Given two groups G and H containing a common subgroup K, one may
wonder whether it is possible to ``glue'' G and H together along
K. To make this precise, assume that there are two monomorphisms
f : K
G and
g : K
H. We want to construst a new group P
and maps
p : G
P and
q : H
P, such that whenever L is an
arbitrary group and
: G
L,
: H
L are
homomorphisms such that
f =
g, there exists a unique
homomorphism
: P
L such that
p
=
and
q
=
.
To begin with consider the simplest example of this situation, namely
where K = {1} and the maps f and g are the obvious ones. In that
case the requirements above are that whenever
and
are
arbitrary homomorphisms into L there is a unique map from the group
P into L, which factors both
and
. The group P
in that case is easy to construct. Namely, consider the group whose
elements are arbitrary words in alphabet from G and H and their
inverses. Multiplication is given by juxtaposition and the only
cancelation rules come from G and H. This group is called the free
product of G and H and is
denoted G*H. Before proceeding we discuss some examples.
The free product of G and H has the feature that if K is an
arbitrary group and there are homomorphisms
f : G
K and
g : H
K then there is a unique homomorphism from G*H to K which agrees
with f and g on the repective images of G and H in the free
product. In some sense the free product of two groups is in the world
of groups corresponds to the wedge operation in the world of pointed
spaces and to disjoint union for unpointed spaces. We have sketched a
proof in the tutorials of the fact that the fundamental group of a
wedge of two pointed spaces is the free product of the respective
fundamental groups. The Van Kampen theorem is just a generalisation of
this obsetvation.
Many times free amalgamated products are easy to write in terms of generators and relations if a presentation for the groups involved is known.
Notice that two isomorphic groups have isomorphic
abelianisations. However, two groups with the same abelianisation may
not be isomorphic. In particular the abelianisation of an abelian
group G is G itself because every element of the form
x-1y-1xy is trivial. Thus
(Gab)ab = Gab but
G
Gab unless G is abelian.
If G is given in terms of generators and relations then Gab is presented by the same generators and for realtions one takes the original relations together with extra relations x-1y-1xy = 1, one for every pair of generators x and y. The reader might like to use the idea of abelianisation to convince himself that the two groups obtained above as free amalgamated products of Z/4Z by Z/6Z over Z/2Z are not isomorphic. The abelianisation of the first is Z/12Z, whereas the second abelianises to Z/2Z x Z/6Z.
The free amalgamated product of G and H over K has the following
property. Let F be an arbitrary group and let
: G
F and
: H
F be homomorphisms such that
g =
h. Let
i : G
G*KH and
j : H
G*KH denote the inclusions. Then
there exists a unique homomorphism
: G*KH
F such that
i =
and
j =
. We are now
ready to state the Van Kampen Theorem.
Before we prove the theorem we explain its statement and examine a few
examples. The theorem states the following. Under the topological
assumptions on U and V, consider the homomorphisms
g :
(U
V, x0)
(U, x0) and
f :
(U
V, x0)
(V, x0).
With respect to those, one can construct the free
amalgamated produc of
(U, x0) and
(V, x0) over
(U
V, x0). Now, there are the maps
u# :
(U, x0)
(X, x0) and
v# :
(V, x0)
(X, x0) and by the mapping property of the free amalgamated
product there exists a unique homomorphism from
:
(U, xo)*
(U
V, x0)
(V, x0) to
(X, x0) such that
i = u# and
j = v#. The
theorem states that
is an isomorphism.
Let us now consider a few examples.
The next family of examples come from connected sums of two closed surfaces. Recall that the connected sum construction is formed by removing an open disk from the respective surfaces and glueing them along the boundaries. Thus the connected sum of two surfaces can be though of as a union of two punctured surfaces intersecting on a circle. By thickening the circle of intersection a bit one can guarantee that the conditions of the theorem are satisfied. The next step is to compute the homomorphism induced by inclusion of the circle as the boundary of the respective punctured surface.
The first and easiest is the case where the surface is a 2-sphere. In that case the punctured surface is contractible and the induced map
Next consider the torus T2. We saw how the torus can be obtained from a square by identifying each pair of parallel edges ini the same orientation.
Next we present a calculation of the fundamental group of the torus using the Van Kampen theorem.
The Klein bottle K2 can be obtained from a square by identifying one pair of parallel edges in the same orientation and the other in opposite orientation.
Similar to the calculation of
(T2) we now present a
calculation of
(K2) using the Van Kampen theorem.
Now at last we are able to show easily that the torus and the Klein bottle are not homeomorphic. In fact they are not even homotopy equivalent.
The real projective plane RP2 is the orbit space of the antipodal action on the 2-sphere.
We are now ready to glue surfaces together. This will be done in a rather random way. Namely, we will not try to be systematic in the way we form connected sums of surfaces and will use various combinations of the 2-sphere, torus, Klein bottle and the real projective plane. Later on we will consider the classification theorem for closed surfaces and will in particular be able to present a calculation using the Van Kampen theorem of all possible fundamental groups of closed surfaces. This will enable us to argue that all the surfaces described in the clasification theorem have distinct homotopy types.