-component of x0 . On the other hand, if x0 and x1 are points
in distinct path components of X , then π1 (X, x 0 ) and π1 (X, x1 ) are not at all related.
If x 0 and x1 lie in the same path component of X , we shall show that the groups
π1 (X, x0 ) and π1 (X, x 1 ) are isomorphic.
Theorem 3.1.18 If X is a path-connected space and x0 , x1 are two distinct points
of X , then the groups π1 (X, x0 ) and π1 (X, x1 ) are isomorphic.
Proof As X is path-connected and x 0 , x1 ∈ X , then ∃ a path u : I → X in X from
x0 to x1 with inverse path ū : I → X defined by ū(t) = u(1 − t) from x 1 to x0 as
shown in Fig. 3.4.
We now define a map
βu : π1 (X, x0 ) → π1 (X, x 1 ), [ f ] → [u ∗ f ∗ u].
Thus βu ([ f ]) = [ū ∗ f ∗ u], ∀ [ f ] ∈ π1 (X, x0 ).

Let g ∈ [ f ]. Then f g rel I˙ =⇒ ū ∗ f ∗ u ū ∗ g ∗ u rel I˙ =⇒
[ū ∗ f ∗ u] = [ū ∗ g ∗ u] =⇒ βu ([ f ]) = βu ([g]) =⇒ βu is well defined. Again
βu is homomorphism. Because, βu ([ f ] ◦ [g]) = βu ([ f ∗ g]) = [ū ∗ ( f ∗ g) ∗ u] =
[ū ∗ f ∗ u ∗ ū ∗ g ∗ u], since u ∗ ū c rel I˙ = [(ū ∗ f ∗ u) ∗ (ū ∗ g ∗ u)] =
[ū ∗ f ∗ u] ◦ [ū ∗ g ∗ u] = βu [ f ] ◦ βu [g], ∀ [ f ], [g] ∈ π1 (X, x0 ).
Fig. 3.4 Isomorphism of x1

fundamental groups in a X
ū
path-connected space
f
x0
Fig. 3.5 Path homotopic X

maps x1
v
u
x0
Finally, βu is an isomorphism with inverse βū : π1 (X, x 1 ) → π1 (X, x 0 ), because

(βū ◦ βu )([ f ]) = βū ([ū ∗ f ∗ u]) = [ū¯ ∗ ū ∗ f ∗ u ∗ ū] = [u ∗ ū ∗ f ∗ u ∗ ū] = [ f ],
since u ∗ ū c rel I˙ =⇒ βū ◦ βu = identity homomorphism. Similarly βu ∗ βū =
identity homomorphism. Hence βu is an isomorphism of groups. ❑
Corollary 3.1.19 If u is a path in X from x 0 to x1 , then u induces an isomorphism
βu : π1 (X, x 0 ) → π1 (X, x1 ).
Remark 3.1.20 If X is path-connected, the group π1 (X, x0 ) is, up to isomorphism,

independent of the choice of the base point x 0 . In this case, the notation π1 (X, x 0 ) is
abbreviated to π1 (X ).
We now consider the following situation.
Proposition 3.1.21 If u and v are two paths in X joining x0 to x1 which are path
homotopic, then their induced isomorphisms βu and βv are identical.
Proof If u and v are path homotopic as shown in Fig. 3.5, then ū and v̄ are also path
homotopic. Therefore, it follows that for any loop f in X based at x0 , ū ∗ f ∗ u is path
homotopic to v̄ ∗ f ∗ v. Consequently, βu ([ f ]) = βv ([ f ]), ∀ [ f ] ∈ π1 (X, x0 ) =⇒
βu = βv . ❑
We now characterize the commutativity of π1 (X, x0 ) for a path-connected space.

Theorem 3.1.22 Let X be a path-connected space and x0 , x1 ∈ X . Then the group
π1 (X, x0 ) is abelian if and only if for each pair of paths u, v from x0 to x1 , βu = βv .
Proof Let the group π1 (X, x0 ) be abelian. Then [u ∗ v̄] ◦ [ f ] = [ f ] ◦ [u ∗ v̄] for each
[ f ] ∈ π1 (X, x 0 ), since u ∗ v̄ is a loop in X based at x0 as shown in Fig. 3.6.
=⇒ u ∗ v̄ ∗ f f ∗ u ∗ v̄ rel I˙
Fig. 3.6 Loop in a

path-connected space
x1 X
v
f u
x0
=⇒ ū ∗ u ∗ v̄ ∗ f ∗ v ū ∗ f ∗ u ∗ v̄ ∗ v rel I˙
=⇒ [v̄ ∗ f ∗ v] = [ū ∗ f ∗ u]
=⇒ βv ([ f ]) = βu ([ f ]), ∀ [ f ] ∈ π1 (X, x 0 )
=⇒ βv = βu .
Conversely, let [ f ], [g] ∈ π1 (X, x0 ). Let u be a path in X joining x 0 to x1 . Then g ∗ u

is also a path in X joining x 0 to x 1 . By hypothesis, βg∗u ([ f ]) = βu ([ f ])
=⇒ [g ∗ u ∗ f ∗ (g ∗ u)] = [ū ∗ f ∗ u]
=⇒ ū ∗ ḡ ∗ f ∗ g ∗ u ū ∗ f ∗ u rel I˙
=⇒ u ∗ ū ∗ ḡ ∗ f ∗ g ∗ u ∗ ū u ∗ ū ∗ f ∗ u ∗ ū rel I˙
=⇒ ḡ ∗ f ∗ g f rel I˙
=⇒ f ∗ g g ∗ f rel I˙
=⇒ [ f ∗ g] = [g ∗ f ]
=⇒ [ f ] ◦ [g] = [g] ◦ [ f ]
=⇒ π1 (X, x0 ) is abelian. ❑
3.1.3 Functorial Property of π1
This subsection studies π1 from the view point of category theory (see Appendix B).
Recall that for every pointed topological space (X, x 0 ) there exists the fundamental
group π1 (X, x0 ). We now show that for every continuous map f : (X, x0 ) → (Y, y0 ),
there exists a homomorphism f ∗ : π1 (X, x 0 ) → π1 (Y, y0 ) satisfying some interesting
properties. The construction of the fundamental group shows that π1 is functorial.
Theorem 3.1.23 Every continuous map f : (X, x0 ) → (Y, y0 ) induces a group

homomorphism f ∗ : π1 (X, x0 ) → π1 (Y, y0 ) with the following properties:
(i) If 1 X : (X, x0 ) → (X, x0 ) is the identity map, then 1 X ∗ : π1 (X, x0 ) →
π1 (X, x0 ) is the identity automorphism;
(ii) If f : (X, x0 ) → (Y, x 0 ) and g : (Y, y0 ) → (Z , x0 ) are two continuous maps,
then (g ◦ f )∗ = g∗ ◦ f ∗ : π1 (X, x0 ) → π1 (Z , z 0 ).(These properties are called
functorial properties);
(iii) If f, g : (X, x0 ) → (Y, y0 ) are two continuous maps such that f g r el {x 0 },
then f ∗ = g∗ ;
(iv) If f : (X, x0 ) → (Y, y0 ) has a left (resp. right) homotopy inverse, then f ∗ :
π1 (X, x0 ) → π1 (Y, y0 ) is a monomorphism (resp. an epimorphism);
(v) If (X, x0 ) and (Y, y0 ) are homotopy equivalent spaces, then the groups π1 (X, x 0 )
and π1 (Y, y0 ) are isomorphic, i.e., if (X, x0 ) (Y, y0 ), then π1 (X, x0 ) ∼ =
π1 (Y, y0 );
(vi) If f : (X, x0 ) → (Y, y0 ) is a homotopy equivalence, then f ∗ : π1 (X, x0 ) →

π1 (Y, y0 ) is an isomorphism.
Proof Define f ∗ : π1 (X, x0 ) → π1 (Y, y0 ) by the rule f ∗ ([u]) = [ f ◦ u]. Then f ∗

is well defined. Because, f ◦ u is a loop in Y based at y0 and if u v rel I˙ by a
homotopy F, then f ◦ u f ◦ v rel I˙ by a homotopy f ◦ F.
We now show that f ∗ is a homomorphism.
Let [u], [v] ∈ π1 (X, x0 ). Then
f ∗ ([u] ◦ [v]) = f ∗ ([u ∗ v]) = [ f ◦ (u ∗ v)]. (3.3)
Now ( f ◦ (u ∗ v))(t) = f ((u ∗ v)(t))

f (u(2t)), 0 ≤ t ≤ 1/2
=
f (v(2t − 1)), 1/2 ≤t ≤1
( f ◦ u)(2t), 0 ≤ t ≤ 1/2
=
( f ◦ v)(2t − 1)), “ “ “ 1/2 ≤ t ≤ 1
= (( f ◦ u) ∗ ( f ◦ v))(t), ∀ t ∈ I
=⇒ f ◦ (u ∗ v) = ( f ◦ u) ∗ ( f ◦ v) (3.4)
Hence from (3.3) and (3.4) it follows that f ∗ ([u] ◦ [v]) = [( f ◦ u) ∗ ( f ◦ v)] =
f ∗ ([u]) ◦ f ∗ ([v]), ∀ [u], [v] ∈ π1 (X, x0 ) =⇒ f ∗ is a group homomorphism.
(i) Let 1 X : (X, x 0 ) → (X, x 0 ) be the identity map. Then 1 X ◦u = u for each loop u
in X based at x0 =⇒ 1 X ∗ ([u]) = [1 X ◦u] = [u], ∀ [u] ∈ π1 (X, x0 ) =⇒ 1 X ∗
is the identity automorphism on π1 (X, x 0 ).
(ii) For any [u] ∈ π1 (X, x 0 ), (g ◦ f )∗ ([u]) = [(g ◦ f ) ◦ u] = [g ◦ ( f ◦ u)] = g∗ ([ f ◦
u]) = g∗ ( f ∗ ([u])) = (g∗ ◦ f ∗ )[u], ∀ [u] ∈ π1 (X, x0 ) =⇒ (g ◦ f )∗ = g∗ ◦ f ∗ .
(iii) Let f g rel {x 0 }. Then ∀ [u] ∈ π1 (X, x0 ), f ◦ u g ◦ u rel {y0 } =⇒
f ∗ ([u]) = [ f ◦ u] = [g ◦ u] = g ∗ ([u]) =⇒ f ∗ = g∗ .
(iv) It follows from (iii).
(v) Suppose (X, x 0 ) (Y, y0 ). Hence ∃ two continuous functions f : (X, x 0 ) →
(Y, y0 ) and g : (Y, y0 ) → (X, x 0 ) such that g ◦ f 1 X and f ◦ g 1Y . Hence
(g ◦ f )∗ = 1 X ∗ and ( f ◦ g)∗ = 1Y ∗ by (iii). Thus it follows from (i) and (ii)
that g∗ ◦ f ∗ = identity automorphism and f ∗ ◦ g∗ = identity automorphism.
Consequently, f ∗ is an isomorphism with g∗ its inverse. For the second part,
proceed as in first part.
(vi) It follows from (v). ❑
Corollary 3.1.24 (i) The fundamental group of a pointed topological space is

invariant under homeomorphisms, and hence it is a topological invariant.
(ii) The fundamental group of a pointed topological space is invariant under homo-
topy equivalences, and hence it is a homotopy invariant.
We now express the results of Theorem 3.1.23 in the language of category theory.
Theorem 3.1.25 (a) π1 is a covariant functor from the category Top∗ of pointed
topological spaces and their base point preserving continuous maps to the cat-
egory Grp of groups and their homomorphisms. Moreover, if f, g : (X, x0 ) →
(Y, y0 ) are continuous maps and f g rel {x0 }, then π1 ( f ) = f ∗ = g∗ = π1 (g).
(b) π1 is a covariant functor from homotopy category Htp∗ of pointed topological
spaces and their homotopy classes of maps to the category Grp.
Proof (a) The object function is defined by (X, x0 ) → π1 (X, x 0 ) and the morphism
function is defined by f → π1 ( f ) = f ∗ . Then (a) follows from Theorem 3.1.23.
(b) follows from (a) and Theorem 3.1.23 (iii). ❑
3.1.4 Some Other Properties of π1
This subsection conveys the behavior of fundamental groups of some special spaces
such as H -spaces, simply connected spaces, product spaces, and some other spaces.
Proposition 3.1.26 Let X be a topological space and A be a subspace of X. If

i : A → X is the inclusion map and r : X → A is a retraction, then r∗ : π1 (X, a) →
π1 (A, a) is an epimorphism for each a ∈ A and i ∗ : π1 (A, a) → π1 (X, a) is a
monomorphism for each a ∈ A.
i r
Proof For each a ∈ A, the composite map (A, a) − → (X, a) − → (A, a) is the
identity map on (A, a) (see Definition 2.7.1 of Chap. 2). Consequently, the composite
homomorphism
i∗ r∗
π1 (A, a) −→ π1 (X, a) −→ π1 (A, a)
is the identity automorphism on π1 (A, a).

Hence r∗ is an epimorphism and i ∗ is a monomorphism. ❑
Proposition 3.1.27 Let A be a strong deformation retract of a space X . Then for

each a ∈ A, the groups π1 (A, a) and π1 (X, a) are isomorphic.
Proof Let A be a strong deformation retract of a topological X . Then ∃ a retraction

r : (X, a) → (A, a) for each a ∈ A such that 1 X i ◦ r rel A =⇒ i ∗ ◦ r ∗ =
id =⇒ i ∗ is an epimorphism. Again r ◦ i = id =⇒ r∗ ◦ i ∗ = id =⇒ i ∗ is a
monomorphism. Consequently, i ∗ : π1 (A, a) → π1 (X, a) is an isomorphism. ❑
Let X be a topological space. Let C be the path component of X containing

x0 . Then π1 (C, x 0 ) = π1 (X, x 0 ), because all loops and homotopies in X based at
x0 lie entirely in the subspace C. Therefore, π1 (X, x 0 ) depends only on the path
component of X containing x0 and provides no information about the set X − C. So
it is usual to deal only with path-connected spaces while studying the fundamental
groups.
Definition 3.1.28 A topological space X is called simply connected if it is path-

connected and π1 (X, x 0 ) = 0 for some x 0 ∈ X (hence for every x0 ∈ X ).
Theorem 3.1.29 If X is a simply connected space, then any two paths in X having
the same initial and final points are homotopic.
Proof Let X be simply connected and u and v be two paths in X from x0 to x1 as

shown in Fig. 3.7. Then u ∗ v̄ is a loop in X based at x 0 .
Since X is simply connected, u ∗ v̄ cx0 =⇒ u ∗ v̄ ∗v cx0 ∗v. Hence it follows
that [(u ∗ v̄) ∗ v] = [cx0 ∗ v] = [v]. But [(u ∗ v̄) ∗ v] = [u ∗ (v̄ ∗ v)] = [u ∗ cx1 ] = [u].
Consequently, [u] = [v] =⇒ u v. ❑
Theorem 3.1.30 Every contractible space is simply connected.
Proof Let X be a contractible space. Then there is a point x0 ∈ X and homotopy

H : X × I → X such that H (x, 0) = x and H (x, 1) = x0 , ∀ x ∈ X .
We claim that X is path-connected. If x ∈ X , the function
σx = H (x, −) : I → X, t → σx (t) = H (x, t)
is a path in X from σx (0) = H (x, 0) = x to σx (1) = H (x, 1) = x0 . Similarly, for

any y ∈ X , σ y is a path from y to x0 and hence σ̄ y (the inverse path of σ y ) is a path in
X from x0 to y. Thus any two paths x and y can be joined by the path σx ∗ σ̄ y in X .
Hence X is a path-connected space. Moreover, π1 (X, x 0 ) = 0 (see Example 3.1.16).
Consequently, X is simply connected. ❑
Corollary 3.1.31 Rn , D n and any convex subset of Rn is simply connected.
Definition 3.1.32 Let X ⊂ Rn be a subspace of Rn . Then X is said to be star convex

if for some x0 ∈ X , all the line segments joining x0 to any other point x of X lie
entirely in X , i.e., (1 − t)x + t x 0 ∈ X, ∀ t ∈ I .
Proposition 3.1.33 Let X ⊂ Rn be a star convex space. Then X is simply connected.
Proof As X is star convex, there is a continuous map H = X × I → X, (x, t) →

(1−t)x +t x0 . Hence X is a contractible space. Consequently, X is simply connected
by Theorem 3.1.30. ❑
We now characterize simply connected spaces as follows:
Theorem 3.1.34 A path-connected space X is simply connected if and only if any

two paths in X having the same initial point and same final point are homotopic.
Fig. 3.7 Paths in a simply X

connected space x1
u
v
x0
Proof Suppose X is simply connected. If u and v be two paths in X from x0 to x1 ,

then u v by Theorem 3.1.29. For the converse let X be path-connected and [ f ] ∈
π1 (X, x0 ). Then by hypothesis, f cx0 =⇒ [ f ] = [cx0 ] =⇒ π1 (X, x0 ) = 0. As
X is path-connected and π1 (X, x0 ) = 0, X is simply connected. ❑
Let X 1 and X 2 be two simply connected spaces. Is X = X 1 ∪X 2 simply connected?

The answer is negative. For an example, we shall show that the unit circle S 1 in the
complex plane can be expressed as the union of two contractible arcs but S 1 is not
simply connected (see Corollary 3.3.12). Note that contractible spaces are simply
connected by Theorem 3.1.30.
We now present the following interesting theorem.
Theorem 3.1.35 Let {Ui : i ∈ } be an open covering of a space X , where each Ui

is simply connected. Then X is itself simply connected if
(a) ∩ Ui = ∅
(b) i = j ∈ , Ui ∩ U j is path-connected.
Proof The space X is path-connected, because by hypothesis each of the open sets Ui
is path-connected and their intersection is nonempty.
It is now sufficient to prove that
π1 (X, x) = 0 for some x ∈ X . Suppose x 0 ∈ Ui . Let u : (I, I˙) → (X, x0 ) be a
i
loop at x0 . Then {u −1 (Ui )} is an open covering of I . Since I is compact, this covering
will have a Lebesgue number μ(say)> 0. This implies that ∃ a partition. 0 = t0 < t1 <
t2 < · · · < tn = 1 of I such that for 0 ≤ j ≤ n−1, u[t j , t j+1 ] is contained in some Ui .
Without loss of generality, we assume that u[t j , t j+1 ] ⊂ U j , 0 ≤ j ≤ n − 1. For each
j, we define a path u j in X by u j (s) = u((1 − s)t j + st j+1 ). Then U j (I ) is contained
in the simply connected open set U j for each j and [u] = [u 0 ∗ u 1 ∗ · · · ∗ u n−1 ].
Clearly, u(t1 ) ∈ U0 ∩ U1 and U0 ∩ U1 is path-connected containing the base point
x0 . Hence, we obtain a path v1 : I → X from x 0 to u(t1 ) such that v1 (I ) ⊂ U0 ∩ U1 .
Similarly, we can find a path v j from x 0 to u(t j ) lying entirely in U j−1 ∩ U j , for
j = 1, 2, . . . , n − 1. If v̄ j denotes the reverse path of v j , then we have
[u] = [u 0 ∗ v̄1 ∗ v1 ∗ u 1 ∗ v̄2 ∗ v2 ∗ u 2 ∗ · · · ∗ v̄n−1 ∗ vn−1 ∗ u n−1 ]

(3.5)
= [u 0 ∗ v̄1 ] ◦ [v1 ∗ u 1 ∗ v̄2 ] ◦ · · · ◦ [vn−1 ∗ u n−1 ]
The first term in the right hand side of (3.5) is a loop based at x0 and lying entirely
in the simply connected space U0 . Similarly the second term is a loop lying entirely
in the simply connected space U1 , and so on. Hence, each term is null homotopic in
U j for some j and so is in X . Consequently, [u] is the zero element of π1 (X, x 0 ).
Since [u] is an arbitrary element of π1 (X, x 0 ), it follows that π1 (X, x0 ) = 0. Hence
X is itself simply connected. ❑
Corollary 3.1.36 The n-sphere S n (n ≥ 2) is simply connected.

Proof Let p = (0, 0, . . . , 0, 1) ∈ Rn+1 be the north pole of S n and q =

(0, 0, . . . , 0, −1) ∈ Rn+1 be the south pole of S n . Then U = S n − { p} and
V = S n − {q} are both open sets homeomorphic to Rn by stereographic projection
and hence they are simply connected. We now prove that U ∩ V (to apply Theorem
3.1.35) is path-connected. Clearly, U ∩ V = S n − { p} − {q} ≈ R n − {0}(under
stereographic projection). We now show that Rn − {0} is path-connected.
Any point x ∈ Rn −{0} can be joined to the point x 0 = (1, 0, . . . , 0) by the straight
line path in Rn −{0}, except for point x of the form (a, 0, . . . , 0), where a < 0. For the
case a < 0, we can take the straight line path from x to x1 = (0, 1, 0, . . . , 0), followed
by the straight line path from x 1 to x 0 (This is possible for n ≥ 2). Consequently, S n
is simply connected by Theorem 3.1.35. ❑
We now give a relation between the fundamental group of a product space and
the fundamental groups of its factors. We recall that if A and B are groups with
operation ‘·’ then the cartesian product A × B can be endowed with a group structure
by the composition (a, b) · (a , b ) = (a · a , b · b ). Moreover, if α : G → A and
β : G → B are group homomorphisms, then the map ψ : G → A × B defined by
ψ(g) = (α(g), β(g)) is a group homomorphism.
Again we recall a basic property of the product topology. Let p1 : (X × Y ) → X,
p2 : X × Y → Y be the canonical projections. Given a pair of continuous maps
f : I → X, g : I → Y , there is a continuous map ( f, g) : I → X × Y defined by
( f, g)(t) = ( f (t), g(t)). Conversely, any continuous map h : I → X × Y defines a
pair of continuous maps p1 ◦ h : I → X, p2 ◦ h : I → Y .
We are now equipped to prove the following Theorem.
Theorem 3.1.37 Let X and Y be two topological spaces with base points x0 ∈ X
and y0 ∈ Y , respectively. Then the fundamental groups π1 (X × Y, (x0 , y0 )) and
π1 (X, x0 ) × π1 (Y, y0 ) are isomorphic.
Proof Let f : (I, I˙) → (X × Y, (x0 , y0 )) be a loop in X × Y at (x0 , y0 ). Then the

canonical projections p1 : (X ×Y ) → X, p2 : (X ×Y ) → Y are continuous maps of
product spaces and hence they induce homomorphisms p1∗ : π1 (X × Y, (x 0 , y0 )) →
π1 (X, x0 ) and p2∗ : π1 (x × Y, (x 0 , y0 )) → π1 (Y, y0 ) defined by p1∗ ([ f ]) = [ p1 ◦ f ]
and p2∗ ([ f ]) = [ p2 ◦ f ]. Then the map ψ = ( p1∗ , p2∗ ) : π1 (X × Y, (x0 , y0 )) →
π1 (X, x0 ) × π1 (Y, y0 ) defined by ψ([ f ]) = ( p∗ ([ f ]), q∗ ([ f ])) = ([ p ◦ f ], [q ◦ f ])
is a group homomorphism.
We claim that ψ is an isomorphism.
ψ is a monomorphism:
ker ψ = {[ f ] ∈ π1 (X × Y, (x0 , y0 )) : ψ([ f ]) = identity element of
π1 (X, x 0 ) × π1 (Y, y0 )}
= {[ f ] ∈ π1 (X × Y, (x0 , y0 )) : p1 ◦ f cx0 and p2 ◦ f = c y0 }.
Let f : (I. I˙) → (X × Y, (x0 , y0 ) be defined by f (t) = (g(t), h(t)), where g :

(I, I˙) → (X, x0 ) and h : (I, I˙) → (Y, y0 ) be the corresponding loops. Let M :
p1 ◦ f cx0 and H : p2 ◦ f c y0 . Then M(t, 0) = ( p1 ◦ f )(t) = p1 ( f (t)) =

g(t), ∀ t ∈ I and M(t, 1) = cx0 (t) = x0 , ∀ t ∈ I .
Similarly, H (t, 0) = h(t) and H (t, 1) = y0 , ∀ t ∈ I . We now define F : I × I →
Y × Y by F(t, s) = (M(t, s), H (t, s)). Then F(t, 0) = (M(t, 0), H (t, 0)) =
(g(t), h(t)) = f (t), ∀ t ∈ I F(t, 1) = (M(t, 1), H (t, 1)) = (x0 , y0 ) =
c(x0 ,y0 ) , ∀ t ∈ I ,
Thus F : f c(x0 ,y0 ) =⇒ [ f ] is the identity element [c(x0 ,y0 ) ]. Consequently,
ker ψ = {0} =⇒ ψ is a monomorphism.
ψ is an epimorphism: Let g : (I, I˙) → (X, x0 ) and h : (I, I˙) → (Y, y0 ) be two
loops. Then ∃ a continuous map (g, h) : (I, I˙) → (X × Y, (x0 , y0 )) defined by
(g, h)(t) = (g(t), h(t)). Let f : (I, I˙) → (X × Y, (x0 , y0 )) be the continuous map
defined by the rule f (t) = (g(t), h(t)). Then ψ([ f ]) = ([ p1 ◦ f ], [ p2 ◦ f ]) =
([g], [h]) =⇒ ψ is an epimorphism.
Consequently, ψ is an isomorphism. ❑
Corollary 3.1.38 The maps
θ : π1 (X, x0 ) × π1 (Y, y0 ) → π1 (X × Y, (x0 , y0 ))
defined by θ([g], [h]) = [g, h] is well defined and is the inverse of ψ.
Proof It follows from the definition of ψ. ❑
We will see that the fundamental groups of arbitrary spaces may be abelian or
nonabelian. But the next theorem shows that the fundamental group of an H-space
is always abelian. We recall that if G and H are groups, x ∈ G and y ∈ H , then in
the direct product
G × H, (x, e) · (e , y) = (x, y) = (e , y) · (x, e ).
holds.
Theorem 3.1.39 If (X, x0 ) is an H -space with multiplication μ and x 0 a homotopy

identity, then π1 (X, x 0 ) is abelian.
Proof The map
θ : π1 (X, x0 ) × π1 (Y, y0 ) → π1 (X × Y, (x0 , y0 )), ([g], [h]) → [(g, h)],
where (g, h) : I → X × Y is defined by t → (g(t), h(t)) is the inverse of the

isomorphism ψ and hence θ is an isomorphism (see Corollary 3.1.38). Let [ f ], [g] ∈
π1 (X, x0 ). Then
[g] = (μ ◦ (c, 1 X ))∗ [g] = μ∗ ([(c, 1 X ) ◦ g]) = μ∗ ([(c ◦ g, g)])

= (μ∗ ◦ θ)([c ◦ g], [g]) = (μ∗ ◦ θ)(e, [g]),
where e is the identity element of π1 (X, x0 ). Similarly, [ f ] = (μ∗ ◦ θ)([ f ], e). Since
μ∗ ◦ θ : π1 (X, x0 ) × π1 (X, x0 ) → π1 (X, x 0 ))
is a homomorphism, we have
(μ∗ ◦ θ)(([ f ], [g])) = (μ∗ ◦ θ)((e, [g]), ([ f ], e))

= (μ∗ ◦ θ)(e, [g]) · (μ∗ ◦ θ)([ f ], e) = [g] ◦ [ f ].
On the other hand, (μ∗ ◦ θ)(([ f ], [g])) = (μ∗ ◦ θ))(([ f ], e), (e, [g])) = (μ∗ ◦
θ)(([ f ], e)) · (μ∗ ◦ θ)(e, [g]) = [ f ] ◦ [g]. Hence
[ f ] ◦ [g] = [g] ◦ [ f ], ∀ [ f ], [g] ∈ π1 (X, x 0 ).
Consequently, π1 (X, x0 ) is abelian. ❑
Corollary 3.1.40 If G is a topological group with identity e then the π1 (G, e) is

abelian.
Proof Since every topological group is an H -group (hence it is an H-space), the

Corollary follows from Theorem 3.1.39. ❑
Remark 3.1.41 If for any pointed space (X, x 0 ), the fundamental group π1 (X, x0 ) is
not abelian (such X exists namely, figure-eight, double torus), then there is no way
to define a multiplication on X making it a topological group. Even we cannot equip
such X with the structure of an H -space. Otherwise, we would have a contradiction
to Theorem 3.1.39 and Corollary 3.1.40.
3.2 Alternative Definition of Fundamental Groups
This section presents an alternative definition of the fundamental group given by

Hurewicz (1904–1956) equivalent to its definition given in Sect. 3.1, which is con-
venient at many situations. The unit circle S1 = {e2πit : 0 ≤ t ≤ 1} in the complex
plane is the prototype of a loop. The elements of π1 (X, x0 ) may be equally well
considered as homotopy classes of maps f : (S 1 , 1) → (X, x0 ).
The basic aim of fundamental groups is to classify all loops up to homotopy
equivalences. This leads to give an alternative definition of fundamental groups.
The fundamental group π1 (X, x 0 ) is now defined as the homotopy classes of maps
f : (S 1 , 1) → (X, x0 ), instead of path classes, which is equivalent to the former
definition.
If f : I → X is a loop in X based at x0 , then f determines a pointed map
f˜ : (S 1 , 1) → (X, x0 ), e2πit → f (t).

3.2 Alternative Definition of Fundamental Groups 125
Conversely a pointed map α : (S 1 , 1) → (X, x0 ) determines a loop f α in X based

at x0 , defined by f α : I → X, t → α(e2πit ). Hence it follows that
Proposition 3.2.1 The function
ψ : π1 (X, x0 ) → [(S 1 , 1), (X, x 0 )], [ f ] → [ f˜]
is a bijection.
Remark 3.2.2 For our subsequence study, we use this identification implicitly and
make no difference between f and
f.
Using this identification map ψ we again prove the group structure of π1 (X, x 0 ).
Theorem 3.2.3 π1 (X, x0 ) is group.
Proof Given two two loops f, g : I → X in X based at x0 , there exist two pointed
maps
[ f˜], [g̃] : (S 1 , 1) → (X, x0 ), e2πit → f (t), g(t).
Define a multiplication in the set [(S 1 , 1) → (X, x0 )] by the rule
[ f˜] ◦ [g̃] = [(
f ∗ g)].
Hence it follows that π1 (X, x0 ) is a group with the constant map c : S 1 → {x 0 }

representing the identity element and the inverse of the map f : (S 1 , 1) → (X, x0 )
is represented by the map g : (S 1 , 1) → (X, x0 ), eiθ → f (e−θ ), 0 ≤ θ ≤ 2π and
the bijection
ψ : π1 (X, x0 ) → [(S 1 , 1), (X, x 0 )], [ f ] → [ f˜]
gives an isomorphism. ❑
Remark 3.2.4 By using the H -cogroup structure of S 1 , the group structure of

π(X, x 0 ) also follows.
Theorem 3.2.5 Let S 0 consist of two points −1 and 1 and let 1 be its based point.
Then the continuous map
λ : I → (S 0 ), t → [−1, t]
induces an isomorphism
λ̃ : [(S 0 ), 1] → π1 (X, x 0 ), [g] → [g ◦ λ].
Proof Since (S 0 ) ≈ S 1 , and S 1 is an H -cogroup, it follows that [(S 1 , 1), (X, x0 )]
is a group. Moreover, the products in [(S 1 , 1), (X, x 0 )] and π1 (X, x0 ) show that λ̃ is a
group homomorphism. Again as λ̃ is a bijection, it follows that λ̃ is an isomorphism

of groups. ❑
Remark 3.2.6 It is sometimes convenient to consider the elements of π1 (X, x0 ) as

homotopy classes of continuous maps f : (S 1 , 1) → (X, x0 ). The circle S 1 is a group
under usual multiplication of complex numbers. The composition law in π1 (S 1 , 1)
can also be defined by using this multiplication.
3.3 Degree Function and the Fundamental

Group of the Circle
This section introduces the concept of degree function of a continuous map f :

(I, I˙) → (S 1 , 1) and develops the necessary tools to compute and study the funda-
mental group of the circle and utilizes degree function to exhibit a space which is not
simply connected, i.e., a space X such that π1 (X, x 0 ) = 0. Since π1 (X, x0 ) consists
of relative homotopy class of maps f : S 1 → X , the space X = S 1 demands its
natural consideration. We consider the particular case of [X, Y ] when X = Y = S 1 .
Let f : S 1 → S 1 be a continuous map. As X moves around S 1 , f (x) will move
around S 1 some integer number of times. This integer is called the degree of f . We
now formalize this concept of degree. For example, the homotopy classes of loops
on S 1 based at 1 is completely characterized by their degrees, which are integers.
Recall that S 1 = {z ∈ C : |z| = 1} is a topological group with 1 as its identity
element under the usual multiplication of complex numbers. The circle S 1 is studied
through the real line R. The homotopy class of a loop is determined by the number
of times it winds around. This leads to the concept of degree function of a continuous
f : (I, I˙) → (S 1 , 1).
We compute π1 (S 1 ) with the concept of degree function. We use a map p : R →
S , called the exponential map, defined by p(t) = e2πit , ∀ t ∈ R as shown in
1
Fig. 3.8. Then p is a continuous onto map which wraps the real line R onto the circle
infinite number of times. p is a group homomorphism from (R, +) to (S 1 , 1) with
ker p = {t ∈ R : p(t) = 1} = {t ∈ R : e2πit = 1} = Z.
Fig. 3.8 Projection of the

real line on a circle in a 2
complex plane
1
p
R 0
1
−1
S1
−2
3.3 Degree Function and the Fundamental Group of the Circle 127
Fig. 3.9 Lifting of a map (R, r0 )

w;
w
w
f˜ w
w p
w
w
w
w
(X, x0 ) / (S 1 , 1)
f
Definition 3.3.1 A continuous map f : (X, x0 ) → (S 1 , 1) is said to have a lifting

f˜ : (X, x0 ) → (R, r0 ), where r0 ∈ ker p = Z if there exists a continuous map f˜ :
(X, x0 ) → (R, r0 ) such that p ◦ f˜ = f , i.e., the triangle in Fig. 3.9 is commutative.
We now show that under suitable conditions, f˜ is unique.
Proposition 3.3.2 Let X be a connected space and f˜, g̃ are two liftings of f :
(X, x0 ) → (S 1 , 1). If f˜(x 0 ) = g̃(x0 ), then f˜ = g̃.
Proof Using the group structure of (R, +), we define a map h : (X, x 0 ) → (R, r0 )
by h(x) = f˜(x) − g̃(x). Now, for any x ∈ X , ( p ◦ h)(x) = p(h(x)) = e2πi h(x) =
˜ ˜
e2πi( f (x)−g̃(x)) = e2πi f (x) /e2πi g̃(x) = ( p ◦ f˜)(x)/( p ◦ g̃)(x) = f (x)/ f (x) = 1 =⇒
h(x) ∈ ker p = Z. Therefore h : X → Z is integral valued. Since X is connected
and h is continuous, then it follows from discreteness of Z that h is constant =⇒
image set h(X ) must be singleton. But by hypothesis, h(x0 ) = f˜(x0 ) − g̃(x) =
0 =⇒ h(x) = 0, ∀ x ∈ X =⇒ f˜(x) = g̃(x), ∀ x ∈ X =⇒ f˜ = g̃. ❑
We now show that any path in S 1 starting at 1 can be lifted to a unique path in R
starting at the origin 0 of R and any homotopy between two given paths in S1 starting
at 1, can be lifted to a unique homotopy between the two lifted paths starting at the
origin 0 of R. These two results follow as corollaries of the following theorem.
Theorem 3.3.3 Let X be a compact convex subset of the Euclidean space R n for
some n. Let f (X, x0 ) → (S 1 , 1) be continuous, and z 0 ∈ Z. Then ∃ a unique
continuous map f˜ : (X, x0 ) → (R, z 0 ) with p ◦ f˜ = f i.e., making the diagram in
Fig. 3.10 commutative.
Proof Since X is a compact metric space, f must be uniformly continuous. Hence

there is an > 0 such that whenever ||x −x || < , then | f (x)− f (x )| < 2. Here ||; ||
is Euclidean norm of Rn and | f (x)| denotes the modulus of the complex number
f (x). We choose 2 = diam S 1 to ensure that f (x) and f (x ) are not antipodal
points i.e., f (x)/ f (x ) = −1. Since X is bounded, ∃ a positive integer n such that
Fig. 3.10 Lifting of f in R (R, z0 )

w;
w
w
f˜ w
w p
w
w
w
w
(X, x0 ) / (S 1 , 1)
f
||x − x0 ||/n < , ∀ x ∈ X . For each x ∈ X , we subdivide the line segment having
end points x0 and x (which entirely lies in X by convexity) into n intervals of equal
length by inserting (uniquely determined) points x0 , x 1 , . . . , xn = x. Then
||x j+1 − x j || = ||x − x 0 ||/n < =⇒ f (x j+1 )/ f (x j ) = −1.
Now for each j with 0 ≤ j ≤ n − 1, the function g j : X → S 1 − {−1} defined by

g j (x) = f (x j+1 )/ f (x j ) is continuous, because multiplication map S 1 × S 1 → S 1
and inversion map S 1 → S 1 are continuous. Then for all x ∈ X , we can write (by
taking j = 0, 1, 2, . . . , n − 1 successively)
f (x 0 )g0 (x) = f (x 1 ), f (x1 )g1 (x) = f (x 2 ), . . . , f (xn−1 )gn−1 (x) = f (x)
and hence f (x) = f (x0 )g0 (x)g1 (x) . . . gn−1 (x) (called telescoping product in S 1 ).
Define a map f˜ : X → R by
f˜(x) = z 0 + log g0 (x) + log g1 (x) + · · · + log gn−1 (x).
Then f˜ is the sum of n continuous functions and hence it is continuous. Moreover,

g j (x0 ) = 1, ∀ j. Consequently, f˜(x0 ) = z 0 and p ◦ f˜ = f . ❑
Corollary 3.3.4 Let f : (I, I˙) → (S 1 , 1) be continuous.

(i) Then there exists a unique continuous map f˜ : I → R with p ◦ f˜ = f and
f˜(0) = 0 (Path lifting Property) (Fig. 3.11).
(ii) If g : (I, I˙) → (S 1 , 1) is continuous and f g rel I˙, then f˜ g̃ rel I˙ (where
p ◦ g̃ = g and g̃(0) = 0). Moreover, f˜(1) = g̃(1). (Homotopy Lifting Property).
Proof (i) follows from the Theorem 3.3.3 by taking in particular X = I = [0, 1] ⊂
R, and Proposition 3.3.2.
(ii) I × I is compact convex. We choose (0, 0) as a base point of I × I . Let F :
f g rel I˙. Then Theorem 3.3.3 gives a continuous map F̃ : I × I → R such
that p ◦ F̃ = F and F̃(0, 0) = 0. We show that F̃ : f˜ g̃ rel I˙ i.e., F can
be lifted. Let ψ0 : I → R be defined by ψ0 (t) = F̃(t, 0). Then p ◦ ψ0 (t) =
p ◦ F̃(t, 0) = F(t, 0) = f (t). Since ψ0 (0) = F̃(0, 0) = 0, uniqueness of
lifting shows that ψ0 = f˜. Again define φ0 : I → R by φ0 (t) = F̃(0, t). Then
proceeding as above, we show that φ0 is the constant function φ0 (t) = 0. Hence
it follows that F̃(0, 1) = 0. Again define ψ1 : I → R by ψ1 (t) = F̃(t, 1). Then
Fig. 3.11 Path lifting of f (R, 0)

<
yy
yy
f˜ yy
yy
yy p
yy
yyy
y
yy
(I, 0) / (S 1 , 1)
f
3.3 Degree Function and the Fundamental Group of the Circle 129
p ◦ ψ1 (t) = F(t, 1) = g(t) and ψ1 (0) = F̃(0, 1) = 0. Hence ψ1 = g̃. Finally,

define φ1 : I → R by φ1 (t) = F̃(1, t). Now p◦φ1 is the constant function c with
value f (1), and φ1 (0) = f˜(1). Hence the constant function at f˜(1) is a lifting
of c, and uniqueness of lifting shows that φ1 (t) ≡ f˜(1), ∀ t ∈ I . Consequently,
g̃(1) = f˜(1) and F̃ : f˜ g̃ rel I˙. ❑
Corollary 3.3.5 Let f, g : (I, I˙) → (S 1 , 1) be two continuous functions such that
f g rel I˙. If w( f ) denotes the winding number of f , then w( f ) = w(g).
Proof w( f ) = f˜(1) − f˜(0) = g̃(1) − g̃(0) (by Corollary 3.3.4) = w(g). ❑
Definition 3.3.6 Let f : (I, I˙) → (S 1 , 1) be a continuous map. The degree of f

denoted deg f is defined by deg f = f˜(1), where f˜ is the unique lifting of f with
f˜(0) = 0.
Remark 3.3.7 It shows by Corollary 3.3.4 that degree of circular maps is an example
of a homotopy invariant.
Remark 3.3.8 deg f is an integer. Because
p ◦ f˜ = f =⇒ ( p ◦ f˜)(1) = f (1) = 1 =⇒ f˜(1)
lies in the ker p = Z =⇒ deg f ∈ Z for every f : (I, I˙) → (S 1 , 1). If f (z) = z m
i.e., if f (t) = p(mt) = e2πimt , then f˜(1) = m. This explains the term degree.
Theorem 3.3.9 The degree function d : π1 (S 1 , 1) → Z, [ f ] → deg f is an iso-

morphism of groups.
Proof By Corollary 3.3.4(ii), d is independent of the choice of the representatives

of the classes [ f ] ∈ π1 (S 1 , 1) and hence the function d is well defined.
d is a homomorphism: Let [ f ], [g] be any two elements of π1 (S 1 , 1) such that
deg f = m and deg g = n. To compute the deg( f ∗ g) we define a path h̃ : I → R
such that p ◦ h̃ = f ∗ g and h̃(0) = 0. Then deg( f ∗ g) = h̃(1).
Let g̃ be the unique lifting of g such that g̃(0) = 0. Define γ̃ : I → R by
γ̃(t) = m + g̃(t). Then γ̃ is a path in R from m to m + n. Again let f˜ be the
lifting of f such that f˜(0) = 0 and f˜(1) = m. Then f˜ ∗ γ̃ is a path in R such that
( f˜ ∗ γ̃)(0) = 0 and ( f˜ ∗ γ̃)(1) = m + n, because f˜ ∗ γ̃ : I → R is defined by

f˜(2t), 0 ≤ t ≤ 1/2
( f˜ ∗ γ̃)(t) =
γ̃(2t − 1), 1/2 ≤ t ≤ 1
We claim that f˜ ∗ γ̃ is the lifting of f ∗ g:

( p ◦ f˜)(2t), 0 ≤ t ≤ 1/2
( p ◦ ( f˜ ∗ g̃))(t) =
( p ◦ γ̃)(2t − 1), 1/2 ≤ t ≤ 1
Now ( p◦ f˜)(t) = f (t), ∀ t ∈ I and ( p◦ γ̃)(t) = p(m + g̃(t)) = e2πim . p(g̃(t)) =

g(t), ∀ t ∈ I =⇒ p◦( f˜∗γ̃) = f ∗g =⇒ deg( f ∗g) = ( f˜∗γ̃)(1) = m+n =⇒ d
is a homomorphism.
d is a monomorphism: Let [ f ] ∈ ker d. Then deg f = 0 =⇒ f˜(1) = 0 =⇒ f˜
is a closed path in R at 0, since f˜(0) = 0. The continuous map p : (R, 0) → (S 1 , 1)
induces a homomorphism p∗ : π1 (R, 0) → π1 (S 1 , 1), defined by p∗ ([ f˜]) = [ p ◦
f˜] = [ f ]. But R is contractible and hence π1 (R, 0) = 0 =⇒ [ f˜] = [c] =⇒ [ f ]
is the identity element of π1 (S 1 , 1) =⇒ ker d is trivial =⇒ d is a monomorphism.
d is an epimorphism: Let n ∈ Z. Define a loop f : I → S 1 by f (t) = e2πint . Then
the path f˜ : I → R defined by f˜(t) = nt, starts at the origin O of R and lifts the
path f . Then deg f = f˜(1) = n =⇒ d is an epimorphism. Consequently, d is an
isomorphism. ❑
Corollary 3.3.10 π1 (S 1 , s) is isomorphic to Z for any s ∈ S 1 .
Proof For any s ∈ S 1 , π1 (S 1 , s) is isomorphic to π1 (S 1 , 1), since S 1 is path-

connected. Hence the Corollary follows. ❑
Definition 3.3.11 A topological space X is called simply connected if it is path-

connected and for every x0 ∈ X , π(X, x0 ) = 0.
Corollary 3.3.12 S 1 is not simply connected.
Proof S 1 is path-connected but π1 (S 1 , s) = 0. Hence S 1 is not simply connected.❑
Remark 3.3.13 The homotopy classes of loops on S 1 based at 1 can be completely

characterized with the help of their degrees.
Theorem 3.3.14 Let f, g : (I, I˙) → (S 1 , 1) be continuous. Then f g rel I˙ if and

only if deg f = deg g.
Proof f g rel I˙ =⇒ f˜ g̃ r el I˙ with f˜(1) = g̃(1) =⇒ deg f = deg g.

Conversely, deg f = deg g =⇒ [ f ] = [g], because the degree function d is
injective =⇒ f g rel I˙. ❑
3.4 The Fundamental Group of the Punctured Plane
This section studies punctured Euclidean plane from homotopy view-point and com-
putes its the fundamental group. It is an important space in geometry and topology.
We now proceed to calculate π1 (R2 −{0}), the fundamental group of the punctured
plane.
Theorem 3.4.1 Let s0 ∈ S 1 . The inclusion map i : (S 1 , s0 ) → (R2 −{0}, s0 ) induces

an isomorphism. i ∗ : π1 (S 1 , s0 ) → π1 (R2 − {0}, s0 ).
3.4 The Fundamental Group of the Punctured Plane 131
Proof r : R2 − {0} → S 1 be the continuous map defined by r (x) = ||x|| x

. The map
r can be depicted as collapsing each radial ray in R − {0} onto the point where the
2
ray intersects S 1 . In particular, it maps each point x ∈ S 1 to itself.

We claim that the induced homomorphism r∗ is the inverse of the induced homo-
morphism i ∗ . We consider the composite map
i r
(S 1 , s0 ) −
→ (R2 − {0}, s0 ) −
→ (S 1 , s0 ).
Then r ◦ i = 1 S 1 =⇒ r∗ ◦ i ∗ = identity automorphism of π1 (S 1 , s0 ). On the other

hand, we also show that i ∗ ◦ r∗ is the identity automorphism of π1 (R2 − {0}, s0 ).
Let f be a loop in R2 − {0} based at s0 . Then (i ∗ ◦ r∗ )([ f ]) = [i ◦ r ◦ f ] where
i ◦ r ◦ f = g : I → R2 − {0} is a loop given by the equation
f (t) f (t)
g(t) = (i ◦ r )( f (t)) = i( )= .
|| f (t)|| || f (t)||
Define F : I × I → R2 − {0} by the equation F(t, s) = s || ff (t)

(t)||
+ (1 − s) f (t). Then
F(t, s) = 0, ∀ (t, s) ∈ I × I . Otherwise, s || ff (t)
(t)||
+ (1 − s) f (t) = 0 would imply
s
f (t)[ + (1 − s)] = 0 =⇒ s + (1 − s)|| f (t)|| = 0,
|| f (t)||
since f (t) = 0, which is not true, since 0 ≤ s ≤ 1.

Then F : f g rel I˙.
Consequently, (i∗ ◦ r∗ )([ f ]) = [i ◦ r ◦ f ] = [g] = [ f ], ∀ [ f ] ∈ π1 (R2 −
{0}, s0 ) =⇒ i ∗ ◦ r∗ is the identity automorphism of π1 (R2 − {0}, s0 ).
Thus we conclude that i ∗ is an isomorphism. ❑
Corollary 3.4.2 S 1 is a strong deformation retract of R2 − {0} and hence π1 (S 1 ) ∼
=
π1 (R2 − {0}) ∼
= Z.
Proof Let i : S 1 → R2 − {0} = X (say). Define a map F : X × I → X by the rule
F(x, t) = (1 − t)x + t ||x||
x
.
Then F(x, t) = 0 for any (x, t) ∈ X × I . Clearly, F is a continuous map such
that F(x, 0) = x, ∀ x ∈ X, F(x, 1) = ||x||
x
∈ S 1 , ∀ x ∈ X, F(a, t) = a, ∀ a ∈ S 1 .
Consequently, S 1 is a strong deformation retract of R2 − {0}. Thus π1 (S 1 ) ∼
=
π1 (R2 − {0}) ∼
= Z. ❑
3.5 Fundamental Groups of the Torus
This section considers the torus and computes its fundamental group. A torus is a
connected 2-manifold homeomorphic to the product of two circles S 1 × S 1 . Surfaces
are very important in geometry, topology, and complex analysis. Recall that a surface
is a Hausdorff space with a countable basis, every point of which has a neighborhood
which is homeomorphic to an open disk in R2 . We shall consider other familiar
surfaces: the sphere S 2 , the projective plane R P 2 , double torus etc. in the next
chapter. The simplest example of a compact surface is the 2-sphere S 2 . We now
consider another example which is the torus. A torus may be considered as any
surface homeomorphic to the surface of a doughnut or a solid ring.
Theorem 3.5.1 The fundamental group of the torus T = S 1 × S 1 is isomorphic to

the group Z × Z.
Proof It follows from Theorems 3.1.37 and 3.3.9. ❑
3.6 Vector Fields and Fixed Points
This section studies vector fields on D 2 and applies the fundamental group to the
following 2 problems:
(a) The existence of vector fields tangent to given surfaces.
(b) Given a topological space X does every continuous map f : X → X necessarily
has a fixed point?
Moreover this section presents a proof of Brouwer fixed point theorem for D 2
using the concept of vector field.
Definition 3.6.1 A vector field on D 2 is an ordered pair (x, v(x)), where x ∈ D 2

and v : D 2 → R2 is a continuous map.
We say that a vector field is nonvanishing if v(x) = 0 for every x ∈ D 2 ; in such

a case, we may consider v : D 2 → R2 − {0} mapping D 2 into R2 − {0}.
Theorem 3.6.2 Given a nonvanishing vector field on D 2 , there exists a point of S 1

where the vector field points directly inward and a point of S 1 where it points directly
outward.
Proof First we show that given a vector field v, ∃ a point of S 1 where v points
directly inward. We consider the map u : S 1 → R1 − {0} obtained by restriction
of v to S 1 . If the assertion is not true, ∃ no point x ∈ S 1 at which v points directly
inward. In other words, ∃ no x ∈ S 1 such that u(x) is a negative multiple of x. It
follows that u is homotopic to the inclusion map i : S 1 → R2 − {0}, under the
homotopy F : S 1 × I → R2 − {0} given by the equation F(x, t) = t x + (1 − t)u(x).
Clearly, F(x, t) = 0, otherwise (1 − t)u(x) = −t x. This is not true for t = 0 or
t = 1, since x ∈ S 1 and u(x) = 0. Again for 0 < t < 1, we have from above,
u(x) = −t x/(1 − t) =⇒ u(x) is a negative multiple of x, which is not true by
assumption. Since u is homotopic to the inclusion map i : S 1 → R2 − {0}, it follows
that u is extendable to the continuous map v : D 2 → R2 − {0}. Hence we reach a
contradiction. Thus we conclude that v points directly inward at some point of S 1 .
3.6 Vector Fields and Fixed Points 133
For the second part, we consider the non-vanishing vector field (x, −v(x)). By the
first part, it points directly inward at some point of S 1 . Then v points directly outward
at that point. ❑
Remark 3.6.3 For nonvanishing continuous vector fields on S n see Chap. 14.
We present a proof of Brouwer Fixed Point Theorem for D 2 by using the concept
of vector field. For an alternative proof see Theorem 3.8.8.
Theorem 3.6.4 (Brouwer Fixed Point Theorem for the disk D 2 ) If f : D 2 → D 2
is continuous, then there exists a point x ∈ D 2 such that f (x) = x.
Proof Suppose f (x) = x for every x ∈ D 2 . Then the map v defined by v(x) =
f (x) − x gives a non-vanishing vector field (x, v(x)) on D2 . But the vector field v
cannot point directly outward at any point x ∈ S 1 , otherwise f (x) − x = ax for
some real a > 0. Then f (x) = (1 + a)x lies outside the unit ball D 2 . Thus we reach
a contradiction. ❑
Corollary 3.6.5 Let M be a 3 × 3 matrix of positive real numbers. Then M has a
positive real eigenvalue.
Proof Let T : R3 → R3 be the linear transformation whose matrix representation is
M relative to the standard basis for R3 . Let O1 = {(x 1 , x 2 , x3 ) ∈ R3 : x1 ≥ 0, x2 ≥ 0,
x3 ≥ 0}, be the first octant of R3 and B = S 2 ∩ O1 . Then B is homeomorphic to
the ball D 2 . Therefore the Brouwer fixed point theorem holds for continuous maps
of B into itself. Now if x = (x1 , x2 , x3 ) ∈ B, then all the components of x are
non-negative and at least one is positive. Since all the entries of M are positive, the
vector T (x) is a vector all of whose components are positive. Consequently, the map
B → B, x → ||TT (x)
(x)||
is continuous and hence it has a fixed point x 0 (say). Therefore,
(x0 )
x0 = ||TT (x 0 )||
shows that T (x0 ) = ||T (x 0 )||x0 . This implies that T has the positive
real eigenvalue ||T (x0 )|| and hence the matrix M has a positive real eigenvalue. ❑
3.7 Knot and Knot Groups
This section conveys a study of knots, which returns to geometry and considers
various ways of embeddings the circle as a subspace of R3 or S3 . The scientists
working in physics and biochemistry find interesting applications of knot theory.
H. Tietze (1880–1964) contributed to the the foundations of knot theory. Fundamental
groups play an important role in the study of knot theory.
Definition 3.7.1 A knot K is a subspace of Euclidean 3-space R3 which is homeo-
morphic to the circle and knot group of K is the fundamental group π1 (R3 − K ) of
the complement of K in R3 .
Remark 3.7.2 Properties of complement of the knot K in R3 are significant, because
it is how the knot is embedded in R3 is crucial.
Definition 3.7.3 The standard embedding K = S 1 ⊂ R2 ⊂ R3 ⊂ S 3 is called the

trivial knot (or unknot).
Remark 3.7.4 A knot is represented by its projection in the plane of the paper. Hence
‘trivial knot’ or ‘unknot’ consists of the unit circle in the x y-plane. Two knots are
said to be same if there exists a homeomorphism of R3 which sends one knot onto
the other knot.
Definition 3.7.5 Two knots K 1 , K 2 are said to be equivalent if there is a homeomor-
phism h : R3 → R3 such that h(K 1 ) = K 2 .
Remark 3.7.6 Two knots K 1 and K 2 are equivalent if one can be continuously
deformed into other.
Definition 3.7.7 A knot K is said to be polygonal if it is made up of a finite number
of line segments. A tame knot is a knot equivalent to a polygonal knot.
For picturing knots and work with them effectively, it needs projecting them into
the plane in a ‘nice’ way in the sense that the projection only crosses itself at a finite
number of points, at most two pieces of the knot meet at such crossing, and they
do so at ‘right angles’ as shown in Figs. 3.12, 3.13 and 3.14. These knots are called
‘trefoil knot’, ‘figure-eight knot’, and ‘square knot’ respectively.
Another important knot called ‘torus knot’ is now described. For any relatively
prime pair of positive integers ( p, q), the image of 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. This is called a torus knot of type ( p, q).
Definition 3.7.8 For relatively prime integers p and q, the torus knot K = K p,q ⊂
R3 is the image of the embedding f : S 1 → S 1 × S 1 ⊂ R3 , z → (z p , z q ), where
the torus S 1 × S 1 is embedded in R3 in the usual way.
Fig. 3.12 Trefoil knot
Fig. 3.13 Figure-eight knot
Fig. 3.14 Square knot

3.7 Knot and Knot Groups 135
To make the above map f injective, the integers p and q are assumed to be
relatively prime.
Remark 3.7.9 Geometrically, 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.
S 3 is considered as the one-point compactification of R3 .
Proposition 3.7.10 Let K be any knot. If R3 − K and S 3 − K are the complements
of K in R3 and 3-sphere S 3 respectively, then the inclusion map i : (R3 − K ) →
(S 3 − K ) induces an isomorphism
i ∗ : π1 (R3 − K ) → π1 (S 3 − K ).
Proof Let K be any knot. Then K is a compact subset of R3 . Moreover, S 3 − K

is the union of R3 − K and an open ball B formed by the compactification point
together with the complement of a large closed ball in R3 containing K . Both B and
B ∩ (R3 − K ) are simply connected. As B ∩ (R3 − K ) is homeomorphic to S 2 × R,
van Kampen theorem (see Chap. 6) implies that the inclusion R3 − K → S 3 − K
induces an isomorphism. Hence if i : (R3 − K ) → (S 3 − K ) is the inclusion map,
then the induced homomorphism
i ∗ : π1 (R3 − K ) → π1 (S 3 − K ).
is an isomorphism. ❑
Remark 3.7.11 The simplest knot is a circle which we may think of as the unit circle
in the x1 x2 -plane. Its knot group is Z.
Definition 3.7.12 A homeomorphism h : R3 → R3 is said to be isotopic to the
identity if there is a homotopy H : R3 × I → R3 such that the map
Ht : R3 → R3 , x → H (x, t)
is a homeomorphism with H0 = 1d and H1 = h.

Remark 3.7.13 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 3.7.14 A homeomorphism h : R3 → R 3 is said to be orientation pre-
serving (or orientation reversing) if its extension homeomorphism h̃ : S 3 → S 3
preserves (or reverses) the orientation of S 3 .
Example 3.7.15 A homeomorphism which is isotopic to the identity is orientation
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 3.7.16 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 3.7.17 Any orientaion preserving homeomorphism of R3 is isotopic to the
identity.
Proposition 3.7.18 Equivalent knots have homeomorphic complements in R3 .
Proof Let K 1 and K 2 be equivalent knots. Then there is a homeomorphism h : R3 →
R3 such that h(K 1 ) = h(K 2 ). Restricting h to R3 − K 1 defines a homeomorphism
h : R3 − K 1 → R3 − K 2 . This shows that equivalent knots have homeomorphic
complements in R 3 . ❑
Remark 3.7.19 As the knot group π1 (R3 − K ) of K is the complement of K in R3 ,
it can be used to distinguish various knots.
Definition 3.7.20 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 3.7.21 Circle knot is a trivial knot. If a knot K is trivial, then the fundamental
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 can not be a trivial knot which means that K can
not be untied.
Remark 3.7.22 For some sort of reasonable presentation for a knot group in terms
of generators and relations the book (Armstrong 1983) is referred.
Remark 3.7.23 There is an interesting link between a knot and an Eilenberg–
Maclane space: see Chap. 11.
3.8 Applications
This section applies fundamental group and degree function to prove some important
theorems such as fundamental theorem of algebra, Brouwer fixed point theorem
for dimension 2, and Borsuk–Ulam theorem. Finally this section applies winding
numbers and exponential map p : R → S 1 , t → e2πit to prove Cauchy integral
theorem of complex analysis.
3.8.1 Fundamental Theorem of Algebra
This subsection provides an alternative proof of fundamental theorem of algebra

given in Theorem 2.10.10. Although algebraic topology is usually ‘algebra serving
topology,’ the role is reversed in the following proof of fundamental theorem of

algebra. The concept of degree function is now utilized to prove this theorem of
algebra.
Theorem 3.8.1 (Fundamental Theorem of Algebra) Every nonconstant polynomial
with coefficients in C has a root in C.
Proof It is sufficient to prove the theorem for a polynomial of the form: p(z) =
a0 + a1 z + · · · + an z n−1 + z n , a0 = 0, n ≥ 1 over C. Suppose the theorem is not true.
Then z → p(z) is a mapping from the complex plane C to C − {0}. The restriction
of this mapping to the different circles: |z| = r for different values of r ≥ 0 are loops
in C − {0}.
Consider the mappings G : I × [0, ∞) → S 1 ⊂ C defined by
p(r e2πit )/ p(r )

G(t, r ) =
| p(r e2πit )/ p(r )|
and F : I × I → S 1 ⊂ C defined by

G(t, s/(1 − s)), 0 ≤ t ≤ 1, 0 ≤ s < 1
F(t, s) =
e2πint , 0 ≤ t ≤ 1, s = 1.
Since G is continuous, lim F(t, s) = lim G(t, s/(1 − s)) = lim G(t, r ) = e2πint
s→1 s→1 r →∞
and hence F is continuous. If F(t, 0) = f 0 (t) and F(t, 1) = f 1 (t), then F : f 0
f 1 rel I˙. Consequently, deg f 0 = deg f 1 . But deg f 0 = 0 and deg f 1 = n. This
implies a contradiction, since n ≥ 1. Thus we conclude that p(z) has a root in C. ❑
Definition 3.8.2 A field F is said to be algebraically closed (or complete) if every
polynomial f (x) over F with degree ≥ 1, has a root in F.
Corollary 3.8.3 The field C of complex numbers is algebraically closed.
Proof It follows from Theorem 3.8.1. ❑
Corollary 3.8.4 The field R of real numbers is embedded in the algebraically closed
field C.
Proof It follows from Corollary 3.8.3. ❑
Remark 3.8.5 The Corollary 3.8.3 proves the algebraic completeness of the field of
complex numbers.
3.8.2 An Alternative Proof of Brouwer Fixed Point Theorem
We now prescribe an alternative proof of Brouwer Fixed Point Theorem given in

Theorem 3.6.4 for dimension 2.
Theorem 3.8.6 The circle S 1 is not a retract of the disk D 2 .
Proof Suppose there is a retraction r : D 2 → S 1 . Then r ◦ i = 1 S 1 , where i :

S 1 → D 2 is the inclusion map. This implies that the composite homomorphism
i∗ r∗
π1 (S 1 , 1) −→ π1 (D 2 , 1) −→ π1 (S 1 , 1) is the identity automorphism by Theorem
3.1.23. But this is not possible, since π1 (D 2 , 1) = 0 and π1 (S 1 , 1) = Z. ❑
Corollary 3.8.7 The identity map 1 S1 : S 1 → S 1 cannot be continuously extended

over D 2 .
Proof Suppose 1 S 1 is continuously extendable over D n . Then there must exist a

continuous map r : D 2 → S 1 such that r | S 1 = 1 S 1 . In other words, r ◦ i = 1 S1 .
Consequently, S 1 is a retract of D 2 . This contradicts Theorem 3.8.6. ❑
Theorem 3.8.8 (Brouwer Fixed Point Theorem for dimension 2) Any continuous
map f : D 2 → D 2 has a fixed point, that is, there exists a point x ∈ D 2 such that
f (x) = x.
Proof Suppose to the contrary that f (x) = x for any x ∈ D 2 . We can define a
map r : D 2 → S 1 by letting r (x) to the point of S 1 , where the ray in R2 starting
from f (x) and passing through x meets S 1 . This is well defined, since the ray meets
S 1 at exactly one point which we call r (x). This means there is a t > 0 such that
x = (1 − t) f (x) + tr (x) =⇒ r (x) = (x−(1−t) t
f (x))
=⇒ r is a continuous function
of x. Clearly, ∀ x ∈ S , r (x) = x =⇒ r is a retraction from D 2 to S 1 =⇒ S 1
1
is a retract of D 2 . This contradicts Theorem 3.8.6. Hence we conclude that f has a

fixed point. ❑
Remark 3.8.9 Brouwer fixed point theorem for D n was first proved and studied by
L.E.J. Brouwer (1881–1967) during 1910–2012. Now this Theorem is proved by
using the homology or homotopy groups. But Brouwer used neither of them, which
had not been invented at that time. Instead, he used the notion of degree of spherical
maps f : S n → S n .
Remark 3.8.10 A generalization of the Corollary 3.6.5 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 3.8.11 If x is a point of S n , then its antipode is the point −x ∈ S n . A

continuous map f : S n → S m is said to be antipode-preserving or antipodal if
f (−x) = − f (x), ∀ x ∈ S n .
Proposition 3.8.12 Let f n : S n → S n be the antipodal map. If there exists a vector

field on S n , then f n 1d .
Proof Let f : S n → Rn+1 be a map such that f (x) = 0 and f (x), x = 0.

Construct a path from x to f n (x) in the plane determined by x and f (x) and on the
sphere:
H (x, t) = α(t)x + β(t) f (x), ||H (x, t)||2 = 1.
This gives the equation
α(t)2 + β(t)2 f (x), f (x) = 1.
Choose α(t) √ = 1 − 2t. Then β(t) = 4(t − t )/|| f (x)|| , since f (x)
2 2 2
√ = 0, gives
β(t) = 2 t − t /|| f (x)||. Consequently, H (x, t) = (1 − 2t)x + 2 t − t 2 f (x)/
2
|| f (x)|| shows that H : 1d f n . ❑
Proposition 3.8.13 If f : S 1 → S 1 is nullhomotopic, then

(i) f has a fixed point;
(ii) f maps some point x ∈ S 1 to its antipode −x.
Proof (i) f : S 1 → S 1 is nullhomotopic =⇒ f has a continuous extension f˜

over D 2 . Consider the map f˜ : D 2 → S 1 ⊂ D 2 . Then by Brouwer Fixed Point
Theorem, f˜ has a fixed point x0 (say). Now Im f˜ ⊂ S 1 and f˜(x 0 ) = x 0 =⇒
x 0 ∈ S 1 , which is clearly a fixed point of f .
(ii) Define a map g : S 1 → S 1 , x → −x. Then g ◦ f : S 1 → S 1 is also null-
homotopic. Hence g ◦ f has a fixed point x 0 (say) by (i). This implies that
x0 = (g ◦ f )(x0 ) = g( f (x 0 )) = − f (x 0 ) =⇒ f (x0 ) = −x0 . ❑
Corollary 3.8.14 There is no antipode-preserving continuous map f : S 2 → S 1 .
3.8.3 Borsuk–Ulam Theorem
We now prove the Borsuk–Ulam theorem for dimension 2. For general case see
Chap. 14.
Theorem 3.8.15 (Borsuk–Ulam Theorem) For every continuous map f : S 2 → R2 ,

there exists a pair of antipodal points x and −x in S 2 such that f (x) = f (−x).
Proof Suppose f (x) = f (−x) for all x ∈ S 2 . Define
f (x) − f (−x)
g : S 2 → S 1 , x → .
|| f (x) − f (−x)||
Then g is a continuous map such that g(−x) = −g(x), ∀ x ∈ S 2 . This contradicts

the Corollary 3.8.14. ❑
Corollary 3.8.16 The 2-sphere S 2 can not be put in the plane R2 .
Proof If possible, let f : S 2 → R2 be an embedding. Then it is continuous and

hence f must map at least one pair of points to a single point of R2 by Theorem
3.8.15. This implies a contradiction, because f cannot be injective. ❑
Example 3.8.17 (Physical interpretation) Consider the earth as 2-sphere S 2 . Let T, P

are 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)).
The the Borsuk–Ulam Theorem 3.8.15 says that 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.
3.8.4 Cauchy’s Integral Theorem of Complex Analysis
This subsection proves Cauchy’s integral theorem (homotopy version) of complex

analysis. This theorem is one of the central theorems in the study of functions of a
complex variable for analytic functions. We now utilize winding number w( f ; a) of
a differentiable loop parametrized by f in the complex plane with respect to a point
a not in Im f (see Corollary 3.3.5) and the exponential map p : R → S 1 to study
complex line integral.
Lemma 3.8.18 Let f be a piecewise differentiable loop in the complex plane and a
be a point in C but not in Im f . Then

1 dz
w( f ; a) = .
2πi f z−a
Proof Let p : R → S 1 be the usual covering map. Define
f (t) − a
g : I → S 1 , t → .
|| f (t) − a||
Then g is a loop in S 1 . Let g̃ : I → R be the unique lifting of g. Define θ(t) by

θ(t) = 2π g̃(t). Then f (t)−a = || f (t)−a||eiθ(t) . Then f z−a
dz
= 2πi[g̃(1)− g̃(0)] =
2πi.w( f ; a). ❑
Theorem 3.8.19 (Cauchy’s Integral Theorem) Let X be an open subset of the com-
plex plane C and f : X → C be an analytic function. If α is a simple closed piecewise
differentiable curve in X ⊂ C such that α is nullhomotopic, then α f = 0.
Proof Let α1 = α and α0 be a constant curve such that α1 α0 . Let F : α1 α0 .

Define h(t) = w(αt ; β), where αt (s) = F(s, t) for 0 ≤ s, t ≤ 1 and β is fixed in
C − X . We claim that h is continuous on I and since h is an integral valued function
and h(0) = 0, it follows that h(t) ≡ 0. In particular, w(α; β) = 0 for all β in C − X .
Hence the theorem follows from Lemma 3.8.18. ❑
Remark 3.8.20 Theorem 3.8.19 follows immediately from Corollary 3.3.5 and
Lemma 3.8.18, since α is nullhomotopic.
3.9 Exercises
1. Let {X i : i ∈ α} be a family of spaces and let for each i ∈ α, xi ∈ X i be a

base
point. Then generalize the Theorem 3.1.37 to prove that π 1 ( X i , (xi )) ∼
=
π1 (X i , xi ).
2. Show that Rn − {0} is simply connected for n > 2.
[Hint: Rn − {0} ≈ S n−1 =⇒ π1 (Rn − {0}) ∼ = π1 (S n−1 ) for n ≥ 3 =⇒
π1 (R − {0}) = 0, ∀ n > 2.]
n
3. Prove that Rn and R2 are not homeomorphic for n > 2.

[Hint: Suppose n > 2. Then deleting a point p from Rn leave Rn − p simply
connected. On the other hand deleting a point from R 2 does not so.]
4. (a) If s0 ∈ S n−1 , show that the inclusion map i : (S n−1 , s0 ) → (Rn − {0}, s0 )
induces an isomorphism of fundamental groups.
[Hint: Proceed as in Theorem 3.4.1.]
(b) Show that the map H : (R n − {0}) → Rn − {0} defined by H (x, t) =
x
t ||x|| + (1 − t)x is a strong deformation of Rn − {0} onto S n−1 .
5. Let A denote the z-axis in R3 . Show that π1 (R3 − A) ∼ = Z.
6. Show that none of the following subspaces A of X are retract of X :
(i) X = S 1 × D 2 and A is its boundary torus S 1 × S 1 .
(ii) X = R3 and A is any subspace of X homeomorphic to S 1 .
[Hint: (i) Using the results that π1 (S 1 ) ∼ = Z, π1 (D 2 ) ∼ = 0, we get
2 ∼ 1 ∼
π1 (S × D ) = Z and π1 (S × S ) = Z × Z, it follows that for any
1 1
homomorphism ψ : Z × Z → Z, ψ((1, 0)) = n and ψ((0, 1)) = m for

some integer m and n. Clearly ψ((m, 0)) = nm and ψ((0, n)) = nm imply
that ψ is not injective. Consequently, there exists no retraction of X to A by
Proposition 3.1.26.
(ii) Use the results that π1 (R3 ) ∼= 0 and π1 (A) ∼
= π1 (S 1 ) ∼
= Z to show that
3
there does not exists any retraction of R to any subspace A homeomorphic
to S 1 .]
7. Show that there is no antipodal map f : S n → S 1 for n > 1.
8. Show that the n-sphere S n has a nonvanishing tangent vector field if and only if
n is odd.(The Brouwer- Poincaré Theorem).
[Hint: If n is odd, then n = 2m + 1. A vector field f on S n can be defined
by f (x 1 , x2 , . . . , x2m+2 ) = (x2 , −x1 , x4 , −x3 , . . . , x2m+2 , −x2m+1 ), ∀ (x1 ,

x 2 , . . . , x2m+2 ) ∈ S n . Then f is a continuous function from S n to S n . For each
x ∈ S n , x, f (x) = (x1 x2 −x 1 x2 )+(x3 x4 −x 3 x4 )+· · ·+(xn xn+1 −x n xn+1 ) = 0.
If n is even, no such vector field exists. See Chap. 11.]
9. Show that the sphere S 2 has no nonvanishing tangent vector field.
10. (a) Let X be a path-connected space. Show that the fundamental group of π1 (X )
based at any point x ∈ X is abelian iff all base point change homomorphisms
βu depend only on the end points of the path u in X .
[Hint: Proceed as in Theorem 3.1.22.]
(b) Let (X, x 0 ) be a pointed space and P(X ) denote the space of paths (I, 0) →
(X, x0 ), i.e., the space of paths in X starting at x0 . If α0 is not the constant
path in X based at x0 , show that π1 (P(X ), α0 ) = 0.
11. Let X be an annulus surrounded by two concentric circles C1 and C2 and C be
any concentric circle lying completely in X . Show that the circle C is a strong
deformation retract of X and π1 (X ) ∼ = Z.
[Hint: From the centre O draw a half line in each direction through a point x as
shown in Fig. 3.15. Each such line meets the circle C in a unique point, r (x)(say).
Let r : X → C be the mapping which maps all points x of the half line to the
corresponding point r (x). Define H : X × I → X by H (x, t) = (1−t)x +tr (x).
Then H is a strong deformation retraction.]
12. (a) Show that there is a bijection
ψ : π1 (X, x0 ) → [(S 1 , 1), (X, x 0 )].
[Hint: Let α : I → X be a loop based at x0 . Then α determines a base point

preserving map
α̃ : (S 1 , 1) → (X, x0 ), e2πit → α(t).
Conversely, a base point preserving continuous map f : (S 1 , 1) → (X, x 0 )

determines a loop α f based at x0 given by α f (t) = f (e2πit ).]
(b) Show that every group is the fundamental of some path-connected topolog-
ical space.
(c) Let X be a path-connected space and let
ψ : π1 (X, x 0 ) → [S 1 , 1], [α] → [α̃] (ignoring base points).
Fig. 3.15 Annulus C2

surrounded by concentric
circles C1
C
O
3.9 Exercises 143
Show that ψ is a surjective map such that if β, δ ∈ π1 (X, x 0 ), then ψ(β) =

ψ(δ) iff β and δ are conjugate elements of π1 (X, x0 ).
13. Let M be the Möbius band. Show that π1 (M) ∼ = Z.
[Hint: The equatorial loop α : I → M such that α(t) = q(t, 12 ), where q :
I × I → M is the canonical identification, represents a generator of π1 (M).]
14. Let f : S 1 → S 1 be a continuous map such that deg f = n. Show that the
homomorphism f ∗ : π1 (S 1 ) → π1 (S 1 ) is given by f ∗ (z) = z n .
15. Let f, g : S 1 → S 2 be the two continuous maps defined by f (z) = z n and
g(z) = z −n . Compute their induced homomorphisms f ∗ and g∗ of the infinite
cyclic group π1 (S 1 , s0 ) into itself.
16. Prove that m-torus T m and n-torus T n are homeomorphic iff m = n.
17. Let A be a deformation retract of a space X . Show that for each a ∈ A, the
groups π1 (X, a) and π1 (A, a) are isomorphic.
18. Show that π1 (R2 − Q2 ) is uncountable.
19. Let X be a finite graph. Define Euler characteristic χ(X ) to be the number of
vertices minus the number of edges. Show that χ(X ) = 1 if X is a tree and the
rank (number of elements in a basis) of π1 (X ) is 1 − χ(X ) if X is connected.
20. Show that π1 (SO(3)) ∼ = Z2 , where SO(3) is the Lie group of rotations in R3 .
21. Prove that the fundamental groups of the following spaces are commutative.
(i) R P n ;
(ii) Vm,n = Hom (Rm , Rn );
(iii) the space of non-degenerate real n × n matrices GL(n, R) = {A : det A =
0};
(iv) the space of orthogonal real n × n matrices O(n, R) = { A : A At = 1d };
(v) the space of special unitary complex n × n matrices SU(n) = { A : A Āt =
1d and det A = 1}.
22. The bouquet of the topological spaces X 1 , X 2 , . . . , X n each with a base point,
is the space obtained from the union X 1 ∪ X 2 ∪ · · · ∪ X n by identifying together
all the base points.
Show that the fundamental group of the bouquet of n circles is the free group on
n generators.
23. Show that the trefoil knot
(i) is not equivalent to the trivial knot;
(ii) cannot be untied.
24. Show that abelianizing a knot group gives the infinite cyclic group.
25. Show that the homomorphism π1 (X (1) ) → π1 (X ) induced by the inclusion of
the one-skeleton of a polyhedron X is surjective.
26. Show that the fundamental groups of a sphere with n handles has the following
presentation:
a1 , b1 , . . . , an , bn : a1 b1 a1−1 b1−1 · · · an bn an−1 bn−1 .

27. Let X be a path-connected space. Show that there is a bijective correspondence

between the conjugacy classes of elements of π1 (X, x 0 ) = [(S 1 , 1) → (X, x0 )]
and the set [S 1 , X ], the free homotopy classes of maps f : S 1 → X (i.e., having
no base points).
28. Let G be a free group and k-elements and let H be a subgroup of G of index n.
Using fundamental group, show that H is a free group on (k − 1)n + 1 elements.
29. Given [ f ] ∈ π1 (S 1 , 1), let γ be the contour { f (t) : t ∈ I } ⊂ C. Define

1 dz
w( f ) = .
2π z
γ
Using fundamental group, show that

(i) w( f ) is an integer;
(ii) w( f ) is independent of the choice of representative f ∈ [ f ];
(iii) w( f ) = deg f .
30. Show that
(i) two continuous maps f, g : S 1 → S 1 are homotopic iff they have have the
same degree;
(ii) the degree function of maps f : S 1 → S 1 sets up a one to one correspon-
dence between the set of homotopy classes [S 1 , S 1 ] and the set Z.
[2] Aguilar, M., Gitler, S., Prieto, C., Algebraic Topology from a Homotopical View
[5] Crowell, R.H., and Fox, R.H., Introduction to Knot Theory, Ginn and Company,
Boston, 1963.
[8] Eilenberg, S., and Steenrod, N., Foundations of Algebraic Topology, Princeton
University Press, Princeton, 1952.
[9] Gray, B., Homotopy Theory, An Introduction to Algebraic Topology, Academic
[10] Hatcher, Allen, Algebraic Topology, Cambridge University Press, 2002.

Cambridge, 1983.
[13] Spanier, E., Algebraic Topology, McGraw-Hill Book Company, New York,
1966.
References
Adhikari, M.R., Adhikari, A.: Basic Modern Algebra with Applications. Springer, New Delhi (2014)
New York (2002)
Bredon, G.: Topology and Geometry. Springer, GTM 139 (1993)
New Delhi (2002)
Crowell, R.H., Fox, R.H.: Introduction to Knot Theory. Ginn and Company, Boston (1963)
Dieudonné, J.: A History of Algebraic and Differential Topology, 1900–1960. Modern Birkhäuser,
Basel (1989)
ton (1952)
(1975)
Munkres, J.R., Topology, A First Course. Prentice-Hall, New Jersey (1975)
Chapter 4
Covering Spaces
This chapter continues the study of the fundamental groups and is designed to
utilize the power of the fundamental groups through a study of covering spaces.
The fundamental groups are deeply connected with covering spaces. Algebraic fea-
tures of the fundamental groups are expressed by the geometric language of covering
spaces. Main interest in the study of this chapter is to establish an exact correspon-
dence between the various connected covering spaces of a given base space B and
subgroups of its fundamental group π1 (B), like Galois theory, with its correspon-
dence between field extensions and subgroups of Galois groups, which is an amazing
result. Historically, the systemic study of covering spaces appeared during the late
19th century and early 20th century through the theory of Riemann surfaces. But its
origin was found before the invention of the fundamental groups by H. Poincaré in
1895. Poincaré introduced the concept of universal covering spaces in 1883 to prove
a theorem on analytic functions.
The theory of covering spaces is of great importance not only in topology but also
in other branches of mathematics such as complex analysis, geometry, Lie groups
and also in some areas beyond mathematics. A covering space is a locally trivial map
with discrete fibers. The objects of this nature can be classified by algebraic objects
related to fundamental groups. The exponential map p : R → S 1 defined by p(x) =
e2πi x , x ∈ R is a powerful covering projection and (R, p) is the universal covering
space of S 1 . Chapter 3 has utilized this map as a tool for computing π1 (S 1 ). Covering
spaces likewise provide useful general tools for computation of fundamental groups.
The fundamental group is instrumental for classifying the topological spaces which
can be covering spaces of a given base space B. For a large class of spaces, the
possible covering spaces of B are determined by the subgroups of π1 (B). Moreover,
the theory of covering spaces facilitates to determine the fundamental groups of
several spaces.
More precisely, this chapter considers a class of mappings p : X → B, called the
‘covering projections’ from a space X , called a covering space, to a space B, called
base space, to which the properties of the exponential map p are extended. Moreover,
DOI 10.1007/978-81-322-2843-1_4
148 4 Covering Spaces
this chapter introduces the concepts of fibrations and cofibrations born in geometry
and topology and proves some classical results such as Borsuk–Ulum theorem and
Hurewicz theorem for a fibration.
For this chapter the books Croom (1978), Hatcher (2002), Rotman (1988), Spanier
(1966), Steenrod (1951), and some others are referred in the Bibliography.
4.1 Covering Spaces: Introductory Concepts and Examples
This section introduces the concept of covering spaces. Covering spaces displays
the first example of the power of the fundamental groups in classifying topological
spaces. Algebraic features of the fundamental groups π1 (B) of the base space B are
expressed in the geometric language of covering spaces of B.
This subsection introduces the concept of a covering space with illustrative examples.
Recall that a topological space X is path-connected if each pair of points in X can be
joined by a path in X . A space that satisfies this property locally is called ‘locally path-
connected.’ If X is a disconnected space, a maximal path-connected subset of the
space X is called a path component and is not a proper subset of any path-connected
subset of X . The path components of a subset B of X are the path components of B
in its subspace topology. For example, each interval and each ray in the real line are
both path-connected and locally path-connected. On the other hand, the subspace
[−1, 0) ∪ (0, 1] of R is not path-connected but it is locally path-connected. The
deleted comb space is path-connected but not locally path-connected. The space of
rationals Q is neither connected nor locally connected.
Definition 4.1.1 Let X and B be topological spaces and let p : X → B be a con-
tinuous surjective map. An open set U of B is said to be evenly covered by p if
p−1 (U ) is a union of disjoint open sets Si , called sheets such that p| Si : Si → U is
a homeomorphism for each i and U is called an admissible open set in B.
Example 4.1.2 Consider the exponential map p : R → S 1 defined by
p(x) = e2πi x = cos 2πx + i sin 2πx, x ∈ R.

Then the open set U = S 1 − {1} is evenly covered by p, since p−1 (U ) = (n −
n∈Z
1 1
, n + ). Clearly, the sheets are open intervals.
2 2
Definition 4.1.3 Let B be a topological space. The pair (X, p) is called a covering
space of B if
4.1 Covering Spaces: Introductory Concepts and Examples 149
(i) X is a path-connected topological space;

(ii) the map p : X → B is continuous;
(iii) each b ∈ B has an open neighborhood which is evenly covered by p.
The map p is called the covering projection and an open set in B which is evenly
covered by p is called p-admissible or simply admissible.
Remark 4.1.4 Some authors do not assume X to be path-connected but assume p to
be surjective while defining a covering space.
Example 4.1.5 The exponential map p : R → S 1 defined in Example 4.1.2 is a cov-
ering projection and hence (R, p) is a covering space of S 1 . Because the open sets
U1 = S 1 − {−1} and U2 = S 1 − {1} are evenly covered by p. Thus each point of
S 1 has an admissible open neighborhood in S 1 . In fact, any proper connected arc
of S 1 is evenly covered by p. The same argument shows that the map p : R → S 1
defined by p(t) = eiαt , where α ∈ R is a fixed nonzero real number, is also a covering
projection.
Example 4.1.6 For any positive integer n, let pn : S 1 → S 1 be the map defined
by pn (z) = z n , z ∈ S 1 . Then (S 1 , pn ) is a covering space of S 1 . Because, in polar
coordinates, pn is given by pn (1, θ) = (1, nθ). The map pn wraps the circle around
itself n times. Let U be an open arc on S 1 subtended by an angle θ, 0 ≤ θ ≤ 2π,
and containing a point x. Then p−1 (U ) consists of n open arcs each determining an
angle θ/n and each containing one nth root of x. Each of these n open arcs is mapped
homeomorphically onto U . Thus any proper arc in S 1 is an admissible neighborhood.
Consequently, (S 1 , pn ) is a covering space of S 1 .
Example 4.1.7 Consider the map f : R2 → S 1 × S 1 from the plane to the torus
defined by f (t1 , t2 ) = (e2πit1 , e2πit2 ), (t1 , t2 ) ∈ R2 . Then (R2 , f ) is a covering space
of S 1 × S 1 .
For any point (z 1 , z 2 ) ∈ S 1 × S 1 , let U be a small rectangle formed by the product
of two open arcs in S 1 containing z 1 and z 2 , respectively. Then U is an admissible
neighborhood whose inverse image consists of a countably infinite family of open
rectangles in the plane R2 . This example is essentially a generalizaton of the covering
projection p : R → S 1 .
Theorem 4.1.8 Let (X 1 , p1 ) be a covering space of B1 , (X 2 , p2 ) be a covering space
of B2 , then (X 1 × X 2 , p1 × p2 ) is a covering space of B1 × B2 , where p1 × p2 :
X 1 × X 2 → B1 × B2 is defined by ( p1 × p2 )(x, y) = ( p1 (x), p2 (y)).
Proof Let (b1 , b2 ) ∈ B1 × B2 and U1 be an open neighborhood of b1 and U2 be
an open neighborhood of b2 which are evenly covered by p1 and p2 , respectively.
Then U1 × U2 is a neighborhood of (b1 , b2 ) in B1 × B2 which is evenly covered by
p1 × p2 . ❑
Example 4.1.9 Consider the exponential map p : R → S 1 defined by
p(x) = e2πi x = cos 2πx + i sin 2πx, x ∈ R.

Then the map ( p, p) : R × R → S 1 × S 1 is a covering projection. In fact for every

positive integer n, the product map p ◦ p ◦ · · · ◦ p = p n : Rn → T n is a covering
n
projection, where T n = S 1 is the n-dimensional torus.
1
Theorem 4.1.10 Let p : (X, x0 ) → (B, b0 ) be a covering projection. If X is path-

connected, then there is a surjection ψ : π1 (B, b0 ) → p−1 (b0 ). If X is simply con-
nected, then ψ is a bijection.
Proof Following the technique for computation of π1 (S 1 , 1) (see Theorem 3.3.9 of

Chap. 3) the theorem can be proved. ❑
Remark 4.1.11 Everytopological space is not necessarily a covering space. The fol-
lowing is an example of a topological space X which is not a covering space of Y .
Example 4.1.12 Let X be a rectangle which is mapped by the projection p onto the
first coordinate to an interval Y . Let U be an interval in Y . Then p −1 (U ) is a strip in
X consisting of all points above U (as shown in Fig. 4.1).
This strip cannot be mapped by p homeomorphically onto U . Hence U is not
evenly covered by p. Consequently, (X, p) is not a covering space of Y .
Example 4.1.13 (Infinite and finite spirals) Let X be an infinite spiral, and p : X →
S 1 be the projection described in Fig. 4.2.
Fig. 4.1 Example of a

neighborhood which is not p−1 (U ) (a, b) X
evenly covered
p(a, b) = a
( ) Y
U
Fig. 4.2 Infinite spiral with

projection p
S1
Fig. 4.3 Finite spiral with

projection p
S1
Each point of X is projected by p to the point on the circle directly below it. Then
(X, p) is a covering space of S 1 . On the other hand, if p : Y → S 1 is a finite spiral
projection as shown in Fig. 4.2, then (Y, p) is not covering space of S 1 , because if x0
and x1 are the end points of the spiral Y , then the points p(x0 ) and p(x1 ) as shown
in Fig. 4.3 have no admissible neighborhoods.
4.1.2 Some Interesting Properties of Covering Spaces
This subsection presents some properties of covering spaces.

Proposition 4.1.14 Let p : (X, x0 ) → (B, b0 ) be a covering space. Then the induced
homomorphism p∗ : π1 (X, x0 ) → π1 (B, b0 ) is a monomorphism and the subgroup
p∗ (π1 (X, x0 )) in π1 (B, b0 ) consists of homotopy class of loops in B based at b0
which lifts to X starting at x0 are loops.
Proof Let an element α ∈ ker p∗ be represented by a loop f˜0 : I → X with a homo-
topy Ft : I → B of f 0 = p ◦ f˜0 to the trivial loop f1 (Fig. 4.4).
Hence there exists a lifted homotopy of loop F t : I → X started at f˜0 and end-
ing with a constant loop (because the lifted homotopy F t is a homotopy of paths

fixing the end points, since t varies each point of Ft gives a path lifting a constant
path, which is therefore constant). Hence [ f˜0 ] = 0 in π1 (X, x 0 ) shows that p∗ is
injective. ❑
We now state the following two other properties of covering spaces whose proofs
are given in Sect. 4.5.2.
Fig. 4.4 Homotopy diagram 8X

qqq
corresponding to a lifting of qqqqq
f˜0
p
qq
f 0 to f˜0 qqq /

I F t
B
Proposition 4.1.15 Let p : (X, x0 ) → (B, b0 ) be a covering space and f : (Y, y0 )

→ (B, b0 ) be a map, where Y is path-connected and locally path-connected. Then
a lift f˜ : (Y, y0 ) → (X, x0 ) of f exists iff f ∗ (π1 (Y, y0 )) ⊂ p∗ (π1 (X, x 0 )).
Proposition 4.1.16 Given a covering space p : X → B and a map f : Y → X with

two lifting f˜1 , f˜2 : Y → X that agree at some point of Y , then if Y is connected,
f˜1 = f˜2 , i.e., f˜1 (y) = f˜2 (y), ∀ y ∈ Y.
4.1.3 Covering Spaces of R P n
This subsection studies covering spaces of real projective spaces R P n and computes
fundamental group of R P 2 .
Definition 4.1.17 (Real projective plane) Let R P 2 be the real projective plane
defined as a quotient space of the 2-sphere S 2 obtained by identifying each point
x of S 2 with its antipodal point −x and p : S 2 → R P 2 be the natural map which
identifies each pair of antipodal points i.e., p maps each x to its equivalence class.
We topologize R P 2 by defining V to be open in R P 2 if and only if p −1 (V ) is open
in S 2 . With this topology R P 2 becomes a topological space.
Theorem 4.1.18 The projective space R P 2 is a surface and (S 2 , p) is a covering

space of R P 2 .
Proof First we show that p : S 2 → R P 2 is a covering map. Given y ∈ R P 2 , we

choose x ∈ p −1 (y). We then choose an -neighborhood U of x in S 2 for some < 1,
using the Euclidean metric d of R 3 . If A : S 2 → S 2 is the antipodal map sending z
to its antipodal point −z, then U contains no pair {z, A(z)} of antipodal points of S 2 ,
since d(z, a(z)) = 2. Consequently, the map p : U → p(U ) is bijective. The antipo-
dal map A : S 2 → S 2 , given by A(z) = −z is a homeomorphism of S 2 and hence
A(U ) is open in S 2 . Since p −1 ( p(U )) = U ∪ A(U ), this set is also open in S 2 . Con-
sequently, p(U ) is open in R P 2 and hence p is an open map. Thus the bijective map
p : U → p(U ) is continuous and open. Hence it is a homeomorphism. Similarly,
p : A(U ) → p(A(U )) = p(U ) is a homeomorphism. The set p−1 ( p(U )) is thus the
union of two open sets U and A(U ), each of which is mapped homeomorphically by
p onto p(U ). Hence p(U ) is a neighborhood of p(x) = y, which is evenly covered
by p. Consequently, (S 2 , p) is a covering space of R P 2 . For the first part, let {Un }
be countable basis of S 2 . Then { p(U )} is a countable basis of R P 2 . Clearly, R P 2 is
a Hausdorff space. Let y1 and y2 be two points of R P 2 . The set p −1 (y1 ) ∪ p −1 (y2 )
consists of four points. Let 2 be the minimum distance between them. Let U1 be
the -neighborhood of one of the points p −1 (y1 ) and U2 be the -neighborhood of
one of the points p −1 (y2 ). Then the sets U1 ∪ A(U1 ) and U2 ∪ A(U2 ) are disjoint.
Consequently, p(U1 ) and p(U2 ) are disjoint neighborhoods of y1 and y2 , respec-
tively, in R P 2 . Since S 2 is a surface and every point of R P 2 has a neighborhood
homeomorphic to an open subset of S 2 , the space R P 2 is also a surface. ❑
A generalization of Theorem 4.1.18 for n > 1 is now given.

Theorem 4.1.19 (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 The sets Ei+ = {(x 1 , x 2 , . . . , xn+1 ) ∈ S n : x i > 0} and E i− = {(x1 , x2 , . . . ,

xn+1 ) ∈ S n : xi < 0} are open sets and cover S n . The map p| Ei+ is 1-1, continuous and
open. Hence if Ui = p(Ei+ ) = p(E i− ), then p −1 (Ui ) = E i+ ∪ E i− . The sets E i+ and
E i− are disjoint open sets, and homeomorphic to Ui . This shows that p : S n → R P n
is a covering space. This asserts that (S n , p) is a covering space of R P n . ❑
Definition 4.1.20 The multiplicity of a covering space (X, p) of B is the cardinal

number of a fiber. If the multiplicity is n, we say that (X, p) is an n-sheeted covering
space of B or that (X, p) is an n-fold cover of B.
Example 4.1.21 (i) (S 2 , p) is a double covering of R P 2 .

(ii) The number of sheets of (R, p) of S 1 is countably infinite.
Because, p identifies pairs of antipodal points, the number of sheets of this
covering in (i) is 2. On the other hand, for the (ii) covering projection p : R → S 1
(see Example 4.1.2) maps each integer and only the integers to 1 ∈ S 1 . Thus
p −1 (1) = Z and hence the number of sheets of this covering is countably infinite.
Theorem 4.1.22 π1 (R P 2 , y) ∼
= Z2 .
Proof The projection p : S 2 → R P 2 is covering map by Theorem 4.1.18. Since S 2

is simply connected, we apply Theorem 4.1.10, which gives a bijective correspon-
dence between π1 (R P 2 , y) and the set p−1 (y). Since p−1 (y) is a two-element set,
π1 (R P 2 , y) is a group of order 2. Since any group of order 2 is isomorphic to Z2 , it
follows that π1 (R P 2 , y) ∼
= Z2 . ❑
Remark 4.1.23 For computing π1 (R P n , y) by using the universal covering space

(S n , q) of R P n , where q identifies the antipodal points of S n , (see Sect. 4.6.2), use
topological group action see Corollary 4.10.4.
4.2 Computing Fundamental Groups of Figure-Eight

and Double Torus
We now consider some topological spaces whose fundamental groups are nonabelian.
This section constructs covering spaces for computation of fundamental groups of
some spaces such as figure-eight and double torus whose fundamental groups are
not abelian. For computing the fundamental group of figure-eight by graph-theoretic
method see Sect. 4.10.6.
Example 4.2.1 (figure-eight) The figure-eight F is the union of two circles A and B
with a point x 0 in common. We now describe a certain covering space X for F.
Let X be the subspace of the plane consisting 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. 4.5.
The projection map p wraps the x-axis around the circle A and wraps the y-axis
around the other circle B; in each case the integer points are mapped by p into the
base point x 0 of F. Then each circle tangent to an integer point on the x-axis is
mapped homeomorphically by p onto B; on the other hand, each circle tangent to
an integer point on the y-axis is mapped homeomorphically onto A; in each case the
point of tangency is mapped onto the point x0 . Then p is a covering map.
Theorem 4.2.2 The fundamental group of the figure-eight is not abelian.

Proof Let f˜ : I → X be the path f˜(t) = (t, 0), going along the x-axis from the
origin (0, 0) to the point (1, 0). Let g̃ : I → X be the path g̃(t) = (0, t), going along
the y-axis from the origin (0, 0) to the point (0, 1). Let f = p ◦ f˜ and g = p ◦ g̃.
Then f and g are loops, in the figure-eight F based at x0 , going around the circles A
and B, respectively. We claim that f ∗ g and g ∗ f are not path homotopic. We lift
each of these paths to a path in X beginning at the origin. The path f ∗ g lifts to a
path that goes along the x-axis from the origin to (1, 0), and then goes once around
the circle tangent to the x-axis at (1, 0). But 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). Since the lifted paths do not end at the same point, f ∗ g
and g ∗ f cannot be path homotopic. Therefore we conclude that [ f ∗ g] = [g ∗ f ]
and hence [ f ] · [g] = [g] · [ f ] proves that the fundamental group of the figure eight
is not abelian. ❑
Remark 4.2.3 For computing the fundamental group of figure-eight by graph-
theoretic method see Sect. 4.10.6.
Corollary 4.2.4 The fundamental group of the double torus T2 is not abelian.
Fig. 4.5 Figure-eight
(0,2) A2
(0,1) A1
B−2 B−1 B1 B2
(-2,0) (-1,0) (1,0) (2,0)
x0
B A
4.2 Computing Fundamental Groups of Figure-Eight and Double Torus 155
Proof Figure-eight F is a retract of T2 ⇒ the inclusion map i : (F, x 0 ) → (T2 , x 0 )

induces a monomorphism i ∗ : π1 (F, x0 ) → π1 (T2 , x0 ) ⇒ π1 (T2 , x0 ) is not abelian,
since π1 (F, x0 ) is not abelian. ❑
Remark 4.2.5 For computing fundamental groups of some orbit spaces see
Sect. 4.10.2.
4.3 Path Lifting and Homotopy Lifting Properties
This section continues the study of covering spaces and displays basic properties of
covering spaces such as path lifting and homotopy lifting properties (PLP and HLP).
We begin with characterization of locally path-connected spaces.
Recall the following definitions.
Definition 4.3.1 A topological space X is said to be locally path-connected if for
each point x ∈ X and every neighborhood Ux of x, there is an open set V with
x ∈ V ⊂ Ux such that any two points in V can be joined by a path in U x .
Definition 4.3.2 A topological space X is said to be semilocally path-connected if

for every point x ∈ X , there is an open neighborhood Ux of x such that every closed
path in Ux at x is nullhomotopic in X .
Proposition 4.3.3 A topological space X is locally path-connected if and only if

each path component of each open subset of X is open.
Proof Left as an exercise. ❑
Theorem 4.3.4 Every covering projection p : X → B is an open mapping for any

locally path-connected space X .
Proof Let X be a locally path-connected space such that p : X → B be a covering

projection and V be an open set in X . We claim that p(V ) is open in B. Let b ∈ p(V )
and x ∈ p−1 (b) and U be an admissible neighborhood for b. Then x is a point of
V such that p(x) = b. Let W be the component of p−1 (U ) which contains x. Since
X is locally path connected, W is open in X by Proposition 4.3.3. Since p maps W
homeomorphically onto U , p maps the open set W ∩ V to the open subset p(W ∩ V )
in B. Then b ∈ p(W ∩ V ) ⊆ p(V ). Since b is an arbitrary point of p(V ), it follows
that p(V ) is a union of open sets and hence p(V ) is an open set. Consequently p is
an open mapping. ❑
Theorem 4.3.5 Let (X, p) be a covering space of B and Y be a space. If f and

g are continuous maps from Y to X for which p ◦ f = p ◦ g, as shown in Fig. 4.6,
then the set A = {y ∈ Y : f (y) = g(y)} (i.e., the set of points of Y at which f and
g agree) is both open and closed in Y . (Y is not assumed to be path-connected or
locally path-connected).
u: X
Fig. 4.6 Triangular diagram
involving f , g and p uu
uuu
f,g uuu
u
u p
uu
uuu
uu
u
uu
Y /B
p◦f =p◦g
Proof To prove that A is open, let y ∈ A and U be an admissible neighborhood

of ( p ◦ f )(y). Then the path component V of p−1 (U ) to which f (y) belongs is
an open set in X and hence f −1 (V ) and g −1 (V ) are open in Y . Since f (y) ∈ V
and f (y) = g(y), then y ∈ f −1 (V ) ∩ g −1 (V ). We claim that f −1 (V ) ∩ g −1 (V ) is a
subset of A and conclude that A is open, since it contains a neighborhood of each of its
points. Let t ∈ f −1 (V ) ∩ g −1 (U ). Then f (t), g(t) ∈ V and ( p ◦ f )(t) = ( p ◦ g)(t).
Since p maps V homeomorphically onto U , it follows that f (t) = g(t) and hence
t ∈ A. Thus it follows that A is an open set.
Next we prove that A is closed. Suppose A is not closed and let t be a limit
point of A not in A. Then f (t) = g(t). The point ( p ◦ f )(t) = ( p ◦ g)(t) has an
elementary neighborhood U such that the points f (t) and g(t) must be in distinct
path components V1 and V2 of p−1 (U ). Since t ∈ f −1 (V1 ) ∩ g −1 (V2 ) which is an
open set in Y, f −1 (V1 ) ∩ g −1 (V2 ) must contain a point y ∈ A. But this implies a
contradiction, since V1 ∩ V2 = ∅ and f (y) = g(y) ∈ V1 ∩ V2 . Hence all limit points
of A must lie in A and therefore A is closed. ❑
Corollary 4.3.6 Let (X, p) be a covering space of B, and let f, g be continuous

maps from a connected space Y into X such that p ◦ f = p ◦ g. If f and g agree at
a point of Y , then f = g.
Proof Let Y be a connected space. Then the only sets that are both open and closed
in Y are Y and ∅. Hence by Theorem 4.3.5 it follows that either A = Y or A =
∅. This implies that either f (y) = g(y) at every y ∈ Y or f (y) = g(y) at every
y ∈ Y . By hypothesis f (y) = g(y) at some y ∈ Y . Thus A = ∅ and hence A = Y .
Consequently, f (y) = g(y), ∀ y ∈ Y shows that f = g. ❑
Remark 4.3.7 The Corollary 4.3.6 gives the uniqueness of the lifting of a map and
generalizes Proposition 3.3.2 of Chap. 3.
We now consider lifting problems. What is lifting problem?

Let p : X → B be a continuous surjective map (not necessarily a covering pro-
jection). Given a subspace A of X and a continuous map f : A → B, does there
exist a continuous map f˜ : A → X such that p ◦ f˜ = f ?
In other words, can we find a continuous map f˜ : A → X making the diagram in
Fig. 4.7 commutative? If such f˜ exists, then f˜ is called a lift of f . The satisfactory
answer is available if p is covering projection.
4.3 Path Lifting and Homotopy Lifting Properties 157
u: X
Fig. 4.7 Lifting of a map f
u
u
f˜
u
u
u p
u
u
u
u
u /B
A f
u: X
Fig. 4.8 Lifting of a path f
u
u
f˜
u
u
u p
u
u
u
u
u /B
I f
Definition 4.3.8 Let (X, p) be a covering space of B and let f : I → B be a path

in B. A path f˜ : I → X in X such that p ◦ f˜ = f , is called a lifting or covering
path of f ,i.e., if it makes the diagram as shown in Fig. 4.8 commutative.
If F : I × I → B be a homotopy, then a homotopy F̃ : I × I → X for which
p ◦ F̃ = F, is called a lifting or covering homotopy of F.
We now generalize Theorem 3.3.3 and its Corollary 3.3.4. of Chap. 3.
Theorem 4.3.9 (The Path Lifting Property) Let (X, p) be a covering space of B
and f : I → B be a path in B beginning at a point b0 ∈ B. If x0 ∈ p −1 (b0 ), then
there is a unique covering path f˜ : I → X as shown in Fig. 4.9 of f beginning at x0
such that p ◦ f˜ = f .
Proof Existence of f˜: Suppose [a, b] ⊂ I is such that f ([a, b]) ⊂ U, where U is
an admissible neighborhood of y = f (a) in B. Let x ∈ f −1 (y). Then x lies in a
unique sheet S (say). Define
g̃ : ([a, b], a) → (X, x), by g̃ = ( p| S )−1 ◦ ( f |[a,b] )
such that p ◦ g̃ = f |[a,b] . Let Ut be an admissible neighborhood of f (t) for each

t ∈ I . Then { f −1 (Ut ), t ∈ I }, being an open cover of the compact metric space I
has a Lebesgue number λ. This shows that if 0 < δ < λ and Y is a subset of I of
diameter less than δ, then Y ⊂ f −1 (Ut ) for some t ∈ I. Thus f (Y ) ⊂ Ut partitions I
with points t1 = 0, t2 , . . . , tk = 1, where ti+1 − ti < δ for 1 ≤ i ≤ k − 1. Then there
is a continuous map g̃1 : [0, t2 ] → X satisfying p ◦ g̃1 = f |[0,t2 ] and g̃1 (o) = x0 .
Fig. 4.9 Path lifting (X, x0 )

r9
property (PLP) r
r
f˜
r
r
r p
r
r
r
r
r
(I, 0) / (B, b0 )
f
Similarly, there is a continuous map g̃2 : [t2 , t3 ] → X satisfying p ◦ g̃2 = f |[t2 ,t3 ]
and g̃2 (t2 ) = g̃1 (t2 ). In this way, for 1 ≤ i ≤ k − 2, there is a continuous map
g̃i+1 : [ti+1 , ti+2 ] → X
satisfying p ◦ g̃i+1 = f |[ti+1 ,ti+2 ] and g̃i+1 (ti+1 ) = g̃i (ti+1 ). Using gluing lemma, and
assembling the functions gi , we obtain a continuous function f˜ : I → X , where
f˜(t) = g̃i (t) if t ∈ [ti , ti+1 ].
The uniqueness of f˜: It follows from Corollary 4.3.6, because I is connected, and
by assumption any two lifts of f agree at the point 0 ∈ I . ❑
Corollary 4.3.10 (Homotopy Lifting Property) Let (X, p) be a covering space of

B and F : I × I → B be a homotopy such that F(0, 0) = b0 . If x0 ∈ p −1 (b0 ), then
there exists a unique homotopy F̃ : I × I → X such that F̃(0, 0) = x0 .
Proof Proceed as in Theorem 4.3.9 by subdividing I × I into rectangles (in place

of I ). ❑
We can prove in a similar way the general form of the Homotopy Lifting Property.
Theorem 4.3.11 (The Generalized Homotopy Lifting Property) Let (X, p) be a

covering space of B and A be a compact space. If f : A → X is continuous and
F : A × I → B is a homotopy starting from p ◦ f , then there is a homotopy F̃ :
A × I → X starting from f and lifts F. Furthermore, if F is a homotopy relative to
a subset A of A, then F̃ is also so.
4.4 Lifting Problems of Arbitrary Continuous Maps
This section gives a necessary and sufficient condition for lifting of an arbitrary
continuous map f : A → X by applying the tools of fundamental groups. More
precisely, given a covering space (X, p) of B and a continuous map f : A → X , can
we find a continuous map f˜ : A → X such that p ◦ f˜ = f ? The answer is positive if
f is a path or a homotopy between paths by the Path Lifting Property (Theorem 4.3.9),
and the Homotopy Lifting Property (Corollary 4.3.10), respectively. To the contrary
the answer is negative for an arbitrary continuous map f . For more results see
Chap. 16.
Example 4.4.1 The exponential map p : R → S 1 defined by p(t) = e2πit is a cov-

ering projection. The identity map 1 S 1 : S 1 → S 1 cannot be lifted to a continu-
ous map ψ : S 1 → R making the triangle in Fig. 4.10 commutative. Otherwise,
p ◦ ψ = 1 S 1 ⇒ ψ is injective ⇒ ψ is an embedding of S 1 into R, since S 1 is compact
⇒ ψ(S 1 ) is a closed interval homeomorphic to S 1 , since any compact connected sub-
set of R must be a closed interval. This is impossible, since a closed interval cannot
be homeomorphic to S 1 .
4.4 Lifting Problems of Arbitrary Continuous Maps 159
t: R
Fig. 4.10 Covering
projection for exponential t
t
t
map p ψ t
t p
t
t
t
t
t
S1 / S1
1S 1
Remark 4.4.2 We now give a necessary and sufficient condition under which an arbi-
trary continuous map f : A → X can be lifted. The methods of algebraic topology
are now applied to solve such problems.
Theorem 4.4.3 (Lifting Theorem) Let (X, p) be a covering space of B. Given a

connected and locally path-connected space A, let f : A → B be any continuous
map. Then given any three points a0 ∈ A, b0 ∈ B and x 0 ∈ X such that f (a0 ) = b0
and p(x0 ) = b0 , there exists a unique continuous map f˜ : A → X satisfying f˜(a0 ) =
x0 such that p ◦ f˜ = f if and only if f ∗ (π1 (A, a0 )) ⊂ p∗ (π1 (X, x0 )).
Proof Suppose that ∃ a continuous map f˜ : A → X satisfying the given con-

ditions. Then the diagram in Fig. 4.11 is commutative. Hence the diagram in
Fig. 4.12 is also commutative (by the functorial property of π1 ). Consequently,
f ∗ (π1 (A, a0 )) = p∗ ( f˜∗ (π1 (A, a0 ))) ⊆ p∗ (π1 (X, x0 )). Conversely, let the algebraic
condition f ∗ (π1 (A, a0 )) ⊂ p∗ (π1 (X, x0 )) holds.
Since A is connected, A has only one component. Again since A is locally path-
connected, this component is a path component. Hence A is path-connected.
Let a ∈ A. We take a path u : I → A such that u(0) = a0 and u(1) = a. Then f ◦
u : I → B is a path such that ( f ◦ u)(0) = f (u(0)) = f (a0 ) = b0 . By path Lifting
Property, Theorem 4.3.9, ∃ a unique path ũ : I → X that lifts f ◦ u in X with ũ(0) =
x0 as shown in Fig. 4.13. Define a map
f˜ : A → X, a → ũ(1).
Fig. 4.11 Lifting of f to f˜ (X, x0 )

q8
q
q
f˜
q
q
q p
q
q
q
q
q
(A, a 0 ) / (B, b 0)
f
Fig. 4.12 Induced π (X, x )

71 0
homomorphisms of f and f˜ p pp
p
f˜∗ p p
p p∗
p
p pp
p
π1 (A, a0 ) / π1 (B, b0 )
f∗
Fig. 4.13 Diagram for 7X

o o~?
lifting theorem o o o ~
~
ũ o o f˜ ~
p
o o o ~
~
o o ~
o o u,v /A
~ f
/B

I
Fig. 4.14 Two paths u and v A

in A a
v
u
a0
To show that f˜ is well defined, choose another path v from a0 to a as shown in

Fig. 4.14. Let ṽ be the unique path in X lifting f ◦ v for which ṽ(0) = x 0 , i.e.,
p ◦ ṽ = f ◦ v and ṽ(0) = x 0 .
Now u ∗ v −1 is a closed path in A at a0 . Then f ◦ (u ∗ v −1 ) = ( f ◦ u) ∗ ( f ◦ v −1 )
is a closed path in B at b0 . Again since
[( f ◦ u) ∗ ( f ◦ v −1 )] = f ∗ [u ∗ v −1 ] ∈ f ∗ π1 (A, a0 ) ⊆ p∗ π1 (X, x 0 )(by hypothesis),
there exists a closed path α in X at x 0 such that
( f ◦ u) ∗ ( f ◦ v −1 ) p ◦ α rel I˙.
Hence
( f ◦ u) ∗ ( f ◦ v −1 ) ∗ ( p ◦ ṽ) ( p ◦ α) ∗ ( p ◦ ṽ) rel I˙;
f ◦ u p ◦ (α ∗ ṽ) rel I˙, since p ◦ ṽ = f ◦ v.
Again by homotopy lifting property (see Corollary 4.3.10)
ũ α ∗ ṽ rel I˙ with ũ(1) = (α ∗ ṽ)(1) = ṽ(1).
This shows that f˜ is well defined.

f˜ is continuous: Let a ∈ A and U be an open neighborhood of f˜(a). To show the
continuity of f˜, we have to find an open neighborhood Va of a with f˜(Va ) ⊂ U. We
take an open admissible neighborhood V of p f˜(a) = f (a) such that V ⊂ p(U ).
Let W be the path component of p −1 (V ) which contains the point f˜(a), and let
V be an open admissible neighborhood of f (a) such that V ⊆ p(U ∩ W ). Then
the path component of p−1 (V ) containing f˜(a) must be contained in U . Since
f is continuous and path-connected A is locally connected, ∃ a path-connected
neighborhood Va of a such that f (Va ) ⊂ V . Then f˜(Va ) ⊂ U . ❑
4.4 Lifting Problems of Arbitrary Continuous Maps 161
Corollary 4.4.4 Let A be simply connected and locally path-connected and f :

(A, a0 ) → (B, b0 ) be continuous. If (X, p) is a covering space of B and if x 0 ∈
p −1 (b0 ), then ∃ a unique lifting f˜ : (A, a0 ) → (X, x0 ) of f .
Proof A is simply connected ⇒ π1 (A, a0 ) = 0 ⇒ p∗ π1 (A, a0 ) = {0} ⊂ p∗ π1

(X, x0 ). Then ∃ a unique lifting f˜ : (A, a0 ) → (X, x 0 ) of f . ❑
Corollary 4.4.5 Let B be a connected and locally path-connected space, and (X, p)
and (Y, q) be covering spaces of B. Let b0 ∈ B and x 0 ∈ X, y0 ∈ Y be base points
with p(x0 ) = b0 = q(y0 ). If p∗ π1 (X, x 0 ) = q∗ π1 (Y, y0 ), then there exists a unique
continuous map f : (Y, y0 ) → (X, x0 ) such that p ◦ f = q.
Example 4.4.6 (S n , p) is a covering space of R P n of multiplicity 2. Since S n

is simply connected for n ≥ 2, it follows that if x0 ∈ p −1 (b0 ), b0 ∈ R P n , then
for any continuous map f : (S n , s0 ) → (R P n , b0 ), there exists a unique lifting
f˜ : (S n , s0 ) → (S n , x0 ).
4.5 Covering Homomorphisms: Their Classifications

and Galois Correspondence
This section defines covering homomorphisms between covering spaces of the base
space B and classify the covering spaces with the help of conjugacy classes of the
fundamental group π1 (B). This classification establishes an exact correspondence
between the various connected covering spaces of a given space B and subgroups of
its fundamental group π1 (B), like Galois theory, with its correspondence between
field extensions and subgroups of Galois groups. There is a natural question: given
a space B, how many distinct covering spaces of B, we can find? Before answering
this question, we explain what is meant by distinct covering spaces of B.
4.5.1 Covering Homomorphisms and Deck Transformations
This subsection introduces the concepts of covering homomorphisms and deck trans-
formations.
Definition 4.5.1 Let (X, p) and (Y, q) be covering spaces of the same space B. A
covering homomorphism h from (X, p) to (Y, q) is a continuous map h : X → Y
such that the diagram in Fig. 4.15 is commutative. If in addition, h is a homeomor-
phism, then h is called an isomorphism. If there is an isomorphism from (X, p) to
(Y, q), then they are called isomorphic or equivalent covering spaces, otherwise, they
are said to be distinct covering spaces. An isomorphism of a covering space onto
itself is called an automorphism or a deck transformation.
Fig. 4.15 Covering X@

h /Y
homomorphism @@ ~~
@@ ~~
@@ ~~
@@ ~~
p @@@ ~~~ q
@@ ~
@@ ~~
~ ~
B
Remark 4.5.2 A homomorphism of covering spaces is a covering projection i.e.,if

h : X → Y is a homomorphism of covering spaces, then (X, h) is a covering space
of Y .
Proposition 4.5.3 Covering spaces of a space B and their homomorphisms form a

category.
Proof We take covering spaces of B as the class of objects and their homomorphisms
as the class of morphisms. Let (X, p) be a covering space of B. Then 1 X : X → X is
a covering homomorphism. If (X, p), (Y, q) and (Z , r ) are covering spaces of B and
h : X → Y, g : Y → Z are covering homomorphisms, then g ◦ h : X → Z is also a
covering homomorphism from (X, p) to (Z , r ). ❑
Isomorphisms in this category are just the isomorphisms of covering spaces as

defined above.
Let Aut (X/B) be the set of all automorphisms of covering space (X, p) of B.
Proposition 4.5.4 (Aut (X/B), ◦) is a group under usual composition of maps.
Proof The identity map 1 X : X → X is itself an automorphism and the inverse of

an automorphism is again an automorphism. Consequently, Aut (X/B) is a group
under usual composition of maps. ❑
Definition 4.5.5 Aut (X/B) is called the automorphism group of the covering space
(X, p) of B. These automorphisms are also known as the covering transformations
or deck transformations of the covering space (X, p) of B.
Let p : X → B, q : Y → B be covering projections. Then (X, p) and (Y, q) are

covering spaces of B. Suppose g, h : X → Y are two covering homomorphisms. We
now consider each of g and h as liftings of the map p : X → B with respect to the
covering projection q : Y → B (Fig. 4.16).
Fig. 4.16 Uniqueness of :Y

uu
lifting uu
uuu
g,h uuu
u q
uu
uuu
uuu
uu p
X /B
4.5 Covering Homomorphisms: Their Classifications and Galois Correspondence 163
Consequently, if X is connected and g and h both agree at a single point of X ,

then g = h. This proves the following proposition.
Proposition 4.5.6 Let g, h : X → Y be two covering homomorphisms from the
covering space (X, p) to the covering space (Y, q) of B. If X is connected and
g(x0 ) = h(x 0 ) for some x 0 ∈ X , then g = h.
4.5.2 Classification of Covering Spaces by Using Group

Theory
This subsection characterizes and classifies covering spaces of a space B with the
help of conjugacy classes of the group π1 (B). The following two results of algebra
are used in this subsection.
(i) If H and K are subgroups of a group G, then they are conjugate subgroups iff
H = g −1 K g for some g ∈ G.
(ii) If H and K are subgroups of a group G, then the G-sets G/H and G/K are
G-isomorphic iff H and K are conjugate subgroups in G.
Theorem 4.5.7 Let (X, p) be a covering space of B, where X and B are path-
connected. If b0 ∈ B, then the groups p∗ π1 (X, y), as y runs over Y = p −1 (b0 ), form
a conjugacy class of subgroups of π1 (B, b0 ).
Proof To prove the theorem we have to prove:
(a) for any y0 , y1 ∈ Y, the subgroups p∗ π1 (X, y0 ) and p∗ π1 (X, y1 ) are conjugate;
(b) any subgroup of π1 (B, b0 ) conjugate to p∗ π1 (X, y0 ) is equal to p∗ π1 (X, y) for
some y ∈ Y .
(a) Let u : I → X be a path from y0 to y1 . Then the function βu : π1 (X, y0 ) →
π1 (X, y1 ) defined by βu ([ f ]) = [ū ∗ f ∗ u], ∀ [ f ] ∈ π1 (X, y0 ), is an isomor-
phism (by Theorem 3.1.18). In particular, βu π1 (X, y0 ) = π1 (X, y1 ) ⇒ ( p∗ ◦
βu )π1 (X, y0 ) = p∗ π1 (X, y1 ). It follows from the definition of βu that ( p∗ ◦
βu )π1 (X, y0 ) = [ p ◦ u]−1 p∗ π1 (X, y0 )[ p ◦ u] ⇒ p∗ π1 (X, y1 ) and p∗ π1 (X, y0 )
are conjugate subgroups of π1 (B, b0 ).
(b) Let H be a subgroup of π1 (B, b0 ) such that H is conjugate to p∗ π1 (X, y0 )
for some [g] ∈ π1 (B, b0 ). Then H = [g]−1 p∗ π1 (X, y0 )[g]. Let g̃ be the unique
lifting of g in X starting at y0 . Then g̃(1) = y(say)∈ Y . Now proceeding as in
(a), we have
p∗ π1 (X, y) = [ p ◦ g̃]−1 p∗ π1 (X, y0 )[ p ◦ g̃]

= [g]−1 p∗ π1 (X, y0 )[g] = H
⇒ p∗ π1 (X, y) = H.
We conclude that the set { p∗ π1 (X, y) : y ∈ Y } forms a complete conjugate class

of subgroups of the group π1 (B, b0 ).
Definition 4.5.8 The conjugacy class of subgroups { p∗ π1 (X, y) : y ∈ Y = p−1 (b0 )}

described above is called the conjugate class determined by the covering space (X, p)
of B.
We now characterize covering spaces of a base space B with the help of conjugacy
classes of subgroups of π1 (B).
Theorem 4.5.9 Let B be path-connected and locally path-connected. Let (X, p)
and (Y, q) be path-connected covering spaces of B; let p(x0 ) = q(y0 ) = b0 . Then
the covering spaces (X, p) and (Y, q) are isomorphic if and only if p∗ π1 (X, x 0 ) and
q∗ π1 (Y, y0 ) are conjugate subgroups of π1 (B, b0 ) (i.e., iff they determine the same
conjugacy class of subgroups of π1 (B, b0 )).
Proof Suppose that the covering spaces (X, p) and (Y, q) are isomorphic. Then there
exists a homeomorphism h : Y → X such that p ◦ h = q i.e., making the diagram
in Fig. 4.17 commutative.
Let h(y0 ) = x 1 . Then h induces an isomorphism h ∗ : π1 (Y, y0 ) → π1 (X, x1 ) ⇒
h ∗ (π1 (Y, y0 )) = π1 (X, x1 ) ⇒ ( p∗ ◦ h ∗ )(π1 (Y, y0 )) = p∗ (π1 (X, x1 )). Hence q∗ (π1
(Y, y0 )) = p∗ π1 (X, x1 ). By Theorem 4.5.7, p∗ π1 (X, x1 ) is a subgroup of π1
(B, b0 ) and conjugate to the subgroup p∗ π1 (X.x 0 ). Consequently, p∗ π1 (X, x0 ) and
q∗ π1 (Y, y0 ) are conjugate subgroups of π1 (B, b0 ). For the converse, let the two
subgroups of π1 (B, b0 ) be conjugate. By Theorem 4.5.7 we can choose a differ-
ent base point y0 in Y such that the two groups are equal. We now consider the
diagram in Fig. 4.18 where q is a covering map. The space X is path-connected;
it is also locally path-connected, being locally homeomorphic to B. Moreover,
p∗ π1 (X, x0 ) ⊆ q∗ π1 (Y, y0 ). In fact, these two groups are equal. By Theorem 4.5.7,
we can lift the map p to p̃ : X → Y such that p̃(x 0 ) = y0 . Then q ◦ p̃ = p.
Reversing the role of X and Y in this discussion, we see that q : Y → B can also
be lifting to q̃ : Y → X such that q̃(y0 ) = x0 as shown in Fig. 4.19.
We claim that p̃ and q̃ as shown in Fig. 4.20 are inverses of each other. Consider
the diagram in Fig. 4.21.
Fig. 4.17 Isomorphisms of h /X

Y II
covering spaces II
II
II
II
II p
q III
II
II
II
$
B
Fig. 4.18 Lifting of p :Y

uu
uu
uuu
p̃ uuu
u q
uu
uuu
uu
u
uu
X /B
p
u: X
Fig. 4.19 Lifting of q
uu
uuu
q̃ uuu
u
uu p
uuu
uuu
u
uu
Y /B
q
Fig. 4.20 Liftings of p and > X

q̃ ~~
q ~~
~~~
~
> Y p
p̃ ~~
~~
~~~
~
X /B
p
Fig. 4.21 Lifting of p :X

uu
involving p̃ and q̃ uu
uuu
q̃◦p̃ uuu
u p
uu
uuu
uu
u
uu
X /B
p
Now q̃ ◦ p̃ : X → X is a lifting of the map p : X → B satisfying the condition

(q̃ ◦ p̃)(x 0 ) = x 0 . The identity map 1 X : X → X is another such lifting of p. Hence
by uniqueness of lifting it follows that q̃ ◦ p̃ = 1 X . Similarly, p̃ ◦ q̃ = 1Y . Conse-
quently, p̃ : X → Y is a homeomorphism and hence the covering spaces (X, p) and
(Y, q) are isomorphic. ❑
Remark 4.5.10 For any covering space (X, p) of B, the subgroups { p∗ (π1 (X, x)) :
x ∈ p −1 (b)} form a conjugacy class of subgroups of π1 (B, b). The above
Theorem 4.5.9 shows that a conjugacy class of a subgroup of π1 (B, b) determines
completely the covering spaces upto isomorphisms.
Recall that
Definition 4.5.11 A topological space X is said to be simply connected if it is path-

connected and π1 (X, x 0 ) = 0 for some x 0 ∈ X (hence for every x0 ∈ X ).
Example 4.5.12 Consider the covering spaces of S 1 . π1 (S 1 , 1) is abelian ⇒ two

subgroups of π1 (B, b0 ) are conjugate if and only if they are equal. Consequently,
two covering spaces of S 1 are isomorphic if and only if they correspond to the same
subgroup of π1 (S 1 ) ∼
= Z. The subgroups of Z are given by < n >, consisting of all
multiples of n, for n = 0, 1, 2, . . . . The covering space (R, p) of S 1 corresponds
to the trivial subgroup of Z, because R is simply connected. On the other hand,
the covering space (S 1 , p) of S 1 defined by p(z) = z n corresponds to the subgroup
< n > of Z. We conclude that every path-connected covering space of S 1 is isomor-
phic to one of these coverings i.e., any covering space of S 1 must be isomorphic
either to (R, p) or to one of the coverings (S 1 , qn ), where qn (z) = z n , z ∈ S 1 wraps
S 1 around itself n times.
u: X
Fig. 4.22 Lifting of f to X
u
u
f˜
u
u
u p
u
u
u
u
u
Y /B
f
Example 4.5.13 Consider the double covering (S 1 , p) over R P 2 . Since S 2 is simply

connected, π1 (S 2 , s) = 0 and hence the conjugacy class contains only the trivial
subgroup.
Example 4.5.14 The plane R2 is simply connected. Consequently, the conjugacy
class of (R2 , r ) over the torus also contains only the trivial subgroup.
Example 4.5.15 Let X denote an infinite spiral and let q : X → S 1 denote the pro-
jection map projecting each point on X to the point on the circle directly beneath
it. Then (X, q) is a covering space of S 1 . Since X is contractible, it has trivial
fundamental group. Consequently, (X, q) determines the conjugacy class of π1 (S 1 )
consisting of only the trivial subgroup. (R, p) also determines the conjugacy class of
π1 (S 1 ) consisting of only the trivial subgroup. We conclude that (X, q) and (R, p)
are isomorphic covering spaces of S 1 by Theorem 4.5.9.
We recall the following proposition (see Sect. 4.1.2).
Proposition 4.5.16 Let p : (X, x0 ) → (B, b0 ) be a covering space. Then the induced
homomorphism p∗ : π1 (X, x0 ) → π1 (B, b0 ) is a monomorphism and the subgroup
p∗ (π1 (X, x0 )) in π1 (B, b0 ) consists of homotopy classes of loops in B based at b0
which lift to X starting at x0 are loops.
Remark 4.5.17 If p : X → B is a covering map, then p is also onto. But its induced
homomorphism
p∗ : π1 (X, x 0 ) → π1 (B, b0 )
need not be an epimorphism. However, p∗ is a monomorphism.

Proposition 4.5.18 Let p : (X, x0 ) → (B, b0 ) be a covering space and f : (Y, y0 )
→ (B, b0 ) be a map, where Y is path-connected and locally path-connected. Then
a lift f˜ : (Y, y0 ) → (X, x 0 ) of f (as shown in Fig. 4.22) exists iff f ∗ (π1 (Y, y0 )) ⊂
p∗ (π1 (X, x0 )).
Proof Since f ∗ ([α]) = ( p ∗ ◦ f˜∗ )[α] ∈ p∗ (π1 (X, x0 )), ∀ [α] ∈ π1 (Y, y0 ), it follows
that f ∗ (π1 (Y, y0 )) ⊂ p∗ (π1 (X, x0 )).
Conversely, let y ∈ Y and β be a path in Y from y0 to y. Then the path f ◦ β in
B starting at b0 has a unique lifting ( f ◦ β) starting at x0 . Define
f˜ : (Y, y0 ) → (X, x 0 ), y → (
f ◦ β)(1).
Clearly, f˜ is well defined and continuous. ❑

Proposition 4.5.19 Given a covering space p : X → B and a map f : Y → X with

two liftings f˜1 , f˜2 : Y → X that agree at some point of Y , if Y is connected, then
f˜1 = f˜2 , i.e., f˜1 (y) = f˜2 (y), ∀ y ∈ Y.
Proof Let y ∈ Y and U be an open neighborhood of f (y) in B such that p−1 (U ) is

a disjoint union of open sets U i each of which is mapped homeomorphically onto U
by p. Suppose U 1 and U2 are the Ui ’s containing f˜1 (y) and f˜2 (y), respectively. By
continuity of f˜1 and f˜2 there is neighborhood N y of y mapped into U 1 by f˜1 and U
2
˜ ˜ ˜
by f 2 . If f 1 (y) = f 2 (y), then U1 = U2 . Hence U1 and U2 are disjoint open sets and
f˜1 = f˜2 throughout the neighborhood N y . Again if f˜1 (y) = f˜2 (y), then U 1 = U2
˜ ˜ ˜ ˜
and hence f 1 = f 2 on N y , because p ◦ f 1 = p ◦ f 2 and p is injective on U1 = U2 .
This shows that the set of points where f˜1 and f˜2 agree is a both open and closed set
in Y . ❑
4.5.3 Classification of Covering Spaces and Galois

Correspondence
This subsection considers the problem of classifying all different covering spaces
of a fixed base space B. The main thrust of this classification is given in the Galois
correspondence between connected covering spaces of B and subgroups of π1 (B).
The Galois correspondence ψ arises from the function that assigns to each covering
space p : (X, x0 ) → (B, b0 ) the subgroup p∗ (π1 (X, x 0 )) of π1 (B, b0 ). By Proposi-
tion 4.5.16, this correspondence ψ is injective. To show that ψ is surjective, we have
to show that corresponding to each subgroup G of π1 (B, b0 ), there is a covering
space p : (X, x 0 ) → (B, b0 ) such that p∗ π1 (X, x0 ) = G.
Definition 4.5.20 A topological space X is said to be semilocally simply connected

if each point x ∈ X has a neighborhood Ux such that the map induced by inclusion
i : Ux → X is trivial, i.e., i ∗ : π1 (Ux , x) → π1 (X, x) is trivial (equivalently, every
closed path in Ux at x is nullhomotopic in X ).
Definition 4.5.21 A topological space X is said to be semilocally path-connected if

for every point x ∈ X , there is an open neighborhood Ux of x such that every closed
path in Ux at x is nullhomotopic in X .
Theorem 4.5.22 Let B be a path-connected, locally path-connected and semilocally

path space. Then for each subgroup G of π1 (B, b0 ) there is a covering space p :
X G → B such that p∗ (π1 (X G , x0 )) = G for some suitable chosen base point x 0 ∈X G .
Proof Let b ∈ B. Since B is semilocally path-connected, there is an open neighbor-

hood Wb of b such that every closed path in Wb at b is nullhomotopic in B. Again
since X is locally path-connected, ∃ an open connected neighborhood Ub of b such
that b ∈ Ub ⊂ Wb . Clearly, every closed path in Ub at b is null homotopic in B and
Ub is evenly covered by p.
Construction of X G : Let P(B, b0 ) be the family of all paths f in B with f (0) =

b0 , topologized by the compact open topology. Define a binary relation f 1 ∼ f 2 mod
G iff f 1 (1) = f 2 (1) and [ f 1 ∗ f 2−1 ] ∈ G. Then ‘∼’ is an equivalence relation. The
equivalence class of f ∈ P(B, b0 ) is denoted by [ f ]. Let X G denote the set of all such
equivalence classes, topologized by the quotient topology. If c0 is the constant path
at b0 , define x 0 = c0 G ∈ X G and p : X G → B, [ f ]G → f (1). Then p(x 0 ) = b0 .
Since any two paths in the basic neighborhoods U[ f1 ]G and U[ f 2 ]G are identified in
X G , the whole neighborhoods are identified. Consequently, the natural projection p :
X G → B is a covering space with p(x0 ) = b0 . Then the image of p∗ : π1 (X G , x 0 ) →
π1 (B, b0 ) is precisely G. Because, for any loop β in B based at b0 , its lifting to X G
starting at x0 = c0 G ends at [β]G and hence the image of this lifted path in X G is
a loop iff [β]G ∼ [c0 ]G (equivalently, [β] ∈ G). ❑
Remark 4.5.23 Every group G can be realized as the fundamental group of the
topological space X G .
Corollary 4.5.24 Let B be a connected, locally path-connected, semilocally simply

connected space. Then every covering space q : Y → B is isomorphic (equivalent)
to a covering spaces of the form p : X G → B.
Proof Let b0 ∈ B be a base point of B and y0 ∈ Y lie in the fiber over b0 . If G =

q∗ π1 (Y, y0 ), then p∗ π1 (X G , x0 ) = G. Hence Theorem 4.5.9 shows that the covering
spaces p : X G → B and q : Y → B are isomorphic. ❑
Corollary 4.5.25 Let B be a connected, locally path-connected, semilocally simply

connected space. If p : X → B is a covering space of B, then every open contractible
set V in B is evenly covered by p.
Proof Since if V is an open path-connected set in B for which every closed path
in V is nullhomotopicin B, then V is evenly covered by p. In particular, if b ∈
V, then p −1 (V ) = (V, x) and contractible open sets are evenly covered in
x∈ p −1 (b)
every covering space of the form p : X G → B. Then the corollary follows from
Corollary 4.5.24. ❑
Corollary 4.5.26 Let B be a connected and locally path-connected space. Then B

has a universal covering space X (i.e., X is simply connected) iff X is semilocally
simply connected.
Proof Theorem 4.5.22 proves sufficiency of the condition. Definition 4.5.20 gives
the necessity of the condition. ❑
Theorem 4.5.27 (Classification theorem) Let B be a path-connected and locally

path-connected space. Then the two coverings p : X → B and q : Y → B are iso-
morphic via a homeomorphism f : X → Y taking a base point x 0 ∈ p −1 (b0 ) to a
base point y0 ∈ q −1 (b0 ) iff p∗ (π1 (X, x 0 )) = q∗ (π1 (Y, y0 )).
Fig. 4.23 Diagram for two f

/Y
X@
isomorphic coverings of B @@ ~~
@@ ~
@@ ~~
@@ ~~
p @@@ ~~~q
@@ ~~
@@ ~~
~ ~
B
Proof Suppose there is a homeomorphism f : (X, x0 )→(Y, y0 ) as shown in Fig. 4.23.

Then the two relations p = q ◦ f and q = p ◦ f −1 show that p∗ (π1 (X, x0 )) =
q∗ (π1 (Y, y0 )).
For the converse, let p∗ (π1 (X, x0 )) = q∗ (π1 (Y, y0 )). Then by the lifting criterion,
we may lift p to p̃ : (X, x0 ) → (Y, y0 ) with q ◦ p̃ = p. Similarly, we obtain q̃ :
(Y, y0 ) → (X, x 0 ) with p ◦ q̃ = q. Then by unique lifting property, it follows that p̃ ◦
q̃ = 1d and q̃ ◦ p̃ = 1d , since these composed lifts fix the base points. Consequently,
p∗ and q∗ are inverse isomorphisms. ❑
Remark 4.5.28 We now present a generalization of the above classification theorem

in the following form.
Theorem 4.5.29 (Classification theorem in general form) Let B be a path-connected,

locally path-connected and semilocally simply connected space. Then there exists a
bijection between the set of base point preserving isomorphism classes of path-
connected covering spaces p : (X, x 0 ) → (B, b0 ) and the set of subgroups of
π1 (B, b0 ), obtained by assigning the subgroups p∗ (π1 (X, x 0 )) to the covering spaces
(X, x 0 ). If the base points are ignored, this correspondence gives a bijection between
isomorphism classes of path-connected covering spaces p : X → B and conjugacy
classes of subgroups of π1 (B, b0 ).
Proof The first part follows from Theorem 4.5.27. For the proof of the second part,
we claim that covering space p : X → B, changing the base point x0 within π −1 (b0 )
corresponds exactly to changing p∗ (π1 (X, x 0 )) to a conjugate subgroup of π1 (B, b0 ).
Suppose x 1 is another base point p −1 (b0 ). Let α̃ is a path from x 0 to x1 . Then α̃
projects to a loop α in B, which represents some element g ∈ π1 (B, b0 ). Define G i
by G i = p∗ (π1 (X, xi )) for i = 0, 1. Then we have an inclusion g−1 G 0 g ⊂ G 1 , since
for f˜ a loop at x 0 ,
γ −1 ∗ f ∗
γ −1 is a loop at x 1 . Similarly, gG 1 g −1 ⊂ G 0 . Using
conjugation the latter relation by g −1 we have G 1 ⊂ g −1 G 0 g and hence g −1 G 0 g =
G 1 . Consequently, changing the base point from x0 to x 1 changes G 0 to the conjugate
subgroup G 1 = g −1 G 0 g. Conversely, to change G 0 to a conjugate subgroup G 1 =
g −1 G 0 g, choose a loop β represents g, that lifts to a path β̃ starting at x0 and let
x1 = β̃(1). The earlier argument proves that G 1 = g −1 G 0 g. ❑
Theorem 4.5.30 (Galois correspondence) Let B be path-connected and locally

path-connected space. The Galois correspondence ψ arising from the function that
assigns to each covering space p : (X, x0 ) → (B, b0 ) the subgroup p∗ (π1 (X, x 0 ))
of π1 (B, b0 ) is a bijection.
Proof ψ is injective: it follows from Proposition 4.5.16.

ψ is surjective: it follows from classification Theorem 4.5.27, since to each sub-
group G of π1 (B, b0 ), there is a covering space p : (X, x 0 ) → (B, b0 ) such that
p∗ π1 (X, x0 ) = G.
Hence this correspondence ψ is a bijection. ❑
Definition 4.5.31 ψ defined in Theorem 4.5.30 is called a Galois correspondence.
4.6 Universal Covering Spaces and Computing π1 (R P n )
This section introduces the concept of a special class of covering spaces, called
universal covering spaces and studies them with the help of fundamental groups of
their base spaces and computes π1 (R P n ).
4.6.1 Universal Covering Spaces
This subsection opens with the concept of universal covering spaces. For a topological
space B, (B, 1 B ) is a covering space over B. This covering space does not create
in general much interest because it corresponds to the conjugacy class of the entire
fundamental group π1 (B, b). On the other hand, the covering space corresponding
to the conjugacy class of the trivial subgroup {0} of π1 (B, b) is interesting. This
covering space, if it exists for some B, is called the ‘universal covering space’.
We now examine the relation between a base space B and its universal covering
space.
Definition 4.6.1 Let B be a topological space. A covering space (X, p) of B for
which X is simply connected (i.e., X is path-connected and π1 (X, x0 ) = 0 for every
x0 ∈ X ) is called the universal covering space of B.
Remark 4.6.2 We now explain the name of the term “universal covering space”.
Theorem 4.6.3 (i) Any two universal covering spaces of the same base space B
are isomorphic.
(ii) If (X, p) is the universal covering space of B and (Y, q) is a covering space of
B, then there is a continuous map
p̃ : X → Y
such that (X, p̃) is a covering space of Y .

Proof (i) Any universal covering space of B determines the conjugacy class of the
trivial subgroup ⇒ any two universal covering spaces of B are isomorphic by
Theorem 4.5.9.
4.6 Universal Covering Spaces and Computing π1 (R P n ) 171
Fig. 4.24 Lifting of p to p̃ :Y

uu
uu
uuu
p̃ uuu
u q
uu
uuu
uu
u
uu
X /B
p
Fig. 4.25 Infinite earring
(ii) We consider the commutative diagram in Fig. 4.24 and choose base points
x0 ∈ X, b0 ∈ B and y0 ∈ Y such that p(x0 ) = q(y0 ) = b0 . Since π1 (X, x0 ) = 0,
p∗ π1 (X, x0 ) ⊂ q∗ π1 (Y, y0 ). Hence Lifting Theorem 4.4.3 shows the existence
of a continuous map p̃ : (X, x 0 ) → (Y, y0 ) such that q ◦ p̃ = p and therefore p̃
is a covering projection. In other words, (X, p̃) is a covering space of Y .
❑
Example 4.6.4 (i) (R, p) is the universal covering space of S 1 , where p(t) = e2πit ,
since the space of real numbers R is simply connected.
(ii) (R 2 , r ) (in Example 4.5.14) is a universal covering space over the torus, since
R2 is simply connected.
(iii) (S 2 , p) is the universal covering space of R P 2 .
(iv) (S n , pn ) is a universal covering space of R P n , where pn : S n → S n is the map
identifying antipodal points of S n for n > 1 (see Theorem 4.1.19).
Remark 4.6.5 A space may not have a universal covering. We now present an exam-
ple of a space which has no universal covering.
Example 4.6.6 (Infinite earring or shrinking wedge of circles) Let Cn be the circle
of radius 1/n in R2 with center at (1/n, 0), for each n ≥ 1. Let X be the subspace
of R2 that is the union of these circles as shown in Fig. 4.25.
Then X is the union of a countably infinite collection of circles. The space X is
called the ‘infinite earring’ or ‘shrinking wedge of circles’ in the plane R2 . Let b0 the
origin. We claim that if U is a neighborhood of b0 in X , then the homomorphism of
Fig. 4.26 Homomorphisms j∗

π1 (Cn , bI0 ) / π1 (X, b0 )
induced by inclusion maps II u:
II u uu
II u
II uuu
II uu
I uu
k∗ III uu i∗
II uu
II uu
I$ uu
π1 (U, b0 )
fundamental groups induced by the inclusion i : U → X is not trivial. To show this,

let n be a given integer, there is a retraction r : X → Cn defined by letting r maps
each circle Ci for i = n to the point b0 . We can choose n sufficiently large such that
inclusion j : U → X and inclusion k : U → U and thus for sufficiently large n, Cn
lies in U . Then in the commutative diagram of groups and homomorphisms induced
by inclusions k∗ and j∗ as shown in Fig. 4.26, j∗ is injective.
Hence j∗ can not be trivial. This asserts that X has no universal covering.
4.6.2 Computing π1 (R P n )
We now present an interesting result of the universal covering space and utilize this
result to compute π1 (R P n ). For an alternative method see Corollary 4.10.4.
Theorem 4.6.7 Let (X, p) be the universal covering space of B and Aut (X/B) be
the group of all automorphisms of (X,B). Then the automorphism group Aut (X/B)
is isomorphic to the fundamental group π1 (B) of B. Moreover, if |π1 (B)| is the order
of the group π1 (B), then |π1 (B)|=number of sheets of the universal covering space.
Proof To prove the first part, let x 0 ∈ X and p(x 0 ) = b0 . We define a map ψ :
Aut (X/B) → π1 (B, b0 ) as follows:
f ∈ Aut (X/B) ⇒ f permutes the points of the fiber p−1 (b0 ). The point f (x 0 ) ∈
p −1 (b0 ), since ( p ◦ f )(x0 ) = b0 . Let u be the path in X joining x 0 and f (x0 ). Then
p ◦ u is a loop in B based at b0 . We define a mapping ψ : Aut (X/B) → π1 (B) given
by ψ( f ) = [ p ◦ u].
ψ is well defined: Let v be any other path joining x0 and f (x0 ). Since X is simply
connected, u is equivalent to v and hence [ p ◦ u] = [ p ◦ u] ⇒ ψ is well defined.
ψ is a homomorphism: Let f, g ∈ Aut (X/B) and u, v be two paths in X joining
x0 to f (x0 ) and to g(x0 ), respectively. Then ψ( f ) = [ p ◦ u] and ψ(g) = [ p ◦ v].
Clearly, f ◦ v is a path joining f (x0 ) to f (g(x 0 )) and hence u ∗ ( f ◦ v) is a path
in X joining x0 to f (g(x0 )). Again ψ( f g) = [ p ◦ (u ∗ ( f ◦ v))] = [( p ◦ u) ∗ ( p ◦
f ◦ v)] = [ p ◦ u][ p ◦ f ◦ v]. Since p ◦ f = p, we have ψ( f g) = [ p ◦ u ∗ p ◦ v] =
[ p ◦ u][ p ◦ v] = ψ( f )ψ(g).
ψ is a monomorphism: Let ψ( f ) = ψ(g). Then [ p ◦ u] = [ p ◦ v], where u, v are
paths in X starting at x 0 and ending at f (x 0 ) and g(x0 ), respectively. Consequently,
4.6 Universal Covering Spaces and Computing π1 (R P n ) 173
Fig. 4.27 Commutativity of (X, α̃(1))

8
the triangle for lifting of p to qqq
qqqqq
h h qqq
qqq p
qqqqq
qqq
q
(X, x0 ) / (B, b0 )
p
p∗ [u] = p∗ [v] ⇒ u and v must have the same terminal point by Monodromy The-
orem 4.9.3 i.e., f (x0 ) = g(x0 ) and hence f = g by Proposition 4.5.6, since X is
connected.
ψ is an epimorphism: Let α ∈ π1 (B, b0 ) and α̃ be the unique lifting of the path α in
X such that α̃(0) = x 0 ∈ X . Consider the commutative diagram in Fig. 4.27 obtained
by applying Lifting Theorem 4.4.3 to define a continuous lifting h of p such that
h(x 0 ) = α̃(1).
Since X is a simply connected covering space of B, there exists a homeomorphism
h : X → X such that h(x0 ) = α̃(1). By the same argument, there is also a homeomor-
phism k : X → X such that k(α̃(1)) = x0 . Since the homeomorphism k ◦ h : X →
X maps x0 to itself and hence by Proposition 4.5.6 it follows that h ◦ k = 1 X . This
implies that h ∈ Aut (X/B) and by definition, ψ(h) = [ p ◦ α̃] = [α]. This shows
that ψ is an isomorphism.
Proof of the last part: Since ψ is one-to-one, it establishes a one-to-one corre-
spondence between p −1 (b0 ) and a subset of π1 (B, b0 ). While proving ψ is onto, we
showed that every homotopy class [α] in π1 (B, b0 ) corresponds to a point α̃(1) in
p −1 (b0 ). Hence it follows that | p −1 (b0 )| = number of sheets of (X, p), is the order
of π1 (B, b0 ). ❑
Remark 4.6.8 Last part of the Theorem 4.6.7 also follows from Theorem 4.10.1(iii),
since π1 (X ) = 0.
Theorem 4.6.9 π1 (R P n ) Z2 for n ≥ 2.
Proof Consider the universal covering space (S n , q) of R P n where q identifies the

antipodal points of S n . Then |π1 (R P n )| = 2 ⇒ π1 (R P n ) ∼
= Z2 . ❑
Theorem 4.6.10 The automorphism group G = Aut (X/B) of a universal covering

space (X, p) of B acts on X freely.
Proof It is sufficient to prove that if g ∈ G and g(x) = x for some x ∈ X , then

g = 1 X . The group homeo(X ) of all homeomorphisms of a space X acts on the set
X by the action defined by g · x = g(x), where g ∈ Homeo (X ) and x ∈ X . Since
the group A(X/B) is a subgroup of Homeo (X ), Aut (X/B) also acts on the space
X by the above action. Thus g, 1 X : X → X are two covering homomorphisms of
corresponding covering projections such that g(x) = x = 1 X (x) for some x ∈ X .
This shows that g = 1 X by Proposition 4.5.6, since every path-connected space is
connected. ❑
Fig. 4.28 Commutativity of

the triangle for the covering
space (X, p)
Corollary 4.6.11 Let (X, p) be covering space of B.

(i) If h ∈ Cov (X/B) = Aut (X/B) and h = 1 X , 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 with h(x) = x; let b = p(x). Consider the commutative
diagram in Fig. 4.28.
Since both h and 1 X complete the diagram in Fig. 4.28, it follows that h = 1 X ,
a contradiction.
(ii) The map h −1 g ∈ Aut (X/B) has a fixed point, namely x and so by (i) h −1 g =
1 X ⇒ h = g. ❑
4.7 Fibrations and Cofibrations
This section gives a systematic approach to the lifting and extension problems through
representation of maps as fibrations or cofibrations which are dual concepts of each
other in some sense and form two important classes of maps in algebraic topology.
They are central concepts in homotopy theory. Every continuous map is equivalently
expressed up to homotopy as a fibration and also as a cofibration. The concept of
fibration first appeared in 1937 implicitly in the work of K. Borsuk (1905–1982).
This concept born in geometry and topology provides important strong mathematical
tools to invade many other branches of mathematics. More precisely, this section
introduces the concepts of fibrations and cofibrations and establishes a connection
between a fibration and a covering projection.
The concept of homotopy lifting property (HLP) is very important in algebraic
topology, specially in homotopy theory. It is the dual concept of the homotopy exten-
sion property (HEP). The concept of HLP leads to the concept of fibration. There is
a dual theory to fibration leading to the concept of cofibration. This is a very nice
duality principle in homotopy theory.
4.7.1 Homotopy Lifting Problems
This subsection discusses homotopy lifting problems of a map. It is an important

problem of algebraic topology and dual to the extension problem. Let p : X → B
4.7 Fibrations and Cofibrations 175
u: X
Fig. 4.29 Lifting of f
u
u
f˜
u
u
u p
u
u
u
u
u
Y /B
f
Fig. 4.30 Homotopy Lifting f

Y ×0 /
s9 X
of H s
s
s
H̃ s s
∩ p
s
s
s
s
s
Y ×I /B
H
Fig. 4.31 Homotopy Lifting f

Y ×0 /
s9 X
Problem s
s
s
H̃ s s
h0 p
s
s
s
s
s
Y ×I /B
H
be a map and Y be a space. If f : Y → B is a map, then the lifting problem for f is

to determine whether there is a continuous map f˜ : Y → X such that the diagram
in Fig. 4.29 is commutative, i.e., f = p ◦ f˜. If there exists such a map f˜ : Y → X,
we say that f can be lifted to X , and f˜ is called a lifting or a lifting of f . To show
that the lifting problem is a problem in the homotopy category, we need the concept
of homotopy lifting property(HLP) which is similar to the concept of HEP.
Definition 4.7.1 A continuous map p : X → B is said to have the HLP with respect
to a space Y , if given maps f : Y → X and H : Y × I → B such that H (y, 0) =
p f (y) for all y ∈ Y , there is a continuous map H̃ : Y × I → X such that H̃ (y, 0) =
f (y) for all y ∈ Y and H = p ◦ H̃ . If f is regarded as a map of Y × 0 to X , the
existence of H̃ is equivalent to the existence of a map represented by the dotted arrow
that makes the diagram in Fig. 4.30 commutative.
Let p : X → B be a map and Y be a space. A homotopy lifting problem is some-
times symbolized by the commutative diagram in Fig. 4.31 where h 0 (y) = (y, 0) for
all y ∈ Y and the maps f : Y → X, H : Y × I → B are said to constitute the data
for the problem in question. The map H is a homotopy of p ◦ f and a solution to
the problem is a homotopy H̃ : Y × I → X of f such that p ◦ H̃ = H . Thus H̃ lifts
the homotopy of H of p ◦ f to a homotopy of f .
Proposition 4.7.2 Let p : X → B has the HLP with respect to a space Y . If f

g : Y → B, then f can be lifted to X iff g can be lifted to X .
Proof Similar to the proof of Corollary 3.3.4 of Chap. 3. ❑

Remark 4.7.3 Let p : X → B and f : Y → B be two continuous maps. Then f can

or cannot be lifted to X is a property of the homotopy class. This implies that the
lifting problem for maps f : Y → B to X is a problem of homotopy category.
4.7.2 Fibration: Introductory Concepts
This subsection introduces the concept of a fibration first implicitly appeared in

the work of K. Borsuk in 1937 but explicitly in the work of Whiteney during 1935–
1940, first on sphere bundles. Fibrations form an important class of maps in algebraic
topology. Covering map is a fibration. The homotopy lifting property leads to the
concept of fibration (or Hurewicz fiber space) (Hurewicz 1955). More precisely, a
continuous map p : X → B has the HLP with respect to a space Y if and only if
every problem symbolized by the commutative diagram in Fig. 4.31 has a solution.
Definition 4.7.4 A pointed continuous map p : X → B is called a fibration (or

fiber map or Hurewicz fiber space) if p has the HLP with respect to every space.
X is called the total space and B is called the base space of the fibration. For b ∈
B, p −1 (b) = F is called the fiber over b. A Serre fibration is map X → B satisfying
HLP with respect to disk D n , ∀ n. It is sometimes called a weak fibration.
p
We use the notation “F → X − → B is a fibration” to mean that p : X → B is a
fibration, F is the fiber space over some specific point of B, and i : F → X is the
inclusion map.
Example 4.7.5 The projection
p : B × F → B, (b, f ) → b
is a fibration.
Definition 4.7.6 A fibration p : X → B is called principal fibration if there is also a

space C and a map g : B → C and a homotopy equivalence (over B, i.e., commuting
through B) of X with mapping path space of g defined by
Pg = {(b, σ) ∈ B × C I : σg (0) = ∗, σg (1) = g(b)},
p1 : Pg → B, (b, σ) → b.
as shown in Fig. 4.32; C is called the classifying space and g is called the classifying
map for the principal fibration.
Theorem 4.7.7 Given a principal fibration p : X → B, a lifting f˜ of f exists if and

only if g ◦ f is homotopic to a constant map, where g : B → C is the classifying
map.
Fig. 4.32 Diagram of the f

Y ×0 /X o Pg
classifying space and {= ~
{ ~~
classifying map {
{
p1 ~~
~~
h0 { p ~~
{ H̃ ~~
{
{
~~~
{ ~ ~
Y ×I /B /C
H g
Fig. 4.33 Existence of p8 X

lifting for a principal f˜ p p
p p p
fibration pp f /

Y NNN B
NNN
NNN g
g◦f NNNN
&
C
Proof Since p is a principal fibration, there is a homotopy equivalence (over B)

X Pg and hence there exist maps
h : X → Pg and k : Pg → X
such that
k ◦ h 1 X and h ◦ k 1 Pg and p1 ◦ h = p, p ◦ k = p1 ,
where p1 : Pg → B, (b, σ) → b.
Given f˜ : Y → X , we obtain a homotopy g ◦ f c, where c : Y → C is the
constant map y → ∗ ∈ C as the composite
H : Y × I → Pg → C, (y, t) → (h ◦ f˜)(y) = ( f (y), σ y ) → σ y (t) = Ht (y)
as shown in Fig. 4.33.

Conversely, let G : g ◦ f c. Define
G y : I → C, t → G(y, t);
f˜ : Y → Pg → X, y → ( f (y), G y ) → k( f (y), G y ).
Hence
( p ◦ f˜)(y) = ( pok)( f (y), G y ) = p1 ( f (y), G y ) = f (y), ∀ y ∈ Y ⇒ p ◦ f˜ = f
❑
Theorem 4.7.8 A lifting f˜ of a principal fibration p : X → B exists iff there exists
a map g̃ : C f → C extending the classifying map g in the diagram as shown in
Fig. 4.34.
Fig. 4.34 Existence of a q8 X

map extending the q qq
f˜
p
q
classifying map g qq f
Y / B / Cf
q qq
q qg̃
g
xq q
C
Proof In the category Top ∗ of pointed topological spaces the mapping cone C f
is obtained from the mapping cylinder M f by identifying Y × {0} ∪ {∗} × I with
∗ in B.
Suppose there is a homotopy
H : c g ◦ f : Y → C,
where c : Y → ∗ ∈ C is the given constant map. Define

(y, t) → H (y, t),
g̃ : C f → C,
b → g(b) for b ∈/ f (Y ).
Then
g̃(y, 0) = ∗ and g̃(y, 1) = g f (y), ∀ y ∈ Y.
This shows that g̃ is the required extension of g.

Conversely, let g̃ be an extension of g. Then there is a homotopy
G : Y × I → C, (y, t) → g̃(y, t.)
Consequently,
G(y, 0) = g̃(y, 0) = ∗
and
G(y, 1) = g̃(y, 1) = (g ◦ f )(y), ∀ y ∈ Y.
Hence g ◦ f c. ❑
Proposition 4.7.9 Let p : X → B be a fibration and α be any path in B such that

α(0) ∈ p(X ). Then α can be lifted to a path α̃ in X .
Proof α can be regarded as a homotopy α : { p0 } × I → B, where { p0 } is a one-

point space. Let x 0 be a point in X such that p(x 0 ) = α(0). Then there exists a map
f : { p0 } → X such that p f ( p0 ) = α( p0 , 0). Hence it follows from the HLP of p
that there exists a path α̃ in X such that α̃(0) = x 0 and p ◦ α̃ = α. This shows that
α̃ is a lifting of α. ❑
Fig. 4.35 Trivial fibration f

Y /B×F
m6
H̃m mm p
h0 m
mm
Y ×I /B
H
Fig. 4.36 Homotopy lifting f

Y /
n7 X
property H̃ n nn
h0 n p
nn
Y ×I /B
H
Example 4.7.10 Let F be any space and p : B × F → B be the projection to the first
factor. Then p is a trivial fibration and for any b ∈ B, the fiber p−1 (b) over b is homeo-
morphic to F. Because, if the diagram in Fig. 4.35 symbolizes homotopy lifting prob-
lem, then the map H̃ : Y × I → B × F defined by H̃ (y, t) = (H (y, t), p f (y)) is a
solution of the lifting problem.
The projection p : B × I → B is said to be a trivial fibration.
Example 4.7.11 For any space X , let P(X ) = M(I, X ) be the space of all paths in
X . Then the map p : P(X ) → X × X , defined by p(α) = (α(0), α(1)) is a fibration.
Again
pi : P(X ) → X, α → α(0), α(1)
for i = 1, 2, respectively, are also fibrations.
Example 4.7.12 Let p : X → Y be a fibration and q : Y → B be also a fibration,

then their composite q ◦ p : X → B is also a fibration.
Theorem 4.7.13 Every covering projection is a fibration.
Proof Let p : X → B be a covering projection and the diagram in Fig. 4.36 sym-
bolizes a homotopy lifting problem. Then for each y ∈ Y , there exists a unique
path α y : I → X such that α y (0) = f (y) and pα y (t) = H (y, t). Then the map
H̃ : Y × I → X, (y, t) → α y (t) is a continuous map and p is a fibration. ❑
Remark 4.7.14 For a covering projection the lifting is unique but it is not true for an
arbitrary fibration.
4.7.3 Cofibration: Introductory Concepts
This subsection conveys the concept of cofibration and studies it in the category Top∗
of pointed topological spaces and pointed maps. Cofibrations form an important class
of maps in topology. Geometrically, the concept of cofibrations is less complicated
than that of fibrations. There is a very nice duality principle in homotopy theory.
Fig. 4.37 Commutative Ht

XA /Z
AA ?
triangle for cofibration AA ~~~
~
f AA ~
~~ H t
Y
Fig. 4.38 Cofibration of f h0

/A×I
A 88
f ~~~ G xxx 88
~ xx
~~ xx 88
~~ g |xx 88 f ×1
X XXXXXXXXX / Y iT 88 d
XXXXXX T T 88
XXXXXX T T FT 8
XXXXXX T T 88
h0 XXXXXXT 8
X,
X ×I
For example, if in the definition of a fibration as a map satisfying homotopy lifting

property we reverse the directions of all rows, we obtain the dual notion, called
a cofibration. This is a continuous map f : X → Y satisfying the property: given
g̃ : Y → Z in Top ∗ and a homotopy H t : Y → Z such that there is a continuous map

Ht : X → Z with the property Ht ◦ f = Ht , i.e., making the triangle in Fig. 4.37
commutative.
In this subsection, we work in Top∗ unless specified otherwise.
Definition 4.7.15 A continuous map f : A → X is said to be a cofibration if for

every topological space Y and given a continuous map g : X → Y and a homotopy
G : A × I → Y starting with g ◦ f , there exists a homotopy F : X × I → Y starting
with g, such that G = F ◦ ( f × 1d ) as shown in Fig. 4.38.
Remark 4.7.16 It follows from Definition 4.7.15 that if A is a subspace of X , the

inclusion map i : A → X is a cofibration if the pair (X, A) has the homotopy exten-
sion property (HEP) with respect to the given space Y .
Example 4.7.17 For any space A in Top∗ , let C A = A × I /A × {1} ∪ {∗} × I be

the cone of A and i : A → C A, a → [a, 0] be the inclusion. Then i is a cofibration.
Proposition 4.7.18 Given a map f : (X, x 0 ) → (Y, y0 ) in Top ∗ , the inclusion i :

Y → Y C X is a cofibration.
f

Proof Let r : I × I → I × {0} ∪ I˙ × I be a retraction. Then given maps f : Y
g
C X → Z in Top ∗ and G : Y × I → Z with G(y, 0) = f (y), ∀ y ∈ Y , define H :
C X × {0} ∪ X × I → Z by the rule H |C X ×{0} = f |C X and H | X ×I = G ◦ (g × 1d ).
Again define F : (Y C X ) × I by the rule F|Y ×I = G and F([s, x], t) = H ([ p1 ◦
g
r (s, t), x], p2 ◦ r (s, t)), ∀ [s, t] ∈ C X, t ∈ I , where p1 , p2 are the restrictions to I ×
{0} ∪ I˙ × I of the projections p1 , p2 : I × I → X . Then F is well defined and is a
continuous map such that F(y, 0) = f (y). ❑
Fig. 4.39 Diagram for f ×1d

X × I kVVVV /Y ×I
cofibration VVVV m m O
m VVV VVVV j0
H mm VVVV
vm mm G j0 VVVV
V
Zo g Y o
f
X
Proposition 4.7.19 If f : X → Y is a cofibration, then f is injective, and in fact it

is a homeomorphism onto its image.
Proof Consider the mapping cylinder M f of f , the quotient space of X × I ∪ Y in

which (x, 1) is identified with f (x). Let Ht : X → M f be the homotopy, mapping
x ∈ X to the image (x, 1 − t) ∈ X × I in M f , and let Ht : Y → M f be the inclusion.
Then the cofibration property of f shows that H t : Y → M f is such that H t ◦ f =
Ht . Restriction to a fixed t > 0, shows that f is injective, since Ht is so. Moreover,
as Ht is a homeomorphism onto its image X × {1 − t}, the relation H t ◦ f = Ht
holds. ❑
There is an equivalent definition of cofibration in Top ∗ .

Definition 4.7.20 A continuous map f : X → Y in Top ∗ is said to be a cofibration
if given a topological space Z , a continuous map g : Y → Z and a homotopy H :
X × I → Z starting from g ◦ f , there exists a homotopy G : Y × I → Z , starting
from g such that H = G ◦ ( f × 1d ) i.e., making the three triangles as shown in
Fig. 4.39 commutative, where j0 (y) = (y, 0), ∀ y ∈ Y and j0 (x) = (x, 0), ∀ ∈ X.
Remark 4.7.21 Let A be a subspace of a topological space X . Then the inclusion map
i : A → X is a cofibration if the pair (X, A) has the absolute homotopy extension
property (see Chap. 2). The converse is not true in general. Because the definition of
a cofibration refers to Top ∗ but the absolute homotopy extension property refers to
maps and homotopies that are not necessarily based.
Theorem 4.7.22 Every continuous map f : X → Y in Top ∗ is the composite of a

cofibration and a homotopy equivalence.
Proof Let M f be the mapping cylinder in Top ∗ obtained from Y and (X × I )/x0 × I
by identifying, for each x ∈ X , the points (x, 1) and f (x). Suppose g : X →
M f , x → [(x, 0)] is the inclusion map. Let h : M f → Y be the map induced by
the identity map 1Y of Y and the map from X × I to Y that sends each [(x, t)] to
f (x). Then f = h ◦ g. We claim that h is a homotopy equivalence and g is a cofi-
bration. We first show that g is a cofibration. Given a map k : M f → Z in Top ∗ , and
a homotopy H : X × I → Z starting from k ◦ g, define maps
G Y : Y × I → Z , (y, s) → k(y), ∀ s ∈ I

k(x, (2t − s)/(2 − s)), 0 ≤ s ≤ 2t
and G X : (X × I ) × I → Z , (x, t, s) →
H (x, s − 2t), 2t ≤ s ≤ 1.
Clearly, G X is continuous and G X (x, 1, s) = k(x, 1) = (k ◦ f )(x)=G Y ( f (x), s).

Hence G X and G Y give together a homotopy G : M f × I → Z such that G starts
from k and G ◦ (g × 1d )(x, s) = G(x, 0, s) = H (x, s). Hence G ◦ (g × 1d ) = H
shows that g is a cofibration.
Finally we show that h is a homotopy equivalence. Define j : Y → M f 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
H : M f × I → M f , (y, s) → y and (x, t, s) → (x, t + s(1 − t)).
Clearly, H is continuous and H : 1d j ◦ h. This shows that h is a homotopy equiv-

alence. ❑
Remark 4.7.23 The dual of the Theorem 4.7.22 is true in the sense that every con-
tinuous map f : X → Y in Top ∗ is also the composite of a homotopy equivalence
and a fiber map.
Theorem 4.7.24 Let A be a closed subset of a topological space X . Then the inclu-
sion i : A → X is a cofibration iff X × {0} ∪ A × I is a retract of X × I .
Proof If i : A → X is a cofibration, then the given continuous maps f : X →

X × {0} ∪ A × I, x → (x, 0) and G : A × I → X × {0} ∪ A × I, (a, t) → (a, t)
determine a map H : X × I → X × {0} ∪ A × I which is a retraction. Conversely,
suppose there is a retraction r : X × I → X × {0} ∪ A × I . Then given a space
Y , a map f : X → Y , and a homotopy G : A × I → Y with the property that
H (a, 0) = f (i(a)), ∀ a ∈ A, define a map

( f ◦ p X ◦ r )(x, t), if (x, t) ∈ r −1 (X × {0})
H : X × I → Y, (x, t) →
(G ◦ r )(x, t), if (x, t) ∈ r −1 (A × I ).
Since X × {0} and A × I are closed in X × I , it follows that H is continuous.

Consequently, i is a cofibration. ❑
4.8 Hurewicz Theorem for Fibration and Characterization

of Fibrations
This section continues the study of fibrations, characterizes path liftings of fibrations
with the help of their fibers and studies 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).
4.8 Hurewicz Theorem for Fibration and Characterization of Fibrations 183
Theorem 4.8.1 (Hurewicz) Let p : X → B be a continuous map. Suppose B is

paracompact and there is an open covering {Vi } of B such that, for each Vi , p| p −1
(Vi ) : p −1 (Vi ) → Vi is a fibration. Then p is a fibration.
Proof The proof is long and complicated. (Dugundji 1966) is referred. ❑
An immediate important consequence of Hurewicz Theorem 4.8.1 gives a suffi-

cient condition for a projection map p : X → B of a fiber bundle (Chap. 5) to be a
fibration.
Corollary 4.8.2 Let p : X → B be the projection of a fiber bundle, and suppose B

is paracompact, then p is a fibration.
Theorem 4.8.3 Let p : X → B be a covering projection and let f, g : Y → X be

liftings of the same map (i.e., p ◦ f = p ◦ g). If Y is connected and f (y0 ) = g(y0 )
for some point y0 of Y , then f = g.
Proof Let A = {y ∈ Y : f (y) = g(y)}. Then A = ∅ and A is an open set in Y . To

show this, let y ∈ A and U be an open neighborhood of p f (y) evenly covered by
p and let Ũ be an open subset of X containing f (y) such that p maps Ũ homeo-
morphically onto U . Then f −1 (Ũ ) ∩ g−1 (Ũ ) is an open subset of Y containing y
and contained in A. Again let B = {y ∈ Y : f (y) = g(y)}. If X is assumed to be
Hausdorff, then B is open in Y . Otherwise, let y ∈ B and U be an open neighborhood
of p f (y) evenly covered by p. Since f (y) = g(y), there are disjoint open sets Ṽ1
and Ṽ2 of X such that f (y) ∈ Ṽ1 and g(y) ∈ Ṽ2 and p maps each of the sets Ṽ1
and Ṽ2 homeomorphically onto U . Then f −1 (Ṽ1 ) ∩ g −1 (Ṽ2 ) is an open subset of Y
containing y and contained in B. Finally, Y = A ∪ B and A and B are disjoint open
sets imply from the connectedness of Y that either A = ∅ or A = Y . By hypothesis
A = ∅ and hence A = Y shows that f = g. ❑
Definition 4.8.4 A continuous map p : X → B is said to have a unique path lifting

if, given paths α and β in X such that p ◦ α = p ◦ β and α(0) = β(0), then α = β.
Proposition 4.8.5 If a continuous map p : X → B has unique path lifting property,

then it has path lifting property for path-connected spaces.
Proof Let p : X → B has unique path lifting property and Y be path connected
space. If f, g : Y → X and maps are such that p ◦ f = p ◦ g and f (y0 ) = g(y0 )
for some y0 ∈ Y, we claim that f = g. Let y ∈ Y and α be a path in Y from y0
to y. Then f ◦ α and g ◦ α are paths in X that are liftings of some path in B and
have the same initial point. Since p has unique path lifting, f ◦ α = g ◦ α and hence
f (y) = ( f ◦ α)(1) = (g ◦ α)(1) = g(y) implies f = g, since α(1) = y. ❑
We now characterize path liftings of fibrations with the help of their fibers.
Theorem 4.8.6 Let p : X → B be a fibration. Then the fibration has unique path
lifting iff every fiber has no nonconstant paths.
Proof Let p : X → B be a fibration with unique path lifting. Let α be a path in

the fiber p−1 (b) and β be the constant path in p−1 (b) such that β(0) = α(0). Then
p ◦ α = p ◦ β ⇒ α = β ⇒ α is a constant path. Conversely, let p : X → B be a
fibration such that every fiber has no nontrivial path. If α and β are paths in X such
that p ◦ α = p ◦ β and α(0) = β(0), then for t ∈ I define a path γt in X by

α((1 − 2t )t), 0 ≤ t ≤ 1/2
γt (t ) =
β((2t − 1)t), 1/2 ≤ t ≤ 1.
Then for each t ∈ I, γt : I → X is a path in X from α(t) to β(t) and p ◦ γt is a closed

path in B, which is homotopic rel I˙ to the constant path at pα(t). By HLP of p, there is
a map H : I × I → X such that H (t , 0) = γt (t ) and H maps 0 × I ∪ I × 1 ∪ 1 ×
I to the fiber p−1 ( pα(t)). Every Since p −1 ( pα(t)) has no nonconstant paths, F maps
0 × I, I × 1 and 1 × I to a single point. Hence it follows that F(0, 0) = F(1, 0).
Consequently, γt (0) = γt (1) and α(t) = β(t). ❑
Proposition 4.8.7 Let X be pointed topological space with base point x0 and P(X )
be the space of paths in X starting at x0 , then the map
p : P(X ) → X, α → α(1)
is fibration with fiber (X ).
Proof Let Y be an arbitrary space. Given maps f : Y → P(X ) and G : Y × I → X

with G 0 = p ◦ f : Y → X , define a function
H : Y × I × I → X,

( f (y))(s(t + 1)), 0 ≤ s ≤ t+1
1
(y, t, s) →
G(y, s(t + 1) − 1), t+1
1
≤s≤1
Then H is continuous and defines a map F : Y × I → X I such that
F(y, t)(0) = f (y)(0) = x0 , ∀ y ∈ Y, t ∈ I.
Hence F ∈ P(X ) and F(y, 0)(s) = f (y)s, ∀ y ∈ Y, ∀ s ∈ I . Consequently, F0 =

f and p ◦ F = G. This implies that F is the required lifting of G. Moreover,
p−1 (x0 ) = {α ∈ P(x) : α(1) = x 0 } = (X ). ❑
4.9 Homotopy Liftings and Monodromy Theorem
This section continues the study of covering spaces by presenting some interesting
applications of the path lifting property and homotopy lifting property of cover-
4.9 Homotopy Liftings and Monodromy Theorem 185
ing projections and proves Monodromy Theorem which provides a necessary and
sufficient condition for two liftings of a covering projection to be equivalent.
4.9.1 Path Liftings and Homotopy Liftings
This subsection discusses path lifting property of a covering projection by using the
homotopy lifting property.
Theorem 4.9.1 Let (X, p) be a covering space of B and b0 ∈ B. If x0 ∈ p −1 (b0 ),

then for any path f : I → B, with f (0) = b0 , there exists a unique path f˜ : I → X
such that f˜(0) = x 0 and p ◦ f˜ = f .
Proof Let A = {a} be a singleton space. We consider the map f : A → B defined

by f (a) = b0 . The path f defines a homotopy F : A × I → B on A given by
F(a, t) = f (t). Then by the Homotopy Lifting Property, ∃ a map F̃ : A × I →
X such that F̃(a, 0) = x0 and p ◦ F̃ = F. Consequently, f˜ : I → X defined by
f˜(t) = F̃(a, t), t ∈ I , is a path in X starting from x0 and having the property:
( p ◦ f˜)(t) = ( p ◦ F̃(a, t)) = F(a, t) = f (t), ∀ t ∈ I i.e., p ◦ f˜ = f.
Clearly the path f˜ : I → X is unique. ❑
Remark 4.9.2 If p : X → B is a covering map, then p is also onto. But its induced
homomorphism
p∗ : π1 (X, x 0 ) → π1 (B, b0 )
need not be a epimorphism. However, p∗ is a monomorphism.
4.9.2 Monodromy Theorem
This subsection gives a criterion for two path liftings in X to be equivalent through
a result known as ‘Monodromy theorem’.
Theorem 4.9.3 (The Monodromy Theorem) Let (X, p) be a covering space of B

and f˜, g̃ are paths in X with same initial point x 0 . Then f˜ and g̃ are equivalent (i.e.,
f˜ g̃ rel I˙) if and only if p ◦ f˜ and p ◦ g̃ are equivalent paths in B.
Proof Let f˜, g̃ be equivalent paths in X . Then ∃ a homotopy F : f˜ g̃ rel I˙ ⇒

p ◦ F : I × I → B is a continuous map such that p ◦ F : p ◦ f˜ p ◦ g̃ rel I˙ and
hence p ◦ f˜ and p ◦ g̃ are equivalent paths in B. Conversely, let p ◦ f˜ and p ◦ g̃ be
equivalent paths in B. Then ∃ a continuous map G : I × I → B such that G : p ◦
f˜ p ◦ g̃ rel I˙. By Homotopy Lifting Property, ∃ a unique homotopy G̃ : I × I →
X such that G̃(0, 0) = x0 and p ◦ G̃ = G. Restricting G̃ on (t, 0), t ∈ I , we have a
path t → G̃(t, 0), starting from x0 and lifting p ◦ f˜. Then t → f˜(t) is also a path in
X starting from x 0 and lifting p ◦ f˜. Hence the uniqueness property of the covering
paths, G̃(t, 0) = f˜(t), ∀ t ∈ I . Similarly, G̃(t, 1) = g̃(t). Again by restricting G̃ on
(0, s), s ∈ I , we have a path s → G̃(0, s), which projects under p to the constant path
at b0 = p(x0 ). A constant path s → x 0 in X also projects under p to the constant
path s → x0 in X . Hence by uniqueness theorem s → G̃(0, s) is a constant path
based at x0 . Similarly, the path s → G̃(s, t) is a constant path based at some point
x1 ∈ p −1 (b0 ). This shows that G̃ is a homotopy between f˜ and g̃ rel I˙. Consequently
f˜ and g̃ are equivalent paths in X . ❑
Corollary 4.9.4 Let (X, p) be a covering space of B and b0 ∈ B, x0 ∈ p −1 (b0 ).

Then the induced homomorphism p∗ : π1 (X, x0 ) → π1 (B, b0 ) is a monomorphism.
Proof Let [ f˜], [g̃] ∈ π1 (X, x 0 ) and [ f˜] = [g̃]. Then p∗ ([ f˜]) = [ p ◦ f˜] and p∗ ([g̃])
= [ p ◦ g̃]. Now p ◦ f˜ p ◦ g̃ rel I˙ ⇔ f˜ g̃ rel I˙. But f˜ g̃ rel I˙ ⇔ p ◦ f˜
p ◦ g̃ rel I , otherwise we arrive at a contradiction by Theorem 4.9.3. This shows that
p∗ is well defined and injective; hence p∗ is a monomorphism. ❑
4.10 Applications and Computations
This section presents applications of covering spaces and computes fundamen-

tal groups of some interesting spaces. Finally it 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, x 0 )) of π1 (B, b0 ).
4.10.1 Actions of Fundamental Groups
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.
Let (X, p) be a covering space of B and b0 ∈ B. We now consider the action of
the fundamental group π1 (B, b0 ) on the fiber p−1 (b0 ) = Y .
Theorem 4.10.1 Let (X, p) be a covering space of B and b0 ∈ B. Let Y = p −1 (b0 )

be the fiber over b0 . Then
(i) π1 (B, b0 ) acts transitively on Y ;
(ii) If x0 ∈ Y , then the isotropy group G x0 = p∗ π1 (X, x 0 ); and
(iii) |Y | = [π1 (B, b0 ) : p∗ π1 (X, x0 )].
Proof First we show that Y is a (right) π1 (B, b0 )-set. We define σ : Y × π1 (B, b0 ) →

Y by the rule σ(x, [ f ]) = x · [ f ] = f˜(1), where f˜ is the unique lifting of f :
(I, 0) → (B, b0 ) such that f˜(0) = x. This definition does not depend on the choice
4.10 Applications and Computations 187
of the representative of the class [ f ] by the Monodromy Theorem 4.9.3. If f is a con-

stant path at b0 , then f˜ is also a constant path at x ∈ Y . Hence x · [ f ] = f˜(1) = x.
Next suppose [ f ], [g] ∈ π1 (B, b0 ). Let f˜ be the lifting of f with f˜(0) = x and g̃
be 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). Consequently, x · [ f ∗ g] = (x · [ f ])[g]. As a result σ is an action
of π1 (B, b0 ) on Y .
(i) σ is transitive: Let x0 ∈ Y and x be any point in Y . Since X is path-connected,
∃ a path λ̃ in X from x0 to x. Then p ◦ λ̃ is a closed path in B at b0 whose lifting
with initial point x 0 is λ̃. Thus [ p ◦ λ̃] ∈ π1 (B, b) ) and x 0 · [ p ◦ λ̃] = λ̃(1) = x.
Hence it follows that π1 (B, b0 ) acts transitively on Y .
(ii) Let f be a closed path in B at b0 and f˜ be the lifting of f with f˜(0) = x0 . Let
G = π1 (B, b0 ). Then G x0 = {[ f ] ∈ π1 (B, b0 ) : x0 · [ f ] = x0 }. Hence [ f ] ∈
G x0 ⇒ x0 · [ f ] = x0 ⇒ f˜(1) = x 0 = f˜(0) ⇒ f˜ ∈ π1 (X, x0 ) and [ f ] = [ p ◦
f˜] ∈ p∗ π1 (X, x0 ) ⇒ π1 (B, b0 ) = G ⊆ p∗ π1 (X, x 0 ). For the reverse inclusion,
assume [ f ] = [ p ◦ g̃] for some [g̃] ∈ π1 (X, x 0 ). Then f˜ = g̃, since both are lift-
ings of f and both have initial point x 0 ⇒ f˜(1) = g̃(1) ⇒ x0 · [ f ] = f˜(1) =
x0 ⇒ [ f ] ∈ G x0 ⇒ p∗ π1 (X, x 0 ) ⊆ G x0 . Consequently, G x0 = p∗ π1 (X, x 0 ).
(iii) Recall that if a group G acts on a set Y and x 0 ∈ Y , then |orbit of x 0 | = [G : G x0 ].
In particular, if G acts transitively, |Y | = [G : G x0 ]. Hence in this case, G =
π1 (B, b0 ) and G x0 = p∗ π1 (X, x0 ) by (ii). Consequently, |Y | = [π1 (B, b0 ) :
p∗ π1 (X, x0 )]. ❑
Corollary 4.10.2 Let (X,p) be the universal covering space of B, then |Y | =
|π1 (B, b0 )|.
Proof The corollary follows from Theorem 4.10.1(iii), since π1 (X, x 0 ) = 0. ❑
Corollary 4.10.3 If n ≥ 2, then π1 (R P n ) ∼
= Z2 .
Proof Since (S n , p) is a covering space of R P n of multiplicity 2, it follows that
[π1 (R P n , x0 ) : p∗ π1 (S n , y0 )] = 2. Again, S n is simply connected for n ≥ 2 ⇒
p∗ π1 (S n , y0 ) = 0 ⇒ |π1 (R P n , x 0 )| = 2 ⇒ π1 (R P n , x0 ) ∼
= Z2 . ❑
Corollary 4.10.4 Let (X, p) be a covering space of b0 ∈ B, x0 ∈ p −1 (b0 ). If p∗ :
π1 (X, x0 ) → π1 (B, b0 ) is onto, then the map p : X → B induces an isomorphism
p∗ : π1 (X, x 0 ) → π1 (B, b0 ).
Proof p∗ is a monomorphism by Corollary 4.9.4 and hence the corollary follows

from the given condition. ❑
4.10.2 Fundamental Groups of Orbit Spaces
This subsection computes the fundamental groups of some important spaces which
are obtained as orbit spaces. For example, projective spaces, lens spaces, figure-
eight and Klein’s bottles are interesting spaces. We represent them as orbit spaces
and compute their fundamental groups. A topological group G with identity e acting
on a topological space X is said to satisfy the condition (A): if for each x ∈ X, ∃ a
neighborhood U x such that, g (Ux ) ∩ Ux = ∅ ⇒ g = e, where
g : X → X, x → gx
is a homeomorphism. This special kind of group action of the group G of home-

omorphisms of X is said to act on X properly discontinuously. For example, any
action of a finite group on a Hausdorff space is properly discontinuous.
Example 4.10.5 The automorphism group Aut (X/B) of (X, p) of B satisfies the
condition (A).
Definition 4.10.6 A covering space (X, p) of B is said to be regular if p∗ π1 (B, b0 )

is a normal subgroup of π1 (B, b0 ).
Example 4.10.7 Let B be a connected, locally path-connected space and G satisfies

the condition (A) on X then (X, p) is a regular covering space of
X mod G, where p : X → X mod G, x → Gx
is the natural projection.
Theorem 4.10.8 If an action of a topological group G on a topological space X

satisfies the condition (A), then
(i) if X is path-connected, then G is the group of deck transformations of the
covering space
p : X → X mod G, x → Gx
(ii) if X is path-connected and locally path-connected, then 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 iso-
morphic.
Proof (i) Let X be path-connected. The deck transformation group contains G as

a subgroup and equals this group, since if f is any deck transformation, then
given any point x ∈ X, x and f (x) are in the same orbit and hence there is some
g ∈ G such that g(x) = f (x). Consequently, f = g, since deck transformation
of a connected covering space are uniquely determined under this situation.
(ii) It follows from Ex. 28 of Sect. 4.11.
(iii) Let y ∈ X mod G. Since X is simply connected, π1 (X, x0 ) = {e}, ∀ x0 ∈
p −1 (y) and hence p∗ π1 (X, x 0 ) = {e}. Consequently, Theorem 4.10.8(iii) fol-
lows from Theorem 4.10.8(ii). ❑
Remark 4.10.9 We first make geometrical constructions of some orbit spaces and
then compute their fundamental groups.
4.10.3 Fundamental Group of the Real Projective Space R P n
This subsection computes the fundamental group of R P n by using group action. We

have computed π1 (Rn ) ∼
= Z2 for n ≥ 2 in Theorem 4.6.9. Here we give an alternative
approach.
Definition 4.10.10 Let S n = {x ∈ Rn+1 : ||x|| = 1} be the n-sphere and R P n be the

n-dimensional real projective n-space. The antipodal map A : S n → S n , x → −x,
generates an action of the two element group G = {+1, −1} given by the relation
(+1)x = x and (−1)x = −x. Then its orbit space S n mod G is R P n , the real pro-
jective n-space.
Theorem 4.10.11 π1 (Rn ) ∼

= Z2 for n ≥ 2.
Proof 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 4.10.8 that the fundamental group of orbit space is
G. Thus π1 (S n mod G) = G ⇒ π(Rn ) = G ∼ = Z2 for n ≥ 2. ❑
Remark 4.10.12 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 non-trivial group that can act freely
on S n if n is even (see Chap. 14).
Remark 4.10.13 A generator for π1 (R P n ) is any loop obtained by projecting a path

in S n connecting two antipodal points.
4.10.4 The Fundamental Group of Klein’s Bottle
This subsection computes the fundamental group of Klein’s bottle. Let G be the
group of transformations of the plane generated by a and b. Consider the action
of G on R2 by a(x, y) = (x + 1, y) and b(x, y) = (1 − x, y + 1), ∀ (x, y) ∈ R2 .
Then a −1 (x, y) = (x − 1, y) and b−1 (x, y) = (1 − x, y − 1). Hence R2 is simply
connected and the action satisfies condition (A), then by Theorem 4.10.8, π1 (R2 mod
G) G. Now
b−1 ab(x, y) = b−1 a(1 − x, y + 1)

= b−1 (2 − x, y + 1) = (1 − 2 + x, y)
= (x − 1, y) = a −1 (x, y), ∀ (x, y) ∈ R2 ⇒ b−1 ab = a −1 .
Therefore R2 mod G is the Klein’s bottle. This gives a representation of Klein’s

bottle as an orbit space whose fundamental group is generated by a and b.
4.10.5 The Fundamental Groups of Lens Spaces
This subsection computes the fundamental group of lens spaces defined by H. Tietze
(1888–1971) in 1908, which are are 3-manifolds. Such spaces constitute an important
class of objects in the study of algebraic topology.
Let m > 1 be an integer space and p be an integer relatively prime to m and
2πi
S 3 = {(z 1 , z 2 ) ∈ C2 : |z 1 |2 + |z 2 |2 = 1} ⊂ C2 . Let ρ = e m be a primitive m-th root
of unity.
Define a map
2πi 2πi p
h : S 3 → S 3 , (z 1 , z 2 ) → (ρz 1 , ρ p z 2 ) = (e m z1, e m z 2 ).
Then h is a homeomorphism of S 3 onto itself of period m, i.e., h m = 1d . Thus h

induces an action of Zm on S n by the rule Zm × S 3 → S 3 , k(z 1 , z 2 ) = h k (z 1 , z 2 ),
where k denotes the residue class of the integer k modulo m, i.e., the action is
2πi
generated by the rotation z → e m z of the unit sphere S 3 ⊂ C2 = R4 . This action
2πir
has no fixed point, because the equation z = e m z, where r is an integer such that
0 < r < m has a solution z = 0 but z = 0 is not a point of S 3 .
The orbit spaces S 3 mod Zm is called a lens space and is denoted by L(m, p). Then
the lens space is the quotient space S 3 /∼ given by an equivalence relation ∼ on S 3 ,
defined by (z 1 , z 2 ) ∼ (z 1 , z 2 ) if there exists an integer k such that k(z 1 , z 2 ) = (z 1 , z 2 ),
i.e., (z 1 , z 2 ) = h k (z 1 , z 2 ).
As S 3 is Hausdorff, Zm is finite and Zm acts on S 3 without fixed point. The
above action of Zm on S 3 satisfies condition (A). Hence Zm π1 (S 3 mod Zm ) =
π1 (L(m, p)). We now extend the method of construction to construct generalized lens
spaces. Let m > 1 be an integer and p1 , p2 , . . . , pn−1 be integers relatively prime to
m and S 2n−1 = {(z 1 , z 2 , . . . , z n ) ∈ Cn : |z 1 |2 + |z 2 |2 + · · · + |z n |2 = 1} ⊂ Cn . Let
2πi
ρ = e m be a primitive mth root of unity. Define
2πi 2πi p1 2πi pn−1
h : 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 )
Then h is a homeomorphism of S 3 with period m, i.e., h m = 1d . Thus as before h

induces an action of Zm on S 2n−1 without fixed point by the rule Zm × S 2n−1 →
S 2n−1 , k(z 1 , z 2 , . . . , z n ) → h k (z 1 , z 2 , . . . , z n ), ∀ h ∈ Zm .
The orbit spaces S 2n−1 mod Z m is called a generalized lens space and is denoted
by L(m, p1 , . . . , pn−1 ). As before π1 (L(m, p1 , . . . , pn−1 )) ∼ = Zm . As a particular
case, for Z2 = {1d , a}, S 2 mod Z2 = L(2, 1), where 1d : S 2 → S 2 is the identity
map and a : S 2 → S 2 is the antipodal map and hence a 2 = 1d .
As the action of Z2 on S 2 yields S 2 mod Z2 = R P 2 , and its fundamental group
π1 (S 2 mod Z2 ) ∼ = Z2 .
4.10.6 Computing Fundamental Group of Figure-Eight

by Graph-theoretic Method
This subsection computes the fundamental group of figure-eight by graph-theoretic

method. We have shown in Theorem 4.2.2 that fundamental group of figure-eight is
not abelian.
For an alternative proof let G be a free group on two letters a and b. Define a
graph X = Graph (G, a, b) as follows:
The vertices of X are the elements of G. Hence the vertices are the reduced words
a and b. The edges of X are of the two types: (g, ga) and (g, gb), g ∈ G. Again
(g, ga), (g, gb), (ga −1 , g) and are the only four edges corresponding to the vertex
g. Now define a map G × X → X , given by h · g = hg, for every g, h ∈ G. Then
h · (g, ga) = (hg, hga) and h · (g, gb) = (hg, hgb), for edges (g, ga) and (g, gb).
Let 1G be the identity element in G. Then 1G · g = g and 1G · (g, ga) = (g, ga)
and 1G · (g, gb) = (g, gb). Again (h 1 h 2 ) · g = h 1 h 2 · g and h 1 · (h 2 · g) = h 1 · (h 2 ·
g) = h 1 h 2 · g, ∀ h 1 , h 2 ∈ G (Since h 2 g be a vertex in X ).
Also h 1 · (h 2 · (g, ga)) = h 1 · (h 2 g, h 2 ga) = h 1 (h 2 g, h 1 (h 2 ga)) = (h 1 (h 2 )) · (g, ga)
= h 1 (h 2 · (g, gb)) = h 1 (h 2 g, h 2 gb)
= h 1 (h 2 g, h 1 (h 2 gb)) = h 1 (h 2 ) · (g, gb).
Clearly, G acts on X . The orbit space X mod G is the Figure-Eight space whose
two loops are the images of the edges (g, ga) and (g, gb). As X is simply connected,
by using Theorem 4.10.8 it follows that π1 (X mod G, ∗) ∼ = G, which is the free
group on two generators.
4.10.7 Application of Galois Correspondence
This subsection presents an interesting application of Galois correspondence.
Theorem 4.10.14 Let (X, p) be a covering space of B. If B is connected, locally

path-connected, and semilocally simply connected, then
(a) The components of X are in one-to-one correspondence with orbits of the action
of π1 (B, b0 ) on the fiber p −1 (b0 );
(b) Under the Galois correspondence between connected covering spaces of B and
subgroups of π1 (X, x0 ), the subgroup corresponding to the component of X
containing a given lift b̃0 = x0 of b0 is the stabilizer group G x0 of x0 ∈ X , whose
action on the fiber leaves x0 fixed.
Proof (a) Let x 0 , x1 ∈ p −1 (b0 ). If they are in different components of X, π1 (X, x 0 )

cannot map one to the other, since there exists no path-connecting them. Claim
that π1 (X, x0 ) acts transitively on each of the components of X to obtain a
bijection. By hypothesis B is locally path-connected, hence X is locally path
connected. Clearly, the notions of connected components and path-connected
components are the same. If x0 and x 1 lie in the same component, there exists a
path α : I → X such that α(0) = x 0 , α(1) = x 1 . Then [ p ◦ α] is an element of
π1 (B, b0 ) whose action on p −1 (b0 ) maps x0 to x1 . Hence this action is transitive.
Then the set of elements of p −1 (b0 ) in a given component constitutes an orbit,
and this produces a bijection.
(b) Choose a given lift x0 of b0 in some component X of X . Under the Galois
correspondence, the subgroup of π1 (B, b0 ) corresponding to X is the image of
G = π1 (X , x0 ) in the inclusion p∗ : G → π1 (B, b0 ). Any loop α ∈ p∗ (G) lifts
back to a loop in X by the unique lifting property. Hence α sends x0 to itself
and is an element of the stabilizer group G x0 of x0 .
Conversely, if β ∈ π1 (B, b0 ) is in the stabilizer group of x 0 , then the lift β̃ of
β is a loop from x 0 to itself and hence β̃ ∈ G, which implies β ∈ p∗ (G). This
shows that p∗ (G) is the stabilizer group of x0 . ❑
4.11 Exercises
1. Assume that f : S n → Rn is a continuous map such that f (−x) = − f (x) for

any x ∈ S n . Show that there exists a point x ∈ S n such that f (x) = 0.
2. Assume that f : S n → Rn is a continuous map. Show that there exists a point
x ∈ S n such that f (x) = f (−x).
3. Prove that no subspace of Rn is homeomorphic to S n .
4. Show that there is no continuous antipode-preserving map f : S 2 → S 1 . Use
this result to prove Borsuk–Ulum theorem for dimension 2.
5. Let X and B be path-connected spaces and (X, p) be a covering space of B. Let
b0 ∈ B and Y = p−1 (b0 ) be the fiber over b0 . Prove the following:
(i) If x0 , x 1 ∈ Y , then p∗ π1 (X, x0 ) and p∗ π1 (X, x 1 ) are conjugate subgroups
of π1 (B, b0 );
(ii) If H is a subgroup of π1 (X, x 0 ) which is conjugate to p∗ π1 (X, x 0 ) for some
x 0 ∈ Y , then there exists a point x1 ∈ Y such that H = p∗ π1 (X, x 1 ).
[Hint: Use Theorem 4.5.7.]
(iii) A covering space (X, p) of B is said to be regular if p∗ π1 (X, x0 ) is a normal
subgroup of π1 (B, b0 ) for every b0 ∈ B. If (X, p) is regular covering space
of B, show that p∗ π1 (X, x0 ) = p∗ π1 (X, x 1 ) for every pair of point x0 , x1 in
the same fiber.
(iv) If X is simply connected, prove that every covering space (X, p) of B is
regular.
(v) If π1 (B, b0 ) is abelian, then every covering space of B is regular.
6. Let B be a connected and locally path-connected space and let b0 ∈ B. Then show
that a covering space (X, p) of B is regular if and only if the group Aut (X/B)
acts transitively on the fiber over b0 .
4.11 Exercises 193
7. Let B be locally path-connected and b0 ∈ B. Show that two covering spaces

(X, p) and (Y, q) of B are isomorphic if and only if the fibers p−1 (b0 ) and
q −1 (b0 ) are isomorphic G = π1 (B, b0 )-sets.
8. Let a group G act transitively as a set Y , and let x, y ∈ Y . Prove that G x = G y
if and only if there exists f ∈ Aut (Y ) with f (x) = y.
9. Show that the graph X described in Sect. 4.10.6 has no cycles.
10. Let (X, p) be a covering space of B, where X is locally path-connected. Let
b0 ∈ B. Given x0 , x1 ∈ Y = p−1 (b0 ), show that there exists an h ∈ Cov (X/B)
with h(x 0 ) = x1 if and only if there exists f ∈ Aut (Y ) with f (x0 ) = x1 .
11. Let (X, p) be a covering space of B, where B is locally path-connected. Let
b0 ∈ B and let the fiber p −1 (b0 ) = y be viewed as a G = π1 (B, b0 )-set. Then
show that ψ : Cov (X/B) → Aut (Y ) defined by ψ(h) = h|Y is isomorphism.
12. Let G be a group acting transitively on a set Y and let y0 ∈ Y . Let N G (G 0 ) denote
the normalizer of the isotropy group G 0 of y0 . Show that Aut (Y ) ∼ = NG (G 0 )/G 0 .
13. Let (X, p) be a covering space of B, where B is locally path-connected. Show
that for b0 ∈ B and x0 ∈ p −1(b0 ) , Aut (X/B) ∼ = N G ( p∗ π(X, x0 ))/ p∗ π1 (X, x 0 ).
Hence show that π1 (S 1 , 1) ∼ = Z.
14. Let (X, p) be a regular covering space of B, where B is locally path-connected.
For b0 ∈ B and x0 ∈ p −1 (b0 ), show that Aut (X/B) ∼ = π1 (B, b0 )/ p∗ π1 (X, x 0 )
by the monodromy group of the regular covering space.
15. Let (X, p) be a universal covering space of B, where B is locally path connected.
Show that for any b0 ∈ B, Aut (X/B) ∼ = π1 (B, b0 ).
16. If B is an H-space, prove that every covering space of B is regular.
[Hint: π1 (B, b0 ) is abelian for b0 ∈ B ⇒ every covering space of B is regular.]
17. Let G be a path-connected topological group and H be a discrete normal sub-
group of G. If p : G → G/H is the natural homomorphism, show that (G, p)
is covering space of G/H .
18. Let (X, p) be a covering space of B and b0 , b1 ∈ B. If F0 and F1 are the fibers over
b0 and b1 , respectively, show that |F0 | = |F1 | and any two fibers of (X, p) a [Hint:
Use Theorem 4.10.1(iii). Since each fiber is discrete and any two fibers have the
same cardinal numbers, it follows that any two fibers are homeomorphic.]
19. Show that the map p : S 1 → S 1 , z → z 2 is a covering map. Generalize to the
map p : S 1 → S 2 , z → z n .
20. If S 1 → S 1 is continuous and antipode preserving, show that f is
not nullhomotopic.
21. Let B be a path-connected and locally path-connected space. Suppose (X, p)
and (Y, q) are covering spaces of B. Let b0 ∈ B, x0 ∈ X and y0 ∈ Y be base
points with p(x0 ) = b0 = q(y0 ). If q∗ π1 (Y, y0 ) ⊂ p∗ π1 (X, x 0 ), show that
(i) there exists a unique continuous map f : (Y, y0 ) → (X, x 0 ) such that p ◦
f = q;
(ii) (Y, f ) is a covering space of X and so X is a quotient space of Y .
22. Let X, B, Y be path-connected and locally path-connected spaces such that
(X, p) is a covering space of B. If x0 ∈ X, y0 ∈ Y and b0 ∈ B with p(x0 ) = b0 ,
show that for every continuous map f : (Y, y0 ) → (B, b0 ) with f ∗ π1 (Y, y0 ) ⊂
p∗ π1 (B, b0 ), there exists a continuous map f˜ : (Y, y0 ) → (X, x 0 ) such that

p ◦ f˜ = f.
23. Let (X, p) be a covering space of B and x0 ∈ X, b0 ∈ b such that p(x0 ) = b0 .
If X is simply connected, show that b0 has a neighborhood U such that the
inclusion map i : U → B induces the trivial homomorphism
i ∗ : π1 (U, b0 ) → π1 (B, b0 ).
24. Let X be a normal space. Show that the inclusion i : A → X is a cofibration iff
the inclusion j : A → U is a cofibration for some open neighborhood U of A
in X .
25. Let p : X → B be a fibration and f : A → B be a continuous map. Show that
there exists a bijection between the homotopy sets C = [g : A → X : p ◦ g =
f ] and D = [g̃ : A → X : p ◦ g̃ f ].
26. Is the map p : (0, 3) → S 1 , x → e2πi x a covering map? Justify your answer.
27. Find nontrivial coverings of Möbius strip by itself.
28. Let B be path-connected, locally path-connected and p : (X, x 0 ) → (B, b0 ) be
a covering space. If H is the subgroup p∗ (π1 (X, x0 )) of π1 (B, b0 ), show that
(i) The automorphism group Aut (X/B) is isomorphic to the quotient group
N (H )/H , where N (H ) = {g ∈ π1 (B, b0 ) : g H g −1 = H } is the normalizer
of H in π1 (B, b0 ).
(ii) The group Aut (X/B) is isomorphic to the group π1 (B, b0 )/H if X is a
regular covering.
(iii) If p : (X, x0 ) → (B, b0 ) is universal covering, then Aut (X/B) ∼
= π1 (B, b0 ).
29. Let (X, p) be a universal covering space of a connected topological space B.
If b0 ∈ B and x0 ∈ X are base points such that x 0 ∈ p−1 (b0 ), show that the
induced homomorphism p∗ : π1 (X, x0 ) → πn (B, b0 ) is an isomorphism for n ≥
2. Hence show that πn (R P m ) ∼ = πn (S m ) for n ≥ 2.
30. Let B be a path-connected space and X be a connected covering space of B. Let
p : X → B be a covering projection. Let b0 ∈ B and x0 ∈ p −1 (b0 ). Show that
for every n ≥ 2, p∗ : π1 (X, x 0 ) → π1 (B, b0 ) is an isomorphism. Hence show
that for every n ≥ 2, πn (S 1 , 1) = 0.
31. Let f : A → X be a continuous map and i : A → M f be the inclusion i(a) =
[a, 0]. Show that the inclusion i : A → M f is cofibration.
[Hint. Use Steenrod theorem, Chap. 2.]

32. Let p : X → B be a fibration with fiber F = p −1 (b0 ) and B be path-connected.
Let Y be any space. Show that the sequence of sets
i∗ p∗
[Y, F] −−−→ [Y, X ] −−−−→ [Y, B]
is exact.
4.11 Exercises 195
33. Let i : A → X be a cofibration, with cofiber X/A and q : X → X/A denote

the quotient map. If Y is any path-connected space, then show that the sequence
of sets
q∗ i∗
[X/A, Y ] −−−→ [X, Y ] −−−→ [A, Y ]
is exact.
[4] Bredon, G.E., Topology and Geometry, Springer-Verlag, New York, Inc.1993.
[5] Davis, J.F. and Kirk, P. Lecture Notes in Algebraic Topology, Indiana University,
Bloomington, IN (http://www.ams.org/bookstore-getitem/item=GSM-35).
[7] Dodson, C.T.J., and Parker, P.E., A User’s Guide to Algebraic Topology, Kluwer,
Dordrecht 1997.
[8] Gray, B., Homotopy Theory, An Introduction to Algebraic Topology, Acamedic
Cambridge, 1983.
[11] Mayer, J. Algebraic Topology, Prentice-Hall, New Jersy, 1972.
[13] Maunder, C.R.F., Algebraic Topology, Van Nostrand Reinhhold, London, 1970.
[15] Munkres, J.R., Elements of Algebraic Topology, Addition-Wesley-Publishing
Company, 1984.
[17] Wallace , A.H., Algebraic Topology, Benjamin, New York, 1980.
References
Aguilar, M., Gliter, S., Prieto, C.: Algebraic Topology from a Homotopical View Point. Springer,
New York (2002)
Armstrong, M. A.: Basic Topology. Springer, New York (1983)
Bredon, G.E.: Topology and Geometry. Springer, New York, Inc (1993)
Davis, J.F., Kirk, P.: Lecture Notes in Algebraic Topology, Indiana University, Bloomington, IN.
http://www.ams.org/bookstore-getitem/item=GSM-35
Boston (1989)
Dodson, C.T.J., Parker, P.E.: A User’s Guide to Algebraic Topology. Kluwer, Dordrecht (1997)
(1975)
Hilton, P.J.: An introduction to Homotopy Theory. Cambridge University Press, Cambridge (1983)
Hurewicz, W.: On the concept of fibre space. Proc. Natl. Acad. Sci., USA 41, 956–961 (1955)
Mayer, J.: Algebraic Topology. Prentice-Hall, New Jersy (1972)
Munkres, J.R.: Elements of Algebraic Topology. Addison-Wesley-Publishing Company, Menlo
Park (1984)
Steenrod, N.: The Topology of Fibre Bundles. Prentice University Press, Prentice (1951)
Wallace, A.H.: Algebraic Topology. Benjamin, New York (1980)
Chapter 5
Fiber Bundles, Vector Bundles and K -Theory
This chapter continues the study of homotopy theory through fiber bundles, vector
bundles, and K -theory. Fiber bundles and vector bundles form special classes of
bundles with additional structures. They are closely related to the homotopy theory
and are important objects in the study of algebraic topology. A fiber bundle is a
bundle with an additional structure derived from the action of a topological group
on the fibers. On the other hand, a vector bundle is a bundle with an additional
vector space structure on each fiber. 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. 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.
A fiber bundle is a locally trivial fibration and has covering homotopy property.
J. Feldbau reduced in 1939 the classification problem of principal fiber bundles with
a given base S n for n ≥ 2 to a problem in homotopy theory (Feldbau 1939). Fiber
bundles carry nice homotopy properties and play a key role in geometry and physics.
(see Chaps. 7, 14 and 17). This subject also marks a return of algebraic topology to
its origin. If we consider the tangent plane at each point of a surface, to get global
information about the surface, we investigate how the planes change as we move the
point on the surface. Again to investigate a higher dimensional smooth geometrical
object such as differential manifold, we consider the linear space tangent at each
point of the manifold. This leads to the concept of tangent bundles of manifolds,
general vector bundles, and fiber bundles.
The concept of fiber bundles arose through the study of some problems in topology
and geometry of manifolds around 1930. Its first general definition was given by
H. Whitney (1907–1989). His work and that of H. Hopf (1894–1971), E. Stiefel
(1909– 1978), J. Feldbau (1914–1945) and many others displayed the importance of
the subject for the application of topology to different areas of mathematics during
DOI 10.1007/978-81-322-2843-1_5
198 5 Fiber Bundles, Vector Bundles, and K -Theory
1935–1940. Since then the subject has attracted general interest because of some
of the finest applications of topology to other fields, and promising many more
applications. On the other hand, the concept of a vector bundle arose through the
study of tangent vector fields to smooth manifolds such as spheres, projective spaces,
and manifolds in general. Although this notion had appeared in the literature before
1955, the definition introduced by W. Hurewicz (1904–1956) in 1955 is much more
general and useful.
K -theory born in connecting the rich structure of vector bundles over a paracom-
pact space B with the set of homotopy classes of continuous maps from B into the
Grassmann manifold G n (F ∞ ) of n-dimensional subspaces in infinite-dimensional
space (F ∞ ) plays a vital role in applications of algebraic topology to analysis, alge-
braic geometry, topology, ring theory, and number theory. The two most surprising
applications of topological K-theory are: J.F. Adams (1930 –1989) solved the Hopf
invariant one problem in 1962 by doing a computation with his Adams operations.
Then he proved an upper bound for the number of linearly independent vector fields
on spheres (see Chap. 17).
More precisely, this chapter studies the theory of fiber bundles with a special
attention to vector bundles with fibers of different dimensions, homotopy classifica-
tion of vector bundles, and K -theory (which is a generalized cohomology theory)
and interlinks vector bundles with homotopy theory. This chapter also studies Hopf
maps, Hopf bundles, and Hurewicz fibering.
Milnor’s construction of a universal fiber bundle for any topological group G
and homotopy classification of numerable principal G-bundles are given and hence
the classification of numerable principal G-bundles has been reduced to homotopy
theory. Finally, it has been shown that for every topological group G, there exists
a topological space BG , called classifying space having the property that for every
pointed topological space B there is a bijective correspondence between isomor-
phism classes of numerable principal G-bundles over B and [B, BG ], the homotopy
classes of base point preserving maps from B to BG . There also exists a bijective
correspondence between the set of isomorphism classes of F-vector bundles over a
paracompact space B and the set [B, G n (F ∞ )] of homotopy classes of continuous
maps from B to Grassmann manifold G n (F ∞ ), which leads to define a group K F (B),
called the K -theory introduced by M. Atiyah and F. Hirzebruch in 1961.
For this chapter the books Gray (1975), Husemöller (1966), Luke and Mischenko
(1984), Nakahara (2003), Spanier (1966), Steenrod (1951), Switzer (1975) and some
others are referred in the Bibliography.
5.1 Bundles, Cross Sections, and Examples
This section introduces the concept of bundles and their cross sections. Fiber bundles
and vector bundles form special classes of bundles with additional structure which
are important in the study of algebraic topology and they are closely related to the
homotopy theory. The recognition of bundles in mathematics was realized during
5.1 Bundles, Cross Sections, and Examples 199
1935–1940 through the work of Whitney, H. Hopf and E. Stiefel and some others.
Since then the subject has created a general interest.
5.1.1 Bundles
This subsection studies the concept of bundles which plays an important role in the
theory of fiber bundles and vector bundles. So, it is natural to introduce the concept
of bundles at the beginning. A bundle is the basic underlying structure leading to the
concepts of fiber bundles and vector bundles. It 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. Roughly speaking, a bundle is a union of fibers parametrized by its base
space and glued together by the topology of the total space.
Let E and B be two topological spaces and p : E → B be a continuous map.
Definition 5.1.1 A bundle ξ = (E, p, B) is an ordered triple consisting of a topo-

logical space E, called the total space of ξ, a topological space B, called the base
space of ξ and a continuous map p : E → B, called the projection of the bundle ξ.
For each b ∈ B, E b = p −1 (b) (it is nonempty as p is onto by assumption), is

called the fiberof ξ overb, which has the topology induced by the inclusion in E.
Clearly, E = Eb = p −1 (b) and every two fibers E b and E b are disjoint if
b∈B b∈B
b = b and hence every point of E lies in exactly one fiber. Sometimes we write
E(ξ) for the total space and B(ξ) for the base space of the bundle ξ to avoid any
confusion.
Example 5.1.2 For the bundle ξ = (E, p, B) as shown in Fig. 5.1, the total space E
is decomposed into fibers of four types: a point, a point together with a segment, two
segments, and a segment.
Fig. 5.1 fibers of bundle

Definition 5.1.3 Let ξ = (X, p, B) be a bundle. If Y be a subspace of X and q =

p|Y : Y → B, then (Y, q, B) is said to be a subbundle of ξ. In particular, if A is a
subspace of B, then the bundle (Y, q, A) is a subbundle of ξ.
Example 5.1.4 Let (X, p, B) be a bundle. If there are inclusion maps f : Y → X

and g : A → B, then the bundle (Y, q, A) is a subbundle of (X, p, B), where q is
the restriction of p over Y. For interesting examples see Sect. 5.3.
Definition 5.1.5 (Induced bundle) Let ξ = (X, p, B) be a bundle and f : A → B

be a continuous map from a topological space A to the topological space B. The
induced bundle f ∗ (ξ) = (Y, q, A) of ξ under f is the bundle (Y, q, A), where Y =
{(a, x) ∈ A × X : f (a) = p(x)}, and q : Y → A, (a, x) → a.
Proposition 5.1.6 Let ξ = (X, p, B) be a bundle and f : A → B be a continuous

map from a topological space A to the topological space B. If f ∗ (ξ) = (Y, q, A) is
the induced bundle of ξ under f and p is an open map, then q is also so.
5.1.2 Cross Sections
We now introduce the concept of cross sections of a bundle. Its importance lies in the
fact that the cross sections of certain bundles are identified with familiar geometric
objects.
Definition 5.1.7 Let ξ = (E, p, B) be a bundle. A cross section (or in brief section)
s of ξ is a continuous map s : B → E such that p ◦ s = 1 B (identity map on B).
For every b ∈ B, ( p ◦ s)(b) = 1 B (b) = b shows that s(b) ∈ p −1 (b), and hence
s(b) lies in the fiber p −1 (b) for each b of the base space B. The condition p ◦ s =
1 B shows that the map s : B → E is injective. Otherwise, if for some b1 = b2 ∈
B, s(b1 ) = s(b2 ), then p(s(b1 )) = b1 and p(s(b2 )) = b2 implies b1 = b2 .
Definition 5.1.8 (Product bundle) The product bundle over B with fiber F is the
bundle (B × F, p, B), where p : B × F → B, (b, x) → b is the projection on the
first factor.
We now determine the cross sections of a product bundle.
Proposition 5.1.9 A cross section s of the product bundle (B × F, p, B) is precisely

of the form s(b) = (b, f s (b)), where f s : B → F is a continuous map uniquely deter-
mined by s.
Proof Let s : B → B × F be a cross section of the product bundle (B × F, p, B).

Then s takes the form s : B → B × F, b → (gs (b), f s (b)) such that p ◦ s = 1 B ,
where gs : B → B and f s : B → F are maps uniquely determined by s.
Hence p(s(b)) = 1 B (b) ⇒ p(gs (b), f s (b)) = b ⇒ gs (b) = b, ∀ b ∈ B ⇒ s(b) =

(b, f s (b)) for every b ∈ B. Conversely, let s : B → B × F be a continuous map such
that s(b) = (b, f s (b)) for every b ∈ B. Then ( p ◦ s)(b) = b, ∀ b ∈ B ⇒ p ◦ s =
1 B ⇒ s is a cross section of the bundle (B × F, p, B). ❑
The map f s : B → F determined in Proposition 5.1.9 gives the following

Corollary.
Corollary 5.1.10 Given a product bundle ξ = (B × F, p, B), let S(ξ) be the set of
all cross sections of ξ and C(ξ) be the set of all continuous maps B → F. Then the
map ψ : S(ξ) → C(ξ), s → f s is a bijection.
Remark 5.1.11 Let η = (X , p , B) be a subbundle of a bundle ξ = (X, p, B). Then

s is a cross section of η iff s(b) ∈ X for every b ∈ B.
5.1.3 Morphisms of Bundles
This subsection introduces the concept of bundle morphisms with an aim to utilize
this concept in the study of fiber and vector bundles. While comparing vector bundles
or fiber bundles over the same or different base spaces the concept of morphisms
becomes necessary like group homomorphisms. A bundle morphism is intuitively a
fiber preserving map and is similar to the concept of a group homomorphism.
Definition 5.1.12 Let ξ = (X, p, B) and η = (Y, q, A) be bundles. A bundle mor-

phism or a fiber map ( f, g) : ξ → η is a pair of continuous maps f : X → Y and
g : B → A such that the diagram in Fig. 5.2 is commutative, i.e., q ◦ f = g ◦ p.
Notation: For the bundle ξ = (X, p, B), we use sometimes the notation E(ξ) for its
total space X .
Remark 5.1.13 The map f in Fig. 5.2 is fiber preserving. Since for every x ∈
X, (q ◦ f )(x) = (g ◦ p)(x) holds, and hence the pair (x, p(x)) is mapped into the
pair ( f (x), g( p(x))) by ( f, g). Consequently, for every b ∈ B we have f ( p−1 (b)) ⊂
q −1 ( f (b)). This implies that f carries fibers of ξ over b into the fibers of η over f (b)
for each b ∈ B.
Remark 5.1.14 The particular case, when ξ and η are both bundles over the same
base space B, is interesting.
Fig. 5.2 Morphism of f

X /Y
bundles
p q
g
B /A
Fig. 5.3 B-morphism of f

XA /Y
bundles AA ~~
AA ~
p AA ~~q
~~
~
B
Definition 5.1.15 Given two bundles ξ = (X, p, B) and η = (Y, q, B) over the
same base space B, a bundle B-morphism f : ξ → η is a continuous map f :
f
X −−−→ Y such that the triangle in the Fig. 5.3 is commutative i.e., p = q ◦ f .
Clearly, q ◦ f = p implies that f ( p−1 (b)) ⊂ q −1 (b) for every b ∈ B. Hence f is a
fiber preserving map.
We now show that a cross section of a bundle carries the general property of a
bundle morphism.
Proposition 5.1.16 The cross sections of a bundle ξ = (X, p, B) over B are pre-
cisely the B-morphisms s : (B, 1 B , B) → (X, p, B).
Proof Let s be cross section of ξ. Then p ◦ s = 1d . This implies the triangle in

Fig. 5.4 is commutative. Hence s is a B-morphism. Conversely, let s : (B, 1d , B) →
(X, p, B) be a B-morphism. Then clearly, s is a cross section of ξ. ❑
Remark 5.1.17 Every general property of bundle morphisms is equally valid for
cross sections also.
Definition 5.1.18 A bundle (Y, q, A) is a subbundle of (X, p, B) if there are inclu-

sion maps f : Y → X and g : A → B. Then the pair ( f, g) : (Y, q, A) → (X, p, B)
is a bundle morphism.
Definition 5.1.19 Let ξ = (X, p, B) and η = (Y, q, A) be two bundles. 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.
Remark 5.1.20 If ( f, g) : ξ → η is a bundle isomorphism, then the pair ( f −1 , g −1 ) :

η → ξ is also a bundle isomorphism such that ( f ◦ f −1 , g ◦ g −1 ) and ( f −1 ◦ f, g −1 ◦
g) are both identity bundle morphisms.
Definition 5.1.21 Let ξ = (X, p, B) be a bundle and A be a nonempty subset of

B. The restricted bundle of ξ to A, denoted by ξ|A, is the bundle (Y, q, A), where
Y = p −1 (A) and q = p|Y .
Example 5.1.22 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.
Fig. 5.4 Cross section of s /X

B@
bundle @@ }}
@@ }
1d @@ }}p
~}
}
B
Definition 5.1.23 Let ξ = (X, p, B) and η = (Y, q, B) be two bundles over the
same base space B. If f : X → Y is a homeomorphism, then f is called a B-
isomorphism. It is said to be locally isomorphic if to every point b ∈ B, there is
an open neighborhood Ub of b and an Ub -isomorphism between the restricted bun-
dles ξ|Ub and η|Ub .
Remark 5.1.24 The relation of being locally isomorphic is an equivalence relation

on the set of all bundles over B and hence if ξ is locally isomorphic to a locally trivial
bundle, then ξ is locally trivial.
Definition 5.1.25 (Canonical morphism) 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 , then the pair of maps ψ : Y → X, (a, x) → x and f : A → B form a bundle
morphism (ψ, f ) : f ∗ (ξ) → ξ called the canonical morphism of the induced bundle.
It has some interesting properties:

Proposition 5.1.26 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 (ψ, f ) : f ∗ (ξ) → ξ
is the canonical bundle morphism, then for each a ∈ A, the restriction ψ|q −1 (a) :
q −1 (a) → p −1 ( f (a)) is a homeomorphism.
Proposition 5.1.27 Let h : C → A and g : A → B be two continuous maps and ξ

be a bundle over B. Then the induced bundle 1∗B (ξ) and the bundle ξ are B-isomorphic
and the induced bundles h ∗ (g ∗ (ξ)) and (g ◦ h)∗ (ξ) are C-isomorphic.
Proposition 5.1.28 Let ( f, g) : (X, p, B) → (Y, q, A) be a bundle morphism and

s : B → X be a cross section of (X, p, B). If the map g : B → A is a homeomor-
phism, then s = f ◦ s ◦ g −1 : A → Y is a cross section of (Y, q, A).
Corollary 5.1.29 Let (X, p, B) and (Y, q, B) be two bundles over the same base
space B and f : (X, p, B) → (Y, q, B) is a bundle B-morphism. Then to every cross
section s of (X, p, B), there exists an induced cross section s of (Y, q, B) given by
s = f ◦ s (Fig. 5.5).
Fig. 5.5 Construction of f

XO /Y
cross section of (Y, q, A)
p s q
g −1
o /A
B g:≈
Fig. 5.6 Induced cross f

X2X22 /Y
section of (Y, q, A) 222 F
2
222
2222 s
s
p 2222

2222 q
2222

B
Proof It follows from Proposition 5.1.28 by taking g = 1 B (see diagram in

Fig. 5.6). ❑
Definition 5.1.30 (Trivial bundle) Let ξ = (X, p, B) be a bundle. A space F is said

to be the fiber space of the bundle ξ if for every b ∈ B, the fiber p−1 (b) (topologized
by the induced topology from X ) is homeomorphic to F. The bundle ξ is said to be a
trivial bundle with fiber F if ξ is B-isomorphic to the product bundle (B × F, q, B),
with the projection map
q : B × F → B, (b, f ) → b.
Remark 5.1.31 A bundle is locally trivial if locally it is a direct product.

All fibers of a locally trivial bundle with connected base are homeomorphic.
Remark 5.1.32 Let X and B be connected spaces. A bundle p : E → B is a covering

space if its fiber is discrete (i.e., if the fiber consists of isolated points which means
that each point is an open set).
5.1.4 Examples
This subsection presents some interesting examples illustrating the concepts dis-
cussed earlier. Here , denotes the inner product in Rn+1 .
Example
√ 5.1.33 (Tangent bundle and normal bundle) Let S n = {x ∈ Rn+1 : x =
x, x = 1} be the n-sphere. The tangent bundle ξT over S n in Rn+1 is the subbundle
(T (S n ), p, S n ) of the product bundle (S n × Rn+1 , p, S n ), whose total space is
T (S n ) = {(b, x) ∈ S n × Rn+1 : b, x = 0},
and projection is
p : T (S n ) → S n , (b, x) → b.
An element of T (S n ) is called a tangent vector to S n at the point b ∈ S n . Clearly,

the fiber p −1 (b) ⊂ T (S n ) is a vector space of dimension n. A cross section of the
tangent bundle ξT over S n is called a tangent vector field (or simply vector field)
over S n .
Similarly, the normal bundle ξ N 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 is
N (S n ) = {(b, x) ∈ S n × Rn+1 : x = tb for some t ∈ R}
and projection is
q : N (S n ) → S n , q(b, x) = b.
An element of N (S n ) is called a normal vector to S n at the point b ∈ S n . Clearly, the

fiber q −1 (b) ⊂ T (S n ) is a vector space of dimension 1. A cross section of the normal
bundle ξ N over S n is called a normal vector field on S n .
Example 5.1.34 (Orthonormal r-frames) The bundle ξr of orthonormal r-frames over

S n for r ≤ n is the subbundle (X, p, B) of the product bundle (S n × (S n )r , p, S n ),
where total space X is defined by
X = {(b, y1 , y2 , . . . , yr ) ∈ S n × (S n )r : b, yi = 0 and yi , y j

= δi j (Kronecker delta), 1 ≤ i, j ≤ r },
where δi j means

1, if i = j
δi j =
0, otherwise
An element (b, y1 , y2 , . . . , yr ) of X is an orthonormal system of r-tangent vectors

to S n at b ∈ S n . A cross section of ξr over S n is called a field of r-frames.
Example 5.1.35 (Canonical vector bundle γ n ) Let Vr (Rn ) be the Stiefel manifold
(variety) of orthonormal r -frames in Rn defined by Vr (Rn ) = {(y1 , y2 , . . . , yr ) ∈
(S n−1 )r : yi , y j = δi j } and G r (Rn ) be the Grassmann manifold (variety) of r-
frames in Rn (r ≤ n). Then G r (Rn ) is the set of r-dimensional subspaces of Rn
with the quotient topology defined by the identification map
π : Vr (Rn ) → G r (Rn ), (y1 , y2 , . . . , yr ) → y1 , y2 , . . . , yr ,
where y1 , y2 , . . . , yr is an r -dimensional subspace in Rn with a basis {y1 , y2 , . . . ,

yr }. The canonical r-dimensional vector bundle γrn = (X, p, G r (Rn )) on G r (Rn )
is the subbundle of the product bundle (G r (Rn ) × Rn , p, G r (Rn )) with the total
space consisting of the subspace of pairs (V, x) ∈ G r (Rn ) × Rn with x ∈ V and
the orthogonal complement vector bundle of γrn , denoted by γr∗ is the subbundle of
n
(G r (Rn ) × Rn , p, G r (Rn )) defined by γr∗ = (Y, p, G r (Rn )), where Y = {(V, x) ∈

n
G r (Rn ) × Rn : V, x = 0 (i.e., x is orthogonal to V )}. In particular, if r = 1, then

γ1n on R P n−1 = G 1 (Rn ), is one-dimensional and is called the canonical line bundle.
∞ ∞
By natural inclusion G r (R ) ⊂ G r (R ) and G r (R ) is defined by G r (R ) =
n n+1
G r (Rn ) with induced (weak) topology. Similarly, G r (Cn ), G r (Hn ) and G r (C∞ )
r ≤n
and G r (H∞ ) are defined, where H is the division ring of quaternions.
If F = R, C or H, then the canonical vector bundle over G r (F∞ ) is denoted by γr .
As G r (Rn ) ⊂ G r (Rn+t ) for integers t ≥ 1, we may view γrn as γrn = γrn+t |G r (Rn ),
which is a restriction of the bundle γrn+t over G r (Rn ). Similarly, the restricted bundle
over G r (Cn ) (or G r (Hn )) is defined.
Example 5.1.36 (Bundle of groups) We now consider a special type of covering

spaces with additional structure on fibers. A bundle of groups ξ = (X, p, B) is a
covering space such that all the fibers p−1 (b) are isomorphic to a fixed group G in
the following sense:
Every point b of B has a neighborhood U for which there exists a homeomorphism
fU : p −1 (U ) → U × G taking each p−1 (b) to {b} × G by a group isomorphism.
Since G is endowed with discrete topology, the projection p is a covering map. The
bundle ξ is called a bundle of groups with fiber G.
5.2 Fiber Bundles: Introductory Concepts
This section considers a class of fibrations, called fiber bundles which are frequently
used in geometry, topology, and theoretical physics. Fiber bundles over paracompact
spaces are always fibrations. Fiber bundles form a special family of topological
spaces in the study of algebraic topology. The concept of fiber bundles arose through
the study of some problems in topology and geometry of manifolds around 1930.
There exists an infinite exact sequence corresponding to any fiber space (see Chap. 7).
A manifold of dimension n is a topological space which looks locally like Rn , but
not necessarily globally so. By introducing a chart, a local Euclidean structure to a
manifold is provided, which facilitates to use the conventional calculus of several
variables. A fiber bundle is a topological space which likewise looks locally a direct
product of two topological spaces.
A fiber bundle with a discrete fiber space is a covering space. Conversely, a
covering space whose all fibers have the same cardinality is a fiber bundle with
discrete fiber. For example, a covering space over a connected space is a fiber bundle
with a discrete fiber. Covering spaces provide tools to study fundamental groups.
Likewise fiber bundles provide tools to study higher homotopy groups. The first
general definition of fiber bundles was given by H. Whitney. The work of H. Whitney,
H. Hopf and E. Stienfel established the importance of fiber bundles for applications
of topology to geometry around 1940. Since then, this topic has created general
interest for its finest applications to other fields such as general relativity and gauge
theories and has promised many more. It also makes a return of algebraic topology to
its origin and revitalized this topic from its origin in the study of classical manifolds.
5.2 Fiber Bundles: Introductory Concepts 207
A covering space is locally the product of its base space and a discrete space.
For the covering space p : R → S 1 , t → e2πit , we say that R is a fiber space over
S 1 . This example introduces the concept of fiber bundles in this section. Roughly
speaking, a fiber bundle looks locally a trivial fibration, because, the total space of
a fiber bundle is locally the product of its base space and its fiber. For the general
theory of fiber bundles see Steenrod (1951).
Definition 5.2.1 A fiber bundle is an ordered quadruple ξ = (X, p, B, F), con-

sisting of a total space X , a base space B, a projection p : X → B such that B
has an open covering {U j } j∈J and for each j ∈ J , there is a homeomorphism
φ j : U j × F → p−1 (U j ), with the property that the composite p ◦ φ j is the pro-
jection
pU j : U j × F → U j
to the first factor (i.e., p ◦ φ j = pU j ).

p−1 (b) is said to be the fiber of ξ over b and p is called a fiber bundle projection.
For a fiber bundle with fiber F, the fiber bundle projection p : X → B and the
projection B × F → B are locally isomorphic (See Definition 5.1.23). A fiber bundle
p
ξ = (X, p, B, F) is sometimes symbolized like F → X −−−→ B. It is called a
covering of B if F is discrete and X is called a covering space over B and p is called
a covering projection. Then p−1 (b0 ) is discrete for b0 ∈ B.
Remark 5.2.2 The space F is homeomorphic to p−1 (b) for each b ∈ B. Usually,
there is a structure group G for ξ consisting of homeomorphisms of F leading to
the concept of G-bundles. If B is a paracompact space, the map p : X → B is a
fibration.
Example 5.2.3 Consider the circle S 1 in the complex plane, i.e., S 1 = {z ∈ C :

|z| = 1}. Define p : R → S 1 , t → e2πit . Then (R, p, S 1 , Z) is a fiber bundle. To
show it, let U1 = S 1 − {1} and U2 = S 1 − {−1}. Then p−1 (U1 ) = R − Z and there is
a homeomorphism φ1 : U1 × Z → p −1 (U1 ) making the triangle in Fig. 5.7 commu-
tative, where φ1 is defined by φ1 (z, n) = n + (1/2πi) log z, and log z is the principal
value of the logarithm function on C − {t ∈ R : t ≥ 0}. Its inverse φ−1 −1
1 : p (U ) →
−1
U1 × Z is defined by φ1 (t) = (e , [t]), where [t] denotes the greatest integer
2πit
< t for t ∈ R − Z = p−1 (U1 ). Similarly we can define φ2 : U2 × Z → p −1 (U2 )

by φ2 (z, n) = n + (1/2πi) log z, where log z is the principal value of the logarithm
function on C − {t ∈ R : t ≤ 0}. Its inverse φ−1 −1
2 is defined by φ2 = (e
2πit
, [t + 12 ]),
−1
for t ∈ p (U2 ).
Fig. 5.7 fiber bundle over φ1

U1 × Z / p−1 (U1 )
S1 G
GG w
GG
G www
pU1 GG w
G# www p
{ w
U1
Example 5.2.4 Let B be a connected space and b0 ∈ B. If p : X → B is a covering

projection, then (X, p, B, p −1 (b0 )) is a fiber bundle.
Example 5.2.5 The n-dimensional torus T n is a fiber bundle. Consider T n defined
by the n-fold cartesian product T n = S 1 × S 1 × . . . × S 1 and the map p : Rn → T n
defined by
(t1 , t2 , . . . , tn ) → (e2πit1 , e2πit2 , . . . , e2πitn ).
Then (Rn , p, T n , F) forms a fiber bundle with fiber F which is the set of integer
lattice points in Rn .
We now consider a continuous action
X × G → X, (x, g) → xg
of a discrete topological group G on a topological space X with an aim to construct

a fiber bundle over the orbit space X mod G.
Definition 5.2.6 A discrete topological group G with identity e is said to act (from
the right) properly discontinuously on a topological X if
PD(i) for every x ∈ X , there is a neighborhood Ux of x such that Ux g ∩ Ux = ∅
implies g = e; and
PD(ii) for every pair of elements x, y ∈ X, y ∈/ x G, there are neighborhoods Vx
and Vy of x and y, respectively, such that Vx g ∩ Vy = ∅, for all g ∈ G.
Remark 5.2.7 The condition PD(ii) shows that the orbit space X/G = X mod G
(topologized by the identification map p : X → X mod G, x → x G) is Hausdorff
and hence the condition PD(i) shows that the projection p : X → X mod G, x →
x G is a covering map.
Proposition 5.2.8 If a discrete topological group G acts on a topological space X
properly discontinuously, then (X, p, X mod G, G) is a fiber bundle.
Proof Let y ∈ X mod G. Then there is some x ∈ X such that p(x) =y and a neigh-
borhood Ux of x in X by PD(i). Let Vy = p(Ux ). Then p −1 (Vy ) = Ux g implies
g∈G
that Vy is open and y ∈ Vy . Moreover, p|Ux : Ux → Vy is a homeomorphism. Define
a homeomorphism
φ y : Vy × G → p −1 (Vy ), (z, g) → (( p|Ux )−1 (z))g,
g ∈ G and z ∈ Vy . Then φ−1

y is defined by
φ−1
y (z g) = ( p(z ), g),
for z g ∈ gUx . Consequently, φ y is a homeomorphism such that p ◦ φ y = pVy . Then

open sets {Vy } from a covering of the space X mod G and hence (X, p, X mod G, G)
is a fiber bundle. ❑
5.2 Fiber Bundles: Introductory Concepts 209
Definition 5.2.6 is now redefined for convenience of future discussion.

Definition 5.2.9 A group G of homeomorphisms of a topological space X is said to
be discontinuous if the orbits in orbit space X/G = X mod G are discrete subsets
and G is said to be properly discontinuous if for x ∈ X, there is a neighborhood, Ux
of x in X such that for g, g ∈ G if Ux g meets Ux g , then g = g . The group is said
to act without fixed points if the only element of G having fixed point is the identity
element.
Remark 5.2.10 The Proposition 5.2.8 prescribes an important way in which fiber
bundles arise.
We now study the local cross sections of fiber bundles.
Definition 5.2.11 A map p : X → B is said to have a local cross section s at a

point b ∈ B if there is a neighborhood Ub in B and a map s : Ub → X is such that
p ◦ s = 1U B .
Remark 5.2.12 Every bundle ξ = (E, p, B) may not have a cross section.
Example 5.2.13 The bundle ξ = (D 2 , p, S 1 ) has no cross section. If possible let

ξ have a cross section s : S 1 → D 2 . Then p ◦ s = 1d implies that s is injective.
Otherwise there exists at least one pair of elements b1 , b2 such that b1 = b2 but
s(b1 ) = s(b2 ). Then p(s(b1 )) = p(s(b2 )) implies b1 = b2 , which is a contradiction.
Again since s : S 1 → D 2 is continuous and injective, it contradicts the result, that
every continuous map f : S 1 → D 2 sends at least one pair of antipodal points of S 1
to the same point (see Borsuk–Ulam Theorem).
Remark 5.2.14 Since every fiber bundle has a cross sections, we assume the existence
of cross section in the following Propositions and Theorem (If G is Lie group and H is
a closed subgroup of G, then the cross section of the natural projection p : G → G/H
exists).
Proposition 5.2.15 Let G be a topological group and H be a closed subgroup of G.

Then the projection p : G → G/H, g → H g has a local cross section at any point
of G/H .
Proof Consider the action σ of G on the space G/H , given by G × G/H →

G/H, (g, g H ) → (gg )H . To prove the proposition it suffices to prove that p has
a local cross section at the coset H . Suppose (U, σ) is a local cross section for p at
H . Then for any other point g H of G/H , the set U g is a neighborhood of g H and
the function σg : g · U → G, given by
g H → g · (σ(g −1 g H )), for g H ∈ g · U,
is continuous and such that p ◦ σg = 1g·U . This proves the proposition.

❑
Theorem 5.2.16 Let G be a topological group and H be a closed subgroup of G. If

p : G → G/H, g → g H has a local cross section at H , then for any closed subgroup
A ⊂ H , the natural projection q : G/A → G/H, Ag → H g, is a fiber bundle with
fiber H/A.
Proof By Proposition 5.2.15 the projection p : G → G/H has a local cross section at
every point of G/H . Let x ∈ G/H and (U, σ) be a local cross section of p at x. Define
ψ : U × H/A → G/A, given by (y, h A) → σ(y). h A for y ∈ U, h ∈ H . Then ψ is a
continuous function such that (q ◦ ψ)(y, h A) = q(σ(y) · h A) = σ(y) · h A = σ(y) ·
h A = σ(y) · H = p(σ(y)) = y = 1U (y) for all y ∈ U and h ∈ H . Again Define φ :
(q)−1 (U ) → U × H A, given by, g A → (g H, σ(g H )−1 · g A) for all g A ∈ q −1 (U ).
Then φ is a continuous map such that φ ◦ ψ = 1d and ψ ◦ φ = 1d . ❑
5.3 Hopf and Hurewicz Fiberings
This section studies the various fiberings of spheres discovered by H.Hopf and
W. Hurewicz.
5.3.1 Hopf Fibering of Spheres
This subsection discusses Hopf fiberings given by Hopf (1931, 1935) and consid-
ers the early examples of bundles spaces: three fiberings of spheres: p : S 2n−1 →
S n , n = 2, 4, 8. The simplest of them is the map p : S 3 → S 2 of the 3-sphere on the
2-sphere given by Hopf in 1935, known as a Hopf map.
Theorem 5.3.1 The 3-sphere is decomposed into a family of great circles, called
fibers, with the 2-sphere as a decomposition space.
Proof Let S 3 be represented in C2 as S 3 = {(z 1 , z 2 ) ∈ C2 : |z 1 |2 + |z 2 |2 = 1} and
S 2 be represented as the complex projective line (i.e., as pairs [z 1 , z 2 ] of complex
numbers, not both zero, with the equivalence relation [z 1 , z 2 ] ∼ [λz 1 , λz 2 ], where
λ(= 0)). Define
p : S 3 → S 2 , (z 1 , z 2 ) → [z 1 , z 2 ]/(|z 1 |2 + |z 2 |2 )1/2 .
Then S 3 is a bundle space over S 2 relative to p. If (z 1 , z 2 ) ∈ S 3 and |λ| = (|z 1 |2 +

|z 2 |2 )1/2 = 1, then (λz 1 , λz 2 ) is also in S 3 and they have the same image point in S 2 .
Conversely, if p(z 1 , z 2 ) = p(z 1 , z 2 ), then (z 1 , z 2 ) = (λz 1 , λz 2 ) for some λ having
|λ| = 1. Hence the inverse image of a point of S 2 is obtained by any point of the
inverse image by multiplying it by eiθ (0 ≤ θ ≤ 2π). Hence the inverse image is just
a great circle of S 3 . This shows that the 3-sphere is decomposed into a family of
great circles, called fibers, with the 2-sphere as a decomposition space. ❑
5.3 Hopf and Hurewicz Fiberings 211
Remark 5.3.2 We now generalize the Hopf map p : S 3 → S 2 through discussion

of some spaces that arise in projective geometry. Let F denote one of the fields R
of real numbers, C of complex numbers or division ring H of quaternions and F n
be the right vector space whose elements are ordered sets of n elements of F. If
x = (x 1 , x2 , . . . , xn ) ∈ F n and λ ∈ F, then xλ = (x1 λ, . . . , xn λ). Define the inner
n
product x and y in F n by x, y = x i yi , where x i is the conjugate of xi . Then
1
y, x = x, y, xλ, y = λx, y, x, (yλ) = x, yλ.
In particular, x, y = 0 iff y, x = 0. This shows that the relation of orthogo-
nality is symmetric. Let S be the unit sphere in F n which is the locus x, x = 1. If
G n is the orthogonal, unitary, or sympletic group according as F = R, C or H, then
each G n is a compact Lie group (see Appendix A). Let F P n be the projective space
associated with F, then it can be thought of the set of all lines through the origin
n+1

in F n+1
= F ⊕ F ⊕ · · · ⊕ F . Then R P n , CP n , and HP n are called n-dimensional
real, complex, and quaternionic projective spaces. We topologize F P n be consider-
ing it as a quotient space of F n+1 − {0}. Then each point of F n+1 − {0} determines
a line through the origin 0. Thus if x and y are nonzero elements of F n+1 , we say
that x ∼ y iff there is an element λ(= 0) ∈ F such that y = xλ. This is an equiva-
lence relation and define F P n to be the set of equivalence classes with the quotient
topology.
There is a natural map F n+1 − {0} → F P n , which is continuous, and gives, on
restriction to the unit sphere of F n , maps
pn : S n → R P n ,
qn : S 2n+1 → CP n ,
and
rn : S 4n+3 → HP n .
Generally we write p in place of pn , qn or rn , where there is no confusion.
Example 5.3.3 (i) (Complex Hopf bundle) ξ = (S 2n+1 , p, CP n , S 1 ) is a trivial

fiber bundle with fiber S 1 .
(ii) (Quaternionic Hopf bundle) γ = (S 4n+3 , p, HP n , S 3 ) is a locally trivial fiber
bundle with fiber S 3 .
(iii) (Real Hopf bundle) η = (S n , p, R P n , Z2 ) is a locally trivial fiber bundle with
fiber Z2 .
Remark 5.3.4 The Example 5.3.3 indicates the importance of bundle theory to com-
pute the homotopy groups of sphere where results are only partly known (see Chap. 7).
5.3.2 Hurewicz Fibering
This subsection studies Hurewicz fibering. This fibering is due to W. Hurewicz and
named after him.
Definition 5.3.5 A Hurewicz fibering is a map p : X → B that has the homotopy

lifting property (HLP) with respect to any space Y .
Let f : Y → Z be continuous. Define E f = {(y, w) ∈ Y × Z 1 : w(0) = f (x)} ⊂

Y × Z 1 with the induced topology. Then there are maps
λ : E f → Z , μ : E f → Y, α : Y → E f ,
defined by λ(y, w) = w(1), μ(y, w) = y, α(y) = (y, c f (y) ), where , c f (y) is the con-
stant path at f (y). Hence the diagram in Fig. 5.8 commutes, i.e., λ ◦ α = f and
μ ◦ α = 1d .
The existence of maps
λ : E f → Z , μ : E f → Y, α : Y → E f ,
proves the following proposition:
Proposition 5.3.6 Let f : Y → Z be continuous. Define E f = {(y, w) ∈ Y × Z 1 :

w(0) = f (x)} ⊂ Y × Z 1 with the induced topology. Then
(i) Y E f .
(ii) λ : E f → Z is a Hurewicz fibering with fiber F f = {(y, w) ∈ Y × Z 1 : w(0) =
f (x), w(1) = ∗}.
Proof (i) Define a homotopy H : E f × I → E f , (y, w, t) → (y, wt ), where

wt (s) = w(st). Then α ◦ μ 1d . Again μ ◦ α = 1d and hence Y E f .
(ii) Consider the commutative diagram in Fig. 5.9. Let h(a) = (h 1 (a), h 2 (a)).
Define β : A × I → E f ⊂ Y × Z 1 , (a, t) → (β1 (a, t), β2 (a, t)), where β1
(a, t) = h 1 (a) and
Fig. 5.8 Commutative Ef

O @@
triangle for Hurewicz @@λ
@@
fibering @
μ
>Z
α
||
||
||| f
|
Y
Fig. 5.9 Commutative h / Ef

A × {0}
rectangular diagram for
λ
Hurewicz fibering
H /Z
A×I
5.3 Hopf and Hurewicz Fiberings 213

h 2 (a)(s(1 + t)), 0 ≤ s ≤ 1/(1 + t)
β2 (a, t)(s) =
H (a, (1 + t)s − 1), 1/(1 + t) ≤ s ≤ 1.
Then β(a, t) ∈ E f , because, the equality ( f ◦ β1 )(a, t) = β2 (a, t)(0) follows

from ( f ◦ h 1 )(a) = h 2 (a)(0). The map β is clearly continuous, since β1 and β2
are also. Moreover, β(a, 0)(s) = h(s) and (λ ◦ β)(a, t) = H (a, t). Clearly,
F f is its fiber.
❑
Theorem 5.3.7 (Hurewicz) Let p : X → B be a covering map. Suppose B is
paracompact, and there is an open covering {U j } of B such that for each Ui ∈
{U j }, p| p−1 (Ui ) : p−1 (Ui ) → Ui is a fibration. Then p is a fibration.
Remark 5.3.8 The proof is long and complicated. We omit the proof referring the
reader to [Dugundji, pp. 400].
An important consequence of Hurewicz theorem.
Proposition 5.3.9 Let p : X → B be the projection of a fiber bundle (X,p,B,F) such
that the base space B is paracompact. Then p is a fibration.
Remark 5.3.10 The decompositions of compact Lie groups modulo their closed
subgroups produce fiber bundles, which serve as a valuable source of examples. We
recommend the survey articles by [H. Samelson, Topology of Lie groups, Bull. Amer.
Math. Soc.58 (1952), 2–37] and Borel (1955).
5.4 G-Bundles and Principal G-Bundles
This section studies G-bundles and principal G-bundles defined by transformation

groups G and shows that if X is a simply connected space and G is a properly discon-
tinuous group of homeomorphisms of X , then the fundamental group π1 (X mod G)
of the orbit space X mod G is isomorphic to G.
A G-bundle is a bundle with an additional structure derived from the action of the
topological group G on a topological space. For a Lie group G named after Sophus
Lie (1842–1899), the principal G-bundles are studied in Sect. 5.11. Transformation
groups obtained by actions of topological groups on topological spaces (see Appendix
A), are now used to study G-bundles defined by transformations groups G.
Definition 5.4.1 Let X be a right G-space. Then the set X mod G = {x G : x ∈
X } is the set of all orbits of X under the action of G on X , equipped with the
quotient topology, which is the largest topology such that the projection map p :
X → X mod G, given by x → x G, is continuous. The quotient space X mod G is
called the orbit space of X .
Clearly, p is an identification map.
Proposition 5.4.2 Let X be a right G-space. Then for each g ∈ G, the map φg : X →
X, x → xg is a homeomorphism and the projection p : X → X mod G, x → x G
is an open map.
Proof Clearly, φg is a homeomorphism with its inverse φg−1 for each g ∈ G. For the

second part, let U be an open subset of X . Then p−1 ( p(U )) = U g is a union of
g∈G

open sets in X and hence U g is an open set in X mod G. Consequently, p(U ) is
g∈G
an open set of X mod G for each open set U of X . ❑
Definition 5.4.3 Let X and Y be two right G-spaces. A map f : X → Y is said to

be a G-morphism if f (x · g) = f (x) · g holds for all x ∈ X and for all g ∈ G.
Clearly, f (x G) ⊂ f (x)G for each x ∈ X .
Definition 5.4.4 A bundle (X, p, B) is said to be G-bundle if the bundles (X, p, B)

and (X, p X , X mod G) are isomorphic for some G-space structure on X by an iso-
morphism (1d , f ) : (X, p X , X mod G) → (X, p, B) i.e., there exists a homeomor-
phism f : X mod G → B making the diagram in Fig. 5.10 commutative.
Proposition 5.4.5 Given G-spaces X and Y and a G-morphism f : X → Y , there

exists a bundle morphism ( f, f˜) : (X, p X , X mod G) → (Y, pY , Y mod G).
Proof Let ξ(X ) = (X, p X , X mod G) be the bundle corresponding to a G-space X .

Then ξ(Y ) = (Y, pY , Y mod G). Clearly, the map f : X → Y induces a quotient
map f˜ : X mod G → Y mod G, given by f˜(x G) → f (x)G. Since the diagram
in the Fig. 5.11 is commutative, the pair ( f, f˜) is a bundle morphism. Clearly,
every G-space X determines a bundle (X, p X , X mod G), where p X : X → X mod
G, x → x G. ❑
Let G be the group of covering transformations(see Chap. 4) of a covering projec-

tion p : X → B. Then the action of G on X is properly discontinuous. Is its converse
true ?
Fig. 5.10 G-bundle (X, p, B) X

1d /X
pX p
f
X mod G /B
Fig. 5.11 G-bundle f

/Y
X
morphism
pX pY
f˜
X mod G / Y mod G
5.4 G-Bundles and Principal G-Bundles 215
Theorem 5.4.6 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. Moreover,
if X is connected, then this covering projection is regular and G is its group of
covering transformations.
Proof Let p : X → X mod G be the projection. Then p is continuous and open

by Proposition 5.4.2. Let U be an open subset of X satisfying the property that
whenever U g meets U g , then g = g . We claim that p(U ) is evenly covered by p. By
assumption, U shows that {U g}g∈G is a disjoint collection of open sets whose union is
p −1 ( p(U )). It is sufficient to prove that p|U g : U g → p(U ) is a bijection. If x ∈ U ,
then p(xg) = p(x) and hence p(U g) = p(U ). If p(xg) = p(x g), with x, x ∈ U ,
then for some s ∈ G, xg = xsg. Consequently, U g meets x sg and g = sg. Hence
s = e and xg = x g. Thus p : U g → p(U ). Since 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. Finally,
as the group of covering transformations is transitive on each fiber, it follows that
the covering projection is regular. ❑
Corollary 5.4.7 Let X be a simply connected space and G be a properly discontin-

uous group of homeomorphisms of X . Then the fundamental group π1 (X mod G) is
isomorphic to G.
Proof It follows by using Theorem 5.4.6 that G is the group of covering transforma-
tions of the regular covering projection p → X mod G. Then the Corollary follows
from Theorem 5.4.6. ❑
We now study from the homotopical view point special G-bundles, known as,
principal G-bundles, which come with an action of some topological G.
Definition 5.4.8 Let G be a topological group with identity e. A (locally trivial)

principal G-bundle is a fiber bundle ξ = (X, p, B, G) with a continuous right action
of G on X , i.e., a continuous map X × G → X, (x, g) → x · g such that there is
an open covering {U j : j ∈ J } of B and for each j ∈ J , there is a homeomorphism
φ j : U j × G → p −1 (U j ) with the conditions:
(i) ( p ◦ φ j )(b, y) = b and
(ii) φ j (b, g) = φ j (b, e) · g for b ∈ U j and g ∈ G.
Example 5.4.9 (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).
We now define the morphisms of G-bundles.

Definition 5.4.10 Let ξ = (X, p, B, G) and ξ = (X , p , B , G) be two principal
G-bundles. A morphism φ : ξ → ξ of principal G-bundles is a pair of maps ( f, h),
where f : X → X , h : B → B are such that the diagram in the Fig. 5.12 is com-
mutative and f (x · g) = f (x) · g for all g ∈ G and x ∈ X.

X / X
G-bundles
p p
h

B / B
Remark 5.4.11 Let ξ = (X, p, B, G) be a principal G-bundle. Then the action of G

on X is free and transitive on each b ∈ B and p(x1 ) = p(x 2 ) iff there exists some g
such that x 2 = x1 g.
Definition 5.4.12 Let ξ = (X, p, B, G) and η = (Y, q, B, G) be two principal G-

bundles over B. Then they are said to be isomorphic or equivalent, denoted by ξ ∼
=η
if there exists a homeomorphism f : X → Y such that
(i) f is equivariant, i.e., f (xg) = f (x)g for all x ∈ X and g ∈ G;
(ii) q ◦ f = p.
Definition 5.4.13 Let ξ = (X, p, B, G) be a G-bundle. Then the family {(U j ,

φ j )} j∈J for ξ given in Definition 5.4.8 is called trivializing cover for ξ and the bundle
ξ is called numerable if there is a partition of unity subordinate to a trivializing cover
for ξ.
Example 5.4.14 Any principal G-bundle over a paracompact space is numerable.
Let B be a space and [ξ] denote the isomorphism class of numerable principal
G-bundle ξ over B. The set of isomorphism classes of numerable principal G-bundles
over B is denoted by K G (B).
Definition 5.4.15 Let ξ = (X, p, B) be a principal G-bundle. If f : A → B is

a map, then f induces a principal G-bundle f ∗ (ξ) = (Y, q, A, G), where Y =
{(a, x) ∈ A × X } such that f (a) = p(x) and as projection q : Y → A, is given
by (a, x) → a and the action of G on Y is given by (a, x) · g = (a, x · g) for
all a ∈ A, x ∈ X and g ∈ G. If f˜ : Y → X, (a, x) → x, then the diagram in the
Definition 5.4.16 Let ξ = (X, p, B) be a principal G-bundle and f : A → B be a

given map. Then the induced G-bundle f ∗ (ξ) = (Y, q, A, G) over A is called the
pull-back of ξ or G-bundle induced by f .
Fig. 5.13 Induced G-bundle f˜

Y /X
q p

A /B
f
5.4 G-Bundles and Principal G-Bundles 217
We recall the following definition of an action of a topological group.

Definition 5.4.17 An action of a group G on X is said to be free if for all g(= e) ∈
G, g · x = x for every x ∈ X . It is called effective if for all g(= e) ∈ G there exists
an element x ∈ X such that g · x = x.
Definition 5.4.18 (Principal G-bundle) Let G be a topological group and B be a

topological space. A principal G-bundle over B consists of a fiber bundle p : X → B
together with an action X × G → X such that
PG(i) the shearing map
T : X × G → X × X, (x, g) → (x, x · g)
maps X × G homeomorphically to its image;

PG(ii) B = X mod G and p : X → X mod G is the quotient map;
PG(iii) for all b ∈ B, ∃ an open neighborhood U of b such that p : p−1 (U ) → U
is G-bundle isomorphic to the trivial bundle q : U × G → U (i.e., there
exists a homeomorphism φ : p−1 (U ) → U × G satisfying q ◦ φ = p,
where (x, g ) · g = (x, gg )).
Remark 5.4.19 The shearing map T is injective iff the action of G on X is free and
hence by PG(i), 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). Moreover, a free
action produces a translation function ρ : 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.
Let X 1 , X 2 , . . . , X n , . . . topological
spaces such that X 1 ⊂ X 2 ⊂ · · · ⊂ X n ⊂
· · · are inclusions. Let X ∞ = lim X n , with weak topology (i.e., A ⊂ X is closed
n→∞
n
iff A ∩ X n is closed in X n for each n).
Proposition 5.4.20 Let G be a topological group. If X 1 ⊂ X 2 ⊂ · · · ⊂ X n · · · be

inclusions such that X ∞ = lim ∪X n . Let σ : X ∞ × G → X ∞ be a free action which
n→∞
restricts to an action of G on X n for each n (i.e., σ(X n × G) ⊂ X n for each n).
Then this action of G on X n is free and if the translation function for this action is
continuous for all n, then the translation function for σ is continuous
Proof Since the action of G on X ∞ is free, the action of G on each X n is free for all
n. Let
T : X n × G → X n × X n , (x, g) → (x, x · g)
be the shearing function for X n and Yn = Im T ⊂ X n × X n . Let

ρn : Yn → G be the
translation function for X n . Then as a topological space Y∞ = Yn = lim Yn and
n→∞
n
ρ∞ |Yn = ρn . This shows that ρ∞ is continuous. ❑
Remark 5.4.21 Let G be a topological group and p : X → B be a principal G-

bundle. Let Y be a G-space on which the action of G is effective. The fiber bundle
with structure group G obtained from p and Y is given by q : (Y × X ) mod G → B,
where (y, x) · g = (y · g, x · g) and q(y, x) = p(x). This is sometimes called “G-
bundle for a fiber bundle with structure group G”. In particular, when Y = R n (Cn ),
and G is the orthogonal group O (n, R) (unitary group U (n.C)) acting in the standard
way, and the restrictions of the trivialization maps to each fiber are linear transforma-
tions, such a fiber bundle is called an n-dimensional real (complex) vector bundle.
A one-dimensional vector bundle is called a line bundle.
5.5 Homotopy Properties of Numerable Principal

G-Bundles
This section studies numerable principal G-bundles over B from the view point
of homotopy theory and constructs a contravariant functor K G : Htp → Set corre-
sponding to a given topological group G. This functor plays an important role in the
study of homotopy theory.
Theorem 5.5.1 Let ξ be a numerable principal G-bundle over B × I . Then the

bundles ξ, (ξ|(B × {1})) × I and (ξ|(B × {0})) × I are G-isomorphic. If h i : B →
B × I is the map h i (b) = (b, i) for i = 0, 1, then the bundles h ∗0 (ξ) and h ∗1 (ξ) are
B-isomorphic.
Proof Let f (b, t) = (b, 1). Then by Ex. 11(b) of Sect. 5.13, f ∗ (ξ) and ξ are isomor-
phic principal G-bundles over B × I . Again f ∗ (ξ) and (ξ|(B × {1}) × I ) are isomor-
phic principal G-bundles over B × I. Hence it follows that ξ and (ξ|(B × {1}) × I )
are also isomorphic principal G-bundles. In a similar way, it follows that ξ and
(ξ|(B × {0}) × I ) are also isomorphic. Again since f ◦ h 0 = h 1 and the bundles
f ∗ (ξ)|(B × {0}) and ξ|(B × {0}) are G-isomorphic, it follows that h ∗1 (ξ) = h ∗0 ◦ f ∗
and h ∗0 (ξ) are isomorphic principal G-bundles. ❑
Corollary 5.5.2 Let ξ = (X, p, B) be a numerable principal G-bundle over B and

f, g : A → B are homotopic maps. Then the principal bundles f ∗ (ξ) and g ∗ (ξ) are
isomorphic over A.
Proof Let Ht : f g : A → B and h i : A → A × I be the map defined by h i (a) =

(a, i) for i = 0, 1. Then H0 ◦ h 0 = f and H1 ◦ h 1 = g. Hence the principal bundles
f ∗ (ξ) and h ∗0 ◦ f ∗ (ξ) are isomorphic by Theorem 5.5.1. ❑
Theorem 5.5.3 For each space B, let K G (B) be the set of isomorphism classes of
numerable principal G-bundles over B. Then K G : Htp → Set is a contravariant
functor from the homotopy category Htp of topological spaces and their homotopy
classes of maps to the category Set of sets and their functions.
5.5 Homotopy Properties of Numerable Principal G-Bundles 219
Proof Define the object function and morphism function as follows:

For each object B in Htp , K G (B) is assigned to be the set of isomorphism classes
of numerable principal G-bundle over B. For a homotopy class [ f ] of f : A → B,
define the morphism function K G ([ f ] : K G (B) → K G (A), by K G ([ f ]) = { f ∗ (ξ)}.
By Corollary 5.5.2, the function K G ([ f ]) is well defined. Consequently, K G ([ f ]) :
K G (B) → K G (A) is a function. Moreover, if f : A → B and g : B → C are two
maps and η is a numerable principal G-bundle over C, then the bundle (g ◦ f )∗ (η)
and f ∗ (g ∗ (η)) are isomorphic for every numerable principal G-bundles η over C.
This shows that K G ([g] ◦ [ f ]) = K G ([ f ]) ◦ K G ([g]). Similarly, η and 1C ∗ (η) are
isomorphic and hence the function K G ([1C ]) is the same as the identity function on
K G ([C]). Consequently, K G is a contravariant functor. ❑
Corollary 5.5.4 Let ξ be a numerable principal G-bundle over B. If f : A → B is
a homotopy equivalence, then K G ([ f ]) : K G (B) → K G (A) is a bijection.
Proof Let f : A → B be a homotopy equivalence. Then there exists a map g :
B → A such that g ◦ f 1 A and f ◦ g 1 B . Hence by using the property of
the contravariant functor K G (see Theorem 5.5.3), it follows that K G ([g ◦ f ]) =
K G ([ f ]) ◦ K G ([g]) is the identity function. This shows that K G ([ f ]) is a surjec-
tive function. Similarly, K G ([ f ◦ g]) = K G ([g]) ◦ K G ([ f ]) is the identity function
implying that K G ([ f ]) is a monomorphism. Hence K G ([ f ]) is a bijection. ❑
Corollary 5.5.5 Let ξ = (X, p, B) be a numerable principal G-bundles over B. If
B is contractible, then ξ is trivial.
Proof Let the space B be contractible. Then the space B is homotopy equivalent
to a point {∗}. Again K G ({∗}) has only one point, the isomorphism class of trivial
bundle. Hence the corollary follows from Corollary 5.5.4. ❑
Let ξ0 = (X, p0 , B0 , G) be a fixed numerable principal G-bundle. For each space
B, define a function φξ0 (B) : [B, B0 ] → K G (B) by φξ0 (B)([ f ]) = [ f ∗ (ξ0 )], [ f ] ∈
[B, B0 ]. This function is well defined by Corollary 5.5.2.
Hence it follows that
Theorem 5.5.6 Given a fixed numerable principal G-bundle ξ0 = (X 0 , p0 ,
B0 , G), φξ0 : [−, B0 ] → K G is a contravariant functor from the category Htp to
the category Set .
We now define a class of principal G-bundles with the help of φξ0 .
Definition 5.5.7 Let ξ0 = (X, p0 , B0 , G) be a principal G-bundle. Then ξ0 is said
to be universal if
(i) ξ0 is numerable; and
(ii) φξ0 : [−, B0 ] → K G is a natural equivalence.
We now characterize universal G-bundles.
Proposition 5.5.8 A numerable principal G-bundle ξ0 = (X 0 , p0 , B0 , G) is univer-
sal iff
(i) for each numerable principal G-bundle ξ = (X, p, B, G), there exists a map
f : B → B0 such that ξ and f ∗ (ξ0 ) are B-isomorphic; and
(ii) if f, g : B → B0 are two maps such that f ∗ (ξ0 ) and g ∗ (ξ0 ) are isomorphic, then
f g.
Proof The condition (i) shows that the function φξ0 (B) : [B, B0 ] → K G (B) is sur-
jective and condition (ii) shows that φξ0 (B) is injective. Hence the proposition
follows. ❑
5.6 Classifying Spaces: The Milnor Construction
This section presents a simple elegant method of construction of a classifying

space and a universal principal fiber space associated with a principal fiber space
(X, p, B, G), the method invented by John Willard Milnor (1931-) in 1956, where
G is any topological group. The classification of numerable principal G- bundles up to
homotopy classes which is a very important problem. This classification asserts that
for every topological group G, there exists a topological space BG such that for
every pointed topological space B, there is a bijective correspondence between the
isomorphism classes of numerable principal G-bundles over B and the homotopy
classes[B, BG ] of base point preserving continuous maps from B to BG . It is an
important example of a numerable principal G-bundle. Milnor was awarded Fields
Medal in 1962 and the Abel Prize in 2011.
Example 5.6.1 (Milnor construction) Let G be a topological group. The universal

fiber space is defined as an infinite join X G = G ∗ G ∗ · · · ∗ G · · · . An element of
X G denoted by x, t is written as
x, t = (t0 x0 , t1 x1 , . . . , tr xr , . . . ),

where each xi ∈ G and ti ∈ [0, 1] such that only a finite number of ti = 0 and ti = 1.
0≤
We say that x, t = x , t in the set X G iff ti = ti for each i and xi = x i for all
i with ti = ti > 0. We note that if ti = ti = 0, then xi and xi may be different but
x, t = x , t in the set X G . We define an action of G from the right X G × G → X G
by the relation
x, tg = xg, t or (t0 x0 , t1 x1 , . . . )g = (t0 x0 g, t1 x1 g, . . . ) (5.1)
for all g ∈ G. We define a topology on X G in such a way that X G becomes a G-space.

We consider two families of functions f i : X G → [0, 1] for 0 ≤ i, which assigns
to the element (t0 x0 , t1 x1 , . . . ) the component ti ∈ [0, 1] and gi : f i−1 (0, 1] → G
for 0 ≤ i, which assigns to the element (t0 x0 , t1 x1 , . . . ) the component xi ∈ G. We
observe that xi cannot be uniquely defined outside f i−1 (0, 1] in a natural way. For
5.6 Classifying Spaces: The Milnor Construction 221
α ∈ X G and g ∈ G, there are the following relations between the action of G and
the functions f i and gi : gi (αg) = gi (α)g and f i (αg) = f i (α). The set X G is made
into a topological space by endowing X G the smallest topology such that each of the
functions fi : X G → [0, 1] and gi : f −1 (0, 1] → G is continuous, where f i−1 (0, 1]
has the subspace topology. From the relations in (5.1), it follows that X G is a G-space
where the G-set structure map X G × G → X G is continuous. We denote the orbit
space X G mod G by BG , the quotient map p : X G → BG and the resulting bundle
ωG = (X G , p, BG ). This is known as Milnor construction and ωG is an example of
principal G-bundle.
We now illustrate this principal G-bundle ωG in some concrete situations.
Example 5.6.2 Let S n be the n-sphere in Rn+1 (n ≥ 1). If a : S n → S n , x → −x is

the antipodal map, then a 2 = a ◦ a is identity 1 S n . The group G = {1, a} = Z2 is a
group of homeomorphisms of S n and the space X G = S n . The action of G on X G is
given by S n × G → S n ,
x · g = x if g = 1
= −x if g = a.
Hence the quotient space X G mod G = BG = R P n . Hence ωG = (S n , p, R P n )

is a numerable principal G-bundle for dimensions ≤ (n − 1).
Example 5.6.3 If G = S 1 and S 2n+1 is the (2n + 1)-sphere in R2n+2 , then the action
of S 1 on S 2n+1 is given by (z 0 , z 1 , . . . , z n )eiθ = (eiθ z 0 , eiθ z 1 , . . . , eiθ z n ). If X G =
G ∗ G ∗ · · · ∗ G ∗ · · · (infinite join), then X G mod G = BG = CP n . Hence wG =
(S 2n+1 , p, CP n ), is a principal numerable G-bundle of dimensions ≤ 2n.
Definition 5.6.4 (Milnor 1956) Let G be a topological group. The functor K G :

Htp → Set is said to be representation if there exists a space BG , called the clas-
sifying space of K G and an element ξG = (X G , pG , BG ), called universal bundle
in K G (BG ), such that there is a natural equivalence between the functor K G and
[−, BG ], defined from the category Htp to the category Set .
Remark 5.6.5 The Definition 5.6.4 implies that for any space, the function ψ(B) :
[B, BG ] → K G (B), [ f ] → [ f ∗ (ξG )]) is a bijection. Moreover, any space homotopy
equivalent to BG is also a classifying space for K G .
Remark 5.6.6 J. Feldbau reduced in 1939 the classification of principal fiber bun-
dles (X, p, B, G) with a given base S n for n ≥ 2 to a problem in homotopy theory
(Feldbau 1939).
We now consider the effect of continuous homomorphism on the corresponding

isomorphism classes of principal bundles.
Let G and H be topological groups and α : H → G be a continuous homomor-
phism. Define an action of H on G, H × G → G, h · g = α(h)g for all h ∈ H and
g ∈ G. Let ξ = (X, p, B, H ) be a principal H -bundle. Define an action:
(X × G) × H → X × G, (x, g) · h → (x · h, h −1 · g).
Let (X × G) mod H be the resulting orbit space and q : (X × G) mod G → B,

[x, g] → p(x) be the projection map. Define another action
((X × G) mod G) × G → (X × G) mod G, ([x, g], g ) → [x, gg ].
Let {φ j : U j × H → p −1 (U j )} be a trivial covering for ξ. Define
ψ j : U j × G → q −1 (U j ), (b, g) → [φ j (b, e), e].
Then the family {U j , ψ j } is a trivializing family for q. Hence we have a principal

G-bundle (X × G) mod (G, q, B, G) and a natural transformation α∗ : K H → K G
under the function
α∗ (B) : K H (B) → K G (B), {ξ} → {α∗ (ξ)}.
Definition 5.6.7 Let ξ = (X, p, B, G) be a principal G-bundle. ξ is said to have an

H -structure if there exists a principal H -bundle η = (Y, q, B, H ) and a continuous
homomorphism α : H → G such that {α∗ (η)} = {ξ}.
We now study the H -structures on ξ from the view point homotopy theory
Proposition 5.6.8 Let H and G be topological groups and α : H → G be a con-

tinuous homomorphism. Then the natural transformation α∗ : K H → K G induces a
unique map(upto homotopy) B H → BG
Proof Clearly, K H (Y ) ∼
= [Y, B H ] and K G (Y ) = [Y, BG ]. Let Bα : B H → BG be
the classifying map for the principal G-bundle (X H × G) mod H → B H . Then we
have the commutative diagram in Fig. 5.14 where Bα∗ [ f ] = [Bα ◦ f ] for any map
f : Y → BH . ❑
We now use the notation of Milnor construction (see Example 5.6.1)
Theorem 5.6.9 (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 See Milnor construction (see Example 5.6.1) and Husemöller (1966). ❑
The above discussion showing the classification of numerable principal G-bundles

reduced to homotopy theory is summarized in the basic and important theorem:
Fig. 5.14 A natural α / KG (Y )

KH (Y )
transformation ∼ ∼
= =
Bα∗

[Y, BH ] / [Y, BG ]
5.6 Classifying Spaces: The Milnor Construction 223
Theorem 5.6.10 Given any topological group G, there exists a topological space
called classifying space BG with the property that for any space B, there exists a
bijection between the set [B, BG ] of homotopy classes of base point preserving maps
from B to BG and the set of isomorphism classes of numerable principal G-bundle
over B.
Example 5.6.11 Let G = S 3 be the multiplicative group nonzero quaternions and
S 4n+1 be the (4n + 1)-sphere in R4n+2 . If X G = G ∗ G ∗ · · · ∗ G (infinite join), then
X G mod G = BG = HP n . Hence wG = (S 4n+1 , p, HP n ) is a principal numerable
G-bundle.
5.7 Vector Bundles: Introductory Concepts
This section conveys introductory concepts of vector bundles. Vector bundles and
their homotopy classifications play a very important role in mathematics and physics.
Vector bundles form a special class of fiber bundles for which every fiber has the
structure of a vector space compatible on neighbouring fibers and the structure group
is a group of linear automorphisms of the vector space. The concept of vector bundles
arose through the study of tangent vector fields to smooth geometric objects such as
spheres, projective spaces, and more generally, manifolds. This section introduces the
concepts of vector bundles, Gauss maps named after C.F. Gauss (1777–1855) and also
gives homotopy classifications of vector bundles. For more homotopy classifications
of vector bundles see Sect. 5.9.
A vector bundle over a topological B is a family of vector spaces continuously
parametrized by B. If the fiber of a vector bundle ξ is R n , then ξ is said to be
finite-dimensional with dimξ =n. On the other hand 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.
Let F denote one of the fields R, C or division ring H of quaternionic numbers.
Definition 5.7.1 An n-dimensional F-vector bundle over B is a fiber bundle ξ =

(X, p, B, F n ) together with the structure of an n-dimensional vector space over F
on each fiber Eb = p −1 (b) such that there is an open covering {U j : j ∈ J } of B and
for each j ∈ J , a homeomorphism φ j : U j × F n → p −1 (U j ) with p ◦ φ j = pU j
and (φ j |{b} × F n ) : {b} × F n → p −1 (b) is an isomorphism of vector spaces over
F for each b ∈ U j . Each φ j is called a coordinate transformation.
Remark 5.7.2 An n-dimensional vector bundle ξ = (X, p, B, F n ) satisfies the fol-

lowing local triviality condition:
Each point b of B has an open neighborhood U and an U -isomorphism ψ :
U × F n → p −1 (U ) with the property that ψ|{b}×F n : {b} × F n → p −1 (b) is a vector
space isomorphism for each b ∈ U . The F-vector bundle ξ is said to be a real, complex
or quaternionic vector bundle according as F = R, C or H. Sometimes we say that
ξ = (X, p, B) is an n-dimensional F-vector bundle over B.
Example 5.7.3 (i) For any space B, the trivial bundle (B × F n , p, B, F n ) is an

n-dimensional F-vector bundle.
(ii) For n ≥ 1, the tangent bundle T (S n ) of the n-sphere S n is the fiber bundle
T (S n ) = (X, p, S n , Rn ), where X = {(x, y) ∈ Rn+1 × Rn+1 : ||x|| = 1 and
x, y = 0}, p : X → S n is the projection given by (x, y) → x, is an
n-dimensional real vector bundle. To see this we take Ui ⊂ S n to be the open
set Ui = {x ∈ R n+1 : ||x|| = 1, xi = 0}, 1 ≤ i ≤ n + 1 and define
φi : Ui × Rn → p −1 (Ui ), (x, y) → (x, fi (y) − x, f i (y)x),
where f i : Rn → Rn+1 is given by
(y1 , y2 , . . . , yn ) → (y1 , y2 , . . . , yi−1 , 0, yi , . . . , yn ).
Then φi are linear maps on each fiber such that φi are homeomorphisms and
satisfy the relation p ◦ φi = pUi .
(iii) For n ≥ 1, the normal bundle N (S n ) over S n is the fiber bundle ξ = (X, q, S n ,
R1 ), where
X = {(x, y) ∈ Rn+1 × Rn+1 : ||x|| = 1, y = r x, r ∈ R}
and q : X → S n , (x, y) → x. Define
φ : S n × R1 → X, (x, r ) → (x, r x)
and
ψ : X → S n × R1 , (x, y) → (x, x, y).
Then φ is a homeomorphism with inverse ψ. Consequently, ξ = N (S n ) is an

1-dimensional real trivial bundle.
(iv) The bundle γrn = (X, p, G r (F n ), F n ), where
X = {(V, y) ∈ G r (F n ) × F n : p(V, y) is the orthogonal projection of y into V },
is an n-dimensional F-vector bundle, where G r (F n ) is the Grassmann mani-

fold of r -dimensional subspaces of F n .
We now show that the local triviality condition of a vector bundle provides the
following continuity proposition.
Proposition 5.7.4 Let ξ = (X, p, B, F n ) be an n-dimensional vectors bundle. Then
p : X → B is an open map. Moreover, the fiber preserving function f : X ⊕ X →
X, (x, x ) → x + x and g : F × X → X, (α, x) → αx, α ∈ F, are continuous.
Proof Let φ : U × F n → p−1 (U ) be a local coordinate of ξ. Then for f |( p−1 (U ) ⊕
p −1 (U )), the above statement is valid. Similarly, for the function restricted to
5.7 Vector Bundles: Introductory Concepts 225

X /Y
vector bundles
p q
g
B /A
p −1 (U ), the above statement is also valid. Since the family of p−1 (U ) is an open
covering of X , the functions f and g are continuous. ❑
We now show that the set of cross sections of an F-vector bundle ξ = (X, p, B)
form a module over the ring R of all F-vector functions continuous on B.
Proposition 5.7.5 Let ξ = (X, p, B) be an n-dimensional F-vector bundle over B.

The cross sections of ξ of forms a module over the ring of continuous F-valued
functions on B.
Proof Let s, s be two cross sections of ξ and f : B → F be a map. Then the function
s + s : B → X , given by
(s + s )(b) = s(b) + s (b)
is a cross section of ξ and the function f s : B → X , given by ( f s)(b) = f (b)s(b)

is also a cross section of ξ for all functions f : B → F. Finally, the map b → 0 ∈
p −1 (b) is a cross section (zero cross section). Let φ : U × F n → p−1 (U ) be a local
coordinate of ξ over U . Suppose φ−1 (s(b)) = (b, g(b)) and φ−1 (s (b)) = (b, g (b))
for all b ∈ B, where g : U → F n and g : U → F n are maps. Then
φ−1 ((s + s )(b)) = (b, g(b) + g (b)), φ−1 (( f s)(b)) = (b, f (b)g(b))
and
φ−1 (0)(b) = (b, 0)
for all b ∈ U. Consequently, s + s , f s and 0 are continuous and hence they are all
cross sections. Consequently, the proposition follows. ❑
Definition 5.7.6 Let ξ = (X, p, B) and η = (Y, q, A) be two vector bundles. A

vector bundle morphism ( f, g) : ξ → η consists of a pair of maps f : X → Y and
g : B → A such that the diagram in the Fig. 5.15 is commutative (i.e., q ◦ f = g ◦ p)
and f | p−1 (b) : p −1 (b) → q −1 (g(b)) is a linear transformation for each b ∈ B. For the
particular case, when B = A, a B-morphism of vector bundles f : ξ → η is defined
by a morphism of the form ( f, 1 B ) : ξ → η.
Remark 5.7.7 If f : ξ → η is a B-morphism of vector bundles, then q ◦ f = p and

f | p−1 (b) : p −1 (b) → q −1 (b) is a linear transformation for each b ∈ B (Fig. 5.16).
Definition 5.7.8 Let ξ = (X, p, B, F n ) and η = (Y, q, A, F n ) be two

n-dimensional F-vector bundles. A vector bundle morphism ( f, g) : ξ → η is said
Fig. 5.16 B-morphism of f

XA /Y
vector bundles AA p q ~~
AA ~
AA ~~
~~ ~
B
to be an isomorphism or an equivalence if f : X → Y is a homeomorphism such

that for each b ∈ B, f | p−1 (b) : p −1 (b) → q −1 (g(b)) is a linear isomorphism. In par-
ticular, f : ξ → η is called an equivalence over B, if B = A and g = 1 B . In this case
we write ξ ∼ = η.
Remark 5.7.9 Equivalent bundles have the same dimension.
Proposition 5.7.10 Relation of being equivalent of bundles over B is an equivalence

relation.
Proof It follows from Definition 5.7.8 reflexivity and transitivity of the given relation.
To show symmetry of the relation, let ξ = (X, p, B, F n ) and η = (Y, q, B, F n ) be
two bundles; f : ξ → η be an equivalence. Then f : X → Y is a continuous 1-1
map. We claim that f is open. To prove this it is sufficient to show that f | p−1 (Uα ) is
open, where {Uα } is an open covering of B. In terms of local coordinates, this is given
by (x, v) → (x, A x v), where A x is a nonsingular linear transformation depending
continuously on x. This map has a continuous inverse, because matrix inversion is
continuous. This shows that f | p−1 (Uα ) is a homeomorphism and hence f is open. ❑
Definition 5.7.11 A vector bundle ξ = (X, p, B, F n ) is said to be trivial if it is

isomorphic to the product bundle B × F n → B.
Example 5.7.12 Let ξ = (B × F n , p, B, F n ) be an n-dimensional product bundle

and η = (B × F m , p, B, F m ) be an m-dimensional product bundle. Then the B-
morphisms f : ξ → η are of the form f (b, x) = (b, g(b, x)), where g : B × F n →
F m is a map such that g(b, x) is linear in x. If L(F n , F m ) denotes the vector space
of all linear transformations T : F n → F m , then the vector spaces L(F n , F m ) and
F mn are isomorphic as vector spaces. Thus g : B × F n → F m is continuous iff the
ψ : B → L(F n , F m ), b → g(b, −) is a continuous function.
Theorem 5.7.13 Let ξ = (X, p, B) be a vector bundle over B and 1 B : B → B be

the identity map. Then the vector bundles 1∗B and ξ are B-isomorphic. Moreover, if
given a pair of maps (continuous)
f g
B2 −−−→ B1 −−−→ B,
the induced vector bundles f ∗ (g ∗ (ξ)) and (g ◦ f )∗ (ξ) are B2 -isomorphic.
Proof We now recall that if g ∗ (ξ) = (E 1 , p1 , B1 ), then
E 1 = {(b1 , x) ∈ B1 × X : g(b1 ) = p(x)}, p1 : E 1 → B1 , (b1 , x) → b1 .

Hence if in particular, g = 1 B : B → B, then 1∗B (ξ) and ξ are B-isomorphic. Again

the B2 -vector bundle induced by the map g ◦ f : B2 → B is given by
E((g ◦ f )∗ (ξ)) = {(b2 , x) ∈ B2 × X : (g ◦ f )(b2 ) = p(x)}.
On the otherhand, the total space of f ∗ (g ∗ (ξ)) is

E( f ∗ (g ∗ (ξ))) = {(b2 , y) ∈ B2 × E 1 : f (b2 ) = p1 (y)}
= {(b2 , (b1 , x)) : g(b1 ) = p(x) and f (b2 ) = p1 (y) = b1 }
= {(b2 , ( f (b2 ), x)) : (g ◦ f )(b2 ) = p(x)}.
Since the map ψ : (g ◦ f )∗ (ξ) → f ∗ (g ∗ (ξ)), given by (b2 , x) → (b2 , ( f (b2 ), x))
is a bundle isomorphism from (g ◦ f )∗ (ξ) to f ∗ (g ∗ (ξ)), it follows that the induced
vector bundles f ∗ (g ∗ (ξ)) and (g ◦ f )∗ (ξ) are B2 -isomorphic. ❑
Corollary 5.7.14 There is a contravariant functor from the category of topological

spaces T op to the category of sets Set which assigns to a topological space B (which
is an object in T op), the vector bundle Vectn (B) equals to the set of isomorphism
classes of n-dimensional vector bundles over B and to any continuous map f : B1 →
B (which is a morphism in category T op), there exists a function m f ∗ : Vectn (B) →
Vectn (B1 ) defined by {ξ} → { f ∗ (ξ)}, where {ξ} denotes the isomorphism class of a
vector bundle ξ over B.
Definition 5.7.15 A Gauss map for an F-vector bundle ξ = (X, p, B) is a contin-

uous map G : X → F m for some m, 1 ≤ m ≤ ∞ such that for each b ∈ B, G| p−1 (b)
is a linear monomorphism.
Proposition 5.7.16 Let ξ = (X, p, B, F n ) be an F-vector bundle of dimension n.

If B is paracompact, then there exists a Gauss map for ξ.
Proof Let U j1 , U j2 , . . . , U jm be a finite collection of coordinate neighborhoods that

cover B. Then we can choose an associated partition of unity f ji .
Define
F ji : X → F n
by the rule
f ji ( p(x)) · qφ−1
ji (y), if p(x) ∈ U ji
F ji (x) =
0, if p(x) ∈
/ U ji ,
where q : U ji × F n → F n is the projection and φ ji are associated coordinate trans-

formations. This shows that F ji is a linear monomorphism on p −1 (b) if f ji (b) = 0.
A Gauss map G : X → F mn is then given by x → (F ji (x), . . . , F jm (x)). ❑
Proposition 5.7.17 Let ξ be a vector bundle over a paracompact space such that
ξ|Ui , i ∈ I (indexing set) is trivial, where {Ui }, i ∈ I , is an open covering of B.
Then there is a countable open covering {V j }, 1 ≤ j, of B such that ξ|V j is trivial.

Moreover, if each b ∈ B is an element of almost n sets Ui , there exists a finite open
covering {V j }, 1 ≤ j ≤ n, of B such that ξ|V j is trivial.
Proof Since B is paracompact, there exists a partition of unity. Let {λi }, i ∈ I be

such that wi = λi−1 (0, 1] ⊂ Ui . Then for each b ∈ B, let T (b) be the finite set of
i ∈ I such that λi (b) > 0. For each finite subset T ⊂ I , let V (T ) be the open subset
of all b ∈ B such that λi (b) > λ j (b) for each i ∈ T and j ∈ / T . If T and T are
two distinct subsets of I each with m elements, then V (T ) ∩ V (T ) = ∅. Let Vm be
the union of all V (T (b)) such that T (b) has m elements. Since i ∈ T (b) gives the
relation V (T (b)) ⊂ Wi , then bundle ξ|V (T (b)) is trivial. Since Vm is a disjoint union,
ξ|Vm is trivial. Moreover, under the last hypothesis, V j = ∅ for j > n. ❑
Theorem 5.7.18 For each n-dimensional vector space ξ = (X, p, B, F n ) over a

paracompact space B, there is a Gauss map g : E(ξ) → F ∞ , where E(ξ) = X . If
B has an open covering of sets {Ui }, 1 ≤ i ≤ m, such that ξ|Ui is trivial, then ξ has
a Gauss map G : E(ξ) → F nm .
Proof By Proposition 5.7.17, we may assume that {Ui } is the countable or finite open
covering of B such that ξ|Ui is trivial. Let ψi : Ui × F n → ξ|Ui be Ui -isomorphisms,
and let {λi } be a partition of unity with closure of λi−1 ((0, 1]) ⊂ Ui . Define a map

g : E(ξ) → F n , g → gi ,
i
where gi | E(ξ|Ui ) is (λi p)( p2 ψi−1 ) and p2 : U × F n → F n is the projection to the sec-
ond factor. The map gi is zero outside E(ξ|Ui ). Since gi : E(ξ) → F n is a monomor-
phism on the fibers of E(ξ) over b with λi (b) > 0 and the images of gi are in
complementary subspaces, the map g is a Gauss map. In general, F n is F ∞ , but
i
if there are only m sets Ui , then F n is F mn . ❑
Corollary 5.7.19 Every n-dimensional vector bundle ξ = (X, p, B, F n ) over a

paracompact B is B-isomorphic to f ∗ (γn∞ ) for some f : B → G n (F ∞ ).
Remark 5.7.20 Theorem 5.7.18 and Corollary 5.7.19 form a homotopy classification
theorem for vector bundles. For other homotopy classification theorems see Sect. 5.9
We now use the following notations for Bott Periodicity Theorem:
Notations: If F = R, G n (F ∞ ) is written B O;
If F = C, G n (F ∞ ) is written BU ;
If F = H, G n (F ∞ ) is written B S p
Theorem 5.7.21 (Bott Periodicity Theorem)

(i) 2 BU BU × Z
(ii) 4 B S p B O × Z
(iii) 4 B O ∼= B Sp × Z
Proof For proof see (Bott 1957, 1959) and Remark 5.7.22 ❑
Remark 5.7.22 All of the basic three methods of proving ‘Bott Periodicity Theorem’
are complicated. The first method prescribed by R. Bott in 1957 and 1959 appeared
in his papers [The stable homotopy groups of classical groups, Proc. Nat. Acad. Sci.,
USA, 43(1957), 933-935 and Ann. of Maths, 70(1959), 313-337] by using Morse
Theory to analyze X for a Lie group X . The other references in this respect are
Milnor (1963), Toda (1959), Dyer and Lashof (1961) and for the general case Atiyah
and Bott (1964).
Remark 5.7.23 An important relation between Bott Periodicity Theorem and stable
homotopy groups of spheres πnS came through the so-called stable J-homomorphism
from the (unstable) homotopy groups of the (stable) classical groups to these stable
homotopy groups πnS . Following the original description given by G. W. Whitehead,
it became the subject of the famous Adams conjecture posed by J.F. Adams in 1963.
This conjecture is true which was proved by Daniel Quillen in 1971.
5.8 Charts and Transition Functions of Bundles
This section studies charts and transition functions of bundles and establishes a one-
one correspondence between the set of equivalence classes of principal G-bundles ξ
over a space B and the set of equivalence classes of transition functions associated
with an atlas of ξ. Let G be a topological group with identity e and F be a G-space.
We assume in this section that all principal bundles are G-bundles and all fibers have
fiber F(or F n ). Given a space B, let (B × F, p, B) denote the product fiber bundle.
Definition 5.8.1 Let ξ = (X, p, B, G) be a principal G-bundle. A chart (U, φ) of ξ is
a pair consisting of an open set U ⊂ B and a homeomorphism φ : U × G → p −1 (U )
such that p ◦ φ = pU and φ(b, g) = φ(b, e) · g, ∀ b ∈ U and ∀ g ∈ G. An atlas is a
family of charts {(U j , φ j ) : j ∈ J } such that {U j : j ∈ J } is an open covering of B
and each homeomorphism φ j : U j × G → p −1 (U j ) is such that p ◦ φ j = pU j and
φ j (b, g) = φ j (b, e) · g, ∀ b ∈ U j and ∀ g ∈ G. An atlas is said to be complete if it
includes all the charts of ξ.
Example 5.8.2 Every principal G-bundle has at least one atlas.
Definition 5.8.3 Given a space B and a topological group G, a set of transition
functions ξ˜ for B and G consists of an open covering {U j : j ∈ J } of B and a family of
maps g jk : U j ∩ Uk → G such that U j ∩ Uk = ∅ and g jl (b) = g jk (b)gkl (b), ∀ b ∈
U j ∩ Uk ∩ Ul (= ∅). Each function g jk is called a transition function defined on
U j ∩ Uk .
For j = k = l, g j j (b) = g j j (b)g j j (b) ∈ G for all b ∈ U j implies that g j j (b) = e

for all b ∈ B. Again for j = l, e = g j j (b) = g jk (b)gk j (b) ∈ G for all b ∈ U j ∩ Uk
implies that g jk (b) = gk−1
j (b) for all b ∈ U j ∩ Uk .
Definition 5.8.4 Let ξ = (X, p, B, G) and ξ = (X , p , B , G) be two principal G-

bundles and ξ˜ = {(U j , g jk ) : j, k ∈ J } and ξ˜ = {(Ua , gab

) : a, b ∈ A} be the sets of
transition functions of ξ and ξ , respectively. A morphism f : ξ˜ → ξ˜ between two

sets of transition functions is a pair consisting of a map f˜ : B → B and a collection

of maps f a j : U j ∩ f˜−1 (Ua ) → G such that

f a j (b) g jk (b) = gab ( f˜(b)) f bk for all b ∈ U j ∩ Uk ∩ f˜−1 (Ua ) ∩ f˜−1 (Ub ).
We now consider the particular case, when f˜ : B → B is 1 B .
Definition 5.8.5 Two sets of transition functions {(U j , g jk ) : j, k ∈ J } and {(Ua ,

gab ) : a, b ∈ A} for the same space B, are said to be equivalent if there exist maps
f a j : U j ∩ Ua → G, such that
−1
gab (b) = f a j (b)g jk (b) f bk (b) for all b ∈ Ui ∩ U j ∩ Ua ∩ Ub , i, j ∈ J and a, b ∈ A.
We now recall the following definition.
Definition 5.8.6 Let ξ = (X, p, B, F n ) and ξ = (X , p , B , F n ) be two

n-dimensional F- vector bundles. A morphism φ : ξ → ξ is a pair of maps φ : X →
X and φ̃ : B → B such that p ◦ φ = φ̃ ◦ p and φ| p−1 (b) : p−1 (b) → p −1 (φ̃(b))
is a linear map for each b ∈ B. The identity maps 1 X : X → X and 1 B : B → B
define an identity morphism 1d : ξ → ξ.
We now establish a relation between the concepts of a vector bundle and a principal
G-bundle. To do this we first show that given an atlas of a principal G-bundle, there
exists a unique set of transition functions.
Proposition 5.8.7 (a) Corresponding to a principal G-bundle over B, there exists

a unique set of transition functions ξ˜ = {(U j , g jk ) : j, k ∈ J } such that the map
φk : Uk × G → p −1 (Uk ) satisfies the relation
φk (b, g) = φ j (b, g jk (b)g), ∀ b ∈ U j ∩ Uk , g ∈ G.
(b) Let ξ = (X, p, B, G) and ξ = (X , p , B , G) be two G-bundles and ξ˜ and ξ˜

be two sets of transition functions as in (a). If φ : ξ → ξ is a bundle morphism,
then φ induces a unique morphism of sets of transition functions f : ξ˜ → ξ˜ such
that
f˜ = φ̃ : B → B and φ ◦ φ j (b, g)
= φa (φ̃(b), f a j (b)g), ∀ b ∈ U j ∩ φ̃−1 (Ua ), g ∈ G.
5.8 Charts and Transition Functions of Bundles 231
Proof (a) Let j, k ∈ J . Then the map
ψ jk = φ j −1 ◦ (φk |(U j ∩ Uk ) × G) : (U j × Uk ) × G → (U j × Uk ) × G
is such that pU j ∩Uk ◦ ψ jk = pU j ∩Uk . Now we can write ψ jk in the form ψ jk (b,
f jk (b, g)) for some f jk : (Ui ∩ Uk ) × G → G. Hence
φk (b, g) = φ j (b, f jk (b, g)), ∀ b ∈ U j ∩ Uk , g ∈ G.
Consequently,
φ j (b, f jk (b, g)) = φk (b, g) = φk (b, e) · g = φ j (b, f jk (b, e)) · g

= φ j (b, f jk (b, e) · g), ∀ b ∈ U j ∩ Uk
and g ∈ G. This shows that f jk (b, g) = f jk (b, e) · g. Hence if we take g jk (b) =

f jk (b, e), ∀ b ∈ U j ∩ Uk , j, k ∈ J , then the requisite condition for φk is satis-
fied. Next let, b ∈ U j ∩ Uk ∩ Ua . Then
φ j (b, g j (b)g) = φa (b, g) = φk (b, gka (b)g) = φ j (b, g jk (b)gka (b)g).
This shows that g ja (b) = g jk (b)gka (b). Hence ξ˜ = {(U j , g jk ) : j, k ∈ J } is a set

of transition functions.
(b) Let j, k ∈ J and a, b ∈ A. Then the maps
−1
φa j = φa ◦ φ(φ j |(U j ∩ φ̃−1 (Ua ) × G) : U j ∩ φ̃−1 (U j ∩ φ̃−1 (Ua )
× G → Ua × G)
are such that pUa ◦ θa j = φ̃ ◦ pU j ∩ φ̃−1 (Ua ), where θa j is of the form θa j (b, g) =
(φ̃(b), h a j (b, g)) for some h a j : (U j ∩ φ̃−1 (Ua )) × G → G. Hence
φ ◦ φ j (b, g) = φa (φ̃(b), h a j (b, g)), ∀ b ∈ U j ∩ φ̃−1 (Ua )
and ∀ g ∈ G. Consequently,
φa (φ̃(b), h a j (b, g)) = (φ ◦ φ j (b, e)) · g = φa (φ̃(b), h a j (b, e) · g)

= φa (φ̃(b), h a j (b, e)g).
Hence it follows that h a j (b, g) = h a j (b, e)g. If we take f a j (b) = h a j (b, e), ∀ b ∈
U j ∩ φ̃−1 (Ua ), j∈ J and a ∈ A, then the requisite condition for (b) is satisfied.
Next suppose that ξ˜ = {(U j , g jk ) : j, k ∈ J } and ξ˜ = {(Ua : gab
) : a, b ∈ A}.
φa (φ̃(b), f a j (b)g jk (b)g) = φ ◦ φ j (b, g jk (b)g) = φ ◦ φk (b, g)

= φb (φ̃(b), f bk (b)g) = φa (φ̃(b), gak

(φ̃(b)) f bk (b)g).

Consequently, it follows that f a j (b)g jk (b) = gal (φ̃(b)) flk (b). Hence { f a j } is a
morphism of sets of transition functions. ❑
Theorem 5.8.8 Given a topological group G, there exists a (1-1)- correspondence
between 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.
Proof Let ξ and ξ be two principal G-bundles over B and {ξ} ˜ be the set of equiv-
alence classes of the sets of transition functions associated with an atlas of ξ given
in Proposition 5.8.7(a). If φ : ξ → ξ is an equivalence of G-bundles, then by Propo-
sition 5.8.7(b), there exists a morphism f (φ) : ξ˜ → ξ˜ . Since here φ̃ : B → B is
1 B , this correspondence is well defined. Next suppose that ξ˜ is any set of transi-
tion functions. Then by Ex. 5 of Sect. 5.13, we have a principal G-bundle ξ and
an atlas {(U j : φ j ) : j ∈ J } of ξ such that ξ˜ is the corresponding set of transition
functions. Hence f (φ) is surjective. To show that f (φ) is injective, let ξ and ξ be
two G-bundles such that the corresponding sets ξ˜ and ξ˜ of transition functions are
equivalent, by an equivalence f : ξ˜ → ξ˜ . Then by Ex. 5 of Sect. 5.13, there is a mor-
phism φ : ξ → ξ inducing f . In particular, φ̃ : B → B is 1 B . Moreover, we have
the morphism f −1 : ξ˜ → ξ˜ defined by
f −1 = { f a−1
j : a ∈ A, j ∈ J }.
Then the corresponding morphism ψ : ξ → ξ of G-bundles is the inverse of φ.

Because,
(ψ ◦ φ ◦ φ j )(b, g) = (ψ ◦ φ)(φ j (b, g)) = ψ ◦ φa (b, fa j (b)g)

= φ j (b, f a−1
j (b) f a j (b)g) = φ j (b, g),
∀ b ∈ U j ∩ Ua and ∀ g ∈ G, j ∈ J, a ∈ A. This shows that ψ ◦ φ = 1d . Similarly,

φ ◦ ψ = 1d . Hence ξ and ξ are equivalent principal G-bundles. Consequently the
correspondence f (φ) is injective. Hence f is a bijection. ❑
Let GL (n, F) = G denote the group of all nonsingular n × n matrices over F.
This is a topological group. We show that there exists a one-one correspondence
between the equivalence classes of n-dimensional F-vector bundles over B and the
equivalence classes of sets of transition functions for B and G = GL (n, F) and also
a one-one correspondence with principal GL (n, F)-bundles.
We now define chart and atlas of a vector bundle in a way analogous to Defini-
tion 5.8.1.
Definition 5.8.9 Let ξ = (X, p, B, F n ) be an n-dimensional F-vector bundle over
B. A chart (U, φ) of ξ is a pair consisting of an open set U ⊂ B and a homeomorphism
5.8 Charts and Transition Functions of Bundles 233
φ : U × F n → p−1 (U ) such that p ◦ φ = pU and φ is linear on all fibers p −1 , b ∈ B.

An atlas is a family {(U j , φ j ) : j ∈ J } of charts such that {U j : j ∈ J } is an open
covering of B.
ξ has at least one atlas.
Construction 5.8.10 Given an atlas of a vector bundle, a method of construction

of a set of transition functions is prescribed:
Let ξ = (X, p, B, F n ) be an n-dimensional F-vector bundle over B and {(U j , φ j ) :
j ∈ J } be a given atlas of ξ over B. We now construct a set of transition functions
{(U j , g jk ) : j, k ∈ J } for B and the group GL (n, F) as follows:
for j, k ∈ J , the maps
ψ jk = φ j −1 ◦ (φk |(U j ∩ Uk ) × F n ) : (U j ∩ Uk ) × F n → (U j ∩ Uk ) × F n
are of the form ψ jk (b, u) = (b, f jk (b, u)) for some f jk : (U j ∩ Uk ) × F n → F n .

Clearly, given a fixed b ∈ U j ∩ Uk , the map f jk (b, −) : F n → F n is a linear iso-
morphism and hence f jk (b, −) ∈ GL (n, F). Taking g jk (b) = f jk (b, −), we have
f jk (b, b) = g jk (b) · u. Hence
φk (b, u) = φ j (b, g jk (b) · u), ∀ b ∈ U j ∩ Uk , u ∈ F n .
Following the above construction, a basic and important theorem is obtained.
Theorem 5.8.11 There exists a one-one correspondence between equivalence

classes of n-dimensional F-vector bundles over a space B and the equivalence classes
of the set of transition functions for B and the group GL (n, F).
5.9 Homotopy Classification of Vector Bundles
This section presents two main theorems on the homotopy classification of vector
bundles. The problems of homotopy classification of vector bundles are very inter-
esting in algebraic topology. This section studies the homotopy classification (see
Theorem 5.9.5 and Corollary 5.9.7) of vector bundles which leads to define K -theory.
The reader is referred to the book Husemöller (1966).
Theorem 5.9.1 Let B be a paracompact space and f, g : B → A be two homotopic

maps. If ξ is a vector bundle over the space A, then the induced bundles f ∗ (ξ) and
g ∗ (ξ) over B are B-isomorphic.
Proof As f g : B → A, there exists a map H : B × I → A such that
H (x, 0) = f (x), H (x, 1) = g(x), ∀ x ∈ B.

Hence f ∗ (ξ) and H ∗ (ξ) are both vector bundles(see Ex. 4 of Sect. 5.13) such that
f ∗ (ξ) and H ∗ (ξ)|(B × {0}) are B-isomorphic. Similarly, g ∗ (ξ) and H ∗ (ξ)|(B × {1})
are B-isomorphic. Since there exists an isomorphism
(α, β) : H ∗ (ξ) : (B × {0}) → H ∗ (ξ)|(B × {1})(see Ex. 4 of Sect. 5.13),
it follows that f ∗ (ξ) and g ∗ (ξ) are B-isomorphic. ❑

Corollary 5.9.2 Every vector bundle over a contractible paracompact space
is trivial.
Proof Let B be a contractible paracompact space. As B is contractible, the identity
map 1 B : B → B and the constant map f : B → B are homotopic. Let ξ be an n-
dimensional vector bundle over B. Then 1∗B (ξ) and ξ are B-isomorphic and f ∗ (ξ)
is B-isomorphic to the product bundle (B × F n , p, B). Since 1 B f, it follows by
Theorem 5.9.1 that ξ is B-isomorphic to the product bundle (B × F n , p, B) which
is a trivial bundle. ❑
We now prove a classification theorem of vector bundles by proving that there
is a (1-1) 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 ∞ ). Let H denote the category of paracompact spaces and
their homotopy classes, and Set denote the category of sets and their functions. Let
Vectn (B) denote the set of B-isomorphic classes of n-dimensional vector bundles over
B. Given an n-dimensional vector bundle ξ over B, let {ξ} denote the B-isomorphism
class in Vectn (B) of ξ and [ f ] denote the homotopy class of f : A → B between
two paracompact spaces A and B. We now study Vect n (−).
Theorem 5.9.3 Vectn : H → Set is a contravariant functor.
Proof The object function is defined by assigning to each object B in H, the set
Vectn (B) of B-isomorphic classes of n-dimensional vector bundles over B which is
an object in the category Set . Define the morphism function
Vectn ([ f ]) : Vectn (B) → Vectn (A)
given by
Vectn ([ f ])({ξ}) = { f ∗ (ξ)}
for every morphism [ f ] in H, where f : A → B is a continuous map between para-

compact spaces. This function is well defined by Theorem 5.9.1. For the identity
map 1 B : B → B, 1∗B (ξ) and ξ are B-isomorphic. This shows that Vectn ([1d ]) is the
identity function. Finally, let [g] denote the homotopy class of g : C → A between
the paracompact spaces C and A. Then f ◦ g : C → B. Hence f induces a vector
bundle over A and g, f ◦ g induce vector bundles over C such that g ∗ ( f ∗ (ξ)) and
( f ◦ g)∗ (ξ) are C-isomorphic. This shows that Vect n ([ f ][g]) = Vectn [g]Vect n [ f ].
Consequently, Vectn is a contravariant functor from H to Set . ❑
5.9 Homotopy Classification of Vector Bundles 235
Let F = R, C or H. Then the natural inclusion G n (F ) ⊂ G n (F

m m+1
) defines a
space G n (F ∞ ) = G n (F ) with induced topology, called Grassmann manifold
m
n≤m
G n (F ∞ ). Let γn∞ denote the n-dimensional vector bundle over Grassmann manifold
G n (F ∞ ). Clearly, φn = [−, G n (F ∞ ]) and Vectn are both contravariant functors from
H to Set .
Theorem 5.9.4 The natural transformation ψ : [−, G n (F ∞ )] → Vectn is a natural
equivalence.
Proof For each object B in the category H, define the function ψ(B) : [B, G n
(F ∞ )] → Vectn (B) given by ψ(B)([ f ]) = { f ∗ (γn∞ )}. Clearly, ψ(B) is well defined.
We claim that ψ is a natural transformation of contravariant functors. Let [ f ] be the
homotopy class of the map f : A → B between paracompact spaces A and B. Then
the diagram in Fig. 5.17 is commutative. To show this, let [g] ∈ [B, G n (F ∞ )]. Then
(Vect n ([ f ]) ◦ ψ(B))([g]) = Vect n ([ f ]){g ∗ (γn∞ )} = {( f ∗ g ∗ (γn )).
On the otherhand,
ψ(A)φ([ f ][g]) = ψ(A)([g ◦ f ]) = {(g ◦ f )∗ (γn∞ )}.
For each B in H, ψ(B) is injective by Ex. 9(b) of Exercises 5.13 and surjective
by Corollary 5.7.19. Hence for each B, ψ(B) is a bijection. Consequently, ψ is an
equivalence.
❑
Theorem 5.9.4 gives the following classification of vector bundles:
Theorem 5.9.5 (Homotopy classification of vector bundles) There exists a one-one
correspondence between isomorphism classes of n-dimensional F-vector bundles
on a paracompact space and the homotopy classes of maps from B to Grassmann
manifold G n (F ∞ ).
Proof For each paracompact space B, the function ψ(B) : [B, G n (F ∞ )] →
V ectn (B) defined by ψ(B)([ f ]) = { f ∗ (γn∞ )} is a bijection. Hence the theorem
follows. ❑
Definition 5.9.6 The natural equivalence ψ : [−, G n (F ∞ )] → V ectn is called a rep-
resentation of the contravariant functor Vectn .
Fig. 5.17 Natural ψ(B)

[B, Gn (F ∞ )] / Vectn (B)
equivalence
φ([f ]) Vectn ([f ])
ψ(A)
[A, Gn (F ∞ )] / Vectn (A)
Corollary 5.9.7 (Classification of vector bundles) Every n-dimensional F-vector

bundle over a paracompact space B is isomorphic to the vector bundle induced by
a map from the base space B to the Grassmann manifold G n (F ∞ ).
Proof Let ξ = (E, p, B) be an n-dimensional F-vector bundle over a paracompact

space B. Since ψ(B) is a bijection, there exists a continuous map f : B → G n (F ∞ )
such that f ∗ ({γn∞ }) = {ξ}. This implies that f ∗ (γn∞ ) and ξ are B-isomorphic. ❑
5.10 K -Theory: Introductory Concepts
This section conveys introductory concept of K -theory (topological) which is a

branch of algebraic topology. This branch created around 1960 by Alexander
Grothendieck (1928–2014) in his study of intersection theory on algebraic varieties
is a surprising theory. Topological K-theory is a branch of algebraic topology. It is the
first example of generalized cohomology theories (see Chap. 15). These are groups
in the sense of abstract algebra. They contain detailed information about the origi-
nal object but are very difficult to compute; for example, an important outstanding
problem is to compute the K-theory of the integers. This theory connects algebraic
topology with algebraic geometry, analysis, ring theory, and number theory.
The concept of K -theory arose through the study of vector bundles. The rich struc-
ture of vector bundles establishes that the set of isomorphism classes of n-dimensional
vector bundles over a paracompact space B has a natural bijective correspondence
with the set of homotopy classes of maps from B into a Grassmann manifold of
n-dimensional subspaces in an infinite-dimensional space. This result motivated M.
F. Atiyah (1929 -) and F. E. Peter Hirzebruch (1927–2012) to introduce K -theory’
in 1961 by using the Grothendick construction. The early work on topological K-
theory is due them. Given a compact Hausdorff space X and F = R or C or H, let
K F (X ) be the Grothendieck group of the abelian monoid of isomorphism classes of
finite-dimensional F-vector bundles over X under Whitney sum. Tensor product of
bundles gives K F -theory a commutative ring structure. K (X ) usually denotes com-
plex K -theory; on the other hand, real K -theory is sometimes denoted by K O(X ).
Whitney sum of two vector bundles, which is a generalization of the concept of
direct sum of vector spaces to vector bundles is now defined.
Definition 5.10.1 Let F = R or C or H and ξ and η be F- vector bundles over B.

Then their Whitney sum ξ ⊕ η is the vector bundle over B such that the fibers of
ξ ⊕ η is the direct sum of the fibers in ξ and η.
Construction 5.10.2 (of ξ ⊕ η) Consider the F- vector bundles ξ = (X, p, B) and

η = (X , p , B). Then ξ × η = (X × X , p × p , B × B). Let
d : B → B × B, b → (b, b).
5.10 K -Theory: Introductory Concepts 237
Define
ξ ⊕ η = d ∗ (ξ × η).
Then its total space is
E(ξ ⊕ η) = {(x, x ) ∈ (X × X ) : p((x) = p (x )}.
If ξ1 , ξ2 and ξ3 are vector bundles over B, then
ξ1 ⊕ ξ2 ∼
= ξ2 ⊕ ξ1 ;
ξ1 ⊕ (ξ2 ⊕ ξ3 ) ∼
= (ξ1 ⊕ ξ2 ) ⊕ ξ3 ;
0 ⊕ ξ = ξ;
where 0 is the 0-plane bundle. Moreover, if ξ1 ∼ = ξ2 and ξ3 ∼= ξ4 , then ξ1 ⊕ ξ3 ∼

=
∼
ξ2 ⊕ = ξ4 . Consequently, the set of B-isomorphism classes of vector bundles over B
is an abelian monoid under the operation ⊕.
There is a standard construction of group (ring) completion of an abelian monoid
(semiring), called Grothendieck Construction given by Grothendieck.
Grothendieck Construction
(i) (Grothendieck group) Let M be an abelian monoid. Take the quotient of the free
abelian group generated by the elements of M bu the subgroup generated by
the set of elements of the form x + y − (x ⊕ y), where ⊕ is the sum on M. The
morphism i : M → G(M) of abelian monoids is universal for any homomor-
phism of monoids f : M → G, where G is an abelian group. Then there is a
unique homomorphism of groups f˜ : G(M) → G such that f˜ ◦ i = f .
(ii) (Grothendieck ring) Let M be a semiring. Then its multiplication induces a
multiplication on G(M) such that G(M) is a ring, called the Grothendieck ring
of M. If the semiring M is commutative, then the ring G(M) is also commutative.
We now use the Grothendieck construction to its group completion as follows:
Let G(B) be the free group generated by isomorphism classes of vector bundles over
B. Let [ξ] be the isomorphism class in G(B) corresponding to the vector bundle ξ.
Let S be the subgroup generated by all elements of the form [ξ] ⊕ [η] − [ξ ⊕ η] and
K F (B) be the quotient group.
Definition 5.10.3 The K -theory on B, denoted by K F (B), is the Grothendieck ring

of the semiring Vect F (B). We write [ξ] for the element of K F (B) determined by a
vector bundle ξ. Grothendieck group K F (B) is also called K -theory on the category
of all F-vector bundles over the base space B.
Example 5.10.4 The K -theory of a point are the integers, because vector bundles
over a point are trivial and hence classified by their rank and the Grothendieck group
of the natural numbers are the integers.
Remark 5.10.5 The relation [ξ] ⊕ [η] − [ξ ⊕ η] = 0 holds. This relation shows that
in the group K F (B) the direct sum corresponds to the group operation.
Remark 5.10.6 For F = R or C or H, K F (B) is the group generated by all real vector
bundles, complex vector bundles or quaternionic vector bundles over B, respectively.
Construction 5.10.7 [of ξ ⊗ F η] Let ξ be an F-vector bundle over B of dimension

m and η be an F-vector bundle over B of dimension n. We construct an F-vector
bundle ξ ⊗ F η over B × B of dimension mn with fibers ξb ⊗ F ηb over (b, b ) as
follows:
take the largest atlases
{(Ua , h a ) : a ∈ A}, {(Vc , h c ) : c ∈ C}
for ξ, η with corresponding system

{Ua , gak : a, k ∈ A}, {Vc , gcd : c, d ∈ C}
of transition functions. Then

{Ua × Vc , fnm ◦ (gak × gcd ) : a, b ∈ A, c, d ∈ C}
is a system of transition functions for GL (nm, F) on B × B , where for the standard

basis e1 , e2 · · · , en of F n ,
f nm : GL (n, F) × GL (m, F) → GL (nm, F)
is the tensor product homomorphism obtained by identifying F n ⊗ F m with F nm by

the unique isomorphism which sends ei × e j to e(i−1)m+ j, for 1 ≤ i ≤ n, 1 ≤ j ≤ m
with f nm (M, N ) to be the usual automorphism
M ⊗ N : Fn ⊗ Fm → Fn ⊗ Fm.
This gives an nm-dimensional vector bundle over B × B , denoted by ξ ⊗ F η.
We now describe the isomorphism class of ξ ⊗ F η as follows:

If we write ξ = ξ (F n ), η = η (F m ), where ξ is a principal GL (n, F)-bundle and η
is a principal GL (m, F)-bundle, then ( f nm )∗ (ξ × η ) is a GL (nm, F)-bundle with
ξ ⊗F η ∼
= (( f nm )∗ (ξ × η ))[F n ⊗ F F m ] (5.2)
The relation (5.2) gives the following properties of ⊗ F .
Proposition 5.10.8 (i) ⊗ F is a functor of two variables.

(ii) (ξ ⊗ F η) ⊗ F ζ ∼
= ξ ⊗ F (η ⊗ F ζ).
(iii) If T : B × C → C × B is the switch map, then T ∗ (η ⊗ F ξ) ∼
= ξ ⊗ F η.
5.10 K -Theory: Introductory Concepts 239
(iv) ξ ⊗ F (η × ζ) ∼= (ξ ⊗ F η) × (ξ ⊗ F ζ).
(v) (g, h)∗ (ξ ⊗ F η) ∼
= g ∗ (ξ) ⊗ F h ∗ (η) for maps g : B → B and h : C → C.
Proof Proposition follows from the formula (5.2). ❑
Remark 5.10.9 Given two F-vector bundles ξ, η over the same base space B, we
apply a diagonal map to obtain an internal tensor product ξ ⊗ F η such that (ξ ⊗ F
η)b = ξb ⊗ F ηb for all b ∈ B.
Remark 5.10.10 Tensor products distribute over the Whitney sum, the group K F (B)
admits also the natural ring structures and the exterior power operations define natural
transformations
λi : K F (B) → K F (B) such that

λ0 (b) = 1, λ1 (b) = b

λk (b + s) = λi (b)λ j (s)
i+ j=k
Details of these constructions can be available in Husemöller (1966).
The above discussion can be summarized in the basic and important result:
Theorem 5.10.11 For F = R or C, K F (B) is the Grothendieck ring whose sum is

induced by [ξ] + [η] = [ξ ⊕ η] and whose product is given by [ξ] · [η] = [ξ ⊗ η].
Moreover, given a continuous map f : B → B, there ia a ring homomorphism f ∗ :
K F (B) → K F (B ) such that f ∗ ([ξ]) = [ f ∗ (ξ)].
Proposition 5.10.12 Let B be the paracompact space. If f g : B → B, then

f ∗ (ξ) = g ∗ (ξ) : K F (B) → K F (B ).
Proof Let ξ = (X, p, B) be a vector bundle, B be a paracompact space and f g :

B → B. Then the induced bundle f ∗ (ξ) and g ∗ (ξ) over B are B-isomorphic (see
Theorem 5.9.1). Hence it follows that f ∗ (ξ) = g ∗ (ξ) : K F (B) → K F (B ). ❑
Corollary 5.10.13 The correspondence B → K F (B) is a functor from the homoto-

topy category of paracompact spaces to the category of rings.
Proof The proof follows from Theorem 5.10.11 and Proposition 5.10.12. ❑
Definition 5.10.14 The functor K F (B) is called K -theory on the category of all
F-vector bundles over the base space B. The element [ξ] of K F (B) is determined
by a vector bundle ξ.
Remark 5.10.15 The K -theory introduced by Atiyah and Hirzebruch in 1961 is the
first example of generalized cohomology theories (see Chap. 15).
Remark 5.10.16 K-theory over contractible spaces is always Z.

5.11 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 5.11.1 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:
(i) E x = p−1 (x), x ∈ M are the orbits for the G-action.
(ii) (Local trivialization). Every point in M has an open neighborhood U and a diffeo-
morphism ψ : p −1 (U ) → U × G such that the diagram in Fig. 5.18 commutes,
i.e., ψx = ψ|E x maps E x 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 (x, g )g = (x, g g).

E is called the total space, M the base space and E x = p −1 (x) the fiber at x ∈ M.
Sometimes we use the notation E to denote the G-bundle (E, p, M).
Remark 5.11.2 (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 → E x given by g → x · g, is a diffeomor-
phism.
(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 5.11.3 (i) For an n-dimensional real vector bundle (V, p, M) the asso-
ciated 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.
Fig. 5.18 Local ψ

p−1 (U ) /U ×G
trivialization FF xx
FF xx
F x
p FFF xx q
F# x| x
U
5.11 Principal G-Bundles for Lie Groups G 241
Fig. 5.19 Isomorphism of ψ

E@ /F
principal G-bundles @@ ~~
@@ ~
p @@ ~~q
~ ~
B
Definition 5.11.4 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. 5.19 commutes i.e., ψb = ψ|E b maps E b = 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 5.11.5 The map ψb : E b → Fb is also a diffeomorphism for each b ∈ B.
5.12 Applications
For applications of fiber bundles to higher homotopy groups, see Chap. 7. For some
other important applications of fiber bundles in determining the existence or nonex-
istence of cross sections of a particular tensor bundle see Steenrod (1951) and to
theoretical physics see Nakahara (2003) and also Chap. 14 of this present book.
Atiyah and Hirzebruch defined in 1961 the K -theory by using stability class of vec-
tor bundles to study manifolds. Since then K -theory is applied for investigation of
manifolds by constructing powerful new topological invariants. For example, J.F.
Adams solved the vector field problems for spheres S n (if n is odd, the problem
is to determine the maximum number of linearly independent nowhere vanishing
vector fields on S n , see Chap. 14), using K -theory see Adams (1962). For further
applications of K -theory, see Atiyah and Singer (1963).
5.13 Exercises
For this section the books Husemöller (1966), Spanier (1966), and Switzer (1975)
are referred.
1. Let ξ = (X, p, B, F n ) and η = (Y, q, B, F n ) be two n-dimensional vector bun-
dles over the space B and f : ξ → η be a B-morphism. Show that f is a vector
bundle isomorphism iff the map f : p −1 (b) → q −1 (b) is a linear isomorphism
for each b ∈ B.
2. Let ξ = (X, p, B, F n ) be an n-dimensional vector bundle over B and f : B1 →
B be a map. Show that the induced bundle f ∗ (ξ) = (Y, q, B, F n ) is a vector
bundle over B1 such that ( f ξ , f ) : f ∗ (ξ) → ξ is a morphism of vector bundles,
where f ξ : Y → X, (b, x) → x.
3. Let ξ = (X, p, B, F n ) be an n-dimensional vector bundle over B = B1 ∪ B2 ,

where B1 = A × [a, c] and B2 = A × [c, b], a < c < b. If ξ|B1 = (X 1 , p1 , B1 )
and ξ|B2 = (X 2 , p2 , B2 ) are trivial bundles, show that ξ is also trivial.
4. Prove the following:
(a) Let p : X → B be the projection of a fiber bundle, and suppose B is para-
compact. Then p is a fibration.
(b) If ξ is a vector bundle over B × I , then there exists an open covering {U j } j∈J
of B such that ξ|(U j × I ) is trivial.
(c) If f : B × I → B × I, (b, t) → (b, 1) is continuous and ξ = (X, p, B ×
I, F n ) is a vector bundle over B × I , where B is a paracompact space, then
there exists a map g : X → X such that (g, f ) : ξ → ξ is a morphism of
vector bundles and g is a linear isomorphism on each fiber.
(d) The vector bundles ξ over B × I and the vector bundle f ∗ (ξ|(B × {1})) are
isomorphic.
(e) The vector bundles ξ and ξ|((B × 1) × I ) over B × I are isomorphic.
(f) There exists an isomorphism ( f, g) : ξ|(B × {0}) → ξ|(B × {1}) of vector
bundles.
5. (a) Let ξ˜ = {(U j , gi j ) : i, j ∈ J } be a set of transition functions for the space
B and the topological group G. Show that there exists a principal G-bundle
ξ = (X, p, B, G) and an atlas {(U j , φ j ) : j ∈ J } for ξ such that ξ˜ is the set
of transition functions for the atlas.
(b) Let ξ = (X, p, B, G) and ξ = (X , p , B , G) be two principal G-bundles
with atlases {(U j , φ j ) : j ∈ J } and {(Ua , φa ) : a ∈ A} and the correspond-
ing sets of transition functions of ξ and ξ . If f : ξ˜ → ξ˜ is a morphism of
sets of transition functions, show that there is a morphism φ : ξ → ξ of
principal G-bundles inducing f .
(c) Let ξ and ξ be two principal G-bundles over the same space B and ψ =
(φ, φ̃) : ξ → ξ is a morphism of principal G-bundles such that φ̃ : B → B
is 1 B . Show that ψ is an equivalence.
6. A Gauss map of an n-dimensional vector bundle ξ = (X, p, B, F n ) in F m (n ≤
m ≤ ∞) is a map f : X → F m such that f | p−1 (b) : p−1 (b) → F n is a linear
monomorphism. Show that for each n-dimensional vector bundle ξ over a para-
compact space B, there is Gauss map f : X → F ∞ and moreover, if B has an
open covering {U j }, 1 ≤ j ≤ m, such that ξ|U j is trivial, then ξ has a Gauss map
f : X → F mn .
7. Show that
(a) A properly discontinuous group of homeomorphisms is discontinuous and
acts without fixed points.
(b) A finite group of homeomorphisms acting without fixed points on a Haus-
dorff space is properly discontinuous.
5.13 Exercises 243
8. Let ξ = (X, p, B, F n ) and η = (Y, q, B, F n ) be vector bundles and f : X → Y

is a 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.
9. (a) Show that every n-dimensional vector bundle over a paracompact space is
B-isomorphic to f ∗ (γn∞ ) for some f : B → G n (F ∞ );
(b) Let f, g : B → G n (F m ) be two maps such that f ∗ (γn m ) and g ∗ (γn m ) are
B-isomorphic. If
i : G n (F m ) → G n (F 2m )
is the natural inclusion, show that the maps i ◦ f and i ◦ g are homotopic
for 1 ≤ m ≤ ∞.
10. 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.
11. (a) Let f : B × I → B × I be the map defined by f (b, t) = (b, 1) and ξ be
a numberable principal G-bundle over B × I . Show that there exists a G-
morphism (g, f ) : ξ → ξ.
(b) Using notation of (a), show that the principal G-bundles ξ and f ∗ (ξ) are
isomorphic over B × I.
12. Let B be a paracompact space. Show that a fiber bundle p : X → B is a fibration.
13. Let H be a closed subgroup of a Lie group G. Show that every natural subgroup
N ⊂ H determines an H/N -fiber bundle with bundle map
p : G/N → G/H, g N → g H.
14. If ξ = (X, p.B) is a principal G-bundle, show that ξ is a bundle with fiber G.
15. 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. Show that
this principal Z2 space gives a principal Z2 - bundle with the real n-dimensional
projective space R P n as its base space.
16. Let p : X → B be a covering projection. Show that it is a principal G-bundle,
where G is the group of covering transformations with the discrete topology.
17. If p : X → B be a fiber bundle, show that p is an open map.
18. For any G-space X , the automorphisms of the trivial G-bundle p2 : X × B → B
are in (1-1)- correspondence with continuous functions f : B → G.
[Hint: Any bundle automorphism of p comes from a morphism of the underlying
trivial G-bundle p2 : G × B → B. If f : B → G is a map, define
φ f : B × G → B × G, (b, g) → (b, f (b)g).
Since (b, g) = (e, b) · g, the map is completely determined by φ f (e, b). Con-
versely given a G-bundle map φ : B × G → B × G, define f (b) to be the sec-
ond component of φ(e, b).]
19. Given a space B, let η n denote the trivial n-dimensional vector bundle p2 :
F n × B → B, where F = R or C.
(a) Let ξ : X → B be a vector bundle which has a cross section s : B → X

such that s(b) = 0 for all b. Show that ξ has a subbundle isomorphic to the
trivial bundle and this ξ ∼
= η 1 ⊕ ξ for some bundle ξ .
[Hint. Define
ψ : F × B → X, (λ, b) → λs(b),
for (λ, b) ∈ F × B under the right hand product on Fb of the scalar λ.

Since s is nowhere zero, Im ψ is a subbundle of ξ and ψ : η 1 → Im ψ is an
isomorphism. Consequently, ξ = η 1 ⊕ ξ , where ξ = (Im ψ)⊥ .]
(b) Let ξ : X → B be a vector bundle which has an m linearly independent cross
sections si : B → X for i = 1, 2, . . . , m. Show that ξ ∼
= η m ⊕ ξ for some

bundle ξ . In particular, if an n-dimensional vector bundle has n-linearly
independent cross sections, show that it is isomorphic to the trivial bundle
ηn .
20. Show that the tangent bundle T (S n ) to S n is trivial only for n = 1, 3 or 7.
21. Let H be a subgroup of a topological map G. Show that the quotient map
X G mod H → BG is locally trivial with fiber homeomorphic to the space G/H
of cosets.
Let G be a topological group.
(a) Show that
(i) the space X G is contractible.
(ii) If ξ = (X G , p , BG ) is any universal G-bundle, then the space X G is
contractible.
(b) Let φg be an inner automorphism of the topological group G. Show that
Bφg 1 BG .
[Hint: Use the Milnor’s universal bundle ξ : X G → BG ].
22. Show that F0 (V ) ⊂ F(V ) is a submanifold and p̃|F0 (V ) : F0 (V ) → M is dif-
ferentiable.
23. Show that there exists a differentiable right O(n)-action F0 (V ) × O(n) →
F0 (V ) such that for b ∈ M, the orbits are the sets F0 (Vb ) for b ∈ M.
24. Represent S 3 as S 3 = {(z 1 , z 2 ) ∈ C2 : z z¯1 + z 2 z¯2 = 1}, S 2 as the complex pro-
jective line. Define p : S 3 → S 2 , (z 1 , z 2 ) → [z 1 , z 2 ]/(|z 1 |2 + |z 2 |2 )1/2 , called
the Hopf map. Show that (S 3 , p, S 2 , S 1 ) is a fiber bundle with fiber S 1 .
25. Generalize the Hopf map p : S 3 → S 2 to the Hopf map of the form
p : S7 → S4
with the fiber as the unit quaternion S 3 = SU (2, H).

26. Let L be the real line bundle over S 1 , such that L is either the cylinder S 1 × R
or the Möbius strip. Show that the Whitney sum L ⊕ L is a trivial bundle.
27. Give two fiber bundles over S1 .
5.13 Exercises 245
p
28. Let ξ : X −−−→ B be a fiber bundle with fiber F and f, g : X → B be homotopic
maps. Show that f ∗ (ξ) = g∗ (ξ).
[Hint. See Steenrod (1951)].
p
29. Let ξ : X −−−→ B be a fiber bundle. Show that ξ is trivial if B is contractible to
a point.
30. Consider a fiber bundle (X, B, p, F) with total space X , base space B, fiber F
and projection p : X → B. Show that the following statements are equivalent:
(i) p is a weak fibration.
(ii) If the base space B is paracompact, then p : X → B is a fibration.
p
31. Let ξ : X −−−→ B be a fiber bundle with base space B a paracompact space.
Show that if p is a weak fibration, then it is a fibration.
32. Let G n,k = G k (Rn ) be the Grassmann manifold of k-planes through the origin
in Rn . Show that
(i) The orthogonal group O(n, R) acts transitively on G n,k ;
(ii) The isotropy group of the standard Rk ⊂ Rn is O(k, R) × O(n − k, R);
(iii) G n,k ∼
= O(n, R)/O(k, R) × O(n − k, R). See Bredon p 464.
33. Let K (B) be the Grothendieck group of all vector bundles over the base space
B. Show that
(i) each element x ∈ K (B) can be represented as a difference of two vector
bundles:
x = [ξ] − [η].
(ii) two vector bundles ξ and η define the same element in the group K (B) iff
there is a trivial bundle θ such that ξ ⊕ θ = η ⊕ θ.
34. Show that given a vector bundle ξ over a finite C W -complex X , there is a vector
bundle η over X such that ξ ⊕ η = θ for some trivial bundle θ.
35. Show that tensor product of vector bundles induces a ring structures in the addi-
tive Grothendieck group K (B). Moreover, show that
(i) if B = {b0 } is a one point space, then the ring K (B) is isomorphic to the
ring Z;
(ii) if B = {b0 , b1 }, then K (B) ∼
= Z ⊕ Z.

[2] Atiyah, M. F., K -Theory, Benjamin, New York, 1967.
[4] Bredon, G.E., Topology and Geometry,Springer-Verlag, New York, Inc.1993.
[6] Dieudonné, J., A History of Algebraic and Differential Topology, 1900-1960,
[7] Dupont, J., Fiber Bundles and Chern-Weil Theory, Aarhus Universitet, 2003.
References
Adams, J.F.: Algebraic Topology: A student’s Guide. Cambridge University Press, Cambridge
(1972)
Atiyah, M.F.: K -Theory. Benjamin, New York (1967)
Atiyah, M.F., Bott, R.: On the periodicity theorem for complex vector bundles. Acta Math. 112,
229–247 (1964)
Atiyah, M.F., Hizebruch, F.: Vector bundles and homogeneous spaces. Proc. Symp. Pure Maths.
Amer. Math. Soc. 7–38 (1961)
Atiyah, M.F., Singer, I.M.: The index of elliptic operators on compact manifolds. Bull. Amer. Math.
Soc. 69, 422–433 (1963)
Arkowitz, Martin: Introduction to Homotopy Theory. Springer, New York (2011)
Borel, A.: Topology of Lie groups and characteristic classes. Bull. Amer. Math. Soc. 61, 397–432
(1955)
Bott, R.: The stable homotopy groups of classical groups. Proc. Nat. Acad. Sci., USA, 43, 933–935
(1957)
Bott, R.: The stable homotopy of the classical groups. Ann. Math. 70, 313–337 (1959)
Bredon, G.E.: Topology and Geometry. Springer, New York (1993)
Dugundji, J.: Topology. Allyn & Bacon, Newtown, MA (1966)
Dyer, E., Lashof, R.: A topological proof of the Bott. Periodicity Theorems. Ann. Math. Pure Appl.
54(4), 231–254 (1961)
Dieudonné, J.: A History of Algebraic and Differential Topology. pp. 1900–1960. Modern
Birkhäuser (1989)
Dupont, J.: Fiber Bundles and Chern-Weil Theory, Aarhus Universitet (2003)
ton (1952)
Feldbau, J.: Sur la classification des espaces fibrés. C. R. Acad. Sci. Paris 208, 1621–1623 (1939)
Gray, B.: Homotopy Theory, An Introduction to Algebraic Topology. Academic Press, New York
(1975)
Hopf, H.: Über die Abbildungen der dreidimensionalen Sphre auf die Kugelflche. Math. Ann.
104(1), 637–665 (1931)
Hopf, H.: Über die Abbildungen von Sphren auf Sphren niedrigerer Dimension. Fund. Math. 25,
427–440 (1935)
Husemöller, D.: Fibre Bundles, McGraw-Hill, New York (1966)
Luke, G., Mischenko, A.: Vector Bundles and their Applications. Kluwer Academic Publisher,
Boston (1984)
References 247

Milnor, J.: Construction of Universal Bundles I, II. Ann. Math. 63(2), 272–284 (1956), 63(2),
430–436 (1956)
Milnor, J.: ‘Morse Theory’, Notes by M. Spivak and R. Wells. Princeton University Press, New
Jersey (1963)
Nakahara, M.: Geometry, Topology and Physics, Institute of Physics Publishing, Bristol (2003)
Steenrod. N.: The Topology of Fibre Bundles. Prentice University Press, Prentice (1951)
Toda, H.: A topological proof of theorems of Bott and Borel-Hirzeburch for homotopy groups of
unitary groups, Memoirs of the College of Science. Univ. Kyoto. Series A: Math. 32(1), 103–119
(1959)
Chapter 6
Geometry of Simplicial Complexes
and Fundamental Groups of Polyhedra
This chapter conveys the geometry of finite simplicial complexes which provides a
convenient way to study manifolds and builds up interesting topological spaces called
polyhedra from these complexes followed by a study of their homotopy properties
with computing their fundamental groups and develops certain analytical geometric
tools for subsequent chapters. These are: simplex, complex, subcomplex, simplicial
map, triangulation, polyhedron, and simplicial approximation. Simplicial complexes
provide useful tools in computing fundamental groups of simple compact spaces.
The combinatorial device, now called abstract complex was systematically used by
W. Mayer (1887–1947) in 1923.
The geometrical objects such as points, edges, triangles, and tetrahedra are exam-
ples of low dimensional simplexes. Many important spaces are constructed from
certain familiar subsets of Euclidean spaces Rn . One of them is a simplex. A simplex
S is just a generalization to n dimensions of a triangle or a tetrahedron and these are
fitted together to form a geometric complex K , known as simplicial complex in such
a way that two simplexes are either disjoint or they meet in a common edge or face
and every proper face of S is also in K . Simplicial complexes provide a convenient
tool for the study of manifolds. For example, Poincaré duality theorem given by
H. Poincaré (1854–1912) in 1895 is one of the earliest theorems in topology. Simpli-
cial complexes form building blocks for homology theory which begins in Chap. 10.
For example, simplicial homology invented and studied by Henry Poincaré during
1895–1904, is one of the most fundamental influential inventions in mathematics.
He started with a geometric object (a space) which is given by combinatorial data
(a simplicial complex). Then the linear algebra and boundary relations by these data
are used to construct simplicial homology groups.
It is easy to define algebraic invariants such as fundamental groups, higher homo-
topy groups and homology groups, etc., of different classes of topological spaces
but difficult to compute them as the supply of useful tools provided by topological
invariants corresponding to arbitrary spaces are quickly exhausted. To facilitate such
computation, this chapter works with topological spaces that can be broken up into
pieces which fit together in a nice way. Such spaces are called triangulable spaces.

DOI 10.1007/978-81-322-2843-1_6
250 6 Geometry of Simplicial Complexes and Fundamental Groups of Polyhedra
A polyhedron is a topological space which admits a triangulation by a simplicial

complex.
Simplicial approximation is an important concept in algebraic topology. It is
sometimes convenient to utilize a good feature of simplicial complexes: arbitrary
continuous maps between their polyhedra can always be deformed to maps that are
linear on the simplexes of some subdivision of the domain complex. This leads to the
concept of simplicial approximation theorem first proved by J.W. Alexander (1888–
1971) in 1915 and later by O. Veblen (1880–1960) in 1922. Its more revised version
was given by E.C. Zeeman (1925–2016) in 1964. This theorem is used to calculate
fundamental groups, and to examine the topological invariance of the homology
groups of a space.
This chapter utilizes the concept of triangulation to solve extension problems and
that of edge group E(K , v) (which is isomorphic to the fundamental group π1 (|K |, v)
for any simplicial complex K ) is applied to graph theory and proves van Kampen
theorem by using graph-theoretic results. It also proves simplicial approximation
theorem given by Brouwer and Alexander around 1920 by utilizing a certain good
feature of simplicial complexes introduced by J.W. Alexander in (1915). This theorem
plays a key role in the study of homotopy and homology theories.
For this chapter the books and papers (Armstrong 1983), (Eilenberg and Steenrod
1952), (Gray 1975), (Hilton and Wylie 1960), (Maunder 1970), (Singer and Thrope
1967), (Veblen 1922), (Zeeman 1964) and some others are referred in Bibliography.
6.1 Geometry of Finite Simplicial Complexes
This section studies geometry of finite simplicial complexes to facilitate the construc-
tion of simplicial complexes which provide a convenient way to study manifolds.
Such study is important in the study of algebraic topology. The term “simplicial
complex” is derived from the term “simplex”. Such complexes are also called geo-
metrical complexes. Historically, the simplicial techniques were gradually modified
until introduction of singular homology by S. Eilenberg (1915–1998) in a topological
invariant manner (see Chap. 10). The concepts of 1-dimensional and 2-dimensional
simplicial complexes date back at least to L. Euler (1707–1783) and their higher
dimensional analog was first studied by J.B. Listing (1808–1882) in 1862.
We start with the concept of a simplex. Let Rn be the Euclidean n-space. It is an
n-dimensional vector space over R. The standard n-dimensional simplex

n = {(x1 , x2 , . . . , xn+1 ) ∈ Rn+1 : 0 ≤ x i ≤ 1, xi = 1}.
More generally, given p + 1 points v0 , v1 , . . . , v p of Rn , they are said to be affinely

independent if the equations
6.1 Geometry of Finite Simplicial Complexes 251

p

p
ai vi = 0, ai ∈ R and ai = 0 imply ai = 0 for all i.
i=1 i=1
We now give the formal definition of “independent points” in Rn .
Definition 6.1.1 A set S of ( p + 1) of distinct points v0 , v1 , . . . , v p in Rn is said

to be(geometrically) independent if the vectors v1 − v0 , v2 − v0 , . . . , v p − v0 are
linearly independent.
This is equivalent to the statement that the equations

p

p
ai vi = 0, ai ∈ R and ai = 0 imply that a0 = a1 = · · · = an = 0.
i=0 i=0
It shows that the Definition 6.1.1 does not depend on the order of the points
v0 , v1 , . . . , v p . Thus if the vectors in S are independent, then no three of them lie on
a line, no four of them lie on a plane, and no (m + 1) of them lie in a hyperplane of
dimension m − 1 or less.
Example 6.1.2 The points v0 , v1 , v2 in Fig. 6.1 are geometrically independent and
the points v0 , v1 , v2 in Fig. 6.2 are geometrically dependent.
Definition 6.1.3 Let {v0 , v1 , . . . , v p } be a set of geometrically independent points

in some Euclidean space Rn . Then a geometric p-simplex s p is the set of points
p
p
ai vi , ai ∈ R such that ai ≥ 0 for all i and ai = 1.
i=0 i=0
Fig. 6.1 Geometrically

independent points in R2
v0
v2
v1
Fig. 6.2 Geometrically

dependent points in R2
v2
v1
v0

p

p
Clearly, s p = {v = ai vi , ai ≥ 0 and ai = 1} ⊂ Rn , and s p is given the
i=0 i=0
subspace topology and is a compact metric space. s p is denoted by s p = v0 v1 v2
. . . v p . The points v0 , v1 , . . . , v p are called vertices of s p and the set Vert(s p ) =
{v0 , v1 , . . . , v p } is called the vertex set of s p and p is called the dimension of s p .

p
The subspace of s p consisting of all those points ai vi with ai > 0 of s p for all
i=0
i is called the interior of s p , denoted by S̊. The particular point in S̊ defined by
1
Ŝ = (v0 + v1 + · · · + v p )
p+1
is called the barycenter of s p .

If {vi0 , vi1 , . . . , vir } is any subset of the set of vertices {v0 , v1 , . . . , v p } of s p , the
subspace of s p consisting of those points linearly dependent of vi0 , vi1 , . . . , vir is
called an r-face of s p . We write s p ≺ σ if s p is a face of σ. Let s p be a p-dimensional
simplex. A face of s p may be empty or s p itself.
Definition 6.1.4 Let s p be a p-simplex. A face of s p is said to be proper if it is

neither empty nor the whole of s p . The number p is called the dimension of s p .
Example 6.1.5 The faces of a 2-simplex s2 = v0 v1 v2 are the 2-simplex s2 itself,
three 1- simplexes v0 v1 , v1 v2 , and v2 v0 ; and three 0-simplexes v0 , v1 and
v2 .
Definition 6.1.6 Let S be a p-simplex. Define an open p-simplex S ◦ by

sp if p = 0
S◦ =
S − Ṡ if p > 0
where Ṡ is the boundary of S, which is the set of all faces of S other than S itself.
Clearly, a simplicial complex is the disjoint union of its open simplexes.
Remark 6.1.7 An open p-simplex s˚p is the interior of s p . For example, an open
1-simplex s˚1 is an open interval.
To build up a geometric complex from a collection of simplexes in a nice way,

we need the following concept.
Definition 6.1.8 Two simplexes st and s p are said to be properly joined if their
intersection is either empty or is a common face.
A geometric complex or a simplicial complex is a finite family K of simplexes

which are properly joined and have the property that each face of a member of K is
also a member of K . Its formal definition is now given.
6.1 Geometry of Finite Simplicial Complexes 253
Fig. 6.3 Example of a

geometrical simplicial
complex
Fig. 6.4 Example which is

not geometrical simplicial
complex
Definition 6.1.9 A geometric finite simplicial complex K is a finite set of simplexes,

all contained in some Euclidean space Rn such that
(i) if s p is a simplex of K , and sq is a face of s p , then sq is in K ;
(ii) if s p and sq are simplexes of K , then s p ∩ sq is either empty, or is a common
face of s p and sq .
The example in Fig. 6.3 is a geomertical simplicial complex and example in

Fig. 6.4 is not so.
6.2 Triangulations and Polyhedra
This section conveys the concepts of triangulations and polyhedra which are very
important in computing fundamental groups and the homology groups of a certain
class topological spaces. It is easy to define the algebraic invariant such as funda-
mental groups but difficult to compute them in general. So it has become necessary
to compute them for at least a reasonably large class of spaces. This problem can be
solved effectively with topological spaces that can be broken up into pieces which
fit together in a nice way. Such spaces are called triangulable spaces obtained by
triangulations. A polyhedron is a topological space which admits a triangulation by
a simplicial complex.
Definition 6.2.1 Given a simplicial complex K , the set of points of Rn that lie in at
least one of the simplexes of K , topologized as a subspace of Rn , is a topological
space, called the polyhedron of K , denoted by |K |.
Remark 6.2.2 A geometrical simplicial complex K is not a topological space. It

is a set whose elements are geometric simplexes. But |K | is a topological space.
It denotes the point set union of the simplexes of K with the Euclidean subspace
topology and is sometimes called a carrier of K or the polyhedron associated with
K or a realization of K
Definition 6.2.3 Let X be a topological space. If there is a geometric complex K

whose carrier |K | is homeomorphic to X , then X is said to be a triangulable space
or a polyhedron and the complex K is called a triangulation of X . The space |K | is
called a realization of K . More precisely, if h : |K | → X is a homeomorphism, the
ordered pair (K , h) is called a triangulation of X and X is said to be a polyhedron.
Definition 6.2.4 Let K be a finite simplicial complex. The dimension of K denoted

by dim K is defined by dim K = sup {dim S}.
S∈K
Thus dim K is the largest positive integer m such that K has an m-simplex; in
particular, an n-simplex has dimension n.
Theorem 6.2.5 (Invariance of dimensions) Let K and M be two finite simplicial
complexes. If there exists a homeomorphism f : |K | → |M|, then dim K = dim M.
Proof Suppose dim K = n and dim M = m. If possible let n > m. Let S = sn be an
n-simplex in K and S̊ = S − Ṡ be its interior. Then S̊ is an open set in |K |. Again,
since f is a homeomorphism, f ( S̊) is open in |M|. Consequently, there exists some
p-simplex σ p in M (where p ≤ m < n) such that f ( S̊) ∩ σ 0p = W (say), a nonempty
open set in |M|. Take a homeomorphism ψ : n → S such that ψ(˙n ) = Ṡ. Define
U by U = (ψ −1 ◦ f −1 )(W ). Then U is an open subset of (n )◦ . Since p < n,
there exists an embedding j : p → (n )◦ such that image of j contains nonempty
open subset of (n )◦ . As U is open and j (W ) is not open but both U and j (W )
are homeomorphic subset (n )◦ . Hence we reach a contradiction. Thus n > m.
Similarly, m > n. Consequently, m = n ❑
Theorem 6.2.5 defines the dimensions of a polyhedron.
Definition 6.2.6 The dimension of a polyhedron X is defined to be the common
dimension of the simplicial complexes associated with triangulations of X .
Definition 6.2.7 Let X be a polyhedron. Then the dimension of X is the common
dimension of the associated simplicial complexes involved in triangulations of X .
Example 6.2.8 For the standard 2-simplex 2 ⊂ R2 , define K to be the family of
all vertices and 1-simplexes of 2 (i.e., which is the family of all proper faces of
2 ). Then K is a simplicial complex such that |K | is the perimeter of the triangle
2 in R2 . If X = S 1 , given distinct points v0 , v1 , v2 ∈ S 1 , define a homeomorphism
h : |K | → S 1 by h(ei ) = vi , for i = 0, 1, 2 and h taking each 1-simplex [ei , ei+1 ]
onto the arc joining vi to vi+1 . Then (K , h) is a triangulation of S 1 and hence S 1 is
a polyhedron.
Example 6.2.9 Let X be the 2-sphere S 2 defined by S 2 = {(x1 , x2 , x3 ) ∈ R3 :
3
xi2 = 1}. Consider a closed 3-simplex S3 = v0 v1 v2 v3 . Then, the complex K
i=1
whose simplexes are the proper faces of S3 is a geometrical simplicial complex.
Clearly, |S3 | = geometric carrier of S3 is the boundary of a tetrahedron, and hence,
it is homeomorphic to S 2 . This shows that S 2 is triangulable with K as one triangu-
lation.
Example 6.2.10 Let K be a family of all proper faces of an n-simplex S. If there is
a triangulation (K , h) of S n−1 , then this simplicial complex K is denoted by Ṡ(this
notation is borrowed, since |K | is the boundary Ṡ, which is homeomorphic to S n−1 ).
6.2 Triangulations and Polyhedra 255
Example 6.2.11 Every p-simplex s p determines a simplicial complex K , consisting

of the family of all (not necessarily proper) faces of s p .
Definition 6.2.12 Let K be a simplicial complex in Rn . A subcomplex L of K is a

subset of K such that L is itself a simplicial complex (i.e., L satisfies the conditions
of Definition 6.1.9. If L is a subcomplex of the complex K , then |L| is called a
subpolyhedron of |K |.
Definition 6.2.13 The n-skeleton K n of a simplicial complex K is that subcomplex

consisting of m-faces of simplexes of K for m ≤ n. By convention, the empty set is
the (−1)-skeleton.
Example 6.2.14 Let s p be a p-simplex in Rn . The simplex s p alone does not from
a simplicial complex. But s p and all faces of s p taken together form a simplicial
complex denoted by K (s p ) such that |K (s p )| = s p . On the otherhand, the set of all
faces of s p other than s p forms the boundary s˙p of s p ; and s p − ṡ p = s̊ p is called the
interior of s p , sometimes it is denoted by Int(s p ).
Definition 6.2.15 A simplicial pair (K , L) consists of a simplicial complex K and

a subcomplex L of K .
Definition 6.2.16 Let K be a simplicial complex and v is a vertex in its vertex set
Vert (K ). Then, the star of v, denoted by st (v), is the subset
st (v) = {s : s is a simplex of K and v is a vertex of s}.
On the other hand, the open star of v denoted by ost(v) is a subset of |K | defined by

ost(v) = s̊ ⊂ K
s∈K
v∈Vert (K )
Example 6.2.17 The open shaded region in Fig. 6.5 consisting of all the open sim-
plexes of which v is a vertex, is the set st(v).
Fig. 6.5 Star of v
v
Proposition 6.2.18 Let K be a simplicial complex in Rn . Then

(i) |K | is a closed compact subspace of Rn ;
(ii) every point of |K | lies in the interior of exactly one simplex of K ;
(iii) if the subspaces of K are taken separately and their union is endowed with the
identification topology, then |K | is exactly obtained;
(iv) if |K | is a connected space, then it is path-connected.
Proof (i) Since the simplicial complex K is finite and each simplex of K is both
closed and bounded, it follows that |K | is compact.
(ii) Let A and B be two simplexes of K such that their interiors overlap. Since K is
a simplicial complex, A and B must meet in a common face. But the only face
of a simplex which contains interior points is the whole simplex itself. This
shows that A = B.
(iii) Since simplexes of K are closed in R n , they are closed subsets of |K |. Hence
if A is a subset of |K |, and if A ∩ B is closed in A for each simplex B of K ,
then A ∩ B is closed in |K |. Consequently, the finite union ∪{A ∩ B : B ∈ K }
is closed in |K |. Thus, the closed subsets of |K | are precisely those which
intersect each simplex of K is a closed set. Hence |K | has the identification
topology.
(iv) Let |K | be a connected space. Given an arbitrary point x ∈ |K |, if L denotes the
subcomplex of K defined by L = {s ∈ K : x ∈ / s} and is the distance from x to
|L|, then for a positive number η < , the set B(x, η) ∩ |K | is path- connected.
This is because any point in the set can be joined to x by a straight line segment
in some simplex of K . This shows that |K | is locally path-connected. Hence it
follows that |K | is path-connected.
❑
Remark 6.2.19 Given a simplicial complex K , the space |K | is topologized as a
subspace of the Euclidean n-space Rn . An alternative description of the topology in
|K | is given.
Proposition 6.2.20 Let K be a finite simplicial complex. Then a subset A of |K | is
closed iff A ∩ s is closed in s, for each simplex s in K .
Proof Since each s is closed in Rn , s is also closed in
|K |. Hence A ∩ s is closed
in s. Consequently, it is also closed in |K |. Thus A = A ∩ s is closed, since K
s∈K
is a finite set of simplexes. The converse part follows trivially. ❑
Corollary 6.2.21 Let K be a finite simplicial complex. Then the topology of |K | as a
subspace of the Euclidean space Rn coincides with the topology of |K |, considered as
the space obtained from simplexes by identifying together the various intersections.
Definition 6.2.22 Let K be a simplicial complex. For each point x ∈ |K |, the sim-
plicial neighborhood of x, N K (x) is the set of all simplexes of K that contain x,
together with all their faces. The link of x ∈ |K |, L K (x) is the subset of simplexes
of N K (x) that donot contain x.
6.2 Triangulations and Polyhedra 257
Definition 6.2.23 Let K be a simplicial complex. For each simplex s of K , the star
of s, denoted by st K (s), is defined to be the union of the interiors of the simplexes
of K that have s as a face.
Remark 6.2.24 N K (x) and L K (x) are subcomplexes of K ; st K (s) is an open set for
each simplex s of K .
6.3 Simplicial Maps
This section considers simplicial maps which are maps from one simplicial complex
to another simplicial complex preserving in some sense the simplicial structures. The
concept of a simplicial map is an analogous concept of a group homomorphism.
Definition 6.3.1 Let K and L be simplicial complexes. A simplicial map f : K →

L is such that
(i) if v is a vertex of a simplex of K , then f (v) is a vertex of a simplex of L;
(ii) if {v0 , v1 , . . . , vn } spans a simplex of K , then { f (v0 ), f (v1 ), . . . , f (vn )} spans
a simplex of L (repetitions among f (v0 ), f (v1 ), . . . , f (vn ) are allowed);
(iii) if x = λi vi is in a simplex v0 .v1 . . . vn of K , then f (x) = λi f (vi ); (i.e.,
f is a linear on each simplex).
Remark 6.3.2 Given simplicial complexes K and L. Some authors define a simplial
map as a function f : |K | → |L| between their corresponding polyhedra which takes
simplexes of K linearly onto simplexes of L
We now use this definition to prove the following proposition which gives the
continuity of a simplicial map.
Proposition 6.3.3 Given simplicial complexes K and L, a simplicial map f :

|K | → |L| between their corresponding polyhedra is continuous.
Proof Let A be a closed subset of |L|. Then A ∩ s is closed in s for each simplex s
of L. Since the restriction of f to any simplex of K is linear, it is continuous. Hence
f −1 (A) ∩ s is closed in s for each simplex s in K . Consequently, f −1 (A) is closed
in |K | by Proposition 6.2.20. Hence f is continuous. ❑
The definition of simplicial map is now extended for simplicial pairs.
Definition 6.3.4 A simplicial pair (K , L) consists of a simplicial complex K and a

subcomplex L of K .
Definition 6.3.5 A simplicial map f : (K , L) → (A, B) for simplicial pairs is just

a simplicial map f : K → A such that f (L) ⊂ B.
Remark 6.3.6 The composite of two simplicial maps is another simplicial map.
6.4 Barycentric Subdivisions
This section introduces the concept of barycentric subdivisions of a simplicial com-

plex. This concept is very significant in providing useful method of changing the
structure of a simplicial complex K without changing the underlying set |K | or its
topology. The aim of barycentric subdivision is to prescribe a process of repeating
it to make the simplexes of a simplicial complex as small as we please according to
our need.
Definition 6.4.1 Let s p = v0 v1 . . . v p be a p-simplex in Rn . Then the barycenter

of s p , denoted by B(s p ), is the point in the open simplex (s p ) defined by
1
p
B(s p ) = vi .
p + 1 i=0
This is the center of gravity of the vertices in the usual sense. Barycenter comes from
the Greek word ‘barys’ meaning heavy.
Definition 6.4.2 A partial ordering is defined in a simplicial complex by si si+1

if si is a face of si+1 . The notation si ≺ si+1 means si si+1 but si = si+1 .
Definition 6.4.3 Let K be a simplicial complex. A barycentric subdivision of K is

a simplicial complex K such that
(i) the vertices of K are the barycenters of simplexes of K ;
(ii) the simplexes of K are the simplexes B(s0 ) . . . B(sm ), where si ≺ si+1 .
If a barycentric subdivision of a complex exists, it is unique. K is called the

first barycentric subdivision of K , denoted by K (1) . For n > 1, the nth barycentric
subdivision K (n) of K is defined inductively by taking K (n) the first barycentric
subdivision of K (n−1) .
Example 6.4.4 A barycentric subdivision K (2) of a 2-simplex is given in Fig. 6.6.
Example 6.4.5 (i) v0 is a 0-simplex and consists of one point, which is its
barycenter.
Fig. 6.6 Barycentric

subdivision K (2)
6.4 Barycentric Subdivisions 259
(ii) The 1-simplex v0 v1 = {tv0 + (1 − t)v1 : t ∈ I } is the closed line segment
with end points v0 , v1 . The barycenter 12 (v0 + v1 ) is the midpoint of the line
segment.
(iii) The 2-simplex v0 v1 v2 is a triangle (with interior) having barycentric subdi-
vision K (2) as shown in Fig. 6.6. This is the center of gravity of the vertices in
the usual sense.
Lemma 6.4.6 Let L be a subcomplex of a simplicial complex K . If K has a barycen-

tric subdivision K , then L has a barycentric subdivision L and it consists of all
simplexes of K which lie in |L|.
Proof The simplexes of K contained in |L| form a subcomplex of K . If si are sim-

plexes of L and s0 ≺ · · · ≺ sm , then B(s0 )B(s1 ) . . . B(sm ) ⊂ sm ⊂ |L| satisfying
the conditions (i) and (ii) of Definition 6.4.3. ❑
Proposition 6.4.7 Let K be a simplicial complex. If a barycentric subdivision K

of K exists, then |K | = |K |.
Proof The simplexes of K form a subcomplex of K . Let si be simplexes of K

and barycentric division of K of K exist. If s = B(s0 ) . . . B(sn ) is a simplex
of K , then s ⊂ sn ⊂ |K |. This shows that |K | ⊂ |K |. Again let x ∈ s ⊂ |K |. We
order the vertices of s = v0 v1 . . . vn so that if x = ri vi , r0 ≥ r1 ≥ · · · ≥ rn . Let
si =v0 v1 . .. vi . Then x = ti B(si ), where ti = (i + 1)(ri − ri+1 ). Hence ti ≥ 0
and ti = ri = 1. Consequently, x ∈ B(s0 ) . . . B(sn ) ∈ K . Thus |K | ⊆ |K |.
Hence |K | = |K |. ❑
Suppose K and M be simplicial complexes in Rn . Let K ∩ M denote the set of

all simplexes which are in both K and M and K ∪ M denote the set of all simplexes
which are in either K or L. Clearly, K ∩ M is a subcomplex of K and M but in general
K ∪ M is not a simplicial complex. We now give a sufficient condition under which
K ∪ M is a simplicial complex.
Proposition 6.4.8 Let K and M be simplicial complexes in Rn . If |K ∩ M| = |K | ∩

|M|, then K ∪ M is a simplicial complex.
Proof Let s ∈ K and t ∈ M. We claim that s ∩ t is a face of both s and t. Let L and N
be subcomplexes of K and M respectively. Then |L ∩ N | ⊂ |L| ∩ |N | ∩ |K ∩ M|.
Suppose x ∈ s ∈ L, x ∈ t ∈ N , and x ∈ w for some simplex w in K ∩ M. Since t
and w are both simplexes in M, t ∩ w ≺ w. Similarly, s ∩ w ≺ s. Hence s ∩ t ∩ w =
(s ∩ w) ∩ (t ∩ w) ≺ s ∩ w ≺ s. Similarly, s ∩ t ∩ w ≺ t. Consequently, x ∈ s ∩ t ∩
w ∈ L ∩ N and x ∈ |L ∩ N |. Again let L be the simplicial complex consisting of
s and all of its faces; N the simplicial complex consisting of t and all of its faces.
Then
|L ∩ N | = s ∩ t ∩ |K ∩ M| = s ∩ t ∩ |K | ∩ |M|( by hypothesis |K ∩ M| = |K | ∩ |M|)

= (s ∩ |K |) ∩ (t ∩ |M|)
=s∩t
As L ∩ N is a subcomplex of L and N , s ∩ t is a subcomplex of s and t. Again

as s ∩ t is convex, it is a face of s and t. Consequently, K ∪ M is a simplicial
complex. ❑
If A is a subset of Rn , then the diameter of A, denoted by diam (A) is defined by
diam (A) = sup{d(x, y) : x, y ∈ A},
where d(x, y) = ||x − y||.

Proposition 6.4.9 Let s be a simplex of positive dimension. Then diam (s) = ||u −
v|| for some pair of vertices u and v of s, i.e., the diameter of s is the length of its
largest edge.

p
Proof Let s = v0 v1 . . . vn and x, y ∈ s. If y = λi vi , where λi are barycentric
i=0
coordinates of y, then

p
p

p
||x − y|| = || λi x − λi vi || = || λi (x − vi )||
i=0 i=0 i=0

p
≤ λi ||x − vi ||
i=0
≤ max(||x − vi || : 0 ≤ i ≤ p) (6.1)
If we replace y by vi , then we have
||x − vi || ≤ max{||v j − vi || : 0 ≤ j ≤ p} (6.2)
Hence it follows from (6.1) and (6.2) that
||x − y|| ≤ max{||v j − vi || : 0 ≤ i, j ≤ p} (6.3)
Consequently, it follows from (6.3) that diam(s) = ||u − v|| for some pair of over-
prices u and v of s. ❑
Definition 6.4.10 Let K ⊂ Rn be a simplicial complex and dim (s) denote the
dimension of a simplex s of K . The mesh of K , written μ(K ), is the maximum of
the diameters of its simplexes i.e., mesh (K ) = max{dim(s) : s is a simplex of K }.
Proposition 6.4.11 If the dimension of a simplicial complex K ⊂ Rm is n, then
μ(K ) ≤ n+1
n
μ(K ).
Proof For proof it needs only measure the length of the 1-simplexes of K . Let
b0 b1 be a 1-simplex with b0 < b1 . Then b1 is the barycenter ofa k-simplex s =
v0 v1 . . . vk in K . Given vectors u 0 . . . , u n , u and numbers ti with ti = 1, we have
6.4 Barycentric Subdivisions 261

u − ti u i = ti (u − u i ) ≤ ti u − u i .

As b0 ∈ v0 v1 . . . vk , b1 − b0 = b1 − ti vi ≤ ti b1 − vi . Again,
v0 + · · · + vk 1 k
vi − b1 = vi − ≤ vi − v j ≤ μ(s).
k+1 k+1 j k+1
Consequently, b1 − b0 ≤ k+1

k
μ(s). As k ≤ n, it follows that k+1
k
≤ n
n+1
.
Hence, we have b1 − b0 ≤ n+1 μ(K ). This proves that μ(K ) ≤
n n
n+1
μ(K ). ❑
Corollary 6.4.12 lim μ(K (r ) ) = 0.

r →∞
n r

Proof By induction, it follows that μ(K (r ) ) ≤ n+1
μ(K ). The corollary follows,
n r
since lim = 0. ❑
r →∞ n + 1
6.5 Simplicial Approximation
This section introduces the concept of simplicial approximation and proves simplicial
approximation theorem, first given by J.W. Alexander in (1915). This theorem is
utilized in calculating fundamental groups, and in the study of topological invariance
of the homology groups of a topological space.
Given topological spaces X and Y with triangulations h : |K | → X and k : |L| →
Y , any continuous map f ; X → Y induces a continuous map k −1 ◦ f ◦ h : |K | →
|L|. Moreover, any continuous map between polyhedra may be approximated by a
simplicial map in the sense that a continuous map g is considered as an ‘approxima-
tion’ to a continuous map f if f g.
Definition 6.5.1 Let f : |K | → |L| be a map between polyhedra. Given a point
x ∈ |K |, the point f (x) lies in the interior of a unique simplex of L. This unique
simplex is called the carrier of f (x).
We now consider simplicial maps which take simplexes to simplexes and are linear
on the corresponding simplexes to define ‘simplicial approximation’ of a simplicial
map.
Definition 6.5.2 Let K and L be simplicial complexes and f : |K | → |L| be a
continuous map between polyhedra. A simplicial map g : K → L is said to be a
simplicial approximation of f : |K | → |L| if g(x) lies in the carrier of f (x) for
each x ∈ |K | (i.e., for every x ∈ |K | and for every simplex t ∈ L , g(x) ∈ t implies
f (x) ∈ t).
In many problems, such as for computation of the fundamental group of a trian-
gulable space, it needs approximate a given map by a simplicial map. We claim that
a simplicial approximation is homotopic to the original map.
Theorem 6.5.3 Let K and L be simplicial complexes and f : |K | → |L| be a con-

tinuous map. Then
(a) any simplicial approximation g to f is homotopic to f ;
(b) the homotopy in (a) is relative to the subspace A = {x ∈ |K | : f (x) = g(x)} of
|K |.
Proof Let g : |K | → |L| be a simplicial approximation to f : |K | → |L|. Sup-

pose |L| lies in Rn , and F : |K | × I → Rn is the straight line homotopy given by
F(x, t) = (1 − t)g(x) + t f (x). Given x ∈ |K |, there is some simplex of L which
contains g(x) and f (x); and as every simplex is convex, all points (1 − t)g(x) +
t f (x), t ∈ I , must lie in this complex. Consequently, image of F is contained in |L|,
and F is a homotopy from f to g. By construction this homotopy is relative to this
subspace A = {x ∈ |K | : f (x) = g(x)} of |K |. ❑
Corollary 6.5.4 Let (K , L) and (B, C) be simplicial pairs and f : (|K |, |L|) →
(|B|, |C|) be a map of pairs. If g is any simplicial approximation to f : |K | → |B|,
then g(|L|) ⊂ |C|, and f g as maps of pairs.
Proof If x ∈ |L|, then f (x) is in the interior of the unique simplex of C. This simplex
also contains g(x) and hence g(x) ∈ |C|. Moreover, the line segment joining f (x)
and g(x) lies entirely in |C|. Hence the corollary follows. ❑
Remark 6.5.5 Simplicial approximations do not always exist.
Example 6.5.6 Let |K | = |L| = [0, 1] with K having vertices at the points 0, 13 , 1
and L having vertices at 0, 23 , 1 as shown in Fig. 6.7.
We shall show that the map f : |K | → |L|, x → x 2 has no simplicial approxi-
mation. Because if g : |K | → |L| is a simplicial approximation to f : |K | → |L|,
then g must agree with f on inverse image of every vertex of L, giving g(0) = 0
and g(1) = 1. But as g is simplicial, g( 13 ) must be 23 . Hence, g takes the closed seg-
ment [0, 13 ] linearly onto [0, 23 ] and [ 13 , 1] linearly onto [ 23 , 1]. Since carrier of f ( 12 )
is [0, 23 ] and this does not contain g( 12 ), we have a contradiction. Similarly, there
is no simplicial approximation to f : |K | → |L|. But simplicial approximation to
f : |K (2) | → |L| exists for its second barycentric subdivision.
Remark 6.5.7 The above example shows that it may be possible to have a simplicial
approximation by replacement of K by a suitable barrycentric subdivision K (m) of
K . Moreover, simplicial approximations are not unique.
Fig. 6.7 Nonexistence of a 1 1

simplicial approximation
2
f (x) = x2 3
1
3 K L
0 0
6.5 Simplicial Approximation 263
We now characterize simplicial complexes with the help of open stars. We recall
that if v is a vertex of K , the open star of v in K is the union of the interiors of those
simplexes of K which have v as a vertex.
Proposition 6.5.8 The vertices v0 , v1 , . . . , vr of a simplicial complex K span (i.e.,

are the vertices of) a simplex of K iff the intersection of their open stars is nonempty.
Proof Let v0 , v1 , . . . , vr be the vertices of the simplex s of K . Then s̊ = Int(s) ⊂

ost(vi ), i.e., the whole of the interior of s lies in ost(vi ) for each i = 0, 1, . . . , r .

r
r
Hence, ∅ = Int(s) ⊂ ost(vi ). Conversely, let ost(vi ) = ∅ and x ∈ ost(vi ).
i=0 i=0 i=0
Then for each i = 0, 1, 2, . . . , r , there is a simplex si in K such that x ∈ Int(si ) and
vi is a vertex of si . Since the set of all interiors of all simplexes of K constitutes a
partition of |K |, there is a unique simplex of K whose interior contains x, which is
the carrier of x. This shows that s0 = s1 = s2 = · · · = sr . Hence v0 , v1 , . . . , vr are
vertices of the simplex s and hence these vertices span some face of s of K . ❑
The following theorem is the most basic form of “Simplicial approximation the-
orem.”
Theorem 6.5.9 (Simplicial approximation) Let f : |K | → |L| be a continuous map
between polyhedra. If r is chosen sufficiently large, then there is a simplicial approx-
imation g : |K (r ) | → |L| to f : |K (r ) | → |L|.
Proof Case I. We first suppose that for each vertex v of K there exists a vertex u of
L such that
f (ost(v)) ⊂ ost(u) (6.4)
Define a function h : Vert (K ) → Vert (L) by choosing a u for each v and assigning
h(v) = u. Then by using Proposition 6.5.8 and the inclusion (6.4) it follows that
if {v0 , v1 , . . . , vr } spans a simplex of K , then {h(v0 ), . . . , h(vr )} spans a simplex
of L. Now extend h linearly over each simplex of K to obtain a simplicial map
h : |K | → |L|. We claim that h is a simplicial approximation to f . For x ∈ |K |,

r
let v0 , v1 , . . . , vr be the vertices of its carrier. Then x ∈ ost(vi ). Hence by the
0

r
inclusion (6.4), f (x) ∈ ost( f (vi )) shows that carrier of f (x) in L has the simplex
0
spanned by h(v0 ), h(v1 ), . . . h(vr ) as a face. Hence it must contain the point h(x).
Case II. For the general case, we replace K by a suitable barycentric subdivision
K (m) and proceed as follows:
Since the open stars of the vertices of L form an open cover of |L| and
f : |K | → |L| is continuous, the inverse images of these open sets under f form an
open cover of |K |. Again as |K | is a compact metric space, there is a Lebesgue num-
ber η of this open cover by Lebesgue Lemma 1.11.5. Choose m sufficiently large such
that μ(K (m) ) < η/2. Given a vertex v of K (m) , the diameter of its open star in K (m) is
less than η. Hence ost(v) in K (m) and ost(u) in |L| are such that ost(v) ⊂ f −1 (ost(u))
for some vertex u of L. Now proceed as in the first part to prove the general
case. ❑
6.6 Computing Fundamental Groups of Polyhedra
This section presents a method of computing fundamental groups of a special class

of topological spaces. More precisely, this section computes fundamental groups of
polyhedra through a study of loops which are closed paths along 1-simplexes called
edge loops.
Let K be a simplicial complex. We now define a group called edge group of K
based at a vertex of K . An edge path in K is a sequence v0 v1 . . . vk of vertices such
that each consecutive pairs vi , vi+1 spans a simplex of K (the possibility vi = vi+1 is
permissible for technical reason). If v0 = vk = v for the edge path v0 v1 . . . vk , then
the edge path is called an edge loop of K based at v. To define the edge group of K ,
we define a simplicial version of the concept of homotopy. Two edge paths are said
to be equivalent if we can obtain one from the other by a finite number of operations
of types:
(i) if vi−1 = vi , replace . . . vi−1 vi . . . by . . . vi , . . . , or conversely, replace . . . vi . . .
by . . . vi vi . . . ; or
(ii) if {vi−1 , vi , vi+1 } spans a simplex of K (not necessarily 2-simplex), replace
. . . vi−1 vi vi+1 . . . by . . . vi−1 vi+1 . . . , or conversely (geometrically this condition
implies that two sides of a triangle can be replaced by the third side and vice-
versa).
This equivalent relation sets up an equivalence relation between edge paths.
We denote the equivalence class of the edge path v0 v1 v2 . . . vk by [v0 v1 . . . vk ]. The
set of equivalence classes of edge loops of K based at a vertex v of K forms a group
under the binary operations of juxtaposition
[vv1 . . . vk−1 v]·[vu 1 · · · u r −1 v] = [vv1 · · · vk−1 vu 1 · · · u r −1 v] (6.5)
The identity element is the equivalence class [v] and the inverse of [vv1 · · · vk−1 v] is
the class [vvk−1 · · · v1 v].
Definition 6.6.1 The set of equivalence classes of edge loops of K based at a vertex
v of a simplicial complex K forms a group under the binary operations given by
(6.5), called the edge group of K based at v, denoted by E(K , v).
This group is closely related to the fundamental group π1 (|K |, v) as given by the
Theorem 6.6.2. It is convenient to present the edge group as a set of generators and
relations.
6.6 Computing Fundamental Groups of Polyhedra 265
Theorem 6.6.2 The group E(K , v) is isomorphic to the fundamental group

π1 (|K |, v) for any simplicial complex K .
Proof Define a function ψ : E(K , v) → π1 (|K |, v) as follows: consider each edge

loop in K as a loop in |K |. Then given an edge loop vv1 . . . vr −1 v, divide I into r
equal segments and let f : I → |K | be the linear extension of
f (0) = f (1) = v, f (i/r ) = vi , 1 ≤ i ≤ r − 1.
Then f is a loop in |K | based at v. Since equivalent edge paths give homotopic loops,
define
ψ : E(K , v) → π1 (|K |, v), [vv1 . . . vr −1 v] → [ f ].
Use simplicial approximation to show that ψ is onto. Then verify that ψ is an iso-
morphism. ❑
6.7 Applications
This section presents some interesting applications. We apply triangulation to prove

an extension problem. Moreover, we apply the concepts of edge path and edge group
E(K , v) (which is isomorphic to the fundamental group π1 (|K |, v)) for any simplicial
complex K to graph theory and prove van Kampen Theorem with the help of graph-
theoretic results obtained.
6.7.1 Application to Extension Problem
One of the basic aims of algebraic topology is to solve extension problems of

continuous maps. This subsection studies the extendability of the continuous map
f : S m → S n to D n+1 for positive integers m and n with m < n.
Theorem 6.7.1 Let T = (K , h) be a triangulation of a space X such that dimension

of K < n. Then every map f : X → S n is inessential, i.e., f is homotopic to a map
of X into a single point of S n .
Proof Choose a triangulation T = (K , h ) of S n with dim K = n. Using simplicial

approximation theorem there exists a map g : X → S n homotopic to f and which
is related to triangulations T (r ) , T , where T (r ) = (K (r ) , h (r ) ). Since the dimension
of a simplicial complex is not changed under barycentric subdivision, it shows that
dim K (r ) < n. Again since g is simplicial, it follows that g(X ) = S n . If possible,
let g be essential. Then by Proposition 2.10.6 of Chap. 2, g(X ) = S n , which is a
contradiction. This shows that g is inessential and hence f is inessential. ❑
Proposition 6.7.2 If m and n are two positive integers such that m < n. Then every
map f : S m → S n is inessential and admits an extension f˜ : D n+1 → S n .
Proof Since dimension of S m is m < n, it follows from Theorem 6.7.1 that the
map f : S m → S n is inessential. Moreover, since (D n+1 , S n ) forms a normal pair it
follows from Proposition 2.10.7 of Chap. 2 that f admits an extension f˜ : D n+1 →
Sn . ❑
Proposition 6.7.3 If (X, A) is a finite triangulable pair, then A has the homotopy
extension property (HEP) with respect to every space.
Proof Let f : X → Y be a given continuous map and Ht : A → Y be a partial
homotopy of f in the sense that f |A = H0 . Consider, the product space P = X × I
and its closed subspace C = (X × {0} ∪ (A × I ) ❑
Define a map

f (x), if x ∈ X, t = 0
H : C → Y, (x, t) →
Ht (x), if x ∈ A, t ∈ I
As (X, A) is a finite triangulable pair, C is a retract of P. Hence there is a retraction

r : P → C. Define a homotopy
G t : X → Y, x → H (r (x, t)).
Then G t is an extension of Ht such that G 0 = f .
6.7.2 Application to Graph Theory
This subsection presents some graph-theoretic results which are used as tools to
prove van Kampen Theorem. A graph is a one-dimensional simplicial complex.
Definition 6.7.4 A one-dimensional subcomplex of a complex K whose polyhedron
is both path-connected and simply connected is called a tree. A tree T is said to be
maximal if T is a tree such that T contains T , then T = T .
Theorem 6.7.5 A maximal tree T in a complex K contains all the vertices of K .
Proof If possible, T does not contain all the vertices of K . Then there exists some
vertex v which is in K − T . Choose a vertex u of T . Since |K | is path-connected,
there is a path joining u and v in |K |. By simplicial approximation Theorem 6.5.9,
this path is replaced by an edge path uv1 v2 . . . vk v. If vi is the last vertex of this
edge path which lies in T , a new subcomplex T is formed by adding the vertex
vi+1 and the edge spanned by vi vi+1 to T . The space |T | is the same as |T | with
a ‘spike’ attached. Clearly, |T | is a deformation retract of |T |. Hence T is a tree,
which contradicts the assumption that T is a maximal tree. ❑
6.7.3 van Kampen Theorem
This subsection proves van Kampen theorem given by van Kampen (1908–1942),
which prescribes a method for computing the fundamental groups of topological
spaces that can be decomposed into simpler spaces whose fundamental groups are
already known. It is convenient to present the edge group as a set of generators and
relations. Some authors call van Kampen theorem Seifert–van Kampen theorem.
This theorem is proved by using the algebraic concept of free product of two groups
(see Chap. 1).
Let L be a subcomplex of K such that |L| is simply connected. Then edge
loops in L has no contribution to E(K , v). This shows that the simplexes of L
may be effectively ignored in calculation of E(K , v). If we list the vertices of K
as v = v0 , v1 , v2 , . . . , vt and denote G(K , L) for the group which is determined by
generators gi j , one for each ordered pair of vertices vi , v j that span a simplex of K
with the relations gi j = 1 if vi , v j span a simplex of L, and gi j g jk = gik if vi , v j , vk
span a simplex of K . If i = j, then gii = 1, and for i = k, g ji = gi−1 j . This implies
that we introduce a generator gi j for each pair of vertices vi , v j which span an edge
of K − L and for which i < j. So we consider only the relations gi j g jk = gik , when-
ever vi , v j , vk span a 2-simplex of K − L and i < j < k. In particular, if any two of
these vertices vi , v j span a simplex of L, we take gi j = 1.
Theorem 6.7.6 The group G(K , L) is isomorphic to the group E(K , v) when |L|
is simply connected.
Proof We construct homomorphisms
ψ : G(K , L) → E(K , v) and θ : E(K , v) → G(K , L)
as follows: we join v to each vertex vi of K by an edge path E i in L, by taking

E 0 = v and define ψ on the generators of G(K , L) by ψ(gi j ) = [E i vi v j E −1 j ]. If
−1
vi , v j span a simplex of L, then E i , vi , v j E j is a loop which lies entirely in L.
Hence, it represents the identity element of E(K , v), since |L| is simply connected.
Moreover, if vi , v j , vk span a simplex of K , then
ψ(gi j )ψ(g jk ) = [E i vi v j E −1 −1
j ][E j v j vk E k ]
= [E i vi vk E k−1 ]
= ψ(gik )
This proves that ψ is a homomorphism.

Define θ : E(K , v) → G(K , L), [vvk vl vm . . . vn v] → gok gkl glm gno . Then θ
defines also a homomorphism such that θ ◦ ψ is the identity. Moreover, ψ ◦ θ is
the identity. Hence ψ is an isomorphism with θ as its inverse. ❑
Remark 6.7.7 Let L , K be two simplicial complexes in the same Euclidean space
which intersect in a common subcomplex. If |L|, |K |, |L ∩ K | are all path-connected
spaces and their fundamental groups are known, then we can calculate the funda-
mental group π1 (|L ∪ K |).
Case 1: If L and K intersect in a single vertex, then any edge loop in L ∪ K based
at this vertex is a product of loops, each of which lies in either L or K . We
now obtain the free product π1 (|L|) ∗ π1 (|K |) for the fundamental group
of |J ∪ K |.
Case 2: In the general case, similar arguments hold, except that the free product
π1 (|L|) ∗ π1 (|K |) effectively counts the homotopy classes of these loops
which lie in |L ∩ K | twice (one in each of π1 (|L|), π1 (|K |)). So we need
in this case some extra relations.
The above discussion leads to prove “van Kampen Theorem.”
Theorem 6.7.8 (van Kampen Theorem) Let L , K be two simplicial complexes in the
same Euclidean space such that |L|, |K | and |L ∩ K | are all path-connected spaces.
If i : |L ∩ K | → |L| and j : |L ∩ K | → |K | are inclusion maps and a vertex v of
L ∩ K is taken as a base point of L ∩ K , then the fundamental group of |L ∪ K |
based at v is the free product π1 (|L|, v) ∗ π1 (|K |, v) with the relations i ∗ (x) = j∗ (x)
for all x ∈ π1 (|L ∩ K |, v).
Proof Let TM be a maximal tree in L ∩ K . We extend it to a maximal tree TM1

in L and a maximal tree TM2 in K . Then TM1 ∪ TM2 is a maximal tree in L ∪ K .
Using Theorems 6.7.5 and 6.7.6, the group π1 (|L ∪ K |) is generated by elements gi j
corresponding to edges of L ∪ K − TM1 ∪ TM2 , with relations gi j g jk = gik given by
the triangles of L ∪ K . But this is precisely the group obtained by taking a generator
bi j for each edge of L − TM1 , a generator ci j for each edge of K − TM2 , with relations
of the form bi j b jk = bik , ci j c jk = cik corresponding to the triangles of L , K with
additional relations bi j = ci j , whenever bi j and ci j correspond to the same edge of
L ∩ K . Since the edges of L ∩ K − TM , when considered as edges of L, give a set of
generators for i∗ (π1 (|L ∩ K |)). Similarly the same edges, when considered as edges
of K give a set of generators for j∗ (π1 (|L ∩ M|)). ❑
Remark 6.7.9 For another form of van Kampen Theorem see Theorem 14.7.1 of
Chap. 14.
6.8 Exercises
1. Let X be a polyhedron and be an arbitrary small positive real number. Show

that there is a simplicial complex K such that |K | = X and μ(K ) < .
[Hint: Let K = X (r ) . If n = dim X , then μ(K ) ≤ ( n+1 ) μ(X ). Choose r suffi-
n r
ciently large to obtain μ(K ) < .]

2. Let L ⊂ K be a subcomplex and f : K → M be a map such that f | L is simpli-
cial. If g is a simplicial approximation to f , show that g f rel L .
6.8 Exercises 269
[Hint: Let L ⊂ Rn . Define Ft : K → Rn by Ft (x) = tg(x) + (1 − t) f (x). Since

both g(x) and f (x) lie in some simplex s, Ft is well defined. Hence Ft (K ) ⊂ L.
Consequently, Ft : f g rel L.]
3. (a) Show that every simplicial complex has a barycentric subdivision.
(b) Let K be a simplicial complex and K be its first barycentric subdivision of
K . Show that each simplex of K is contained in a simplex of K .
[Hint: (a) Use double induction: first on the dimension of the complex and second
on the number of simplexes.
(b) Let s be a simplex of K and barycenters B(s0 ), B(s1 ), . . . , B(sk ), are the
vertices of s, where vi ∈ K and v0 < v1 < · · · < vr . Hence all vertices s of s lie
in vr . Hence s is contained in s. si ∈ K and s0 < s1 < · · · < sk . Hence all these
vertices of s lie in sk . This implies that s ⊂ sk .]
4. Let s p be a geometric p-simplex in Rn . Show that
(a) s p is a closed convex connected subspace of Rn , is the closure of its interior;
(b) faces of s p is a closed subspace of sn ;
(c) s p determines its vertices;
(d) any two simplexes are identical iff they have the same set of vertices.
[Hint: Use the result that a point of s p is a vertex iff it is not a point of an
open line segment lying inside s p ].
5. Let s p = v0 v1 . . . v p be a geometric p-simplex in Rn and s p = v0 , v1 , . . . , v p
be a geometric p-simplex in Rm . Show that s p and s p are linearly homeomorphic.
p
p
[Hint: Define f : s p → s p ,
ai vi → ai vi for all points of s p . Then f is
i=0 i=0
a linear homeomorphism].
6. Show that a geometric p-simplex s p is completely characterized by its dimension.
[Hint: Use Ex 5].
7. Let K be a simplicial complex in Rn and A be a subset of |K |. Show that
(a) A is closed in |K | iff A ∩ S is closed in S for every simplex S of K .
(b) if S is a simplex in K of largest dimension, then S̊ = S − Ṡ is an open subset
of |K |.
(c) The topology of |K | as a subspace of Rn , coincides with the topology of |K |,
considered as the space obtained from its simplexes by identifying together
with the various intersections.
8. Let (K , L) be a simplicial pair in Rn . Show that
(i)|K | is a closed and compact subspace of Rn ;
(ii)every point of |K | is in the interior of exactly one simplex of K ;
(iii)|L| is a closed subspace of |K |;
(iv) if C is another subcomplex of K , then L ∪ C and L ∩ C are both subcom-
plexes of K ;
(v) If |K | is a connected space, then it is path-connected.
9. Let K be a simplicial simplex. Prove that for each vertex v of Vert (K ),

(i) st (v) is an open subset of |K |;
(ii) the family of all such stars is an cover of |K |,
(iii) if x ∈ st (v), then the line segment with end points x and v is contained in
st (v).
10. Given a simplex s of K and any point x in the interior of s, show that st K (s) =
|N K (x)| − |L K (x)|.
11. If v0 , v1 , . . . , vm are in the vertex set Vert (K ) of a simplicial complex K . Show

m
that {v0 , v1 , . . . , vn } spans a simplex of K iff st (vi ) = ∅.
i=0
[Hint. See Proposition 6.5.8]
12. Using van Kampen Theorem, show that
(i) π1 (S n ) = 0 for n > 1;
(ii) If X is the union of two circles in R2 with one point in common, then π1 (X )
is the free group on two generators.
(Compare the fundamental group of ‘figure-eight’ described in Chap. 4).
13. Show that a group G is finitely presented iff there exists a polyhedron X such
that G ∼= π1 (X, x 0 ).
14. (Tietze) If X is a connected polyhedron, show that fundamental group π( X, x 0 ) is
finitely presented in the sense that π1 (X, x 0 ) has a presentation with only finitely
many generators and finitely many relations.
[1] Croom, F.H., Basic Concepts of Algebraic Topology, Springer-Verlag, New York,
[3] Lahiri B. K., First Course In Algebraic Topology, Narosa Publishing House,
New Delhi, 2005.
[4] Rotman, J.J., An Introduction to Algebraic Topology,Springer-Verlag, New York,
1988.
[5] Spanier, E., Algebraic Topology, McGraw-Hill Book Company, New York, 1966.
[7] Wallace, A.H., Algebraic Topology, Benjamin, New York, 1980,
References 271
References
Alexander, J.W.: A proof of the invariance of certain constants of analysis situs. Trans. Amer. Math.
Soc. 16, 148–154 (1915)
ton (1952)
Gray, B.: Homotopy Theory: An Introduction to Algebraic Topology. Academic, New York (1975)
Hilton, P.J., Wylie, S.: Homology Theory. Cambridge University Press, Cambridge (1960)
Lahiri, B.K.: First Course In Algebraic Topology. Narosa Publishing House, New Delhi (2005)
Maunder, C.R.F.: Algebraic Topology. Van Nostrand Reinhold, London (1970)
Singer, I.M., Thrope, J.A.: Lecture Notes on Elementary Topology and Geometry. Springer, New
York (1967)
Veblen, O.: Analysis Situs, vol. 2. Part II. American Mathematical Society Colloquium Publications,
New York (1922)
Zeeman, E.C.: Relative simplicial approximation. Proc. Camb. Phil. Soc. 60, 39–43 (1964)
Chapter 7
Higher Homotopy Groups
This chapter continues to study homotopy theory displaying construction of a

sequence of covariant functors πn given by W. Hurewicz (1904–1956) in 1935 from
topology to algebra by extending the concept of fundamental group, which is the first
influential functor of homotopy theory invented by H. Poincaré (1854–1912) in 1895.
It also studies Hopf map and Freudenthal suspension theorem. Prior to Hurewicz, the
idea of higher homotopy groups was originated by E.Čech (1893–1960) in 1932 but
the notation used by Hurewicz has become standard and it is followed. The higher
dimensional homotopy groups provide fundamental tools of classical homotopy the-
ory and are the most powerful basic invariants in algebraic topology. There is an
infinite exact sequence of homotopy groups associated with a fiber space which is
utilized to study Hopf fibering and to compute higher homotopy groups of certain
spaces. Weak fibration has a key role in the study of higher homotopy groups.
The basic problem of n-dimensional homotopy groups is to classify all contin-
uous maps from S n to pointed topological spaces X up to homotopy equivalence.
For the study of pointed topological spaces X of low dimension, the fundamen-
tal group π1 (X ) is very useful. But it needs refined tools for the study of higher
dimensional spaces. For example, fundamental group can not distinguish spheres
S n with n ≥ 2. Such a limitation of low dimension can be removed by consider-
ing the natural higher dimensional analogs of π1 (X ). The complete determination
of higher homotopy groups of spheres is still one of the major unsolved problems
in topology. The classification problem of continuous maps of an n-sphere S n to
a given pointed topological space (X, x0 ) up to homotopy equivalence led to the
discovery of ‘homotopy groups.’ For n = 1, recall that given a pointed topological
space (X, x0 ), π1 (X, x 0 ) defined by [(I, I˙), (X, x 0 )] as the set of homotopy classes
of loops f : (I, I˙) → (X, x0 ) admits a group structure, called the fundamental group
of (X, x 0 ). For each integer n > 1, the definition of the nth (absolute) homotopy
group πn (X, x0 ) is strictly analogous to that of the fundamental group. This means
that πn (X, x 0 ) = [(I n , ∂ I n ), (X, x0 )], where ∂ I n is the boundary of the n-cube I n .
An element of πn (X, x0 ) can be also well defined as a homotopy class relative to
DOI 10.1007/978-81-322-2843-1_7
274 7 Higher Homotopy Groups
sn of the continuous maps f : (S n , sn ) → (X, x0 ), where S n is the unit sphere and

sn = (1, 0, . . . , 0) ∈ Rn+1 is regarded as the base point of S n . For n > 1, there is a
rotation of the n-sphere S n providing a homotopy interchanging its two hemispheres.
It implies the interesting property that πn (X, x0 ) is abelian for n > 1.
More precisely, this chapter defines nth (absolute) homotopy group and gener-
alizes it to (relative) homotopy group of a triplet and studies algebraic, functorial
and fibering properties with exactness of homotopy sequence of fibering along with
Hopf maps introduced by H. Hopf (1894–1971) in 1935 for investigation of certain
homotopy groups of Sn . This chapter also presents action of π1 on πn , Freudenthal
suspension theorem given by H. Freudenthal (1905–1990) in 1937 for investiga-
tion of the homotopy groups πm (S n ) for 0 < m < n and the nth cohomotopy
set π n (X, A) on which K. Borsuk (1905–1982) endowed in 1936 an abelian group
structure under certain conditions on (X, A). This chapter also discusses some inter-
esting applications of higher homotopy groups. Computing the homotopy group
πn (X, x0 ) is very difficult even in the cases of some simple spaces, such as certain sub-
spaces of Euclidean spaces. These computations were the object of investigations of
many prominent topologists, like E. Cartan (1869–1951), S. Eilenberg (1915–1998),
J.P. Serre (1926–) and many others.
For this chapter, the books Croom (1978), Gray (1975), Hatcher (2002),
Hu (1959), Switzer (1975) and some others are referred in Bibliography.
7.1 Absolute Homotopy Groups: Introductory Concept
This section defines and studies absolute homotopy groups πn (X, x 0 )(n > 1), which
are generally called higher homotopy groups of a pointed topological space (X, x 0 ).
For an alternative approach given by W. Hurewicz see Sect. 7.2.
Let I n be the topological product of n-copies of I for n > 1. The nth absolute
homotopy group πn (X, x 0 ) is defined in a way analogous to the construction of the
fundamental group π1 (X, x0 ) of a pointed topological space (X, x 0 ) by replacing I
by n-cube I n and I˙ by the boundary ∂ I n of I n . Every point t ∈ I n is represented by
t = (t1 , t2 , . . . , tn ), ti ∈ I . The real number ti is called the i-th coordinate of t. Thus,
I n = {(t1 , t2 , . . . , tn ) : 0 ≤ ti ≤ 1}.
An (n − 1) - face of I n is given by setting some coordinates ti to be 0 or 1. The union

of all (n − 1) - faces of I n is called the boundary of I n , denoted by I˙n or ∂ I n . Thus
∂ I n = I˙n = {(t1 , t2 , . . . , tn ) ∈ I n : some ti ∈ I˙}, which is topologically equivalent
to (n − 1)- sphere S n−1 . Let Int (I n ) denote the interior of I n and it is considered as
a subspace of I n .
Proposition 7.1.1 Int (I n ) = {(t1 , t2 , . . . , tn ) : 0 < ti < 1}.
Proof Let 0 < ti < 1 for each i and = mini (1 − ti , ti ). Then a disk of radius with
center t = (t1 , t2 , . . . , tn ) is contained in I n . This shows that t ∈ Int (I n ). Again if
7.1 Absolute Homotopy Groups: Introductory Concept 275
for some i, ti = 1 or 0, then a disk of radius r (however small it may be) with center
t contains points with ti > 1. This implies that such points are not in Int (I n ). ❑
Remark 7.1.2 We write ∂ I n = I n − Int (I n ).
We now consider the set
Fn (X, x0 ) = { f : (I n , ∂ I n ) → (X, x 0 ) : f is continuous }.
If Fn (X, x0 ) is topologized by compact open topology (see Chap. 1), then the homo-
topy set [(I n , ∂ I n ), (X, x0 )] relative to ∂ I n , denoted by πn (X, x 0 ) is the set of all
path- components of the space Fn (X, x 0 ).
We recall the definition of relative homotopy.

Definition 7.1.3 Let f, g be two continuous maps in Fn (X, x 0 ). Then f, g are said
to be homotopic related to ∂ I n denoted by f g r el ∂ I n if there exists a continuous
map F : I n × I → X such that
F(t1 , t2 , . . . , tn , 0) = f (t1 , t2 , . . . , tn ), F(t1 , t2 , . . . , tn , 1) = g(t1 , t2 , . . . , tn ), ∀ (t1 , t2 , . . . , tn ) ∈ I n
and F(t1 , t2 , . . . , tn , s) = x 0 , ∀ (t1 , t2 , . . . , tn ) ∈ ∂ I n and s ∈ I .

We have shown in Chap. 2 that the relative homotopy relation is an equivalence
relation. If [ f ] denotes the equivalence class of f ∈ Fn (X, x 0 ), then πn (X, x0 ) is
the set defined by πn (X, x 0 ) = {[ f ] : f ∈ Fn (X, x 0 )}. We are now in a position to
endow the set πn (X, x0 ) with a group operation.
Define a composition ∗ in Fn (X, x0 ) as follows : For f, g ∈ Fn (X, x 0 ), f ∗ g is

defined by

f (2t1 , t2 , . . . , tn ), if 0 ≤ t1 ≤ 1/2
( f ∗ g)(t) =
g(2t1 − 1, t2 , . . . , tn ), if 1/2 ≤ t1 ≤ 1
for every t = (t1 , t2 , . . . , tn ) ∈ I n .

At t1 = 12 , f (1, t1 , . . . , tn ) = x0 and g(0, t2 , . . . , tn ) = x 0 , since (1, t1 , . . . , tn ),
(0, t2 , . . . , tn ) ∈ ∂ I n . Again for the point (t1 , t2 , . . . , tn ) ∈ ∂ I n , the points (2t1 , t2 ,
. . . , tn ) and (2t1 −2, t2 , . . . , tn ) ∈ ∂ I n and hence ( f ∗g)((t1 , t2 , . . . , tn ) = x0 for any
(t1 , t2 , . . . , tn ) ∈ ∂ I n . By pasting lemma (see Chap. 1), f ∗ g is continuous. Clearly,
f ∗ g ∈ Fn (X, x0 ).
Lemma 7.1.4 Given f 1 , f 2 ∈ [ f ] and g1 , g2 ∈ [g], there is a homotopy F : I n ×
I → X such that
F : f 1 ∗ g1 f 2 ∗ g2 rel ∂ I n .
Proof As f 1 f 2 rel ∂ I n and g1 g2 rel ∂ I n , ∃ homotopies G and H such that
G : f 1 f 2 rel ∂ I n and H : g1 g2 rel ∂ I n .
Define F : I n × I → X by the rule

G(2t1 , t2 , . . . , tn , s), if 0 ≤ t1 ≤ 1/2
F(t1 , t2 , . . . , tn , s) = .
H (2t1 − 1, t2 , . . . , tn , s), if 1/2 ≤ t1 ≤ 1.
Then F : f 1 ∗ g1 f 2 ∗ g2 rel ∂ I n . ❑
Define a composition ‘◦’ in πn (X, x0 ) by the rule
[ f ] ◦ [g] = [ f ∗ g] (7.1)
This composition ‘◦’ is clearly independent of the choice of the representatives

of the classes and hence it is well defined.
Theorem 7.1.5 πn (X, x 0 ) is a group under the composition ‘◦’ for n ≥ 1.
Proof For n = 1, π1 (X, x 0 ) has been shown (in Theorem 3.1.12 Chap. 3) to be a
group, called the fundamental group of (X, x0 ) based at x0 . For n > 1, proceed in a
similar way to show that πn (X, x 0 ) is a group under the composition defined in (7.1).
The zero element of the group in the homotopy class of the unique constant map
c : I n → x0 . The inverse element of [ f ] ∈ πn (X, x 0 ) is the homotopy class [ f −1 ],
of the composite map f and ψ, where ψ : I n → I n , t → (1 − t1 , t2 , . . . , tn ), ∀ t ∈
t = (t1 , t2 , . . . , tn ) ∈ I n , i.e., where f −1 : I n → X is the map defined by
f −1 (t1 , t2 , . . . , tn ) = f (1 − t1 , t2 , . . . , tn ), ∀ (t1 , t2 , . . . , tn ) ∈ I n .
To show this it is sufficient to prove that f ∗ f −1 c rel ∂ I n . Consider the continuous

map F : I n × ∂ I n defined by
⎧
⎪
⎪ x0 , if 0 ≤ t1 ≤ s/2
⎪
⎨ f (2t − s, t , . . . , t , s),
1 2 n if s/2 ≤ t1 ≤ 1/2
F(t1 , t2 , . . . , tn , s) =
⎪ f −1 (2t1 + s − 1, t2 , . . . , tn , s),
⎪ if 1/2 ≤ t1 ≤ (1 − s)/2
⎪
⎩
x0 , if (1 − s)/2 ≤ t1 ≤ 1
Clearly, f ∗ f −1 c rel ∂ I n . ❑
Definition 7.1.6 The group πn (X, x0 ) is called the (absolute) homotopy group of
(X, x0 ) for n ≥ 1. For n = 1, this group is called the fundamental group.
Remark 7.1.7 The homotopies defined in the proof of the set πn (X, x 0 ) to be a group
for n ≥ 1 are precisely the same homotopies defined in the proof for the fundamental
group π1 (X, x0 ). While defining the homotopies for πn (X, x0 ) all the actions are in
t1 by keeping the other coordinates unchanged.
7.1 Absolute Homotopy Groups: Introductory Concept 277
Proposition 7.1.8 Let (X, x 0 ) be a pointed topological space and X 0 be the path
component of X containing x 0 , then π1 (X 0 , x0 ) ∼
= πn (X, x0 ) for n ≥ 1.
Proof Since I n is path- connected, the proof is immediate. ❑
7.2 Absolute Homotopy Groups Defined by Hurewicz
This section conveys an alternative method of construction of higher homotopy

groups introduced by Hurewicz (Hurewicz 1935). Given a pointed topological space
X , he defined in 1935 higher homotopy groups by endowing a group structure on
the set πn (X ) = [S n , X ] to classify continuous maps f : S n → X up to homo-
topy equivalence. This classification made the discovery of homotopy groups. His
approach of construction was: if the boundary ∂ I n of the n-cube I n is identified to a
point, a quotient space which is homeomorphic to an n-sphere S n with a base point
sn ∈ S n is obtained and hence an element of πn (X, x 0 ) can be equally well defined
as a homotopy class relative to sn ∈ S n of the maps f : (S n , sn ) → (X, x0 ). This
motivated to define the absolute homotopy groups πn (X, x 0 ). The group π1 (X, x 0 )
is in general nonabelian. On the other hand the groups πn (X, x0 ) are all abelian for
all n ≥ 2.
Definition 7.2.1 For every pointed space (X, x0 ) and every integer n ≥ 0, the
n-th homotopy set is denoted by the set πn (X, x0 ) = [(S n , sn ); (X, x0 )]. For n ≥
2, πn (X, x0 ), it is a group called a (higher) homotopy group and is sometimes called
an absolute homotopy group.
Remark 7.2.2 The higher homotopy group has some interesting properties different
from fundamental group. If n > 1, there exists a rotation of S n which keeps its base
point sn fixed and interchanges the two hemispheres of S n . This implies intuitively
that for n > 1, the group πn (X, x 0 ) is abelian. Its analytical proof is given in Theorem
7.2.5. Moreover, we show as a consequence that every homotopy group of a pointed
space can be expressed as a fundamental group of some other space (see Corollary
7.2.4). Let (X, x 0 ) be a pointed space. The fundamental group π1 (X, x0 ) is not in
general commutative (but commutative for some spaces).
We now use the concepts of loop spaces and suspension spaces defined in the
Chap. 2.
Theorem 7.2.3 For every integer n > 1, the groups πn (X ) and πn−r (r X ) are
isomorphic for all 1 ≤ r ≤ n − 1 for any pointed space X, where X is the loop
space of X and r = × × . . . × (r-times).
Proof If is the suspension functor, then and are adjoint functors (see Propo-
sition 2.5.1). Now
πn (X ) = [S n , X ] ∼
= [S n−r , r X ] = πn−r (r X ).
❑
Corollary 7.2.4 If X is a pointed space, then for n ≥ 2, the groups πn (X ) and

π1 (n−1 X ) are isomorphic.
Proof It follows from Theorem 7.2.3 as a particular case. This shows that for n ≥ 2,
πn (X ) can be equivalently defined by πn (X ) = π1 (n−1 X ) = [S 1 , n−1 X ]. ❑
Theorem 7.2.5 If X is a pointed space, then the group πn (X ) is abelian for all
n ≥ 2.
Proof For n ≥ 2, n−1 X is a loop space of X and hence n−1 X is an Hopf group
(H-group). Hence by Theorem 3.1.39 of Chap. 3, π1 (n−1 X ) is abelian. ❑
The following theorem plays an important role in algebraic topology. This is

sometimes referred as ‘Dimension Axiom.’
Theorem 7.2.6 If X is a singleton space, then πn (X ) = 0 for all n ≥ 0.
Proof As X is a singleton space, there exists only one map c : S n → X , which is the
constant function. Hence [S n , X ] has only one element [c], denoted by 0, ∀ n ≥ 0.
❑
Given pointed spaces (X, x0 ) and (Y, y0 ), like fundamental groups, the groups
πn (X, x0 ), πn (Y, y0 ) and πn (X × Y, (x0 , y0 )) have similar relations.
Theorem 7.2.7 The groups πn (X, x 0 ) × πn (Y, y0 ) and πn (X × Y, (x 0 , y0 )) are iso-
morphic for all n ≥ 1.
Proof Define a function ψ : πn (X, x 0 ) × πn (Y, y0 ) → πn (X × Y, (x 0 , y0 )),

([ f ], [g]) → [( f, g)]. It can be shown that ψ is a homomorphism and is also a
bijection. ❑
7.3 Functorial Properties of Absolute Homotopy Groups
This section presents functorial properties of higher homotopy groups by defining

homomorphisms induced by a base point preserving continuous map between two
pointed topological spaces to the corresponding higher homotopy groups.
Theorem 7.3.1 If f : (X, x0 ) → (Y, y0 ) is continuous, then f induces a homomor-

phism f ∗ : πn (X, x0 ) → πn (Y, y0 ) for n ≥ 1.
Proof For any α ∈ Fn (X, x 0 ), the composition f ◦α ∈ Fn (Y, y0 ) and the assignment
α → f ◦ α defines a map f : Fn (X, x 0 ) → Fn (Y, y0 ). The continuity of f shows
that f carries the path- components of Fn (X, x 0 ) into the path components of
Fn (Y, y0 ). Hence f determines an induced transformation
f ∗ : πn (X, x0 ) → πn (Y, y0 ), [α] → [ f ◦ α], ∀ [α] ∈ πn (X, x 0 ).

7.3 Functorial Properties of Absolute Homotopy Groups 279
Clearly, f ∗ sends the zero element of πn (X, x 0 ) into the zero the element of πn (Y, y0 ).
Moreover, for any two maps α, β ∈ Fn (X, x 0 ), it can be shown that f (α ∗ β) =
( f ◦ α) ∗ ( f ◦ β in Fn (Y, y0 ). Hence f ∗ : πn (X, x0 ) → πn (Y, y0 ) defined by
f ∗ ([α ◦ β]) = ( f ∗ [α]) ◦ [ f ∗ [β]] is a group homomorphism. ❑
Definition 7.3.2 The homomorphism f ∗ : πn (X, x 0 ) → πn (Y, y0 ) defined in Theo-

rem 7.3.1 is called the homomorphism induced by f : (X, x0 ) → (Y, y0 ).
Corollary 7.3.3 Let f, g : (X, x0 ) → (Y, y0 ) be two homotopic maps, then their
induced transformations f ∗ , g∗ : πn (X, x0 ) → πn (Y, y0 ) are the same for every n.
Corollary 7.3.4 If f : (X, x0 ) → (X, x0 ) is the identity map, then f ∗ : πn (X, x 0 ) →

πn (X, x0 ) is the identity transformation on πn (X, x 0 ) for every n.
Corollary 7.3.5 Let f : (X, x0 ) → (Y, y0 ) and g : (Y, y0 ) → (Z , z 0 ) be two base

point preserving maps, then (g ◦ f )∗ = g∗ ◦ f ∗ : πn (X, x 0 ) → πn (Z , z 0 ) for every n.
Corollary 7.3.6 If f : (X, x0 ) → (Y, y0 ) is a homotopy equivalence, then f ∗

induces an isomorphism f ∗ : πn (X, x 0 ) → πn (Y, y0 ) of groups for every n ≥ 1.
Proof It follows from Corollaries 7.3.3–7.3.5. ❑
Corollary 7.3.7 (Homotopy Invariance) If two pointed spaces (X, x0 ) and (Y, y0 )
have the same homotopy type, then for each n ≥ 1, there is a group isomorphism
ψ : πn (X, x0 ) → πn (Y, y0 ).
Proposition 7.3.8 If X is a path-connected space, then for any two points x0 and
x1 , there is an isomorphism ψ : πn (X, x0 ) → πn (X, x1 ) for every n ≥ 1.
Proof Consider a continuous curve x t in X (t ∈ I ), which connects x0 and x1 . Define

a family of maps from I n to X , which takes the boundary of I n along the curve. This
induces a family of homomorphisms
h t : πn (X, x0 ) → πn (X, x t )
for every n ≥ 1 and t ∈ I .

Similarly, we have another family of homomorphisms
gt : πn (X, x t ) → πn (X, x 0 ).
Hence h 1 ◦ g1 = 1d and g1 ◦ h 1 = 1d . Consequently, there is an isomorphism
ψ = h 1 : πn (X, x 0 ) → πn (X, x 1 )
for every n ≥ 1. ❑
Corollary 7.3.9 If the topological spaces X and Y are path- connected and homo-
topy equivalent, then groups πn (X ) and πn (Y ) are isomorphic for every n ≥ 1.
Proof Let f : X → Y be a homotopy equivalence and x0 ∈ X, y0 ∈ Y . Then f

induces an isomorphism f ∗ : πn (X, x 0 ) → πn (Y, y0 ) for every n ≥ 1. As X is path-
connected, the groups πn (X ) ∼
= πn (Y ) for every n ≥ 1. ❑
We summarize the above discussion in the basic and important functorial property
of πn like π1 .
Theorem 7.3.10 Let Htp ∗ be the homotopy category of pointed topological spaces
and Ab be the category of abelian groups and homomorphisms, then πn : Htp ∗ → Ab,
is a covariant functor for each n > 2. For n = 1, π1 is a covariant functor from Htp ∗
to the category Grp of groups and homomorphisms.
Like fundamental groups, higher homotopy groups of a pointed topological space

and higher homotopy groups of its covering spaces have a close relation.
Theorem 7.3.11 Let ( X̃ , p) be a covering space of X and x0 ∈ X . If x̃0 ∈ p −1 (x0 ),

then p induces an isomorphism p∗ : πn ( X̃ , x̃0 ) → πn (X, x 0 ) for every n ≥ 2.
Proof Let p : X̃ → X be a covering map such that p(x̃0 ) = x0 and [ f˜] ∈

πn ( X̃ , x̃0 ) = [S n , X̃ ]. Then p∗ : πn ( X̃ , x̃0 ) → πn (X, x 0 ) given by p∗ [ f˜] = [ p ◦ f˜] ∈
πn (X, x0 ) is a homomorphism. We claim that p∗ is an isomorphism.
Monomorphism: Let [ p ◦ f˜] = [ p ◦ g̃], where f˜, g̃ : S n → X̃ are pointed maps.

Then p ◦ f˜ p ◦ g̃. By using covering homotopy theorem, f˜ g̃ and hence
[ f˜] = [g̃].
Epimorphism: Let [ f ] ∈ πn (X, x 0 ). We now consider the diagram in Fig. 7.1

As S n is simply connected for n ≥ 2, there exists a unique lifting f˜ : S n → X̃ of
f : S n → X such that p ◦ f˜ = f . Hence p∗ [ f˜] = [ f ]. ❑
Theorem 7.3.12 πn (S 1 ) = 0 for all n ≥ 2.
Proof Recall that p : R → S 1 , where p(t) = e2πit is covering map. Thus by using
Theorem 7.3.11, p induces an isomorphism p∗ : πn (R) → πn (S 1 ) for every n ≥ 2.
Since the space R is contractible, πn (R) = 0 for every n ≥ 2. This shows that
πn (S 1 ) = 0 for every n ≥ 2. ❑
Fig. 7.1 Lifting of f q8 X̃

f˜
q qq p
q
qq
S n / X
f
7.4 The Relative Homotopy Groups: Introductory Concepts 281
7.4 The Relative Homotopy Groups: Introductory

Concepts
This section generalizes the concept of absolute homotopy groups πn (X, x0 ) intro-
duced in Sect. 7.1 by defining the relative homotopy groups πn (X, A, x0 ) (with n ≥ 2)
for any triplet (X, A, x0 ) which is an important concept in homotopy theory. This
generalization was given by Hurewicz (1904–1956) and studied by him during 1935–
1936.
Definition 7.4.1 A triplet (X, A, x0 ) consists of a topological space X , a nonempty
subspace A of X and a point x 0 ∈ A. If x0 is the only point of A, then the triplet is
simply (X, x 0 ).
Construction of πn (X, A, x0 )
Let I n be the cartesian product of the unit interval I = [0, 1] for n > 0, called
the n-cube. Thus I n = {(t1 , t2 , . . . , tn ) : ti ∈ I, i = 1, . . . , n}. The initial (n − 1)-
face of I n defined by tn = 0, is identified with I n−1 hereafter. The union of all
remaining (n − 1)-faces of I n is denoted by J n−1 . Hence ∂ I n = I n−1 ∪ J n−1 and
∂ I n−1 = I n−1 ∩ J n−1 .
A map f : (I n , I n−1 , J n−1 ) → (X, A, x0 ) is a continuous function f : I n → X
such that f (I n−1 ) ⊂ A and f (J n−1 ) = x0 . In particular, f : (∂ I n , ∂ I n−1 ) →
(A, x 0 ) is a continuous function. Let Fn = Fn (X, A, x0 ) be the set of all such maps. It
is topologized by the compact open topology. Two maps f, g ∈ Fn (X, A, x0 ) are said
to be homotopic relative to the system {I n−1 , A; J n−1 , x 0 } if there exists a homotopy
Ht : I n → X, t ∈ I such that H0 = f, H1 = g and Ht ∈ Fn (X, A, x0 ). In notation,
f g rel {I n−1 , A; J n−1 , x1 }. Let πn (X, A, x0 ) be the set of homotopy classes of all
these maps relative to the system {I n−1 , A; J n−1 , x0 }). Let [ f ] denote the homotopy
class of f ∈ Fn and 0 the homotopy class of the constant map c(I n ) = x0 . Then
the set πn (X, A, x0 ) is the set of all path components of the space Fn . For n ≥ 1,
define a composition ‘∗’ in Fn (X, A, x0 ) as follows: For f, g ∈ Fn (X, A, x0 ), 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.
for every t = (t1 , t2 , . . . , tn ) ∈ I n .

Then f ∗ g is continuous and f ∗ g ∈ Fn (X, A, x 0 ).
Define a composition in πn (X, A, x 0 ) by the rule [ f ] ◦ [g] = [ f ∗ g] for all
[ f ], [g] ∈ πn (X, A, x 0 ). This composition is independent of the choice of the repre-
sentatives of the classes and hence it is well defined.
Remark 7.4.2 For f, g ∈ Fn (X, A, x 0 ), f ∗ g may 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.
So, it is immaterial which coordinate we use to define f ∗ g.

Theorem 7.4.3 πn (X, A, x 0 ) is a group under the composition ‘◦’ for n ≥ 1.
Proof Proceed as in the case of πn (X, x0 ) (see Theorem 7.1.5). ❑
Remark 7.4.4 If J n−1 is identified to a point s0 , then the triplet (I n , I n−1 , J n−1 )
admits a configuration equivalent to the triplet (D n , S n−1 , s0 ) consisting of the unit
n-cell D n , its boundary S n−1 , and a base point s0 ∈ S n−1 . This shows that an element
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 : (D n , S n−1 , s0 ) → (X, A, x0 ).
Proposition 7.4.5 For n > 2, the group πn (X, A, x0 ) is abelian.
Proof An element of πn (X, A, x0 ) is a homotopy class of a map f : (D n , S n−1 , s0 ) →
(X, A, x0 ). If n > 2, there exists a rotation of D n which leaves the point s0 fixed and
interchanges two halves of D n . Hence the group πn (X, A, x 0 ) is abelian for n > 2.
❑
We now search the maps f : (D n , S n−1 , s0 ) → (X, A, x 0 ) which define the zero
element of πn (X, A, x0 ).
Proposition 7.4.6 If an element [ f ] ∈ πn (X, A, x 0 ) is represented by a map f ∈
Fn (X, A, x0 ) such that f (I n ) ⊂ A, then [ f ] = 0.
Proof Since f ∈ Fn (X, A, x 0 ) and f (I n ) ⊂ A, there exists a homotopy Ht ∈
Fn (X, A, x0 ), t ∈ I defined by
Ht (t1 , t2 , . . . , tn−1 , tn ) = f (t1 , . . . , tn−1 , t + tn − ttn ).
Then H0 = f , and H1 (I n ) = x0 . Hence [ f ] = 0. ❑
7.5 The Boundary Operator and Induced Transformation
This section conveys the algebraic properties of boundary operator and induced
transformation by defining boundary operator ∂ : πn (X, A, x0 ) → πn−1 (A, x0 ) for
n > 0 and the transformation f∗ : πn (X, A, x 0 ) → πn (Y, B, y0 ) induced by a
continuous map f : (X, A, x0 ) → (Y, B, x 0 ). They play a central role in the study
of homotopy sequence.
7.5.1 Boundary Operator
This subsection studies the boundary operator ∂ which is an important concept in

higher homotopy groups.
7.5 The Boundary Operator and Induced Transformation 283
Definition 7.5.1 Let (X, A, x 0 ) be a triplet and n > 0 be an integer. If [ f ] ∈

πn (X, A, x0 ), then it is represented by a continuous map f : (I n , I n−1 , J n−1 ) →
(X, A, x 0 ). If n = 1, f (I n−1 ) is a point of A which gives a path component [g] ∈
πn−1 (A, x0 ) of A. If n > 1, then the map f | I n−1 : (I n−1 , ∂ I n−1 ) → (A, x0 ) represents
an element [g] ∈ πn−1 (A, x0 ), independent of the choice of the representative map
f . The boundary operator ∂ : πn (X, A, x 0 ) → πn−1 (A, x0 ) for n > 0 is defined by
setting ∂([ f ]) = [g].
We now give some properties of ∂.

Proposition 7.5.2 The boundary operator ∂ sends the zero element of πn (X, A, x 0 )
into the zero element of πn−1 (A, x0 ).
Proof It follows from definition of ∂ : πn (X, A, x 0 ) → πn−1 (A, x0 ). ❑
Proposition 7.5.3 For n > 1, the boundary operator ∂ : πn (X, A, x0 ) → πn−1

(A, x0 ) is a homomorphism.
Proof It follows from the definition of ∂. ❑
7.5.2 Induced Transformations
This subsection generalizes the concept of induced transformations given in Theo-

rem 7.3.1 and relates it to the boundary operator ∂ for triplets. Let f : (X, A, x0 ) →
(Y, B, y0 ) be a continuous map which means that f is a continuous map from
X to Y such that f (A) ⊂ B and f (x0 ) = y0 . Then f sends the path com-
ponents of X into the path components of Y . Hence f induces a transformation
f ∗ : π0 (X, x0 ) → π0 (Y, y0 ), which sends the zero element of π0 (X, x 0 ) into the zero
element of π0 (Y, y0 ).
Theorem 7.5.4 If f : (X, A, x 0 ) → (Y, B, y0 ) is continuous, then f induces a

homomorphism f ∗ : πn (X, A, x0 ) → πn (Y, B, y0 ) for n ≥ 1.
Proof If n ≥ 1, then for any map g ∈ Fn (X, A, x 0 ), the composite map f ◦ g ∈

Fn (Y, B, y0 ) and hence the corresponding g → f ◦ g defines a continuous func-
tion f : Fn (X, A, x0 ) → Fn (Y, B.y0 ) such that f sends the path components
of Fn (X, A, x0 ) to the path components of Fn (Y, B, y0 ). Hence f induces a trans-
formation f ∗ : πn (X, A, x 0 ) → πn (Y, B, y0 ), which carries the zero element of
πn (X, A, x0 ) to the zero element of πn (Y, B, y0 ). For n = 1, A = {x0 }, B = {y0 }
or for n > 1, πn (X, A, x 0 ) and πn (Y, B, y0 ) are groups and f ∗ is a homomorphism,
called the homomorphism induced by f . ❑
Corollary 7.5.5 The identity map 1d : (X, A, x 0 ) → (X, A, x 0 ) induces the identity
transformation 1d∗ : πn (X, A, x0 ) → πn (X, A, x0 ) for all n ≥ 0.
Proof It follows from the induced transformation 1d∗ . ❑

Fig. 7.2 Commutativity of f∗

πn (X, A, x0 ) / πn (Y, B, y0 )
the rectangle for ∂ and
∂ ∂
induced transformations
g∗
πn−1 (A, x0 ) / πn−1 (B, y0 )
Corollary 7.5.6 If f : (A, X, x0 ) → (Y, B, y0 ) and g : (Y, B, y0 ) → (Z , C, z 0 )

are maps of triples, then for every n ≥ 0, (g ◦ f )∗ = g∗ ◦ f ∗ : πn (X, A, x 0 ) →
πn (Z , C, z 0 ).
Proof It follows from the definition of induced transformations f ∗ and g∗ . ❑
We now give a relation between the boundary operator and the induced transfor-
mation.
Proposition 7.5.7 If f : (X, A, x0 ) → (Y, B, y0 ) is a continuous function and if
f | A = g : (A, x0 ) → (B, y0 ) is the restriction of f , then the diagram in Fig. 7.2 is
commutative for every n ≥ 1, i.e., ∂ ◦ f∗ = g∗ ◦ ∂ holds.
Proof It follows from definitions of ∂, induced transformations f ∗ , and g∗ that the
diagram in Fig. 7.2 is commutative. ❑
7.6 Functorial Property of the Relative Homotopy Groups
This section continues to study homotopy theory by displaying the functorial prop-
erties of the relative homotopy groups πn (X, A, x0 ) (n ≥ 2) and homotopy proper-
ties of maps f ∈ Fn (X, A, x0 ) and considers the homotopy equivalence of a map
f ∈ Fn (X, A, x 0 ).
Definition 7.6.1 Let f, g : (X, A, x0 ) → (Y, B, b0 ) be two continuous maps. They
are said to be homotopic relative to the system {A, B, x 0 , y0 } (or simply homotopic)
if there exists a map Ht : (X, A, x0 ) → (Y, B, b0 ), t ∈ I such that H0 = f and
H1 = g.
Proposition 7.6.2 Let f, g : (X, A, x0 ) → (Y, B, b0 ) be two homotopic maps. Then
their induced transformation are equal, i.e., f ∗ = g∗ : πn (X, A, x 0 ) → (Y, B, b0 ) for
every n.
Proof To prove it, we have to show that f ∗ (α) = g∗ (α), ∀ α ∈ πn (X, A, x 0 ).
Case 1: If n = 0, A = x0 and B = y0 , then α is a path component of X . If x ∈ α,
then f ∗ (α) and g∗ (α) are path components of Y containing the points f (x) and g(x),
respectively. Let Ht : f g. Define a path β : I → Y, t → Ht (x). Then the path β
joins f (x) to g(x) and hence f ∗ (α) = g∗ (α), ∀ α ∈ πn (X, A, x 0 ). This shows that
f ∗ = g∗ .
7.6 Functorial Property of the Relative Homotopy Groups 285
Case 2: If n > 0, choose a map m ∈ Fn (X, A, x 0 ) such that [m] = α. Then

f ∗ (α) and g∗ (α) are represented by the maps f ◦ m and g ◦ m, respectively. Hence
Ht ◦ m : f ◦ m g ◦ m shows that f ∗ ([m]) = g∗ ([m]). Consequently, f ∗ (α) =
g∗ (α), ∀ α ∈ πn (X, A, x 0 ) implies f ∗ = g∗ . ❑
Definition 7.6.3 A map f : (X, A, x 0 ) → (Y, B, b0 ) is said to be a homotopy
equivalence if there exists a map g : (Y, B, b0 ) → (X, A, x0 ) such that g ◦ f is
homotopic to the identity map on (X, A, x0 ) and f ◦ g is homotopic to the identity
map on (Y, B, b0 ).
Proposition 7.6.4 If f : (X, A, x 0 ) → (Y, B, b0 ) is a homotopy equivalence, then
the induced transformation f ∗ : πn (X, A, x0 ) → πn (Y, B, b0 ) is an isomorphism
for every n > 1.
Proof It follows from Definition 7.6.3 and Proposition 7.6.2. ❑
Corollary 7.6.5 If the topological spaces X and Y are path - connected and homo-
topy equivalent, then the groups πn (X, A, x 0 ) and πn (Y, B, y0 ) are isomorphic for
every n > 1.
Proof Proceed as in Corollary 7.3.9. ❑
Corollary 7.6.6 πn (X, A, x0 ) depends on the homotopy type of (X, A, x0 ).
Corollary 7.6.7 If n ≥ 1 and A is a strong deformation retract of X , then i∗ :
πn (A, x0 ) → πn (X, x0 ) is an isomorphism.
Proof The corollary follows from Corollary 7.6.6. ❑
Theorem 7.6.8 Let Htp 2 be the homotopy category of triplets and their cotinuous
maps, and Ab be the category of abelian groups and homomorphisms. Then πn :
Htp 2 → Ab is a covariant functor for each n > 2.
Proof It follows Theorem 7.5.4, Proposition 7.6.2 and Corollaries 7.5.5–7.5.6. ❑
7.7 Homotopy Sequence and Its Exactness
This section defines homotopy sequence and proves its exactness, which provides
powerful tools for the study of homotopy theory, specially for computing homotopy
groups of certain spaces and proves also some immediate consequences of exactness
of homotopy sequences.
7.7.1 Homotopy Sequence and Its Exactness
Given a triplet (X, A, x 0 ), the inclusion maps i : (A, x 0 ) → (X, x0 ), and j :

(X, x0 ) → (X, A, x0 ) induce transformations i ∗ : πn (A, x0 ) → πn (X, x0 ), j∗ :
πn (X, x0 ) → πn (X, A, x 0 ). The transformations
i ∗ , j∗ and ∂ give rise to a beginningless sequence:

j∗ ∂ i∗ j∗
· · · → πn+1 (X, x 0 ) −−−−→ πn+1 (X, A, x 0 ) −−−→ πn (A, x 0 ) −−−→ πn (X, x 0 ) −−−−→ πn (X, A, x0 )
∂ ∂ i∗
−−−→ · · · π1 (X, A, x0 ) −−−→ π0 (A, x 0 ) −−−→ π0 (X, x 0 ) (7.2)
called the homotopy sequence of the triplet (X, A, x0 ), denoted by π(X, A, x 0 ).

Every set in (7.2) has its zero element and transformation in (7.2) sends the zero
element into the zero element.
Definition 7.7.1 The sequence (7.2) of any triplet (X, A, x0 ) is said to be exact if the
kernel of each transformation is the same as the image of the preceding transformation
Theorem 7.7.2 (Exactness of homotopy sequence) The homotopy sequence (7.2) of

any triplet (X, A, x0 ) is exact.
Proof To prove the exactness, we show that

(i) Im j∗ = ker ∂, i.e., Im j∗ ⊆ ker ∂ ⊆ Im j∗ ;
(ii) Im ∂ = ker i ∗ , i.e., Im ∂ ⊆ ker i ∗ ⊆ Im ∂;
(iii) Im i ∗ = ker j∗ , i.e., Im i ∗ ⊆ ker j∗ ⊆ Im i ∗ .
(i) Im j∗ = ker ∂: For each n > 0, let [ f ] ∈ πn (X, x 0 ). Then for each rep-
resentative f ∈ Fn (X, x 0 ), (∂ ◦ j∗ )([ f ]) is determined by the restriction
j ◦ f | I n−1 = f | I n−1 . Since f (I n−1 ) = x0 , it follows that ∂ ◦ j∗ = 0. Hence
Im j∗ ⊆ ker ∂. For the reverse inclusion, let n > 1 and f ∈ Fn (X, A, x 0 )
represent [ f ] ∈ πn (X, A, x0 ). Then ∂[ f ] = 0 shows that there exists a
homotopy Ht : I n−1 → A such that H0 = f | I n−1 , H1 (I n−1 ) = x 0 and
Ht (∂ I n−1 ) = x0 , ∀ t ∈ I . Define a map

Ht (s), if s ∈ I n−1
Ft : ∂ I n → A, s →
x0 , if s ∈ J n−1
for all t ∈ I . Then F0 = f |∂ I n . Hence by HEP, the homotopy Ft has a extension

t : I n → X such that F
F 0 = f . Again since F 1 (∂ I n ) = F1 (∂ I n ) = x 0 , F
1

represents an element [g] ∈ πn (X, x 0 ). Since Ft ∈ Fn (X, A, x 0 ), it follows
that j∗ ([g]) = [ f ]. If n = 1, [ f ] is represented by a path f : I → X such
that f (0) ∈ A and f (1) = x0 . The given condition ∂([ f ]) = 0 shows that
f (0) is contained in the same path-component of A as x 0 . Hence there exists
a homotopy Ft : I → X such that F0 = f, Ft (0) ∈ A, Ft (1) = x 0 and
F1 (0) = x0 . Consequently, F1 represents an element [g] ∈ π1 (X, x 0 ) and hence
the homotopy Ft shows that j∗ ([g]) = [ f ].Thus ker ∂ ⊆ Im j∗ . Consequently,
Im j∗ = ker ∂.
(ii) Im ∂ = ker i ∗ : For each n > 0, let [ f ] ∈ πn (X, A, x 0 ) and f ∈ Fn (X, A, x 0 )
represent [ f ]. Then the element (i ∗ ◦ ∂)([ f ]) is given by g = f | I n−1 . Define a
homotopy
7.7 Homotopy Sequence and Its Exactness 287
G t : I n−1 → X, (t1 , . . . , tn−1 ) → f (t1 , t2 , . . . , tn−1 , t).
Then G 0 = g, G 1 (I n−1 ) = x0 and G t ∈ Fn−1 (X, x 0 ) for n > 1. Hence

(i ∗ ◦ ∂)([ f ]) = 0 implies i ∗ ◦ ∂ = 0. This shows that Im ∂ ⊆ ker i ∗ . For the
converse inclusion, first let n > 1 and [ f ] ∈ πn (A, x) be represented by f ∈
Fn−1 (A, x0 ) such that i ∗ [ f ] = 0. Then there exists a homotopy Ft : I n−1 → X
such that F0 = f, F1 (I n−1 ) = x0 and Ft (∂ I n−1 ) = x0 . Define a map
g : I n → X, (t1 , t2 , . . . , tn−1 , tn ) → Ftn (t1 , t2 , . . . , tn−1 ).
Then g ∈ Fn (X, A, x 0 ) represents an element [g] ∈ πn (X, A, x0 ). Since g| I n1 =

f, it follows that ∂([g]) = [ f ]. The remaining part for n = 1 is left as an
exercise. Hence ker i∗ ⊆ Im ∂. Consequently, Im ∂ = ker i ∗ .
(iii) Im i ∗ = ker j∗ : For n ≥ 1, we claim that j∗ ◦ i ∗ = 0. Let [ f ] ∈ πn (A, x0 ) and
f ∈ Fn (A, x0 ) represent [ f ]. Then the element ( j∗ ◦ i ∗ )([ f ]) ∈ πn (X, A, x 0 )
is represented by j ◦ i ◦ f ∈ Fn (X, A, x 0 ). Since ( j ◦ i ◦ f )(I n ) ⊂ A, it follows
by Proposition 7.4.6 that ( j∗ ◦ i ∗ )([ f ]) = 0, ∀ [ f ] ∈ πn (A, x0 ). This shows
that j∗ ◦ i ∗ = 0. Hence Im i ∗ ⊆ ker j∗ . To show the reverse inclusion, let [ f ] ∈
πn (X, x0 ) be such that j∗ ([ f ]) = 0. Choose f ∈ Fn (X, x 0 ) a representative of
[ f ]. Then j∗ ([ f ]) = 0 shows that there is a homotopy Ft : I n → X such that
F0 = f, F1 (I n ) = x 0 , and Ft ∈ Fn (X, A, x0 ). Define a homotopy

F2tn (t1 , t2 , . . . , tn−1 , 0), if 0 ≤ 2tn ≤ t
G t : I n → X, (t1 , t2 , . . . , tn−1 , tn ) → n −t
Ft (t1 , t2 , . . . , tn−1 , 2t2−t ), if t ≤ 2tn ≤ 2.
Then G 0 = f, G 1 (I n ) ⊂ A and G t (∂ I n ) = x 0 for all t ∈ I . Hence G 1 repre-

sents an element [g] ∈ πn (A, x0 ) and the homotopy G t shows that i ∗ ([g]) = [ f ].
Hence ker j∗ ⊆ Im i ∗ . Consequently, Im i ∗ = ker j∗ . ❑
7.7.2 Some Consequences of the Exactness of the Homotopy

Sequence
This subsection presents some immediate consequences of the exactness of the homo-
topy sequence (7.2).
Proposition 7.7.3 Let (X, A, x0 ) be a triplet, A be a retract of X and x 0 ∈ A.
Then πn (X, x0 ) ∼
= πn (A, x0 ) ⊕ πn (X, A, x 0 ) for any n ≥ 2 and the inclusion map
i : A → X induces a monomorphism i ∗ : πn (A, x0 ) → πn (X, x0 ) for any n ≥ 1.
Proof Let r : X → A be a retraction. Then r ◦i = 1 A shows that r∗ ◦i ∗ is the identity
automorphism on πn (A, x0 ) for every n ≥ 1. Consequently, i ∗ is a monomorphism
and r∗ is an epimorphism for every n ≥ 1. Again for n ≥ 2, since the group πn (X, x 0 )
is abelian, it follows from r∗ ◦ i ∗ = 1d that the group πn (X, x0 ) decomposes into
the direct sum πn (X, x 0 ) = B ⊕ C, where B = Im i ∗ and C = ker r∗ . Since i ∗ is a

monomorphism, B ∼ = πn (A, x0 ). Again it follows from the exactness of the homotopy
sequence of (X, A, x0 ) that j∗ : πn (X, x 0 ) → πn (X, A, x0 ) is an epimorphism for
any n ≥ 2, and ker j∗ = Im i ∗ = B and j∗ maps C isomorphically onto πn (X, A, x 0 ).
Consequently, C ∼ = πn (X, A, x0 ). ❑
Remark 7.7.4 For a given triplet (X, A, x0 ), if A is a retract of X , then the group
π2 (X, A, x0 ) is abelian.
Definition 7.7.5 Let (X, A, x0 ) be a triplet. Then X is said to be deformable into

A relative to a point x 0 ∈ A, if there exists a homotopy Ht : X → X such that
H0 (x) = x, H1 (x) ∈ A and Ht (x0 ) = x0 , ∀ x ∈ X, t ∈ I.
Proposition 7.7.6 Let (X, A, x 0 ) be a triplet and X be deformable into A relative

to a point x0 ∈ A. If i : A → X is the inclusion map, then πn (A, x0 ) ∼= πn (X, x0 ) ⊕
πn+1 (X, A, x0 ) for every n ≥ 2 and i ∗ : πn (A, x0 ) → πn (X, x0 ) is an epimorphism
for every n ≥ 1.
Proof Since X is deformable into A relative to the point x0 ∈ A, there exists a

homotopy Ht : X → X such that H0 (x) = x, H1 (x) ∈ A and Ht (x0 ) = x0 , ∀ x ∈ X
and t ∈ I . Define a map
f : (X, x0 ) → (A, x0 ), x → H1 (x).
Clearly, i ◦ f = H1 H0 rel x0 . Hence i ∗ ◦ f ∗ is the identity automorphism on

πn (X, x0 ). Hence f ∗ is a monomorphism and i ∗ is an epimorphism for every n ≥ 1.
If n ≥ 2, then the group πn (A, x0 ) is abelian. Hence i ∗ ◦ f∗ = 1d implies that
πn (A, x0 ) = Im f ∗ ⊕ ker i ∗ , where i ∗ is an epimorphism and f ∗ is a monomorphism
for any n ≥ 1. This shows that Im f ∗ ∼ = πn (X, x 0 ). Again from exactness of the
homotopy sequence of (X, A, x0 ) it follows that ∂ : πn+1 (X, A, x0 ) → πn (A, x0 ) is
a monomorphism. Consequently, ker i ∗ = Im ∂ ∼ = πn+1 (X, A, x0 ). ❑
Definition 7.7.7 Let (X, A) be a pair of topological spaces. It is called 0-connected

if every path component of X intersects A. The pair (X, A) is called n-connected if
(X, A) is 0-connected and πm (X, A, a) = 0 for 1 ≤ m ≤ n and for all a ∈ A. A
topological space X is called n-connected if πm (X, x) = 0 for 0 ≤ m ≤ n and for
all x ∈ X .
We now characterize n-connected spaces with the help of inclusion maps.
Proposition 7.7.8 Let (X, A) be a pair of topological spaces. Then (X, A) is n-

connected (n ≥ 0) iff for the inclusion map i : (A, x0 ) → (X, x0 ), the induced map
i ∗ : πm (A, x0 ) → πm (X, x 0 ) is a bijection for m < n and a surjection for m = n,
and for all x0 ∈ A.
Proof It follows from the homotopy sequence (7.2) of the triplet (X, A, x0 ). ❑
7.8 Homotopy Sequence of Fibering and Hopf Fibering 289
7.8 Homotopy Sequence of Fibering and Hopf Fibering
This section studies the homotopy sequence of fibering which exists corresponding
to a fiber space. and describes Hopf fibering: p : S 2n−1 → S n , for n = 2, 4, 8. They
provide tools in computing higher homotopy groups of certain topological spaces.
H. Hopf (1894–1975) described various fiberings of spheres by spheres in his paper
(Hopf 1935).
7.8.1 Homotopy Sequence of Fibering
This subsection discusses the homotopy sequence of a fibering. Let p : X → B be

a projection, b0 ∈ B and F = p −1 (b0 ) = ∅ be the fiber space of p. If x 0 ∈ F,
then (X, F, x 0 ) forms a triplet. Since p(F) = b0 , the map p : X → B defines
a map q : (X, F, x 0 ) → (B, b0 ) such that p = q ◦ j, where j : (X, x0 ) →
(X, F, x0 ) is the inclusion map.Then by fibering property of p (see Ex.9 of Sect. 7.14),
q∗ : πn (X, F, x0 ) → πn (B, b0 ) is a bijection for n ≥ 1. Then q∗−1 exists. Define
d∗ = ∂ ◦ q∗−1 : πn (B, b0 ) → πn−1 (F, x0 ) for n ≥ 1. Hence the homotopy exact
sequence (7.2) produces the following exact sequence
p∗ d∗ i∗ p∗
· · · −−−−→ πn+1 (B, b0 ) −−−→ πn (F, x0 ) −−−→ πn (X, x0 ) −−−−→ πn (B, x0 )
d∗ p∗ d∗ i∗
−−−→ · · · −−−−→ π1 (B, b0 ) −−−→ π0 (F, x0 ) −−−→ π0 (X, x 0 )
(7.3)
called the homotopy sequence of the fibering p : X → B based at x0 .
Proposition 7.8.1 If the fiber F in sequence (7.3) is totally disconnected, then p∗ :

πn (X, x0 ) → πn (B, b0 ) is an isomorphism for n ≥ 2 and p∗ is a monomorphism for
n = 1.
Proof By hypothesis, πn (F, x0 ) = 0 for n ≥ 1. Hence the proposition follows from

the exactness of the homotopy sequence (7.3). ❑
Proposition 7.8.2 Let p : X → B be a fibering such that it admits a cross section

s : B → X . Then for every b0 ∈ B and x0 = s(b0 ) ∈ F = p−1 (b0 ), πn (X, x 0 ) ∼ =
πn (B, b0 )⊕πn (F, x0 ) for n ≥ 2 and p∗ : πn (X, x 0 ) → πn (B, b0 ) is an epimorphism
for n ≥ 1.
Proof p ◦ s = 1 B implies that p∗ ◦ s∗ : πn (B, b0 ) → πn (B, b0 ) is the identity

automorphism. Hence s∗ : πn (B, b0 ) → πn (X, x 0 ) is a monomorphism and p∗ :
πn (X, x0 ) → πn (B, b0 ) is an epimorphism for n ≥ 1. For n ≥ 2, πn (X, x0 ) is
abelian and hence p∗ ◦ s∗ = 1d shows that πn (X, x0 ) is the direct sum
πn (X, x0 ) = Im s∗ ⊕ ker p∗ .
Since s∗ is a monomorphism, Im s∗ = ∼ πn (B, b0 ). Again as p∗ is an epimorphism for

every n ≥ 1, it follows from (7.3) that i ∗ is a monomorphism for every n ≥ 1. Hence
ker p∗ = Im i ∗ ∼= πn (F, x0 ). ❑
7.8.2 Hopf Fiberings of Spheres
This subsection discusses the problem whether a continuous map p : S m → S n

for m > n > 1 is necessarily nullhomotopic which was not known until 1935.
The problem was solved in 1935 by H. Hopf with the discovery of his famous
map p : S 3 → S 2 , now called Hopf map. This section studies Hopf fibering: p :
S 2n−1 → S n , for n = 2, 4, 8. Then in each case, the fiber F is an (n − 1)-sphere S n−1
and is contractible in S 2n−1 . Hence πm (S n ) ∼
= πm (S 2n−1 ) ⊕ πm−1 (S n−1 ) for every
n = 2, 4, 8 and every m ≥ 2.
(i) In particular, if n = 2, then πm (S 2 ) ∼
= πm (S 3 ) for m ≥ 3.
(ii) If n = 4 or 8, then
πm (S 4 ) ∼
= πm−1 (S 3 ), for 2 ≤ m ≤ 6,
πm (S 8 ) ∼
= πm−1 (S 7 ), for 2 ≤ m ≤ 14,
π7 (S 4 ) ∼
= Z ⊕ π6 (S 3 ),
π15 (S 8 ) ∼
= Z ⊕ π14 (S 7 ),
7.8.3 Problems of Computing πm (Sn )
This subsection studies the problems of computing the homotopy groups πm (S n ).

The spheres S n are perhaps the simplest noncontractable spaces (see Chap. 10).
The homotopy groups πm (S n ) are not completely determined. The homotopy group
πm (S n ) for m ≤ n are known. Computing the homotopy groups completely is one
of the major unsolved problems. It has been shown that
(i) πm (S n ) = 0, for m < n, by Corollary 7.10.4.
(ii) πm (S 1 ) = 0, for m > 1, by Theorem 7.3.12.
(iii) π1 (S 1 ) ∼
= Z (see Chap. 3).
(iv) πn (S n ) ∼
= Z (see Corollary 7.10.5).
Remark 7.8.3 One may conject that πm (S n ) = 0 for m > n. But this is not true.
H. Hopf first disproves this conjecture in 1931 by showing that π3 (S 2 ) is not trivial
7.8 Homotopy Sequence of Fibering and Hopf Fibering 291
Table 7.1 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 Z 2 × 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
(see Theorem 7.9.1). It is isomorphic to Z. Many other examples for πm (S n ) (for

m > n) are known for particular pair of integers m and n (see a sample Table 7.1)
but not known in all possible cases.
7.9 More on Hopf Maps
This section studies in general, continuous maps p : S 2n−1 → S n for n = 2, 4, 8

introduced by H. Hopf in 1935 (Hopf 1935), now called Hopf maps while investigat-
ing certain homotopy groups of spheres. These three are the early examples of bundle
spaces such that p is not homotopic to a constant map which 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 showing that π3 (S 2 ) = 0.
Theorem 7.9.1 π3 (S 2 ) = 0.
Proof To prove this, we consider Hopf map p : S 3 → S 2 . Define S 3 = {(z, ω) ∈
C × C : |z|2 + |ω|2 = 1}. Let ρ be an equivalence relation on S 3 defined by
(z, ω)ρ(z , ω ) ⇔ (z, ω) = (λz , λω ) for some λ ∈ C such that |λ| = 1. Let M =
S 3 /ρ be the quotient space topologized by the quotient topology and p : S 3 → M be
the projection map, (z, ω) → z, ω. For z, ω ∈ M, p −1 (z, ω), called the fiber
over z, ω, is a great circle of S 3 . Clearly, M is homeomorphic to S 2 . Hence replacing
M by S 2 , we obtain the Hopf map p : S 3 → S 2 . In this way, S 3 is decomposed into
a family of great circles with S 2 as a quotient space.
We claim that p is not homotopic to a constant map. If possible, let H : S 3 × I →
2
S be a homotopy between p and a constant map c. Then it allows a homotopy
H : S 3 × I → S 3 such that the triangle in Fig. 7.3 is commutative. The map H is a
homotopy between the identity map on S and a constant map. This shows that S 3
3
is contractible, which is a contradiction. Hence π3 (S 2 ) = 0. ❑

Fig. 7.3 Commutative 7 S3

nnn
nnn
diagram for Hopf map p H
n nn p
n
nnn
3
S ×I / S 2
H
Theorem 7.9.2 π7 (S 4 ) = 0.
Proof To prove this theorem, we consider the Hopf map p : S 7 → S 4 , where S 7 is

represented by the unit sphere S 7 given by
S 7 = {(z, ω) ∈ H × H : ||z||2 = 1},
where H is the division ring of quaternions. Let D be the unit disk in H defined by
D = {z ∈ H : ||z|| ≤ 1}.
Identifying the boundary of D to a single point we obtain the quotient space M

of the unit disk D. Since the real dimension of D is 4, the quotient space M is
homeomorphic to S 4 . Proceeding as in Theorem 7.9.1 the Hopf map p : S 7 → S 4
proves that π7 (S 4 ) = 0. ❑
Theorem 7.9.3 π15 (S 8 ) = 0.
Proof To prove this theorem we consider the Hopf map p : S 15 → S 8 . In R16 , we

can perform a similar construction as in Theorems 7.9.1 and 7.9.2 which shows that
π15 (S 8 ) = 0. ❑
Theorem 7.9.4 If m > 1, then πm (S 1 , s0 ) = 0.
Proof Let p : R → S 1 be a covering map. Any continuous map f : S m → S 1 lifts

to a map f˜ : S m → R by the lifting property, since S m is simply connected. Again
since S m is simply connected, f˜ is nullhomotopic. Projecting this homotopy (to a
constant map) to S 1 we can define a homotopy of f to a constant map.This implies
that πm (S 1 , s0 ) = 0. ❑
7.10 Freudenthal Suspension Theorem and Table

of πi (Sn ) for 1 ≤ I, n ≤ 8
This section studies Freudenthal suspension theorem with its immediate conse-
quences and displays a table showing the values of πi (S n ) for 1 ≤ i, n ≤ 8. One of
the deepest problems in homotopy theory is computing homotopy groups πn+m (S n ).
Hans Freudenthal was partially successful in 1937 in solving such problems.
Freudenthal suspension theorem is a fundamental theorem in algebraic topology.
7.10 Freudenthal Suspension Theorem and Table of πi (S n ) for 1 ≤ I, n ≤ 8 293
It demonstrates the behavior of simultaneously taking suspensions and increasing

the index of the homotopy groups of the space in question. The impact of suspen-
sion functor comes from the classical theorem of Freudenthal which facilitates to
study the deepest problems of homotopy theory to compute the homotopy groups
of spheres. Moreover, this theorem implies the concept of stabilization of homotopy
groups and ultimately leads to stable homotopy theory which is a generalized coho-
mology theory (see Chap. 15). Stable homotopy groups of spheres are one of the
most important objects in algebraic topology. Moreover, this section gives Table 7.1
displaying a small sample of πi (S n ) extracted from the paper (Toda 1962).
7.10.1 Freudenthal Suspension Theorem
This subsection studies Freudenthal suspension theorem proved by Hans

Freudenthal in 1937 which establishes the stable range for homotopy groups.
Freudenthal observed that the suspension operation on topological spaces shifts
by one their low-dimensional homotopy groups. This observation was important
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 (see Chap. 15.) More
precisely, for each pair m and n of positive integers, there is a natural homomor-
phism E : πm (S n ) → πm+1 (S n+1 ). This homomorphism is called the Freudenthal
suspension homomorphism defined by H. Freudenthal in 1937 (Freudenthal 1937).
Definition 7.10.1 Consider πm (S n ) as homotopy classes of maps f : (S m , 1) →
(S n , 1) and S n as the subspace of S n+1 consisting of all points of S n+1 with last
coordinate 0. This means S n is the equator of S n+1 . Define the point (0, 0, . . . , 1) ∈
S n+1 as the north pole and the point (0, 0, . . . , −1) ∈ S n+1 as their south pole.
Let [ f ] ∈ πm (S n ). Then f : S m → S n is a continuous map. Extend f to a
continuous function f˜ : S n+1 → S n+1 as follows: f˜| S m = f 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. 7.4.
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 f (x). This defines the map f˜ on the northern hemisphere.
For the southern hemisphere it is similarly defined. The extended map f˜ is called
the suspension of f .
We are now in a position to define ‘Suspension homomorphism’ E.
Fig. 7.4 Radially extended

map
t f (t)
x Sm S n f (x)
S m+1 S n+1
Definition 7.10.2 (Suspension homomorphism) The natural homomorphism
E : πn (S n ) → πn+1 (S n+1 ), [ f ] → [ f˜]
is called the suspension homomorphism.
H. Freudenthal proved the following suspension theorem in 1937 which is very

important to study the homotopy theory. In his honor this suspension theorem is
known as ‘The Freudenthal suspension theorem.’
Theorem 7.10.3 (The Freudenthal suspension theorem) The suspension homor-

phism
E : πm (S n ) → πm+1 (S n+1 )
is an isomorphism for m < 2n − 1 and is onto for m ≤ 2n − 1.
Proof See [Freudenthal, 1937]. ❑
An immediate consequence of the Theorem 7.10.3 is the following.

Corollary 7.10.4 (Hurewicz) The homotopy groups πm (S n ) = 0 for 0 < m < n.
Proof For any positive integer p < m, the integer m + p + 1 < 2n. Hence m −
p < 2(n − p) − 1 proves by Freudenthal Suspension Theorem that πm (S n ) ∼ =
πm−1 (S n−1 ) ∼
= ... ∼
= π1 (S n−m+1 ). Since n − m + 1 > 1 for m < n, it follows that
π1 (S n−m+1 ) = 0 and its isomorphic image πm (S n ) is also 0. ❑
Corollary 7.10.5 (Hopf) For every integer n ≥ 1, πn (S n ) ∼

= Z. (This result is
known as ‘Hopf degree theorem’).
Proof It follows from previous argument that π1 (S 1 ) ∼ = π2 (S 2 ). Since π1 (S 1 ) ∼

=
2 ∼
Z, π2 (S ) = Z. If n ≥ 2, then n < 2n − 1 and hence by Frudenthal Suspension
Theorem π2 (S 2 ) ∼
= π3 (S 3 ) ∼
= π4 (S 4 ) ∼
= ... ∼
= πn (S n ). ❑
7.10.2 Table of π i (Sn ) for 1 ≤ I, n ≤ 8
Remark 7.10.6 Table 7.1 displays a small sample of the values of the groups πi (S n )
extracted from the paper (Toda 1962).
7.11 Action of π1 on πn
This section considers an important action of π1 on πn which provides tools to make

the abelian group πn (X, x0 ) a module over the group ring Z[π1 (X, x0 )] for n > 1, and
is used to prove the homological version of Whitehead theorem. The fundamental
7.11 Action of π1 on πn 295
group π1 (x, x0 ) acts on πn (X, x0 ) as a group of automorphisms for n ≥ 1. Let X be

a path- connected space and x 0 , x1 ∈ X . Given a path γ : I → X from x0 = γ(0)
and x 1 = γ(1), define a base point changing homomorphism
βγ : πn (X, x1 ) → πn (X, x 0 ), [ f ] → [γ ◦ f ],
which is an isomorphism with inverse βγ , where γ is the inverse path of γ defined

by γ(s) = γ(1 − s). Hence it follows that if X is a path-connected space, different
choices of base point x0 give isomorphic homotopy groups, written simply as πn (X ).
So without loss of generality we assume that γ is a loop in X based at x 0 . Since
βγη = βγ ◦ βη , there is a homomrphism
ψ : π1 (X, x0 ) → Aut (πn (X, x0 )), [γ] → βγ .
This is called the action of π1 on πn and each element of π1 acts as an automorphism

[ f ] → [γ◦ f ] of πn . In particular, for n = 1, this is the action of π1 on itself by inverse
automorphisms. If n > 1, the action makes the abelian group πn (X, x 0 ) a module over
the group ring Z[π1 (X, x 0 )], whose elements are finite sums ηi γi , where n i ∈ Z
i
and γi ∈ π1 , multiplication being defined by distributivity
and the multiplication
in
π1 . The module structure of πn (X, x 0 ) is given by ( ηi γi ) · α = ηi (γ · α) for
i i
α ∈ πn (X, x0 ). Sometimes we say that πn is a π1 -module instead of Z[π1 (X, x0 )]-
module.
Definition 7.11.1 A topological space with trivial π1 action on πn is called

‘n-simple’ and is called ‘simple’ if it is n-simple for all n.
Definition 7.11.2 A topological space X is said to be abelian if it has trivial action

on all πn , since when n = 1, this is the condition that π1 to be abelian.
Remark 7.11.3 The concept of action of π1 on πn is used to prove the homological

version of Whitehead theorem.
7.12 The nth Cohomotopy Sets and Groups
This section studies the concept of cohomotopy sets of pointed topological spaces
and pairs of spaces, which form groups under suitable situations. There is a dual-
ity between homotopy groups πm (X, A, x 0 ) and cohomotopy groups π m (X, A.x 0 ).
More precisely, given a pair (n, m) of integers such that πm (X, A, x0 ) and πn (X, A)
are abelian groups, there is a homomorphism ψ : πm (X, A, x0 ) ⊗ π n (X, A) →
πm (S n , s0 ). In particular, if m = n, there exists a homomorphism ψ : πn (X, A, x0 ) ⊗
π n (X, A) → Z. K. Borsuk endowed the abelian group structure in 1936 on the set
π n (X, A) under certain conditions on (X, A) (Borsuk 1936).
Let Top ∗ and Top 2 denote the category of pointed topological spaces and the
category of pairs of topological spaces, respectively.
Definition 7.12.1 For any pointed space (X, x0 ) in Top ∗ , the nth-cohomotopy set
π n (X, x0 ) is defined to be the set of all homotopy classes [(X, x0 ), (S n , s0 )] of con-
tinuous maps f : (X, x0 ) → (S n , s0 ) ∈ Top ∗ .
This definition is generalized for (X, A) in Top 2 .
Definition 7.12.2 For any pair of topological spaces (X, A) in Top 2 , the nth-
cohomotopy set π n (X, A) is defined to be the set of all homotopy classes
[(X, A), (S n , s0 )] relative to A of continuous maps f : (X, A) → (S n , s0 ) ∈ Top 2 . In
particular, if A = ∅, π n (X, A) is defined by π n (X ) and called the nth-cohomotopy
set of the topological space X .
Remark 7.12.3 π n (X, A) has a distinguished element namely, the homotopy class
of the constant map c : X → s0 . The element [c] is denoted by the symbol 0 and
called the zero element of π n (X, A).
Definition 7.12.4 π 0 (X, A) is defined to be the set of all open and closed subspaces
of X not intersecting A. The zero element of π 0 (X, A) is the empty subspace of X.
Definition 7.12.5 Every map f : (X, A) → (Y, B) in Top 2 induces a transformation
f ∗ : π n (Y, B) → π n (X, A), [α] → [α ◦ f ], called the transformation induced by f .
Remark 7.12.6 Clearly, f ∗ is well defined and sends the zero element of π n (Y, B)
to the zero element of π n (X, A).
Proposition 7.12.7 π n : Top 2 → Set is a contravariant functor.
Proof Consider the object function: (X, A) → π n (X, A) and every f : (X, A) →
(Y, B) in Top 2 . Then f ∗ satisfies the functorial properties. ❑
Let (X, A) ∈ Top 2 be such that A = ∅. If we identify A to a point ∗, we obtain
the quotient space X/A with base point ∗. If q : (X, A) → (X/A, ∗) is the natural
projection, then q maps X − A homeomorphically onto X/A − {∗}.
Proposition 7.12.8 The induced map q ∗ : π n (X/A, ∗) → π n (X, A) is a bijection.
Proof q ∗ is surjective: Let α ∈ π n (X, A) be represented by a map f : (X, A) →
(S n , s0 ). Define a function

f (q −1 (z)), if z ∈ X/A − {∗}
θ : (X/A, ∗) → (S n , s0 ), z →
s0 , if z = ∗.
Clearly, θ is continuous as f is so. Hence β = [θ] ∈ π n (X/A, ∗). Again θ ◦ q = f

shows that q ∗ ([β]) = α. Hence q ∗ is onto.
q ∗ is injective: Let γ, δ : (X/A, ∗) → (S n , s0 ) be two maps such that q ∗ ([γ]) =
q ([δ]). Then γ ◦ q ∼
∗
= δ ◦ q rel A. Hence there exists a continuous map H : (X ×
7.12 The nth Cohomotopy Sets and Groups 297
I, A× I ) → (S n , s0 ) such that H (x, 0) = (γ ◦q)(x), H (x, 1) = (δ ◦q)(x), ∀ x ∈ X

and H (a, t) ∈ A, ∀ a ∈ A, t ∈ I .
Define a map

H (q −1 (z), t), if z ∈ X/A − {∗}
F : (X/A × I, {∗} × I ) → (S , s0 ), (z, t) →
n
s0 , if z = ∗.
Then F : γ δ rel {∗} shows that [γ] = [δ]. Hence q ∗ is injective.

Consequently, q ∗ is a bijection. ❑
Remark 7.12.9 The Proposition 7.12.8 shows that while computing cohomotopy
sets of a given pair (X, A) ∈ Top 2 , without loss of generality, we may assume that
if A = ∅, the subset A consists of a single point {a} say. Then for n > 1, since
S n is simply connected, the inclusion map j : X → (X, a) induces a one-to-one
transformation j ∗ : π n (X, a) → π n (X ), provided that X satisfies certain homotopy
extension properties (which hold in particular, if X is a paracompact Hausdorff space
and a is any given point of X ).
7.13 Applications
This section gives some interesting applications of higher homotopy groups. Homo-
topy groups play a key role in algebraic topology. For more applications see Exercises
7.14 of this chapter, Chap. 14 and also Chap. 17.
Proposition 7.13.1 Let (X, A) be a pair of path- connected spaces. Then π1 (X,
A, x 0 ) can be identified with the set of cosets a H of the subgroup H of π1 (X, x 0 )
represented by loops in A at x 0 .
Proof π1 (X, A, x0 ) is the set of homotopy classes of paths in X from a varying point
in A. Define a map ψ : π1 (X, x 0 ) → π1 (X, A, x0 ) by considering a loop at x0 as an
element of π1 (X, A, x 0 ). Since A is path- connected, every element of π1 (X, A, x 0 )
is homotopic to a loop at x 0 . Hence ψ is surjective. Again two loops α, β ∈ π1 (X, x 0 )
are homotopic rel A iff [α−1 ∗β] is represented by a loop in A. Hence, we can identify
π1 (X, A, x0 ) with the set of cosets a H . ❑
Proposition 7.13.2 Let (X, A, x 0 ) be a triplet such that A is a strong deformation
retract of X . If i : (A, x0 ) → (X, x0 ) is the inclusion, then i ∗ : πn (A, x0 ) →
πn (X, x0 ) is an isomorphism for all n > 1.
Proof Let A ⊂ X be a retract with retraction r : X → A. Since A is a strong defor-
mation retract of X, i ◦ r 1 X rel A. Then the inclusion i a homotopy equivalence
and hence i ∗ : πn (A, x0 ) → πn (X, x 0 ) is an isomorphism for all n > 1. ❑
Corollary 7.13.3 If (X, A, x0 ) is a triplet such that A is a strong deformation retract
of X . Then πn (X, A, x0 ) = 0 for any integer n > 0.
Proof Since i ∗ : πn (A, x0 ) → πn (A, x0 ) is an isomorphism for n > 0,the corollary

follows from the exact sequence (7.2) of the triplet (X, A, x 0 ). ❑
Definition 7.13.4 A map p : E → B is said to have polyhedra covering homotopy

property (PCHP) if it has the covering homotopy property for every triangular space
X . If p has PCHP, then p is said to be a fibering.
Proposition 7.13.5 Let a map f : (X, A, x 0 ) → (Y, B, y0 ) be given. If f :

X → Y is a fibering and A = f −1 (B), then the induced transformations
f ∗ : πn (X, A, x0 ) → πn (Y, B, y0 ) is an isomorphism for every n > 1.
Proof f ∗ is a monomorphism: Clearly, f ∗ is a homomorphism for n > 1. Let

[g], [h] ∈ πn (X, A, x 0 ) be such that f ∗ ([g]) = f ∗ ([h]). Then their representatives
g, h ∈ πn (X, A, x 0 ) are such that f ◦ g and f ◦ h represent the same element of
πn (Y, B, y0 ). Consequently, there exists a map H : (I n × I, I n−1 × I, J n−1 × I ) →
(Y, B, y0 ) such that H (ω, 0) = ( f ◦ g)(ω) and H (ω, 1) = ( f ◦ h)(ω), ∀ ω ∈ I n .
Let C = (I n × 0) ∪ (J n−1 × I ) ∪ (I n × 1). Then, C is a closed subspace of I n × I .
Define a map
⎧
⎪
⎨g(ω), if ω ∈ I n , t = 0
F : C → X, (ω, t) → x0 , if ω ∈ J n−1 , t ∈ I
⎪
⎩
h(ω), if ω ∈ I n , t = 1.
Then f ◦ F = H |C . Since C is a strong deformation structure of I n × I, F has

an extension F n : I n × I → X such that f ◦ F = H. Again since H maps I n−1 × I
−1
into B and A = f (B), then condition f ◦ F = H shows that H (I n−1 × I ) ⊂ A.
Consequently, the map H : (I × I, J
n n−1
× I, J n−1
× I ) → (X, A, x 0 ) is such that
H(ω, 0) = g(ω) and H (ω, 1) = h(ω) for all ω ∈ I n . This implies that g and h
represents the same element of πn (X, A, x 0 ). In other words, [g] = [h]. This shows
that f ∗ is a monomorphism.
f ∗ is an epimorphism: Let [g] ∈ πn (Y, B, y0 ) be an arbitrary element. Then [g]
is represented by a map g : (I n , I n−1 , J n−1 ) → (Y, B, y0 ). Since J n−1 is a strong
deformation retract of I n , it follows that ∃ a map h : I n → X such that f ◦ h = g
and h(J n−1 ) = x0 . Again since A = f −1 (B), f ◦ h = g, h(I n−1 ) ⊂ A, it gives a
map h : (I n , I n−1 , J n−1 ) → (X, A, x0 ). Clearly, f ∗ ([h]) = [g] shows that f ∗ is an
epimorphism. ❑
Proposition 7.13.6 If X = {x 0 } is a topological space consisting of a single point

x0 , then πn (X, x0 ) = 0 for every n ≥ 0.
Proof If X = {x0 }, then for each n, the map f : I n → X is the only map of I n onto
X , which is a constant map. Hence πn (X, x0 ) = 0. ❑
Proposition 7.13.7 For 0 < m < n, the m-th homotopy group πm (S n ) = 0.

Proof Let [ f ] ∈ πm (S n ) be represented by a map f : (S m , 1) → (S n , 1). If we

represent S m and S n as the boundary complex of the simplexes of dimension m + 1
and n + 1, f has a simplicial approximation cannot map a simplex onto a simplex
of higher dimension, g cannot be onto. Let s0 be a point of S n which is not the range
of g. Then, S n − {s0 } is homeomorphic to Rn and hence S n − {s0 } is contractible
and g is a map such that g(S m ) is contained in a contractible. Consequently, g is a
nullhomotopic, i.e., homotoic to a constant map c. Hence, [ f ] = [g] = [c] = 0. ❑
Theorem 7.13.8 Let (X, A, x 0 ) be a triplet and A be contractible relative to a point

x 0 ∈ A. Then πn (X, A, x 0 ) ∼
= πn (X, x0 ) ⊕ πn−1 (A, x0 ) for any n ≥ 3 and i∗ maps
πn (A, x0 ) into the zero element of πn (X, x 0 ) for any n ≥ 1.
Proof By the given condition, ∃ a homotopy Ht : A → X such that H0 = i : A →

X, H1 (A) = x0 , and Ht (x0 ) = x0 . Hence i ∗ = 0 for any n ≥ 1. For n ≥ 2, use the
exactness of the homotopy sequence (7.2) of (X, A, x 0 ). Then i ∗ = 0 shows that j∗
is a monomorphism and ∂ is an epimorphism. Consequently, πn (X, A, x 0 ) may be
considered as an extension of πn (X, x 0 ) by πn−1 (X, x0 ). Define a homomorphism f ∗ :
πn−1 (A, x0 ) → πn (X, A, x 0 ) for each n ≥ 2. Let [g] ∈ πn−1 (A, x0 ) be represented
by a map g : (I n−1 , ∂ I n−1 ) → (A, x0 ). Define a map
h : (I n , I n−1 , J n−1 ) → (X, A, x 0 ), (t1 , t2 , . . . , tn ) → (Htn ◦ g)(t1 , t2 , . . . , tn−1 ).
Set f ∗ ([g]) = [h]. Since h| I n−1 = g, it follows that ∂ ◦ f ∗ is the identity automorphism
on πn−1 (A, x0 ). This implies that f ∗ is a monomorphism for every n ≥ 2. For n ≥ 3,
the group πn (X, A, x 0 ) is abelian. Hence ∂ ◦ f ∗ = 1d implies that
πn (X, A, x 0 ) = Im f ∗ ⊕ ker ∂
∼
= πn−1 (A, x0 ) ⊕ πn (X, x0 ),
because f ∗ and j∗ are monomorphisms and ker ∂ = Im j∗ . ❑
Theorem 7.13.9 Let F → E → B be a fiber bundle such that the inclusion F → E

is homotopic to a constant map. Show that the long exact sequence of homotopy
groups breaks up into split short exact sequences producing isomorphisms
πn (B) ∼
= πn (E) ⊕ πn−1 (F).
In particular, for the Hopf bundles S 3 → S 7 → S 4 and S 7 → S 15 → S 8 this gives

isomorphisms
πn (S 4 ) ∼
= πn (S 7 ) ⊕ πn−1 (S 3 ); πn (S 8 ) ∼
= πn (S 15 ) ⊕ πn−1 (S 7 ).
Proof The maps i ∗ : π1 (F, x0 ) → πn (E, x0 ) in the long exact sequence (7.3) for the
Serre fibration are induced by the inclusion i : F → E. Hence if this is homotopic
to a constant map, then i ∗ = 0. Thus for all n > 0, we have a short exact sequence
p∗ d∗
0−
→ πn (E, x 0 ) −−−−→ πn (B, b0 ) −−−→ πn−1 (F, x0 ) −
→ 0.
Again since p : E → B has the homotopy lifting property with respect to all disks,
it follows that there is a splitting map πn (B, b0 ) → πn (E, x0 ) such that the above
short exact sequence splits. ❑
Corollary 7.13.10 The groups π7 (S 4 ) and π15 (S 8 ) contain Z summands.
Proof The corollary follows from Theorem 7.13.9. The corollary also follows from
Hopf fiberings of spheres (see section 7.8.2). ❑
Theorem 7.13.11 If S k → S m → S n is a fiber bundle, then k = n − 1 and

m = 2n − 1.
Proof Consider the relations n ≤ m and k ≤ m and k + n = m. If k = m, then

n = 0, and S 0 is not connected. This contradicts that S m → S n is a surjection. So
k < m, and hence S k → S m is homotopic to a constant map. Then it follows that
πi (S n ) ∼
= πi (S m ) ⊕ πi−1 (S k ), ∀ i > 0. This implies k > 0 and hence m > n. In
particular, taking values of i = 1, 2, . . . , n, we have πi (S k ) = 0 if i < n − 1 and
πn−1 (S k ) ∼
= Z. Consequently, k = n − 1. Hence m = 2n − 1. ❑
7.14 Exercises
1. Show that
(i) πn (Rm ) = 0 for every positive integer n and m. Because, Rm is homotopy
equivalent to a point.
(ii) πn (D m ) = 0 for m-disk D m for every positive integers n and m.
2. Let X be a path- connected space. Show that πn (X, x 0 ) ∼ = πn (X, x1 ) for all
x 0 , x1 ∈ X for n ≥ 1.
[Hint. There is a homeomorphism h : (X, x0 ) → (X, x1 ). Hence there exists a
homotopy equivalence between (X, x 0 ) and (X, x1 ).]
3. If X is a contractible space, show that πn (X, x0 ) = 0 for all n ≥ 0.
[Hint: As X is contractible, it is homotopy equivalent to a singleton space. Use
Dimension Theorem 7.2.6.]
4. (a) Let X be a connected covering space of a path- connected space B with
covering projection p : (X, x0 ) → (B, b0 ) such that p(x 0 ) = b0 . Show that
the induced homomorphism p∗ : πn (X, x 0 ) → πn (B, b0 ) is an isomorphism
for any n ≥ 2.
(b) Show that 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 is 0, (b) follows.]
5. Let (X, x 0 ), (Y, y0 ) be two topological spaces in Top ∗ . Show that πn ((X ×
Y ), (x0 , y0 )) ∼
= πn (X, x 0 ) ⊕ πn (Y, y0 ), for n ≥ 1.
7.14 Exercises 301
6. Show that the 4-manifold S 2 ×S 2 is simply connected, but it is not homeomorphic

to S 4 .
[Hint: Use the results : π2 (S 2 × S 2 ) ∼
= Z ⊕ Z and π2 (S 4 ) = 0.]
7. Let (X, x0 ) and (B, b0 ) be pointed topological spaces. Consider the homotopy
exact sequence (7.3) of the fibering p : X → B. Show that
(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, x 0 ) ⊕ 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 7.7.3 and 7.7.6 and exactness property of the
homotopy sequence of a fibering p : X → B.]
8. (a) Show that for any triplet (X, A, x 0 ) 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, x 0 )
lies in the entire of π2 (X, A, x 0 ).
9. Let p : (X, A, x0 ) → (Y, B, b0 ) be a given continuous map. If p : X →
Y is a fibering and A = p−1 (B), show that the induced transformation p∗ :
πn (X, A, x0 ) → πn (Y, B, y0 ) is a bijection for every n > 0.
10. Show that a continuous map f : (D n , 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 : (D n , S n−1 , s0 ) →
(X, A, x 0 ) such that g(D n ) ⊂ A.
11. Let p : X → B be a weak fibration with p(x 0 ) = 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.]
12. 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.
13. Let p : X → B be a covering of X with discrete fiber F. Suppose b0 ∈ B and
x 0 ∈ p −1 (b0 ), show that
(i) p∗ : πn (X, x 0 ) → πn (B, b0 ) is an isomorphism for all n > 1 and a
monomorphism 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, x 0 )) in π1 (B, b0 ).
[Hint: Since F is discrete, πn (F, x0 ) = πn ({x0 }, x0 ) = 0 for all n ≥ 1.]
14. 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 determi-
nant 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.]

15. 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.
16. 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.
17. Let X be a path-connected space with a base point x0 ∈ X and f : S n → X
n+1 map such that f (s0 ) = x0 , where s0 is a base point of S . If
n
be a continuous
Y = X D , 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, x 0 ), where α ∈ π1 (X, x 0 ).
[3] Barratt, M.G., Track groups I, Proc. Lond. Math. Soc. 5(3), 71–106, 1955.
[4] Borel, A., Topology of Lie groups and characteristic classes, Bull. Amer. Math.
Soc. 61(1955), 397–432.
[5] Bott, R., The stable homotopy groups of classical groups, Proc. Nat. Acad. Sci.,
USA, 43(1957), 933–935.
[6] Bott, R., The stable homotopy of the classical groups, Annals of Mathematics.
Ann. of Math, 70(1959), 313-337.

Cambridge, 1983.
[13] Husemöller, D., fiber Bundles, McGraw-Hill, Inc, 1966.
Company, 1984.
[18] Rotman, J.J., An Introduction to Algebraic Topology,Springer-Verlag, New
York, 1988.
[19] Spanier, E., Algebraic Topology, McGraw-Hill Book Company, New York,
1966.
[20] Steenrod. N., The Topology of fiber Bundles, Prentice University Press, Prentice,
1951.
[21] Wallace, A.H., Algebraic Topology, Benjamin, New York, 1980.
References
Adhikari, M.R., Adhikari, A.: Basic Modern Algebra with Applications. Springer, Heidelberg
(2014)
Borel, A.: Topology of lie groups and characteristic classes. Bull. Am. Math. Soc. 61, 397–432
(1955)
Bott, R.: The stable homotopy groups of classical groups. Proc. Nat. Acad. Sci. USA 43, 933–935
(1957)
Borsuk, K.: Sur les groupes des classes de transformations continues, pp. 1400–1403. C.R. Acad.
Sci, Paris (1936)
Bott, R.: The stable homotopy of the classical groups. Ann. Math 70, 313–337 (1959)
New Delhi (2002)
Croom, F.H.: Basic Concepts of Algebraic Topology. Springer, Heidelberg (1978)
Dugundji, J.: Topology. Allyn & Bacon, Boston (1966)
Dieudonné, J.: A History of Algebraic and Differential Topology. Modern Birkhäuser, Boston
(1989). 1900–1960
ton (1952)
Freudenthal, H.: Über die Klassen von Sphärenabbildungen. Compositio Mathematica 5, 299–314
(1937)
(1975)
Hilton, P.J.: An introduction to Homotopy Theory. Cambridge University Press, Cambridge (1983)
Hopf, H.: Üeber die Abbildungen von Sphären niedriger dimension. Fund. Math. 25, 427–440
(1935)
Hurewicz, W.: Beitrage der Topologie der Deformationen, Proc. K.Akad.Wet., Ser.A 38, 112–119,
521–528 (1935)
Husemöller, D.: Fibre Bundles. McGraw-Hill, Inc, New York (1966)
Massey, W.S.: A Basic Course in Algebraic Topology. Springer, Heidelberg (1991)
Munkres, J.R.: Elements of Algebraic Topology. Addison-Wesley Publishing Company, Boston
(1984)
Steenrod, N.: The Topology of Fibre Bundles. Prentice University Press, Prentice (1951)
Toda, H.: Composition methods in homotopy groups of spheres. Ann. of Math. Stud. 49, (1962)
Switzer, R.M.: Algebraic Topology-Homotopy and Homology. Springer, Heidelberg (1975)
Whitehead, G.W.: Elements of Homotopy Theory. Springer, Heidelberg (1978)
Chapter 8
C W -Complexes and Homotopy
This chapter conveys a study of a special class of topological spaces, called

C W -complexes introduced by J.H.C. Whitehead (1904–1960) in 1949 with their
homotopy properties to meet the need for development of algebraic topology. This
class of spaces is broader and has some better categorical properties than simplicial
complexes, but still retains a combinatorial nature that allows for computation (often
with a much smaller complex). Algebraic topologists now feel that the category
of C W -complexes is a good category for homotopy and homology theories. So a
study of C W -complexes should enter in a basic course of algebraic topology and
this study should move up to the theorem that every continuous map between C W -
complexes is homotopic to a cellular map. More precisely, this chapter studies the
basic aspects of C W -complexes and relative C W -complexes with their homotopy
properties and proves Whitehead theorem, Freudenthal suspension theorem (general
form) and cellular approximation theorem with their applications.
The concept of C W -complexes is introduced as a natural generalization of the
concept of polyhedra by relaxing all ‘linearity conditions’ in simplicial complexes,
instead cells are attached by arbitrary continuous maps starting with a discrete set,
whose each point is regarded as a 0-cell. Simplicial structure does not behave well
with respect to the usual topological operations such as products and quotients
of spaces. There is a natural question: what is the good category of topological
spaces in which homotopy theory works well? So one of the problems for systematic
study of algebraic topology is to decide a suitable category of topological spaces.
J.H.C. Whitehead constructed a new category which is now called the category
of C W -complexes and studied it in his two papers (Whitehead 1949a, b). A C W -
complex is a Hausdorff space built up by successive adjunctions of cells of dimensions
1, 2, 3, . . . ; such spaces form an extensive class of topological spaces suitable for
the study of algebraic topology, where a weak homotopy equivalence is necessarily
a homotopy equivalence.

DOI 10.1007/978-81-322-2843-1_8
306 8 C W -Complexes and Homotopy
There are many advantages of C W -complexes over polyhedra: one is that a poly-
hedron can be regarded as a C W -complex with fewer cells than there were simplexes
originally and another advantage is the permissibility of many constructions such as
the product of two polyhedra is a C W -complex in a natural way, since the product
of two simplexes is a cell, but not a simplex, in general. Since all C W -complexes
are paracompact and all open coverings of a paracompact space are numerable, the
results on the homotopy classification of principle G-bundles discussed in Chap. 5
apply to all locally trivial principal G-bundles over a C W -complex. Moreover, C W -
complexes readily lend themselves to study homotopy, homology and cohomology
theories in a relatively convenient way.
One of the main features of C W -complexes is that it is possible to define a
continuous map f : K → X from a C W -complex K into a topological space X step
by step by defining them in succession on the n-skeletons K (n) of K. The construction
of a C W -complex is made by stages by successive attachments of cells. Despite
the fact that every topological space is not a C W -complex, it is sufficient for many
important purposes to consider only C W -complexes (instead of arbitrary topological
spaces) by a theorem of Whitehead which says that given any topological space X,
there exists a C W -complex K and a weak homotopy equivalence f : K → X .
The following terminology and notations for any integer n ≥ 1 are used in this
chapter.
Rn = {(x1 , x2 , · · · , x n ) ∈ Rn : xi ∈ R} (n-dimensional Euclidean space with
norm x).
D n = {x ∈ Rn : ||x|| ≤ 1} (closed n-dimensional disk or ball).
en = {x ∈ Rn : ||x|| < 1} (n-dimensional cell or open n-dimensional disk or ball).
S n−1 = {x ∈ Rn : ||x|| = 1} ((n − 1)-dimensional sphere).
n−1 = {(x1 , x2 , · · · , x n ) ∈ Rn : 0 ≤ xi ≤ 1, xi = 1} (n-dimensional simplex).
For this chapter the books and papers Gray (1975), Hatcher (2002), Maunder
(1970), Rotman (1988), Spanier (1966), Switzer (1975), Whitehead (1978), and the
papers Blakers, A.L. and Massey (1952), Whitehead (1949b), and some others are
referred in the Bibliography.
8.1 Cell-Complexes and C W -Complexes: Introductory

Concepts
This section introduces the concepts of cell-complexes and C W -complexes. A

C W -complex X is a cell complex X satisfying two additional conditions: CW(4)
and CW(5) (i.e., having weak topology and satisfying closure finite property).
C W -complexes form an important class of topological spaces that contains all simpli-
cial complexes. A simplicial complex is built up successively by attaching simplexes
along their boundaries. A simplex and its boundary form a triangulation (D n , S n−1 )
for some n. Thus, a polyhedron is built up successively by attaching simplexes by
maps of their boundaries. On the other hand, a C W -complex is built up successively
8.1 Cell-Complexes and C W -Complexes: Introductory Concepts 307
by attaching a family of 1-cells to a discrete space; attaching a family of 2-cells to

the result; attaching a family of 3-cells to the result and so on (even allowing more
than a finite number of cells). A C W -complex generalizes the concept of polyhedra,
because the cells are attached by arbitrary continuous maps starting with a discrete
set, whose points are regarded as 0-cells.
8.1.1 Cell-Complexes
This subsection introduces the concept of cell-complexes, given by J.H.C. Whitehead

in 1949, which is easier to handle than simplicial complexes at many situations. This
concept generalizes the notion of simplicial complexes.
Definition 8.1.1 An n-cell is a pair (X, A) of topological spaces homeomorphic to

the pair (D n , S n−1 ).
Example 8.1.2 Let Int n−1 = {x = (x1 , x2 , . . . , xn ) ∈ n−1 : 0 < xi < 1}. Then
(n−1 , ∂n−1 ) is an (n − 1)-cell, where ∂n−1 = n−1 − Int n−1 .
Definition 8.1.3 A cell complex X is a Hausdorff space which is the union of disjoint
subspaces eα (α ∈ A) called cells satisfying
(i) to each cell, an integer n ≥ 0 is assigned. This integer is called its dimension. If
the cell eα has dimension n, we use the notation eαn for this cell.
The union of all cells eαk with k ≤ n, denoted by X (n) is called the n-skeleton
of X .
(ii) If eαn is an n-cell, there is a characteristic map ψα : (D n , S n−1 ) → (X, X (n−1) )
such that its restriction ψα | Dn −S n−1 is a homeomorphism from D n − S n−1 onto
eαn .
Remark 8.1.4 Some authors prefer to call a cell complex X as a cell complex or
simply a complex K on X .
Definition 8.1.5 A continuous map f : X → Y between two cell-complexes is said

to be cellular if f (X (n) ) ⊂ Y (n) for all n ≥ 0.
Example 8.1.6 The polyhedron of any finite geometric simplicial complex is a cell
complex. Each open n-simplex is an n-cell, and in this case, the maps ψα are all
homeomorphisms.
Example 8.1.7 The n-sphere S n is a cell complex with two cells e0 , en , where e0 =
{(1, 0, 0, . . . , 0)} and en = S n − e0 . The cell en is called the standard n-cell and it is
thought of as n-sphere minus its ‘east point’ e0 = {(1, 0, 0, . . . , 0)}.
Example 8.1.8 (i) R P n is a cell complex with one cell of dimension k for each
k ≤ n. It is represented symbolically as R P n = e0 ∪ e1 ∪ · · · ∪ en .
(ii) CP n is a cell complex with one cell of dimension 2k for each k ≤ n. It is

represented symbolically as CP n = e0 ∪ e2 ∪ · · · ∪ e2n ;
(ii) HP n is a cell complex with one cell of dimension 4k for each k ≤ n. It is
represented symbolically as HP n = e0 ∪ e4 ∪ · · · ∪ e4n .
Definition 8.1.9 Let X be a cell complex with characteristic ψα and A ⊂ X . Then
A is said to be subcomplex of X if A is a union of cells eα and ēα ⊂ A, where
ψα (D n ) = ēα .
Example 8.1.10 X (n) is a subcomplex of a cell complex X for every n ≥ 0. Because,
ēαn = ψα (D n ).
Example 8.1.11 Let X be a cell complex and K (A) be the intersection of all subcom-
plexes of X containing A. If A ⊂ B, then K (A) ⊂ K (B) and K (A) is a subcomplex
of the cell complex X .
Example 8.1.12 An n-skeleton X (n) is a subcomplex of a cell complex X.
Definition 8.1.13 A pair (X, A) is called a relative cell complex if X is a Hausdorff
space and X − A is a union of disjoint subspaces eαn (α ∈ A) called cells satisfy (i)
and (ii) of Definition 8.1.3.
Remark 8.1.14 A relative cell complex generalizes the concept of a cell complex.
Definition 8.1.15 A polytope is the union of finitely many simplices, with the addi-
tional property that, for any two simplices that have a nonempty intersection, their
intersection is a vertex, edge, or higher dimensional face of the two.
Remark 8.1.16 A polytope is an important concept and this term was generally used
before the creation of C W -complex by J.H.C. Whitehead in 1949.
8.1.2 C W -Complexes
This subsection studies an important class of topological spaces, called

C W -complexes, which describe cell complexes X which are closure finite and have
the weak topology. Such topological spaces constitute a most useful class of spaces
in which homotopy theory works well. C W complexes meet the need of homotopy
theory and develop both homotopy and homology theories. This class of spaces is
broader and has some better categorical properties than simplicial complexes, but still
retains a combinatorial nature that allows for computation (often with a much smaller
complex). Roughly speaking, a C W complex is made of basic building blocks called
cells. An n-dimensional closed cell is the image of an n-dimensional closed ball
under an attaching map. For example, a simplex is a closed cell, and more generally,
a convex polytope is a closed cell.
One of the problems for systematic study of algebraic topology is to decide a suit-
able category of topological spaces. J.H.C. Whitehead introduced in 1949 (Whitehead
1949b) a suitable category, which is now called the category of C W -complexes. For a
long time the term C W -complexes was not in regular use. Many later authors contin-
ued to refer to ‘polyhedra’ which are now called finite C W -complexes. Computing
homotopy and homology groups is in general a difficult problem. One of the diffi-
culties is that given the arbitrary spaces X, Y it is not easy to construct continuous
maps f : X → Y . If we pay our attention to a class of spaces obtained step by step
out of simple building blocks such as simplicial complexes, then we have a better
chance for constructing maps step by step, extending them over the building blocks
one at a time. This motivated Whitehead to define C W -complexes in 1949.
A C W -complex is built up by successive adjunctions of cells of dimensions
1, 2, 3, . . . . The precise definition asserts how the cells may be topologically glued
together. A C W complex is a Hausdorff space X together with a partition of X
into open cells (of varying dimensions) which satisfy some additional conditions
prescribed below.
Definition 8.1.17 A C W -complex is a Hausdorff space X , together with an indexing

set An for each integer n ≥ 0, and maps ψin : D n → X for all n ≥ 0, i ∈ An , are such
that the following conditions are satisfied: If en = {x ∈ R n : d(x, 0) < 1}(n ≥ 1),
then
CW(1) X = ∪ψin (ein ), for all n ≥ 0, i ∈ An (e0 and D 0 are each considered as a
single point);
CW(2) ψin (en ) ∩ ψ mj (em ) = ∅, unless n = m and i = j; and ψi |en is (1-1) for all
n ≥ 0, i ∈ An ;
CW(3) if X (n) = ∪ψim (em ), for all m, 0 ≤ m ≤ n and i ∈ An , then ψi (S n−1 ) ⊂
X (n−1) for each n ≥ 1 and i ∈ An ;
CW(4) A subset Y of X is closed if and only if (ψin )−1 (Y ) is closed in D n , for each
integer n ≥ 0 and i ∈ An ;
CW(5) for each integer n ≥ 0 and i ∈ An , the subspace ψin (D n ) is contained in the
union of a finite number of sets of the form ψ mj (em ).
Definition 8.1.18 The maps ψin are called characteristic maps for the C W -complex
X , and the subspaces ψin (D n ) are called n-cells of X ; X (n) is called the n-skeleton
of X . If n is the smallest integer such that X (n) = X , then X is said to be finite
dimensional with dimension n or simply n-dimensional. Otherwise, i.e., if there
exists no such n, X is said to be infinite dimensional.
Remark 8.1.19 The condition CW(4) says that X has the union topology; frequently
called the weak topology and the condition CW(5) says that X is closure finite in
the sense that each closed cell is covered by a finite union of open cells.
Remark 8.1.20 The original reason for the term ‘C W -complex’: the symbol C
stands for ‘closure-finite’ and the symbol W stands for ‘weak topology’. Hence
‘C W -complex’ stands for ‘closure-finite and weak topology.’
Remark 8.1.21 A C W -complex X is a cell complex X satisfying two additional

conditions CW(4) and CW(5).
Remark 8.1.22 A simplicial complex is a set of simplexes but not a topological space.
On the other hand, a C W -complex is itself a topological space Thus a simplicial
complex K is not a C W -complex but the polyhedron |K | is a C W -complex.
Given a simplicial K , the spaces |K | provide an extensive class of C W -complexes.
Proposition 8.1.23 Let K be a simplicial complex. Then |K | is a C W -complex.
Proof |K | is a subspace of some Euclidean space. Hence |K | is a Hausdorff space.
For each n-simplex sn of K , let s˙n denote the boundary of sn and ψsnn : (D n , S n−1 ) →
(sn , s˙n ) be a relative homeomorphism. If An denotes the set of all n-simplexes of K ,
then the characteristic maps ψsnn make |K | into a C W -complex. This is because, since
the properties CW(3) and CW(5) are obvious. The properties CW(1) and CW(2)
and CW(4) follow from the properties of the simplicial complex K . ❑
Remark 8.1.24 Some authors prefer the following alternative definition of a
C W -complex instead of Definition 8.1.17.
Definition 8.1.25 A C W -complex is an ordered triple (X, E, ψ), where X is a
Hausdorff space, E is a family of cells in X , and ψ = {ψe : e ∈ E} is a family of
maps, such that
(i) X = ∪{e : e ∈ E} (disjoint union);
(ii) for each k-cell e ∈ E, the map ψe : (D k , S k−1 ) → (e ∪ X (k−1) , X (k−1) ) is a
relative homeomorphism;
(iii) if e ∈ E, then its closure ē is contained is a finite union of cells in E;
(iv) X has the weak topology determined by {ē : e ∈ E}.
If (X, E, ψ) is a C W -complex, then X is called a C W -space and or sometimes a
C W -complex. The pair (E, ψ) is called a C W -decomposition of X , and ψe ∈ ψ is
called the characteristic map of e.
Remark 8.1.26 (a) One may consider a C W -complex space X as a generalized
polyhedra and (E, ψ) as a generalized triangulation of X .
(b) 1. Axiom (i) indicates that the cells E partition X .
2. Axiom (ii) indicates that each k-cell e arises from attaching a k-cell to X (k−1)
through the attaching map ψe | S k−1 .
3. Axiom (iii) is called closure-finiteness.
Definition 8.1.27 Let (X, E, ψ) be a C W -complex. It is said to be finite if E is
a finite set. A C W -complex X is regarded as a generalized polyhedron and a pair
(E, ψ) as a generalized triangulation of X .
Definition 8.1.28 A pair (X, E) is a union of disjoint subspaces eα (α ∈ A) called
cells satisfying the conditions (i) and (ii) of Definition 8.1.25.
Proposition 8.1.29 Let X be a C W -complex, and let Y be any space. A function f :
ψin f
X → Y is continuous, iff each composite function f ◦ ψin : D n −−−−→ X −−−→ Y
is continuous, for each n ≥ 0 and i ∈ An .
Proof Clearly, each f ◦ ψin is continuous if f is continuous. Conversely, let V be

a closed subset of Y . Then each subset (ψin )−1 ( f −1 (V )) is closed in D n . Hence
f −1 (A) is closed in X by CW(4). Consequently, f is continuous. ❑
Example 8.1.30 S n is a C W -complex for n ≥ 0. Consider S n as a subspace of Rn+1 .

For each n ≥ 1 define

f : (D n , S n−1 ) → (S n , s0 ), x → (2 (1 − ||x||2 )x, 2||x||2 − 1),
where s0 = (0, 0, . . . , 1) ∈ S n . Let ei denote an i-cell. Then the map f shows that S n
is a C W -complex with E = {e0 , en }. Clearly, for n = 0, S 0 has a C W -decomposition
with two 0-cells, which are {e10 , e20 }.
Example 8.1.31 S n has a C W -decomposition with 2 i-cells in every dimension

0 ≤ i ≤ n. Let E +n be the upper closed hemisphere and E −n be the lower closed
hemisphere of S n . Then S n = E +n ∪ E −n and E +n ∩ E −n = S n−1 (the equator). Hence
there are two n-cells e1n and e2n with e1n = E +n and e2n = E −n . This shows that S n has a
C W -decomposition with two i-cells in every dimension 0 ≤ i ≤ n.
Example 8.1.32 The real number space R has the standard C W -structure with
0-skeletons the integers Z and as 1-cells the intervals {[n, n + 1] : n ∈ Z}.
Example 8.1.33 The space Rn has the standard C W -structure with cubical cells
which are products of the 0-cells and 1-cells from R.
Example 8.1.34 The torus T is a C W -complex. To show this consider T , formed

from the square ABCD by identifying the edges AB, DC, and DA, CB as shown in
Fig. 8.1.
Consider the maps f 0 , f 11 , f 21 , f 12 defined as follows : f 0 : D 0 → T sends D 0 to
the point where the four vertices A, B, C and D are identified; f 11 , f 21 : D 1 → T are
such that f 11 sends D 1 to DC and f 21 sends D 1 to DA, respectively, and f 12 : D 2 → T
is defined by mapping D 2 homeomorphically onto the square ABCD and composing
this map with the identification map onto T . Clearly, characteristic maps make T
into a C W -complex with one 0-cell, two 1-cells, and one 2-cell.
Fig. 8.1 Representation of D C

the torus
A B
8.1.3 Examples of Spaces Which Are Neither

C W -Complexes Nor Homotopy Equivalent
to a C W -Complex
This subsection now presents some examples of topological spaces which are neither
C W -complexes nor homotopy equivalent to a C W -complex. As there exist spaces
which are not Hausdorff, every space is not a C W -complex.
Example 8.1.35 (i) Let sn (n > 1) be an n-simplex. Then ∂sn is regarded as a

‘0-dimensional cell complex’: K 0 , whose cells are the points of ∂sn . This is
closure finite but does not have the weak topology.
(ii) sn (n > 1) regarded as a complex K n = K 0 ∪ en , where en = sn − ∂sn and
K 0 = ∂sn as in (i). This has the weak topology, since s n = K n , but is not
closure finite.
(iii) Every cell complex is not a C W -complex. For example, consider a simplicial
complex K , which has a metric topology but which is not locally finite (e.g.,
a complex covering coordinate axes in Hilbert space). The weak topology in
such a complex K cannot be metricized, see [J.C Whitehead, 1939].
Remark 8.1.36 Every topological space is not homotopy equivalent to a C W -

complex.
Example 8.1.37 Let X be a subspace of R1 consisting of points 0 and 1/n for all
integers n ≥ 1. The path components of X are just the single points (since each point
1/n is both open and closed). The topological space X is not homotopy equivalent to
a C W -complex. If X were homotopy equivalent to a C W -complex Y , then Y would
have an infinite number of path components. Suppose f : X → Y is a homotopy
equivalence, then f (X ) is compact and hence f (X ) is contained in a finite subcom-
plex of Y . This shows that f (X ) is contained in the union of a finite number of path
components, this contradicts the assumption that f is a homotopy equivalence. This
implies that the topological space X is not homotopy equivalent to a C W -complex.
Example 8.1.38 An infinite-dimensional Hilbert space is not a C W complex. Since

it is a Baire space, it cannot be expressed as a countable union of n-skeletons, each
of which being a closed set with empty interior.
8.2 Cellular Spaces
This section conveys the concept of cellular spaces which reconciles the intuitive
notion of a topological space built up by attaching cells with formal definition of a
C W -complex. For this purpose the precise meaning of a ‘space built up by attaching
cells’ is first given.
Definition 8.2.1 A cellular space is a topological space X , with a sequence of sub-

spaces
X 0 ⊂ X 1 ⊂ X 2 ⊂ · · · ⊂ X,
8.2 Cellular Spaces 313

such that X = X n , with the following properties:
n=0
CS(1) X 0 is a discrete space.

CS(2) for each positive integer n, there is an indexing set An , and there exist con-
tinuous maps ψin : S n−1 → X n−1 for each i ∈ An . Moreover, X n is the space
obtained from X n−1 and (disjoint) copies Din of D n (one for each i ∈ An ) by
identifying the points x and ψin (x) for each x ∈ Sin−1 and each i ∈ An .
CS(3) A subset Y of X is closed iff Y ∩ X n is closed in X n , for each n ≥ 0.
Remark 8.2.2 If X is a finite dimensional C W complex, then CS(3) holds automat-
ically, because all sets An are empty for sufficiently large n.
Example 8.2.3 Every C W -complex is a cellular space. Its converse is also true:
every cellular space is a C W -complex (see Ex. 25. of Sect. 8.9)
8.3 Subcomplexes of C W -Complexes
We now introduce the concept of a subcomplex of a C W -complex which creates

interest in many situations. We use the notation of Sect. 8.1.
Definition 8.3.1 Let X be a C W -complex and Y be a subspace of X . Then Y is said
to be a subcomplex of X if, for each integer n ≥ 0, there exists a subset Cn of An
such that
(i) Y = ∪ψin (en ) for all n ≥ 0 and i ∈ Cn ;
(ii) ψin (D n ) ⊂ Y for all n ≥ 0 and i ∈ Cn .
If Y contains only a finite number of cells, Y is called a finite subcomplex.
Remark 8.3.2 (i) Arbitrary unions of subcomplexes of a C W -complex X is a sub-
complex of X .
(ii) Arbitrary intersections of subcomplexes of a C W -complex X is also a subcom-
plex of X .
Example 8.3.3 Let X be a C W -complex. Its every n-skeleton X (n) is a subcomplex
of X .
Proposition 8.3.4 Let X be a C W -complex. For each integer n ≥ 0 and i ∈ An ,
ψin (D n ) is contained in a finite subcomplex of X .
Proof As X is a C W -complex, it follows from condition CW(5) that ψin (D n ) is
contained in the union Y of a finite number of sets of the form ψ mj (em ). Since Y
may not satisfy CW(2), Y may not be a subcomplex of X . On the other hand, if
ψ mj (em ) is a set of Y such that ψ mj (em ) is not contained in Y , then by using CW(3)
and CW(5), we can adjoin a finite number of sets ψkn (eq ) to Y with q < m so as to
include ψ mj (S m−1 ). By going on working down in dimension, a finite number of sets
ψrj (er ) can be added to Y until Y becomes a finite subcomplex of X . ❑
8.4 Relative C W -Complexes
This section introduces the concept of relative C W -complexes which is a general-

ization of the notion of absolute C W -complexes.
Definition 8.4.1 A relative C W -complex (X, A) consists of a topological space X ,
a closed subspace A, and a sequence of closed subspaces (X, A)k for k ≥ 0 such that
(i) (X, A)◦ is obtained from A by adjoining 0-cells;
(ii) For k ≥ 1, (X, A)k is obtained from (X, A)k−1 by adjoining k-cells;
(iii) X = ∪(X, A)k ;
(iv) X has the weak topology with respect {(X, A)k }k (i.e., X has a topology coherent
with {(X, A)k }k ).
In this case, (X, A)k is called the k-skeleton of X relative to A. If X = (X, A)n for
some n, we say that dim(X − A) ≤ n.
Remark 8.4.2 An absolute C W -complex X is a relative C W -complex (X, ∅) and

its k-skeleton is denoted by X (k) .
Definition 8.4.3 A relative C W -structure on a pair (X, A) is a sequence
A = (X, A)−1 ⊂ (X, A)0 ⊂ · · · ⊂ (X, A)n ⊂ (X, A)n−1 ⊂ · · · ⊂ X
that (X, A) is obtained from (X, A) by attaching n-cells, n ≥ 0, X =

n n−1
such
(X, A)n−1
and X has the weak topology. We say dim(X, A) = n if (X, A)n = X
n≥−1
and (X, A)n−1 = X .
Remark 8.4.4 If A = {x 0 } in (X, A), then X is a C W -complex. Conversely, if X is

a C W -complex and A ⊂ X is any subcomplex, then (X, A) a relative C W -complex.
So one may replace the phrase ‘relative C W -complex (X, A)’ by the more restricted
notion of a C W -complex pair (X, A), where X is a C W -complex and A is a sub-
complex’.
Example 8.4.5 If (K , L) is a simplicial pair, then there is a C W -complex pair

(|K |, |L|), with (|K |, |L|)n = |K (n) ∪ L|.
Example 8.4.6 I is a C W -complex, with I (0) = {0, 1} = I˙ and I (n) = I for some
n ≥ 1.
Definition 8.4.7 Let (X, A) and (Y, B) be relative C W -complexes and f : (X, A) →
(Y, B) be a continuous map. Then f is said to be cellular if f ( X̄ (n) ) ⊂ Ȳ (n) , where
X̄ = X (n) ∪ A and Ȳ = Y (n) ∪ B for every integer n.
8.5 Homotopy Properties of C W -Complexes, Whitehead Theorem … 315
8.5 Homotopy Properties of C W -Complexes, Whitehead

Theorem and Cellular Approximation Theorem
This section conveys homotopy properties of C W -complexes, proves Whitehead

theorem and cellular approximation theorem which play a key role in algebraic
topology. Every continuous map between C W -complexes is homotopic to a cellular
map and every two homotopic cellular maps are cellularly homotopic. This result
leads to the C W -approximation theorem. On the other hand, Whitehead’s theorem
states that the continuous maps between C W -complexes that induce isomorphisms
on all homotopy groups are actually homotopy equivalences.
8.5.1 Homotopy Properties of C W -Complexes
This subsection studies homotopy properties of C W -complexes. A C W -complex is

a homotopy-theoretic generalization of the concept of a simplicial complex.
Theorem 8.5.1 Let (X, A) and (Y, B) be relative C W -complexes and f : (X, A) →
(Y, B) be continuous. Then f g rel A, for some cellular map g.
Proof Define homotopies H (n) : X̄ (n) × I → Y inductively such that

(i) H (n) | X (n−1) ×I = H (n−1) ;
(ii) H (n) (x, 0) = f (x);
(iii) H (n) (x, 1) ∈ Ȳ (n) ;
(iv) H (n) (a, t) = f (a), ∀ a ∈ A.
Then πn (Y, Ȳ (n) , ∗) = 0 by Ex.12 of Sect. 8.9 for any choice of ∗. Define H : X ×
I → Y by H | X (n) ×I = H (n) and take g(x) = H (x, 1). By (iii), g(x) ∈ Ȳ (n) and by
(ii) and (iv), it follows that H : f g rel A, where g is a cellular map. ❑
Corollary 8.5.2 Let f, g : (X, A) → (Y, B) be cellular maps and f g rel A.

Then there is a cellular homotopy relative to A between them in the sense that there
(n) (n+1)
exists a continuous map H : (X, A) × I → (Y, B) such that H (X × I ) ⊂ Y .
Proof It is sufficient to show that there is a cellular homotopy H : (X, A) × I →

(n) (n+1)
B such that H (X × I ) ⊂ Y . Clearly, (X × I, X × {0} ∪ A × I ∪ X × 1) is
(n−1)
a relative C W -complex with n-skeleton X × {0} ∪ X × I ∪ X × 1. Applying
Theorem 8.5.1 to the given homotopy the required homotopy is obtained. ❑
Corollary 8.5.3 Every continuous map between C W -complexes is homotopic to a

cellular map and any two homotopic cellular maps between two C W -complexes are
cellularly homotopic.
Proof It follows from Theorem 8.5.1 and Corollary 8.5.2. ❑
Theorem 8.5.4 Let (X, A) be a pair of spaces such that the inclusion map i : A →
X is a weak homotopy equivalence. If K is a C W -complex, with a 0-cell as base
point, then for any choice of base point in A, the induced map i ∗ : [K , A] → [K , X ]
is a bijection.
Proof i ∗ is onto: Let f : K → X be a based continuous map. We show by induction

on the skeletons of K that f can be deformed into A. The map f regarded as a
map K × {0} → X can be extended to a continuous map f : K × I → X such that
f (K × 1) ⊂ A, and if L is a subcomplex of K , which is mapped by f into A, then
f (L × I ) ⊂ A. Let L be such a subcomplex of K and Y (n) = K (n) ∪ L. The map
f is now extended as the constant homotopy to (K × {0}) ∪ (L × I ). If x is any
0-cell of K − L, there exists a path u : I → X such that u(0) = f (x) and u(1) ∈ A.
Then we can continuously extend f to Y (0) × I by setting f (x, t) = u(t), 0 ≤ t ≤ 1.
We now start extension on n. We assume that f has been continuously extended to
a map f : (K × {0}) ∪ (Y (n−1) × I ) → X, such that f (Y (n−1) × 1) ⊂ A. Then for
each cell ψαn (D n ) of K − L, the composite map
ψαn ×1d f
(D n × {0}) ∪ (S n−1 × I ) −−−−−−→ (K × {0}) ∪ (Y (n−1) × I ) −−−→ X,
sends S n−1 × 1 to A. Define a homeomorphism

⎧
⎪
⎪ (x, 0) → (x/2, 0), ∀ x ∈ D n ;
⎪
⎨(x, t) → ( 1 (1 + t)x, 0), ∀ x ∈ S n−1 , t ∈ I ;
g : D n × I → D n × I, 2
⎪
⎪(x, 1) → (x/||x||, 2 − 2||x||), x ∈ D , ||x|| ≥ 2
n 1
⎪
⎩
(2x, 1), x ∈ D n : ||x|| ≤ 12 .
Then, the map f ◦ (ψαn × 1d ) ◦ g −1 : (D n , S n−1 ) → (X, A) represents an element

of πn (X, A), with some base point. But by the exactness of the homotopy sequence, it
follows that πn (X, A) = 0. Consequently, f ◦ (ψαn × 1d ) ◦ g −1 can be continuously
extended to a map of D n × I that sends D n × 1 and S n−1 × I to A. Again apply-
ing the homeomorphism g, the map f ◦ (ψαn × 1d ) can be extended to a continuous
map of D n × I that sends D n × 1 to A. This process defines a continuous exten-
sion f : (K × {0}) ∪ (Y (n) × I ) → X such that f (Y (n) × 1) ⊂ A. Hence it gives
a continuous extension f : K × I → X such that f (K × 1) ⊂ A. This shows that
i ∗ : [K , A] → [K , X ] is onto.
i ∗ is injective: Let f, h : K → A be based continuous maps such that i ◦ f
i ◦ h by a based homotopy H : K × I → X . Since K × I is a C W -complex and
(K × {0}) ∪ (k0 × I ) ∪ (K × 1) is a subcomplex, H can be deformed to a map
F : K × I → A such that F coincides with H on (K × {0}) ∪ (k0 × I ) ∪ (K × 1).
Hence F is a based homotopy between f and h. This implies that i∗ is injective.

Consequently, i ∗ is a bijection. ❑
We now extend the Theorem 8.5.4 with the help of mapping cylinder when f :
A → X be a weak homotopy equivalence.
Corollary 8.5.5 Let f : A → X be a weak homotopy equivalence and K be a C W -

complex. Then f ∗ : [K , A] → [K , X ] is a bijection (where K has a 0-cell as base
point, and A, X have any base points that correspond under f ).
g
Proof The map f : A → X may be considered as the composite A → M f −−−→ X,
where M f is the mapping cylinder, i : A → M f is an inclusion map and g is a
homotopy equivalence by using Theorem 4.7.22 of Chap. 4. Since both f and g
are weak homotopy equivalences, i is also so. Hence i ∗ : [K , A] → [K , M f ] is a
bijection by Theorem 8.5.4. But as g∗ : [K , M f ] → [K , X ] is a bijection, f ∗ is also
so and is such that f ∗ = g∗ ◦ i ∗ . ❑
Remark 8.5.6 This Corollary 8.5.5 leads to Whitehead theorem.
8.5.2 Whitehead Theorem
This subsection answers the question when the concepts of weak homotopy equiv-
alence and homotopy equivalence coincide. Every homotopy equivalence is a weak
homotopy equivalence. Is its converse true? Its answer is found in Whitehead theo-
rem which asserts that a weak homotopy equivalence, for connected C W complexes,
is a homotopy equivalence. So it has become necessary to introduce the concept of
‘weak homotopy equivalence’ at the beginning.
Definition 8.5.7 Let f : (X, x0 ) → (Y, f (x 0 )) be a continuous map in the category

of pointed topological spaces. Then f is called a weak homotopy equivalence if
its induced map f ∗ : π0 (X, x 0 ) → π0 (Y, f (x 0 )) is a (1-1) correspondence and f ∗ :
πm (X, x0 ) → πm (Y, f (x 0 )) is an isomorphism for all m ≥ 1 and all points x 0 ∈
X . The continuous map f is said to be an n-equivalence (for some n ≥ 1) if f ∗ :
πm (X, x0 ) → πm (Y, f (x 0 )) is an isomorphism for 0 < m < n and an epimorphism
for m = n for all points x0 ∈ X .
We are now in a position to present Whitehead Theorem, proved by

J.H.C. Whitehead in his classical paper (Whitehead, Combinatorial Homotopy
I, Bull Amer Math Soc 55(1949), 213–245), where the concept of C W -complex
was first defined.
Theorem 8.5.8 (Whitehead) Let f : K → L be a weak homotopy equivalence of

C W -complexes. Then f is a homotopy equivalence.
Proof By Corollary 8.5.5, the induced function f ∗ : [L , K ] → [L , L] is a bijection.

Hence there exists a continuous map g : L → K such that f ◦ g 1 L . Then, g is also
a weak homotopy equivalence. Similarly, there exists a continuous map f : K → L
such that g ◦ f 1 K . Now,
f ( f ◦ g) ◦ f = f ◦ (g ◦ f ) f
shows that g ◦ f 1 K . Hence f is a homotopy equivalence with g a homotopy

inverse of f . ❑
There is another form of Whitehead theorem given below.
Theorem 8.5.9 (Alternative form of Whitehead theorem) Let K and L be

C W -complexes and f : K → L be a continuous map such that its induced homo-
morphisms
f ∗ : πn (K ) → πn (L)
are isomorphism for all n ≥ 1. Then f is a homotopy equivalence.
Remark 8.5.10 (i) If X and Y are path-connected spaces, it is sufficient for f to

be a weak homotopy equivalence that f ∗ : πm (X, x0 ) → πm (Y, f (x 0 )) is an
isomorphism for all m ≥ 1 and for just one point x 0 ∈ X .
(ii) Every homotopy equivalence is a weak homotopy equivalence. Is its converse
true? Whitehead theorem proves that the converse is also true if X and Y are
both C W -complexes. Hence the concepts of weak homotopy equivalences and
homotopy equivalences coincide for C W -complexes.
(iii) Whitehead theorem does not hold for general topological spaces or even for
all subspaces of Rn . For example, the Warsaw circle W S 1 , which is a compact
subset of Euclidean plane and obtained by closing up a topologist’s sine curve
with an arc, has all homotopy groups zero. On the other hand any continuous
map f : W S 1 → A is not a homotopy equivalence, where A is a one-pointic
space.
(iv) For possible generalizations of Whitehead theorem to more general topological
spaces, one may study ‘Shape theory.’
8.5.3 Cellular Approximation Theorem
This subsection proves a key theorem in algebraic topology: ‘Cellular Approximation

Theorem’ which is an analogue for C W -complexes of the ‘Simplicial Approximation
Theorem 6.5.9 of Chap. 6 for simplicial complexes.
Theorem 8.5.11 (Cellular approximation theorem) Let X and Y be C W -complexes,

and f : X → Y be a continuous map such that f | A is cellular for some subcomplex
A of X (possibly empty). Then there exists a cellular map g : X → Y such that
g| A = f | A and g f rel A.
Proof We use induction on the skeletons X (n) of X and can define a homotopy
H : X × I → Y that starts with f , ends with a cellular map, and is the constant
homotopy on A × I . Let x be a 0-cell of X − A. Then there is a path in Y from
f (x) to a point of Y (0) . We can now define a map H on X (0) × I ∪ A × I . Suppose
that H has been extended to X (n−1) × I , and H (X (n−1) × 1) ⊂ Y (n−1) . Then H
can be extended to each n-cell of X − A, since πn (Y, Y (n) ) = 0. The result gives a
continuous extension such that H (X (n) × 1) ⊂ Y (n) . Then by inductive process, the
required homotopy H : X × I → Y is obtained. ❑
Remark 8.5.12 Whitehead theorem shows that despite every topological space is not
a C W -complex, it is sufficient for many purposes to consider only C W -complexes
instead of arbitrary topological spaces.
8.6 More on Homotopy Properties of C W -Complexes
This section proves an interesting property of C W -complexes which leads to the con-
cept of Eilenberg–MacLane Spaces. Such spaces are discussed in details in Chap. 11.
These spaces establish interlink between homotopy and cohomology theories (see
Chaps. 15 and 17).
Theorem 8.6.1 Let X be a C W -complex and n ≥ 0 be an integer. Then there exists

a C W -complex Y , having X as a subcomplex such that if i : X → Y is the inclusion,
then
(i) i ∗ : πm (X ) → πm (Y ) are isomorphisms for 0 < m < n;
(i) πn (Y ) = 0.
Proof Let G be a set of generators for the group πn (X ). For each α ∈ G, take a
based representative map ψαn : S n → X , which may be assumed to be cellular by
Theorem 8.5.11. Let the space Y be obtained from X by attaching cells eαn+1 by the
maps ψαn , one for each α ∈ G. Then the space Y is a C W -complex and X is a cellular
space. Hence Y is also so, since the maps ψαn send S n into X (n) . Moreover, X is a
subcomplex of Y and i ∗ : πm (X ) = πm (Y (n) ∪ X ) → πm (Y ) is an isomorphism for
0 < m < n, and onto for m = n. But for each α ∈ G, i ∗ (α) ∈ πn (Y ) is represented
by the map i ◦ ψαn : S n → Y and this is homotopic to the constant map, since Y has
an (n + 1)-cell attached by ψαn . Consequently, πn (Y ) = 0. ❑
Remark 8.6.2 Given a C W -complex X and integer n ≥ 0 the process described

above can be iterated to “kill off” the higher homotopy groups πm (X ) for all m ≥ n.
8.7 Blakers–Massey Theorem and a Generalization

of Freudenthal Suspension Theorem
This section presents Freudenthal suspension theorem obtained as a consequence of

Blakers–Massey theorem (see Blakers and Massey 1952) and also a generalization
of Freudenthal suspension theorem. One of the main problems of homotopy theory
is to determine the homotopy groups πr (S n ) of spheres explicitly for r ≥ n. Such
search has discovered many techniques developing algebraic topology. Freudenthal
invented in 1937 the concepts of the suspension X of a pointed space X and the
homotopy suspension map E. He defined a map E
E : πr (X, x0 ) → πr +1 ( X, ∗), n ≥ 0, [ f ] → [1 S 1 ∧ f ], (8.1)
as a natural transformation from the functor πr to the functor πr +1 ◦ which is a

homomorphism for all (X, x0 ) and n ≥ 1, called Freudenthal suspension homomor-
1S1 ∧ f
phism, where S r +1 ≈ S 1 ∧ Sr −−−−−−→ S 1 ∧ X = X . In particular, for X = Sr
and Y = S n , E gives group homomorphisms
E : πr (S r ) → πr +1 (S n+1 ) for r ≥ 1, n ≥ 1 (8.2)
E is a bijection for 1 ≤ r ≤ 2n − 1 and surjection for r = 2n − 1.
Remark 8.7.1 Blakers–Massey theorem is a fundamental result in algebraic topol-

ogy. It is used to prove a general form of Freudenthal suspension theorem.

Remark 8.7.2 Given a triad (X, A, ∗), let X 2 = A en , X 1 = A em . If X = X 1 ∪
α β
X 1 , such that A = X 1 ∩ X 2 and i : (X 1 , A) → (X, X 2 ) is the inclusion, then X 1 −
A ≈ X − X 2 ≈ Rm .
Theorem 8.7.3 (Blakers–Massey theorem) LetX be a C W -complex and A be a sub-

complex of X such that X 2 = A en , X 1 = A em . If X = X 1 ∪ X 1 , A = X 1 ∩ X 2
α β
and i : (X 1 , A) → (X, X 2 ) is the inclusion, then i ∗ : πr (X 1 , A, ∗) → πr (X, X 2 , ∗)
is an isomorphism for r < m + n − 2 and is onto if r = m + n − 2.
Proof See Blakers and Massey 1952 or Gray 1975. ❑
Definition 8.7.4 A pointed topological space (X, x 0 ) is said to be n-connected if

πi (X, x0 ) = 0 for all i ≤ n.
Remark 8.7.5 A 0-connected space means path-connected and 1-connected space

means simply connected.
Theorem 8.7.6 (General Form of Freudenthal Suspension Theorem) If X is an

n-connected C W -complex (n ≥ 1), then the suspension homomorphism
8.7 Blakers–Massey Theorem and a Generalization of Freudenthal … 321
E : πr (X, x 0 ) → πr +1 ( X, ∗)
is an isomorphism for all 1 ≤ r ≤ 2n, and an epimorphism for r = 2n − 1.

Proof It follows from Blakers–Massey theorem. ❑
Corollary 8.7.7 For all n ≥ 1, πn (S n ) is isomorphic to Z.
Proof As S n is (n − 1)-connected, the suspension map : πr (S n ) → πr +1 (S n+1 )
is an isomorphism for r < 2n − 1. For n ≥ 2, n + 1 < 2n and hence, in particular,
: πn (S n ) → πn+1 (S n+1 ) is an isomorphism for n ≥ 2. As π1 (S 1 ) and π2 (S 2 ) are
both isomorphic to Z, hence it follows that πn (S n ) is isomorphic to Z for all n. ❑
Remark 8.7.8 If we take X = S n in Theorem 8.7.6, then the Freudenthal suspension
Theorem 7.10.3 of Chap. 7 follows. In this sense, the Theorem 8.7.6 is said to be a
general form of Freudenthal suspension theorem.
8.8 Applications
This section presents some interesting applications by utilizing the main features of
C W -complexes.

n ∼ 0, m < n
Theorem 8.8.1 πm (S ) =
Z, m = n
Proof Suppose m < n. Let [ f ] ∈ πm (S n ). By cellular approximation theorem we

may assume that f : S m → S n is cellular. S n is a C W -complex with a 0-cell e0 and
an n-cell en under the standard cellular structure. Hence its m-skeleton is just the
base point e0 , because for m < n there is only the 0-cell e0 . This shows that the
map f : S m → S n is the trivial map. If m = n, then it follows from Freudenthal
Suspension Theorem (Chap. 7) that for all n, πn (S n ) ∼
= Z. ❑
Remark 8.8.2 (Eilenberg–MacLane Space) It follows from Theorem 8.8.1 that there
exists a C W -complex X such that πm (X ) = 0 for m = n, and πn (X ) ∼
= Z. Such a
C W -complex is called an Eilenberg–MacLane Space K (Z, n). Eilenberg–MacLane
spaces are studied in the Chap. 11. The importance of such spaces is two-fold:
the spaces develop homotopy theory as well as the cohomology theory of C W -
complexes.
Theorem 8.8.3 Let X be a C W -complex. Then
(i) every path component of X is a C W -complex, hence it is closed;
(ii) the path components of X are closed and open;
(iii) path components of X are the components of X ;
(iv) X is connected iff X is path-connected.
Proof (i) As X is a disjoint union of cells, each of which is path-connected, it

follows that each path component Y of X is a union of cells. If e is an n-cell with
e ⊂ Y , ē = ψe (D n ) is also path- connected and hence ē ⊂ Y . Consequently, Y
is a C W -complex and hence Y is closed.
(ii) Let Y be a path component of X and Z be the union of other path components.
Since Z is a union of C W -subcomplexes, it is a C W -subcomplex and it is
closed. Again since Z is the complement of Y in X , it follows that Y is open.
(iii) Let Y be a path component of X and Z be the complement of X containing Y .
Since Y is closed and open, it is connected. Hence Y = Z .
(iv) follows from (iii).
❑
Theorem 8.8.4 Let (X, E) be a C W -complex.

(i) If e ∈ E is an n-cell (n > 0) with characteristic map ψe , then ē = Im ψe =
ψe (D n );
(ii) If E is a finite subset of E, then |E | is a C W -subcomplex iff |E | is closed.
(iii) If e ∈ E, then its closure ē is contained in a finite C W -subcomplex;
(iv) Every compact subset A of X lies in a finite C W -subcomplex and hence a
C W -space X is compact iff (X, E) is a finite C W -complex for every C W -
decomposition E;
(v) A subset A of X is closed iff A ∩ Y is closed in Y for every finite C W -subcomplex
Y in X ;
(vi) If E (possibly infinite) is a subset of E, then |E | is a C W -subcomplex iff |E |
is closed.
Proof (i) Since ψe is continuous, it follows that
ψe (D n ) = ψe (D n − S n−1 ) ⊂ ψe (D n − S n−1 ) = ē =⇒ ψe (D n ) ⊂ ē.
For the reverse inclusion, using the compactness of D n , it follows that ψe (D n )

is compact. Again as the space X is Hausdorff, ψe (D n ) is a closed subset of X
containing e = ψe (D n − S n−1 ) and hence ē ⊂ ψe (D n ).
(ii) Let E be a finite subset of E. Assume that |E | is closed and e ∈ E . Then
e ⊂ |E | and ē ⊂ |E |. Consequently, |E | is a C W -subcomplex. Conversely,
let |E | be a C W -subcomplex. Then ē ⊂ |E | for every e ∈ E . Consequently,
|E | = ∪{e : e ∈ E } = ∪{ē : e ∈ E } is a finite union of closed set, and hence
|E | is closed.
(iii) Let dimension of e be n. Then for n = 0, the proof obvious. We now use induc-
tion on n. If n > 0, then (i) shows that ē − e = ψe (D n ) − e ⊂ (e ∪ X (n−1) ) −
e ⊂ X (n−1) . Now by axiom (iii) of Definition 8.1.25, ē meets only finitely many
cells other than e; e1 , e2 , . . . , er (say). Then dim(ei ) ≤ n − 1 for all i. By induc-
tion, there is a finite C W -subcomplex X i containing ēi , for i = 1, . . . , r and
each X i is closed by (ii). But ē ⊂ e ∪ X 1 ∪ · · · ∪ X r . Hence it is a union of
finitely many cells and hence it is closed and so it is a finite C W -complex.
(iv) Suppose A ∩ e = ∅ for every e ∈ E. Choose a point ae ∈ A ∩ e. Let Y be the

set of all such points ae . Then for each e ∈ E, by (iii) there exists a finite C W -
complex X e containing ē. Hence Y ∩ ē ⊂ Y ∩ X e is a finite set and hence is
closed in ē. Again since X has the weak topology, Y is closed in X . Again the
same reasoning implies that every subset of Y is closed in X and hence Y is dis-
crete. Since Y is compact, as it is a closed subset of A. Consequently, Y is finite
and hence A meets only finitely many e ∈ E : e1 , . . . , er (say). Then by using
(iii), there are finite C W -subcomplexes X i with ēi ⊂ X i , for i = 1, 2, . . . , r .
Hence it follows that A is contained in the finite C W -subcomplex ∪ X i .
(v) Let A be closed in X . Then A ∩ Y is closed in Y . Conversely, for each e ∈ E, let
X e be a finite subcomplex containing ē. Then by hypothesis, A ∩ X e is closed
in X e . Hence A ∩ ē = (A ∩ X e ) ∩ ē is closed in X e and is also closed in the
smaller set ē. Consequently, A is closed in X , since X has the weak topology
determined by all ē.
(vi) Let |E | be closed and e ∈ E . Then ē ⊂ |E | and |E | is a C W -subcomplex.
Conversely, let |E | = Y be a C W -subcomplex. Then by (v) it is sufficient to
prove that Y ∩ A is closed in A for every finite C W -subcomplex A of X . Clearly,
Y ∩ A is a finite union of cells: Y ∩ A = e1 ∪ e2 ∪ · · · ∪ er (say). As Y ∩ A is
a C W -subcomplex, ēi ⊂ Y ∩ Y for all i. Hence Y ∩ A = ē1 ∪ ē2 ∪ · · · ∪ ēr .
This shows that Y ∩ A is closed in A (also in X ). ❑
8.9 Exercises
1. Let X be a C W -complex. Show that

(i) the n-skeleton X n ⊂ X is closed for every n;
(ii) X is locally path-connected;
(iii) if X connected, then X is path-connected.
(iv) if A is a C W -subcomplex of X, then the inclusion map i : A ⊂ X is a
cofibration.
2. Let X be a C W -complex. Show that the following statements are equivalent:
(i) X is path-connected;
(ii) X is connected;
(iii) the 1-skeleton X is connected;
(iv) the 1-skeleton X is path-connected.
[Use the fact that every C W -complex is locally path-connected.]
3. Let X be a C W -complex and f : S n → X n ⊂ X . Show that X en+1 is a
f
C W -complex.
and ein be an n-cell. Show that A = X n − ein is a
4. Let X be a C W -complex
skeleton of X and X = A
n
en for some f : S n−1 → X n−1 .
f
5. Let X be a C W -complex. Show that arbitrary unions and intersection of sub-

complexes of X are again subcomplexes of X .
6. If X and Y are finite C W -complexes, show that the product space X × Y is also
a C W -complex.
[Hint. Let {ei } and {ek } be cellular decompositions of X and Y , respectively. Then
the family {ei × ek } is a cellular decomposition of X × Y . Again, if f and g are
characteristic maps of ei and ek , respectively, then f × g is also a characteristic
map ei × ek .]
7. If X is a C W -complex and A is a subcomplex of X , show that the quotient space
X/A is also a C W -complex.
8. Let X be a C W -complex and A ⊂ X be compact. Show that
(i) A ⊂ X (n) for some n;
(ii) A ⊂ Y for a subcomplex of X , such that Y has only a finite number of cells.
9. Let X be a C W -complex. Prove that
(i) X is a T1 -space;
(ii) X is a normal space.
10. Let (X, A) be a relative C W -complex. Show that it has the absolute homotopy
extension property.
11. Let (X, A) be a relative C W -complex and X̄ = X (n) ∪ A. Let H : X̄ (n−1) × I ∪
X̄ (n) × 0 → Y . If πn (Y, B, ∗) = 0 for any choice of ∗ and H (x, 1) ∈ B. Then
show that, H has a continuous extension π̃ : X̄ (n) × I → Y such that π̃(x, 1)∈B.
12. Let (X, A) be a relative C W -complex with cells only in X − A with dimensions
≥ n. Show that πk (X, A, ∗) = 0 for k < n.
13. Let X be an n-dimensional C W -complex, en be an n-cell in X and p be a point
in e. Prove that X − e is a strong deformation retract of X − { p}.
[Hint. For n = 0, it is trivial. For n > 0, show that there is a retraction r :
X − { p} → X − e and a homotopy H : (X − { p}) × I → X − { p} such that
H (x, 0) = x, H (x, 1 = r (x))].
14. Let (X, E) be a C W -complex, for some fixed n > 0, let E be a family of n-cells
in E. Show that
(i) X = |E | ∪ X (n−1) is closed in X ;
(ii) every n-skeleton X (n) is closed in X for n > 0;
(iii) every n-cell e is open in X (n) ;
(iv) X (n) − X (n−1) is an open subset of X (n) ;
15. Let X be a Hausdorff space. Show that there exists a C W -complex Y and a
weak homotopy equivalence f : Y → X , where Y is uniquely determined upto
16. (Serre) Let p : X → B be a Serre fibration (i.e., p has the homotopy lifting
property with respect to every cube I n , n ≥ 0, where I 0 is a singleton by def-
inition). Show that it has the homotopy lifting property with respect to every
C W -complex K .
8.9 Exercises 325
Fig. 8.2 HLP with respect fñ

/
K (n) n6 X
to C W -complex K F̃n n n n
n p
n n F
K (n)
×I
n
/B
[Hint. As every C W -complex has the weak topology determined by its skeletons,
it is sufficient to show that there exists a map F̃n , for every n ≥ 0, making the
diagram in Fig. 8.2 commutative, where f˜ : K → X and F : K × I → B are
given and fñ and Fn are appropriate restrictions. Then use induction on n.]
17. A space X is called compactly generated if X is a Hausdorff space and it has the
weak topology determined by its compact subsets (see Appendix B). Show that
every C W -complex is compactly generated.
18. If (X, E) is a C W -complex, show that X (0) is a discrete closed subset of X .
19. Show that Klein bottle has a decomposition of the form {e0 , e11 , e21 , e2 }, i.e., with
one 0-cell, two 1-cells, and one 2-cell.
20. Define the dimension of a C W -complex (X, E) to be
dim X = sup{dim(e) : e ∈ E}.
Show that dim X is independent of C W -decomposition of X .

21. Show that a C W -complex is connected iff its 1-skeleton X (1) is connected.
22. Let A be a subcomplex of the C W -complex X . If πn (X, A) = 0 for all n, show
that A is a string deformation retract of X .
23. If X has the homotopy type of a C W -complex and is a Lindelöf space, show
that X has the homotopy type of a countable C W -complex.
24. Show that any continuous map f : K → L between C W -complexes is homo-
topic to a cellular map and any two cellular maps g1 , g2 that are homotopic are
also homotopic by a cellular homotopy.
25. Show that every C W -complex is a cellular space and every cellular space is a
C W -complex.
26. Let X and Y be C W complexes. Show that the function space F (X, Y ) of all
continuous maps from X to Y (with the compact-open topology) is not a C W
complex in general, but if X is locally finite, then F (X, Y ) is homotopy equivalent
to a C W complex.
[Hint: Milnor, 1959.]
1. Adams, J.F., Algebraic Topology: A student’s Guide, Cambridge University

2. Aguilar, Gitler, S., Prieto, C., Algebraic Topology from a Homotopical View
3. Arkowitz, Martin, Introduction to Homotopy Theory, Springer, New York, 2011.

4. Dieudonné, J., A History of Algebraic and Differential Topology, 1900–1960,
5. Dold,A., Lectures on Algebraic Topology, Springer-Verlag, New York, 1972.
6. Dugundji, J., Topology, Allyn & Bacon, Newtown, MA, 1966.
7. Dyer, E., Cohomology Theories, Benjamin, New York, 1969.
8. Eilenberg, S., and Steenrod, N., Foundations of Algebraic Topology, Princeton
9. Hilton, P.J., An introduction to Homotopy Theory, Cambridge University Press,
Cambridge, 1983.
10. Hilton, P. J. and Wyle, S. Homology Theory, Cambridge University Press, Cam-
bridge, 1960.
11. Husemöller, D., Fibre Bundles, Springer, New York, 1975.
12. Hu, S.T., Homotopy Theory, Academic Press, New York, 1959.
13. Hu, S.T., Homology Theory, Holden Day, Oakland CH, 1966.
14. Mayer, J. Algebraic Topology, Prentice-Hall, New Jersy, 1972.
15. Massey, W.S., A Basic Course in Algebraic Topology, Springer-Verlag, New
16. Milnor, J., On Spaces having the Homotopy Type a C W -complex, Trans. Amer.
Maths. Soc. 90 (1959), 272–280.
17. Munkres, J.R., Elements of Algebraic Topology, Addition-Wesley-Publishing
Company, 1984.
18. Singer, I.M., and Thrope, J.A., Lecture Notes on Elementary Topology and Geom-
etry, Springer-Verlag, New York, 1967.
19. Steenrod. N., The Topology of Fibre Bundles, Prentice University Press, Prentice,
1955.
20. Steenrod, N., A Convenient Category of Topological Spaces, Mich. Math J 14
(1967), 133–152.
21. Whitehead, J.H,C. A certain exact sequence, Ann. of Math. 52 (1950), 51–110.
References
(1972)
New York (2002)
Arkowitz, Martin: Introduction to Homotopy Theory. Springer, New York (2011)
Blakers, A.L., Massey, W.S.: The homotopy groups of a triad II. Ann. Math. 55(2), 192–201 (1952)
Boston (1989)
Dold, A.: Lectures on Algebraic Topology. Springer, New York (1972)
Dyer, E.: Cohomology Theories. Benjamin, New York (1969)
ton (1952)
References 327
Gray, B.: Homotopy Theory. An Introduction to Algebraic Topology. Academic Press, New York
(1975)
Hilton, P.J., Wyue, S.: Homology Theory. Cambridge University Press, Cambridge (1960)
Hu, S.T.: Homology Theory. Holden Day, Oakland CH (1966)
Milnor, J.: On spaces having the homotopy type a CW-compex. Trans. Am. Math. Soc. 90, 272–280
(1959)
Munkres, J.R.: Elements of Algebraic Topology. Addition-Wesley-Publishing Company, Menlo
Park (1984)
Singer, I.M., Thrope, J.A.: Lecture Notes on Elementary Topology and Geometry. Springer, New
York (1967)
Spanier, E.: Algebraic Topology. McGraw-Hill, New York (1966)
Steenrod. N.: The Topology of Fibre Bundles. Prentice University Press, Prentice (1955)
Steenrod, N.: A convenient category of topological spaces. Mich. Math. J. 14, 133–152 (1967)
Whitehead, J.H.C.: Combinatorial homotopy. I. Bull. Am. Math. Soc. 55, 213–245 (1949a)
Whitehead, J.H.C.: Combinatorial homotopy. II. Bull. Am. Math. Soc. 55, 453–496 (1949b)
Whitehead, J.H.C.: A certain exact sequence. Ann. Math. 52(2), 51–110 (1950)
Chapter 9
Products in Homotopy Theory
This chapter continues to study homotopy theory through different products defined
between homotopy groups such as the Whitehead product introduced by
J.H.C. Whitehead in 1941, the Samelson product introduced by H. Samelson in
1953 and the mixed product introduced by McCarty in 1964. Moreover, this chapter
finds a generalization of Whitehead product and a relation between Whitehead and
Samelson products. These products are used to solve several problems in algebraic
topology. Computing the homotopy groups of even simple spaces is one of the basic
problems in homotopy theory. The problem of computing the homotopy groups of
n-sphere is not completely solved. In most cases, it is not known whether the homo-
topy groups are trivial or not. Different products are used to solve such problems.
For example, Whitehead product provides methods for computing nonzero elements
of homotopy groups of spheres. Throughout this chapter we consider topological
spaces with base points. The base point is denoted by ∗ (unless otherwise stated),
and not often explicitly mentioned.
For this chapter the books and papers Gray (1975), Hatcher (2002), Hu (1959),
James (1971), Maunder (1980), Spanier (1966), Whitehead (1941, 1944) and some
others are referred in Bibliography.
9.1 Whitehead Product Between Homotopy Groups

of C W -Complexes
This section studies Whitehead product defined by J.H.C. Whitehead (1904–1960)

in 1941 (Whitehead 1941) between two homotopy groups to study homotopy groups
of pointed C W -complexes X . This product associates with each pair of elements
α ∈ π p (X, x0 ) and β ∈ πq (X, x0 ) an element denoted [α, β] ∈ π p+q−1 (X ), called
Whitehead product in his honor. This product provides a technique at least in some

DOI 10.1007/978-81-322-2843-1_9
330 9 Products in Homotopy Theory
cases for constructing nonzero elements of π p+q−1 (X ). He also defined generalized

products involving the rotation groups. The concept of Whitehead product is utilized
in this chapter as well as in Chap. 17 to solve several problems in algebraic topology.
Definition 9.1.1 Let X be a C W -complex and α ∈ π p+1 (X ), β ∈ πq+1 (X ). Then

α and β can be represented by the maps
•
f : (E 1 , E 1 ) → (X, ∗)
•
g : (E 2 , E 2 ) → (X, ∗)
where E 1 and E 2 are oriented cells of dimensions p + 1 and q + 1, respectively, with

• •
boundaries E 1 and E 2 . Hence E 1 × E 2 is a cell, oriented by the product of the given
orientations of E 1 and E 2 ; the base point of E 1 × E 2 is the point (∗, ∗). Its boundary
• •
S = (E 1 × E 2 )• = E 1 × E 2 ∪ E 1 × E 2 is an oriented ( p + q + 1)- sphere and the
map h : (S, ∗) → (X, ∗) defined by
•
f (x), if x ∈ E 1 and y ∈ E 2 ,
h(x, y) = •
g(y), if x ∈ E 1 and y ∈ E 2 .
represents an element [α, β] ∈ π p+q+1 (X ), called Whitehead product of α and β.

This product depends only on the homotopy classes α, β of f, g, respectively.
Remark 9.1.2 The binary operation
π p+1 (X, x 0 ) × πq+1 (X, x 0 ) → π p+q+1 (X, x0 ), (α, β) → [α, β]
is natural in the sense that if
ψ : X → Y is continuous, and α ∈ π p+1 (X, x 0 ), β ∈ πq+1 (X, x 0 ), then
ψ∗ ([α, β]) = [ψ∗ (α), ψ∗ (β)] ∈ π p+q+1 (Y ).
If in particulars, p = 0 = q, then [α, β] = αβα−1 β −1 ∈ π1 (X, x 0 ) is the commu-

tator of α and β (see Proposition 9.1.12). This justifies the notation ‘[ , ]’ of the
Whitehead product.
Remark 9.1.3 α and β can also be represented by maps
f : (S p+1 , ∗) → (X, x0 )
g : (S q+1 , ∗) → (X, x0 )
where S p+1 and S q+1 are oriented ( p + 1)-sphere and (q + 1)-sphere, respectively.
9.1 Whitehead Product Between Homotopy Groups of C W -Complexes 331
Definition 9.1.4 A continuous map h : S p × S q → X is said to have type (α, β) if

h|S p ×{∗} represents α ∈ π p (X ) and h|{∗} × S q represents β ∈ πq (X, x 0 ).
We can characterize maps of types (α, β) with the help of Whitehead products.
Theorem 9.1.5 Let α ∈ π p (X, x0 ), β ∈ πq (X, x0 ). Then there exists a continuous

map S p × S q → X of type (α, β) if and only if the Whitehead product [α, β] = 0
Proof Let f : (S p , ∗) → (X, ∗) and g : (S q , ∗) → (X, ∗) be representatives of α

and β, respectively. Clearly, (S p × S q , S p ∨ S q ) is a relative C W -complex with just
one cell; a characteristic map for this cell is
ω p,q = ω p × ωq : (E p × E q , (E p × E q )• ) → (S p × S q , S p ∨ S q ),
the attaching map for this cell is a representative of the Whitehead product [i 1 , i 2 ]
of the homotopy classes of the inclusion maps i 1 : S p → S p ∨ S q , and i 2 : S q →
S p ∨ Sq .
Let k = ( f, g) : S p ∨ S q → X be the map determined by f and g. Then there
exists a continuous map S p × S q → X of type (α, β) if and only if the map k can be
extended over S p × S q . Since ω p,q is a relative homeomorphism, it is so iff the map
ψ = k ◦ ω p,q |(E p × E q )• can be extended over E p × E q , i.e., iff ψ is nullhomotopic.
But the homotopy class of ψ is
k∗ [i1 , i 2 ] = [k ◦ i 1 , k ◦ i 2 ] = [α, β].
This proves the theorem. ❑
Corollary 9.1.6 Let X be an H -space with continuous multiplication μ. Then

[α, β] = 0 for any α ∈ π p (X, ∗) and β ∈ πq (X, ∗).
Proof Let f, g be any representatives of α, β, respectively. Consider the map
h = μ ◦ ( f × g) = (S p × S q , (∗, ∗)) → (X, ∗).
Then h has type (α, β) for all representative f, g of α, β, respectively. Consequently

the corollary follows from Theorem 9.1.5.
Theorem 9.1.7 Let α ∈ πm (X, x 0 ), β ∈ πn (X, x 0 ) be such that their Whitehead

product [α, β] = 0. If ξ ∈ π p (S m ), η ∈ πq (S n ), then the Whitehead product [α ◦
ξ, β ◦ η] = 0.
Proof Since [α, β] = 0, by Theorem 9.1.5, there is a map h : (S m × S n , (∗, ∗)) →

(X, ∗) of type (α, β).
Let
f : (S p , ∗) → (S m , ∗)
g : (S q , ∗) → (S n , ∗)
be the representatives of ξ and η, respectively. Then the map h ◦ ( f × g) : (S p ×

S q , (∗, ∗)) → (X, ∗) defined by (h ◦ ( f × g))(x, y) = h( f (x), g(y)) is of type (α ◦
ξ, β ◦ η). Hence it follows by Theorem 9.1.5 that the Whitehead product [α ◦ ξ, β ◦
η] = 0.
Definition 9.1.8 A bunch of spheres is a topological space which is homeomorphic

to a union of spheres with a single point in common.
Theorem 9.1.9 If X has the trivial Whitehead product and X has the homotopy
type of a bunch of spheres, then X is homotopy commutative.
Proof Let Y denote the appropriate union of spheres and g : Y → X be a

homotopy equivalence. Let h : Y ∨ Y → X be the map d ◦ g on each factor, where
d : X → X is defined by d(t, λ) = λ(t), ∀ λ ∈ X . Since all Whitehead prod-
ucts vanish in X , we can extend h|S p ∨ S q to S p × S q for each pair of spheres S p ,
S q contained in Y . Hence we can extend h to Y × Y . Using a homotopy inverse to
g, we obtain the required map f : X × X → X .
An equivalent definition of Whitehead product is given in Definition 9.1.10 which
is sometimes convenient for use.
Definition 9.1.10 Let X be a pointed topological space with base point x0 ∈ X ,

m ≥ 1 and n ≥ 1 be given integers. Let α ∈ πm (X, x 0 ) and β ∈ πn (X, x 0 ) be two
given elements. We construct an element [α, β] ∈ πm+n−1 (X, x 0 ), called Whitehead
product of α and β as follows:
Let α and β be represented by the continuous maps f : (I m , ∂ I m ) → (X, x 0 )
and g : (I n , ∂ I n ) → (X, x0 ) respectively. Clearly, I m+n ≈ I m × I n , ∂ I m+n =
(I m+n )• = (I m × ∂ I n ) ∪ (∂ I m × I n ) and the intersection of these two sets is ∂ I m ×
∂ I n.
Define a function

f (s), s ∈ I m , t ∈ ∂ I n
h : ∂I m+n
→ X, (s, t) → (9.1)
g(t), t ∈ I n , s ∈ ∂ I m
Then h is well defined and continuous, since f (s) = g(t) = x 0 ∀ (s, t) ∈ ∂ I m ×

∂I .
n
As the point 0 = (0, 0, . . . , 0) of ∂ I m+n is in ∂ I m × ∂ I n , it follows that h(0) = x0 .

Again since ∂ I m+n ≈ S m+n−1 , h represents an element γ ∈ πm+n−1 (X, x0 ), which
is the homotopy class of h. As γ depends only on the elements α and β define [α, β]
by setting [α, β] = γ.
Remark 9.1.11 [α, β] = γ depends only on the elements α, β. Hence [α, β] is well
defined.
Proposition 9.1.12 If α ∈ π1 (X, x 0 ) and β ∈ π1 (X, x 0 ), then [α, β] = αβα−1 β −1

is the commutator of π1 (X, x 0 ).
9.1 Whitehead Product Between Homotopy Groups of C W -Complexes 333
Fig. 9.1 Boundary of a 0×R

square
f
g g
0 R×0
f
Proof Suppose m = 1 = n. Then (I × I )• is the boundary of the unit square I × I

in the plane R2 , with clockwise orientation, and with the origin as base point as
shown in Fig. 9.1.
The maps f, g are loops representing α, β, respectively. Hence it follows that
[α, β] = αβα−1 β −1 ∈ π1 (X, x 0 ). ❑
Remark 9.1.13 The notation [α, β] for the Whitehead product is consistent with our
standard notation for the commutator of two elements in a group.
Remark 9.1.14 Given base point preserving maps f : S m → X and g : S n → X, let

[ f, g] : S m+n−1 → X be the composite
f ∨g
S m+n−1 −
→ S m ∨ S n −−−−−→ X,
where the first map is the attaching map of the (m + n)-cell of S m × S n with its
usual C W -structure. Since homotopies of f or g give rise to homotopies [ f, g], we
have a well-defined product πm (X, x0 ) × πn (X, x 0 ) → πm+n−1 (X, x 0 ). The notation
[ f, g] is used since for m = n = 1, this is just the commutator product in π1 (X, x 0 ).
Clearly, for m = 1 and n > 1, [ f, g] is the difference between g and its image under
the π1 -action of f .
9.2 Whitehead Products Between Homotopy

Groups of H-Spaces
This section studies Whitehead products between the homotopy groups of H -spaces
(Hopf’s spaces) and topological groups. Let X be a given H -group and x0 ∈ X be a
homotopy unit of X . Then the group operation in πn (X, x0 ) is closely related to the
multiplication in X .
Definition 9.2.1 Let X be a given H -group and x0 ∈ X be a homotopy unit of X . For

n > 0, let α, β ∈ πn (X, x0 ) be represented by the maps f, g : (I n , ∂ I n ) → (X, x 0 ),
respectively. Define h : (I n , ∂ I n ) → (X, x 0 ), t → f (t)·g(t), t ∈ I n (the right hand
product is the usual multiplication in X ). Then h represents the element α + β ∈

πn (X, x0 ). In particular, if X is a topological group and x0 is the neutral element of
X , then define
k : (I n , ∂ I n ) → (X, x0 ), t → f (t)·[g(t)]−1 , t ∈ I n ,
where right hand multiplication is the usual multiplication in the topological group
X . Then k represents the element α − β ∈ πn (X, x 0 ).
Theorem 9.2.2 Let X be a topological group. Then for every pair of elements α ∈
πm (X, x0 ) and β ∈ πn (X, x0 ), their Whitehead product [α, β] = 0.
Proof Let X be topological group and α ∈ πm (X, x 0 ) and β ∈ πn (X, x0 ) be rep-

resented by f : (I m , ∂ I m ) → (X, x 0 ), and g : (I n , ∂ I n ) → (X, x0 ), respectively.
Define
h : I m+n → X, (s, t) → f (s)·g(t), s ∈ I m , t ∈ I n ,
where the right hand product is the usual multiplication in X . Then h|∂ I m+n represents
the Whitehead product [α, β]. This shows that [α, β] = 0. ❑
Theorem 9.2.3 Let X be an H -space with multiplication μ. Then for every pair of
elements α ∈ πm (X, x0 ) and β ∈ πn (X, x 0 ), their Whitehead product [α, β] = 0.
Proof ‘Let
f : (S m , ∗) → (X, x0 )
g : (S n , x0 ) → (S n , x0 )
be the representatives of α and β, respectively. Consider the map
h = μ ◦ ( f × g) : (S m × S n , (∗, ∗)) → (X, x0 ).
Then h is of type (α, β) for any representative f, g of α, β, respectively. This con-

cludes by Theorem 9.1.5 that the Whitehead product [α, β] = 0 ❑
An important characterization of a continuous map h : S m × S n → X to be of

type (α, β) with the help of Whitehead product.
Theorem 9.2.4 Let X be an H -group. If α ∈ πm (X, x 0 ), β ∈ πn (X, x 0 ), then there

exists a map S m × S n → X of type (α, β) iff [α, β] = 0.
Proof (S m × S n , S m ∨ S n ) is a relative C W -complex with just one cell. A charac-

teristic map for this cell is ψm,n = ψm × ψn : (D m × D n , (D m × D n )• ) → (S m ×
S n , S m ∨ S n ); the attaching map for the cell is a representative of the Whitehead
product [i 1 , i 2 ] of the homotopy classes of the inclusion maps
Sm → Sm ∨ Sn , Sn → Sm ∨ Sn .
9.2 Whitehead Products Between Homotopy Groups of H -Spaces 335
Let α, β be represented by f, g and k = ( f, g) : S m ∨ S n → X be the map deter-

mined by them. Then there is a continuous map S m × S n into X of type (α, β) iff k
can be extended over S m × S n . Since ψm,n is a relative homeomorphism, it is true iff
the map k ◦ ψm,n |(D n ×Dn )• can be extended over D m × D n , i.e., it is nullhomotopic.
But the homotopy class of the latter map is k∗ [i 1 , i 2 ] = [k∗ ◦ i 1 , k∗ ◦ i 2 ] = [α, β].
Corollary 9.2.5 If X is an H -group, then [α, β] = 0 for every α ∈ πm (X, x0 ), β ∈

πn (X, x0 ).
Proof Let X be an H -group with multiplication μ and α, β be represented by f and

g respectively. Then μ ◦ ( f × g) : S m × S n → X has type (α, β). Then by Theorem
9.2.4, it follows that [α, β] = 0. ❑
9.3 A Generalization of Whitehead Product
This section studies the generalized Whitehead product which is obtained by a gen-
eralization of Whitehead product. The set [ X, Y ] has the group structure under
the product of two continuous maps: for f, g : X → Y their product denoted by
f.g : X → Y , is defined by

f (x, 2t), 0 ≤ t ≤ 1/2,
( f ·g)(x, t) = (9.2)
g(x, 2t − 1), 1/2 ≤ t ≤ 1,
where x ∈ X and t ∈ I .
The inverse of a map f : X → Y denoted by f −1 : X → Y and defined by
f −1 (x, t) = f (x, 1 − t), where x ∈ X and t ∈ I. (9.3)
Let A and B be polyhedra, X be any pointed topological space and α ∈ [ A, X ]

be represented by f : A → X and β ∈ [ B, X ] be represented by g : B → X .
If p1 : A × B → A and p2 : A × B → B are the respective projection maps, define
f = f ◦ p1 : (A × B) → X
g = g ◦ p2 : (A × B) → X
and the commutator k = ( f −1 .g −1 ).( f .g ) : (A × B) → X ,

where the products and inverses come from the suspension structure of (A × B)
given by (9.2) and (9.3). Then k |(A∨B) 0, since k |(A×∗) 0 and k |(∗×B) 0.
Using the homotopy extension property for the polyhedral pair ((A × B), A ∨
B), there is a map k : (A × B) → X such that k k and k|(A ∨ B) 0. Hence
k induces a map k̃ : (A ∧ B) = (A × B)/(A ∨ B) → X with the property k =
k̃ ◦ q, where q : A × B → A ∧ B is the projection. As [k̃] does not depend on the

choice of its representatives, it follows that [k̃] is well defined.
Definition 9.3.1 The generalized Whitehead product of α = [ f ] ∈ [ A, X ] and

β = [g] ∈ [ B, X ] denoted by [α, β]GW is defined by setting [α, β]GW = [k̃] ∈
[(A ∧ B), X ]
Remark 9.3.2 The classical definition of Whitehead product is obtained from the the
generalized Whitehead product when A = S p and B = S q . In this sense the above
[α, β]GW is called generalized Whitehead product.
Proposition 9.3.3 If X is an H -group, then [α, β]GW = 0, ∀α ∈ [ A, X ] and

β ∈ [ B, X ].
Proof If X is an H -group, then the group [(A × B), X ] is abelian and hence
the commutator map k = [ f, g] is nullhomotopic. If q : A × B → A ∧ B is the
projection, then it follows that k̃ ◦ 2 q 0. Hence k ◦ k̃ ◦ q 0 : (A ×
B) → X . Consequently, k ◦ k̃ 0 and hence k̃ 0. This implies that
[α, β]GW = 0. ❑
9.4 Mixed Products in Homotopy Groups
This section studies mixed products introduced by McCarty in 1964 (McCarty 1964)
associated with pointed topological spaces and fibrations. It is basically a part of
theory of Hopf construction.
9.4.1 Mixed Product in the Homotopy Category of Pointed

Topological Spaces
This subsection defines mixed product in the homotopy category of pointed topolog-
ical spaces.
Definition 9.4.1 Let X and Y be pointed topological spaces with base point denoted
by ∗ and h : X → Y be a base point preserving continuous map. Let A be a pointed
space and m : A × X → Y be a continuous map such that

m(a, ∗) = ∗, ∀a ∈ A
(9.4)
m(∗, x) = h(x), ∀ x ∈ X
Then given p, q ≥ 1, a product of π p (A) with πq (X ) to π p+q (Y ) is defined as

follows:
9.4 Mixed Products in Homotopy Groups 337
Fig. 9.2 Anticommutativity πp (A) × πq (X) / πp+q (Y )

of the mixed product
1d ×Σ∗ Σ∗

πp (A) × πq+1 (ΣX) / πp+q+1 (Y )
Let f : S p → A, g : S q → X represent α ∈ πq (A), β ∈ πq (X ), respectively.

Consider m ◦ ( f × g) : S p × S q → Y satisfying (9.4) with h replaced by h ◦ g :
S q → Y . Then m ◦ ( f × g) agrees with g ◦ ρ on the subspace
S p ∨ Sq = S p × ∗ ∪ ∗ × Sq ⊂ S p × Sq ,
where ρ : S p × S q → S q denotes the right projections. Thus the separation element

d(μ ◦ ( f × g), g ◦ ρ) ∈ π p+q (Y ) is defined in the natural way. The element d(μ ◦
( f × g), g ◦ ρ) denoted by α, βm is referred as the mixed product associated with
the given map m.
Remark 9.4.2 It is convenient for formal reasons to define
β, αm = (−1) pq+1 α, βm .
Each element of A determines via m, a map of X into Y . If we suspend this map

and reverse the process we obtain a map k : A × X → Y satisfying (9.4) with h
replaced by h : X → Y . Then ∗ α, β = −α, ∗ βm , where ∗ denotes the
Freudenthal suspension. In otherwords, the diagram in Fig. 9.2 is anticommutative,
where the upper row is given by the mixed product associated with m and the lower
row by the mixed product associated with k.
Proof See G.W. Whitehead (1944). ❑
9.4.2 Mixed Product Associated with Fibrations
This subsection defines mixed product corresponding to a fiber space and a topo-
logical transformation group acting on it. F be a fiber with base point e, and let H
be a topological transformation group acting on F. We denote the transformation of
x ∈ F under g ∈ H by g·x. Suppose that
g·e = e, ∀ g ∈ H (9.5)
Let α ∈ π p (H ) and β ∈ πq (F). Their mixed product α, βm ∈ π p+q (F) is
defined as follows:
Take representatives u : S p → H , v : S q → F of α, β, respectively, and let h, k :

S p × S q → F be the maps given by
h(ξ, η) = u(ξ).v(η), where . represents the action of H on G.
k(ξ, η) = v(η), where ξ ∈ S p , η ∈ S q . Since h and k agree on S p ∨ S q by (9.5), their
separation element d(h, k) ∈ π p+q (F) is defined. We denote d(h, k) ∈ π p+q (F) by
α, βm , this is well defined because it is independent of the choice of representatives
and is called the mixed product of α, β associated with fiber F.
9.5 Samelson Products
This section studies Samelson product, generalized samelson product and iterated
Samelson product given by Hans Samelson (1916–2005).
9.5.1 The Samelson Product
This subsection presents Samelson product associated with Hopf groups.
Definition 9.5.1 Let X be an H -group or a topological group with multiplication μ

and two-sided homotopy inversion φ. Define continuous maps k : X → X, x → x −1
and ψ : X × X → X, (x, y) → (x, y)(x −1 y −1 ). Let ψ be a map homotopic to ψ and
restricts to the trivial group on X ∨ X and thus factors through X ∧ X. For the based
spaces P and Q, define the generalized Samelson product
−, − : [P, X ] ⊗ [Q, X ] → [P ∧ Q, X ]
by α, β = [ψ ◦ f ∧ g], where α and β are represented by f and g, respectively.

In particular, for P = S p and Q = S q ,P ∧ Q can be identified with S p+q , and the
Samelson product becomes a map
α, β : π p (X ) ⊗ πq (X ) → π p+q (X )
Remark 9.5.2 The map S p × S q → S q × S p which interchanges the coordinates

induces a map from S p+q into itself which has degree (−1) p+q .
Theorem 9.5.3 The map (α, β) → α, β is bilinear and so defines a pairing π p ⊗
πq → π p+q . Moreover, β, α = (−1) pq+1 α, β.
Proof Left as an exercise.
Remark 9.5.4 One important application of Samelson product is that it can be

considered as an obstruction to homotopy commutativity.
9.5 Samelson Products 339
9.5.2 The Iterated Samleson Product
This subsection presents the Samelson product and the generalized Samelson product
the associated with topological groups. Let G be a topological group. Corresponding
to each element α ∈ π1 (G), there exists an operator
αG : πn (G) → πn+1 (G), n = 1, 2, . . .
defined by taking the Samelson product with α. From the Jacobi identity, each of these
operations constitutes a derivation with respect to the Samelson product in π∗ (G).
For a Lie group G, π2 (G) = 0 and then these derivations form an anticommuting set
of operations, and in particular 2α2G = 0. Clearly, αG = 0 if α can be represented by
a loop within the center of G.
Definition 9.5.5 Let Rn represent the group of rotations of Euclidean n-space,

where n = 1, 2, . . . . Let D : πr (Rn ) → πr +1 (Rn ) be the operator defined by tak-
ing Samelson product with the generator α ∈ π1 (Rn ). In general the operator
D t : πr (Rn ) → πr +t (Rn ) is defined in a similar way for t ≥ 1.
Remark 9.5.6 For n = 2, D is trivial.
Theorem 9.5.7 For n > 2 and n ≡ 2 mod 4, the operator D 2 : πr (Rn ) → πr +2 (Rn )
is trivial.
Proof See James (1971). ❑
Similar result holds for all values of n.
Theorem 9.5.8 The operator D 6 : πr (Rn ) → πr +6 (Rn ) is trivial.
Remark 9.5.9 The operator D 4 : πr (Rn ) → πr +4 (Rn ) is trivial for n = 3 or 4 (see

James (1971)).
Remark 9.5.10 If the topological group is homotopy commutative, then the

Samelson product α, β is trivial. Hence α, β can be thought of an obstruction to
homotopy commutativity. Samelson used this criterion to show that the unitary group
U (2.C) in two variables is not homotopy abelian. One of the results of Samelson
asserts that if α is a generator of π3 (U (2, C)), then α, β is nonzero. The paper
(James and Thomas 1959) asserts that among the classical compact groups G only
the truely commutative ones are homotopy commutative. This method is utilized
again to find elements α of πn (G) such that α, β is nonzero.
9.6 Some Relations Between Whitehead

and Samelson Products
This section gives certain relations between Samelson and Whitehead product in
homotopy groups. We first compare the Whitehead product in homotopy groups of a
pointed topological space X with the Samelson product in homotopy groups of X .
By using the adjointness relation, we have
π p+1 (X ) = [S p+1 , X ] = [ S p , X ] ∼
= [S p , X ] = π p (X ).
We utilize this isomorphism ρ = ρ p : π p+1 (X ) → π p (X ) to define ρ explicitly.
Proposition 9.6.1 Let f : (I p+1 , ∂ I p+1 ) → (X, ∗) represent α ∈ π p+1 (X, ∗). Then
the map
ρ( f ) : (I p , ∂ I p ) → (X, ∗), (x1 , x 2 , . . . , x p )(t) → f (t, x1 , . . . , x p )
represents ρ(α) ∈ π p (X ).
Proof It follows from the above discussion by using the relative homeomorphism
γ p : (I p , ∂ I p ) → (S p , ∗).
Theorem 9.6.2 If α ∈ π p+1 (X ), β ∈ πq+1 (X ), then ρ[α, β] = (−1) p ρ(α), ρ(β)

∈ π p+q (X ).
Proof See Whitehead (1978, pp. 476–478). ❑
Corollary 9.6.3 If α1 , α2 ∈ π p+1 (X ), β ∈ πq+1 (X ) and p > 0, then
[α1 + α2 , β] = [α1 , β] + [α2 , β], [β, α1 + α2 ] = [β, α1 ] + [β, α2 ].
Corollary 9.6.4 Given positive integers p, q, r , let α ∈ π p+1 (X ), β ∈ πq+1 (X ), γ ∈

πr +1 (X ). Then
(−1)r ( p+1) [α, [β, γ]] + (−1) p(q+1) [β, [γ, α]] + (−1)q(r +1) [γ, [α, β]] = 0.
Theorem 9.6.5 If α ∈ π p (X ), β ∈ πq (X ), γ ∈ πm (S p ), δ ∈ πn (S q ), and if [α, β] =

0, then [α ◦ γ, β ◦ δ] = 0.
Proof By hypothesis, [α, β] = 0. Hence there exists a continuous map f : S p ∧

S q → X of type (α, β). Let g : S m → S p , h : S n → S q be representatives of γ and
δ, respectively. Then f ◦ (g × h) : S m × S n → X is a continuous map of the type
(α ◦ γ, β ◦ δ). Hence [α ◦ γ, β ◦ δ] = 0. ❑
Let G be a topological group with a subgroup H and quotient space G/H =

Y (say). Consider the standard action of H on Y given by I.M. James in 1971. This is a
9.6 Some Relations Between Whitehead and Samelson Products 341
pointed action and hence every H -bundle with fiber Y gives a canonical cross section.
The classes of H -bundles over S n correspond to an element α ∈ πn−1 (H ). Consider
the bundle E with fiber Y and base S n which corresponds to an element α ∈ πn−1 (H ).
Let i ∗ : π∗ (Y ) → π∗ (E) be the homomorphism induced by the inclusion i : Y → E.
Let ξ ∈ πn (E) denote the class of the canonical cross section.
Theorem 9.6.6 Under the above notations, for any element β ∈ πq (Y ), the relation
i ∗ α, β = [ξ, i ∗ β] holds, (9.6)
where the brackets on the left denote the relative Samelson product corresponding
to the standard action of H on G/H and those on the right denote the Whitehead
product in π∗ (E).
Remark 9.6.7 The relation (9.6) gives an interesting relation between the relative
Samelson product and Whitehead product, since the existence of a cross section
implies i ∗ is injective.
9.7 Applications
This section presents some interesting applications of different products in homotopy

theory.
9.7.1 Adams Theorem Using Whitehead Product
This subsection conveys the most important application of Whitehead product

appearing in the study of Hopf invariant (see Chap. 17). For example, if τ2n is a gen-
erator of the group π2n (S 2n ) and [τ2n , τ2n ] ∈ π4n−1 (S 2n ) is the Whitehead product.
then the its Hopf invariant H ([τ2n , τ2n ]) = 2, which shows the existence of nonzero
Hopf invariant. There is a natural question: does there exist an element in π4n−1 (S 2n )
with the Hopf invariant 1? This problem has several reformulations. One of them is:
for what values of n the real vector space Rn+1 admit a structure of real division alge-
bra with a unit. Frank Adams solved this problem in 1960. Section 17.5 of Chap. 17
(The Hopf Invariant and Adams Theorem) of this book is referred. Again, if τn is a
generator of the group πn (S n ) represented by the identity map 1d : S n → S n , Adams
proved in 1960 in his paper Adams (1960) that [τn , τn ] = 0 only if n = 1, 3, 7, so
that the n-sphere S n is an H -space only for these values of n.
Remark 9.7.1 The importance of Whitehead product can be realized by the results
given in Ex.16, Ex.18 and Ex.19 of Sect. 9.8.
9.7.2 Homotopical Nilpotence of the Seven Sphere S7
This subsection studies homotopical nilpotence of S 7 . Let [X, Y ] denote the set of
all homotopy classes of base point preserving continuous maps from X to Y . We will
not distinguish notationally between a map and its homotopy class. The multipli-
cation and inversion in the unit Cayley numbers induce the standard multiplication
μ : S 7 × S 7 → S 7 and two-sided inverse φ : S 7 → S 7 . Then μ ∈ [S 7 × S 7 , S 7 ] and
φ ∈ [S 7 , S 7 ]. For the H -space (S 7 , m, φ) define a commutator map ψ : S 7 × S 7 →
S 7 , (x, y) → (x y)(x −1 y −1 ) using the multiplication μ and inversion φ.
Recall that Cayley multiplication is not associative but is disassociative in the sense
that any two elements generate an associative subalgebra. Define inductively the n-
fold commutator map m : (S 7 )n → S 7 by m n = m ◦ (m n−1 × 1d ), where m 1 = 1d ,
the identity map on S 7 . Then m n induces a unique homotopy class kn : ∧n S 7 → S 7
with kn ◦ qn = m n , where ∧n S 7 is the n-fold smash product of S 7 (≈ S 7n ) and qn :
(S 7 )n → ∧n S 7 is the projection map. The homotopical nilpotence of the H -space
(S 7 , μ, φ) denoted by nil (S 7 , μ, φ) is the least integer n such that m n+1 (and hence
kn+1 ) is nullhomotopic.
Theorem 9.7.2 nil (S 7 , μ, φ) = 3.
Proof See Gilbert (1972). ❑
9.8 Exercises
In this exercise, let [α, β] and α, β denote Whitehead product and Samelson product
of α and β respectively. Prove the following:
1. If m > 1, α ∈ πm (X, x 0 ) and β ∈ π1 (X, x0 ), then [α, β] is the element βα − α ∈
πm (X, x0 ).
2. If m > 1, then the assignment α → [α, β] for a given β ∈ πn (X, x0 ) defines a
homomorphism β∗ : πm (X, x0 ) → πm+n−1 (X, x0 ).
3. If m + n > 2, given α ∈ πm (X, x 0 ) and β ∈ πn (X, x 0 ), [β, α] = (−1)mn [α, β].
4. Let α ∈ πm (X, x 0 ), β∈πn (X, x 0 ) and γ ∈ π p (X, x 0 ). Then (−1)mp [[α, β], γ] +
(−1)nm [[β, γ], α] + (−1) pn [[γ, α], β] = 0.
mra
5. Let X be an arbitrary homotopy abelian H - space and α be an arbitrary element
in π3 (X ). Show that α, α = 0 iff 2(α ∧ α)∗ (β) = 0 in π6 (X ) for any β ∈
[X ∧ X.X ].
6. Let τ2n be a generator of the group π2n (S 2n ). Show that
(i) the group π4n−1 (S 2n ) is infinite for any n ≥ 1;
(ii) [τ2n , τ2n ] ∈ π4n−1 (S 2n ) is in the kernel of the suspension homomorphism
: π4n−1 (S 2n ) → π4n (S 2n+1 ).

9.8 Exercises 343
7. Let X be a topological group and α ∈ π p (X ) and β ∈ πq (X ). Show that

(i) for a fixed α, the map β → [α, β] is a homomorphism of groups and hence
for n ≥ 2,
[α, β1 + β2 ] = [α, β1 ] + [α, β2 ].
(ii) for m + n ≥ 3, [β, α] = (−1)mn [α, β] and for m + np ≥ 4, there is a

Jacobi identity,
(−1)mp [[α, β], γ] + (−1)nm [[β, γ], α] + (−1) pn [[γ, α], β] = 0.
(iii) for any continuous map between two topological groups
ψ : (X, x 0 ) → (Y, y0 ), ψ∗ ([α, β]) = [ψ∗ (α), ψ∗ (β)].
8. For n ≥ 2, show that the operator D 4 : πr (S n ) → πr +4 (S n ) given in Definition

9.5.5 is trivial.
9. Let X be an H -space and A, B be polyhedra. Given α ∈ [ A, X ] and β ∈
[β, X ], show that [α, β] = 0 iff there is a map μ : A × B → X such that
[μ| A ] = α and [μ| B ] = β.
10. Show that all generalized Whitehead products vanish in a pointed space X iff
[ P, X ] (equivalently, [P, X ]) is abelian for all polyhedra P.
11. Let A be a polyhedron and τ denote the class of identity map of A. If the
generalized Whitehead product [τ , τ ] = 0, show that A is an H -space.
12. If X is an H -space, show that all Whitehead products are trivial on X .
13. If q ≤ 3n − 3, homomoporphism
E : πq (S n ) → πq+1 (S n+1 )
is generated by all Whitehead products [α, β] with α ∈ πr (S n ), β ∈ πr (S n ), r +

s = q + 1.
mra
14. (a) Let μ : S 3 × S 3 → S 3 be a homotopy associative multiplication of the 3-
sphere S 3 and β ∈ π3 (S 3 ) ∼
= Z be a generator. Show that Samelson (com-
mutator) product β, βμ ∈ π6 (S 3 ) ∼ = Z12 .
(b) Let X be a finite C W -complex which is also an H -space with π3 (X ) ∼ =Z
and β ∈ π3 (X ) be a generator. Show that
(i) if m is any homotopy associative multiplication on X , then the Samelson
product
β, βm generates π6 (X ).
(ii) given a multiplication m on S 3 , there exists a multiplication μ on the
H -space X such that the generator β : (S 3 , m) → (X, μ) is an H -map.
(iii) given any multiplication μ on X , there exists a multiplication m on S 3
such that the generator β : (S 3 , m) → (X, μ) is an H -map.
[Hint: See Arkowitz and Curjel (1969) for (a) and Stephen (1978) for (b).]
15. Examine the validity of following statements:

(i) Let μ : S 3 × S 3 → S 3 be any given multiplication and X be a pointed space.
Then there exists a multiplication m : X × X → X such that β : (S 3 , μ) →
(X, m) is an H -map;
(ii) Let m : X × X → X be any given multiplication and μ : S 3 × S 3 → S 3 be
a multiplication such that such that β : (S 3 , μ) → (X, m) is an H -map.
16. Show that the n-sphere S n ia an H -space iff the Whitehead product [τn , τn ] = 0,
where τn ∈ πn (S n ) is represented by the identity map.
17. Using Samelson product prove that among the classical compact groups G,
the only commutative ones are those which are homotopy abelian. Hence find
elements α ∈ πn (G) such that the Samelson product α, α = 0 (James and
Thomas 1959).
18. Let X = S 2n × S 2n / ∼ be the quotient space obtained by identifying the points
(x, x0 ) and (x 0 , x), where x0 is the base point of S
2n
. Show that the space X =
S × S / ∼ is homeomorphic to the space S
2n 2n 2n
D 4n , where f is the map
f
defining the Whitehead [τ2n , τ2n ], where τ2n is a generator of the group π2n (S 2n ).
19. Let τn ∈ πn (S n ), αk ∈ πk (S k ) be the generators given by the identity maps 1d :
S n → S n and 1d : S k → S k respectively. Show that the
(i) Whitehead product [τn , αk ] ∈ πn+k−1 (S n ∨ S k ) has infinite order;
(ii) group πn+k−1 (S n ∨ S k ) is infinite.

[2] Arkowitz, M., Whitehead products as images of Pontrjagin products, Trans.
Amer. Math. Soc. 158, 453–463, 1971.
[3] Blakers, A.L., and Massy, W.S., Products in homotopy theory, Ann. of Math.
58, 295–324, 1953.
[5] Hilton, P.J., Homotopy Theory and Duality, Nelson, London, 1965.
[6] Hilton, P.J., Note on the Jacobi identity for Whitehead products, Proc. Cam-
bridge Philos. Soc., 57, 180–182, 1961.
[7] James, I.M., and Thomas E., Which Lie Groups are homotopy abelian?, Proc.
Nat. Acad. Sc (USA) 45(5), 131–140, 1959.
[8] Massey, W.S., and Uehara,H., The Jacobi identity for Whitehead products,
Princeton Univ. Press, 1957.
[9] May, J.P., A Concise Course in Algebraic Topology, University of Chicago
Press, 1999.
[10] McCarty, G.S.,Products between homotopy groups and J -morphism, Quart.J.

of Math. Oxford 15(2), 362–370, 1964
[11] Miyazaki, H., On realizations of some Whitehead products, Tohoku Math. J.,
12, 1–30, 1960.
[12] Nakaoka, M., and Toda, H., On Jacobi identity for Whitehead products, J. Inst.
Polytech, Osaka City Univ. Ser.A.5, 1–13, 1954.
[13] Samelson H., A connection between the Whitehead product and Pontragin
product, Amer. J. Math 75, 744–752,1953.
[14] Toda, H., Generalized Whitehead products and homotopy groups of spheres,
J. Inst. Polytech. Osaka City Univ. Ser. A,3, 43-48, 1953.
References
Adams, J.F.: On the non-existence of elements of Hopf invariant one. Ann. Math. 72, 20–104 (1960)
(1972)
Arkowitz, M.: Whitehead products as images of Pontrjagin products. Trans. Am. Math. Soc. 158,
453–463 (1971)
Arkowitz, M., Curjel, C.R.: Some properties of the exotic multiplications on the three-sphere. Q. J.
Math. Oxf. Ser. 20(2), 171–176 (1969)
Blakers, A.L., Massy, W.S.: Products in homotopy theory. Ann. Math. 58, 295–324 (1953)
Dieudonné, J.: A History of Algebraic and Differential Topology, pp. 1900–1960. Modern
Birkhäuser, Boston (1989)
Gilbert, W.J.: Homotopical nilpotence of the seven sphere. Proc. Am. Math. Soc. 32, 621–622
(1972)
Gray, B.: Homotopy Theory. An Introduction to Algebraic Topology. Acamedic Press, New York
(1975)
Hilton, P.J.: Homotopy Theory and Duality. Nelson, London (1965)
Hilton, P.J., Whitehead, J.H.C.: Notes on the Whitehead product. Ann. Math. 58(2), 429–442 (1953)
Hilton, P.J.: Note on the Jacobi identity for Whitehead products. Proc. Camb. Philos. Soc. 57,
180–182 (1961)
James, I.M.: Products between homotopy groups. Compositio Mathematica 23, 329–45 (1971)
James, I.M., Thomas, E.: Which Lie groups are homotopy abelian? Proc. Nat. Acc. Sc. USA 25,
131–140 (1959)
Massey, W.S., Uehara, H.: The Jacobi Identity for Whitehead Products. Princeton University Press,
Princeton (1957)
Maunder, C.R.F.: Algebraic Topology. Van Nostrand Reinhold Company, London (1980)
May, J.P.: A Concise Course in Algebraic Topology. University of Chicago Press, Chicago (1999)
McCarty, G.S.: Products between homotopy groups and J -morphism. Q. J. Math. Oxf. 15(2), 362–
370 (1964)
Miyazaki, H.: On realizations of some Whitehead products. Tohoku Math. J. 12, 1–30 (1960)
Nakaoka, M., Toda, H.: On Jacobi identity for Whitehead products. J. Inst. Polytech, Osaka City
Univ. Ser A 5, 1–13 (1954)
Samelson, H.: A connection between the Whitehead product and Pontragin product. Am. J. Math
75, 744–752 (1953)
Stephen, J.S.: A Samelson product and homotopy associativity. Proc. Am. Math. Soc 70(2), 189–195
(1978)
Toda, H.: Generalized Whitehead products and homotopy groups of spheres. J. Inst. Polytech. Osaka
City Univ. Ser. A,3, 43-48 (1953)
Whitehead, G.W.: On products in homotopy groups. Ann. Math. 47(2), 460–475 (1944)
Whitehead, J.H.C.: On adding relations to homotopy groups. Ann. of Math. 42(2), 409–428 (1941)
Chapter 10
Homology and Cohomology Theories
This chapter opens with homology and cohomology theories which play a key role in
algebraic topology. Homology and cohomology groups are also topological invari-
ants like homotopy groups and Euler characteristic. Homology (cohomology) the-
ory is a sequence of covariant (contravariant) functors from the category of chain
(cochain) complexes to the category of abelian groups (modules). A key feature
of these functors is their homotopy invariance in the sense that homotopic maps
induce the same homomorphism in homology (cohomology). In particular, topo-
logical spaces of the same homotopy type have isomorphic homology (cohomolgy)
groups.
Homotopy groups are easy to define but very difficult to compute in general.
For example, for spheres the computation of πm (S n ) for m > n faces serious prob-
lems. Fortunately, there is a more computable alternative approach to homotopy
groups, the so-called homology groups Hn (X ) of a topological space X . For exam-
ple, for spheres, the homology groups Hm (S n ) are isomorphic to the homotopy
groups πm (S n ) for 1 ≤ m ≤ n and Hm (S n ) = 0 for all m > n, which is an advantage
of homology groups. Historically, homology groups came earlier than homotopy
groups. Homology invented by H. Poincaré in 1895 is one of the most fundamen-
tal influential invention in mathematics. Homology groups are refinements, in some
sense, of Euler characteristic.
Chapter 12 presents another approach, known as an axiomatic approach to homol-
ogy and cohomology theories defined on the category of spaces having homotopy
type of finite CW -complexes. This approach is essentially due to Eilenberg and Steen-
rod and is the most important contribution to algebraic topology since the invention
of the homology groups by Poincaré. Homotopy and homology groups have some
close relations at least for a certain class of topological spaces.
The aim of homology theory is to assign a group structure to cycles that are not
boundaries. The basic tools such as complexes and incidence numbers for construct-
ing simplicial homology groups were given by Poincaré in 1895. The basic idea of
his construction is that it starts with a geometric object (a space) which is given
DOI 10.1007/978-81-322-2843-1_10
348 10 Homology and Cohomology Theories
by combinatorial data (a complex). Then the linear algebra and boundary relations
determined by this data are used to construct homology groups. It took more than
thirty years to develop homology theory (Hn ) applicable to curvilinear polyhedra,
embodying the notions given by Poincaré in 1895. The functor Hn measures the
number of ‘n-dimensional holes’ in the simplicial complex (or in the the space),
which means that the n-sphere S n has exactly one n-dimensional hole and there is no
m-dimensional hole if m = n. A 0-dimensional hole is a pair of points in different
path components which asserts that H0 measures path connectedness. The simplicial
techniques in the simplicial homology theory prescribed by Poincaré were gradu-
ally generalized to singular homology using the algebraic properties of the singular
complex. The cohomology groups of a topological space were not recognized until
1930.
After setting up the basic apparatus, H. Poincaré (1854–1912) constructed the
homology groups of a polyhedron in 1895. These homology groups have several
generalizations to singular homology groups of an arbitrary topological space made
by S. Lefschetz (1884–1972) in 1933, S. Eilenberg (1915–1998) in 1944, E. Čech
(1893–1960) in 1932, and for compact metric spaces by L. Vietoris (1891–2002)
in 1927. Their approaches for constructing homology and cohomology theories and
choice of methods are often dictated by the nature of the problems. For example,
singular homology and cohomology theories are defined for all topological spaces.
The idea of Poincaré on homology theory was generalized in two directions
(i) from complexes to more general topological spaces where the homology groups
are not characterized by numerical invariants;
(ii) from the group Z to arbitrary abelian groups.
There exist different homology theories such as simplicial homology, singular
homology, Čech homology, cellular homology, etc., and their corresponding coho-
mology theories with different constructions. But Eilenberg–Steenrod theorem uni-
fied them by showing that any two homology theories with isomorphic coefficient
groups on the category of all compact polyhedral pairs are isomorphic (see Chap. 12).
The cohomology groups (modules) of a topological space were not recognized until
S. Lefschetz formulated a simplified method of the duality theorem for manifolds in
the 1930s.
Homology theory Hn and cohomology theory H n are dual to each other in some
sense: there is a bilinear pairing of chains and cochains and Hn is a covariant functor
but H n is a contravariant functor. The basic property of cohomology which dis-
tinguishes it from homology is the existence of a natural multiplication called cup
product which makes the direct sum of all cohomology modules with coefficient in a
ring R into a graded R-algebra. This extra structure is more subtle than the additive
structure of homology module (group) of the space.
The most important homology theory in algebraic topology is the singular homol-
ogy. The simplicial techniques are gradually modified until the creation of singular
homolgy by S Eilenberg which is a topological invariant. Simplicial homology is
the primitive version of singular homology. To inaugurate a simplicial homology
10 Homology and Cohomology Theories 349
theory, H. Poincaré started in 1895 with a geometric object (a space) which is given
by combinatorial data (a simplicial complex). Then the linear algebra and bound-
ary relations by these data are used to construct homology groups. Using these tools
Poincaré defined directly the Betti numbers invented by E. Betti (1823–1892) and tor-
sion numbers which are numerical invariants and characterize the homology groups
having coefficient group Z of integers.
Attention for shift from numerical invariants to groups associated with homology
theories was successfully made during the period 1925–1935. This shift is partly
due to Emmy Noether (1882–1935). Her algebraic approach to homology conveys
a major contribution to the geometrical approach to homology given in 1895 by
H. Poincaré. There is a natural question: how to relate the groups C p (K ; G),
Z p (K ; G), and B p (K ; G) defined in Sect. 10.2 to the topological spaces whose tri-
angulation is K ? Is it possible for C p (K ; G), to express any property which remains
unchanged under homeomorphism? Homology groups provide the desired topolog-
ical invariant. Cohomoogy theory invented by J.W. Alexander (1888–1971) and A.
Kolmogoroff (1903–1987) independently in 1935 is dual to homology theory. E.Čech
and H. Whitney (1907–1989) developed simplicial cohomology theory during 1935–
1940. The terms ‘coboundary’, ‘cocycle’, ‘cochain’, and ‘cohomology’ were given
by E.Čech.
More precisely, this chapter conveys constructions of simplicial, singular, Čech
and cellular homology theories, and their dual cohomology theories. Moreover, this
chapter studies basic properties of homology and cohomology theories, Euler char-
acteristic (a numerical topological invariant) and Betti number from the viewpoint
of homology theory, Hurewictz theorem, Mayer–Vietoris sequences, Jordan curve
theorem, and universal coefficient theorem and also discusses cohomology theory.
For this chapter the books Croom (1978), Dold (1972), Gray (1975), Hatcher
(2002), Maunder (1970), Rotman (1988), Spanier (1966) and some others are referred
in Bibliography.
10.1 Chain Complexes
This section studies chain complexes with their basic properties needed for con-
structing homology groups. W. Mayer (1887–1947) studied in 1929 chain complex,
boundary, cycle from purely algebraic viewpoint.
Definition 10.1.1 A sequence C = {Cn , ∂n }, n ∈ Z of abelian groups and their

homomorphisms
∂n : Cn → Cn−1
such that ∂n ◦ ∂n+1 = 0, for all n, is called a chain complex and ∂n is called a boundary
homomorphism. More precisely,
∂n+1 ∂n
C : · · · → C n+1 −−−−−→ Cn −−−−→ Cn−1 → · · · (10.1)
is called a chain complex if for all n ∈ Z, the equality ∂n ◦ ∂n+1 = 0 holds. The group
Cn is called the n-dimensional chain group of the complex C and elements of Cn are
called n-chains for C.
Definition 10.1.2 The elements of Z n = ker ∂n and elements of Bn = Im ∂n+1 in the

sequence (10.1) are called n-cycles and n-boundaries for the complex C respectively.
Proposition 10.1.3 For any chain complex C in the sequence (10.1), Bn = Im ∂n+1
is a subgroup of Z n = ker ∂n .
Proof It follows from the condition of a chain complex C that ∂n ◦ ∂n+1 = 0, for all
n ∈ Z. ❑
Definition 10.1.4 The quotient group Z n /Bn for any chain complex C is called the
n-dimensional homology group of the chain complex C, denoted by Hn (C) or simply
Hn . The complex C is said to be acyclic if Hn (C) = 0 for all n. The elements of
Hn = Z n /Bn are called homology classes, denoted by [z] for every z ∈ Z n .
Remark 10.1.5 For an acyclic complex C, Hn (C) = 0 for all n implies that the
sequence (10.1) is exact at Cn for all n and hence it makes the sequence (10.1) exact.
This shows that the homology group of a chain complex measures its deviation from
the exactness of the sequence (10.1).
Example 10.1.6 Consider the chain complex
∂2 ∂1
C : · · · → 0 → Z ⊕ Z −−−−→ Z −−−→ 0 → · · · ,
where the chain groups are given by
C1 = Z, C2 = Z ⊕ Z, C n = 0 for n = 1, 2,
and the homomorphism ∂2 is defined by ∂2 (x, y) = 3x + 3y. The group Z 2 of 2-

cycles is isomorphic to Z and the group Z 1 of 1-cycles is C1 = Z. All other groups
of cycles are zero. On the other hand, the groups Bn of n-boundaries are zero except
for n = 1 and B1 is generated by the element 3 of the group Z. Hence H1 (C) =
Z3 , H2 (C) = Z and Hn (C) = 0 for n = 1, 2.
Definition 10.1.7 Let C = {Cn , ∂n } and C = {Cn , ∂n }, n ∈ Z be two chain com-
plexes of abelian groups. A sequence f = { f n : Cn → Cn }, n ∈ Z of homomor-
phisms is called a chain map from C to C if these homomorphisms commute with
the boundary homomorphisms, i.e., if each square in the Fig. 10.1 is commutative,
i.e.,

f n ◦ ∂n+1 = ∂n+1 ◦ f n+1 , ∀ n ∈ Z.
We abbreviate the entire above collection to f : C → C and call f a chain map.

10.1 Chain Complexes 351
Fig. 10.1 Chain map ∂n+1 ∂n

··· / Cn+1 / Cn / Cn−1 / ···
fn+1 fn fn−1

∂n+1
∂n

··· / C / C / C / ···
n+1 n n−1
Proposition 10.1.8 Let C = {Cn , ∂n } and C = {Cn , ∂n } be two chain complexes of
abelian groups and f = { f n : Cn → Cn } be a chain map. Then f n maps n-cycles of
C into n-cycles of C and n-boundaries of C into n-boundaries of C for all n ∈ Z,
i.e., f n (Z n ) ⊂ Z n and f n (Bn ) ⊂ Bn for every n.
Proof The proof follows from the commutativity of each square in Fig. 10.1. ❑
Theorem 10.1.9 Let f : C → C be a chain map between two chain complexes

C = {Cn , ∂n } and C = {Cn , ∂n }. Then for each integer n, f n : Cn → Cn induces a
homomorphism.
Hn ( f n ) = f n ∗ : Hn (C) → Hn (C ), [z] → [ f n (z)].
Definition 10.1.10 Hn ( f n ) = f n ∗ : Hn (C) → Hn (C ) defined in Theorem 10.1.9 is

called the homomorphism in homology induced by f n for every integer n.
Remark 10.1.11 f and f ∗ are written in places of f n and f n ∗ respectively, unless

there is any confusion.
Proposition 10.1.12 (a) Let f : C → C and g : C → C be two chain maps.

Then their composite g ◦ f : C → C is also a chain map such that (g ◦ f )∗ =
g∗ ◦ f ∗ : Hn (C) → Hn (C ).
(b) If 1C : C → C is the identity chain map, then (1C )∗ : Hn (C) → Hn (C) is the
identity automorphism.
Definition 10.1.13 Let C = {Cn , ∂n } and C = {Cn , ∂n } be two chain complexes
and f, g : C → C be two chain maps. Then f is said to be chain homotopic to g,

denoted by f g, if there is a sequence {Fn : Cn → Cn+1 } of homomorphism such
that

∂n+1 Fn + Fn−1 ∂n = f n − gn : Cn → Cn , ∀ n ∈ Z
holds.
In particular, a chain map f : C → C is called a chain homotopy equivalence if
there exists a chain map g : C → C such that g ◦ f 1C and f ◦ g 1C .
Proposition 10.1.14 The relation of chain homotopy on the set S(C, C ) of all chain
maps from C to C is an equivalence relation.
Theorem 10.1.15 Two homotopic chain maps f, g : C → C induce the same

homomorphisms in the homology, i.e., if f g : C → C , then the homomorphisms
f ∗ = g∗ : Hn (C) → Hn (C ) for every n.
Proof Let f g : C → C . Then there exists a chain homotopy {Fn : Cn → Cn+1

}.
Let [z] ∈ Hn (C). Then ∂n ([z]) = 0 shows that f n ([z]) − gn ([z]) = ∂n+1 Fn ([z]) is a
boundary. Hence [ f n [z]] = [g, [z]] implies that f n ∗ ([z]) = gn ∗ ([z]), ∀ [z] ∈ Hn (C).
Consequently, f n ∗ = gn ∗ for all n. Hence f ∗ = g∗ . ❑
Proposition 10.1.16 (a) All chain complexes and chain maps form a category
denoted by Comp.
(b) For each n ∈ Z, Hn is a covariant functor from the category Comp of chain com-
plexes and chain maps to the category Ab of abelian groups and homomorphisms.
Proof (a) The objects here are taken chain complexes and the morphisms are taken
chain maps. The composition of chain maps is defined coordinatewise: {gn } ◦
{ f n } = {gn ◦ f n }. Hence they form a category written Comp.
(b) The object function is defined by assigning to each chain complex the sequence
of its homology groups, and morphism function is defined by assigning to each
chain map f between chain complexes the induced map f ∗ between their homol-
ogy group. This shows that for each n, Hn : Comp → Ab is a covariant functor
by Proposition 10.1.12. ❑
10.2 Simplicial Homology Theory
This section begins with the simplicial homology theory invented by H.Poincaré in
1895 on the category of simplicial pairs starting with construction of the homology
groups of a simplicial complex in two steps: first by assigning to each simplicial
complex a certain complex, called chain complex followed by assigning to the chain
complex its homology group. This theory stems from Poincaré’s seminal mathemat-
ics paper ‘Analysis situs’ and five supplements to the paper around the turn of the
nineenth century to the beginning of the twentieth century (between 1895 and 1904).
This theory characterizes topological spaces which look like polyhedra. These can
be used to cover a manifold by a process called triangulation. The most advantage
of this theory is: it is easier to visualize geometrically than other homology theories.
10.2 Simplicial Homology Theory 353
10.2.1 Construction of Homology Groups of a Simplicial

Complex
This subsection constructs homology groups of oriented simplicial complexes K and

it is shown that they coincide with the homology groups of |K |. This construction
assigns a group structure to cycles that are not boundaries. This subsection associates
to every simplicial map a homomorphism on the corresponding simplicial homology
groups and presents functorial properties of simplicial homology. The problem is:
how to relate C p (K ), Z p (K ) and B p (K ) to the topological properties of the spaces
whose triangulation is K ? This subsection gives its answer.
Definition 10.2.1 An oriented simplicial complex K is a simplicial complex and a

partial order on vertex set Vert(K ) whose restriction to the vertices of any simplex
in K is a linear order.
An oriented n-simplex for n ≥ 1, is obtained from an n-simplex sn = v0 v1 . . . vn
by assigning an ordering of its vertices. The two equivalence classes of permutations
of the ordering of its vertices determine the orientations of sn . The equivalence class of
even permutations of the chosen ordering determines the positively oriented simplex
+sn . On the other hand the equivalence class of odd permutations determines the
negatively oriented simplex −sn . A vertex is considered a zero simplex positively
oriented.
Example 10.2.2 In the 2-simplex s2 = v0 v1 v2 , if the ordering v0 < v1 < v2 is

assigned, then +s2 = v0 v1 v2 and −s2 = v2 v1 v0 .
Definition 10.2.3 Let K be an oriented simplicial complex with simplexes s p+1

and s p , whose dimensions differ by 1. Then for each such pair (s p+1 , s p ) a number
[s p+1 , s p ], called incidence number is assigned as follows:
If s p is not a face of s p+1 , then [s p+1 , s p ] = 0.
If s p is a face of s p+1 , then vertices v0 , v1 , . . . , v p of s p are labeled so that
+s p = v0 v1 . . . v p .
Let v denote the vertex of s p+1 which is not vertex of s p . Then s p+1 = ±v0 v1 . . . v p .
If +s p+1 = +v0 v1 . . . v p , then [s p+1 , s p ] = +1.
If +s p+1 = −v0 v1 . . . v p , then [s p+1 , s p ] = −1.
Remark 10.2.4 If [s p+1 , s p ] = +1, then s p is a positively oriented face of s p+1 and
if [s p+1 , s p ] = −1, then s p is a negatively oriented face of s p+1 .
The choice of a positive ordering of vertices of s p+1 clearly induces a natural
ordering of the vertices in each face of s p+1 . Thus an orientation of s p+1 induces a
natural ordering of its vertices. Hence the Definition 10.2.3 implies that if s p is a face
of s p+1 , then the incidence number [s p+1 , s p ] is positive or negative according as the
chosen orientation of s p+1 agrees or disagrees with orientation of s p respectively.
Example 10.2.5 If +s2 = v0 v1 v2 , σ1 = v0 v1 , ρ1 = v0 v2 , as shown in Fig. 10.2,

then [s2 , σ1 ] = +1 but [s2 , ρ1 ] = −1.
Fig. 10.2 Orientations v2

}> AAA
involving +s2 , σ1 and ρ1 . }}} AA
} AA
ρ1 }} AA
}}} AA
}} AA
}} AA
}} A
v0 σ1
/ v1
Theorem 10.2.6 Let K be an oriented complex, s p be an oriented p-simplex of K

and s p−2 be a ( p − 2)- face of s p . Then

[s p , s p−1 ][s p−1 , s p−2 ] = 0.
s p−1 ∈K
Proof Order the vertices v0 , v1 , . . . , v p−2 so that +s p−2 = v0 . . . v p−2 . Then s p
has two additional vertices a and b(say). Assume that +s p = abv0 . . . v p−2 . Every
( p − 2)-simplex in s p is a face of exactly two ( p − 1)-faces of s p , which are taken
+s 1p−1 = av0 . . . v p−2 and +s 2p−1 = bv0 . . . v p−2 (say). Thus the nonzero terms
in the sum
[s p , s p−1 ][s p−1 , s p−2 ]
s p−1 ∈K
occur for only +s 1p−1 and +s 2p−1 .

Case 1 Suppose +s 1p−1 = +av0 . . . v p−2 and +s 2p−1 = +bv0 . . . v p−2 .
Then [s p , s 1p−1 ] = −1, [s 1p−1 , s p−2 ] = +1, [s p , s 2p−1 ] = +1 and [s 2p−1 , s p−2 ] =
+1.
Hence [s p , s p−1 ][s p−1 , s p−2 ] = (−1)(+1) + (+1)(+1) = 0.
s p−1 ∈K
Case 2 Suppose +s 1p−1 = +av0 . . . v p−2 and +s 2p−1 = −bv0 . . . v p−2 .
Then [s p , s 1p−1 ] = −1, [s 1p−1 , s p−2 ] = +1, [s p , s 2p−1 ] = −1 and [s 2p−1 , s p−2 ] =
−1.
Hence [s p , s p−1 ][s p−1 , s p−2 ] = 0. The remaining other cases may be proved
s p−1 ∈K
similarly. ❑
To define the homology groups of an oriented simplicial complex K , we need the

concepts of p-chains, p-cycles and p-boundaries corresponding to K .
Definition 10.2.7 Let K be an oriented simplicial complex and p be a positive inte-

ger. A p-dimensional chain or a p-chain is a function c p from the family of oriented
p-simplicial of K to Z such that for each p-simplex s p , c p (−s p ) = −c p (+s p ). The
family of p-chains forms a group called the p-dimensional chain group of K with
coefficients in Z under usual functional addition:
(c1p + c2p )(s p ) = c1p (s p ) + c2p (s p ), (10.2)

where the addition on right-hand side (RHS) of (10.2) is the usual addition in Z.
This group denoted by C p (K , Z) is called the p-dimensional chain group of K with
coefficients in Z.
The Definition 10.2.7 can be extended for an arbitrary abelian group G in place
of Z.
Definition 10.2.8 A p-dimensional chain or a p-chain of an oriented simplicial

complex K with coefficients in an additive abelian group G is a function c p from
the family of oriented p-simplexes of K to G such that if c p (+s p ) = g for some
g ∈ G, then c p (−s p ) = −g. The family of such p-chains forms a group denoted by
C p (K ; G) under usual functional addition
(c1p + c2p )(s p ) = c1p (s p ) + c2p (s p ), (10.3)
where the addition in RHS of (10.3) is the usual group operation in G. As G is an

abelian group, C p (K ; G) is also an abelian group. The group C p (K ; G) is called the
p-dimensional chain group of K with coefficients in G.
Remark 10.2.9 If the oriented complex K has no p-simplex for some p, then we
take C p (K , G) = 0, the trivial group consisting of the identity element 0 of G only.
Remark 10.2.10 If the oriented complex K is infinite, then c p (s p ) = 0 for all but a
finite number of p-simplexes of K .
Definition 10.2.11 An elementary p-chain on an oriented complex K is a p-chain

c p for which there is a p-simplex s p in K such that c p (+s p ) = g and c p (t p ) = 0,
whenever, s p = ±t p . Such a p-chain is denoted by a formal product g·s p , when
g = c p (+s p ).

An arbitrary p-chain d p on K can be written as a formal finite sum d p = gi ·s ip of
elementary p-chains where gi = c p (+s ip ) and the index i ranges over all p-simplexes
of K . This notation justifies the use of the word coefficient.
Proposition 10.2.12 Let K be a finite oriented simplicial complex. Then

(i) If c p = f i ·s ip and d p = gi ·s ip are two p-simplexes on K , then
c p + d p = ( f i + gi )·s ip ;
(ii) The additive inverse of d p in the group C p (K ; G) is the chain −d p = (−gi )·s ip .
Proof (i) and (ii) follow from the definition of p-chains. ❑
Theorem 10.2.13 Let K be a finite oriented simplicial complex and n p be the num-
ber of p-simplexes in K . Then the chain group C p (K ; G) is isomorphic to the direct
sum of n p -copies of G.
Proof Let C p (K ; G) be the p-dimensional chain group of K with coefficients in G.

Define a map

ψ : C p (K ; G) → G ⊕ G ⊕ · · · ⊕ G, gi ·s ip → (g1 , g2 , . . . , gn p ).
s p ∈K
Then ψ is a group isomorphism. ❑
Remark 10.2.14 If the coefficient group G is taken to be a commutative ring or field,

then C p (K ; G) is a module or a vector space. Poincaré original definition was given
in terms of integers.
To define the homology groups of a finite oriented simplicial complex K with

coefficient group G, we introduce the following concepts. Let K be a finite oriented
simplicial complex and G be an abelian group.
Definition 10.2.15 If g·s p is an elementary p-chain with p ≥ 1, then the boundary

of g·s p , denoted by ∂(g·s p ), is defined by

∂(g·s p ) = [s p , s ip−1 ]g·s ip−1 (10.4)
s ip−1 ∈K
The boundary operator ∂ is now extended by linearity to a homomorphism

∂ : C p (K ; G) → C p−1 (K ; G), c p → ∂(gi ·s ip−1 ) (10.5)
The boundary of a 0-chain is taken to be 0.
The boundary operator ∂ has an interesting property.
Theorem 10.2.16 Let K be a finite oriented simplicial complex and p ≥ 2. Then

the composite homomorphism
∂p ∂ p−1
C p (K ; G) −−−→ C p−1 (K ; G) −−−−−→ C p−2 (K ; G)
is trivial, i.e., ∂ p−1 ◦ ∂ p = 0 for all p ≥ 2.
Proof To prove the theorem it is sufficient to prove the result for an elementary p-
chain g·s p for p ≥ 2. We claim that for such a p-chain g·s p , the composite ∂ p−1 ◦
∂ p = 0. Now
(∂ p−1 ◦ ∂ p )(g·s p ) = ∂ p−1 (∂ p (g·s p ))

⎛ ⎞

= ∂ p−1 ⎝ [s p , s ip−1 ]g·s ip−1 ⎠ by (10.4)
s p−1 ∈K
j j

= [s p , s ip−1 ][s ip−1 , s p−2 ]g·s p−2 (10.6)
s ip−1 ∈K s p−2
j
∈K
j
Reversing the order of summation and collecting coefficients of each simplex s p−2 ∈
K , we have
⎛ ⎞

∂ p−1 (∂ p (g·s p )) = ⎝ [s p , s p−1 ][s p−1 , s p−2 ]g·s p−2 ⎠
i i j j
(10.7)
j
s p−2 ∈K s ip−1 ∈K
= 0 by Theorem 10.2.6 for all elementary p-chains g·s p .
Hence from (10.7) it follows that ∂ p−1 ◦ ∂ p = 0. ❑
Corollary 10.2.17 For any oriented simplicial complex K and an abelian group G,
the groups C p (K ; G) and the homomorphisms ∂ p : C p (K ; G) → C p−1 (K ; G) form
a chain complex, denoted by C(K ; G).
Corollary 10.2.18 Im ∂ p is a subgroup of ker ∂ p−1 .
Definition 10.2.19 Let K be an oriented simplicial complex and G be an an abelian

group. For p ≥ 0, a p-chain b p ∈ C p (K ; G) is called a p-dimensional boundary or a
p-boundary on K if there is a ( p + 1)-chain c p+1 ∈ C p+1 (K ; G) such that ∂(c p+1 ) =
b p . The set of all p-boundaries is the homomorphic image ∂(C p+1 (K ; G)) and is a
subgroup of C p (K ; G), called the p-dimensional boundary group of K , denoted by
B p (K ; G).
Remark 10.2.20 Let K be an oriented n-dimensional simplicial complex. Then there

are no p-chains on K for p > n. As a result C p (K ; G) = 0 for p > n. Thus there
exists no (n + 1)-chain on K and hence Cn+1 (K ; G) = 0; so Bn (K ; G) = 0.
Definition 10.2.21 Let K be an oriented complex. If p is a positive integer, a p-

dimensional cycle or p-cycle on K is a p-chain denoted by z p such that ∂ p (z p ) = 0.
The set of all p-cycles is denoted by Z p (K ; G).
Proposition 10.2.22 Z p (K ; G) is a subgroup of C p (K ; G), and B p (K ; G) is a

subgroup of Z p (K ; G) for each p such that 0 ≤ p ≤ n, where n is the dimension
of K .
Proof It follows trivially from Theorem 10.2.16. ❑
Remark 10.2.23 The subgroup Z p (K ; G) is the kernel of the homomorphism
∂ p : C p (K ; G) → C p−1 (K ; G).
Definition 10.2.24 The group Z p (K ; G) is called the p-dimensional cycle group of

K . The group Z 0 (K ; G) of 0-cycles is the group C0 (K ; G) of 0-chains.
Remark 10.2.25 Intuitively, a p-cycle on K is a linear combination of p-simplexes

which makes a complete circuit. The p-cycles enclosing ‘holes’ of K are those cycles
which are not boundaries of ( p + 1)-chains. A p-cycle which is the boundary of a
( p + 1)-chain was called by Poincaré ‘a cycle homologous to zero’.
Definition 10.2.26 Two p-cycles c p and d p on an oriented complex K are said to

be homologous, denoted by c p ∼ d p if there is a ( p + 1)-chain e p+1 such that
∂(e p+1 ) = c p − d p .
If a p-cycle f p is the boundary of a ( p + 1)-chain, f p is said to be homologous

to zero, denoted by f p ∼ 0.
Remark 10.2.27 As the relation of being homologous for p-cycles is an equivalence

relation, it partitions Z p (K ; G) into homology classes
[z p ] = {c p ∈ Z p (K ; G) : c p ∼ z p }.
The homology class [z p ] is the coset
z p + B p (K ; G) = {z p + ∂(c p+1 ) : ∂(c p+1 ) ∈ B p (K ; G)}.
Definition 10.2.28 For p ≥ 1, Z p (K ; G) = ker ∂ p is called the group of p-cycles

of K with coefficients in G.
Remark 10.2.29 Since we take the boundary of every 0-chain to be 0, we define

0-cycle to be 0-chain. Thus for p = 0, Z 0 (K ; G) of 0-cycles is the group C0 (K ; G)
of 0-cycles.
Definition 10.2.30 If p ≥ 0, the image Im ∂ p+1 is a subgroup of C p (K ; G) and is

called the group of p-dimensional boundaries or p-boundaries of C p (K ; G), denoted
by B p (K ; G).
Remark 10.2.31 For any chain complex C p (K ; G), the group of boundaries B p is a
subgroup of the group of cycles Z p by Theorem 10.2.16. For the converse if B p ⊂ Z p
for all p, then the corresponding sequence of groups and their homomorphisms is a
chain complex, i.e., ∂ p ◦ ∂ p+1 = 0
Definition 10.2.32 Let K be an oriented simplicial complex and G be an abelian

group. Then the simplicial homology group of the corresponding chain com-
plex C(K ; G), denoted by H p (K ; G) is the quotient group ker ∂ p /Im ∂ p+1 , i.e.,
H p (K ; G) = Z p /B p .
Remark 10.2.33 The homology classes [z p ] are actually members of the simplicial
homology group H p (K ; G).
Remark 10.2.34 For p < 0 or p > dim K , we take C p (K ; G) = 0. Hence H p
(K ; G) = 0 for all such p. This group H p (K ; G) is sometimes called an ‘absolute’
simplicial homology group.
The following natural questions arise:
Do the homology groups Hn (K ; G) depend on the choice of an orientation of K ?
Is it possible for C p (K ) to express any property which remains unchanged under
homeomorphism.
To solve such problems consider two copies K 1 and K 2 of the given simplicial
complex K endowed with distinct orientations.
Consider the map ψ : C(K 1 ; G) → C(K 2 ; G),

+s, if the two orientations of s coincide

ψ(s) =
−s, otherwise.
Then ψ is an isomorphism. In other words, given K and an abelian group G,

C(K , G) is uniquely determined up to isomorphism and hence Hn (K ; G) is uniquely
determined up to isomorphism.
More precisely, we prove that
Theorem 10.2.35 Let K 1 and K 2 denote the same simplicial complex K endowed
with distinct orientations. Then given an abelian group G, H p (K 1 ; G) ∼
= H p (K 2 ; G)
for all p ≥ 0.
Proof Let p ≥ 0 be an integer. For a p-simplex s p of K , let 1 s p and 2 s p denote
the p-simplex in the oriented complex K 1 and K 2 respectively. Define a function α
on the simplexes of K such that α(s p ) = ±1 and 1 s p = α(s p )· 2 s p . Then define a
sequence ψ = {ψ p } of homomorphisms

ψ p : C p (K 1 ; G) → C p (K 2 ; G), gi · 1 s ip → α(s ip )gi · 2 s ip ,
1 i
where gi · s p is a p-chain on K 1 . Hence ψ p is well defined. Clearly, the dia-
gram in Fig. 10.3 is commutative. Hence it follows that if z p ∈ Z p (K 1 ; G), then
∂ p ψ p (z p ) = ψ p−1 ∂ p (z) = 0 shows that ψ p (z p ) ∈ Z p (K 2 ; G). This means that ψ p
maps a p-cycle of K 1 into a p-cycle of K 2 . Similarly, ψ p maps a p-boundary
of K 1 into a p-boundary of K 2 . Consequently, ψ p induces a homomorphism
Fig. 10.3 Diagram ψp

/
Cp (K1 ; G) Cp (K2 ; G)
involving ψ p and ∂ p
∂p ∂p
ψp−1
Cp−1 (K1 ; G) / Cp−1 (K2 ; G)
ψ p∗ : H p (K 1 ; G) → H p (K 2 ; G), [z p ] → [ψ p (z p )],
for all homology classes [z p ] ∈ H p (K 1 ; G). Reversing the roles of K 1 and K 2 we

have a sequence {φ p } of homomorphisms
φ p : C p (K 2 ; G) → C p (K 1 ; G)
such that
φ p ◦ ψ p = 1d (identity automorphism of C p (K 1 ; G)) (10.8)
and
ψ p ◦ φ p = 1d (identity automorphism of C p (K 2 ; G)) (10.9)
(10.8) shows that (φ p ◦ ψ p )∗ = φ p∗ ◦ ψ p∗ = identity 1d and

(10.9) shows that (ψ p ◦ φ p )∗ = ψ p∗ ◦ φ p∗ = identity 1d . Consequently, ψ p∗ : H p
(K 1 ; G) → H p (K 2 ; G) is an isomorphism of groups. ❑
Remark 10.2.36 The structure of 0-dimensional homology group H0 (K ; G) shows
whether the polyhedron |K | is connected: there is no torsion in dimension 0 and the
rank of the free abelian group H0 (K ; G) is the number of components of |K |.
10.2.2 Induced Homomorphism and Functorial

Properties of Simplicial Homology
This subsection associates to every simplicial map a homomorphism on the corre-

sponding simplicial homology groups and presents functorial properties of simplicial
homology.
Definition 10.2.37 Let K and L be oriented simplicial complexes and f : K → L
be a simplicial map. For each p ≥ 0, define
f : C p (K ) → C p (L), v0 , v1 , . . . , v p → f (v0 ), . . . , f (v p ).
If some f (vi ) is repeated, then the term on the right is zero.

Proposition 10.2.38 If f : K → L is a simplicial map, then
f : C∗ (K ) → C∗ (L)
is a chain map, which means that f ◦ ∂ = ∂ ◦ f .

Proof It follows from the definitions of f and ∂. ❑
Proposition 10.2.39 If f : K → L is a simplicial map, then the induced homomor-
phism
f ∗ : Hn (K ) → Hn (L), z + Bn (K ) → f (z) + Bn (L)
is a homomorphism of groups.
Proposition 10.2.40 Chain complexes and chain maps form a category under usual
composition. This category is denoted by Comp.
Proof The objects here are taken chain complexes and the morphisms are taken
chain maps. The composition of chain maps is defined coordinatewise: {gn } ◦ { f n } =
{gn ◦ f n }. Hence they form a category. ❑
Proposition 10.2.41 Hn is a covariant functor from the category Comp of chain
complexes and chain maps to the category Ab of abelian groups and homomorphisms
for each n ∈ Z. Moreover, Hn is a topological invariant.
Proof The object function is here defined by assigning to each chain complex the
sequence of its homology groups, and morphism function is defined by assigning
to each chain map f between chain complexes the induced map f ∗ between their
homology groups. This shows that for each n, Hn : Comp → Ab is a covariant functor
which is a topological invariant by Proposition 10.1.12. ❑
10.2.3 Computing Homology Groups of Polyhedra
The subsection considers the problem: how to relate C p (K ; G), Z p (K ; G), and
B p (K ; G) to the topological spaces whose triangulation is K ? Let X be a polyhedron
and G be an abelian group. For calculation of the homology groups we use the
following steps:
Step 1: Triangulate X .
Step 2: Choose an orientation for the simplicial complex K thus obtained by trian-
gulation.
Step 3: Calculate the chain group Cn (K ; G).
Step 4: Describe the boundary homomorphisms ∂n .
Step 5: Calculate the groups of cycles Z n (K ; G).
Step 6: Calculate the groups of boundaries Bn (K ; G).
Step 7: Calculate the quotient group Hn (K ; G) = Z n (K ; G)/Bn (K ; G).
Example 10.2.42 Consider the simplicial complex K having only one vertex v with
Z as the coefficient group. As there is only one possible orientation on K and with
that orientation, C p (K ; Z) = 0 for all p = 0 and C0 (K ; Z) is the free abelian group
on the single generator v. Hence

0, p = 0
H p (K ; Z) =
Z, p = 0.
Example 10.2.43 Let S 2 be the 2-sphere. Then H2 (S 2 ; G) = G. Consider the sim-

plicial complex K consisting of all 2-simplexes, 1-simplexes and 0-simplex that are
faces of a single 3-simplex s3 , where s3 is not in K . Geometrically, this is the surface
of a tetrahedron and this surface is homeomorphic to S 2 . The simplex S 2 is precisely
the two skeletons of the complex K . We orient the complex K by chosing a fixed
ordering of its vertices: a0 < a1 < a2 < a3 . It will induce the positive orientation of
the simplexes. In this way, we have the following oriented simplexes of K :
1-simplexes: +s11 = a2 a3 , +s12 = a1 a3 , +s13 = a0 a3 , +s14 = a1 a2 , +s15 =
a0 a2
2-simplexes: s21 = a1 a2 a3 , +s22 = a0 a2 a3 , +s23 = a0 a1 a3 , +s24 = a0 a1 a2 .
Then the only 2-cycles on S 2 are the chains of the form
g·s21 − g·s22 + g·s23 − g·s24 , g ∈ G. Hence Z 2 (S 2 ; G) ∼
= G. Since there are no 3-
simplexes in S 2 , the chain C3 (S 2 ; G) = 0 and hence ∂C3 (S 2 ; G) = 0 gives B2
(S 2 ; G) = 0. Consequently, H2 (S 2 ; G) = Z 2 (S 2 ; G) ∼
= G.
Remark 10.2.44 The homology groups of any polyhedron do not depend on any
particular choice of its triangulation. Because if K and L are two triangulations of
a polyhedron X , then given an abelian group G, H p (K ; G) ∼ = H p (L; G) for each
p ≥ 0.
Remark 10.2.45 The functor H p measures the number of ‘ p-dimensional holes’

in the simplicial complex, in the sense that the p-sphere S p has exactly one p-
dimensional hole and no m-dimensional hole if m = p. A 0-dimensional hole is a pair
of points in different path components and hence H0 measures path connectedness.
10.3 Relative Simplicial Homology Groups
This section extends the concept of absolute simplicial homology groups to the con-
cept of relative simplicial homology groups. If K is an oriented simplicial complex
and L is a subcomplex, then L is also oriented in the induced orientation, by the
partial order on Vert(L) inherited from that on Vert(K ).
Definition 10.3.1 Let L be a subcomplex of an oriented simplicial complex K . The
relative chain group C p (K , L) of the pair (K , L) is defined to be the free abelian
group freely generated by by all p-simplexes with interiors in K − L .
Definition 10.3.2 For p ≥ 1, the boundary operator ∂ p : C p (K , L) → C p−1 (K , L)

is defined by the formula (10.5), where the summation is taken over all those ( p − 1)-
simplexes s ip−1 of K whose interiors do not intersect L.
Proposition 10.3.3 ∂ p ◦ ∂ p+1 = 0
Proof It is left as an exercise. ❑

10.3 Relative Simplicial Homology Groups 363
Definition 10.3.4 The relative chain groups C p (K , L) and the operators ∂ p form a
chain complex C(K , L), called the relative chain complex of the pair (K , L).
Definition 10.3.5 A simplicial map of pairs f : (K , L) → (K 1 , L 1 ) is a simplicial

map f : K → K 1 such that f (L) ⊂ L 1 .
Definition 10.3.6 The group of cycles Z p (K , L) is the kernel of the homomor-

phism ∂ p and the group of boundaries B p (K , L) is the Im ∂ p+1 in the chain complex
C(K , L).
Definition 10.3.7 The relative homology group H p (K , L) is the quotient group

Z p (K , L)/Bk (K , L).
Remark 10.3.8 H p satisfies the functorial properties.
Proposition 10.3.9 Let (K , L) and (K 1 , L 1 ) be two oriented simplicial pairs. Then

any simplicial map f : (K , L) → (K 1 , L 1 ) induces homomorphisms
f ∗ : H p (K , L) → H p (K 1 , L 1 ) such that
(a) if g : (K 1 , L 1 ) → (K 2 , L 2 ) is a simplicial map, then (g ◦ f )∗ = g∗ ◦ f ∗ : H p
(K , L) → H p (K 2 , L 2 );
(b) if 1(K ,L) : (K , L) → (K , L) is the identity map, then 1(K ,L)∗ : H p (K , L) →
H p (K , L) is the identity automorphism.
Corollary 10.3.10 H p is a covariant functor from the category of all relative sim-
plicial chain complexes and chain maps to the category of abelian groups and homo-
morphisms.
Proof It follows from Proposition 10.3.9. ❑
Remark 10.3.11 The relative homology groups H p (K , ∅) of an oriented simplicial

complex K with respect to empty set ∅ coincide with the absolute ones, i.e., H p (K )
and H p (K , ∅) are always isomorphic.
Theorem 10.3.12 Let the simplicial complexes K 1 and K 2 intersect along a simpli-
cial complex K 3 , which is a subcomplex in both K 1 and K 2 . Then the embedding i
of the pair (K 1 , K 3 ) to the pair (K 1 ∪ K 2 , K 2 ) induces isomorphisms, for all p ≥ 0,
i ∗ = ψ p : H p (K 1 , K 3 ) → H p (K 1 ∪ K 2 , K 2 ).
Proof The chain complexes of the pairs (K 1 , K 3 ), (K 1 ∪ K 2 , K 2 ) coincide, because

they are constructed over the same set of simplexes. Hence the theorem follows. ❑
10.4 Exactness of Simplicial Homology Sequences
This section conveys the relations between absolute simplicial homology groups of
simplicial chain complexes and the relative simplicial homology groups of relative
simplicial chain complexes using the language of exact sequences and shows that
the relative simplicial homology groups H p (K , L) for any pair (K , L) of simplicial
complexes fit into a long exact sequence.
We recall that a sequence of groups and homomorphisms
f n+1 fn
· · · → Cn+1 −−−−−→ Cn −−−−→ Cn−1 → · · · (10.10)
is called exact if ker f n = Im f n+1 for all n.

This definition shows that any exact sequence of groups of the form (10.10) is
a chain complex. Since ker f n = Im f n+1 , all the homology groups of that complex
are trivial. The converse result is also true: any chain complex with trivial homology
groups is an exact sequence. This shows that the homology groups of a chain complex
give a measure of its inexactness in some sense.
The definition of relative chain complex C(K , L) of the pair (K , L) of simplicial
complexes shows that the sequence
i∗ j∗
→ C(L) −−−→ C(K ) −−−→ C(K , L) −
0− →0 (10.11)
with homomorphism i∗ induced by the embedding i of L in K and the homomorphism

j∗ obtained by forgetting those the simplexes of K that are contained in L, is exact.
Theorem 10.4.1 Let C, C1 and C2 be chain complexes related by the short exact
sequence
i j
0−
→ C1 −−−→ C −−−→ C2 −
→0 (10.12)
Then there are homomorphisms ∂ making the long sequence of homology groups
i∗ j∗ ∂
· · · → H p (C1 ) −−−→ H p (C) −−−→ H p (C2 ) −−−→ H p−1 (C1 ) → · · · (10.13)
exact.
Proof Left as an ecxercise. ❑
An immediate consequences of Theorem 10.4.1 is that the relative homology

groups H p (K , L) for any pair (K , L) of simplicial complexes fit into a long exact
sequence.
Theorem 10.4.2 (Exact sequence of the pair) For any pair (K , L) of simplicial
complexes, the sequence of the homology groups of these complexes and the relative
homology groups of the pair (K , L)
10.4 Exactness of Simplicial Homology Sequences 365
i∗ j∗ ∂
· · · → H p (L) −−−→ H p (K ) −−−→ H p (K , L) −−−→ H p−1 (L) → · · ·
is exact.
Proof Since the chain complexes C(K ), C(L), C(K , L) form a short exact sequence,
the theorem follows from Theorem 10.4.1. ❑
Theorem 10.4.3 If K 1 and K 2 are subcomplexes of a simplicial complex K such

that K = K 1 ∪ K 2 , then there exist homomorphisms
∂ : H p (K 1 ∪ K 2 ) → H p−1 (K 1 ∩ K 2 )
such that the homology groups of the complexes L = K 1 ∩ K 2 , K 1 , K 2 and K form

the long exact sequence
i∗ j∗ ∂
· · · → H p (L) −−−→ H p (K 1 ) ⊕ H p (K 2 ) −−−→ H p (K ) −−−→ H p−1 (L) → · · ·
Remark 10.4.4 In addition to the long exact sequence of homology groups for the
pair (K , L) of simplicial complexes, there is another long exact sequence, known as
Mayer–Vietoris sequence, which is convenient at many situations.
Theorem 10.4.5 (Mayer–Vietoris) If K 1 and K 2 are subcomplexes of a simplicial
complex K such that K 1 ∪ K 2 = K , then there is an exact sequence
∂ i∗ j∗ ∂
· · · → H p+1 (K ) −−−→ H p (K 1 ∩ K 2 ) −−−→ H p (K 1 ) ⊕ H p (K 2 ) −−−→ H p (K ) −−−→ H p−1 (K 1 ∩ K 2 )
10.5 Simplicial Cohomology Theory: Introductory

Concepts
This section conveys the concept of simplicial cohomology theory. A cohomol-

ogy group is the dual to homology group. Given an oriented simplicial complex K ,
cochain complex, cocycle, coboundary, cohomology class and the cohomology group
are defined dually. In contrast with homology theory, a cohomology theory is a con-
travariant functor. In some sense, these two theories are adjoint (or dual) to each other.
Historically, J.W. Alexander and A. Kolmogoroff invented simplicial cohomology
in 1935 independently. E.Čech and H. Whitney developed it during 1935–1940. The
terms ‘coboundary’, ‘cocycle’, ‘cochain,’ and ‘cohomology’ were given by E.Čech.
The advantage of cohomology over homology is that the cohomology group admits
an additional structure making it a ring.
Definition 10.5.1 A sequence C ∗ = {C n , δ n }, n ∈ Z of additive abelian groups C n

together with a sequence of group homomorphisms δ n : C n−1 → C n , such that
δ n+1 ◦ δ n = 0, ∀ n ∈ Z, is called a cochain complex and δ n is called a coboundary
homomorphism.
More precisely,
δn δ n+1
C∗ : · · · −
→ C n−1 −−−→ C n −−−−−→ C n+1 −
→ ··· (10.14)
is called a cochain complex if δ n+1 ◦ δ n = 0 ∀ n ∈ Z.
Definition 10.5.2 The elements of Z n = ker δ n+1 are called n-cocycles and the ele-
ments of B n = Im δ n are called n-coboundaries of the cochain complex C ∗ (10.14).
Proposition 10.5.3 Z n and B n form groups for all n for the cochain complex C ∗
(10.14).
Proof It follows from respective definitions. ❑
Proposition 10.5.4 B n is a subgroup of Z n for all n for the cochain complex C ∗

(10.14).
Proof It follows from the property of the cochain complex C ∗ (10.14) that δ n+1 ◦
δ n = 0. ❑
Definition 10.5.5 The quotient group Z n /B n for any cochain complex C ∗ (10.14),
denoted by H n (C ∗ ) (or simply H n ), is called the n-dimensional cohomology group
of the cochain complex C ∗ .
For an oriented simplicial complex K , we define cohomology groups of K as

follows:
Definition 10.5.6 Let K be an oriented simplicial complex. Then the cohomology

groups of the corresponding cochain complex C ∗ (K ) are called the cohomology
groups of K and are denoted by H n (K ).
Definition 10.5.7 Let C ∗ = {C n , δ n } and C ∗ = {C n , δ n } be two cochain com-

plexes. Then a sequence f = { f n : C n → C n }, n ∈ Z is called a cochain map from
C ∗ to C ∗ if the diagram in Fig. 10.4
commutes,.i.e., f n+1 ◦ δ n = δ n ◦ f n holds for all n.
Fig. 10.4 Cochain map / Cn δn / C n+1 δ n+1 / C n+2 / ···

···
fn f n+1 f n+2
δn
δ n+1

··· / C n / C n+1 / C n+2 / ···
10.5 Simplicial Cohomology Theory: Introductory Concepts 367
Each simplicial map f : K → K between simplicial complexes induces homo-

morphisms:
f ∗ : H n (K ) → H n (K )
of the cohomology groups.
Proposition 10.5.8 A simplicial cohomology theory H ∗ = {H n , δ n } is a sequence of

contravariant functors from the category of simplicial complexes and the simplicial
maps to the category of abelian groups and homomorphisms.
Proof Each simplicial map f : K → K between simplicial complexes induces

homomorphisms
f ∗ : H n (K ) → H n (K )
of their cohomology groups and these homomorphisms act in the opposite direction.
Hence it is easy to prove the theorem. ❑
Remark 10.5.9 The homology groups and cohomology groups of a simplicial com-
plex are closely related. If we know homology groups of K , we can find the coho-
mology groups of K and conversely, provided the corresponding chain complex is
finitely generated and free.
Recall that for a finitely generated abelian group G, if F and T represent free part
and torsion part of G respectively, then G is always isomorphic to the group F ⊕ T
(Adhikari and Adhikari 2014).
Theorem 10.5.10 For the simplicial homology groups Hn (C) and the cohomolgy
groups H n (C) of any finitely generated free chain complex C,
(i) Free part of H n (C)= Free part of Hn (C);
(ii) Torsion part of H n (C)= Torsion part of Hn−1 (C);
(iii) H n (C) is isomorphic to the direct sum
F ⊕ T , where F is the free part of Hn (C) and T is the torsion group of Hn−1 (C).
10.6 Simplicial Cohomology Ring
This section defines cup product of cochains with an eye to endow the direct sum of
all the cohomolgy groups of a simplicial complex (with coefficients in a commutative
ring) a ring structure. This algebraic structure given independently by Alexander and
Kolmogroff in 1935 by defining a product, now called cup product of cochains, has
wide applications in algebraic topology. But this does not fit for homolgy groups.
This is the advantage of cohomolgy theory over homology theory.
Let K be simplicial complex and : |K | → |K | × |K | be the diagonal map

defined by (x) = (x, x). It induces homomorphisms ∗ : H n (K × K ) → H n (K ).
We are now in a position to define the cohomology product.
Definition 10.6.1 Let x ∈ H n (K ), y ∈ H m (K ) be two elements of cohomology

groups of a simplicial complex K . Then their cup product x ∪ y is defined by the
rule x ∪ y = ∗ i(x ⊗ y), where x ⊗ y defines embedding

i: H n (K ) ⊗ H n (K ) → H n+m (K × K ).
n+m
This definition shows that to find the product of two elements of cohomology
groups we first consider their tensor product as an element of the corresponding
cohomology group of the direct product K × K and take its image under the homo-
morphism ∗ induced by the diagonal map.
Let R be a ring. The cup product is associative and distributive, it is natural to
try to make the cup product the multiplication in a ring structure on the cohomology
groups of a simplicial complex K . If we define H ∗ (K ; R) to be the direct sum of
the groups H n (K ; R), then the elements of H ∗ (K ; R) are the finite sum xi with
xi ∈ H i (K ; R) and the product of two such sums is defined to be
⎛ ⎞

xi ⎝ yj⎠ = xi y j .
i j i, j
This makes H ∗ (K ; R) into a ring with identity if R has an identity.
Example 10.6.2 For the real projective plane R P 2 , H ∗ (R P 2 , Z2 ) consists of all the
polynomials a0 + a1 x + a2 x 2 with coefficients a0 , a1 , a2 ∈ Z2 and hence
H ∗ (R P 2 , Z2 ) is the quotient ring Z2 [x]/< x 3 >.
Remark 10.6.3 Geometrically, the cup product on manifolds is interpreted as

‘intersection numbers’, see Dold (1972, Chap. 7), Munkres (1984, Chap. 8).
10.7 Singular Homology
This section presents singular homology using the algebraic properties of singular
chain complexes. Singular homology theory generalizes the simplicial homology
theory. The former is easier to work while the latter is easier to visualize geometri-
cally. These two theories are related by the basic result that the singular homology of
a polyhedron is isomorphic to the simplicial homology of any of its triangulated sim-
plicial complexes. Again for any pointed topological space X , the homotopy groups
πn (X ) are very important invariants. It is easy to define πn (X ) but difficult to com-
pute them. Only for a few CW -complexes their homotopy groups are known. So this
section defines singular homology groups Hn (X ) which are different invariants from
10.7 Singular Homology 369
simplicial homology groups. The algebraic properties of singular chain complexes

lead to singular homology theory.
10.7.1 Singular Homology Groups
This subsection defines singular homology groups (modules). We have defined in

Chap. 6 the standard n-simplex
n = {(t0 , t1 , . . . , tn ) ∈ Rn+1 : 0 ≤ ti ≤ 1, ti = 1} ⊂ Rn+1 .
It has vertices v0 = (1, 0, . . . , 0), v1 = (0, 1, 0, . . . , 0), . . . , vn = (0, 0, . . . , 0, 1)

in the space Rn+1 . In particular, it defines orientation of n . The simplex n has the
ith face
n−1 (i) = {(t0 , tn , . . . , tn ) : ti = 0}.
It is a standard (n − 1)-simplex in the space Rn .
Definition 10.7.1 A singular n-simplex of the space X is a continuous map f :

n → X .
Definition 10.7.2 A singular n-chain is a finite linear combination n i f i , where

each f i : n → X is a singular n-simplex, n i ∈ Z. The group of n-chains Cn (X ) is
a free abelian group generated by all singular n-simplexes of X .
Definition 10.7.3 The boundary homomorphism is defined by
∂ : Cn (X ) → C n−1 (X ), f → i=0
n
(−1)i ei ( f ),
where ei is given by ei ( f ) = f |n−1 (i) , the restriction of f on the ith face n−1 (i).
Lemma 10.7.4 The composite ∂ 2 = ∂ ◦ ∂ = 0.
Proof It is sufficient to prove that the composite homomorphism
∂n+1 ∂n
Cn+1 (X ) −−−−−→ Cn (X ) −−−−→ Cn−1 (X )
is trivial, i.e., Im ∂n+1 ⊂ ker ∂n . Clearly it follows from the definition:

e j−1 (ei ( f )), for j > 1,

(ei ◦ e j )( f ) = ei (e j ( f )) =
e j (ei+1 ( f )), for j ≤ 1 ❑
The remaining part of the proof is left as an exercise.

Proposition 10.7.5 The complex C(X ) defined by
∂n+1 ∂n ∂1
· · · −→ Cn+1 (X ) −−−−−−→ Cn (X ) −−−−→ C n−1 (X ) −→ · · · −→ C 1 (X ) −−−−→ C0 (X ) −→ 0.
(10.15)
is a chain complex.
Proof It follows from the Lemma 10.7.4. ❑
Definition 10.7.6 The group Z (X ) = ker ∂n is called the group of cycles, and the
group Bn (X ) = Im ∂n+1 is called the group of boundaries.
Proposition 10.7.7 Bn (X ) is a subgroup of Z n (X ).
Proof It follows from the Lemma 10.7.4 that Bn (X ) is a subgroup of Z n (X ). ❑
Definition 10.7.8 The nth homology group (module) Hn (X ) of X is defined by

Hn (X ) = ker ∂n /Im ∂n+1 , i.e., Hn (X ) = Z n (X )/Bn (X ).
Remark 10.7.9 Convention: The group H0 (X ) = C0 (X )/Im ∂1 and Hn (X ) = 0 for

n < 0.
Definition 10.7.10 Let c1 , c2 ∈ Cn (X ). Then the chain c1 is said to be chain homo-

topic to the chain c2 if c1 − c2 = ∂n+1 (d) for some d ∈ Cn+1 (X ). The class [c]
(under the chain homotopy relation)∈ Hn (X ) is called a homological class of the
cycle c ∈ Cn (X ).
Remark 10.7.11 Chain homotopic maps induce the same homomorphism on homol-
ogy.
Remark 10.7.12 The group Hn (X ) is an abelian group which is a module. If it is

finitely generated, then
Hn (X ) ∼
= Z ⊕ Z ⊕ · · · ⊕ Z ⊕ Zn 1 ⊕ · · · ⊕ Zn t (10.16)
Definition 10.7.13 The rank of Hn (X ) is the number of Z’s in the decomposi-

tion (10.16) and it is called the Betti number of the space X . In symbol, βn (X ) =
rank(Hn (X )).
Remark 10.7.14 The Betti numbers are named after E. Betti (1823–1892) and gen-
eralize the connectivity number which he used while studying curves and surfaces.
Euler characteristic of X is closely related to its Betti number βn (X ) (see Sect. 10.17).
Remark 10.7.15 Hn (X ) = Hn (C(X )).
Definition 10.7.16 Let g : X → Y be a continuous map. Then g induces the homo-

morphism g : Cn (X ) → Cn (Y ), which maps a singular simplex f : n → X to
a singular simplex g ◦ f : n → Y . It defines a homomorphism g∗ : Hn (X ) →
Hn (Y ), called the homomorphism induced by g.
Theorem 10.7.17 Let f g : X → Y . Then f ∗ = g∗ : Hn (X ) → Hn (Y ) for all

n ≥ 0.
Proof As f g : X → Y , there exists a homotopy H : X × I → Y such that

H (x, 0) = f (x) and H (x, 1) = g(x). Then for any singular simplex s : n → X ,
we have a map H ◦ (s × I ) : n × I → Y . The cylinder n × I has a canonical sim-
n+1 (i), i = 0, 1, . . . , n
plicial structure. We divide n × I into (n + 1)-simplices
as follows:
n+1 (i) = {(t0 , t1 , . . . , tn , α) ∈ n × I : t0 + · · · + ti−1 ≤ α ≤ t0 + · · · + ti }.

Then the map G : H ◦ (s × I ) : n × I → Y defines (n + 1)-singular simplices of

n
dimension (n + 1). We now define k as k(s) = (−1)i G| n+1 (i) .
i=0
Clearly, the homomorphisms
kn : Cn (X ) → C n+1 (Y ), n i si → n i kn (si ), n i ∈ Z
define a chain homotopy k : C(X ) → C(Y ). Hence the theorem follows. ❑
Corollary 10.7.18 Homotopic maps induce the same homomorphism in homology

groups.
Corollary 10.7.19 Homotopy equivalent spaces have isomorphic homology groups.
Proof Let X and Y be two homotopy equivalent spaces. Then Hn (X ) ∼

= Hn (Y ) for
all n ≥ 0 by Corollary 10.7.18. ❑
Corollary 10.7.20 If X is contractable, then Hn (X ) = 0 for all n > 0.
Proof Since X has the homotopy type of one-point space. The corollary follows
from Corollary 10.7.19 and dimension axiom (Exercises 6 of Sect. 10.21). ❑
Remark 10.7.21 The characteristic of homology groups is a homotopy invariant and

hence it is a fundamental feature of homology groups.
Theorem 10.7.22 Let X be a topological space and A be a subspace of X . Then
the sequence of homology groups
i∗ j∗ ∂ i∗
··· −
→ Hn (A) −−−→ Hn (X ) −−−−→ Hn (X, A) −−−→ Hn−1 (A) −−−→ Hn−1 (X ) −→ · · ·
is exact.

10.7.2 Reduced Singular Homology Groups
This subsection defines reduced singular homology groups (modules) H ∗ (X ) on

a nonempty topological space X and shows that reduced homology modules are
completely determined in terms of singular homology modules of X and vice-versa.
Let S∗ (X ) be the singular chain complex of X and C p be an abstract chain complex
such that C 0 = R, the R-module R and C p = 0 for all p = 0. Then there is a chain
map
ri , if p = 0
f : S∗ (X ) → C∗ , ri xi →
0, if p > 0.
Then f is an onto map and ker f is a sub-chain complex of S∗ (X ), written as

S∗ (X ),
called the reduced singular chain complex of X . As S∗ (X ) is a chain complex of
free R-modules, the augmented map f splits and hence S0 (X ) = S0 (X ) ⊕ R and
S p (X ) =
S p (X ) for all p > 0.
Definition 10.7.23 The reduced p-dimensional homology module of a nonempty

p (X ) is defined to be the homology module of the chain
space X , denoted by H

complex S∗ (X ).
As f (Im ∂1 ) = 0, it follows that Im ∂1 ⊂ ker f =

S0 (X ). Hence
S0 (X )
S0 (X )
H0 (X ) = = 0 (X ) ⊕ R,
⊕R=H
Im ∂1 Im ∂1
p (X )
H p (X ) = H
for all p > 0

p = H p except H
Remark 10.7.24 The relation H 0 ⊕ R = H0 shows that the reduced
singular homology modules are completely characterized in terms of the reduced
singular homology modules and play an important role in computing homology
groups.
Remark 10.7.25 If G is the coefficient group of a homology theory H∗ , then for the
unique map f from a topological space X to a one-point space, the kernel of the
homomorphism
f ∗ : H0 (X ) → G
0 (X )
is H
10.7.3 Relative Singular Homology Groups
This subsection generalizes the concept of singular homology groups of a nonempty

space defined in Sect. 10.7.1 by introducing the concept of relative singular homology
groups in the following way:
Let X be a space and A be a subspace of X . Then Cn (A) ⊂ Cn (X ), and
∂n (Cn (A)) ⊂ C n−1 (A) and each generator of the group Cn (A) maps to a genera-
tor of the group C n (X ). Let Cn (X, A) be the quotient group Cn (X )/C n (A), which
is a group of relative n-chains of the space X modulo the subspace A. The group
Cn (X, A) is a free abelian group with generators g : n → X, f (n ) ∩ (X \A) = ∅.
Since the boundary operator ∂n : Cn (X ) → C n−1 (X ) takes Cn (A) to C n−1 (A), it
induces a quotient boundary map ∂n : Cn (X, A) → Cn−1 (X, A). Hence it produces
a sequence of boundary maps by varying n
∂n ∂n−1 ∂2 ∂1
· · · −→ Cn (X, A) −−−−→ C n−1 (X, A) −−−−−−→ · · · −−−−−→ C1 (X, A) −−−−→ C0 (X, A) −→ 0.
Just as before, a calculation shows that ∂ 2 = 0. Hence {Cn (X, A), ∂n } forms a chain
complex.
Definition 10.7.26 The resulting homology groups Hn (X, A) are called relative sin-
gular homology groups of (X, A).
Remark 10.7.27 By considering the definition of the relative boundary map ∂, we
see:
(i) elements of Hn (X, A) are represented by relative cycles:
n-chains α ∈ C n (X ) are such that ∂α ∈ Cn−1 (A);
(ii) a relative cycle α is trivial in Hn (X, A) off it is a relative boundary, i.e.,
α = ∂β + γ for some β ∈ Cn+1 (X ) and γ ∈ C n (A).
These properties present the intuitive idea precisely that Hn (X, A) is ‘homology
of X modulo A’.
Theorem 10.7.28 (Exact sequence of the pairs of spaces) Let (X, A) be a pair of
spaces. Then the sequence of homology groups
∂ i∗ j∗ ∂ i∗
· · · −−−→ Hn (A) −−−→ Hn (X ) −−−→ Hn (X, A) −−−→ Hn−1 (A) −−−→ · · ·
is exact, where i : A → X and j : X → (X, A) are inclusion maps.

Proof See Rotman (1988, pp. 96). ❑
Proposition 10.7.29 (Homotopy property for pairs of spaces) Let f, g : (X, A) →
(Y, B) be two maps homotopic through maps of pairs (X, A) → (Y, B) of spaces.
Then f ∗ = g∗ : Hn (X, A) → (Y, B).
10.8 Eilenberg–Zilber Theorem and Künneth Formula
This section gives Eilenberg–Zilber theorem and Künneth formula which are used
for computing homology or cohomolgy of product spaces.
10.8.1 Eilenberg–Zilber Theorem
This subsection gives Eilenberg–Zilber Theorem.

Theorem 10.8.1 For topological spaces X and Y , there is a (natural) chain equiv-
alence
ψ : C∗ (X × Y ) → C ∗ (X ) ⊗ C∗ (Y )
which is unique up to chain homotopy. Moreover,
Hn (X × Y ) ∼
= Hn (C∗ (X ) ⊗ C∗ (Y ))
for all n ≥ 0.
10.8.2 Künneth Formula
This subsection gives Künneth formula which gives a split exact sequence with
middle term as given in Theorem 10.8.2.
Theorem 10.8.2 (Künneth formula) For every pair of topological spaces X and Y
and for every integer n ≥ 0,

Hn (X × Y ) ∼
= Hi (X ) ⊗ H j (Y ) ⊕ Tor (H p (X ), Hq (Y ))
i+ j=n p+q=n−1
Remark 10.8.3 (Original version of Künneth formula) If X and Y are compact poly-
hedra, then
bn (X × Y ) = bi (X )b j (Y ),
i+ j=n
where bi (X ) is the ith Betti number of X . It follows from the Theorem 10.8.2. Because
for any f ·g (finitely generated) abelian groups A and B, the group Tor (A, B) is finite
and hence it has no contribution to the calculation of the Betti numbers.
10.8 Eilenberg–Zilber Theorem and Künneth Formula 375
Example 10.8.4 For positive integers m, n and m = n,

Z, if p = 0, m, n, m + n
H p (S × S ; Z) ∼
m n
=
0, otherwise.
If m = n, then
⎧
⎪
⎨Z, if p = 0, 2n
n n ∼
H p (S × S ; Z) = Z ⊕ Z, if p = n
⎪
⎩
0, otherwise.
For m = n = 1, this gives the homology groups of the torus S 1 × S 1 .
Example 10.8.5 If X = S 1 ∨ S 2 ∨ S 3 (wedge), then

Z, if p = 0, 1, 2, 3
H p (X ; Z) ∼
=
0, otherwise.
Remark 10.8.6 It follows from Examples 10.8.4 and 10.8.5 that the spaces
S 1 × S 2 and S 1 ∨ S 2 ∨ S 3 are the same homology groups, but they are not homotopy
equivalent
10.9 Singular Cohomology
This section introduces the concept of singular cohomology which is an algebraic

variant of homology. The basic difference between them is that cohomology groups
are contravariant functors but homology groups are covariant functors. The homology
groups determine the corresponding cohomology groups but its converse is true if
the homology groups are finitely generated. The cohomology groups (modules or
rings) of a topological space were not recognized until late 1930, when S. Lefschetz
formulated a simplified proof of the duality theorem for orientable n-manifold with
boundary.
Definition 10.9.1 Given a topological space X and an abelian group G, the singular
n-cochain group C n (X ; G) with coefficients in G is defined to be the dual group
given by C n (X ; G) = Hom (Cn (X ; G), G) of the singular chain group C n (X ; G).
Remark 10.9.2 An n-cochain α ∈ C n (X ; G) assigns to each n-simplex σ : n → X

a value α(σ) ∈ G. Since the singular n-simplexes form a basis of C n (X ), these
values can be assigned arbitrarily. Hence n-cochains are precisely the functions from
singular n-simplices to G. Again C n (X ; G) is isomorphic to the direct product of as
many copies of G as there are n-simplexes in X .
Definition 10.9.3 Let G be an abelian group. Given a cochain α ∈ C n (X ; G), the

element α : C n (X ; G) → G is a a homomorphism. Then the coboundary map δ :
C n+1 → G is defined by δ(α) = α ◦ ∂, which is the composite map
∂ α
Cn+1 (X ; G) −−−→ Cn (X ; G) −−−→ G.
This implies that for a singular (n + 1)-simplex

σ : n+1 → X, δα(σ) = (−1)α (σ|v0 ···v̂i ···vn+1 ),
i
where ‘hat’ symbol over vi indicates that this vertex is deleted from the sequence
v0 , v1 , . . . , vn .
Clearly, δ 2 = δ ◦ δ = 0.
Definition 10.9.4 The cohomology group H n (X ; G) with coefficient group G is
defined to be the quotient group ker δ/Im δ at C n (X ; G) in the cochain complex
δn δ n−1
· · · ←− C n+1 (X ; G) ←−−−− C n (X ; G) ←−−−−−− C n−1 (X ; G) ←− · · · ←− C 0 (X ; G) ←− 0
The group of n-cocycles is ker δ n and is denoted by Z n (X ; G) and the group of n-

boundaries is Im δ n−1 and is denoted by B n (X ; G).
Definition 10.9.5 An element of H n (X ; G) is a coset β + B n (X ; G), where β is an

n-cocycle. This is called the cohomology class of β, denoted by [β] ∈ H n (X ; G).
Remark 10.9.6 For a cochain α, δ ◦ α = α ◦ δ = 0. Hence α vanishes on n bound-

aries.
10.10 Relative Cohomology Groups
The relative cohomology groups H n (X, A; G) for a pair (X, A) with coefficient
group G are defined by dualizing the short exact sequence
i j
0−
→ C n (A; G) −−−→ C n (X ; G) −−−→ Cn (X, A; G) −
→0
and by applying Hom (−, G) functor to obtain
i j
− C n (A; G) ←−−−− C n (X ; G) ←−−−− C n (X, A; G) ←
0← −0
where by definition C n (X, A; G) = Hom (Cn (X, A), G). This sequence is exact.
10.10 Relative Cohomology Groups 377
Relative coboundary operators
δ : C n (X, A; G) → C n+1 (X, A; G)
are defined by restriction of absolute δ’s, and hence relative cohomology groups
H n (X, A; G) are obtained.
The maps i and j commute with δ, since i and j commute with δ. Hence the
short exact sequence of cochain groups is part of short exact sequence of cochain
complexs, which give rise to an associated long exact sequence of cohomology
groups
j∗ i∗ δ
··· −
→ H n (X, A; G) −−−−→ H n (X ; G) −−−−→ H n (A; G) −−−→ H n+1 (X, A; G) −
→ ···
10.11 Hurewicz Homomorphism
This section establishes a close connection between homotopy and homology

groups of a certain class of topological spaces through Hurewicz homomorphism
defined by Hurewicz (1904–1956) in 1935 in his paper (Hurewicz 1935). He first
asserted that for a simplicial pair (K , L) if πr (K , L) = 0 for 1 ≤ r < n(n ≥ 2), then
πr (K , L) → Hr (K , L) is an isomorphism (Original version of Hurewicz theorem).
This paper cast light for the first time onto the relationship between homological and
homotopical invariants. A series of four papers of Hurewicz published during 1935–
1936, has greatly influenced the development the modern homotopy theory. In one
sense homology is an approximation to homotopy. For more results see Sects. 17.1.2
and 17.1.3 of Chap. 17.
The original version of Hurewicz theorem is subsequently refined and is now
studied.
Definition 10.11.1 Let X be a pointed topological space with a base point x0 ∈ X

and sn be the standard generator of Hn (S n ), n = 1, 2, . . . , If f : S n → X repre-
sents an α ∈ πn (X, x 0 ), then the induced homomorphism f ∗ : Hn (S n ) → Hn (X )
defines an element f ∗ (sn ) ∈ Hn (X ). Set h(α) = f ∗ (sn ). Since h(α) ∈ Hn (X ) does
not depend on f , define a correspondence h : πn (X, x 0 ) → Hn (X ), α → h(α), n =
1, 2, . . . . Then h is a homomorphism, called the Hurewicz homomorphism.
Theorem 10.11.2 (Hurewicz) Let (X, x0 ) be a pointed topological space, such that
π0 (X, x0 ) = 0, π1 (X, x0 ) = 0, . . . , πn−1 (X, x0 ) = 0, where n ≥ 2. Then
H1 (X ) = 0, H2 (X ) = 0, . . . , Hn−1 (X ) = 0,
and the Hurewicz homomorphism h : πn (X, x 0 ) → Hn (X ) is an isomorphism.

Fig. 10.5 Rectangular h /

πn (∨Sαn ) Hn (∨Sαn )
diagram related to h (γβ )∗ (γβ )∗

h
πn (Sβn ) / Hn (Sβn )
Proof By Ex. 15 of sect. 8.9 of Chap. 8, there exists a CW -complex Y and a weak
homotopy equivalence f : Y → X . Then f induces an isomorphism in homology
groups. So without loss of generality we may assume that X is a CW -complex.
This means by the given conditions of the theorem that X is an (n − 1)-connected
CW -complex. Then up to homotopy equivalence X may be chosen so that X has a
single 0-cell, and it does not have any cells of dimensions 1, 2, . . . , n − 1. Hence
H1 (X ) = 0, H2 (X ) = 0. The nth skeleton X (n) is a wedge of
0, . . . , Hn−1 (X ) =
(n)
spheres, i.e., X = Sα . Let gα : Sα →
n n
Sαn be the embedding of the αth sphere,
α α

n
and let kβ : S n → be the attaching maps of the (n + 1)-cells eβn+1 . Then the maps
α
gα determine the generators of the group πn (X (n) ). Let γβ ∈ πn (X (n) ) be the element
determined by the maps kβ .
The first nontrivial homotopy group πn (X ) is given as the factor group of the
homotopy group πn (X (n) ) ∼
= Z ⊕ · · · ⊕ Z by the subgroup generated by γβ . The
cellular chain group

C(X ) = Hn (X (n) ) = Hn ( Sαn ) and Hn (X ) = C(X )/Im ∂n+1 .
α
Since the Hurewicz homomorphism h : πn (S n ) → Hn (S n ) is an isomorphism,

the diagram in Fig. 10.5 is commutative, where the horizontal homomorphisms are
isomorphisms. Hence h induces an isomorphism πn (X, x0 ) → Hn (X ). ❑
Corollary 10.11.3 Let X be a simply connected topological space, and H1

(X ) = 0, H2 (X ) = 0, . . . , Hn−1 (X ) = 0. Then π1 (X ) = 0, π2 (X ) = 0, . . . , πn−1
(X ) = 0 and the Hurewicz homomorphism h : πn (X, x0 ) → Hn (X ) is an isomor-
phism.
Definition 10.11.4 A pointed topological space X with base point x0 is said to be

n-connected if πi (X, x 0 ) = 0, for all i ≤ n.
Remark 10.11.5 0-connected space means path-connected and 1-connected means

simply connected.
Corollary 10.11.6 If X is a path-connected topological space, then there is an

epimorphism
h 1 : π1 (X ) → H1 (X ; Z)
which induces an isomorphism
h 1∗ : π1 (X )/ ker h 1 → H1 (X ; Z)
10.11 Hurewicz Homomorphism 379
with ker h 1 the commutator subgroup of π1 (X ).
Corollary 10.11.7 If X is path-connected, then h 1 is an isomorphism iff the group

π1 (X ) is abelian.
Example 10.11.8 If X is the figure-eight, then its fundamental group is the free group
on two generators by van Kampen theorem. Hence by Corollary 10.11.7, H1 (X ; Z)
is the free abelian group on 2 generators (i.e., Z × Z).
Theorem 10.11.9 If X is an n-connected topological space with n ≥ 2, then

H q (X ) = 0 for all q ≤ n, and the Hurewicz map h : πn+1 (X ) → H
n+1 (X ) is an
isomorphism.
Proof It follows from Theorem 10.11.2 and Corollary 10.11.3. ❑
Remark 10.11.10 A complete proof can be found in Spanier (1966), Maunder

(1970).
Corollary 10.11.11 πn (S n ) ∼
= Z.
Proof Since S n is (n − 1)-connected, πn (S n ) ∼

= Z. ❑
Remark 10.11.12 An equivalent formulization of Hurewicz theorem with its gener-

alization has been discussed in Chap. 17.
10.12 Mayer–Vietoris Sequences
This section introduces the concepts of Mayer–Vietoris sequences in singular and

simplicial homology theories. Let A, B be subspaces of a topological space X .
Mayer–Vietoris Sequence prescribes a method to compute the singular homology
groups of A ∪ B if we know the homology groups of A, B, and A ∩ B.
10.12.1 Mayer–Vietoris Sequences in Singular Homology

Theory
This subsection displays Mayer–Vietoris sequences in singular homology theory.

Definition 10.12.1 Let X be topological space. For a pair of subspaces A, B ⊂
X such that X = Int (A) ∪ Int (B), in addition to the long exact sequence of the
homology groups of the pair (X,A) of spaces, there is an exact sequence of the form
∂ i∗ j∗ ∂
· · · −−−→ Hn (A ∩ B) −−−−→ Hn (A) ⊕ Hn (B) −−−−→ Hn (X ) −−−→ Hn−1 (A ∩ B) −→
· · · −→ H0 (X ) −→ 0 (10.17)
The sequence (10.17) is called a Mayer–Vietoris sequence in singular homology.

Let A, B be two subspaces of X such that A ∩ B = ∅ and A ∪ B = Int (A) ∪
Int (B). Then the sequence
··· −
→H n (A ∩ B) −
→H n (A) ⊕ H
(B) −
→H n (A ∪ B) −
→H n−1 (A ∩ B) −
→
(10.18)
is exact and is called reduced singular Mayer–Vietoris exact sequence for reduced
singular homology.
Remark 10.12.2 There is another form of Mayer–Vietoris sequence which is some-

times more convenient to apply.
Definition 10.12.3 Let X 1 and X 2 be two subspaces of X such that X is the union of
interiors of X 1 and X 2 . If f i : X 1 ∩ X 2 → X i and gi : X i → X are inclusion maps
for i = 1, 2, define
φ : Hn (X 1 ∩ X 2 ) → Hn (X 1 ) ⊕ Hn (X 2 ), α → ( f 1∗ (α), f 2∗ (α))
ψ : Hn (X 1 ) ⊕ Hn (X 2 ) → Hn (X ), (α1 , α2 ) → g1∗ (α) − g2∗ (α)
Then there exists a long exact sequence
φ ψ
· · · → Hn+1 (X ) −−−→ Hn (X 1 ∩ X 2 ) −−−→ Hn (X 1 ) ⊕ Hn (X 2 ) −−−→ Hn (X ) −−−→ Hn (X 1 ∩ X 2 ) −
→ ···
(10.19)
The sequence (10.19) is also called the Mayer–Vietoris sequence and the homomor-
phisms are called the connecting homomorphisms.
Remark 10.12.4 The sequence (10.18) can be viewed as an analog of the von
Kampen theorem, since if A ∩ B is path-connected, then H1 terms of this sequence
gives an isomorphism
ψ : H1 (X ) → H1 (A) ⊕ H1 (B)/Im f.
It is the abelianized statement of the von Kampen theorem, and H1 is the abelianized
of π1 for the path-connected spaces.
10.12.2 Mayer–Vietoris Sequences in Simplicial Homology

Theory
If K 1 , K 2 are any two subcomplexes of a simplicial complex K , then the following

sequence
10.12 Mayer–Vietoris Sequences 381
∂ i∗ j∗ ∂
· · · −−−→ Hn (K 1 ∩ K 2 ) −−−→ Hn (K 1 ) ⊕ Hn (K 2 ) −−−−→ Hn (K 1 ∪ K 2 ) −−−→ Hn−1 (K 1 ∩ K 2 ) −
→ ···
(10.20)
is exact, and is called Mayer–Vietoris sequence in simplicial homology.
Example 10.12.5 If X = S n with A = E n+ and B = E n− , the north and south hemi-

sphere of S n , then A ∩ B = S n−1 . Consequently, in the reduced Mayer–Vietoris
sequence (10.18), the terms
n (B) = 0
i (A) ⊕ H
H
give isomorphisms
i (S n ) ∼
H =H i−1 (S n−1 ).
Example 10.12.6 (One-point union of n-cells) Let X = X 1 ∪ X 2 ∪ · · · ∪ X n be the

one-point union of n-spaces X i , each of which is homomorphic to S 1 . Then X has
a triangulation, which is homeomorphic to the union of n triangle Ti , T2 , . . . , Tn all
having one vertex v (say) is common. Then
⎧
⎪
⎨Z, if p = 0
H p (X 1 ∪ X 2 ∪ · · · ∪ X n ; Z) ∼
= ⊕n Z, if p = 1
⎪
⎩
0, otherwise.
10.13 Computing Homology Groups
The section computes homology groups of some spaces.
10.13.1 Homology Groups of a One-Point Space
Let P be a one-point space. Then there is a unique map f n : n → P for any n. Hence
the chain complex corresponding to the point P which is viewed as a 0-dimensional
simplex, is C n (P) = Z for all n ≥ 0.
Clearly,
Z, if n = 0
Hn (P; Z) ∼=
0, otherwise
Definition 10.13.1 A topological space X having the same homology groups of a

one- point space is called an acyclic space.
Example 10.13.2 Every contractible space X is acyclic with


Z , if n = 0
Hn (X ; Z) ∼
=
0, otherwise.
10.13.2 Homology Groups of CW-complexes
The main aim of this subsection is to develop a technique to compute homology

groups of CW -complexes. The singular chain complex is too large to compute. We
construct here a cellular chain complex E(X ) which is smaller than C(X ). We start
into computations of homology groups of spheres. The reduced homology groups
H n (X ) are defined viz the augmented chain complex with coefficients Z:
ES
→ C0 (X ) −−−−→ Z −
··· − →0
with E S defined by summing coefficients.

n Z, if p = n
Theorem 10.13.3 (Reduced Homology groups of spheres) H p (S ) =
0, otherwise .
Proof Consider a long exact sequence of the pair of spaces (D n , S n−1 ):
··· −
→H p (D n ) −
→ H p (D n , S n−1 ) −
→H p (S n−1 ) −
→H p−1 (D n ) · · · −
→.
Hence H p−1 (D n ) = 0. Consequently, H p (D n , S n−1 ) ∼

p (D n ) = 0, H =H p (S n−1 ).
The theorem follows by induction on n. ❑
n+1 ( X ) ∼
Theorem 10.13.4 Let X be a C W -complex. Then H =H n (X ) for each n.
Proof X = C + X ∪ C − X described as in Fig. 10.6.

We now consider a long exact sequence in homology for the pair of spaces
(C + X, X ):
··· −
→H n (C + X ) −
→ Hn (C + X, X ) −
→H n (X ) −
→H n−1 (C + X ) −
→ ···
∗ (C + X ) = 0. Consequently, Hn (C + X, X )
Since the cone C + X is contractible, H
∼
=H n (X ). Again since (C X, X ) is always a Borsuk pair, it follows that
+
Fig. 10.6 Diagram for

Suspension
C +X
shaded
C +X
C −X
X
10.14 Cellular Homology 383
Hn (C + X, X ) ∼
=H n (C + X/ X ) ∼
=H n (C + X ∪ C −1 X ) = H
n ( X ).
10.14 Cellular Homology
This section presents cellular homology which is a homology for the category of CW -
complexes. It agrees with singular homology, and can provide an effective means for
computing homology groups. This section introduces cellular homology theory that
reflects the cellular structure of a C W -complex X. This theory is most suitable for
computing homology groups of CW -complexes. Given a CW -complex decomposi-
tion E of a CW -complex X , a chain complex is defined whose group of n-chains,
is a free abelian group for each n ≥ 0, whose rank is the number of n-cells in E.
The cellular chain complex is defined directly in terms of singular homology groups.
Cellular homology is a homology functor from the category of C W -complexes and
cellular maps to the category of abelian groups and homomorphisms. On the category
of C W -complexes there is a natural equivalence from the cellular homology to the
singular homology.
Definition 10.14.1 Let X be a cellular space and n ≥ 0. Define Wn (X ) = Hn (X (n) ,

X n−1 )(singular homology) and dn : Wn (X ) → Wn−1 (X ) as the composite dn = j∗ ◦
∂ as shown in Fig. 10.7.
where
j : (X (n−1) , ∅) → (X (n) , X (n−1) )
is the inclusion map and ∂ is the connecting homomorphism arising from the long
exact sequence of the pairs (X (n) , X (n−1) ).
Proposition 10.14.2 For a cellular space, (W∗ (X ), d) is a chain complex.

Proof It is sufficient to show that dn ◦ dn+1 = 0. Clearly, dn ◦ dn+1 is the composite.
Hn+1 (X (n+1) , X (n) ) −

→ Hn (X (n) ) −
→ Hn (X (n) , X (n−1) ) −
→ Hn−1 (X (n−1) ) −
→ Hn−1 (X (n−1) , X (n−2) ).
This is zero, since the middle two arrows are adjacent arrows is the long exact
sequence of the pair (X (n) , X (n−1) ). ❑
Fig. 10.7 Commutative Hn (X (n) , X (n−1)

P
)
PPP
triangle involving ∂, j∗ and PPP
PPP
dn PPPdn
∂ PPP
PPP
PPP
PPP
P(
Hn−1 (X (n−1)) /H (X (n−1) , X (n−2) )
j∗ n−1
Definition 10.14.3 If X is a cellular space and n ≥ 0, then its cellular homology

group is defined as usual
Hn (W∗ (X )) = ker dn / Im dn+1
Remark 10.14.4 Cellular homology in algebraic topology is a homology for the

category of CW -complexes. It agrees with singular homology, and can provide an
effective means for computing homology groups.
Remark 10.14.5 One sees from the cellular chain complex that the n-skeleton X (n)
determines all lower-dimensional homology modules:
Hk (X ) ∼
= Hk (X (n) ) for k < n.
Remark 10.14.6 An important consequence of this cellular chain complex is that
if a CW -complex has no cells in consecutive dimensions, then all of its homology
modules are free. For example, the complex projective space CPn has a cell structure
with one cell in each even dimension; it follows that for 0 ≤ k ≤ n, H2k (CPn ; Z) ∼
=Z
and H2k+1 (CPn ; Z) = 0.
Remark 10.14.7 Generalization The Atiyah-Hirzebruch spectral sequence is the

analogous method of computing the (co)homology of a CW -complex, for an arbitrary
extraordinary (co)homology theory.
10.15 Čech Homology and Cohomology Groups
This section introduces Čech homology and cohomology groups. The homology
theory constructed by Čech is called Čech homology theory after his name. Čech
homology group of X with coefficient group G is denoted by Ȟi (X ; G). This theory
defines homology invariants on topological spaces which are more general than
polyhedra. The Čech cohomology theory is dual to the Čech homology theory.
In simplicial homology theory, certain topological properties of a polyhedra |K |
are expressed in algebraic terms. The algebraic information is obtained through
the arrangement of the complex K into simplexes as faces of each other. In Čech
homology theory corresponding to every finite open cover U of a topological space
X , there is assigned a simplicial complex.
Definition 10.15.1 Let U be a finite open cover of a topological space X . The nerve
of U denoted by N (U), is the abstract complex whose vertices are members of U and
whose simplexes are those subfamilies of U which have a nonempty intersection.
Example 10.15.2 The n-dimensional simplexes are those subfamilies of U with

(n + 1) elements which have a nonempty intersection.
Remark 10.15.3 The term ‘the complex U’ is used in place of ‘the nerve defined by
the finite open cover U’.
10.15 Čech Homology and Cohomology Groups 385
A homology theory attains its full height only if it is defined for a pair of topological
spaces. Let (X, A) be a topological pair and U be a cover of X . If U is a cover of A,
then (U, U ) is a pair of complexes but it may not be a simplicial pair, because U is
not in general a subcomplex of U. So we assume that (X, A) is a compact pair.
Let (X, A) be a compact pair and U be an open cover of X . Then there is a sim-
plicial pair (U, U A ), where U A is subfamily of U consisting of those sets which meet
A. Then this pair defines a homology group Ȟ ((U, U A ); G), called Čech homology
group of (X, A) with coefficient in G, where G is either a topological group or a
vector space over a field.
Definition 10.15.4 Let U = {Uα } be an open covering of X and V = {Vβ } be a
refinement of U in the sense that Vβ is contained in some Uα . Then these inclusions
induce a simplicial map
N (V ) → N (V ),
which is unique up to homotopy. The direct limit group lim H i (N (U); G) with
→
respect to finer and finer open cover U is called the Čech cohomologyindex vCech
cohomology group Ȟ (X ; G).
Remark 10.15.5 Relative Čech cohomology groups are defined in a way analogous
to the Definition 10.15.4.
For full exposition of Čech homology and cohomology groups for an arbitrary
pair (X, A) over a coefficient group G see Eilenberg and Steenrod (1952).
10.16 Universal Coefficient Theorem for Homology

and Cohomology
This section studies universal coefficient theorem for homology and cohomology
theories. The basic need for such study comes from the fact that homology and
cohomolgy theories with coefficients in different abelian groups are frequently con-
venient than the corresponding theories in integral coefficients. For example, in the
group Z2 the elements 1 and −1 coincide. This shows that there is no need to consider
orientations of simplexes but it is simplicial to consider unoriented complexes. This
makes many definition simpler. If the coefficient group is a field of characteristic 0
such as field R or field Q, then there is no torsion and any homology group has the
form F ⊕ F ⊕ · · · ⊕ F, which is completely determined by its rank.
10.16.1 Homology with Arbitrary Coefficient Group
This subsection discusses homology groups Hn (X ; G) with an arbitrary coefficient

group G (an abelian group), which is a natural generalization of Hn (X ) = Hn (X ; Z).
It is sometimes gained by this generalization. It has been working so far with homol-
ogy groups of chain complexes in which the chain groups are free abelian groups.
a topological space X , each element of S(X ) is a formal linear combi-
Thus given
nation m i si , where each si is a singular simplex and m i ∈ Z. Given an abelian
i
group G. it is sometimes helpful to make a generalization in which m i ∈ G. The new
born complex is as usual denoted S(X ; G) and the corresponding homology groups
Hn (X ; G) are called the homology group of X with coefficients in G. In this sense
Hn (X ; G) is a generaization of Hn (X ) = Hn (X ; Z).
There is a natural question: how are the homology groups with coefficients in
an arbitrary abelian group G and those with coefficients in Z related. Universal
coefficient theorem gives its answer.
Theorem 10.16.1 (Universal coefficient theorem for homology)

(i) For any space X and any abelain group G, there are exact sequences for all
n ≥ 0:
α
→ Hn (X ) ⊗ G −−−→ Hn (X ; G) −
0− → Tor (Hn−1 (X ; G) −
→0 (10.21)
where α : Hn (X ) ⊗ G −
→ Hn (X ; G), [z]‘ ⊗ g → [z ⊗ g].
(ii) The sequence (10.21) splits:
Hn (X ; G) ∼
= Hn (X ) ⊗ G ⊕ Tor (Hn−1 (X ); G) (10.22)
where ‘⊗’ denotes the usual tensor product of two groups and ‘ Tor’ denotes the
usual torsion product of two abelian groups.
Corollary 10.16.2 If Tor (Hn−1 (X ), G) vanishes in (10.22), then
Hn (X ) ⊗ G ∼
= Hn (X ; G).
Proof See Rotman (1988, pp. 264) ❑
Theorem 10.16.3 (Universal Coefficient Theorem for simplicial homology) Let K

be an oriented simplicial complex. Then for any integer n,
Hn (K ; G) ∼
= Hn (K ; Z) ⊗ G ⊕ Tor (Hn−1 (K ; Z) ∗ G.
Proof It follows from Theorem 10.16.1(ii). ❑
Remark 10.16.4 It is sometimes used to write simply Hn (X ) in place of Hn (X ; Z),

which is a special case of Hn (X ; G).
As the groups Hn (X ; G) are completely determined by the groups Hn (X ) (see

Universal Coefficient Theorem) this generalization cannot offer new information
10.16 Universal Coefficient Theorem for Homology and Cohomology 387
about X . But one can gain in this generalization: H∗ (X ; G) may be easier sometimes
to handle than H∗ (X ). For example, if K is a simplicial complex and G = F is a field,
then Hn (K ; F) is a finite dimensional vector space for each n over F, and hence it
is determined up to isomorphism by dimension of this vector space. Another gain in
this case is that the homorphisms induced by continuous maps in the corresponding
homology groups are linear maps of vector spaces which are utilized to obtain many
interesting results.
Another convenience is for example, in the group Z2 the elements 1 and −1 coin-
cide. This implies that there is no need of keeping track of orientations of simplexes
and unoriented complexes need to be considered.
Remark 10.16.5 There is a natural relation between homology groups with coeffi-
cients in an arbitrary abelian group G and homolgy groups with integral coefficients
given by Universal Coefficient Theorem.
10.16.2 Universal Cohomology Theorem for Cohomology
We now give the universal coefficient theorem for cohomology corresponding to the
universal coefficient theorem of homology.
Definition 10.16.6 A chain complex C∗ is said to be of finite type if each of its terms
C n is finitely generated.
Definition 10.16.7 A topological space X is said to be of finite type if each of its

homology groups Hn (X ) is finitely generated.
Example 10.16.8 (i) Every compact polyhedra in a space of finite type;

(ii) Every compact CW -complex is a space of finite type;
(iii) R P ∞ is a space of finite type but it is not compact.
Theorem 10.16.9 (Universal coefficient theorem for cohomology)

(i) Let X be a topological space of finite type and G be an abelian group. There is
an exact sequence for every n ≥ 0;
α
0−
→ H n (X ) ⊗ G −−−→ H n (X ; G) −
→ Tor (H n+1 (X ), G) −
→ 0. (10.23)
Here α : H n (X ) ⊗ G −→ H n (X ; G), [z] × g → [zg], where zg : σ → z(σ)g

for an n-simplex σ in X (as z(σ) ∈ Z, z(σ)g has a meaning).
(ii) The sequence (10.23) splits:
H n (X ; G) ∼
= H n (X ) ⊗ G ⊕ Tor (H n+1 (X ), G).

10.17 Betti Number and Euler Characteristic
This section studies Betti number and Euler characteristic of a polyhedron, which
are closely related and revisits Euler characteristic from the viewpoint of homology
theory. The Swiss mathematician Leonhard Euler gave a formula for comparing
geometrical objects mathematically which relates the number of vertices V , the
number of edges E, and the number of faces F of a polyhedron in an alternating
sum V − E + F = 2 of a 3-dimensional polyhedron P. This result was given by
Euler in 1752. Poincaré gave first application of his homology theory, which is a
generalization of Euler formula to general polyhedra. The characteristic χ(P) of P
is defined by χ(P) = V − E + F.
Definition 10.17.1 If G = R in the homology group Hn (K ; G) of a finite simplicial
complex K with coefficient group G, then the group Hn (K ; G) is a real vector space.
If its dimension is q, then q is called the called the qth Betti number of K , denoted
by βq .
Euler characteristic of a topological space is also a topological invariant readily
computable by ‘polyhedronization’ of the space. Homology groups are refinements
of the Euler characteristic in some sense. Euler characteristic of a space X is an
integer. It is different from other topological invariants such as compactness or con-
nectedness which reflects geometrical properties of X . In this section we study Euler
characteristic using vector spaces, graph theory, and algebraic topology.
10.17.1 Euler Characteristics of Polyhedra
Euler characteristic is a numerical invariant which can be used to distinguish topolog-

ically nonequivalent spaces. The search of other invariants has established a remark-
able connection between two branches of modern mathematics: topology and modern
algebra. There are many algebraic invariants associated with topological spaces: most
commonly, we associate a group to a space, in such a way that topologically equiva-
lent spaces have isomorphic groups. Historically, Euler’s theorem asserts: if K is any
polyhedron homeomorphic to S 2 , with V vertices, E edges, and F 2-dimensional
faces, then V − E + F = 2.
Definition 10.17.2 The Euler characteristic of a simplicial complex K of dimension
n is defined by the alternative sum
χ(K ) = α0 (K ) − α1 (K ) + α2 (K ) + · · · + (−1)n αn (K )

n
= (−1)i αi (K ),
i=0
where αi (K ) is the number of simplexes of dimension i in K .

10.17 Betti Number and Euler Characteristic 389
Fig. 10.8 Cube
Fig. 10.9 Tetrahedron
Definition 10.17.3 Let X be a subset of R3 , which is homeomorphic to a polyhedron

K . Then this Euler characteristic χ(X ) of X is defined by
χ(X ) = (number of vertices in K)-(number of edges in K) + (number of faces in K).
Example 10.17.4 The Euler characteristic of the cube as shown in Fig. 10.8 is given
by
χ(cube) = 8 − 12 + 6 = 2.
Example 10.17.5 The Euler characteristic of the tetrahedron X as shown in Fig. 10.9
is given by
χ(X ) = 4 − 6 + 4 = 2.
Poincaré-Alexander theorem ensures that χ(X ) is independent of the polyhedron

K provided K is homeomorphic to X .
Theorem 10.17.6 (Pincaré-Alexander) The Euler characteristic is independent of
the polyhedron K as long as K is homeomorphic to X .
Proof See Nakahara (2003). ❑
Example 10.17.7 (i) If X is a one-point space, then χ(X ) = 1.
(ii) If X is a line, then χ(X ) = 2 − 1 = 1, since a line has two vertices and an edge.
(iii) If X is a triangular disk, then χ(X ) = 3 − 3 + 1 = 1.
(iv) The simplest polyhedron which is homeomorphic to S 1 has three edges of a
triangle. Then χ(S 1 ) = 3 − 3 = 0.
(v) The simplest polyhedron which is homeomorphic to S 2 has the faces of a
tetrahedron in Fig. 10.9. Then χ(S 2 ) = 4 − 6 + 4 = 2.
Theorem 10.17.8 Let X and Y be two geometrical objects in R3 . If X is homeo-
morphic to Y , then χ(X ) = χ(Y ). In particular, if χ(X ) = χ(Y ), then X cannot be
homeomorphic to Y .
Example 10.17.9 S 1 cannot be homeomorphic to S 2 , since χ(S 1 ) = 0 but

χ(S 2 ) = 2.
Remark 10.17.10 Two figures, which are not homeomorphic to each other may have
the same Euler characteristic.
For example, χ(X ) = 1 = χ(l), where l is a line and if X = R P 2 is the real

projective plane, then the Euler characteristic χ(R P 2 ) = 1. But l and R P 2 are not
of the same homotopy type. There is a general result which ensures that its converse
is true.
Proposition 10.17.11 If X and Y are two geometric objects such that they are of
the same homotopy type, then χ(X ) = χ(Y ).
10.17.2 Euler Characteristic of Finite Graphs
This subsection studies Euler characteristic of a finite graph.
Definition 10.17.12 Let X be a finite graph. The Euler characteristic χ(X ) is defined
to be the number of vertices minus the number of edges.
Theorem 10.17.13 Let X be a finite graph. Then χ(X ) = 1 if X is a tree and the
rank of π1 (X ) is 1 − χ(X ) if X is connected.
10.17.3 Euler Characteristic of Graded Vector Spaces
Definition 10.17.14 Let {Vn } be a given graded vector space with Vn = 0 for only
finitely many values of n. Euler characteristic χ({Vn }) of {Vn } is defined to be the
alternating sum (−1)n dim Vn , i.e.,

χ({Vn }) = (−1)n dim Vn .
n
Example 10.17.15 If {Vn , ∂n } is a chain complex with {Vn } as above, then
χ({Vn }) = χ(H ({Vn , ∂n })).

10.17 Betti Number and Euler Characteristic 391
10.17.4 Euler–Poincaré Theorem for Finite C W -complexes
This subsection gives the Euler–Poincaré theorem and studies Euler characteristic
of a finite C W -complex by generalizing the celebrated formula: number of vertices
− number of edges + number of faces for 2-dimensional complexes. Betti number
and Euler characteristic have a close relation. For certain spaces such as spherical
complexes, the homology groups with integral coefficients are finitely generated.
Definition 10.17.16 For any topological space X , if the homology group Hn (X ) is

finitely generated, then its rank is called the nth ‘Betti number’ of X , denoted by
βn (X ).
Remark 10.17.17 Given a topological space X , its nth Betti number βn (X ) is the
number of free generators of Hn (X ). Thus βn (X ) is the number of copies of Z in
Hn (X ). The Euler characteristic χ(X ) of X is defined as the alternating sum of its
Betti numbers:
∞
χ(X ) = (−1)n βn (X ).
n=0
Definition 10.17.18 Let X be a finite CW -complex.

Then the Euler characteristic
χ(X ) is defined to be the alternating sum (−1)n cn , where cn is the number of
n
n-cells of X .
Remark 10.17.19 This χ(X ) can be defined purely in terms of homology of X and
hence depends only on the homotopy type of X . This shows that χ(X ) is independent
of the choice of CW -structure of X . The rank of the finitely generated abelian group
is the number of Z summands when the group is expressed as a direct sum of cyclic
groups.
Theorem 10.17.20 Let X be a finite CW-complex. Then the Euler characteristic

χ(X ) = (−1)n rank Hn (X ).
n
To prove this theorem we use the fact that if
0−
→A−
→B−
→C −
→0
is a short exact sequence of finitely generated abelian groups, then rank B = rank
A+ rank C.
dm d1
Proof Let 0 −→ Cm −−−−→ C m−1 − → ··· −
→ C1 −−−→ C 0 − → 0 be a chain com-
plex of finitely generated abelian groups with cycles Z n = ker dn , boundaries
Bn = Im dn+1 , and homology Hn = Z n /Bn . Then we have short exact sequences
0−
→ Zn −
→ Cn −
→ Bn−1 −
→0
→ Bn −
and 0 − → Zn −
→ Hn −
→ 0.
Consequently, it follows that rank Cn = rank Z n + rank Bn−1 and rank Z n =

rank Bn + rank Hn .
Hence it implies that

(−1)n rank Cn = (−1)n rank Hn .
n n

Consequently, χ(X ) = (−1)n rank Hn (X ). ❑
n
Corollary 10.17.21 Let X be a finite CW-complex. Then the Euler characteristic

χ(X ) is a homotopy invariant.
Proof Let X be an n-dimensional C W -complex. Then by Theorem 10.17.20

n
χ(X ) = (−1)r rank Hr (X ).
r =0
As each Hr (X ) is a homotopy invariant, it follows that χ(X ) is a homotopy invariant.

❑
Theorem 10.17.22 (The Euler–Poincaré theorem) Let K be an oriented simplicial

complex of dimension n and for q = 0, 1, 2, . . . , n, let αq denote the number of
q-simplexes of K . Then
n n
(−1)q αq = (−1)q βq ,
q=0 q=0
where βq denotes the qth Betti number of K .
Proof The theorem follows from Theorem 10.17.20 and Definition 10.7.13. ❑
10.18 Cup and Cap Products in Cohomology Theory
This section conveys the basic concepts of cup and cap products. A basic property
of cohomology which distinguishes from homology is the existence of a natural
multiplication called cup product. This product makes the direct sum of all coho-
mology groups into a graded ring. This product is used to study ‘duality’ theorem
on manifolds.
Recall that given a topological space X and an abelian group G, the singular n-
cochain group C n (X ; G) with coefficients in G is defined to be the dual group given
10.18 Cup and Cap Products in Cohomology Theory 393
by C n (X ; G) = Hom (Cn (X ), G) of the singular chain group Cn (X ; G). Instead of

G, here we take commutative ring R with identity 1.
10.18.1 Cup Product
This subsection introduces the concept of cup product in cohomology theory.

Definition 10.18.1 Given a topological space X and a commutative ring R with 1, let
C n (X ; R) = Hom (Cn (X ; R), R) and C ∗ (X ; R) = C n (X ; R). If ψ ∈ C n (X ; R)
n≥0
and c ∈ C n (X ; R), their associated element in R denoted by [c, ψ] is defined by
setting [c, ψ] = ψ(c) ∈ R. In particular, if c ∈ cn+1 (X ), then [c , δ(ψ)] = [∂c , ψ].
Definition 10.18.2 Let σ be any singular m + n simplex given by σ = v0 , . . . vn . . .
vm+n . Consider the affine maps
λn : n → m+n , ρm : m → m+n
given by λn = v0 . . . vn and λm = vn . . . vm+n . Given ψ ∈ [C n (X ; R)] and θ ∈

C m (X ; R), their associated element in R denoted by [σ, ψ ∪ θ] is defined by setting
[σ, ψ ∪ θ] = [σλn , ψ] · [σρm , θ] (10.24)
where the right hand multiplication in (10.24) is the usual multiplication of scalars
already defined in R.
Definition 10.18.3 Let X be a topological space and R be a commutative ring. If
ψ ∈ C n (X ; R) and θ ∈ C m (X ; R), then their cup product ψ ∪ θ ∈ C m+n (X ; R) is
defined by
[σ, ψ ∪ θ] = [σλn , ψ][σρm , θ]
for every singular (m + n)-simplex σ in X , where a right hand multiplication is the

usual product of two elements in the ring R.
Remark 10.18.4 The relation in (10.24) shows that ψ operates on the front n-face of
σ and θ operates on the back n-face of σ and the result is obtained by multiplication
in R.

Remark 10.18.5 If ψ = ψi , θ = θ j are arbitrary elements of C ∗ (X ; R), then
i j
the cup product defines a function

C ∗ (X ; R) × C ∗ (X ; R) → C ∗ (X ; R), ( ψi ) ∪ ( θj) = ψi ∪ θ j ,
i j i, j
where ψi ∈ C i (X ; R) and θ j ∈ C j (X ; R).

Proposition 10.18.6 Letn X be a topological space and R be a commutative ring.

Then C ∗ (X ; R) = C (X ; R) is a graded ring under the cup product.
n≥0
Proof Let ψ ∈ C n (X ; R), θ ∈ C m (X ; R) and φ ∈ C p (X ; R). If σ is an

(n + m + p)-singular simplex in X , then by definition of cup product it follows
that [σ, ψ ∪ (θ ∪ φ)] = [σ, (ψ ∪ θ) ∪ φ] for all σ. This shows that cup product is
associative. Again, if ψ ∈ C n (X ; R), θ ∈ C m (X ; R) and φ ∈ C m (X ; R) and σ is an
(n + m)-singular simplex in X , then
[σ, ψ ∪ (θ + φ)] = [σλn , ψ][σλm , θ + ψ]

= [σ, ψ ∪ θ] + [σ, ψ∪] for all σ.
This shows that the left distributivity holds. The right distributivity is similarly
proved.
If R contains 1, define c ∈ C 0 (X ; R) by [x, c] = 1 for all x ∈ X (use the fact
that 0-simplexes in X are identified with the points of X ). Then c is a both-sided
identity in C ∗ (X ; R). Consequently, it follows from the definition of cup product
that C ∗ (X ; R) is a graded ring. ❑
Proposition 10.18.7 The cup product in C ∗ (X ; R) is bilinear.
Proof If follows from the definition of cup product and the above distributive
laws. ❑
This shows that one may regard cup product as a map
∪ : C ∗ (X ; R) ⊗ C ∗ (X ; R) → C ∗ (X ; R).
Remark 10.18.8 The coboundary operator is a derivation of the graded ring C ∗ (X ; R)

in the sense that δ(ψ ∪ θ) = δψ ∪ θ + (−1)n ψ ∪ δθ for ψ ∈ C n (X ; R), θ ∈
C m (X ; R).
We summarize the above discussion in the basic and important result

Theorem 10.18.9 Let X be a topological space and R be a commutative ring. Then

H ∗ (X ; R) = H n (X ; R).
n≥0
Proof The direct sum Z ∗ (X ; R) of the cocycles is a subring of C ∗ (X ; R) and the

direct sum B ∗ (X ; R) of the coboundaries is a two-sided ideal in Z ∗ (X ; R). Hence
passing of cup product to the quotient, the direct sum H ∗ (X ; R) of the cohomology
rings becomes a graded ring. ❑
Definition 10.18.10 If X is a topological space and R is a commutative ring, then

the cohomology ring with coefficients in R is
10.18 Cup and Cap Products in Cohomology Theory 395

H ∗ (X ; R) = H n (X ; R).
n≥0
Remark 10.18.11 The ring C ∗ (X ; R) has several demerits

(i) Its vast size makes it difficult to compute;
(ii) It does not satisfy the homotopy axioms;
(iii) It is not always commutative.
The ring structure on C ∗ (X ; R) inherited by H ∗ (X ; R) overcomes these defects.
Remark 10.18.12 For more properties of cup product see Ex.9 of Sect. 10.21.
10.18.2 Cap Product
This subsection discusses the cap product ∩, which is an adjoint operation of cup
product ∪.
Definition 10.18.13 For an arbitrary topological space X , coefficient ring R, and

each pair of integers m, n an R- bilinear cap product map
∩ : Cn+m (X : R) × C n (X ; R) → Cm (X ; R).
is defined by the following rule:

For ψ ∈ C n (X ; R), z ∈ C m+n (X ; R), z ∩ ψ is the unique m-dimensional chain such
that
[z ∩ ψ, θ] = [z, ψ ∪ θ] (10.25)
for every m-cochains θ, i.e., for any singular (m + n)-simplex σ, set σ ∩ ψ =

[σλn , ψ]σρm and extend to arbitrary (m + n)-chain by linearity relation (10.25).
Using a further extension by linearity there is a pairing:
∩ : C∗ (X ; R) ⊗ C ∗ (X ; R) → C∗ (X ; R) (10.26)
Proposition 10.18.14 The pairing
∩ : C∗ (X ; R) ⊗ C ∗ (X ; R) → C∗ (X ; R)
makes C∗ (X ; R) a graded commutative ring with identity 1.
Proof It follows from the definition of the pairing cap product map. ❑
Corollary 10.18.15 There is a bilinear pairing
∩ : Hm (X ; R) ⊗ H n (X ; R) → Hm−n (X ; R)
Proof It follows by passage to the quotients of the bilinear pairing in Proposition

10.18.14. ❑
10.19 Applications
This section presents some interesting applications such as Jordan curve theorem,
Euler charact eristic of a cellular complex.
10.19.1 Jordan Curve Theorem
A homeomorphic image of a circle is called a Jordan curve. This section studies

Jordan curve theorem with a homological proof of the theorem. A homeomorphic
image of a circle is called a Jordan curve.This theorem is one of the most classical
theorem on topology. It asserts that a subspace of R2 homeomorphic to S 1 separates
R2 into two complementary components. This statement of ‘Jordan curve theorem’
appears to be an intuitive one, but Jordan asserted that intuition is not a proof.
So it needs rigorous proof. The first rigorous proof was given by Veblem (1905).
R. Maehara proved Jordan curve theorem using Brouwer fixed point theorem in
1984 Maehara (1984). But we prove this theorem in a different way.
Jordan curve theorem is one of the first problems of a purely topological nature,
is related to connectedness, in particular to continuum theory. One version of this
theorem says that a simple closed curve J in the Euclidean plane divides the plane
into two regions and J is their common boundary. This result was announced and
discussed by C. Jordan in 1887, but this proof was not rigorous.
Lemma 10.19.1 (i) Let A be a subspace of S n homomorphic to D k for some k ≥ 0.

Then H i (S n − A) = 0 for all i;
(ii) Let X is a subspace of S n homomorphic to S k for some k with 0 ≤ k < n. Then

i (S n − X ) ∼ Z, if i = n − k − 1
H =
0, otherwise .
Proof (i) We apply first induction on k. If k = 0, S n − A is homeomorphic to

Rn . So the proof follows trivially. Next let h : I k → A be a homeomorphism.
Consider the open sets D = S n − h(I k−1 × [0, 1/2]) and S = S n − h(I k−1 ×
[1/2, 1]), with D ∩ S = S n − A and D ∪ S = S n − h(I k−1 × {1/2}). By induc-
tion Hi (D ∪ S) = 0 for all i. Hence by Mayer–Vietoris sequence, there are
isomorphisms ψ : H i (S n − A) → H i (D) ⊕ H i (S) for all i. The two compo-
nents of ψ are induced by the inclusions S − A → D and S n − A → S. Hence
n
if there exists an i-dimensional cycle α in S n − A which is not a boundary in

S n − A, then α is also not a boundary in at least one of D and S. If i = 0, ‘cyclic’
here is considered in the sense of augmented chain complexes, which are deal-
ing with reduced homology. In a similar way, we can further subdivide the last
I factor of I k into quarters, eights, . . . to obtain a nested sequence of closed
subintervals I1 ⊃ I2 ⊃ · · · with intersection one point p ∈ I , such that α is not
a boundary in S n − h(I k−1 × Im ) for any m. By induction on k, α is the bound-
ary of a chain β in S n − h(I k−1 × { p}). Hence β is a finite linear combination
of singular simplices with compact image in S n − h(I k−1 × { p}). The union of
these images is covered by the nested sequence of open sets S n − h(I k−1 × Im ).
Hence by compactness β must be a chain in S n − h(I k−1 × Im ) for some m.
This contradiction implies that α is a boundary in S n − A. This completes the
induction steps.
(ii) We prove it by induction on k, starting with the trivial case k = 0, when
S n − X ≈ S n−1 × R. We represent X as a union of A1 ∪ A2 , where A1 and
A2 homeomorphic to D k and A1 ∩ A2 is homeomorphic to S k−1 . We now use
Mayer–Vietoris sequence for A = S n − A1 and B = S n − A2 , both of which
have trivial reduced homology groups by (i). Hence there exist isomorphisms
H i (S n − X ) ∼
=H i+1 (S n − (A1 ∩ A2 )) for all i.
❑
Theorem 10.19.2 (Jordan curve) The complement in the plane R2 of a Jordan curve
J consists of two open components, each of which as J as its boundary.
Proof It follows from (ii) of Lemma 10.19.1 that a subspace of S 2 homeomorphic to
S 1 separates S 2 into two complementary complements, i.e., into two path components
since open subsets of S n are locally path-connected. To complete the proof use now
R2 in place of S 2 , since deleting a point from an open set in S 2 does not affect its
connectedness. ❑
Theorem 10.19.3 (Generalized Jordan curve theorem) A subspace of S n homeo-
morphic to S n−1 separates it into two components, and these components have the
same homology group as a point. In particular, both complementary regions are
homeomorphic to open balls.

10.19.2 Homology Groups of Sαn
α∈a
Theorem 10.19.4 Let A be an indexing set, and Sαn be a copy of the n-sphere, α ∈ A.
Then
⊕Z(α), if p = n
p
H Sα ≡
n
α∈A
0, otherwise ,

where Z(α) is a free abelian group with generators α ∈ A.
α∈A

Proof The spaces ( Sαn ) and Sαn = Sαn+1 are homotopy equivalent.
α∈A α∈A α∈A
Hence the theorem follows. ❑
Remark 10.19.5 For more generalization of the result in Sect. 10.19.2 see Ex. 9 of
Sect. 10.21.
10.20 Invariance of Dimension
This section establishes the homotopy invariance of dimensions of spheres in the

sense that the spheres S m and S n are not homotopy equivalent if m = n and proves
that the dimension of a Euclidean space is a topological invariant.
Proposition 10.20.1 If m and n are two distinct nonnegative integers, then the
spheres S m and S n are not homotopically equivalent.
Proof Without loss of generality, we assume that 0 ≤ m < n. Let G be the nontrivial
coefficient group of the ordinary homology theory. Then
Hn (S m ) = 0 and Hn (S n ) ∼
=G (10.27)
Since the group G is nontrivial, it follows from (10.27) that the spheres S m and S n
cannot be homotopy equivalent. ❑
Corollary 10.20.2 If m = n, then the spheres S m and S n are not homeomorphic.
Proof It follows from Proposition 10.20.1. ❑
Proposition 10.20.3 If m and n are two distinct nonnegative integers, then the
Euclidean spaces Rm and Rn are not homeomorphic.
Proof If possible, the Euclidean spaces R m and Rn are homeomorphic. Then there
exists a homeomorphism
f : Rm → Rn (10.28)
Hence the image point f (0) of the origin of Rm is a point v in Rn . Let g be the
translation of the Euclidean space Rn defined by
g : Rn → Rn , x → x − v.
Then the homeomorphism

k = g ◦ f : Rm → R n (10.29)
carries the origin of Rm into the origin of Rn . This implies that there exists a homeo-
morphism k = k|Rm −{0} : Rm − {0} → Rn − {0}. Then using the fact that Rn − {0}
10.20 Invariance of Dimension 399
is homotopy equivalent to S n−1 , we find that the spaces Rm − {0} and Rn − {0} are
homotopy equivalent to the spheres S m−1 and S n−1 respectively. This shows that
the spheres S m−1 and S n−1 are homotopy equivalent. This contradicts Proposition
10.20.1 as m = n. ❑
Remark 10.20.4 The dimension of a Euclidean space is a topological invariant.
10.21 Exercises
1. (Three utilities problem) Suppose there are three houses on a plane and each
requires to be connected to the gas, water, and electricity lines. Show that there
is no way to make all the nine connections without any of the lines crossing each
other.
[Hint: Use Jordan curve theorem.]
2. Calculate the homology groups of the chain complex all of whose homology
groups are 0 except the groups C0 = 2Z, C1 = 4Z, C2 = 3Z, C3 = Z, and the
boundary homomorphism ∂1 is given by the (2 × 4)-matrix composed of the
rows (1 1 1 1) and (−1 −1 −1 −1);
∂2 is given by the the (4 × 3)-matrix composed of the rows (1 1 1),

(1 − 1 − 1), (−1 − 1 1), and (−1 1 −1) and ∂3 maps the whole group C3
to 0.
(Geometrically, the above chain complex corresponds to a cell decomposition of
the closed orientable 3-manifold which is obtained by taking the quotient of the
sphere S 2 by the linear action of the group {±1, ±i, ± j, ±k} of the quaternion
units. Hence the homology groups coincide with those of that manifold.)
3. For each fixed integer n ≥ 0 and an abelian group G, show that cohomology is
a contravariant functor
H n (−; G) : Top → Ab.
4. Calculate the cohomolgy groups of the Klein bottle.

5. Show that every continuous map f : (I n , ∂ I n ) → (X, x 0 ) induces by passage to
the quotient a singular n-simplex associated with an n-simplex, each of whose
faces is the constant map on x 0 . Hence deduce a homomorphism
h n : πn (X, x 0 ) → Hn (X ; Z) for n ≥ 1.
Further, show that these homomorphisms are functorial in the sense that every
continuous map f : (X, x0 ) → (Y, y0 ) induces a commutative diagram as shown
in Fig. 10.10.
6. (Dimension axiom) If X is a one-point space,and G is an abelian group, show
that
Fig. 10.10 Functional πn (X, x0 )

hn
/ Hn (X; Z)
representation of h n
πn (f ) Hn (f )
hn

πn (Y, y0 ) / Hn (Y ; Z)
(i)
G, if n = 0
Hn (X ; G) ∼
=
0, if n > 0.;
(ii)
∼ G, if n = 0
H (X ; G) =
n
0, if n > 0.
7. (Homotopy axiom) If f, g : X → Y are homotopic maps. show that they induce

the same homomorphism
f ∗ = g ∗ : H n (Y ; G) → H n (X ; G).
8. If n is a positive integer, show that

Z, if m = n
Hn (S ) ∼
n
=
0, otherwise.;
[Hint. Let X 1 = {x ∈ S n : xn > − 12 } and X 2 = {x ∈ S n : xn < 12 }. Then X 1

and X 2 are contractible spaces. Use induction on n and apply Mayer–Vietoris
sequence.]
9. (X α , xα ) be based spaces, where α ∈ A. Assuming that (X α , xα ) is a Borsuk
pair for each α ∈ A, show that

n (
H X α) = n (X α ).
H
α∈A α∈A
10. (Cup Product) Let R be a ring such as Z, Zn , and Q. For cochains α ∈ C k (X ; R)

and β ∈ C r (X ; R), their cup product α ∪ β ∈ C k+r (X ; R) is the cochain whose
value on a singular simplex σ : k+r → X, (α ∪ β)(σ) = α(σ|v0 , · · · , vk )·
β(σ|vk , . . . , vk+r ), where the right-hand side is the usual product in R. Show
that
10.21 Exercises 401
(i) δ(α ∪ β) = δα ∪ β + (−1)k α ∪ δβ;

(ii) the cup product of two cocycles is again a cocycle;
(iii) the cup product of a cocycle and a coboundary in either order, is a cobound-
ary;
(iv) there is an induced cup product
∪
H k (X ; R) × H r (X ; R) −−−→ H k+r (X ; R);
(v) the cup product ‘∪’ is associative and distributive;

(vi) If R has an identity element, then there is an identity element for cup product;
11. Let Htp be the homotopy category of topological spaces and GR be the category
graded rings. If R is a commutative ring, show that

H ∗ ( ; R) = H n ( ; R) : Htp → GR
n≥0
is a contravariant functor.

[Hint. Use Z ∗ (X ; R) = Z n (X ; R) and B ∗ (X ; R) = B n (X ; R).]
n≥0 n≥0
12. (Cohomology cross product) Given CW -complexes X and Y ∈ C0 , define a cross
product of cellular cochains α ∈ C k (X ; R) and β ∈ C r (Y ; R) by setting
(α × β)(eik × erj ) = α(eik )β(erj )
and letting α × β take the value 0 on (k + r )-cells of X × Y which are not the
product of a k-cell of X with an r -cell of Y .Prove that
(i) δ(α × β) = δα × β + (−1)k α × δβ for cellular cochains α ∈ C k (X ; R)
and β ∈ C r (X ; R) (coboundary formula);
(ii) given a definition of cross product there is a cup product (agreeing with the
original definition) as the composite
× ∗
H k (X ; R) × H r (X ; R) −−−→ H k+r (X × X ; R) −−−−→ H k+r (X ; R),
where
: X → X × X, x → (x, x)
is the diagonal map.

13. (Exact homology sequence) Given CW -complexes X and A ∈ C 0 and (X, A) ∈
C show that the sequence
i∗ j∗ ∂ i∗
· · · −→ Hn (A) −−−−→ Hn (X ) −−−−→ Hn (X, A) −−−→ Hn−1 (A) −−−−→ Hn−1 (X ) −→
··· −
→ H0 (X, A) −
→0
is exact, when the boundary operator
∂ : Hn (X, A) → Hn−1 (A)
is defined as follows:
If an element [ f ] ∈ Hn (X, A) is represented by a relative cycle f , then ∂[ f ] is
the class of the cycle ∂α in Hn−1 (A).
14. Show that for all i > 0,

Z, for i = n
Hi (D n , ∂ D n ) ∼
=
0, otherwise .
[Hint.Use the long exact sequence of reduced homology groups for the pairs
i−1 (S n−1 ) are isomor-
(D n , ∂ D n ), the homomorphisms ∂ : Hi (D n , ∂ D n ) → H

phic for all i > 0, since the remaining terms Hi (D ) = 0 for all i.]
n
15. Show that
2, if n is even
χ(S ) =
n
0, if n is odd.
16. Let T 2 be the torus. Show that Euler characteristic χ(T 2 ) = 0.
[Hint. χ(T 2 ) = β0 (T 2 ) − β1 (T 2 ) + β2 (T 2 ) = 1 − 2 + 1 = 0.]

17. Let B be a finite CW -complex and p : X → B be an n-sheeted covering space.
Show that χ(X ) = nχ(B).
18. Let R P 2 be the real projective plane. Show that the Euler characteristic
χ(R P 2 ) = 1.
19. Show that for any simplicial complex K the 0-dimensional homology group
H0 (K ; Z) is the free abelian group whose rank is the same as the number of
connected components of K .
20. For any connected closed triangulable manifold M of dimension n, show that

Z, if M is orientable
Hn (M; Z) ∼
=
0, if M is not so.
21. Let K be a simplicial complex. Show that the groups H p (K ; Z)do not depend
on the choice of an orientation of K .
22. Let X be a nonempty space and x ∈ X . Show that the inclusion map X →
(X, {x}) determines an exact homology sequence.
23. Show that
10.21 Exercises 403

∼ Z, if p = 0, n, m, n, n + m
H (S × S ; Z) =
p n m
0, otherwise.
24. Show that
p (S n ; Z) ∼ Z, if p = n
H =
0, otherwise.
n+1 ( X ) ∼
25. Let X be a nonempty space. Show that H =H n (X ) for each n.
26. Show that
(i) the Euler characteristic is additive:
for any cellular space and its finite cellular subspaces A and B,
χ(A ∪ B) = χ(A) + χ(B) − χ(A ∩ B)
(ii) Euler characteristic is multiplicative:
for any finite cellular spaces X and Y ,
χ(X × Y ) = χ(X )χ(Y ).
27. Show that a finite connected cellular space X of dimension 1 is homotopy equiv-
alent to the bouquet of 1 − χ(X ) circles.
28. Show that the fundamental group of S2 with n points removed is a free group of
rank n − 1.
29. Show that given any knot K , the homology groups H0 (R3 − K ; Z) and H1 (R3 −
K ; Z) are isomorphic.
30. Let K be a simplicial complex. Then K is said to be connected if K is not the
union of two nonempty subcomplexes of K which have no subcomplexes in
common. Show that
(i) K is connected iff the polyhedra |K | is connected;
(ii) if K is a connected complex, then H0 (X ; Z) ∼
= Z.
31. Let K be a (finite) oriented simplicial complex of dimension m. Show that
(i) Hn (K ) is finitely generated ( f ·g) for every n ≥ 0;
(ii) Hn (K ) = 0 for all n > m;
(iii) Hm (K ) is free abelian, possibly zero.
[Hint. (i) Cn (K ) is f ·g and hence its subgroup Z n (K ) is also so.
(ii) Cn (K ) = 0 for all n > m.
(iii) As Cm+1 (K ) = 0, Bm (K ) = 0 and hence Hm (K ) = Z m (K ). Use the result
that a subgroup of a free abelian group is also free abelian (see Sect. 14.8).]
32. Let f, g : X → Y be homotopic maps between two polyhedra. Show that the
induced homomorphism
f ∗ = g∗ : Hn (X ) → Hn (Y )
for all n. Hence show that the homotopy equivalence of polyhedra X and Y have
isomorphic homology groups.
[Hint. As f and g are homotopic maps, any simplicial approximation to f is at

the same time a simplicial approximation to g.]
33. Let K and L be two simplicial complexes. If f : |K | → |L| and g : |L| → |M|
are continuous maps of polyhedra, show that
( f ◦ h)n∗ : Hn (K ) → Hn (M)
satisfying the relation

( f ◦ h)n∗ = f n∗ ◦ h n∗
in each dimension.
34. Let X be a topological space with base point x0 . Show that
(i) Hn (X, x0 ) ∼
= Hn (X ) for all n ≥ 1;
(ii) Hn (X, x0 ) ∼
=H n (X ) for all n ≥ 0.
35. Let (X, A) be a pair of topological spaces with X compact Hausdorff and A
closed in X , where A is a strong deformation retract of some closed neighbour-
hood of A in X . If p : (X, A) → (X/A, y) is the identification map, show that
its induced homomorphism
p∗ : Hn (X, A) → Hn (X/A, y)
is an isomorphism for all n.

36. If (X, A) is a compact Hausdorff pair of topological spaces for which A is a
strong deformation retract of some compact neighbourhood of A in X . Show
that Hn (X, A) ∼ =H n (X/A) for all n.
37. If X is cellular space and n ≥ 0, show that Hn (W∗ (X )) ∼
= Hn (X, X (n−1) ), where
(n)
Wn (X ) = Hn (X , X ). n−1
38. If X is a cellular space with X (−1) = ∅, then for all n, Hn (X ) ∼

= Hn (W∗ (X )),
where Wn (X ) = Hn (X (n) , X n−1 ).
39. Show that cellular homology is a homology functor from the category of CW -
complexes and cellular maps to the category of R-modules and R-linear maps.
40. Show that there is a natural equivalence from the cellular homology theory to
the singular homology theory.

[3] Dieudonné, J., A History of Algebraic and Differential Topology, 19001960,
[5] Fulton, W., Algebraic Topology, A First Course, Springer-Verlag, New York,
1975.
[6] Greenberg, M.J., and Harper. J.R., Algebraic Topology, A First Course, Ben-
jamin/Cummings Pub.Company, London, 1981.
[7] Hilton, P.J., and Wylie, S. Homology Theory, Cambridge University Press,
Cambridge, 1960.
[8] Hu, S.T., Homology Theory, Holden Day, Oakland, CA, 1966.
[9] Lefschetz, S., Algebraic Topology, Amer. Math. Soc., New York, 1942.
[10] Matveev, Sergey V., Lectures on Algebraic Topology, European Math. Soc.,
2006
References
(1972)
Adhikari, M.R., Adhikari A.: Basic Modern Algebra with Applications. Springer, Heidelberg (2014)
Čech, E.: Théorie génerale de l’homologie dans un espace quelconque’. Fundam. Math. 19, 149–183
(1932)
Croom, F.H.: Basic Concepts of Algebraic Topology. Springer, Berlin (1978)
Dieudonné, J.: A History of Algebraic and Differential Topology, pp. 1900–1960. Modern
Birkhäuser, Boston (1989)
ton (1952)
Fulton, W.: Algebraic Topology, A First Course. Springer, New York (1975)
(1975)
Greenberg, M.J., Harper J.R.: Algebraic Topology, A First Course. Benjamin/Cummings Publisher
Company, London (1981)
Hu, S.T.: Homology Theory. Holden Day, Oakland (1966)
Hurewicz, W.: Beitrage der Topologie der Deformationen. Proc. K. Akad. Wet. Ser. A 38, 112–119,
521–528 (1935)
Lefschetz, S.: Algebraic Topology. American Mathematical Society, New York (1942)
Maehara, R.: The Jordan curve theorem via the Brouwer fixed point theorem. Am. Math. Mon. 91,
641–643 (1984)
Mateev, S.V.: Lectures on Algebraic Topology. European Mathematical Society (2006)
Mayer, J.: Algebraic Topology. Prentice-Hall, Englewood Cliffs (1972)
Massey, W.S.: A Basic Course in Algebraic Topology. Springer, Heidelberg (1991)
Munkres, J.R.: Elements of Algebraic Topology. Addision- Wesley, Reading (1984)
Nakahara, M.: Geometry, Topology and Physics. Taylor and Francis, Boca Raton (2003)
Switzer, R.M.: Algebraic Topology-Homotopy and Homology. Springer-Verlag, Heidelberg (1975)
Veblem, O.: Theory of plane curves in nontrivial analysis situs. Trans. Am. Math. Soc. 6, 83–98
(1905)
Chapter 11
Eilenberg–MacLane Spaces
This chapter conveys homotopy theory through an important class of C W -complexes

called Eilenberg–MacLane spaces introduced by S. Eilenberg (1915–1998) and
S. MacLane (1909–2005) in 1945. An Eilenberg–MacLane space is a C W -complex
having just one nonzero homotopy group G in dimension n (G is abelian if n > 1),
denoted by K (G, n). The spaces K (G, 1) had been studied by W. Hurewicz (1904–
1956) before Eilenberg and MacLane took up K (G, n) as a general case. The impor-
tance of Eilenberg–MacLane spaces is twofold. First, they develop homotopy the-
ory. Secondly, they are closely linked with the study of cohomology operations (see
Chap. 15). This chapter proves that given an Eilenberg–MacLane space K (G, n),
and a C W -complex X , the group [X, K (G, n)] is the cohomology group H n (X ; G).
This interesting result relates cohomology theory with homotopy theory.
More precisely, this chapter constructs Eilenberg–MacLane spaces, and finally
studies their homotopy properties and relates cohomology theory with homotopy
theory. The construction process of Eilenberg–MacLane spaces K (G, n) for all pos-
sible (G, n) is very interesting and depends on a very natural class of spaces, called
Moore spaces of type (G, n), denoted by M(G, n). This chapter also studies Post-
nikov towers to meet the need for the construction of Eilenberg–MacLane spaces.
For this chapter the books and papers Eilenberg and MacLane (1945a), Gray
(1975), Hatcher and Allen (2002), Hu (1959), Maunder (1980), Seree (1951), Spanier
(1966), and some others are referred in Bibliography.
11.1 Eilenberg–MacLane Spaces: Introductory Concept
This section presents Eilenberg–MacLane spaces with some interesting examples.

Such spaces interlink different concepts in algebraic topology and present some
amazing results.

DOI 10.1007/978-81-322-2843-1_11
408 11 Eilenberg–MacLane Spaces
Definition 11.1.1 A pointed C W -complex is called an Eilenberg–MacLane space if

it has only one nontrivial homotopy group. If G is a group and n is a positive integer,
the Eilenberg–MacLane space of type (G, n) is a pointed C W -complex X whose
homotopy groups vanish in all dimensions except n, where G = πn (X ) and G is to
be abelian for n > 1.
Remark 11.1.2 We use the notation K (G, n) for a C W -complex which represents
an Eilenberg–MacLane space of type (G, n). This is well defined, since there is only
one space of type (G, n) up to homotopy equivalence. For a group G, K (G, 0) is
defined to be the group G with the discrete topology.
Definition 11.1.3 A path-connected space whose fundamental group is isomorphic
to a given group G and which has a contractible universal covering space is called a
K (G, 1) space.
Example 11.1.4 We look at the following examples:

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.
where RP∞ , CP∞ , S 1 denote the infinite dimensional real projective space, infinite
dimensional complex projective space, unit circle in C respectively. Consequently,
K (Z2 , 1) = RP∞ (infinite dimensional real projective space).
K (Z, 2) = CP∞ (infinite dimensional complex projective space)
K (Z, 1) = S 1 (unit circle in C), but S 2 is not an Eilenberg–MacLane space of type K (Z, 2).
Example 11.1.5 K (Zm , 1) is an infinite dimensional lens space l ∞ (m) = S ∞ mod

Zm , where Zm acts on S ∞ , regarded as the unit sphere in C∞ , by scalar multi-
plication by the mth roots of unity, this action being the map (z 1 , z 2 , . . . , ) −→
e2πi/m (z 1 , z 2 , . . . , ).
Remark 11.1.6 The infinite dimensional lens space K (Zm , 1) cannot be replaced by
any finite dimensional complex.
Example 11.1.7 Given a closed connected subspace K of S 3 which is nonempty, the
complement S 3 − K is an Eilenberg–MacLane space. In particular, if K is the trivial
knot, then S 3 − K is an Eilenberg–MacLane space K (Z, 1).
11.2 Construction of Eilenberg–MacLane Spaces K (G, n) 409
11.2 Construction of Eilenberg–MacLane Spaces K (G, n)
This section prescribes a process of construction of Eilenberg–MacLane spaces

K (G, n) up to homotopy equivalence. An alternative proof of uniqueness of K (G, n)
up to homotopy equivalence is given in Theorem 15.11.14. Eilenberg–MacLane
spaces are pointed C W -complexes X for which πr (X, x0 ) = 0 except for one value
n ≥ 1 of r . The existence of such spaces was shown by J.H.C. Whitehead (1904–
1960) in 1949 by using the properties of C W -complexes. The proof is done by
induction on m > n, the inductive assumption being that there exists a C W -complex
X (m) such that

G for r = m,
πr (X (m) , x0 ) ∼
=
0 for 1 ≤ r ≤ m and r = m
(for m = n + 1 the result of πr is above).
11.2.1 A Construction of K (G, 1)
This subsection gives a construction of K (G, 1). It can be obtained as an orbit space.
Let G be for an arbitrary group G (not necessarily abelian). Let q be the q-simplex
with ordered vertices (g0 , g1 , . . . , gq ) of elements of G and C(G) be the complex
obtained as a quotient space of the collection of disjoint simplices q by identifying
their certain faces by canonical linear homeomorphism, preserving the ordering of the
vertices. This attaches the q-simplex q to the (q −1) simplexes (g0 , . . . , gi , . . . , gn ),
where the notation gi indicates that this vertex is omitted. The group G acts on
C(G) by the left multiplication: g.(g0 , g1 , . . . , gq ) = (g.g0 , g.g1 , . . . , g.gq ). Let
C(G)mod G be its quotient space. This action of G on C(G) is a covering space
action. Hence the quotient map p : C(G) → C(G)mod G is the universal covering
of the orbit space BG = C(G)mod G, which is a K (G, 1) space.
Remark 11.2.1 The homotopy type of a C W -complex K (G, 1) is uniquely defined

by G.
11.2.2 A Construction of K (G, n) for n > 1
This subsection conveys detailed process of construction of K (G, n) for all possible
(G, n) (group G is abelian for n > 1). This construction process is done on the
following stages:
(i) Construction of ‘Moore spaces’ M(G, n);

(ii) Using ‘killing homotopy’;
(iii) Applying ‘Postnikov decomposition’;

(iv) Construction of K (G, n) uniquely determined up to homotopy equivalence by
G and n.
11.2.3 Moore Spaces
This subsection introduces the concept of a Moore space, which is a generalization of

Eilenberg–MacLane space. We claim that the Eilenberg–MacLane spaces K (G, n)
exist for all proper (G, n) and each one of them is unique up to homotopy type.
The following class of spaces plays an important role in the construction process of
K (G, n).
Definition 11.2.2 A C W -complex X with one 0-cell, all of its other cells are in
dimensions n and n + 1, and is such that πn (X ) ∼
= G, is called a Moore space of
type (G, n), denoted by M(G, n), G is abelian for n > 1.
Remark 11.2.3 For existence of Moore spaces see Exercise 8 of Sect. 11.4.
Example 11.2.4 (i) S n is a Moore space M(Z, n).

(ii) R P 2 is a Moore space M(Z2 , 1).
(iii) CP 1 is a Moore space M(Z, 2).
Remark 11.2.5 There is a homological version of a Moore space.
Definition 11.2.6 Given an abelian group G and an integer n ≥ 1, there is a C W -

complex written M(G, n) such that Hn (X ) ∼= G and H̃i (X ) = 0 for i = 0. The
space M(G, n) is also called a Moore space of type (G, n).
11.2.4 Killing Homotopy Groups
This subsection conveys a process of construction for obtaining an Eilenberg–

MacLane space K (G, n) from a Moore space M(G, n) by killing all homotopy
groups above the nth.
Definition 11.2.7 The process of construction to obtain an Eilenberg–MacLane

space K (G, n) from a Moore space M(G, n) by trivializing all homotopy groups
beyond the nth is called ‘killing homotopy groups’.
Theorem 11.2.8 (Killing homotopy theorem) Given a C W -complex X and an inte-

ger n > 0, there exists a relative C W -complex (X
, X ) with cells in dimension (n +1)
only such that
(i) πn (X
) = 0;
(ii) πm (X
) ∼= πm (X
) for m < n,

here X = (X α∈A Dαn+1 )/ ∼, where x is identified to f (x) for x ∈ ∂ D n+1 ,

i.e., x ∼ f (x), ∀ x ∈ ∂ Dαn+1 .
Proof Let { f α : S n → X, α ∈ A} represent a set of generators of πn (X ). Then

for each α∈ A, take an (n + 1)-ball Dαn+1 and attach it by f α to X to obtain
X
= (X α∈A Dαn+1 )/ ∼, where x is identified to f (x) for all x ∈ ∂ D n+1 ,
i.e., x ∼ f (x), ∀ x ∈ ∂ Dαn+1 . This shows that the relative C W -complex (X
, X ) has
only these Dαn+1 as (n + 1)-cells, which precisely make the generators of πn+1 (X )
inessential. Clearly, below dimension n, the space X
has the same homotopy groups
as the space X by the inclusion i : X → X
. ❑
Remark 11.2.9 X
resembles X below dimension n but at dimension n, πn (X
) = 0.
Example 11.2.10 A 4-ball D 4 may be attached to S 2 by Hopf map S 3 → S 2 to kill

the group π3 (S 2 ).
11.2.5 Postnikov Tower: Its Existence and Construction
This subsection studies Postnikov tower (or Postnikov system) which gives a way
of constructing a topological space given by M. Postnikov (1927–2004) in 1951
(Postnikov 1951).
Theorem 11.2.11 Any C W -complex X admits a decomposition into a tower or a

system of C W -complex pairs (X [n] , X ), called a Postnikov tower or a system of X
described in Fig. 11.1 with
(i) cells in dimension (n + 1) and above only;
(ii) πm (X [n] ) = 0 for all m > n;
(iii) i n∗ πm (X ) ∼
= πm (X [n] ) for m ≤ n, where i n : X → X [n] is the inclusion.
Proof If n ≥ 0 is a fixed integer,

apply killing homotopy Theorem 11.2.8 to kill
(1) (n+2)
πn+1 (X ). Then X n = (X Dα )/ ∼, with πn+1 (X n(1) ) = 0. Again apply the
α∈A
same procedure to X n(1) to get X n(2) with πn+1 (X n(2) ) = πn+2 (X n(2) ) = 0. Recursively,

we obtain X n(r ) . Then the required space X [n] is obtained by setting X [n] = X n(r )
r ≥1
with weak topology. This gives (i) and (ii). (iii) follows from the preservation of πm
by direct limits for C W -complexes. ❑
Definition 11.2.12 X [n] is called the nth Postnikov section of X in the Postnikov
tower which is defined uniquely up to homotopy equivalence.
Fig. 11.1 Postnikov tower

homotopy fibres
of a CW-complex X ..
.

[n] o
X
<
K(πn (X), n)
x
inxx
xxx
x
xx
X FF
FF in−1
FF
FF
F#
? _ K(πn−1 (X), n − 1)
X [n−1] o

..
.
Remark 11.2.13 Postnikov sections X [n] of X can be viewed as successive approx-

imations to X and are considered dual to cellular skeletal approximations X (n) .
Each inclusion i n : X → X [n] is convertible up to homotopy with a fibration.
pn
X (n) → X̄ (n) −−−−→ X [n] ,
where jn : X (n) → X̄ (n) is an inclusion and X̄ (n) = {(x, f ) ∈ X × (X [n] ) I : f (0) =

i n (x)} X, pn : (x, f ) → f (1) and fiber X (n) = {(x, f ) ∈ X̄ (n) : f (0) = f (1)}.
Definition 11.2.14
X (n) is called the n-connected covering space of X , which is also
usually denoted by X (n) .
Using the above notations we have the following theorem:
Theorem 11.2.15 Let X be a C W -complex. Then

0 for m ≤ n
(n)
πm ( X )∼
=
πm (X ) for m > n.
Proof It follows from construction of X (n) that the only nontrivial homotopy groups
(n)
in πm ( X ) are those above n and are isomorphic to those of X by homomorphisms
jn∗ induced by jn :
X (n) → X̄ (n) . ❑
Remark 11.2.16 Postnikov tower is also given as follows:

For any C W -complex X , there is a sequence of fibrations
Yn → Yn−1 → · · · → Y2 → Y1 = K (π1 (X ), 1)
with the fiber of Ym → Ym−1 being K (πm (X ), m); and connecting maps f m : X →
Ym such that the homomorphisms
( f m )∗ : πi (X ) → πi (Ym )
are isomorphism for i ≤ m.
11.2.6 Existence Theorem
This subsection shows that given an abelian group G and an integer n > 1, there
exists a C W -complex K (G, n) determined uniquely up to homotopy equivalence by
G and n.
Theorem 11.2.17 (Existence Theorem) Given an abelian group G and an integer
n > 1, there exists a C W -complex K (G, n). The homotopy type of K (G, n) is
uniquely determined by G and n.
Proof Construction: The construction of K (G, n) is completed by setting
K (G, n) = M(G, n)[n] . Uniqueness: The C W -complex X of type (G, n) having
all homotopy groups equal to 0 except for πr (X ) = G is uniquely determined up to
homotopy equivalence, because if A and B are both Eilenberg–MacLane spaces of
type (G, n), then the identity homomorphism 1d : G → G induces a map h : A → B
by Ex. 6 of Sect. 11.4. Then h is a homotopy equivalence. ❑
Remark 11.2.18 Let G = Z. Then given an arbitrary n, a natural construction of

K (Z, n) starts with S n for which πn (S n ) ∼ = Z. Then attach (n + 2)-cells to kill
πn+1 (S n ) and iterate this process of attaching higher cells to kill higher homotopy
groups. The resulting space is a K (Z, n).
11.3 Applications
This section presents some important applications of Eilenberg–MacLane spaces.

Given an Eilenberg–MacLane space K (G, n), and a C W -complex X , the group
[X, K (G, n)] is the cohomology group H n (X ; G). This amazing result relates coho-
mology theory to homotopy theory by admitting a group structure on the set of homo-
topy classes of continuous maps from a C W -complex to an Eilenberg–MacLane
space.
Theorem 11.3.1 (Whitehead theorem) Let G be an abelian group and K (G, n) be

an Eilenberg–MacLane space. Then there is weak homotopy equivalence
αn : K (G, n) → K (G, n + 1),
which is also a homotopy equivalence.
Proof Since πn (K (G, n + 1)) = ∼ πn+1 (K (G, n + 1) =∼ G and πn (K (G, n)) ∼ =
πn+1 (K (G, n + 1) ∼
= G, it follows that πn (K (G, n + 1)) ∼
= πn (K (G, n) for every
n ≥ 1. Consequently, there is a continuous map
αn : K (G, n) → K (G, n + 1)
such that its induced homomorphism
αn∗ : πn (K (G, n)) → πn (K (G, n + 1))
is an isomorphism. Again since all other groups are trivial, αn is a weak homotopy
equivalence. Moreover, K (G, n + 1) has the homotopy type of a C W -complex.
Hence it follows that αn is a homotopy equivalence. ❑
Proposition 11.3.2 There is a natural group structure on [X, K (G, n)].
Proof The space K (G, n) is homotopy equivalent to a loop space by Theorem 11.3.1.
Since every loop space is an H -group and the set of homotopy classes of maps
from any pointed space to an H -group admits a group structure (see Chap. 2), the
proposition follows. ❑
Theorem 11.3.3 (Hopf theorem) If X is a path-connected n-dimensional C W -

complex, then H n (X ; Z) ∼
= [X, S n ] for an ordinary cohomology H ∗ .
Proof Construct an Eilenberg–MacLane space K (Z, n) from S n by attaching cells

of dimensions ≥ n + 2. Then the inclusion i : S n → K (Z, n) induces a function
i ∗ : [X, S n ] → [X, K (Z, n)].
i ∗ is injective: Suppose i ∗ ( f ) = i ∗ (g). Then there is a homotopy H : X × I →
K (Z, n) between f and g. By cellular approximation, it can be made to have an
image inside of (n + 1)-skeleton of K (Z, n), which is S n . This implies that f g.
i ∗ is surjective: Since the C W -complex X is n-dimensional, it follows by cellular
approximation that i ∗ is surjective. ❑
Corollary 11.3.4 Given an abelian group G, an integer n ≥ 1 and a path-connected

n-dimensional C W -complex X , the group [X, K(G,n)] is the cohomology group
H n (X ; G).
Proof The corollary follows likewise Theorem 11.3.3. ❑

Proposition 11.3.5
Z, if i = n,
πi (S P (S )) ∼
∞ n
=
0, if i = n.
Proof Since S P ∞ (S n ) is the Eilenberg–MacLane space K (Z, n), the proposition

follows. ❑
11.4 Exercises
1. Show that the Eilenberg–MacLane space K (G, n) is an H -space iff the group
G is abelian.
2. Show that for n > 1 the spaces K (G, n) and K (G, n − 1) are homotopy
equivalent.
3. Given a topological group G, show that the classifying space BG for K G (see
Chap. 5) and the Eilenberg–MacLane space K (G
, n) are homotopy equivalent
iff G and K (G
, n − 1) are homotopy equivalent for n > 1 and in particular, BG
and K (G, 1) are homotopy equivalent iff G is a discrete group.
4. Show that the Klein bottle is a K (G, 1)-space, where G is the group with two
generators a, b and one relation given by aba = b. Is any other surface an
Eilenberg–MacLane space K (G, n)? Justify your answer.
5. For any continuous map f : X → Y , show that a weak decomposition
exists. If π1 (X ) operates trivially on πn ((M f , X )) for all n, show that f has a
Moore–Postnikov decomposition, where M f is the mapping cylinder of f .
6. Let G be an abelian group and n ≥ 1. If g : G → H is a homomorphism
of groups, show that there exists a map h : K (G, n) → K (H, n) such that
h ∗ = g : πn (K (G, n)) → πn (K (H, n)).
7. Let G be the free group on k generators. Show that the wedge sum of k unit
k
circles S i is a K (G, 1) space.
i=1
8. Given an integer n > 1 and an abelian group G, show that there is a Moore
space M(G, n) which is a C W -complex with one 0-cell and all other cells in
dimensions n and n + 1 are such that πn (M(G, n)) ∼= G.
9. Show that the complement to any knot K in three-dimensional sphere S 3 is of
type K (G, 1), where G is a group depending on K .
10. (Whitehead tower) For any C W -complex X , there is sequence of fibrations:
· · · → X n → X n−1 → · · · → X 1 → X
where the fiber of X n → X n−1 is K (πn (X ), n − 1), and X 1 is the universal cover
of X . Show that πi (X n ) = 0 for all i ≤ n and the map f n : X n → X induces
isomorphisms f n ∗ : πi (X n ) → πi (X ) for i > n.
11. Given Eilenberg–MacLane spaces K (G, n) and K (H, n) show that the product
space K (G, n) × K (H, n) is a K (G × H, n). Hence show that the n-torus T n is
an example of K (Zn , 1).
12. Let X be a C W -complex of the form Sαn eβn−1 for some n ≥ 1. Show that
α β
for every homomorphism ψ : πn (X ) → πn (Y ) with Y path-connected, there
exists a map f : X → Y such that f ∗ = ψ. Hence show that given an integer n
and a group G (G is abelian if n > 1), the Eilenberg–MacLane space K (G, n)
is unique up to homotopy equivalence.

[4] Barratt, M.G., Track groups I, Proc. Lond. Math. Soc. 5(3), 71–106, 1955.
York, Heidenberg, Berlin, 1978.
[7] Dold, A., Lectures on Algebraic Topology, Springer-Verlag, New York, 1972.
[8] Eilenberg, S., and MacLane, S., Relations between homology and homotopy
groups, Proc. Natn. Acad. Sci. USA, 29, 155–158, 1943.
groups of spaces I, Ann. of Math., 46, 480–509, 1945.
1975.
Cambridge, 1983.
[12] Hilton, P.J. and Wylie, S. Homology Theory, Cambridge University Press,
Cambridge, 1960.
[14] Mayer, J., Algebraic Topology, Prentice-Hall, New Jersy, 1972.
Company, 1984.
[16] Rotman, J.J., An Introduction to Algebraic Topology, Springer-Verlag, New
York, 1988.
References 417
References
(1972)
New York (2002)
Arkowitz, M.: Introduction to Homotopy Theory. Springer, Berlin (2011)
Armstrong, A.: Basic Topology. Springer, New York (1983)
Dieudonné, J.: A History of Algebraic and Differential Topology, 1900–1960, Modern Birkhäuser
(1989)
Eilenberg, S., MacLane, S.: Relations between homology and homotopy groups. Proc. Nat. Acad.
Sci. USA 29, 155–158 (1943)
Eilenberg, S., MacLane, S.: General theory of natural equivalences. Trans. Amer. Math. Soc. 58,
231–294 (1945a)
Eilenberg, S., MacLane, S.: Relations between homology and homotopy groups of spaces I. Ann.
Math. 46, 480–509 (1945b)
ton (1952)
(1975)
Maunder, C.R.F.: Algebraic Topology, Van Nostrand Reinhhold, London, 1970. Dover, Reprinted
(1980)
Munkres, J.R.: Elements of Algebraic Topology. Addition-Wesley-Publishing Company, Reading
(1984)
Postnikov, M.: Determination of the homology groups of a space by means of the homotopy invari-
ants. Doklady Akademii Nauk SSSR 76, 359–362 (1951)
Seree, J.-P.: Homologie singuliére des espaces fibrés. Ann. of Math. 54, 425–505 (1951)
Chapter 12
Eilenberg–Steenrod Axioms for Homology
and Cohomology Theories
This chapter presents an approach formulating axiomatization of homology and coho-

mology 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 first appeared in their cele-
brated book “The Foundations of Algebraic Topology” in 1952. This approach clas-
sifies and unifies different homology (cohomology) groups, and is the most important
contribution to algebraic topology since the invention of the homology groups by
Poincaré in 1895 and is called the axiomatic approach for homology theory given by
a set of seven axioms by S. Eilenberg and N. Steenrod.
This axiomatic approach simplifies the proofs of many lengthy and complicated
theorems and escapes the avoidable difficulty to motivate the students who are learn-
ing homology and cohomology theories for the first time as their systematic study.
This approach gives the subject conceptual with coherence and elegance. It pro-
vides a quick approach for computing homology and cohomology groups. It uni-
fies different homology groups (modules) on the category of compact triangulable
spaces. It also inaugurates its dual theory called cohomology theory. This approach
did not contain the term C W -complex whose definite study was first given by
J.H.C Whitehead (1904–1960) in 1949. But it is thought today that this approach is
considered on the category of finite C W -complexes.
Homology invented by Henry Poincaré in 1895 was studied by him during
1895–1904. This homology called simplicial homology is one of the most fundamen-
tal powerful inventions in mathematics. He started with a geometric object (a 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. There
are other homology theories:
(i) homology groups for compact metric spaces introduced by L. Vietoris
(1891–2002) in 1927;
(ii) homology groups for compact Hausdorff spaces introduced by E. Cech
(1893–1960) in 1932;

DOI 10.1007/978-81-322-2843-1_12
420 12 Eilenberg–Steenrod Axioms for Homology and Cohomology Theories
(iii) singular homology groups first defined by S. Lefschitz (1884–1972) in 1933.

Čech, Vietoris, Lefschitz constructed these homology theories in different meth-
ods and designed these tools for solving some specific problems. Initially, all
these theories lived in isolation in the sense that no relation among them was
established.
Algebraic topologists started around 1940 comparing various definitions of
homology and cohomology given in the previous years. Eilenberg and Steenrod
initiated a new approach by taking a small number of their properties (not focussing
on machinery used for construction of homology and cohomology groups) as axioms
to characterize a theory of homology and cohomology. The most interesting result
is the proof that on the category of all topological pairs having homotopy type of
finite CW-complex pairs all homology and cohomology theories satisfying these
axioms have isomorphic groups. This result concludes that there is only one concept
of homology and (cohomology) in that category.
For this chapter the books Eilenberg and Steenrod (1952), Gray (1975), Hu (1966),
Maunder (1970), Rotman (1988), Spanier (1966) and some others are referred in the
Bibliography.
12.1 Eilenberg–Steenrod Axioms for Homology Theory
This section conveys an axiomatic approach to homology theory announced by

Eilenberg and Steenrod in 1945 but published in Eilenberg and Steenrod (1952).
Let C0 be the category of all based topological spaces having homotopy type of finite
CW-complexes and C be the category of all topological pairs having homotopy type
of finite CW-complex pairs.
A homology theory H on the category C consists of three functions H = {H, ∗, ∂}
which satisfy the following axioms:
(i): The first function H assigns to each topological pair (X, A) in C and each
integer p, (positive, negative, or 0), an abelian group H p (X, A), called the
p-dimensional homology group of the topological pair (X, A) in the homol-
ogy theory H. In particular, for A = ∅, it is called p-dimensional (absolute)
homology group of the space X .
(ii): The second function ∗ assigns to each continuous map f : (X, A) → (Y, B)
in C and each integer p a homomorphism
f ∗ = f p∗ : H p (X, A) → H p (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 C and an integer
p a homomorphism
12.1 Eilenberg–Steenrod Axioms for Homology Theory 421
∂ = ∂ p : H p (X, A) → H p−1 (A),
called the boundary operator on the group H p (X, A) in the homology theory
H.
Moreover, these functions satisfy the following seven axioms H(1)-H(7), called
the Eilenberg–Steenrod axioms for homology theory H on C;
Axiom H(1)(Identity Axiom). If 1 X : (X, A) → (X, A) is the identity map on a
topological pair (X, A) in C, then the induced homomorphism 1 X ∗ : H p (X, A) →
H p (X, A) is the identity automorphism of the homology group H p (X, A) for every
integer p.
Axiom H(2)(Composition Axiom). If f : (X, A) → (Y, B) and g : (Y, B) →
(Z , C) are continuous maps in C, then
(g ◦ f ) p∗ = g p∗ ◦ f p∗ : H p (X, A) → H p (Z , C)
for every integer p.
Remark 12.1.1 The above axioms H(1) and H(2) show that for every fixed integer
p, the functions H p form a covariant functor from the category C to the category Ab
of all abelian groups and their homomorphisms. We use the notation H p ( f ) = f p∗ .
H p is called the homology functor in the homology theory H.
Axiom H(3)(Commutativity Axiom). If f : (X, A) → (Y, B) is a continuous

map in C and if g : A → B is a continuous map in C defined by g(x) = f (x)
for all x ∈ A, then the diagram in Fig. 12.1 is commutative, i.e., g∗ ◦ ∂ = ∂ ◦ f ∗
This axiom connects the homology functor in the homology theory H with bound-
ary operator ∂ and induced homomorphisms.
Axiom H(4)(Exactness Axiom). If (X, A) is a topological pair in C and i : A →
X, j : X → (X, A) are the inclusion maps, then the sequence
i∗ j∗ ∂
· · · → H p (A) −−−→ H p (X ) −−−→ H p (X, A) −−−→ H p−1 (A) → · · ·
of groups and homomorphisms, called the homology sequence of (X, A), is exact.
Fig. 12.1 Diagram Hp (X, A)

f∗
/ Hp (Y, B)
connecting boundary
operator ∂ and induced
∂ ∂
homomorphisms in H
g∗

Hp−1 (A) / Hp−1 (B)
Remark 12.1.2 The above axioms H(1)–H(4) are algebraic axioms.
Axiom H(5)(Homotopy Axiom). If two continuous maps f, g : (X, A) → (Y, B)

in C are homotopic in C, then f p∗ = g p∗ for every integer p.
Axiom H(6)(Excision Axiom). If U is an open set of a topological space X where
closure U is contained in the interior Å of a subspace A of X (i.e., U ⊂ Å) and if
the inclusion map
i : (X − U, A − U ) → (X, A)
is in C, then the induced homomorphism
i ∗ : H p (X − U, A − U ) → H p (X, A)
is an isomorphism for every integer p.

The inclusion map i is called the excision of the open set U and i ∗ is called its
p-dimensional excision isomorphism.
Axiom H(7)(Dimension Axiom). The p-dimensional homology group H p (X ) of
a one-point space X = { point} in the homology theory H consists of a single
element for every integer p = 0, in symbol, H p ( point) = 0, for p = 0.
This completes the definition of a homology theory H on the given category C.
If H satisfies only the first six axioms H(1)–H(6), then H is called a generalized
homology theory on the category C.
The 0-dimensional homology group
G = H0 ( point)
is called the coefficient group of the homology theory H. Consequently, the dimen-
sion axiom locates the coefficient group at the right dimension.
Remark 12.1.3 The Eilenberg and Steenrod axioms for homology functors provide
an elegant and quick access to the further study of algebraic topology.
Remark 12.1.4 The construction of simplicial homology theory and its development
are given in Chap. 10. This homology theory applies to the category 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 .
12.2 The Uniqueness Theorem for Homology Theory
This section gives the uniqueness theorem for the axiomatic approach to homol-
ogy theory which deals with two homology theories in C with isomorphic coeffi-
cient groups. The most interesting result is the proof that on the category C of all
12.2 The Uniqueness Theorem for Homology Theory 423
topological pairs having homotopy types of finite CW-complex pairs all homology
theories satisfying the Eilenberg–Steenrod axioms have isomorphic groups. This
result concludes that there is only one concept of homology in that category. This
uniqueness theorem is very important in the development of algebraic topology.
Eilenberg and Steenrod proved that any two homology theories with isomorphic
coefficient groups on the category of all compact polyhedral pairs are isomorphic.
Let H = {H, ∗, ∂} and H = {H , , ∂ } be two arbitrary homology theories in
C. Suppose G = H0 ( point), G = H0 ( point) are their coefficient groups.
Definition 12.2.1 Let H and H be two homology theories on C. An isomorphism

(natural) ψ : H → H is a sequence of natural equivalences
ψn : Hn → Hn ,
for all n ≥ 0 such that the diagram in Fig. 12.2 is commutative for all pairs (X, A)
in C and for all n ≥ 0.
Theorem 12.2.2 Let G and G be abelian groups and h : G → G be a homomor-

phism. Then for every pair (X, A) in C and every integer n, there exists a unique
homomorphism
h n : Hn (X, A) → Hn (X, A)
such that
(i) h 0 = h on G = H0 ( point);
(ii) for every map f : (X, A) → (Y, B) in C and every integer n, the diagram in
Fig. 12.3 is commutative, i.e., h n ◦ f ∗ = f ◦ h n
(iii) for every pair of spaces (X, A) in C and every integer n, the diagram in Fig. 12.4
is commutative, i.e., h n−1 ◦ ∂ = ∂ ◦ h n .
Proof For proof see (Hu 1966, pp. 51). ❑
Fig. 12.2 Isomorphism of ∂ / Hn−1 (A)

Hn (X, A)
homology functors
ψn ψn−1

∂ / H (A)
Hn (X, A) n−1
Fig. 12.3 Diagram f∗

Hn (X, A) / Hn (Y, B)
involving h n and induced
hn hn
homomorphisms
f
Hn (X, A) / H (Y, B)
n
Fig. 12.4 Diagram ∂ / Hn−1 (A)

Hn (X, A)
connecting boundary
homomorphisms with h n
hn h n−1

∂ / H (A)
Hn (X, A) n−1
Remark 12.2.3 The unique homomorphism
h n : Hn (X, A) → Hn (X, A)
is an isomorphism of groups.
Theorem 12.2.4 (The Uniqueness Theorem) Let G and G be two abelian groups
and h : G → G be an isomorphism of groups. Then
h n : Hn (X, A) → Hn (X, A)
is also an isomorphism for every pair of spaces (X, A) in C and every integer n.
Proof Let k : G → G be the isomorphism of groups defined by k = h −1 . Then by
Theorem 12.2.2, there exists a unique homomorphism
k n : Hn (X, A) → Hn (X, A)
satisfying the conditions (i)–(iii) of Theorem 12.2.2 for every pair of topological
spaces (X, A) in C and every integer n. This shows that
(i) kn ◦ h n = Identity automorphism of the groups Hn (X, A);
(ii) h n ◦ kn = Identity automorphism of the groups Hn (X, A).
Consequently, h n is an isomorphism of groups. ❑
Remark 12.2.5 Given a coefficient group G, there exists only one homology theory
in the category C. Hence the name ‘The Uniqueness Theorem’ is justified.
12.3 Eilenberg–Steenrod Axioms for Cohomology Theory
This section presents an axiomatic approach to cohomology theory given by

Eilenberg and Steenrod, which is dual (parallel) to their homology theory. In fact
these two theories differ in only one point: homology functors are covariant; on the
other hand, cohomology functors are contravariant. Hence one can expect a dual
theorem in cohomology theory for every theorem established in homology theory.
The Eilenberg–Steenrod axioms for homology and cohomology functors make the
subject algebraic topology elegant and provides a quick access to further study.
12.3 Eilenberg–Steenrod Axioms for Cohomology Theory 425
A cohomology theory K on the category C consists of three functions K =

{H, ∗, δ} satisfying the following axioms:
(i) The first function H assigns to each topological pair (X, A) in the category
C and each integer p (positive, negative, or 0) an abelian group H p (X, A)
which is called the p-dimensional cohomology group of the topological pair
(X, A) in the cohomology theory K. In particular, for A = ∅, it is called the
p-dimensional (absolute) cohomology group of the space X .
(ii) The second function ∗ assigns to each f : (X, A) → (Y, B) in C and each
integer p a homomorphism
f ∗ = f p∗ : H p (Y, B) → H p (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 C and an integer
p a homomorphism
δ = δ(X, A, p) : H p−1 (A) → H p (X, A),
called the coboundary operator on the group H p−1 (A) in the cohomology
theory K.
Moreover, these three functions satisfy the following axioms C(1)–C(7), called
the Eilenberg–Steenrod axioms for cohomology theory on C:
Axiom C(1)(Identity Axiom). If 1 X : (X, A) → (X, A) is the identity map on a
topological pair (X, A) in C, then the induced homomorphisms
1∗X : H p (X, A) → H p (X, A)
is the identity automorphism of the cohomology group H p (X, A) for every inte-
ger p.
Axiom C(2)(Composition Axiom). If f : (X, A) → (Y, B) and g : (Y, B) →
(Z , C) are maps in C, then
(g ◦ f )∗p = f p∗ ◦ g ∗p : H p (Z , C) → H p (X, A)
Remark 12.3.1 The above axioms C(1) and C(2) show that for every fixed integer
p, the function H p forms a contravariant functor from the category C to the cate-
gory Ab. We use notation H p ( f ) = f p∗ . The functor H p is called the p-dimensional
cohomology functor in the cohomology theory K.
Axiom C(3)(Commutativity Axiom). If f : (X, A) → (Y, B) is a map in C and

if g : A → B is the map in C defined by g(x) = f (x) for all x ∈ A, then the
Fig. 12.5 Diagram g∗

H p−1 (B) / H p−1 (A)
connecting coboundary
operator δ with induced
homorphisms in K δ δ
f∗

H p (Y, B) / H p−1 (X, A)
diagram in Fig. 12.5 is commutative, i.e., δ ◦ g ∗ = f ∗ ◦ δ for every integer p.

This axiom connects the cohomology functors in the cohomology theory K with
the coboundary operator δ and induced homomorphisms.
Axiom C(4)(Exactness Axiom). If (X, A) is a topological pair in C and i : A →
X, j : X → (X, A) are inclusion maps, then the cohomology sequence
δ j∗ i∗
· · · → H p−1 (A) −−−→ H p (X, A) −−−−→ H p (X ) −−−→ H p (A) → · · ·
of the topological pair (X, A) is exact.
Remark 12.3.2 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 C are
homotopic in C, then
f p∗ = g ∗p

Axiom C(6)(Excision Axiom). If U is an open set of a topological space X whose
closure U is contained in the interior Å of a subspace A of X (i.e., U ⊂ Å) and if
the inclusion map i : (X − U, A − U ) → (X, A) is in C, then the induced homo-
morphism i ∗ : H p (X, A) → H p (X − U, A − U ) is an isomorphism for every
integer p.
The inclusion map i is called the excision of the open set U and i ∗ is called its
p-dimensional excision isomorphism.
Axiom C(7)(Dimension Axiom). The p-dimensional cohomology group H p (X )
of a one-point space X = { point} consists of a single element for every integer
p = 0, in symbol,
H p ( point) = 0, for p = 0.
This completes the definition of a cohomology theory K on the given category C.

If K satisfies only the first six axioms C(1)–C(6), then K is called a generalized
cohomology theory on the category C.
The 0-dimensional cohomology group G = H 0 ( point) is called the coefficient
group of the cohomology theory K.
Remark 12.3.3 The Uniqueness Theorem for cohomology theory is similar to that
of homology theory.
12.3 The Uniqueness Theorem for Homology Theory 427
Remark 12.3.4 Section 12.3 conveys an axiomatic approach to cohomology the-

ory announced by Eilenberg and Steenrod in 1945 but published in Eilenberg and
Steenrod (1952). The most interesting 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 result concludes that
there is only one concept of cohomology in that category.
12.4 The Reduced 0-dimensional Homology

and Cohomology Groups
This section conveys the concepts of 0-dimensional homology groups. Let P0 denote
a fixed reference point and also the space consisting of this single point in C0 . The
group H0 (P0 ) is as usual called the coefficient group of the given homology theory
H and is denoted by G.
Definition 12.4.1 Let G be a coefficient group of a homology theory H on C0 . Let
X and P0 be in C0 . If x ∈ X and g ∈ G, let (Gx) X denote the image of G in H0 (X )
under the homomorphism f ∗ induced by map f : P0 → X defined by f (P0 ) = x.
The image of G in H0 (X ) under f ∗ is denoted by (Gx) X .
Definition 12.4.2 If the unique map f : X → P0 is in C0 , then space X is said to
be collapsible. In such a case the kernel of the homomorphism f ∗ : H0 (X ) → G is
defined. It is called the reduced 0-dimensional homology group of X , denoted by
0 (X ).
H
Definition 12.4.3 If a topological space X in C0 is collapsible in the sense that the
unique map f : X → P0 is in C0 , then the image of G in H 0 (X ) under f ∗ is denoted
0 (X ) = H 0 (X )/G X is called the reduced 0-dimensional
by G X . The factor group H
cohomology group of X .
Definition 12.4.4 Let x ∈ X, h ∈ H 0 (X ) and f : P0 → X be given by f (P0 ) = x.
Then f ∗ (x) ∈ G is denoted by h(x). The kernel of f ∗ : H 0 (X ) → G is denoted by
x0 (X ).
H
12.5 Applications
This section presents some interesting applications of homology theory H = {H, ∗,

∂} and cohomology theory K = {H, ∗, δ} in the category C of all topological pairs
having homotopy types of finite CW-complex pairs. Let C0 be the category of all based
topological spaces having homotopy type of finite CW-complexes. Then C0 is a small
subcategory of the category C. For some direct consequences of Eilenberg–Steenrod
axioms see Chap. 13 and for various applications of homology and cohomology
theories see Chaps. 14 and 17.
12.5.1 Invariance of Homology Groups
This subsection proves invariance of homology groups in the sense that homeo-
morphic pairs of topological spaces in the category C have isomorphic homology
groups.
Theorem 12.5.1 A homeomorphism f : (X, A) → (Y, B) in the category C induces

isomorphisms
f ∗ : Hn (X, A) → Hn (Y, B), for every integer n.
Proof Since f −1 f = 1d , ( f −1 f )∗ = ( f −1 )∗ f ∗ = 1d . Similarly, f ∗ ( f −1 )∗ = 1d .

Consequently, f ∗ is an isomorphism with its inverse ( f ∗ )−1 = ( f −1 )∗ . ❑
Remark 12.5.2 Hn (X, A) is a topological invariant. It is also a homotopy invariant

in the sense that if f : (X, A) → (Y, B) in the category C is a homotopy equivalence,
then it induces isomorphisms
f∗ : Hn (X, A) → Hn (Y, B), for every integer n.
12.5.2 Invariance of Cohomology Groups
This subsection proves invariance of cohomology groups in the sense that homeo-
morphic pairs of topological spaces in the category C have isomorphic cohomology
groups.
Theorem 12.5.3 A homeomorphism f : (X, A) → (Y, B) in the category C induces

isomorphisms
f ∗ : H n (Y, A) → H n (X, B), for every integer n.
Proof The proof is similar to that of Theorem 12.5.1. ❑
Remark 12.5.4 H n (X, A) is a topological invariant. It is also a homotopy invariant.
12.5.3 Mayer–Vietoris Theorem
This subsection presents an application of “Excision Axiom” which provides a tech-

nique to compute homology groups. For example, Mayer–Vietoris theorem (see
Sect. 10.12) leads in this respect.
α1 γ1 δ1
· · · −→ Hn (A) / Hn (X1 ) / H( X1 , A) / Hn−1 (A) −→ · · ·
α2 β1 α α2
β2
γ2

δ2
· · · −→ Hn (X2 ) / Hn (X) / H ( X 1 , X2 ) / Hn−1 (X2 ) −→ · · ·
Fig. 12.6 Four lemma diagram
Definition 12.5.5 A topological triad (X ; A, B) consists of a topological space X

together with an ordered pair (A, B) of subspaces A and B of X . The topological
triad (X ; A, B) is said to be proper 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 12.5.6 (Mayer–Vietoris Theorem) Let X, X 1 , X 2 and A be topological

spaces in C0 such that X = X 1 ∪ X 2 , A = X 1 ∩ X 2 . If the inclusion (X 1 , A) →
(X, X 2 ) is an excision, then there is a long exact sequence in homology, called
Mayer–Vietoris sequence of the proper topological triad (X ; X 1 , X 2 ) :
α β
→ Hn (A) −−−→ Hn (X 1 ) ⊕ Hn (X 2 ) −−−→ Hn (X ) −−−→ Hn−1 (A) −
··· − → ···
Proof Consider the commutative diagram with two long exact homology sequences
provided by axiom H(4), where by assumption α : Hn (X 1 , A) → Hn (X, X 2 ) is an
isomorphism by Excision Axiom H(6). Then use four lemma to complete the proof
(Fig. 12.6). ❑
12.6 Exercises
In this section we use the notations described in Sect. 12.4.

1. If f : X → Y ∈ C0 , x ∈ X, y = f (x) and g ∈ G, then show that f ∗ maps (Gx) X
onto (Gy)Y .
· · · −→ Hn (A) / Hn (X) / H( X, A) / Hn−1 (A) / Hn−1 (X) −→ · · ·

· · · −→ Hn (B) / Hn (Y ) / H( Y, B) / Hn−1 (B) / Hn−1 (Y ) −→ · · ·
Fig. 12.7 Five lemma diagram
2. If f : X → Y ∈ C0 and Y is collapsible, show that X is collapsible. If (X, A) ∈ C

and X is collapsible, show that A is collapsible and the map (X, A) → (P0 , P0 ) ∈
C. If P is a space consisting of a single point, show that H0 (P) = 0 and H0 (P) =
(G, P) P .
3. If X is collapsible and x ∈ X , show that H0 (X ) decomposes into the direct sum
H0 (X ) = H 0 (X ) ⊕ (Gx) X and the correspondence g → (gx) X maps G isomor-
phically onto (Gx) X .
0 (X ) =
4. If X is a space consisting of a single point, show that H 0 (X ) = G X and H
0.
5. Let f : X → Y, x ∈ X, y = f (x), h ∈ H 0 (Y ). Show that
(i) ( f ∗ h)(x) = h(y);
(ii) f ∗ maps H y0 (Y ) into H
x0 (X );
(iii) Hy (Y ) contains the kernel of f ∗ .
0
6. If X is a collapsible space and x is in X , show that

x0 (X ) ⊕ G X ;
(i) H 0 (X ) decomposes into H 0 (X ) = H
(ii) The map f : X → P0 induces an isomorphism from G onto G X .
7. Let G be a coefficient group of a homology theory H on C0 . If f : X → Y in
C0 is collapsible, x ∈ X, and y = f (x), show that X is collapsible and f ∗ maps
H0 (X ) into H
0 (Y ) and maps (Gx) X isomorphically onto (Gy)Y .
8. Let f : (X, A) → (Y, B) be a map of pair of spaces. If both f : X → Y and g :
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. 12.7, and use five lemma result]


[4] Hilton, P.J. and Wylie, S. Homology Theory, Cambridge University Press, Cam-
bridge, 1960.
[5] Switzer, R.M., Algebraic Topology–Homotopy and Homology, Springer-Verlag,
References
(1972)
Dieudonné, J.: A History of Algebraic and Differential Topology, 1900–1960. Modern Birkhäuser
Classics. Birkhäuser, Basel (1989)
ton (1952)
Hu, S.T.: Homology Theory. Holden Day, Oakland (1966)
Maunder, C.R.F.: Algebraic Topology. Van Nostrand Reinhold, London (1970)
Rotman, J.J.: An Introduction to Algebraic Topology, Springer, New York, 1988
Chapter 13
Consequences of the Eilenberg–Steenrod
Axioms
This chapter continues the study of homology and cohomology theories by

considering some immediate consequences of the Eilenberg–Steenrod axioms:
H(1)–H(7) and C(1)–C(7) given by Eilenberg and Steenrod for homology and
cohomology theories described in Chap. 12. Finally, this chapter establishes a close
connection between cofibrations and homology theory, and computes the ordinary
homology groups of S n with coefficients in an abelian group G.
For this chapter, the books (Eilenberg and Steenrod 1952; Gray 1975; Hu 1966;
Spanier 1966) and some others are referred in Bibliography.
13.1 Immediate Consequences
This section deals with some properties of homology and cohomology groups which
directly follow from the Eilenberg and Steenrod axioms of homology and cohomol-
ogy theories. Let H = {H, ∗, ∂} denote an arbitrary given homology theory on the
category C of topological pairs having homotopy type of finite CW -complex pairs.
Then C is a full subcategory of the category of pairs of topological spaces and maps
of pairs and this admits the construction of mapping cones. Let C0 denote the cate-
gory of pointed spaces having homotopy of finite C W -complexes. Throughout this
chapter, it is assumed that (X, A) is in C and X is in C0 unless stated otherwise.
We first establish the homotopy invariance of the homology groups in H.
Theorem 13.1.1 Let f : (X, A) → (Y, B) in C be a homotopy equivalence. Then
the induced homomorphism
f ∗ : Hn (X, A) → Hn (Y, B)
is an isomorphism for every n.

DOI 10.1007/978-81-322-2843-1_13
434 13 Consequences of the Eilenberg–Steenrod Axioms
Proof By hypothesis, f is a homotopy equivalence. Hence there exists a map

g : (Y, B) → (X, A) in C such that g ◦ f is homotopic to the identity map on the
topological pair (X, A) and f ◦ g is homotopic to the identity map on the topological
pair (Y, B).
Consequently, by Axioms H(1) and H(2) it follows that
g∗ ◦ f ∗ = (g ◦ f )∗ : Hn (X, A) → Hn (X, A)
and
f ∗ ◦ g∗ = ( f ◦ g)∗ : Hn (Y, B) → Hn (Y, B)
are the identity automorphisms of the groups Hn (X, A) and Hn (Y, B), respectively.
This shows that f ∗ is an isomorphism with g∗ as the inverse of f ∗ for every integer
n. ❑
Corollary 13.1.2 Let X and Y be topological spaces in C0 such that X is homotopy

equivalent to Y . Then the groups Hn (X ) and Hn (Y ) are isomorphic for every integer
n.
Proof The corollary follows from Theorem 13.1.1 by taking in particular A = ∅ and
B = ∅. ❑
Remark 13.1.3 For every integer n, the n-dimensional homology groups Hn (X ) of

space X in C0 and the n-dimensional homology groups Hn (X, A) of (X, A) in C are
both homotopy invariants.
Corollary 13.1.4 Let the group G (abelian) be the coefficient group of the homology
theory H. If a topological space X is contractible, then
H0 (X ; G) ∼
= G,
and
Hn (X ; G) = 0 for n = 0.
Proof By hypothesis X is contractible. Hence X is homotopically equivalent to the

distinguished singleton space {∗}. Hence the corollary follows from Corollary 13.1.2.
❑
We now consider some consequences of the Exactness Axiom: H(4) of Chap. 12.
Proposition 13.1.5 Let X be a topological space in C0 and A be a subspace of X .

If the inclusion map i : A → X is a homotopy equivalence, then Hn (X, A) = 0 for
every integer n.
Proof As i : A → X is a homotopy equivalence it follows from Theorem 13.1.1

that the induced homomorphism
13.1 Immediate Consequences 435
i ∗ : Hn (A) → Hn (X )
is an isomorphism for every integer n.

We now consider the homology sequence of the pair of topological spaces (X, A)
in C.
i∗ j∗
· · · → Hn (A) −−−→ Hn (X ) −−−→ Hn (X, A)
∂ i∗
−−−→ Hn−1 (A) −−−→ Hn−1 (X ) → · · · (13.1)
Since the two homomorphisms i ∗ in (13.1) are isomorphisms, if follows from exact-
ness of this sequence that Hn (X, A) consists of a singleton element for every integer
n. In other words, Hn (X, A) = 0 for every n. ❑
Corollary 13.1.6 Given any topological space X in C0 ,
Hn (X, X ) = 0
for every integer n.
Proof The proof follows from Proposition 13.1.5 by taking in particular,

A = X. ❑
For the pair (X, A) of topological spaces, we now establish some relations between
homology groups of (X, A), X and A.
Proposition 13.1.7 (a) If X is a topological space and A is a retract of X , then

(i) the inclusion map i : A → X induces a monomorphism
i∗ : Hn (A) → Hn (X ),
for every integer n;

(ii) the inclusion map j : X → (X, A) induces an epimorphism
j∗ : Hn (X ) → Hn (X, A),
for every integer n;

(iii) the boundary operator
∂ : Hn (X, A) → Hn−1 (A)
is a trivial homomorphism for every integer n;

(b) Hn (X ) ∼
= Hn (A) ⊕ Hn (X, A)
Proof (a) By hypothesis, A is a retract of X . Hence there exists a retraction r :

X → A such that r (x) = x for every x ∈ A. This shows that r ◦ i : A → A
is the identity map on A. By using axioms H(1) and H(2), it follows that the
composite homomorphism
i∗ r∗
Hn (A) −−−→ Hn (X ) −−−→ Hn (A) (13.2)
is the identity automorphism of the group Hn (A) for every integer n. This implies
from (13.2) 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∗
We now consider the homology sequence of the pair (X, A):
i∗ j∗
· · · → Hn (A) −−−→ Hn (X ) −−−→ Hn (X, A)
∂ i∗
−−−→ Hn−1 −−−→ Hn−1 (X ) → · · · (13.3)
As i ∗ : Hn−1 (A) → Hn−1 (X ) is a monomorphism, by using the exact sequence

(13.3), we find that ∂ is a trivial homomorphism and j∗ is an epimorphism.
(b) Let n be an arbitrary given integer. Since i ∗ : Hn (A) → Hn (X ) is a monomor-
phism, it follows that Im i ∗ ∼= Hn (A). Again from the exact sequence (13.3), it
follows that ker j∗ = Im i ∗ . Since Hn (X ) = Im i ∗ ⊕ ker r∗ and j∗ is an epimor-
phism, it follows that
ker r∗ ∼
= Hn (X )/Im i ∗ = Hn (X )/ker j∗
by Isomorphism Theorem as shown in Fig. 13.1.

This shows that
Hn (X ) = Hn (A) ⊕ Hn (X, A)
for every integer n. ❑
Corollary 13.1.8 Let G be the coefficient group of the homology theory H. Then in
H, for every point x0 of a topological space X ,
Fig. 13.1 Diagram j∗

Hn (X) / Hn (X, A)
OOO n7
involving isomorphism of OOO
O nnnnn
homology groups p OOOO nnn∼
' nnn =
Hn (X)/ker j∗
H0 (X ) ∼
= G ⊕ H0 (X, x0 ),
Hn (X ) ∼
= Hn (X, x0 ) for n = 0.
Proof Since every singleton subspace of a topological space X is a retract of X , the

corollary follows from Propositions 13.1.7 and 13.1.4. ❑
We now study the effect of deformation retraction on the corresponding homology

groups.
Proposition 13.1.9 Let a topological X be deformable into a subspace A of X . Then

for every integer n,
(a) (i) the inclusion map i : A → X induces an epimorphism
i∗ : Hn (A) → Hn (X );
(ii) the inclusion map j : X → (X, A) induces trivial homomorphism
j∗ : Hn (X ) → Hn (X, A);
(iii) the boundary operator
∂ : Hn (X, A) → Hn−1 (A)
is a monomorphism.
(b) Hn (A) ∼= Hn (X ) ⊕ Hn+1 (X, A).
Proof (a) As X is deformable into its subspace A, there exists a homotopy
h t : X → X, for all t ∈ I
such that h 0 = 1 X and h 1 (X ) ⊂ A. Define a map
h : X → A, x → h 1 (x).
Then the composite map i ◦ h
h
X −−−→ A → X
is homotopic to h 0 = 1 X . By using axioms H(1), H(2), and H(5), it follows that

the composite homomorphism i ∗ ◦ h ∗
h∗ i∗
Hn (X ) −−−−→ Hn (A) −−−→ Hn (X )
is the identity automorphism of the group Hn (X ) for every integer n. Conse-

quently it implies that i ∗ is an epimorphism and h ∗ is a monomorphism and the
abelian group Hn (A) decomposes into the direct sum
Hn (A) = Im h ∗ ⊕ ker i ∗
We now consider the homology sequence of the pair (X, A)
∂ i∗
→ Hn+1 (X, A) −−−→ Hn (A) −−−→ Hn (X )
··· −
j∗ ∂
−−−→ Hn (X, A) −−−→ Hn−1 (A) −
→ ··· (13.4)
As i ∗ is an epimorphism, it follows from the exact sequence (13.4) that j∗ is a

trivial homomorphism and ∂ is a monomorphism.
(b) Let n be an arbitrary given integer. Then the induced homomorphism
h ∗ : Hn (X ) → Hn (A)
is a monomorphism. Hence Hn (X ) ∼
= Im h ∗ . Again for the exact sequence (13.4),
it follows that
ker i ∗ = Im ∂.
Again since ∂ : Hn+1 (X, A) → Hn (A) is a monomorphism, it implies that

Im ∂ ∼
= Hn+1 (X, A). Hence it follows that
Hn (A) = Im h ∗ ⊕ ker i ∗ ∼
= Hn (X ) ⊕ Hn+1 (X, A)
for every integer n. ❑
Corollary 13.1.10 Let G be the coefficient group of the homology theory H and X
be a contractible space. If A is a nonempty subspace of X , then
H0 (A) ∼
= G ⊕ H1 (X, A),
Hn (A) ∼
= Hn+1 (X, A), n = 0.
Proof Since A = ∅ is a subspace of a contractible space X , it follows that X is

deformable into A. Consequently the corollary follows from the Proposition 13.1.9
and Corollary 13.1.4. ❑
We are now in a position to study the Excision Axiom H(6). Here X − U is written
as X \U .
Theorem 13.1.11 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 C the excision
e : (X \U, A\U ) → (X, A)
e∗ : Hn (X \U, A\U ) → Hn (X, A)
for every integer n if there exists an open set V of 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.
Proof By hypothesis V ⊂ U ⊂ A. Hence V ⊂ Å (interior of A). Then by Excision

Axiom H(6), it follows that the excision
ẽ : (X \V, A\V ) → (X, A)
ẽ∗ : Hn (X \V, A\V ) → Hn (X, A)

Again as
i : (X \U, A\U ) → (X \V, A\V )
is a homotopy equivalence, it induces an isomorphism
i ∗ : Hn (X \U, A\U ) → Hn (X \V, A\V )
for every integer n. We now consider the composite map
ẽ
e = ẽ ◦ i : (X \U, A\U ) → (X \V, A\V ) −−−→ (X, A).
Then by Axiom H(2), it follows that
e∗ = ẽ∗ ◦ i ∗ : Hn (X \U, A\U ) → Hn (X, A)
is an isomorphism for every integer n. ❑

13.2 Applications
This section presents some applications derived as further consequences of Eilenberg

and Steenrod axioms such as relation between cofibrations and homology. 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. Throughout this section it is assumed that (X, A)
is in C and X is in C0 unless stated otherwise.
13.2.1 Cofibration and Homology
There is close relation between cofibrations and homotopy theory. This subsection
establishes some relations between cofibrations and homology theory H = {H, ∗, ∂}.
Theorem 13.2.1 If i : A → X is a cofibration and a ∈ A, then the projection
p : (X, A) → (X/A, {a})
induces isomorphisms in homology.
Proof Since i is a cofibration, its mapping cone Ci is homotopy equivalent to X/A.

Again since Ci = X̊ ∪ C A, we have an inclusion j : (X, A) → (Ci , C A). This
induces isomorphisms in homology. But (C A, {a}) is contractible and Ci /C A is
homeomorphic to X/A. Hence it follows that the projection
p : (X, A) → (X/A, {a})
induces isomorphisms
Hn (X, A) ∼
= Hn (Ci , C A) ∼
= Hn (X/A, {a}).
❑
Theorem 13.2.2 If i : {∗} → X is a cofibration, then Hn (X ) ∼

= Hn (X, {∗}) ⊕
Hn ({∗}).
Proof As i : {∗} → X is a cofibration, there exists a map p : X → {∗} such that

p ◦ i = 1d . Hence the sequence
→ Hn ({∗}) −
··· − → Hn (X ) −
→ Hn (X, {∗}) −
→ ···
splits for every n and hence the the desired result is proved. ❑
13.2.2 Computing Ordinary Homology Groups of Sn
This subsection computes the ordinary homology groups of S n with coefficients in

an abelian group G.
Theorem 13.2.3 If (X, {x0 }) ∈ C, then for all n ∈ Z, there is an isomorphism
n : Hn (X, {x 0 }) → Hn+1 ( X, {∗}).

σn =
∼ Hn (X, {x0 }), and compose these

Proof Consider the isomorphisms Hn+1 (C X, X ) =
with those homomorphisms which are induced by the homeomorphisms C X/ X ≈
X and the projection
(C X, X ) → (C X/ X, {∗}).
Theorem 13.2.4 For m, n ∈ Z with m ≥ 0,
Hn (S m , {∗}) ∼
= Hn+1 (S m+1 , {∗}) ∼
= Hn−m (S ◦ , {∗}) ∼
= Hn−m ({∗}).
Proof The first two isomorphisms follows from suspensions S m+1 = S m =

m+1 S ◦ . The other isomorphism follows from the triad (S ◦ , {−1}, {1}) and inclusion
({−1}, ∅) → (S ◦ , {+1}). ❑
Theorem 13.2.5 Let H∗ be an ordinary homology 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.
Proof Use Theorem 13.2.2 to obtain Hn (S m ) ∼ = Hn (S m , {∗}) ⊕ Hn ({∗}) = Hn

(S , {∗}) for n = 0. Again from Theorem 13.2.4, we have Hn (S m , {∗}) ∼
m
= Hn−m ({∗}).

G, if n = m
Hn (S m , {∗}) ∼
=
0, otherwise.
Combining these, we have H0 (S ◦ ) ∼

= G ⊕ G, and for m = 0, H0 (S m ) ∼
=
m ∼
Hm (S ) = G. ❑
13.3 Exercises
In this section H = {H, ∗, ∂} denotes an arbitrary given homology theory on the

category C of topological pairs having homotopy type of finite CW -complex pairs.
1. If f : X → Y is a continuous map homotopic to a constant map, show that for

every integer n(n = 0) the induced homomorphisms
f∗ : Hn (X ) → Hn (Y )
on the homology groups are trivial.

2. Let f : (X, A) → (Y, B) be deformable into the subspace B of Y . Show that for
every integer n, the induced homomorphism
f ∗ : Hn (X, A) → Hn (Y, B)
in the given homology theory is trivial.

3. By using homology prove Brouwer fixed point theorem: Every continuous map
from the closed n-ball to itself, for n ≥ 1, has a fixed point.
4. If A → X is a weak retract, prove that for all n ∈ Z, Hn (X ) ∼ = Hn (A) ⊕
Hn (X, A).
5. For a weak deformation retract A of X , show that Hn (X, A) = 0. Further show
that Hn (X, X ) = 0.
6. Show that for n = m, the spheres S n and S m cannot be homeomorphic.
(Hint: For n = m, S n and S m have different homotopy types and hence they
cannot be homeomorphic.)
7. Show that for n = m, Rn and Rm cannot be homeomorphic.
(Hint: Add one point to compactify each Rn and Rm to obtain S n and S m , which
would be homeomorphic if Rn and Rm were homeomorphic.)
8. Let G be the coefficient group of a given homology theory H. Show that for any
singleton space X

0, p = 0
H p (X ; G) =
G, p = 0.
9. Let X be a discrete space with n distinct points. Show that

Gn , if p = 0
H p (X ; G) ∼
=
0, if p = 0,
where G n denotes the direct sum of n copies of the coefficient group G in the
homology theory H.
10. If a continuous map f : X → Y has a left (right) homotopy inverse, show that
the induced homomorphism
13.3 Exercises 443
f ∗ : Hn (X ) → Hn (Y )
in the homology theory H is a monomorphism (an epimorphism) for each n.

References
(1972)
Dieudonné, J.: A History of Algebraic and Differential Topology, 19001960, Modern Birkhäuser
(1989)
ton (1952)
(1975)
Hu, S.T.: Homology Theory. Holden-Day Inc, San Francisco (1966)
Switzer, R.M.: Algebraic Topology-Homotopy and Homology. Springer, Heidelberg (1975)
Whitehead, G.W.: Elements of Homotopy THeory. Springer, New York (1978)
Chapter 14
Applications
In earlier chapters some applications of algebraic topology have been discussed.

This chapter conveys further applications to understand the scope and power of
algebraic topology displaying the great beauty of the subject. Some concepts ini-
tially introduced in homology and homotopy theories to solve problems of topology
have found fruitful applications to other areas of mathematics. More precisely, this
chapter conveys some interesting applications of homotopy and homology theories.
For example, Hopf classification theorem, Borsuk–Ulam theorem, Hairy Ball the-
orem, Ham Sandwich theorem, Lusternik–Schnirelmann theorem, Lefschetz fixed
point theorem, van Kampen theorem are proved and also some results related to
graphs, Mayer–Vietoris sequence, fixed points of continuous maps, vector fields
and applications to algebra are studied in this chapter. Algebraic topology is now
witnessing potential applications to various areas of science and engineering. This
chapter also indicates some applications of algebraic topology to physics, chemistry,
economics, biology, medical science, and engineering with specific references.
For this chapter the books Armstrong (1983), Croom (1978), Dodson and Parker
(1997), Gray (1975), Hatcher (2002), Nakahara (2003), Spanier (1966) with some
other books and papers are referred in Bibliography.
14.1 Degrees of Spherical Maps and Their Applications
This section introduces the concept of ‘degree of a spherical map’ and applies it to
prove Brouwer degree theorem for an arbitrary degree, Hopf’s classification theorem
and Brouwer fixed point theorem. The degree of a spherical map f : S n → S n was
defined and studied by L.E.J. Brouwer (1881–1967) during 1910–1912 to examine
whether given two spherical maps are homotopic or not. He took the first step towards
connecting the two basic concepts: homotopy and homology in topology by using
his concept of degree of a spherical map which offers interesting applications. The
DOI 10.1007/978-81-322-2843-1_14
446 14 Applications
classical definition of the degree of a spherical map f : S n → S n given by Brouwer

prior to the rigorous development of homology theory, is more intuitive than its
definition from the view point of homology theory. The latter definition is more
elegant but the geometric flavor is perhaps lost.
The concept of winding number of a curve with respect to a point in complex
analysis or the concept of index of a vector field given by H. Poincaré (1854–1912)
contained implicitly the idea of degree of a continuous map prior to Brouwer. This
concept of degree of a spherical map is used to solve some problems. After having
generalized Brouwer’s result to an arbitrary dimension, H. Hopf (1895–1971) under-
took a systematic study of the problem of classifying the continuous mappings of a
polytope into a polytope (see Chap. 18).
14.1.1 Degree of a Spherical Map
This subsection introduces the concept of the degree function of spherical maps
f : S n → S n through homology and characterizes homotopy property of spherical
maps by their degrees. Recall that a group homomorphism f : Z → Z is completely
determined by the image f (1) of its generator 1 ∈ Z; i.e., f is simply multiplication
by the integer f (1). This leads to the following concept of the degree of a spherical
map.
Definition 14.1.1 Given a continuous map f : S n → S n and a triangulation K of S n

for an integer n ≥ 1, there is a homeomorphism h : |K | → S n ; and a homomorphism
ψ : Hn (K ; Z) → Hn (K ; Z) defined by
ψ = (h −1 ◦ f ◦ h)∗ : Hn (K ; Z) → Hn (K ; Z).
Since Hn (K ; Z) is isomorphic to Z, there exists a unique integer m with the property

that ψ(x) = m·x for x ∈ Hn (K ; Z). This unique integer m is called the degree of f ,
and is denoted by deg f . The function d : f → deg f is called degree function.
Remark 14.1.2 Given a continuous map f : S 1 → S 1 , if x moves around S 1 , then

its image f(x) moves around S 1 , some integral number of times. This integer is called
the degree of f. The degree function d sets up a (1-1) correspondence between the
set [S 1 , S 1 ] of homotopy classes of continuous maps f : S 1 → S 1 and the set Z of
integers.
Let m be the degree of a spherical map f : S n → S n obtained by a triangulation

K of S n . A natural question arises: does this degree depend on a particular choice of
the triangulation of S n ?.
To get its answer consider another triangulation L of S n . Then there exists another
homeomorphism k : |L| → S n . This shows that β = k −1 ◦ h : |K | → |L| and
β −1 : h −1 ◦k : |L| → |K | are both homeomorphisms. Hence (k −1 ◦ f ◦k) : |L| → |L|
14.1 Degrees of Spherical Maps and Their Applications 447
is a continuous map and its induced homomorphism (k −1 ◦ f ◦ k)∗ : Hn (L; Z) →

Hn (L; Z) is such that
(k −1 ◦ f ◦ k)∗ (x) = ((k −1 ◦ h) ◦ (h −1 ◦ f ◦ h) ◦ (h −1 ◦ k))∗ (x)

= (β∗ ◦ (h −1 ◦ f ◦ h)∗ ◦ β∗−1 )(x)
= (β∗ (m(β∗−1 ))(x))
= β∗ ◦ (β∗−1 )(m ·x) for all x ∈ Hn (L; Z)
= m ·x for all x ∈ Hn (L; Z).
This shows that deg f of a map f : S n → S n does not depend on a particular

choice of triangulations of S n . Consequently, it follows that the deg f of a spherical
map is well defined.
We now present some interesting properties of degree functions of spherical maps
and classify such maps with the help of their degrees.
Theorem 14.1.3 (Brouwer’s degree theorem) Let f, g : S n → S n be two homotopic
maps. Then deg f = deg g., i.e., homotopic spherical maps have the same degrees.
Proof Let f, g : S n → S n be two homotopic maps. As their degrees do not depend
on the triangulation k : |K | → S n , we fix a triangulation k. Then the maps
ψ = h −1 ◦ f ◦ k and φ = k −1 ◦ g ◦ k are homotopic. This shows by the homo-
topy axiom H(5) of Eilenberg–Steenrod that ψ∗ = φ∗ : Hn (K ; Z) → Hn (K : Z)
(see Chap. 12). This implies by Definition 14.1.1 that deg f = deg g. ❑
Proposition 14.1.4 (i) The identity map 1 S n : S n → S n has degree +1;
(ii) If f : S n → S n and g : S n → S n are continuous maps, then deg(g ◦ f ) =
deg g·deg f ;
(iii) If f : S n → S n is a homeomorphism, then deg f = ±1.
Proof (i) The identity map 1 S n : S n → S n induces the identity homomorphism in
homology groups. Hence (i) follows.
(ii) Given a triangulation K of S n , there is a homeomorphism k : |K | → S n . If
deg f = n 1 and deg g = n 2 , then k −1 ◦ g ◦ f ◦ k : |K | → |K | is a continuous
map and its induced homomorphism
(k −1 ◦ g ◦ f ◦ k)∗ : Hn (K ; Z) → Hn (K : Z)
is such that
(k −1 ◦ g ◦ f ◦ k)∗ (x) = ((k −1 ◦ g ◦ k) ◦ (k −1 ◦ f ◦ k))∗ (x)

= (k −1 ◦ g ◦ k)∗ (n 1 ·x)
= n 2 ·(n 1 ·x)
= (n 2 n 1 )·x for all x ∈ Hn (K ; Z).
by using the definition of degree of a spherical map.
Hence deg(g ◦ f ) = deg g·deg f.

448 14 Applications
(iii) Let h : S n → S n be a homeomorphism. Then h −1 ◦ h = 1 S n : S n → S n implies

by (i) and (ii) that
deg(h −1 ◦ h) = 1 = deg h −1 ·deg h.
Since both deg h and deg h −1 are integers, it follows that deg h = deg h −1 is
either +1 or −1.
❑
Remark 14.1.5 The classical definition of the degree of a spherical map f : S n → S n

given by Brouwer is more intuitive than its definition from the view point of homology
theory. Brouwer defined deg f as the number of times that the domain sphere wraps
around the range sphere. His definition shows that if f : S 1 → S 1 , z → z n , then
deg f = n; if f : S n → S n , n ≥ 1 is a constant map, then deg f = 0 and if
f : S n → S n , n ≥ 1 is the identity map, then deg f = +1.
We are now in a position to study homotopy properties of spherical maps and

their fixed points with the help of their degrees.
Proposition 14.1.6 If ri : S n → S n is the reflection map defined by
ri : (x1 , x2 , . . . , xi , . . . , x n+1 ) → (x1 , x2 , . . . , −xi , . . . , x n+1 ),
then deg ri = −1.
Proof We first consider the case i = 1. Let g = r1 : S n → S n , be the continuous

map defined by
g(x 1 , x2 , . . . , xi , . . . , x n+1 ) = (−x 1 , x2 , . . . , x i , . . . , xn+1 )
Let k : S n → S n be the homeomorphism which interchanges the coordinates x1

and xi . Then deg k = ±1. Hence it follows that ri = k ◦ g ◦ k. This shows that
deg ri = deg k·deg g·deg k = deg g. we now claim that deg g = −1. For n = 0, it is
trivial and for all n ≥ 1, the result follows by iterated suspension. ❑
Theorem 14.1.7 (Antipodal degree) The antipodal map A : S n → S n , x → −x

has degree (−1)n+1 for n ≥ 1.
Proof Using the notation of Proposition 14.1.6, we see that A = r1 ◦ r2 ◦ · · · ◦ rn+1

and deg ri = −1 for each i. Hence it follows that deg A = (−1)n+1 . ❑
Corollary 14.1.8 If a continuous map f : S n → S n has no fixed point, then f has

degree (−1)n+1 for n ≥ 1.
Proof Let f : S n → S n be a continuous map having no fixed point. Then the line
segment (1 − t) f (x) − t x does not pass through the origin for any t in I and any x
in S n . Consider now the continuous map
(1 − t) f (x) − t x
F : S n × I → S n : (x, t) → .
||(1 − t) f (x) − t x||
As the map A : S n → S n , x → −x is the antipodal, it follows that F : f A..

Hence deg f = deg A = (−1)n+1 . ❑
Corollary 14.1.9 If a continuous map f : S 2n → S 2n is such that f 1 S 2n , then f

has a fixed point.
Proof By hypothesis f 1 S 2n . Then deg f = +1. If possible, f has no fixed

point, then Corollary 14.1.8 implies that deg f = (−1)2n+1 = −1. Hence we have a
contradiction. In other words, f has a fixed point. ❑
14.1.2 Hopf Classification Theorem
This subsection provides a complete homotopy classification of spherical maps with

the help of their degrees which are integers. H. Hopf generalized this result of
Brouwer to an arbitrary dimension n and made a systematic study of the classi-
fication problems of continuous mappings between certain class of spaces, called
polytopes (see Chap. 18). H. Hopf proved in 1927 that the converse of Brouwer
degree theorem is also true for arbitrary dimension n. These two combined results
are known as ‘Hopf Classification Theorem’; of course Brouwer proved a partial
converse for n = 2: if f and g are continuous maps on the 2-sphere which have the
same degree, then f g.
Definition 14.1.10 Two continuous maps f and g are said to belong to the same
homology class if they induce identical homomorphisms of homology groups (for
all dimensions and all coefficient groups), and they are said to belong to the same
homotopy class if they can be embedded into a common one-parameter continuous
family of mappings.
Theorem 14.1.11 (Hopf Classification Theorem) Let f, g : S n → S n be two con-

tinuous maps. Then f g iff deg f = deg g.
Proof Suppose that deg f = deg g. We claim that f g. By using induction on n,

we prove the theorem. For n = 1, a map with degree m is representable as a periodic
real function on R which increases by m each time as its argument increases by 1.
Clearly, any two such maps are homotopic. Next suppose that the result is valid for
n − 1. Then the two maps f and g (where deg f = deg g) admit representation
as the suspensions of maps of S n−1 to itself. As the suspension map preserves the
degree, then deg f = deg( f ) and deg g = deg(g) by Ex. 12 of Sect. 14.11.
Hence it follows by induction hypothesis that the maps f and g are homotopic.
Conversely, let f and g be homotopic maps. Then deg f = deg g by Brouwer’s
degree Theorem 14.1.3. ❑
450 14 Applications
Hopf extended Brouwer’s definition of degree to maps from polyhedra to spheres

and extended his classification theorem in 1933 to such maps:
Theorem 14.1.12 (Extended Hopf classification theorem) If X is a polyhedron of

dimension not exceeding n and f, g : X → S n are two given continuous maps. Then
f g iff deg f = deg g.
Proof See Spanier (1966). ❑
14.1.3 The Brouwer Fixed Point Theorem
This subsection proves Brouwer fixed point theorem and its immediate consequences
by using homology theory. L.E.J. Brouwer 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. The papers of H. Poincaré during 1895–1904
can be considered as blue prints for theorems to come. The results of Brouwer during
1910–1912 may be considered as the first one of the proofs in algebraic topology. He
proved the celebrated theorem ‘Brouwer fixed point theorem’ by using the concept
of degree of a continuous spherical map defined by Brouwer himself.
Proposition 14.1.13 The n-sphere S n is not contractible for any finite n ≥ 0.
Proof Let 1 S n : S n → S n be the identity map. If possible, S n is contractible. Then

1 S n c for some constant map c. But 1 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 and
hence S n is not contractible for n ≥ 1. Again for n = 0, S 0 = {−1, 1} is a discrete
space. Hence S 0 can not be contractible. Consequently, the proposition follows for
every n ≥ 0. ❑
Remark 14.1.14 The infinite dimensional sphere S ∞ is contractible (see Chap. 2)

but S n is not so for any finite n ≥ 0.
Theorem 14.1.15 (Brouwer no retraction theorem) There exists no continuous onto

map f : D n+1 → S n which leaves every point of S n fixed for each integer n ≥ 0.
Proof If possible, there exists a continuous onto map f : D n+1 → S n for every
n ≥ 0 such that f (x) = x for all x ∈ S n . Define a map.
H : S n × I → S n , (x, t) → f ((1 − t)x)
Then H is a continuous map such that H : f c for some constant map c. This
implies that S n is contractible, which is a contradiction. ❑
Theorem 14.1.16 (Brouwer fixed point theorem) Every continuous map f :

D n+1 → D n+1 has a fixed point for every integer n ≥ 0.
Proof If possible, f : D n+1 → D n+1 has no fixed point for every integer n ≥ 0.
This implies in this case that f (x) and x are distinct points and hence f (x)
= x
for all x ∈ D n+1 . If n = 0, it an immediate contradiction. Hence it is well assumed
from now that n ≥ 1. By assumption, for each x ∈ D n+1 , the points x and f (x)
are distinct. For any x ∈ D n+1 we now consider the half-line in the direction from
f (x) to x. Let g(x) denote the point of intersection of this ray with S n . Then we
may consider g : D n+1 → S n as a continuous map. Moreover, g(x) = x for every
x ∈ S n . This contradicts “Brouwer no retraction theorem”. This asserts that f (x)
has a fixed point. ❑
14.2 Continuous Vector Fields
This section studies nonvanishing vector fields on S n . A vector field v on S n is a

continuous function which associates to each vector x of unit length in Rn+1 a unit
vector v(x) in Rn+1 such that x and v(x) are orthogonal (for n = 1, it is geometrically
described in Fig. 14.1). If we imagine that v(x) begins at the point x ∈ S n , then v(x)
must be tangent to the circle.
Definition 14.2.1 Let x be a point of S n . If a vector v in Rn+1 beginning at x is

tangent to S n at x and whose endpoint v(x) varies continuously in Rn+1 as x moves
in S n , then v : S n → Rn+1 is called a continuous vector field on S n . Moreover, if
v(x)
= 0 for all x ∈ S n , the vector field v is said to be nonvanishing.
Remark 14.2.2 A vector field v : S n → Rn+1 on S n is a continuous map such that

for each x ∈ S n , the vector v(x) is orthogonal to the vector x.
Theorem 14.2.3 The n-sphere S n admits a continuous nonvanishing vector field iff
n is odd.
Proof Suppose n is odd and n = 2m − 1 and x = (x1 , x 2 , . . . , x2m ) is a point of S n .

Define a map
Fig. 14.1 Geometrical

description of a vector field v
x
v(x)
y
v(y)
452 14 Applications
v : S 2m−1 → R2m , (x1 , x2 , . . . , x2m−1 , x 2m ) → (x2 , −x1 , . . . , x2m , −x2m−1 ).
Then v is a vector field on S 2m−1 (= S n ) with the desired property.

For the converse, let v be a nonvanishing vector field on S n . Define
H : S n × I → S n , (x, t) → cos πt ·x + sin πt ·v(x).
Then H : 1 S n A, where map 1 S n : S n → S n is the identity map and
A : S n → S n , x → −x.
is the antipodal map. This implies that the map A has degree 1. Consequently, by
Theorem 14.1.7 n is odd. ❑
Remark 14.2.4 If n is odd, the difficult problem of determining the maximum num-
ber of linearly independent nowhere vanishing vector fields on S n was solved by
Adams (1962) by using K -theory.
Theorem 14.2.5 There is a continuous nonvanishing tangent vector field on S n iff

the antipodal map
An : S n → S n , x → −x
is homotopic to the identity map 1S n on S n .
Proof Let v : S n → Rn+1 be a nonvanishing tangent vector field on S n . Define a

continuous map

H : S n × I → S n , (x, t) → (1 − 2t)x + 2 t − t 2 v(x)/||v(x)||.
Then H (x, 0) = x = 1 S n (x) and H (x, 1) = −x = An (x) for all x ∈ S n . This shows
that H : 1 S n An and hence An 1 S n . Conversely, a homotopy H : 1 S n An can
be approximated by a differentiable homotopy. This gives tangent curve elements
and hence a nonzero tangent field of directions, because the tangent line to the curve
αx (t) = H (x, t) at t = 0 contains a unit vector pointing in the direction of increasing
t, which is tangent to the sphere, and is nonvanishing. ❑
Theorem 14.2.6 If the integer n > 0 is even, then there exists no continuous unit
tangent vector field on the n-sphere S n .
Proof If possible, let f : S n → S n be a continuous unit tangent vector field on S n

(n is even). Define a homotopy
Ht : S n → S n , x → x cos(tπ) + f (x) sin(tπ).
Then ||Ht (x)|| = 1 for every t ∈ I and every x ∈ S n . Now H0 (x) = x and
H1 (x) = −x for every x ∈ S n show that H0 is the identity map and H1 is the
14.2 Continuous Vector Fields 453
antipodal map of S n which are homotopic. Hence deg(H0 ) = 1, deg(H1 ) = (−1)n+1 .

As H0 H1 , then (−1)n+1 = 1. This asserts that the integer n must be odd. This
gives a contradiction. ❑
The absence of any nonvanishing vector field on S 2 presents a popular result called
Hairy Ball Theorem.
Theorem 14.2.7 (Hairy Ball Theorem) A hairy ball can not be combed flat.
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 Theorem 14.2.3 for n = 2, we conclude that a hairy ball can
not be combed flat. ❑
Remark 14.2.8 The hairy torus is the only orientable hairy surface that can be
combed smoothly.
14.3 Borsuk–Ulam Theorem with Applications
This section proves Borsuk–Ulam Theorem theorem in a general form for all finite
dimensions, which is a generalization of this theorem for 2-dimensional case proved
in Chap. 3. This theorem was first formulated by S. Ulam (1909–1984) but its first
proof was given by K. Borsuk (1905–1982) in 1933. Since then various alternative
proofs have appeared in literature. Moreover, this section conveys Ham Sandwich
Theorem and Lusternik–Schnirelmann Theorem as applications of Borsuk–Ulum
theorem.
14.3.1 Borsuk–Ulam Theorem
Theorem 14.3.1 (Borsuk–Ulam Theorem) Let m, n be integers such that

m > n ≥ 0. Then there does not exist a continuous map f : S m → S n preserving
the antipodal points.
Proof If possible, let there be a continuous map f : S m → S n such f (x) = f (−x)

for all x ∈ S m . Suppose S n is obtained from S m by giving the last m − n coordinates
equal to zero and i : S n → S m is the usual inclusion map. Then the composite
i ◦ f : S m → S m also preserves the antipodal points. Hence deg(i ◦ f ) must an odd
integer. Again the composite homomorphism
f∗ i∗
(i ◦ f )∗ = i ∗ ◦ f ∗ : Hm (S m ; Z) −−−−→ Hm (S n ; Z) −−−→ Hm (S m ; Z)
454 14 Applications
is the trivial homomorphism, since Hm (S n ; Z) = 0 in this situation. This shows that

deg(i ◦ f ) = 0, which is a contradiction. ❑
Remark 14.3.2 Borsuk– Ulam theorem asserts that any continuous map f : S n →
Rn must identify a pair of antipodal points of S n .
Corollary 14.3.3 If a continuous map f : S n → Rn preserves antipodal points,

then there exists a point x ∈ S n such that f (x) = 0.
Proof If possible, let f (x)

= 0 for all x ∈ S n . Define a continuous map
f (x)
h : S n → S n−1 , x → .
|| f (x)||
As f preserves antipodal points, it follows that h also preserves antipodal points.

But this contradicts the Borsuk–Ulam Theorem 14.3.1. ❑
Corollary 14.3.4 S n cannot be embedded in Rn .
Proof S n cannot be homeomorphic to a subspace of Rn by Remark 14.3.2. ❑
14.3.2 Ham Sandwich Theorem
This theorem is proved by applying Borsuk–Ulam Theorem.
Theorem 14.3.5 (Ham Sandwich Theorem) Let A1 , A2 , . . . , An be n bounded con-

vex subsets of Rn . Then there exists a hyperplane which simultaneously bisects all
of the Ai ’s.
Proof For n = 3, given a three-layered ham sandwich, it can be divided with one
cut that each of the three pieces is divided into two equal parts. To prove this result
for n = 3, construct the continuous map f : S 3 → R3 as follows:
for x ∈ S 3 , take a hyperplane Px perpendicular to x passing through the point
(0, 0, 0, 1/2). Let vi (x) be the volume of that part of Ai , which lies on the same side
of the hyperplane Px at the point x. Now construct the continuous map
f : S 3 → R3 , x → (v1 (x), v2 (x), v3 (x)).
Hence by Borsuk–Ulam Theorem, there exists a point x0 ∈ S 3 such that f (x 0 ) =

f (−x0 ). This implies that v1 (x0 ) = v1 (−x 0 ), v2 (x0 ) = v2 (−x 0 ) and v3 (x0 ) =
v3 (−x0 ). For n > 3 the same procedure is taken. ❑
14.3 Borsuk–Ulam Theorem with Applications 455
14.3.3 Lusternik–Schnirelmann Theorem
This subsection proves the Lusternik–Schnirelmann theorem by using Borsuk–Ulam

Theorem.
Theorem 14.3.6 (Lusternik–Schnirelmann) Let S n be covered by n + 1 closed sets

A1 , A2 , . . . , An+1 . Then one of them must contain a pair of antipodal points.

n+1
Proof By hypothesis, A j = S n ⊂ Rn+1 . Define a continuous map
j=1
f : S n → Rn , x → (d(x, A1 ), . . . , d(x, An )),
where d(x, Ai ) is the distance of x from the closed set Ai . Then f must identity a
pair of antipodal points by Borsuk–Ulam Theorem. Consequently, there is a point x 0
in S n such that d(x0 , A j ) = d(−x0 , A j ) for 0 ≤ j ≤ n.
Now only two cases arise:
Case I: If d(x 0 , A j ) = 0 for some j, then both the points x0 , −x0 ∈ A j , since each
A j is a closed set.
Case II: If d(x0 , A j ) > 0 for all j = 1, 2, . . . , n, then x 0 , −x0 ∈ An+1 , since the
A j ’s form a cover of S n . ❑
14.4 The Lefschetz Number and Fixed Point Theorems
This section conveys the concept of Lefschetz number which is an integral homotopy
invariant and generalizes the Euler characteristic. This number is closely related to
the degree of a spherical map. S. Lefschetz (1884–1972) published the first version
of his fixed point formula in 1923 which asserts that given a closed manifold M and
a map f : M → M, for each q there is an induced homomorphism on homology
with rational coefficients Q
f q ∗ : Hq (M; Q) → Hq (M; Q).
For each q we may choose a basis for the finite-dimensional rational vector space
Hq (M; Q) and we write f q ∗ as a matrix with respect to this basis.
Lefschetz number is an important concept introduced by Lefschetz in 1923. It is
a number associated with each continuous map f : |K | → |K | from a polyhedron
into itself and the number is denoted by f . It is also closely related to the Euler
characteristic formula. It proves a powerful fixed point theorem known as Lefschetz
fixed point theorem, which is an important application of homology. This theorem
generalizes Brouwer fixed point theorem. Moreover, some other results on fixed
points follow as its applications.
456 14 Applications
Definition 14.4.1 (Lefschetz number) Let K be a fixed triangulation of a compact

triangulable space X . Suppose n is the dimension of K and f : X → X is a
continuous map. Then there is a homeomorphism k : |K | → X such that if n is the
dimension of K , then each homology group Hq (K ; Q) with rational coefficients Q
is vector space over Q and each homomorphism
f qk∗ = (k −1 ◦ f ◦ k)∗ : Hq (K ; Q) → Hq (K ; Q)
is a linear transformation. The trace of the corresponding matrices does not depend
n
on a particular choice of the basis. The alternating sum (−1)q trace fqk∗ of the
q=0

n
traces of f qk∗ of these linear transformations denoted by f = (−1)q trace f qk∗ , is
q=0
called the Lefschetz number of f .
Remark 14.4.2 The number f does not dependent on the triangulation of X . Hence
f is well defined.
Remark 14.4.3 If f g, then f ∗ = g∗ in homology. Hence it follows that f = ∧g

whenever f is homotopic to g.
Definition 14.4.4 The rank of the free part of the abelian group Hq (K ; Z) of a finite
complex K is called the Betti number of K , denoted by βq .
Remark 14.4.5 The Lefschetz number is an integer and generalizes the Euler char-
acteristics of an oriented complex.
Definition 14.4.6 (Fixed Point Property) If a topological space X is such that every
continuous map f : X → X has a fixed point, then X is said to be a space with fixed
- point property.
Example 14.4.7 The topological space X = [0, 1] in the real line R is a space with
the fixed point property. Every closed interval [a, b] in the real line R has also the
fixed-point property.
The following theorem shows that f is the ‘obstruction’ to f being fixed point
free.
Theorem 14.4.8 (Lefschetz Fixed Point Theorem) Let X be a compact triangulable

space and f : X → X be a continuous map with f its Lefschetz number.
(a) If f
= 0, then f has a fixed point;
(b) If X has the same rational homology groups as a point, then X has the fixed
point property.
(c) (Brouwer Fixed Point Theorem): Any contractible compact triangulable space
has the fixed point property.
14.4 The Lefschetz Number and Fixed Point Theorems 457
(d) If the identity map 1 X of X is homotopic to a fixed point free map f : X → X ,

then the Euler characteristic χ(X ) = 0;
(e) If X = S n , then f = 1 + (−1)n deg f ;
Proof (a) Case I: Let K be a finite simplicial complex having X as its polyhedron.
Suppose the simplicial map f : |K | → |K | has no fixed point. Then there exists
a simplex σ in K such that f (σ)
= σ. Now delete each q-simplex of K to obtain
a basis over Q for the vector space Cq (K ; Q) in such a way that with respect to
this basis the linear transformation f q : Cq (K ; Q) → Cq (K ; Q) will represent
a matrix having zero along its diagonal, and hence having trace zero.
Hence

n
n
(−1)q trace f q = (−1)q trace fq ∗
q=0 q=0
gives that f = 0. In other words, if f

== 0, then f has a fixed point.
Remark 14.4.9 We may calculate the Lefschetz number of f at homology level or

at chain level according to our convenience.
Case II: For general case see Gray (1975) or Armstrong (1983).
(b) By hypothesis, X has only one component. Hence the only nonzero rational
homology group is H0 (X ; Q) ∼ = Q, because Hq (X ; Q) = 0, for q > 0. By
using the definition of Lefschetz number, for any map f : X → X , the induced
homomorphisms
f qk∗ : Hq (X ; Q) → Hq (X ; Q)
are all zero for q > 0. On the other hand, f 0k∗ : Q → Q is the identity linear
transformation. This proves that f = 1(
= 0) and hence f has a fixed pint.
(c) It follows from (b).
(d) Let f be fixed point free. Then f = 0. By hypothesis f is homotopic to 1 X .
Then it induces identity homomorphisms on homology groups and the trace of
an identity linear map is the dimension of its domain:

n
n
f = (−1)q trace(1q ∗ ) = (−1)q dim Hq (X ; Q) = χ(X ) = 0.
q=0 q=0
(e) The only nonzero rational homotopy groups of S n are Q in dimensions 0 and n,
i.e.,
H0 (S n ; Q) ∼
= Hn (S n ; Q) ∼
=Q
458 14 Applications
and hence
f = trace ( f 0∗ ) + (−1)n tr ( f n ∗ ) = 1 + (−1)n deg f.
Remark 14.4.10 The Lefschetz number of f : X → X may be calculated at either

homology level or at chain level according to our convenience.
14.5 Application of Euler Characteristic
This section coveys some interesting applications of Euler characteristic χ(X ) of a

finite C W -complex X . As it is defined in terms of homology of X , it depends only
on the homotopy type of X .
Recall the definitions of Euler characteristic:
Definition 14.5.1 The Euler characteristic of a finite C W -complex X of dimension

n is defined to be the alternating sum

n
χ(X ) = (−1)q αq ,
q=0
where αq denotes the number of q-cells of X , generalizing the formula of Euler

characteristic: number of vertices − number of edges + number of faces for 2-
dimensional complexes.
Definition 14.5.2 The Euler characteristic of a finite simplicial complex K of dimen-

sion n is defined to be the alternating sum

n
χ(K ) = (−1)q βq ,
q=0
where βq denotes the number of q-dimensional simplices of K .
Remark 14.5.3 If G = R in the group Hn (K ; G), then the group Hn (K ; G) is a real

vector space. If its dimension is q, then q is called the called the qth Betti number
of K , denoted by βq .
Theorem 14.5.4 Let X be a finite C W -complex with χ(X )

= 0, and ψt : X → X
be a flow. Then there exists a point x 0 ∈ X such that ψt (x0 ) = x0 , ∀ t ∈ R.
14.5 Application of Euler Characteristic 459
Proof Under the given hypothesis, ψt 1 X (see Chap. 2) and hence ∧ψt = ∧1 X =
χ(X )
= 0. Consequently, there exists a fixed point x0(t) of ψt for each t ∈ R. Define
X n = {x ∈ X : ψ1/2n (x) = x} for each natural number n. Then X n ⊃ X n+1 , and
each X n is a nonempty closed set and hence X ∞ = X n
= ∅. Let x ∈ X ∞ . Then
n
x is a fixed point for any ψm/2n . Since the numbers m/2n are dense in R, x is a fixed
point of ψt for any t ∈ R. ❑
Remark 14.5.5 Given a finite C W -complex X and continuous map f : X → X ,
the relation f = ∧1 X = χ(X )
= 0 (where f is the Lefschetz number of f and
χ(X ) is the Euler characteristic of X ) shows that a flow ψt : X → X has a fixed
point. We now claim that there exists a fixed point common to every ψt . We prove
this by induction and the infinite intersection property as follows; corresponding to
each rational n points, assign X n to be the set of fixed points of x 1/2n . Then X n is a
nonempty closed set. Since X n+1 ⊆ X n , the set X ∞ = ∩n X n
= ∅. This shows that
X ∞ is a set of points fixed under all rational numbers of dyadic form r/2n . Since the
set of rational numbers are dense in the real number space R, it follows that every
element in X ∞ is fixed under ψt for all t ∈ R.
Lefschetz Fixed Point Theorem 14.4.8 gives the following corollary.
Corollary 14.5.6 (a) ∧1 X = χ(X );
(b) For the antipodal map A : S n → S n , x → −x, ∧ A = 0.
(c) If f : S n → S n is not a homeomorphism, then f must have a fixed point.
Definition 14.5.7 A platonic solid is a special polyhedron having the property that
its faces are congruent regular polygons and each vertex belongs to the same number
of edges. It is sometimes called a regular simple polyhedron.
Theorem 14.5.8 There are only five regular simple polyhedra.
Proof Let P be a regular simple polyhedron with V number of vertices, E number of
edges, and F number of faces and m be the number of edges meeting at each vertex
and n be the number of edges of each face. For n ≥ 3, by counting vertices by edges
we obtain 2E = mV . Again by counting faces by edges we obtain 2E = n F. Then
from Euler formula V − E + F = 2 for P it follows that m2 E − E + n2 E = 2. This
shows that E = 2( m1 + n1 − 1)−1 , which must be a positive integer. Hence m1 + n1 > 1.
The possibilities are only:
(i) If m = 5, n = 3, then E = 30, V = 12 and F = 20. Hence P is the regular
icosahedron see Fig. 14.2.
(ii) If m = 4, n = 3, then E = 12, V = 6 and F = 8, show that P is the regular
octahedron.
(iii) If m = 3, n = 3, then E = 6, V = 4 and F = 4 show that P is a regular
tetrahedron.
(iv) If m = 3, n = 4, then E = 12, V = 8 and F = 6 show that P is the cube.
(v) If m = 3, n = 5, then E = 30, V = 20 and F = 12 show that P is the regular
dodecahedron see Fig. 14.3.
460 14 Applications
Fig. 14.2 Icosahedron v1
v6 v7
v2
v12 v10
v8
v9
v5 v11 v3
v4
Fig. 14.3 Dodecahedron v1

v10 v2
v16 v11
v17
v9 v12 v3
v15
v18
v20
v8 v13 v4
v14 v19
v7 v5
v6
Proposition 14.5.9 Let B be a finite C W -complex and p : X → B be an m sheeted

covering space. Then the Euler characteristic χ(X ) = mχ(B).
Proof By hypothesis, it follows that X admits a C W -complex structure with the

j-cells of X , which are the lifts to X of j-cells of B. As p is m-sheeted, the number
of distinct such lifts is m. Hence it follows that the number of j-cells of X is exactly
m times that of j-cells of B. This concludes that χ(X ) = mχ(B). ❑
14.6 Application of Mayer–Vietoris Sequence
This section conveys some interesting applications of Mayer–Vietoris sequence (see

Chap. 10) defined by W.Mayer (1887–1947) and L. Vietoris (1891–2002) such as
in computing homology groups of S n . In addition to the long exact sequence of
homology groups of a pair (X, A) of spaces, there is another long exact sequence of
14.6 Application of Mayer–Vietoris Sequence 461
homology groups, called a Mayer–Vietoris sequence. This sequence is sometimes

more convenient to apply.
Let U1 and U2 be two subspaces of X such that X is the union of the interiors of
U1 and U2 . If f i : U1 ∩ U2 → Ui and gi : Ui → X are inclusion maps for i = 1, 2,
define
φ : Hn (U1 ∪ U2 ) → Hn (U1 ) ⊕ Hn (U2 ), α → ( f 1∗ (α), f 2∗ (α)),

and ψ : Hn (U1 ) ⊕ Hn (U2 ) → Hn (X ), (α1 , α2 ) → g1∗ (α1 ) − g2∗ (α2 )
Under this situation one obtains an exact sequence
φ
· · · → Hn+1 (X ) −−−→ Hn (U1 ∪ U2 ) −−−→ Hn (U1 ) ⊕ Hn (U2 )
ψ
−−−→ Hn (X ) −−−→ Hn (U1 ∩ U2 ) → · · · (14.1)
The sequence (14.1) is called the Mayer–Vietoris sequence and the homomor-
phisms are called connecting homomorphisms.
Mayer–Vietoris sequence is now applied to compute the homology groups of the
n-sphere S n .
Theorem 14.6.1 Given a positive integer n,

Z, if m = 0 or n
Hm (S ; Z) ∼
n
=
0, otherwise
Proof Let U1 = {x ∈ S n : x n > − 12 } and U2 = {x ∈ S n : x n < 12 }. Then U1 and

U2 are contractible spaces. Use induction on n and apply Mayer–Vietoris sequence
to compute Hm (S n ). ❑
14.7 Application of van Kampen Theorem
This section presents an important application of van Kampen theorem which asserts
that for n > 1, S n is simply connected. The van Kampen Theorem has been studied
in Chap. 6. An alternative form of van Kampen Theorem is now given.
Theorem 14.7.1 (van Kampen theorem) Let X = X 1 ∪ X 2 and X 1 , X 2 and A =

X 1 ∩ X 2 are all open path-connected subsets of X . If x0 ∈ A, then π1 (X, x 0 ) is the
free product:
π1 (X, x 0 ) ∼
= π1 (X 1 , x0 ) ∗π1 (A,x0 ) π1 (X 2 , x0 ).
462 14 Applications
In other words, if i 1 : A → X 1 and i 2 : A → X 2 are inclusions, then π1 (X, x0 ) is

isomorphic to the free product π1 (X 1 , x 0 ) ∗π1 (A,x0 ) π1 (X 2 , x 0 ) modulo the relations
i 1∗ (α) = i 2∗ (α) for every α ∈ π1 (A, x0 ).
Proof See [Gray, pp 40–41]. ❑
Corollary 14.7.2 Under the hypothesis of van Kampen theorem,

(i) If X 1 and X 2 are simply connected, then X is also so;
(ii) If A is simply connected, then π1 (X, x0 ) = π1 (X 1 , x0 ) ∗ π1 (X 2 , x0 ).
(iii) If X 2 is simply connected, then π1 (X, x0 ) = π1 (X, x0 )/N (π1 (A, x0 ))
where N (π1 (A, x0 )) is the normalizer of π1 (A, x0 ).
Corollary 14.7.3 For n > 1, S n is simply connected.
Proof Let X 1 = S n − N , where N = (0, 0, . . . , 0, 1) and X 2 = S n − S, where S =

(0, 0, . . . , 0, −1). Clearly, X 1 and X 2 are both homeomorphic to Rn . Consequently,
they are path-connected spaces and simply connected. Again as n > 1, X 1 ∩ X 2 is a
path-connected space. Hence the corollary follows from Corollary 14.7.2(i). ❑
14.8 Applications to Algebra
Algebraic topology generally utilizes algebraic techniques to get topological infor-

mation, but this direction is sometimes reversed in a convenient way. For example,
it is proved by using the concepts of fundamental group and covering space that a
subgroup of a free group is free. The other example is a proof of the algebraic result
by using the concept of degree function that if n is an even integer, then Z2 is the
only nontrivial group that can act freely on S n .
Theorem 14.8.1 Every subgroup of a free group is free.
Proof Given a free group F, we can construct a graph B corresponding to a basis of

F such that π1 (B) ∼
= F. This construction is possible. For example, one may take B
to be the graph which is a wedge of circles corresponding to a basis for F for such a
construction. Consequently, for each subgroup G of F, there exists a covering space
p : X → B having induced monomorphism
p∗ : π1 (X ) → π1 (B)
such that p∗ (π1 (X )) = G by Ex. 14 of Sect. 14.11. Since p∗ is a monomorphism,

π1 (X ) ∼
= G. Again as X is a graph by Ex. 16 of Sect. 14.11, the group G ∼
= π1 (X ) is
free by Ex. 15 of Sect. 14.11. ❑
14.8 Applications to Algebra 463
Recall that the antipodal map
A : S n → S n , x → −x
generates an action of Z2 on S n with orbit space the real projection n-space R P n .

There is a natural question: does there exist any finite group that acts freely on
S n ? If n is an even integer, Z2 is the only nontrivial group that can act freely on S n
(see Ex. 23 of Sect. 14.11).
Given a topological space X , if a topological group G acts on X (from the left), then
for every g ∈ G, the map ψg : X → X defined by ψg (x) = gx is a homeomorphism
(see Appendix A). Hence for each g ∈ G, ψg is an element of the group Homeo(X )
of homeomorphisms of X . This action is said to be free if the homeomorphism ψg in
the group Homeo(X ) corresponding to each nontrivial element g of G has no fixed
points.
The above discussion may be summarized in an important result.
Proposition 14.8.2 Under the above situation the antipodal map A : Sn →
S n , x → −x generates a free action of Z2 .
Remark 14.8.3 For more applications of algebraic topology to algebra see

Sect. 14.11.
14.9 Application of Brown Functor
Brown functor defined in Appendix B plays a key role in the study of algebraic
topology.
Definition 14.9.1 Let F be Brown functor and X be a topological space in the cate-
gory C0 . Then an element u ∈ F(X ) is said to be m-universal if the homomorphism
ψu : πk (X ) → F(S k ), [ f ] → F([ f ])(u)
is an isomorphism for k < m and an epimorphism for k = m. An m-universal

element u is said to be a universal element if it is m-universal for all m ≥ 1.
Proposition 14.9.2 Let F be Brown functor in the category C0 . Then it is repre-

sentable in the sense that there is a pointed topological space X ∈ C0 which is
determined up to homotopy equivalence such that there exists a natural equivalence
of functors
ψ : [−; X ] → F.
464 14 Applications
Proof For every pointed topological space Y ∈ C0 , there is a bijection
ψY : [Y ; X ] → F(Y ), [ f ] → F([ f ])(u),
where u ∈ F(Y ) is a universal element. ❑
Remark 14.9.3 For more results associated with Brown functor see Sect. 14.11.
14.10 Applications Beyond Mathematics
Algebraic topology has also interesting applications in some areas other than math-
ematics. The present book is beyond the scope of the study of these applications,
except to give some references. However, the author is preparing a new book ‘Topics
in Topology with Applications’ (unfinished).
14.10.1 Application to Physics
Algebraic topology has made a revolution in mathematical physics in the second half
of the twentieth century. For example, fiber bundles and vector bundles constitute
an extensive special class of manifolds, and play a key role in some theories of
physics, general relativity, and gauge theories. Moreover, algebraic topology plays
an important role in condensed matter physics, statistical mechanics, elementary
particle theory, and some other branches of physics. Homotopy theory is specially
used in the study of solitons, monopoles, and condensed system. Many interesting
topological spaces appear in physics at different situations. For example, the phase
space of a quantum system with n pure states can be considered as the complex
projective space CP n−1 . Each state (wave function) is a nonzero vector in Cn , but
the states that differ only by multiplicative factors are physically indistinguishable,
and hence they are identified with each other. Another example: the phase space of
a classical mechanical system with nondegenerate Lagrangian belongs to the same
homotopy type of the configuration space.
For this subsection the following books are referred.
[1] M.F. Atiyah, The geometry and physics of knots, Cambridge University Press,
Cambridge, 1990.
[2] M. Monastyrsky, Topology of Gauge Fields and Condensed Matter, Plenum,
New York, 1993.
[3] M. Nakahara, Geometry, Topology and Physics, Taylor and Francis, 2003.
[4] C. Nash and S. Sen Topology and Geometry for Physicists, Academic Press,
London, 1983.
14.10 Applications Beyond Mathematics 465
[5] Schwartz, A. S., Quantum field theory and topology, Springer, Berlin, 1993.
[6] N. E. Steenrod, The Topology of Fibre Bundles, Princeton University Press,
1951
14.10.2 Application to Sensor Network
Algebraic topology is now used to solve coverage problems by integrating local data
about sensor networks into global information and utilizes its strong tools to deter-
mine whether there is any hole in a sensor coverage. For example, certain topological
invariants such as Euler characteristic, fundamental groups and higher homotopy
groups, homology and cohomology groups play a key role in solving the coverage
problems.
For this subsection the following papers are referred.
[1] E. W. Chambers, J. Erickson, and P. Worah, Testing contractibility in planar Rips
complexes, in Proc. 24th Annu. Symp. Computat. Geom., College Park, MD,
pp. 251–259. 2008.
[2] J. Cortes, S. Martinez, T. Karatas, and F. Bullo, Coverage control for mobile
sensing networks, in Proc. IEEE Int. Conf. Robot. Autom., Washington, DC,
Vol. 2, pp. 1327–1332, 2002.
[3] de Silva and Robert Ghrist, Homological Sensor Networks, Notices of AMS 54
(1) pp 1–11, 2007.
14.10.3 Application to Chemistry
Topology and graph theory have strongly influenced the recent development of chem-
istry through their applications in nonroutine mathematical methods. For example,
“chemical topology”, “invariance of molecular topology”, “chemical applications
of topology and graph theory” and “topological methods in chemistry” are now
outstanding developments of chemistry which are closely related to topology.
For this subsection the following books are referred:
[1] E. V. Babaev, The Invariance of molecular topology, Moscow State University,
Moscow, 1994.
[2] Bonchev D., Rouvray R., (Eds) Chemical Topology: Introduction and Funda-
mentals, Gordon and Breach Publ., Reading, 1999.
[3] R.B. King, (Ed.)., Chemical Applications of Topology and Graph Theory; Stud-
ies in Physical and Theoretical Chemistry, Vol. 28, Elsevier, Amsterdam, 1983.
[4] H. E. Simmons, Topological Methods in Chemistry, Wiley Interscience, New
York, 1989.
466 14 Applications
14.10.4 Application to Biology, Medical Science

and Biomedical Engineering
The fields of biological & medical physics and biomedical engineering are now
emerging as a multidisciplinary area connecting topology with different areas of
physics, biology, chemistry, medicine, and some of their closely related fields. For
example, knot theory, a branch of topology, is used in biology to study the effects of
certain enzymes on DNA. Algebraic topology addresses the growing need for this
multidisciplinary research. For example, recent investigation in molecular biology,
theory of protein and DNA involves application of algebraic topology, which is a
stimulating feature.
For this subsection following books are referred.
[1] I. Darcy and D.Mners, Knot Theory, Polish Academy of Sciences, Warszawa,
1998
[2] M.I. Monastyrsky (Ed.) Topology in Molecular Biology, Springer-Verlag Berlin
Heidelberg, 2007.
14.10.5 Application to Economics
The Brouwer fixed point theorem given by L.E.J. Brouwer in 1912 is one of the stim-
ulating events in the history of topology. Since then this theorem has been extending
its influence to diverse areas of mathematics, mathematical economics and related
fields. For example, in economics Brouwer fixed point theorem plays a key role
in studying general equilibrium theory and in the most basic and general models
of economists. The other example, is the ‘social choice’ model which is a model
for decision making in mathematical economics and social science which is closely
related to homotopy problems. The author of this book earned an inspiration on the
topic ‘social choice and topology’ by attending a lecture of Eckman (2003) at ETH,
Zurich.
The following books and papers are referred for this subsection.
[1] Kim C.Border, ‘ Fixed Point Theorems with Applications to Economics and
Game Theory,’ Cambridge University Press, 1985.
[2] B. Eckman, ‘Social Choice and Topology: A Case of Pure and Applied Mathe-
matics,’ 2003.
[3] A. Granas and J. Dugundji, Fixed Point Theory, Springer, New York, 2003.
14.10 Applications Beyond Mathematics 467
14.10.6 Application to Computer Science
Algebraic topology has recently found some surprising fruitful results in computer
science by establishing a close relation between the theory of concurrent computation
and the theories of algebraic and combinatorial topology.
For this subsection the following references are given.
[1] Edelsbrunner, H. and Harer J. L., Computational Topology. An Introduction.
Amer. Math. Soc., Providence, Rhode Island, 2009.
[2] Gyulassy A., Natarajan V., Pascucci V., Bremer P.T., and Ann B.H., A topological
approach to simplification of three-dimensional scalar functions. IEEE Trans.
Vis. Comput. Graph.Vol 12, 474–484. 2006.
14.11 Exercises
This section conveys some interesting results through different exercises.

1. Let K be a finite complex. Show that
(a) the qth Betti number βq of K is the dimension of Hq (K ; Q) as a vector
space over Q;
(b) the complexes K whose polyhedra are homotopy equivalent have the same
Euler characteristic
2. (Hopf Trace Theorem). Let K be a finite simplicial complex of dimension n,
and ψ : C(K ; Q) → C(K ; Q) be a chain map. Show that

n
n
(−1)q trace ψq = (−1)q trace ψq ∗ ,
q=0 q=0
where ψq ∗ is the homomorphism induced by chain map ψq in the corresponding

homology groups.
3. Show that the only compact closed surfaces with Euler characteristic zero are
the Klein bottle, which is non orientable, and the torus, which is orientable.
[Hint: Only these two compact closed surfaces admit a fixed-pint free map which
is homotopic to the identity map. A fixed point free map on the torus can be
obtained by a flow along a nowhere-zero (nonvanishing) tangent vector field.]
4. Use Euler characteristic to show that a sphere cannot be homotopy equivalent to
a point.
5. Prove that the homotopy classes of maps of a sphere to itself can be characterized
with the help of integers.
[Hint: Use Hopf classification theorem.]

468 14 Applications
6. Show that no subspace of Rn can be homeomorphic to S n .
[Hint: Use Borsuk–Ulam theorem to show that no continuous map from S n to

Rn is injective].
7. Show that while wrapping a soccer ball with three pieces of papers, one must
contain a pair of antipodal points.
[Use Theorem 14.3.6 (Lusternik–Schnirelmann).]

8. Show that the following statements are equivalent:
(i) A continuous map f : D n → D n has a fixed point (Brouwer fixed point
theorem).
(ii) There does not exist a continuous map r : D n → S n−1 which is identity on
S n−1 (Retraction theorem).
(iii) The identity map 1d : S n−1 → S n−1 is not nullhomotopic (Homotopy
theorem).
9. Show that every nullhomotopic map f : S n → S n has at least one fixed point.
10. If f : (D n+1 , S n ) → (D n+1 , S n ) is a continuous map, show that deg f =
deg( f | S n ).
[Hint: Consider the commutative diagram in Fig. 14.4]

11. Show that every continuous map f : S n → S n with deg f = d induces the
homomorphism f ∗ : πn (S n ) → πn (S n ) which is a map multiple by d.
12. Let f : S n → S n be a continuous map for n ≥ 1 and f : S n+1 → S n+1 be its
suspension map. Show that deg f = deg( f ) for all n ≥ 1.
13. Given a continuous map f : Sn → S n , show that f : S n → S n has degree
d iff the map f : S n → S n has degree d.
14. Let B be a path-connected, locally path-connected, and semi locally simply
connected space. Show that for every subgroup G of π1 (B, b0 ), there is a covering
space p : X → B such that p∗ (π1 (X, x0 )) = G for a suitable base point x0 ∈ X .
15. Let X be a connected graph with maximal tree M. Show that π1 (X ) is a free
group with basis the classes [bi ] corresponding to the edges ei of X − M.
16. Show that every covering space of a graph is also a graph, with vertices and
edges the lifts of the vertices and edges in the base graph.
17. Let X = S n and E n+ , E n− be the north and south hemispheres of S n . Then
A ∩ B = S n−1 . Show that
Fig. 14.4 Rectangle f∗

Hn+1 (Dn+1 , S n ) / Hn+1 (D n+1 , S n )
involving ( f | Sn )∗
∂∗ ∼
= ∂∗ ∼
=

(f |S n )∗
n (S n )
H
/H
n (S
n
)
14.11 Exercises 469
i (E n+ ) ⊕ H
(i) H i (E n− ) = 0;
i (S ) ∼
(ii) H n
=H i−1 (S n−1 ).
18. Let T be a Brown functor and {∗} denote the one-point space. Show that T ({∗})
is a set consisting of a single element.
19. Let T be a Brown functor, X 0 be a pointed space. If u 0 ∈ T (X 0 ), show that there
is a pointed space X obtained from X 0 by attaching together with a universal
element u ∈ T (X ) such that u| X 0 = u 0 .
20. Let T be a Brown functor. If X and Y are pointed C W -complexes with uni-
versal element u ∈ T (X ) and u ∈ T (Y ). Show that there exists a homotopy
equivalence f : X → Y such that T ([ f ])(u ) = u.
21. Let T be a Brown functor and u ∈ T (X ) be a universal element. Show that if X
is a pointed C W -complex, then the map
ψu : [Y, X ] → T (Y ), [ f ] → T ([ f ])
is a bijection.
22. Show that Z p × Z p cannot act freely on S n for a prime integer p.
23. If n is an even integer, show that Z2 is the only nontrivial group that can act
freely on S n .
24. Let f : S 2n → S 2n be any continuous map. Show that there is a point x ∈ S 2n
such that either f (x) = x or f (x) = −x.
25. Show that every continuous map f : R P 2n → R P 2n has a fixed point.
26. Construct continuous maps g : R P 2n−1 → R P 2n−1 without fixed points from
linear operators R2n → R2n without eigenvectors.
27. (Poincaré duality) Let K be a triangulation of a homology n-manifold. If |K | is
orientable, show that there exists an isomorphism
(i)
h : H r (K ; Z) → Hn−r (K ; Z)
for all r ;
(ii)
h : H r (K ; Z2 ) → Hn−r (K ; Z2 )
for all r.
[Hint: See Maunder (1996).]

28. Let M be a compact manifold of odd dimension. Show that its Euler character-
istic χ(M) = 0.
[Hint: Since Hr (M; Z2 ) is finite dimensional, the Euler characteristic χ(M) is

well defined. Again since the dim Hr (M; Z2 ) = dim Hn−r (M; Z2 ), it follows
that
470 14 Applications

χ(M) = (−1)r dim Hr (M; Z2 ) = 0].
r
29. For any knot K , show that the homology groups H0 (R3 − K ; Z) and H1 (R3 −
K ; Z) are isomorphic to Z and the remaining homology groups of the comple-
ment R3 − K are trivial. Hence shows that the trefoil knot is nontrivial.
[Hint: Use the result that if two knots are equivalent, then their complements
have isomorphic homology groups.]
30. T be an orthogonal transformation of Rm such that T n = 1d and T k
= 1d
for 0 < k < n. Show that the group G = {T 0 , T 1 , · · · , T n−1 } acts freely on
S m−1 ⊂ Rn if every eigenvalue λ of T satisfies λk
= 1 for 0 < k < n. Compute
the fundamental group and homology groups of the orbit space S m−1 mod G,
when m = 4 and G acts freely on S 3 . (A orbit space of this type is called a lens
space).
[Hint: For a free action, the group π1 (S m−1 mod G) = G. Use Hurewicz theorem
and Poincaré duality for respective computations.]
31. Let G n,k be the Grassmann manifold of k-planes in Rn . Show that
πi (G n,k ) = πi−1 (O(k)) for i < n − k.
[Hint: Use fibration (Vn,k , G n,k , O(k))].

32. Apply the Lefschetz fixed point theorem to prove that
(i) if n is even, every continuous map f : CP n → CP n has a fixed point;
(ii) if n is odd, there is a fixed point unless f ∗ (β) = −β for β, a generator of
H 2 (CP n ; Z).
[Hint: Use the results that f ∗ : H ∗ (CP n ; Z) → H ∗ (CP n ; Z) is a ring

homomorphism, each cohomology group is even degree ≤ n has rank 1,
and each cohomology group in odd degree is 0.]
33. Using cup products, prove that every continuous map S m+n → S m × S n induces
a trivial homomorphism Hm+n (S m+n ) → Hm+n (S m × S n ), for positive integers
m and n.
[Hint: Use Künneth formula.]

34. Use homotopy groups to show that R P m is not a retract of R P m for n > m > 0.
35. Show that πi (Vn,k ) = 0, for i < n − k.
[Hint: Use the fibration (Vn,k+1 , Vn,k , S n−k−1 ).]

14.11 Exercises 471
36. Show that for any closed surface other than the sphere and the projective plane,
the homotopy groups of dimensions
greater than 1 are trivial.
∼ Z, for n is even
37. Show that πn−2 (Vn , 2) =
Z2 , for n is odd.
[Hint: Use the exact homotopy sequence of the fibration (Vn−2 , S n−1 , S n−2 ).]
[1] Adams, J.F.,Algebraic Topology: A student’s Guide, Cambridge University

[4] Dold, A.,Lectures on Algebraic Topology, Springer-Verlag, New York, 1972.
[6] Eschrig, H., Topology and Geometry for Physics, Springer-Verlag Berlin Hei-
delberg 2011.
1975.
Cambridge, 1983.
[9] Hilton, P. J. and Wylie, S. Homology Theory, Cambridge University Press,
Cambridge, 1960.
[12] Pearl Elliot (ed), Open Problems in Topology II, Elsevier, Amsterdam, 2007.
[13] Schwartz, A. S., Quantum field theory and topology, Springer, Berlin„ 1993.
472 14 Applications
References
(1972)
Adhikari, M.R., Adhikari, A.: Basic Modern Algebra with Applications, Springer, New Delhi, New
York, Heidelberg (2014)
Atiyah, M.F.: The geometry and physics of knots. Cambridge University Press, Cambridge (1990)
Babaev, E.V.: The Invariance of molecular topology. Moscow State University, Moscow (1994)
Bonchev D., Rouvray R. (eds.): Chemical Topology: Introduction and Fundamentals. Gordon and
Breach Publ., Reading, (1999)
Border, K. C.: Fixed Point Theorems with Applications to Economics and Game Theory. Cambridge
University Press, Cambridge (1985)
Chambers, E. W., Erickson, J., Worah, P.: Testing contractibility in planar Rips complexes. In:
Proceedings of the 24th Annual Symposium of Computation Geometry. College Park, MD, pp.
251–259 (2008)
Cortes J., Martinez S., Karatas T., Bullo, F.: Coverage control for mobile sensing networks. In:
Proceedings of the IEEE International Conference on Robotics Automation. Washington, DC,
Vol.2, pp. 1327 –1332. (2002)
Croom, F.H.: Basic Concepts of Algebraic Topology. Springer, New York, Heidelberg, Berlin (1978)
Darcy, I., Mners, D.: Knot Theory. Polish Academy of Sciences, Warszawa (1998)
Dieudonné, J.: A History of Algebraic and Differential Topology, 1900–1960, Modern Birkhäuser
(1989)
Dodson, C.T.J., Parker P.E.: A User’s Guide to Algebraic Topology, Kluwer Academic Publishers,
Berlin (1997)
Eckman, B.: “Social Choice and Topology A Case of Pure and Applied Mathematics”, (2003).
http://www.elsevier.de/expomath
Edelsbrunner, H., Harer, J.L.: Computational Topology. An Introduction. Amer. Math. Soc, Provi-
dence, Rhode Island (2009)
ton (1952)
Eschrig, H.: Topology and Geometry for Physics. Springer, Berlin (2011)
Fulton, W.: Algebraic Topology. A First Course. Springer, New York (1975)
Granas, A., Dugundji, J.: Fixed Point Theory. Springer, New York (2003)
Gray, B.: Homotopy Theory. An Introduction to Algebraic Topology. Acamedic Press, New York
(1975)
Gyulassy, A., Natarajan, V., Pascucci, V., Bremer, P.T., Ann, B.H.: A topological approach to
simplification of three-dimensional scalar functions. IEEE Trans. Vis. Comput. Graph 12, 474–
484 (2006)
King R.B.(ed.), Chemical Applications of Topology and Graph Theory; Studies in Physical and
Theoretical Chemistry, Vol. 28, Elsevier, Amsterdam (1983)
Maunder, C.R.F.: Algebraic Topology, Van Nostrand Reinhold Company, London, 1970. Dover,
Reprinted (1996)
Monastyrsky, M. (ed.): Topology in Molecular Biology. Springer, Berlin (2007)
Monastyrsky, M.: Topology of Gauge Fields and Condensed Matter. Plenum, New York (1993)
Nash, C., Sen, S.: Topology and Geometry for Physicists. Academic Press, London (1983)
References 473
Nakahara, M.: Geometry. Institute of Physics Publishing, Taylor and Francis, Bristol, Topology and
Physics (2003)
Elliot, Pearl (ed.): Open Problems in Topology II. Elsevier, Amsterdam (2007)
Schwartz, A.S.: Quantum field theory and topology. Springer, Berlin (1993)
de Silva., Ghrist, R.: Homological Sensor Networks. Notices of AMS 54(1), 1–11 (2007)
Simmons, H.E.: Topological Methods in Chemistry. Wiley Interscience, New York (1989)
Steenrod N.E.: The Topology of Fibre Bundles. Princeton University Press, Princeton (1951)
Switzer, R.M.: Algebraic Topology: Homotopy and Homology. Springer, Berlin (1975)
Chapter 15
Spectral Homology and Cohomology
Theories
This chapter continues to study homology and cohomology theories through the con-
cept of a spectrum and constructs its associated homology and cohomology theories,
called spectral homology and cohomology theories. It also introduces the concept of
generalized (or extraordinary) homology and cohomology theories. Moreover, this
chapter conveys the concept of an -spectrum and constructs a new -spectrum A,
generalizing the Eilenberg–MacLane spectrum K (G, n). It constructs a new general-
ized cohomology theory h ∗ ( ; A) associated with this spectrum A, which generalizes
the ordinary cohomology theory of Eilenberg and Steenrod. This chapter works in
the category C whose objects are pairs of spaces having the homotopy type of finite
CW -complex pairs and morphisms are continuous maps of such pairs. This is a
full subcategory of the category of pairs of topological spaces and maps of pairs,
and this admits the construction of mapping cones. Let C0 be the category whose
objects are pointed topological spaces having the homotopy type of pointed finite
CW -complexes and morphisms are continuous maps of such spaces. There exist
the (reduced) suspension functor : C0 → C0 and its adjoint functor : C0 → C0
which is the loop functor.
The idea of spectrum was originated by F. L. Lima (1929–) in 1958 and has
been proved to be very useful. This chapter studies cohomology theories associated
with different spectra, Brown representability theorem, stable homotopy groups,
homotopical description of cohomology theory, and the cohomology operations.
Around 1959, several topologists attempted to consider systems of covariant func-
tors h n : C → Ab (contravariant functors h n : C → Ab ), where n takes all values in
N or Z and Ab is the category of abelian groups and homomorphisms. These func-
tors satisfy all the axioms of Eilenberg and Steenrod for homology (cohomology)
with the exception of dimension axiom. The theory associated with such functors is
now known as generalized (or extraordinary) homology and cohomology theories
to distinguish them from ordinary homology and cohomology theories. There are
several functors from (based) topological spaces to graded abelian groups such as
stable homotopy groups or more generally, homology of a space with coefficients
DOI 10.1007/978-81-322-2843-1_15
476 15 Spectral Homology and Cohomology Theories
in a spectrum. These are examples of generalized homology (cohomology) theories

which are constructed in two equivalent methods: reduced and unreduced theories.
The former one is a functor on C0 and the latter one is a functor on C.
The cohomology groups H n (X ; G) of a space X ∈ C0 with coefficients in an
abelian group G can be characterized as the group of homotopy classes of maps of
X into the Eilenberg–MacLane space K (G, n). This asserts that the cohomology
theory with coefficients in G can be described in this way. The spaces K (G, n)
are considered as the components of a spectrum. Given a cohomology theory, one
may define the corresponding homology groups as the cohomology groups of the
complement of X in a sphere in which X is embedded.
For this chapter, the books (Eilenberg and Steenrod 1952), (Gray 1975), (Hatcher
2002), (Maunder 1980), (Spanier 1966) and papers (Brown 1962), (Whitehead 1962)
and some others are referred in the Bibliography.
15.1 Spectrum of Spaces
This section conveys the concept of a spectrum E = {E n , αn } of spaces introduced

by Lima in 1958. Let Ab be the category of abelian groups and homomorphisms.
Special spectra, called -spectra play a key role in algebraic topology. For exam-
ple, Eilenberg–MacLane spectrum is an -spectrum and it relates cohomology with
homotopy (see Theorem 15.5.4). Each spectrum E = {E n , αn } produces two differ-
ent sequences of functors h n (; , E) and
h n (; E) from the category C0 to the category
Ab , the first one is covariant and the second one is contravariant. These are called the
spectral homology and cohomology functors associated with the spectrum E, and in
brief abbreviated E n and
En.
Definition 15.1.1 A spectrum E = {E n , αn } of spaces in C0 is a sequence {E n } of
topological spaces in C0 together with a sequence of continuous maps
αn : E n → E n+1 in C0
(equivalently,
αn : E n → E n+1 in C0 ).
.
Definition 15.1.2 A spectrum E = {E n , αn } in C0 is is said to be an -spectrum if
αn : E n → E n+1 , n ∈ Z is a base point preserving weak homotopy equivalence
Definition 15.1.3 (Eilenberg–MacLane spectrum) The spectrum X = {X n , αn }

given by X n = K (Z, n) and αn : K (Z, n) → K (Z, n + 1), a base point preserving
weak homotopy equivalence, is called an Eilenberg–MacLane spectrum. In general,
the Eilenberg–MacLane spectrum X = {X n , αn }, denoted by H G, is defined by tak-
ing X n = K (G, n), where K (G, n) is an Eilenberg–MacLane space of type (G, n).
15.1 Spectrum of Spaces 477
Fig. 15.1 Map between α

n
ΣEn / En+1
spectra
Σfn fn+1
βn
ΣFn / Fn+1
Definition 15.1.4 (Suspension spectrum) For X ∈ C0 , the spectrum X = {X n , αñ }

defined by taking X n = n X , and α̃n : ( n X ) → n+1 X to be the natural home-
omorphism is called a suspension spectrum. Any suspension spectrum is clearly
of this form ‘up to weak homotopy equivalence’, where X = X 0 . This spectrum is
denoted by X.
Definition 15.1.5 (Sphere spectrum) If X n = S n in the suspension spectrum X =
{X n , αn }, then the spectrum is called sphere spectrum and it is abbreviated as S. Thus
the sphere spectrum S = {S n , αn }, where
αn : S n → S n+1
is the identity map.

Definition 15.1.6 (Unitary spectrum) Let U be the infinite unitary group. There is
a canonical homotopy equivalence f : U → 2 U. Suppose

U, if n is odd
En =
U, if n is even.
Let n be odd and vn : U → U be the map corresponding to f . Let n be even

and vn : U → U be the identity map. The resulting spectrum U is called the
unitary spectrum.
Definition 15.1.7 Let E = {E n , αn } and F = {Fn , βn } be two spectra. A map f :
E → F between spectra is a sequence of continuous maps
f n : E n → Fn
such that the diagram in Fig. 15.1 is homotopy commutative for each n.
15.2 Spectral Reduced Homology Theory
This section constructs spectral reduced homology theory on C0 associated with a

given spectrum E .
Definition 15.2.1 A graded abelian group is a sequence {G n } of abelian groups,
defined for each integer n. A homomorphism f : {G n } → {G n } of graded groups is
a sequence { f n } of homomorphisms f n : G n → G n .
Remark 15.2.2 Sometimes, one writes G ∗ for {G n }. Similar definitions are given
for graded R-modules, or graded sets. Such objects and homomorphisms form a
category denoted by μZ∗ (or Ab ), μ R , S∗ in the cases of graded abelian groups,
graded R-modules, and graded sets respectively.
Example 15.2.3 The sequence G n = πn (X, ∗) for n ≥ 1 and G n = 0 if n ≤ 0 is a

graded abelian group if G 1 = π1 (X, ∗) is abelian.
Definition 15.2.4 A reduced (spectral) homology theory on C0 associated with a

n } = {h n } from
spectrum E = {E n , αn } is a sequence of covariant functors written { E
C0 to Ab satisfying:
RH(i)(Homotopy axiom): Let f : X → Y be in C0 , and f ∗ : E m (X ) → E m (Y )
be the induced homomorphism. If f g in C0 then f ∗ = g∗ .
RH(ii)(Suspension axiom): There is a natural transformation σ : E m (X ) →
m+1 ( X ), which is an isomorphism.
E
RH(iii)(Exactness axiom): If (X, B) ∈ C and ∗ ∈ B, then the sequence
n (B) −−i−
E
∗
n (X ) −−−
→E
p∗
n (X/B)
−→ E
n (X ) for each n, where p : X → X/B is the map collapsing B to a

is exact at E
point and i : B → X is the inclusion map.
Each spectrum E = (E n ,
αn ) gives rise to a reduced homology theory on C0 , called
spectral reduced homology theory.
Definition 15.2.5 Given X ∈ C0 , consider the direct system
βn
→ πn+m (X ∧ E n ) −−−−→ πn+m+1 (X ∧ E n+1 ) −
··· − → ··· ,
where the homomorphisms βn are the composites
E (1d ×
αn )∗
πn+m (X ∧ E n ) −−−→ πn+m+1 (X ∧ E n ∧ S 1 ) −−−−−−−→ πn+m+1 (X ∧ E n+1 )
for n ≥ m, and E is the Freudenthal suspension homomorphism.

m (X ) = lim{πn+m (X ∧ E n ), βn }.
Define E
−
→
m } constitutes a reduced homology theory on C0 .
Theorem 15.2.6 { E
Proof Let f : X → Y ∈ C0 . To define its induced homomorphism f ∗ consider the

commutative diagram as shown in Fig. 15.2.
m (X ) → E
It gives rise to a map f ∗ : E m (Y ) such that 1d ∗ =1dentity, and ( f ◦
g)∗ = f ∗ ◦ g∗ . We now verify RH(i)–RH(ii). Since f g, f ∧ 1d g ∧ 1d and
hence
( f ∧ 1d )∗ = (g ∧ 1d )∗ : πn+m (X ∧ E n ) → πn+m (Y ∧ E n ).
15.2 Spectral Reduced Homology Theory 479
E (1d ×
αn )∗
πn+m (X ∧ En ) /π 1 / πn+m+1 (X ∧ En+1 )
n+m+1 (X ∧ En ∧ S )
(f ∧1d )∗ (f ∧1d )∗ (f ∧1d )∗
(1d ×
αn )∗

E / πn+m+1 (Y ∧ En ∧ S 1 ) / πn+m+1 (Y ∧ En+1 )
πn+m (Y ∧ En )
Fig. 15.2 Diagram for reduced homology theory
This implies RH(i).

Next define σ as follows. For any space X ∈ C0 , define
: πn (X ) → πn+1 (S 1 ∧ X ), [θ] → [1d ∧ θ].
Consider the commutative diagram in Fig. 15.3.

Replacing S 1 ∧ X by X , it follows that induces a natural transformation
σ:E m (X ) → E m+1 ( X ). Verify that σ is an isomorphism. For this verification
consider the diagram in Fig. 15.4 which commutes up to sign, where TX : S 1 ∧ X →
X ∧ S 1 = X is the natural homeomorphism. Hence RH(ii) follows. Verification
m } is a reduced homology theory
of RH(iii) is left as an exercise. Consequently, { E
on C0 . ❑
Example 15.2.7 For any CW -complex X , the homology group

S m (X ) associated
with the sphere spectrum S is given by the direct limit
E
πn+m (X ∧ S n ) −−−→ πn+m+1 (X ∧ S n+1 ) −
→ ··· .
S
These homology groups are also written πm (X ) for any CW -complex X and are
trivial if m < 0.
Fig. 15.3 Diagram πn+m (X ∧ En )

Σ /π 1
n+m+1 (S ∧ X ∧ En )
involving suspension
E E
homomorphisms
1 Σ /π 1
πn+m+1 (X ∧ En ∧ S ) n+m+2 (S ∧ X ∧ En ∧ S 1 )
(1d ∧
αn )∗ (1d ∧
αn )∗

Σ /π 1
πn+m+1 (X ∧ En+1 ) n+m+2 (S ∧ X ∧ En+1 )
Fig. 15.4 Diagram for Σ / π (S 1 ∧ X)

πn (X) n
construction of σ
(TX )∗
E E

Σ /π
πn (X ∧ S 1 ) n+1 (S
1
∧ X ∧ S 1)
15.3 Spectral Reduced Cohomology Theory
This section gives rise to spectral reduced cohomology theory which is dual to spectral
reduced homology theory. A cohomology theory is not just a collection of cohomol-
ogy functors. This needs connecting homomorphisms relating h n = [ , An ] with
h n+1 = [ , An+1 ]. Such a construction can be made using the map
αn : An → An+1
for some -spectrum A = { An , αn }, where the spaces An are unique up to homotopy

equivalence with h n+1 = [−, An+1 ].
Definition 15.3.1 A reduced (spectral) cohomology theory on C0 associated with a

n } ={h n }
spectrum E = {E n , αn } is a sequence of contravariant functors written { E
from C0 to Ab together with a sequence
σ n : h n+1 o → h n
of natural transformations such that

(i) Homotopy axiom: If f 0 , f 1 ∈ C0 and f0 f 1 in C0 , then h n ( f 0 ) = h n ( f 1 ) (some-
times written as f ∗ = g ∗ ) for all n;
(ii) Suspension axiom: σ n (X ) : h n+1 ( X ) → h n (X ) is an isomorphism for all

X ∈ C0 ; and
(iii) Exactness axiom: If (X, B) ∈ C and ∗ ∈ B, then the sequence
h n ( p) h n (i)
h n (X/B) −−−−−→ h n (X ) −−−−−→ h n (B)
is exact at h n (X ) for each n, where p : X → X/B is the map collapsing B to a point

and i : B → X is the inclusion map.
Furthermore, the theory is called an ordinary cohomology theory (reduced), if the

following axiom is also satisfied:
(iv) Dimension axiom: For the o-sphere S 0 , h n (S 0 ) = 0 if n = 0.
Definition 15.3.2 The graded group {h n (S 0 )} is called the coefficient system of the
cohomology theory {h n }.
Definition 15.3.3 The cohomology group h n (X ; A) of X ∈ C0 associated with the

spectrum A is defined to be direct limit group of the sequence of groups and homo-
morphisms
αn ∗ αn ∗
· · · → [X, An ] −−−−→ [X, An+1 ] −−−−−→ [X, 2 An+2 ] → · · ·
15.3 Spectral Reduced Cohomology Theory 481
i.e., h n (X ; A) = lim [X, k An+k ].

k→∞
Theorem 15.3.4 Let A = { An , αn } be an -spectrum and X be a C W -complex.

Then h n (X ; A) = [X, An ].
Proof As each αk : Ak → Ak+1 in C0 is weak homotopy equivalence, it follows

that h n (X ; A) = [X, An ]. ❑
15.4 Generalized Homology and Cohomology Theories
If a homology theory H satisfies only the first six axioms H(1)–H(6) (see Chap. 12),
then H is called a generalized (or extraordinary) homology theory on the category
C. On the other hand, a cohomology theory which satisfies only the first six axioms
C(1)–C(6) (see Chap. 12) is called a generalized (or extraordinary) cohomology
theory on the category C.
The 0-dimensional homology group
G = H0 ( point)
is called the coefficient group of the homology theory H. Consequently, the dimen-
sion axiom locates the coefficient group at the right dimension.
Remark 15.4.1 There is a natural question: whether there is a dual theory of coho-
mology theory for homology theory. The integral homology groups of a space X
can be described by the Dold–Thom theorem, as the homotopy groups of the infinite
symmetric product of X . However, the duality between homology and cohomology
is not apparent from this description, nor is it clear how to generalize it. Examples of
generalized homology theories are known; for instance, the stable homotopy groups
(see Sect. 15.10).
15.5 The Brown Representability Theorem
This section studies the Brown representability theorem which relates homotopy
theory with generalized cohomology theory. This theorem plays a key role in the
applications of homotopy theory to other areas. Moreover, Brown proved that under
certain conditions, any cohomology theory satisfying Eilenberg–Steenrod axioms
can be obtained in the form [, Y ] for some suitable space Y . More precisely, E.H.
Brown (1962) proved in his paper that if H satisfies certain axioms, there is a space
Y , unique up to homotopy type, such that H is naturally equivalent to the functor
which assigns to each CW -Complex X with base point, the set of homotopy classes
of maps of X into Y . More precisely, Brown representability theorem asserts that

there exist connected CW -complexes An with base point and natural equivalences
h n (X ) ∼
= [X, An ],
where X runs over connected CW -complexes with base point (see Brown 1962). So
we obtain a collection of spaces An (n ∈ Z).
If we divert attention from the reduced h n (X ) to relative groups h n (X, Y ) we
should divert attention from suspension isomorphisms
∼
=
h n (X ) −−−→
σ : h n+1 ( X ).
to the coboundary maps δ as a cohomology theory does not consists only of functors
h n ; they are connected by coboundary maps.
Let H n be a sequence contravariant functors from the category of pairs of finite
CW -complexes to the category of abelian groups and δq : H q (A) → H q+1 (X, A)
be a sequence of natural transformations. Furthermore, suppose H q and δ q satisfy
all the Eilenberg–Steenrod axioms except the dimension axiom which is replaced by
the condition that H q on a point be countable.
E.H. Brown prescribed a very simple set of conditions on a functor H in his land
mark paper (Brown 1959) that the functor H to be representable in the sense that H
is naturally equivalent to [, Y ] for some space Y . This space Y is called a classifying
space for the functor H . Every -spectrum represents a generalized cohomology
theory. Is its converse true? Do all cohomology theories arise in this way from an
-spectrum? E.H Brown gave necessary and sufficient conditions in 1962 under
which a contravariant functor T has the form [, Y ] for some fixed space Y. It shows
that there is a close relation between generalized cohomology theory and homotopy
theory.
Theorem 15.5.1 If A = { An , αn } is an -spectrum, then the functors X → h n (X ) =
[X, An ], n ∈ Z, define a reduced cohomology theory on the category of pointed CW-
complexes and base point preserving maps.
Proof See Brown (1962). ❑
Remark 15.5.2 Theorem 15.5.1 shows that every -spectrum A = { An , αn } on the
category C0 defines a cohomology theory given by h n (X ) = [X, An ], n ∈ Z, on the
category C0 . Is the converse of this theorem true? Brown proved in 1962 that all
cohomology theories arise in this way from an -spectrum A = { An , αn } on the
category C0 , where the spaces An are unique up to homotopy equivalence. In other
words, Brown proved in 1962 that there is an -spectrum A = { An , αn } such that
H n (X ) is naturally equivalent to the group of homotopy classes of maps of X into An .
Remark 15.5.3 There are natural isomorphisms between the groups h n (X ; Z) and
H n (X ; Z) for the integral coefficients. In an analogous way, given an abelian
group G, the singular cohomology group H n (X ; G) can be defined as H n (X ; G) =
[X ; K (G, n)]. In particular,
15.5 The Brown Representability Theorem 483
(i) for n = 1, H 1 (X ; Z) = [X, S 1 ];

(ii) for n = 2, H 2 (X ; Z) = [X ; CP ∞ ]. This result implies that the complex line
bundles are classified by the elements of H 2 (X ; Z).
Theorem 15.5.4 (Brown Representability Theorem) Every reduced cohomology

theory h ∗ on the category C0 has the form h n (X ) = [X, An ] for some -spectrum
A = { An , αn }, where the spaces An are unique up to homotopy equivalence.
Proof See Brown (1962). ❑
Corollary 15.5.5 Ordinary cohomology is representable as maps into Eilenberg–

MacLane spaces.
Proof Let H ∗ (−; G) be an ordinary cohomology theory. Then πi (An ) = [S i , An ] =

H n (S i ; G), where
G if i = n
H n (S i ; G) =
0 otherwise
This shows that An is an Eilenberg–MacLane space K (G, n). This implies that
H n (−; G) = [−, K (G, n)]. ❑
Remark 15.5.6 Consider the singular ordinary cohomology group H n (X ; G) for

a C W -complex X with coefficients in an abelian group G. Then by Remark 15.5.3
H n (X ; G) is the set of homotopy classes of maps from X to K (G, n), the Eilenberg–
MacLane space of type (G, n). The corresponding spectrum is the Eilenberg–
MacLane spectrum H G has the nth space K (G, n).
Remark 15.5.7 Let h ∗ be a generalized cohomology theory defined on C W -complex

pairs. Then
h n (X ) = h n (X, point) ⊕ h n ( point)
and hence define h n (X ) = h n (X, point). Now applying Brown representability the-
orem, there exists a connected C W -complex An with a base point and natural equiv-
alences such that

h n (X ) = [X, An ]
where X runs over connected C W -complexes with base points. In this way, a
sequence of spaces {An } is obtained. However, a cohomology theory does not contain
only the functors; also contains coboundary maps connecting them.
15.6 A Generalization of Eilenberg–MacLane Spectrum

and Construction of Its Associated Generalized
Cohomology Theory
In this section the author of the present book constructs a new -spectrum general-
izing the Eilenberg–MacLane spectrum and also presents its associated cohomology
theory. The motivation of this construction comes from the Corollary 15.5.5 which
asserts that the singular ordinary cohomology groups of a C W -complex can be
identified with the groups of homotopy classes of continuous maps into Eilenberg–
MacLane spectrum spaces. This section investigates a new generalized cohomol-
ogy theory, constructed by replacing the Eilenberg–MacLane spectrum by a new
-spectrum.
This section works in the category C whose objects are pairs of spaces having the
homotopy type of finite CW -complexes and morphisms are maps of such pairs. This
is a full subcategory of the category of pairs of topological spaces and maps of pairs,
and this admits the construction of mapping cones. In particular, there exists the
(reduced) suspension functor : C → C and its adjoint functor : C → C which is
the loop functor. Let C0 be the full subcategory of C, whose objects are spaces with
base points, Ab be the category of abelian groups and homomorphisms.
15.6.1 Construction of a New -Spectrum A
This subsection constructs a new -spectrum generalizing the Eilenberg–MacLane

spectrum. Recall that an -spectrum A is a sequence of spaces An , n ∈ Z, in C0 ,
together with a sequence of maps αn : An → An+1 , n ∈ Z, in C0 , where each
αn is a weak homotopy equivalence. Let (X, ∗) ∈ C0 and X n , n > 0 be the n-fold
cartesian product of X and Sn be the group of permutations of {1, 2, 3, . . . , n}. Define
a right action
X n × Sn → X n , (x 1 , x2 , . . . , xn ).α = (X α(1) , X α(2) , . . . , X α(n) ).
The orbit space S P n (X ) = X n mod Sn is the n-fold symmetric product of X

i.e., it is the quotient space of the n-fold product X n obtained by identifying all n-
tuples (x 1 , x 2 , . . . , xn ) that differ only by a permutation of their coordinates. Thus a
typical point of S P n (X ) is an unordered n-tuples (x1 , x2 , . . . , xn ), xi ∈ X . There is a
natural inclusion X n ⊂ X n+1 , (x 1 , x2 , . . . , xn ) → (∗, x 1 , x2 , . . . , x n ). This induces
an inclusion S P n (X ) ⊂ S P n+1 (X ). This gives rise to an ascending sequence of
spaces of S P n (X )’s with the weak topology
S P n (X ) ⊂ S P n+1 ⊂ · · · ⊂ · · ·
15.6 A Generalization of Eilenberg–MacLane Spectrum … 485
S P ∞ (X ) is defined by the union of this ascending sequences i.e., S P ∞ (X ) =

lim S P n (X ).
n→∞
Dold and Thom showed that S P ∞ (S n ) = K (Z, n) which is the Eilenberg–
MacLane space with homotopy groups πr (K (Z, n)) are all zero except for r = n
(Dold and Thom 1958). We now utilize homotopy properties of infinite sym-
metric product (see Sect. 2.9 of Chap. 2): If f : (X, ∗) → (Y, ∗) is in C0 , then
S P n ( f ) : S P n (X ) → S P n (Y ) is the map defined by passing to the quotient, and
S P ∞ ( f ) : S P ∞ (X ) → S P ∞ (Y ) is the map defined by passing to the limit. S P ∞
(and also S P n for each n ≥ 0) is a covariant functor C0 → C0 that if f : (X, ∗) →
(Y, ∗) is a homotopy equivalence, so is S P ∞ ( f ). Moreover, if X is a connected
CW -complex, there is a weak homotopy equivalence
ρ : S P ∞ (X ) → S P ∞ ( X )
defined by
[ρ(x 1 , x2 , . . . , x n )](t) = ((x 1 , t), (x2 , t), . . . , (xn , t)), t ∈ I ).
Using these facts, the author of the present book has constructed a new -spectrum
A = { An , αn } generalizing the Eilenberg–MacLane spectrum in the following way.
Let Y be a connected CW -complex, define

−n S P ∞ (Y ) if n < 0
An =
S P ∞ ( n+1 Y ) if n ≥ 0
The homotopy equivalence αn : An → An+1 is defined by

identity if n < −1
αn =
◦ ρn if n ≥ −1
where ρn : S P ∞ ( n+1 Y ) → S P ∞ ( n+2 Y ), if n ≥ −1, is a homotopy equiva-

lence. Note that ρ−1 = ρ.
If we take in particular Y = S 0 , the 0-sphere, then An = K (Z, n). Consequently,
for Y = S 0 , the above -spectrum becomes an Eilenberg–MacLane spectrum. In
this way, we obtain an -spectrum A = { An , αn } which is a generalization of the
Eilenberg–MacLane spectrum whose sequence of spaces are K (Z, n).
15.6.2 Construction of the Cohomology Theory Associated

with A
The author of this book constructs the cohomology theory associated with his
-spectrum A = { An , αn } defined in the previous subsection by generalizing the
Eilenberg–MacLane spectrum. Since αk is weak homotopy equivalence
h n (X ; A) = [X, An ].
The coefficient system of the theory can be obtained in the following way: If
n < 0,
h n (S 0 , A) = [S 0 , −n S P ∞ (Y )] = [ −n S 0 , S P ∞ (Y )]
= [S −n , S P ∞ (Y )]
= π−n (S P ∞ (Y ))
= H−n (Y ; Z),
the −nth singular homology group of Y with coefficients in Z.

If n ≥ 0,
h n (S 0 , A) = [S 0 , S P ∞ ( n+1 Y )] = [ S 0 , S P ∞ ( n+1 Y )]
= [S 1 , S P ∞ ( n+1 Y )]
= π1 (S P ∞ ( n+1 Y ))
= H1 ( n+1 Y ; Z),
the one-dimensional singular homology group of n+1 Y with coefficients in Z.

In general, one may calculate the cohomology groups of the n-sphere S n in the
theory. If n < 0,
h n (S k , A) = [S k , −n S P ∞ (Y )] = [ −n S k , S P ∞ (Y )]
= [S k−n , S P ∞ (Y )]
= πk−n (S P ∞ (Y ))
= Hk−n (Y ; Z),
If n ≥ 0,
h n (S k , A) = [S k , S P ∞ ( n+1 Y )] = [ S k , S P ∞ ( n+1 Y )]
= [S k+1 , S P ∞ ( n+1 Y )]
= πk+1 (S P ∞ ( n+1 Y ))
= Hk+1 ( n+1 Y ; Z).
Remark 15.6.1 The relations displayed above show that our generalized cohomol-
ogy theory has some close relations with the ordinary singular homology theory with
integral coefficients.
15.7 K -Theory as a Generalized Cohomology Theory 487
15.7 K -Theory as a Generalized Cohomology Theory
This section presents an exposition of K -theory (instead its development), which

is the first example of generalized cohomology theories. It plays a centrally impor-
tant role in connecting algebraic topology to analysis and algebraic geometry. For
example, an outstanding purely algebraic result in K-theory is that the only possible
dimensions of a real (not necessarily associative) division algebra are 1, 2, 4, and 8
proved by J.F. Adams and M. Atiyah to solve ‘Hopf invariant one problem’. This
result asserts that R, C, H and Cayley numbers (an eight-dimensional non-associative
algebra) are the only real division algebras (see Chap. 17).
Recall that the Grothendieck group K (X ) is defined in Chap. 5. Given a finite
dimensional C W -complex X this group K (X ) has the natural structure of a com-
mutative ring stemming from the tensor product of vector bundles. If f : X → Y is
a continuous map, then f induces a ring homomorphism
f ∗ : K (Y ) → K (X ),
where to each vector bundle ξ over X , there is the induced bundle f ∗ (ξ) over Y .
Moreover,
if f g : X → Y, then f ∗ = g ∗ : K (Y ) → K (X ).
Hence the correspondence

X → K (X )
defines a homotopy invariant functor from category C0 to the category of rings.

Let (X, x0 ) be a CW -complex. The natural inclusion {x0 } → X induces a homo-
morphism of rings
K (X ) → K ({x0 }) = Z, [ξ] − [η] → dim ξ − dim η ∈ Z (15.1)
Let K 0 (X, x0 ) denote the kernel of the homomorphism (15.1)
K 0 (X, x 0 ) = ker(K (X ) → K (x0 )).
The elements of the subring K 0 (X, x0 ) are represented by differences [ξ] − [η]
for which dim ξ = dim η.
Let (X, Y ) be a finite CW -pair of spaces and K 0 (X, Y ) denote the ring
K 0 (X, Y ) = K 0 (X/Y ).
For negative integer −n, let
K −n (X, Y ) = K 0 ( n X, n Y ),
where n X = (S n × X )/(S n ∨ X ).
The above discussion can be summarized in the basic and important result.
Theorem 15.7.1 K ∗ (−) forms a generatized cohomology theory on C.
Remark 15.7.2 For deeper properties and applications of vector bundles, it is sug-
gested to study Bott periodicity theorem as the main tool for calculation of K -theory,
linear representations and cohomology operations in K -theory and Aitiyah–Singer
formula for calculation of the indices of elliptic operators on compact manifolds and
for this purpose the book (Luke and Mishchenko 1998) is referred.
15.8 Spectral Unreduced Homology

and Cohomology Theories
This section generalizes the concepts of spectral homology and cohomology groups
defined on C0 by introducing the concepts of spectral homology and cohomology
groups defined on C. The domain of our theories from the category C0 to C is trans-
ferred by a simple transformation. Homology and cohomology theories defined on
pairs (X, A) are called unreduced homology and cohomology theories (sometimes
the word ‘unreduced’ is dropped).
Definition 15.8.1 Let E = {E n ,
αn } be a spectrum on C. For (X, A) ∈ C, set
m (X ∪ C A), E m (X, A) = E
E m (X, A) = E m (X ∪ C A),
where the vertex of the cone is taken as a base point. In particular, if A = ∅, we

interpret X ∪ C A as the space X with a point added, which is used as base point.
{E m } and {E m } are called the spectral unreduced homology and cohomology theories
associated with the spectrum E.
Definition 15.8.2 An unreduced homology theory on C is a sequence of covariant
functors E m : C → Ab (category of abelian groups) for m ∈ Z satisfying the axioms
URH(i)(Homotopy axiom): Let f, g : (X, A) → (Y, B) ∈ C, and f g. Then
f ∗ = g∗ : E m (X, A) → E m (Y, B);
URH(ii)(Excision axiom): If U is open and U ⊂ Int A, then the inclusion map
e : (X − U, A − U ) → (X, A) induces isomorphisms in homology;
URH(iii)(Exactness axiom): There are natural transformations ∂ : E m (X, A) →
E m−1 (A) which fit up to exact sequence
∂
→ E m (A) −
··· − → E m (X ) −
→ E m (X, A) −−−→ E m−1 (A) −
→ ···
15.8 Spectral Unreduced Homology and Cohomology Theories 489
An unreduced cohomology theory associated to a spectrum is dual to the corre-

sponding unreduced homology theory.
Definition 15.8.3 An unreduced cohomology theory on C is a sequence of con-
travariant functors E m : C → Ab for m ∈ Z satisfying the axioms
URC(i)(Homotopy axiom): Let f, g : (X, A) → (Y, B) ∈ C, and f g. Then
f ∗ = g∗ : E m (Y, B) → E m (X, A);
URC(ii)(Excision axiom): If U is open and U ⊂ Int A, then the inclusion map
e : (X − U, A − U ) → (X, A) induces isomorphisms in cohomology;
URC(iii)(Exactness axiom): There are natural transformations δ : E m (A) →
E m+1 (X, A) which fit into exact sequence
δ
− E m (A) ←
··· ← − E m (X ) ←
− E m (X, A) ←−−− E m−1 (A) ←
− ···
Remark 15.8.4 E m (X, ∅) is abbreviated E m (X ) and E m (X, ∅) is abbreviated E m (X ).
15.9 Cohomology Operations
This section studies cohomology operations which form an important topic in alge-
braic topology. The technique utilized for developing the algebraic structure of the
cohomology ring has substantially enriched homotopy theory with some surprising
results. Eilenberg–MacLane spaces are closely linked with the study of cohomology
operations. The cohomology of an Eilenberg–Maclane space K (G, n), depending
on n and G has the surprising property. A cohomology operation is a natural trans-
formation of cohomology functors

H n ( ; G) → H n (−; G ).
Given integers n, m and abelian groups G, G , in general, a cohomology operation

is a natural transformation
ψ : H n (X, Y ; G) → H m (X, Y ; G )
subject to one axiom only:
if f : (X, Y ) → (X , Y ) is continuous and h ∈ H n (X , Y ; G), then ψ( f ∗ h) =

f ∗ (ψh).
On the other hand, a stable cohomology operation is a collection of cohomology
operations
ψn : H n (X, Y ; G) → H n+r (X, Y ; G ).
Here n runs over Z, but r, G and G are fixed and each ψn is a natural trans-
formation. Moreover, it is also necessary the following diagram in Fig. 15.5 to be
commutative for each n.
Fig. 15.5 Diagram for δ / H n+1 (X, Y ; G)

H n (X, Y ; G)
stable cohomology operation
ψn ψn+1

δ / H n+r+1 (X, Y ; G )
H n+r (X, Y ; G )
Example 15.9.1 Let G = G = Z2 . Then ψn is called the Steenrod square denoted

by Sqr .
Remark 15.9.2 A cohomology operation is a concept which can be applied in any

dimension.
Remark 15.9.3 Given a cohomology operation
ψ : H n (X, Y ; G) → H m (X, Y ; G ),
it need not be necessary to appear as the nth term of any stable cohomology operation.
15.9.1 Cohomology Operations of Type (G, n; T, m)

and Eilenberg–MacLane Spaces
Ordinary cohomology of CW -complexes is representable by an Eilenberg–MacLane

space. This section identifies the set of all cohomology operations of type (G, n; T, m)
with the mth ordinary cohomology group of the Eilenberg–MacLane space K (T, n)
with coefficient group G.
Definition 15.9.4 Let G, T be abelian groups and m, n be nonnegative integers.

A cohomology operation θ of type (G, n; T, m) is a natural transformation (Nat)
θ : H n (−; G) → H m (−; T ) of functors defined on the category C0 of pointed CW -
complexes.
Example 15.9.5 For each n, and each ring R the operation
H n (X ; R) → H 2n ; R), x → x ∪ x = x 2
(cup product) is a cohomology operation. This is not generally a homomorphism.
The set of all cohomology operations Nat(H n (−; G), H m (−; T )) of type (G, n;
T, m) is denoted by O (G, n; T, m). We now identify the set O (G, n; T, m) with
the ordinary cohomology groups of the Eilenberg–MacLane spaces.
Theorem 15.9.6 Nat(H n (−; G), H m (−; T )) = H m (K (T, n); G).

15.9 Cohomology Operations 491
Proof Ordinary cohomology of CW -complexes is representable by an Eilenberg–

MacLane space. Hence using Yoneda lemma (see Appendix B), a cohomology opera-
tion of type (n, m, G, T ) is given by a homotopy class of maps K (G, n) → K (T, n).
Again, by representability, the cohomology operation is given by an element
H m (K (G, n); T ), because,
Nat(H n (−; G), H m (−; T )) = Nat([−, K (G, n)], [−, K (T, m)])
= [K (G, n), K (T, m)]
= H m (K (G, n); T )(mth ordinary cohomology group
of K(G, n) with coefficient group T ). ❑
The above discussion is summarized in the basic surprising result.

Theorem 15.9.7 Given an Eilenberg–MacLane space X = K (G, n) there is a (1-1)
correspondence between H m (K (G, n); T ) and the set of all cohomology operations
from H n (X ; G) to H m (X ; T ) for some Eilenberg–MacLane space K (T, m).
Remark 15.9.8 For more study on cohomology operations (Mosher and Tangora
1968), (Steenrod and Epstein 1962) and (Spanier 1966) are referred.
15.9.2 Cohomology Operation Associated with a Spectrum
This subsection studies cohomology operations in the new cohomology theory

h ∗ ( ; A) constructed in Sect. 15.6.2. There also exist relations between the cohomol-
ogy operations and the general cohomology groups of some spaces in this general
cohomology theory. This subsection establishes some such relations. More precisely,
we prove that the abelian group of all cohomology operations of degree k for the
cohomology theory h ∗ ( ; A) is isomorphic to the group h n+k (S P ∞ ( n Y ); A) and
the graded abelian group of all stable cohomology operations of degree k for the
cohomology theory h ∗ ( ; A) is isomorphic to the group lim h n+k (S P ∞ ( n Y ); A).
←
∗
Definition 15.9.9 A cohomology operation in h ( ; A) of degree k is a natural
transformation φm : h m ( ; A) → h m+k ( ; A).
Proposition 15.9.10 Let Om k be the set of all cohomology operations of degree k of

the type φm for the cohomology theory h ∗ ( ; A). Then Om k forms an abelian group.
Proof We define an addition ‘+’ on Om k by the rule: (φm + ψm )(X )(x) = φm (X )

(x) + ψm (X )(x), for all x ∈ h m (X ; A) and for all X ∈ C0 , where the right-hand side
addition is the addition in the additive abelian group h m+k (X ; A). Then (Om k , +) is
an abelian group. ❑
Remark 15.9.11 The group Om k has some interesting properties. For example, in
the ordinary singular cohomology theory H ∗ ( ; Z), the group Om k is isomorphic to
the group H m+k (K (Z, m); Z), where K (Z, m) is an Eilenberg–MacLane space.
Theorem 15.9.12 The homomorphism
λ : Om k → h m+k (S P ∞ ( m Y ); A) defined by λ(φ) = φ[I d],
where [I d] is the homotopy class of the identity map
Id : S P ∞ ( m Y ) → S P ∞ ( m Y ),
is an isomorphism of groups.
Proof Let x ∈ h m (X ; A) be represented by a map g : X → S P ∞ ( m Y ) in C0 . We

now define a map
μ : h m+k (S P ∞ ( m Y ); A) → Om k by the rule μ(α)(x) = [α ◦ g] = g ∗ (α),
for all α ∈ h m+k (S P ∞ ( m Y ); A). Now μ(α + β)(x) = g∗ (α + β) = g ∗ (α) +

g ∗ (β). This implies that μ is a homomorphism. Moreover, λ ◦ μ =identity and μ ◦ λ
= identity. Hence λ is an isomorphism with its inverse μ. ❑
Corollary 15.9.13 For the ordinary singular cohomology theory H ∗ ( ; Z), the
group Om k is isomorphic to the group H m+k (K (Z, m); Z), where K (Z, m) is an
Eilenberg–MacLane space.
Proof For Y = S 0 (0-sphere), the space S P ∞ ( m Y ) becomes the Eilenberg–

MacLane space K (Z, m) and the general cohomology theory reduces to the ordi-
nary singular cohomology theory H ∗ ( ; Z). Hence the Corollary follows from
Theorem 15.9.12. ❑
15.9.3 Stable Cohomology Operations
Definition 15.9.14 For the cohomology theory h ∗ ( ; A), a stable cohomology oper-
ation of degree k is a sequence φm : h m ( ; A) → h m+k ( ; A) of cohomology oper-
ations of degree k such that the following diagram commutes, i.e., σ m+k (φm (x)) =
φm+1 (σ m (x)), ∀x ∈ h m (X ; A) and ∀X ∈ C0 , where σ m is the suspension isomor-
phism in h ∗ ( ; A).
k
Remark 15.9.15 Let {O m } be the set of all stable cohomology operations of degree
k
k for the cohomology of h ∗ ( ; A). We denote a sequence {φm } ∈ {O m } by a single
letter φ (Fig. 15.6).
Fig. 15.6 Diagram for φm

hm ( ; A) / hm+k+1 ( ; A)
cohomology operation of
degree k
σm σm+k
φm+1

hm+k (Σ, A) / hm+k+1 (Σ, A)
k
Proposition 15.9.16 {O m } forms an additive abelian group.
k
Proof We define an addition ‘+’ on {O m } by the rule (φ + ψ)(x) = φ(x) + ψ(x),
k
∀x ∈ h ∗ (X ; A) and ∀X ∈ C0 . Then {O m } becomes an additive abelian group. ❑
k
15.9.4 A Characterization of the Group {O m }
For this purpose, we define a sequence of homomorphisms
γm+k : h m+k+1 (S P ∞ ( m+1 Y ); A) → h m+k (S P ∞ ( m Y ); A)
as the product of homomorphisms
(ρm )∗
h m+k+1 (S P ∞ ( m+1 Y ); A) −−−−−→ h m+k+1 ( S P ∞ ( m Y ); A)
−1
(σ m+k )
−−−−−−−→ h m+k (S P ∞ ( m Y ); A),
−1
i.e., γm+k = (σ m+k ) ◦ (ρm )∗ , where (ρm )∗ is the homomorphism induced by the
adjoint map
ρm of ρm : S P ∞ ( m Y ) → S P ∞ ( m+1 Y )
given by Spanier. Hence the following sequence of abelian groups and homomor-
phisms
γm+k
→ h m+k+1 (S P ∞ ( m+1 Y ); A) −−−−−→ h m+k (S P ∞ ( m Y ); A) −
··· − → ···
form an inverse system of groups and homomorphisms. Let lim

←h
m+k
(S P ∞ ( m Y ); A)
denote the inverse limit group of the above system.
k
Theorem 15.9.17 The graded abelian group {O m } of all stable cohomology oper-
ations of degree k for the cohomology theory h ∗ ( ; A) is isomorphic to the group
lim m+k
← h (S P ∞ ( m Y ); A).
Proof It follows from the definition of lim h m+k (S P ∞ ( m Y ); A) that an element
←
of lim h m+k (S P ∞ ( m Y ); A) is a sequence of elements xm ∈ h m+k (S P ∞ ( m Y ); A)
←
−1
such that γm+k (xm+1 ) = xm . Hence (σ m+k ) ◦ (ρm )∗ (xm+1 ) = xm , i.e., (ρm )∗
(xm+1 ) = σ m+k (xm ).
We now show that to each sequence of elements xm ∈ h m+k (S P ∞ ( m Y ); A),
k
there corresponds a stable cohomology operation of degree k in {O m } and con-
versely.
Let μ : h m+k (S P ∞ ( m Y ); A) → Om k be the homomorphism defined in
Theorem 15.9.12. Let μ(xm ) = φm . We now show that {φm } is a stable cohomol-
ogy operation of degree k in h ∗ ( ; A).
Let x ∈ h m (X ; A) be represented by a map f : X → S P ∞ ( m Y ). Then σ m (x)
is represented by the composite map
f ρm
X −−−−→ S P ∞ ( m Y ) −−−−→ S P ∞ ( m+1 Y ).
Again φm (x) = μ(xm )(x) = f ∗ (xm ). Hence we have
φm+1 (σ m (x)) = μ(x m+1 )(σ m (x)) = (ρm ◦ f )∗ (xm+1 ) = ( f )∗ ◦ (ρm )∗ (xm+1 )
= ( f )∗ (σ m+k (xm )) = σ m+k ( f ∗ (xm )) = σ m+k (φm (x)), ∀x ∈ h m (X ; A)
k
⇒ φm+1 ◦ σ m = σ m+k ◦ φm ⇒ {φm } ∈ {O m }.
φm+1 (σ m (x)) = μ(xm+1 )(σ m (x)) = (ρm ◦ f )∗ (xm+1 )

= ( f )∗ ◦ (ρm )∗ (x m+1 )
= ( f )∗ (σ m+k (xm ))
= σ m+k ( f ∗ (x m )) = σ m+k (φm (x))∀x ∈ h m (X ; A)
k
⇒ φm+1 ◦ σ m = σ m+k ◦ φm ⇒ {φm } ∈ {O m }
k
Conversely, let {φm } ∈ {O m }. Then φm+1 (σm (x)) = σ m+k (φm (x)), ∀x ∈ h m
(X ; A). Let λ(φm ) = xm , where λ is defined in Theorem 15.9.12. Then xm ∈ h m+k
(S P ∞ ( m Y ); A). Hence μ(x m ) = φm . Consequently,
φm+1 (σ m (x)) = μ(x m+1 )(σ m (x)) = (ρm ◦ f )∗ (xm+1 ) = ( f )∗ ◦ (ρm )∗ (xm+1 ).
Again,
σ m+k (φm (x)) = σm+k (μ(xm )(x)) = σ m+k ( f ∗ (xm )) = ( f )∗ ◦ σ m+k (xm ).
Hence it follows that corresponding to each sequence φm , there exists a sequence

of elements x m ∈ h m+k (S P ∞ ( m Y ); A) such that σ m+k (xm ) = (ρm )∗ (xm+1 ). ❑
k
Corollary 15.9.18 If Y = S n , then the graded abelian group {O m } of all sta-
ble cohomology operations of degree k in h ∗ ( ; A) is isomorphic to lim H m+k+n
←
(K (Z, m + n); Z), where H ∗ ( ; Z) is the ordinary cohomology theory with coeffi-
cients in Z.
Proof
h m+k (S P ∞ ( m S n ); A) = h m+k (K (Z, m + n); A) = H m+k+n (K (Z, m + n); Z).
Hence the Corollary follows from Theorem 15.9.17. ❑

k
Corollary 15.9.19 The graded abelian group {O m } of all stable cohomology oper-
ations of degree k in h ∗ ( ; A) is isomorphic to lim
← H
m+k
(B(N, m Y ); A), where
B(N, Y ) is the reduced monoid of the singular o-chain of m Y with coefficients
m
in N (multiplicative monoid of positive integers), i.e., B(N, m Y ) is the set of all

functions (X, ∗) → (N, 1) with finite support.
Proof In this case, h m+k (S P ∞ ( m Y ); A) = h m+k (B(N, m Y ); A). Hence the

Corollary follows from Theorem 15.9.17. ❑
Definition 15.9.20 Let E = {E n , en } and E = {E n , en } be two spectra. Let αn :

E n → E n be defined for n ≥ n 0 such that the diagram in Fig. 15.7 is commutative.
Such a sequence {αn } is called a map of spectra of degree r .
Remark 15.9.21 Clearly, spectra form a category with this definition of morphism.
Moreover, if α = {αn } : E → E is a map of spectra of degree r , then it induces
natural homomorphisms of homology and cohomology theories
m (X ) → E
α:E m−r (X ), α : E
mn (X ) →
E
m+r
(X )
for all m commuting with the suspension isomorphism, i.e., α(σ(x)) = σ(α(x)).
Definition 15.9.22 The above transformation α is called a stable homology or coho-

mology operation associated with spectrum. The simplest examples of such opera-
tions are coefficient transformations.
Fig. 15.7 Diagram for a Σαn

ΣEn / ΣE
n+r
map of spectra of degree r
en en+r
αn+1

En+1 / E
n+r+1
15.10 Stable Homotopy Theory and Homotopy Groups

Associated with a Spectrum
This section conveys the concept of ‘stable homotopy groups’ introduced in 1937 as
a natural generalization of Freudenthal suspension theorem. In algebraic topology
we use the word ‘stable’ when a phenomenon 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 develop-
ments in the late 1950s. One is the invention 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 classifying manifolds up to cobordism to
a problem, a solvable problem in stable homotopy theory. Moreover, this section
studies homotopy groups of a spectrum.
15.10.1 Stable Homotopy Groups
This subsection conveys a study of stable homotopy groups. If X is an n-connected

complex, then the suspension map : πr (X ) → πr +1 (X ) is an isomorphism for
r < 2n + 1. In particular, is an isomorphism for r ≤ n. This shows that X is an
(n + 1)-connected CW -complex. Hence it follows that m X is (n + m)-connected.
Consider the sequence of groups
πr (X ) −
→ πr +1 ( X ) −
→ ··· −
→ πr +m ( m X ) −
→ ··· (15.2)
Since m X is (n + m)-connected, the map 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).
Hence for m sufficiently large, the sequence of maps in (15.2) are all isomor-
phisms. The resulting group is called the stable homotopy group denoted by πrs (X ).
Since adding any finite number of terms to the beginning of (15.2) does not affect
the resulting stable homotopy group, πrs (S n ) ∼
= πrs (S 0 ). This shows that the only sta-
ble homotopy groups of spheres are the ones πrs (S 0 ) for some value of r , which is
simply denoted by πrs . These stable homotopy group classify the mapping of (r + m)-
dimensional spheres onto m-dimensional spheres, for sufficiently large value of m.
Stable phenomena had of course appeared implicitly before 1937; reduced homol-
ogy and cohomology are examples of 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.
We recall that the suspension homomorphism asserts: πi (S n ) → πi+1 (S n+1 ) is
an isomorphism for i < 2n − 1 and an epimorphism for i = 2n − 1. More gener-
ally, this holds for the suspension πi (X ) → πi+1 ( X ), whenever X is an (n − 1)-
15.10 Stable Homotopy Theory and Homotopy Groups Associated … 497
connected CW -complex for n ≥ 1. If q is small relative to n, then πn+q (S n ) is inde-

pendent of n.
Example 15.10.1 Consider the homotopy groups πn+r (S n ) of spheres. We have the
suspension homomorphism:
E : πn+r (S n ) → πn+r +1 (S n+1 ).
The Freudenthal suspension theorem says that this homomorphism is an isomor-

phism for n > r + 1.
For example, πn+1 (S n ) is isomorphic to Z2 for n > 2. The groups πn+r (S n ) (n >
r + 1) are called the stable homotopy groups of spheres.
More generally, let X and Y be two CW -complexes with base point which is
assumed to be a 0-cell. The suspension X is the reduced suspension: either S 1 ∧ X
or X ∧ S 1 which are homeomorphic. If f : X → Y is a map between CW -complexes
with base point, its suspension f is to be 1d ∧ f : S 1 ∧ X → S 1 ∧ Y (or f ∧ 1d :
X ∧ S 1 → Y ∧ S 1 ). Suspension defines a function
S : [X, Y ] → [ X, Y ]
Theorem 15.10.2 Suppose Y is n- connected for n ≥ 1. Then S is onto if dim X ≤

2n + 1 and a bijection if dim X < 2n.
Proof See Spanier (1966, pp 458). ❑
Definition 15.10.3 An element of [X, Y ] (defined under the above situation) is called
a stable homotopy class of maps.
Consider the exact homotopy sequence of the fibration

p∗
p : S 3 → S 2 : · · · → π3 (S 1 , s0 ) −
→ π3 (S 3 , s0 ) −−−−→ π3 (S 2 , s0 )
−
→ π3 (S 1 , s0 ) → · · ·
Since π3 (S 1 , s0 ) = π2 (S 1 , s0 ) = 0, p∗ : π3 (S 3 , s0 ) → π3 (S 2 , s0 ) is an isomor-
phism. Consequently, π3 (S 2 , s0 ) ∼ = Z, the first example, where πm (S n , s0 ) = 0 for
m > n. Since π3 (S 3 , s0 ) is graded by [1S 3 ], it follows that π3 (S 2 , s0 ) is generated by
[ p]. The map p is called the Hopf map.
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
homotopy group.
Consider the sequence of groups πn+q ( n X, ∗) for n = 0, 1, . . . , and the suspen-

sion homomorphisms between them. If X is a CW -complex, then n X has no cells
in dimension < n except for a 0-cell. Hence it is (n − 1)-connected, and
πn+q ( n X, ∗) ∼
= πn+q+1 ( n+1 X, ∗) ∼
= · · · if n > q + 1 (15.3)
Consequently, for large n, the sequence (15.3) stabilizes in the sense that all the
groups in this sequence are isomorphic.
Definition 15.10.4 The stable value in the sequence (15.3) is called the qth stable
homotopy group of X or q-stem of X denoted πqS (X ).
Remark 15.10.5 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 classifying manifolds up to
cobordism to a problem, a solvable problem in stable homotopy theory (see Gray pp.
324–357).
Remark 15.10.6 Higher algebraic K -theory introduced by Quillen in the early 1970
earns deep recognition by Segal and others.It can be viewed as a construction in
stable homotopy.
The coefficient groups πqS (S 0 ) are called the stable stems πqS . These groups are
known only through a finite range of n > 0 (note πnS = 0 for n < 0, π0S ∼ = Z). For
details study see James (1995).
If X and Y are finite CW -complexes and f : X → Y induces the zero homomor-
phism: πqS (X ) → πqS (Y ), then k f is nullhomotopic for some k.
15.10.2 Homology Groups Associated with a Spectrum
This subsection returns to homology theory associated with a given spectrum in

C0 . Such theories are closely related with stable homotopy theory and the study of
spectra. Let A = { An , αn } be an arbitray spectrum. Then the spaces X ∧ An form a
spectrum X ∧ A and the resulting groups are given by
h r (X ; A) = lim πn+r (X ∧ An )
n→∞
for a C W -complex X which form a homology theory, called homology theory asso-
ciated with the spectrum A (Whitehead 1962).
15.10 Stable Homotopy Theory and Homotopy Groups Associated … 499
15.10.3 Homotopy Groups Associated with a Spectrum
This subsection defines homotopy groups associated with a spectrum. We work in

category C0 . First we define the homotopy groups of a spectrum A = { An ,
αn }. These
are really stable homotopy groups. We have the following homomorphisms:
(
αn )∗
πn+r (An ) → πn+r +1 ( An+1 ) −−−−−→ πn+r +1 (An ) (15.4)
Define πr (A) = lim πn+r (An ); hence the homomorphisms of the direct system
n→∞
are those displayed in (15.4). If A is an -spectrum then the homomorphism
πn+r (A) → πn+r +1 (An+1 )
is an isomorphism for n + r ≥ 1; the direct limit is obtained. Hence we have πr (A) =

πn+r (An ) for n + r ≥ 1.
Example 15.10.7 For the Eilenberg–MacLane spectrum A = {K (G, n), αn },

G, if r = 0
πr (A) =
0, if r = 0.
Example 15.10.8 For A = BU -spectrum,

Z, if r is even
πr (A) =
0, if r is odd
by the Bott periodicity theorem (Bott 1959).
Example 15.10.9 For the suspension spectrum S = { An ,

αn },

n X, if n ≥ 0
An =
point, if n < 0.
Hence πr (S) = lim πn+r ( n X ). The limit is attained for n > r + 1. The homotopy
n→∞
groups of the spectrum S are stable homotopy groups of X .
15.11 Applications
This section presents some interesting applications associated with spectra.

μ∧1d i∧1d 1d ∧i
E∧E∧E /E∧E S0 ∧ E /E∧E / E ∧ S0
1d ∧μ μ μ
l r
μ
E∧E /E E
Fig. 15.8 Ring spectrum
15.11.1 Poincaré Duality Theorem
This subsection presents Poincaré duality theorem in the language of spectral homol-
ogy E ∗ (M; E) and spectral cohomology E ∗ (M; E) associated with a ring spectrum
E and also in some other forms. Manifolds generally arise in many problems of
analysis. Historically, homology theory was first applied to manifolds by Poincaré,
giving a key result, known as Poincaré duality theorem. He first stated this theorem
in 1895 in terms of Betti numbers. At that time there was no concept of cohomol-
ogy, which was invented in 1930s. Poincaré stated that the mth and (n − m)th Betti
numbers of a closed (i.e., compact and without boundary) orientable n-manifold are
equal. But there are at present its different forms given below.
Definition 15.11.1 A ring spectrum is a spectrum E with a product μ : E ∧ E → E
(i.e., a map of spectra) and identity i : S 0 → E such that the diagrams in Fig. 15.8
commute up to homotopy, where l and r are natural homotopy equivalences.
The product μ is said to be commutative if the diagram in Fig. 15.9 commutes
up to homotopy.
Definition 15.11.2 An orientation of an n- manifold M at x with respect to a ring
spectrum E is a choice of E 0 (P) module generators of E n (M, M − x). Given a
collection {X } of subsets of M, M is said to be consistently oriented along {X }
with respect to E, if there is a chosen collection of classes [X ] ∈ E n (M, M − x))
satisfying

(i) ρ XX ∩Y ∗ [X ] = ρYX ∩Y ∗ [Y ];
X
(ii) ρ X ∗ [X ] is an orientation of X , where
ρ BA : (M, M − A) → (M, M − B)
is the inclusion.
Fig. 15.9 Commutative T /E∧E

E∧E
multiplication
μ μ
E
Definition 15.11.3 A manifold M is said to be oriented with respect to E, if it can

be consistently oriented along all compact subsets. A collection of such classes is
called an E-orientation.
Example 15.11.4 R P n is orientable iff n is odd but CP n is orientable for each n.
Theorem 15.11.5 (Poincaré duality theorem) Let E = {E n , αn } be a ring spectrum

on C0 . If M is a compact n-manifold oriented with respect to E, then for the coho-
mology and homology associated with E
Ẽ m (M; E) ∼
= Ẽ n−m (M; E).
Proof See Gray (1975). ❑
Remark 15.11.6 If E is the Eilenberg–MacLane spectrum, then applying Poincaré

duality theorem to ordinary homology, for any compact manifold M, it follows that
Hm (M; Z p ) ∼
= H n−m (M; Z p ) and Hm (M; Z p ) is a finite dimensional vector space.
We now give another form Poincaré duality theorem connected with a finite sim-
plicial complex K .
Definition 15.11.7 A path-connected space X is said to be a homology n-manifold
if there exists a triangulation K of X such that for each point x ∈ |K |, and for each
integer m, the homology groups Hm (L K (x)) and Hm (S n−1 ) are isomorphic, where
L K (x) is the link of x in K .
Definition 15.11.8 A homolopy n-manifold X is said to be orientable if there exists

a triangulation K of X , for which the n-simplexes of K can be identified with the
elements of the chain group C n (K ) in such a way that if σ is any (n − 1)-simplex, and
σ1 and σ2 are two n-simplexes that contain σ as a face, then σ occurs with opposite
signs in ∂(σ1 ) and ∂(σ2 ).
Example 15.11.9 (i) A homology 0-manifold (a point) is orientable;

(ii) S 1 is orientable.
Theorem 15.11.10 (Another form Poincaré duality theorem) Let K be a triangula-

tion of a homology n-manifold and K be the first barycentric subdivision of K . If
|K | is orientable, there is an isomorphism
(i) ψ K : H m (K ) → Hn−m (K ) for all integers m;
(ii) There is an isomorphism ψ K : H m (K ; Z2 ) → Hn−m (K ; Z2 ) for all integers m.
Proof See Maunder (1980). ❑
Remark 15.11.11 Poincaré duality theorem is not true for all homology n-manifolds,
unless coefficient group is Z2 is used. Those homology manifolds for which the
theorem is true for coefficient group Z are precisely those that are orientable.
One of the most important applications of cap product defined in Sect. 10.18 of
Chap. 10 is the present form of Poincaré duality theorem.
Theorem 15.11.12 (An alternative form of Poincaré duality theorem) If M is a com-

pact connected oriented n-manifold with generator z ∈ Hn (M; R) for the ordinary
homology with coefficient ring R, then the map
ψ M : H m (M; R) → Hn−m (M; R), u → u ∩ z
is an isomorphism for all integers n.
Proof See Vick (1994). ❑
Remark 15.11.13 H m is a contravariant functor. On the other hand, Hn−m is a covari-

ant functor. The family of isomorphisms
ψ M : H m (M; R) → Hn−m (M; R)
is natural in following sense: if f : M → N is a continuous map between oriented n-

manifolds which are compatible with orientation, then ψ N = f ∗ ◦ ψ M ◦ f ∗ , where f ∗
and f ∗ are homomorphisms induced by f in homology and cohomology respectively.
15.11.2 Homotopy Type of the Eilenberg–MacLane

Space K (G, n)
We have discussed Eilenberg–MacLane spaces K (G, n) in Chap. 11 and proved in

Theorem 11.2.17 that it is uniquely determined by G and n. We prove the same result
in an alternative way.
Theorem 15.11.14 Let G be an abelian group. Then the homotopy type of the
Eilenberg–MacLane space K (G, n) is completely by the group G and the integer n.
Proof Using the result of Ex. 4 of Sect. 15.12, any isomorphism G → G is induced
by a continuous map
f : K (G, n) → K (G, n).
Since all other groups are trivial, the map induces isomorphism in all homotopy
groups. This asserts by Whitehead theorem (Theorem 8.5.9 of Chap. 8) that f is a
homotopy equivalence. ❑
15.11.3 Application of Representability Theorem of Brown
Let h ∗ be a generalized cohomology theory associated with an -spectrum A =

{An , αn }. We now observe that there are the following natural equivalences, at least
if X is connected.
[X, An ] ∼
=h n (X ) ∼
=h n+1 ( X ).
∼
= [ X, An+1 ] ∼= [X, An+1 ].
This natural equivalence is induced by a weak homotopy equivalence
αn : An → An+1 .
This shows that the above sequence of spaces forms a spectrum.

Example 15.11.15 Let H ∗ be ordinary cohomology : H n (X, Y ) ∼ = H n (X, Y ; G).
The corresponding spectrum A is the Eilenberg–MacLane spectrum for the group
G; the nth space is the Eilenberg–MacLane space of type (G, n). That is, we have

G, if r = n
πr (An ) = [S , An ] ∼
n n
= H (S ; G) = r
0, if r = n.
Example 15.11.16 (a) Let K ∗ be complex K -theory. The corresponding spectrum

is called the BU -spectrum. Each even term A2n is the space BU or Z × BU ,
depends on whether we choose to work with connected spaces or not. Each odd
term A2n+1 is the space U .
(b) Let K ∗ be real K -theory. The corresponding spectrum is called the B O-
spectrum. Every eighth term A 8n is the space B O or Z × B O, depends on
whether we chose to work with connected spaces or not. Each term A2n+1 is the
space U .
Remark 15.11.17 All spectra are not -spectra.
Example 15.11.18 Given a CW -complex X , let

n X, if n ≥ 0
An =
point, if n < 0
with the obvious map. Define a spectrum S to be a suspension spectrum or S-spectrum

of X if
ψn : An → An+1
is a weak homotopy equivalence for n sufficiently large. Then this spectrum is called
the ‘suspension spectrum’ S is usually not an -spectrum. In particular, let sphere
spectrum S is the suspension spectrum of S 0 ; it has nth term S n for n ≥ 0.
15.11.4 More Applications of Spectra
This subsection presents more connections of homotopy theory with cohomology

theory through spectra.
Theorem 15.11.19 Let E = {E n , αn } be a spectrum and X be a CW-complex. Then
m (X ) ∼
there exists a natural isomorphism E = [X, E m ].
Proof Consider the diagram in Fig. 15.10 where αn : E n → E n+1 is the adjoint to

αn : E n → E n+1 and ψ is a bijective correspondence. Hence
m (X ) = [ n−m X, E n ] ∼
E = [X, n−m E n ] ∼
= [X, E m ],
as and are adjoint functors (see Chap. 2). ❑

Remark 15.11.20 Theorem 15.11.19 interlinks homotopy theory with cohomology
theory.
m (S k ) ∼
Theorem 15.11.21 For any spectrum E, E =Ek (S m ).
m (S ◦ ) ∼
Proof By property RH(ii), it is sufficient to show that E =E−m (S ◦ ). Consider
the direct limit of the sequence as shown in Fig. 15.11:
Hence the theorem follows ❑
Definition 15.11.22 The direct limit in Theorem 15.11.21 is called the mth homo-
topy group of the spectrum E and is sometimes abbreviated πm (E). It is also called
∗ on C0 .
∗ and E
the group of coefficients for the theories E
Definition 15.11.23 Given an -spectrum E = (E n , αn ) and a CW -complex pair
(X, A), the cohomology groups of (X, A) associated with E are defined by E n (X,
A; E) = [X/A, E n ] with multiplication induced by αn .
We write E ∗ (X, A; E) = ⊕E m (X, A; E). The corresponding reduced groups are
; E) = E n (X, x0 ; E), where x 0
given by A = ∅; E n (X ; E) = E n (X, ∅; E) and E(X
is the base point(assumed to be a 0-cell).
Fig. 15.10 Diagram α

n
[Σn−m X, En ] / [Σn−m+1 X, E
n+1 ]
associated with a spectrum E
αn ψ
[Σn−m X, ΩEn+1 ]
Fig. 15.11 Triangular · · · → πn+m (En ) / πn+m+1 (En+1 ) → · · ·

diagram for groups and
homomorphism
E (
αn )∗
πn+m+1 (ΣEn )
Remark 15.11.24 E n (X ; E) = [X + , E n ], the set of unbased homotopy classes of

n (X ; E) = [X ; E n ].
maps X into E n and E
n (X ; E))
n (X, A; E)(and hence also E n (X ; E), E
Theorem 15.11.25 The groups E
are all abelian.
Proof E n (X, A; E) ∼
= [X/A, E n+1 ] ∼
= [X/A, (E n+2 )], which is an abelian
group. ❑
The exactness of the cohomology sequence proves the following:
Theorem 15.11.26 Let E = {E n , αn } be an -spectrum on C0 . Then
n (X ; E) ∼
E =En (X ; E) ⊕ E
n (S ◦ ; E),
where X is a CW-complex.
Proof Let x 0 be the base point of X . Now consider the exact cohomology sequence
of the pointed space (X, x 0 ):
j∗ i∗
n (x0 ; E) −
→ E n (X, x0 ; E) −−−→ E n (X ; E) −−−→ E
··· − → ···
The above sequence splits, because, there is a map p : X → x0 such that p ◦ i =

1d : x0 → x0 . Hence i ∗ ◦ p∗ = 1d . Consequently,
E n (X ; E) ∼
= E n (X, x 0 ; E) ⊕ E n (x0 , E).
But
n (X ; E) and E n (x0 ; E) = E
E n (X, x0 ; E) = E n (S 0 ; E).
15.11.5 Homotopical Description of Singular

Cohomology Theory
Eilenberg–MacLane spaces are used in giving homotopical description of singular

cohomology theory. The Eilenberg–MacLane spaces K (Z, n) fit together via the
loop functor

K (Z, 0) ←−−− K (Z, 1) ←−−− K Z, 2) ←−−− · · ·
to form a spectrum, called the Eilenberg–MacLane spectrum. Given a connected

CW -complex X define cohomology theory by setting
H n (X ; Z) = [X, K (Z, n)].
In general, given a CW -complex X , there is a bijective correspondence between

the cohomology group H n (X ; G) and the homotopy classes of maps X to K (G, n).
Using this fact it can be shown that the cohomology operations are completely clas-
sified by the cohomology groups of K (G, n)’s.
15.12 Exercises
1. Let X = ∪X α = lim X α has the weak topology and assume

−
→
(i) for all α, β ∈ A, there exists δ ∈ A such that X α ∩ X β = X δ ;
(ii) for all α ∈ A, {β ∈ A : β ≤ α} is finite (ordering β ≤ α is defined by β ≤ α
iff X β ⊂ X α ). Then show that Em (lim X α ) ∼ m (X α ).
= lim E
−
→ −
→
2. Show that for a spectrum E, the homology and cohomology groups satisfy the
following properties:
m (X 1 ∨ · · · ∨ X n ) ∼
(i)
E =Em (X 1 ) ⊕ · · · ⊕ E
m (X n );
∼
E (X 1 ∨ · · · ∨ X n ) = E (X 1 ) ⊕ · · · ⊕ E
(ii)m m m (X n ).

3. Let (X α , ∗α ) ∈ C0 be indexed by a set A and X α denote the quotient space

α∈A
X α /∗α ∼ ∗β . If the set A is finite, this is the one-point union X α1 ∨ · · · ∨ X αn .
α∈A
Show that

m (
(i) E X α) ∼ m (X α );
= ⊕E
α∈A
(ii) if (X α , ∗α ) are CW -complexes, and E = {E n ,
αn } is an -spectrum, them

m (
E X α∈A ) ∼ m (X α ).
= α∈A E
α∈A
(This result is sometimes called the wedge axiom.)

4. Given abelian groups G and T , show that there is a bijection
ψ : [K (G, n), K (T, n)] → H om(G, T ).

15.12 Exercises 507
5. Given an abelian group G, show that there exists an -spectrum E such that

K (G, n), n ≥ 0
En =
point, otherwise.
6. Let H∗ ( ; G) denote the singular homology with coefficients in the abelian

G and {G r }r ∈ Z be a sequence of abelian groups. Show that h n (X, A) =
group
Hm (X, A; G r ) defines a homology theory with coefficient group
m+r =n
h r (point)=G r for all r .
7. Let
f g
0−
→ G −−−→ H −−−→ K −
→0
be a short exact sequence of abelian groups. Construct natural long exact

sequences
··· −
→Em (X ; G) −−−
c
→E m (X, K ) −−α−→ E
m (X ; H ) −−d−→ E m+1 (X ; G) −
→ ··· ,
··· −
→Em (X ; G) −−−
c
→E m (X ; K ) −−α−→ E
m (X ; H ) −−d−→ E m−1 (X ; G) −
→ ···
This is called the Bockstein sequence and α is called the Bockstein homomor-
phism.
8. A spectrum E is called properly convergent if αn : E n → E n+1 is a (2n + 1)-
isomorphism for each n. Show that
(i) if E n is a pointed topological space for each n and E is properly convergent,
then E n is (n − 1)-connected;
(ii) if E is an -spectrum and E n is connected for each n, then E is properly
convergent.
9. If X is a CW -complex, show that the reduced ordinary homology and cohomol-
ogy groups H m (X ; G) and H
m (X ; G) with coefficient groups G are trivial for
m < 0.
[Hint: X ∧ K (G, n) is a CW -complex with all cells in dimension n and larger,
except for 0-cells. Hence πm+n (X ∧ K (G, n)) = 0 for m < 0 and n > 1. This
shows that H m (X ; G) = 0 for m < 0. Since n−m X has all cells in dimension
n − m and larger, except for 0-cells, [ n−m X, K (G, n)] = 0, if m < 0.]
10. Show that the sphere spectrum S is a ring spectrum and every spectrum E is a
module over S.
∗ (S 0 ) is a graded commutative ring with unit
11. If E is a ring spectrum, show that E
∗ ∗ 0
and E (X ) is a module over E (S ) for every X ∈ C0 .

[2] Adams, J.F., Stable Homotopy and Generalized Homology, University of
Chicago, 1974.
[4] Bott, R., The stable homotopy of the classical groups, Ann. Math. Second Series
70: 313–337, 1959.
York, Heidelberg, Berlin, 1978.
[7] Dold, A., Lectures on Algebraic Topology, Springer-Verlag, New York, 1972.
[8] Dold, A., Relations between ordinary and extraordinary homology, Algebraic
Topology Colloquium Aarhus, 2–9, 1962.
1975.
Cambridge, 1983.
[11] Hilton, P. J. and Wylie, 1945, 1942 S. Homology Theory, Cambridge University
[12] Huber, P. J., Homotopy theory in general categories, Math. Ann. 144, 361–385,
1961.
[13] Mayer, J., Algebraic Topology, Prentice-Hall, New Jersey, 1972.
[15] Mitra, Shibopriya, A study of some notions of algebraic topology through homo-
topy theory, PhD Thesis, University of Calcutta, 2007.
[16] Mosher, R. and Tangora, M.C., Cohomology Operations and Applications in
Homotopy Theory, Harper and Row, New York, 1968.
[17] Steenrod, N.E. and Epstein, D. B. A., Cohomology operations, Princeton Uni-
versity Press, Princeton, 1962.
References 509
References
(1972)
Adams, J.F.: Stable Homotopy and Generalized Homology. University of Chicago, Chicago (1974)
Brown, E.H.: Cohomology theories. Ann. Math. 75, 467–484 (1962)
Bott, R.: The stable homotopy of the classical groups. Ann. Math. Second Series 70, 313–337 (1959)
Dieudonné, J.: A History of Algebraic and Differential Topology, 1900-1960. Modern Birkhäuser,
Boston (1989)
Dold, A.: Relations between ordinary and extraordinary homology. Algebraic Topology Collo-
quium, Aarhus, pp. 2–9 (1962)
Dold, A., Thom, R.: Quasifaserungen und unendliche symmetrische Produkte. Ann. Math. 67(2),
239–281 (1958)
ton (1952)
(1975)
Huber, P.J.: Homotopy theory in general categories. Math. Ann. 144, 361–385 (1961)
James, I.M. (ed.): Handbook of Algebraic Topology. North Holland, Amsterdam (1995)
Luke, G., Mishchenko, A.S.: Vector Bundles and Their Applications. Kluwer Academic Publishers,
Boston (1998)
Maunder, C.R.F.: Algebraic Topology. Cambridge University Press, Cambridge (1980)
Mayer, J.: Algebraic Topology. Prentice-Hall, New Jersey (1972)
Mitra, S.: A study of some notions of algebraic topology through homotopy theory. Ph.D. thesis,
University of Calcutta (2007)
Mosher, R., Tangora, M.C.: Cohomology Operations and Applications in Homotopy Theory. Harper
and Row, New York (1968)
Steenrod, N.E., Epstein, D.B.A.: Cohomology Operations. Princeton University Press, Princeton
(1962)
Vick, J.W.: Homology Theory: Introduction to Algebraic Topology. Springer, New York (1994)
Whitehead, G.W.: Generalized homology theories. Trans. Am. Math. Soc. 102, 227–283 (1962)
Chapter 16
Obstruction Theory
This chapter studies a theory known as “Obstruction Theory” by applying cohomol-

ogy theory to encounter two basic problems in algebraic topology such as extension
and lifting problems. Obvious examples are the homotopy extension and homotopy
lifting problems. The homotopy classifications of continuous maps together with the
study of extension and lifting problems, play a central role in algebraic topology.
Obstruction theory leads to make an attempt to find a general solution. This theory
was originated in the classical work of H. Hopf (1894–1971), S. Eilenberg (1915–
1998), N. Steenrod (1910–1971) and M. Postnikov (1927–2004) around 1940. The
term “obstruction theory” refers to a technique for defining a sequence of cohomol-
ogy classes that are obstructions to finding solution to the extension, lifting, or relative
lifting problems. More precisely, this chapter studies certain sets of cohomology ele-
ments, called obstructions which are associated with both a single map in the case
of extension and with a pair of maps in the case of homotopies. These are invariants
depending only on the spaces and mappings. In polyhedra, these are the characteris-
tics for the existence or nonexistence of the desired extensions and homotopies. The
underlying idea of associating cohomology elements with mappings was implicitly
used by H. Whitney (1907–1989) and first explicitly formulated by S. Eilenberg. Let
X be a compact triangulable space and f : X → X be a continuous map. Then f
has a fixed point if the Lefschetz number f of f is a nonzero integer. This implies
that f is the “obstruction” to f being fixed point free. Such an example displays
the basic objective of obstruction theory.
Extension problems play a central role in topology. Most of the basic theorems
of topology together with their successful applications in other areas in mathemat-
ics are solutions of particular extension problems. The most successful results have
been obtained using the tools of algebraic topology, which offers a conversion of
the geometric problem into an algebraic problem. The homotopy classification prob-
lem is closely related to extension and lifting problems. Many extension and lifting
problems are still unsolved.

DOI 10.1007/978-81-322-2843-1_16
512 16 Obstruction Theory
N.E. Steenrod wrote in his excellent paper published in Steenrod (1972):

“Many of the basic theorems of topology, and some of its most successful appli-
cations in other areas of mathematics, are solutions of particular extension problems.
The deepest results of this kind have been obtained by the method of algebraic
topology. The essence of the method is a conversion of a geometric problem into
an algebraic problem which is sufficiently complex to embody the essential features
of the geometric problem, yet sufficiently simple to be solvable by standard alge-
braic methods. Many extension problems remain unsolved, and much of the current
development of algebraic topology is inspired by the hope of finding a truly general
solution”.
For this chapter the books and papers Arkowitz (2011), Davis and Kirk (2001),
Dodson and Parkar (1997), Eilenberg and Steenrod (1952), Gray (1975), Hatcher
(2002), Hu (1959), Maunder (1970), Spanier (1966), Steenrod (1951, 1972) and
some others are referred in Bibliography.
16.1 Basic Aim of Obstruction Theory
This section conveys the aim of obstruction theory and describes a technique for
studying various homotopy problems such as extension problems, lifting problems,
and relative lifting problems which are basic problems in algebraic topology. To earn
the basic objective of obstruction theory we start with a simple example: given a
group homomorphism f , ker f is an algebraic indicator which is an obstacle to f
for being injective. We normally use homotopy theory to yield algebraic indicators
for obstacles to extension and lifting problems of continuous maps. For example,
a continuous map f : S n → X has a continuous extension over the (n + 1)-ball
D n+1 bounded by S n iff f is nullhomotopic (see Theorem 2.10.1 of Chap. 2). Hence
in this case the obstacle for extension of f is precisely [ f ] = 0(∈ πn (X )). Again
the obstacle to lifting problems in a principal fibration is a constant map. Hence
this problem can be expressed as the homotopy class of the map into the classifying
space.
There are several techniques to develop obstruction theory to extension problems
using the tools of algebraic theory. Obstructions are built step by step using the tools
of cohomology theory. The most useful technique is to associate certain sets of coho-
mology elements with a single map in case of extension, and with a pair of maps in
case of homotopies. These cohomology elements are called obstructions. This idea
of associating cohomolgy elements with mappings was first found implicitly in the
work of Whitney and explicitly in the work of S. Eilenberg. The latter theory is tra-
ditionally called Eilenberg obstruction theory in his honor. It involves cohomology
groups with coefficients in certain homotopy groups to define obstructions to exten-
sion problems. But there are several cohomology theories such as cellular, singular,
simplicial, Čech cohomology, etc. The uniqueness theorem of cohomology theory
asserts that any two cohomology groups having the same coefficient group coin-
cide on finite C W -complexes, which implies that the cohomology groups of finite
16.1 Basic Aim of Obstruction Theory 513
CW-complexes are completely determined by the coefficient group. We study here

obstruction theory for C W -complex (see (Hu 1959)) but Olum studied this theory
for polyhedra (see (Olum 1950)) in almost identical techniques in view of the fact
that polyhedron admits a C W structure.
Recall the concepts of cochain and cochain group which are applied in obstruction
theory:
Definition 16.1.1 Given a topological space X and an abelian group G, the singular
n-cochain group C n (X ; G) with coefficients in G is defined to be the dual group
given by C n (X ; G) = Hom (Cn (X ; G), G) of the singular chain group Cn (X ; G).
An element of C n (X ; G) is called a cochain.
Remark 16.1.2 An n-cochain α ∈ C n (X ; G) assigns to each n-simplex σ : n → X

a value α(σ) ∈ G. Since the singular n-simplexes form a basis of C n (X ; G), these
values can be assigned arbitrarily. Hence n-cochains are precisely the functions from
singular n-simplexes to G. Again C n (X ; G) is isomorphic to the direct product of
as many copies of G as there are n-simplexes in X .
The basic aim of obstruction theory is to study mainly the following four types
of problems:
(i) The extension problem;
(ii) The lifting problem;
(iii) The relative lifting problem;
(iv) Cross section problem.
16.1.1 The Extension Problem
This subsection explains the extension problem. Given a CW-complex pair (X, A)
with inclusion map i : A → X , and a continuous map f : A → E, does there exist
a continuous map h : X → E (represented by dotted arrow) such that the triangle in
Fig. 16.1 is commutative? This is called an extension problem. If such h exists, then
h is called an extension of f . For understanding the technique of obstruction theory
in extension problem, we first consider the Example 16.1.3.
Example 16.1.3 Let X be a finite CW-complex. If we need construction of a contin-

uous function on X , we use induction: if the function is defined on r -skeleton X (r ) ,
we attempt to extend it over the (r + 1)-skeleton X (r +1) . Then the obstruction to
extending it over an (r + 1)-cell is an element of πr .
Fig. 16.1 Representing

extension problem

lifting problem
16.1.2 The Lifting Problem
This subsection explains the lifting problem: given a fibration p : E → B and

a continuous map g : X → B, does there exist a continuous map h : X → E
(represented by dotted arrow) called a lift or lifting of g, such that the triangle in
Fig. 16.2 is commutative? This problem is called a lifting problem. If such h exists,
then h is called a lift or lifting of g. Thus an extension problem is the question of
finding a criterian to make the triangle in Fig. 16.2 commutative.
Example 16.1.4 (Path lifting Property) Let f : (I, I˙) → (S 1 , 1) be continuous.

Then there exists a unique continuous map f˜ : I → R with p f˜ = f and f˜(0) = 0.
Hence f˜ is the unique lifting of f (see Chap. 3).
16.1.3 Relative Lifting Problem
This subsection explains the relative lifting problem which combines both the exten-
sion and lifting problems into a single problem described in the diagram in Fig. 16.3
with commutative square, called the extension-lifting square. In other words, a rel-
ative lifting problem is the question of finding a criterian to make the diagram in
Fig. 16.3 commutative.
Thus given a C W -complex pair (X, A) if i : A → X is the inclusion map, then
the problem is to determine a continuous map h; X → E (it exists) such that hi = f
and ph = g. This h (if it exists) is called a solution of the extension-lifting problem,
called a relative lift. If we take in particular, B = {∗}, we obtain the extension
problem, and for A = {∗} we obtain the lifting problem.
Remark 16.1.5 In most cases, these obstructions are in the cohomology groups that
are all zero, which gives a solution. On the contrary, if the obstructions are nonzero,
it can be used to encounter the problem expressed in cohomology terms.

relative lifting problem
16.1 Basic Aim of Obstruction Theory 515
16.1.4 Cross Section Problem
It is a particular case of the relative lifting problem when X = B and f : X → B is

the identity map on B in the relative lifting case. Again the relative lifting problem
reduces to determining a cross section of the induced bundle f ∗ ξ of the bundle
ξ : p : E → B under a continuous map f : X → B. For step-by-step extension of
cross section of fiber bundle see Sect. 16.4.2.
Remark 16.1.6 The above observation asserts that the relative lifting and cross
section problems are equivalent.
16.2 Notations and Abbreviations
In obstruction theory, for simplicity we consider only finite cell complexes X of

dimension n and continuous maps from X into a path-connected n-simple space Y .
The standard notations and abbreviations used in obstruction theory are given with
their meanings.
X : finite CW-complex
A : subcomplex of X
X (n) : n-skeleton of X

Xn : A ∪ X (n)
Y : path-connected n-simple space
πn (Y ) : nth homotopy group of Y
Cn+1 (X, A; πn (Y )) : (n + 1)th chain group of X modulo A with coefficient group πn (Y )
C n+1 (X, A; πn (Y )) : (n + 1)th cochain group of X modulo A with coefficient group πn (Y )
cn+1 ( f ) : obstruction of cochain of f in C n+1 (X, A; πn (Y ))
∂ : boundary homomorphism
δ : coboundary homomorphism
fg : f ◦ g (composite map)
∂∂ : ∂ ◦ ∂ = 0 (0-homomorphism)
Z n (X, A; πn (Y )) : group of n-cycles of X modulo A with coefficient group πn (Y )
B n (X, A; πn (Y )) : group of n-boundaries of X modulo A with coefficient group πn (Y )
H n (X, A; πn (Y )) : nth cohomology group of X modulo A with coefficient group πn (Y )
[cn+1 ( f )] : γ n+1 ( f ) (cohomology class of cn+1 ( f )) in H n+1 (X, A; πn (Y ))
ψ# : cochain homomorphism induced by ψ
ψ∗ : homomorphism induced by ψ in cohomology
σ : cell of the CW-complex X
|K | : underlying space of a simplicial complex K
ξ : fiber bundle.
16.3 The Obstruction Theory: Basic Concepts
This section presents the basic concepts of obstruction theory and uses the notations
of Sect. 16.2. Before conveying the basic concepts of ‘obstruction theory’ we recall
that a path-connected topological space is said to n-simple if there is point x0 ∈ X
such that π1 (X, x0 ) acts trivially on πn (X, x 0 ) in the sense that each element of
π1 (X, x0 ) acts on πn (X, x 0 ) as the identity. For example, every H -space is n-simple
for all n. If X is n-simple, then the homoptopy groups πn (X, x 0 ) do not depend on
its base point. Hence if X is n-simple, then the groups πn (X, x0 ) is simply denoted
by πn (X ).
This section considers extension problem: Given a subcomplex A of a finite cell
complex X , a path-connected n-simple space Y and a continuous map f : A → Y ,
the extension problem for f over the whole X is to determine whether or not f
can be continuously extended over X . In obstruction theory, an attempt is made to
extend the given map f step by step over the subcomplexes X̃ n , n = 0, 1, 2, · · · . of
X , where X̃ n = A ∪ X (n) . This process is continued till some obstruction for further
extension is met. Then the traditional technique is to measure this obstruction and
to change the previously constructed partial extension of f so that this obstruction
vanishes and hence further extension of f might be possible.
Definition 16.3.1 Given an integer n ≥ 0, a subcomplex A of a finite cell complex
X , and a path-connected n-simple space Y , a continuous map f : A → Y is said
to be n-extensible over X if f has a continuous extension over the subcomplex X̃ n
of X .
Example 16.3.2 Every continuous map f : X → Y is 1-extendable.
Definition 16.3.3 The supremum of n for which f is n-extensible is called the
extension index of f over X .
Proposition 16.3.4 Homotopic maps have the same extension index.
Proof Let f, g : A → Y be two maps such that f g. Let f˜ be an extension of f .
Define g̃ which coincides with f˜ on X n − A and coincides with g on A. Then there
is a homotopy between g̃ and f˜, so there is a homotopy between g and f on X n − A.
Hence the index of g ≤ index of f . Similarly, index of f ≤ index of g. This asserts
that index of f = index of g. ❑
Corollary 16.3.5 The extension index of a continuous map is a homotopy invariant
in the sense that homotopic maps have the same extension index.
Proof The corollary follows from Proposition 16.3.4. ❑
16.3.1 The Obstruction Cochains and Cocycles
This subsection conveys the concepts of obstruction cochain and cocycle. It is impor-
tant that cycles and coboundaries are primary concepts of a cohomology group. These
16.3 The Obstruction Theory: Basic Concepts 517
are linked to an obstruction to extension problems. Consider a given continuous map

f : X̃ n → Y . Then f determines a cochain cn+1 ( f ) in C n+1 (X, A; πn (Y )) as fol-
lows.
Let σ be an arbitrary (n +1)-cell of X . Then σ is in X (n+1) − X (n) and its boundary
∂σ is an oriented n-sphere. As ∂σ ⊂ X (n) ⊂ X̃ n , the restriction map f σ of f over
∂σ determines an element [ f σ ] ∈ πn (Y ). Define a map
cn+1 ( f ) : X (n+1) − X (n) → πn (Y ), σ → [ f σ ].
Definition 16.3.6 The cochain cn+1 ( f ) : X (n+1) − X (n) → πn (Y ), σ → [ f σ ]. is

called the obstruction cochain of f : X̃ n → Y .
We claim that cn+1 ( f ) is a cocycle i.e., it vanishes on boundaries.
Theorem 16.3.7 Let X be a finite C W -complex and A be a subcomplex of X . If Y
is a path-connected n-simple space, then given a continuous map f : X̃ n → Y , its
cochain cn+1 ( f ) is a cocycle of X modulo A.
Proof To prove the theorem it is sufficient to prove that cn+1 ( f ) ∈ Z n+1 (X, A;
πn (Y )). Let σ be an arbitrary (n + 2)-cell of X . Then ∂σ ∈ X̃ n+1 . We claim that
[δcn+1 ( f )](σ) = 0. Let B be the subcomplex ∂σ of X and B (n) be the n-skeleton of
B. Then we have the following results:
(i) there is a homomorphism ∂ : Cn+1 (B; πn (Y )) → Z n (B : πn (Y )) = Z n (B (n) :
πn (Y )), since there is no n-cycle in the (n + 1)-skeleton of B;
(ii) there is an isomorphism g : Z n (B (n) ; πn (Y )) → Hn (B (n) ; πn (Y )) by quotient
group definition of homology group;
(iii) there is a natural homomorphism h : πn (B (n) ) → Hn (B (n) ; πn (Y )) by Hurewicz
theorem;
(iv) there is a homomorphism k∗ : πn (B (n) ) → πn (Y ) induced by k = f | B (n) .
Combining the above homomorphisms, we have the homomorphisms
∂
Cn+1 (B; πn (Y )) −−−→ Z n (B, πn (Y )) = Z n (B (n) , πn (Y ))
h k∗
= Hn (B (n) ; πn (Y )) ←−−− πn (B (n) ) −−−−→ πn (Y )
If n > 1 then B (n) is (n − 1)-connected and hence h is an isomorphism by

Hurewicz theorem. Again if n = 1, h is an epimorphism and ker h is contained in
ker k∗ , since the group πn (Y )) is abelian. Consequently, in either case, there is a well
defined homomorphism
ψ : k∗ h −1 : Z n (B; πn (Y )) → πn (Y ).
Since Cn−1 (B, πn (Y )) is a free abelian group, the kernel Z n (B; πn (Y )) of ∂ :

Cn (B, πn (Y )) → Cn−1 (B, πn (Y )) can be expressed as a direct summand of
Cn (B, πn (Y )). This asserts that the homomorphism
ψ : Z n (B; πn (Y )) → πn (Y )
has an extension
ψ̃ : Cn (B; πn (Y )) → πn (Y ).
Again for every (n + 1)-cell σ in B, the element [cn+1 ( f )](σ ) is represented by

the partial map k|∂σ . Hence it follows that [cn+1 ( f )](σ ) = k∗ h −1 (∂σ ) = ψ̃(∂σ ).
This asserts that [δcn+1 ( f )](σ) = [cn+1 ( f )](∂σ) = ψ̃(∂∂(σ )) = 0. This implies
that cn+1 ( f ) ∈ cn+1 ( f ) ∈ Z n+1 (X, A; πn (Y )). ❑
Remark 16.3.8 Since cn+1 ( f ) ∈ Z n+1 (X, A; πn (Y )), cn+1 ( f ) is called the obstruc-
tion cocyle of f . The reason for naming obstruction cocycle is given in Proposition
16.3.9.
We assume that somehow an extension of a map over X̃ n = A ∪ X n has been
achieved for some n. We now consider its extension problem over X̃ n+1 .
Proposition 16.3.9 A continuous map f : X̃ n → Y has a continuous extension
over X̃ n+1 iff cn+1 ( f ) = 0.
Proof Let σ be an arbitrary (n + 2)-cell of X and f˜ : X̃ n+1 → Y be a given contin-
uous extension of f . Then cn+1 ( f ) = cn+1 ( f˜|∂σ) = [ f˜|∂(∂σ)] = 0. Conversely,
suppose cn+1 ( f ) = 0. Then f is a coboundary on X . Hence there is a continuous
map f : X̃ n+1 → Y such that f = δ f˜ = f˜δ. This asserts that for any (n + 1)-cell
σ of X , f (σ) = f˜(∂σ). This shows that f˜ agrees on its boundary. Consequently, f˜
is an extension of f . ❑
Example 16.3.10 Let M be the Möbius band and f be the continuous map which
defines the complex structure of M. If σ1 is the boundary circle and σ2 is the interior,
then [c1 ( f )](σ1 ) = 0 shows that the extension of all 1-cells is admitted. The second
obstruction [c2 ( f )](σ2 ) = [ f |∂σ2 ] = [ f | S 1 ]. It shows that there is an obstruction to
the attaching discs to M. This result shows that the Möbius band M is non-orientable.
Proposition 16.3.11 Let f, g : X̃ n → Y be two homotopic maps. Then cn+1 ( f ) =
cn+1 (g).
Proof Let σ be an (n + 1)-cell of X and f, g : X̃ n → Y be two homotopic maps.
Then f |∂σ g|∂σ. Hence it shows that [ f |∂σ] = [g|∂σ] in πn (Y ). This implies
that cn+1 ( f ) = cn+1 (g). ❑
Remark 16.3.12 The obstruction cocycle cn+1 ( f ) of f : X̃ n → Y is a homotopy
invariant in the sense that if f, g : X̃ n → Y are homotopic maps, then cn+1 ( f ) =
cn+1 (g).
Definition 16.3.13 Given a continuous map f : X̃ n → Y , its obstruction cocy-
cle cn+1 ( f ) determines the cohomolgy class [cn+1 ( f )] ∈ H n+1 (X, A; πn (Y )). This
cohomology class [cn+1 ( f )] is abbreviated as γ n+1 ( f ) and cn+1 ( f ) is said to repre-
sent it.
16.3.2 The Deformation and Difference Cochains
This subsection conveys the concepts of deformation and difference cochains and
considers the problem of constructing homotopies between two given continuous
maps f, g : X̃ n → Y which are assumed to be homotopic on X̃ n−1 (called partial
homotopic on X̃ n−1 ). We claim that the difference of the obstruction cocycles cn+1 ( f )
and cn+1 (g) is a coboundary.
First we consider a partial homoptopy Ht : X̃ n−1 → Y such that H0 = f | X̃ n−1
and H1 = g| X̃ n−1 . Again the topological product P = X × I is also a cell complex.
Let P (n) be the n-dimensional skeleton of P. As usual notation
(n) = (A × I ) ∪ P (n)
P
= (
X n × {0}) ∪ (
X n−1 × I ) ∪ (
X n × {1}).
Define a map
⎧
⎨ f (x), x ∈
⎪ Xn, t = 0

h : P → Y, (x, t) → Ht (x), x ∈ X n−1 , t ∈ I
n
⎪
⎩
g(x), x ∈ Xn, t = 1
Then h is continuous and determines an obstruction cocycle cn+1 (h) of the complex
P modulo A × I with coefficient group πn (Y ). Then it follows from the definition of
h that cn+1 (h) agrees with cn+1 ( f ) × 0 on X × {0} and with cn+1 (g) × 1 on X × {1}.
Let B be the subcomplex (X × {0}) ∪ (A × I ) ∪ (X × {1}) of X × I = P.
Hence cn+1 (h) − cn+1 ( f ) × 0 − cn+1 (g) × 1 is a cochain of X × I modulo B
with coefficient group πn (Y ). Again the map σ → σ × I establishes a bijective
correspondence between the n-cells of X − A and (n + 1)-cells of P − B. This
correspondence defines an isomorphism
ψ : C n+1 (X, A; πn (Y )) → C n+1 (P, B; πn (Y )).
This determines a unique cochain, denoted by d n ( f, g; Ht ) in C n (X, A; πn (Y )) such

that
ψd n ( f, g; Ht ) = (−1)n+1 (cn+1 (h) − cn+1 ( f ) × 0 − cn+1 (g) × 1).
Definition 16.3.14 The unique cochain d n ( f, g : Ht ) ∈ C n (X, A : πn (Y )) is called

the deformation cochain of f and g. In particular, if f | X̃ n−1 = g| X̃ n−1 and Ht (x) =
f (x) = g(x) for all x ∈ X̃ n−1 and for all t ∈ I , then d n ( f, g : Ht ) is abbreviated in
brief as d n ( f, g), and is called the difference cochain of f and g.
Remark 16.3.15 As d n ( f, g) = cn+1 ( f ) − cn+1 (g), d n ( f, g) is called the difference

cochain of f and g.
Proposition 16.3.16 Given two continuous maps f, g : X̃ n → Y , and a partial

homotopy Ht : X̃ n−1 → Y has a continuous extension H̃t : X̃ n → Y over X̃ n with
the property that H̃0 = f and H̃1 = g iff d n ( f, g : Ht ) = 0.
Proof It needs extensions of f and g to have an extension of Ht over X̃ n+1 . This

is possible iff cn+1 ( f ) = 0 = cn+1 (g), and cn+1 (Ht ) = 0. Then ψd n ( f, g; Ht ) =
(−1)n (0 − 0 + 0) = 0. Again since ψ is an isomorphism, the only element of ker ψ
is the zero element. This implies that Ht has a continuous extension over X̃ n+1 iff
d n ( f, g : Ht ) = 0. ❑
Remark 16.3.17 d n ( f, g : Ht ) plays a key role in the study of obstruction theory

because of its coboundary formula Proposition 16.3.18.
Proposition 16.3.18 (Coboundary formula) δd n ( f, g : Ht ) = cn+1 ( f ) − cn+1 (g).
Proof As δ I = 0, isomorphism ψ commutes with δ, and hence it follows that
ψδd n ( f, g : Ht ) = δψd n ( f, g : Ht ) (16.1)
Again, as cn+1 (h), cn+1 ( f ), cn+1 (g) are all cocycles and δ0 = −I and δ1 = I ,
applying δ to both sides of (16.1) we have δψδd n ( f, g : Ht ) = cn+1 ( f ) × I −
cn+1 (g) × I . This shows that ψδd n ( f, g : Ht ) = ψ(cn+1 ( f ) − cn+1 (g)). Since ψ is
an isomorphism, the proposition follows. ❑
16.3.3 The Eilenberg Extension Theorem
This subsection proves Eilenberg extension theorem which is a key result in obstruc-
tion theory when stepwise extension process faces an obstruction. Recall that given a
continuous map f : X̃ n → Y , its obstruction cocycle cn+1 ( f ) determines an element
γ n+1 ( f ) ∈ H n+1 (X, A; πn (Y )).
Theorem 16.3.19 (Eilenberg extension theorem) Given a continuous map f :

X̃ n → Y , the element γ n+1 ( f ) ∈ H n+1 (X, A; πn (Y )) vanishes iff there exists a
continuous map h̃ : X n+1 → Y such that h̃| X̃ n−1 = f | X̃ n−1 .
Proof First suppose that element γ n+1 ( f ) = 0. If cn+1 ( f ) ∈ C n+1 (X, A : πn (Y ))

is a representative of γ n+1 ( f ), then cn+1 ( f ) is homotopic to a constant map rel A.
Hence there exists a continuous map h : X̃ n → πn (Y ) such that f | X̃ n−1 = h| X̃ n−1
and cn+1 (h) = 0. This implies that h has a continuous extension h̃ : X̃ n+1 → Y
by Proposition 16.3.9. Conversely, assume that there is a continuous extension
h̃ : X̃ n+1 → Y of h = h̃| X̃ n . Then cn+1 (h) = 0. Again since f | X̃ n−1 = h| X̃ n−1 , the
difference cochain d n ( f, h) is defined. Hence by Proposition 16.3.18, cn+1 (h) = 0.
It implies that cn+1 ( f ) is the coboundary of d n ( f, h). This shows that γ n+1
( f ) = 0. ❑
Remark 16.3.20 conveys the significance of Eilenberg extension theorem.

Remark 16.3.20 Assume that f˜ is a continuous extension of a given map f : A →
Y . If the obstruction cocycle cn+1 ( f ) = 0, then it follows from Proposition 16.3.9 that
f˜ cannot be extended over X̃ n+1 . This gives an obstruction in stepwise extending
process. The importance of the Eilenberg extension theorem is that if cn+1 ( f ) is
nullhomotopic rel A, then the obstruction can be removed by modifying the values
of f on the open n-cells in X − A only.
16.3.4 The Obstruction Set for Extension
This subsection conveys the concept of ‘obstruction set’ which plays a key role in
extensibility of a map f : A → Y over the whole space X .
Definition 16.3.21 Given a finite C W -complex pair and a path-connected n-simple
space, a continuous map f : X̃ n → Y determines a cocycle cn+1 ( f ) up to homotopy
and hence determines an element γ n+1 ( f ) ∈ H n+1 (X, A; πn (Y )), called an (n + 1)-
dimensional obstruction element of f .
Definition 16.3.22 The set of all (n + 1)-dimensional obstruction elements of f :

X̃ n → Y forms a subset of H n+1 (X, A; πn (Y )), called an obstruction set of f ,
denoted by On+1 ( f ).
Definition 16.3.23 A continuous map f : A → Y is said to be n-extensible over X

if there exists a continuous extension f˜ : X̃ n → Y of f .
Proposition 16.3.24 If f, g : X̃ n → Y are two homotopic maps, then On+1 ( f ) =

On+1 (g).
Proof Let f, g : X̃ n → Y be two maps such that f g. We claim that On+1 ( f ) =

On+1 (g). As f g, cn+1 ( f ) = cn+1 (g) by Proposition 16.3.11. Hence they define
the same equivalence class in H n+1 (X, A; πn (Y )). This implies that On+1 ( f ) =
On+1 (g). ❑
Recall that if (X, A) and (X , A ) be two cellular pairs and ψ : (X, A) → (X , A )

is a cellular map, then given a continuous map f : X̃ n → Y , the composite map
g = f ψ : X̃ n → Y induces a unique cochain homomorphism
ψ # : C n+1 (X, A; πn (Y )) → C n+1 (X , A ; πn (Y ))
such that cn+1 (g) = ψ # cn+1 ( f ). Hence ψ induces a homomorphism
ψ ∗ : H n+1 (X, A; πn (Y )) → H n+1 (X , A ; πn (Y ))

Proposition 16.3.25 Let ψ : (X, A) → (X , A ) be a cellular map and f : A → Y

be a continuous map. If g = f ψ : A → Y , then the induced homomorphism
ψ ∗ : H n+1 (X, A; πn (Y )) → H n+1 (X , A ; πn (Y ))
sends On+1 ( f ) into On+1 (g).

Proof Using the relation cn+1 (g) = ψ # cn+1 ( f ), it follows that for each obstruction
of f , there is a corresponding obstruction of g, which is the image of ψ # . Conse-
quently, ψ # induces a map between the cohomology groups and hence in particular,
in obstruction sets. ❑
Proposition 16.3.26 A continuous map f : A → Y is (n + 1)-extensible over X iff
On+1 ( f ) = {0}.
Proof f : A → Y is (n + 1)-extensible over X iff cn+1 ( f ) = 0. This asserts that the
nullity class of cn+1 ( f ) and the cohomology class γ n+1 ( f ) are equal. This shows
that On+1 ( f ) = {0} in H n+1 (X , A ; πn (Y )). ❑
Theorem 16.3.27 If Y is n-simple and H n+1 (X, A; πn (Y )) = {0}, then for every
integer m such that m ≤ n < r, then the m-extensibility of f : A → Y over X
implies its r -extensibility over X .
Proof The theorem follows by repeated applications of the results of Exs. 12 and 13
of Sect. 16.5. ❑
Theorem 16.3.28 Let Y be n-simple and H n+1 (X, A; πn (Y )) = {0} for every inte-
ger n ≥ 1. Then every continuous map f : A → Y has a continuous extension
over X .
Proof Clearly, if X − A is of dimension not exceeding r given in Theorem 16.3.27.
Then it asserts that every continuous map f : A → Y has a continuous extension
over X iff it is n-extensible over X . This implies the theorem. ❑
16.3.5 The Homotopy Index
This subsection gives the concept of homotopy index which is important in obstruc-
tion theory. Homotopy problem is a special case of extension problem. So the tech-
niques of obstruction theory can be naturally applied.
Definition 16.3.29 Two maps f and g are said to be n-homotopic rel A if f | X̃ n
and g| X̃ n are homotopic relative to A. If f g rel A, then they are automatically
n-homotopic.
Definition 16.3.30 The supremum of n such that f and g are n-homotopic is called
the homotopy index of the pair of maps ( f, g) rel A.
Proposition 16.3.31 Every pair of maps f, g : X → Y such that f | A = g| A are

0-homotopic
Proof By assumption Y is path-connected. Hence the proposition follows. ❑
Remark 16.3.32 If (X, A) is a simplicial pair, then the homotopy index of any pair
of maps f, g : X → Y rel A is a topological invariant. For more results see exercises
in Sect. 16.5.
16.4 Applications
This section applies obstruction theory to solve some problems of algebraic topology
and proves some key results.
16.4.1 A Link between Cohomolgy and Homotopy

with Hopf Theorem
Theorem 16.4.1 Given a CW-complex X and an abelian group G, there is a bijec-

tion
ψ : [X, K (G, n)] → H n (X ; G), [ f ] → f ∗ (τn ).
Proof ψ is surjective: Let β ∈ H n (X ; G). Choose a cocycle g : C n (X ) → G

which represents β. Then g assigns an element g(σin ) ∈ πn (K (G, n)) = G. Let
h i : S n → K (G, n) be representatives of the elements of g(σin ). Let f (n) | X (n−1) be a
constant map. Define a map f (n) : X (n) → K (G, n) as the composite
∨h i
f (n) : X (n) → X (n) / X (n−1) = Sin −−−−→ K (G, n).
i
such that f (n) | X (n−1) is a constant map. Since g is a cocycle and g coincides with the
distinguishing cochain d(∗, f (n) ), it follows that
0 = δg = δd(∗, f (n) ) = c( f (n) ) − c(∗) = c( f (n) ).
This shows that there is an extension of the map f (n) : X (n) → K (G, n) to a map
(n+1)
f : X (n+1) → K (G, n). Then the further obstructions to extend the map
f (n+1) : X (n+1) → K (G, n)

to the skeletons X (n+q) are in the corresponding groups
C n+q (X ; πn+q−1 K (G, n)) = 0 for q ≥ 2.
This shows that ψ is a surjection.

ψ is injective: Let f, g : X → K (G, n) be two continuous maps such that
f ∗ (τn ) = g ∗ (τn ) in the cohomology group H n (X ; G). Then by cellular approxi-
mation theorem, we assume that f | X (n−1) = g| X (n−1) = ∗. Hence the element f ∗ (τn )
coincides with the cohomology class of the distinguishing cocycle d(∗, f ). Then
f ∗ (τn ) = [d(∗, f )] and g ∗ (τn ) = [d(∗, g)] give
[d( f, g)] = [d( f, ∗)] + [d(∗, g)] = − f ∗ (τn ) + g ∗ (τn ) = 0.
Consequently, there exists a homotopy
Ht : f | X (n) g| X (n)
relative to the skeleton X (n−2) . Hence all obstructions to extend this homotopy to the
skeletons X (n+q) are all zero. ❑
Remark 16.4.2 If X is a CW-complex of infinite dimension, using the intervals

p
2 − 1 2 p+1 − 1 1 1
, = 1 − , 1 − ,
2p 2 p+1 2p 2 p+1
we construct a homotopy between f | X (n+ p) and g| X (n+ p) .
Theorem 16.4.3 (Hopf) Let X be a CW-complex of dimension n. Then there is a

bijection ψ : H n (X ; Z) → [X, S n ].
Remark 16.4.4 The Theorem 16.4.1 asserts that the cohomology groups of a C W -
complex can be identified with the groups of homotopy classes of continuous maps
into Eilenberg-MacLane spaces.
16.4.2 Stepwise Extension of A Cross Section
This subsection considers the problems of constructing a cross section of a fiber

bundle. Throughout this subsection it is assumed that for the fiber bundle ξ : F ⊂
E → X , the base space X is a finite C W -complex. Suppose that A is a subcomplex
of X and the fiber space F is path-connected and n-simple. We use the notations of
Sect. 16.2. If A does not contain all of the 0-dimensional skeleton X (0) of X , then we
assume that we have a partially defined cross section f : X̃ n → E, the problem is
to extend it over X̃ n+1 . In such a problem, obstruction may appear. Indeed, if σ is an

(n + 1)-cell of X − A, the cross section f |∂σ might describe a nontrivial element in
πn (F) and in this case f will not have a continuous extension over σ. Consider the
ψ f : {(n + 1)-cells σ of X } → πn (F), σ → [ f |∂σ].
As by hypothesis, F is n-simple, and f |∂σ is a topological n-sphere, the function

ψ f , is well defined and ψ f sends an (n + 1)-cell σ to an element of πn (F) which
is determined by f |∂σ through some random trivialization ξ|σ ≈ σ × F. Then ψ f
can be extended by linearity and ψ f can be regarded as a πn (F)-valued cochain,
abbreviated cn+1 ( f ) ∈ C n+1 (X, A : πn (F)). Again for every (n + 2)-cell σ, we have
(δψ f )(σ) = ψ f (∂σ) = [ f |∂∂σ] = 0. This shows that cn+1 ( f ) is a cocycle. Hence
its cohomology class [cn+1 ( f )] is an element of H n+1 (X, A : πn (F)), the element
is usually abbreviated γ n+1 ( f ).
Remark 16.4.5 Because cn+1 ( f ) being the zero indicator that all of these elements
of H n+1 (X, A : πn (F)) vanish, it asserts that the given partially cross section can be
extended to X̃ n+1 using the homotopy between f |∂σ and the constant map.
If we start with a different partially defined cross section g that agrees with f | X̃ n−1 ,
then the resulting cocycle cn+1 (g) would differ from cn+1 ( f ) by a coboundary. This
asserts that there is a well-defined element of the cohomology group H n+1 (X, A :
πn (F)) such that if a partially defined cross section on X̃ n+1 exists that agrees with
the given choice on X̃ n−1 , then the cohomology class γ n+1 ( f ) must be trivial. Its
converse is also true as homotopy section is in the sense p f 1d .
Definition 16.4.6 Given an (n + 1)-cell σ of X , a function c( f, σ) is defined by the
rule c( f, σ) = ψ f (σ) = [ f |∂σ] ∈ πn (F).
Definition 16.4.7 The function ψ f of σ given ψ f (σ) = [ f |∂σ] = c( f, σ) ∈ πn (F)
is called the obstruction cocycle of f and is sometimes denoted by c( f ).
The above discussion with corresponding notations leads to the the following
important results.
Theorem 16.4.8 A cross section over A ∪ X (n) can be extended over A ∪ X (n+1) iff
c( f, σ) vanishes for each (n + 1)-cell σ of X .
Proof Let xσ be a base point of a given (n +1)-cell σ of X and Fσ be the fiber over xσ .
Choose an orientation of σ. Consider σ the oriented cell and ∂σ oriented boundary.
If f is a cross section f : X̃ n → E, define c( f, σ) = ψ f (σ) = [ f |∂σ] ∈ πn (F). As
F is n-simple, the theorem follows. ❑
Corollary 16.4.9 Given a fiber bundle p : E → X whose fiber F is a path-
connected n-simple space, then the element γ n+1 ( f ) ∈ H n+1 (X : πn (F)) vanishes
iff there are cross sections of p : E → X defined over the n-skeleton of X that extend
over the (n + 1)-skeleton.
Remark 16.4.10 For more results see Exercises in Sect. 16.5.
16.4.3 Homological Version of Whitehead Theorem
This subsection presents a homological version of the Whitehead theorem (see The-
orem 8.5.8) which asserts that if X and Y are connected abelian CW-complexes and
if f : X → Y induces isomorphisms on all homotopy groups, then f is a homotopy
equivalence.
Theorem 16.4.11 (Homological version of Whitehead theorem) Let X and Y be

simply connected CW-complexes. If a continuous map f : X → Y induces isomor-
phisms
f ∗ : Hn (X ) → Hn (Y )
on all homology groups, then f is a homotopy equivalence.
Proof It can be proved by applying the Hurewicz theorem and obstruction theory
to extend the homological version of Whitehead theorem to C W -complexes with
trivial action of π1 on all homotopy groups. ❑
Remark 16.4.12 For an alternative proof see Theorem 17.2.1 of Chap. 17.
16.4.4 Obstruction for Homotopy Between Relative Lifts
This subsection uses obstruction theory to obtain obstruction for homotopy between
relative lifts. The extension problems have closed connection with homotopy prob-
lems. We now want to define obstruction for homotopy of maps. Given continuous
two maps f 0 , f 1 : X → Y a homotopy between them is an extension of the map
X × ∂ I ∪ {∗} × I → Y determined by f 0 and f 1 to the space X × I . Hence we can
use obstruction theory to obtain obstruction for homotopy.
Let (X, A) be a C W -complex pair with i : A → X inclusion map and p : E → B
be a fiber map with fiber F. If h 0 , h 1 : X → E are two relative lifts of f as shown in
diagram Fig. 16.4, then h 0 , h 1 , g and f determine continuous maps

h t (x) if t = 0, 1 and x ∈ X
H : X × ∂ I ∪ A × I → E, (x, t) →
g(x) if x ∈ A
and
G : X × I → B, (x, t) → f (x).
Then there is a commutative square shown in Fig. 16.5 where j is inclusion. A rela-
tive lift L : X × I → E for this diagram is a homotopy between h 0 and h 1 such that
L(a, t) = g(a) and ( pL)(x, t) = f (x), ∀ a ∈ A and ∀ x ∈ X and ∀ t ∈ I . If C j is
the mapping cone of j, then the obstructions to this relative lift are in

relative lifting
Fig. 16.5 Commutative

square involving H and G
H n+1 (C j ; πn (F)) ∼
= H n+1 (X
× I /(X × ∂ I ∪ A × I ); πn (F))
∼
= H n+1 ( X/ A; πn (F))
∼
= H n+1 ( X/A; πn (F))
∼
= H n (Ci ; πn (F)),
where Ci is the mapping cone of the inclusion map i : A → X . This gives an

obstruction for homotopy between relative lifts.
16.5 Exercises
1. Let f : X → Y and g : Y → Z be two continuous maps in C0 . Show that

g f : X → Z is inessential iff g extends to g̃ : C f → Z .
2. Let X be a simple space (i.e., n-simple for every n ≥ 1). Show that πr (X ) = 0
for r ≤ n − 1, where n ≥ 1, iff for every CW-complex K of dimension ≤ n − 1,
any continuous map f : K → X is nullhomotopic.
3. Let f : X → Y be a map, where X is the homotopy type of a CW-complex.
Taking a Postnikov system for Y , the map f is said to be the n-trivial if the
composite map
f in
i n f : X −−−→ Y −−−→ Y [n]
is nullhomotopic. This is well defined, since it does not depend on the choice of
Y [n] . f is said to be trivial if it is n-trivial for all n.
If X is an n-dimensional CW-complex, show that f is n-trivial iff f is trivial.
4. Let X be a compact triangulable space and f : X → X be a continuous map.
Show that Lefschetz number f of f is the “obstruction” to f being fixed point
free.
[Hint. Use Lefschetz fixed point theorem.]
5. Let f : X → Y be (n − 1)-trivial. For the exact sequence in S
α∗
[X, K (πn (Y ), n)] −−−−→ [X, Y [n] ] −
→ [X, Y [n−1] ],
let the set

n (X ; πn (Y )).
On ( f ) = α∗−1 [(i n f )] ⊂ H
Show that the set On ( f ) is nonempty iff f is n-trivial.
(On ( f ) is called the n-dimensional obstruction set to f being essential).

6. Let K be a finite CW-complex. If πn (X ) = 0 for all n ≥ 0, show that any two
continuous maps f, g : |K | → X are homotopic.
7. Show that
(i) a polyhedron |K | is contractible iff πn (|K |) = 0 for all n ≥ 0.
(ii) Let K be an m-dimensional simplicial complex. If n is a positive integer
and πn (Y ) is zero for all m = n (and if n = 1, Y is 1-simple), then there
is a bijective correspondence between the homotopy classes of maps of |K |
into Y and the elements of H n (K ; πn (Y )).
8. Let Y be an n-connected space and K be a CW-complex of dimension n. Show
that [K , Y ] = {∗}.
9. Show that the extension index of a continuous map f : A → Y is a topological
invariant.
10. Let (X, A) be a relative CW-complex and f, g : X (n) → Y be two continuous
maps such that f | X (n−1) g| X (n−1) rel A. Show that a choice of homotopy defines
a difference cochains d ∈ C n (X, A : πn (Y )) such that δd = θn+1 ( f ) − θ n+1 (g).
11. Let (X, A) be a relative CW-complex, n ≥ 1, and Y be an (n−1)-connected space
(π1 (Y ) is assumed to be abelian for n = 1). Let f, g : X → Y be two continuous
maps such that f | A = g| A . Show that f | X (n−1) g| X (n−1) rel A and the obstruction
in H n (X, A; πn (Y )) to extending this homotopy to X (n) is independent of the
choice of homotopy on X (n−1) and depends only on the homotopy classes of f
and g relative to A.
12. Show that a continuous map f : A → Y is n-extensible over X iff On+1 ( f ) is
nonempty.
13. Show that a continuous map f : A → Y is (n +1)-extensible over X iff On+1 ( f )
contains the zero element of H n+1 (X, A; πn (Y )).
14. Let Y be a path-connected and n-simple space. If H n+1 (X, A; πn (Y )) = 0
for every integer n ≥ 1, show that every continuous map f : A → Y has a
continuous extension over X .
15. (Poincaré-Hopf theorem) Show that a closed oriented n-manifold has a nowhere
zero vector field iff its characteristic is zero.
16. Given a fiber bundle ξ : F ⊂ E → X , the base space X is a finite C W -complex.
Suppose that A is a subcomplex of X and the fiber space F is path-connected
and n-simple. If ξ admits a cross section f : X̃ n → F, using the notations used
in Sect. 16.4.2 show that
16.5 Exercises 529
(i) f has a continuous extension to a cross section f˜ : X̃ n+1 → F iff cn+1 ( f ) =

0;
(ii) If f, g : X̃ n → F are homotopic cross sections, then cn+1 ( f ) = cn+1 (g);
(iii) cn+1 ( f ) is a cocycle and its cohomology class γ n+1 ( f ) ∈ H n+1 (X, A;
πn (F);
(iv) γ n+1 ( f ) is a topological invariant.
17. If (X, A) is a simplicial pair, show that the homotopy index of any pair of maps
f, g : X → Y rel A is a topological invariant.
18. Let X be a CW-complex, and Y be a path-connected (n − 1)-simple space. Given
a continuous map f : X (n−1) → Y , define
c( f ) : Hom (Cn (X ), πn−1 (Y )) → πn−1 (Y ), σ → [ f ψαn ],
where the characteristic map ψαn is regarded as map of S n−1 to X (n−1) . Show that
(i)c( f ) depends only on the homotopy class of f ;
(ii)c( f ) = 0 iff f has a continuous extension to a map f˜ : X (n) → Y ;
(iii)δc( f ) = 0;
(iv) if γ( f ) represents the homotopy class of c( f ) in H n (X ; πn−1 (Y )), then
γ( f ) = 0 iff there exists a continuous map g : X (n) → Y such that f = g
on X (n−2) .
(v) Let B be a CW-complex and A ⊂ B be its subcomplex. Suppose X̃ n =
B (n) ∪ A, where B (n) is the nth skeleton of B. Let σ = en+1 be an (n +1)-cell
of B, which does not belong to A and ψσ : S n → X n be the attaching map
corresponding to A and ψσ : S n → X n be the attaching map corresponding
to the cell σ. Any map f : X̃ n → Y , where Y is homotopically simple (in the
sense that π1 (Y ) acts trivially on πn (Y ) for each n), defines a cochain c( f )
by taking the value c( f ) on the generator σ, given by c( f )(σ) = [ f ψσ ] ∈
πn (Y ), where the composite is given by
ψσ f
f ψσ : S n −−−−→ X̃ n −−−→ Y.
Show that the cochain c( f ) is a cocycle, i.e., δc( f ) = 0.

(vi) Using the notation of the above Ex. 18 (v), show that a continuous map
f : X̃ n → Y has a continuous extension f˜ : X n+1 → Y iff c( f ) = 0.
(vii) If f f and g g rel A, then the pair ( f , g ) has the same homotopy
index rel A as the pair ( f, g).
(viii) Let Y be a homotopically simple space, (X, A) a CW-pair and X̃ n = X (n) ∪ A
for n = 0, 1, . . . . If f : X̃ n → Y is a continuous map, show that there exists
a continuous map g : X̃ n+1 → Y such that g| X̃ n−1 = f | X̃ n−1 iff [c( f )] = 0
in H n+1 (B, A; πn (Y )).
[1] Adams, J.F., On the non-existence of elements of Hopf invariant one, Ann of
Math. 72(2), 20–104, 1960.
[2] Adams J.F., Algebraic Topology: A student’s Guide, Cambridge University
[3] Dieudonné J., A History of Algebraic and Differential Topology, 1900–1960,
[4] Dold A., Lectures on Algebraic Topology, Springer-Verlag, New York, 1972.
[5] Hilton P.J., An introduction to Homotopy Theory, Cambridge University Press,
Cambridge, 1983.
[6] Hilton P. J. and Wylie S., Homology Theory, Cambridge University Press, Cam-
bridge, 1960.
[7] Massey W.S., A Basic Course in Algebraic Topology, Springer-Verlag, New
[8] Mayer J. Algebraic Topology, Prentice-Hall, New Jersy, 1972.
[9] Olum P., Obstructions to extensions and homotopies, Ann of Math. 25, pp 1–25,
1950.
[10] Switzer R.M., Algebraic Topology-Homotopy and Homology, Springer-Verlag,
[11] Whitehead, G.W., On mappings into group like spaces, Comment. Math. Helv.
28, 320–328. 1954.
[12] Whitehead G.W., Elements of Homotopy Theory, Springer-Verlag, New York,
References
Adams, J.F.: On the non-existence of elements of Hopf invariant one. Ann. Math. 72(2), 20–104
(1960)
(1972)
Davis, J.F., Kirk, P.: Lecture Notes in Algebraic Topology. Indiana University, Bloomington (2001).
http://www.ams.org/bookstore-getitem/item=GSM-35
Dieudonné, J.: A History of Algebraic and Differential Topology, 1900–1960. Modern Birkhäuser
Classics. Birkhäuser, Basel (1989)
Dodson, C.T.J., Parkar, P.E.: User’s Guide to Algebraic Topology. Kluwer Academic Publishers,
Dordrecht (1997)
Eilenberg, S., Steenrod, N.: Foundations of Algebraic Topology. Princeton University Press,
Princeton (1952)
References 531

Olum, P.: Obstructions to extensions and homotopies. Ann. Math. 25, 1–25 (1950)
Samelson, H.: Groups and spaces of loops. Comment. Math. Helv. 28, 278–286 (1954)
Steenrod, N.E.: The Topology of Fibre Bundles. Princeton University Press (1951)
Steenrod, N.E.: Chohomology operations and obstructions to extending continuous functions. Adv.
Math. 8, 371–416 (1972)
Whitehead, G.W.: On mappings into group like spaces. Comment. Math. Helv. 28, 320–328 (1954)
Chapter 17
More Relations Between Homology
and Homotopy
This chapter displays some similarities and further interesting relations between
homology and homotopy groups of topological spaces in addition to some relations
between these theories discussed earlier. The concept of homotopy presents a math-
ematical formulation of the intuitive idea of a continuous transition between two
geometrical configurations. On the other hand, the concept of homology presents
a mathematical precision to the intuitive idea of a curve bounding an “area” or a
surface bounding a “volume.” L.E.J. Brouwer (1881–1967) first connected these
two basic concepts of algebraic topology in 1912 by proving 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 and only if they are equivalent
from the view point of homology theory). Hopf’s classification theorem generalizes
Brouwer’s result to an arbitrary dimension.
The homotopy groups resemble the homology groups in many respects under
suitable situations as shown by Hurewicz in his celebrated “Equivalence Theorem.”
Homotopical and homological versions of Whitehead theorem are similar. Since
homology groups are in general more computable than homotopy groups, the homo-
logical version of Whitehead theorem is often convenient to apply. Cohomology
groups of a C W -complex are dual to homotopy groups in the sense that cohomology
groups of a C W -complex can be identified with the groups of homotopy classes
of continuous maps into Eilenberg–MacLane spaces. By replacing the Eilenberg–
MacLane spaces by suitable spaces, “generalized cohomology theories” are con-
structed in Chap. 15.
There is also a lack of similarities between these two theories essentially due to
absence in higher homotopy groups the excision property for homology and also
absence in higher homotopy groups a theorem analogous to van Kampen theorem
for fundamental group. This chapter continues to study Eilenberg–MacLane spaces,
Moore spaces, Dold–Thom theorem, Hopf invariant and Adams classical theorem
on Hopf invariant.

DOI 10.1007/978-81-322-2843-1_17
534 17 More Relations Between Homology and Homotopy
In this chapter, C 0 denotes the category of pointed topological spaces having homo-
topy type of finite pointed C W -complexes and C denotes the category of topological
pairs of spaces having homotopy type of finite C W -complex pairs.
For this chapter, the books Adams (1972), Gray (1975), Hatcher (2002), Maunder
(1980), Spanier (1966), the papers Eilenberg and MacLane (1945), Steenrod (1949)
and some others are referred in the Bibliography.
17.1 Some Similarities and Key Links
Higher homotopy groups which are the natural higher-dimensional analogue of

the fundamental groups carry certain similarities and key links with homology
groups.
17.1.1 Some Similarities
This subsection shows that the homotopy groups resemble the homology groups in
many aspects.
Example 17.1.1 The fundamental groups π1 (X ) are not always abelian but the
groups πn (X ) are always abelian for n ≥ 2. On the other hand, homology groups
Hn (X ) are always abelian for n ≥ 1.
Example 17.1.2 The relative homotopy groups give a long exact sequence like long
exact sequence of homology groups.
17.1.2 Hurewicz Homomorphism Theorem: A Key Link
This subsection establishes a key link between homotopy and homology groups
with the help of Hurewicz homomorphism given by Withold Hurewicz (1904–1956)
during 1934–1936. His classical result known as Hurewicz theorem says that for
n ≥ 1 the first nonzero homotopy group πn (X ) of a simply connected space X is
isomorphic to the first nonzero ordinary homology group Hn (X ) for n > 1. Their
relative version is also similar.
Recall that for any topological space X and positive integer m there exists a group
homomorphism
h ∗ : πm (X ) → Hm (X )
called the Hurewicz homomorphism from its m-th homotopy group to its m-th homol-
ogy group (with integer coefficients). For m = 1, the fundamental group is not abelian
in general but its abeliazation is the first homology group
17.1 Some Similarities and Key Links 535
H1 (X ) ∼
= π1 (X )/[π1 (X ), π1 (X )],
where [π1 (X ), π1 (X ), ] is the commutator subgroup of π1 (X ).

Definition 17.1.3 (Hurewicz homomorphism) Let m > 0 and H∗ be the ordinary
m (X ) ∼
homology theory. The homomorphism h : πm (X ) → H = Hm (X ) defined as
the composite
πm (X ) −−−→ πm+1 (X ∧ S 1 ) ∼ m (X ),
E
= πm+1 (X ∧ K (Z, 1)) −
→H
(use the result that S 1 K (Z, 1)) is a natural homomorphism, called the Hurewicz
homomorphism, where E is the Freudenthal suspension homomorphism
Remark 17.1.4 An equivalent formulation of Hurewicz theorem given
Theorem 10.11.2 of Chap. 10 is now presented.
Theorem 17.1.5 (Hurewicz) Let X be a simply connected pointed topological space.
Then the following statements are equivalent:
(i) πi (X ) = 0, if 1 ≤ i < n(n ≥ 2);
i (X ) = 0, if 1 ≤ i < n(n ≥ 2).
(ii) H
Either implies that h : πr (X ) → Hr (X ) is an (n + 1)-isomorphism.
Proof Since h is natural, and X is well pointed we may assume that X is a C W -
complex. First suppose that πi (X ) = 0 for i < n. Then E : πr (X ) → πr +1 ( X ) is
an (n + 1)-isomorphism, since n > 1. Consider the composite map βm :
E (1d ∧h m )∗
πr +m (X ∧ K (Z, m)) −−−→ πr +m+1 (X ∧ K (Z, m) ∧ S 1 ) −−−−−−−−→ πr +m+1 (X ∧ K (Z, m + 1)).
Again since X ∧ K (Z, m) is (m + n − 1)-connected, E is an isomorphism for r <

m + 2n − 1 and is onto if r = m + 2n − 1. Let f : X → Y be continuous and f˜ :
X → Y be adjoint to f . Consider the commutative diagram as shown in Fig. 17.2,
Exercise 1 of Sect. 17.6.
Assume that (K (Z, m + 1), K (Z, m)) is a relative C W -complex with cells
in dimensions greater than 2m + 1. This implies that (X ∧ K (Z, m + 1), X ∧
K (Z, m) ∧ S 1 ) is a relative C W -complex with cells in dimensions > n + 2m + 1.
Hence it follows that (1d ∧ h m )∗ is an isomorphism if r + m + 1 < n + 2m + 1 and
is onto if r + m + 1 = n + 2m + 1. Hence, βm is an isomorphism if r < m + n
and is onto if r = m + n. Consequently, h is an (n + 1)-isomorphism. This h is an
(n + 1)-isomorphism under condition (i). Hence it follows that the statement (i) is
equivalent to the statement (ii). ❑
Corollary 17.1.6 Let X be a simply connected space. If Hi (X ) = 0 for all i <
n, then πi (X, ∗) = 0 for i < n and the Hurewicz homomorphism h : πn (X, ∗) →
Hn (X ) is an isomorphism for every integer n ≥ 2.
Proof It follows from Theorem 17.1.5 that h : πm (X ) ∼
= Hm (X ) for the smallest m
such that πk (X ) = 0 for 1 ≤ k < m. ❑
Remark 17.1.7 The first nontrivial homotopy group of a simply connected space
X and the first nontrivial homology group of the same space X occur in the same
dimension and they are isomorphic under Hurewicz homorphism h.
Remark 17.1.8 For n = 1, the Hurewicz homomorphism h : π1 (X ) → H1 (X ) has
as kernel the commutator subgroup of π1 (X ).
17.2 Relative Version of Hurewicz Homomorphism

Theorem
This section conveys relative version of Hurewicz homomorphism theorem. Let

(X, A) ∈ C and A
= ∅. Define k : πi (X, A) → Hi (X, A) to be the composite
( p A )∗ h
i (X/A) = Hi (X, A),
πi (X, A) −−−−−→ πi (X/A, ∗) −−−→ H
where h is the Hurewicz homomorphism.

Theorem 17.2.1 (Relative Hurewicz theorem) Let A be simply connected, and
π1 (X, A) = 0. Then the following statements are equivalent:
(i) πi (X, A) = 0 for 1 ≤ i < n(n ≥ 2);
(ii) Hi (X, A) = 0 for 1 ≤ i < n(n ≥ 2).
Either implies that k : πi (X, A) → Hi (X, A) is an isomorphisms for i ≤ n and onto
for i = n + 1.
Proof We may assume that (X, A) is a C W -pair. As in the case of Hurewicz Theorem
17.1.5, it follows that
( p A )∗
(i) implies the final condition. But πi (X, A) −−−−−→ πi (X/A) is an (n + 1)-
isomorphism. Hence the theorem follows. ❑
Definition 17.2.2 A map f : X → Y between C W -complexes is said to be an h-
equivalence if its induced homomorphisms
f ∗ : πm (X ) → πm (Y )
are isomorphism for each m ≥ 1 and X is said to be h-equivalent to Y if there exists

an h-equivalence f : X → Y .
Theorem 17.2.3 Let X be simply connected C W -complex such that Hm (X ) ∼ =
Hm (S n ) for n ≥ 2, with integral coefficients. Then X is h-equivalent to S n .
Proof By Corollary 17.1.6, πn (X ) ∼ = Hn (X ), and this group is assumed to be iso-
morphic to Z. Let f : S n → X represent a generator. Then f ∗ : πn (S n ) → πn (X )
is an isomorphism in homotopy and also f ∗ : Hn (S n ) → Hn (X ) is so in homology.
Then by Whitehead theorem, f is a homotopy equivalence and hence is h-equivalent
to S n ❑
17.4 Dold–Thom Theorem 537
Remark 17.2.4 Theorem 17.2.3 has interesting applications. For example, a closed
connected n-manifold of the homotopy type of the n-sphere S n is homeomorphic to
the n-sphere (compare with Poincaré conjecture given in Sect. 18.1). Consequently,
these spheres are characterized by invariants of algebraic topology.
Example 17.2.5 The spaces S n ∨ S n ∨ S 2n and S n × S n are simply connected for

n ≥ 2 and have isomorphic homology groups. As their cohomology rings are differ-
ent, they cannot be h-equivalent.
17.3 Alternative Proof of Homological Version

of Whitehead Theorem
This section conveys an alternative proof of Theorem 16.4.11 of homological version

of the well known classical Whitehead theorem given by J.H.C. Whitehead (1904–
1960) in homotopy theory saying that a continuous map between C W -complexes
which induces isomorphisms on all homotopy groups is a homotopy equivalence.
Since homology groups are easier to compute in general than homotopy groups, the
homological version of Whitehead theorem is often convenient to use.
Theorem 17.3.1 (Whitehead theorem in homological form) Let X and Y be both

simply connected C W -complexes and f : X → Y be a continuous map. If the
induced homomorphism f ∗ : Hm (X ) → Hm (Y ) is an isomorphism for each m, then
Proof Let C f = Z be the mapping cylinder of f . We may consider f to be an incu-

sion X → Z . Again since the spaces X and Y are both simply connected, it follows
that π1 (Z , X ) = 0. The relative version of Hurewicz homomorphism Theorem 17.2.1
asserts that the first nonzero homotopy group πm (Z , X ) is isomorphic to the first
nonzero homology group Hm (Z , X ) for m > 1. All the groups Hm (Z , X ) are zero
from the long exact sequence of homology. This shows that all the groups πm (Z , X )
also vanish. This implies that the inclusion X → Z induces isomorphisms on all
homotopy groups. Consequently, this inclusion is a homotopy equivalence. Hence it
follows from the diagram in Fig. 17.1 that f = p ◦ i is a homotopy equivalence. ❑
Remark 17.3.2 Theorem 17.3.1 gives an alternative proof of Theorem 16.4.11.

Whitehead theorem
17.4 Dold–Thom Theorem
This section studies Dold–Thom Theorem given by Albrecht Dold (1928–2011)

and René Thom (1923–2002) in 1958 and its immediate consequences. This theo-
rem presents a homotopy theoretic definition of homolgy. More precisely, let C0 be
the category of base pointed finite C W -complexes and S P ∞ be the infinite sym-
metric product functor defined in Sect. B.2.5 of Appendix B on C0 . The functor
S P ∞ has the interesting property that it can be used to define Eilenberg–MacLane
spaces: S P ∞ (S n ) is an Eilenberg–MacLane space K (Z, n) and for a Moore space
M(G, n), the space S P ∞ (M(G, n)) is a K (G, n). In particular, S P ∞ (S 2 ) = CP ∞
is an Eilenberg–MacLane space K (Z2 , 2). Dold–Thom theorem establishes a sur-
prising close connection between πi (S P ∞ (X )) and Hi (X ; Z) for every space X in
C0 .
Theorem 17.4.1 (Dold–Thom Theorem) The functor X → πi (S P ∞ (X )) for i ≥ 1
coincides with the functor X → Hi (X ; Z) on C0 .
Proof See Dold and Thom (1958). ❑
Corollary 17.4.2 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.
Corollary 17.4.3 (i) S P ∞ (S n ) is a K (Z, n);
(ii) For a Moore space M(G, n), S P ∞ (M(G, n)) is a K (G, n).
Proof (i) It follows from Dold–Thom Theorem by taking in particular, X = S n .
(ii) It follows from Dold–Thom Theorem by taking X = M(G, n). ❑
Corollary 17.4.4 A path-connected, commutative, associative H -space X with a
strict identity element has the weak homotopy type of a product of Eilenberg–
MacLane spaces.
Proof Left as an exercise.
Corollary 17.4.5 The functor S P ∞ gives Eilenberg–MacLane spaces.
∞
-complex, then S P (X ) is path-connected and has the weak
Proof If X is a C W
homotopy type of n K (Hn (X ), n). Hence the corollary follows from Corollary
17.4.4. ❑
Remark 17.4.6 The map πn (X ) → πn (S P ∞ (X )) = Hn (X ; Z) induced by the inclu-
sion X = S P 1 (X ) → S P ∞ X is the Hurewicz homomorphism. Using the Hurewicz
homomorphism and naturality this reduces to the case X = S 1 , where the map
S P n (S n ) → S P ∞ (S n ) induces on π1 a homomorphism Z → Z, which is an isomor-
phism. The suspension isomorphism makes a further definition to the case n = 1,
where the inclusion S P 1 (S 1 ) → S P ∞ (S 1 ) is a homotopy equivalence and hence it
induces an isomorphism on π1 .
17.4 Dold–Thom Theorem 539
Remark 17.4.7 Dold–Thom theorem asserts that πn (S P ∞ )(X ) = ∼ Hn (X ) on C0 for

all n ≥ 1. Hence πn (S P ∞ (S 2 )) = 0 for all n > 2. On the other hand
π3 (2 2 S 2 ) ∼
= π5 (S 4 ) ∼
= Z2 ,
which is generated by the double suspension of the Hopf map p : S 3 → S 2 (Dold

and Thom 1958).
17.5 The Hopf Invariant and Adams Theorem
This section defines Hopf invariant using cup product and discusses Adams theorem
which provides a solution of vector field theorem.
17.5.1 Hopf Invariant
H. Hopf (1894–1971) introduced in 1935 the concept of an invariant, now called

Hopf invariant. The Hopf invariant has been generalized by G.W. Whitehead in 1950
to a homomorphism
H : πm (S n ) → πm (S 2n−1 ) (m ≤ 4n − 4)
and by P.J. Hilton in 1951 to a homomorphism
H : πm (S n ) → πm+1 (S 2n ) (m > 0)
Homology is used to define the degree of a spherical map f : S n → S n which

distinguishes different homotopy classes of maps f . Cup products can be used to
define something similar concept for maps S 2n−1 → S n . Hopf did this using more
geometric constructions prior to the invention of cohomolgy and cup products. There
are several definitions of the Hopf invariant H ( f ) for a continuous map f : S 2n−1 →
S n . We define H ( f ) here as one of the most remarkable applications of cup product
in topology.
Recall that S n and S 2n−1 may be given the structure of finite C W -complexes,
each having only two cells. Given a continuous map f : S 2n−1 → S n for n ≥ 2, let
S nf denote the space obtained by attaching a 2n-cell to S n via f . Then S nf is a finite
C W -complex with these cells: one is of dimension 0, one is of dimension n and one
is of dimension 2n. Since n > 1, the cohomology of S nf is given by

∼ Z, if m = 0, n, 2n
H m
(S nf ) =
0, otherwise.
Definition 17.5.1 Let a ∈ H 2n (S nf ; Z) and b ∈ H n (S nf ; Z) be a chosen pair of gen-

erators. The integer H ( f ), called Hopf invariant is defined to be the integer for which
the cup product b ∪ b = b 2 = H ( f )·a in H 2n (S nf ; Z).
Example 17.5.2 If n is odd then, H ( f ) = 0.
Example 17.5.3 Consider the exact homotopy sequence of the fibration p : S 3 → S 2

p∗
· · · → π3 (S 1 , s0 ) −
→ π3 (S 3 , s0 ) −−−−→ π3 (S 2 , s0 ) −
→ π2 (S 1 , s0 ) → · · ·
Since π3 (S 1 , s0 ) = π2 (S 1 , s0 ) = 0, p∗ : π3 (S 3 , s0 ) → π3 (S 2 , s0 ) is an isomorphism.
Consequently, π3 (S 2 , s0 ) ∼ = Z, the first example, where πm (S n , s0 )
= 0 for m > n.
Since π3 (S , s0 ) is generated by [1 S 3 ], it follows that π3 (S 2 , s0 ) is generated by [ p].
3
The map p is called the Hopf map.
For each q, consider the isomorphisms

→∼
π2q+2 (S q+2 , s0 ) − →∼
=π2q+3 (S q+3 , s0 ) − →∼
=··· − →∼
=πq+n (S n , s0 ) − =···
For each q > 1 the common group πn+q (S n , s0 ) is denoted by by πqS . It is called the
kth stable homotopy group. For example, π1S ∼ = π4 (S 3 , s0 ) ∼
= Z2 and is generated by
[ p], where p : S 3 → S 2 is the Hopf map.
The exceptional case π4n−1 (S 2n , s0 ) invites attraction in many respects. The homo-
morphism
H : π4n−1 (S 2n , s0 ) → Z, f → H ( f )
defined by Hopf is now called the Hopf invariant.
Remark 17.5.4 If n is odd, then H ( f ) = 0, because of anticommutativity of the cup

product.
Definition 17.5.5 Let X be the C W -complex obtained by attaching a (2n + 2)-cell

to S n using f as the attaching map. Then

∼ Z, if m = 0, n + 1 and 2n + 2
H (X ) =
m
0, otherwise
Remark 17.5.6 The elements of π3 (S 2 ) may be given a geometrical interpretation

by assigning an integer (Hopf integer) to each element of π3 (S 2 ).
An alternative definition of Hopf invariant given by N. Steenrod in 1949 is now

conveyed.
Definition 17.5.7 (Steenrod) Given an element α ∈ π2n−1 (S n )(n ≥ 1), the Hopf
invariant of α is also defined as follows: Represent α by a map f : S 2n−1 → S n and
17.5 The Hopf Invariant and Adams Theorem 541
let Y = C f . Then H n (Y ) = ∼ Z, H 2n (Y ) =
∼ Z, the generators being τn , τ2n , where
∗ ∗
f 1 (τn ) = sn , τ2n = f 2 (s2n ), and sn , s2n are the generators of H n (S n ), H 2n (S 2n )
respectively. The Hopf invariant of α denoted by H (α) is the defined by
τn2 = τn ∪ τn = H (α) · τ2n .
H (α) is an integer, called the Hopf invariant of α. This integer H (α) depends only
on α in the sense that it does not depend on the choice of its representative f and
hence it is well defined.
Remark 17.5.8 Consider a continuous map f : S 2n−1 → S n for n ≥ 1. Then there

exists a unique integer integer H ( f ), called the Hopf invariant of f .It depends only
on the homotopy class of f . The assignment
H : π2n−1 (S n ) → Z, f → H ( f )
is a homomorphism such that for n = 2, 4, 8, H ( f ) = 1. Its converse is also true:

up to homotopy, the Hopf maps are the only ones of Hopf invariant 1. This proves
the purely algebraic theorem that C, H,, and Cayley numbers are the only nontrivial
real division algebra. (see Remark 17.5.11).
Remark 17.5.9 The Definition 17.5.7 is due to Steenrod given in 1947. It is some
what different from Hopf original definition.
17.5.2 Vector Field Problem and Adams Theorem
The problem for which n there exists a continuous map f : S 2n−1 → S n with Hopf
invariant H ( f ) = 1 was solved by J.F. Adams (1930–1989) in his papers (Adams
1958, 1960). This theorem, also called Adams’ theorem, is a deep theorem in homo-
topy theory which states that the only n-spheres which are H -spaces are S 0 , S 1 , S 3 ,
and S 7 . This relates to the existence of division algebra structure on Euclidean space
Rn . A division algebra is a finite dimensional real vector space together with a bilin-
ear multiplication having both-sided identity and such that each nonzero element
has a both-sided multiplicative inverse. The real numbers, complex numbers, the
real quaternions, and the Cayley numbers are examples of real division algebras. J.F.
Adams proves that there are no other examples. Corresponding to each continuous
map f : S 2n−1 → S n one can associate an integer H ( f ), called Hopf invariant. This
means that to each element of π2n−1 (S n ), one can assign an integer which is its Hopf
invariant.
Theorem 17.5.10 (Adams) There exists a continuous map f : S 2n−1 → S n with

Hopf invariant one only when n = 2, 4, 8.
Proof See Adams (1958, 1960). ❑

Remark 17.5.11 The Definition 17.5.7 is due to Steenrod given in 1947. It is some
what different from Hopf original definition. For any integer n > 0, that there exists
continuous S 4n−1 to S 2n of arbitrary even Hopf invariant. Does there exist maps
having odd Hopf invariant? Hopf maps S 3 → S 2 and S 7 → S 4 are each of Hopf
invariant one. Using Cayley numbers, one can define an analogous map from S 15 →
S 8 of Hopf invariant one. Adams showed in 1952 that there exist maps f : S 4n−1 →
S 2n of odd Hopf invariant using cohomology operations only when n is a power of 2.
If π2n−1 (S n ) contains no element of Hopf invariant one, then there is no real division
algebra of dimension n. Adams proved in 1960 that such elements exist precisely for
n = 1, 2, 4, 8, whose simpler proof is given by (Atiyah 1967) using K -theory.This
result asserts purely algebraic theorem that R, C, H and Cayley numbers are the
only real division algebras (see Steenrod and Epstein 1962).
17.6 Exercises
1. Let f : X → Y be a continuous map and f˜ : X → Y be adjoint to f . Show

that the diagram in Fig. 17.2 commutes.
2. Show that the diagram in Fig. 17.3 commutes up to sign.
3. If X is path-connected, show that there is an epimorphism
h : π1 (X ) → H1 (X ; Z).
Fig. 17.2 Commutative

square involving f˜∗ and f ∗
Fig. 17.3 Diagram

involving E and σ
17.6 Exercises 543
4. If X is path-connected, show that the homomorphism
h : π1 (X ) → H1 (X ; Z)
h ∗ : π1 (X )/ker h → H1 (X ; Z)
with ker h, the kernel of h the commutator subgroup of π1 (X ).

5. If X is a simply connected C W -complex, and

n (X ) = Z, if n = m
H
0, if n
= m,
show that X S m .
6. Given an element α ∈ π2n−1 (S n )(n ≥ 1), let H (α) be the Hopf invariant of α.
Prove that
(i) H (α) depends only on α, and not on the choice of its representative f :
S 2n−1 → S n ;
(ii) H : π2n−1 (S n ) → Z, f → H ( f ) is a homomorphism;
(iii) H ( f ) = 0 if n is odd;
(iv) if n is even, H [βn , βn ] = ±2, where βn is the generator of πn (S n ) represented
by the identity map;
(v) Deduce that if n is even, then 2 ∈ Im H and π2n−1 (S n ) has an element of
infinite order, and S n can not be a Hopf space.
7. For any integer n > 0, show that there exists a continuous map f from S 4n−1 to
S 2n of arbitrary even Hopf invariant H ( f ).
8. Show that two continuous maps from S 3 to S 2 are homotopic iff they have the
same Hopf invariant.
9. Let f : X → Y be a continuous map of simply connected C W -complexes. Show
that f is a homotopy equivalence if and only if f ∗ : Hi (X ) ∼
= Hi (Y ) for all i.
10. Let (X, A) be a pair of simply connected C W -complexes. If Hm (X, A) = 0
for m < n, n ≥ 2, then show that πm (X, A) = 0 for m < n and the Hurewicz
homomorphism
h : πn (X, A) → Hn (X, A)
is an isomorphism.
11. (Homological description of Moore space) Given an abelian group G and an
integer n ≥ 2, a pointed C W -complex X is said to be a Moore space of type
(G, n) if X is n-connected and

Hi (X ) = G, if i = n .
0, if i
= n
A Moore space of type (G, n) or any space homeomorphic to it is denoted by

M(G, n). For example, S n is a Moore space of type (G, n).
Show that
(i) The Hurewicz homomorphism
h n : πn (M(G, n)) → Hn (M(G, n))
is an isomorphism.
(ii) πn (M(G, n)) = G.
(iii) If f : G → H is a homomorphism of abelian groups, then there exists a
continuous map
ψ : M(G, n) → M(H, n)
such that its induced homomorpshim
ψ∗ : Hn (M(G, n)) → Hn (M(H, n))
coincides with f.
12. Let f : X → Y be a map between simply connected spaces. If f ∗ : Hm (X ) →
Hm (Y ) is bijective if m < n and surjective if m = n(n ≥ 2), show that
f ∗ : πm (X ) → πm (Y )
is bijective for m < n and surjective for m = n.

13. Show that for an integer n ≥ 1, there exist maps
f : S 4n−1 → S 2n
of arbitrary even Hopf invariant H ( f ).

14. Show that the image of Hopf invariant H : π4n−1 (S 2n ) → Z is either Z or 2Z.
15. (Hopf) For each continuous map f : S 3 → S 2 , show that there exists an integer
H ( f ) with the properties:
(i) if f, g : S 3 → S 2 are homotopic, then H ( f ) = H (g);
(ii) If h : S 2 → S 3 is a continuous map with degree n, then H ( f oh) = n.H ( f );
(iii) There exists a map g : S 3 → S 2 such that H (g) = 1;
(iv) If there exists a Hopf invariant H ( f ) with the above properties, then the
map
ψ : π3 (S 2 ) → Z, [ f ] → H ( f )
is surjective.
16. Let τ2n be a generator of the group π2n (S 2n ) and [τ2n , τ2n ] ∈ π4n−1 (S 2n ) be the
Whitehead product. Show that
17.6 Exercises 545
(i) the Hopf invariant H ([τ2n , τ2n ]) = 2;

(ii) the element [τ2n , τ2n ] has infinite order;
(iii) the group π4n−1 (S 2n ) is infinite for every n ≥ 1.

[3] Switzer, R.M., Algebraic Topology:Homotopy and Homology, Springer-Verlag,
References
Adams, J.F.: On the nonexistence of elements of hopf invariant one. Bull. Am. Math. Soc. 64,
279–282 (1958)
Adams, J.F.: On the nonexistence of elements of hopf invariant one. Ann. Math. 72, 20–104 (1960)
(1972)
Atiyah, M.F.: K -Theory. Benjamin, Elmsford (1967)
Basel (1989)
Dold, A., Thom, R.: Quasifaserungen und unendliche symmetrische Produkte. Ann. Math. 67(2),
239–281 (1958)
Eilenberg, S., MacLane, S.: Relations between homology and homotopy groups of spaces, Ann.
Math. 46(2) (1945)
ton (1952)
(1975)
Maunder C.R.F: Algebraic Topology, Van Nostrand (1970). Cambridge University Press,
Cambridge (1980)
Steenrod, N.: Cohomology Invariants Mapp. 50, 954–988 (1949)
Steenrod, N., Epstein, D.B.A.: Cohomolgy operations, Annals of Mathematics Studies, vol. 50.
Prineton University Press, Princeton (1962)
Chapter 18
A Brief History of Algebraic Topology
This chapter focuses the history on the emergence of the ideas leading to new areas of
study in algebraic topology and conveys the contributions of some mathematicians
who introduced new concepts or proved theorems of fundamental importance or inau-
gurated new theories in algebraic topology starting from the creation of homotopy,
fundamental group, and homology group by H. Poincaré (1854–1912) in 1895, which
are the first most profound and far reaching inventions in algebraic topology. This
subject arose through the study of the problems in mathematical analysis and geome-
try in Euclidean spaces, particularly, through Poincaré’s work in the classification of
algebraic surfaces. An important feature in the history of algebraic topology is that the
concepts initially introduced in homology and homotopy theories for applications to
problems of topology have found fruitful applications to other areas of mathematics
and have become the starting points of various theories: category theory, homologi-
cal algebra and K -theory are outstanding examples. The term “topology” was given
by J.B. Listing (1802–1882) in 1862 instead of previously used “Analysis situs”.
The subject “topology” was studied by C. Felix Klein (1849–1925) in his “Erlangen
Program” in 1872 and considered the invariants of arbitrary continuous transfor-
mation, a kind of geometry. He classified geometries by their underlying symmetry
groups, and this classification greatly influenced the synthesis of the mathematics.
J.W. Alexander (1888–1971) used ‘topological’ in the titles of his research papers in
the twenties.
The basic problem in algebraic topology is to devise ways to assign various alge-
braic objects such as groups, rings, modules to topological spaces and homomor-
phisms to the corresponding structures in a functorial way. The literature on alge-
braic topology is very vast. Properties and characteristics which are shared by home-
omorphic spaces are called topological properties and topological invariants; on the
other hand those by homotopy equivalent spaces are called homotopy properties and
homotopy invariants. The Euler characteristic invented by L. Euler (1707–1783) in
1752 is an integral invariant, which distinguishes non-homeomorphic spaces. The
search of other invariants has established connections between topology and modern
DOI 10.1007/978-81-322-2843-1_18
548 18 A Brief History of Algebraic Topology
algebra in such a way that homeomorphic spaces have isomorphic algebraic struc-
tures. Historically, the concept of fundamental group introduced by Poincaré in 1895
is the first important invariant of homotopy theory which came from such a search.
His work explained the difference between curves deformable to one another and
curves bounding a larger space. The first one led to the concepts of homotopy and
fundamental group and the second one led to homology. Some concepts studied
now in algebraic topology had been found in the work of B. Riemann (1826–1866),
C. Felix Klein and H. Poincaré. But the foundation of algebraic (combinatorial topol-
ogy) was laid in the decade beginning 1895 by H. Poincaré through publication of his
famous series of memoirs ‘Analysis Situs’ during the years 1895–1904. His moti-
vation was to solve specific problems involving paths, surfaces, and geometry in
Euclidean spaces. His vision of the key role of topology in all mathematical theories
began to materialize from 1920.
For this chapter the books and papers Adams (1960, 1962, 1972), Atiyah (1967),
Barratt (1955), Brown (1962), Dieudonné (1960), Dold and Thom (1958), Dold
(1962), Eilenberg and MacLane (1942, 1945a), Eilenberg and Steenrod (1952),
Freudenthal (1937), Hopf (1935), Hurewicz (1935), James (1999), Maunder (1970),
Poincaré (1895, 1900, 1904), Whitehead (1941, 1949, 1950), Whitehead (1950,
1953, 1962) and some others are referred in the Bibliography.
18.1 Poincaré and his Conjecture
H. Poincaré born in France is the first mathematician who systemically attacked

the problems of assigning topological invariants to topological spaces. He also first
introduced the basic concepts and invariants of combinatorial topology, such as Betti
numbers and the fundamental group. He proved a formula relating the number of
edges, vertices and faces of n-dimensional polyhedron (the Euler-Poincaré theo-
rem) and gave the first precise formulation of the intuitive notion of dimension. The
monumental work of Poincaré embodied in “Analysis situs”, Paris, 1895 introduced
the concepts of homotopy, fundamental group and homology. He is the first mathe-
matician who applied algebraic objects in homotopy theory. His work organized the
subject algebraic topology for the first time which has been discussed in earlier chap-
ters. At the beginning his work did not attract mathematicians to a great extent but his
promising work with his vision attracted mathematical community since 1920. This
subject is an interplay between topology and algebra and studies algebraic invariants
provided by homotopy and homology theories. The twentieth century witnessed its
greatest development.
Poincaré 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
an ellipse or even to an arbitrary closed curve, but it not equivalent to a straight line
18.1 Poincaré and his Conjecture 549
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.”
The idea of homotopy for the continuous maps of unit interval was originated by
C. Jordan (1838–1922) in 1866 and that of for loops was introduced by H. Poincaré
in 1895 to define an algebraic invariant called the fundamental group. H. Poincaré
may be regarded as the father of algebraic topology. The concept of fundamental
groups invented by H. Poincaré in 1895 conveys the first transition from topology to
algebra by assigning an algebraic structure on the set of relative homotopy classes
of loops in a functorial way. Fundamental group is a basic very powerful invariant in
algebraic topology and is the first of a series of algebraic invariants πn associated with
a topological space with a base point. Historically, the concept of fundamental group
introduced by Poincaré is the first important algebraic invariant of homotopy theory.
This group provides information about the basic shape, or holes, of the topological
space. His work in algebraic topology is mainly in geometric terms.
Poincaré posed in 1904 a conjecture which is called Poincaré conjecture. This
conjecture asks whether a simply connected compact n-manifold having the same
homology groups as S n is homeomorphic to S n ? It is not hard to show that the
conjecture is true for n = 2.
(i) For n = 4 the conjecture was proved to be true by M. Freedman (1951-) in 1982
and he was awarded the 1986 Fields medal for this work.
(ii) For n = 5 the conjecture was demonstrated by Christopher Zeeman (1925–
2016) in 1961.
(iii) For n = 6 the conjecture was proved to be true by John R. Stallings (1935–2008)
in 1962.
(iv) For n ≥ 7 the conjecture was established by Stephen Smale (1930-) in 1961. He
subsequently extended his proof for all n ≥ 5 independently. He was awarded
the Fields Medal in 1966 for this work.
(v) For n = 3, (its equivalent statement asserts that a compact 3-manifold homo-
topically equivalent to S 3 is homeomorphic to S 3 ) the problem has been solved
by Grigori Yakovlevich Perelman (1966-) in year 1994. He was offered the 2006
Fields Medal for his contributions to geometry and his revolutionary insights
into the analytical and geometric structure of the Ricci flow. But he declined
to accept the award or to appear at the Spain ICM 2006. The scientific journal
“Science” declared on 22nd December 2006 Perelman’s proof of the Poincaré
conjecture as the scientific “Breakthrough of the Year 2006”, such recognition
is possibly the first in the area of mathematics.
Poincaré also made significant contributions in algebra, differential equations, com-
plex analysis, algebraic geometry, celestial mechanics, mathematical physics, phi-
losophy of mathematics and popular science through his publication of 30 books
and over 500 papers. Of course, some of the ideas which Poincaré developed had of
course their origins prior to him, with L. Euler, and B. Riemann above all.
18.2 Early Development of Homotopy Theory
This section conveys the early development of homotopy theory arising 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. Topol-
ogists regard Poincaré as the founder and H. Hopf and Hurewicz as the cofounders
of algebraic topology. The concept of homotopy presents a mathematical formulation
of the intuitive idea of a continuous transition between two geometrical configura-
tions. On the other hand, the concept of homology presents a mathematical precision
to the intuitive idea of a curve bounding an ‘area’ or a surface bounding a ‘volume’.
Algebraic topology attempts to measure degrees of connectivity by using homology
and homotopy groups.
The concept of homotopy, at least for maps of the unit interval I was given by
C. Jordan in 1886. The word ‘homotopy’ was first given by Max Dehn (1878–1952)
and Paul Heegaard (1871–1948) in 1907. The Jordan Curve Theorem, a classical
theorem, was first stated by Jordan in 1892. His paper contained some gaps. Its first
rigorous proof given by Oswald Veblen (1880–1960) in 1905 is one of the greatest
developments of algebraic topology.
The importance of homotopy theory was realized in 1930 with the discovery of
the Hopf map with his striking result π3 (S 2 ) = 0. Prior to him homotopy theory
was used as a secondary tool for the homology theory except for the fundamental
group. Hopf fibering given by H. Hopf through his work during 1935–1941 plays an
important role in the study of algebraic topology. The Hopf maps p : S 2n−1 → S n for
n = 2, 4, 8 introduced by Hopf in 1935 are utilized to study certain homotopy groups
of spheres such as π3 (S 2 ) = 0, π7 (S 4 ) = 0 and π15 (S 8 ) = 0. Homotopy theory is
used in solving many of the old problems of c1assical topology. The fundamental
groups are deeply connected with covering spaces. Historically, the systemic study of
covering spaces appeared during the late 19th century and early 20th century through
the theory of Riemann surfaces. But its origin was found before the invention of the
fundamental groups by H. Poincaré in 1895. Poincaré introduced the concept of
universal covering spaces in 1883 to prove a theorem on analytic functions.
Some analytical geometric tools are required for development of algebraic
topology. These are: simplex, complex, subcomplex, simplicial map, triangula-
tion, polyhedron and simplicial approximation. Simplicial complexes introduced by
J.W. Alexander provide useful tools in computing fundamental groups of simple
compact spaces as well as for the study of manifolds. For example, Poincaré dual-
ity theorem is one of the earliest theorems in topology. Simplicial complexes form
building blocks of homology theory. The simplicial approximation theorem given by
Brouwer and Alexander around 1920 by utilizing a certain good feature of simplicial
complexes plays a key role in the study of homotopy and homology theories. The
combinatorial device, now called abstract complex was systematically used by W.
Mayer (1887–1947) in 1923.
18.2 Early Development of Homotopy Theory 551
It is easy to define algebraic invariants such as fundamental groups, higher homo-

topy groups and homology groups, etc., but difficult to compute them for different
classes of topological spaces. To facilitate such computation, Chap. 6 conveys a spe-
cial class of topological spaces that can be broken up into pieces which fit together
in a nice way. Such spaces are called triangulable spaces. The concept of triangu-
lation is also utilized to compute homology groups of triangulable spaces and to
solve extension problems. The concept of edge-group E(K , v) (which is isomorphic
to the fundamental group π1 (|K |, v) for any simplicial complex K ) is applied to
graph theory. For example, van Kampen theorem for fundamental groups given by
E. van Kampen (1908–1942) in 1933 is proved by using graph-theoretic results.
The extension of ‘Euler characteristic’ was given by A.L. Cauchy (1789–1857)
in 1813 and H. Poincaré in 1895. The study of 1-dimensional and 2-dimensional
simplicial complexes dates back at least to Euler and that of higher-dimensions
first appeared in the work of J. Listing (1808–1882) in 1862. Listing is the first
mathematician who used the word ‘topology’ in his famous article published in
1847, although he had used the term in some of his previous correspondence.
Except for fundamental group, the subject of homotopy was first found in the work
of L.E.J. Brouwer who made the first step in 1912 towards connecting homology and
homotopy groups of certain spaces which are the two basic concepts of algebraic
topology by showing 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 (that is, 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 view point 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 celebrated results of Brouwer include the proof of the
topological invariance of dimension of Rn , Brouwer fixed point theorem (the theorem
is named after his name) and the simplicial approximation theorem. His results are
important in the foundations of algebraic topology.
After having generalized Brouwer’s result of degree mapping of a continuous map
of a two-dimensional sphere into itself to an arbitrary dimension, H. Hopf made a
systematic study of the problem of classifying the continuous maps of polytopes (a
polytope is the union of finitely many simplices, with the additional property that, for
any two simplices that have a nonempty intersection, their intersection is a vertex,
edge, or higher-dimensional face of the two). His work is mainly based on highly
geometric intuitions like Brouwer. He continued the work of Brouwer by using the
degree mapping and the homotopy class of a map as tools. Hopf characterized homo-
topy class of self maps of a sphere by their degrees. His definition of the degree of a
spherical map f : S n → S n is more intuitive than its definition from the view point of
homology theory. His definition shows that if f : S n → S n , n ≥ 1 is a constant map,
then deg f = 0 and if f : S n → S n , n ≥ 1 is the identity map, then deg f = +1.
He proved its homotopy invariance and gave a complete homotopy classification of

mappings of n-dimensional polytopes into the n-dimensional sphere S n .
W. Hurewicz made remarkable contributions to algebraic topology. His inven-
tion of the higher homotopy groups πn in 1935–1936 is a natural generalization of
the fundamental group to higher-dimensional analogue of the fundamental group.
πn is a sequence of functors given by W. Hurewicz in 1935 from topology to alge-
bra by extending the concept of fundamental group. Historically, Hurewicz intro-
duced higher homotopy groups by defining a group structure in πn (X ) = [S n , X ].
He showed that the higher homotopy groups πn (X ) are all abelian for n ≥ 2 though
π1 (X ) is in general not abelian. By an action of π1 on πn it is also proved that for a
simply connected space X , the group πn (X, x 0 ) ∼ = [S n , X ].
Hurewicz establishes a close connection between homotopy and homology groups
of a certain class of spaces through Hurewicz homomorphism defined by Hurewicz
in 1935 in his paper Hurewicz (1935). He first asserted that for a simplicial pair
(K , L) if πr (K , L) = 0 for 1 ≤ r < n(n ≥ 2), then πn (K , L) → Hn (K , L) is an
isomorphism. This work cast light for the first time onto the relationship between
homological and homotopical invariants. A series of four papers of Hurewicz pub-
lished during 1935–1936, has greatly influenced the development of the modern
homotopy theory. More precisely, Hurewicz is the first mathematician who estab-
lished a connection in 1935 between homology groups Hn (X ) and homotopy groups
πn (X ) for (n − 1)-connected spaces X , when n ≥ 2. by defining group homomor-
phisms h : πn (X ) → Hn (X ) for all n. This result known as Hurewicz theorem asserts
that the first nonzero homotopy groups πn (X ) of a simply connected space X is iso-
morphic to the first nonzero homology group of Hn (X ) for n ≥ 2. Their relative
version is also similar. The one-dimensional case of this theorem was already given
by Poincaré, who established that the homology relation appears after abelianiz-
ing the homotopy relation. His celebrated result is the “Equivalence theorem”: if
a space X has homotopy groups πn (X ) = 0 for 1 ≤ n ≤ n 0 for some n 0 > 1, then
Hn (X ) = 0 for 1 ≤ n ≤ n 0 and πn 0 +1 (X ) ∼
= Hn 0 +1 (X ). He also introduced the con-
cepts of homotopy equivalence and homotopy equivalent spaces.
For the study of spaces X of low dimension, the fundamental group π1 (X ) is very
useful. But it needs refined tools for the study of higher-dimensional spaces. For
example, fundamental group can not distinguish spheres S n with n ≥ 2. Such limita-
tion of low dimension can be removed by considering the natural higher-dimensional
analogues of π1 (X ) given by Hurewicz. His another invention is the long exact
homotopy sequence for fibrations in 1941, and the fundamental theorem, known as
Hurewicz theorem which connects homotopy and homology groups in 1935. His
work led to homological algebra. The homotopy extension property (HEP) and its
dual the homotopy lifting property (HLP) play critical role in homotopy theory.
There are various notions of fibrations in algebraic topolgy but the work of Hurewicz
shows that a fiber map is simply a continuous map which has the HLP for arbitrary
topological spaces.
18.2 Early Development of Homotopy Theory 553
The homotopy equivalent relation is much coarser than the relation of homeo-
morphism and hence is more accessible to classification. For example, the disk D n
is of the same homotopy type of a single point { p} ⊂ D n but D n is not homeo-
morphic to { p}. The higher homotopy groups and homology groups are invariants
of the the homotopy equivalence class of a space. This concept has offered a new
foundation for the development of combinatorial invariants of spaces and manifolds.
His contributions made breakthrough in the field of topology.
By a synthesis of the work of Hopf and Hurewicz, H. Freudenthal proved the
completeness of Hopf ’s classification and discovered the suspension map in around
1935. Since then the study of homotopy of spheres comes up a challenging field
of research of many topologists including Hopf. Freudenthal made a breakthrough
in algebraic topology by establishing a theorem in 1937 known as Freudenthal sus-
pension theorem while investigating the homotopy groups πm (S n ) for 0 < m < n.
J.H.C. Whitehead (1904–1960) introduced the concept of simple homotopy theory,
which has developed through algebraic K -theory.
Freudenthal also studied the nth cohomotopy set π n (X, A) on which K. Borsuk
(1905–1982) endowed in 1936 the abelian group structure under certain conditions
on (X, A) (Borsuk 1936). For each integer p > n+1 2
, the cohomotopy groups of a
compact pair (X, A) of finite dimension n, the pth cohomotopy group π p (X, A)
is defined by π p (X, A) = [(X, A), (S p , s0 )], which is an abelian group. The set
π p (X, A) is defined for p ≥ 0, but addition operation in it is defined for p > n+1
2
. If
f : (X, A) → (Y, B) is continuous, it induces maps
f ∗ : π p (Y, B) → π p (X, A)
for all p, which are homomorphisms when both sides are groups. Again a map-
ping δ : π p−1 (A) → π p (X ) is defined for p > 0, which is a homomorphism when
both sides are groups. The basic difference between cohomology and cohomotopy
is the lack of a group structure in π p (X, A) for p ≤ n+1 2
. So it is not possible to
compute cohomotopy group by an induction on p starting with p = 0. The coho-
motopy groups defined by K Borsuk in 1936 was deeply studied by Spanier in his
paper Spanier (1949). Spanier has shown that with the induced homomorphism and
the coboundary operator, the cohomotopy groups satisfy all the Eilenberg–Steenrod
axioms for cohomology theory, and emphasized the importance of the cohomotopy
groups. His investigation, however, has been restricted to the case of compact spaces.
Lens spaces defined by H. Tietze (1880–1964) in 1908 form an important
class of 3-manifolds in the study of their homotopy classification. Subsequently,
K. Reidemeister (1893–1971) established their topological classification in 1935
and J.W.C. Whitehead gave their homotopical classification in 1941. Tietze gave
a finite presentation for the fundamental group and proved the topological invari-
ant of fundamental groups. He also contributed to the development of knot theory,
Jordan curves, cell complexes and even general topology which has now entered in
the premises of analysis.
18.3 Category Theory and C W -Complexes
This section conveys the concepts of category theory and C W -complexes which
play a key role in the development of algebraic topology. Category theory is
very important in mathematics to unify different concepts in mathematics. It con-
veys a key language in algebraic topology. The concepts of categories, functors,
natural transformations and duality introduced and studied during 1942–1945 by
S. Eilenberg (1913–1998) and S. Mac-Lane (1909–2005) form category theory
(Eilenberg and MacLane 1942, 1945). Originally, the purpose of these notions was
to provide a technique for classifying certain concepts such as that of natural iso-
morphism. The whole idea of category theory arose through the field of algebraic
topology. The first and the simplest realization of this idea is the fundamental group
(or Poincaré group) of a pointed space. Many concepts of algebraic topology are
unified and explained by category theory, and it plays a key role for the study of
homotopy, homology and cohomology theories which constitute the basic text of
this book, in addition to adjoint functor, representable functor, abelianization functor,
Brown functor, and infinite symmetric product functor which are important functors
in the study of algebraic topology.
J.H.C. Whitehead constructed a new category in 1949 in his paper Whitehead
(1949), which is now called the category of C W -complexes. The concept of C W -
complexes is a natural generalization of the concept of polyhedra, where cells are
attached by arbitrary continuous maps starting with a discrete set, whose points are
each regarded as a 0-cell. C W -complexes built up by successive adjunctions of cells
of dimensions 1, 2, 3, . . . . C W -complexes form an extensive class of topological
spaces for the study of algebraic topology, where a weak equivalence is necessar-
ily a homotopy equivalence. C W -complexes carry excellent combinatorial proper-
ties which are flexible than simplicial complexes. The existence of Eilenberg and
MacLane spaces was shown by J.H.C. Whitehead in 1949 by using the properties of
C W -complexes.
C W -complexes give a convenient setting for homotopy theory. It is surprising
that the homotopy groups of C W -complexes supply a vast information. For example,
Whitehead theorem asserts that if a continuous map f : X → Y between connected
C W -complexes induces isomorphisms f ∗ : πn (X ) → πn (Y ) for all n, then f is a
homotopy equivalence. Again every space is not a C W -complex but for many pur-
poses it is sufficient to consider only C W -complexes. This conclusion follows from a
theorem of Whitehead established in 1950 that says that given any topological space
X , there exists a C W -complex K and a weak homotopy equivalence f : K → X .
The cellular approximation theorem is an analogue result of simplicial approximation
theorem.
The origin of C W -complexes closely relates to the birth of many concepts and
development of algebraic topology in general. There are many advantages of C W -
complexes over polyhedra. Since all C W -complexes are paracompact and all open
coverings of a paracompact space are numerable, the results on the homotopy clas-
sification of principal G-bundles hold for locally trivial principal G-bundles over a
18.3 Category Theory and C W -Complexes 555
C W -complex. Algebraic topologists feel that the category of C W -complexes is a

good category for homotopy and homology theories.
J.H.C. Whitehead defined a product between two groups in 1941 to study homo-
topy groups of pointed topological spaces X . This product associates with elements
α ∈ π p (X, x0 ) and β ∈ πq (X, x 0 ) an element [α, β] ∈ π p+q−1 (X ), called Whitehead
product in his honor. This product provides a technique at least in some cases for
constructing nonzero elements of π p+q−1 (X ). He also defined generalized products
involving the rotation groups. Whitehead product is used to solve several problems
proving many amazing results in algebraic topology.
18.4 Early Development of Homology Theory
This section conveys early development of homology theory starting from sim-
plicial homology defining the homology groups of a polyhedron by Poincaré in
1895, followed by several generalizations of his homology beyond polyhedra during
the period 1927–1944. The extensions were made by several topologists such as
L. Vietoris (1891–2002) in 1927 (for compact metric spaces), E. Čech (1893–
1960) in 1932 (for compact Hausdorff spaces), S. Lefschetz (1884–1972) in 1933,
S. Eilenberg in 1944 (for arbitrary topological spaces). Singular homology of an
arbitrary topological space constructed by Eilenberg in 1944 is the most powerful
homology.
Homology theory plays a key role in algebraic topology. The basic tools such as
complexes and incidence numbers are necessary for constructing simplicial homol-
ogy groups as defined by Poincaré. To inaugurate a homology theory, Poincaré started
in 1895 with a geometric object (a space) which is given by combinatorial data (a
simplicial complex). Then the linear algebra and boundary relations by these data
are used to construct homology groups, called simplicial homology groups. This the-
ory stemmed from his ‘Analysis Situs’. Using these tools Poincaré defined directly
the Betti numbers invented by E. Betti (1823–1892) and torsion numbers which are
numerical invariants and characterized the homology groups based on the coefficient
group Z of integers. The concept of relative homology (modulo a subcomplex) was
given by Lefschetz in 1927 and the operator ∂ was used by Lefschetz. Attention
for shift from numerical invariants to groups associated with homology theories was
successfully made during the period 1925–1935. This shift is partly due to Emmy
Noether (1882–1935). The algebraic approach of Noether to homology is a fruitful
contribution to the geometrical approach of Poincaré. Inspired by the above approach
of Noether, P. Alexandroff (1896–1982) and H. Hopf gave jointly the first detailed
study of homology theory from the view point of algebra in 1935.
There are two directions of generalizations of simplicial homology invented by
Poincaré in 1895:
(i) from complexes to more general spaces where the homology groups are not
characterized by numerical invariants;
(ii) from the group Z to arbitrary abelian groups.
Several homology theories are constructed other than homology invented by

Henry Poincaré in 1895. They include
(i) Homology groups for compact metric spaces introduced by L. Vietoris (1891–
2002) in 1927;
(ii) Homology groups for compact Hausdorff spaces introduced by E.Čech in 1932;
(iii) Singular homology groups are first defined by S. Lefschitz in 1933.
(iv) Cellular homology groups for C W -complexes (see Sect. 10.14).
All these homology theories and their dual theories called cohomology theories lived
in isolation in the sense that their interrelations were not established for a long time.
Cohomology is dual to homology and it arises from the algebraic dualization of
the construction of homology. The origin of the concept of cohomology groups is
the duality theorem given J.W. Alexander in 1935. Cohomology theory given by
S. Lefschetz in 1930 was further developed by J.W. Alexander in 1936, H. Whitney
in 1938 and Lefschetz himself in 1942. L. Pontryagin (1908–1988) proved in 1934
the complete group invariant form of the duality theorem in his paper Pontryagin
(1934). Alexander gave the first formal definition of the cohomology groups in 1936
at the Moscow conference.
The cohomotopy groups defined by K. Borsuk in 1936 resemble to cohomology
groups in some sense. E. Spanier (1921–1996) deeply studied cohomotopy groups in
his paper (Spanier 1949) and has shown that the cohomotopy groups satisfy analogues
of all the cohomology axioms under situations when they are meaningful.
Simplicial cohomology constructed by J.W. Alexander and A. Kolmogoroff
(1903–1987) in 1935 was developed by E. Čech and H. Whitney during 1935–1940.
Čech cohomology is a cohomology theory based on the intersection properties of
open covers of a topological space. It is named after E.Čech. De Rham cohomolgy
defined for smooth manifolds has many deep results including direct relationships
to solutions of differential equations on manifolds. The homology and cohomology
groups for C W -complexes can be directly calculated from the cellular structure like
simplicial structure in the simplicial homology and cohomology groups of a polyhe-
dron and they provide the most useful tools in algebraic topology. Cohomology has
a multiplicative structure making it a ring (algebra). This advantage of cohomolgy
over homology facilitates more development of cohomology than homology.
Historically, homology theory was first applied to manifolds by Poincaré, giving
a result known as Poincaré duality theorem. He first stated this theorem in 1895 in
terms of Betti numbers. At that time there was no concept of cohomology, which
was invented in 1930s. Poincaré stated that the m-th and (n − m) th Betti numbers
of a closed (i.e., compact and without boundary) orientable n-manifold are equal.
But there are at present its different forms.
Lefschetz number is an important concept introduced by Lefschetz in 1923. It is a
number associated with each continuous map f : |K | → |K | from a polyhedron into
18.4 Early Development of Homology Theory 557
itself and the number is denoted by f . It is also related to the Euler characteristic
formula and proves a powerful fixed point theorem known as Lefschetz fixed point
theorem, which is a classical application of homology and generalizes Brouwer fixed
point theorem. Moreover, some other results on fixed points follow as its applications.
The proof of homotopy type invariance of homology was given by J.W. Alexander
in 1915 and 1926 and by O. Veblen in 1922 through their work in terms of simplicial
homology groups of a polyhedron. The concept of induced homomorphisms f ∗ was
used since the time of Poincaré but it had neither any name nor any status for at least
next 35 years. On default of formal recognition of boundary operator ∂ and f ∗ for
such period the homology groups earned no formal status for a long period. The first
formal recognition of the homology sequence and its exactness was found in 1941
in the work of Hurewicz.
The excision property does not hold in general for homotopy groups. This failure
makes homotopy groups so much harder for computing than homology groups. How-
ever, Fredenthal suspension theorem shows that in some special cases there is a range
of dimensions in which excision property holds. This leads to the concept of stable
homotopy groups, which begins with stable homotopy theory. Computation of these
groups even for simple spaces is a difficult problem. An interesting conjecture posed
by Freyd also seems to be very hard (Gray, pp. 145). The stable homotopy groups of
spheres are fundamental objects in algebraic topology and attempts are going on for
their calculation. Stable homotopy groups πn s (X ) define a reduced homology theory
on the category of pointed C W -complexes.
Higher homotopy groups have certain similarities with homology groups. For
example, πn (X ) are always abelian for n ≥ 2 and there are relative homotopy groups
which give a long exact sequence like long exact sequence of homology groups. The
higher homotopy groups are easier to define but harder to compute than either homol-
ogy groups or fundamental groups essentially due to absence in higher homotopy
groups the excision property for homology and also absence in higher homotopy
groups a theorem analogous to van Kampen theorem for fundamental group. In spite
of these computational difficulties, homotopy groups are of great importance. For
example, Whitehead theorem given by J.H.C. Whitehead which says that a contin-
uous map between C W -complexes which induce isomorphisms on all homotopy
groups is a homotopy equivalence. The homological version of Whitehead theorem
is similar.
18.5 Hopf Invariant
This section presents Hopf invariant which is an important concept invented by

Hopf to solve the problem when a map f : S m → S n for m > n > 1 is necessarily
nullhomotopic. This problem was resolved by Hopf with the discovery of his famous
map f : S 3 → S 2 . Hopf in his celebrated paper (Hopf 1931) studied the space of
homotopically nontrivial continuous mappings of spheres: S 3 → S 2 . He showed
in 1931 that π3 (S 2 , s0 ) is nonzero. He later solved the general problem when a

continuous mapping f : S m → S n for m > n > 1 is necessarily nullhomotopic.
The basic problem which led to the discovery of homotopy groups was to classify
homotopically the maps of an n-sphere S n into a given space. Hopf introduced
the concept of an invariant H ( f ) of f , now called Hopf invariant of f , which
depends only on the homotopy class of f . Hopf invariant for certain class of mappings
and Hopf group (which is a generalization of topological group) are two important
inventions of H. Hopf. It is proved that two continuous maps from S 3 to S 2 are
homotopic iff they have the same Hopf invariant. Hopf developed vector field theory.
His work has earned a permanent place in the history of algebraic topology.
The main thrust in homotopy theory appears to centralize on the problems of
determining homotopy groups of spheres. The basic tools used in this search are of
algebraical nature, like “generalized Hopf invariants” studied by G. Whitehead, or
“cup products” introduced by N. Steenrod. Homology is used to define the degree of
spherical maps f : S n → S n which distinguish different homotopy classes of maps
f . Cup products can be used something similar for maps f : S 2n−1 → S n , i.e., for
elements of π2n−1 (S n ), Hopf did this by using more geometric constructions prior to
the invention of cohomolgy and cup products. The Hopf invariant H ( f ) defined now
by using cup product is as one of the most remarkable applications of cup product
in topology.
The problem for which n there exists a continuous map f : S 2n−1 → S n with Hopf
invariant H ( f ) = 1 was solved by Adams (Adams 1960, 1962). This theorem, also
called Adams theorem, is a deep theorem in homotopy theory which states that
the only n-spheres which are H -spaces are S 0 , S 1 , S 3 , and S 7 . This relates to the
existence of division algebra structure on Euclidean space Rn . The real numbers,
complex numbers, the real quaternions, and the Cayley numbers are examples of real
division algebras. J.F. Adams proves that there are no other examples. Corresponding
to each continuous map f : S 2n−1 → S n one can associate an integer H ( f ) its Hopf
invariant. This means that to each element of π2n−1 (S n ), one can assign an integer
which is its Hopf invariant. If π2n−1 (S n ) contains no element of Hopf invariant one,
then there is no real division algebra of dimension n. Adams theorem shows that
such elements exist precisely for n = 1, 2, 4, 8, whose simpler proof is given by
Atiyah (1967) by using K -theory introduced by M.F. Atiyah (1929-) and F.E. Peter
Hirzebruch (1927–2012) in 1961.
18.6 Eilenberg and Steenrod Axioms
This section presents the axiomatic approach of homology and cohomolgy theories
given by S. Eilenberg and N.E. Steenrod (1910–1971) in 1945 (Eilenberg and Steen-
rod 1945) as axioms to characterize a theory of homology and cohomology (see
Chap. 12). The usual approach to homology arises through the complicated notion
of a complex. Many of the ideas used in constructions, such as orientation, chain
and algebraic boundary seem to be artificial (see Chap. 10). The motivation of these
18.6 Eilenberg and Steenrod Axioms 559
concepts appears only in retrospect. Several homology theories were constructed

during 1927–1933 which are different from simplicial homology theory invented by
Poincaré in 1895. Since their constructions are complicated and different, for greater
logical simplicity, algebraic topologists started around 1940 comparing various def-
initions of homology and cohomology given in the previous years. The construction
of a homology theory and proofs of its main properties are extremely complicated.
To avoid these problems, S. Eilenberg and N.E. Steenrod initiated a new approach
in 1945 by taking a small number of their properties (not focusing on machinery
used for construction of homology and cohomology groups) as axioms to charac-
terize a theory of homology and cohomology. This axiomatic approach has greatly
influenced later developments of algebraic topology. The axioms reveal that the first
six axioms carry a very general character, while the seventh axiom, which is the
“Dimension Axiom” is very specific. There exist many such theories such as stable
homotopy, various K -theories and bordism theories. The author of the present book
has constructed a new generalized cohomology theory (see Chap. 15).
This axiomatic approach given by a set of seven axioms of S. Eilenberg and N.
Steenrod, announced in 1945 and published in their book in 1952 (Eilenberg and
Steenrod 1952) with the proof of their uniqueness is the most important contribu-
tion to algebraic topology since the invention of the homology groups by Poincaré.
The uniqueness theorem of homology theory asserts that any two homology groups
having the same coefficient group coincide on finite C W -complexes, which implies
that the homology groups of finite C W -complexes are completely determined by the
coefficient group, and hence are computable from the axioms. This approach classi-
fies and unifies different homology groups on the category of compact triangulated
spaces. An analogous approach given by them also inaugurated its dual theory called
cohomology theories.
The exactness of the homology sequence of a pair of topological spaces was
formalized by Eilenberg and Steenrod in 1945 while giving “Axiomatic Approach
to Homology Theory”. On the other hand, the form of the Mayer–Vietoris sequence
(exact) of a triad which is now used was also given by Eilenberg and Steenrod in
1952, although formulae for the homology groups of the union of two polyhedra
were prescribed by W. Mayer (1887–1948) in 1929 and L. Vietoris in 1930.
18.7 Fiber Bundle, Vector Bundle, and K -Theory
This section conveys the early development of fiber bundles, vector bundles and
K -theory. Fiber bundles and vector bundles are special bundles with additional struc-
ture and are closely related to the homotopy theory. The recognition of bundles in
mathematics was realized during 1935–1940 through the work of H. Whitney, H.
Hopf and E. Stiefel and some others. Since then the subject has created a general
interest. There is a link-up between the study of vector bundles and homotopy the-
ory. The K -theory studied in 1959 by Atiyah and Hirzebruch (1959) connects vector
bundles with homotopy theory and is a generalized cohomolgy theory.
One of the most important notions in topology is the notion of fiber spaces which is
the most fruitful generalization of covering spaces. Although this notion had appeared
in the literature before 1955, the definition introduced by Hurewicz in 1955 (Hurewicz
1955) is much more general and useful. The concept of fiber bundles arose through
some problems in topology and geometry of manifolds around 1930. Its first general
definition was given by H.Whitney. His work and that of H. Hopf, E. Stiefel (1909–
1978), J. Feldbau (1914–1945), and many others displayed the importance of the
subject for the application of topology to different areas of mathematics and to other
fields also (see Chaps. 14 and 17). This subject also marks a return of algebraic
topology to its origin.
Covering spaces provide tools to study the fundamental groups. Fiber bundles
provide likewise tools to study higher homotopy groups (which are generalization of
fundamental groups). 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.
The concept of fiber bundle arose through some problems in topology and geom-
etry of manifolds around 1930. Fiber bundles form a nice class of maps in topology,
and many naturally emerging maps are fiber bundles. Fiber bundles are fibrations
and fibrations are a natural class of maps in algebraic topology. The notion of fiber
bundles plays a central role to study spaces up to homotopy. A fiber bundle is a
bundle with an additional structure derived from the action of a topological group on
the fibers. On the other hand, a vector bundle is a bundle with an additional vector
space structure on each fiber.
The concept of a vector bundle came from the study of tangent vector fields
to smooth manifolds, such as spheres, projective spaces etc. A fiber bundle is a
locally trivial fibration and has covering homotopy property. Theory of fiber bundles
including classifying theorem, with a special attention to vector bundles with fibers
of different dimensions and K -theory (which is generalized cohomology theory)
interlinks vector bundles with homotopy theory.
The concept of fibration plays a key role in the study of homotopy theory, which
appeared implicitly in 1937 in the work of Borsuk but explicitly in the work of
Whiteney during 1935–1940, first on sphere bundles. This concept led to general
fiber bundles. Hurewicz and Steenrod made the first attempt in 1940 to formulate
the homotopy-theoretic properties latent in the notion of fiber bundles and gave a set
of sufficient conditions to establish that a large class of homotopy lifting problems
always has a solution. More precisely, if p : X → B is a continuous map, the condi-
tion for a homotopy lifting problem consists of a map f : Y → X and a homotopy
G : Y × I → B of its projection p ◦ f . A solution of this problem is a homotopy
H : Y × I → X of f such that p ◦ H = G.
J.P. Serre (1926-) studied fibrations and showed in 1950 that a continuous map
to be a fibration iff every homotopy lifting problem with X a finite complex has a
solution. This result characterizes a map to be a fibration and may be considered as a
definition of a fibration. Hurewicz modified Serre’s definition in 1955 by removing
18.7 Fiber Bundle, Vector Bundle, and K -Theory 561
all restrictions on X . Hurewicz established that the projection of every fiber bundle
with paracompact base space is in particular a fibration according to his definition.
J.W. Milnor (1931-) invented a new method in 1956 for giving a classifying
space and a universal principal fiber space associated with principal fiber bundle.
The most celebrated published work of Milnor is his proof in 1956 of the existence
of 7-dimensional sphere with nonstandard differential structure. He constructed a
universal fiber bundle for any topological group G and homotopy classification of
principal G-bundles. The relations between G and a classifying space BG can be
readily displayed using a geometric analogue of the resolution of homological alge-
bra. The above homotopy classification of vector bundles, Milnor’s construction of
a universal fiber bundle for any topological group G with homotopy classification
of numerable principal G-bundles and corresponding to the set of the isomorphism
classes of F-vector bundles over a paracompact space B, the group K F (B) called
the K -theory introduced by M.F. Atiyah (1929-) and Hirzebruch in 1961 are very
powerful results.
18.8 Eilenberg–MacLane Spaces and Cohomology

Operations
This section conveys the concept of Eilenberg and MacLane spaces introduced by
S. Eilenberg and S. MacLane during 1942–1943 which plays a central role in alge-
braic topology. The importance of Eilenberg–MacLane spaces is twofold. First, they
are important in homotopy theory. Second, they are closely linked with the study
of cohomology operations (invented by Serre). They carry close connection with
cohomology. The cohomology classes of a C W -complex have a bijective corre-
spondence with the homotopy classes of continuous maps from the complex into an
Eilenberg–MacLane space. This gives a strict homotopy-theoretic interpretation of
cohomology. In this sense cohomology groups may be considered ‘dual’ to homotopy
groups for C W -complexes. Moreover, every topological space has the homotopy type
of an iterated fibration of Eilenberg–MacLane spaces (called a Postnikov system).
Given an abelian group G and an integer n > 0, Eilenberg and MacLane con-
structed a space X = K (G, n) with nth homotopy group G and all other homotopy
groups vanish. Such spaces K (G, n) are now called an Eilenberg–MacLane spaces.
The homotopy sets [X, Y ] were first systematically studied by Barratt in 1955 while
studying ‘Track groups’. Eilenberg and MacLane studied the homological and coho-
mological structures of the complex K (G, n). These complexes were defined in a
purely algebraic fashion for every abelian group G and any integer n = 1, 2, . . .. The
topological significance of these complexes X = K (G, n) are on the fact that homo-
topy groups πn (X ) ∼ = G and πi (X ) = 0 for i = n. There are many other important
topological and also algebraic applications of these complexes.
The concept of cohomology operations introduced by Serre is a natural transfor-
mation of functors

H n ( ; G) → H n (−; G ).
Steenrod defined operations from one cohomology group to another (the so-called
Steenrod squares) that generalized the cup product. The additional structure made
cohomology a finer invariant. More precisely, Steenrod defined a family of new
operations Sqi : H n ( ; Z2 ) → H n+1 ( ; Z2 ) which is a sequence of operations, one
for each dimension, and behaves well with respect to suspension and they are the
components of a stable operation. These operations form a (noncommutative) algebra
under composition, known as the Steenrod algebra.
The method of cohomology operations can be used to study homology groups of
spheres. Let α ∈ πn (S m ) and ψ : H m ( ; G) → H n+1 ( ; L) be a cohomology opera-
tion. The mapping cone C α of α is a complex with one 0-cell, the base point (may
be ignored), one m-cell, and one (n + 1)-cell. We say that α is determined by ψ iff
the operator
ψ : H m (Cα ; G) → H n+1 (Cα ; L)
is nonzero. For example, if α is a Hopf map, then Cα is CP2 and the operation
Sq2 : H 2 → H 4 is the cup square, which is nonzero in Cα .
18.9 Generalized Homology and Cohomology Theories
This section presents certain functors which satisfy all the axioms of Eilenberg and
Steenrod for homology (resp. cohomology) with the exception of dimension axiom.
The theory of such functors is known as the generalized (or extraordinary) homology
(resp. cohomology). These theories first appeared in print in 1952 (Eilenberg and
Steenrod 1952). Several such functors have been found to be very useful. For example,
K -theory, various forms of bordism and cobordism theories, stable homotopy and
cohomotopy theories are their outstanding examples.
Around 1959 several algebraic topologists, working in different directions, con-
sidered systems of covariant functors
h n : C0 → Ab
(resp. contravariant functors)

h n : C0 → Ab
from the category C0 whose objects are pointed topological spaces having the homo-
topy type of pointed finite C W -complexes and morphisms are maps of such spaces
to the category Ab of abelian groups and their homomorphisms. These functors sat-
isfy all the axioms of Eilenberg and N. Steenrod for homology (resp. cohomology)
with exception of dimension axiom. The notions initially introduced in homology
and homotopy theories for applications to problems of topology have found fruitful
18.9 Generalized Homology and Cohomology Theories 563
applications to other parts of mathematics such as algebra, analysis, geometry, graph

theory. Homological algebra and K -theory are their outstanding examples. Among
the various homology theories, ordinary homology theory H∗ is the most useful, it is
usually much easier in most of the cases, to compute the ordinary homology groups
of a given space X than computing h ∗ (X ) for some other homology theory h ∗ . The
first step to computing h ∗ (X ) usually consists of computing H∗ (X ). In this sense,
ordinary homology theory is the fundamental homology theory.
18.10 -Spectrum and Associated Cohomology Theories
This section conveys the concept of -spectrum and its associated cohomology
theories. The notion of spectrum introduced by F. L. Lima (1929-) in 1958 has proved
to be very useful.
A spectrum A = { An , αn } in C0 is a sequence {An } of spaces in C0 together with
a sequence of continuous maps
αn : An → An+1 in C0
(equivalently,
αn : An → An+1 in C0 ).
and it is said to be an -spectrum if αn : An → An+1 , n ∈ Z is a base point pre-

serving weak homotopy equivalence for every integer n. There is a special sequence
{An } of spaces, An = K (G, n) together with a sequence {αn } of homotopy equiva-
lences, αn : An → An+1 , relating the cohomology groups of Eilenberg–MacLane
spaces to the homotopy groups of spaces by the relation H n (X ; G) = [X, K (G, n)],
the homotopy classes of continuous maps from X to K (G, n), which admits a natural
group structure. For example, the spectrum A = { An , αn } given by An = K (Z, n)
and αn : K (Z, n) → K (Z, n + 1), a base point preserving weak homotopy equiva-
lence, is called an Eilenberg–MacLane spectrum. In general, the Eilenberg–MacLane
spectrum A = { An , αn }, denoted by H G, is defined by taking An = K (G, n), where
K (G, n) is an Eilenberg–MacLane space of type (G, n).
Each spectrum A={An , αn } produces two different sequences of functors h n (; , A)
and h n (; A) from the category C0 to the category Ab, the first one is covariant and the
second one is contravariant. These are called the spectral homology and cohomology
functors associated with the spectrum A, and in brief abbreviated An and An .
The homology theory associated with a spectrum was first defined by
G.W. Whitehead (1918–2004) in 1962, now called a spectral homology theory asso-
ciated with spectrum. Its dual theory is called a spectral cohomology theory. For
example, the Eilenberg–MacLane space K (G, n) form a spectrum K (G), such that its
associated cohomology group H n (Y ; K (G)) ∼ = H n (Y ; G) for a finite C W -complex
Y . In particular, as the infinite symmetric product S P ∞ (S n ) of the n-sphere S n is
the Eilenberg–MacLane space K (Z, n), it follows that H n (X, Z) = [X, K (Z, n)].
Using this fact the author of the book has constructed in Chap. 15 a new -spectrum
A, generalizing the Eilenberg–MacLane spectrum K (G, n) and also constructed its

associated cohomology theory h ∗ ( ; A) which generalizes the ordinary cohomology
theory of Eilenberg and Steenrod.
Instead of defining products axiomatically, G W Whitehead defined products
directly from the -spectrum in 1962. For example, for a finite C W -complex X ,
if X = {X n , f n }, where X n = n X and f n is the natural homeomorphism X n →
X n+1 , then it is a spectrum and if Y is a complex, let H n (Y ; X ) be the distinct limit
of the groups [ m Y, X n+m ] under the composite maps
E fn∗
[ m Y, X n+m ] −−−→ [ m+1 Y, X n+m ] −−−−→ [ m+1 Y, X n+m+1 ].
The functors H n ( ; X) behave very much like cohomology group; indeed, they
satisfy the Eilenberg–Steenrod axioms with the exception of the ‘Dimension Axiom’,
which says that the cohomology groups of point vanish except in dimension zero.
There are more interesting examples of cohomology theories derived from spec-
trum: for example, if S is the suspension functor, the cohomolgy theory associated this
spectrum is H n (Y ; S) which is just the stable cohomotopy groups. Other important
cohomology theories are various Bordism and K -theories. Atiyah and Hirzebruch
made a study of the group K (X ) in 1959 from the category of complex vector bun-
dles over a finite dimensional C W -complex. They developed their study of K (X )
in 1961 into a generalized cohomology. K -theory carries many similarities to ordi-
nary cohomology theory and plays a key role in many areas of mathematics such as
modern algebra and number theory.
18.11 Brown Representability Theorem
This section presents a surprising theorem proved by E.H. Brown (1926-) in 1962,
now known as Brown representability theorem. This theorem solves the problem:
every -spectrum defines a cohomology theory. Is its converse true? Brown proves
all cohomology theories on the category of C W -complexes arise from -spectra.
This theorem relates homotopy theory with generalized cohomology theory and
plays a key role in the applications of homotopy theory to other areas. Brown proved
that under certain conditions, any cohomology theory satisfying Eilenberg–Steenrod
axioms can be obtained in the form [ , Y ] for some suitable space Y . Brown rep-
resentability theorem presents necessary and sufficient conditions under which a
contravariant functor on X has the form [X,Y] for some fixed Y . This shows that
there is a close relation between generalized cohomology theory and homotopy the-
ory, which plays a key role in the later development of algebraic topology.
More precisely, E.H. Brown proved in his paper Brown (1962) that, if H satisfies
certain axioms, there is a space Y , unique up to homotopy type, such that H is
naturally equivalent to the functor which assigns to each C W -Complex X with base
point the set of homotopy classes of maps of X into Y . Thus Brown representability
18.11 Brown Representability Theorem 565
theorem asserts that there exist connected C W -complexes An with base point and
natural equivalences
h n (X ) ∼
= [X ; An ],
where X runs over connected C W -complexes with base point. In this way, Brown
constructed a new contravariant functor in his paper (Brown 1962). In other words,
he proved that every reduced cohomology theory on the category of C W -complexes
with base points and base point preserving maps has the form
h n (X ) = [X ; An ]
for some -spectrum A = { An , αn }. This functor now called Browns functor. This
theorem in homotopy theory presents a necessary and sufficient condition for a con-
travariant functor on the homotopy category of pointed connected C W -complexes,
to the category of sets, to be a representable functor. The representability theo-
rem of Brown shows that the set of all cohomology operations of above type is

in bijective correspondence with the group H m (K (G, n); G ). These groups were
studied intensively by Eilenberg and Steenrod during 1950–1952 and determined by
Henri Paul Cartan (1904–2008) in 1953. This theorem has made a turning point in
algebraic topology. Brown showed in 1963 that many of the most important functors
in algebraic topology are essentially homotopy functors and hence accessible to the
methods of homotopy theory. It is proved that cohomology theory is to a large extent
a branch of stable homotopy theory.
18.12 Obstruction Theory
This section conveys the early development of obstruction theory by using coho-
mology theory which describes a technique for studying various homotopy prob-
lems such as extension problems, lifting problems and relative lifting problems.
These are basic problems in algebraic topology. The origin of obstruction theory
is found in the classical works of H. Hopf, S. Eilenberg, N.E. Steenrod and M.
Postnikov (1927–2004). It appears in most textbooks on algebraic topology with
different approaches. For example, Steenrod studied obstruction on fiber bundles
in his book (Steenrod 1951), Spanier in his book (Spanier 1966) and Whitehead
in his book (Whitehead 1978) used the concept of obstruction to solve extension
and classification problems for continuous maps of a C W -complex into a topolog-
ical space. One application of obstruction theory is to define characteristic classes.
N.E. Steenrod wrote in his excellent paper (Steenrod 1972):
“Many of the basic theorems of topology, and some of its most successful appli-
cations in other areas of mathematics, are solutions of particular extension problems.
The deepest results of this kind have been obtained by the method of algebraic
topology. The essence of the method is a conversion of a geometric problem into
an algebraic problem which is sufficiently complex to embody the essential features
of the geometric problem, yet sufficiently simple to be solvable by standard alge-

braic methods. Many extension problems remain unsolved, and much of the current
development of algebraic topology is inspired by the hope of finding a truly general
solution”.
[1] Adams J.F., Stable Homotopy Theory, Springer-Verlag, Berlin, 1964.

[2] Alexander, J.W., On the chains of a complex and their duals, Proc. Nat. Acad.
Sci., USA 21, 509–511, 1935.
[3] Aull, C.E., and Lowen R., (Eds) Handbook of the History of General Topology,
Volume 3, Kluwer Academic Publishers, 2001.
[4] Bott, R., The stable homotopy of the classical groups, Ann. of Math., 70, 313–
337 1959.
[5] Čech,E., Théorie génerale de l’homologie dans un espace quelconque’, Fund.
Math. 19, 149–183, 1932.
groups of spaces, Ann. of Math., 46(2), 1945.
[7] Freedman, M., The topology of four-dimensional manifolds, Journal of Differ-
ential Geometry17, 357–453, 1982.
[8] Glenys, G.L., and Mishchenko, A.S., Vector bundles and their applications,
Mathematics and its Applications, 447, Kluwer Academic Publishers,
Dordrecht, 1998.
[10] Hilton, P.J., An introduction to Homotopy Theory, Cambridge University, 1983.
[11] Hilton, P.J., and Wylie, S., Homology Theory, Cambridge University Press,
Cambridge, 1960.
[12] Kampen van. E., On the connection between the fundamental groups of some
related spaces, American Journal of Mathematics, 55, 261–267, 1933.
[13] Mayer, J., Algebraic Topology, Prentice-Hall, New Jersey, 1972.
[14] Milnor, J.W., Spaces having the homotopy type of a C W -complex, Trans Amer.
Math. Soc., 90, 272–280, 1959.
[15] Olum, P., Obstructions to extensions and homotopies. Ann. of Math., 25, 1–
50,1950.
[16] Poincaré, H., Papers on Topology: Analysis Situs and its Five Supplements,
Translated by J. Stillwell, History of Mathematics, 37, Amer. Math. Soc, 2010.
[17] Smale, S., Generalized Poincaré’s conjecture in dimensions greater than 4,
Ann. of Math., 74 (2), 391–406. 1961.
[18] Spanier, E., Algebraic Topology, McGraw-Hill, 1966.
[19] Spanier, E., Infinite symmetric products, function spaces and duality, Ann. of
Math., 69, 142–198, 1959.
[20] Stallings, J.R., On fibering certain 3-manifolds, Topology of 3-manifolds and

related topics, Proc. The Univ. of Georgia Institute, 1961, Prentice Hall, 95–100,
1962.
[21] Steenrod, N.E., The Topology of Fibre Bundles, Princeton, 1951.
[24] Zeeman, C., The generalised Poincaré conjecture, Bull. Amer. Math. Soc 67,
270, 1961.
References
Adams, J.F.: On the non-existence of elements of Hopf invariant one. Ann. Math. 72, 20–104 (1960)
Adams, J.F.: Stable Homotopy Theory. Springer, Berlin (1964)
(1972)
Alexander, J.W.: On the chains of a complex and their duals. Proc. Natl. Acad. Sci., USA 21,
509–511 (1935)
Atiyah, M.F.: K-theory. Benjamin, New York (1967)
Atiyah, M.F., Hirzebruch, F.: Riemann-Roch theorems of differentiable manifolds. Bull. Am. Math.
Soc. 65, 276–281 (1959)
Atiyah, M.F., Hirzebruch F.: Vector bundles and homogeneous spaces. Proc. Symp. Pure Math.,
Am. Math. Soc. 7–38 (1961)
Aull, C.E., Lowen R. (eds.): Handbook of the History of General Topology, vol. 3. Kluwer Academic
Publishers, Berlin (2001)
Borsuk, K.: Sur les groupes des classes de transformations continues, pp. 1400–1403. C.R. Acad.
Sci, Paris (1936)
Bott, R.: The stable homotopy of the classical groups. Ann. Math. 70, 313–337 (1959)
Brown, E.H.: Cohomology theories. Ann. Math. 75, 467–484 (1962)
Čech, E.: Théorie génerale de l’homologie dans un espace quelconque’. Fund. Math. 19, 149–183
(1932)
Boston (1989)
Dold, A.: Relations between ordinary and extraordinary homology. Algebr. Topol. Colloq. Aarhus
2–9 (1962)
Dold, A., Thom, R.: Quasifaserungen und unendliche symmetrische Produkte. Ann. Math. Second
Series 67, 239–281 (1958)
Eilenberg, S., MacLane, S.: Natural isomorphism in group theory. Proc. Natl. Acad. Sci. USA 28,
537–544 (1942)
Eilenberg, S., MacLane, S.: General theory of natural equivalence. Trans. Am. Math. Soc. 58,
231–294 (1945a)
Eilenberg, S., MacLane, S.: Relations between homology and homotopy groups of spaces. Ann.
Math. 46(2) (1945b)
Eilenberg, S., Steenrod, N.: Axiomatic approach to homology theory. Proc. Natl. Acad. Sci. USA
31, 117–120 (1945)
Eilenberg, S., Steenrod, N.: Foundations of Algebraic Topology. Princeton University Press,
Princeton (1952)
Freedman, M.: The topology of four-dimensional manifolds. J. Differ. Geom. 17, 357–453 (1982)
Freudenthal, H.: Über die Klassen von Sphärenabbildungen. Compositio Mathematica 5, 299–314
(1937)
Hopf, H.: Über die Abbildungen der 3-Sphäre auf die Kugelfleche. Math. Ann. 104, 637–665 (1931)
Hopf, H.: Über die Abbildungen von Sphären auf Sphären niedrigerer Dimension. Fund. Math. 25,
427–440 (1935)
Hurewicz, W.: Beitrage der Topologie der Deformationen. Proc. K. Akad. Wet., Ser. A 38, 112–119,
521–528 (1935)
Hurewicz, W.: On duality theorems. Bull. Am. Math. Soc. 47, 562–563 (1941)
Hurewicz, W.: On the concept of fibre space. Proc. Natl. Acad. Sci., USA 41, 956–961 (1955)
James, I.M.: (ed.) History of Topology. North-Holland, Amsterdam (1999)
Kampen van. E.: On the connection between the fundamental groups of some related spaces. Am.
J. Math. 55, 261–267 (1933)
Mayer, J.: Algebraic Topology. Prentice-Hall, New Jersey (1972)
Milnor, J.W.: Spaces having the homotopy type of a C W -complex. Trans. Am. Math. Soc. 90,
272–280 (1959)
Olum, P.: Obstructions to extensions and homotopies. Ann. Math. 25, 1–50 (1950)
Poincaré, H.: Analysis situs. J. Ecole polytech 1(2), 1–121 (1895)
Poincaré, H.: Second complément à l’analysis situs. Proc. Lond. Math. Soc. 32, 277–308 (1900)
Poincaré, H.: Cinquiéme, complément à l’analysis situs. Rc. Cir. Mat. Palermo. 18, 45–110 (1904)
Poincaré, H.: Papers on topology: analysis situs and its five supplements (Translated by J. Stillwell,
History of Mathematics). Am. Math. Soc. 37 (2010)
Pontryagin, L.: The general topological theorem of duality for closed sets. Ann. Math. 35(2), 904–
914 (1934)
Smale, S.: Generalized Poincaré’s conjecture in dimensions greater than 4. Ann. Math. 74(2), 391–
406 (1961)
Spanier, E.H.: Borsuk’s cohomotopy groups. Ann. Math. 50, 203–245 (1949)
Spanier, E.: Infinite symmetric products, function spaces and duality. Ann. Math. 69, 142–198
(1959)
Stallings, J.R.: On fibering certain 3-manifolds, topology of 3-manifolds and related topics. In:
Proceedings of The University of Georgia Institute, 1961, pp. 95–100. Prentice Hall, Englewood
Cliffs (1962)
Steenrod, N.E.: The Topology of Fibre Bundles. Princeton University Press, Princeton (1951)
Steenrod, N.E.: Chohomology operations and obstructions to extending continuous functions. Adv.
Math. 8, 371–416 (1972)
Whitehead, J.H.C.: On adding relations to homotopy groups. Ann. Math. 42, 409–428 (1941)
Whitehead, J.H.C.: Combinatorial homotopy: I. Bull. Am. Math. Soc. 55, 213–245 (1949)
Whitehead, J.H.C.: A certain exact sequence. Ann. Math. 52(52), 51–110 (1950)
Whitehead, G.W.: A generalization of the Hopf invariant. Ann. Math. 51, 192–237 (1950)
Whitehead, G.W.: On the Freudenthal theorem. Ann. Math. 57, 209–228 (1953)
Whitehead, G.W.: Generalized homology theories. Trans. Am. Math. Soc. 102, 227–283 (1962)
Zeeman, C.: The generalised Poincaré conjecture. Bull. Am. Math. Soc. 67, 270 (1961)
Appendix A
Topological Groups and Lie Groups
This appendix studies topological groups, and also Lie groups which are special
topological groups as well as manifolds with some compatibility conditions. The
concept of a topological group arose through the work of Felix Klein (1849–1925)
and Marius Sophus Lie (1842–1899). One of the concrete concepts of the the-
ory of topological groups is the concept of Lie groups named after Sophus Lie.
The concept of Lie groups arose in mathematics through the study of continuous
transformations, which constitute in a natural way topological manifolds. Topo-
logical groups occupy a vast territory in topology and geometry. The theory of
topological groups first arose in the theory of Lie groups which carry differential
structures and they form the most important class of topological groups. For exam-
ple, GL (n, R), GL (n, C), GL (n, H ), SL (n, R), SL (n, C), O(n, R), U(n, C), SL
(n, H ) are some important classical Lie Groups. Sophus Lie first systematically
investigated groups of transformations and developed his theory of transformation
groups to solve his integration problems.
David Hilbert (1862–1943) presented to the International Congress of Mathe-
maticians, 1900 (ICM 1900) in Paris a series of 23 research projects. He stated in
this lecture that his Fifth Problem is linked to Sophus Lie theory of transformation
groups, i.e., Lie groups act as groups of transformations on manifolds. A translation
of Hilbert’s fifth problem says “It is well-known that Lie with the aid of the concept
of continuous groups of transformations, had set up a system of geometrical axioms
and, from the standpoint of his theory of groups has proved that this system of axioms
suffices for geometry”.
For this appendix, the books Bredon (1993), Chevelly (1957), Pontragin (1939),
Sorani (1969), Switzer (1975) and some others are referred in Bibliography.
A.1 Topological Groups: Definitions and Examples
This section introduces the concept of topological groups with illustrative examples.
A topological group is simply a combination of two fundamental concepts: group and
topological space and hence the axiomatization of the concept of topological groups
DOI 10.1007/978-81-322-2843-1
570 Appendix A: Topological Groups and Lie Groups
is a natural procedure. O. Schreior (1901–1929) gave a formal definition of modern

concept of topological groups in 1925 and F. Leja (1885–1979) in 1927 in terms of
topological spaces. Topological groups are groups in algebraic sense together with
continuous group operations. This means that the topology of a topological group
must be compatible with its group structure.
Definition A.1.1 A topological group G is a Hausdorff topological space together

with a group multiplication such that
TG(1) group multiplication m : G × G → G, (x, y) → x y is continuous;
TG(2) group inversion inv : G → G, x → x −1 is continuous.
The continuity in TG(1) and TG(2) means that the topology of G must be compatible
with the group structure of G. The conditions TG(1) and TG(2) are equivalent to
the single condition that the map
G × G → G, (x, y) → x y −1
is continuous.
Remark A.1.2 Some authors do not assume ‘Hausdorff property’ for a topological
group.
Example A.1.3 Rn (under usual addition) and S 1 = {z ∈ C : |z| = 1} (under usual

multiplication of complex numbers) are important examples of topological groups.
We now describe some classical topological groups GL(n, R), SL(n, R), O(n, R),
SO(n, R) and their complex analogues.
Definition A.1.4 (General linear group) GL(n, R) is the set of all n × n non-
singular matrices with entries in R. It is a group under usual multiplication of matri-
ces, called general linear group over R.
Definition A.1.5 (Special linear group) SL(n, R) defined by SL(n, R) = { A ∈

GL(n, R) : det A = 1} is a subgroup of GL(n, R).
Definition A.1.6 (Orthogonal group) O(n, R) defined by O(n, R) =

{A ∈ GL(n, R) : A At = I } is a subgroup of GL(n, R).
Definition A.1.7 (Special orthogonal group) SO(n, R) defined by SO(n, R) =

{A ∈ O(n, R) : det A = 1} is a group.
Theorem A.1.8 The general real linear group GL (n, R) of all invertible n × n
matrices over R is a topological group. This group is neither compact nor connected.
Proof Let Mn (R) be the set of all n × n real matrices. Let A = (ai j ) ∈ Mn (R). We
2
can identify Mn (R) with the Euclidean space Rn by the mapping
2
f : Mn (R) → Rn , (ai j ) → (a11 , a12 , . . . , a1n , a21 , a22 , . . . , a2n , . . . , an1 , an2 , . . . , ann ).
Appendix A: Topological Groups and Lie Groups 571
This identification defines a topology on Mn (R) such that the matrix multiplication
m : Mn (R) × Mn (R) → Mn (R)
is continuous.
Let A = (ai j ) and B = (bi j ) ∈ Mn (R). Then the i jth entry in the product m(A, B)
n
is aik bk j . As Mn (R) has the topology of the product space R1 × R1 × . . . , ×R1
k=1
(n 2 copies), and for each pair of integers i, j satisfying 1 ≤ i, j ≤ n, we have a
projection pi j : Mn (R) → R1 , which sends a matrix A to its i jth entry. Then m is
continuous if and only if the composite maps
m pi j
Mn (R) × Mn (R) −
→ Mn (R) −→ R1

n
are continuous. But pi j m(A, B) = aik bk j , which is a polynomial in entries of A
k=1
and B. Hence the composite maps pi j ◦
m are continuous.
GL (n, R) topologized as a subspace of the topological space Mn (R) is such that
the matrix multiplication
GL (n, R) × GL (n, R) → GL (n, R), (A, B) → AB
is continuous. We next claim that the inverse map
inv : GL (n, R) → GL (n, R), A → A−1
is continuous. The map
inv : GL (n, R) → GL (n, R) ⊂ R1 × R1 × . . . × R1 (n 2 copies)
is continuous if and only if all the composite maps
inv p jk
GL (n, R) −→ GL (n, R) −→ R1 , 1 ≤ j, k ≤ n
are continuous. But each composite map p jk ◦ inv sends a matrix A to the jkth
element of A−1 , which is (1/ det A) (k jth cofactor of A), where det A = 0 ∀ A ∈
GL (n, R). Hence the composite maps p jk ◦ inv are continuous. Consequently,
GL (n, R) is a topological group.
The group GL (n, R) is not compact: Clearly, GL (n, R) is the inverse image of
nonzero real numbers under the determinant function
det : Mn (R) → R.
The determinant function is continuous, since it is just a polynomial in the matrix co-
efficient. Hence the inverse image of {0} is a closed subset of Mn (R). Its complement
is the set of all nonsingular n × n real matrices is an open subset of Mn (R). Hence
GL (n, R) is not compact.
The group GL (n, R) is not connected: Clearly, the matrices with positive and
negative determinants give a partition of GL (n, R) into two disjoint nonempty open
sets. Hence GL (n, R) is not connected. ❑
Definition A.1.9 GL (n, C) is the set of all n × n nonsingular matrices with complex
entries. It is a group under usual multiplication of matrices, called the general complex
linear group.
Theorem A.1.10 GL (n, C) is a topological group. It is not compact.
Proof Every element A ∈ GL (n, C) is a nonsingular linear transformation of Cn
over C. If {z 1 , z 2 , . . . , z n } is a basis of Cn , then {x1 , y1 , . . . , xn , yn } is a basis of R2n ,
where z i = xi + i yi . Every element A ∈ GL (n, C) determines a linear transforma-
tion Ã ∈ GL (2n, R) into a subset of GL (n, R). Since GL (n, C) is an open subset
of a Euclidean space, it is not compact. ❑
Corollary A.1.11 The set U(n, C) = { A ∈ GL (n, C) : A A∗ = I }, is a compact
subgroup of GL (n, C), where A∗ denotes the transpose of the complex conjugate
of A.
Remark A.1.12 dimC GL (n, C) = n 2 .
Definition A.1.13 A homomorphism f : G → H between two topological groups
G and H is a continuous map such that f is a group homomorphism. An isomorphism
f : G → H between two topological groups is a homeomorphism and is also a group
homomorphism between G and H .
Example A.1.14 The special orthogonal group SO (2, R) and the circle group S 1 are
isomorphic topological groups under an isomorphism f of topological groups given
by

cos θ − sin θ
f : SO (2, R) → S 1 , → eiθ .
sin θ cos θ
Remark A.1.15 For quaternionic analogue see the sympletic group SU (n, H) =
{A ∈ GL (n, H) : A A∗ = I } (Ex. 10 of Sect. A.4).
A.2 Actions of Topological Groups and Orbit Spaces
This section introduces the concept of actions of topological groups and studies
some important orbit spaces (thus obtained) with an eye to compute their funda-
mental groups. Real and complex projective spaces, torus, Klein bottle, lens spaces,
and figure-eight are important objects in geometry and topology and they can be
represented as orbit spaces.
Definition A.2.1 Let G be a topological group with identity element e and X a

topological space. An action of G on X is a continuous map σ : G × X → X , with
the image of (g, x) being denoted by gx such that
(i) (gh)x = g(hx);
(ii) ex = x,
∀ g, h ∈ G and ∀ x ∈ X . The pair (G, X ) with the given action σ is called a topo-
logical transformation group and in brief we call X a G-space.
Remark A.2.2 If we change any one G, X or σ, then we get a different transformation

group. If we forget the topologies from the space, then the group G and the action σ
give together the concept of G-set.
Orbit spaces are closely related with G-spaces. Let X be a G-space. Two elements
x, y ∈ X are said to be G-equivalent if ∃ an element g ∈ G such that gx = y. The
relation of being G-equivalent is an equivalence relation and the set {gx : g ∈ G}
denoted by Gx, the equivalence class determined by x, is called the orbit of x. If
the group G is compact and the space X is Hausdorff, then the orbits are closed sets
of X and in this case, the coset space G/Gx is homeomorphic to the orbit Gx. The
action of G on X is said to be free if Gx = {e}, ∀ x ∈ X . Two orbits in X are either
identical or disjoint. The set of all distinct orbits of X , denoted by X mod G, with
the quotient topology induced from X , is called the orbit space of the transformation
group.
Proposition A.2.3 Let X be a G-space. For every g ∈ G, the map ψg : X → X

defined by ψg (x) = gx is a homeomorphism.
Proof Since the action of G on X given by x → gx is continuous, ψg is continuous

for each g ∈ G. Moreover, ψg ◦ ψg−1 = ψgg−1 = ψe = 1 X and ψg−1 ◦ ψg = ψg−1 g =
ψe = 1 X ⇒ for each g the map ψg : X → X is a homeomorphism. ❑
Let Homeo (X ) denote the group of all homeomorphisms of X under usual compo-
sition of mappings.
Proposition A.2.4 Let X be a G-space. Then the action of G on X induces a homo-

morphism f : G → Homeo (X ).
Proof For each g ∈ G, the map f : G → Homeo (X ) defined by f (g) = ψg is a

homeomorphism by Proposition A.2.3 ⇒ f is well defined. Moreover, for g, h ∈
G, f (gh) = f (g) f (h) ⇒ f is a homomorphism. ❑
Definition A.2.5 Let X be a G-space. The action of G on X is said to be effective if

the map f : G → Homeo (X ), g → ψg is a monomorphism i.e., g = e ⇒ ψg = 1 X .
Proposition A.2.6 Every effective action of a topological group G on a topological

space X induces an embedding f : G → Homeo (X ).
Proof Let σ : G × X → X be an effective action. Then g = e ⇒ ψg = 1 X . Con-

sider the homomorphism f : G → Homeo (X ) defined by f (g) = ψg , where
ψg : X → X is given by ψg (x) = σ(g, x) = gx. Then for g = h(g, h ∈ G) ⇒ h −1 g
= e ⇒ ψh −1 g = 1 X ⇒ ψh −1 g (x) = x, ∀ x ∈ X ⇒ (h −1 g)x = x, ∀ x ∈ X ⇒ gx =
hx, ∀ x ∈ X ⇒ ψg (x) = ψh (x), ∀ x ∈ X ⇒ ψg = ψh ⇒ f (g) = f (h) ⇒ f : G
→ Homeo (X ) is an embedding. ❑
Corollary A.2.7 Every topological group G can be viewed as a group of homeo-

morphisms of a topological space X . This is an analogue of Caley’s Theorem for
group.
Remark A.2.8 Geometrical interpretation of free action: The action of G on X

is free if and only if each g(= e) ∈ G moves every point of X . Given a group G
of homeomorphisms of X , we can always define an action σ of G on X by taking
σ(g, x) = g(x). What is the topology of G? Since G ⊆ Homeo (X ), any topology on
Homeo (X )(viz. the compact open topology) will induce a topology on G. Now the
problem is: whether or not the map σ : G × X → X defined above is continuous. If
X is a locally compact Hausdorff space, then the compact open topology gives σ a
continuous action (see Dugundji, pp 259).
Example A.2.9 (Real Projective Space R P n ) Let f : S n → S n (n ≥ 1) be the antipo-

dal map i.e., f (x) = −x. Then f ◦ f = f 2 = 1 S n . Thus G = {1 S n , f } ∼= Z2 is a
group of homeomorphisms of S n and so it acts on S n . The action is free and the orbit
space S n mod G denoted by R P n , is called the real projective n-space. The space
R P n is a compact, connected manifold of dimension n.
Example A.2.10 (Complex Projective Space CP n ) Let S 1 = {z ∈ C : |z| = 1} be

the circle group. Then S 1 is a topological group. Let S 2n+1 = {(z 0 , z 1 , . . . , z n ) ∈
n
Cn : |z i |2 = 1}, be the (2n + 1)-dimensional unit sphere. Then S 1 acts S 2n+1
i=0
continuously under the action defined by z ·(z 0 , z 1 , . . . , z n ) = (zz 0 , zz 1 , . . . , zz n ).
This action is free. The orbit space S 2n+1 mod S 1 , denoted by CP n is called the
complex projective n-space. This space is a compact , connected manifold of real
dimension 2n.
Example A.2.11 (Torus) Let R be the real line and f : R → R be a homeomor-

phism defined by f (x) = x + 1, which is a translation. Then for each integer
n, h n : R → R, x → x + n is also a homeomorphism. These are just translations by
integer amounts. Then the cyclic group < f > generated by the homeomorphism h of
R is the infinite cyclic group Z. Endowing Z the discrete topology, we find that Z acts
as a group of homeomorphisms on R. Again the action is free and the orbit space R
mod Z = R/Z is homeomorphic to the circle group S 1 . We now consider the prod-
uct action of the discrete group Z × Z on R × R. Let f, g be two homeomorphisms
R → R. Define the action of ( f, g) on R × R by ( f, g)(x, y) = (x + 1, y + 1). Then

( f m , g n )(x, y) = (x + m, y + n) for every pair of integers. Again the action is free
and the orbit space S 1 × S 1 is the 2-torus. An n-torus is obtained similarly as an orbit
space of Rn for n ≥ 2.
A.3 Lie Groups and Examples
This section introduces the concept of Lie groups with illustrative examples. Lie
groups play an important role in geometry and topology. A Lie group is a topological
group having the structure of a smooth manifold for which the group operations are
smooth functions. Such groups were first considered by Sophus Lie in 1880 and
are named after him. He developed his theory of continuous maps and used it in
investigating differential equations. The fundamental idea of his Lie theory was
published in his paper Lie (1880) and his later book with F. Engel published in 1893
(Lie and Engel 1893). Lie classified infinitesimal groups acting in dimensions 1 and
2 up to analytic coordinate changes. Lie displayed the key role of his Lie theory as
a classifying principle in geometry, mechanics and ordinary and partial differential
equations. This theory made a revolution in mathematics and physics.
A.3.1 Lie Group: Introductory Concepts
Definition A.3.1 A topological group G is called a real Lie group if

(i) G is a differentiable manifold;
(ii) the group operations (x, y) → x y and x → x −1 are both differentiable.
Definition A.3.2 A topological group G is called a complex Lie group if

(i) G is a complex manifold;
(ii) the group operations (x, y) → x y and x → x −1 are both holomorphic.
Definition A.3.3 The dimension of a Lie group is defined to be its dimension as a

manifold.
Remark A.3.4 A Lie group is not necessarily connected. Given a Lie group, let G ◦
denote the connected component of G which contains the identity element of G. Then
G ◦ is a closed subgroup of G. Any other connected component of G is homeomorphic
to G ◦ . This shows that if G is a Lie group, the (real or complex) dimension of G is
well defined and it is the dimension of the manifold G ◦ .
Definition A.3.5 Let G and H be Lie groups. A differentiable map f : H → G is

called a homomorphism if f is a group homomorphism and a regular analytic map.
f (H ) is a subgroup of G and a submanifold of G.
Definition A.3.6 Let G be a Lie group and f : R → G be a homomorphism of Lie

groups. Then f (R) is called a one parameter subgroup of G.
Remark A.3.7 It was believed before 1956 that a topological space may admit only
one differentiable structure. Examples show that differentiable structure of a topo-
logical space may not be unique. For example, John Willard Milnor (1931-) proved
in 1956 that the 7-sphere S 7 admits 28 different differentiable structures (Milnor
1956). Milnor was awarded the Fields Medal in 1962 for his work in differential
topology. He was also awarded the Abel Prize in 2011. For another example, Sir
Simon Kirwan Donaldson (1957-) proved in 1983 that R4 admits an infinite num-
ber of different differentiable structures (Donalson 1983). He was awarded a Fields
Medal in 1986.
A.3.2 Some Examples of Lie Groups
This subsection presents important examples of Lie groups.

(i) The real line R is a Lie group under usual addition.
(ii) The classical Lie groups:
GL(n, R), SL(n, R), O(n, R), SO(n, R) are all manifolds, because for each
point x of any one of these groups, there exists an open neighborhood home-
omorphic to a Euclidean space. All of them are real Lie groups, called clas-
sical Lie groups. GL(n, R) is a real Lie group of dimension n 2 , SL(n, R) is
a real Lie group of dimension n 2 − 1, O(n, R) is a real Lie group of dimen-
sion n(n − 1)/2, SO(n, R) is a real Lie group of dimension n(n − 1)/2. Their
complex and quaternionic analogues are also Lie groups.
A.4 Exercises
1. Show that the circle group S 1 in the complex plane is a Lie group. (This group
is denoted by U (1, C) or by simply 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 S L(n.R) defined by SL (n, R) = {X ∈
GL (n, R) : det X = 1} is a noncompact connected topological group and is a
real Lie group of dimension n 2 − 1.
[ Hint: SL (n, R) is a subgroup of GL (n, R). It is a hypersurface of GL (n, R).]
4. Prove that the special complex linear group SL (n, C) given by SL (n, C) =
{X ∈ GL (n, C) : det X = 1} is a noncompact connected topological group and
is a complex Lie group of dimension n 2 − 1.
[Hint: SL (n, C) is a subgroup of GL (n, C).]

5. Show that the orthogonal group given by O(n, R) = { A ∈ GL (n, R) : A At =
I = At A} is a compact non-connected topological group and is a real Lie group
of dimension n(n−1)2
.
[Hint: O(n, R) is a closed subspace of GL (n, R). It contains matrices A ∈
GL (n, R) with det A = ±1. Thus O(n, R) is a bounded closed subset of the
2
Euclidean space Rn . Hence it is compact. Thus O(n, R) is a compact subgroup
of GL (n, R). Since it contains matrices with determinant equal to 1 and matrices
with determinant equal to −1, it non-connected.]
6. Prove that the special orthogonal group SO (n, R) given by SO (n, R) =
O(n, R) ∩ SL (n, R) is a real compact connected topological group and is a
real Lie group of dimension n(n−1)2
.
7. Prove that the general (complex) linear group GL(n, C) is a topological group
and is a connected, noncompact complex Lie group of dimension n 2 .
8. Show that the unitary group U (n, C) defined by U (n, C) = { A ∈ GL (n, C) :
A A∗ = A∗ A = I } is a connected compact topological group, and is a real Lie
group of dimension n 2 , 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).
As a subspace 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) ∩
z w
SL (n, C). Show that the group SU (2, C) = { A = : z, w ∈ C and
−w̄ z̄
|z|2 + |w|2 = 1} is isomorphic to S 3 .
[Hint. Use the form of A.]
10. The quaternionic analogue of orthogonal unitary groups is the sympletic group
SU (n, H) = { A ∈ GL (n, H) : A A∗ = 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 R P 3 and SO (3, R) are homeo-
morphic.
12. Show that
(i) Let M be a manifold. A flow ψt : M → M is an action of R on M. If a
flow is periodic with period p, it may be regarded as an action of U (1, C)
or SO (2, R) on M.
(ii) Let A ∈ GL (n, R) and x ∈ Rn . The action σ of GL (n, R) on Rn is defined
by the usual matrix action on a vector
σ : GL (n, R) × Rn → Rn , (M, x) → M ·x
13. Show that the right translation R : (a, g) → Ra g and left translation L : (a, g)
→ L a g of a Lie group are free and transitive actions.
14. Let a Lie group G act on a manifold M. Show that
(i) the isotropy group G x of any x ∈ M is a Lie subgroup;
(ii) if G acts on M freely, then the isotropy group G x of any x ∈ M is trivial.

15. Show that the orthogonal group O (n + 1, R) acts on R P n transitively from left.
16. Show that orthogonal group O (n, R) acts transitively on the Grassmann manifold
(see Chap. 5) G n,r (r ≤ n).
17. Show that the special orthogonal group SO (n, R) acts transitively on the Stiefel
manifold (see Chap. 5) Vn,r = Vr (Rn ), (r ≤ n).
A.5 Additional Reading
tions, Springer, New Delhi, New York, 2014.
[2] Borel, A., Topology of Lie groups and characteristic classes, Bull. Amer. Math.
Soc.61 (1955), 397–432.
[4] Dupont, J., Fibre Bundles, Aarhus Universitet, 2003.
[5] Nakahara, M., Geometry, Topology and Physics, Institute of Physics Publishing,
Taylor and Francis, Bristol, 2003.
[6] Samelson, H., Topology of Lie groups, Bull. Amer. Math. Soc. 52 (1952), 2–37.
[7] Spanier, E., Algebraic Topology, McGraw-Hill Book Company, New York, 1966.
References
Adhikari, M.R., Adhikari, A.: Basic Modern Algebra with Applications. Springer, New Delhi
(2014)
Bredon, G.: Topology and Geometry. Springer, Heidelberg (1993) (GTM 139)
Borel, A.: Topology of Lie groups and characteristic classes. Bull. Am. Math. Soc. 61, 397–432
(1955)
Chevelly, C.: Theory of Lie Groups I. Princeton University Press, Princeton (1957)
Boston (1989)
Donalson, S.: Self-dual connections and the topology of smooth 4-manifolds. Bull. Am. Math. Soc.
8(1), 81–83 (1983)
Dupont, J.: Fibre Bundles. Aarhus Universitet, Denmark (2003)
Lie, S.: Theorie der Transformations gruppen. Math. Annalen 16, 441–528 (1880)
Lie, S., Engel, F.: Theorie der Transformations gruppen, Bd III. Teubner, Leipzig (1893)
Milnor, J.: On manifolds homeomorphic to the 7-sphere. Ann. Math. 64(2), 399–405 (1956)
Nakahara, M.: Geometry, Topology and Physics, Institute of Physics Publishing. Taylor and Francis,
Bristol (2003)
Pontragin, L.: Topological Groups. Princeton University Press, Princeton (1939)
Samelson, H.: Topology of Lie groups. Bull. Am. Math. Soc. 52, 2–37 (1952)
Sorani, G.: An Introduction to Real and Complex Manifolds. Gordon and Breech, Science Pub.,
Inc, New York (1969)
Appendix B
Categories, Functors and Natural
Transformations
This appendix conveys category theory through the study of categories, functors, and
natural transformations with an eye to study algebraic topology which consists of the
constructions and use of functors from certain category of topological spaces into
an algebraic category. Algebraic topology studies techniques for forming algebraic
images by mechanisms that create these images which are known as functors. They
have the characteristic feature that they form algebraic images of spaces and project
continuous maps into their corresponding algebraic images. Category theory plays
an important role for the study of homotopy, homology and cohomology theories
which constitute the basic text of algebraic topology. So the readers of algebraic
topology can not escape learning the concepts of categories, functors and natural
transformations.
The present book uses category theory and conveys a study of some important
functors such as homotopy, homology and cohomology functors in addition to adjoint
functor, representable functor, abelianization functor, Brown functor, and infinite
symmetric product functor. All constructions in algebraic topology are in general
functorial. Fundamental groups, higher homotopy groups, homology and cohomol-
ogy groups are not only algebraic invariants of the underlying topological space, in
the sense that two topological spaces which are homeomorphic have the isomorphic
associated groups (or modules) but also they are homotopy invariants in the sense
that homotopy equivalent spaces have isomorphic algebraic structures. Moreover,
corresponding to a continuous mapping of topological spaces the induced group (or
module) homomorphism on the associated groups (modules) can be used to show
the non-existence (or much more deeply, existence) of a continuous mapping of the
spaces.
Historically, the whole idea of category theory arose through the field of alge-
braic topology. The first and the simplest realization of this idea is the fundamental
group (or Poincaré group) of a pointed topological space. The concepts of categories,
functors, natural transformations and duality were introduced during 1942–1945 by
S. Eilenberg (1913–1998) and S. MacLane (1909–2005).1 Originally, the purpose
1 (i)
Natural isomorphism in group theory, Proc. Nat. Acad Sc., USA 28 (1942), 537–544.
DOI 10.1007/978-81-322-2843-1
582 Appendix B: Categories, Functors and Natural Transformations
of these notions was to provide a technique for classifying certain concepts, such as
that of natural isomorphism. Many concepts of algebraic topology are unified and
explained by category theory which is a very important branch of modern mathemat-
ics. This branch has been quite rapidly growing both in contents and applicability to
other branches of mathematics.
For this chapter, the books and papers Adhikari and Adhikari (2014), Eilenberg
and MacLane (1942, 1945), Eilenberg and Steenrod (1952), Gray (1975), Hatcher
(2002), MacLane (1972), Rotman (1988), Spanier (1966), Steenrod (1967) and some
others are referred in Bibliography.
B.1 Categories: Introductory Concepts
This section introduces the concept of ’category’ to specify a class of objects for
their study. It is observed that to define a new class of mathematical objects in mod-
ern mathematics, it becomes necessary to specify certain types of functions between
the objects such as topological spaces and continuous maps, groups and homomor-
phisms, modules and module homomorphism. A formulation of this observation
leads to the concept of ‘categories’. A ‘category’ may be thought roughly as consist-
ing of sets, possibly with additional structures, and functions, possibly preserving
additional structures. More precisely, a category can be defined with the following
characteristics.
Definition B.1.1 A category C consists of

(a) a class of objects X, Y, Z , . . . denoted by ob(C);
(b) for each ordered pair of objects X, Y , a set of morphisms with domain X and
range Y denoted by C(X, Y ) or simply mor (X, Y ); i.e., if f ∈ mor (X, Y ), then
X is called the domain of f and Y is called the co-domain (or range) of f : one
f
also writes f : X → Y or X −−−→ Y to denote the morphism from X to Y ;
(c) for each ordered triple of objects X, Y and Z and a pair of morphisms f : X → Y
and g : Y → Z , their composite denoted by g f : X → Z i.e., if f ∈ mor (X, Y )
and g ∈ mor (Y, Z ), then their composite g f ∈ mor (X, Z ) and satisfies the fol-
lowing two axioms:
(i) associativity: if f ∈ mor (X, Y ), g ∈ mor (Y, Z ) and h ∈ mor (Z , W ), then
h(g f ) = (hg) f ∈ mor (X, W );
(ii) identity: for each object Y in C, there is a morphism 1Y ∈ mor (Y, Y ) such
that if f ∈ mor (X, Y ), then 1Y f = f and if h ∈ mor (Y, Z ), then h1Y = h.
Clearly, 1Y is unique.
If the class of objects is a set, the category is said to be small.
(ii) General theory of natural equivalence, Trans Amer. Math. Soc., 58 (1945), 231–294.
Appendix B: Categories, Functors and Natural Transformations 583
Example B.1.2 (i) Sets and functions form a category denoted by Set .
(Here the class of objects is the class of all sets and for sets X and Y, mor (X, Y )
equals the set of functions from X to Y and the composition has the usual
meaning, i.e., usual composition of functions).
(ii) Finite sets and functions form a category denoted by Set F .
(iii) Groups and homomorphisms form a category denoted by Grp .
(Here the class of objects is the class of all groups and for groups X and
Y, mor (X, Y ) equals the set of homomorphisms from X to Y and the compo-
sition has the usual meaning).
(iv) Abelian groups and homomorphisms form a category denoted by Ab .
(v) Rings and homomorphisms form a category denoted by Ring .
(vi) R-modules and R-homomorphisms form a category denoted by Mod R .
(vii) Exact sequences of R-modules and R-homomorphisms form a category.
(viii) Topological spaces and continuous maps form a category denoted by Top .
(ix) Topological spaces and homotopy classes of maps form a category denoted
by Htp .
(Here the class of objects is the class of all topological spaces and for topolog-
ical spaces X and Y, mor (X, Y ) equals the set of homotopy classes of maps
from X to Y and the composition has the usual meaning.)
Definition B.1.3 A subcategory C ⊂ C is a category such that

(a) the objects of C are also objects of C, i.e., ob(C ) ⊂ ob(C);
(b) for objects X and Y of C , C (X , Y ) ⊂ C(X , Y );
(c) if f : X → Y and g : Y → Z are morphisms of C , their composite in C
equals their composite in C.
Definition B.1.4 A subcategory C of C is said to be a full subcategory of C if for

objects X and Y in C , C(X , Y ) = C (X , Y ).
The category in Example B.1.2(ii) is a full subcategory of the category in Example

B.1.2(i).
Remark B.1.5 The category Set F in Example B.1.2(ii) is a subcategory of the cate-
gory Set in Example B.1.2(i). On the other hand, the categories in Example B.1.2(iii)–
(vii) are not subcategories of the category in Example B.1.2(i), because each object
of one of the former categories consists of a set, endowed with an additional structure
(hence different objects in these categories may have the same underlying sets).
Remark B.1.6 In category Example B.1.2(ix), the morphisms are not functions and
so this category is not a subcategory of the category in Example B.1.2(i).
Let C be a category and A, B, C, . . . be objects of C.
Definition B.1.7 A morphism f : A → B in C is called a coretraction if there is a

morphism g : B → A in C such that g f = 1 A . In this case g is called a left inverse
of f and f is called a right inverse of g and A is called a retract of B.
Dually we say that f is a retraction if there is a morphism g : B → A such that

f g = 1 B in C. In this case g is called a right inverse of f .
Definition B.1.8 A two-sided inverse (or simply an inverse) of f is a morphism

which is both a left inverse of f and a right inverse of f .
Lemma B.1.9 If f : A → B in C has a left inverse and a right inverse, they are
equal.
Proof Let g : B → A be a left inverse of f and g : B → A a right inverse

of f , then g f = 1 A and f g = 1 B . Now g = g 1 B = g ( f g ) = (g f )g = 1 A
g = g . ❑
Definition B.1.10 A morphism f : A → B is called an equivalence (or an isomor-

phism) in a category C denoted by f : A ≈ B if there is a morphism g : B → A
which is a two-sided inverse of f .
Example B.1.11 An equivalence in the category Top of topological spaces and their
continuous maps is a homeomorphism and that in category Htp of topological spaces
and their homotopy classes of maps is a homotopy equivalence.
Remark B.1.12 An equivalence f : A ≈ B has a unique inverse denoted by f −1 :

B → A and f −1 is also an equivalence.
Definition B.1.13 Two objects A and B in C are said to be equivalent denoted by

A ≈ B if there is an equivalence f : A ≈ B in C.
Remark B.1.14 As the composite of equivalences is an equivalence, the relation of

being equivalent is an equivalence relation on any set of objects of a category C.
B.2 Functors: Introductory Concepts and Examples
This section introduces the concept of functors and studies functors of different nature
such as covariant and contravariant functors, adjoint functor, representable functor,
abelianization functor, Brown functor, and infinite symmetric product functor which
play a key role in algebraic topology. The main interest in category theory is in the
maps from one category to another. Those maps which have the natural properties of
preserving identities and composites are called functors (covariant or contravariant).
An algebraic representation of topology is a mapping from topology to algebra.
Such a representation, formally called a functor, converts a topological problem into
an algebraic one. The concept of algebraic functors is very important in algebraic
topology. For example, homotopy and homology theories provide a sequence of
covariant (algebraic) functors. On the other hand, cohomology theory provides a
sequence of contravariant (algebraic) functors.
B.2.1 Functors: Introductory Concepts
Definition B.2.1 Let C and D be categories. A covariant functor (or contravariant

functor) T from C to D consists of
(i) an object function which assigns to every object X of C an object T (X ) of D;
and
(ii) a morphism function which assigns to every morphism f : X → Y in C, a mor-
phism T ( f ) : T (X ) → T (Y ) (or T ( f ) : T (Y ) → T (X )) in D such that
(a) T (1 X ) = 1T (X ) ;
(b) T (g f ) = T (g)T ( f ) (or T (g f ) = T ( f )T (g)) for g : Y → W in C.
Example B.2.2 (i) (Forgetful functor) There is a covariant functor from the cate-
gory of groups and homomorphisms to the category of sets and functions which
assigns to every group its underlying set. This functor is called a forgetful functor
because it forgets the structure of a group.
(ii) (Hom R functor) Let R be a commutative ring. Given a fixed R-module M0 ,
there is a covariant functor π M0 (or contravariant functor π M0 ) from the category
of R-modules and R-homomorphisms to itself which assigns to an R-module
M the R-module Hom R (Mo , M) (or Hom R (M, Mo )) and if α : M → N is an
R-module homomorphism, then
π M0 (α) : Hom R (M0 , M) → Hom R (M0 , N ) is defined by π M0 (α)( f ) = α f
∀ f ∈ Hom R (M0 , M)
(π M0 (α) : Hom R (N , M0 ) → Hom R (M, M0 ) is defined by
π Mo (α)( f ) = f α ∀ f ∈ Hom R (N , M0 )).
(iii) (Dual functor) Let C be any category and C ∈ ob(C). Then there is a covariant
functor h C : C → Set (category of sets and functions), where the object func-
tion is defined by h C (A) = C(C, A) (set of all morphisms from the object C to
the object A in C) ∀ objects A ∈ ob(C) and for f : A → B in C, the morphism
function h C ( f ) : h C (A) → h C (B) is defined by h C ( f )(g) = f g ∀ g ∈ h C (A)
(the right hand side is the composite of morphisms in C).
Its dual functor h C defined in a usual manner is a contravariant functor.
Remark B.2.3 A functor from a category C to itself is sometimes called a functor on

C. Any contravariant functor on C corresponds to a covariant functor on C 0 and vice
versa. Thus any functor can be regarded as a covariant (or contravariant) functor on
a suitable category. In spite of this, we consider covariant as well as contravariant
functors on C.
Definition B.2.4 A functor T : C → D is called

(i) faithful if the mapping T : C(A, B) → D(T (A), T (B)) is injective;
(ii) full if the mapping T : C(A, B) → D(T (A), T (B)) is surjective; and
(iii) an embedding if T is faithful and T (A) = T (B) =⇒ A = B.
Definition B.2.5 A category C is called concrete if there is a faithful functor T :

C → Set .
Theorem B.2.6 Let F : C1 → C2 be a functor from a category C1 to a category C2 .

If two objects X and Y are isomorphic in C1 then the objects F(X ) and F(Y ) are
isomorphic in C2 .
Proof Let f : X → Y be a isomorphism in C1 . Then there exists a morphism g :

Y → X in C1 such that g ◦ f = 1 X and f ◦ g = 1Y . If F is a covariant functor, then
F(g) ◦ F( f ) = 1 F(X ) and F( f ) ◦ F(g) = 1 F(Y ) , which imply that the objects F(X )
and F(Y ) are isomorphic in C2 . If F is a contravariant functor, then the proof is
similar. ❑
Remark B.2.7 Consider a functor from the category Top to another category, say
the category Grp of groups. Let X and Y be objects in Top . Then F(X ) and F(Y )
are objects in Grp . If F(X ) and F(Y ) are not isomorphic, then X and Y can not
be homeomorphic. To the contrary if F(X ) and F(Y ) are isomorphic, then X and
Y may not be homeomorphic. For example, for the covariant functor πm , πm (S 2 ) ∼ =
πm (S 3 ) for m ≥ 3 but S 2 and S 3 are not homeomorphic (see Hopf fibering of spheres,
Chap. 7).
Definition B.2.8 (Bifunctor) A bifunctor is a mapping T : C1 × C2 → C, defined

on the Cartesian product of two categories C1 and C2 with values in the category C,
which assigns to each pair of objects in A1 ∈ C1 and A2 ∈ ×C2 , some object A ∈ C,
and to each pair of morphisms
α : A1 → A1 , β : A2 → A2
the morphism
T (α, β) : T (A1 , A2 ) → (A1 , A2 )
The following conditions
T (1 A1 , 1 A2 ) = 1T (A1 ,A2 )
must also be satisfied. In such a case one says that the functor is contravariant with
respect to the first argument and covariant with respect to the second.
Remark B.2.9 Similarly one can define bifunctors contravariant in both arguments
and covariant in the first and contravariant in the second argument. Thus a bifunctor
is a functor whose domain is product category.
Example B.2.10 Hom functor is of the type C op × C → Set . It is a bifunctor in two

arguments. The Hom functor is a natural example; it is contravariant in one argument,
covariant in the other.
B.2.2 Abelianization Functor
This subsection defines abelianization functor. The concept of abelianized groups is

utilized to establish a connection between the fundamental group and first homol-
ogy group of a pointed space. If G is a group (not necessarily abelian), then the
commutator subgroup [G, G] is a normal subgroup of G and the quotient group
G/[G, G] is called the abelianized group of G, and denoted by G ab : If f : G → H
is a homomorphism, it induces a homomorphism
f ∗ : G/[G, G] → H/[H, H ].
This gives a functor G → G ab , f → f ∗ which is called ‘abelianization functor’

from the category of groups and homomorphisms to the category of abelian groups
and homomorphisms.
B.2.3 Adjoint Functor
This subsection discusses adjoint functors which are very important in the study
of homotopy theory. Recall that the loop functor is a covariant functor from
the category of pointed topological spaces and continuous maps to the category of
H -groups and continuous homomorphisms such that the functor also preserves
homotopies (see Chap. 2). In the same chapter we have also described suspension
space which is dual to the loop space. Let X be a pointed topological space with
base point x0 . Then the suspension space of X , denoted by X , is defined to be the
quotient space of X × I in which (X × {0}) ∪ {x0 } × I ) ∪ (X × 1) is identified to a
single point. If (x, t) ∈ X × I, [x, t] denotes the corresponding points of X under
the quotient map X × I → X . The point [x0 , 0] ∈ X is also denoted by x0 and
X becomes a pointed space with base point x 0 . If f : X → Y is a continuous map,
then f : X → Y is defined by ( f )([x, t]) = [ f (x), t]. Hence is a covariant
functor from the category of pointed topological spaces and continuous maps to itself.
Recall that is also a covariant functor from the category of pointed topological
spaces and continuous maps to the category of H -cogroups and homomorphisms
(see Chap. 2).
Definition B.2.11 The functors and defined from the category of pointed topo-
logical spaces and continuous maps to itself form a pair of functors, called an adjoint
pair in the sense that for pointed topological spaces X and Y , there is an equivalence
between mor ( X, Y ) ≈ mor (X, Y ), where both sides are interpreted as the set
of morphisms in the category of pointed topological spaces and continuous maps.
If f : X → Y is a morphism, then the corresponding morphism f˜ : X → Y is
defined by f˜([x, t]) = f (x)(t) for all x ∈ X and t ∈ I .
Definition B.2.12 Corresponding to the adjoint pair of functors and , the con-
tinuous maps f : X → Y and f˜ : X → Y in the category pointed topological
spaces and continuous maps are said to be adjoint to each other if the morphism
f˜ : X → Y is defined by f˜([x, t]) = f (x)(t) for all x ∈ X and t ∈ I .
Remark B.2.13 The adjoint relation mor ( X, Y ) ≈ mor (X, Y ) holds, because
base point preserving maps X → Y are exactly the same as the base point preserv-
ing maps X → Y , the correspondence is given by assigning to f : X → Y the
family of loops by restricting f to the images of the segment {x} × I in X .
B.2.4 Brown Functor
This subsection defines Browns functor which is used to prove Brown representability
theorem.
Definition B.2.14 Let T : Htp ∗ → Set ∗ be a contravariant functor from the homo-
topy category of pointed topological spaces and homotopy classes of their maps to
the category of pointed sets and their maps. If i : X → Y in Htp ∗ , u ∈ T (Y ) and
u| X denotes the element T ([i])(u) ∈ T (X ), then T is called a Brown functor if it
satisfies following two axioms:
B(1): Wedge axiom If {X j } is a family of pointed topological spaces and i j :
X j → ∨X j is the inclusion, then T (i j ) : T (∨ j X j ) → j T (X j ) is an equiva-
lence of sets;
B(2): Mayer–Vietoris axiom For any excisive triad (X ; A, B) (i.e., X = Int
(A) ∪ Int (B)) and for any u ∈ T (A) and v ∈ T (B) such that u| A∩B = v| A∩B ,
there exists an element w ∈ T (X ) such that w| A = u and w| B = v.
B.2.5 Infinite Symmetric Product Functor
This subsection defines ‘infinite symmetric product functor’. This functor is impor-
tant to prove Dold–Thom theorem, which is a very key result of algebraic topology.
Let X be a pointed topological space with base point x 0 and X n = X × × . . . × X
be its nth cartesian product for n ≥ 1. If Sn denotes the symmetric groups of the set
{1, 2, . . . , n}, then there is a right action on X n , which permutes the coordinates,
i.e., for σ ∈ Sn , we define (x 1 , x 2 , . . . , xn )·σ = (xσ(1) , xσ(2) , . . . , xσ(n) ), xi ∈ X . The
orbit space X n mod Sn of this action denoted by S P n X and is called the nth
symmetric product of X . The equivalence class of (x1 , x2 , . . . , xn ) is denoted by
[x 1 , x2 , . . . , xn ].
Define inclusions S P n X → S P n+1 X, [x1 , x2 , . . . , xn ] → [x0 , x1 , . . . , x n ] for
n ≥ 1 and form the union S P ∞ X = S P n X equipped with the union topology
n
Fig. B.1 Commutative

diagram for symmetric
product
(weak topology), which means that a subset A ⊂ S P ∞ X is closed iff A ∩ S P n X is

closed for each n ≥ 1.
Definition B.2.15 For a pointed space X the space S P ∞ X is called the infinite
symmetric product of X .
Remark B.2.16 For a pointed space X the elements of S P n X may be considered as
unordered n-tuples [x 1 , x 2 , . . . , xn ], where n ≥ 1. Then S P ∞ X is a pointed space
with the base point 0 = [x 0 ]. There exists a natural inclusion i : X → S P ∞ X , where
X = S P 1 X.
Let f : X → Y be a base point preserving continuous map. Then f induces
maps f ∗(n) : S P n X → S P n Y , which are compatible with the action of the group Sn .
Moreover, these maps make the diagram in Fig. B.1 commutative and hence induce
a map f˜∗ : S P ∞ X → S P ∞ Y .
Remark B.2.17 S P ∞ satisfies the functorial properties.
Proposition B.2.18 Let Top ∗ be the category of pointed topological spaces.
(i) If f = 1 X in Top ∗ , then f ∗ = 1S P ∞ (X ) ;
(ii) If f : X → Y and g : Y → Z are in Top ∗ , then (g ◦ f )∗ = g∗ ◦ f ∗ : S P ∞ X →
S P∞ Z.
Hence it follows that S P ∞ is a covariant functor. It proves the following proposition.
Proposition B.2.19 S P ∞ : Top ∗ → Top ∗ is a covariant functor.
Example B.2.20 (i) For each integer n ≥ 1, S P ∞ S n is an Eilenberg–MacLane
space K (Z, n), which is a C W -complex having just one nontrivial homotopy
group πn (K (Z, n)) ∼
= Z.
(ii) S P n S 2 ≈ CP n .
(iii) S P ∞ S 2 ≈ CP ∞ .
B.3 Natural Transformations
This section introduces the concept of natural transformations. In some occasions

we have to compare functors with each other. We do this by means of suitable
maps between functors, called natural transformations which are very important in
algebraic topology. For example, this concept is used in homology and cohomology
theories, to compare homotopy and homology groups and applies to Yoneda lemma
and ‘representable functor’.
B.3.1 Introductory Concepts
Definition B.3.1 Let C and D be categories. Suppose T1 and T2 are functors of the
same variance (either both covariant or both contravariant) from C to D. A natural
transformation φ from T1 to T2 is a function from the objects of C to morphisms of D
such that for every morphism f : X → Y in C the appropriate one of the following
conditions hold:
φ(Y )T1 ( f ) = T2 ( f )φ(X ) (when T1 and T2 are both covariant functors)
or φ(X )T1 ( f ) = T2 ( f )φ(Y ) (when T1 and T2 are both contravariant functors).
Definition B.3.2 Let C and D be categories and T1 , T2 be functors of the same

variance from C to D. If φ is a natural transformation from T1 to T2 such that φ(X )
is an equivalence in D for each object X in C, then φ is called a natural equivalence.
Example B.3.3 Let R be a commutative ring and Mod be the category of R-modules
and R-homomorphism, M and N be objects in Mod . Suppose g : M → N is a
morphism in Mod . So by Example B.2.2(ii), π M , π N are both covariant functors
and π M , π N are both contravariant functors from Mod to itself. Then there exists a
natural transformation g∗ : π N → π M , where g ∗ (X ) : π N (X ) → π M (X ) is defined
by g∗ (X )(h) = hg for every object X in Mod and for all h ∈ π N (X ); and a natural
transformation
g∗ : π M → π N , where g∗ (X ) is defined in an analogous manner.
If g is an equivalence in Mod , then both the natural transformations g∗ and g ∗ are
natural equivalences.
Example B.3.4 If f : X → Y is a morphism in the category of pointed topological

spaces and continuous maps, then its corresponding adjoint morphism f˜ : X → Y
is defined by f˜([x, t]) = f (x)(t) for all x ∈ X and t ∈ I . The equivalence f ↔ f˜
comes from a natural equivalence from the functor mor (−, −) to the functor
(−, −).
B.3.2 Yoneda Lemma
This lemma provides important tools in algebraic topology. Moreover, it defines a

new functor called ‘representable functor’.
Lemma B.3.5 (Yoneda) Let C be any category and T be a covariant functor from
C to Set (category of sets and set functions). Then for any object C in C, there is
an equivalence θ = θC,T : (h C , T ) → T (C), where (h C , T ) is the class of natural
transformations from the set valued functor h C to the set valued functor T such that
θ is natural in C and T .
Proof Let η : h C → T be a natural transformation. Define θ : (h C , T ) → T (C),

η → η(C)(1C ) ∈ T (C) and ρ : T (C) → (h C , T ), given by ρ(x)(X )( f ) = T ( f )(x)
∀ x ∈ T (C), X ∈ C and f ∈ mor (C, X ). Then ρ(x) ∈ (h C , T ). Moreover, ρ ◦ θ =

identity and θ ◦ ρ = identity show that θ is an equivalence. it is left as an exercise to
show that θ is natural in C and T . ❑
Remark B.3.6 For detailed proof of Yoneda lemma see Adhikari and Adhikari
(2014). For dual of the result of Yoneda lemma see Ex. 11 of Sect. B.5.
Definition B.3.7 Let T : C → Set be a representable (contravariant) functor and C

is an object in C. If η : h C → T is natural equivalence then under the equivalence
θ given by Yoneda lemma, θ : (h C , T ) → T (C) the associated element θ(η) of η is
defined by θ(η) = η(C)(1C ) ∈ T (C).
Example B.3.8 Let Grp be the category of groups and homomorphisms; Set be the
category of sets and functions and S : Grp → Set be the forgetful functor which
assigns to each group G its underlying set SG. Then
(i) there is a natural equivalence from the covariant functor h Z (= H om(Z, −)) to
the covariant functor Set ; and
(ii) there is an equivalence θ : (S, S) → SZ.
B.3.3 Representable Functor
This functor is defined to prove Brown Representability Theorem. Let C be a category.

Then each object C of C defines a contravariant functor h C : C → Set which assigns
to an object X of C, the set mor (X, C) (the set of all morphisms for the object X to
the object C in C) and for the morphism f : X → Y in C, h C ( f ) = f ∗ : h C (Y ) →
h C (X ) is defined by f ∗ (g) = g ◦ f, ∀ g ∈ h C (Y ).
Definition B.3.9 Let C be a category. A contravariant functor T : C → Set is said

to be representable if there is an object C in C and a natural equivalence ψ : h C → T .
The object C is called a classifying space for T and C is said to be representable.
Definition B.3.10 If T is a representable functor and θ : h C → T is a natural equiv-

alence, then the associated element θ(η) = η(C)(1C ) ∈ T (C) is called the universal
element for T and C is called the representable object.
B.4 Convenient Category of Topological Spaces
This section describes an important category called the category of compactly gen-
erated Hausdorff spaces convenient for the study of homology and cohomology
theories. This category introduced by N. Steenrod (1910–1971) in 1967, includes
almost all important spaces in topology. For this section the paper Steenrod (1967)
and the book Gray (1975) are referred.
Definition B.4.1 A Hausdorff space is said to be a compactly generated Hausdorff

space if each subset which intersects every compact set in a closed set is itself closed.
The category of compactly generated Hausdorff spaces and their continuous maps
is denoted by CG.
Proposition B.4.2 Let X be a Hausdorff space and if for each subset S and each
limit point x of S, there exists a compact set C in X such that x is a limit point of
S ∩ C, then X ∈ CG.
Proof To prove this proposition it is sufficient to show that if each limit relation in
X lies in some compact subset of X , then X ∈ CG. Let S meet each compact set in a
closed set, and let x be a limit point of S. Then by hypothesis, there exists a compact
set C such that x is a limit point of S ∩ C. Since S ∩ C is closed, x ∈ S ∩ C and
hence x ∈ S. This shows that S is closed and hence X ∈ CG. ❑
Definition B.4.3 A topological space X is said to satisfy the first axiom of count-
ability if there exists a countable open base about every point in X and X is then
called a first countable space.
Example B.4.4 Every metrizable space is a first countable space.
Proposition B.4.5 The category CG contains all locally compact spaces and all
topological spaces satisfies the first axiom of countability.
Proof If X is locally compact, we take C to be the compact closure of a neighborhood

of x ∈ X . If X is first countable, we take C to consist of x and a sequence in S
converging to x. ❑
Corollary B.4.6 The category CG includes all metrizable spaces.
Remark B.4.7 All Hausdorff spaces are not in CG.
Example B.4.8 Let X be the set of all ordinal numbers preceding and including the
first noncountable ordinal . Endow X the topology defined by its natural order. Let
S be the subspace of the topological space X obtained by deleting all limit ordinals
except . Since each infinite set contains a sequence converging to a limit ordinal
of second kind, the only compact subsets of S are finite sets. This shows that the set
S − meets each compact set in a closed set, but is not closed in S, because it has
as a limit point.
Remark B.4.9 The Example B.4.8 shows that a subspace S of a compactly generated
topological space X need not be compactly generated.
Proposition B.4.10 Let X ∈ CG and Y be a Hausdorff space. If f : X → Y is a

quotient map, then Y ∈ CG.
Proof Let B ⊂ Y meet each compact set of Y in a closed set. If C is a compact

set in X , then f (C) is compact and hence B ∩ f (C) is closed. This shows that
f −1 (B ∩ f (C)) is closed and thus f −1 (B ∩ f (C)) ∩ C = f −1 (B) ∩ C is closed.
This implies that f −1 (B) meets each compact set of X in a closed set. As X ∈
CG, f −1 (B) is closed. Again since f is a quotient map, B must be closed in Y .
Consequently, Y ∈ CG. ❑
Remark B.4.11 It follows from the preceding results that the category CG is larger
in the sense that it contains almost all important spaces in topology.
For example, it contains all continuous maps between any two of its spaces.
The following proposition facilitates to examine the continuity of a function.
Proposition B.4.12 Let X ∈ CG and Y be a Hausdorff space and f : X → Y be

continuous on each compact subset of X . Then f is continuous.
Proof Let B ⊂ Y be closed and A be compact in X . Since Y is a Hausdorff space

and f | A is continuous, f (A) is compact and hence f (A) is closed in Y . This shows
that B ∩ f (A) is closed, and hence also
( f | A )−1 (B ∩ f (A)) = ( f −1 (B)) ∩ A.
Since X ∈ CG, f −1 (B) is closed in X . This proves the continuity of f . ❑
Definition B.4.13 Let X be a Hausdorff space. The associated compactly generated

topological space k(X ) is the set X with the topology defined as follows:
a closed set of k(X ) is a set that meets each compact set of X in a closed set.
If f : X → Y is a mapping of Hausdorff spaces, k( f ) denotes the same function
k(X ) → k(Y ).
Theorem B.4.14 (Steenrod) Let X be a Hausdorff space and k(X ) be its associated
compactly generated space. Then
(i) the identity function 1d : k(X ) → X is continuous;
(ii) k(X ) is a Hausdorff space;
(iii) k(X ) and X have the same compact sets;
(iv) k(X ) ∈ CG;
(v) if X ∈ CG, then 1d : k(X ) → X is a homeomorphism;
(vi) if f : X → Y is continuous on compact sets, then k( f ) : k(X ) → k(Y ) is
continuous;
(vii) 1d∗ : πn (k(X ), ∗) → πn (X, ∗) establishes a 1-1 correspondence for all n and
all ∗, where πn (X, ∗) is the nth homotopy group of (X, ∗).
Proof (i) Let A ⊂ X be closed in X and C be compact in X . Then C is closed in X

and hence A ∩ C is also closed in X . This implies that A is also closed in k(X ).
(ii) Since X is a Hausdorff space, (ii) follows from (i).
(iii) Let A be a compact subset in k(X ). Then by (i), A is also compact in X . If C

is compact in X , and C̃ is the same set C with its relative topology from k(X ).
Then the identity map C̃ → C is continuous by (i). We claim that its inverse
is also continuous. Let B be a closed set of C̃. Then by definition B meets
every compact set of X in a closed set. This shows that B ∩ C = B is closed
in C and hence the identity map C → C̃ is also continuous. Consequently, C̃
is compact.
(iv) Let a set A meet each compact set of k(X ) in a closed set. Then by (iii), A
meets each compact set of X in a compact set and hence in a closed set. This
shows that A is closed in k(X ).
(v) If follows from (iv).
(vi) It is sufficient to prove by Proposition B.4.12 that k( f ) is continuous on each
compact set of k(X ). Let C̃ be a compact set in k(X ). Then by (iii), the set C̃
endowed with the topology induced from X , which is C say, is compact and
the identity map C̃ → C is a homeomorphism. Since f |C is continuous, f (C)
is compact, and hence by (iii), f (C) is the same set f (C̃) with its topology in
k(Y ). Consequently, the map k( f )|C̃ : C̃ → f (C) factors into the composition
of f |C and the two identity maps
C̃ → C → f (C) → f (C̃).
Consequently, k( f )|C̃ : C̃ → f (C) is continuous and hence k( f ) : k(X ) →

k(Y ) is continuous.
(vii) Since the maps of closed cells into X coincide with the maps into closed cells
into k(X ) by (vi), then (vii) follows from (vi) as the sets under consideration
are derived from such mappings. ❑
Definition B.4.15 Let X and Y be in CG and their product X × Y (in CG) be k(X ×c
Y ), where ‘×c ’ denotes the product under the usual product topology.
Remark B.4.16 Given X, Y ∈ CG, X ×c Y may not be in CG but this product satisfies
the universal property given in Fig. B.2.
There are continuous projections p1 : X × Y → X and p2 : X × Y → Y such
that if f : Z → X and g : Z → Y are continuous, and Z is in CG, then there exists
a unique map F : Z → X × Y such that f = p1 ◦ F and g = p2 ◦ F.
Theorem B.4.17 Let X and Y be in CG. There are continuous projections p1 :

X × Y → Y and p2 : X × Y → Y such that if f : Z → X and g : X → Y are con-
tinuous, and Z is in CG, there exists a unique map F : Z → X × Y with f = p1 ◦ F
and g = p2 ◦ F as shown in Fig. B.2.
Proof Since the identity function
X × Y → X ×c Y
Fig. B.2 Existence of the

unique map F : Z → X × Y
is continuous by Theorem B.4.14 and the projections X ×c Y into X and Y are

also continuous, their compositions projecting X × Y into X and Y are continuous.
Consequently they are in CG. Let Z ∈ CG and f : Z → X, g : Z → Y be in CG.
Then f and g are the components of the unique mapping ( f, g) : Z → X ×c Y .
Since k(Z ) = Z and k(X ×c Y ) = X × Y , there exists a unique mapping k( f, g) :
Z → X × Y , satisfying the required properties. ❑
B.5 Exercises
1. Let VF be the category of vector spaces over a field F and their linear transfor-
mations. If D : VF → VF is given by D(V ) = V ∗ and D(T ) = T ∗ , where V ∗
is the dual space of V and T ∗ is the adjoint of T , show that D is a contravariant
functor.
2. Let Grp be the category of groups and their homomorphisms and C(G) be
the commutator subgroup of G (i.e., the subgroup generated by [g1 , g2 ] =
g1 g2 g1−1 g2−1 ). Show that C : Grp → Grp is a covariant functor.
3. Let Ab denote the category of abelian groups and their homomorphisms.
(a) For an abelian group G, let T (G) denote its torsion group. Show that T :
Ab → Ab defines a functor if T ( f ) is defined by T ( f ) = f |T (G) for every
homomorphism f in Ab such that
(i) f is a monomorphism is Ab implies that T ( f ) is also so;
(ii) f is an epimorphism in Ab does not always imply that T ( f ) is also so.
(b) Let p be a fixed prime integer. Show that T : Ab → Ab defines a functor,
where the object function is defined by T (G) = G/ pG and the morphism
function T ( f ) is defined by T ( f ) : G/ pG → H/ p H , x + pG → f (x) +
p H for every homomorphism f : G → H in Ab such that
(i) f is an epimorphism in Ab implies that T ( f ) is also so;
(ii) f is a monomorphism in Ab does not always imply that f is also so.
4. Show that the equivalences in the category

(i) Set are bijections of sets;
(ii) Grp are isomorphisms of groups;
(iii) Ring are isomorphisms of rings;

(iv) Mod R are isomorphisms of modules;
(v) Top are homeomorphisms of topological spaces.
5. Let X, Y be Hausdorff spaces and C(X, Y ) be the topological space of all contin-
uous maps X → Y with compact open topology. Define Y X = kC(X, Y ), where
kC(X, Y ) is the associated compactly generated topological space of C(X, Y ).
Show that the evaluation map E : C(X, Y ) ×C X → Y, ( f, x) → f (x) is con-
tinuous on compact sets. Moreover, if X and Y are in CG, show that E is con-
tinuous as a mapping Y X × X → Y .
6. Let X, Y and Z be in CG. Show that Z Y ×X ≈ (Z Y ) X .
7. If α : A → B is a retraction and also a monomorphism in a category C, prove
that α is an isomorphism.
8. Show that if α : A → B is an epimorphism in Set , then α0 ∈ Set 0 is a monomor-
phism.
9. Let X and Y be objects of a category Set and let g : X → Y be a morphism in
Set . Show that there is a natural transformation g∗ from the covariant functor
h Y to the covariant functor h X and a natural transformation g∗ from the con-
travariant functor h X to the contravariant functor h Y . Further show that if g is
an equivalence in Set , both these natural transformations g∗ and g ∗ are natural
equivalences.
10. Let A and C be objects of a category C. Using Yoneda Lemma, show that
(h C , h A ) ≈ C(A, C).
[Hint. Take T = h A . Then by Yoneda’s Lemma (h C , h A ) ≈ h A (C) = C(A, C).]
11. (Dual of Yoneda lemma) Let C be any category and T a contravariant functor
from C to Set . Show for any object C ∈ C, there is an equivalence θ : (h C , T ) →
T (C) such that θ is natural in C and T , where (h C , T ) is the class of natural
transformations from the set valued functor h C to the set valued functor T such
that θ is natural C and T .
12. (a) Let C be a category and T : C → Set be a contravariant functor. Show that
for each object C in C there is a one-to-one correspondence θ between natural
transformations η : h C → T and elements x ∈ T (C).
[Hint: Use dual result Yoneda lemma.]
(b) Let C be a category and T : C → Set be a representable functor. If C and C
are representable objects for T with universal elements x, x respectively.
Show that there is an isomorphism ψ : C → C such that T (ψ)(x ) = x.
13. (a) If T is a Brown functor and ∗ is a one-point space, show that T (∗) is a set
which also consists of a single element.
(b) Let T be a Brown functor and X = Y be the suspension for some topologi-
cal space Y . Show that T (X ) admits a group structure with the distinguished
element in the pointed set T (X ) as identity element.
B.6 Additional Reading
[3] Atiyah, M. F., K -Theory, Benjamin, New York, 1967.
References
Adhikari, M.R., Adhikari, A.: Basic Modern Algebra with Applications.

Springer, New Delhi (2014)
New York (2002)
Atiyah, M.F.: K -Theory. Benjamin, New York (1967)
Eilenberg, S., Mac Lane, S.: Natural isomorphism in group theory. Proc. Natl. Acad. Sci., USA 28,
537–544 (1942)
Eilenberg, S., Mac Lane, S.: General theory of natural equivalence. Trans. Am. Math. Soc. 58,
231–294 (1945)
ton (1952)
(1975)
Hatcher, A: Algebraic Topology. Cambridge University Press, Cambridge (2002)
MacLane, S.: Categories for the Working Mathematician. Springer, Heidelberg (1972)
Steenrod, N.: A Convenient Category of Topological Spaces. Michigan Math J. 14, 133–152 (1967)
List of Symbols
∅ : empty set
X ⊂ Y or Y ⊃ X : set-theoretic containment (not necessarily
proper)
N : set of natural numbers (or positive integers)
Z : ring of integers (or set of integers)
R : field of real numbers (or set of real numbers)
Q : field of rational numbers (or set of rational
numbers)
C : field of complex numbers (or set of complex
numbers)
H : division ring of quaternions (or set of quater-
nions)
pp(or p.) : particular page of reference
×, : product of sets, groups, modules, or spaces
∼
= : isomorphism between groups
≈ : homeomorphism between topological spaces
iff : if and only if
|X | : cardinal of a set X
Zn : ring of integers modulo n (or residue classes
of integers modulo n), 2
GL(V ) : general linear group on V , 7
G = Gi : direct product of family {G i : i ∈ I } of
i∈I
groups, 9
G =⊕i∈I G i : direct sum of family {G i : i ∈ I } of
groups, 10
G⊕H : direct sum of groups, 10
G∗H : free product of groups, 13
G x = {g ∈ G : g·x = x} : isotropy group or the stabilizer group of x, 16
G⊗H : tensor product of modules ,18
Cn : complex n-space, 22

DOI 10.1007/978-81-322-2843-1
600 List of Symbols
I : closed unit interval [0,1], 22

In : n-cube = {x ∈ Rn : 0 ≤ xi ≤ 1 for 1 ≤ i ≤ n}
for x = (x1 , x2 , . . . , xn ), 22
Dn : n-disk or n-ball = {x ∈ Rn : x ≤ 1}, 22
Sn : n-sphere = {x ∈ Rn+1 : x = 1} = ∂ D n+1
(the boundary of the (n + 1)-disk D n+1 ), 22
RPn : real projective space = quotient space of S n
with x and −x identified, 22
CP n : complex projective space = space of all com-
plex lines through the origin in the complex
space Cn+1 , 22
∂ I n = I˙n = I n − Int I n : boundary of I n , 26
|| || : X → R : norm function, 27
X/A : quotient space (endowed with quotient topol-
ogy), 33
Mf : mapping cylinder of f , 35
C
f : mapping cone of f , 35
: disjoint union of sets or spaces, 36
X ∨Y : wedge (one-point union), 36
X ∧Y : smash product, 36
X : suspension space, 36
Y X = F(X, Y ) : function space (endowed with compact open
topology), 37
Vr (Rn ) : Stiefel manifold of orthonormal r-frames in
Rn , 39
G r (Rn ) : Grassmann manifold of r-planes of Rn through
the origin, 39
GL(n, R) : general real linear group, 41
F : f g : F is a homotopy between f and g, 47
f g : homotopic maps, 47
F : f g : F is a path homotopy between f and g, 48
p
[X, Y ] : set of homotopy classes of maps from
X to Y, 53
Htp∗ : homotopy category of pointed spaces, 53
Htp : homotopy category of spaces, 54
(X, A) : topological pair, 55
X Y : homotopy equivalent spaces, 59
H -space : Hopf space, 66
H -group : Hopf group, 67
(X ) : loop space of pointed space X , 75
n X : iterated loop space of pointed space X , 75
:X∨X →X : folding map, 77
X : suspension space of a pointed space X , 78
: loop functor, 79
List of Symbols 601
: suspension functor, 80
R∞ : infinite dimensional Euclidean space, 85
C∞ : infinite dimensional complex space, 85
S∞ : infinite dimensional sphere, 86
HEP : homotopy extension property, 90
N DR : neighborhood retract, 95
DR : deformation retract, 96
S Pn : finite symmetric product functor, 96
S P∞ : infinite symmetric product functor, 96
(X, x 0 ) : set of all loops in X based at x0 , 110
π1 (X, x0 ) : fundamental group (Poincaré group) of X
based at x0 , 114
π1 : fundamental group functor, 119
p : R → S1 : exponential map, 126
f : lifting of f , 127
w( f ) : winding number of f , 129
deg f : degree of f , 129
d : [ f ] → deg f : degree function, 129
π1 (R2 − {0}) : fundamental group of punctured plane, 130
(x, v(x)) : vector field, 132
π1 (R3 − K ) ∼ = (π1 (S 3 − K )) : knot group of K , 133
(X, p, B) : covering space of B, 148
n
T n = S1 : n-dimension torus, 150
1
(S n , p, R P n ) : covering space of R P n , 153
Aut (X/B) : automorphism group of covering space of
(X, p) of B, 162
HLP : homotopy lifting property, 175
p
F → X −
→B : p : X → B is a fibration with F fiber
space, 176
Top ∗ : category of pointed topological spaces, 180
L(m, p) : lens space, 190
ξ : bunndle, 199
f ∗ (ξ) : induced bundle, 200
γn : canonical vector bundle, 205
δi j : Kronecker delta, 205
(S n , p, R P n , Z2 ) : real Hopf bundle, 211
(S 2n+1 , p, CP n , S 1 ) : complex Hopf bundle, 211
(S 4n+3 , p, HP n , S 3 ) : quaternionic Hopf bundle, 211
(X G , p, BG ) : numerable principal G-bundle, 222
G r (F n ) : Grassmann manifold of r-dimensional sub-
spaces of the vector space F n , 224
BO : G n (R∞ ), 228
BU : G n (C∞ ), 228
602 List of Symbols
B Sp : G n (H∞ ), 228
G L(n, F) : group of nonsingular n × n matrices
over F, 232
V ectn (B) : B-isomorphic classes of n-dimensional vector
bundles over B, 234
G n (F ∞ ) = G n (F ∞ ) : Grassmannian manifold, 235
n≤m
γn∞ : n-dimensional vector bundle over
G n (F ∞ ), 235
ξ⊕η : Whitney sum of vector bundles over B, 236
K F (B) : K -theory on F-vector bundles over B, 237
K (B) : Grothendieck group of vector bundles
over B, 245
n : standard n-dimensional simplex, 250
|K | : polyhedron of simplicial complex K , 253
st(v) : star of v, 255
ost(v) : open star of v, 255
N K (x) : simplicial neighbourhood of x ∈ |K |, 256
L K (x) : link of x ∈ |K |, 256
(K , L) : simplicial pair, 257
K : barycentric subdivision of K , 258
B(sp ) : barycenter of sp , 258
K (n) : nth barycentric subdivision of K , 258
μ(K ) : mesh of simplicial complex K , 260
E(K , v) : edge group of K based at v, 264
πn (X, x0 ) : nth homotopy groups of (X, x0 ), 274
πn (X, A, x0 ) : nth relative homotopy group of
(X, A, x0 ), 282
∂In : boundary of I n , 274
f g rel ∂ I n : f is homotopic to g rel ∂ I n , 275
∂ : boundary operator in homotopy theory, 283
Htp 2 : homotopy category of triplets, 285
πn : homotopy functor, 285
πn : cohomotopy functor, 296
PC H P : polyhedra covering homotopy property, 298
en : open n-cell, 306
X (n) : n-skeleton of cell complex, 307
CW-complex : cell complex having closure finite and weak
topology, 307
ψin : characteristic maps for CW-complex, 309
W S1 : Warsaw circle, 318
[α, β] : Whitehead product of α and β, 330
[α, β]GW : generalized Whitehead product of α
and β, 336
α, β!m : mixed product of α and β, 338
List of Symbols 603
α, β! : Samelson product of α and β, 338

τn : generator of πn (S n ), 341
nil (S 7 , μ, φ) : homotopical nilpotence of the H -space
(S 7 , μ, φ), 342
Zn : group of n-cycles, 350
Bn : group of n-boundaries, 350
Hn : homology functor, 352
Comp : category of chain complexes, 352
Ab : category of abelian groups, 352
cp ∼ dp : c p and d p are homologous p-cycles, 358
Z p (K ; G) : group of p-cycles of K with coefficients in G
for p ≥ 1, 358
B p (K ; G) : group of p-boundaries of K with coefficients
in G for p ≥ 1, 358
H p (K ; G) : p-dimensional simplicial homology group of
K with coefficients in G for p ≥ 1, 359
Z n = ker δ n+1 : group of n-cocycles, 366
B n = Im δ n : group of n-coboundaries, 366
Hn : n-dimensional cohomology, 366
H ∗ = {H n , δ n } : cohomology theory, 367
"n : standard n-simplex, 369
βn (X ) : βn (X ) = rank(Hn (X )) = nth Betti number
of X , 370
H∗ : reduced singular homology, 372

Hn (X, A) : relative singular homology groups of
(X, A), 373
C n (X ; G) = H om(Cn (X ; G), G) : singular n-cochain group, 375
H n (X, A) : relative singular chomology groups of
(X, A), 376
H n (X, A; G) : relative singular chomology group with coef-
ficient group G, 377
W∗ (X ) : cellular chain complex, 383
Hn (W∗ (X )) : cellular homology group, 383
Ȟ ∗ (X ; G) : Čech cohomology group, 385
χ(K ) : Euler characteristic of a simplicial complex
K , 388
ψ∪θ : cup product of ψ ∈ C n (X ; R) and θ ∈ C m
(X ; R), 393
σ∩ψ : cap product of σ ∈ Cm+n (X ; R) and ψ ∈
C n (X ; R), 395
GR : category of graded rings, 401
α×β : cohomology cross product of cellular
cochains, 401
l ∞ (m) = S ∞ mod Zm : infinite dimensional lens space, 408
604 List of Symbols
K (G, n) : Eilenberg-MacLane space of type (G, n), 408

M(G, n) : Moore space of type (G, n), 410
X [n] : nth Postnikov section of X in the Postnikov
tower (X [n] , X ), 411
S P ∞ (S n ) (or S P ∞ S n ) : infinite symmetric product of S n , 414
H : homology theory, 420
H(1) − H(7) : Eilenberg-Steenrod axioms for homology
theory, 421
C(1) − C(7) : Eilenberg-Steenrod axioms for cohomology
theory, 425
0 (X )
H : reduced 0-dimensional cohomology group of
X , 427
(X ; A, B) : topological triad, 429
deg f : degree of spherical map f , 447
f : Lefschetz number of f , 456
f : suspension map of f , 468
G n,k : Grassmann manifold of k-planes in Rn , 470
Vn,k : orthonormal k-frames in Rn , 470
E = {E n , αn } : spectrum of spaces, 476
HG : Eilenberg-MacLane spectrum, 476
n }
{E : spectral homology theory associated with the
spectrum E = {E n , αn }, 478
n }
{E : spectral cohomology theory associated with
the spectrum E = {E n , αn }, 480
K (X ) : Grothendieck group of finite dimensional
C W -complex X , 487
K ∗ (−) : K -theory (a generalized cohomology theory),
488
{E n } : spectral unreduced homology theory, 488
{E n } : spectral unreduced cohomology theory, 488
πrs (X ) : r th stable homotopy group (or r -stem)
of X , 498
h r (X ; A) : homology theory associated with the spec-
trum A, 498

Xn : X n = A ∪ X (n) (A is subcomplex of a finite
C W -complex X ), 515
C n+1 (X, A; πn (Y )) : (n + 1)th chain group of X modulo A with
coefficient group πn (Y ), 517
cn+1 ( f ) : obstruction cochain of f : X̃ n → Y , 517
H n (X, A; πn (Y )) : nth cohomology group of X modulo A with
coefficient group πn (Y ), 518
γ n+1 ( f ) : cohomology class of cn+1 ( f )) in H n+1
(X, A; πn (Y )), 518
d n ( f, g) = d n ( f, g : Ht ) : difference cochain of f and g, 519
On+1 ( f ) : obstruction set of f , 521
List of Symbols 605
c( f ) : obstruction cocycle of f , 525

H( f ) : Hopf invariant of f : S 2n−1 → S n , 540
A = { An , αn } : spectrum of spaces, 563
SL(n, R) : special linear group, 570
O(n, R) : orthogonal group, 570
SO(n, R) : special orthogonal group, 570
GL(n, C) : general complex linear group, 572
X mod G : orbit space of G-space X , 573
SU(n, H) : sympletic group, 577
C : category, 582
Grp : category of groups, 583
≈ : equivalence between objects in category
theory, 584
A≈B : A and B are equivalent objects in a category,
584
[G, G] : commutator subgroup of G, 587
G/[G, G] : abelianized group of G, 587
& : pair of adjoint functors, 587
S P n X (or S P n (X )) : nth symmetric product of X , 588
CG : category of compactly generated Hausdorff
spaces, 592
Author Index
A F
Adams, J.F. (1930–1989), 73, 198, 541 Feldbau, J. (1914–1945), 197, 560
Alexander, J.W. (1888–1971), 250, 349, 547 Freedman, M. (1951–), 549
Alexandroff, P.S. (1896–1982), 31, 555 Freudenthal, H. (1905–1990), 274, 550
Atiyah, M.F. (1929–), 236, 558
G
B Gauss, C.F. (1777–1855), 223
Betti, E. (1823–1892), 11, 349, 555 Grothendieck, Alexander (1928–2014), 236
Borsuk, K. (1905–1982), 174, 453, 553
Brouwer, L.E.J. (1881–1967), 138, 445, 533,
550 H
Brown, E.H. (1926–), 564 Heegaard, Poul (1871–1948), 47, 550
Hilbert, David (1862–1943), 569
Hilton, Peter John (1923–2010), 75
Hirzebruch, F.E. Peter (1927–2012), 236,
C
558
Cantor, George (1885–1918), 1
Hopf, H. (1894–1971), 31, 46, 197, 274, 290,
Cartan, E. (1869–1951), 274
511, 539, 550
Cartan, Henri Paul (1904–2008), 565
Hopf, H. (1895–1971), 446
Cauchy, A.L. (1789–1857), 551
Hurewicz, W. (1904–1956), 124, 198, 273,
Čech, E. (1893–1960), 273, 348, 419, 555
377, 534, 550
D J
Dehn, Max (1878–1952), 47, 550 Jordan, C. (1838–1922), 46
Dold, Albrecht (1928–2011), 538
Donaldson, S.K. (1957–), 576
K
Kan, Daniel (1927–2013), 79
E Klein, C.F. (1849–1925), 34, 547, 569
Eckmann, Beno (1917–2008), 75 Kolmogoroff, A. (1903–1987), 349, 556
Eilenberg, S. (1913–1998), 554, 581
Eilenberg, S. (1915–1998), 250, 274, 348,
407, 419, 511 L
Euler, L. (1703–1783), 107 Lefschetz, S. (1884–1972), 348, 420, 455,
Euler, L. (1707–1783), 46, 250, 547 555
DOI 10.1007/978-81-322-2843-1
608 Author Index
Leja, F. (1885–1979), 570 Schreior, O. (1901–1929), 570

Lie, Marius Sophus (1842–1899), 569 Serre, J.P. (1926–), 274, 560
Lie, Sophus (1842–1899), 213 Smale, Stephen (1930–), 549
Lima, F.L. (1929–), 475, 563 Stallings, John R. (1935–2008), 549
Listing, J.B. (1802–1882), 547 Steenrod, N. (1910–1971), 95, 419, 511,
Listing, J.B. (1808–1882), 250, 551 558, 591
Stiefel, E. (1909–1978), 197, 560
M
MacLane, S. (1909–2005), 407, 554, 581 T
Mayer, W. (1887–1947), 249, 349, 460, 550 Thom, René (1923–2002), 538
Mayer, W. (1887–1948), 559 Tietze, H. (1880–1964), 553
Milnor, J.W. (1931–), 220, 561, 576 Tietze, H. (1888–1971), 190
Möbius, A.F. (1970–1868), 34
U
N Ulam, S. (1909–1984), 453
Noether, Emmy (1882–1935), 349, 555
P V
Perelman, G.Y. (1966–), 549 van Kampen, E. (1908–1942), 267, 551
Poincaré, H. (1854–1912), 46, 107, 249, 273, Veblen, O. (1880–1960), 250, 550
348, 446, 547 Vietoris, L. (1891–2002), 348, 419, 460,
Pontryagin, L. (1908–1988), 556 555, 556
Postnikov, M. (1927–2004), 411, 511, 565
W
R Whitehead, G.W. (1918–2004), 563
Reidemeister, K. (1893–1971), 553 Whitehead, J.H.C. (1904–1960), 553
Riemann, B. (1826–1866), 46, 548, 549 Whitney, H. (1907–1989), 197, 349
S Z
Samelson, H. (1916–2005), 338 Zeeman, E.C. (1925–2016), 250, 549
Subject Index
A Boundary operator, 283

Abelianization functor, 587 Bouquet of spaces, 143
Abelianization of knot group, 143 Brouwer fixed point theorem, 133, 138, 451,
Action of a group, 15 456
Action of π1 on πn , 294 Brouwer no retraction theorem, 450
Action of topological group, 572 Brouwer’s degree theorem, 447
Acyclic space, 381 Brown functor, 463, 588
Adams theorem, 542 Brown representability theorem, 481, 564
Adjoint functor, 79–81, 587 Browns functor, 565
Adjoint pair of functors, 587 Bundle, 199
Admissible open set, 148 Bundle of groups, 206
Algebraically closed (complete) field, 137
Algebraic completeness, 137
Analysis situs, 548 C
Anticommutativity of the cup product, 540 Canonical line bundle, 205
Antipodal degree theorem, 448 Canonical morphism, 203
Antipodal map, 138 Canonical vector bundle, 205
Application of Galois correspondence in Cantor’s theorem, 4
covering spaces, 191 Cap product, 395
Applications beyond mathematics, 464 Category, 582
Aspherical space, 99 Category of compactly generated Hausdorff
Automorphism group of the covering spaces, space, 591
162 Category theory, 581
Cauchy’s integral theorem, 140
Čech homology, 385
B Cell complex, 307
Baire space, 29 Cellular approximation theorem, 319
Banach space, 27, 85 Cellular homology, 383
Barycentric subdivision, 258 Cellular homotopy, 315
Base space of a bundle, 199 Cellular map, 307
Betti number, 11, 456 Cellular space, 312
Bifunctor, 586 Chain complex, 349
B-isomorphism, 203 Chain group, 350
Blakers-Massey theorem, 320 Chain map, 350
Bockstein sequence, 507 Characteristic map, 309
Borsuk–Ulam theorem, 139, 453 Chart of principal G-bundle, 229
Bott periodicity theorem, 229 Circle group, 576
DOI 10.1007/978-81-322-2843-1
610 Subject Index
Circle knot, 136 D

Classical Lie groups, 576 Deformable into, 91
Classical topological groups, 570 Deformation, 91
Classification of covering spaces, 167 Deformation retraction, 88, 92
Classification of vector bundles, 233, 235 Degree function, 129, 446
Classification theorem of covering spaces, Degree of circular map, 129
169 Determinant function, 571
Classifying map, 176 Difference cochain, 519
Classifying space, 176, 221 Differential manifold, 38
Closure-finite, 309 Dimension of a finite simplicial complex,
Coboundary formula, 520 254
Dimension of a Lie group, 575
Cochain complex, 366
Dimension of CW-complex, 325
Cofibration, 180, 440
Disjoint union, 22
Cohomology class, 376, 518
Division algebra, 542
Cohomology cross product, 401 Dodecahedron, 460
Cohomology group, 376 Dold–Thom theorem, 538, 539
Cohomology operation, 489 Double suspension, 539
Cohomology operation of degree k, 491 DR-pair, 96
Cohomology ring, 394 Dual functor, 585
Comb space, 87 Dual of Yoneda lemma, 596
Commutator subgroup, 378, 535
Compactly generated space, 29
Compact open topology, 37 E
Compact space, 29 Edge group, 264
Complete metric space, 27 Effective action, 573
Complex Hopf bundle, 211 Eigenvector, 21
Complex Lie group, 575 Eilenberg extension theorem, 520
Complex projective space, 22, 574 Eilenberg–MacLane space, 408, 538
Complex n-space, 22 Eilenberg–MacLane spectrum, 476, 563
Comultiplication, 78 Eilenberg obstruction theory, 512
Concrete category, 586 Eilenberg–Steenrod axioms for cohomology
Conjugacy class, 163 theory, 425
Connected space, 28 Eilenberg–Steenrod axioms for homology
0-connected space, 288 theory, 421
Continuous extension, 97 Eilenberg-Zilber theorem, 374
Continuous vector field, 451 Elementary p-chain, 356
Contour, 144 Equivalence in category, 584
Equivalence theorem, 552
Contractible space, 83
Equivalent knots, 134
Contraction, 83
Equivalent objects, 584
Contravariant functor, 585
Equivalent transition functions, 230
Coretraction, 583 Essential map, 61
Countably compact space, 30 Euclidean n-space, 22
Covariant functor, 585 Euler characteristic, 388, 458
Covering (deck) transformations, 162 Euler characteristic of a finite graph, 143
Covering homomorphism, 161 Euler–Poincaré theorem, 392
Covering projection, 149 Evaluation map, 596
Covering space, 148 Evenly covered, 148
Cross section of a bundle, 200 Exact homology sequence, 364, 373, 401
Cross section problems, 515 Exactness of homotopy sequence, 286
Cup product, 368, 393, 400 Exact sequence of cohomology groups, 377
CW-complex, 309 Excision axiom, 422
Subject Index 611
Exponential map, 126 Generalized Jordan curve theorem, 397

Extension index, 516 Generalized Whitehead product, 336
Extension index of map, 516 General linear group, 570
Extension-lifting square, 514 Graded ring, 394
Grassmann manifold, 39, 205
Grothendieck construction, 237
F Grothendieck group, 237, 245
Faithful functor, 585 Grothendieck ring, 237
Fiber bundle, 207 Group of n-boundaries, 376
Fiber of a bundle, 199 Group of n-cocycles, 376
Figure-eight knot, 134 Group of p-cycles, 358
Finite symmetric product, 96 G-space, 573
First countable space, 592
Five lemma, 12
Fixed point property, 456
H
Flow of maps, 54
Hairy Ball Theorem, 453
Forgetful functor, 585
Ham Sandwich Theorem, 454
Four lemma, 19
H -cogroup, 76
Free action, 16, 217, 574
h-equivalent spaces, 536
Free group, 8, 462
H -group (or Hopf’s group), 67
Free homotopy class, 144
Hilbert’s fifth problem, 569
Free product, 13
Hilbert space, 27
Freudenthal suspension homomorphism,
Homeomorphism, 25, 26
535
Freudenthal suspension theorem, 294 Homological description of Moore space,
Full functor, 585 543
Functorial properties of homotopy groups, Homological version of Whitehead theorem,
278 526
Functorial properties of relative homotopy Homology theory associated with a
groups, 284 spectrum, 498
Functorial property of π1 , 117 Homomorphism of H -cogroups, 77
Fundamental group of figure-eight, 154 Homotopic maps, 47
Fundamental group of figure-eight by graph- Homotopic paths, 48
theoretic method, 191 Homotopical description of cohomology
Fundamental group of Klein’s bottle, 189 theory, 505
Fundamental group of lens space, 190 Homotopical nilpotence, 342
Fundamental group of orbit space, 189 Homotopy abelian, 67
Fundamental group of the double torus, 154 Homotopy associative, 72
Fundamental theorem of algebra, 99, 137 Homotopy axiom, 400
Fundamental group (Poincaré group), 114 Homotopy category, 54
Homotopy class, 53
Homotopy equivalence, 58
G Homotopy equivalent spaces, 59
Galois correspondence theorem for covering Homotopy extension property (HEP), 90
spaces, 169 Homotopy group πn (X, x0 ), 276
Gauss map, 227 Homotopy groups associated with a
G-bundle, 214 spectrum, 499
General complex linear group, 572 Homotopy homomorphism, 68
Generalization of Eilenberg–MacLance Homotopy index, 522
spectrum, 485 Homotopy invariance, 279
Generalized cohomology theory, 422, 560 Homotopy invariance of dimensions of
Generalized (extraordinary) cohomology spheres, 398
theory, 481 Homotopy invariant, 118
Generalized homotopy lifting property, 158 Homotopy inverse, 67
612 Subject Index
Homotopy lifting property, 158 Isomorphic (equivalent) covering space, 161

Homotopy sequence, 286 Isomorphic (equivalent) principal G-
Homotopy sequence of fibering, 289 bundles, 216
Homotopy type, 60 Isomorphism of bundles, 202
Hom R functor, 585 Isotopic, 135
Hopf classification theorem, 449, 450
Hopf degree theorem, 294
Hopf fibering, 210, 290 J
Hopf integer, 540 Jordan curve theorem, 397
Hopf invariant, 539, 541
Hopf invariant one problem, 487
Hopf map, 290 K
Hopf theorem, 414, 524 K (G, 1) space, 408
Hopf trace theorem, 467 K (G, n) space, 408
H -space (or Hopf’s space), 66 Killing homotopy theorem, 410
Hurewicz fiber space (fibration), 176 ‘Klein’ bottle, 34
Hurewicz fibering, 212 Knot, 133
Hurewicz homomorphism, 377, 535 Knot group, 133
Hurewicz theorem, 183, 377, 535 Kronecker delta, 205
Hurewicz theorem for fibration, 213 K -theory, 237, 487, 560
Hurewicz theorem on homotopy groups, 294 K -theory of a point, 237
Künneth formula, 374
I L
Icosahedron, 460 Lebesgue lemma, 31
Identification map, 33 Lefschetz number, 456
Incidence number, 353 Lie group homomorphism, 575
Infinite complex projective space, 68 Lie group of rotations, 143
Infinite dimensional complex projective Lifting map, 127
space, 408 Lifting problem, 158, 514
Infinite dimensional complex projective Lifting theorem, 159
space CP ∞ , 30 Link between Cohomolgy and Homotopy,
Infinite dimensional Euclidean space, 85 523
Infinite-dimensional Hilbert space, 312 Local cross section, 209
Infinite dimensional lens space, 408 Locally compact space, 29
Infinite dimensional real projective space, Locally connected space, 28
408 Locally path-connected space, 155
Infinite dimensional real projective space Locally trivial principal G-bundle, 215
R P ∞ , 30 Loop space, 74
Infinite dimensional sphere, 86 Lusternik–Schnirelmann, 455
Infinite dimensional sphere S ∞ , 30
Infinite dimensional unitary space, 85
Infinite earing, 171 M
Infinite (finite) spirals, 150 Manifold, 38
Infinite real projective space, 68 Map between spectra, 477
Infinite symmetric product, 96 Mapping cone, 35
Infinite symmetric product functor, 538, 588 Mapping cylinder, 35
Infinite symmetric product space, 589 Maximal tree, 266
Invariance of cohomology groups, 428 Mayer–Vietoris axiom, 588
Invariance of dimensions of simplicial com- Mayer–Vietoris sequence, 380, 461
plexes, 254 Mayer–Vietoris theorem, 365, 429
Invariance of homology groups, 428 Mesh of a simplicial complex, 260
Invariant factors, 11 Metric space, 27
Subject Index 613
m-extensibility of a map, 522 P

Milnor construction, 220 Paracompact space, 29, 31
Milnor’s universal bundle, 221 Partial homotopic mpas, 519
Mixed product, 338 Partition of unity, 31
Möbius band (or strip), 34 Pasting (or gluing lemma), 49
Monodromy theorem, 185 Path, 109
Moore space, 410, 538 Path homotopy, 48
Morphism of bundles, 201 Path homotopy classes, 50
Path lifting property, 128, 514
π1 acts trivially on πn , 516
N Platonic solid, 459
Natural equivalence, 79, 590 Poincaré-Alexander theorem, 389
Natural topology, 23 Poincaré conjecture, 549
Natural transformation, 589 Poincaré duality, 469
n-boundaries, 350 Poincaré duality theorem, 501
n-cell, 307 Polyhedron, 253
n-coboundaries, 366 Polytope, 308
n-cocycles, 366 Postnikov section, 411
n-connected space, 288, 378 Postnikov tower, 411
n-cube, 22 Principal (differentiable) G-bundle, 240
n-cycles, 350 Principal fibration, 176
n-dimensional cohomology group, 366 Product bundle, 200
n-dimensional homology group, 350 Projection map of a bundle, 199
n-dimensional torus, 150 Properly discontinuous, 208
n-disk, 22 p-simplex, 251
NDR-pair, 95
n-extensible map, 516
Q
Normal pair, 99
q-stem (qth stable homotopy group), 498
Normal vector, 205
qth Betti number, 388
Normal vector field, 205
Quaternionic Hopf bundle, 211
Normed linear space, 27
Quotient space, 33
Notations in obstruction theory, 515
n-sheeted (n-fold) covering space, 153
n-simple space, 516 R
n-skeleton, 255, 307, 309 rank Hn (X ), 392
n-sphere, 22 Real division algebra, 21
nth cohomotopy set, 296 Real Hopf bundle, 211
Nullhomotopic, 97 Real Lie group, 575
Nullhomotopic (or inessential map), 61 Real projective plane, 34, 152
Numerable principal G-bundle, 216 Real projective space, 22, 574
Reduced cohomology theory, 480
Reduced homology groups of spheres, 382
O Reduced homology theory, 478
Obstruction cochain, 517 Reduced singular homology group, 372
Obstruction cocycle, 525 Regular covering space, 188
Obstruction for homotopy between relative Regular simple polyhedra, 459
lifts, 526 Relative cohomology group, 377
Obstruction set, 521 Relative CW-structure, 314
One parameter subgroup, 576 Relative homology group, 363
One-point union of n-cells, 381 Relative homotopy group, 281
Open p-simplex, 252 Relative Hurewicz theorem, 536
Orbit space, 573 Relative lifting problem, 514
Orthogonal group, 570 Relative singular homology group, 373
614 Subject Index
Representable functor, 565, 591 Suspension space, 36, 78

Representation of functor V ectn , 235 Suspension spectrum, 477
Restricted bundle, 202 Symmetric product space, 588
Retract, 88 Sympletic group, 577
Retraction, 88
Riemann sphere, 3
Ring spectrum, 500 T
Table of πi (S n ) for 1 ≤ i, n ≤ 8, 291
Tangent bundle, 204
S Tangent vector, 204
Samelson product, 338 Tangent vector field, 204, 452
Semilocally path-connected, 167 Tensor product, 14
Semilocally simply connected, 167 Three lemma, 19
Serre fibration, 176, 324 Three utility problem, 399
Shearing map, 217 Tietze extension theorem, 32
Simplicial approximation theorem, 263 Topological invariant, 118, 361, 399
Simplicial cohomology ring, 367 Topological pair, 55
Topological triad, 429
Simplicial cohomology theory, 367
Topologist’s sine curve, 41
Simplicial complex, 253
Torsion product, 14
Simplicial homology group, 358
Torus, 33, 574
Simplicial map, 257
Torus knot, 135
Simplicial pair, 257
Total space of a bundle, 199
Simply connected, 120
Transition function, 229
Singular n-cochain group, 513
Trefoil knot, 134
Smash product (or reduced product), 36
Triangulable space, 253
Special linear group, 570
Triangulation, 253
Special orthogonal group, 570
Trivial bundle, 204
Spectral cohomology, 476
Trivial knot (unknot), 134
Spectral homology, 476
Trivial vector bundle, 226
Spectral unreduced cohomology theory, 488
Spectral unreduced homology theory, 488
Spectrum of spaces, 476 U
Sphere spectrum, 477 Unitary spectrum, 477
Sphere with n handles, 143 Universal coefficient theorem for homology,
Square knot, 134 386, 387
Stabilizer group, 16 Universal covering space, 168, 170
Stable cohomology operation, 489 Universal element, 463, 591
Stable cohomology operation of degree k, Universal principal G-bundle, 219, 222
492 Unreduced cohomology theory, 488
Stable cohomology operation of type Unreduced homology theory, 488
(G, n; T, m), 490 Untied, 136
Stable homotopy group, 496 Urysohn lemma, 32
Star convex, 120
Steenrod algebra, 562
Steenrod theorem, 104, 593 V
Stepwise extension of a cross section, 524 van Kampen theorem, 268
Stiefel manifold, 39, 205 Vector bundle, 223
Strong deformation retract, 92 Vector field, 132
Structure theorem, 11
Subcategory, 583
Subcomplex of a CW-complex, 313 W
Subpolyhedron, 255 Warsaw circle, 318
Suspension functor, 80 Weak deformation retract, 92
Subject Index 615
Weak retract, 89 Whitehead tower, 415

Weak topology, 30, 309 Whitney sum, 236
Wedge (or one-point union), 36 Winding number, 129
Wedge axiom, 588
Whitehead product, 329
Whitehead theorem, 317, 414
Whitehead theorem in homological form, Y
537 Yoneda lemma, 590

Mahima Ranjan Adhikari - Basic Algebraic Topology and Its Applications-Springer (2016)

Uploaded by

Copyright:

Available Formats

Mahima Ranjan Adhikari - Basic Algebraic Topology and Its Applications-Springer (2016)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mahima Ranjan Adhikari - Basic Algebraic Topology and Its Applications-Springer (2016)

Uploaded by

Copyright:

Available Formats

Mahima

Basic Algebraic Topology

ISBN 978-81-322-2841-7 ISBN 978-81-322-2843-1 (eBook)

Library of Congress Control Number: 2016944483

© Springer India 2016

Printed on acid-free paper

This Springer imprint is published by Springer Nature

Algebraic topology is one of the most important creations in mathematics which

applications of homotopy in dealing with some extension problems, lifting problems,

CW-complexes is a suitable category for a systematic study of algebraic topology.

Chapter 14 presents further interesting applications of the homotopy, homology,

Chapter 17 presents some similarities and interesting relations among homotopy,

(i) Homology groups for compact metric spaces introduced by L. Vietoris

more deeply, existence) of mappings. So the readers of algebraic topology cannot

Kolkata, India Mahima Ranjan Adhikari

1 Prerequisite Concepts and Notations . . . . . . . . . . . . . . . . . . . . . . . 1

1.12 Separation Axioms, Urysohn Lemma, and Tietze

3.6 Vector Fields and Fixed Points . . . . . . . . . . . . . . . . . . . . . . 132

4.10.3 Fundamental Group of the Real Projective

6.7.1 Application to Extension Problem. . . . . . . . . . . . . . 265

8.3 Subcomplexes of CW-Complexes . . . . . . . . . . . . . . . . . . . . . 313

10.3 Relative Simplicial Homology Groups . . . . . . . . . . . . . . . . . 362

10.20 Invariance of Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . . 398

15.9.2 Cohomology Operation Associated

17 More Relations Between Homology and Homotopy . . . . . . . . . . . . 533

Appendix A: Topological Groups and Lie Groups . . . . . . . . . . . . . . . . 569

Appendix B: Categories, Functors and Natural Transformations. . . . . . 581

List of Symbols. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599

Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607

Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609

Mahima Ranjan Adhikari, Ph.D. is the former Professor of Pure Mathematics at

1.1 Set Theory

N (set of natural numbers/positive integers)

© Springer India 2016 1

Definition 1.1.1 Let X be a nonempty set and ρ be an equivalence relation on X .

For i = j, X i may be equal to X j . The collection of sets {X i : i ∈ I } is finite or

Definition 1.1.5 A relation ρ on X is said to be antisymmetric if (x, y) ∈ ρ and

Remark 1.1.7 Zorn’s lemma is indispensible to prove many results of mathematics.

Definition 1.1.8 A map f : X → Y is said to be

Definition 1.1.9 Let f : X → Y and g : Y → Z be two maps. Their composite map

Proposition 1.1.10 Let f : X → Y and g : Y → Z be two maps. Then g ◦ f has

The bijections of sets define equivalent sets.

Definition 1.1.11 Two sets X, Y are said to be equivalent denoted by X ∼ Y if there

Example 1.1.12 Let S 2 = {(x, y, z) ∈ R3 : x 2 + y 2 + z 2 = 1} be the unit sphere in

Example 1.1.13 A permutation of a set X is a bijection from X onto itself.

Definition 1.1.14 A set X is said to be finite if either X is empty or X ∼ Zn for

Example 1.1.17 Z, N, R, C are infinite sets but Zn is a finite set.

Definition 1.1.18 A set X is said to be countable if either X is finite or there exists

Proposition 1.1.19 Every infinite set contains a countable set.

The notion of counting is extended by assigning to every set X (finite or infinite)

1.2 Groups and Fundamental Homomorphism Theorem

An arbitrary subset of a group forming a group, called a subgroup contained in a

Proposition 1.2.6 A nonempty subset H of a group G is a subgroup of G if and only

Theorem 1.2.9 (Lagrange Theorem) The order of a subgroup of a finite group

Theorem 1.2.11 Let G be a finite group of order n and m be a positive divisor of n.

We now define a special map between groups, called a homomorphism which

Definition 1.2.12 Let G and H be groups. Then a map f : G → H is said to be a

Particular homomorphisms carry special names having interesting properties.

Definition 1.2.13 Let f : G → H be a homomorphism of groups. Then

Definition 1.2.22 If S is a nonempty subset of a group G, then S is called the

Example 1.3.12 (i) Zn = x : x n and Z = x : ∅.

Example 1.3.14 The cyclic group Zn has the presentation x : x n .

b ∈ B, the cyclic group b is infinite; and

Fig. 1.2 Free product G ∗ H G; ∗ HcG