You are on page 1of 18

The completion conjecture in equivariant cohomology by J. P.

May

Consider an RO (G)-graded cohomology theory

kG.

We shall not insist on a

detailed definition; suffice it to say that there is a suspension isomorphism for each real representation of equivariant K-theory
,

G. and

The first examples were real and complex K G. The next example was equivariant stable

KO

cohomotopy theory

~G"

There are RO(G)-graded ordinary cohomology theories with

coefficients in Mackey functors. The study of these theories is still in its infancy. They can all be

defined for arbitrary compact Lie groups, but we shall restrict our attention to finite groups. When we localize away from the order of G, there are very
*

powerful algebraic devices for the reduction of the calculation of nonequivariant calculations. G,
*

kG

to

If we localize at a prime dividing the order of

there are techniques G.

for reducing calculations to consideration of p-groups

contained in kG at p

There are no known general procedures for the calculation of G. Largely for this reason, the reservoir of known

for p-groups

calculations Let

is almost empty. be the Burnside ring of finite G-sets. Then ~ takes values in

A(G)

the category of A(G)-modules. interest in the R0(G)-grading.

For some purposes, this is the ~ain reason for The assertion is false for Z-graded equivariant Let EG be a

cohomology theories which fail to extend to RO(G)-graded theories. free contractible G-CW complex and let e: EG ÷ * = {pt}

be the trivial rap.

We have an induced homomorphism of A(G)-modules

e : kG(*)

~ kG(EG). The completion conjecture

(We use unreduced theories until otherwise specified.) for kG asserts that, on integer gradings, e

becomes an isomorphism upon

completion with respect to the topology given by the powers of the augmentation ideal of A(G). As we shall shortly m~ke precise, when kG, G is a p-group and the kG is computable

completion conjecture holds for in nonequivariant ter~s.

the p-adie completion of

The first theorem in this direction was due to Atiyah and Segal since G [5].

[hi ; see also Atiyah

Their results are stated in terms of representation rings but,

is finite, the Burnside ring gives the same topology.

Theorem i. theory.

The completion conjecture holds for real and complex equivariant K-

621

However, the completion conjecture

certainly

fails to hold in general.

Counterexample H;(?;Z__) gradings,

2.

The completion

conjecture fails for ordinary cohomology coefficient system Z_Z. On integer the ordinary

with coefficients

in the constant whereas

H;(*;Z__) = H~(*;Z__) = Z,

HG(EG;Z ~) = H*(BG;Z), BG.

integral cohomology of the classifying

space

The Segal conjecture theory.

is the completion

conjecture

for equivariant cohomotopy

The central step in its proof has been supplied in a beautiful piece of [61.

work by Gunnar Carlsson

Theorem 3. theory.

The completion conjecture holds for equivariant

stable cohomotopy

There is an equivalent implied generalization

nonequivariant

reformulation

and an interesting

that I will discuss at the end. my instinct was that it was
.

When I first heard about the Segal conjecture, unlikely to be true in general.

I also felt that it was a much less important kG the completion

problem than the general one of explaining for which theories conjecture would or would not hold. However,

Carlsson's work not only completed

the proof of Theorem 3, it also led to very substantial progress on the general problem. This development is Joint work of Jeff Caruso and myself of Carlsson's work 19] and

follows up our simplification

[8], which was undertaken in

hopes of just such a generalization.

MY understanding of these matters also owes with Jeremy Gunawardena, Gaunce

a great deal to Joint work and conversations Lewis, Stewart Priddy, and Stefan Waner.

My purpose here is to explain Carlsson's work,

our generalization

of it, and

related matters in conceptual terms, without getting bogged down in details of proofs. None of the steps presents any great difficulty any more, all of the but there are quite a few steps. We

work lying very close to the foundations,

shall stick to the main line of development,

and this means that a great deal of Adams Ill suuun~rized

earlier work on the Segal conjecture will go unn~ntioned. what was known late in 1980.

We shall first explain the force of the completion conjecture and its reduction to a question about p-groups, For each subgroup associated to inclusion ~ H of G, following May and McClure theory [19]. k~

there is an RO(H)-graded H

(depending not Just on In particular, k*. with

as an abstract group but on the there is an associated nonequivin terms of restriction

HcG).

H = e,

ariant cohomology theory

Modulo interpretation

622

RO(G)--~RO(H),

The projection
w

G

x

Y + y

indtmes a map Y.

~: k*(Y) + k*(Y)
w

of Z-graded

cohomology theories on spaces ~: k (Y) + kG(Y)

We say that

kG

is split if there is a map ~ is the identity. If kG We have

of cohomology theories

such that

a notion of an RO(G)-graded ring-valued cohomology theory. theory, so is each k H. We say that
*

is a ring ~ is a map

is a split ring theory if kH

of ring-valued cohomology theories, and each theory.
.

is then also a split ring
q

We say that

kG

is of finite type if each

kH(*), q ~ Z kG(X) complexes

and

H C G,

is a

finitely generated Abelian group. generated for all ~ s RO(G)

This ensures that G-CW

is finitely X. The most

and all finite

interesting exa~Eoles are split ring theories of finite type. The connection between equivariant and nonequivariant established by the following observation I19, lenm~ 121. cohomology is

Lemma 4. gradings,

If

kG

is split and is natural]y

Y

is a free

G-CW

complex, then, on integer

kG(Y)

isomorphic to

k*(Y/G).

Thus the completion conjecture relates

kG(*)

to

k*(BG). kG(X)

The relevance of is given by the

the conKoletion conjecture to the general calculation of following result ll9, Prop. 151.

Proposition

5.

Assume that for each H C G

is a split ring theory of finite type and that (where kH BH n denotes the n-skeleton of H, BH).

limlk*(BH n) = 0 If the c o ~ l e t i o n EG x X + X

conjecture holds for

for each

then the projection

induces an isomorphism k~(X)

k~(EG × X)
G-CW complexes X, where the left side is

for all

a s RO(G)

and all finite q(X)

the completion of

with respect to the Burnside ring topology.

Of course, by the lemma, kq(EG For clarity of exposition, integer gradings o

x

X) - - kq(EG -

xG

X)

for

q E Z.

we shall henceforward

restrict attention to

The completed Burnside ring Green functor satisfies to the subgroups of G of prime power order.

induction with respect

Don't werry if you don't it leads to a

understand the previous sentence.

As a matter of pure algebra,

623

proof that the Burnside

ring rapidly disappears [14,151.

from the picture.

See [19, Thin

13 and Prop 14] and also Laitinen

Theorem 6. subgroups H

The completion conjecture holds for of prime power order.

kG

if it holds for

kH

for all

Proposition 7.

If

G

is a p-grcup and kG if and only if

~

is split, then the conloletion g*-• kG(*) + kG(EG) * * induces an

conjecture holds for

isomorphism upon passage to p-adic conpletion.

Henceforward groap; k

(until otherwise specified near the end),

G p.

is to be a pTo avoid constant kG,

and co@hate symbols will indicate coni01etion at

repetition of hypotheses, given or constructed,

we assume once and for all that all theories

are split and of finite type.

We are at the starting point of Carlsson's work, and some preliminary philosophical comments are in order. The stable part of algebraic topology has theory on spaces, homotopy theory Carlsson's preprint [6] was written

three main branches:

homology and cohomology

on spectra, and infinite loop space theory. from the second point of view. Specifically,

Carlsson worked in the stable [17]. This was done with

category of G-spectra constructed nV encouragement, advice.

by Lewis and mFself

and I n~st apologise to Carlsson for giving him very bad simplifies considerably when

As Caruso and I discovered, the mathematics

the first point of view is taken, and the changed point of view is crucial to our generalization Carlsson's of Carlsson's work that is the theme of this paper.

theorem asserts that the Segal conjecture holds for all p-groups of the

if it holds for all elementary Abelian p-groups, that is, for all p-groups form (Zp) n. The Segal conjecture was proven by Lin [18] for [13] for Zp, and by Adams, Gunawardena, Z2, by

Gunawardena with n _> 2.

and Miller

[31 for

(Zp) n

These authors actually prove the equivalent nonequivariant of the conjecture to be discussed later. My work with Caruso led for

reformulation

to work with Priddy that gives a geodesic proof of the Segal conjecture (Zp) n within Carlsson's context
*

([20] plus later corrections). Abelian groups

We shall see that a version of the reduction to elementary goes through for all theories that kq(*) = 0 kG such that sn~ll k, q.

is bounded below, in the sense (This hypothesis serves only to

for all sufficiently

ensure the convergence

of certain Adams spectral sequences. )

While we shall be

able to shed some light on the elementary Abelian case, the general picture is still obscure. We shall describe a satisfactory necessary and sufficient k. is cohomologically

condition for the completion conjecture to hold when

624

boanded above,

in the sense that

Hq(k) = 0

for all sufficiently

large

q,

but

without this unpleasant truth". theory Here
*

hypothesis it seems that "calculation

is the way to the

k

denotes the spectrum which represents the nonequivariant

k . to give content to the discussion. Let U be a

We need some definitions

countably infinite dimensional real G-inner product space which contains infinitely ma w U = R~ ~) U', fix copies of each irreducible where U' representation of G. We take and so

contains no copies of the trivial representation

R s C U, s > 0.

By an indexing G-space, U. kGV here

we understand a finite dimensional kG, namely a

G-inner product subspace of collection of based G-spaces ~: z W - V ~ v V of in V; W. --~ kGW for

We assume given a G-prespectrum for indexing G-spaces W-V where V

and based G - m p s

V C W;

denotes the orthogonal complement of SV is the 1-point con~Dactification sV--~X. We require in no X r = 0

As usual,

zVx = X ^ SV, 2Vx

similarly,

is the G-space of based maps kGV and raps ~,

technical conditions

on the spaces

but these result G-CW

loss of generality and need not concern us here. and or Y and for an integer q, write q = r-s,

For based where

complexes and

r > 0, s _> 0,

s = 0

(to avoid separate cases) and define kG(x;Y) V s = [T rx, colim 2 - 2 (Y^kGV)] VDR s G" k G. It specializes to

This is the Z-graded bitheory associated to
and

=

:

These are reduced theories, homotopy and cohomotopy, space is S V, and

to which we switch h e n c e f o r ~ r d . kG

To define stable its

we take

to be the sphere G-prespectrum; is the evident identification.

o: zW-VsV--~ S W

Carlsson's

first step was joint work with Cusick

[7] and involved reduction

from a problem in cohomotopy to a more tractable problem in homotopy. Independently and concurrently, Caruso and Waner arrived at an extremely elegant

way of carrying out essentially the same reduction. [i0] that a model for certain zero in to EG

They had observed earlier V of

could be obtained by taking the union over with boundary M(V) M(V)

smooth compact G-manifolds V.

embedded with codimension M(V)/SM(V) is equivalent

The essential property of where TV

is that Sv
0

sV/T v,

denotes the singular set of
^'~

(namely the set of points
^* +

with non-trivial Whitehead

iso~r~py subgroup).

Using this model and an easy Spaniere: kG(S ) --~kG(EG ) can be

duality argument,

one finds that

identified with a certain natural map of inverse limits

625

A simple cofibration argument then gives the following conclusion.

Proposition

8.

The completion conjecture

holds for

kG

if and only if

lim ~ ( s V S )
For Carlsson's

: 0

second step, it is convenient to introduce the notation

J
Here Y is a G-CW complex, W is a representation of G, and the inverse limit is taken over the h o ~ m o r p h i s m s kG(s(J+llW;y ) q where e: S O ÷ SW

(I^ e).

, kG(sJW^ sW;y^ S W) ~ k~(sJW;Y), q The case Y = S0 is particularly

is the evident inclusion.

important. involving

By clever use of interchange smash products,

of limits applied to bi-indexed limits

Carlsson uses the criterion of the previous proposition

to obtain the following conclusion.

Proposition H of G,

9. then

If the completion conjecture holds for kH for all subgroups ^G k.(W) = 0 for all W # 0. Conversely, if the completion
* kH

conjecture holds for any one W # 0

for all proper subgroups W G = {0},

H

of

G

and

AG

k~(W__) = 0

for

such that

then the completion

conjecture holds for

Carlsson fixes a good choice of Let ~

W

with

WG = 0,

which we shall call of of G and take G. Z

Z. to be

be the elementary Abelianization G

G/[G,G] ~ Zp

the pullback to of Z

of the reduced regular representation H

The restriction

to any proper subgroup

contains a copy of the trivial representation, due to Quillen and Venkov if G is not elementary 126])

and a theorem of Serre

[29] (or its consequence a(Z) is nilpotent

implies that the Euler class Abelian. Proposition

9 has the following consequence.

Corollar~
proper

i0.

Assume that the completion conjecture holds for
H of G. Then t h e completion conjecture

kH

for all
for kG if and

subgroups

holds

only if

~(Z)

= 0.

At this point, Carlsson introduces unreduced suspension of Equivalently, EG EG

the key simplification.

Let

EG

be the

with one of its cone points as basepoint. EG + + S 0, and one obtains

is the cofibre of the evident map

626

the fundamental
^

long exact sequence

...

kO(EG÷

iOCZ q_

" +

....

Carlsson assumes inductively that the Segal conjecture holds for all proper subquotients if the a of G and p r o v e s t h a t both This ( E ° ~ ~) = 0 implies that and g . ( g G ^ ~) = 0 = 0 his and t h u s t h a t vanishing theorems of a direct
-0

is not elementary Abelian. holds for G.

~(~)

Segal conjecture G

CarZsson observes that

fail when

is elementary Abelian and suggests the possibility a

proof that the connecting homomorphism Priddy and I provide such a proof. The generalization associated in N. For to ~ J X

is an isomorphism in this case.

requires us to introduce a bitheory of and G. Y Here and J = N/H, q = r-s where H is normal

to any subquotient J-CW complexes

(as above), we define

kJ(x;y) = Izrx,colim aVH-SS(y ^ (kGV)H]j
q V DR s

and specialize as before to obtain works equally well with general, write J = e) h, ~ depends on ~ H

k~(Y)

and ~.

kj(X).

W

Everything said so far

replaced by and N

It is vital to recognize that, in J. In particular, H = N we

and not Just on

for the nonequivariant

bitheory obtained by taking we write and k. H = G, and k~

(and thus

in the above definition; so obtained from

for the nonequivariant An example may

bitheories

H = e

respectively,

clarify the definition.

Example ii. zVx and with

For a G-space

X,

let

SGX

be the G-prespectrtun with identification. ~: X + XG

V~-

space

a: zW-VzVx --~zWx SGX

the evident

The theory whose composite

represented by

is split if there exists a map xG-*x J is hemotopic VH

with the inclusion representation of

to the identity.

Since every V, we conclude Note the

occurs in (SGX)~

for some indexing G-space

by cofinality that particular case

is just a copy of the bitheory

(sjxH),.

X = SO .

Now Carlsson's vanishing theorems

generalize as follows.

Theorem 12. for kj

If

G

is not elementary Abelian and the completion conjecture holds J of G, then (EG^Z) = O.

for all proper subquotients

Theorem 13.

If

G

is not elementary Abelian and

k,

is bounded below, then

kO.(EG+^ Z_) = 0.

627

We have stated these differently rather than inductive.

since the proof of the second is direct k~ as

The same results are valid if we start with some

ambient theory, and we de4nce the following generalization by induction. finite type. Remember that all theories

of Carlsson's theorem

in sight are assumed to be split and of

Theorem 14.

If

G

is not elementary Abelian,
*

all

h.

are bounded below, and

the completion conjecture holds for of G,

kj

for all elementary Abelian subquotients kG and all other kj.

then the completion conjecture holds for

Before discussing the proofs of Theorems 12 and 13, we describe what happens ^ pn(n-l)/2 in the elementary Abelian case. Let M n denote the free Zp-module on generators, over Zp where ~
P

denotes the p-adic integers;

we take tensor products

belcw.

In the case of stable homotopy,

the following theorem is more

or less implicit in Carlsson's work. n~self.

The general case is due to Caruso and

Theorem 15.

If

G = (Zp) n J of

and the completion conjecture holds for G, then M n ~) Z l-n k~(S0).

kj

for all

proper subquotients

kG('EGAZ) ~

The following theorem is due to Priddy and nvself, although most of the work is in an Ext calculation due to others and discussed belcw.

Theorem 16. above, then

If

G = (Zp) n

and

k.

is bounded below and cohomologically

bounded

^G k.(EO +A Z_) % Mn ® z-n k.S O. In the absence of the bounded above hypothesis, there is an inverse limit of Adams spectral sequences such that E 2 = ExtA(H*(BG)[a(z)-ll and {E r} converges to ® H*(k),Zp)

k,a(Ea ^ Z).
e(Z) e H2(pn-I)(BG) has an obvious explicit

+

Here the Euler class description. At this point, that kG,

is a split ring theory with unit

the virtues of naturality ~anifest themselves. Suppose O G e,: ~, -J* k,. We have the following

628

commutative

square. ^G ~ ~,(EG^Z) a ^G + , ~,(EG ^ Z) [ e,
a

e.

To de@ace the Segal conjecture for G = (Zp) n, it suffices^ to^ find a theory for which the completion conjecture holds and mod p. Indeed, ~ on the top is a morphism of e*:~o(S0) ÷ k'(S0)v

kG is non-trivial and.

~o(sO)-~e s 91

whose d o m i n

target are each freely generated by a copy of the free Zp-mOdule M n.

Given kG,

Corollary 10, a bit of calculation along the lines of Theorem 15, and a chase of the diagram show that ~ restricts to an isomorphism between the respective copies of M~and is therefore an isomorphism. In [20] , we thought that equivariant

K-theory would do for kG, but Costenohle has since proven the astonishing fact that in this case --k,(S = 0. 0) kG(X) = HG(EG × X;ZD). * With k~ A choice which does work is specified by

The completion conjecture holds trivially for theories so
^ is one of k,(S O )-

constructed by crossing with EG. a split ring theory, the exact sequence (*)

modules and the isomorphisms of Theorems 15 and 16 are isomorphisms of ~,(sO)modules. The general case of the diagram and the now established truth of the

Segal conjecture lead to the second part of the following result, the first part being evident.

Theorem 17.

If

G = (Zp)n, k.

is bounded below and cohomologically bounded
*

above, and the completion conjecture holds for J of G, (i) If then the following conclusions hold.
A

kj

for all proper subquotients
A

k.(S 0)

and

k~(S 0)

are not isomorphic as Zp-modules, then the ~.
^

completion conjecture fails for (ii) If kG

is a split ring theory and

k.(S 0)

and

~ ( S 0)

are isomorphic ~.

as k.(S0)-modules,

then the completion conjecture holds for

By Example ii and part (i). we conclude that the completion conjecture generally fails for the theories with non-trivial action by infinite examples, G-CW G. (SGX)* when X is a finite G-CW complex

We shall later point out an important class of

~.(S 0)

cor~olexes for which the completion conjecture holds. In these ^ and k~(S 0) are not isomorphic, so we cannot expect the

isomorphism of Theorem 16 to hold without the bounded above hypothesis. We have another rather startling example of the failure of the isomorphism of Theorem 16. For any ~, there is an associated connective G-cohomology h are

theory for which all of the associated nonequivariant theories

629

connective,

hq(S O) = 0

for

q < O.

When

G = (Zp) n

and

q

is connective,

kG(~G^Z) : O q
Counterexam~le

for q<l-n.
18. The completion conjecture K-theory k0 G and kU G fails for real and complex when G = Z 2. Here the domain

connective equivariant of ~

is zero in negative degrees, but Davis and Mahowald

[II] have calculated

its target groups and shown that they are periodic.

The co~01etion conjecture i nve st igat ion.

for equivariant

cobordism theories

is under

We turn to the proofs of the four calculational theorems starting point for Theorems

above.

The

12 and 15 is Carlsson's observation that elementary

obstruction theory implies a natural isomorphism [X,~G ^ Y]G = [TX'Y]G where space. X is a finite Thus G-CW complex, TX ' Y is any G-

is its singular set, and J

kG(EG A__Z) is the inverse limit over q of the coli~its over V D R s of the groups

of the p-adic completions

(**)
The hard work in Carlsson's Ounawardena's

[T(zrsJZ ^ sV-R s) ,kGV] G
preprint is his analysis of TX. Following up

insistence that Quillen's work on posets ought to be relevant,

Caruso and I found an almost trivial w~y of carrying out essentially the same analysis. Let a be the poset of non-trivial [25], B~ elementary Abelian subgroups of In fact, 6Z G. As

observed by Quillen

is contractible. B~

is a G-category via

conjugation of subgroups and G-category 67[X] ~[X]

is G-contractible.

We construct a topological X; the objects of O and

by parametrizing (A,x), where A

(Z by fixed points of

are pairs

is an elementary Abelian subgroup of ~[X] to TX + TX

x ~ XA.

There is an evident projection funetor from

(regarded as is a G-

a trivial topological O-category) , and the induced map hcmotopy equivalence point categories. by a simple application

B~X]

of Quillen's theorem A [24] to fixed G-space B~[*]

It is convenient to factor out the contractible TX. Thus set Now AX AX = B6Z[X]/B~ *].

since it raps trivially to G-ho,sDtopy equivalence filtration. If we set that

We have a natural

AX + TX. BmX =

comes with an evident natural finite then we find by immediate inspection

FmAX/Fm_IAX,
= V

of definitions

BmX

+ ) o AN( ~ )z~xA(~

@

Here

[~] runs through the G-orbits of strictly ascending chains ~ = (A0C . - - C % )

630

of non-trivial

elementary Abelian subgrcups of ~ under the action of G

G, A(~) = Am, (of which A(~)

and

N(m)

is the

isotropy group of subgroup).

is a normal

We plug this analysis into the description of replace the fnnctor completion, T by the functor A in

kG(EG^Z) above. We may q (**) and then pass to colimits, A gives a finite sequence of Bm for

and inverse lilaits.

The filtration of

long exact sequences in which the third terms come by substitution of T in (**). Inspection of definitions Bm contributes a copy of shows that the [~] th

wedge summand of

the functor If J(~) # e,

kJ(~)(Z A(~)) where J(~) = N(~)/A(~). q+m -then these groups are zero by Proposition 9 and the induction J(~) = e and A(~) # G, then these groups are zero since so that the maps null

hypothesis. ZA(~)

If

contains a copy of the trivial representation,

e: S O + SZA(m) homotopic. with those definitions

which give the inverse limit system are nonequivariantly immediately.

Theorem 12 follows ~ such that

For Theorem 15, we are still left Since Zo = {0} , a check of Here we

A(~0) = G = (Zp) n. ~

shows that the

wedge stmum~nd contributes

At 0 kq+m($ ).

view our finite sequence of long exact sequences as an exact couple and obtain a spectral sequence converging to (reindexed) building copy of k~(S 0) k,(~OAZ). ~ Its El-term is the direct sum of a with A(~) = G. Recall that the Tits

for each

Tits(G) G.

is the classifying space of the poset of non-trivial We regard ~ as a chain of Tits(G) by forgetting

proper A(~) = G

subgroups of

and obtain an isomorphism E m,q-m = Cm-i ~) k~ +m(S 0 ) 1 + where carries Tits C, + dI denotes the augmented simplicial chains of to d @ I Tits(G). The isomorphism

and Theorem 15 now follows from standard facts about

buildings

[25,32]. 13 and 16 are based on the existence of certain

The proofs of Theorems (nonequivariant)

spectra with good properties.

Theorem 19.

There exist spectra

BG -V

and maps

f: BG -W + BG -V

for

V C W

which satisfy the following properties. (i) (2) k.(BG -V) is isomorphic to k~(sV;EG+).

The following diagram c o m ~ t e s .

k.(m -w)

........

f.

k.(m -v)

k,(S

GW

,~o *, ;EG+) (i ^ e), kG¢~V^ sW-V;EG+a sW_V ) N kG(s V ;EG+)

631

(3) (4)

H*(BG -v)

is a free H*(BG)-module

on one generator

IV

of degree -dim V. determined by W-V.

f*: H*(BG -V) + H*(BG -W) f (I V ) = a(W-V)Iw , where

is the morphism of H*(BG)-modules a(W-V) is the Euler class of

Carlsson constructed

such spectra by a kind of double dualization argument Spanier-Whitehead duality,

based on certain assumed facts about equivariant proofs of which are given by Adams stable contexts. construction A more conceptual,

I2] and by Lewis and ~yself but also more technically and n~yself.

I171 in different difficult,

is due to Lewis, Steinberger,

Some years ago, we used of

our stable category of G-spectra to construct a spectrum level generalization the familiar twisted half-smash product construction on spaces. spectra may be specified by BG -V : EG ~ G S-V" The required properties generalizations products. BG -V over are then immediate consequences of spectrum level The desired

of familiar space level properties of twisted half s m s h Thorn spectra, Lewis and I have shown that -V

By use of equivariant

can also be described as the Thom spectrum of the virtual bundle BG. With this description, the eohomological properties

in Theorem 19 are

consequences

of the Thom isomorphism. (i) and (2) of Theorem 19 give that ^G + k . ( m ^ __Z) : lim ~ . ( ~ - j Z ) .

Properties

Properties

(3) and (4) give that colin H*(BG-JZ) = H*(BG)[L-1], L = a(Z).

It is true quite generally that passage to inverse limits from an inverse sequence of convergent Adams spectral sequence [20]. In particular, sequences gives a convergent spectral

passage to inverse limits from the Adams spectral where k represents k*, gives a spectral

sequences of the spectra sequence converging from

BG-JZ^ k,

E 2 = ExtA(H*(BG)[L -1] ® H*(k),Zp) to
a(Z),

^G + k.(EG ^ Z).
hence

For Theorem 13, the localization
and thus ^G. k . [ E G+^ Z ) = O.

is zero by the nilpotency of
The first statement of Theorem

E2 = 0

16 follows by convergence

and an easy comparison of spectral sequences argument and Miller.

from the following homological calculation of Adams, Gunawardena,

632

Theorem 20.

If

G = (Zp) n

and

K

is an A-module which is bounded above (and not

in general otherwise), then ExtA(H*(BG) [L-I] @ K,Zp) -= ExtA(N n ® K,Zp) where degree Nn -n is a free Zp-module on p n(n-l)/2 generators which is concentrated in

and has trivial A-~ction.

The general case follows from the case K = Zp and the fact that ExtA(Q,Z p) = 0 implies EXtA(~ ® K,Zp) = 0 when K = Zp. K is bounded above. We sketch very briefly the

key steps in the proof for construction s: R+M ÷ M. R+

Singer (and Li) [30,311 introduced a basic M, there is an augmentation EXtA(e,l) Ext isomorphism

on A-modules.

For an A-module

Adams, Gunawardena, and Miller 13,13] proved that

is always an isomorphism. (Z-IR+)n(zp) + z-nZp. A -no dul e s

It follows inductively that there is an

Singer and Li [30,31] proved that there is an isomorphism of

(Z-IE+) (Zp) ~ (H*(BG) [L-I] )Bn.

Here the general linear group

GL(n,Zp)

acts on the localization, and

Bn

denotes

the Borel subgroup of upper triangular matrices.

To obtain Theorem 20, one climbs Ext isomorphism

up from the invariants to the entire localization to obtain an H*(BG) [L-If + Nn.

This last step is carried out by direct indnctive calculation up A conceptual,

a chain of parabolic subgroups by Adams, G u n a ~ r d e n a , and Miller I3]-

but less elementary, ar@ament which highlights the role played by the Steinberg module is given by Priddy and Wilk~rson [231. (I have oversimplified slightly; when

p > 2, both [3] and [23] replace R+ by Gunawardena's enlarged analog [131 which is related to H (BZp) as R+ is to H (B~p).) To close, we return to the Segal conjecture and describe its nonequivariant equivalent and an implied generalization, We return to general finite grcups computed by several people [28,121. G. following Lewis, May, and McClure [161. G 0 The coefficient groups w,(S ) have been

The answer is

(A)

G ~.(S 0 )%

Z ~.(BW~ + ) (H)
(H) of subgroups H and WH = NH/H, G. It is natural to ask for an interpretaThe connection is best ~G(S )
* 0

Here the sum ranges over conjuga~y classes where NH is the normalizer of H in

tion of the S e ~ l

conjecture in terms of this isomorphism. [17].

explained in terms of spectra and G-spectra
* +

The groups

are the homoThe

topy groups of the fixed point spectrum (So)G of the sphere G-spectrum S G.

groups WG(EG ) are the homotopy groups of the fixed point spectrum of the dual G-spectrum DG(EG+). spectra Of course, DG(S0) = SG, and s+: EG + + SO induces a map of

633

(B) where the equivalence

e : (SG)G

~ DG(EG+) O = D(BG+), The Segal conjecture m~y be viewed as when G is a p-group, this map

comes from Lemma 4. In particular,

a statement about this map.

induces an equivalence upon p-adic completion. (A) comes from an equivalence

The isomorphism

(c)
Tom Dieck's proof of in t e ~ of which WH (A) s

~: v ~(sw~ +) (H)
o ~

~ (so)G

in [12] leads to an explicit description of can be evaluated; see [16, Thms i and 8]. of the G-set G/H, so

Observe that that G x WH

is the group of auton~rphisms

acts on

GIH.
component of the composite c o ~: V Z~(BWH +) (H) ~D(BG +) where T is the transfer

Theorem 21.

The

H~-

is the adjoint of the element associated to the natural cover

T(1) g w0((BG × BWH)+),

(O/H) ×G×wH(EO × EWH) - - - ~ B O
If G is a p-group, then ~ o ~ induces

× BWH.
upon completion at p.

an equivalence

In view of Theorem 6 and Proposition equivalent to the Segal conjecture,

7, the last statement

is in fact

and it is this formulation that was studied

in special cases in the papers

I18, 13, 27, 3, 211, among others.

Thus the Segal conjecture gives a description of the function spectrum

D(BG +) = F(BG+,S) = F(BG+Z~S0).
It is natural to ask more generally function spectrum F ( BG + ,Z~BH + ) = F (Z=BG + ,Z=BH +) for finite groups answered when Miller [3]. G G and and H H. The question was raised by Adams and Miller and and implies if there is an analogous description of the

are elementary Abelian by Adams, Gunawardena, and I [16] proved that the Segal conjecture and H.

Lewis, MeClure, 0

an answer for arbitrary Let B(G,H)

be the classifying G-space for principal SGB(G,H) + of Example ii.

(G,H)-bundles.

We

have the G-prespectrum

Its cohomolog~ theory is

634

always split and is a split ring theory if main result of [16].

H

is Abelian.

The following is the

Theorem 22.

The Segal conjecture (SGB(G,H)+) *

implies the completion conjecture for all finite groups G and
H.

for the G-

cohomology theories

Of course, the Segal conjecture

itself is the case

E = e.

Again, the coefficient groups have been computed by tom Dieck [12]; indeed, G + n.(X ) in nonequivariant terms for any G-space X. When he computes X = B(G,~), Prop 5l. his result leads to the following description; see [16, Thm I and

(A')

~GB(G,H) + -

Z

~ ~w(BWp+).
(H) (p) of subgroups of homomorphisms H of G and WHthe

(H) T(p)]
Here the sums run over conjugacy orbits [(p)] of H-conjugacy classes classes p: H + H;

groups are Wp = NG×H(Ap)/A@, Let associated ZG where Ap = {(h,p(h))lh c H} C O × H. functor; ZGX is the G-spectrum

denote the suspension G-spectrum SGX,

to the G-prespectrum

and Theorem 22 may be viewed as a

statement about the map of fixed point spectra (B') e * : IS B(G,E)+I G

;

~ + ~F(EG + ,ZGB(G,H) ) G= F(BG+,Z ® BE +) ,

where t h e e q u i v a l e n c e
from an equivalence (C')

a g a i n comes from Lermr~ 4.

The i s o m o r p h i s m

(A')

comes

~:

V (H)

V Z~(BWp +) [(p)]

~ [Z~B(O,~)+] G.
k~

Theorem 21 generalizes

as follows.

See [16, Thms 1 and 8].

Theorem 23.

The

~-W

component of the composite o ~: V V Z~(BWp +) ~ F(BG+,Z~B~ +)

e

(~) [(p)]
is the adjoint of the following composite. Z~(~ Here Bp = Ep/H, where x BWp) + T --Z ( ~ ) ~Z~(BE +) is the classifying

Ep = [(G x H)/Ap] XGxWp(EG × EWp), p and T

map of the H-bundle

F~o + Bp,

is the transfer associated to the cover

635
W

Bp ÷ BG x BWp. completion at

If p.

G

is a p-group, then

e

o ~

induces an equivalence upon

On the ~0-1evel, Theorems 22 and 23 lead to a complete description of the group of stable maps level information. (G x H)-sets. [Z BG ,Z BH I Let A(G,E) in terms of purely algebraic Burnside ring

be the Grothendieck group of H-free finite (G x H)-sets

It is free Abelian on the set of transitive

S = G × ~/Ap, p: H + 9, S the stable map

appearing in the previous theorem, and we associate to

e(s): z~(s~ +)
If I: BG + BG × BWp

~ • Z®(B~ +) ~®(BP+)~ Z®(B~+).
BWp, then is an

is obtained by choosing a basepoint in by an easy verification. Observe that

e(S) = Z~p + o T o Z~I + A(G)-module and let by the augmentation

A(G,E)

A(G,H) ideal of

be its completion with respect to the topology given A(G).

Corollary 24.

There is a natural isomorphism

When

G

is a p-group, we may use completion at As observed by Nishida

p

provided we also

complete the right-hand side. following striking

I221, this has the

(and easy) consequence.

Corollary 25.

If

G

and

H G

are finite groups such that and H

BG

and

B~

are

stably p-equivalent, then

have isomorphic p-Sylow subgroups.

Bibl iography i. J.F. Adams. Graeme Segal's Burnside ring conjecture. Bull. Amer. Math.

Soc. 6 (1982), 201-210. 2. J.F. Adams. Prerequisites for Carlsson's lecture. Proceedings 1982

Aarhus conference in algebraic topology. 3. J.F.

This volume. The Segal conjecture for

Adams, J. H. Gunawardena, and H. Miller. I, II, III.

elementary Abelian p-groups, 4. M.F. Atiyah.

In preparation. Inst. Hautes

Characters and cohomology of finite groups.

Etudes Sci. Publ. Math. No 9 (1961), 23-64. 5M.F. Atiyah and G. B. Segal. Equivariant K-theory and completion. J.

Diff. Geometry 3 (196!)), 1-18.

636

6.

G. Carlsson. conjecture.

Equivariant Preprint,

stable homotopy and Segal's Burnside

ring

1982. The skeletal filtration of equivariant Preprint, 1982. proof of the Segal stable

7.

G. Carlsson and L. Cusick.

homotopy and Segal's conjecture. 8. J. Caruso and J. P. May. conjecture. 9.

An exposition of Carlsson's manuscript, 1982.

Privately distributed

J. Caruso and J. P. May. distributed manuscript,

The completion 1982. An approximation

conjecture for

g,

I.

Privately

i0.

J. Caruso and S. Waner.

theorem for equivariant 1980.

loop

spaces in the compact Lie case. ii. 12. D.M. Davis and M. Mahowald. 0rbittypen

Preprint,

The spectrum

PL^bo.

Preprint.

1983.

T. tom Dieck. Mathe~tik

und aquivariante

Homologie II.

Archiv de r

26 (1975), 650-662. Cohomotopy of some classifying 1981. ring and stable cohomotopy of a finite spaces. Thesis.

13.

J.H.

Gunawardena.

Cambridge University. 14. E. Laitinen° group. 15.

On the Burnside

Math. Scand. 44 (1979), 37-72. Burnside ring and Segal's conjecture, Proceedings 1980, 397-403. Classifying G-spaces and TopologF. CMS it ~ S c a n d i n a v i a n

E. Laitinen.

Congress of Mathematicians 16. L.G.

Lewis Jr., J. P. May, and J. E. McClure.

the Segal conjecture. Conference 17. L.G. Proceedings

Current trends in Algebraic Vol 2 part 2, 165-179.

Canadian Math. Soe. 1982. Equivariant In

Lewis Jr., J. P. May, J. E. McClure, and M. Steinberger. Springer Lecture Notes in Mathematics.

stable homotopy theory. preparation. 18. W.H. Lin.

On conjectures

of Mahowald,

Segal and Sullivan.

Math. Proc.

C~ab. Phil. Soe. 87 (1980), 449-458. 19. J.P. May and J. E. McClure. A reduction of the Segal conjecture. CMS Conference Proceedings Current

Trends in Algebraic Topology. 209-222. 20. J.P.

Vol 2 part 2,

Canadian Math. Soc. 1982. The Segal conjecture manuscript. 1982. for periodic groups. To appear. for elementary Abelian p-

May and S. B. Priddy. Privately circulated

groups. 21.

H. Miller and C. Wilkerson. Proceedings 1982 Northwestern

On the Segal conjecture conference

in algebraic topology.

22.

G. Nishida. groups.

On stable p-equivalences 1983.

of classifying

spaces of finite

Preprint.

23.

S.B.

Priddy and C. Wilkerson. for elementary

Hilbert's theorem 90 and the Segal Preprint. 1983.

conjecture

Abelian p-groups.

637

24.

D. Quillen. Mathematics

Higher algebraic K-theory Vol. 341, 85-147. 1973.

I.

Springer Lecture Notes in

25.

D. Quillen. a group.

Homotopy properties

of the poset of nontrivial p-subgroups

of

Advances in Math. 28 (1978), 101-128. Cohomolo~y of finite groups and elementary

26.

D. Quillen and B. B. Venkov. Abelian subgroups.

Topology ii (1972), 317-318. for cyclic groups and its

27.

D.C.

Ravenel.

The Segal conjecture Amer. J. Math. Equivariant

consequences. 28. G.B. Segal.

To appear. Proceedings of the Gauthier-

stable homotopy theory.

International Villars. 29.

Congress of Mathematicians

1970, Vol 2, 59-63.

1971. Sur la dimension cohomologique des groupes profinis.

J.-P. Serre.

Topology 3 (1965), 413-420. 30. W.M. Singer. A new chain complex for the h o m o l o ~ of the Steenrod

algebra. 31. W.M.

Math. Proc. Camb. Phil. Soc. 90 (1981), 279-292. Resolutions of modules over the Steenrod Math. Z. 181 (1982), 269-

Singer and H. H. Li.

algebra and the classical theory of invariants.

286.
32. R. Steinberg. Prime power representations of finite linear groups II.

Can. J. Math. 18 (1956), 580-591.