R O C Z N IK I PO LSK T EG O TO W A RZY STW A M A TEM A TYCZNEGO Séria I : P E A C E M A TEM A TY CZN E X I X (1977)
|WOJCIECH PUblKOWSKll
Equivariant homology theories indexed by representations
CONTENTS
Introduction... 343
0. Notations and introductory re m a rk s... 343
1. Index-systems for (?-th.eories... 344
2. Unreduced U-theories... 346
3. Spectra and related ^ -th e o rie s... 348
4. Transformations between U-theories... 350
5. Some rem arks... 352
Introduction. This paper is a slightly modified version of Chapter I of my dissertation written under prof. K. Gçba. I am greatful to him and to prof. A. Jankowski for helpful conversations.
In literature there are many examples of equivariant analogues of “generalized homology and cohomology theories” , hut there was no general definition which covered all of them. Such a definition is given in Section 1 — and next some properties of “equivariant generalized homology and cohomology theories” (called shortly “G-theories” ) are proved — with special attention to differences between equivariant and non-equivariant case. This paper has thus rather unifying and ideol
ogical character — no hard and deep theorems are proved.
0. Notations and introductory remarks. In the sequel G will denote a compact Lie group, BG — additive group of the Grothendieck ring of real representations of G (denoted usually by BO(G)), BG+ — additive semi
group of representations (BG+ <= BG), neBG+ — ^-dimensional trivial representation of G.
E em ark . We do not distinguish between representation V and corresponding element in BG+. To be precise — we must choose repre
sentatives Va for all aeBG+ and fix isomorphisms
such that diagrams
У a + Vfi+y
Ш + у р X \ ^ a , P + y
/ ^
r .+ V p +Уу v
<Pa,/S+^\
а+Р+У
у / fa+PiV
a+P
JrV v
\
commute.
We can do it (for example) as follows:
1. Let {У0 = 1 , Vlf V2, ...} be a set of irreducible representations of G (one for each isomorphism class).
For any aeRO(G)+ there exists exactly one representative Va = V0+ . . . + F 0 + F x + . . . . + 7 ! + ... + V 3+ ... + 7 S
-<-*
qtim es-* times-»-
*-k8 times-»-2. Define cpa>ft to be a permutation of summands, which sends F^-sum- mands of Vp after 7 fc-summands of Va (k = 0 , 1 , . . . ) .
The category T(G) of G-spaces in which we work is not very impor
tant. It must satisfy certain usual conditions. The best such category seems to be the category of О-spaces of O-homotopy type of finite G-GW- complexes in the sense of Matumoto [12]. T. (G) denotes corresponding category of pointed О-spaces, and T2(0) — of О-pairs. hT(G), hT.(G) and hTz(G) are suitable homotopy categories (i.e., morphisms areO-homo- topy classes of O-maps).
To be reasonably short — we use some standard conventions and notations from algebraic topology (for example writting X instead of (X , 0) in T 2(0) or using the same letter for a О-map and its O-homotopy class).
Notice one useful generalization of “suspension functor” to the equivariant case :
Let VeRG+. Denote by 8 v eT.(G) its one-point compactification (with distinguished point “at infinity” ), and by
2JV: JiT.{G)->hT.{G)
a functor defined on objects Z V{X) — 8 V
aX (smash product with 8 V),
called V-suspension.
1. Index-systems for 0-theories. In all known examples of G-theories —
homology functors are indexed either by integers (as various (?-bordism
theories [8], [9], [20], [21], [23] and singular ^-homology [2], [5], [10],
[11], [13]), or by cyclic group (H#-theories [18]), or by representations of G (stable (r-homotopy [19], Ебг-bordism [14], [15], [16]). These examples are either theories on T2{G) — as geometric (r-bordism [20], [21], [8], [9], [14], [15], [23] and singular G-homology [2], [5], [10], [11], [13], or reduced theories (i.e., theories on T.(G )) — and in this case suspen
sions are indexed by (all or some) representations of G — as homotopical (r-bordism [7], [8], [16], stable (x-homotopy [19], K G-theories [18].
In Section 1 we restict ourselves to the case of reduced G-theories.
^-theories on T 2(G) will be considered in Section 2.
Let us introduce the definition:
D e f in it io n 1.1. Index-system (for (x-theories) is a diagram / = ( M - ^ B B - ^ B ) ,
where M, В are abelian groups, and <p, rp — group-homomorphisms.
We are going to index (x-homology functors by elements of В and suspensions by elements of M. To be more precise:
D e f in it io n 1.2. Let # — (M -£-* BG-X> B) be an index-system.
A (reduced) — (x-theory (indexed by / ) consists of:
(a) a family of functors
hr: hT.(G)->Ab, reB , satisfying :
hr(Y)->hr{X)-+hr(XfY ) exact for any (x-cofibration Y-+X, (b) a family of natural isomorphisms
oa: hr->hr+y)(Pal > , ae(p~1(BG+), satisfying
oaob = oa+b under natural isomorphisms of functors
ZVajrVb = 2 7
«+
П=
2 7<0+Ь).
R em ark . “Inverting the arrows” in Definition 1.2 we get a defi
nition of a (reduced) (x-cohomology theory. Remarks of this type are usually dropped in the rest of this paper. Cohomological versions are used with no comments in some places.
E x a m p l e s . 1.1. Homotopical (x-bordism of T. Tom Dieck (a) unoriented [9], with index-system
B G ^ -> B G — >Z ,
where dim denotes virtual dimension.
(b) unitary [8], with index-system
Z ® BU (G )-^> BG — >Z , where BU(G) is the unitary representation ring,
<p\RU( 0 ) ~ “forgetting a complex structure”, <p\z — imbedding onto trivial real representations.
1.2. Stable (x-homotopy (and (x-cohomotopy) of G. Segal [19] and homotopical E(x-bordism of [16] are (x-theories with index-system
B G ~ > B G — >BG .
1.3. ÜL^-theory (complex) [18] has as its index-system Z ® B U ( G ) - ^ B G ^ ^ Z 2,
where y — as in example 1.1(b), dim2 — virtual dimension modulo 2.
1.4. For other examples see Section 2, where we show how to get a reduced (x-theory from the unreduced one — and give examples of unreduced (x-theories.
2. Unreduced (x-theories. Unreduced G- theory is defined on the category J i T 2(G). It has boundary homomorphisms instead of suspensions.
To index the homology functors we use — as previous — a homo
morphism
ip: BG~>B.
D e f in it io n 2.1. An unreduced (x-homology theory consists of:
(a) a family of functors
hr : hT2(G)->Ab, reB , (b) a family of natural transformations
d: hr-+hr_vl-Q, r e B ,
where Q: hT2(G)->JiT2(G) is a restriction, i.e., Q (X , Y) = Y subjected to usual axioms:
I. E x c isio n . Let U = Int( U) c JJ c Int(A), (X , A), ( X —U, A —
— U)eT2{G). к: ( X — U, A — U) c —> (X , A) induces isomorphisms к*: hr(X — U , A — U)->hr(X ,A ), reB .
II. E x a c t seq u e n ce s. For any reB , ( X , A )eT2{G) a sequence hr{A)— > hr(X )— > hr{ X ,A ) — * h r_vl(A )— + ...
is exact (non-named arrows are induced by inclusions).
R em ark. As in non-equivariant case — one can deduce from these axioms exact sequence of a triple, Mayer-Yietoris sequence and similar general properties.
E x a m p l e s . 2.1. Singular (or G-cellular) G-homology and ^-coho
mology theory [2], [5], [10], [11], [13]. In this case ip = dim: RG-+Z.
2.2. Geometric G-bordism: unoriented [9], [20], [21], oriented [23]
and unitary [8] — are (unreduced) G-theories with ip — dim.
2.3. Geometric jRG-bordism theory [14], [15], [16] is a G-theory on hT2{G) indexed by id: RG-+RG.
Now we are going to explain a relationship between reduced and unreduced G-theories.
In the non-equivariant case — reduced and unreduced theories are in one-to-one correspondence [22], but in equivariant case it is not so.
Not every reduced G-theory can be obtained from some unreduced one.
T h e o r e m 2.1. There is one-to-one correspondence between unreduced G-theories indexed by ip: RG-^R and reduced G-theories with index-system Z ^ R G - ^ R , i{l) = 1 .
P roof. We give a short description of the correspondence men
tioned in Theorem, omitting the verifications, which are as in [22].
Let {hr, d} be an unreduced G-theory for ip: RG— >R. The correspond
ing reduced G-theory {hr, on} is defined as follows:
hr(X) = hr( X ,{ * } ) , where * is the distinguished point in X.
To define u’s we use exact sequence of the triple (GX, X , {*}), which gives an isomorphism
hr{CX, X)-^-> hr_vl(X , {*}) (G — reduced cone).
Combining d-1 with the isomorphism hr{GX, X)->hr{GXJX^ {X}) and G-homeomorphism C X /X — Z1X we obtain
a1: h ^ X ^ b ^ X ) . on are defined by iterations.
Now let us start with a reduced G-theory {hr1 an} indexed by Z — RG — > R .
Put hr(X ,A ) = hr(X/A).
To define d : hr(X , A)->hr_vl(A), look at the homomorphism p *: hr{X v C A )-*h r{X u C A IX u {*}),
X kj CA is regared as a pointed G space with distinguished point — the vertex * of the cone G A.
12 _ p r ace M a te m a tyczn e 19 z . 2
But X u (L 4 /X u {*} is O-homeomorphic to 2'1(JL u {*}) = U1(A /0), thus we can put
d = ( a 1) - 1 'p * .
Now it is clear why not every reduced О-theory can be obtained from some unreduced one: boundary homomorphisms allow us to define only trivial suspensions. (One can easily find an example of two different reduced О-theories which give the same unreduced O-theory.)
The next point in which there is a difference between equivariant and non-equivariant theories — are coefficients of the theory. In the non-equivariant case coefficients are simply the groups hr (point) (or hr(8°) for reduced theory). In the equivariant case the role of coefficients plays the restriction of the О-theory to the “category of canonical orbits”
0(G). 0(G) is a category with objects GjR for all closed subgroups R of G and with O-homotopy classes of О-maps as morphisms (see [2]).
D e f in it io n 2.2. Coefficients of an unreduced G-theory {hr, 0} are groups hr(GJH) and homomorphisms hr(GIR)-+Jir(G/K) induced by (r-maps G/R-^G/K.
For reduced О-theory we must use hr[(G/R)+) instead of Tir(GjR).
3. Spectra and related О-theories. In non-equivariant situation any (co-)homology theory is representable by suitable spectrum [4], [1].
We define a O-spectrum and show how it leads to G-theory — but we cannot prove the “representability theorem” for О-theory with general index-system. For index-system Z — > B G — > Z it is a simple consequence of E. H. Brown theory [4] and was proved in [13].
D e f in it io n 3.1. Let f = ( M — > JR G ~ > R ) be the index-system.
AG-spectrum A consists of:
(a) a family of pointed (z-spaces AreT.(G), reB, (b) a family of 6r-maps (in T.(G))
£“ : S*aAr-*A r+v9a, aç(p~1(BG+ ), reB , satisfying
pb a
__a+b
cr + y>q>a c r c r
The definition of (z-theory with coefficients in 6r-spectrum is almost the same as in non-equivariant case — except one point about the group- structure. Namely the set of (z-homotopy classes of O-maps
Аг+у)<Ра\д.
usually lias no natural group-structure when <pa contains no trivial sum
mand.
L em m a 3.1. Let X , Y, ZcT.(G), YeRG+.
(a) A set [Yv X , Y]G has natural group structure provided V has a trivial summand.
(b) Group structure from (a) is commutative provided V has two-di
mensional trivial summand.
(c) Suppose V has a trivial summand. The mapping A lz : [ S VX , Y]g- * [ S r X A Z , Y a ZI s is then a group homomorphism.
P roof, (a) Let V = W +1. S vjS w is (т-homeomorphie to S r v S r which allows ns to define 6r-equivariant i?-cogroup structure on S v
V: S r ->Sr v S v .
The map V defines in usual way group structure in [L v X , Y"\a .
(b) Using two-dimensional trivial summand in V we can write a G- homotopy between V and rV, r: S v v S r ~>Sv v S v being “a transposition” , r{xvy) = (y v x).
Thus S v is a ^-equivariant H - commutative LT-cogroup. Commutati
vity of [YVX , Y~\ q follows then in usual way.
(c) This part of the lemma follows directly from the definition of group structures in both sets. Q.E.D.
For X , YeT.(G) put {X , Y}e = lim ([X ,
n
By Lemma 3.1 it has natural abelian group structure. Now we are ready to define a .G-cohomology theory with coefficients in (х-spectrum A : D e f in it io n 3.2. (a) hr(X ; A) = lim {{X ^ X , Ar+ }Q, A“+&),
oeç)- 1( B G + )
where Л«+6 = ( ^ +We) . - i ^ : W i l d ,
(b) cr“ : hr{X\ A)-^hr+w4>a{E4>^X\ A) is the identity (both groups are the limit of the same direct system — but with translated indexes in the second case).
L e m m a 3.2. Functors hr, rcR and transformations o a , aep~1(RG+) form a reduced G-cohomology theory. *
P roof. Standard.
D e f in it io n 3.3. (a) hr{X , A) = lim ({/S'9’*1, X /\A _ r}Q, p%+b),
aeq>-l(№ +)
where
K +b = № * 4 ^ ) , - ^ : {№ , Х л Х „ а_л Ь { в * С + » ) ,Х л Х ^ .+м_,}в,
(b) oa: hr{X , A)-+hr+vv (XVaX } A) is induce dby maps HVa\ {S Vb, 1 л M a \ n - r}a-
L em m a 3.3. Functors hr,r e R and transformations oa, ae<p~1(RG+) form a reducèd G-homology theory.
P ro of. Standard.
B em ark . In Definitions’ 3.2 and 3.3 we can use (x-homotopy classes [ > ] g instead of { , }G provided <pM containes a representation with non-zero trivial summand.
E x a m p l e s . 3.1. Equivariant Thom spactra [8], [9], [16] define
“homotopical (x-bordism and (x-cobordism theories”: These spactra are constructed in usual way from “classifying (x-vector bundles” [7].
3.2. Stable (x-homotopy and (x-cohomotopy [19] are (x-theories de- fined by (x-spectrum of spheres 8 = ($ F, ow}, ow being canonical G- homeomorphism between E wS v and 8 r+w. (For v^RG+ we must put 8 V
— point.)
3.3. Eilenberg-Maclane (x-spectrum K (Z) = {K {Z , V), awj was de
fined in [19].
Index-system for examples 3.2 and 3.3 is / = ( R G ^ 'R G ~ > R G ) .
4. Transformations between (x-theories. This paragraph is devoted to transformations between G-theories with different index-systems. To speak about such transformations — we first introduce a notion of map
ping of index-system.
D e f in it io n 4.1. Let / = {M -Z+ R G -^> R), Ж = R G ~ > 8 ) be two index-systems.
Mapping (a, ft): #->Ж is a commutative diagram M--- ^— » RG---- ?----> R
a p
N ---- ^---- >RG---* S
where a and /? are homomorphisms of abeliam groups.
Now we are ready to define a transformation of (x-theories over some mapping of index-systems (in both directions — see Definitions 4.2 and 4.3). Let Ж, (a, /?) be as in Defefinition 4.1. Let {hr, oa), {gs, rb}
be (x-theories in dexed by f an d Ж respectively.
-
D e f in it io n 4.2. A tran sfo rm atio n ê : {hr, oa}->{gs, t 6} over the m ap p in g (a, /?): # is a fam ily of tran sfo rm atio n s of fun ctors
Kv~+g*v, VeRG,
such that following diagrams commute:
aab rb
Y Y
Ъ . . &
у+Ф n . уФ ,vy>V+y>Mb
" ^ ifvv+ гф "veEG, b€fi~1(BG+ ).
E x a m p l e s . 4.1. F o r an y in dex-system = ( M ~ ^ B G - ^ R) there exists a m ap p in g « / - > / , n am ely
+ R G --- - --- >BG
л
<р
и — ?— > д а ----*--- » E .
Moreover, stable 6r-homotopy [19] is universal among (r-theories with index-systems of the from (RG-^-> BG — » B). The same about G-co- homotopy. For precise formulation see Section 5.
4.2. ^-graded ^ -th e o ry has natural transformation into Z2-graded one — over the mapping of index-systems
B G ---->BG— — — >Z 4
id
BG --- -— > BG dim- - > Z%
This transformations ê is the identity for every veRG. Let J f ,
( a , ft), { \ , or*}, {g3, г6} be as in Definition 4.2.
D e f i n i t i o n 4 .3. A transformation %: {gs, r 6} ->{hr, or®} over the mapping (a, ft): is a family of transformations of functors
Xv* 9 t v €RG)
such that following diagrams commute:
Xv yw
n°b
Qvv+v/лЬ
Y Y
y/ib xv+fib
г . у ф
" ^ Г1/у)К+у>ф **
VcRG, Ь€/г-1(да+).
E x a m p l e s . 4.3. “Stabilized (x-bordism” ) and transformation 9l*( )-^$R?’s( ) over the mapping
BO A
i
z l
- --- ->BG
г- ---- >BG
id dim
rrу --- > z was defined in [6].
4.4. Pontriagin-Thom construction gives a transformation from geo
metric to homotopical (x-bordism [8] — over the mapping from Example 4.3.
E e m ark . We can also define maps of (x-spectra (over some map
ping between their index-systems) and show that they lead to trans
formations of corresponding ^-theories.
5. Some remarks.
A. Multiplicative structures. All ^-theories considered in examples of Sections 1 and 2 are in natural way “multiplicative ^-theories” . Gen
eral definition of multiplicative (x-theory needs some assumption about index-system (to formulate anticommutativity of multiplication).
Namely, suppose that for index-system # — (M — > BG-^-> B) there exists a factorization
B G —4—> B
\ /
dim 2\ ^ }/в
iD e f in it io n 5.1. Let {h r , a a] be a (reduced) (x-theory with index-sys
tem £/ . Suppose there exists a factorization as above or all elements of groups hr(X) (re B , XeT.(G)) are of order 2.
Then multiplicative structure on the theory {hr, oa} is a family of pairings (for all X , YeT.(G))
hr(X) ® hs( Y)->hr+s{XA Y) satisfying usual axioms:
(i) naturality (with respect to induced maps), (ii) associativity,
(iii) identity (there exists an element l e h Q( S ° ) such that for every
a}ehr ( X ) it holds 1 "X = oc),
(iv) stability (suspension ca = multiplication by <ra(l)),
(v) anticommutativity:
For xehr(X), yehs (Y) it holds
Х ‘ У = ( — l f r'QSy ' X
(or os'y = y -os in the case of all elements of order 2).
R em ark . Evident reformalation gives a definition of multiplicative unreduced O-theory.
D e f i n i ti o n
5.2. Let # and q be as above. Multiplicative G-spectrum is a 6r-spe$trum {Ar, e®} with the additional structure:
(i) id e n tity : For any VeBG+ given a map i v : SlF->AvV,
(ii) m u ltip lic a tio n : For any r, seB given a map yr s : Ar л As->Ar+s such that usual (see [22]) diagrams commute.
R em ark . Standard procedure shows that multiplicative ^-spectra lead to multiplicative 6r-theories.
E x a m ple s.
Multiplicative structures in all the examples of O-theories mentioned in this paper are defined in the same way as in corresponding non-equivariant theories.
B. Universality of the stable G-homotopy. We shall prove that stable O-homotopy of = {nl, a v} is universal among O-homology theories with index systems of the form = (BG— > BG-X> B). Cohomological version (i.e., universality of the stable O-cohomotopy) is also true, with similar proof.
T h eo ee m