The goal of this section is to introduce the notion of classifying space of a (Hausdorff) topological group . This space has the property that isomorphism classes of principal -bundles over a paracompact Hausdorff base space correspond bijectively to homotopy classes of maps from the base to , hence the name. More explicitly this correspondence is given by pullback of a universal principal -bundle .
The whole world in an example[edit | edit source]
We begin by considering the special case of . But first we need some preliminaries on Grassmannians. Denote by the Grassmannian of -dimensional linear subspaces of . We then have a tautological short exact sequence of vector bundles on :
denotes the tautological subspace bundle on
is the trivial bundle with fibre
is the obvious quotient bundle. If we define
we obtain a vector bundle
which we can think of as the tautological quotient bundle on the infinite quotient Grassmannian
Now let be a complex vector bundle of rank over a manifold ; for the present discussion it is irrelevant whether this is a smooth vector bundle or just a topological one. Denote by the trivial vector bundle with fibre over , where is the space of sections of . We then have a short exact sequence of vector bundles
where the map
is the obvious one given by evaluation of sections. Thinking of
as the vector space
, this expresses
as a quotient of
, i.e. every fibre of
-dimensional quotient of
. This corresponds to having a map (continuous or smooth, depending on the chosen setting)
is the pullback
of the tautological quotient bundle
is the tangent bundle of
we can think of this map
as a generalised Gauss map for the manifold. Denote by
the set of isomorphism classes of complex vector bundles of rank
be the set of homotopy classes of maps
. Since pullbacks under homotopic maps are isomorphic we have sketched the proof of a bijective correspondende
For the present case of
Consider the case . By choosing inner products on the (e.g. the obvious ones) we obtain an isomorphism . Combining this with what we have seen so far we obtain a bijective correspondence
Now let be the unit -sphere. One can show by means of clutching functions that corresponds bijectively to . But by what we have seen now we also know that corresponds to . Thus we conclude that there is an isomorphism
(actually we only know that it is a bijection, we will however see shortly that it is indeed an isomorphism). Note that we are explicitly working with
here, but the same argument does work for general classifying spaces once we have defined them.
Note that The whole world in an example doesn't quite fit into the general statement given in the introduction: there are no principal bundles mentioned here. It turns out however that there is a somewhat canonical correspondence between complex vector bundles of rank and principal -bundles (this works for other fields too). Namely we define the frame bundle of a given vector bundle of rank to be the principal -bundle whose fibre over a point is the space of isomorphisms from to the fibre of over that point. Note that there is an obvious right action of on which preserves the fibres and acts freely and transitively on them. Noting that pulling back commutes with the frame bundle construction we can translate what we have seen so far into a correspondence
The correspondence here is given by pulling back the frame bundle of
. We will denote it by
and say that it is the universal principal -bundle
(precisely because of the above correspondence).
We now want to see that the space is weakly contractible. This will be true in the general setting and is part of the allure of the universal principal bundle. Knowing that is paracompact we obtain a long exact sequence of homotopy groups:
All the maps except
are the obvious ones. To construct
start by picking an element
and choosing a representative
. We can think of this map as a map
on the closed unit
-disk that is constant on its boundary
. From the fact that
is a fibre bundle with paracompact base space we know that there is a lift
. Since the image of
is a single point we know that its image under
must be contained in a fibre. Thus we have
and we can define
The fact that this is well-defined follows from the homotopy lifting property for fibre bundles over a paracompact base. If we now knew that
is an isomorphism we'd be done. This is indeed true as the following exercise shows:
Meditate on the fact that is the same map as the one in (MISSING).
Think about what the above construction of looks like explicitly in the case of the Hopf fibration .
Let be the group of integers under addition. Show that and that is with the map being the complex exponential. Do the above by noting that principal -bundles on are in correspondence with and that the latter is in correspondence with .
Think about what the total space of the pullback of under the double cover looks like.
To define the classifying space and universal bundle for a general (Hausdorff) group we can start by constructing a principal -bundle with contractible. To do this we use the topological join construction:
Let be topological spaces. The join is then defined to be , where the equivalence relation is generated by and for and .
Note that the join of two spaces is just the space obtained by connecting every point of one space by the unit interval to every point of the other space. Moreover the construction is associative and commutative. We now define the universal bundle by setting and . One has to verify that this is indeed a principal -bundle (in particular that it has local trivialisations), but that follows by thinking in an appropriately elegant way about .
We have seen how to construct . If has a faithful linear representation then note that acts freely on and so we can just define and .
We now have the following correspondence in general:
Let be a paracompact Hausdorff space and let be a Hausdorff topological group. There is a bijective correspondence
where the map from right to left is given by pulling back
along a representative.
For Example 5.3 to fit into this general picture we have to mention another result:
Let be a Hausdorff topological group. Then every principal -bundle with contractible total space is a universal bundle in the sense of Theorem 5.1. In particular its total and base spaces are unique up to homotopy.
Although a rigorous proof of Theorem 5.1 in the general topological setting requires some care regarding technicalities, we can easily argue why this should be true for the case of a smooth manifold and a compact Lie group. Namely let be a principal -bundle and consider the following diagram:
Here where for , and . Now note that the bundle is the pullback of under the obvious map , and also that it is the pullback of under the projection . But we also have that is a fibre bundle with fibre which is weakly contractible and so it has a global section . Finally it follows from what we have said so far that is the pullback of under .