A BESTIARY OF TOPOLOGICAL OBJECTS

N. P. STRICKLAND

Contents
1. Introduction
2. Wedges of circles
3. The conguration spaces and
4. Projective spaces
5. Grassmannians and ag varieties
6. Milnor hypersurfaces
7. Unitary groups
8. Projective unitary groups
9. Lens spaces and
10. Fermat hypersurfaces
11. Two-dimensional manifolds
12. Three-dimensional manifolds
13. Simply-connected four-dimensional manifolds
14. Moore spectra
15. Eilenberg-MacLane spaces
16. Wilson spaces
17. The building for
18. Looped spheres
19. Doubly looped spheres
20. Homology of Eilenberg-MacLane spaces
21. Things left to do
References

2
3
4
7
10
14
15
17
18
21
24
26
29
31
32
34
35
37
40
44
48
50

N. P. STRICKLAND

1. Introduction
These notes are intended to give graduate students in algebraic topology an overview of the
range of examples that are understood. I have also added a long list of other examples which I have
not had time to write anything about as yet. It takes quite a lot of theory to justify everything
that I say about the examples I discuss, but I hope that it will not take too much theory to
understand most of it.
These notes are new and not yet thoroughly debugged. Any comments, suggestions or corrections would be gratefully received.
1.1. Background Reading. In this section, we list some sources which provide the theory necessary to justify the examples which follow.
Maunders book [27] is a pleasant introduction to general homology and homotopy theory.
Dolds book [13] contains more information about homology, and Whiteheads book [44] is a
very good reference for the more classical parts of homotopy theory. Simplicial methods are an
indespensible tool in much of algebraic topology; the canonical reference is [28]. The book [19]
also contains a good treatment of simplicial sets and CW complexes, together with a lot of useful
background material from point-set topology. Another indespensable tool is the theory of spectral
sequences, which is explained in a pleasant and approachable way in [31]. Lie groups crop up
in topology in a number of ways; Adams exposition in [1] is very elegant. His book [3] also
provides an excellent survey of the theory of innite loop spaces. For some purposes, one needs to
approach homotopy theory from a more abstract point of view (to clarify the relationship between
the homotopy theory of spaces and that of chain complexes, for example). One way to do this is
to use Quillens theory of closed model categories; there is a nice exposition in [16]. (There are a
number of other useful survey articles in the same collection.)
For stable homotopy theory, the student should start by reading about the -category in [21,
Chapter 16] (actually, the whole book is highly recommended) and then read Adams Chicago
lecture notes [2, Part III]. Adams construction of the smash product should be ignored, however,
for reasons to be explained shortly. The rst part of Margolis book [26] is a very good treatment
of the formal properties of the stable homotopy category, although it is almost entirely devoid of
geometry. Kochmans new book [23] looks very promising, although I have not seen a copy yet.
One of the most important themes in stable homotopy theory is the interection with the algebraic
theory of formal groups via complex cobordism. This is explained in [2], and taken further in
Ravenels book [36]. The latter also contains valuable surveys of a number of areas of topology (as
well as some large and elaborate calculations of stable homotopy groups). Another important (and
related) theme is the exploitation of the Nilpotence Theorem of Hopkins, Devinatz and Smith.
This is explained in another book by Ravenel [37], which does a remarkably good job of presenting
a very deep result in a way which is comprehensible with a minimum of prerequisites.
There are now ways of setting up the foundations of stable homotopy theory which are more
satisfactory in a number of ways than those known to Adams. In particular, Elmendor, Kriz,
Mandell and May have costructed a category whose homotopy category is the same as the one
considered by Adams, but which has much better properties before passage to homotopy. This
is explained in [17], which is unfortunately not light reading. The rather easier construction
described in [30, Chapter XII] is just as good for most purposes. Incidentally, the book [30] also
contains a wealth of other fascinating material.

A BESTIARY OF TOPOLOGICAL OBJECTS

2. Wedges of circles
Let be a wedge of circles, = 1 . . . 1 . We shall show that is the quotient of

a contractible space by a free action of the free group on generators. It will follow that
= 0 unless = 1, and that 1 = . For this and related material, see [41, Chapter 2].
The inclusion : 1 of the th wedge summand in can be thought of as an element

1 (for = 0, . . . , 1). Recall the denition of the free group generated by the
elements : an element of is a sequence (1 , . . . , ) (possibly empty) in which each term
is either an or an 1 for some , and never occurs next to 1 . Thus () and (3 , 1 , 1 )

2
are elements of but (1 , 2 , 2 ) and (2 , 1 ) are not. The group operation is dened by joining
3
2
sequences together and discarding adjacent pairs of the form ( , 1 ) or (1 , ) until there are

none left. There is an obvious homomorphism 1 sending a sequence (1 , . . . , ) to the

product 1 2 . . . .

We next dene to be the (innite) simplicial complex with vertices and edges joining to
for each . More explicitly, write = {0 , . . . , 1 } and

= ( ( [0, 1]))/
(, , 0)

(, , 1) .
.
We take the element () as the basepoint of

We now dene a truncation map : as follows. The map : of vertices sends

(1 , . . . , ) to (1 , . . . , 1 ) (and 0 = () to itself). Clearly, if and are joined by an edge then

the same is true of () and (), so extends to give a simplicial map . Moreover, it

is easy to see that is homotopic to the identity, so we have a family of maps : with
0 = 1 and 1 = . We may also assume that (0) = 0.
We now extend this and dene () for all 0 by the formula () = ( ()),
where = + with and 0 1, and () = 0. One can check that this gives a

continuous map : [0, ] , where [0, ] is topologised as the one-point compactication
of [0, ). Using a homeomorphism [0, 1] [0, ] (sending to /(1 ), say), we conclude that

0 = 1 is homotopic to = 0, so that is contractible as claimed, and thus = 0 for all .

. To do this, think of as ( [0, 1])/ and

We now dene a map :

as (0 [0, 1])/ , where in the latter case (, 0) (, 1) 0 for all . In this picture,
sends to 0 and [0, 1] to [0, 1] by the projection. One can check that this is a bundle

projection, and in fact it displays as /, where acts on the left on in an obvious way.
that 1 = . (As yet, we
It follows from the long exact sequence of the bration

have only proved that this is a bijection, not an isomorphism of groups, but that is also true). It
also follows that = 0 for > 1.

N. P. STRICKLAND

3. The configuration spaces and

Dene
= { injective maps : {0, . . . , 1} }

= {(0 , . . . , 1 ) = for = }
We take the usual inclusion {0, . . . , 1} as the basepoint in .

Let be the group of permutations of the set {0, . . . , 1}. This acts on on the right
by composition (i.e. sends to ). We dene
= / = { subsets of of order }.
Clearly is a noncompact open submanifold of and thus has real dimension 2. The
nite group acts freely (i.e. = implies = 1), which implies that the orbit space
is also a manifold, and that the projection is a covering map.

Another nice description of is as follows. Let be the set of monic polynomials of degree
over in one variable . This can be identied with in an obvious way. Dene a map

by sending (0 , . . . , 1 ) to the polynomial ( ). This is clearly independent of the order of

the s, so it induces a map : . One can check that this is a homeomorphism of

with an open subset of .

Now write = {0, . . . , 1} (with as the basepoint). It is not hard to see that is
homotopy equivalent to a wedge of circles, so that = 0 for = 1 and 1 is a free group
on generators.
3.1. Fibrations. We next consider the map : +1 dened by

(0 , . . . , ) = (0 , . . . , 1 ).
We claim that this is a bre bundle projection with bre , in other words that we can cover
with open sets and nd homeomorphisms : 1 such that = : 1

(such an is called a local trivialisation). First, note that this is at least plausible; the preimage of
a point under is just the set of pairs (, ) with {0 , . . . , 1 }, which certainly
looks rather like .
To make this precise, we rst let be the set of -tuples (0 , . . . , 1 ) such that = +
and 0 < 1 < . . . < 1 (so ). Given a point , dene a homeomorphism
: as follows. The map will send vertical lines to vertical lines, it will send the point

to , and it will send the vertical band +1 to the band + 1. Explicitly, for
0 we dene ( + ) = + 0 , for 1 we dene ( + ) = + 1 + 1,
and for +1 we dene = ( )/(+1 ) and
( + ) = + + ( (1 ) +1 ).
Next, observe that 1 is the set of pairs (, ) where and {0 , . . . , 1 }. Dene
1
: 1 by (, ) (, ()). This is a bijection, with inverse (, ) (, ()).

With a bit more work, one can check that it is a homeomorphism, as required.
So far, we have only trivialised over the open set ; we need to do the same over a covering
family of open sets. For each permutation of {0, . . . , 1} and each complex number with
= 1, write (, ) = { }. Dene : 1 (, ) (, ) by

(, ) = (, ()).
One can check that this gives a trivialisation over (, ), so we need only check that these sets
cover . For any , choose any angle that does not occur as arg( ) for any , < ,
and set = . The real parts of the numbers 0 , . . . , 1 are then all distinct, so there is a
unique permutation such that ((0) ) < . . . < ((1) ). We thus have (, ), so the
sets (, ) cover as claimed.
This implies that the projection : +1 is a bration with bre . The resulting

long exact sequence shows (by induction on ) that = 0 for > 1.

A BESTIARY OF TOPOLOGICAL OBJECTS

3.2. Cohomology. Recall that

[, 1 ] = [, ] = 1 .
Thus, for each = we have a class 1 corresponding to the map . Because
the map 1 : is homotopic to the identity, we see that = .

To nd some relations among these classes, consider a triple < < . This gives a map
3 by ( , , ). Clearly , and are in the image of this map in cohomology.

There is a homeomorphism
3 = {0, 1}
given by
(, , ) (, , ( )/( )).
It follows that 3 1 ( 1 1 ). Using this, one can check that
+ + = 0.
In fact, we have
= [ 0 < < ]/( + + < < )
and this has a basis consisting of the monomials 1 1 . . . such that 1 < 2 < . . . and <
for all . This is proved by making a similar statement for and verifying it inductively
using the Serre spectral sequences of the evident brations
1,+1 , 1, .

3.3. The action of the symmetric group. The rst stage in computing is to understand the action of on . The character of this representation of is the map

:
trace( : ).

Let : be the complex conjugation map. This has = and thus = (1)

on . It follows from the Lefschetz xed point theorem that

() =
(1) trace( : ) = ( xed point set of ).

To understand the right hand side, suppose that has a xed point, say = for some
. We then have 2 = and thus 2 = 1, so that is a product of disjoint transpositions,
say of them. It is not hard to check that the xed-point set of is homeomorphic to a disjoint
union of 2 copies of ( ) 2 , where is the upper half plane, which is homeomorphic to
. Using our earlier calculation of we conclude that when > 1, we have ( ( )) = 0
and thus () = 0. From this we conclude that

!
if = 1

() = 2( 2)! if is a transposition

0
otherwise.
From this it is easy to check that = 2 ind (1). It follows that there is an isomorphism of
2
ungraded [ ]-modules
( ; ) = [ ] [2 ] ( ).
More delicate analysis shows that = 1 is a module over [ ] that is free of rank
one as a module over the subring [1 ]. We can also describe the rest of as a direct
sum of modules of the form [ ] [] (1 . . . ).
Using this and the covering spectral sequence
( ; ) =
one
can show that ( ; ) = [] and ( ; ) = [], where 1 ( ; ) maps
1

to
< in ( ; ). The groups ( ; ) are in principle known for all , but are
complicated.

N. P. STRICKLAND

3.4. References. Most of the material here is originally due to Fred Cohen [11]. His later survey
article [10] does many things more cleanly, and also gives further references. I rst learnt the above
calculation of the character of as a -module from Erich Ossa, although apparently a
similar approach had been independently discovered by Atiyah and Jones some time previously. I
have also written but not published an exposition with more details.

A BESTIARY OF TOPOLOGICAL OBJECTS

4. Projective spaces
Let be a complex vector space of nite dimension , equipped with a Hermitian inner product.
The group acts by multiplication on {0}, and the subgroup 1 = { = 1} preserves
the sphere ( ) = { = 1}. The associated projective space is dened as
= ( {0})/ = ( )/ 1 .
This is a compact complex manifold of complex dimension 1. Clearly, the points of biject
with one-dimensional subspaces of via [] . We can also identify with the space
:= { End ( ) 2 = = , trace() = 1} :
an endomorphism corresponds to a line = if = ( ) or equivalently
() = , /, for all . We write 1 for ( ).
There is a tautological line bundle with total space

() = {(, []) }.
If dim( ) dim( ), then the space of linear embeddings is connected, so we get a

canonical homotopy class of maps .

If is an innite-dimensional vector space (topologised as the colimit of its nite-dimensional

subspaces) then we also dene = ( 0)/ , and topologise it as the colimit of the s, where
runs over nite-dimensional subspaces. In particular, we will consider the innite-dimensional
projective space = lim . However, in the rest of this section we shall assume that

has nite dimension unless we explicitly state otherwise.

4.1. (Generalised) Cohomology. For any linear embedding : 2 , we get an embedding

2
1
2 such that is the usual generator
: = . There is a unique element

of 2 2 = , for every such . Moreover,

= []/+1 .
Dually, we have
= { 0 < }

with , = , .
In the limit, we get
= [[]]
= { 0}.
If we use coecients (with > 2) then the action of the Steenrod algebra is
(
)

=
+(1)
= 0.

We can also describe the complex -theory of (compare [4]). We have an element =

[ 1] 0 , in terms of which we have

0 = []/
= []/ = [ 1 , ]/
2 .
The action of the Adams operations is [] = [ ] and thus () = (1 + ) 1.
Let be a multiplicative generalised cohomology theory. A strict complex orientation of is

a class 2 such that the image of in 2 1 = 2 2 is the usual generator. We say

that is complex-orientable if it admits such an orientation; one can show that this is the case if
is torsion-free or concentrated in even degrees. We say that is complex-oriented if we have
chosen an orientation. In particular, and (with = /) are complex-oriented. See [2, 36]
for more information about such cohomology theories.
If is complex-oriented, then we have
= []/
= [[]]

N. P. STRICKLAND

= { 0 < }
just as with ordinary homology.
4.2. Dierential geometry. Let be a one-dimensional subspace of , corresponding to a
point of . Let be the set of lines that are not orthogonal to , which is an open neighbourhood of in . There is a homeomorphism
: Hom(, )
dened by
() = graph() = .
The maps are charts for a holomorphic atlas on , making it a complex manifold.
This construction also shows that the tangent bundle of is given by
= Hom(, ) = Hom(, ) =
(using the fact that End( ) is canonically isomorphic to ). As is a trivial bundle of rank ,
this gives
= 1 1 0 .
Note also that this isomorphism gives a natural Riemannian metric on the tangent bundle.
There is a natural embedding of in a Euclidean space by the map
: = ( )/ 1 End( )

([])() = , .
This gives a homeomorphism of with the set
{ End( ) 2 = = and trace() = 1}.
4.3. Bundles, brations and Thom spaces. The bration 1 ( ) can be easily

identied with the evident bration 1 () , where () is the circle bundle associated

to the line bundle .

More generally, let denote the th tensor power of . There is thus a bration
1 ( ) .

This can be identied as follows. Let be the cyclic subgroup of 1 generated by exp(2/).
Let 1 act on ( )/ by .[] = [ 1/ ]. We then have a bration
1 ( )/ ( )/ 1 = ,

which is isomorphic to the bration above.

Next, we claim that the Thom space is homeomorphic to ( ). Indeed, the Thom
space is obtained from the disc bundle
() = {(, []) and 1}
by collapsing out (). We dene a map
()/() ( )

by
(, []) [, ,

1 2 ].

It is easy to check that this is a homeomorphism.

A more general result can be obtained in a slightly dierent way; we claim that

= ( )/

or equivalently (using the conjugate-linear isomorphism 1 given by the metric, and taking
= +1 and = )
( ) = + / 1 .

This space is often called a stunted projective space and written .

A BESTIARY OF TOPOLOGICAL OBJECTS

4.4. Cell structure. The main result is that has a cell structure
= 2 4 . . . 2 .
Indeed, we have seen above that ( ) is the cobre of a map 2+1 = ( ) , so that

+1 = 2+2 as required. Note that this also gives a cell structure

= 2 2+2 . . . 2+2 .
+

(However, + is not usually homotopy equivalent to 2 , as one sees by examining the

action of the Steenrod algebra on the mod cohomology.)
The cell structure on can also be obtained by Morse theory (compare [32, Section I.4]).
Dene a map

2
)/2 .
:

[] = (
=0

One can show that this is a Morse function on , in other words that all of its critical points
are non-degenerate. The critical points are precisely the basis vectors [ ] (for = 0, . . . , ). The
map () = [ + ] is a dieomorphism of with a neighbourhood of [ ], and

() = [ ] +
( ) 2
=

to second order in . This shows that the (real) Hessian of at [ ] has 2 negative eigenvalues
(and none of the eigenvalues are zero, which is what nondegeneracy means). Thus, [ ] has index
2. Morse theory tells us that has a cell structure with one cell of dimension for each
critical point of index , just as before.
4.5. Product structure. We can regard +1 as the space of polynomials of degree at most ,

via the bijection (0 , . . . , ) . With this identication, multiplication of polynomials

gives a bilinear map +1 +1 ++1 , which in turn induces a map +

(using the fact that a product of nonzero polynomials is nonzero).

In particular (using commutativity of []) we get a map
( 1 ) / .

The fundamental theorem of algebra implies that this is a bijection (and hence a homeomorphism,
as the source is compact and the target Hausdor).
We also get a map : , which makes into an Abelian topological

monoid. Up homotopy, this also has inverses, given by the map : sending
to

[ ] to [ ].
4.6. Homotopy groups. Using the bration 1 ( ) and the fact that 1 =

concentrated in degree 1, we see that

{

if = 2

=
2+1

if = 2
In particular, we have = 0 for = 1 or 2 < < 2 + 1 and 2+1 = . In the limiting
case we get
{
if = 2

=
0 otherwise
4.7. References. The material in this section is all well-known, and can mostly be found in [44]
or [21] (for example).

10

N. P. STRICKLAND

5. Grassmannians and flag varieties

We again let be a complex vector space of nite dimension . We let denote the set
of subspaces such that dim( ) = (so = 1 ). If we let ( , ) denote the
space of inner-product preserving linear maps : , then the construction ( ) gives

a bijection ( , )/ () , and we topologise so as to make this a homeomorphism.

We can also identify with the space

( ) := { End ( ) 2 = = , trace() = } :

an endomorphism corresponds to a subspace = ( ) if = ( ) or

equivalently = .
There is a tautological bundle with total space

( ) = {(, ) ( ) }.
If dim( ) dim( ), then the space of linear embeddings is connected, so we get a

canonical homotopy class of maps .

If is an innite-dimensional vector space (topologised as the colimit of its nite-dimensional

subspaces) then we also dene

= { < dim( ) = } =

< , dim( )<

In particular, we consider the spaces

() = ( ) = ( , )/ ().
One checks that the space ( , ) is contractible and that () acts freely on it, so () is
indeed a classifying space for principle ()-bundles, as indicated by the notation.
5.1. Schubert cells. Given 0 , we let () denote the set of sequences
0 = 0 . . . = .
We dene : () by unnished.

Given , we dene
5.2. Innite Grassmannians.
5.3. Fock spaces. Let be a complex universe. Given subuniverses , we write
i /( ) and /( ) are both nite-dimensional. This is easily seen to be an equivalence
relation. We say that is standard if both and are innite-dimensional, and = .
If is standard and then is also standard. A polarisation of is an equivalence class
of standard subuniverses. Let be a polarisation.
Note that if , then + and also lie in .
Denition 5.1. Given , with , we put
(, ) = { }.
This is naturally identied with the Grassmannian of subspaces of / . The set can thus be
regarded as a ltered colimit of projective varieties.
Denition 5.2. Given , we dene dim(, ) = dim( / ) dim(/ ), for any
with . This is easily seen to be independent of , and to satisfy dim(, ) = 0 and
dim(, ) + dim( , ) = dim(, ).
Denition 5.3. Given a vector space , we write for the th exterior power. We also write
= dim( ) , for any nite-dimensional vector space . Finally, we write det( ) = .
We note that when there is an isomorphism det( ) = det( ) det( / ), which is
natural in the pair (, ).

A BESTIARY OF TOPOLOGICAL OBJECTS

11

Denition 5.4. Given , we dene

det(, ) = det(/( )) det( /( )) = Hom(det(/( )), det( /( )))
(which is a one-dimensional complex vector space).
Proposition 5.1. The set can be made into a category, with det(, ) as the morphisms from
to . Moreover, the composition map
det( , ) det(, ) det(, )

is an isomorphism.
Proof. First, for any we put
det(, ; ) = Hom(det(/ ), det( / )).
If then we have canonical isomorphisms det(/ ) = det(/ )det(/ ) and det( / ) =
det( / ) det(/ ). As det(/ ) is invertible, these induce an isomorphism det(, ; )
det(, ; ). These isomorphisms compose in the obvious way. Thus, we can replace det(, )
by det(, ; ) for any convenient . Now take , and put = / , = /
and = / . We have
det( , ) det(, ) = det( ) det( ) det( ) det()
= det( ) det()
= det(, ).
This identication is easily seen to be independent of , and to be associative.

Denition 5.5. For any , we dene the Fock space () = (, ) as follows. For any
, with , we put
(; , ) = det(/ ) +/ (/ ) = Hom(det(/ ), +/ (/ )).
Now suppose we have . On the one hand, we have det(/ ) =
det(/ ) det(/ ). On the other hand, the ring structure of ( / ) gives a map
( )
( )
( )
( )

= /
+/
+/

.
: det(/ ) +/

Let be the ideal in (/ ) generated by / 1 (/ ). Then det(/ ) = 0 and

(/ )/ = (/ ). Our map thus induces a map
( )
( )

+/
/
: det(/ )

and thus a map

(; , ) = Hom(det(/ ), / (/ ))
Hom(det(/ ) det(/ ), det(/ ) / (/ ))
Hom(det(/ ), det(/ ) / (/ ))

Hom(det(/ ), / ( / ))

= (; , ).
It is easy to see that these maps are injective, and that they compose together in the obvious way.
We can thus dene
() = lim (; , ).

Proposition 5.2. There are natural isomorphisms

( ) = det( , ) dim( ,) ()
for all , .

12

N. P. STRICKLAND

Proof. Put = dim( , ). It will suce to give compatible isomorphisms ( ; , )

det( , ) (; , ) for all , with and + . Put = dim(/ ),
so dim( / ) = . We then have det( , ) = det( / ) det(/ ), so
( ; , ) = det( / ) + (/ )
= det(, ) det(/ ) + (/ )
= det(, ) (; , )
as required.

Denition 5.6. Given , we put 0 () = { dim(, ) = 0}. Given , with

we put
0 (; , ) = 0 () (, ) = { and dim(, ) = 0}.
We also let () denote the line bundle over 0 () with bre det( , ) at
Proposition 5.3. There is a natural isomorphism
(0 (; , ); ()) = 0 (; , )
(where (, ) denotes the space of algebraic sections).
Proof. Put = dim(/ ), so dim( / ) = for 0 (). Let be the bundle over (, )
with bre / at . The restriction of () to 0 (; , ) is det(/ ) det( ) , so
(0 (; , ); ()) = det(/ ) (0 (; , ); det( ) ).
On the other hand, we have
0 (; , ) = det(/ ) ( ) .
The claim now follows from Lemma 5.4 below.

Lemma 5.4. Let be a nite-dimensional complex vector space, and let be the tautological
bundle over Grass ( ) (the Grassmannian variety of subspaces of ). Then (Grass( ); det( ) ) =
( ) .
Proof. Suppose we have an element ( ) . For Grass ( ) we let () denote the
restriction of to det( ) = , so () is an element of det( ) , which is the fbre
of the bundle det( ) at the point . Thus, we can regard () as a section of ( ) , which is
easily seen to be algebraic. Thus, we have a map
: ( ) (Grass ( ); ( ) ).

If = dim( ) then Grass ( ) = { } and is obviously bijective. We therefore suppose that

< dim( ).
Now suppose we have (Grass ( ); ( )) . Let be the set of linearly independent lists
= (1 , . . . , ) in . Given , we dene
= span() Grass ( )
()() = (1 . . . ).

One checks that has codimension + 1 2 in . As () is a rational function

that is regular away from a closed subvariety of codimension at least two, it extends uniquely as
a globally dened polynomial function. We also see from the denition that
(
)

()(1 1 , . . . , ) =
()(1 , . . . , )

()((1) , . . . , () ) = sgn() ()(1 , . . . , ),

showing that () is alternating and multilinear. It can thus be regarded as an element of ( ) .
It is easy to see that the maps and are mutually inverse isomorphisms.

A BESTIARY OF TOPOLOGICAL OBJECTS

13

5.4. Flag varieties. For {0, . . . , }, put

Flag ( ) = {
< whenever < }.

If = {} for some , then Flag ( ) is the projective space associated to a certain vector
bundle over Flag ( ).
Note also that the space Flag( ) = Flag{0,...,} ( ) is a bre bundle over ( ), with bre
Flag( ) Flag( ).
5.5. Presentations as homogeneous spaces. The group () acts freely and transitively on
the space ( , ). Given a set
= {1 < < } {0, . . . , },
we put
() = { () ( ) = for = 1, . . . , },
which is the stabiliser of the point
(1 < . . . < ) Flag ( ),
and is isomorphic to
(1 ) (2 1 ) ( 1 ) ( ).
We nd that there is a canonical homeomorphism
Flag ( ) = ( , )/ ().
If = we have ( , ) = () and so Flag ( ) = ()/ (). In particular, we have
= Flag{} ( ) = ()/( () ( ))
= 1 = ()/( (1) ( 1)).
5.6. (Generalised) Cohomology. Let be an even periodic cohomology theory
5.7. Bott periodicity.
5.8. References. The material in this section is all well-known, and can mostly be found in [44]
or [21] (for example).

14

N. P. STRICKLAND

6. Milnor hypersurfaces
The Milnor hypersurface (where ) is the complex variety

= {([], [])
= 0}.
=0

There are two evident line bundles over :

([],[]) =
([],[]) = .
We write = () and = ( ) for the Euler classes.
Theorem 6.1. We have
= [, ]/(+1 , 1 + . . . + () ).
We now outline a proof of this. It is easy to see that +1 = 0. Consider the bundle 0 over
with bres

+1
0

= 0}.
([],[]) = {
=0

Over , the identity map gives an inclusion 0 and thus a nowhere-vanishing section of

Hom(, 0 ) = Hom(, + 1 ) = ( + 1) , or dually a nowhere-vanishing section

of ( + 1) . It follows that the Euler class (( + 1) ) maps to zero in .
To compute this Euler class, we work with the evident analogous bundles over . Here
+1 = (( + 1) ) = (( + 1) )( + ).
As + is not a zero-divisor in ( ) and
( + )( 1 + . . . + () ) = +1 +1 = +1 ,
we can conclude that (( + 1) ) = 1 + . . . + () . We thus get a map
= [, ]/(+1 , 1 + . . . + () ) = ,

and the claim is that this is an isomorphism. To see this, consider the Serre spectral sequence for
the evident bration 1 . This has the form

( ; 1 ) = [, ]/(+1 , ) = .
The 2 term is concentrated in even bidegrees, so all dierentials must be zero, and = 2 .
Thus, the associated graded ring of under the Serre ltration is [, ]/(+1 , ).
We can now lter by
= ([]/+1 ){1, , . . . , /2 }.
We nd that the associated graded ring is just [, ]/(+1 , ). The Serre ltration on
always stops with , so trivially 2 has Serre ltration at most two. We thus get
a ltration-preserving map : and a resulting map of associated graded rings. It is easy

to see that this sends to , and thus that it is an isomorphism. From this we can conclude that
: is an isomorphism, as required.

A BESTIARY OF TOPOLOGICAL OBJECTS

15

7. Unitary groups
Let be a complex vector space of nite dimension , equipped with a Hermitian inner product.
We dene
( ) = { unitary automorphisms of } = { End( ) = 1}
( ) = { antihermitian endomorphisms of } = { End( ) + = 0}.
It is not hard to see that ( ) is a subgroup and a closed submanifold of Aut( ), and thus a
Lie group.
7.1. Dierential geometry. The map exp() + gives a dieomorphism of a
neighbourhood of 0 in ( ) with a neighbourhood of in ( ), and thus an isomorphism
( ) ( ). This shows that the tangent bundle of ( ) is trivial (which is in fact true
for any Lie group). Note also that
dim ( ) = dim ( ) = 2 .
We also write
( ) = { ( ) det() = 1}
( ) = { ( ) trace() = 0}
We again have ( ) = ( ) and
dim ( ) = dim ( ) = 2 1.
7.2. Complex reection maps. Given a line and 1 < we dene
(, ) = 1 End( ) = End( )
or equivalently
(, [])() = + ( 1),
(when ( ) = { = 1}). It is easy to see that (, ) ( ) and that det((, )) =
. Given a chosen basepoint 0 we dene
(, ) = (, )(, 0 )1 ( ).
These constructions give continuous maps
: (+ ) ( )

: ( )

det

Moreover, the map (, 0 ) gives a splitting of the extension ( ) ( ) 1 and

thus a homeomorphism ( ) = ( ) 1 (which does not respect the group structure).

There is a commutative diagram as follows.
1
+

()

w +

w w 2+1

w ( + 1)

w w 2+1

Here 0 () = (0 ), where 0 is the rst basis vector in +1 . The top copy of 2+1 should be

thought of as 1 ( ) and the bottom one as { +1 = 1}. The right hand vertical
map sends to 0 + ( 1)(0 + )/(1 + 2 ). Its inverse sends ( + , ) to ( + ) ,
where
= /(( 1) + )
= (1 + 2 ) 2
= (1 + 2 ).
The top line of the diagram is a cobration, and the bottom one is a bration.

16

N. P. STRICKLAND

7.3. Cell structure. The top cell of gives a map 2+1 + ( + 1), which in

2+1
2+1
2+1
turn gives a homeomorphism
/

.
From this we get a map 2+1 () ( + 1) sending 2 () to (). By induction,

we get a cell structure on ( + 1) with cells indexed by the evident basis for the exterior algebra
[1 , 3 , . . . , 2+1 ].

7.4. (Generalised) cohomology. For any complex-oriented multiplicative homology theory

(in particular, for = ) we have

( ) = [ + ] = [1 , . . . , 21 ]

( ) = [ ] = [3 , . . . , 21 ]

( ) = [ + ] = [1 , . . . , 21 ]

( ) = [ ] = [3 , . . . , 21 ]
All these rings are actually Hopf algebras, because ( ) and ( ) are groups. All the generators
are actually primitive, in other words the coproduct is ( ) = 1 + 1 and similarly for

. Here + is embedded in ( ) by , and + is embedded in ( ) by the

( ) + .
inverse of isomorphism Prim ( )

The Steenrod operations in mod cohomology (with > 2) are given by

(
)

2+1 =
2(+(1))+1

2+1 = 0.
7.5. The innite unitary group. Write = lim (). This has

= [ + ] = [2+1 0]

= [ + ] = [2+1 0]

= (0, , 0, , 0, , . . .)
This shows that = +2 = 2 . In fact, we have a homotopy equivalence 2 ; this
is called Bott periodicity.
7.6. The Miller splitting.
7.7. References. Most of the above can be found in [39, Chapter IV]. The homology ring is also
calculated in [44, Section VII.4]. For the theory of Lie groups in general, see [1].

A BESTIARY OF TOPOLOGICAL OBJECTS

17

8. Projective unitary groups

Lemma 8.1. () ts into diagrams as follows, in which each square is a homotopy-pullback:
The second diagram maps in an obvious way to the rst.
Corollary 8.2. () is the homotopy bre of the map () classifying , or

equivalently the frame bundle ( ).

This means we can approach () using the tower
() = ( ) 1 ( ) 1 ( ) = 1 0 ( ) = .

One can also check that ( ) = ( ) = ( )/ 1 . This has an evident action of (),
with orbit space ( ).
The pullback of to ( ) has a -dimensional trivial summand, and we write for its
orthogonal complement; then +1 ( ) is the sphere bundle in .
If we have a complex orientation 2 then the Chern polynomial of is ( ) .
In ( ) this becomes divisible by , and the Chern polynomial of is ( ) / , so the
(
)

Euler class is
. Now put

() = (, , ) = ( ).
We deduce that there are short exact sequences
(
)
(
)

()/
( + 1) 221 ann(

, ()).

Note also that acts on () by translation, and () is connected, so each element of

acts by a map homotopic to the identity. It follows that the projection : () () and

the associated transfer ! : ()+ ()+ satisfy ! = .1 () and ! = .1 () .

Thus, if is invertible in then the map

: () () = 1

is an isomorphism.

18

N. P. STRICKLAND

9. Lens spaces and

Fix an integer and an odd prime . Write = and = = 1 . Regard as
the subgroup { = 1} of generated by 2/ = . The lens space = is the quotient
( )/ = 21 / . The action of on 21 is free, so is a manifold.
Let be the tautological line bundle over , with bres [] = . We can also regard as
the sphere bundle of via the map ( ) sending [] to ([], ).

9.1. Cell structure. We now describe a cell structure on ( ) that is compatible with the action
of and thus induces a cell structure on . This comes from [39, Section V.5] or [44, Section
II.7]. Write
= { 1 and 0 arg() 2/}

= {(0 , . . . , , 0, . . . , 0) [0, 1]}

2+1 = {(0 , . . . , , 0, . . . , 0) }
One can check that is a closed -cell and that the cells with and 0 < 2 give a
CW structure on . The dierential in the cellular chain complex is

(2 ) =
21

2+1

) = 2 2 .

This gives a cell structure = 0 . . . 21 with (2 ) = 21 and (2+1 ) = 0. It follows

that
(; ) = {0 , . . . , 21 }.
9.2. Generalised cohomology. As = ( ), the cobre of the evident projection is

the Thom space . With this identication, the natural map from to the cobre becomes

the zero-section . For any generalised cohomology theory , this cobration gives us a

long exact sequence

+1 +1 .

If is a complex-oriented multiplicative theory then is a free module over =

[]/ generated by the Thom class ( ). Under the zero-section , this Thom

class pulls back to give the Euler class ( ) = [](()) = []().

In particular, if = (; ) then []() = 0. Putting this into our long exact sequence we nd that the restriction maps 2 ( ; ) 2 (; ) and the coboundary maps

2+1 (; ) 2+2 ( ; ) are isomorphisms. From this, we conclude that

(; ) = []/ [],
where 2 (; ) is the image of the usual generator of 2 , and 1 (; ) is the unique
class with () = , or equivalently the unique class whose Bockstein is () = .
The Steenrod action is
(
)

=
+(1)

= 0
= 0
= .
It follows that the Milnor Bockstein operations are given by
= 0
=

A BESTIARY OF TOPOLOGICAL OBJECTS

19

We can also be more explicit in the case of complex -theory. Write = 1 0 ; it is

not hard to check that
0 = []/( , (1 + ) 1)
1 =
9.3. Homotopy groups. Using the evident bration 21 , we nd that

0 = 0
1 =
= 0

for 1 < < 2 1

21 = .
9.4. The innite case. Write = lim = / . As is contractible, the long exact

sequence for the bration gives

=
0

if = 1
otherwise.

In other words, we see that is a model for the classifying space . We have
( ; ) = [] []
( ; ) = []/
0 = []/((1 + ) 1) [ ].

9.5. The intermediate case. It is also useful to consider the space = 2 , so that

+1 . This has cohomology

( ; ) = {1, , , . . . , 2 } = [] []/(2+1 , 2 ).

9.6. Thom spectra. We next consider the Thom spectrum for . When 0 this
is a space, and it has a nice alternative description: there is a homeomorphism

= + / = 2 . . . 2(+)1 .

To see this, observe that a point in the total space () over has the form ([], 1 , . . . , ),
with 21 and for each . We dene a map : + 0 2(+)1 by () =

/, and a map : () + by

([], 1 , . . . , ) = [(, , 1 , . . . , , )].

One can check that is a homeomorphism. By passing to the one-point compactications, we

obtain a homeomorphism = + / as claimed. This naturally identies ( ; )

+1
+1

with the subquotient { , , . . . ,

,
} of ( ; ) = [] [].

In fact, for all there is a similar identication of ( ; ) as a subquotient of

1
[, ] []. There is a unique action of the Steenrod algebra on [, 1 ] [] compatible
with this, given by
(
)
+
=
+(1)
for 0.

The spectra also have an interesting periodicity property:

(+)

Proposition 9.1.

2 whenever 1 divides .

Proof. It is enough to show that 1 is stably isomorphic to the trivial bundle 1 over
, or equivalently that 1 = 0 in 0 = []/( , (1 + ) 1). In this ring we have
+ ( 1)/2 2 + . . . + = 0 so ( 2 ) so 1 ( ) = 0 as required.

Note that we really only needed 1 ( 1) = 0 as stable spherical brations rather than as
vector bundles, or in other words that 1 ( 1) = 0 in Adams group ( ) rather than in
( ). See [21] for more about the groups ().

20

N. P. STRICKLAND

9.7. Lins theorem. One can assemble the spectra into a tower

. . . .
2
By applying the functor (; ), we get a sequence of inclusions
.1 .2 .3 . . . ,

where = [] [] and thus every term is a subgroup of = [1 ] []. If we let be

the homotopy inverse limit of the tower, we might thus expect that (; ) = . However,
contains many elements that are in the image of the Steenrod operation for an unbounded set

of s, and this cannot happen in the cohomology of a spectrum, so we cannot have (; ) =

Instead, we have the celebrated theorem of Lin, which says that = ( 1 ) , and thus

(; ) = {1 }. For a modern proof in its proper context, see [30, Corollary XX.6.2].
(The notational conventions in force there are that = (/) and that all spectra are implicitly
completed at .)
9.8. Stable splitting. Another interesting phenomenon is the stable splitting of = .
Note that is homotopy equivalent to ([] 0)/ , so we take this as our model. For any 0
we dene : by [ ] = [ ]. If = (mod ) then the map [ ] = [ + (1 ) ]

is well-dened, so . It is also clear that = . Thus, the maps give an action of

Aut( ) = on .

0
One can check that the spectrum = is -complete, so we get an action of the
-adic group ring [ ] on . For each there is a unique Teichmller representative
u

= lim such that = (mod ) and 1 = 1. The elements

1
[] [ ]
=

2
are orthogonal idempotents with
=0 = 1, so we have a stable splitting =
=0 .
Moreover, we have
( ; ) = { + = (mod 1)}.
In particular, we have
(0 ; ) = [1 ] [2 ],
and in fact there is a homotopy equivalence 0 = ( + ) .

It is a very common phenomenon for classifying spaces of nite groups to split stably as a wedge
of simpler pieces. See [7] for a survey.

A BESTIARY OF TOPOLOGICAL OBJECTS

21

10. Fermat hypersurfaces

Fix an odd integer and an odd prime . Write = +1 and = = . Let :

be the map () = =0 . We shall study the Fermat hypersurface

= {[] () = 0}.
Proposition 10.1. is a smooth complex manifold.
Proof. Suppose that = 0 and () = 0. Note that the map sending to [ + ] is an

open embedding. It is thus enough to check that { ( + ) = 0} looks like a manifold

near 0 , and thus enough to show that ( + ) = 0 to rst order in , for some . Let
be the vector (1 , . . . , 1 ), so that

0

, =
= () = 0

and thus . One can check directly that (+) = 2 to rst order in , as required.

Proposition 10.2. The normal bundle to in is .

proof ?
10.1. Morse theory. Consider the function : dened by

[] =
/
.

It is clear that 0 1, and that () = 0 if and only if .

We next consider the space

= { 2+1 [0, 1] for 0 }.

This has a free action of , and we write = / . It is not hard to see that the evident map
+1 / 1 = is injective, so we think of as a subspace of . Explicitly, it is the subspace

of all points [0 : . . . : ] such that all the numbers have the same argument. By the triangle
inequality, we thus have
= {[] [] = 1}.
We next claim that is continuously dierentiable on , with no critical points in ( ).
To see this, note that for any , with = 0 we have + = + Re(/) to rst
order in . Thus, when [] = 0 we have


1
+ 1 Re( )

[ + ] =
+ Re( 2 )
(to rst order). To simplify this, write
1 = ( 1 , . . . , 1 )

0
2 = (0 0 2 , . . . , 2 )

=


=
.
We nd that
(

, 1 2 , 2 .

This shows that is continuously dierentiable. If [] is a critical point then the real part of the
expression in brackets must vanish for all . As we can replace by , we see that the expression
in brackets must itself vanish for all . It follows that 1 and 2 must be linearly dependent,
say 1 = 2 for some and all . It is easy to see that = 1, and multiplying by

gives ( ) = 0. This means that , as required.

[ + ] = [] + Re

22

N. P. STRICKLAND

We now apply the ideas of Morse theory. Write = 1 { 1 } . By standard results on

2
the solution of dierential equations on manifolds, there is a continuously dierentiable function
: (0, 1) with ( 1 , ) = and

2
)
(

((, )).
(, ) =

2
It follows that ((, )) = , and that the map : (, ) gives a homeomorphism
1 {}.
For small > 0, one expects that 1 {} should be homeomorphic to the sphere bundle of
the normal bundle to = 1 {0} in . This can be made precise, and in fact we obtain a
homeomorphism between the total space of the normal bundle and . After recalling that the
normal bundle is and adding a point at innity, we obtain a homeomorphism

= /.

Note that is homotopy equivalent to the total space of , which is obtained from by
removing . It is easy to see that has codimension in , and so transversality arguments
show that the map is an isomorphism for < 1 and an epimorphism for = 1.

10.2. Cohomology. We need to consider the following three maps:

: (1)/2

[] = [ : ]

(inclusion)
: 1

[0 : . . . : ] = [0 : . . . : 1 ].

The denition of is legitimate because whenever (0 , . . . , ) = 0 and = 0 we must have
(0 , . . . , 1 ) = 0. Write for the usual generator of 2 , for any > 0. It is easy to see
that and are homotopic to the usual inclusions, so that = = . We also write
for 2 .
Next, observe that is a smooth embedding of real codimension 2 2 ( 3) = + 1, so
we have a Gysin map ! : (3)/2 ++1 , using which we dene = ! (1) +1 .

More explicitly, we can collapse out a tubular neighbourhood of (3)/2 in to get a map
from to the Thom space of an ( + 1)-dimensional complex bundle, and we dene to be the
pullback of the Thom class.
Note that ! is a map of modules over , so we have (3)/2 = ! ((3)/2 ). On the other
hand, (3)/2 is the fundamental class in (3)/2 , so it is equal to ! (1), where is the
inclusion of a point in (3)/2 . It follows that (3)/2 = ! ! (1) is the fundamental class in
.
Using this, Poincar duality, and the fact that is ( 1)-connected, we conclude that
e
= {1, , . . . , (3)/2 , , , . . . , (3)/2 } 1 .
Using the fact that has degree , we see that 1 is times the fundamental class (3)/2 ,
and thus that (+1)/2 = .
10.3. The middle dimension. We now analyse the middle-dimensional cohomology 1 .
For this, we need to understand the cohomology of .
Proposition 10.3. There is a stable splitting = / , where is a nite wedge of copies
of .
Proof. Recall that = / , where

= { 2+1 [0, 1] for 0 }.

We can identify with the ( + 1)-fold join . . . , so the usual formula ( ) =

)1 shows that = 0 ( )(+1) , concentrated in degree .

(
Note that both and are -dimensional, (1)-connected CW complexes with a free action
of . The usual proof of the essential uniqueness of gives canonical ( 1)-connected maps

/ , and non-canonical maps / compatible with and .

A BESTIARY OF TOPOLOGICAL OBJECTS

23

As : / is surjective, we conclude that () : / / is

the identity and thus that is a homotopy equivalence. After replacing by ()1 if necessary,
we may assume that = 1. It follows that there is a stable splitting = / for some
. Because , , and are ( 1)-connected and is -dimensional, we see that the homology
of with any coecients is concentrated in dimension and thus that is a wedge of copies of
.

Proposition 10.4. The number of spheres in is = (( 1)+1 1)/.

Proof. In this proof, all cohomology has coecients . It will be enough to show that the rank
of is + 1.
We will need a little structure theory of modules over the ring = [ ]. Write = [2/ ]

[1]. One can check that = []/ and that [] = 1 . It follows that the reduced
cohomology of the discrete space is isomorphic to = /1 as an -module, and thus that

the reduced cohomology of is = (+1) . By applying Schanuels lemma to the short

1

exact sequences and we nd that

= ( 1) .
By the Krull-Schmidt theorem, we can cancel to get
( 2) .
(In fact, it is not too hard to just write down an isomorphism here.) It follows easily that
= (+1) is the direct sum of with a free module over . By counting dimensions,
the rank of must be = (( 1)+1 1)/.
Now consider the covering spectral sequence
( ; ) = + .
The line = 0 is just = [] []. The line = is a free module of rank one
0
over on one generator 2 , together with the group (which is also located in
0
2 and is annihilated by and ). All other lines are zero, so the only possible dierential is
+1 . The image of must be annihilated by , which acts regularly on , so we must
have +1 ( ) = 0. Because has dimension , the element (+1)/2 must be hit, so +1 ()
must be a unit multiple of (+1)/2 . From this it is easy to see that the page consists of
0
{1, , . . . , (1)/2 } on the bottom line, together with in . It follows that has
dimension + 1, as required.

Corollary 10.5. The group 1 is free Abelian of rank + 2.

Proof. Our earlier study of the cohomology of showed that it is concentrated in even degrees,
and it is easy to see that this is valid with any eld coecients. This implies that is a free
Abelian group concentrated in even degrees, so we need only check that 1 ( ; ) has rank
+ 2. For the rest of this proof, all cohomology has coecients .

The cobration together with the Thom isomorphism 2

gives an exact sequence

0 = 1 +1 +1 = 0.

Here the rst term is zero because is odd, and the last is zero because has dimension . It
follows easily that 1 has rank + 2.

10.4. References. Most of the material here comes from [24]. The fact that the inclusion

is ( 1)-connected is a special case of the Lefschetz hyperplane theorem in complex algebraic

geometry. For a Morse-theoretic proof, see [32, Section I.7].

24

N. P. STRICKLAND

11. Two-dimensional manifolds

In this section we consider compact connected closed oriented smooth manifolds of dimension
two (which we shall refer to as surfaces). For more discussion, see [41, Chapter I]. It is well-known
that surfaces are classied by their genus, which is a nonnegative integer. A surface of genus zero
is homeomorphic to 2 , and a surface of genus one is homeomorphic to the torus 1 1 .
The simplest model of a surface of genus > 0 is given by identifying sides in a polygon. Let
= be a 4-gon in 2 . We give each edge a label and a direction. The labels are
1 , 1 , 1 , 1 , 2 , 2 , 2 , 2 , . . . , , , ,
The rst edge with each label is oriented anticlockwise, and the second one is oriented clockwise.
We illustrate the scheme in the case = 2.
1
w





[
[[

[
^
[

u 1

[[
][
[






2

Let = be the quotient space of where edges with the same label are identied together
(and thus all the vertices of are identied to a single point, which we take as the basepoint). It is
trivial to check that each point of other than the basepoint has a neighbourhood homeomorphic
to a disc, and only a little harder to check that this holds for the basepoint as well. This means
that is a topological manifold, and one can easily give it a smooth structure.
Note that the image of each edge or in is a closed loop passing through the basepoint,
so we can regard it as an element of 1 . One can show that
{
1 , . . . , , 1 , . . . [1 , 1 ] . . . [ , ] = 1 if = 1
=
0
otherwise.
This shows that the universal cover of
Moreover, we have

= 0

= =1

=2

is contractible; in fact, it is homeomorphic to 2 .

{1 , . . . , , 1 , . . . , }

The multiplicative structure is as follows: we have = for all , and all other products of s
and s vanish.
Let 2 be a wedge of 2 circles, so that 1 2 is a free group on 2 generators, which we can
call 1 , . . . , , 1 , . . . , . Write
= [1 , 1 ] . . . [ , ] 1 2 = [ 1 , 2 ].
It is not hard to see that there is a cobration

1 2 2 .

Note that any smooth Riemann surface is a surface of the type we are considering, so it will be
homeomorphic to for some . Every Riemann surface of genus is holomorphically isomorphic

A BESTIARY OF TOPOLOGICAL OBJECTS

25

to a projective complex variety of the form

= {[ : : ] 2 2 =

+1

( )},

=0

for suitable complex numbers 0 , . . . , +1 (with no two the same).

One can also tesselate the open unit disc with 4-gons whose sides are geodisics in the
hyperbolic metric, and let be the group of holomorphic automorphisms of this tesselation. This
can be done in such a way that each polygon in the tesselation is a fundamental domain and /
is holomorphically isomorphic to .

26

N. P. STRICKLAND

12. Three-dimensional manifolds

In this section we consider compact connected closed oriented smooth manifolds of dimension
three, which we refer to simply as 3-manifolds. Most of the material here comes from John
Hempels book [20].
Let be such a thing, and write = 1 , which is a nitely-presented group. Using Poincar
e
duality and the Hurewicz theorem, we have
0 = 0 =
1 = 2 = /[, ] = 1
2 = 1 = Hom(, ) = 1
3 = 3 = .
Note that 2 is a nitely generated free Abelian group.
12.1. Finite fundamental groups. If = 0 then the above shows that 1 = 2 = 0 and
3 = 3 = (by the Hurewicz theorem). It follows that the generator of 3 is a homotopy
equivalence 3 . In fact, in this situation is always homeomorphic to 3 . This was origi
nally conjectured by Poincar and eventually resolved as a consequence of Perelmans celebrated
e
work on Thurstons geometrization conjecture.

Now suppose instead that is merely nite. Then the universal cover is a simply connected
3
compact closed 3-manifold, and is therefore homeomorphisc to as above. It is interesting to
ask which groups can occur like this. If a nite group admits a map (4) such that

the resulting action on 3 4 is free, then we can take = 3 /. Let be the set of groups
which admit such a map. The identication of is a problem in classical representation theory,
which is completely understood; we will not give the details here.
It is not the case that every 3-manifold with nite fundamental group arises as a quotient of
3 by a linear action of a nite group, as discussed above. Nonetheless, it is conjectured that is
the same as the set of nite fundamental groups of 3-manifolds. To see how one might approach
this, let be a 3-manifold with nite fundamental group and consider the covering spectral
sequence

(; ) = .

As is oriented, the action of on 3 = is trivial. The spectral sequence therefore looks

like this:
u

44
44
44
44
44
4
6

The only possible dierential is 4 . The target ring is concentrated in degrees less than or
equal to 3 (because is a 3-manifold); it follows that 4 : +4 must be an isomorphism

0,3
for all > 0. Let be the generator of 2 = 0 = , and write = 4 () 4 . Using the
multiplicative properties of the spectral sequence, we see that multiplication by is an isomorphism
+4 for > 0. The existence of such an element is a strong constraint on the structure

of the nite group . For example, one can show quite easily that for each odd prime the Sylow
-subgroups of are cyclic.

A BESTIARY OF TOPOLOGICAL OBJECTS

27

12.2. Connected sums. Given two 3-manifolds and , we can form their connected sum
# by cutting small open 3-balls out of and to get two manifolds with boundary 2 ,
and gluing them together along the boundary in a way that respects the orientations. It turns out
that the result is well-dened up to dieomorphism. It is easy to see that 3 # = .
We say that a 3-manifold is prime if it cannot be written as a connected sum of two
manifolds not dieomorphic to 3 . It is natural to try to decompose an arbitrary 3-manifold
as a connected sum = 1 # . . . # , where each is prime. It turns out that this is always
possible, and that the decomposition is essentially unique. Moreover, we have
1 = 1 1 . . . 1
2 = 2 1 . . . 2
1 = 1 1 . . . 1 .
We next attempt a crude classication of prime 3-manifolds. It is easy to see that if 1 is
nite and the Poincar conjecture holds then is prime. If 1 = then it can be shown that
e
= 1 2 , which is prime.

Now suppose that is prime and that = 1 is innite and not cyclic. We claim that
is a simply-connected non-compact 3-manifold, so 1 = 3 = 0,

is contractible. Indeed,

so the only possible reduced homology is 2 = 2 = 2 . If 2 = 0 then a result called

the Sphere Theorem tells us that there is an embedding 2 which represents a nontrivial

element of 2 . It is at least plausible that we could cut along this embedded sphere and
thus write as a connected sum. The example of 1 2 shows that this is a little too naive,
but it turns out that that is essentially the only counterexample, and we have excluded it by our

conditions on 1 . Thus, we must have 2 = 0 and thus = 0 and thus is contractible

as claimed. This implies that is homotopy equivalent to the classifying space .
We now return to the case of a general 3-manifold , decomposed as 1 # . . . # say, with
each prime. It is no longer the case that the higher homotopy groups of vanish, but we can
still calculate 2 . For simplicity, we assume that each group 1 ( ) is innite and not cyclic.
Note that 2 has a natural action of = 1 , so it is a module over the group ring []. For
each we can choose a based embedding : 2 which separates from the other s,

and we also write for the corresponding element of 2 . It turns out that
2 = []{1 , . . . , }/(1 + . . . + ).
For a proof, see [42].
12.3. Swarups homotopy classication. Let be the category of 3-manifolds with a given
basepoint. The morphisms from to are homotopy classes of pointed maps which have degree
one (in other words, we require that [ ] = [ ] 3 ).
Let be the category of pairs (, ), where is a group and 3 . The morphisms from
(, ) to ( , ) are homomorphisms : such that ( ) = .

Given a manifold , there is an obvious map : 1 , and we can dene

= [ ] 3 1 . We thus get an object = (1 , ) of , and it is easy to see that

this gives a functor : . Swarup proved that this functor is full [43]. It follows that

is isomorphic to if and only if is homotopy equivalent to , by an orientation preserving

equivalence. In other words, is a complete invariant of the homotopy type of .
12.4. Heegaard splittings. An important way of constructing 3-manifolds is via Heegaard splittings. Let be an oriented surface of genus > 0, and the usual solid with = . Write
for the inclusion , and let : be an orientation-preserving dieomorphism. We can

then form a pushout diagram

w .

28

N. P. STRICKLAND

It is not hard to check that is a 3-manifold. Such a decomposition of a given 3-manifold is

called a Heegaard splitting of genus . It can be shown that any 3-manifold admits such a splitting
for some .
Note that
1 = 1 , . . . , , 1 , . . . [1 , 1 ] . . . [ , ] = 1
1 = 1 , . . . ,

(and = 0)

1 = 1 , . . . , (

) ( ) = 1.

12.5. Surgery on knots. Another important construction of 3-manifolds is by surgery on knots

and links. We explain a version which creates homology 3-spheres, in other words manifolds
with = (, 0, 0, ) = 3 . (It is equivalent to say that 1 is perfect.)
Let be a knot in 3 (that is, a smoothly embedded copy of 1 ). Let be a tubular
neighbourhood of . Then is an oriented 1 -bundle over 1 , so that 1 .
Write = 3 , which is homotopy equivalent to 3 . By Alexander duality we have

= 2 , so = concentrated in degree one. Moreover, the Mayer-Vietoris sequence

0 = 2 3 1 1 1 1 3 = 0

shows that 1 = 1 1 = 2 . There is thus a basis 1 = {, } such that 0

in 1 and 0 in 1 . Here is parallel to , and is a bre of , so it winds once

around .
For any integer , there is an automorphism of 1 1 that xes and sends to
+ . We can form a pushout diagram as follows:

w .

In other words, is obtained from 3 by removing the solid torus , twisting it times around
and gluing it back in. As = , the Mayer-Vietoris sequence for = shows that
= 3 as claimed.
Any 3-manifold may be obtained from 3 by a related procedure in which is allowed to be
a link and more general twisting is permitted.
12.6. Hyperbolic 3-manifolds. Write this
12.7. Other constructions. Other important classes of 3-manifolds include bre bundles

1 , with a surface (so = [0, 1]/((, 0) ( , 1)) for some automorphism of

). Similarly, one has bundles 1 , which are classied by 2 . More generally,

one can start with such a bundle and perform a certain kind of surgery on a nite set of bres to
obtain a so-called Seifert-bred manifold.

A BESTIARY OF TOPOLOGICAL OBJECTS

29

13. Simply-connected four-dimensional manifolds

In this section we consider compact connected closed oriented topological manifolds of dimension
four, which we refer to simply as 4-manifolds. The following result gives a complete classication
of simply-connected 4-manifolds up to oriented homotopy equivalence. It is essentially the same
as [22, Theorem 2.1].
Theorem 13.1. Let be a simply-connected 4-manifold. Then
0 = 4 = 0 = 4 =
1 = 3 = 1 = 3 = 0.
Moreover, 2 is a nitely-generated free Abelian group. There is a natural unimodular symmetric bilinear form on 2 . The set of oriented homotopy types of 4-manifolds bijects with the
set of isomorphism classes of nitely-generated free Abelian groups equipped with a unimodular
symmetric bilinear form.
We now sketch the proof of this fact. Let be a simply-connected 4-manifold. By Poincar
e
duality and the Hurewicz theorem we have
0 = 4 =
1 = 3 = 0
2 = 2 .
We next consider the universal coecient sequence
Ext(1 , ) Hom( , ).
Using this we nd that
0 =
1 = 0
2 = Hom(2 , )
It follows that 2 is a nitely-generated torsion-free Abelian group, and thus isomorphic to
for some . We have already seen that 2 = 2 , so 2 = also.
The intersection form is dened by (, ) = , [ ]. This is easily seen to be a symmetric
bilinear form, and the Poincar duality isomorphism 2 2 tells us that is unimodular.
e
Clearly, if : is an orientation-preserving homotopy equivalence of simply-connected 4
manifolds then induces an isomorphism ( 2 , ) ( 2 , ). We thus get a map from the
set of oriented homotopy types of simply-connected 4-manifolds to the set of isomorphism types
of free Abelian groups with a unimodular form.
Now let be the result of removing a small open ball from . We have a cobration

3 4,

in which the last map has degree one. It follows that is concentrated in degree 2, where

it is isomorphic to 2 = 2 = say. Choose maps 1 , . . . , : 2 giving a basis for

this group. It is easy to see that the resulting map : = 2 . . . 2 is a homotopy

equivalence. Using this, we conclude that

3 = {[ , ] < } { }.
Here [ , ] denotes the Whitehead product, and : 3 2 is the Hopf map. In particular, we

have

[ , ] +

=
<

for some integers . One can show that the matrix ( ) is the same as the matrix of the bilinear
form written in terms of the basis dual to { }. It follows that determines the homotopy type
of .

30

N. P. STRICKLAND

Conversely, suppose that we start with a unimodular form on . We can then let denote

the inclusion of the th summand in = 2 . . . 2 and dene a map = < [ , ] +

: . We let be the cobre of . It is easy to check that is a nite CW

complex which looks homologically like a simply-connected 4-manifold with intersection form .
It can be shown that there is a topological manifold homotopy equivalent to . This essentially
completes the proof of the theorem.
It turns out that there are far fewer smooth 4-manifolds than topological 4-manifolds. Work of
Seiberg and Witten (which simplies an earlier proof due to Donaldson) shows that the symmetric
bilinear forms arising from smoothable 4-manifolds are all diagonalisable over , which rarely the
case for an arbitrary form. This is the end of a long story, which starts by considering how
quantum electrodynamics would work if spacetime were dieomorphic to a given 4-manifold .
For Donaldsons theory, see [15]. I dont know an appropriate reference for the Seiberg-Witten
theory.

A BESTIARY OF TOPOLOGICAL OBJECTS

31

14. Moore spectra

Let be an odd prime. Let be the category of -local spectra such that = 0 for
< 0 and = 0 for > 0. Such spectra are called Moore spectra. Note that if is a Moore
spectrum then the Hurewicz theorem gives 0 = 0 and = 0 for = 0. Let be the
category of -local Abelian groups, so that 0 .
Theorem 14.1. The functor 0 : is an equivalence of categories. If , then

[, ] = Hom(0 , 0 )
[, ] = Ext(0 , 0 )
[, ] = 0

for > 1.

To prove this, let be a

-local Abelian group. Choose a set { } of generators of , and
thus an epimorphism = () , with kernel say. As is a subgroup of a free module
over () , it is free. (This is well-known if is nitely-generated. Essentially the same proof works
in general, except that one has to choose a
well-ordering of and use transnite recursion.) We
may therefore choose an isomorphism = () and thus a short exact sequence

()
() .

0
()

0
() ,

Dene =
and =
so we have given isomorphisms 0 = and
0 = . For any -local spectrum , the universal property of the wedge gives

0
[, ] =
[() , ] = Hom(, 0 ).

In particular, we have a map whose eect on 0 is the given map . Write

for the cobre of this map. It is not hard to see that and that there is a canonical
isomorphism 0 = . This shows that the functor 0 : is essentially surjective.

From the dening cobration and the fact that 1 = /2, we nd easily that 1 = /2,
which is zero because is -local and is odd.
Consider again an arbitrary -local spectrum . As above, we have [, ] = Hom(, 0 )
and similarly [, ] = Hom(, ) and [, ] = Hom(, 1 ) and [, ] = Hom(, 1 ).
Using this and the cobration
,

we obtain a short exact sequence

Ext(, 1 ) [, ] Hom(, 0 ).
In particular, if then there exists a map (0 ) inducing the identity on 0 = 0 ,

and thus on all homology groups (because the other ones are zero). The map is thus an equivalence.
It follows that all spectra in have the form for some .
Given that 1 = 0, the above short exact sequence shows that [, ] = Hom(, ), and
thus that [, ] = Hom(0 , 0 ) for all , . This shows that the functor 0 :

is also full and faithful, and thus an equivalence of categories. The other claims are now easy to
check.
14.1. References. For this and much other material, the books [26] and [2] are good references.

32

N. P. STRICKLAND

15. Eilenberg-MacLane spaces

Consider an integer > 1 and write
= { pointed CW complexes such that = 0 for all = }.
We make this into a category, with homotopy classes of maps as morphisms. We also consider the
category of Abelian groups.
The following theorem is proved in [44, Section V.7] (for example):
Theorem 15.1. The functor : is an equivalence of categories. If is a CW complex

and then there is a natural isomorphism [, ] = (; ).

It follows that for any Abelian group there is an essentially unique CW complex (, )
equipped with an isomorphism (, ) = . This is called the Eilenberg-MacLane space of
type (, ). We also have
[, (, )] = (; )
[(, ), (, )] = Hom(, ).
There are a number of ways of constructing (, ).
(a) Let be a topological group which is a CW complex. Then Milgrams classifying-space
construction [25, Chapter 1] gives a CW complex which is a covariant functor of ,
together with a natural map which is a homotopy equivalence. Moreover, there

is a natural homeomorphism ( ) = .
If is an Abelian topological group then the addition map : is a group

homomorphism so we can apply to get a map of spaces = ( ) .

It is easy to see that this makes into an Abelian topological group with = .
Thus, given a discrete Abelian group , we can dene (, ) = ; it is straightforward to check that (, ) = concentrated in degree , as required.
(b) For any pointed space , we dene the th symmetric power SP = / . The
map +1 sending (1 , . . . , ) to (1 , . . . , , 0) induces an inclusion SP

SP+1 , so we can dene SP = lim SP . This is also the free Abelian topological

monoid generated by modulo one relation, which sets the basepoint equal to the identity. In other words, for any Abelian topological monoid we have a bijection between
continuous monoid maps SP and pointed maps .

The Dold-Thom theorem [14] says that if is a connected CW complex then there is
a homotopy equivalence

SP =
( , ).
>0

In particular, we have
(, ) = SP ( ).
We outline a proof of the Dold-Thom theorem. There is a well-known homeomorphism

SP = , which sends [1 , . . . , ] to the list of coecients of the polynomial ( ).

This restricts to give a homotopy equivalence
SP 1 SP 1 1 .
In the limit we get SP 1 = 1 = (, 1) (which is a special case of the theorem). One
can also show that the functor SP converts cobrations to quasibrations, and thus that
the functor SP is a homology theory. The rest is general nonsense.
(c) There is an essentially unique space equipped with a map that induces

an isomorphism ( ) ( ). If is odd then a theorem of Serre tells us that

= , concentrated in degree . Thus = (, ) when is odd.
We recall the construction of : it is the telescope of the sequence
2

. . . .

A BESTIARY OF TOPOLOGICAL OBJECTS

33

Here we write : for any map of degree . The telescope of any sequence

0
1
0 1 . . . is the space [, + 1] / , where ( + 1, ) is identied with

( + 1, ) for all . One can show that Tel( ) = lim , and that

Tel( ) = lim ( ).

There is a well-developed and satisfying theory of spaces whose homotopy groups are
all rational vector spaces. The original reference is [35], and there are some more recent
survey articles in [18].
(d) By methods similar to those of Section 14, one can construct a simply-connected space
with = , concentrated in degree . It follows that = 0 for < ,
and that = . However, is not an Eilenberg-MacLane space, because the
homotopy groups for > will not vanish in general. Nonetheless, these higher
homotopy groups can be killed o by attaching extra cells of dimension greater than ,
and after doing this we are left with an Eilenberg-MacLane space. This sort of procedure
is explained in [44, Section V.2].

34

N. P. STRICKLAND

16. Wilson spaces

Let be an odd prime, and > 0 an integer. In this section we discuss the properties of certain
space which was rst studied by Steve Wilson [45, 46]. See also [8, Sections 22-23].
The space is a ( 1)-connected -local -space, such that for all the groups
and are nitely generated free modules over () . Moreover, the rst nontrivial group is
= = () .
The space is also atomic: if : is such that : ( ; ) ( ; ) is nonzero,

then is a homotopy equivalence. It follows that is indecomposable: it cannot be written as

with and both not contractible.
Wilson proved that the above facts characterise uniquely: if is an indecomposable ( 1)connected -local -space of nite type such that and are free over () and = () ,
then is (noncanonically) homotopy equivalent to .
One construction of the spaces is that = , where is characterised by
< ( 1) +1 , where = 2( 1).
It follows that the space can be made into a ring up to homotopy, and we have
+ = = () [ 0 < < ( 1)].
16.1. Bipolynomial Hopf algebras. The Hopf algebras 2 and 2 (with coecients
() ) are both polynomial rings; in other words, is a bipolynomial Hopf algebra. The
structure theory of such Hopf algebras is known [38]; we will give a brief outline.
There is a unique Hopf algebra structure on = () [ 0] such that the elements

= + are primitive. Indeed, we have = [ 0], so it is clear that there

is a unique coproduct on such that the elements are primitive. We need only check
that the elements ( ) actually lie in . This can be done by induction on
. We call with this coproduct the Witt Hopf algebra. It can be shown that is self-dual,
and thus bipolynomial.
Given any integer > 0, we can make into a connected graded Hopf algebra by putting
in degree 2 . We call this graded algebra (). It can be shown that any connected
graded bipolynomial Hopf algebra of nite type is noncanonically isomorphic to a tensor product

for certain (uniquely determined) integers 0.

>0 ()
In the case of , the integers are implicitly given by

= () ( )

where
() =

(1 1 )1 ,

where the product runs over all > 0 for which 1 < ( 1).
16.2. Torsion-free spaces. Let be a connected -local CW complex of nite type, and suppose
that and are free over () (or equivalently, torsion-free). Suppose also that is a
product of Eilenberg-MacLane spaces (which follows from the other assumptions if is an

space). Then it can be shown that >0 for certain uniquely determined integers 0.
Using this, we nd that the connected components of the innite loop spaces in the spectra
, , , , and () are all -locally products of Wilson spaces.

A BESTIARY OF TOPOLOGICAL OBJECTS

35

17. The building for

Let be a nite partially ordered set. Recall that a subset is called a chain i it is
totally ordered, in other words for all , we have either or . Given a map
: [0, 1], we write supp() = { () > 0}. Write

= { : [0, 1] supp() is a chain }.

This is a simplicial complex with vertex set . It has one -simplex for each chain of length + 1.
For more details, see [6, Chapter 6].
Let be a vector space over of nite dimension , and write
Sub( ) = { proper nonzero subspaces of }.
This is a partially ordered set under inclusion, so we can form the simplicial complex Sub( ).
This is called the Bruhat-Tits building for the group Aut( ) .
The following theorem is due to Quillen, and is also proved as [6, Theorem 6.8.5].
Theorem 17.1. Sub( ) is homotopy equivalent to a wedge of (1)/2 copies of the sphere
2 .
Note that Aut( ) acts on Sub( ) and thus on = 2 ( Sub( ); () ). Thus is a module
over the group ring = () [Aut( )], called the Steinberg module. It crops up in a number of
places in topology: see [30, Section XX.6] for a connection with the Segal conjecture, and [33] for
an application to the stable homotopy theory of nite complexes.
Before discussing the properties of , we recall some facts about . Let be the subgroup
of matrices with ones on the diagonal and zeros below it (so = (1)/2 ). Equivalently,
if and only if ( ) = and acts as the identity on /1 , for 1 .

It is well-known that
= ( 1)( ) . . . ( 1 ) = (1)/2 ( 1)(1 1) . . . ( 1).
(Consider the number of possible choices for ( ) after (1 ), . . . , (1 ) have been chosen.)
Theorem 17.2.
(a) is free of rank one as a module over () [ ].
(b) is an indecomposable projective module over () [ ]; in fact, there is an idempotent () [ ] such that () [ ].
(c) is a simple module over [ ].
Proof of Theorem 17.1. The claim is trivial when = 2, so suppose that > 2. Choose a line
< , and let be the set of hyperplanes < such that = . If 0 then any
is the graph of a unique map 0 , and thus = Hom(0 , ) = 1 .

Next, observe that a map : of nite posets induces a map : . Explicitly,

we have

()() =
().
()=

If 0 , 1 : are two such maps and 0 () 1 () for all then there is an obvious way

to construct a map of posets {0, 1} , and one can check that {0, 1} = [0, 1] ,

so 0 and 1 are homotopic.

Now take = Sub( ) . Dene , , : by

( ) =
( ) = +
( ) = .
We then have , so the identity map is homotopic to the constant map , so is
contractible. We thus have a homotopy equivalence Sub( ) Sub( )/ .
For each , write = { Sub( ) : () > 0}. These spaces are disjoint, and their
union is Sub( ) . Moreover, one can see that = (0, 1] Sub()/ , where all points of

the form (1, ) are identied together. Using this, we see that Sub( )/ = Sub().

36

N. P. STRICKLAND

Here denotes the unreduced suspension [0, 1] / , where all the points (0, ) are identied
to one point, and all the points (1, ) are identied to a dierent point. By induction, we know

that Sub() is a wedge of (1)(2)/2 copies of the sphere 3 . As 3 = 2 and

(1)(2)/2
(1)/2
(1)/2

=
, we see that Sub( ) is a wedge of
copies of 2 , as
claimed.

We now take = , and write for the evident copy of in , so we have a ag 0 =

0 < . . . < = . Let be the chains 1 < . . . < 1 such that = for all
. By examining carefully the induction in the above proof, we see that Sub( ) is homotopy
equivalent to the space obtained by collapsing out all simplices not contained in . Moreover,
and its complement are invariant under the subgroup . It follows that is -equivariantly
isomorphic to () {}. On the other hand, one can check that permutes freely and transitively,
so that () {} () [ ]. This proves the rst part of Theorem 17.2. For the second part, we
simply record the formula

1
()[] () [ ].
=
: ,
Here runs over < , and runs over the group = { ( ) = for all }.
We shall not check here that this has the properties claimed.

A BESTIARY OF TOPOLOGICAL OBJECTS

37

18. Looped spheres

Let > 0 be an integer, and consider the loop space 2+1 .
18.1. Homology and cohomology. The adjunction [, ] = [, ] shows that 2+1 =
+1 2+1 . In particular, the identity element of 2+1 2+1 gives a map : 2 2+1 ;

explicitly, we have ()() = . The Hurewicz image of is a class 2 2+1 .

Proposition 18.1. is a primitive element, and 2+1 = [] (as an algebra under the
Pontrjagin product).
Proof. Write for the usual generator of 2+1 2+1 , so that 2+1 = []. The bration
2+1 2+1 2+1 gives a Serre spectral sequence

[] 2+1 = .
The 2 term is concentrated in columns 0 and 2 + 1, so the only possible dierential is
0,
2+1 : 2 2
+1,2 .

As 2+2 = = 2+1 = , this dierential must be an isomorphism for > 0. By

induction on degree, we see that 2+1 = { 0} with = 2 and = 1 and
0

2+1 ( ) = for > 0.

1
As +1 is a derivation, we see by induction on + that = (, ) , and thus that

+
2+1 is the divided-power algebra [ ]. Moreover, is primitive for dimensional reasons,
1
1
so that

( ) = ( ) /! =
( ) /! ( ) /! =
.

1
1
1

=+

=+

We next consider the dual Hopf algebra 2+1 . By the Hurewicz theorem, we know that
2 2+1 is freely generated by , so we must have , = 1. It follows easily that , =
1

, and thus that 2+1 = [].

We could also have proved the above theorem using the Eilenberg-Moore spectral sequence [31]

Tor,

2+1

(, ) = 2+1 .

To compute the Tor groups, we consider the dierential graded algebra = [ ] [] with
1
dierential
(( )[] ) = ( )[1]
1
1
(( )[] ) = 0.
1
(Note that this is a derivation.) There is an evident augmentation : . Clearly is a

resolution of by free []-modules, so Tor[] (, ) is just ( [] ) = [ ], with ( )[]

1
1
in bidegree (, (2+1)). As the 2 page is so sparse, there is no room for dierentials. As before,
we conclude that 2+1 = [ ].
1
18.2. The James model. Let 2 be the free monoid generated by the set 2 , modulo the
relation that identies the basepoint of 2 with the identity element of the monoid. There is an
evident map ( 2 ) 2 , and we write 2 for its image. We topologise 2 as a quotient

of ( 2 ) , and 2 as the colimit of the spaces 2 . This makes 2 into the free topological
monoid generated by 2 modulo the basepoint, so that for any topological monoid there is a
natural bijection between maps 2 of monoids, and pointed maps 2 of spaces.

We would like to apply this with = 2+1 , but unfortunately this is only a monoid up to
homotopy. Th x this, we consider instead the measured loop space
2+1 = {(, ) 0, : [0, ] 2+1 , (0) = (1) = }.

38

N. P. STRICKLAND

See [44, Section III.2] for more detailed discussion of this space. It can be given a topology in a
natural way. It is a topological monoid with product (, ).(, ) = ( + , ), where
{
()
if 0
() =
( ) if + .
One can also show that 2+1 is homotopy equivalent to 2+1 . The unit map : 2

2+1 2+1 thus extends to give a map 2 2+1 2+1 of -spaces.

It is not hard to see that 2 = 2 4 . . . and thus that 2 = [] = 2+1 ,

and thus that our map 2 2+1 is a weak equivalence. In fact, it can be shown that

2+1 has the homotopy type of a CW complex, and thus that our map is an actual homotopy
equivalence. All this is discussed in more detail in [44, Section VII.2].
18.3. The James-Hopf maps. We can use the James model to construct maps out of 2+1
by combinatorial means. The most important example is the James-Hopf map : 2+1

2+1 . Consider an element 1 . . . 2 , so 2 (we allow = ). For any set

= {1 < . . . < } {1, . . . , } of order , we write = 1 . . . 2 . We order the
collection of all such subsets lexicographically, to obtain a list 1 , . . . , say. Write
(1 . . . ) = 1 . . . 2 .
One can check that this is well-dened (it does not change if we drop some s with = ) and
that it gives a continuous map 2 2 . It is easy to see that sends 1 2 to the

basepoint, and that the induced map 2 = 2 /1 2 2 is the usual inclusion.

It follows that = 2 2 , and thus that

1
()!
2 2 .
!(!)

The James-Hopf maps are discussed in [5, Section II.2]; this book also contains much other
material of the same kind.
18.4. Fibrations. We can use the James-Hopf maps to construct some interesting -local bra
tions. Let be an odd prime, and let : 2+1 2 +1 be a map which induces a injection

in mod cohomology. The formulae of the previous section show that this applies when = ,
or more generally when = 1 . . . with 1 + . . . + = .

Proposition 18.2. There is a -local bration 1 2 2+1 2

+1

Proof. Let be the homotopy bre of . Because 1 2 = 2 . . . 2( 1) is (2 1)connected, we see that is null on 1 2 . Thus, the inclusion 1 2 2+1 factors

(uniquely) through a map : 1 . We claim that is a -local equivalence.

To see this, consider the Eilenberg-Moore spectral sequence (with mod coecients)

Tor

+1

( , 2+1 ) = .

Note that 2 +1 = [ ], with = 2 . It is well-known that such a divided power

1
1
algebra over is isomorphic to a tensor product of truncated polynomial algebras [] = []/ .
Specically, we have

2 +1 =
[ ]
with = 2+

and similarly
2+1 =

[ ]

with = 2 .

As is injective, we see that

2+1 = 2

is a unit multiple of + . It follows that

+1

=0

[ ] = ( 2

+1

){ 0 < }.

A BESTIARY OF TOPOLOGICAL OBJECTS

39

As this is a free module over 2 +1 , we see that the 2 page of the Eilenberg-Moore spectral

sequence is concentrated on the line = 0, where we just have { 0 < }, which is

2
2
isomorphic to 1 . It follows easily that our map 1 is an equivalence in

mod cohomology, as required.

Related methods give a -local bration

21 1 2 21 .

I think that this is explained in [34]. These brations, together with the ones in the proposition,

give a maze of interlocking exact sequences relating the homotopy groups () and 1 () ,
which can be assembled into the so-called EHP spectral sequence. This is an eective tool for

calculating () . For more discussion, see [36].

40

N. P. STRICKLAND

19. Doubly looped spheres

We now consider the double loop space 2 2+1 . Let be an odd prime. All our (co)homology
groups will have mod coecients.
19.1. Homology. We will outline a proof of the following theorem:
Theorem 19.1. There are elements 2 1 2 2+1 (for 0) and 2 2 2 2+1
(for > 0) such that
2 2+1 = [ > 0] [ 0].
We remark that much stronger results are known; in fact, Fred Cohen has described
as a functor of (see [11]). Our discussion follows Cohens, but is much simpler because we
restrict attention to the case = 2 and = 21 .
There is an evident unit map : 21 2 2+1 , which carries a homology class 0

21 2 2+1 .
We next need to dene a certain natural (nonadditive) operation : 21 2 21 2 ,

which will help us to produce all the algebra generators for 2 2+1 .
Consider a list 1 , . . . , of distinct points in , and write = min= /3. Consider also
a list 1 , . . . , of pointed maps 2 , for some space . We dene a new map : {} =

2 as follows. Let be the closed disc of radius about . Dene () = 0 for all

outside , and let the restriction of to be the evident composite

/ 2 .

One can check that this gives a continuous map

= (1 , . . . , ; 1 , . . . , ) : 2 ,

and moreover that this construction gives a continuous map

: (2 ) 2 .

This construction is probably due to Boardman, but has been elaborated and extended by many
people. See [3] for a pleasant survey of these ideas.
We next need to construct a certain singular chain 1 . Consider the point =
(0, 1, . . . , 1) , and let be the convex hull of its orbit under the action of . This is

a cell of dimension ( 1), contained in the hyperplane = ( 1)/2. Of course, is not

contained in . However, it turns out that one can choose a map : with () = 0

such that the map () = + () gives an embedding : . Moreover, if is obtained

from by identifying points in that lie in the same orbit under , then the natural map
is a homotopy equivalence. All this can be done completely explicitly, using a certain

combinatorial triangulation of . In particular, one can show that the degree of acting on
= 2 is the signature ().
We let denote the image of under the map mentioned above. Consider a cycle
21 2 . One can consider as a chain on (2 ) . The action of on

is free, so we have = (). for some chain 2 . It follows that

( ) =
() . = = 0 (mod ).

Here we have used the fact that . = () , because is odd.

This shows that is a cycle, dening a homology class in dimension (1)+(2 1) =
2 1. We can thus dene
() = [ ] 21 2 .
One can check that this only depends on the homology class of , so we get a well-dened operation
: 21 2 21 2 .

A BESTIARY OF TOPOLOGICAL OBJECTS

41

We will need to know how this interacts with the homology suspension map : 1 =
, which is induced by the evaluation map . A special case of the Kudo

transgression theorem says that for any 21 2 we have

() = () 2 .
Recall that 0 was dened as the obvious generator of 21 2 2+1 . For > 0 we dene
= 0 2 1 2 2+1
= 0 2 2 2 2+1 .
Note that (0 ) is the usual generator of 2+1 = [], so the Kudo theorem tells us that

( ) = . The usual relation between and dierential in the Serre spectral sequence for

the bration 2 2+1 2+1 2+1 shows that survives to the 2 th page and

that 2 ( ) = . Related arguments show that 2 (1) ( (1) = . We can thus set

up a model spectral sequence with

2
= [ ] [ ] [ > ]

2 ( ) =

2 (1) (

(1)

(1)

= [

+1

[ > ] [ > ]{1,

(1)

) = .

If denotes the Serre spectral sequence, then we get a map of rst quadrant spectral
2
sequences. This is an isomorphism on the bottom line at , and also on the whole page. A
standard comparison result now tells us that it is an isomorphism on the whole 2 page, and we
conclude that
2 2+1 = [ > 0] [ 0]
as claimed.

19.2. The Snaith splitting. In this section, we outline a proof that 2 2+1 splits stably as
a wedge of nite spectra. Our proof follows [12]. We start with the following combinatorial
approximation to 2 2+1 . Let 21 denote the set of (discontinuous) maps : 21

such that there are at most points for which () is not the basepoint. There is an
evident surjective map ( 21 ) 21 : it sends a point [1 , . . . , ; 1 , . . . ] of

to the map : 2 with ( ) = and () = 0 for all points other than s. We

the
topologise 21 as a quotient of ( 21 ) . We also write 21 = 21 ,
and we topologise this as the direct limit.
The map : (2 ) 2 that we discussed earlier, together with the unit map

: 21 2 2+1 , gives a map : 21 2 2+1 , and in the limit a map : 21

2 2+1 . It can be shown that this is a homotopy equivalence [29].

We next write for the quotient 21 /1 21 . One can see directly that
= ( )+ (21) .
This is also the Thom space ( )(21) , where is the vector bundle over .
Moreover, the bundle 2 = = is trivial. To see this, let be the space of
complex polynomials of degree at most , and note that the map sending

([1 , . . . , ], ) to [1 , . . . , ; (0 ), . . . , ( )] is an isomorphism (by Lagrange interpolation, for

example). We conclude that
= 2(1) ( ) .
We next show that the ltration of 21 by the spaces 21 is stably split. We rst
generalise the denition of 21 : for any space and any pointed space we let (, )
denote the set of discontinuous maps : such that () = 0 for all but nitely many points

. Thus 21 = (, 21 ). The same procedure that gave our map 21 2 2+1

gives maps ( , ) . After some minor adjustments we can make these compatible

42

N. P. STRICKLAND

as varies, and thus get a map ( , ) = lim in the limit. (The cleanest

way to make these adjustments is to use the constructions in [40].) It can be shown that this is a
homotopy equivalence when is a connected CW complex.
We now dene a map : 21 ( , ) as follows. We take the polynomial ring []

as our model for , and 21 {} as our model for 21 . There is a natural identication
= {(, ) , = , : 21 } {}.

Suppose that 21 , and let be the nite set points such that () = . For
of
each subset with = , we dene () = ( ), so that []. We then
dene a map : [] by setting ( ) = (, ) for each subset as above, and ( ) =

for all other polynomials . Thus is a point of ( , ), and we dene () = . One can
check that this gives a continuous map : 21 ( , ).

If we identify ( , ) with and take adjoints, we get a map of spectra

: 21 .

It is easy to see that 1 21 = 0 and that the induced map

: = 21 /1 21

is just the identity. Using this, it is not hard to conclude that

2+1
2 +

.
0

(We need to observe here that the connectivity of = 2(1) ( ) tends to as does,
so that the wedge of the spectra is the same as their product.)
19.3. More about homology. One can show that the loop sum map 2 2+1 2 2+1

2 2+1 splits stably as a wedge of maps + , and that the operation sends

to . This is enough to tell us how the Snaith splitting splits the homology ring
2 2+1 = [ > 0] [ 0]. Explicitly, we can dene a weight function on
monomials by = = , and = + . It follows easily from the above that

is the span of those monomials such that = . Note that this is zero unless is 0 or 1

mod , and that +1 = 0 . In fact, there is a -local equivalence +1 = 21

and is -locally stably contractible unless {0, 1} (mod ). The lowest degree generator on

is 1 2(1) .
We can also determine the (co)action of the (dual) Steenrod algebra on 2 3 . There is a

general formula for () (for 21 2 ), which involves the so-called Browder bracket
operation (, ). In the case = 2+1 one can show that vanishes. Given this, the formula
reduces to
{
/
if divides

() =
0
otherwise.

If we write () = 0 and () = >0 then we nd that

( ()) () = ()
( ()) () = () + ()
() = () ( )(())
() = ()(()) = 1 (()).
One can also show that

= /(( ) + ) ,

as modules over the Steenrod algebra . Here 2(1) is dual to 1 . If does not
= /(( ) > /), which means that is a so-called
divide then we have
Brown-Gitler spectrum. A great deal is known about these [9].

A BESTIARY OF TOPOLOGICAL OBJECTS

43

The element 1 2(1) is carried by a stable map 2(1) , which we also call 1 .

Multiplication by 1 gives maps

...
0 2 .

It turns out that the telescope of this sequence is just the Eilenberg-MacLane spectrum .
Now take = 1, so we consider 2 3 . The evident map 3 (, 3) gives a map : 2 3

2
(, 3) = 1 of -spaces, whose composite with the unit map : 1 2 3 is the identity.

Using this, we see that the bre of is an -space, and that 2 3 is homotopy equivalent
to 1 (but not as -spaces). It can be shown that there is a -local sphere bundle over
whose Thom spectrum is the Eilenberg-MacLane spectrum () .
19.4. References. A very important application of the above theory is to the proof of the nilpotence theorem of Hopkins, Devinatz and Smith. See [37] for a discussion of this.

44

N. P. STRICKLAND

20. Homology of Eilenberg-MacLane spaces

Let be an odd prime. In this section, we discuss the mod homology of the Eilenberg-MacLane
spaces ( , ). The key to an ecient computation of these groups is to t them together into
an elaborate algebraic object called a Hopf ring and compute them all simultaneously. This is
discussed in more detail in [47].
20.1. The Hopf ring . We will write for the Eilenberg-MacLane spectrum with =
. The th space in this spectrum is = ( , ). We will need to consider the following
maps:
:

: +
:

: 1
The map induces the addition map in the group = [, ]. The map induces the
multiplication map

= [, ] [, + ] = + .
The maps and are the diagonal map and the constant map.
These maps induce
(a) A product
: + ,

written ( ) = or just .
(b) Another product

: + + ,
written ( ) = .
(c) A coproduct

= :

.
=+

(d) An augmentation

= : .
Note that 0 = [ ] = {[] }. It is easy to see that
[] [] = [ + ]
[] [] = []
[] = [] []
[] = 1.
Next, we consider 1 = . It is well-known that = [] [], where = 1,
= 2, and both and are primitive. It follows that 1 = [] []. The second factor
here is a divided-power algebra, spanned by elements [] (to be thought of as /!) with
(
)
+
[] []
=
[+]

([] ) =
[] [] .
=+

If we write = [ ] then [] is isomorphic as an algebra to 0 [ ], where [] means the

truncated polynomial algebra []/ .
Similarly, we nd that

(, 2) = = [] =
[ ],
0

A BESTIARY OF TOPOLOGICAL OBJECTS

45

where = [ ] 2 . We also write for the image of in ( , 2) = 2 .

We can combine the elements , and using our two products and to get many more
elements of ; in fact, it turns out that we get all of this way. To keep this process under
control, we need to understand some of the properties of these products. The relevant properties
can be summarised by saying that they make into a graded ring in the category of graded
coalgebras (otherwise known as a Hopf ring). More explicitly: and are both associative
products, with units [0] and [1] respectively. The -product is graded-commutative, but we have
= (1) [(1) ]
if and . Both products are compatible with and , in the sense that

( ) =
(1)
( ) = ()(),
and similarly for . The rst equation is written in the usual Hopf algebra notation, with

() = and () = . Finally, we have a distributivity formula

( ) =
(1) ( ) ( ).
(Note that the obvious formula ( ) = ( ) ( ) is both dimensionally inconsistent and
nonlinear in . Instead of the obvious diagonal map , we need to use the coproduct
map , which is induced by the diagonal map of spaces.)
For any sequence of integers = (0 , 1 , . . .) with 0 < and = 0 for 0, we dene

= 0 0 1 . . . .
1

It is convenient to set = [1] [0] if = 0. This ensures that ( ) = 0 and ( ) = 0 for all
. Similarly, given a sequence = (0 , 1 , . . .) with {0, 1} and = 0 for 0, we dene
= 0 1 . . . .
1
0
The main result is that

[ ] [ ].

The subring consists of those factors [ ] for which

+ 2 = , together

with those factors [ ] for which

+ 2 = 1.
The proof uses the Rothenberg-Steenrod spectral sequence (also called the bar spectral sequence)

Tor ( , ) = +1 .
The 2 term is easy to compute (given that is as claimed) using the formulae
Tor ( , ) = Tor ( , ) Tor ( , )

Tor[] ( , ) = [] =
[[] ]
with Tor1,2
0
[]

Tor

( , ) = [] []

with Tor1, and Tor2, .

With more work, one can show that the spectral sequence always collapses, and deduce that
+1 is as claimed.
20.2. Stable homology. We next consider the homology of the spectrum , in other words the
ring

= lim + .

This is called the dual Steenrod algebra. The colimit implicitly uses a map

: + ++1 +1 .

This is just the homology suspension map, which is the same as the map . This kills decomposable elements, which implies that is spanned by the images of the elements

46

N. P. STRICKLAND

and . It turns out that we do not need the latter elements, because = 0 . Let

2 2 and 2 1 be the images of 2 2 and 2 1 respectively.

It turns out that 0 = 1 and that
= [ > 0] [ 0].
This ring has a natural Hopf algebra structure, arising from the map
(11)

= ( ) ( ) = .

For more discussion of this, see [36, Section 2.2] or [2, Part III, Section 12]. It turns out that the
coproduct is

( ) =

=+

( ) = 1 +

=+

20.3. Automorphisms of the additive group. The Hopf algebra [ > 0] < can
be described in the following more conceptual way. For any -algebra , write

() = { nilpotent elements of },
considered as a group under addition. Dene

Aut( )() = { isomorphisms : () (), natural for -algebras }.

One can show that any Aut( )() has the form () = for uniquely determined

coecients , with 0 . We write Aut1 ( )() for the subgroup where 0 = 1.

Given a map : [ > 0] , we obtain an element : 0 ( ) of Aut1 ( )().

This clearly gives a bijection

Rings( [ > 0], ) = Aut1 ( )().

The coproduct on [ > 0] makes the left hand side into a group, the right hand side is a
group under composition, and the coproduct is the unique one for which these two group structures
coincide.
If is a space and is an -algebra, we can dene () = Rings( , ). Because of the

above relationship of [ ] with and Aut1 ( ), it turns out that there is a natural action
)() on the set (). It turns out that for many popular spaces , there
of the group Aut1 (

is a simple and conceptual description of () as a set with action of Aut1 ( )().

20.4. Cohomology. We can dualise the above calculation to describe and . Firstly,

we let () be dual to = >0 with respect to the obvious monomial basis, and let
be dual to . It follows that () is dual to , and that these elements form a basis
for the Hopf algebra (which is called the Steenrod algebra). The coproduct is given by

() =
() ()
=+

= 1 + 1 .
There is an explicit but elaborate formula for the product () (). See [26] for a discussion of
this, and an extensive structure theory of modules over .
Let be the tautological generator of . It can be shown (I think) that is the free

graded-commutative algebra generated by the elements () for which >0 ( + 2 ) <

(note that 0 does not contribute). This element has degree + 0 (2 1) + >0 (2 2) ,
and it generates an exterior (resp. polynomial) algebra if the degree is odd (resp. even). The algebra
is the tensor product of all these exterior and polynomial algebras.

A BESTIARY OF TOPOLOGICAL OBJECTS

47

An alternative basis for is given by the so-called admissible monomials [39]. Write = 0
and = (, 0, 0, . . .). These elements satisfy the Adem relations:
(
)
/

( 1)( ) 1
=
(1)+
+

=0

(1)

=0

( 1)( )

(1)/

=0

(1)

+1

+ +

( 1)( ) 1
1

+ .

A sequence = (0 , 1 , 1 , . . . , , , 0, 0, . . .) is said to be admissible if {0, 1} and

+1 + for all . We write
= 0 1 1 2 . . . .
It can be shown that these elements form a basis for , as runs over the set of all admissible
sequences. Moreover, we have
{
1
2
1 if =
...1

, =
0 otherwise
{
1
2
1 if =
. . . 1
, =
0 otherwise
20.5. References. Find a more precise reference for being a Hopf algebroid. Add a reference
for free unstable modules.

48

N. P. STRICKLAND

21. Things left to do

(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)

(14)
(15)
(16)
(17)
(18)
(19)
(20)
(21)
(22)
(23)
(24)
(25)
(26)
(27)
(28)
(29)
(30)
(31)
(32)
(33)
(34)
(35)
(36)
(37)
(38)
(39)
(40)
(41)
(42)
(43)

Spaces of long knots

The Kummer surface.
Brieskorn manifolds.
The dodecahedral manifold.
Knot complements, the Massey product in the complement of the Borromean rings.
Grassmannians, Stiefel manifolds, (bounded) ag varieties.
Buhstaber-Ray manifolds representing generators for and
Matrices of bounded rank, intersection loci.
Projective unitary groups.
Loop groups, algebraic loops, central extensions, representations.
Toric varieties.
Moduli of stable marked curves of genus zero.
Manifolds of isospectral tridiagonal matrices. (See papers by Gaifullin, Tomei, M. Davis
(Some aspherical manifolds). Fix a monic polynomial () of degree + 1 with distinct
real roots. Let be the space of ( + 1) ( + 1) real symmetric matrices such that
= 0 when > 1 and det( ) = (). Then is a (, 1) and a compact
oriented -manifold. It can be written as a union of 2 permutahedra.)
Springer varieties
.
Projective unitary groups
Spaces related to degeneracy loci
The connective covers .
General things about ; some stable splittings.
Stable retracts of , ([ ])/ , (Mitchells complexes with cohomology free over ())
and ( ) (Je Smiths examples for the smash product theorem).
Stable retracts of and .

The spectra (), (), (), 0 occuring in Nick Kuhns work on the Whitehead
conjecture.
Bruhat-Tits building for .
The contractible spaces , , and the quotient space = .
The space (, ) of linear isometries, and the associated operad.
Spaces of embeddings of manifolds. I seem to remember that Goodwillie and/or Waldhausen have some stu about this. Contractibility of the space of embeddings in .
Some generalities about topology of manifolds: Atiyah duality, Morse theory, connected
sums, handlebodies, elementary ideas about surgery.
Brown-Gitler spectra.
The Brown-Comenetz spectrum .
Map(, ).
Generalised Moore spectra (), /.
Spectra realising small -modules.
, .
, .
, , , and all that.
Surgery spectra, Waldhausen -theory
, (), (), (), () etc.
Tate spectra , , .
(), (), and the nilpotence theorem.
Spectra associated with the Adams conjecture and etale homotopy theory.
Spaces occuring in unstable -periodic homotopy theory and Brayton Grays theory of
cospectra.
and stu about the projective dimension of .
Elliptic spectra, , and so on.

A BESTIARY OF TOPOLOGICAL OBJECTS

(44)
(45)
(46)
(47)
(48)
(49)
(50)
(51)
(52)
(53)
(54)
(55)
(56)
(57)
(58)
(59)
(60)
(61)
(62)
(63)
(64)

49

( ) and related objects.

Spectra occuring in the disproof of the telescope conjecture.
, string topology, the cactus operad.
Floer pro-spectra.
Seiberg-Witten and Yang-Mills moduli spaces.
Teichmller space, and compactications.
u
Spaces of representations of discrete groups in 2 ().
Calabi-Yau manifolds.
Examples from rational homotopy theory.
Groups of homotopy spheres.
Fuchsian groups.
Something about Goodwillie calculus.
Something about -algebras.

1 , Madsen-Weiss
Aut(Free ), outer space, associated -theory spectra.
-compact groups. () as a homotopy colimit of -toral groups, Sullivans classifying
spaces for 23 , -compact groups with Weyl group
Anick spaces, Grays delooping of the bre of the double suspension.
Examples from the theory of LS-category.
Simultaneous conjugacy spaces ( ()ad ) / ().
Hom( , ())
The space of maximal tori in ().

50

N. P. STRICKLAND

References
[1] J. F. Adams. Lectures on Lie groups. W.A. Benjamin, 1969.
[2] J. F. Adams. Stable Homotopy and Generalised Homology. University of Chicago Press, Chicago, 1974.
[3] J. F. Adams. Innite Loop Spaces, volume 90 of Annals of Mathematics Studies. Princeton University Press,
1978.
[4] M. F. Atiyah. -Theory. Advanced Book Classics. Addison Wesley, 1989.
[5] H. J. Baues. Commutator calculus and groups of homotopy classes, volume 50 of London Mathematical Society
Lecture Notes. Cambridge University Press, 1981.
[6] D. Benson. Representations and Cohomology II, volume 31 of Cambridge Studies in Advanced Mathematics.
Cambridge University Press, 1991.
[7] D. Benson. Stably splitting . Bulletin of the American Mathematical Society, 33:189198, 1996.
[8] J. M. Boardman, D. C. Johnson, and W. S. Wilson. Unstable cohomology operations. In I. M. James, editor,
Handbook of Algebraic Topology. Elsevier, Amsterdam, 1995.
[9] E. H. Brown and R. Cohen. The Adams spectral sequence of 2 3 and Brown-Gitler spectra. In W. Browder,
editor, Algebraic topology and algebraic -theory, number 113 in Annals of Mathematics Studies. Princeton
University Press, 1987.
[10] F. Cohen. Artins braid groups, classical homotopy theory and sundry other curiosities. In J. Birman and
A. Libgober, editors, Braids, volume 78 of Contemporary Mathematics, pages 167206. American Mathematical
Society, 1989.
[11] F. R. Cohen, T. Lada, and J. P. May. The Homology of Iterated Loop Spaces, volume 533 of Lecture Notes in
Mathematics. SpringerVerlag, 1976.
[12] F. R. Cohen, J. P. May, and L. Taylor. Splitting of certain spaces . Mathematical Proceedings of the
Cambridge Philosophical Society, 84:465496, 1978.
[13] A. Dold. Lectures on Algebraic Topology. SpringerVerlag, New York, 1980.
[14] A. Dold and R. Thom. Quasifaserungen und unendliche symmetrische produkte. Annals of Mathematics,
67:239281, 1958.
[15] S. K. Donaldson and P. B. Kronheimer. The geometry of four-manifolds. Oxford Mathematical Monographs.
Oxford University Press, 1990.
[16] W. G. Dwyer and J. Spalinski. Homotopy theories and model categories. In I. M. James, editor, Handbook of
Algebraic Topology. Elsevier, Amsterdam, 1995.
[17] A. D. Elmendorf, I. Kriz, M. A. Mandell, and J. P. May. Rings, Modules and Algebras in Stable Homotopy
Theory, volume 47 of Amer. Math. Soc. Surveys and Monographs. American Mathematical Society, 1996.
[18] Y. Felix. Algebraic topology and rational homotopy theory, volume 1318 of Lecture Notes in Mathematics.
Springer Verlag, 1988.
[19] R. Fritsch and R. A. Piccinini. Cellular structures in topology, volume 19 of Cambridge studies in advanced
mathematics. Cambridge University Press, 1990.
[20] J. Hempel. 3-Manifolds, volume 86 of Annals of Mathematics Studies. Princeton University Press, 1976.
[21] D. Husemoller. Fibre Bundles, volume 33 of Graduate Texts in Mathematics. SpringerVerlag, 1975.
[22] R. C. Kirby. The Topology of 4-Manifolds, volume 1374 of Lecture Notes in Mathematics. SpringerVerlag,
1989.
[23] S. O. Kochman. Bordism, stable homotopy and Adams spectral sequences, volume 7 of Fields Institute monographs. American Mathematical Society, 1996.
[24] Kulkarni and Wood. Topology of nonsingular complex hypersurfaces. Advances in Mathematics, 35:239263,
1980.
[25] I. Madsen and R. J. Milgram. The Classifying Spaces for Surgery and Cobordism of Manifolds, volume 92 of
Annals of Mathematics Studies. Princeton University Press, 1979.
[26] H. R. Margolis. Spectra and the Steenrod Algebra. North-Holland, 1983.
[27] C. R. F. Maunder. Algebraic Topology. Cambridge University Press, 1980.
[28] J. P. May. Simplicial Objects in Algebraic Topology, volume 11 of Van Nostrand Mathematical Studies. Van
Nostrand Reinhold, 1967.
[29] J. P. May. The Geometry of Innite Loop Spaces, volume 271 of Lecture Notes in Mathematics. Springer
Verlag, 1972.
[30] J. P. May. Equivariant homotopy and cohomology theory. Published for the Conference Board of the Mathematical Sciences, Washington, DC, 1996. With contributions by M. Cole, G. Comezaa, S. Costenoble, A. D.
n
Elmendorf, J. P. C. Greenlees, L. G. Lewis, Jr., R. J. Piacenza, G. Triantallou, and S. Waner.
[31] J. McCleary. A Users Guide to Spectral Sequences, volume 12 of Mathematics Lecture Series. Publish or
Perish, 1985.
[32] J. W. Milnor. Morse Theory, volume 51 of Annals of Mathematics Studies. Princeton University Press, 1963.
[33] S. A. Mitchell. Finite complexes with ()-free cohomology. Topology, 24(2):227246, 1985.
[34] J. C. Moore and J. A. Neisendorfer. Equivalence of Toda-Hopf invariants. Israel J. Math., 66(1-3):300318,
1989.
[35] D. G. Quillen. Rational homotopy theory. Annals of Mathematics, 90:205295, 1969.
[36] D. C. Ravenel. Complex Cobordism and Stable Homotopy Groups of Spheres. Academic Press, 1986.

A BESTIARY OF TOPOLOGICAL OBJECTS

51

[37] D. C. Ravenel. Nilpotence and Periodicity in Stable Homotopy Theory, volume 128 of Annals of Mathematics
Studies. Princeton University Press, 1992.
[38] D. C. Ravenel and W. S. Wilson. Bipolynomial Hopf algebras. Journal of Pure and Applied Algebra, 4:4145,
1974.
[39] N. E. Steenrod and D. B. Epstein. Cohomology operations, volume 50 of Annals of Mathematics Studies.
Princeton University Press, 1962.
[40] R. Steiner. A canonical operad pair. Mathematical Proceedings of the Cambridge Philosophical Society, 86:443
449, 1979.
[41] J. Stillwell. Classical topology and combinatorial group theory, volume 72 of Graduate Texts in Mathematics.
Springer-Verlag, second edition, 1993.
[42] G. A. Swarup. On embedded spheres in 3-manifolds. Math. Ann., 203:89102, 1973.
[43] G. A. Swarup. On a theorem of c.b. thomas. Journal of the London Mathematical Society, 8:1321, 1974.
[44] G. H. Whitehead. Elements of Homotopy Theory, volume 61 of Graduate Texts in Mathematics. Springer
Verlag, 1978.
[45] W. S. Wilson. The -spectrum for Brown-Peterson cohomology, part I. Commentarii Mathematici Helvetici,
48:4555, 1973.
[46] W. S. Wilson. The -spectrum for Brown-Peterson cohomology, part II. American Journal of Mathematics,
97:101123, 1975.
[47] W. S. Wilson. Brown-Peterson Homology: An Introduction and Sampler, volume 48 of Regional Conference
Series in Mathematics. American Mathematical Society, 1982.