HOMOLOGY, COHOMOLOGY, AND THE DE RHAM THEOREM

CHAD BERKICH

Abstract. This paper is devoted to several commonly used homology and co-
homology theories, as well as an important result which links them together—
the de Rham theorem. We introduce singular homology, singular cohomology
as well as de Rham cohomology in the first few sections. Then we state and
prove the de Rham theorem. Finally, an application of the de Rham theorem
will be provided.

Contents
1. Introduction 1
2. Singular Homology 2
3. Exterior Differentiation and De Rham Cohomology 4
4. Singular Cohomology and Smooth Singular Homology 8
5. The de Rham Theorem 9
6. An Application of the de Rham Theorem 14
Acknowledgments 14
References 15

1. Introduction
The primary idea behind (co)homology is trying to identify ”n-dimensional
holes” in a topological space. Common examples are the hole in the center of
a torus and the inside of a hollow sphere. Holes are interesting to measure because
the mechanisms used to measure them, their homology and cohomology groups, are
algebraic topological invariants.
There are two different ways to do this: singular homology and de Rham co-
homology. Singular homology groups are the quotients of the groups of ”closed
singular chains” and ”boundary singular chains.” Closed chains are chains sent to
zero by the boundary operator. Boundary chains are singular chains which are
obtained by applying the boundary operator to other chains, and are a subgroup
of closed chains. De Rham cohomology groups are the quotient of the groups of
”closed forms” and ”exact forms.” Closed forms are differential forms whose exte-
rior derivative is zero. Exact forms are differential forms which are obtained by
applying the exterior derivative to other forms and are a subgroup of closed forms.
These concepts will be introduced in section 2 and section 3, respectively.
While the methods are different, they have a similar logic behind them. The
de Rham Theorem proves an isomorphism between singular cohomology groups,

Date: AUGUST 28, 2021.

1
2 CHAD BERKICH

the dual of the singular homology groups, and the de Rham cohomology groups.
Section 4 and section 5 are devoted to detailed explanations of them.
Finally, in section 6 an application of the de Rham theorem is given.

2. Singular Homology
This first section will be concerned with developing the theory of singular ho-
mology. A standard introduction to this topic is Chapter 2 of [Hat02]. One can
also find a brief summary at the beginning of Chapter 18 of [Lee12].
To begin, we define the k-simplex.
Definition 2.1. A (geometric) k-simplex is a subset of Rn of the form
Xk k
X
{ ti vi |0 ≤ ti ≤ 1, ti = 1},
i=0 i=0
where {v0 , v1 , · · · , vk } are k+1 affinely independent points in Rn . We use [v0 , v1 , · · · ,
vk ] to denote the k-simplex with vertices v0 , v1 , · · · , vk .
The dimension of the simplex is defined as the integer k, or one less than the
amount of vertices. This is a generalization of a triangle or tetrahedron to k di-
mensions.
∆k := [0, e1 , · · · , ek ] ⊂ Rk is called the standard k-simplex, where ei is the i-th
standard basis vector of Rk .
Definition 2.2. A simplex whose vertices are a non-empty subset of the vertices
{v0 , v1 , · · · , vk } is called a face of the simplex. Subsets with m + 1 vertices are
m-simplices, ∆m . (k − 1)-dimensional faces are the boundary faces.
Having defined the k-simplex, we will now turn our attention to how it can
be used in relation to topological spaces, specifically through the creation of the
singular k-simplex.
Definition 2.3. For a topological space X, a singular k-simplex in X is a contin-
uous map σ : ∆k → X.
Definition 2.4. The free abelian group generated by all singular k-simplices in X,
denoted as Ck (X), is called the singular chain group of X in degree k. An element
in
P Ck (X) is called a singular k-chain. It is is a finite formal linear combination
i ai σi , where ai ∈ Z.

Having defined both the singular k-simplex and the singular chain group, we
now turn our attention to the boundary operator, which decomposes k-simplices
into a sum of its (k − 1)-faces.
Definition 2.5. The boundary operator ∂k : Ck → Ck−1 is defined as
k
X
∂k (σ) = (−1)i σ|[v0 ,v1 ,··· ,vi−1 ,vi+1 ,··· ,vk ]
i=0
where σ|[v0 ,v1 ,··· ,vi−1 ,vi+1 ,··· ,vk ] is the restriction of σ to the boundary face [v0 , v1 , · · · ,
vi−1 , vi+1 , · · · , vk ] ({vi } are the vertices of the simplex). For any ∂m such that
m > k or m ≤ 0 is defined as the zero map.
Proposition 2.6. For any k-chain, the boundary of its boundary is trivial, or
∂k−1 (∂k (σ)) = 0.
 HOMOLOGY, COHOMOLOGY, AND THE DE RHAM THEOREM 3

Proof. As previously stated,

k
X
∂k (σ) = (−1)i σ|[v0 ,v1 ,··· ,vi−1 ,vi+1 ,··· ,vk ]
i=0
Now consider ∂k−1 (∂k (σ)). This gives

X k Xi−1
∂k−1 (∂k (σ)) =  (−1)j (−1)i σ|[v0 ,v1 ,··· ,vj−1 ,vj+1 ,··· ,vi−1 ,vi+1 ,··· ,vk ]
i=0 j=0

k
X
+ (−1)j−1 (−1)i σ|[v0 ,v1 ,··· ,vi−1 ,vi+1 ,··· ,vj−1 ,vj+1 ,··· ,vk ] 
j=i+1

Since the i-th term was already removed by the first boundary operator, it cannot
be removed again. This means the next term to be removed by the second boundary
operator is the (i + 1)-th term. The sign is (−1)j−1 as the next sign would have
been (−1)i . The sums cancel as they have opposite signs of (−1)i+j and (−1)i+j−1
and they cover the same terms. For example, for the first sum when i = m and
j = n this gives
(−1)m+n σ|[v0 ,v1 ,··· ,vn−1 ,vn+1 ,··· ,vm−1 ,vm+1 ,··· ,vk ] .
Then, for the second sum, when i = n and j = m, this gives
(−1)m+n−1 σ|[v0 ,v1 ,··· ,vn−1 ,vn+1 ,··· ,vm−1 ,vm+1 ,··· ,vk ]
Thus, ∂k−1 (∂k (σ)) = 0. 
Corollary 2.7. Im(∂k+1 ) ⊆ Ker(∂k )
Proof. Suppose σ is a k-chain in Im(∂k+1 ). Then σ = ∂k+1 (ν) where ν is a (k + 1)-
chain. By Proposition 2.6,
∂k (σ) = ∂k (∂k+1 (ν)) = 0
Thus, σ ∈ Ker(∂k ) and Im(∂k+1 ) ⊆ Ker(∂k ). 
Definition 2.8. The singular homology group is defined as Hk (X) = Ker(∂k )/Im(∂k+1 ).
We will now prove some important properties of the singular homology group.
Proposition 2.9. For a collection of path-connected topological spaces {Xj }, where
X = j Xj , an isomorphism Hk (X) ∼
` L
= j Hk (X j ) is induced by the inclusion
maps ιj : Xj → X.
Proof. Because they are continuous, singular simplices have
Lpath-connected images,
so the inclusion maps induce an isomorphism Ck (X) ∼ = j Ck (Xj ). Because the
boundary operator is a homomorphism, this relationL is preserved. Thus, the same
will apply to the homology groups, and Hk (X) ∼ = j Hk (Xj ). 
Before we prove the next property, an important notion must be defined.
Definition 2.10. A k-chain is homologous to 0 if it is the boundary of an k + 1-
chain, or σ = ∂ν for σ ∈ Ck (X) and ν ∈ Ck+1 (X). This is denoted as σ ∼ 0. If
σ −  ∼ 0, where σ and  are k-chains, then σ ∼  and they are homologous.
Proposition 2.11. For a path-connected, non-empty topological space X, H0 (X) ∼
=
Z.
4 CHAD BERKICH

Proof. H0 (X) = Ker(∂0 )/Im(∂1 ) = C0 /Im(∂1 ) byP definition asP∂0 is the zero map.
Let I : C0 → Z be a homomorphism such that I( i ai σi ) = i ai . I is called the
index of a chain. Since X is non-empty, this is surjective. If Ker(I) = Im(∂1 ) when
X is path-connected, then P I induces an isomorphism H0 (X) P → Z. P
Let σ ∈ C1 (X) and σ = i ai σi . ByPdefinition,
P ∂(σ) = i ai σi |[v1 ] − i ai σi |[v0 ] .
Applying the index gives I(∂(σ)) = i ai − i ai = 0. Thus, Im(∂1 ) ⊂ Ker(I)
and if σ ∼ 0, then I(σ) = 0. Next, let x and y be points in X. Since X is path-
connected, there exists a path between them of singular 1-simplices σi = (pi , pi + 1)
k−1
X
where 0 ≤ i ≤ k − 1, p0 = x, and pk = y. Let σ = aσi . Applying the boundary
i=0
operator gives
k−1
X k−1
X
∂(σ) = a∂(σi ) = a(pi+1 − pi ) = a(y − x)
i=0 i=0

Thus, I(∂(σ)) = 0. Because ∂(σ) is a boundary, there exists a 0-chain ν such that
ν = ∂(σ) ∼ 0. This implies that a(y − x) ∼ 0 and thus ay ∼ ax. Thus, any 0-chain
in X is homologous to ax. Since ν ∼ 0 means that I(ν) = 0, homologous chains
have equal indices. ν ∼ ax implies that I(ν) = I(ax) = a and ν ∼ I(ν)x. So, if
I(ν) = 0, then ν ∼ 0. Thus, Ker(I) ⊂ Im(∂1 ). 

Corollary 2.12. H0 (X) = ∼ L

j Z where j is the number of path-connected compo-
nents of X.
Proof. This follows from Propositions 2.9 and 2.11. 

Proposition 2.13. For a one-point space {p}, H0 ({p}) = Z and Hk ({p}) = 0 for
k 6= 0.
Proof. TherePexists a unique singular k-simplex σk : ∆k → {p} for every k such
that ∂σk = i (−1)i σk−1 . Since this is a sum of k + 1 terms, it is 0 when k + 1
is even and σk−1 when k + 1 is odd and k 6= 0. This results in the following chain
complex of alternating isomorphisms and trivial maps:
∼ ∼ ∼
···
0 /Z = /Z 0 /Z = /Z 0 /Z = / ···

Since Ker(∂k ) = Im(∂k+1 ), this gives Hk ({p}) = 0 for k 6= 0. Finally, since the
one-point space only has one path-connected component, H0 ({p}) = Z follows from
Proposition 2.11. 

3. Exterior Differentiation and De Rham Cohomology

The main references for this section are Chapter 7 of [Tu11] and Chapter 17
of [Lee12]. Some familiarity with smooth manifolds and differential forms will be
assumed. Readers are recommended to look up basic notions and conclusions in
Chapter 5 of [Tu11] and Chapter 14 of [Lee12].
Before defining it on manifolds, we will define the exterior derivative on Euclidean
spaces.
Definition 3.1. Consider a smooth differential k-form ω = I ωI dxI on an open
P
subset U ⊆ Rn . dxI refers to dxi1 ∧ dxi2 ∧ · · · ∧ dxim for a multi-index I =
 HOMOLOGY, COHOMOLOGY, AND THE DE RHAM THEOREM 5

(i1 , i2 , · · · , im ). The exterior derivative of ω is defined as

X X X ∂ωI
dω = dωI ∧ dxI = j
dxj ∧ dxI ,
j
∂x
I I

where j iterates {1, 2, · · · , n}.

The exterior derivative is linear over R and is a (k + 1)-form.
Definition 3.2. A differential form ω is closed if dω = 0.
Definition 3.3. A differential form ω is exact if there exists a (k − 1)-form η such
that ω = dη.
Now, some important properties of the exterior derivative will be proven.
Proposition 3.4. Consider a smooth k-form ω and smooth l-form η on an open
subset of U ⊆ Rn . The following property holds:
d(ω ∧ η) = dω ∧ η + (−1)k ω ∧ dη
This is the product rule for the exterior derivative.
Proof. Let ω = I fI dxI and η = J gJ dxJ . Applying the exterior derivative on
P P
their wedge product gives
n  
XX ∂fI ∂gJ
d(ω ∧ η) = m J
g + f dxm ∧ dxI ∧ dxJ
m I
m=1
∂x ∂x
I,J

To
 more easily manipulate
 the sums, they will be split into two different sums over
∂fI ∂gJ ∂fI
g
∂xm J + f
∂xm I . Consider the first sum with ∂x m gJ . By linearity, it can be
reordered into this form:
n
XX ∂fI
m
dxm ∧ dxI ∧ gJ dxJ = dω ∧ η
m=1
∂x
I,J

Now consider the second sum. By the anti-commutativity of the wedge product,
dxm ∧ dxI = (−1)k (dxI ∧ dxm ), and by linearity, it can be reordered into this form:
n
XX ∂gJ m
(−1)k fI dxI ∧ dx ∧ dxJ = (−1)k ω ∧ dη
∂xm
I,J m=1

Thus, adding the two sums together gives d(ω ∧ η) = dω ∧ η + (−1)k ω ∧ dη. 

Proposition 3.5. Every exact form is closed. In another word, d2 = 0.

Proof. Consider a differential k-form ω = I fI dxI . Then, applying the exterior
P
derivative gives
n
XX ∂fI l
dω = dx ∧ dxI
∂xl
I l=1
Applying the exterior derivative again gives
n Xn
X X ∂ 2 fI
d2 ω = ( m dxl
dxm ∧ dxl ∧ dxI ).
m=1
dx
I l=1
6 CHAD BERKICH

Consider the sum in the parenthesis on the right hand side for a fixed I. When
m = l, these terms just go to 0 since ∂m ∧ ∂m = 0. Thus, the sum can split into
two sections:
n m−1
X ∂ 2 fI n n
X
m l I
X X ∂ 2 fI
m dxl
dx ∧ dx ∧ dx + m dxl
dxm ∧ dxl ∧ dxI
m=1
dx m=1
dx
l=1 l=m+1
Since l = m + 1 for the second sum, this means that l − 1 = m. Thus, the second
sum can be re-indexed as:
n Xl−1
X ∂ 2 fI
m dxl
dxm ∧ dxl ∧ dxI .
m=1
dx
l=1

Since dx ∧ dx = −(dx ∧ dxm ) this allows the second sum to be re-indexed by

m l l

switching l and m. This only requires that a negative be introduced to account for
the anti-commutativity of the wedge product. Thus, this gives
Xn m−1
X ∂ 2 fI
− m dxl
dxm ∧ dxl ∧ dxI
m=1
dx
l=1

Combining this with the other sum gives 0. Thus, d2 ω = 0. 

Moreover, from the definition of the exterior derivative one can easily verify the
following:
Proposition 3.6. For a vector field X on U and f ∈ C ∞ (U ), we have (df )(X) =
Xf .
In fact, the exterior derivative on U can be characterized uniquely by the three
previous properties. To be more concrete, we have the following proposition:
Proposition 3.7. Let D : Ω∗ (U ) → Ω∗ (U ) be a linear map satisfying:
(1) D(ω ∧ η) = Dω ∧ η + (−1)k ω ∧ Dη, where k is the degree of the form ω;
(2) D2 = 0;
(3) For a vector field X on U and f ∈ C ∞ (U ), we have (Df )(X) = Xf .
Then D ≡ d on U .
Proof. Since every k-form on U is a combination of terms like f dxi1 ∧ · · · ∧ dxik ,
it suffices to show that D coincides with d on such terms. By (iii), Df = df for
f ∈ C ∞ (U ). And Ddxi = DDxi = 0 by (ii). From (i) we know that for all k and
all multi-indices I of length k, we have
D(dxI ) = D(dxi1 ∧ · · · ∧ dxik ) = 0.
Finally, for f dxI = f dxi1 ∧ · · · ∧ dxik ,
D(f dxI ) = (Df ) ∧ dxi + f D(dxI ) = (df ) ∧ dxI = d(f dxI ).
Thus, D coincides with d on U . 
Now we can expand our horizon to arbitrary smooth manifolds instead of Eu-
clidean spaces. There is a generalization of the exterior derivative on smooth man-
ifolds:
Theorem 3.8. Let M be a smooth manifold, with or without boundary. Let Ωk (M )
denote the vector space of smooth k-forms for M . There exists a R-linear map
d : Ωk (M ) → Ωk+1 (M ) such that the following properties hold:
 HOMOLOGY, COHOMOLOGY, AND THE DE RHAM THEOREM 7

(1) d(ω ∧ η) = dω ∧ η + (−1)k ω ∧ dη

(2) d2 ω = 0
This is the exterior derivative on smooth manifolds.
Proof. Let ω be a smooth differential k-form on M and let p be a point in PM . Let
(U, x1 , x2 , · · · , xk ) be a smooth chart on M where p ∈ U . Suppose ω = I fI dxI
on U . Let dU be the exterior derivative on U . The exterior derivative of ω on U is
X
dU ω = dfI ∧ dxI
I
Let dω|p = dU ω|p . We need to show that this definition of dω|p is independent of
the choice
P of open an U . Thus, let (V, y 1 , y 2 , · · · , y k ) be a smooth chart on M and
J
let ω = J gJ dy on V . Assume U ∩ V is non-empty and let p ∈ U ∩ V . On U ∩ V ,
X X
ω= fI dxI = gJ dy J
I J
Additionally, from above we know that there is a unique exterior derivative dU ∩V :
Ω∗ (U ∩ V ) → Ω∗ (U ∩ V ) such that
X X
dU ∩V ( fI dxI ) = dU ∩V ( gJ dy J ),
I J
or X X
dfI ∧ dxI = dgJ ∧ dy J .
I J
This implies that they also agree when applied at p, so dω|p is well defined, inde-
pendently of the chart. The properties follow from the proofs of Propositions 3.4
and 3.5. 
One can also prove that the exterior derivative on manifolds is unique. A proof
can be seen in section 19.4 of [Tu11].
Definition 3.9. Let M be a smooth manifold with or without boundary, and let
k ∈ Z and k ≥ 0. The exterior derivative d : Ωk (M ) → Ωk+1 (M ) is R-linear, so its
subspaces are also R-linear. Thus, define
Z k (M ) = Ker(d : Ωk (M ) → Ωk+1 (M ))
or the closed smooth k-forms on M . Then, define
B k (M ) = Im(d : Ωk−1 (M ) → Ωk (M ))
or the exact smooth k-forms on M . If k > dim(M ) or k < 0, then Ωk is the zero
vector space.
Since each closed form is exact as d2 ω = 0 for any k-form, this implies that
B (M ) ⊆ Z k (M ). With this fact in mind, we now define the de Rham cohmology
k

groups.
Definition 3.10. The kth de Rham cohomology group of M is defined to be
k
HdR = Z k (M )/B k (M )
The use of the subscript dR is used to denote that this is the de Rham cohomology
group1.
1Despite the use of the word ”group” here, the de Rham cohomology group is actually a real
vector space.
8 CHAD BERKICH

We will now prove an important result about de Rham cohomology groups.

Proposition 3.11. For a countable collection of smooth k-manifolds {M
Qj } with or
k
`
without boundary, where M = j Mj , there exists an isomorphism from j HdR (Mj )
k
to HdR (M ) and which is induced by the inclusion maps ιj : Mj → M .
Proof. The pullback of the inclusion maps ι∗j : Ωk (M ) → Ωk (Mj ) induce an isomor-
phism from Ωk (M ) to j Ωk (Mj ), sending ω to (ι∗1 ω, ι∗2 ω, · · · ) = (ω|M1 , ω|M2 , · · · ).
Q
If a smooth k-form is zero on each Mj , then it is zero on M , so the map is injective.
The map is surjective because defining a k-form on each Mj defines a k-form on
M . By passing to quotients we obtain an isomorphism on de Rham cohomology
groups. This must be countable because smooth manifolds are second-countable.
Thus, in order for M to be smooth, it must be second countable, which is only
accomplished when Mj is countable. 

4. Singular Cohomology and Smooth Singular Homology

We will now define a new cohomology theory, called singular cohomology. A
standard introduction to this topic is Chapter 3 of [Hat02]. For our purposes, we
will follow the first few sections of Chapter 18 of [Lee12], which are only concerned
with coefficients on R.
Definition 4.1. Let G be an abelian group. Consider C k (X; G) = Hom(Ck (X); G),
which is called the group of singular k-cochains with coefficients in G. The differ-
ential map δk : C k (X; G) → C k+1 (X; G) is dual to the boundary operator ∂k on
chains. The spaces of singular cochains and δ form a cochain complex. The kth
singular cohomology group is defined as H k (X; G) = Ker(δk )/Im(δk+1 ).
When G = R, H k (X; G) is naturally isomorphic to Hom(Hk (X), R). So from
now we just take Hom(Hk (X), R) as the definition of H k (X; G).
Proposition 4.2. Properties of Singular Cohomology:
a. For a single point {q}, H k ({q}; R) is one dimensional when k = 0 and trivial
when k 6= 0. `
b. For a collection of topological spaces {Xj }, where X = j Xj , the inclusion map
ιj : Xj ,→ X induces an isomorphism H k (X; R) ∼ = j H k (Xj ; R)
Q

Proof.
a. By Proposition 2.13, H0 ({q}) ∼ = Z and Hk ({q}) = 0 for k 6= 0. Since a homo-
morphism from Hk ({q}) to R determines and is determined by the image of the
generator 1 ∈ Z under it, we have H 0 ({q}; R) ∼
= R and H k ({q}; R) = 0. L
b. By Proposition 2.9, the inclusions maps induce an isomorphism Hk (X) ∼= j Hk (Xj )
∼ k ∼ L
and thus Hom(Hk (X), R) = H (X; R) = Hom( j Hk (Xj ), R). By Proposition
B.572 in [Lee12], Hom( j Hk (Xj ), R) ∼= j Hom(Hk (Xj ), R) ∼ = j H k (Xj ; R).
L Q Q

Thus, H k (X; R) ∼= j H k (Xj ; R).

Q


We will now begin to define smooth singular homology, which will correspond
closely with our original definitions of singular homology.

2Part of Appendix B
 HOMOLOGY, COHOMOLOGY, AND THE DE RHAM THEOREM 9

Definition 4.3. For a smooth manifold M , a smooth k-simplex in M is defined to

be a smooth map σ : ∆k → M .
Definition 4.4. The smooth chain group in degree k is denoted as Ck∞ (M ) ⊆
Ck (M ). Elements in the group, which are called smooth chains, are finite formal
linear combinations of smooth k-simplices.
Now, we will use the fact that the boundary of a smooth simplex is a smooth
chain to define another homology group.
Definition 4.5. The kth smooth singular homology group of M is defined as
Ker(∂k : Ck∞ (M ) → Ck−1
∞
(M ))
Hk∞ (M ) = ∞ ∞
Im(∂k+1 : Ck+1 (M ) → Ck (M ))
where the boundary operator is the same as defined in Definition 2.5 as Ck∞ (M ) ⊆
Ck (M ).
Theorem 4.6. For a smooth manifold M , the map ι∗ : Hk∞ (M ) → Hk (M ), which
is induced by inclusion, is an isomorphism.
Remark 4.7. The proof of Theorem 4.6 is too long to put here. The readers can
find it in Page 478 of [Lee12] (proof of Theorem 18.7).
This conclusion is an important bridge in the proof of the de Rham theorem.
We will see the role it plays in the well-definedness of Definition 5.5.

5. The de Rham Theorem

Having defined all the necessary homology and cohomology groups, we will now
build up some necessary results to prove the de Rham theorem. The approach we
follow is due to Glen E. Bredon and can be found in the last section of Chapter 18
of [Lee12].
The proof of the de Rham Theorem partially relies on Mayer-Vietoris sequences.
While these diagrams will not be explicitly covered, they can also be found in
[Lee12]: Theorem 18.4 for singular homology, Theorem 17.20 for de Rham coho-
mology, and Theorem 18.6 for singular cohomology.

First, we introduce an important tool for proving isomorphisms of (co)homology

theories.
Proposition 5.1. The Five Lemma:
A
f1
/B f2
/C f3
/D f4
/E

h1 h2 h3 h4 h5
    
A0
g1
/ B 0 g2 / C 0 g3 / D 0 g4 / E 0
If the diagram is commutative, the horizontal rows are exact, and h1 , h2 , h4 , and h5
are isomorphisms, then h3 is also an isomorphism.
Proof. First we will prove surjectivity. Let c0 ∈ C 0 . h4 is surjective, so there exists
d ∈ D such that g3 (c0 ) = h4 (d). Since g4 is exact, g4 (g3 (c0 )) = 0. Because the
diagram is commutative, h5 (f4 (d)) = g4 (h4 (d)). Since g3 (c0 ) = h4 (d), h5 (f4 (d)) =
g4 (h4 (d)) = g4 (g3 (c0 )) = 0. Thus, f4 (d) = 0.
For that c, consider the equation g3 (c0 )−g3 (h3 (c)). Since the diagram is commu-
tative, g3 (h3 (c)) = h4 (f3 (c)) = h4 (d) and thus g3 (c0 ) − g3 (h3 (c)) = g3 (c0 ) − h4 (d) =
10 CHAD BERKICH

0. This implies that g3 (c0 − h3 (c)) = 0. Because the rows are exact, there exists
b0 ∈ B 0 such that c0 −h3 (c) = g2 (b0 ). Because h2 is surjective, there exists b ∈ B such
that b0 = h2 (b). Thus, h3 (c +f2 (b)) which is equal to h3 (c) + h3 (f2 (b)) because h3 is
a homomorphism. Then h3 (c) + h3 (f2 (b)) = h3 (c) + g2 (h2 (b)) = h3 (c) + g2 (b0 ) = c0
so h3 is surjective.
Now we will prove injectivity. Suppose h3 (c) = 0. Then g3 (h3 (c)) = 0. Since
the diagram is commutative, g3 (h3 (c)) = h4 (f3 (c)) = 0. Since h4 is injective, this
means f3 (c) = 0. Since the rows are exact, there exists b ∈ B such that c = f2 (b)
and h3 (f2 (b)) = g2 (h2 (b)) = 0. Thus, there exists a0 ∈ A0 such that h2 (b) = g1 (a0 ).
Since h1 is surjective, there exists a ∈ A such that a0 = h1 (a). Since h2 is injective,
h2 (f1 (a) − b) = h2 (f1 (a)) − h2 (b) = g1 (h1 (a)) − h2 (b) = g1 (a0 ) − h2 (b) = 0. Thus,
f1 (a) − b = 0 and c = f2 (b) = f2 (f1 (a)) = 0. Since the kernel of h3 is trivial, h3 is
injective. 
Now, we will introduce a variant of Stokes’ Theorem for differential forms, Stokes’
Theorem for Chains. Recall that the original Stokes’ Theorem is
Theorem 5.2. Let M be an oriented smooth manifold of dimension n, with bound-
ary ∂M . Let ω be a compactly supported smooth (n − 1)-form on M . Then we have
Z Z
dω = ω.
M ∂M

To give the variant, we need to define the following operation:

Definition 5.3. For a smooth manifold M , a smooth differential k-form ω, and a
smooth k-simplex σ, the integral of ω over σ is defined as
Z Z
ω= σ∗ ω
σ ∆k
k
X
Additionally, for a smooth k-chain c = ci σi , the integral of ω over c is defined
i=1
to be
Z k Z
X
ω= ω
c i=1 σi

Theorem 5.4 (Stokes’ Theorem for Chains). For a smooth k-chain c in a smooth
manifold M and a smooth (k − 1)-form ω on M,
Z Z
ω = dω
∂c c

The readers can find a proof of Theorem 5.4 on page 481 of [Lee12].
Finally, we will develop the core construction in Bredon’s proof of the de Rham
theorem, which bridges the de Rham cohomology and singular homology explicitly.
k
Definition 5.5. A linear map ` : HdR (M ) → H k (M ; R) is defined such that for
∼
any [ω] ∈ HdR (M ) and [c] ∈ Hk (M ) = Hk∞ (M ),
k
Z
`[w][c] = ω
c̃
where c̃ is a smooth k-cycle that represents the homology class [c]. This is called
the de Rham homomorphism.
 HOMOLOGY, COHOMOLOGY, AND THE DE RHAM THEOREM 11

Proposition 5.6. The de Rham homomorphism is well-defined.

Proof. For two smooth k-cycles that represent the same homology class, c̃ and c̃0 ,
by Theorem 4.6 there is a smooth (k + 1)-chain b̃ such that c̃ − c̃0 = ∂ b̃. Thus, from
Theorem 5.4 we know that
Z Z Z Z
ω− ω= ω = dω = 0
c̃ c̃0 ∂ b̃ b̃
If ω is exact, then Z Z Z
ω= dη = η=0
c̃ c̃ ∂c̃
∂c̃ = 0 and dω = 0 because they represent a homology and cohomology class,
respectively. 
Proposition 5.7 (Naturality of the de Rham homomorphism). Let M, N be a
smooth manifolds and k be a non-negative integer.
a. For a smooth map F : M → N , the diagram commutes:
p F∗ / H p (M )
HdR (N ) dR

` `
 
H p (N ; R) / H p (M ; R)
F∗
b. For open subsets U, V ⊂ M where U ∪ V = M , the diagram commutes:
p−1
HdR (U ∩ V )
δ / H p (M )
dR

` `
 
H p−1 (U ∩ V ; R) / H p (M ; R)
∂∗
where δ is the connecting homomorphism of the Mayer–Vietoris sequences for
de Rham cohomology while ∂ ∗ is it for singular cohomology.
Proof. a. By definition, for a smooth k-simplex in M , σ, and a smooth k-form on
N , ω, the following holds:
Z Z Z Z
F ∗ω = σ∗ F ∗ ω = (F ◦ σ)∗ ω = ω
σ ∆k ∆k F ◦σ
which means that
`(F ∗ [ω])[σ] = `([ω])[F ◦ σ] = `([ω])(F∗ [σ]) = F ∗ (`[ω])[σ]
and thus the diagram commutes.
b. Since the previous diagram commutes, `(δ[w])[e] = (δ ∗ `[w])[e] for any
p−1
[w] ∈ HdR (U ∩ V ) and for any [e] ∈ Hp (M ). Since
Hom(Hp (M ), R) ∼ = H p (M ; R)
this is equal to
`(δ[w])[e] = `([w])(∂∗ [e])
For a smooth k-form representing
R Rδ[ω], σ, and a smooth (k−1)-chain representing
∂∗ [e], c, this equates to e σ = c ω. By the characterization property of the
connecting homomorphism ∂∗ , we can let c = ∂f , where f is a smooth k-chain
in U and f 0 in V , so that f + f 0 and e represent the same homology class.
12 CHAD BERKICH

Similarly, by the characterization property of the connecting homomorphism δ,

we can pick η ∈ Ωp−1 (U ) and η 0 ∈ Ωp−1 (V ) so ω = η|U ∩V − η 0 |U ∩V . Then let
σ be the k-form that is equal to dη and dη 0 on U and V , respectively. Because
∂f + ∂f 0 = ∂e = 0 and dη|U ∩V − dη 0 |U ∩V = dω = 0, the following occurs
Z Z Z Z Z Z
0
ω= ω= η− η = η+ η0
c ∂f ∂f ∂f ∂f ∂f 0
Z Z Z Z Z
= dη + dη 0 = σ+ σ= σ
f f0 f f0 e
such that the diagram commutes.


Eventually, the destination is right in front of us:

Theorem 5.8 (The de Rham Theorem). For any smooth manifold M and non-
k
negative integer k, the de Rham homomorphism ` : HdR (M ) → H k (M ; R) is an
isomorphism.
Proof. We call a smooth manifold M a de Rham manifold if the de Rham homo-
k
morphism ` : HdR (M ) → H k (M ; R) is an isomorphism for any non-negative integer
k. Additionally, we call an open cover {Ui } of a smooth manifold M a de Rham
n
\
cover if each Ui is a de Rham manifold and if a finite intersection Ui is de Rham.
i=1
Finally, a de Rham basis is a de Rham cover that is also a basis for the topology of
M.
By Proposition 5.7, the de Rham homomorphism commutes with cohomology
groups induced by smooth maps, so if manifold is diffeomorphic to a de Rham
manifold, then it is also de Rham. The goal of the proof is to show that every
smooth manifold is de Rham. This is achieved via the following six steps:
`
Step 1 If {Mj } is a countable collection of de Rham manifolds, then j Mj is de
Rham.
By Propositions 3.11 and 4.2, isomorphisms are induced between the
direct product of the`cohomology groups of the manifolds Mj`and the co-
homology groups of j Mj by the inclusion maps ιj : Mj ,→ j Mj . As a
result of Prop `
5.7, these isomorphisms commute with the de Rham homo-
morphism for j Mj . So the latter is itself an isomorphism.
Step 2 Any convex, open subset U ⊂ Rn is de Rham.
k
By the Poincaré Lemma of differential forms, if k 6= 0, then HdR (U ) = 0.
By Proposition 4.2, the singular cohomology groups of U are also trivial
when k 6= 0 because of the homotopy equivalence of U and a one-point
0
space. If k = 0, then HdR (U ) is the space consisting of constant functions
and is one-dimensional. Because H0 (U ) is generated by a singular 0-simplex,
H 0 (U ; R) = Hom(H0 (U ); R) is one dimensional. Let f be the constant
function 1 and σ : ∆0 → M be a singular, smooth 0-simplex. Then,
Z
`[f ][σ] = σ ∗ f = (f ◦ σ(0)) = 1
∆0
Since the de Rham homomorphism is not the zero map for k=0, it is an
isomorphism.
 HOMOLOGY, COHOMOLOGY, AND THE DE RHAM THEOREM 13

Step 3 If M has a finite de Rham cover, then M is de Rham.

[p
Let M = Ui where {Ui } is a finite de Rham cover. The proof will be
i=1
by induction.
• The p = 1 case is trivial.
• Next, let M have a de Rham cover {U, V }, which consists of two sets.
Using the Mayer–Vietoris sequence for de Rham and singular cohomol-
ogy together, the following diagram is obtained:

k−1
HdR k−1
(U ) ⊕ HdR (V ) / H k−1 (U ∩ V ) / H k−1 (U ∩ V ) /
dR dR

a b c
  
H k−1 (U ; R) ⊕ H k−1 (V ; R) / H k−1 (U ∩ V ; R) / H k−1 (M ; R) /

k
HdR k
(U ) ⊕ HdR (V ) / H k (U ∩ V )
dR

d e
 
H k (U ; R) ⊕ H k−1 (V ; R) / H k (U ∩ V ; R)

(the lower part continues from where the upper part lefts off). In
the diagram, the rows are exact sequences, while the columns are de
Rham homomorphisms. The diagram is commutative by Proposition
5.7. From the Five Lemma, we know that if the columns a, b, d, and e
are isomorphisms, then so is c. Thus, M is de Rham.
• Assume the claim is true for all smooth manifolds having a de Rham
cover with p open sets (p ≥ 2). Consider M with a de Rham cover
p
[
{Ui }i=1,2,··· ,p+1 . Let U = Ui and V = Up+1 . By the induction
i=1
hypothesis, both U and V are de Rham, and so is U ∩ V because of its
[p
de Rham cover Ui ∩ Up+1 . Thus, by the previous argument, M is
i=1
de Rham by induction.
Step 4 If M has a de Rham basis, then M is de Rham.
Let {Uα } be a de Rham basis for a smooth manifold M . Let f : M → R
be an exhaustion function3. For each integer m, let
Am = {q ∈ M : m ≤ f (q) ≤ m + 1} ⊂ M
A0m = {q ∈ M : m − 1/2 < f (q) < m + 3/2} ⊂ M
For every q ∈ Am , there is a basis open subset containing q which is con-
tained in A0m . The collection of these sets is an open cover of Am . Because
f is an exhaustion function, Am is compact and is covered by finitely many
basis sets. Let Bm be the union of these basis sets. The union of the sets
in Bm is a de Rham cover of Bm , so Bm is de Rham by the previous step.
3An exhaustion function for a topological space M is a continuous function f : M → R with
the property that the set f −1 ((−∞, c]) is compact for each c ∈ R. One can prove that every
smooth manifold has a smooth positive exhaustion function (see Proposition 2.28 in [Lee12]).
14 CHAD BERKICH

0
m ⊂ Am , Bm ∩ Bl 6= ∅[
Because B[ if l = m − 1, m, or m + 1. Thus, let
U = Bm and V = Bm . Thus, U and V are disjoint
m=2a+1,a∈Z m=2a,a∈Z

a so they are de Rham by the first step. U ∩ V

unions of de Rham manifolds,
is de Rham because it is Bm ∩ Bm+1 and each Bm and Bm+1 has a
m∈Z
finite de Rham cover. Thus, by Step 3, M , which is equal to U ∪ V , is de
Rham.
Step 5 An open subset of Rn is de Rham.
Let U ⊆ Rn be an open subset. Thus, U has a basis of open balls. Each
ball is convex and therefore is de Rham. A finite intersection of balls is
convex, so they are also de Rham. Thus, U has a de Rham basis, and is de
Rham.
Step 6 Every smooth manifold is de Rham.
A smooth manifold has a basis of smooth coordinate domains, which are
diffeomorphic to an open subset of Rn , along with their finite intersections.
These form a de Rham basis, and thus, by step 4, every smooth manifold is
de Rham.


6. An Application of the de Rham Theorem

Since the de Rham theorem gives an isomorphism between singular and de Rham
cohomology, we can obtain some information about differential forms on manifolds
merely from topological data. For example, consider the 2-torus, T 2 : we have
(
k 2 R for k = 0, 2
HdR (T ) = .
R2 for k = 1
k
Since ` : HdR (T 2 ) → H k (T 2 ; R) is an isomorphism by the de Rham theorem, and
none of the de Rham cohomology groups are 0, closed forms that are not exact
exist on T 2 .
In contrast, consider the n-sphere S n : we have
(
k n R for k = 0, n
HdR (S ) = .
0 for 0 < k < n
k
Since ` : HdR (S n ) → H k (S n ; R) is an isomorphism by the de Rham theorem, and
k n
HdR (S ) = 0, all closed k-forms are exact when 0 < k < n.

Acknowledgments
It is a pleasure to thank my mentor, Yao Luo, for his guidance and insight
throughout the program. I would also like to thank Daniil Rudenko for his orches-
tration of the Apprentice Program. Finally, thank you to Peter May and all of those
involved in running this incredible program.
 REFERENCES 15

References
[Hat02] Allen Hatcher. Algebraic topology. Cambridge University Press, Cam-
bridge, 2002.
[Nad07] Prerna Nadathur. “An Introduction to Homology”. In: The University of
Chicago REU (2007).
[Tu11] Loring W. Tu. An Introduction to Manifolds. Springer., 2011.
[Lee12] John M. Lee. Introduction to Smooth Manifolds. Springer, 2012.
[Bur18] Alexander Burka. “Elementary Homology Thoery With Computations”.
In: The University of Chicago REU (2018).
[Cha18] Nick Chaiyachakorn. “De Rham’s Theorem, Twice”. In: The University
of Chicago REU (2018).