Cech and sheaf cohomology

Greg Arone

November 11, 2021

An example of Cech cohomology
As usual, let I = [0, 1]. Let us calculate Ȟ∗ (I , A). Consider the
cover U = {[0, 43 ), ( 12 , 1]}. The groups Ȟ∗U (I ; A) are calculated by
the following complex:
n 0 1 2 ···
[1,−1]
CUn A ⊕ A −−−→ A → 0 · · ·
ȞnU A 0 0 ···
Now let U = {[0, b1 ), (a2 , b2 ), . . . , (an , 1]}, where for each k,
ak < bk−1 < ak+1 . Then the Cech complex has the following form
n 0 1 2 ···
CUn n
A −→ An−1 → 0 ···
ȞnU A 0 0 ···
Every open cover of I has a refinement of this type. From here one
can easily reach the (unsurprising) conclusion:

A n=0
Ȟn (I ; A) =
0 n>0
Homotopy invariance
“Similarly” to our calculation of Ȟn (I ; A), one can prove the
following
Lemma
Let X be a paracompact space. The canonical inclusions
i0 , i1 : X ,→ X × I induce isomorphisms (which are necessarily the
same)
∼
=
i0∗ , i1∗ : Ȟ∗ (X × I ; A) −
→ Ȟ∗ (X ; A).

As a consequence we have the following important theorem

Theorem (Homotopy invariance)
Let f , g : X → Y be homotopic maps between paracompact
spaces. Then they induce the same homomorphism

Ȟ∗ (Y ; A) → Ȟ∗ (X ; A).

The Mayer-Vietoris sequence
Suppose X = U1 ∪ U2 , where Ui are open subsets. There is a LES
(i 0 ,i 0 ) (j 0 −j 0 )
0 → Ȟ0 (X ; A) −−1−−
2
→ Ȟ0 (U1 ; A) ⊕ Ȟ0 (U2 ; A) −1−−−
− 2
→ Ȟ0 (U1 ∩ U2 ; A) →
(i11 ,i21 ) (j11 −j21 )
→ Ȟ1 (X ; A) −−−−→ Ȟ1 (U 1 ; A) ⊕ Ȟ1 (U 2 ; A) −
−−−−
→ Ȟ1 (U 1 ∩ U2 ; A) ···

Proof: A denotes an abelian group, and also the sheaf represented

by A: O 7→ AO . Let A1 = i1∗ i1! A be the sheaf O 7→ AO∩U . Def.
A2 , A12 similarly. We get a sort of exact sequence of presheaves:
(i1 ,i2 ) j1 −j2
0 → A −−−→ A1 ⊕ A2 −−−→ A12 (99K 0).
This is a left exact sequence of presheaves and it is short exact
when evaluated on a set O that is contained either in U1 or in U2 .
It follows that if O is an open cover of X , that is subordinate to
{U1 , U2 }, there is a short exact sequence of Cech complexes
(i1∗ ,i2∗ ) j1∗ −j2∗
∗ (X ;A)−
0→CO −−−→CO∗ (U1 ;A)⊕CO∗ (U2 ;A)−−−→CO∗ (U1 ∩U2 ;A)→0.
From here we get a LES of cohomology groups Ȟ∗O (−; A). The
poset of subordinate open covers is cofinal, so passing to colimits
we obtain a LES of cech cohomology groups
Examples

Using homotopy invariance, the Mayer-Vietoris sequence, and

induction, one can easily calculate the cohomology groups of a
sphere: 
n d A n = 0, or d
Ȟ (S ; A) =
0 otherwise

Remark
It is often convenient to consider reduced cohomology:
e ∗ (X ; A) := ker Ȟ∗ (X ; A) → Ȟ∗ (∗; A) .


H

Reduced cohomology is almost the same as unreduced cohomology,

except that it knocks of a factor of A in degree zero. For example
e ∗ (∗; A) = 0, and H
H e ∗ (S d ; A) only has a copy of A in degree d.
The cohomology of CW-complexes is often tractable. Let X be a
CW-complex. For each d, let X d be the d-skeleton of X . Thus X
is filtered by subspaces X d−1 ⊂ X d ⊂ · · · . Let Cn be the set of
n-dimensional cells of X .
Lemma
For each d, there is a long exact sequence

· · · → Ȟn (X d ; A) → Ȟn (X d−1 ; A) → H

e n (S d−1 ; A)Cd → Ȟn+1 (X d ; A) · · · .

In fact, one can organize things even better by splicing the exact
sequences. One obtains a chain complex of the following form

AC0 → AC2 → · · · → ACd → · · ·

This is called the cellular chain complex of X . Its cohomology is

isomorphic to the cech cohomology (or any other cohomology
satisfying the Eilenberg-Steenrod axioms).
Example: projective spaces
The complex projective space CP n is obtained from CP n−1 by
attaching a cell of dimension 2n. Thus CP n has a cell structure
with a single cell in each even dimension. It follows immediately
that the cohomology of CP n is given by the following formula:

i n A i = 2l ≤ 2n
Ȟ (CP ; A) =
0 otherwise

The real projective space requires a little more work. The space
RP n is obtained by attaching a cell of dimension n to RP n−1 .
Thus RP n has a cell structure with a single cell in each dimension.
It follows that the cellular chain complex of RP n has the following
form
0 1 2 ··· n
d0 d1 d2 d n−1
A −→ A −→ A −→ · · · −−−→ A

Lemma
d i is multiplication by 2 for odd i and is zero for even i.
products
The lemma about the cohomology of X × I is a special case of a
general result about the cohomology of products. Before we state
the result, let us review tensor products of complexes.
Definition
Let A• , B • be (co)chain complexes. We define their tensor product
(A ⊗ B)• as follows. In degree n,
M
(A ⊗ B)n = Ai ⊗ B j .
i+j=n

The differential is defined as follows. Let am ∈ Am , b n ∈ B n . Then

d(am ⊗ b n ) = dA (am ) ⊗ b n + (−1)m am ⊗ dB (b n ).

Similarly we define the tensor product of graded abelian groups,

such as H ∗ (X ) ⊗ H ∗ (Y ).
Algebraic Kunneth theorem

Theorem
Suppose A• , B • are chain complexes of free R-modules. There is a
natural homomorphism:

H ∗ (A) ⊗R H ∗ (B) → H ∗ (A ⊗ B).

This homomorphism is an isomorphism if A• is a chain complex of

free R-modules, and H n (A) is also a free R-module for every n.
Topology
Let X , Y be topological spaces. Let U, V be open covers of X and
Y respectively. Let U × V be the open cover of X × Y consisting
of all products of elements of U and V. Notice that there is a
homomorphism

CU0 (X ; A) ⊗ CV0 (Y ; B) → CU0 ×V (X × Y ; A ⊗ B)

The homomorphism sends a pair of functions

(f : U → A) ⊗ (g : V → A) to the function
f ⊗ g : U × V → A ⊗ B. It turns out, that this homomorphism can
be extended to a map of chain complexes
Theorem
There exists a natural homomorphism - in fact a chain homotopy
equivalence

CU∗ (X , A) ⊗ CV∗ (Y , B) → CU∗ ×V (X × Y ; A ⊗ B).

The Kunneth formula

The chain homomorphism induces a natural homomorphism

Ȟ∗ (X , A) ⊗ Ȟ∗ (Y , B) → Ȟ∗ (X × Y ; A ⊗ B).

In particular, if A = B = R is a ring, one gets a homomorphism

Ȟ∗ (X , R) ⊗R Ȟ∗ (Y , R) → Ȟ∗ (X × Y ; R).

Theorem
This homomorphism is an isomorphism if, for example, X is a CW
complex with finitely many cells in each dimension, and Ȟ∗ (X , R)
is a free R-module (or more generally, flat).
Internal ring structure on cohomology
Suppose X is a space and R is a commutative ring. Then we have
natural homomorphisms

Ȟ∗ (X , R) ⊗R Ȟ∗ (X , R) → Ȟ∗ (X × X ; R) → Ȟ∗ (X , R).

This endows Ȟ∗ (X , R) with the structure of a graded commutative

ring. For whatever historical reasons, the multiplication in
cohomology is denoted by the symbol ∪, and is called “cup
product”.
Example
Ȟ∗ (S m × S n , Z) ∼
= Z[um , un ]/(um2
, un2 )
Ȟ∗ (CP n , Z) ∼
= Z[u2 ]/(u n+1 ).
2
Ȟ∗ (RP n , Z/2) ∼
= Z[x]/(x n+1 ).

Applications: The Borsuk-Ulam theorem, non-existance of division

algebra structure on Rn for n 6= 2k .