Bordism Skript

A lecture course on Cobordism Theory
Johannes Ebert
Mathematisches Institut Universität Münster Einsteinstr. 62
48149 Münster, Germany
Winter term 2011/12; Manuscript typeset by Felix Haas
Contents
1 Differential Topology 2
1.1 Transversality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Orientations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 The cobordism relation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2 Orientation and (Co)homology 7

2.1 Orientation and Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Homological Interpretation of the Mapping Degree . . . . . . . . . . . . . . . . . . . . . 9
2.3 The Leray - Hirsch theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Thom-Isomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 Fiber Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.6 Homotopy Invariance of Fiber Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.7 Classification of G-Bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3 Characteristic Classes 26
3.1 Definition and Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Universal Classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3 Cohomology of classfying spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4 Multiplicative Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.5 Bordism vs. Homotopy: The Pontrjagin-Thom Construction . . . . . . . . . . . . . . . 37
3.6 Pontrjagin-Thom Construction and Homology . . . . . . . . . . . . . . . . . . . . . . . 38
4 Spectra and the Bordism Ring 39

4.1 Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 Calculation of the Oriented Bordism Ring . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3 The Signature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
COBORDISM THEORY 2
1 Differential Topology
Definition 1.0.1 Regular and Singular Value Let M m , N n , be two smooth manifolds and f ∶
M → N be a smooth map. An element x ∈ N is a regular value if for all y ∈ M with f (y) = x, the
differential Ty f ∶ Ty M → Tx N is surjective. A singular value is an x ∈ N which is not a regular
value.
Remark 1.0.2 If m < n, then the condition that x ∈ N is a regular value means that x does not
lie in the image of f .
Definition 1.0.3 Let M m ⊂ N n be a submanifold. The codimension of M is the number n − m.
Proposition 1.0.4 If x ∈ N is a regular value of the map f ∶ M m → N n , then f −1 (x) ⊂ M is a

submanifold of codimension n.
We now come to the most important fundamental result of differential topology.
Definition 1.0.5 Let M m be a smooth manifold and C ⊂ M be a subset. We say that C has
φ
measure zero if for each chart M ⊃ U → Rn , the set φ(C ∩ U ) ⊂ Rn has Lebesgue measure zero.
This notion is well defined because of two facts: a C 1 -diffeomorphism of open subsets of Rn maps
sets of measure zero to measure zero; and any manifold has a countable atlas. Note that the union
of countably many sets of measure zero has again measure zero and that the complement of a set
of measure zero is never empty (unless M itself is empty). The following theorem is fundamental
for all of differential topology.
Theorem 1.0.6 Sard′ s Theorem Let f ∶ M m → N n be a smooth map. Then the set of singular
values of f Crit(f ) ⊂ N has measure zero.
For the proof, see [1], p. 58 ff. The most immediate application, or rather, the simplest special
case, of Sard’s theorem is:
Corollary 1.0.7 If n > m and f ∶ M m → N n is smooth, then f (M ) ⊂ N has measure zero, hence
N ∖ f (M ) ≠ ∅.
Theorem 1.0.8 The Whitney Embedding Theorem Let M m , N n be manifolds and A ⊂ M

be closed. Assume that n ≥ 2m + 1. Let N carry a complete metric. Let f ∶ M → N be smooth
and assume that there exists a neighborhood A ⊂ U such that f ∣U is an injective immersion. Let
∶ M → (0, ∞) be a continuous function. Then there exists an injective immersion g ∶ M → N such
that g∣A = f ∣A and such that d(f (x), g(x)) ≤ (x).
We discussed the proof that is given in [2], section 15.7, but only in the case N = Rn . Recall that a
proper injective immersion is an embedding. In particular, if we started with a proper map f and
a bounded , the resulting g will be an embedding. Each manifold admits a proper map to R: pick
a countable, locally finite cover by relatively compact open sets Ui of M , let fi ∶ M → R≥0 be a
function with compact support and fi ∣Vi = 1, where Vi ⊂ Ui are smaller subsets that cover X. Then
the function f = ∑i ifi is proper. So any manifold has a proper map to Rn for each n. This shows
that we can realize any manifold as a closed submanifold of Rn for n ≥ 2m + 1. In particular, any
manifold admits a complete metric.
Definition 1.0.9 Let f ∶ M → N be an immersion. The normal bundle of f is the vector bundle
ν f ∶= f ∗ T N /T M on M . If f is the inclusion of a submanifold, we write νM
N
or simply νM for the
normal bundle.
Remark 1.0.10 Suppose that N has a Riemann metric. Then the normal bundle is isomorphic
to the orthogonal complement of T M in the Riemannian vector bundle f ∗ T N .
COBORDISM THEORY 3
Before we define tubular neighborhoods, we add a small remark on vector bundles. Let π ∶ E → M
be a smooth vector bundle. There is a canonical bundle monomorphism η ∶ E → T E∣M , sending
e ∈ E to the equivalence class of the curve R → E; t ↦ te. Moreover, the differential of π yields a
vector bundle epimorphism T π ∶ T E∣M → T M . These maps yield an exact sequence
0 → E → T E∣M → T M → 0, (1)
which admits a canonical splitting, namely the differential of the zero section ι ∶ M → E.
Definition 1.0.11 Tubular Map Let f ∶ M → N be an embedding. A tubular map for f is a

smooth map t ∶ ν f → N with the following properties:
1. The restriction of t to the zero section is equal to f .
2. There is an open neighborhood U ⊂ ν f of the zero section such that t∣U is an embedding.
η Tt
3. The composition ν f → T ν f → f ∗ T N → ν f (the first map was described before the definition,
the third map is the quotient map) is the identity map.
Theorem 1.0.12 Each embedding has a tubular map.
We followed the proof given in [2] and we used a lemma from [1]. To gain some flexibility in dealing
with tubular maps, we need two lemmata.
Lemma 1.0.13 There exists a smooth map σ ∶ R+ × R+ × [0, 1] × Rn → Rn , (r, , t, x) ↦ σtr, (x) such
that:
1. for each fixed r, , the map (t, x) ↦ σtr, (x) is an isotopy.
2. σ0r, = id.
3. σtr, ∣Br (0) = id.
4. σ1r, (Rn ) = Br+ (0).
Using this lemma, we show: if t ∶ ν f → N is a tubular map, then there exists a tubular map
t̃ ∶ ν f → N that coincides with f on a neighborhood of the zero section and that embeds the whole
of ν f into N . The image of such a tubular map is called a tubular neighborhood. The following
theorem explains in which sense a tubular neighborhood is uniquely determined by f .
Theorem 1.0.14 Let f ∶ M → N be an embedding and t0 , t1 ∶ ν f → N be tubular maps. Then

there exists an isotopy from t0 to t1 .
An application of the Whitney embedding theorem and tubular maps is the following theorem.
Theorem 1.0.15 Smooth approximation theorem Let M, N be manifolds, N with a complete

Riemann metric. Let f ∶ M → N be a continuous map.
1. There exists a function ∶ M → (0, ∞), such that any function g ∶ M → N which satisfies
d(f (x), g(x)) < (x) is homotopic to f .
2. For any continuous function ∶ M → (0, ∞), there is a smooth map g ∶ M → N with
d(f (x), g(x)) < (x). If there is a closed A ⊂ M and f is already smooth on a neighbor-
hood of A, we can pick g to coincide with f on A.
COBORDISM THEORY 4
1.1 Transversality
The most important notion in differential topology is transversality.
Definition 1.1.1 Let M ; X; Y be smooth manifolds, f ∶ X → M , g ∶ Y → N be smooth maps. Then

f and g are called transverse if for all x, y with f (x) = g(y) = z, we have that
(Tx f )(Tx X) + (Ty g)(Ty Y ) = Tz Z.
We write f ⋔ g when f and g are transverse. If g is the inclusion of a submanifold Y ⊂ M , we

write f ⋔ Y .
One of the important properties of transversality is:
Proposition 1.1.2 Let M ; X; Y be smooth manifolds, f ∶ X → M be a smooth map and Y ⊂ N .

Assume that f ⋔ Y . Then f −1 (Y ) ⊂ X is a submanifold and the normal bundles are related by
νfX−1 (Y ) ≅ f ∗ νYM .
The proof is an exercise. The following result is one of the key results of differential topology:
Theorem 1.1.3 The Transversality Theorem Let f ∶ M → N and g ∶ L → N be smooth maps.

Let N be endowed with a Riemann metric and ∶ M → (0, ∞) be a function. Then there exists a
map h ∶ M → N with d(f (x), h(x)) < (x) such that h ⋔ g. If A ⊂ M is closed and if f is transverse
to g at all points of A, then we can choose h to coincide with f on A. Moreover (due to Theorem
1.0.15), if is suitably small, then h will be homotopic to f relative to A.
The proof presented in the lecture was taken from [1], §14. The first application of transversality
is to computations of some homotopy groups (see any book on algebraic topology, e.g. [2], for the
definition of homotopy groups).
Theorem 1.1.4 Let M m and N n be smooth manifolds, g ∶ N n → M m be a smooth map. Let

x ∈ M ∖ g(N ) and j ∶ M ∖ g(N ) → M be the inclusion. The map πi (M ∖ g(N ), x) → πi (M, x)
induced by j is an epimorphism if j ≤ m − n − 1 and an isomorphism if j < m − n − 1.
By Theorem 1.0.15, any homotopy class has a smooth representative, which can be chosen to
be tranverse to g. A dimension count and a similar argument, applied to homotopies, gives the
proof.
Example 1.1.5 M = S m , N = ∗. Then M ∖ N ≅ Rm and hence contractible. We obtain that

πi (S m ) = 0 for i ≤ m − 1.
Example 1.1.6 Let Mon(Rn , Rm ) ⊂ Matm,n (R) be the subspace of injective linear maps. It is an
open subset; and the complement Matm,n ∖ Mon(Rn , Rm ) is the union ⋃n−1
r=0 Mr , where Mr is the
subset of all matrices of rank r. Mr is a submanifold of dimension (m − r)r + rn and therefore
πi (Mon(Rn , Rm ) = 0 for i < m − n.
Example 1.1.7 Let M be an m-dimensional manifold. The ordered configuration space of M

is the manifold F k (M ) ∶= {(xi , . . . , xk ) ∈ M k ∣i ≠ j ⇒ xi ≠ xj }. Let ∆i,j ∶= {(x1 , . . . , xk )∣xi = xj },
a submanifold of M k of codimension m. Then F k (M ) = M k ∖ ⋃i≠ ∆i,j . Therefore the inclusion
F k (M ) → M k is highly connected. In particular, for M = Rm , we obtain πi (F k (M )) = 0 as long as
i ≤ m − 2 (this does not depend on k).
COBORDISM THEORY 5
1.2 Orientations
For many purposes in differential topology, orientations are quite important. We begin by fixing
definitions and conventions. Let V be a finite-dimensional real vector space. An orientation of V
is an equivalence of bases of V , where two bases are equivalent if their transformation matrix has
positive determinant. Another way (and in some sense better) to define orientations is by means
of the exterior algebra.
Definition 1.2.1 Let V be an n-dimensional real vector space. Let Λn V be the top exterior power;
a one-dimensional vector space. An orientation of V is one of the two components of Λn V ∖ {0}.
The standard orientation of Rn is the component containing the element e1 ∧ . . . en . If V and W
are two oriented vector spaces, we orient their sum V ⊕ W so that, if (v1 , . . . , vn ) is an oriented
basis of V and (w1 , . . . , wm ) is an oriented basis of W , then (v1 , . . . , vn , w1 , . . . , wm ) is an oriented
basis of V ⊕ W .
The problem with this definition is that the vector spaces V ⊕ W and W ⊕ V are canonically
isomorphic, but that the isomorphism is not orientation preserving (its determinant is (−1)mn ). It
can be shown that it is impossible to remove this problem: direct sum of oriented vector spaces is
not a symmetric operation, or in more fancy terms: oriented vector spaces do not form a symmetric
monoidal category.
Definition 1.2.2 Let V → X be a real vector bundle. An orientation of V is a family x , x ∈ X,

where x is an orientation of the fibre Vx , such that for each y ∈ X, there exists a neighborhood U of
y and a bundle map h ∶ V ∣U → U ×Rn , such that for each x ∈ U , the linear isomorphism hx ∶ Vx → Rn
is orientation- preserving, where Rn is endowed with the standard orientation.
Again, a more elegant definition can be given using the exterior algebra. Let Λn V → X be the
bundle of top exterior powers. It is a real line bundle. We form Or(V ) ∶= (Λn V ∖0)/R>0 → X; which
is a twofold covering. We define an orientation to be a cross-section of Or(V ) → X. The above
orientation conventions prompt an orientation convention for submanifolds. If M is an oriented
manifold and N ⊂ M a submanifold, we say that orientations of N and the normal bundle νN M
are
compatible if the (almost) canonical isomorphism T M ∣N ≅ T N ⊕ νN is orientation-preserving.
M
1.3 The cobordism relation

Definition 1.3.1 (Co)bordism Let Mi , i = 0, 1 be closed manifolds. A bordism (or cobordism)
from M0 to M1 is a triple (W, φ0 , φ1 ), where W is a compact manifold with boundary ∂W =
∂0 W ∐ ∂1 W , which is decomposed into open a disjoint union of closed and open subsets. The maps
φi ∶ ∂i W → Mi are diffeomorphisms.
Let M be a manifold with boundary. If x ∈ ∂M , the tangent space Tx ∂M ⊂ Tx M is a codimension

1 subspace. A vector v ∈ Tx M ∖ Tx ∂M is an inward vector if there is a smooth curve c ∶ [0, 1) → M ,
c(t) ∈ ∂M ⇒ t = 0, c(0) = x and dt
d
c(0) = v. A vector v is an outward vector if −v is an inward one.
The subset of all inward vectors in an open half-space in the tangent space.
Definition 1.3.2 Normal Vector Field Let (W, φ0 , φ1 ) be a bordism from M0 to M1 . A normal
vector field is a section v ∶ ∂M → T M ∣∂M such that for all x ∈ ∂0 W , the vector v(x) is an inward
vector and for all x ∈ ∂1 W , v(x) is outward.
One can show that there is always a normal vector field and the space of normal vector fields
is convex (a partition of unity argument). Any normal vector field gives rise to a vector bundle
isomorphism
R ⊕ T ∂W ≅ T W ∣∂W ; (u, t) ↦ u + tv. (2)
Using this isomorphism, we define the notion of an oriented bordism.
COBORDISM THEORY 6
Definition 1.3.3 Oriented Bordism Let Mi , i = 0, 1 be oriented closed manifolds. An oriented

bordism from M0 to M1 is a bordism (W, φ0 , φ1 ), where W is in addition oriented, the above vector
bundle isomorphism and φi are orientation-preserving.
Definition 1.3.4 Bordism Group The unoriented bordism group of n-dimensional manifolds is
the set
Nn ∶= {M n ∣ M closed smooth n − manifold}/ ∼
with
M0 ∼ M 1 ∶⇔ there is a bordism from M0 to M1 .
The oriented bordism group is the set
Ωn ∶= {M n ∣ M closed oriented smooth n − manifold}/ ∼
where the equivalence relation is now taken to be only by oriented bordisms.
Theorem 1.3.5 Bordism and oriented bordism are equivalence relations. The assigment [M0 ] +
[M1 ] ↦ [M0 ∐ M1 ] turns both Nn and Ωn into abelian groups. Moreover, defining [M0 ][M1 ] ∶=
[M0 × M1 ], turns N∗ ∶= ⊕n≥0 Nn into a commutative ring and Ω∗ ∶= ⊕n≥0 Ωn into a graded-
commutative ring.
Proof: We only prove that being bordant is an equivalence relation for Ωn (Nn is easier).
• Reflexivity: [0, 1] × M is a bordism from M to M .
• Symmetry: If W ∶ M0 → M1 is a bordism, then so is −W ∶ M1 → M0 .
• Transitivity: If W01 ∶ M0 → M1 and W12 ∶ M1 → M2 are bordisms, then by gluing along M1

we get a bordism W01 ∪M1 W12 ∶ M0 → M2 .
• Neutral element: ∅ ∐ M = M ∐ ∅ = M
• Inverse: We have [M ] + [−M ] = ∅, which can be seen, by considering the cylinder over M
and bending one end over the other.
Due to the classification of 1-dimensional manifolds, the zero dimensional bordism groups are
easy to calculate. The result is that Ω0 ≅ Z and N0 ≅ Z/2. This calculation already gives rise to
interesting results. Consider the following situation. Let M m be an oriented manifold, N n ⊂ M m
be a submanifold. Let furthermore K k be a third oriented and closed manifold and f ∶ K → M be
a smooth map. Assume that f ⋔ N . Then f −1 (N ) ⊂ K is a submanifold of K; it has dimension
k − m + n and the normal bundle is νfK−1 (N ) ≅ f ∗ νN
M M
. Since M and N are oriented, νN has an
−1
orientation by the above convention and we orient f (N ) by this orientation convention. Taking
the bordism class of f −1 (N ), we obtain an element
−1
♯(f ; N ) ∶= [f (N )] ∈ Ωk−m+n .
That this definition makes sense is the content of the next result. The proof gives a first insight of
the connection between homotopies and bordism and is an important motivation for the
Pontrjagin-Thom construction.
Proposition 1.3.6 If fi ∶ K → M , i = 0, 1, are both transverse to N and if f0 ∼ f1 , then

[f0−1 (N )] = [f1−1 (N )] ∈ Ωk−m+n .
COBORDISM THEORY 7
The simplest special case of the invariant ♯(f ; N ) is when k − m + n = 0; in this case we get an
element in Ω0 = Z. We discuss two special cases in more detail. The first is the mapping degree.
Let f ∶ M n → N n be a smooth map between closed oriented manifolds of the same dimension. Let
z ∈ N be a regular value. Let 1 " f ∶ M → M × N be the map x ↦ (x, f (x)) (the graph embedding).
Definition 1.3.7 Mapping Degree The mapping degree at a regular value z ∈ N of a smooth
map f ∶ M → N is the element degz (f ) ∶= ♯(1 " f ; M × z) ∈ Ω0 = Z.
Proposition 1.3.8 The mapping degree has the following properties:
1. If N is connected, then degz (f ) does not depend on the choice of the regular value z.
2. If f0 ∼ f1 , then deg(f0 ) = deg(f1 ).
3. deg f g = deg(f ) deg(g).
The first two properties are not entirely trivial consequences of the homotopy invariance. For the
first, we consider the submanifold graph(f ) = {(x, f (x)} ⊂ M × N and the map ιz ∶ M → M × N ,
2
x ↦ (x, z) and note that ♯(1 " f ; M × z) = (−1)n ♯(ιz , graph(f )). Now we can give a purely
differential-topological proof of one of the pivotal results of homotopy theory. Let M n be a closed
oriented connetced n-manifold. Let [M ; S n ] be the set of homotopy classes of continuous maps.
Given f ∶ M → S n , we know that f is homotopic to a smooth map. The homotopy invariance of
the mapping degree shows that [f ] ↦ deg(f ) is a well-defined map [M ; S n ] → Z.
Theorem 1.3.9 Hopf The map deg ∶ [M ; S m ] → Z is a bijection for each closed oriented
connected n-manifold M .
Corollary 1.3.10 There is an isomorphism of groups πn (S n ) ≅ Z.
Another important special case of the intersection index is the Euler number of a vector bundle.
Let V → M be a rank k oriented vector bundle over a compact oriented manifold of dimension n.
Then the Euler number Eul(V ) is defined to be the element ♯(s; M ) ∈ Ωn−k , where M ⊂ V is the
zero section and s ∶ M → V is any section transverse to the zero section. Surely Eul(V ) does not
depend on the choice of s and if V has a section without zero, then Eul(V ) = 0. Computation of
some examples (spheres, complex-projective spaces, oriented surfaces) give credibility to the
following theorem.
Theorem 1.3.11 If M is a closed oriented manifold, then
n
Eul(T M ) = χ(M ) = ∑ (−1)k dim Hk (M ; Q).
k=0
The proof can be given either using Morse theory [6] or Poincaré duality (see Theorem 2.4.15
below).
2 Orientation and (Co)homology

2.1 Orientation and Duality
Definition 2.1.1 Orientation Conventions The singular 1-simplex
∆1 Ð→ R,
(1 − t)e0 + te1 z→ 2t − 1
gives a cycle in H1 (R, R ∖ 0; Z) whose equivalence class is a generator, denoted µ1 . We set
µ1 ∈ H1 (R, R − 0; Z) ≅ Z, µn ∶= µ1 × ⋯ × µ1 ∈ Hn (Rn , Rn − 0; Z) ≅ Z
and define µn ∈ H n (Rn , Rn − 0; Z) by the requirement ⟨µn ; µn ⟩ = 1. Then we have
µm × µn = µm+n , µn × µm = (−1)mn µm+n , ⟨µn , µn ⟩ = 1.
COBORDISM THEORY 8
Definition and Lemma 2.1.2 Fundamental Class Let M n be a topological manifold and
x ∈ K ⊂ M n with K compact. Then there is exactly one class [M ]K ∈ Hn (M, M − K), such that
ϕ
for all x ∈ K and every oriented chart x ∈ U → Rn with ϕ(x) = 0, we have that
/ Hn (M, M − x) ≅ / Hn (U, U − x) ≅ / Hn (Rn , Rn − 0).
Hn (M, M − K)
[M ]K / µn
Theorem 2.1.3 Poincare − Lefschetz − Alexander Duality Let M n be an oriented topological

manifold with L ⊆ K ⊆ M and K, L compact and nice, then the following is an isomorphism
≅
(−) ∩ [M ]K ∶ H p (K, L) Ð→ Hn−p (M − L, M − K)
Remark 2.1.4 Neighborhood retracts (e.g. submanifolds) are ”nice”. If K, L are not nice, then
the above isomorphism still holds, if one replaces H p (K, L) by the Cech-cohomology
Ȟ ∶= colimU,V H p (U, V ),
where L ⊂ U ⊂ V ⊃ K and U, V are open in M .
Example 2.1.5
1. Let M m , N n be oriented topological manifolds and K ⊆ M, L ⊆ N compact, then
[M × N ]K×L = [M ]K × [N ]L ∈ Hm+n ((M × N ), (M × N ) − (K × L))
due to the uniqueness of the fundamental class.
2. Let M0 , M1 be oriented, compact n-manifolds and (W, id0 , id1 ) an oriented bordism. We glue
necks onto W :
̂ ∶= W ∪M ((−∞, 0] × M0 ) ∪M ([0, ∞) × M1 ),
W 0 1
≅
̂, W
consider the isomorphism ϕ ∶ Hn+1 (W ̂ − W ) → Hn+1 (W, ∂W ) and define
̂
[W ] ∶= φ∗ [W ]W ∈ H (W, ∂W ).
n+1
Theorem 2.1.6 Let W be a bordism with boundary ∂W = ∂0 W ⊔ ∂1 W . Consider the homology

sequence of the pair (W, ∂W )
j
⋯ / Hn (W, ∂W ) ∂ / Hn (∂W ) / Hn (W ) /⋯
It holds that
∂[W ] = −[∂0 W ] + [∂1 W ].
Proof: See [7], Lemma VI.9.1.

◻
Corollary 2.1.7 Let W be a bordism with boundary ∂W = ∂0 W ⊔ ∂1 W and x ∈ H n (W ).

Consider the inclusions ji ∶ ∂i W ↪ W , then
⟨j0∗ x, [∂0 W ]⟩ = ⟨j1∗ x, [∂1 W ]⟩ .
Proof: Let j = j0 ⊔ j1 , then

⟨j0∗ x, [∂0 W ]⟩ − ⟨j1∗ x, [∂1 W ]⟩ = ⟨j ∗ x, ∂[W ]⟩ = ⟨x, (j∗ ∂)[W ]⟩ = 0,
where in the first step we have used the previous theorem and in the last step we used that
(j∗ ∂) = 0, due to the exactness of the homology sequence.
◻
COBORDISM THEORY 9
2.2 Homological Interpretation of the Mapping Degree

Theorem 2.2.1 Let M n , N n be closed connected oriented topological manifolds and f ∶ M → N a
continuous mapping, then due to Hn (M ) ≅ Z ≅ Hn (N ), we have that
∃! d ∈ Z ∶ f∗ [M ] = d[N ], and deg(f ) = d.
Proof: W.l.o.g. let f be smooth, x ∈ N be a regular value. Let further x ∈ U ⊆ N with U open
and all u ∈ N be regular values. Then
f ∣f −1 (U ) ∶ f −1 (U ) Ð→ U
is a covering, i.e. f −1 (U ) = ⊔ri=1 Vi and f ∣Vi → U is a diffeomorphism. Let further ϕ ∶ U → Rn be a

map with ϕ(x) = 0 and let f −1 (x) = {y1 , . . . , yr }, then consider the following commutative diagram
f∗
Hn (M ) / Hn (N )
≅
f∗
Hn (M, M − {y1 , . . . , yr }) / Hn (N, N − x)
ϕ∗

Hn (⊔ri=1 Vi , ⊔ri=1 (Vi − {yi })) Hn (Rn , Rn − 0)
4
≅
(a1 ,...,ar )↦∑i ai
⊕ri=1 Hn (Rn , Rn − 0)
where in the left column Hn (M ) ∋ [M ] z→ (ε1 µn , . . . , εr µn ) ∈ ⊕ri=1 Hn (Rn , Rn − 0) with

εi ∶= sgn(det dyi f ) and in the right column Hn (N ) ∋ [N ] z→ µn ∈ Hn (Rn , Rn − 0).
Due to ⊕ri=1 Hn (Rn , Rn − 0) ∋ (ε1 µn , . . . , εr µn ) z→ ∑ri=1 εi µn = deg(f )µn ∈ Hn (Rn , Rn − 0), and the
commutativity of the diagram, we indeed get f∗ [M ] = deg(f )[N ].
2.3 The Leray - Hirsch theorem

Remark 2.3.1 Thom − Isomorphism Special Case Let M m be an oriented closed manifold
and V → M an oriented n-vector bundle. Consider the disc bundle W ∶= {v ∈ V ∣ ∣v∣ ≤ 1} which is a
compact manifold and gives rise to the following commutative diagram
H k (M ) / H n+k (W, ∂W )
≅ ≅

≅ / Hm−k (W )
Hm−k (M )
which lets the above map be an isomorphism called Thom-isomorphism. The image of 1 ∈ H 0 (M )
under this map is referred to as the Thom class.
For the general situation, we consider a principal ideal domain R, a topological space X and a
fiber bundle π ∶ E → X with fiber F x ∶= π −1 (x) and a subbundle E0 ⊂ E. We make the following
definitions and remarks:
• H ∗ (E, E0 ) ∶= H ∗ (E, E0 ; R) is a H ∗ (X)-module via x ⋅ y ∶= π ∗ (x) ∪ y, where

x ∈ H ∗ (X), y ∈ H ∗ (E, E0 ).
• Let F0x ∶= F x ∩ E0 and jx ∶ (F x , F0x ) ↪ (E, E0 )

COBORDISM THEORY 10
The Leray-Hirsch theorem makes a statement on the structure of H ∗ (E; E0 ) as a module over
H ∗ (X) under the following assumption.
Assumption 2.3.2 There is a set B ⊆ H ∗ (E, E0 ) of homogeneous elements, such that:

• For every n ∈ N there are only finitely many b ∈ B of degree n.
• For all x ∈ X the set {jx∗ b ∣ b ∈ B} is a R-basis of H ∗ (F x , F0x ).
Theorem 2.3.3 Leray − Hirsch Theorem Let everything be as above and assume 2.3.2, then
≅
Φ ∶ H ∗ (X) ⊗R RB Ð→ H ∗ (E, E0 ; R)
x ⊗ b z→ π ∗ x ⋅ b
is an isomorphism of H ∗ (X)-modules (but not a ring isomorphism!).
Proof:
1.) Formalities
a) Let f ∶ Y → X be a continuous function. Consider

π∣f
(f ∗ E, F ∗ E0 ) /Y
f˜ f

(E, E0 )
π /X
and we set f ∗ B ∶= {f˜∗ b ∣ b ∈ B} ⊆ H ∗ (f ∗ E, f ∗ E0 ; R), then (f ∗ E, F ∗ E0 ) → Y and f ∗ B

fulfill the requirements of the theorem.
b) Φ is a natural transformation of functors
TopX Ð→ R − Mod
where TopX has as objects (f, Y ) where f ∶ Y → X is a continuous function and the
morphisms are commutative triangles
f1
Y1 /X
>
f2
Y2
Φ is given by
Φ ∶ H ∗ (Y ) ⊗R RB Ð→ H ∗ (f ∗ E, f ∗ E0 ; R)
y ⊗ b z→ π∣∗ y ⋅ f˜∗ b f
2.) Let now X be a zero dimensional CW complex, then
H ∗ (X; R) ⊗R RB ≅ ( ∏ R) ⊗R RB, H ∗ (E, E0 ; R) ≅ ∏ H ∗ (F x , F0x ; R),

x∈X x∈X
where the first isomophism is given by r ⊗ b ↦ (rx )x∈X ⊗ b and the second by b ↦ (bx )x∈X .
Under these identifications, Φ becomes
∗
Φ ∶ ( ∏ R) ⊗R RB Ð→ ∏ H (F , F0 ; R),
x x
x∈X x∈X
(rx )x∈X ⊗ b z→ (rx bx )x∈X
which is an isomorphism.
COBORDISM THEORY 11
3.) We check the Mayer-Vietoris property. Take A, B ⊆ X open, and assume the theorem holds
for A, B, A ∩ B, then we shall prove that the theorem also holds for A ∪ B. We introduce the
notation
Ln (A) ∶= (H ∗ (A) ⊗R RB)n (the piece of degree n); K n (A) ∶= H n (E∣A , E0 ∣A ; R);
Φ becomes a natural transformation of functors Ln → K n from spaces over X to R-modules.

Now consider the following diagram of Mayer-Vietoris sequences
Ln−1 (A) ⊕ Ln−1 (B) / Ln−1 (A ∩ B) ∂ / Ln (A ∪ B) / Ln (A) ⊕ Ln (B) / Ln (A ∩ B)
≅ Φ ≅ Φ ≅ Φ ≅ Φ ≅ Φ

K n−1 (A) ⊕ K n−1 (B) / K n−1 (A ∩ B) ∂ / K n (A ∪ B) / K n (A) ⊕ K n (B) / K n (A ∩ B)
All the above squares commute, only commutativity of the square containing the boundary
maps ∂ needs to be checked. Commutativity here amounts to
π ∗ (∂x) ∪ b = ∂(π ∗ x ∪ b).
This holds, since considering open subsets U1 , U2 ⊆ Y such that U1 ∪ U2 = Y , we have a

commuative diagram of cochain complexes
0 / C ∗ (Y ) / C ∗ (U1 ) ⊕ C ∗ (U2 ) / C ∗ (U1 ∩ U2 ) /0

U
×z ×z ×z

0 / C ∗ (Y × Z) / C ∗ (U1 × Z) ⊕ C ∗ (U2 × Z) / C ∗ (U1 ∩ U2 × Z) /0
U ×Z
with z ∈ C ∗ (Z) and ∂z = 0, where the subscript denotes the complex of small simplices with
respect to the covers U = {U1 , U2 } and U × Z = {U1 × Z, U2 × Z}. These complexes are chain
equivalent to the full singular cochain complex by the small simplex theorem. Thus the
boundary map ∂ from the Mayer-Vietoris sequence commutes with the cohomology cross
product ×. Thus we indeed have the above commutativity.
≅
4.) Let f ∶ Y → X be a weak homotopy equivalence (i.e. fn ∶ πn (Y ) Ð→ πn (X) ∀n). We assume
that the theorem holds for Y and show that this also lets the theorem hold for X.
The map f induces a map between the long exact homotopy sequences:
π∗ (F x ) / π∗ (f ∗ E) / π∗ (Y ) / π∗−1 (F x ) /⋯
≅ ≅ ≅ ≅

π∗ (F x ) / π∗ (E) / π∗ (X) / π∗−1 (F x ) /⋯
Thus by the 5-Lemma, f˜ ∶ f ∗ E → E is a weak homotopy equivalence and the same argument
shows that f˜ ∶ f ∗ E0 → E0 is weak homotopy equivalence. By the Hurewicz theorem and the
5-Lemma again,
f˜∗ ∶ H∗ (f ∗ E; f ∗ E0 ) Ð→ H∗ (E; E0 )
is an isomorphism and the universal coefficient theorem proves that
≅
H ∗ (E, E0 ; R) Ð→ H ∗ (f ∗ E, f ∗ E0 ; R)
≅
H ∗ (X) Ð→ H ∗ (Y )
is an isomorphism.
COBORDISM THEORY 12
5.) The theorem holds for all finite dimensional CW complexes X. This is proved by induction
on the dimension n of X. We have already seen the case n = 0. So now let dim(X) = n > 0
and let D ⊂ X contain exactly one point out of every open n-cell. The theorem holds for D.
Let
A ∶= ⋃ {open n−cells} ≃ D, B ∶= X ∖ D ≃ X n−1 ;
this gives a cover A ∪ B = X and A ∩ B = ∪ ({open n−cells} ∖ {∗}) ≃ ∪S n−1 . By induction
hypothesis and step (4), the theorem holds for A, B and A ∩ B and so by step (3), it follows
for X.
6.) We can now prove the theorem for an arbitrary CW complex X. We set E (n) ∶= π −1 (X (n) )
and have a commutative diagram
H ∗ (X) ⊗R RB
Φ / H ∗ (E, E0 ; R)
≅ (i)

(lim H ∗ (X (n) )) ⊗R RB ≅ (ii)
←Ð
= (iii)
lim Φ(n)
lim(H (Xn (n)
) ⊗R RB) ←
Ð / lim H ∗ (E (n) , E (n) )
←Ð ←Ð 0
where (i) is an isomorphism by the CW-homology theorem. The map (ii) is an isomorphism
since E (n) → E is a π∗ -isomorphism for ∗ < n and thus a homology isomorphism in small
degrees. (iii) is an isomorphism because the inverse system is eventually constant.
7.) For a general space X we now chose a CW-approximation f ∶ Y → X with Y a CW-complex

and f a weak homotopy equivalence
2.4 Thom-Isomorphism
Definition 2.4.1 Thom − Class Let π ∶ V → X be a real vector bundle of rank n and set
V0 ∶= V ∖ 0 where 0 is understood as the image of the zero section. A Thom-class is an element
τ ∈ H n (V, V0 ; R) ∶ jx∗ τ ∈ H n (V x , V0x ; R) ≅ R is a generator ∀ x ∈ X.
Corollary 2.4.2 ofTheorem2.3.3 Let τ be a Thom-class, then

≅
th ∶ H ∗ (X; R) Ð→ H ∗+n (V, V0 ; R)
x z→ π ∗ x ∪ τ
is an isomorphism.
Remark 2.4.3
• If τ is a Thom-class for V → Y and f ∶ X → Y , then
f˜∗ τ ∈ H n (f ∗ V, f ∗ V0 )
is a Thom-class also.
• Consider two vector bundles V n → X and W m → Y with Thom-classes τV , τW , then
τV × τW ∈ H n+m (V × W, (V × W )0 )
is a Thom-class also.
COBORDISM THEORY 13
Definition 2.4.4 Thom − Space Let V → X be a vector bundle with metric, then consider the
associated disc and sphere bundles
DV ∶= {v ∈ V ∣ ∣v∣ ≤ 1}, SV ∶= {v ∈ V ∣ ∣v∣ = 1}.
The Thom-space of V is defined as
Th(V ) ∶= DV /SV ≅ P(V ⊕ R)/PV,
where PV is the projective bundle of V .
Remark 2.4.5
• If X is compact, then the Thom-space coincides with the one-point-compactification:
Th(V ) = V + .
• We have
̃ ∗ (Th(V )) ≅ H ∗ (V, V0 ),
H Th(V × W ) ≅ Th(V ) ∧ Th(W ).
Definition 2.4.6 Compatible Thom − Class Let V → X be an oriented vector bundle. A

Thom-class τ is called compatible with the orientation, iff for all x ∈ X and every orientation
preserving isomorphism ϕ ∶ Rn → V x , we have that
jx∗ ϕ∗
H n (V, V0 ; R) Ð→ H n (V x , V0x ; R) Ð→ H n (Rn , Rn ∖ 0; R), τ ↦ µn .
Remark 2.4.7 If V n → X and W m → Y are oriented vector bundles with compatible

Thom-classes τV , τW , then (−1)nm τV × τW is a compatible Thom-class for V × W.
Theorem 2.4.8 Let V → X be a vector bundle of rank n, then set HV ∶= ∐x∈X H n (V x , V0x ; R)
(which has a unique topology that turns it into a bundle (of discrete abelian groups) over X).
1.) The following is an isomorphism of abelian groups:
≅
ψ ∶ H n (V, V0 ; R) Ð→ Γ(X; HV ).
α z→ (x ↦ jx∗ α)
2.) We have H k (V, V0 ; R) = 0 for all k < n.
Proof:
The proof follows a similar pattern as the proof of the Leray-Hirsch theorem.
1.) ψ is natural w.r.t. bundle maps. We have Hf ∗ X = f ∗ HV and the following commutes for
f ∶Y →X
ψ
H n (V, V0 ; R) / Γ(X; HV )
ψ
H n (f ∗ V, f ∗ V0 ; R) / Γ(Y ; Hf ∗ V )
2.) Let V = X × Rn , then HV ≅ X × H n (Rn , Rn − 0) and the following commutes

= / H 0 (X) ⊗ H n (Rn , Rn − 0)
H n (X × (Rn , Rn − 0))
ψ =

= / H 0 (X) ⊗ H n (Rn , Rn − 0)
Γ(HV ) = H 0 (X, H n (Rn , Rn − 0))
where we have used the Künneth theorem and Γ(X; HV ) = C 0 (X, H n (Rn , Rn − 0)). So the
theorem holds for trivial bundles.
COBORDISM THEORY 14
3.) We now run a Mayer-Vietoris argument. Let X = A ∪ B with A, B open. We assume that
the theorem holds for A, B, A ∩ B. We get the commutative diagram
⋯0 / H n (V, V0 ) / H n (V ∣A , V0 ∣A ) ⊕ H n (V ∣B , V0 ∣B ) / H n (V ∣A∩B , V0 ∣A∩B )
= (i) ≅ ≅

⋯0 / Γ(HV ) / Γ(A, HV ) ⊕ Γ(B, HV ) / Γ(A ∩ B, HV ).
The upper row is the Mayer-Vietoris sequence, and we have used that
H n−1 (V ∣A∩B , V0 ∣A∩B ) = 0 by assumption. The exactness of the bottom sequence is clear.
With the 5-lemma we get that (i) is also an isomorphism. As in the proof of Theorem 2.3.3,
we conclude that the statement is true for bundles over CW complexes.
Theorem 2.4.9 Let V → X be a vector bundle of rank n, then
1.) if V is oriented, then there is exactly one compatible Thom-class τV ∈ H ∗ (V, V0 ; Z),
2.) if V is arbitrary, then there is exactly one compatible Thom-class τV ∈ H ∗ (V, V0 ; Z2 ).
Proof:
2.) For R = Z/2 we have H n (V x , V0x , Z/2) = Z/2, then HX (V, V0 , Z/2) → X is a Z/2
group-bundle (not a principal bundle!) and thus trivial. Considering the isomorphism in
(2.4.8):
≅
ψ ∶ H n (V, V0 ; R) Ð→ Γ(HV ; R), α z→ (x ↦ jx∗ α)
it follows that α is a Thom class, iff ψ(α)(x) is a generator for all x ∈ X, but in a Z/2 there
is only one such generator.
1.) In the bundle H n (V, V0 ; R) we have the subbundle UV of all generators, which is a Z/2
principal bundle and thus not necessary trivial. Considering the orientation bundle Or(V )
we have the isomorphism
≅
Or(V ) Ð→ UV ,
oX z→ ϕ∗x un
where ϕx ∶ V x → Rn is orientation preserving.
Definition 2.4.10 Euler − Class Let π ∶ V → X be a vector bundle of rank n and

s ∶ (X, ∅) → (V, V − 0) be the zero section. We define
1.) e(V ) ∶= s∗ τV ∈ H n (X, Z/2) the mod 2 - Euler-class.
2.) If V is oriented, we define the Euler-class e(V ) ∶= s∗ τV ∈ H n (X, Z).
Theorem 2.4.11 Properties of the Euler − Class Let V, V1 , V2 be vector bundles over X of
rank n, n1 and n2 respectively. Let f ∶ Y → X be a continuous function, then the following hold
1.) e(f ∗ V ) = f ∗ (e(V )),
2.) e(V1 ⊕ V2 ) = (−1)n1 n2 e(V1 ) ∪ e(V2 ),
3.) if there is a nowhere vanishing section z ∶ X → V , then e(V ) = 0,

COBORDISM THEORY 15
4.) if n = rank (V ) is odd, then 2e(V ) = 0 and thus e(V1 ⊕ V2 ) = e(V1 ) ∪ e(V2 ) is always true.
Proof:
1.) Follows from naturality of the Thom-class τV .
2.) Follows from (2.4.7).
3.) Since two sections are alway homotopic, we have that z is homotopic to the zero section s, so
e(V ) = s∗ τV = z ∗ τV .
Since z factors over (X, X):

(X, ∅)
z / (V, V0 )
9

(X, X)
and H n (X, X) = 0, we have e(V ) = 0.
4.) We denote V with opposite orientation by V − and consider the orientation reversing map
f (v) = −v. With the commuting diagram
f
V / V,
~
X
we have that V ≅ V − as oriented bundles and thus e(V ) = e(V − ) = −e(V ).
Lemma 2.4.12 Thom − Classes of Submanifolds Let M m be an oriented manifold and

N n ⊆ M m an oriented submanifold. Let K ⊆ N be compact and j ∶ N ↪ M the inclusion. We
consider a tubular neighborhood U ⊇ N with the projection π ∶ U → N and set V ∶= π −1 (N ∖ K).
M
With the normal bundle νN , due to excision, it holds that
≅
H m−n (U, U ∖ N ) Ð→ H m−n (M, M ∖ N ),
τν M z→ τN
M
N
with [M ]K ∈ Hm (M, M ∖ K) and j∗ ∶ Hn (N, N ∖ K) → Hn (M, V ), we have
M
τN ∩ [M ]K = (−1)n(m−n) j∗ [N ]K ∈ Hn (M, V )
If both M and N are compact and N = K, then

M
τN ∩ [M ]K = (−1)m(m−n) j∗ [N ].
M
We can express this loosely by saying that the fundamental class of N and the Thom-class of νN
are Poincaré dual.
Proof: We can restrict to the following case
• M = U due to excision,
• K = {∗} because of the definition of the fundamental class,

COBORDISM THEORY 16
• N = Rn × {0} ⊆ Rm and K = {0}.

This gives us
τN = 1 × um−n ∈ H m−n (Rn × (Rm−n , Rm−n
0 ))
and we have [M ] = u1 × ⋯ × u1 ∈ Hm (Rm , Rm
0 ), j∗ [N ] = u1 × ⋯ × u1 × 1 ∈ Hn ((R , R0 ) × R
n n m−n
),
which gives us
M
τN ∩ [M ] = (1 × um−n ) ∩ (u1 × ⋯ × u1 )
= (−1)n(m−n) (1 ∩ (u1 × ⋯ × u1 )) × (um−n ∩ (u1 × ⋯ × u1 ))
´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹¸ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶ ´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹¸ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶
n m−n
= (−1) n(m−n)
(u1 × ⋯ × u1 ) × 1 = (−1) n(m−n)
j∗ [N ]
´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹¸ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶
n
Corollary 2.4.13 Let M m be oriented and compact, π ∶ V → M a vector bundle of rank k and
s ∶ M → V a section that is transversal to the zero section: s ⋔ 0. Then K m−k ∶= s−1 (0) ⊆ M is a
submanifold with νZM = V ∣Z . It follows that
j∗ [Z] = (−1)(m−k)k τZM ∩ [M ] = (−1)(m−k)k e(Z) ∩ [M ].
Corollary 2.4.14 Poincare − Hopf With the same conditions as in the previous corollary, but
k = m and thus [Z] ∈ Ω0 ≅ Z, it holds that
⟨e(V ), [M ]⟩ = e(V ) ∩ [M ] = j∗ [Z] = Eul(V )
Theorem 2.4.15 Topological Gauss − Bonnet Let M n be a closed oriented manifold, then
χ(M ) = ⟨e(T M ), [M ]⟩ = Eul(T M )
Proof:
• If n is odd, then 2e(T M ) = 0 and thus ⟨e(T M ), [M ]⟩ = 0, but also χ(M ) = 0, due to
Poincaré duality and the universal coefficient theorem, i.e.
dim Hk (M ; Q) = dim H n−k (M ; Q) = dim Hn−k (M ; Q).
• We thus consider n even and chose B a basis of H∗ (M ; Q), where all α ∈ B are homogeneous
and {α# ∣ α ∈ B} is a basis for H ∗ (M ; Q) with
⟨α# ∪ β, [M ]⟩ = δα,β ∀ α, β ∈ B.
Note that such elements α# ∈ H ∗ (M ; Q) exist, due to Poincaré duality. Now we consider
the diagonal
∆ ∶= {(x, x) ∈ M × M } ⊂ M × M, T ∆{(v, −v) ∈ T M × T M } ≅ T M,

M ×M M ×M
with the Thom-class τ = τ∆ ∈ H n (M × M ; Q) of ν∆ . By the Künneth theorem there
exists unique cγδ ∈ Q such that
τ = ∑ cγδ (γ # × δ), γ, δ ∈ B.
γ,δ
We now see that
⟨(α# × β) ∪ τ, [M × M ]⟩ = ⟨α# × β, τ ∩ [M × M ]⟩ = (−1)n ⟨α# × β, j∗ [M ]⟩

= ⟨j ∗ (α# × β), [M ]⟩ = ⟨α# ∪ β, [M ]⟩ = δαβ ,
COBORDISM THEORY 17
and analogously get
⟨(α# × β) ∪ τ, [M × M ]⟩ = ∑ cγδ ⟨(α# × β) ∪ (γ # × δ), [M × M ]⟩ = cβα (−1)∣β∣ ,

γ,δ
which gives cαβ = (−1)∣α∣ δαβ and thus τ = ∑α (−1)∣α∣ (α# × α), which in turn lets us calculate
⟨e(T M ), [M ]⟩ = ⟨j ∗ τ, [M ]⟩ = ∑(−1)∣α∣ ⟨j ∗ (α# × α); [M ]⟩

α
∣α∣
= ∑(−1) ⟨α ∪ α; [M ]⟩ = ∑(−1)∣α∣ = χ(M )
#
α α
2.5 Fiber Bundles

In the following, let G be a topological group, X, F be topological spaces and F be a left G-space.
Definition 2.5.1 Fiber Bundle with Structure Group A fiber bundle over X with structure
group G and fiber F is a triple
π q
(E Ð→ X, P Ð→ X, ϕ)
• π ∶ E Ð→ X is a fiber bundle,
• q ∶ P Ð→ X is a G-principal fiber bundle,

≅
• ϕ ∶ P ×G F Ð→ E is an isomorphism of fiber bundles.
Remark 2.5.2 Alternative Definition In the literature one usually finds the above definition
in terms of an open, trivializing cover {Ui }i∈I of X with trivializations
hi ∶ π −1 (Ui ) Ð→ Ui × F,
giving rise to trivialization changes
ϕij ∶= hi ○ h−1
j ∶ (Ui ∩ Uj ) × F Ð→ (Ui ∩ Uj ) × F
(x, f ) z→ (x, gij (x)f )
with the functions gij ∶ (Ui ∩ Uj ) → G fulfilling the cocycle condition gij gjk = gik . Definition 2.5.1
agrees with that provided G ⊂ Homeo(F ), i.e. G acts faithfully on F .
Example 2.5.3
• G = Gln (R), F = Rn gives a vector bundle.
• G = O(n), F = Rn gives a riemannian vector bundle.
• G and F discrete gives a covering.
• G = Homeo(F ) with KO-topology (here we need X to be compact, so that G is a topological

group), F compact and metrizable gives a fiber bundle with fiber F .
• G = Diff(X) with C ∞ -topology and X a compact manifold gives a smooth bundle with fibre
X.
Theorem 2.5.4 Ehresmann′ s Fibration Theorem Let E, X be manifolds, f ∶ E → X be a

smooth, proper submersion, then f ∶ E → X is a fiber bundle with structure group Diff(F ) and
fiber F .
COBORDISM THEORY 18
Remark 2.5.5 Classifying Space We define
PrinG (X) ∶= {π ∶ P → X ∣ π principal G bundle}/Iso
which gives the contravariant functor PrinG (−) ∶ Top Ð→ Set. Since f0 ≅ f1 gives f0∗ P ≅ f1∗ P , the
functor PrinG (−) in fact factorizes over HoTop ∶
PrinG (−) ∶ HoTop Ð→ Set.
Now the question is, if there is a G-bundle π ∶ E → B, such that
[−, B] Ð→ PrinG (−), f ↦ f ∗E
is a bijection. The answer to this question is affirmative, if we restrict ourselves to base spaces X
that are CW-complexes, and is given by the so called classifying space.
2.6 Homotopy Invariance of Fiber Bundles

Theorem 2.6.1 Let X be paracompact and P → X × [0, 1] a G-principal bundle. Consider
r ∶ X × [0, 1] → X × [0, 1], r(x, t) ∶= (x, 1), then there is a bundle map
P
R /P

X × [0, 1]
r / X × [0, 1]
with R∣P ∣X×1 = id.
Before we come to the proof of the theorem, let us state a few corollaries.
Corollary 2.6.2 Let X, P, G be as in the theorem and Pt ∶= P ∣X×{t} . Then there is a bundle
isomorphism
ϕ
P1 × [0, 1] /P

X × [0, 1]
id / X × [0, 1]
with ϕ∣P1 ×[0,1] = id.
Proof: P1 × [0, 1] = r∗ P
Corollary 2.6.3 With X, P, G as in the theorem, it holds that P0 ≅ P1 .

R
Proof: P0 ⊆ P Ð→ P is a bundle map over x ↦ (x, 1) and thus P0 ≅ P1 .
Corollary 2.6.4 Let f0 , f1 ∶ X → Y be homotopic and P → Y a G-principal bundle, then
f0∗ P ≅ f1∗ P.
Remark 2.6.5 The proof of Theorem 2.6.1 is not very intuitive, and so we give a few examples
to get a feeling of what is involved.
COBORDISM THEORY 19
1.) Let π ∶ E Ð→ X × [0, 1] a Galois cover, then

E0 × 0 /7 E
π

E0 × [0, 1] / X × [0, 1]
π0 ×id
where the diagonal map exists and is unique because of the homotopy lifting theorem.
2.) Vector bundles on compact spaces. Let V → X × [0, 1] be a vector bundle and X a compact
Hausdorff space. Let further pr ∶ X × [0, 1] → X be the projection onto the first component
and ιt ∶ X → X × [0, 1] the embedding ιt (x) ∶= (x, t) for which we define Vt ∶= ι∗t V, then it
holds that
V ∣X×{t} ≅ (pr∗ Vt )∣X×{t}
By the Tietze extension theorem, there is a neighborhood X × {t} ⊆ U ⊆ X × I such that
V ∣U ≅ (pr∗ Vt )∣U .
Since X is compact there is a ε > 0, such that X × (t − ε, t + ε) ⊆ U and thus
Vt ≅ Vu , for ∣u − t∣ < ε.
Since I is compact and connected, we get V0 ≅ V1 . (Lebesgue Lemma).
3.) Smooth fiber bundles with compact fibers. Let M be a manifold and f ∶ X → M × I a smooth
fiber bundle with compact fiber. ∂t is a vector field on M × I. There is a vector field V on X
with
(Tx f )(Vx ) = ∂t ∣f (x) .
In the following the fact that f is proper is essential. We take the flow ϕt of V and get the
maps
ϕt φ
X /X X0 × [0, 1] /X
f f

M ×R
m / M ×R M × [0, 1]
id / M × [0, 1]
with m(x, s) = (x, s + t) and φ(x, t) ∶= ϕt (x).

4.) Smooth G-principal bundle for a Lie group G. A connection ω on P is a G-equivariant
splitting ω ∶ p∗ T M → T P of
dp
0 Ð→ P × g Ð→ T P Ð→ p∗ T M Ð→ 0
such that ω(p∗ T M ) ⊂ T P is a subbundle. A connection defines a unique lift of any vector
field on M to a G-equivariant vector field on P . As in the previous example, the flow of this
lifted vector field gives the desired isomorphism.
Proof: Of the theorem
• Let (x, t) ∈ X × [0, 1], then there is a cover Ux,t × Ix,t with x ∈ Ux,t and t ∈ Ix,t , such that
P ∣Ux,t ×Ix,t is trivial. Since [0, 1] is compact, we find an open U ⊆ X and a partition
0 = t0 ≤ ⋯ ≤ tr = 1, such that P ∣U ×[ti ,ti+1 ] is trivial. The number r can be reduced by the
following property. Let
g ∶ P ∣U ×[a,b] Ð→ U × [a, b] × G, h ∶ P ∣U ×[b,c] Ð→ U × [b, c] × G

be trivializations, then there is a function f ∶ U → G, such that for x ∈ P ∣U ×[a,b] , we have
g(h−1 (u, b, xf (b))) = (u, b, x).
Now we can change the trivialization h using f to obtain a trivialization of P over U × [a, c].
COBORDISM THEORY 20
• There is an open cover (Uj )j∈N of X such that PUj ×[0,1] is trivial. Let λj be a partition of
unity subordinate to (Uj )j∈N , and set
λ(x) ∶= max λj (x), µj (x) ∶= λj (x)/λ(x), supp(µj ) ⊆ Uj , max µj (x) = 1.

j∈N j∈N
We now define
rj ∶ X × [0, 1] Ð→ X × [0, 1].

(x, t) z→ (x, max(µj (x), t))
There is a bundle map Rj over rj :
Rj
P ∣Uj ×[0,1] / P∣
Uj ×[0,1]
≅ ≅
(rj ,idG )
Uj × [0, 1] × G / Uj × [0, 1] × G
and, outside of supp(λj ), we have that Rj is the identity. We can thus extend Rj to all of
P . Now, due to the local finiteness of the partition of unity, the following expressions are
well defined and give the maps stated in the theorem
R ∶= ⋯ ○ Rj ○ Rj−1 ○ ⋯ ○ R2 ○ R1 ,
r ∶= ⋯ ○ rj ○ rj−1 ○ ⋯ ○ r2 ○ r1 .
2.7 Classification of G-Bundles

In the following let G be a topological group and let P → X, E → B be G-principal bundles. Let
E be weakly contractible (i.e. πi (E) = 0 ∀i). In this section we want to show that
PrinG (X) ≅ [X; BG],
where we restrict ourself to the case where X is a CW-complex. The proof of the general case can
be found in [2], chapter 14.
Theorem 2.7.1 Let X be a CW-complex, A ⊆ X a subcomplex, P → X and E → B be

G-principal bundles with E ≅ ∗, then for each bundle map
f
P ∣A /E

A /B
there is an extension to a bundle map g ∶ P → E, such that g∣P ∣A = f.
Proof: The following set is partially ordered by inclusion:
{(Y, g) ∣ A ⊆ Y ⊆ X, Y complex g ∶ P ∣Y → E bundle map g∣P ∣A = f },
and with Zorn’s lemma there exists a maximal element. It now suffices to study the case where
we just add one cell, i.e. X is the pushout in
COBORDISM THEORY 21
ϕ
S n−1 /Y
ψ
Dn /X
Now given the bundle map f ∶ PY → E, we recall the homotopy theorem and get
π ∗ P ≅ Dn × G.
Under this isomorphism, the map f becomes
f˜
S n−1 × G Ð→ E,
which is G-equivariant and thus determined by f˜∣S n−1 ×1 → E. Now finally due to πn−1 (E) = 0, we
see that we can extend f˜ and thus also f .
◻
Corollary 2.7.2 Let X be a CW-complex, P → X and E → B be G-principal bundles where E is

weakly contractible. Then mapG (P ; E) is weakly contractible.
Proof: For a space Z to be weakly contractible, it is equivalent, that for all k ≥ 0 and every
map S k−1 → Z factors over Dk :
S k−1 /Z
=

Dk
For Z = mapG (P ; E), this amounts to
P × S k−1 /E
;

P × Dk
which is solved by the last theorem.
◻
Corollary 2.7.3 Let X be a CW-complex, P → X and E → B be G-principal bundles where E is

weakly contractible. Then
∃ f ∶ X Ð→ B ∶ f ∗ B ≅ P,
and f is unique up to homotopy.
Proof: The equivariance of the diagram

f
P /E
f /G
X /B
gives f ∗ E ≅ P . It only remains to show uniqueness. Let f0 , f1 be two such maps, then define
f = (f0 , f1 ) ∶ X × {0, 1} → E and we have the isomorphism
ϕ ∶ P × {0, 1} Ð→ f ∗ E Ð→ E.
So we have a bundle P × [0, 1] → X × [0, 1] and the G-map ϕ, thus the extension of f to X × [0, 1]
exists and we have f0 ≅ f1 .
COBORDISM THEORY 22
Definition 2.7.4 Universal G − Bundle A G-principal bundle E → B with B a CW-complex

and E ≃ ∗ is called universal G-bundle. For such a bundle we write
EG ∶= E, BG ∶= B.
Example 2.7.5
1.) Let G = Σn be the symmetric group, and
F n (Rm ) ∶= {(x1 , . . . , xn ) ∈ (Rm )n ∣ xi ≠ xj , for i ≠ j}
called the ordered configuration space, which has the action
Σn × F n (Rm ) Ð→ F n (Rm ), (σ, (x1 , . . . , xn )) ↦ (xσ(1) , . . . , xσ(n) ),
and thus gives rise to the unordered configuration space C n (Rm ) ∶= F n (Rm )/Σn . Together
they form the Σn -principal bundle
F n (Rm ) Ð→ C n (Rm ).
The configuration space has vanishing lower homotopy groups: πi (F n (Rm )) = 0 for i < m − 1
by Theorem 1.1.4, thus the direct limit
F n (R∞ ) ∶= lim F n (Rm )

Ð→
m
is weakly contractible: πi (F n (R∞ )) = 0 ∀i and we have
EΣn ≅ F n (R∞ ), BΣn ≅ C n (R∞ ) = F n (R∞ )/Σn .
Let now π ∶ Y → X be any n-fold covering (i.e. a fiber bundle with structure group Σn and
fiber n = {1, . . . , n}), then considering
F n (R∞ ) = {f ∶ n → R∞ ∣ f injective}, C n (R∞ ) = {S ⊆ R∞ ∣ #S = n},
and an embedding j:
j
Y / X × R∞
π
pr1
z
X
we can define fj ∶ X → C (R ), fj (z) ∶= j(π −1 (z)), which gives
n ∞
Y ≅ (fj∗ F n (R∞ )) ×Σn n.
2.) Let G = Gln (R). Recall the definition of the Stiefel manifolds
n ∶= {f ∈ Matm,n ∣ f injective},
Stm
for the Stiefel manifolds it also holds that πi (Stm

n ) = 0, i < m − n by an application of
Theorem 1.1.4 and there is a Gln (R)-action
Gln (R) × Stm

n Ð→ Stn ,
m
(g, f ) ↦ f ○ g,
by which we can define the Grassmann manifolds Grm n ∶= Stn /Gln (R). Again we work with
m
their direct limits

St∞
n ∶= lim Stm , Gr∞ ∞
n = Stn /Gln (R) = lim Grm ,
Ð→ n Ð→ n
m m
COBORDISM THEORY 23
which form the Gln (R)-principal bundle St∞ ∞

n Ð→ Grn . We also get vector bundles
Vn,m ∶= Stm
n ×Gln (R) R Ð→ Grn
n m
whose total space can also be written as Vn,m = {(V, v) ∣ V ∈ Grm

n , v ∈ V }.
For a vector bundle π ∶ V → X of rank n over a paracompact base X, we take the local
trivializations
≅
ϕi ∶ Ui × Rn Ð→ π −1 (Ui ),
with the open cover (Ui )i∈N , for which we find a subordinate partition of unity (λi )i∈N , which
induces a bundle map
ϕ
V / Gr∞ × R∞
n
f
X / Gr∞
n
with ϕ(v) ∶= (π(v), (λ1 (π(v))ϕ−1 −1 −

1 (v), λ2 (π(v))ϕ2 (v), . . . )) and f (x) ∶= ϕ(π 1x). By the
Gauss map we now have f ∗ Vn,∞ ≅ V and thus conclude
St∞
n ≅ E(Gln (R)), Gr∞
n ≅ B(Gln (R)).
3.) We take any closed subgroup G ⊂ Gln (R). Then G is a Lie group and
Gln (R) Ð→ Gln (R)/G
is a smooth G-principal bundle. Analogous to the previous case, we have
BG ≅ St∞ ∞
n /G = (Stn ×Gln (R) (GLn (R)/G))
4.) The previous examples are folklore; this one I found in the book [3]. We consider the
diffeomorphism group G = Diff(M ) with C ∞ -topology of a compact manifold M . We
consider the Fréchet submanifold of embeddings Emb(M, R∞ ) ⊆ C ∞ (M, Rm ), and the action
Diff(M ) × Emb(M, Rm ) Ð→ Emb(M, Rm ), (f, j) ↦ j ○ f.
We will check that this gives the Diff(M )-principal bundle
Emb(M, Rm ) Ð→ Emb(M, Rm )/Diff(M ).
Fix a j0 ∶ M ↪ Rm and w.l.o.g. let this be the inclusion. We consider a tubular neighborhood
r
U → M ↪ Rm and set
U ∶= {j ∶ M → Rm ∣ j(M ) ⊆ U, r ○ j ∶ M → M diffeomorphism} ⊆ Emb(M, Rm ).
Diff(M ) ⊆ C ∞ (M, M ) is open (note that for the homeomorphisms Homeo(M ) ⊆ C 0 (M, M )
is not open, since being a homeomorphism is not an open condition). The mapping
U Ð→ U, j ↦ j ○ (rj)−1
is Diff(M )-equivariant, since for h ∈ Diff(M ) also j ○ h ↦ j ○ (rj)−1 , and thus induces
ψ ∶ U/Diff(M ) Ð→ U.
Its inverse is ψ −1 (j) = Im(j), and we indeed get local trivializations
U/Diff(M ) × Diff(M ) Ð→ U.
([j], h) z→ ψ([j]) ○ h
So just as in the finite dimensional cases, we get
B(Diff(M )) ≅ Emb(M, R∞ )/Diff(M )

COBORDISM THEORY 24
We now turn to the general construction of classifying spaces. There are several posiibilities; we
use the one introduced by Milnor.
Definition 2.7.6 Join Let J be a set and Xj be a topological space with j ∈ J. The join of
{Xj }j∈J is defined as
∗j∈J Xj ∶= {(tj , xj )j∈J ∣ tj ∈ [0, 1], xj ∈ Xj , ∑ tj = 1 and only finitely many tj ≠ 0}/ ∼,
j
with
(tj , xj )j∈J ∼ (tj ′ , xj ′ )j ′ ∈J , ∶⇔ ∀ j ∈ J ∶ tj = tj ′ and if tj ≠ 0, then xj = xj ′ .
The join X ∶= ∗j∈J Xj is given the coarsest topology such that the following projection maps are
continuous
tj ∶ X Ð→ [0, 1], xj ∶ t−1
j (0, 1] Ð→ Xj .
Theorem 2.7.7 Milnor Construction of the Classifying Space The following is a functor
G z→ (EG → BG),
i.e. every topological group G has a universal G-bundle.
Proof: We shall see, that the following choices give the desired properties of the classifying
space
EG ∶= ∗i∈N G, BG ∶= EG/G,
where the latter arises from the free right G-action (tj , gj ) ⋅ g ∶= (tj , gj ⋅ g) on EG.
• We need to check local triviality of π ∶ EG → BG. The sets Ui ∶= t−1

i (0, 1] ⊆ EG are G-stable
and open, which also lets π(Ui ) ⊆ BG be open. The local trivializations are
Ui
h / π(Ui ) × G h((tj , gj )j∈J ) ∶= (π((tj , gj )j∈J ), gi )
π
y
π(Ui )
which are G-equivariant. The inverse map is induced by a section
s ∶ π(Ui ) Ð→ Ui , [(tj , gj )j∈J ] ↦ (tj , gj gi−1 )j∈J .
• EG ≅ ∗ is contractible. Set 0 ∶= (0, g). The two maps EG → EG that associate to (tn , gn )n∈N
the images
(0, (t1 , g1 ), (t2 , g2 ), . . . ), ((1, e), 0, 0, . . . ),
respectively, are homotopic, via the homotopy
((t, e), ((1 − t)t1 , g1 ), ((1 − t)t2 , g2 ), . . . ).
Moreover, the formula ((tt1 , g1 ), ((1 − t)t1 , g1 ), (t2 , g2 ), ⋯) defines a homotopy H1 from the
map (0, (t1 , g1 ), (t2 , g2 ), . . . ) to the identity. In a similar way, we define homotopies
H2 , H3 , . . .
H1 H2 H3
(0, (t1 , g1 ), (t2 , g2 ), . . . ) z→ ((t1 , g1 ), 0, (t2 , g2 ), . . . ) z→ ((t1 , g1 ), (t2 , g2 ), 0, (t3 , g2 ) . . . ) z→ ⋯
If we reparametrize the homotopies, such that the first takes time [0, 12 ], the second [0, 14 ]
and so on, then the infinite concatenation is well defined and gives a homotopy from
identity to the map (0, (t1 , g1 ), (t2 , g2 ), . . . ).
COBORDISM THEORY 25
Theorem 2.7.8
• Let π ∶ P → X be a G-principal bundle and X be paracompact, then there is a countable

cover {Ui }i∈N , with a subordinate partition of unity {λi }i∈N and a family of G-equivariant
maps {ϕi }i∈N , where
≅
ϕi ∶ P ∣Ui Ð→ G, and (π, ϕi )ϕi ∶ P ∣Ui Ð→ Ui × G,
that induce the map
P Ð→ EG.
p z→ (λi (π(p)), ϕi (p))i∈N
• Vice versa, if f ∶ P Ð→ EG is G-equivariant and f −1 (t−1 j (0, 1]) G-stable and open in P , then
Vj ∶= π(f −1 (t−1
j (0, 1])) is an open cover of X and the following commutes
f tj
P / EG 6/ R
π
λj
X
COBORDISM THEORY 26
3 Characteristic Classes
3.1 Definition and Basic Properties
Characteristic classes correspond to the cohomology of BG.
Definition 3.1.1 Characteristic Class Let G be a topological group and R a commutative ring.
A characteristic class for G-bundles with coefficients in R of degree n is a natural transformation
of functors
c ∶ PrinG (−) Ð→ Hsing
n
(−; R),
where PrinG (−), Hsing
n
(−; R) ∶ Top → Set.
Remark 3.1.2 Yoneda Lemma and Cohomology of BG The Yoneda lemma gives
PrinG (−) ≅ [−; BG],
which lets a characteristic class take values in the cohomology of BG:
c(EG → BG) ∈ H n (BG; R).
I.e. the characteristic classes form a ring, which is H ● (BG; R).
Remark 3.1.3 Conventions In the following, we shall consider K-vector bundles V → X for
K = R, C, H. When possible, we shall treat them on a common ground and thus adopt the
conventions listed in the following table.
K Structure Group Coefficient Ring F d Characteristic Class

R O(n) Z2 1 Stiefel − Whitney
wi (V ) ∈ H i (X, Z2 )
C U (n) Z 2 Chern
ci (V ) ∈ H 2i (X, Z)
H Sp(n) Z 4 Pontrjagin
Be aware however, that the Pontrjagin classes arising in the quaternionic case K = H, are not the
ones one usually finds in the literature.
In the complex case, we shall make use of the complex base (v1 , . . . , vn ), which induces the
positively oriented real base (v1 , iv1 , . . . , vn , ivn )
Remark 3.1.4 Goals The goal of this section is to calculate H ∗ (X; R) for the following pairs
(BO(n); Z2 ), (BU (n); Z), (BSO(n); Z2 ), (BSO(n); Q), (BO(n); Q).
Also we shall try to better understand the way in which elements of H ∗ (BG; R) correspond to
characteristic classes.
Definition and Lemma 3.1.5 Tautological Bundle For K = R, C, H, we set

H ∶= {(l, v) ∣ l ∈ KP ∞ , v ∈ l} and the canonical projection gives us the so called tautological bundle
H Ð→ KP ∞ .
Its dual L ∶= H ∗ gives the generator x ∶= e(L) ∈ H d (KP ∞ ; F) of the cohomology ring
H ● (KP ∞ ; F) ≅ F[x].
Definition 3.1.6 Outer Tensor Product Given two vector bundles V → X, W → Y , we define
their outer tensor product as
V ⊠ W ∶= [pr∗X V ⊗ pr∗Y W ] Ð→ X × Y.
COBORDISM THEORY 27
The following definition of the characetristic classes is due to Dold and can be found in the book
[5].
Definition 3.1.7 Chern − and Stiefel − Whitney Classes Consider the K = R, C vector
bundle V → X, then we have the Künneth isomorphism
≅ n
H dn (X × KP ∞ ; F) Ð→ ⊕ [H dk (X; F) ⊗F H d(n−k) (KP ∞ ; F)]
k=0
n
e(V ⊠ L) z→ ∑ [ak (V ) ⊗ xn−k ]
k=0
which defines the characteristic classes ak (V ) ∈ H dk (X; F), which we call
K=R∶ wk (V ) ∶= ak (V ) Siefel − Whitney Classes,

K=C∶ ck (V ) ∶= ak (V ) Chern Classes.
Setting the higher classes to zero:
if k > n, then wk (V ) ∶= 0, ck (V ) ∶= 0,
we define the total Stiefel-Whitney and Chern classes as

∞ ∞
w(V ) ∶= ∑ wk (V ), c(V ) ∶= ∑ ck (V ).
k=0 k=0
Theorem 3.1.8 Properties of the Chern − Classes Let V → X and W → X be R vector

bundles, then the Chern classes of V, W have the following properties:
1.) f ∗ c(V ) = c(f ∗ V ) for all continuous maps f ∶ Y → X,
2.) c0 (V ) = 1,
3.) c(V ⊕ W ) = c(V )c(W ),
4.) c(L) = 1 + x.
Remark 3.1.9 Analogous statements with analogous proofs hold for wi (V ) in the real case.
Proof:
1.) Follows from the naturality of the Künneth isomorphism.
2.) Let p ∈ X, then we can consider it as a map p ∶ {∗} → X and it suffices to show
p∗ c0 (V ) = 1 ∈ H 0 (∗). But we know p∗ c0 (V ) = c0 (p∗ X) = c0 (Cn → {∗}), and
e(Cn ⊠ L) = e(⊕ni=1 L) = e(L)n = xn . Now equating the coefficients gives the desired result.
3.) Let V have rank n and W rank m, then on one hand we have
n,m
e((V ⊕ W ) ⊠ L) = e((V ⊠ L) ⊕ (W ⊠ L)) = ∑ (ck (V ) × xn−k ) ∪ (cl (W ) × xm−l ),
k,l=0
∗
on the other, we also have e((V ⊕ W ) ⊠ L) = ∑n+m
k=0 ck (V ⊕ W ) × x
n+m−k
.
4.) Consider e(L ⊠ L) ∈ H 2 (CP ∞ × CP ∞ ) and set x1 ∶= e(L ⊠ C), x2 ∶= e(C ⊠ L), which gives
e(L ⊠ L) = c0 (L) × x2 + c1 (L) × 1 = 1 × x2 + zx1 × 1, z ∈ Z.
Take the interchanging map t ∶ CP ∞ × CP ∞ → CP ∞ × CP ∞ , t(x, y) ∶= (y, x), then
t∗ x1 = x2 , t∗ x2 = x1 t∗ (L ⊠ L) ≅ L ⊠ L,
and thus z = 1.
COBORDISM THEORY 28
Remark 3.1.10
• In the literature (e.g. [4]), the above properties (1-4) are usually stated as axioms.
• We have CP 1 ≅ S 2 and ⟨c1 (H); [CP 1 ]⟩ = −1. This convention was introduced by Hirzebruch,
in order to recover the usual Riemann-Roch formula for Riemann surfaces.
Theorem 3.1.11 Further Properties
5.) If V, W are line bundles, then c1 (V ) + c1 (W ) = c1 (V × W ).
6.) If rank (V ) = n, then cn (V ) = e(V ).
Proof:
6.) Let z ∶ {∗} → CP ∞ be the basepoint, then

n n
(1 × z)∗ e(V ⊠ L) = (1 × z)∗ ( ∑ ck (V ) × xn−k ) = ∑ ck (V ) × z ∗ xn−k = cn (V ),
k=0 k=0
on the other hand we have (1 × z)∗ e(V ⊠ L) = e(V ⊠ z ∗ L) = e(V ).
5.) Let f, g ∶ X → CP ∞ be continuous with f ∗ L = V , g ∗ L = W , then due to naturality, it suffices

to prove c1 (L ⊠ L) = c1 (L) × 1 + 1 × c1 (V ), and we indeed have
(6)
c1 (L ⊠ L) = e(L ⊠ L) = c1 (L) × 1 + 1 × c1 (V ).
Corollary 3.1.12 We have CP ∞ ≅ K(Z, 2) giving rise to the group isomorphism
H 2 (X; Z) ≅ [X, K(Z, 2)] ≅ [X, CP ∞ ] ≅ (PrinU (1) (X), ⊗)

c1 (L) z→ L.
Similarly, in the real case
H 1 (X; Z/2) ≅ [X, K(Z/2, 2)] ≅ [X, RP ∞ ] ≅ (PrinO(1) (X), ⊗)

c1 (L) z→ L.
Corollary 3.1.13 A real line bundle L → X is trivial iff w1 (L) = 0. Similarly, a complex line
bundle L → X is trivial iff c1 (L) = 0.
The following is a topological version of the technique, omnipresent in the theory of compact Lie
groups, to reduce statements on Lie groups to the maximal torus.
Theorem 3.1.14 Splitting principle Let V → X be a K-vector bundle of rank n, then there
exists a space Q and a map f ∶ Q → X, such that
• f ∗ is injective in H ∗ (−; F),
• f ∗ V = V1 ⊕ ⋯ ⊕ Vn with rank (Vi ) = 1.

COBORDISM THEORY 29
Proof: Consider the projective bundle q ∶ P V → X and the associated tautological line bundle
q ∗ V ⊇ HV ∶= {(l, v) ∣ l ∈ P V, v ∈ l} Ð→ P V.
Take the set B ∶= {1, c1 (HV ), c1 (HV )2 , . . . , c1 (HV )n−1 } ⊆ H ∗ (P V ), then for all x ∈ X the inclusion
jx ∶ P Vx ↪ P V lets jx∗ B be a base of H ∗ (P Vx = CP n−1 ) (Leray-Hirsch). Thus we have that q ∗ is
injective in cohomology and we conclude
q ∗ ≅ HV ⊕ q ∗ V /HV .
Induction over the rank of V now gives the claim.
◻
Remark 3.1.15
• We can iterate the above construction
q4 q3 q2 q
⋯ Ð→ P (q2∗ (q ∗ V /HV )/Hq∗ V /HV ) Ð→ P (q ∗ V /HV ) Ð→ P V Ð→ X
l1 ⊆ l2 ⊆ l3 z→ l1 ⊆ l2 z→ l
with dim(li ) = i. After n-steps, we arrive at Q = {(x, lo ⊆ ⋯ ⊆ ln ) ∣ x ∈ X, li ⊆ Vx }, thus each
fiber is a flag manifold
Fk (Cn ) ∶= {0 ⊆ X1 ⊆ ⋯ ⊆ Kk ⊆ Ck ∣ dim(Xi ) = i},
on which U (n) acts transitively and has the stabilizer subgroup
⎛ z1 0 0 ⎞
⎜ ⋱ ⋮ ⎟
⎜ ⎟.
⎜ 0 zk 0 ⎟
⎝ 0 ... 0 (∗) ⎠
If k = n, then Fn (Cn ) = U (n)/T (n), where T (n) is the maximal torus.
For a GLn (C)-principal bundle
P ×GLn (C) Fn (Cn ) Ð→ X
is injective in cohomology and gives
EU (n) ×U (n) U (n)/T (n) / BU (n)
5
=

BT (n)
• Let c1 (HV ), then p∗ V = HV ⊕ p∗ V /HV and c(p∗ V ) = (1 + x)c(p∗ V /HV ), which gives
∞
c(p∗ V /HV ) = ∑ xk c(p∗ V ).
k=0
• We always have #Weyl(G) = χ(G/T ), which is proved with Poincaré-Hopf, and thus
H ∗ (BG; Q) ≅ H ∗ (BT ; Q)#Weyl(G) .
Theorem 3.1.16 Uniqueness The splitting principle implies that if c̃ is a characteristic class
for C-vector bundles, which satifies (1 − 4), then c̃ = c.
Theorem 3.1.17 Let V be an R-vector bundle of rank n, then w1 (Λn V ) = w1 (V ).
Proof: The splitting principle shows, that it is sufficient to prove the claim for a sum of line
bundles V = V1 ⊕ ⋅ ⊕ Vn , and then w(V ) = ∏ni=1 (1 + w1 (Vi )), and
w1 (V ) = w1 (V1 ) + ⋯ + w1 (Vn ) = w1 (Λ1 V1 ⊗ ⋯ ⊗ Λ1 Vn ) = w1 (Λn (V1 ⊕ ⋯ ⊕ Vn )) = w1 (Λn V ).
◻
Corollary 3.1.18 Orientability A rank n vector bundle V → X is orientable, iff w1 (V ) = 0.
COBORDISM THEORY 30
3.2 Universal Classes

Lemma 3.2.1 Let W → X be a C-vector bundle, then
ck (W ) = (−1)k ck (W ).
Proof: Let W be one dimensional, then W ≅ W ∗ and
c1 (W ∗ ⊗ W ) = c1 (W ∗ ) + c1 (W ) = c1 (C) = 0.
If V = ⊕ni=1 Wi with Wi being line bundles, then

n n
c(V ) = ∏(1 + c1 (Wi )), c(V ) = ∏(1 − c1 (Wi ))
i=1 i=1
and the general case follows from the splitting principle.

◻
Remark 3.2.2 For the complexification VC ∶= V ⊗R C of an R-vector bundle V → X it also

follows that c2l+1 (V C ) = −c2l+1 (VC ), and thus
2c2l+1 (V C ) = 0.
Definition 3.2.3 Pontrjagin − Classes The k-th Pontrjagin-class of an R-vector bundle V → X

is defined as
pk (V ) ∶= (−1)k c2k (VC ) ∈ H 4k (X; Z).
If V has rank n, we define p0 (V ) = 1 and
∞
p(V ) ∶= ∑ pk (V ).
k=0
Theorem 3.2.4 Properties of Pontrjagin − Classes

1.) p(f ∗ V ) = f ∗ p(V )
2.) 2p(V ⊕ W ) = 2p(V )p(W )
3.) For rankV = 2m and V oriented: e(V )2 = pm (V ).
4.) If we start with a C-vector bundle V , then
pk (VR ) = (−1)k ∑ cp (V )cq (V ).

p+q=k
Remark 3.2.5 From 4.), we see that for complex vector bundles, Pontrjagin classes do not give
any new information.
Proof:
1.) Follows from the naturality of the Chern classes.
2.)
∞ ∞
2p(V ⊕ W ) = 2 ∑ pk (V ⊕ W ) = ∑ (−1)k 2c2k (VC ⊕ WC )
k=0 k=0
∞
= ∑ (−1)k 2 ∑ cp (VC )cq (WC )
k=0 p+q=2k
∞
= ∑ (−1)k 2 ∑ c2i (VC )c2j (WC ) = 2p(V )p(W )
k=0 i+j=k
COBORDISM THEORY 31
3.) The isomorphism V ⊕ V ≅ VC given by (v, w) ↦ v + iw is not orientation preserving. We

have the basis
(v1 , . . . , v2m , iv1 , . . . , iv2m ) of V ⊕ V,

(v1 , iv1 , . . . , v2m , iv2m ) of VC .
That is, we need 2m(2m − 1)/2 = m(2m − 1) transpositions, in order to identify the two basis
and we thus get the following sign:
e(VC ) = (−1)m(2m−1) e(V ⊕ V ).
Now we can get to the actual calculation:
pm (V ) = (−1)m c2m (VC ) = (−1)m e(VC ) = (−1)m+m(2m−1) e(V ⊕ V )

= e(V ⊕ V ) = e(V )2
where we have used m + m(2m − 1) ≡ 0 (mod2).
4.) p(VR ) = ∑∞ ∞
k=0 (−1) c2k (VC ) = ∑k=0 (−1) c2k (V )c2k (V ).
k k
Proof:
”⇒” This is easy, due to the fact, that a section of Λn V is a volume form.
”⇐” This follows immediately from Corollary 3.1.13.
It will be of crucial importance to understand the characteristic classes of the tangent bundle of
the complex projective space. For that, we need to determine the tangent bundle of the projective
space first.
Theorem 3.2.6 Let V be a finite-dimensional C-vector space, then there is a bundle

isomorphism
T (P V ) ≅ Hom(H, V )/C, V ∶= X × V.
Proof: Let n + 1 be the dimension of V . Consider the bundle maps
φ ∶ P V × End(V ) Ð→ T (P V )
(l, f ) z→ dt ∣t=0 (e
d tf
⋅ l) ∈ Tl (P V )
and
∣H
End(V ) Ð→ Hom(H, V ) Ð→ Hom(H, V )/Hom(H, H).
(3)
given by restriction and quotienting. Consider the kernel of φ. If φ(l, f ) = 0, then l ⊂ V is an

eigenspace of f and thus f -invariant and hence contained in the kernel of φ. By a dimension
count, φ is surjective and the kernels of φ and ψ agree. As ψ is an epimorphism, the result
follows.
◻
COBORDISM THEORY 32
Lemma 3.2.7 It holds that p(T CP n ) = (1 + x2 )n+1 .
Proof: We have C ⊕ T CP n = L ⊕ ⋯ ⊕ L and thus p(T CP n ) = p(L)n+1 , where p(L) = (1 + x2 ),

´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹¸ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶
n+1
since c(LC ) = (1 − x)(1 + x) = (1 − x2 ).
Definition 3.2.8 Characteristic Numbers Let M n be R-oriented and c ∈ H n (BSO(n); R) a

characteristic class, then its associated characteristic number is defined as
⟨c(T M ), [M ]⟩ =∶ c([M ]).
Theorem 3.2.9 Consider the inclusion j ∗ ∶ H n (BSO(n + 1); R) Ð→ H n (BSO(n); R) and let
c ∈ H n (BSO(n + 1); R) be a characteristic class. If M0 , M1 are oriented bordant, then
(j ∗ c)([M0 ]) = (j ∗ c)([M1 ]).
Remark 3.2.10 Let M n be a manifolds, whose tangent bundle arises as the pullback by
fT M ∶ M → BSO(n), then (fT M )∗ [M ] ∈ Hn (BSO(n)), and
⟨c, (fT M )∗ [M ]⟩ = ⟨(fT M )∗ c, [M ]⟩ = c([M ]).
Remark 3.2.11 The theorem can be applied to polynomials of Pontrjagin classes, but not to the
Euler class.
3.3 Cohomology of classfying spaces

We need one final ingredient for the computation of the cohomology of classifying spaces, namely
the transfer for finite coverings.
Remark 3.3.1 Let G be a finite group. For a G-principal bundle p ∶ X Ð→ Y , we consider a

singular simplex c ∶ ∆n → Y . Because the simplex is simply-connected, there exist ∣G∣ different lifts
c̃ of c to X and two lifts differ by some g ∈ G.
Definition 3.3.2 Transfer Map Let G be a finite group and let p ∶ X Ð→ Y be a G-principal
bundle. We define the transfer map
trf p ∶ C∗ (Y ) Ð→ C∗ (X)
by the following formula

trf p (c) ∶= ∑ c̃.
c̃ lift of c
This equals ∑g∈G gc̃ if a specific lift is chosen.
Remark 3.3.3 Transfer Map Since trf p is a chain map, we get the induced map
trf p ∶ Hn (Y ; R) Ð→ Hn (X; R),
Considering the induced map of chain complexes
p∗ ∶ C∗ (X) Ð→ C∗ (Y ),
we have
trf p ○ p∗ (c) = ∑ gc, p∗ ○ trf p (c) = ∣G∣c. (4)
g∈G
COBORDISM THEORY 33
By dualizing, we get an induced map on cochain complexes (and hence in cohomology)
trfp ∗ ∶ C ∗ (X; R) Ð→ C ∗ (Y ; R).
Equation 4 implies
trf ∗p ○ p∗ (y) = ∣G∣y, p∗ ○ trf ∗p (y) = ∑ gy
g∈G
on the cochain level. Because (pg)∗ y = p∗ y, p∗ maps into the G-invariant part
p∗
C ∗ (Y ; R) Ð→ C ∗ (X; R)G .
This map is injective (true for any covering), and it surjective if R = Q because
1 1 ∗ 1
x= ∑ gx = p ○ trf ∗p (x) = p∗ ( trf ∗p (x)),
∣G∣ g ∣G∣ ∣G∣
Because G is finite and we are considering Q-vector spaces, H ∗ (C ∗ (X; Q)G ) ≅ H ∗ (X; Q)G .
Altogether, we have proven that
p∗ ∶ H ∗ (Y ; Q) → H ∗ (X; Q)G
is an isomorphism.
Theorem 3.3.4 Ring − Isomorphisms The following ring homomorphisms are isomorphisms
1.) Z[c1 , . . . , cn ] Ð→ H ∗ (BU (n); Z),

2.) Z[c2 , . . . , cm ] Ð→ H ∗ (BSU (n); Z),
3.) Z2 [w1 , . . . , wn ] Ð→ H ∗ (BO(n); Z2 ),
4.) Z2 [w2 , . . . , wn ] Ð→ H ∗ (BSO(n); Z2 ),
5a.) Q[p1 , . . . , pm ] Ð→ H ∗ (BO(2m + 1); Q),
5b.) Q[p1 , . . . , pm ] Ð→ H ∗ (BO(2m); Q),
6.) Q[p1 , . . . , pm ] Ð→ H ∗ (BSO(2m + 1); Q),
Q[p1 , . . . , pm , e]
7.) Ð→ H ∗ (BSO(2m); Q).
(pm − e ) 2
Proof: All the proofs are by induction over n or m respectively.

1.) The key fact needed for the inductive proof is the existence of a homotopy commutative
diagram
≃ / SVn
BU (n − 1)
p

BU (n)
id / BU (n),
where SVn is the unit sphere bundle of the universal vector bundle and the left vertical map
is induced by the inclusion U (n − 1) → U (n). A similar statement is true for the orthogonal
groups.
Now we evoke the Gysin-Sequence
(∗) p!
Z[c1 , . . . , cn−1 ] H ∗ (BU (n − 1)) / H ∗ (BU (n))
hh
p∗
∃j ⋅cn
, v
∗
H (BU (n))
COBORDISM THEORY 34
where (∗) is the induction hypothesis and j exists, since p∗ is surjective (because the Chern
class ci on BU (n − 1) extends to BU (n)) and every surjection on a polynomial ring has a
split.
Due to the surjectivity of p∗ , we have that p! = 0 and thus ⋅cn is injective, which means for a
x ∈ H ∗ (BU (n)), that we have
x = cn ⋅ y + jp∗ (x)
for some y if smaller degree. We can now repeat this argument with y.
2,3,4) Are proved similarly.
6,7) (i) Here we obtain from the Gysin sequence and the inductive assumption a commutative
diagram
p!
Q[p1 , . . . , pm−1 ] H ∗ (BSO(2m − 1)) / H ∗ (BSO(2m))
ii
p∗
⋅e
u
∗
H (BSO(2m))
Again, p∗ is surjective and a similar argument as before applies.

(ii) For p ∶ BSO(2m) Ð→ BSO(2m + 1) the induced map p∗ is not surjective, since e is not
in the image. Instead, p∗ is injective.
p!
Q[p1 ,...,pm ,e]
H ∗ (BSO(2m)) / H ∗ (BSO(2m + 1))
(p2m −e) i
p∗
⋅e=0
u
H ∗ (BSO(2m + 1))
We need to determine the kernel of the Gysin map p! . For that, one uses the identity
p! (p∗ x ⋅ y) = x ⋅ p! (y). This implies that if x is a polynomial in the Pontrjagin classes,
then p! (x) = 0 and p! (ex) = p! (e)x. Therefore the kernel is the ideal eH ∗ (BSO(2m)),
as was to be shown.
5.) BO(n) ≅ EO(n) ×O(n) O(n)/SO(n), so BSO(n) → BO(n) is a two sheeted covering.
Idenifying
BSO(n) = {(V, o) ∣ V ⊆ R∞ , dim(V ) = n, o orientation of V },
we have a Deck-transformation
T ∶ BSO(n) Ð→ BSO(n), T (V, o) = (V, −o).
We have a Z2 action Z2 × H ∗ (BSO(n); Q) → H ∗ (BSO(n); Q) and get the isomorphism

≅
H ∗ (BO(n); Q) Ð→ H ∗ (BSO(n); Q)Z2 .
The fact that T (e) = −e and T (pi ) = pi , finishes the proof.
◻
COBORDISM THEORY 35
3.4 Multiplicative Sequences

Theorem 3.4.1 Let V → X be an oriented vector bundle and
F (V ) ∈ H ∗∗ (X; Q) ∶= ∏ H k (X; Q)
k≥0
such that
F (X × R) = 1, F (V ⊕ W ) = F (V )F (W ),
then the following is a ring homomorphism
Ω∗ Ð→ Q.
[M ] z→ ⟨F (T M ), [M ]⟩
Definition 3.4.2 Multiplicative Characteristic Class A multiplicative characteristic class
with values in Q is a natural transformation
F ∶ PrinG (−) Ð→ H ∗∗ (−; Q).
In the following we shall need the set of multiplicative characteristic classes of K-vector bundles
with values in Q that fulfill certain additional requirements. We set
MCCK ∶= {F ∶ PrinGln (K) (−) Ð→ H ∗∗ (−; Q) ∣ F natural, F (X × K) = 1, F (V ⊗K W ) = F (V )F (W )}
Remark 3.4.3 Let F ∈ MCCC and consider the dual canonical bundle L → CP ∞ , then
f (x) ∶= F (L) ∈ H ∗∗ (CP ∞ ; Q) = Q[[x]] and
f (x) = 1 + f1 x + f2 x2 + ⋯ ∈ 1 + xQ[[x]].
Theorem 3.4.4 Hirzebruch The following maps are bijections. In the complex case
MCCC Ð→ 1 + xQ[[x]],
F z→ F (L)
and in the real case
MCCR Ð→ 1 + x2 Q[[x2 ]].
F z→ F (LR )
Proof: We shall only prove the complex case.
• First we prove injectivity. Let F0 (L) = F1 (L), then due to the splitting principle, we have
F0 (p∗ V ) = F0 (L1 ⊕ ⋯ ⊕ Ln ) = F0 (L1 )⋯F0 (Ln ) = F1 (L1 )⋯F1 (Ln ) = F1 (p∗ V ).
• Now we come to surjectivity. Let f ∈ 1 + xQ[[x]] and V = ⊕ni=1 Li for line bundles Li . Then
we define
n n n
F (V ) ∶= ∏ F (Li ) ∶= ∏ f (c1 (Li )) ≡ ∏ f (xi ).
i=1 i=1 i=1
The main theorem on symmetric polynomials, says that for monomials σi of even order,
there is a Krn (σ1 , . . . , σn ) such that
n
Krn (σ1 (x1 , . . . , xn ), . . . , σn (x1 , . . . , xn )) = (∏ f (xi )) ∈ H 2r (X; Q),
i=1 2r
and
n n n
c(V ) = ∏ c(Li ) = ∏(1 + xi ) = ∑ σi (x1 , . . . , xn )
i=1 i=1 i=1
so we have ci = σi (x1 , . . . , xn ) and thus
n
Krn (c1 (V ), . . . , cn (V ) = (∏ f (xi )) .
i=1 2r
COBORDISM THEORY 36
Remark 3.4.5 So for every f ∈ 1 + x2 Q[[x2 ]], we get an element F ∈ MCCR and thus the ring
homomorphism
∗ ⊗ Q → Q,
ΩSO [M ] ↦ F ([M ]).
In the following, we shall consider the new series
∞ ∞ ∞
⊕(n+1)
g(t) ∶= ∑ F ([CP ])t = ∑ ⟨F (T CP ), [CP ]⟩ t = ∑ ⟨F (L
n n n n n
), [CP n ]⟩
n=0 n=0 n=0
∞ ∞
= ∑ ⟨f (x)
n+1
, [CP n ]⟩ tn = ∑ ϕ(f
n t
) n
n=0 n=0
(f )
where ϕn is the n-th coefficient of f (x)n+1 .
Theorem 3.4.6 Hirzebruch Let f and g be as above, then setting

x
q(x) ∶= , h(x) ∶= q −1 (x) i.e. h(q(x)) = x,
f (x)
it holds that
g(x) = h′ (x).
(f )
Proof: We take f to have a positive convergence radius. Then, since ϕn was the n-th
coefficient of f (x)n+1 , we have
∞ ∞
(f ) n 1 f (x)n+1 1 1 f (x)
g(t) = ∑ ϕn t = ∑ ( ∫ n+1
dx) tn = ∫ f (x)t
dx
n=0 n=0 2πi c x 2πi c 1 − x
x
1 1 1 1 dx 1 h′ (z)
= ∫ dx = ∫ = ∫ dz = h′ (z)
2πi c 1 − g(x) q(x)
t 2πi c q(x) − t 2πi c z − t
Corollary 3.4.7
1.) There is exactly one f ∈ 1 + x2 Q[[x2 ]], such that for given an ∈ Q and a0 = 1, we have
ϕ(f )
n = an ∀n ∈ N.
2.) The following homomorphism is injective
Q[y1 , y2 , . . . ] Ð→ Ω∗ ⊗ Q
yi z→ [CP 2i ].
(f )
3.) If ϕn = 1 for all n ∈ N, then g(t) = ∑∞k=0 t
2k
= 1−t
1
2 , which gives h(t) = arctanh(t),
q(x) = tanh(x) and
x x
f (x) = = .
q(x) tanh(x)
COBORDISM THEORY 37
3.5 Bordism vs. Homotopy: The Pontrjagin-Thom Construction

Definition 3.5.1 Let M m be a smooth manifold, V → X a smooth vector bundle of rank k and
A ⊆ M a closed subset. We define
≅
L(M, A; V ) ∶= {(N, g, ϕ) ∣ N ⊆ M closed codim k, N ∩ A = ∅, g ∶ N → X, ϕ ∶ νN
M
→ g ∗ V }.
Let everything be as in the definition. Consider the Thom space Th(V ) ∶= D1 V /SV with
∞ = {SV }. To any given continuous map
f ∶ (M, A) Ð→ (Th(V ), ∞),
we can associate an element (f −1 (0), f ∣f −1 (0) , ϕ) ∈ L(M, A; V ). However we need to say what ϕ is
supposed to be. For that, we consider a tubular neighborhood U = D1/2 V of M in D1 V where
f −1 (U ) ⊆ M ∖ A. Making use of a smooth approximation, we can assume, that f ∣f −1 (0) is smooth.
If f is transverse to the zero section: f ⋔ 0, we have
f −1 (0) ⊂ M ∖ A, dim(f −1 (0)) = m − k, M

νN = (f ∣N )∗ ν0U ≅ (f ∣N )∗ V.
≅
So there is a canonical map ϕ ∶ νN
M
Ð→ (f ∣N )∗ V .
Lemma 3.5.2 There exists a map
{f ∶ (M, A) → (Th(V ), ∞) ∣ f ⋔ 0} Ð→ L(M, A; V ),

f z→ (f −1 (0), f ∣f −1 (0) , ϕ).
Let two smooth functions f0 , f1 ∈ {f ∶ (M, A) → (Th(V ), ∞) ∣ f ⋔ 0} be homotopic relative to A

with the homotopy
F ∶ [0, 1] × (M, A) → (Th(V ), ∞)
being constant on [0, ε], [1 − ε, 1] and w.l.o.g. F ∣F −1 (U ) smooth and F ⋔ 0. We further consider the
codimension k submanifold W ∶= F −1 (0) ⊆ [0, 1] × M for which W ∩ ([0, 1] × A) = ∅ together with
[0,1]×M
maps G ∶ W → X and ψ ∶ νW → G∗ M , such that
[0,1]×M
G∣Ni = fi ∣Ni , νW ∣Ni = νN
M
i
, ψ∣Ni = ϕi ,
where (Ni , gi , ϕi ) are the data obtained from fi under the above map.
Lemma 3.5.3 There is an induced map
T ∶ [(M, A), (Th(V ), ∞)] Ð→ L(M, A; V )/ ∼,
where (N0 , g0 , ϕ0 ) ∼ (N1 , g1 , ϕ1 ), iff there exists a closed codimension k submanifold

[0,1]×M
W ⊆ [0, 1] × M with W ∩ ([0, 1] × A) = ∅ and maps G ∶ W → X, ψ ∶ νW → G∗ M , such that
G∣Ni = gi ∣Ni , ψ∣Ni = ϕi .
Theorem 3.5.4 Pontrjagin − Thom T is bijective.
Proof: We will construct an inverse of T :
P ∶ L(M, A; V ) Ð→ [(M, A), (Th(V ), ∞)].
Now (N, f, ϕ) is given. We chose a tubular neighborhood N ⊂ U ⊆ M ∖ A and consider the map
⎧
⎪
⎪x, if x ∈ U
M Ð→ U /∂U = M /(M ∖ U ), x↦⎨
⎪
⎩∞, if x ∉ U.
⎪
COBORDISM THEORY 38
≅ ≅
For a tubular map νN M
Ð→ U we get the induced map U /∂U Ð→ Th(νN
M
), which induces the first
of the following two maps
(M, A) Ð→ (Th(νN
M
), ∞) Ð→ (Th(V ), ∞),
≅
the second one being induced by g ∶ N → X and ϕ ∶ νN M
Ð→ g ∗ V. Defining
L̃(M, A; V ) ∶= {(N, g, ϕ, t) ∣ (N, g, ϕ) ∈ L(M, A; V ), t ∶ νN → M ∖ A tubular map}, the above
M
concatenation induces a map
L̃(M, A; V )/ ∼ Ð→ [(M, A), (Th(V ), ∞)],
where (N0 , g0 , ϕ0 , t0 ) ∼ (N1 , g1 , ϕ1 , t1 ), iff there is a (W, G, ψ, T ) with (W, G, ψ, ) as above and
[0,1]×M
T ∶ νW → [0, 1] × (M ∖ A) with a tubular map such that T ∣Ni = ti .
Adding the fact that L̃(M, A; V ) ≅ L(M, A; V ), proves the claim.
Example 3.5.5 Let M = B × S n and A = B × {∗}, then we have
L(M, A; V )/ ∼≅ [B × (S n , ∗); (Th(V ), ∞)] = [(S n , ∗); map(B, Th(V ))] = πn (map(B, Th(V )))
3.6 Pontrjagin-Thom Construction and Homology

Take V → X to be an oriented vector bundle of rank k and f ∶ S n+k → Th(V ) with f ⋔ 0. Setting
M ∶= f −1 (0), we get an induced map c ∶ S n+k → Th(νM
S n+k
). We orient M , such that
S n+k S n+k
T M ⊕ νM = M × R . With τ ∈ H (Th(νM )), it holds that
n+k k
τ ∩ c∗ [S n ] ∈ Hn (ν) ≅ Hn (M ),
which gives τ ∩ c∗ [S n ] = (−1)nk [M ].
Corollary 3.6.1 Let W be an oriented bordism from M0 to M1 and ji ∶ Mi ↪ W the

corresponding inclusions, then
(j0 )∗ [M0 ] = (j1 )∗ [M1 ].
Proof: Embedding W in [0, 1] × Rn for some n ∈ N, gives c ∶ ([0, 1] × Rn )+ → Th(νW

n+1
R
) and we
have
([0, 1] × Rn )+ ≅ [0, 1]+ ∧ S n ≅ ([0, 1] × S n )/([0, 1] × {∗}).
Definition 3.6.2 Hurewicz − Homomorphism Let V → X be a vector bundle of rank k, then

the Hurewicz-homomorpism hur is defined as the following map, factoring over the
Thom-isomorphism (x ↦ τ ∩ x):
πn+k Th(V )
hur / Hn (X)
8
≅
'
Hn+k Th(V )
That is for f ∶ S n+k → Th(V ) with f ⋔ 0 and M ∶= f −1 (0), it holds that
hur([f ]) = (τ ∩ f∗ [S n ]) = (−1)nk f∗ [M ].
COBORDISM THEORY 39
4 Spectra and the Bordism Ring

4.1 Spectra
Definition 4.1.1 Stable Vector Bundle A vector bundle V → X is called stable, iff it has the
following additional data:
1.) A filtration X0 ⊂ X1 ⊂ ⋯ ⊂ X of X,
2.) rank n vector bundles Vn → Xn ,
3.) isomorphisms εn ∶ Vn ⊕ R → Vn+1 ∣Xn .
Two stable vector bundles V0 , V1 over X are called equivalent (concordant), iff there is a stable
vector bundle V over X × [0, 1], such that VX×0 = V0 and VX×1 = V1 .
Example 4.1.2 Stable normal bundle.
Definition 4.1.3 (Ω) − Spectrum A spectrum is a sequence of pointed spaces {Xn }n∈N and
maps
εn ∶ ΣXn → Xn+1 .
A spectrum {Xn , εn }n∈N is called Ω − spectrum, iff εn is a homotopy equivalence for all n ∈ N.
Remark 4.1.4 An equivalent definition for a spectrum is, to prescribe maps Xn → ΩXn+1 .
Definition 4.1.5 Thom Spectrum Let V → X be a stable vector bundle, The Thom spectrum
Th(V ) is defined by setting Th(V )n ∶= Th(Vn ) and taking the maps
ε∗n
Σ(Th(V )n ) = Th(Vn ⊕ R) Ð→ Th(Vn+1 ) = Th(V )n+1 .
where εn ∶ Vn ⊕ R → Vn+1 ∣Xn is the isomorphism, which is part of the data of the given stable
vector bundle.
Definition 4.1.6 Eilenberg − MacLane Spectrum En = K(Z; n)
Theorem 4.1.7 Generalized (Co)homology Theories Let {En , εn }n∈N be a spectrum and X
be a space, then the following maps
(Σ∧id)∗ (εk ∧id)∗
πn+k (Ek ∧ X+ ) Ð→ πn+k+1 (ΣEk ∧ X+ ) Ð→ πn+k+1 (Ek+1 ∧ X+ )
make it possible to define the generalized homology theory
En (X) ∶= lim πn+k (Ek ∧ X+ ).

Ð→
k
For a pair (X, A), we can define the generalized (co)homology theories
En (X, A) ∶= lim πn+k (Ek ∧ X/A),

Ð→
k
E (X, A) ∶= lim[(X, A), (Ωk−n Ek , ∗)].
n
Ð→
k
Example 4.1.8 By the above construction, the Eilenberg-Mclane spectrum En = K(Z; n), gives
E n (X, A) = lim[(X, A), (Ωk−n K(Z; k), ∗)] = lim[(X, A), K(Z; n)] = H n (X, A; Z).
Ð→ Ð→
k k
COBORDISM THEORY 40
Definition 4.1.9 Homotopy Groups of a Spectrum Let E = {En , εn }n∈N be a spectrum, then
we define its homotopy groups by
πn (E) ∶= lim πn+k (Ek ) = En (∗).

Ð→
k
Theorem 4.1.10 Brown Representation Theorem On the category on CW-complexes, all

generalized (co)homology theories have an associated spectrum, which gives rise to them by the
above construction.
Definition 4.1.11 Thom Spectrum We consider the universal bundle ESO(n) → BSO(n) and
set
Vn ∶= ESO(n) ×SO(n) Rn ,
which lets V ∶= lim Vn be a stable vector bundle. The Thom spectrum is now defined as
Ð→n
M SO ∶= Th(V ) = lim M SO(n),
Ð→
n
with M SO(n) = Th(Vn ).
Definition 4.1.12 The to the Thom spectrum associated generalized (co)homology theories are
denoted
M SOn (X) ∶= lim πn+k (M SO(k) ∧ X+ ), M SOn (X) ∶= lim[X+ , Ωk−n M SO(k)].
Ð→ Ð→
k k
Remark 4.1.13 Our goal is to calculate the bordism ring. In order to do so, the following
isomorphism will come in very handy
π∗ (M SO) ≅ ΩSO
∗ ,
and explains our interest in the Thom spectrum.
Theorem 4.1.14 Pontrjagin − Thom Let M n be a closed manifold, then
πn+k (M SO(k)∧X+ ) = {(M n , (f, g), ϕ) ∣ M ⊆ Rn+k , f ∶ M → BSO(k), g ∶ M → X, ϕ ∶ νM ≅ f ∗ VR }/ ∼,
and
lim πn+k (M SO(k) ∧ X+ ) = {(M n , g, o) ∣ g ∶ M → X, o orientation on M n }/ ∼ .
Ð→
k
Where in both cases ∼ means bordant.
Remark 4.1.15
1.) The following holds:
M SOn (X) = lim[X+ , Ωk−n M SO(k)] = lim[Σk−n X+ , M SO(k)]

Ð→ Ð→
k k
= lim[(S k−n × X; {∗} × X), M SO(k)]
Ð→
k
PT
= lim L((S k−n × X; {∗} × X), Vk )
Ð→
k
= lim{M ⊆ Rk−n ∣ dim(M ) = dim(X) − n, νM oriented pr ∶ M → X proper}/ ∼
Ð→
k
= {p ∶ M → X ∣ p proper, oriented}/ ∼ .
2.) If M n is closed and oriented, then

≅
M SOk (M ) Ð→ M SOn−k (X).
COBORDISM THEORY 41
4.2 Calculation of the Oriented Bordism Ring

Theorem 4.2.1 The following map is an isomorphism
Q[y1 , y2 , . . . ] Ð→ Ω∗ ⊗ Q.
yn z→ [CP n ]
Proof: We have already seen injectivity. Let p(n) be the number of partitions of n ∈ N, then
⎧
⎪
⎪0, if i ≠ 4N
dim Q[y1 , y1 , . . . ]i = ⎨
⎪
⎪ if i ∈ 4N.
⎩p(i/4),
The surjectivity can be seen as follows:
Ωn ⊗ Q = πn (M SO) ⊗ Q
hur Thom
= lim πn+k (M SOk ) ⊗ Q Ð→ lim Hn+k (M SOk ; Q) ≅ lim Hn (BSO(k); Q).
Ð→ Ð→ Ð→
k k k
Now Hn (BSO(k); Q) is dual to H (BSO(k); Q), which has the same rank as Q[y1 , y2 , . . . ].
n
4.3 The Signature

In the following, let M 4k be a closed and oriented manifold. We consider the following symmetric,
bilinear, non-degenerate form
β ∶ H 2k (M ; R) × H 2k (M ; R) Ð→ R.
(x, y) z→ ⟨x ∪ y, [M ]⟩
With Sylvester’s theorem, we know, that there is a basis {vi }i∈I , such that
β = diag(1, . . . , 1, −1, . . . , −1),

´¹¹ ¹ ¹ ¹ ¸ ¹ ¹ ¹ ¹ ¹¶ ´¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¸¹¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¹ ¶
r s
with r, s well defined.
Definition 4.3.1 Signature Let everything be as above, however M n be of arbitrary dimension

n. Then we define the signature of M to be
⎧
⎪
⎪sign(β), if n ∈ 4N
sign(M ) ∶= ⎨ with sign(β) ∶= r − s.
⎪
⎪ if n ≠ 4N,
⎩0,
In order to prove some of the properties of the signature, we shall need a little linear algebra.
Lemma 4.3.2 Let (V, β) be a R-vector space with symmetric bilinear form and U ⊆ V be an
isotropic subspace (i.e. β∣U ×U = 0). If dim U = 12 dim V, then sign(β) = 0.
Proof: We chose a complement W of U and a basis w1 , . . . , wn of W , such that (wi , wj ) = δij ai

with ai ∈ {±1, 0}. Now (ui , wj ) exists due to non-degeneracy and we get the matrix
⎛ 1 ⎞
⎜ −1 1 ⎟
⎜ ⎟
⎜ 0 ⎟
⎜ ⎟
⎜ ⎟
⎜ ⎟
⎜ ⎟
⎜ 1 0 ⎟
⎝ ⎠
thus sign(V ) = 0.
COBORDISM THEORY 42
◻
Lemma 4.3.3 Let (V0 , ω0 ), (V1 , ω1 ) be symplectic vector spaces, (V0 ⊗ V1 , ω0 ⊗ ω1 ) symmetric,
then
sign(V0 ⊗ V1 , ω0 ⊗ ω1 ) = 0.
Proof: Take an isotropic subspace U ⊆ V0 with dim U = 1
2 dim V0 , then (ω0 ⊗ ω1 )∣U ⊗V1 is
isotropic.
◻
Lemma 4.3.4 Properties of the Signature
1.) sign(−M ) = −sign(M ),
2.) sign(M ⊔ N ) = sign(M ) + sign(N ),
3.) sign(CP 2n ) = 1,
4.) sign(∂M ) = 0,
5.) sign(M × N ) = sign(M ) ⋅ sign(N ).
Proof:
1.)-3.) These are easy.
4.) Let dim(∂M ) = 4n and j ∶ ∂M ↪ M be the inclusion. Then
j∗
H 2n (M ) / H 2n (∂M ) δ / H 2n+1 (M, ∂M )
≅ ≅
j∗
H2n (∂M ) / H2n (M )
and we have ⟨j ∗ x ∪ j ∗ y, [M ]⟩ = ⟨x ∪ y, j∗ [∂M ]⟩ = 0, since j∗ [∂M ] = 0. Thus Im(j ∗ ) is

isotropic. Also we have
dim Imj ∗ = dim ker δ = dim ker j∗ = dim H2n (∂M ) − dim Imj∗ = dim H2n (∂M ) − dim Imj ∗ ,
which gives dim Imj ∗ = 1
2 dim H2n (∂M ).
5.) Let dim M = n and dim N = 4k − n. We define As ∶= H s (M ) ⊗ H 2k−s (N ), which gives

H 2k (M × N ) = ⊕ (As ⊕ An−2 ) ⊕ An/2 .
0=s<n/2
The form is zero on As × At , if s + t ≠ n, thus the above decomposition is orthogonal. Now

As ⊆ As ⊕ An−s is isotopic and thus
sign(H 2k (M × N )) = sign(An/2 ).
– If n is odd we have An/2 = 0 and thus sign(M × N ) = 0 = sign(M ) ⋅ sign(N ).
– If n = dim M = 4p + 2, dim(N ) = 4q + 2, we have p + q − 1 = k and
sign(H 2k (M × N )) = sign(A2p+1 ) = sign(H 2p+1 (M ) ⊗ H 2q+1 (N )) = 0
since both are symplectic forms.
– If dim M = 4p, dim N = 4q, we have
sign(M × N ) = sign(H 2p (M ) ⊗ H 2q (N )) = sign(M ) ⋅ sign(N ).
◻
Theorem 4.3.5 Hirzebruch Let L ∈ H ∗ (BSO; Q) associated to x
tanh x , then
sign(M ) = ⟨L(T M ), [M ]⟩ .
COBORDISM THEORY 43
References
[1] T. Bröcker, K. Jänich: Einführung in die Differentialtopologie, Springer-Verlag, 1973.
[2] T. tom Dieck: Algebraic Topology. EMS Textsbooks in Mathematics, Zürich, 2008.
[3] Andreas Kriegl, Peter W. Michor: The Convenient Setting of Global Analysis. AMS
Mathematical Surveys and Monographs Volume 53, 1997.
[4] John Milnor, James D. Stasheff: Characteristic Classes, Princeton University Press (AM-76),
1974.
[5] Matthias Kreck: Differential Algebraic Topology, AMS Graduate Studies in Mathematics
Volume 110, 2010.
[6] Hirsch: Differential topology
[7] G. E. Bredon: Topology and Geometry

Bordism Skript

Uploaded by

Copyright:

Available Formats

Bordism Skript

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

Bordism Skript

Uploaded by

Copyright:

Available Formats

A lecture course on Cobordism Theory

2 Orientation and (Co)homology 7

4 Spectra and the Bordism Ring 39

Definition 1.0.3 Let M m ⊂ N n be a submanifold. The codimension of M is the number n − m.

Proposition 1.0.4 If x ∈ N is a regular value of the map f ∶ M m → N n , then f −1 (x) ⊂ M is a

We now come to the most important fundamental result of differential topology.

Theorem 1.0.8 The Whitney Embedding Theorem Let M m , N n be manifolds and A ⊂ M

Definition 1.0.11 Tubular Map Let f ∶ M → N be an embedding. A tubular map for f is a

1. The restriction of t to the zero section is equal to f .

Theorem 1.0.12 Each embedding has a tubular map.

1. for each fixed r, , the map (t, x) ↦ σtr, (x) is an isotopy.

3. σtr, ∣Br (0) = id.

4. σ1r, (Rn ) = Br+ (0).

Theorem 1.0.14 Let f ∶ M → N be an embedding and t0 , t1 ∶ ν f → N be tubular maps. Then

Theorem 1.0.15 Smooth approximation theorem Let M, N be manifolds, N with a complete

Definition 1.1.1 Let M ; X; Y be smooth manifolds, f ∶ X → M , g ∶ Y → N be smooth maps. Then

(Tx f )(Tx X) + (Ty g)(Ty Y ) = Tz Z.

We write f ⋔ g when f and g are transverse. If g is the inclusion of a submanifold Y ⊂ M , we

One of the important properties of transversality is:

Proposition 1.1.2 Let M ; X; Y be smooth manifolds, f ∶ X → M be a smooth map and Y ⊂ N .

Theorem 1.1.3 The Transversality Theorem Let f ∶ M → N and g ∶ L → N be smooth maps.

Theorem 1.1.4 Let M m and N n be smooth manifolds, g ∶ N n → M m be a smooth map. Let

Example 1.1.5 M = S m , N = ∗. Then M ∖ N ≅ Rm and hence contractible. We obtain that

Example 1.1.7 Let M be an m-dimensional manifold. The ordered configuration space of M

Definition 1.2.2 Let V → X be a real vector bundle. An orientation of V is a family x , x ∈ X,

1.3 The cobordism relation

Let M be a manifold with boundary. If x ∈ ∂M , the tangent space Tx ∂M ⊂ Tx M is a codimension

Definition 1.3.3 Oriented Bordism Let Mi , i = 0, 1 be oriented closed manifolds. An oriented

Ωn ∶= {M n ∣ M closed oriented smooth n − manifold}/ ∼

where the equivalence relation is now taken to be only by oriented bordisms.

• Reflexivity: [0, 1] × M is a bordism from M to M .

• Symmetry: If W ∶ M0 → M1 is a bordism, then so is −W ∶ M1 → M0 .

• Transitivity: If W01 ∶ M0 → M1 and W12 ∶ M1 → M2 are bordisms, then by gluing along M1

Proposition 1.3.6 If fi ∶ K → M , i = 0, 1, are both transverse to N and if f0 ∼ f1 , then

2 Orientation and (Co)homology

Theorem 2.1.3 Poincare − Lefschetz − Alexander Duality Let M n be an oriented topological

Theorem 2.1.6 Let W be a bordism with boundary ∂W = ∂0 W ⊔ ∂1 W . Consider the homology

Proof: See [7], Lemma VI.9.1.

Corollary 2.1.7 Let W be a bordism with boundary ∂W = ∂0 W ⊔ ∂1 W and x ∈ H n (W ).

Proof: Let j = j0 ⊔ j1 , then

2.2 Homological Interpretation of the Mapping Degree

∃! d ∈ Z ∶ f∗ [M ] = d[N ], and deg(f ) = d.

is a covering, i.e. f −1 (U ) = ⊔ri=1 Vi and f ∣Vi → U is a diffeomorphism. Let further ϕ ∶ U → Rn be a

where in the left column Hn (M ) ∋ [M ] z→ (ε1 µn , . . . , εr µn ) ∈ ⊕ri=1 Hn (Rn , Rn − 0) with

2.3 The Leray - Hirsch theorem

• H ∗ (E, E0 ) ∶= H ∗ (E, E0 ; R) is a H ∗ (X)-module via x ⋅ y ∶= π ∗ (x) ∪ y, where

• Let F0x ∶= F x ∩ E0 and jx ∶ (F x , F0x ) ↪ (E, E0 )

Assumption 2.3.2 There is a set B ⊆ H ∗ (E, E0 ) of homogeneous elements, such that:

• For all x ∈ X the set {jx∗ b ∣ b ∈ B} is a R-basis of H ∗ (F x , F0x ).

is an isomorphism of H ∗ (X)-modules (but not a ring isomorphism!).

a) Let f ∶ Y → X be a continuous function. Consider

and we set f ∗ B ∶= {f˜∗ b ∣ b ∈ B} ⊆ H ∗ (f ∗ E, f ∗ E0 ; R), then (f ∗ E, F ∗ E0 ) → Y and f ∗ B

2.) Let now X be a zero dimensional CW complex, then

H ∗ (X; R) ⊗R RB ≅ ( ∏ R) ⊗R RB, H ∗ (E, E0 ; R) ≅ ∏ H ∗ (F x , F0x ; R),

Φ becomes a natural transformation of functors Ln → K n from spaces over X to R-modules.

Ln−1 (A) ⊕ Ln−1 (B) / Ln−1 (A ∩ B) ∂ / Ln (A ∪ B) / Ln (A) ⊕ Ln (B) / Ln (A ∩ B)

π ∗ (∂x) ∪ b = ∂(π ∗ x ∪ b).

This holds, since considering open subsets U1 , U2 ⊆ Y such that U1 ∪ U2 = Y , we have a

0 / C ∗ (Y ) / C ∗ (U1 ) ⊕ C ∗ (U2 ) / C ∗ (U1 ∩ U2 ) /0

1. for each fixed r, , the map (t, x) ↦ σtr, (x) is an isotopy.

3. σtr, ∣Br (0) = id.

4. σ1r, (Rn ) = Br+ (0).