Simplicial functors and stable homotopy theory
Manos Lydakis
Fakult"at f"ur Mathematik, Universit"at Bielefeld, 33615 Bielefeld, Germany
manos@math206.mathematik.uni-bielefeld.de
May 6, 1998
1. Introduction
The problem of constructing a nice smash product of spectra is an old
and well-known problem of algebraic topology. This problem has come to
mean the following: Find a model category, which is Quillen-equivalent
to the model category of spectra, and which has a symmetric monoidal
product corresponding to the smash product of spectra. Two solutions
were found recently, namely the smash product of S-modules of [EKMM],
and the smash product of symmetric spectra of [HSS]. Here we present an-
other solution, the smash product of simplicial functors. As a byproduct,
we obtain simpler and more descriptive constructions of the model struc-
tures on spectra. As another byproduct, we obtain a model structure on
simplicial functors which should be thought of as a model structure on
homotopy functors. As a final byproduct, we obtain a model-category
version of the part of Goodwillie's calculus of homotopy functors having
to do with "the linear approximation of a homotopy functor at *" [G1],
without the restriction on the relevant homotopy functors of being "sta-
bly excisive".
We say a few words about particular features that distinguish simplicial
functors from S-modules and symmetric spectra. First of all, an advan-
tage of simplicial functors is their simplicity. Also, the monoids under
the smash product of simplicial functors (these are what one would call
"algebras over the sphere spectrum", after thinking of simplicial func-
tors as spectra by using the results of this paper) are well known: They
are the FSPs, which were introduced in 1985 by B"okstedt [B]. There
1
are interesting constructions with FSPs, e.g., topological cyclic homol-
ogy, which can be considerably simplified and conceptualized using the
smash product of simplicial functors [LS]. Note however that it is not
clear that an E1 -FSP has a commutative model, although the analogous
statement is true for S-modules and symmetric spectra. This has, e.g.,
the disadvantage, that it is not clear how to give a model category struc-
ture to (or even how to define) MU-algebras using simplicial functors
(although this can be done for modules over any any FSP, and algebras
over any commutative FSP, by similar methods as in [Sch]). Returning
to examining the special features of simplicial functors, in constrast to
S-modules and symmetric spectra, simplicial functors have not only an
interesting unstable theory (modeling certain homotopy functors, cf. re-
mark 4.6 and section 8), but interesting "metastable", as well as a whole
tower of "higher stable" theories, which are related to the calculus of
homotopy functors [G1], [G2], [G3] (this will appear in a future paper).
Finally, simplicial functors can be not only smashed, but also composed,
with each other. There is a canonical map from the smash product to the
composition product, which is an isomorphism in an important special
case (proposition 5.13), a fact that provides a very concrete description of
the smash product. There is a close connection between simplicial func-
tors and -spaces [L2]. In fact, -spaces may be identified with a special
kind of simplicial functors (see convention 2.11 of [L2]) in such a way
that the smash product of -spaces corresponds to the smash product of
simplicial functors under this identification.
We have tried to make this paper reasonably self-contained. We develop
from scratch the part of the theories of simplicial categories, of model
categories, and of stable homotopy, that we need. In particular, no back-
ground is assumed from the reader, other than familiarity with simplicial
sets (and some basic algebraic topology).
This paper is organized as follows. In section 2, we establish some conven-
tions on notation and terminology, and we introduce homotopy cartesian
squares of simplicial sets. In section 3, we introduce simplicial categories.
In section 4, we introduce spectra and simplicial functors. In section 5,
we introduce the smash product of simplicial functors and we prove its
category-theoretic properties. In section 6, we introduce model cate-
gories. In section 7, we construct a certain model structure on simplicial
categories of a certain form, which includes the simplicial categories of
spectra and simplicial functors. The resulting model structures for spec-
tra and simplicial functors form stepping stones for the construction of
later, more interesting, model structures, as follows. In section 8, we con-
2
struct a model structure on simplicial functors, which should be thought
of as a model structure on certain homotopy functors. In section 9, we
construct the stable model structure on simplicial functors, which we
later show is Quillen-equivalent to the stable model structure on spectra.
The latter was introduced in [BF], but we give a self-contained account
of it in section 10, together with all the facts of stable homotopy theory
that we need. In section 11, we compare the stable model structures.
In section 12, we compare the smash products of spectra and simplicial
functors. Finally, in section 13, we give the connection to the calculus of
homotopy functors.
2. Preliminaries
We denote the category of pointed simplicial sets by S0. We write A+ for
the pointed simplicial set obtained from the simplicial set A by adding
a disjoint basepoint. We denote the basepoint of any pointed simplicial
set K by 0K , or simply by 0 when no confusion can arise. Let * be the
pointed set {0}, let S0 be the pointed set {0; 1}, and let S1 be the pointed
simplicial set 1=@1. We fix once and for all a countable category Sfin0
equivalent to the full subcategory of S0 consisting of the finite pointed
simplicial sets, such that it contains *, S0, S1, and all (n)+ . We choose
the smash product in S0 so that it preserves Sfin0, so that smashing (either
on the left or on the right) by S0 is the identity, and so that smashing
by * always yields *. Given a simplicial set A and a pointed simplicial
set K, we write A K for A+ ^K, and K A for K ^A+ .
We work within some fixed universe. A set is small if it is in the universe,
and it is a class if it is a subset of the universe. From now on, unless
the contrary is explicitly stated, whenever we say "set" we really mean
"small set". All our categories are assumed to have a class of objects and
a class of maps, and to have sets of maps between any two fixed objects.
When we say that a category has all limits and colimits, we really mean
only these indexed by small categories (those with a set of objects and
a set of maps). Finally, we need to explain how we use the word "map":
For us, a map always preserves all the available structure. For example,
if K and L are pointed simplicial sets, then a map K ! L is the same
thing as a pointed simplicial map K ! L.
Let D be the pointed simplicial set obtained by choosing a basepoint
for 1. If A is a simplicial set, A0 denotes the singular complex of the
realization of A. If f : A ! uK is a map of simplicial sets, where uK
denotes the underlying simplicial set of the pointed simplicial set K, the
homotopy fiber hf(f) of f is the pullback of A0! uK0 u hom(D; K0),
3
where the last map is obtained by evaluating on the non-basepoint vertex
of D. The basic properties of the homotopy fiber are the following. There
is a natural map F ! hf(f), where F is the fiber of f, which is a weak
equivalence if f is a fibration. Further, given a map K0 ! K1 of pointed
simplicial sets and a commutative square
A0 ! A1
f0 # # f1
uK0 ! uK1
of simplicial sets with both horizontal maps weak equivalences, the in-
duced map of homotopy fibers hf(f0) ! hf(f1) is also a weak equivalence.
A commutative square of simplicial sets
A0 ! A1
f0 # # f1
B0 ! B1
is called homotopy cartesian if, for every choice of a basepoint of B0
(i.e., for every possibility of viewing the bottom horizontal map as a
map of pointed simplicial sets), the induced map hf(f0) ! hf(f1) is a
weak equivalence. In case f0 and f1 are fibrations, this condition is
equivalent to the following: The canonical map from A0 to the pullback
of B0 ! B1 A1 is a weak equivalence.
A commutative square of pointed simplicial sets is called homotopy carte-
sian if and only if so is the underlying square of simplicial sets. In this
case, if the underlying square is as above, with Ai = uKi, Bi = uLi, and
K1 = *, then we call the sequence K0 ! L0 ! L1 a fibration sequence.
3. Simplicial categories
In this section we introduce the part of the theory of simplicial categories
that we need.
3.1. Definition: [L1, p. 74; see also the bottom of p. 75] Let C and D be
functors from a small category I to categories, with C small, i.e., taking
values in small categories. The maps from C to D are themselves objects
of a category hom(C; D). Given two objects X and Y of hom(C; D), the
maps from X to Y are those families fi, parameterized by the objects i
of I, satisfying the following two conditions. First, fiis a map Xi ! Yiin
the category of functors from Cito Di. Second, given a map ff : i ! j in I
and an object K in Ci, the two maps from Xff*K (which equals ff*XK)
to Y ff*K given by fj and ff*fi are equal.
4
3.2. Definition: A simplicial category is a functor from op to cate-
gories. Given simplicial categories C and D with C small, the category
of simplicial functors from C to D is the category hom(C; D).
3.3. Convention: For the rest of this paper, unless the contrary is
explicitly stated, all simplicial categories C will have a discrete simplicial
class of objects. Thus the class of objects of Cn is independent of n, and
we refer to it as the class of objects of C.
If C is a simplicial category, we sometimes say "map in C" instead of
"map in C0".
3.4. Definition: A simplicial category is called pointed, if there is an
object of C which is a zero object, i.e., both initial and final, in Cn for
all n. Thus we may view C(K; L) as a pointed simplicial set, for all
objects K and L of C. A simplicial functor between pointed simplicial
categories is called pointed, if it takes zero objects to zero objects.
3.5. Example: We define the pointed simplicial category S. Its objects
are the pointed simplicial sets. The maps from K to L in Sn are given
by the maps from n K to L in S0. The composition vu in Sn is
given by the composition v(n u)(d K) in S0, where d is the diagonal
n ! n x n, and K is the source of u in Sn. Note that S has a
small full pointed simplicial subcategory Sfin, whose value at [0] is the
category Sfin0which we have defined in the previous section.
3.6. Remark: According to our conventions, a pointed simplicial cate-
gory is essentially a category "enriched over S0". Most of this section, and
section 5, could be generalized by using "category theory enriched over a
symmetric monoidal category" (see [D]). Our approach does not require
enriched category theory, and avoids "coherence conditions". There is
also the equivalent approach of defining a simplicial category C to be
a class of objects, a family of simplicial sets C(K; L) defined for each
pair (K; L) of objects, and an assosiative and unital composition. In these
terms, a simplicial functor X from the simplicial category C to the simpli-
cial category D is a family of maps of the form C(K; L) ! D(XK; XL),
that preserve composition and units. This is the approach of [Q] (see
section II.1), but the definition given there is too restrictive for our pur-
poses. (There it is required that all simplicial categories have functors ?K
and ? K, behaving like the functors with the same name in proposi-
tion 3.16. This rules out the simplicial category Sph , which is used in
5
the next section to represent spectra as simplicial functors.)
3.7. Definition: Let P : S x S ! S be the pointed simplicial functor
given on objects by the smash-product (its definition on maps of Sn
involves the diagonal n ! n x n). Given K in S, define K* : S ! S
by K*L = P (L; K). Given a pointed simplicial functor X : C ! S,
define X ^K : C ! S to be the composition K* O X. Given a simplicial
set A, define X A to be X ^A+ .
3.8. Example: Given a small pointed simplicial category C, the dual
pointed simplicial category C* is defined as follows. Let C*0be the cat-
egory of pointed simplicial functors from C to S. Let the objects of C*
be the objects of C*0. Finally, let the maps in C*nfrom X to Y be the
maps in C*0from X n to Y .
3.9. Convention: Given X in C*, we write X(n) for the functor
from Cn to Sn it determines. We denote by XK the value of X(n)
at K, which does not depend on n. We denote by XnK the pointed set
of n-simplices of XK.
3.10. Definition: Let E be the category of pointed sets. Given X in C*
and [n] in , we define a pointed functor Xn : Cn ! E. The value of Xn
on objects has already been defined above. Given u : K ! L in Cn and x
in XnK, we define u*x in XnL to be u*(x; 1), where we wrote u* also for
the map XK n ! XL, and 1 denotes the identity map in nn.
3.11. Proposition: Fix a collection of pointed functors Xn from Cn to
pointed sets, and fix, for all K in C, a choice of assembling the XnK to
a pointed simplicial set. Then this data determines a pointed simplicial
functor if and only if, given ff : [m] ! [n] in , the induced maps
XnK ! Xm K assemble to a natural map Xn ! Xm ff*. Further, given X
and Y in C*, a collection of maps Xn ! Yn determines a map X ! Y
if and only if, for all K in C, the maps assemble to a map of pointed
simplicial sets XK ! Y K.
Proof. The conclusion follows from the following two observations.
First, given ff : [m] ! [n], u : K ! L in Cn, X in C*, and x in Xm K,
the equality u*(x; ff) = (ff*u)*(x; 1) holds. For, since X is simplicial,
ff*(u*) = (ff*u)*. But ff*(u*)(x; 1) = u*(x; ff).
Second, a map f : X ! Y in C*0is a family of maps XK n ! Y K,
6
satisfying a condition stated in definition 3.1. This condition implies
that this family is determined by a family of maps XK ! Y K, by
precomposing with the projections XK n ! XK. Equivalently,
__
given ff : [m] ! [n] and x in Xm K, we have f(x; ff) = f(x; 1). |_|
3.12. Proposition: All colimits and limits exist in C*0, and are com-
puted degreewise.
__
Proof. This follows from proposition 3.11. |_|
3.13. Definition: (Representable simplicial functors) Given an ob-
ject K of the small pointed simplicial category C, we define, using propo-
sition 3.11, an object K* of C*. Given L in C, the pointed simplicial
set K*L is defined to be C(K; L). The pointed functor K*n: Cn ! E
is defined to be the one represented by K. Then these data assemble to
a pointed simplicial functor, and the assignment K 7! K* extends to a
pointed simplicial functor Cop ! C*.
Given X in C*, define XK in C* to be the composition K* O X.
3.14. Proposition: Given X in C* and K, L in C, the pointed
sets S0(L; XK) and C*0(K* ^L; X) are isomorphic.
Proof. By the Yoneda lemma, the maps from K*n^Ln to Xn are in bi-
jective correspondence with the maps from Ln to XnK. The conclusion
__
follows from the description in proposition 3.11 of the maps in C*. |_|
3.15. Corollary: Given X in C* and K in C, XK ~=C*(K*; X).
3.16. Proposition: Given K, L in S and X, Y in C*, there are
isomorphisms
S(L; Y K) ~= S(L ^K; Y ); and
C(X; Y K) ~= C*(X ^K; Y ) ~= S(K; C*(X; Y ))
__
natural in all the variables involved. |_|
3.17. Proposition: Given X and Y in C*, the pointed set C*0(X; Y )
is isomorphic, via the obvious map, to the limit of the diagram
Y Y
S0(XK; Y K) --!! S0(XK n; Y L);
K u
7
where the first product is indexed by all objects K of C, and the second by
all u in the sum over n; K; L of Cn(K; L). The two maps in the diagram
are defined as follows. Given a family of maps fM : X(M) ! Y (M) and
a map u : K ! L in Cn, we have to define two maps XK n ! Y L.
The first is u*(fK n) and the second is fLu*.
Proof. By definition, a map of simplicial functors f : X ! Y is a family
of maps XK n ! Y K, satisfying two conditions. First, for fixed n,
this family should give a map of functors Cn ! Sn. The second condition
is found in the proof of proposition 3.11. These conditions are equivalent
to the following single condition. Given u : K ! L in Cn, the equality
u*(fK n) = fLu* holds. The conclusion follows immediately from
__
this. |_|
3.18. Corollary: Given X; Y in C*, the pointed simplicial set C*(X; Y )
is isomorphic, via the obvious map, to the limit of the diagram
Y Y
S(XK; Y K) --!! S(XK n; Y L):
K u
3.19. Definition: The canonical map K* ^XK ! X, given X in C*
and K in C, is the map corresponding to the identity of XK under the
isomorphism of proposition 3.14.
3.20. Corollary: Given X in C*, the canonical maps K* ^XK ! X
induce an isomorphism between X and the colimit of the diagram
_ _
L*^ XK n --!! K* ^XK;
u K
where the first map is induced from u* : XK n ! XL, and the second
map is induced from u* : L* n ! K*.
Proof. Let Z be the colimit. Propositions 3.17, 3.14, and 3.16, imply
__
that the functors C*0(X; ?) and C*0(Z; ?) are isomorphic. |_|
3.21. Definition: Let Q : C ! D be a pointed simplicial functor be-
tween small pointed simplicial categories. The pointed simplicial functor
Q* : C* ! D* takes X to the colimit of the diagram
_ _
(QL)*^ XK n --!! (QK)*^ XK;
u2W n;K;LCn(K;L) K
8
where the maps are similar to the ones in the previous corollary.
3.22. Definition: Given pointed simplicial functors L : C ! D and
R : D ! C, the pair (L; R) is called an adjoint pair, provided that the
pointed simplicial functors Cop x D ! S, taking (X; Y ) to C(X; RY )
and D(LX; Y ), are isomorphic.
3.23. Proposition: In the situation of definition 3.21, let the pointed
simplicial functor Q* : D* ! C* be given by precomposing with Q.
Then (Q*; Q*) is an adjoint pair.
Proof. By corollaries 3.18, 3.15, and proposition 3.16, the pointed sim-
__
plicial functors C*(X; Q*Y ) and D*(Q*X; Y ) are isomorphic. |_|
3.24. Proposition: In the situation of proposition 3.23, the pointed
simplicial functors Q*K* and (QK)* are isomorphic, for all K in C.
Proof. By proposition 3.23 and corollary 3.15, the pointed simplicial
__
functors D(Q*K*; ?) and D((QK)*; ?) are isomorphic. |_|
4. Spectra and simplicial functors
In this section we consider two simplicial categories of the form C*. One
of them turns out to be one of the standard models for the simplicial
category of spectra (e.g., the spectra of [BF]). The other one, denoted
by SF , is closely related to homotopy functors (see remark 4.6).
4.1. Definition: Given K in S, let K = S1 ^K, and K = S(S1; K).
1
Given a small pointed simplicial category C and X in C*, let X = XS .
A spectrum E consists of a sequence of pointed simplicial sets En and
a sequence of maps E(n) : En ! En+1 , for n = 0; 1; : :.: A map
f : E ! E0 of spectra is a sequence of maps fn : En ! E0n such that
fn+1 O E(n) = E0(n) O fn. Given a spectrum E and a pointed simplicial
set K, the pointwise definition of E ^K does give a spectrum, thus spectra
form a pointed simplicial category Sp .
4.2. Definition: If n is a positive integer and K is in S, define an
equivalence relation ~ on the n-fold cartesian product Kxn by x ~ x0 if
and only if either x = x0 or xi = 0 and x0j= 0 for some i and j. Let
K^n = Kxn = ~. We denote the class of the sequence (x1; x2; : :;:xn)
in K^n by x1 . x2 . .x.n. Define K^0 = S0. We have an isomorphism
9
^(K; m; n) : K^m ^ K^n ! K^(m+n) , denoted by x ^y 7! x . y, induced by
concatenation of sequences. Thus, for any non-negative integers m1, m2,
and m3, and any x1, x2, and x3, with xiin K^mi , the elements (x1.x2).x3
and x1 . (x2 . x3) of K^(m1+m2+m3) are equal.
We now define a pointed simplicial subcategory K^ of S, whose unique
zero-object is * and whose non-zero objects are given by the K^n , as
follows. The pointed simplicial sets of maps from K^n to K^(m+n) are
trivial if m is negative, else they are isomorphic to K^m , and they are
given by the image of K^m in S(K^n ; K^(m+n) ) under the adjoint to the
map K^m ^ K^n ! K^(m+n) described in the previous paragraph.
Finally, define the standard n-sphere Sn recursively, as S1 ^Sn-1 (note
that Sfin contains all Sn). There is an isomorphism in : Sn ! (S1)^n,
defined by in = ^(S1; 1; n - 1) O (1S1^ in-1). There is a unique pointed
simplicial subcategory Sph of S with objects all Sn, such that the maps in
assemble to an isomorphism Sph ! (S1)^.
4.3. Proposition: There is an isomorphism (not just an equivalence)
of pointed simplicial categories from Sph * to Sp , taking the pointed sim-
plicial functor X to the spectrum E with En = XSn, and with the
map S1 ! S(En; En+1) given by the composition of the isomorphism
S1 ! Sph (Sn; Sn+1 ) with the map Sph (Sn; Sn+1 ) ! S(XSn; XSn+1 )
__
induced by the pointed simplicial functor X. |_|
From now on, we identify Sp and Sph *, using the previous proposition.
4.4. Definition: Let SF be the dual of Sfin, and, more generally,
let SF (n)be the dual of the n-fold cartesian product of Sfin with itself.
An object X of SF can be evaluated on an arbitrary K in S, by defin-
ing XK as the colimit, over all L ! K with L in Sfin, of XL. We
obtain a simplicial functor S ! S commuting with filtered colimits and,
conversely, any such simplicial functor can be canonically identified with
one coming from SF , in fact, coming from its restriction to SF . A map
X ! Y in SF is called a a weak equivalence of associated homotopy func-
tors provided that, for all fibrant K in S, the map XK ! Y K is a weak
equivalence (this terminology is explained in remark 4.6). We denote the
class of these maps by SF hw.
4.5. Convention: For the remainder of this paper, unless the contrary
is explicitly stated, the term "simplicial functor" means "object of SF ".
10
4.6. Remark: A functor between two categories, both of which have
some fixed class of weak equivalences, is called a homotopy functor, pro-
vided that it preserves the weak equivalences. Thus the category of ho-
motopy functors from finite pointed simplicial sets to pointed simplicial
sets is an interesting category, and so is its full subcategory of pointed
homotopy functors (these are the functors preserving zero objects). Note
that the category of homotopy functors (between any categories with
weak equivalences) has itself a natural class of weak equivalences. It is,
however, not clear how to define limits and colimits of homotopy func-
tors, let alone how to put a model structure on them. We already know
that SF has all limits and colimits, and we shall see in section 8 that
there is a model structure on SF whose class of weak equivalences is
given by SF hw. The connection between SF and homotopy functors is
explained below.
For the remainder of this remark the term "homotopy functor" means
"pointed homotopy functor from Sfin0to S0". Any simplicial functor X
has an associated homotopy functor, defined by K 7! XRK, where R
denotes the simplicial functor "singular complex of the realization". Thus
a map of simplicial functors is in SF hw if and only if it induces a weak
equivalence of associated homotopy functors. Further, the construction
F 7! F of [W, p. 402] shows that every homotopy functor comes from a
simplicial functor, up to weak equivalence. This suggests that the right
way to view homotopy functors (at least for purposes of putting a model
structure on them) is as simplicial functors.
4.7. Definition: The assignment K 7! K* of definition 3.7 gives a
pointed simplicial functor Sfin ! SF (this involves using the natural
isomorphism between K ^L and L ^K in the special case L = (n)+ ).
The pointed simplicial functor Hom : SF op x SF ! SF is defined by
Hom(X; Y )(K) = SF (X; Y O K*).
4.8. Proposition: The pointed simplicial functors from Sfin x SF
to SF , taking (K; X) to Hom(K*; X), resp. to X O K*, are isomorphic.
__
Proof. This follows immediately from corollary 3.15. |_|
5. The smash product of simplicial functors
In this section we construct a symmetric monoidal pointed simplicial
functor SF x SF ! SF , written (X; Y ) 7! X ^Y and called the smash
11
product of simplicial functors, satisfying the expected adjoint functor
properties.
5.1. Definition: The external smash product SF x SF ! SF (2) is
defined by
X e^Y = P O (X x Y );
where P is as in definition 3.7, and X x Y : Sfin x Sfin ! S x S.
Note that P induces a pointed simplicial functor Sfinx Sfin ! Sfin,
denoted again by P . The smash product SF x SF ! SF is defined by
X ^Y = P*(X e^Y ), with P* as in definition 3.21.
5.2. Proposition: There is an isomorphism
SF (X; Hom(Y; Z)) ~=SF (X ^Y; Z);
natural for X, Y , and Z in SF .
Proof. By proposition 3.23, it suffices to show that the left-hand side
is naturally isomorphic to SF (2)(X e^Y; P *Z). Let [1]0 be the category
with two objects 0 and 1 and two non-identity maps, both from 0 to 1.
By corollary 3.18, the left-hand side is the limit of`a diagram D defined
on [1]0x[1]0, with Dijdefined as follows. Let Sij= m;nNiSfinmxNjSfinn,
where "N" stands for "nerve", i.e., "N0" stands for "objects" and "N1"
stands for "maps". Every element of Sijdetermines m and n, as well as
maps u and v in Sfin, with u : K0 m ! K1, and v : L0 n ! L1,
such that if i = 0 then m = 0 and u is the identity, and similarly for
j = 0. Then Dijis the product over all elements of Sijof
S(XK0^ Y L0 m n; Z(K1^ L1)):
Note that the limit of a functor F from I to (pointed simplicial) sets
injects, via the canonical map, in F i, provided that i is an object of I
that maps to any other object. We use this to identify such a limit with
the corresponding subobject of F i. We want to identify the limit of D as
the limit`of a functor D0 defined on [1]0, with D0idefined as follows. Let
Si = m Ni(Sfinmx Sfinm). Every element of Si determines m, as well as
maps u and v of finite pointed simplicial sets, with u : K0 m ! K1,
and v : L0 m ! L1, such that if i = 0 then m = 0 and u, v
are identity maps. Then D0iis the product over all elements of Si of
S(XK0^ Y L0m ; Z(K1^ L1)). Thus the first limit is a subobject of the
second,Qand we want to show that this injection is also a surjection. Fix f
in (K;L)S(XK ^Y L; Z(K ^L)), and suppose that f is an element of the
12
second limit. Fix u : K0 m ! K1 and v : L0 n ! L1. Since f is in
the second limit, it satisfies the corresponding naturality statement with
respect to (u; 1L0), as well as with respect to (1K1; v) (here 1L0 denotes
the identity map of L0 in Sfinm, and 1K1 denotes the identity map of K1
in Sfinn). Thus f n satisfies the corresponding naturality statement
with respect to (u; 1L0). Half of the argument that f is in the first limit
involves considering the corresponding square, and the square expressing
the naturality of f with respect to (1K1; v), as well as the composition
of these two squares. The other half involves using the pairs (1K0; v),
__
and (u; 1L1). |_|
5.3. Corollary: There exists a natural isomorphism
Hom (X ^Y; Z) ~=Hom (X; Hom (Y; Z)):
Proof. The simplicial functors Hom (Y; Z) O K* and Hom (Y; Z O K*) are
__
naturally isomorphic. The conclusion follows from proposition 5.2. |_|
We show below that, just like the smash product of pointed sets, this
smash product is what we call "strongly symmetric monoidal". This is
a property stronger than being symmetric monoidal, and also easier to
define (the definition does not involve coherence conditions).
5.4. Definition: Fix a functor F : C x C ! C, which is associative,
commutative, and unital, up to isomorphism. We say that F is is strongly
symmetric monoidal provided that, for any non-negative integer n, the
n-th iterate F (n): Cn ! C has trivial automorphism group. Here F (n)
is defined as follows: It equals the constant functor with value some
fixed unit if n = 0, it equals the identity if n = 1, and, for n > 1,
F (n)= F (1C x F (n-1)).
5.5. Definition: Let S be the object of SF given by the inclusion of Sfin
in S.
__
5.6. Proposition: There exists a canonical isomorphism (S0)* ! S. |_|
5.7. Lemma: The simplicial functors (K ^L)* and K* ^L* are isomor-
phic, for any K and L in Sfin.
__
Proof. This is a special case of proposition 3.24. |_|
13
5.8. Remark: Recall from [L2] the smash product of -spaces, and
that -spaces may be identified with a special kind of simplicial functors
(namely, these simplicial functors which can be evaluated degreewise,
cf. convention 2.11 of [L2]) in such a way that representable -spaces
correspond naturally to representable simplicial functors (in fact, the
-space n of [L2] corresponds to the simplicial functor K*, with K a
pointed set of cardinality n + 1). The smash product of -spaces corre-
sponds to the smash product of simplicial functors under this identifica-
tion. This is so for representable objects by the above lemma, and every
object is a colimit of representables, cf. the proof of proposition 2.16
in [L2].
5.9. Theorem: The smash product of simplicial functors is strongly
symmetric monoidal, with unit S.
5.10. Lemma: The smash product of simplicial functors is associative
and commutative, up to natural isomorphism, and S acts as a unit, up
to natural isomorphism.
Proof of theorem 5.9. Given lemma 5.10, it remains to show that, given a
non-negative integer n and simplicial functors X1; : :;:Xn, every natural
automorphism of X1^ . .^.Xn equals the identity. The case n = 2 should
suffice to explain the proof, so we assume n = 2 below, and we write
X and Y instead of X1 and X2.
Let OE = OE(X;Y )be an automorphism of X ^Y , natural in the simplicial
functors X and Y . Consider the special case X = K*and Y = L* for
some K; L. By lemma 5.7 and corollary 3.15, SF (K* ^L*; K* ^L*) ~=
S(K ^L; K ^L). Let OE = f* for some f : K ^L ! K ^L in Sfin. We
show that all k ^l in K0^ L0 are fixed under f. Note that this is trivially
true if K = L = S0, since OE is an automorphism. Let ^k: S0 ! K and
^l: S0 ! L correspond to k and l. It suffices to show that f O(^k^^l) = ^k^^l.
But
f O (^k^^l)= (f O (^k^^l))*1K ^L
= (^k^^l)*f*1K ^L
= (^k^^l)*OE1K ^L
= (^k*^^l*)OE1K ^L
= OE(^k*^^l*)1K ^L
= OE(^k^^l)
= ^k^^l;
where the fourth equality follows from identifying (K ^L)* with K* ^L*
14
using lemma 5.7, the fifth equality follows from naturality, and the last
equality follows since OE(S0;S0)= 1.
By definition of X ^Y , there is a map onto X ^Y from the sum, over
all (K; L), of (K ^L)*^ XK ^Y L. The map
(K ^L)*0M ^X0K ^Y0L ! (X ^Y )0M
can be described as follows, after identifying (K ^L)* and K* ^L*. It
takes f ^a ^b to (^a^^b)f, where a 2 X0K and ^ais the corresponding map
K* ! X, and similarly for b, and f 2 (K ^L)*0M. Thus every vertex
of (X ^Y )M is in the image of a map K* ^L* ! X ^Y . It follows that OE
is the identity on vertices.
Given Z in SF , there is a natural map Z(LK ) ! (ZL)K , where LK =
S(K; L). The adjoint map Z(LK ) ^K ! ZL is itself adjoint to the
composition of K ! S(LK ; L) ! S(Z(LK ); ZL), where the first map
is adjoint to the evaluation map (the evaluation map is adjoint to the
identity of LK ), and the second map is the one induced by the simplicial
functor Z. This map is an isomorphism if Z is representable. Since OE is
n)+
the identity on vertices, it follows that OE( is the identity on vertices
for X and Y representable, i..e, that OE is the identity on n-simplices for X
and Y representable. It follows that OE is the identity on n-simplices for X
and Y of the form (representable)n, because a simplex of n x n is
determined by its vertices.
Finally, for any X and Y , the canonical map (K ^L)*^ XK ^Y L ! X ^Y
is isomorphic to the smash product of K* ^XK ! X and L*^ Y L ! Y .
Thus every simplex of X ^Y is in the image of a map X0^ Y 0! X ^Y ,
__
with the primed simplicial functors of the form (representable)n. |_|
Proof of lemma 5.10. We claim that X ^S ~= X. This follows from
proposition 5.2, proposition 5.6, and the fact that, by proposition 4.8,
Hom ((S0)*; Z) is isomorphic to Z.
To check that X ^Y ~= Y ^X, we check that SF (X ^Y; Z) ~=SF (Y ^X; Z)
for all Z in SF . This follows from the isomorphisms SF (2)(X e^Y; P *Z) ~=
SF (2)(Y e^X; P *Z O T ) and P *Z ~= P *Z O T , where T is the obvious
involution of Sfinx Sfin.
Finally, we compare both (X ^Y ) ^Z and X ^(Y ^Z) to a more symmet-
ric simplicial functor X ^Y ^Z. We will give only one comparison (the
other is similar). Here X ^Y ^Z is defined as P*(3)(X e^Y e^Z), where P (3)
is as in definition 5.4 (i.e., it takes (K; L; M) to K ^L ^M) and the ex-
ternal smash product X e^Y e^Z is defined in the obvious way (to simplify
15
the exposition, we write as if the smash product in S was associative,
instead of associative up to unique natural isomorphism). The isomor-
phism SF ((X ^Y ) ^Z; W ) ~= SF (X ^Y ^Z; W ) is obtained by observing
that
SF ((X ^Y ) ^Z; W ) ~= SF (X ^Y; Hom (Z; W ))
~= SF (2)(X e^Y; P *Hom (Z; W ))
~= SF (3)(X e^Y e^Y; (P (3))*W );
where the last isomorphism is similar to the one in the proof of proposi-
__
tion 5.2. |_|
5.11. Definition: We define a natural map X ^Y ! X O Y , which we
call the assembly map. This map is an isomorphism when Y is repre-
sentable (proposition 5.13).
We first define a natural map X(K) ^L ! X(K ^L) . This is the map
that most resembles other assembly maps in the literature, and it is the
special case Y = S ^L. Its adjoint L ! S(X(K); X(K ^L) is defined as
the composition L ! S(K; K ^L) ! S(XK; X(K ^L)), where the first
map is adjoint to the identity, and the second map is the one defined by
the simplicial functor X. There is a similar map K ^X(L) ! X(K ^L) .
We use the name special assembly map to refer to any of these two maps.
To handle the general case, by definition of the smash product, it is
enough to specify a natural map XK ^Y L ! XY (K ^L)). This is de-
fined as the composition
XK ^Y L ! X(K ^Y L) ! XY (K ^L);
where the first map is a special assembly map, and the second is given
by applying X to a special assembly map.
5.12. Remark: We are now able to identify the monoids in this smash
product. Surprisingly enough, these turn out to be well-known (although
this is no surprise to the reader familiar with the smash product of -
spaces, cf. remark 2.13 of [L2]).
Let us say that a simplicial functor X is a monoid, if there are maps
: X ^X ! X and j : S ! X satisfying the usual associativity and unit
conditions. Then corresponds to e : X e^X ! P *X , i.e., to a natural
map e : XK ^XL ! X(K ^L) . In fact, X is a monoid if and only if
it is an FSP, as defined by B"okstedt in [B], under e, j, and the special
assembly map of definition 5.11.
16
5.13. Proposition: The assembly map X ^Y ! X O Y is an isomor-
phism, whenever Y is representable.
Proof. Fix a representable Y . The case X is representable follows from
lemma 5.7. The functor X O Y preserves all limits and colimits (recall
that Y is fixed). The functor X ^Y preserves all colimits, since it is a
left adjoint. Similarly, both functors commute with smashing by pointed
__
simplicial sets. The conclusion now follows from corollary 3.20. |_|
5.14. Remark: The composition product (X; Y ) 7! X OY is associative
and unital (with unit S), up to natural isomorphism (it is not strictly
associative, because X is originally defined only on Sfin), and the as-
sembly map is compatible with associativity and unit isomorphisms. In
the language of monoidal categories, the assembly map makes the iden-
tity functor of SF a lax monoidal functor from the monoidal category
(SF ; O; S) to the monoidal category (SF ; ^; S).
5.15. Remark: If the restriction of the smash product functor P
to Sph x Sph would lift to Sph (which it does not) then the meth-
ods of this section would give a symmetric monoidal product of spectra
(such a product seems not to exist).
A symmetric monoidal product on the pointed simplicial category of
symmetric spectra is constructed in [HSS]. Let Sph be the smallest
pointed simplicial subcategory of S containing both Sph and, for all n,
the obvious copy of n inside S(Sn; Sn). Then the restriction of P
to Sph x Sph does lift to Sph , and the methods of this section give
a symmetric monoidal product in (Sph )*. In fact, the pointed simpli-
cial category (Sph )* is isomorphic to the pointed simplicial category
of symmetric spectra, and the symmetric monoidal products correspond
under this isomorphism.
A symmetric monoidal product on the pointed simplicial category of -
spaces is constructed in [L2]. Let Sdfinbe the full pointed simplicial
subcategory of Sfin with objects the discrete objects of Sfin. Then the
restriction of P to Sdfinx Sdfindoes lift to Sdfin, and the methods of
this section give a symmetric monoidal product in (Sdfin)*. In fact, the
pointed simplicial category (Sdfin)* is equivalent to the pointed simplicial
category of -spaces, and the symmetric monoidal products correspond
under this equivalence.
17
6. Model categories
Model categories were introduced by Quillen in [Q]. In this section, we
give an exposition of the parts of the general theory that are needed in
this paper.
6.1. Definition: Fix a category C and a commutative diagram
A ! C
# #
B ! D
in C. If the diagram is a pushout, then we say that the map C ! D
is a cobase change of the map A ! B (along the map A ! C). If the
diagram is a pullback, then we say that the map A ! B is a base change
of the map C ! D (along the map B ! D). If there exists a map
B ! C such that both resulting triangles commute, then we say that the
map A ! B has the left lifting property with respect to the map C ! D,
and we say also that the map C ! D has the right lifting property with
respect to the map A ! B. Given a class S of maps in C, Sl denotes
the class of maps in C that have the left lifting property with respect to
all the maps in S, and Sr denotes the class of maps in C that have the
right lifting property with respect to all the maps in S. If S = {f}, we
abbreviate Sl and Sr by fl and fr.
Given a commutative diagram
A ! C ! A
# # #
B ! D ! B
in C, such that both the composite A ! A and the composite B ! B are
identities, we say that the map A ! B is a retract of the map C ! D.
A limit E1 of a diagram E = E1 E2 . .i.n C is called a sequential
limit, and a colimit E*1 of a diagram E* = E*1! E*2! . . .in C is
called a sequential colimit. The canonical map between E(*)1and E(*)1
will be called the composition of the diagram E(*). Let pn be the map
En+1 ! En, and in be the map E*n! E*n+1. We denote the composition
of E by p1p2 . .,.and the composition of E* by . .i.2i1. A class S of maps
in C is called closed under sequential (co)limits if, for all diagrams E(*)
as above, it contains any composition of E(*)whenever it contains all the
maps of E(*).
18
A class S of maps in C is called closed under retracts if, for all maps f
and g in C with f a retract of g and g in S, the map f is also in S.
Finally, S is called closed under (co)base change if, for all maps f and g
in C with f a (co)base change of g and g in S, the map f is also in S.
6.2. Proposition: Let S be any class of maps in a category C. Then Sl
is closed under sums, cobase change, sequential colimits, and retracts.
Also Sr is closed under products, base change, sequential limits, and re-
__
tracts. |_|
6.3. Proposition: Let f; p; i be maps in a category C with f = pi.
Then f 2 pl implies that f is a retract of i. Also f 2 ir implies that f is
__
a retract of p. |_|
6.4. Definition: An object X of a category C is called small if, for
every diagram E in C of the form E1 ! E2 ! . .,.the canonical map
colim C(X; E) ! C(X; colim E) is an isomorphism.
Given a class S of maps of a category C with small domains, the class S0
of cofibrations generated by S is defined to be the class of all retracts of
all compositions X0 ! X1 ! . .i.n C satisfying the following property.
For all n 0, there exists a set I,`a family {si : Ai ! Bi | i 2 I} of
maps in S indexed by I, a map f : iAi ! Xn,`and a pushout square
expressing Xn ! Xn+1 as the cobase change of isi along f.
6.5. Proposition: (The small object argument) Let C be a category
having all colimits and f : X ! Y be a map in C. Given a set S of
maps in C with small domains, there exists a factorization f = pi in C,
such that p is in Sr and i is in the class of cofibrations generated by S.
Proof. Given s 2 S, say As is the domain of s and Bs is the codomain of s.
Given a map q : Z ! Y in C, let C(s; q) be the set of all pairs g = (g1; g2)
in C(As; Z) x C(Bs; Y ) such that qg1 = g2s. Consider the following
pushout ` `
s g2C(s;q)Ag ! Z
` ` # #
s g2C(s;q)Bg ! W (q)
where, for g 2 C(s; q), Ag = As and Bg = Bs. The map Ag ! Z
is g1, and the map Ag ! Bg is s. Denote by i0(q) the map Z ! W (q),
and by h(g) the map Bg ! W (q). Note that h(g)s = i0(q)g1. By the
definition of C(s; q), we have a map p0(q) : W (q) ! Y with q = p0(q)i0(q)
19
and g2 = p0(q)h(g).
Let X0 = X and p0 = f. For n 1, let Xn = W (pn-1), pn = p0(pn-1),
and in-1 = i0(pn-1). Let i : X0 ! X1 be the composition . .i.1i0. Let
jn : Xn ! X1 be the canonical map, so that jn+1in = jn and i = j0.
Since pnin-1 = pn-1, there exists p : X1 ! Y by pjn = pn. In particular,
f = pi.
To show that p 2 Sr, let G 2 C(s; p). Since As is small, there exists n
with jng1 = G1. Thus, setting g2 = G2, we obtain g 2 C(s; pn). We
claim that jn+1h(g) is the required lifting. For pjn+1h(g) = pn+1h(g) =
__
g2 = G2, and jn+1h(g)s = jn+1ing1 = jng1 = G1. |_|
6.6. Proposition: In the situation of proposition 6.5, suppose also
that S is countable, that the codomain of every map in S is small, that f
is a map between small objects, and that the set of maps between any
two small objects in C is finite. Then the map i can be factored as a
composition X0 ! X1 ! . . .of maps in S0, and each X0 ! Xn can
be factored as a composition X0 = X(n;0)! X(n;1)! . . .of maps in S0
between small objects.
Proof. Recall from the proof of proposition 6.5 that i is the composition
of maps in S0 of the form X = X0 ! X1 ! . .,.with f factoring as a
composition X ! Xn ! Y . Recall also that there is a construction fac-
toring any q : Z ! Y as Z ! W (q) ! Y , such that Xn+1 = W (Xn ! Y )
and such that the map Xn ! Xn+1 is part of the resulting factorization
of Xn ! Y . Thus it suffices to prove that, given maps in S0 of the
form Z0 ! Z1 ! . . .of small objects with colimit Z, and given a map
q : Z ! Y , then the composition Z0 ! Z ! W (q) can be factored as a
composition Z00! Z01! . .o.f maps in S0 between small objects. But, if
{s1; s2; : :}:is an enumeration of S, then Z0ncan be taken to be the value
__
of W (Zn ! Y ) obtained by taking S = {s1; : :;:sn}. |_|
6.7. Definition: Let C be a category having all finite limits and col-
imits. Given an ordered triple M = (Mw ; Mf; Mc) of classes of maps
of C, we define two more classes of maps by Maf = Mw \ Mf and
Mac = Mw \ Mc. We call the maps in these classes, respectively, weak
equivalences, fibrations, cofibrations, acyclic fibrations, and acyclic cofi-
brations in M. If the map from an initial object (an initial object exists
in C, since the empty diagram has a colimit) to an object X of C is
a cofibration, then we say that X is cofibrant (in M). If the map to a
terminal object from an object X of C is a fibration, then we say that X
20
is fibrant (in M).
Given such C and M, we say that M is a closed model structure on C,
if it satisfies the following conditions.
1. For any maps f and g in C with gf defined, if two of the three maps f,
g, and gf are in Mw , so is the third.
2. The classes Mw , Mf, and Mc are closed under retracts.
3. The inclusions Mc (Maf)l and Mac (Mf)l hold.
4. For all f 2 C there exist p 2 Maf, i 2 Mc, q 2 Mf, and j 2 Mac,
with f = pi = qj.
A closed model category is a category C with some fixed closed model
structure. For a closed model category C, the distinguished classes of
maps will be denoted simply by Cw , Cf, and so on.
6.8. Proposition: For any closed model category C, the equalities
Claf= Cc, Clf= Cac, Crc= Caf, and Crac= Cf hold.
Proof. By definition, Claf Cc. Conversely, a map f in Clafis a retract
of a map i in Cc, by proposition 6.3 and condition 4 of definition 6.7.
Finally, f is in Cc by condition 2 of definition 6.7. The proofs of the
__
remaining equalities are similar (note that T Sl , S T r). |_|
6.9. Definition: Let C be a pointed simplicial category such that C0
has all limits and colimits. We say that C has all limits and colimits,
provided that we have fixed maps C x S ! C, (X; K) 7! X ^K, and
Sopx C ! C, (K; X) 7! XK , and natural isomorphisms C(X0; XK1) ~=
C(X0^ K; X1) ~=S(K; C(X0; X1).
6.10. Definition: Let C be a pointed simplicial category having all
limits and colimits. Given maps p : X ! Y and i : W ! Z in C, and a
map j : K ! L in S, let pj be the canonical map from XL to XK xY KY L.
Let pi be the canonical map from C(Z; X) to C(W; X) xC(W;Y )C(Z; Y ).
Let i__^j be the canonical map from W ^L [W ^K Z ^K to Z ^L.
6.11. Definition: Let C be a pointed simplicial category having all
limits and colimits. A model structure on C consists of a closed model
structure on C0, satisfying the following condition. Given a cofibration i
and a fibration p in C, the map pi is a fibration in S (see example 6.13),
which is a weak equivalence if so is i or p.
21
6.12. Remark: We compare definition 6.11 with the definitions of [Q].
There is a forgetful functor from pointed simplicial categories to non-
pointed simplicial categories. There are also obvious versions of defini-
tions 6.9, 6.10 (in the non-pointed case, we write X A instead of X ^K,
and A is now a (non-pointed) simplicial set) and 6.11, which apply to non-
pointed simplicial categories. The forgetful functor from pointed simpli-
cial categories to non-pointed simplicial categories takes model structures
to model structures. In the situation of the non-pointed version of def-
inition 6.11, C becomes a closed simplicial model category, in the sense
of section II.2 of [Q]. Conversely, any closed simplicial model category
whose underlying category has all limits and colimits (i.e., not just the
finite ones) and for which the objects X A and XA are defined for all A
(i.e., not just for the finite ones) and depend functorially in X and A, is
a special case of the non-pointed version of definition 6.11. Any closed
simplicial model category is in particular a closed model category, which
in turn is in particular a model category, but we will not need any of
these weaker structures. All model structures found in this paper are
special cases of (the pointed version of) definition 6.11.
6.13. Example: There is the standard model structure on S, for which
all special classes of maps are defined by checking on underlying simplicial
sets (in particular, the cofibrations are the injective maps). This follows
immediately from the corresponding result for (non-pointed) simplicial
sets [Q, section II.3]. Since we do not need any other model structure
on S, we write simply Sw , Sf, etc., for the corresponding classes of maps
in S. We need to know [Q, section II.3] that the fibrations can already be
defined as (Sfinac)r, where Sfinac= Sac\ Sfin0, and, similarly, Saf = (Sfinc)r.
6.14. Proposition: Let C be a pointed simplicial category having all
limits and colimits, and let M be a closed model structure on C0. Con-
sider the following conditions.
1. Given a fibration p in C and a cofibration j in S, the map pj is a
fibration in C, which is a weak equivalence if so is p or j.
2. Given a cofibration i in C and a cofibration j in S, the map i__^j is a
cofibration in C, which is a weak equivalence if so is i or j.
Each of these conditions is equivalent to M being a model structure on C.
Proof. The adjoint functor properties of ? ^K and ?K imply that the
statements j 2 (pi)l, i 2 (pj)l, and i__^j 2 pl, are equivalent. The conclu-
__
sion follows from proposition 6.8 and example 6.13. |_|
22
7. The pointwise model structure
In this section we construct a model structure on pointed simplicial cat-
egories of the form C*, called the pointwise model structure. The "strict
model category structure on spectra" of [BF] is a special case of this
construction.
7.1. Definition: Let C be a small pointed simplicial category. A map
X ! Y in C* is called a pointwise weak equivalence, resp. a pointwise
fibration, provided that, for all K in C, the map XK ! Y K is a weak
equivalence, resp. a fibration, in S. We denote the classes of these maps
by C*pw, resp. C*pf. Let C*paf= C*pw\ C*pfand C*pc= (C*paf)l.
We define two more classes of maps of C*, namely (C*pc)g and (C*pac)g
(the notation (C*(a)c)g intends to suggest "generating pointwise (acyclic)
cofibrations in C*"; see proposition 7.2 below). Define (C*p(a)c)g to be
the class of all maps of the form X ^f with X a representable simplicial
functor in C* and f a(n acyclic) cofibration of pointed finite simplicial
sets.
7.2. Proposition: (i) The classes C*pw, C*pf, and C*pcare part of a
model structure C*pfor C*.
(ii) The class of cofibrations generated by (C*pc)g equals C*pc. The class
of cofibrations generated by (C*pac)g equals C*pac.
7.3. Definition: In case C = Sfin, we write SF p instead of C*p. We
refer to SF p as the pointwise model structure on simplicial functors. In
case C = Sph , we write Sp p instead of C*p. We refer to Sp p as the
pointwise model structure on spectra.
The pointwise model structure on spectra is also constructed in [BF]
(proposition 2.2), where it is called "the strict model structure", and
similarly the pointwise weak equivalences are called "strict weak equiva-
lences", and so on.
Proof of proposition 7.2. The results of section 3 show that C* has all
limits and colimits, and condition 1 of proposition 6.14 is true in our
case. By proposition 6.14, if the four conditions of definition 6.7 are
satisfied, then the proof of part (i) is complete. Condition 1 is true. By
proposition 6.2, C*pcis closed under retracts. Thus condition 2 is also
true.
Let the ordered pair of symbols (d; e) be either (c; af) or (ac; f). Fix
23
a map f in C* and apply proposition 6.5 with S = (C*pd)g. We obtain
a factorization f = pi with p 2 ((C*pd)g)r and i 2 ((C*pd)g)0. By corol-
lary 3.15 and proposition 3.16, the map in S obtained by evaluating p on
any K is in (Sfind)r, which equals Se by example 6.13. Thus, by defini-
tion, p is in C*pe. By corollary 3.15, proposition 3.16, and example 6.13,
(C*pc)g C*pc. Further, (C*pac)g C*pw, and (C*ac)g C*pc, thus (C*pac)g
C*pac. By proposition 6.2, ((C*pc)g)0 C*pc, thus ((C*pac)g)0 C*pc. Since
((C*pac)g)0 C*pw, we have ((C*pac)g)0 C*pac, and the verification of
condition 4 is complete.
Before we verify condition 3, we prove that ((C*pd)g)0 C*pd, completing
the proof of part (ii). Let f in C*pd, and factor f = pi with i 2 ((C*pd)g)0
and p 2 C*pe. In case d = c, f 2 C*pc= (C*paf)l, also p 2 C*paf, and
proposition 6.3 implies that f is a retract of i, thus f 2 ((C*pc)g)0. In
case d = ac, p is in C*pw, since so are i and f. Thus p 2 C*paf. But
f 2 C*pac C*pc= (C*paf)l, and proposition 6.3 implies that f is a retract
of i, thus f 2 ((C*pac)g)0.
Finally, we verify condition 3. We have to check that C*pac (C*pf)l,
and we know that C*pac= ((C*pac)g)0. The conclusion follows from propo-
sition 6.2, and the fact that, by corollary 3.15, proposition 3.16, and
__
example 6.13, (C*pac)g (C*pf)l. |_|
8. The homotopy-functor model structure
In this section we construct a model structure on SF , the homotopy-
functor model structure of simplicial functors, for which the weak equiv-
alences are the maps of simplicial functors inducing weak equivalences of
associated homotopy functors. This model structure should be thought
of as a model structure on pointed homotopy functors from Sfin to S
(cf. remark 4.6).
8.1. Definition: Given a map f : X ! Y in a pointed simplicial
model category C, the mapping cylinder Zf of f is defined to be the
colimit of X 1 X ! Y , where the first map is the composition of
X ~=X 0 and X @1 and the second map is f. Consider the colimit
of X 1 X @1 ! X _ Y , where the first map is X (@1
1), and the second map is defined as follows. It is the composition of
X @1 ~=X _ X and 1X _ f, where the inverse isomorphism X _ X ~=
X @1 takes the first copy of X to the image of X @0. Then these
two colimits are naturally isomorphic. Thus there is a canonical map
X _ Y ! Zf, which is a cofibration if X is cofibrant, since it is a cobase
24
change of X (@1 1), which is a cofibration by proposition 6.14.
On the other hand, the canonical map X ! X _ Y is a cofibration if Y
is cofibrant. There is also a canonical map Zf ! Y , induced from the
composition of the projection X1 ! X with f, together with 1Y . This
map is a homotopy equivalence, i.e., it has a homotopy inverse (which is
a section, and is given by the canonical map, and the homotopy in the
other direction is given by using the homotopy between 11 and the
map 1 ! 1 that projects on the image on @1). Thus, if X and Y
are cofibrant, we have a factorization of f as X ! Zf ! Y , with the
first map a cofibration, and the second a homotopy equivalence. If C
is given by the pointwise structure on SF , and if X = L*, Y = K*,
and f is induced from a map g : K ! L, we denote the canonical
cofibration X ! Zf by cg. We define the class SF extrahacto consist of
all cg__^h, with g 2 Sfinwand h 2 Sfinc.
8.2. Definition: Let SF hc = SF pc, SF ghac= SF gpac[ SF extrahac, and
SF hf = (SF ghac)r.
8.3. Theorem: (i) The classes SF hw, SF hf, and SF hc form a model
structure SF h for SF .
(ii) The class of cofibrations generated by SF ghacequals SF hac.
8.4. Definition: We refer to SF h as the homotopy-functor model struc-
ture of simplicial functors (cf. remark 4.6).
8.5. Lemma: The classes of acyclic fibrations for SF h and SF p coin-
cide.
Proof of theorem 8.3. Condition 2 of proposition 6.14 is true in our
case, by proposition 7.2 and 6.14, since SF hc = SF pc (the part where i
is an acyclic cofibration follows since i__^j is in SF hw if so is i). By
proposition 6.14, if the four conditions of definition 6.7 are satisfied, then
the proof of part (i) is complete. Condition 1 is true. We already know
that SF hcis closed under retracts, and proposition 6.2 implies that so is
SF hf. Thus condition 2 is also true.
Fix a map f in SF . Proposition 7.2 and lemma 8.5 give the factorization
by a cofibration and an acyclic fibration. Apply proposition 6.5 with
S = SF ghac. We obtain a factorization f = pi with p 2 (SF ghac)r = SF hf
and i 2 (SF ghac)0. The inequalities SF ghac SF hw, and SF ghac SF hc
hold. By proposition 6.2, (SF ghac)0 SF hc. Since (SF ghac)0 SF hw, we
25
have (SF ghac)0 SF hac.
Before we verify condition 3, we prove that (SF ghac)0 SF hac, completing
the proof of part (ii). Let f in SF hac, and factor f = pi with i 2 (SF ghac)0
and p 2 SF hf. Then p is in SF hw, since so are i and f. Thus p 2 SF paf,
by lemma 8.5. But f 2 SF hac SF pc = SF lpaf, and proposition 6.3
implies that f is a retract of i, thus f 2 (SF ghac)0.
Finally, we verify condition 3. We have to check that SF hac SF lhf, and
we know that SF hac = (SF ghac)0. The conclusion follows from proposi-
__
tion 6.2, since, by definition, SF ghac SF lhf. |_|
8.6. Definition: Let R be the object of SF taking K 2 Sfin to the
singular complex of the realization of K.
8.7. Proposition: The class SF hf consists of these maps X ! Y
in SF pf for which the square
XK ! XRK
# #
Y K ! Y RK
is homotopy cartesian for each K in S.
__
Proof of lemma 8.5. This follows immediately from proposition 8.7. |_|
8.8. Proposition: The class SF hf consists of these maps f : X ! Y
in SF pf for which the square
XK ! XL
# #
Y K ! Y L
is homotopy cartesian for each weak equivalence K ! L in Sfin.
Proof of proposition 8.7. Fix a map X ! Y is in SF hf. Given K in Sfin,
let K1 be the target of the map i in the factorization of K ! * given
by proposition 6.5 with S = Sfinac. We first reduce to proving the version
of the conclusion obtained by replacing RK by K1 : Since i is an acyclic
cofibration and RK is fibrant, the map j : K ! RK factors as ki with
k : K1 ! RK . Further, k is a homotopy equivalence, since K1 is
also fibrant. Thus both X(k) and Y (k) are weak equivalences, and the
reduction follows.
26
Proposition 6.6 gives that K1 is the filtered colimit of certain Kn with
K0 = K, and each Kn is the filtered colimit of certain finite K(n;m), with
K(n;0)= K, and all the maps involved are weak equivalences. We have to
show that the map from XK to Y K xY K1 XK1 is a weak equivalence.
Since Sw and Sf are closed under filtered colimits, and since pullbacks
in S commute with filtered colimits, we may assume that K1 equals
some K(n;m). The conclusion now follows from proposition 8.8. The
general case, where K is not necessarily finite, follows since K is the
filtered colimit of its finite subobjects, and since K 7! RK commutes
with filtered colimits by the definition of R.
Conversely, assume that f gives such a homotopy cartesian square for
every K, and fix K ! L in Sfinw. Suppose first that K ! L is in Sfinac.
Factor the map K ! RK through the acyclic cofibration which is the
composition of K ! L and L ! RL. The resulting map RL ! RK is
a homotopy equivalence. As above, we get that its associated square is
homotopy cartesian. Since, in a composed square, the "source" square
is homotopy cartesian if so are the "target" and the composed square,
it follows that the square associated to K ! L is homotopy cartesian.
In the general case of a map K ! L in Sfinw, similar arguments (using
mapping cyclinders to reduce to the case of an acyclic cofibration, and
recalling that the map from the mapping cylinder to L is a homotopy
equivalence) show that the square associated to K ! L is homotopy
__
cartesian. Thus f 2 SF hf, by proposition 8.8. |_|
Proof of proposition 8.8. Fix f : X ! Y in SF pf and g : K ! L in Sfinw.
Recall that cg is a cofibration from L* to Z, where Z is the mapping
cylinder of g*. There is a map from the square in the statement of the
proposition to
F(Z; X) ! XL
# #
F(Z; Y ) ! Y L
which is the identity on the right column, and on the left column it
is induced from the homotopy equivalence Z ! K*. Thus the map is a
weak equivalence on all four corners, therefore the square in the statement
of the proposition is homotopy cartesian if and only if the square above
is, i.e., if and only if the fibration fcg is acyclic. But, by the proof of
proposition 6.14, this happens if and only if f 2 (cg__^h)r for all maps h
__
in Sfinc. |_|
27
9. The stable model structure of simplicial functors
In this section we construct a model structure on SF , the stable model
structure of simplicial functors, which we show in section 11 is equivalent
to the "standard" model structure of spectra (i.e., to the stable model
structure of spectra introduced in [BF]).
9.1. Definition: Given K in Sfin, the canonical map K ! K
is a vertex in S(K; K), which is canonically isomorphic to K*K.
Let s0K denote the map from (K)*^ S1 to K* corresponding to the
above vertex in K*K under the isomorphisms of corollary 3.15 and
proposition 3.16. Let sK denote the canonical map from (K)*^ S1 to
the mapping cylinder of s0K. We define the set SF extrasacto consist of the
maps sK with K in Sfin.
Given X in SF , define T X in SF by K 7! XK, so that there is
a canonical map X ! T X. Define T 1X as the colimit of the se-
quence X ! T X ! T (T X) ! . .,.and define the stabilization Xs
of X as T 1RXR. A map f : X ! Y in SF is called a a stable weak
equivalence of simplicial functors provided that the map fs is a point-
wise weak equivalence of simplicial functors. We denote the class of
these maps by SF sw. We set SF sc= SF pc, SF gsac= SF ghac[ SF extrasac, and
SF sf= (SF gsac)r.
9.2. Theorem: (i) The classes SF sw, SF sf, and SF sc form a model
structure SF s for SF .
(ii) The class of cofibrations generated by SF gsacequals SF sac.
9.3. Definition: We refer to SF s as the stable model structure of sim-
plicial functors.
9.4. Lemma: The classes of acyclic fibrations for SF s and SF p coin-
cide.
9.5. Definition: Given non-negative integers n and l, we say that a
simplicial functor X in SF (n) is o(l) if there exists an integer c such
that, for all integers k and all k-connected K1; : :;:Kn in S, the pointed
simplicial set X(K1; : :;:Kn) is (lk - c)-connected. we say that a map f
in SF (n)is o(l) if so is its pointwise homotopy fiber. In case n = 1, we
say that f is eo(l) if fR is o(l).
9.6. Proposition: Every map in SF gsacis eo(2).
28
Proof. This is non-trivial only for maps of the form sK , and it is sufficient
to consider the maps s0K. In case K = n=@n, the conclusion follows
by estimating the connectivity of the canonical map nRL ! nRL.
__
The general case follows by induction on the skeleta of K. |_|
9.7. Proposition: If f is eo(2), then f is a stable weak equivalence.
Proof. In fact, there exists an integer c such that, for all n, the map
obtained by evaluating T nRfR on any pointed simplicial set K is (n-c)-
__
connected. |_|
Proof of theorem 9.2. The proof of theorem 8.3 applies almost word for
word, except that two points need an extra argument. First, the claim
that i__^j is in SF sw if i : X ! Y is in SF sacand j : K ! L is in Sc.
Second, the claim that (SF gsac)0 SF sw. To prove the second claim,
note that the class of stable weak equivalences is closed under retracts
and sequential colimits, that every map in SF gsacis eo(2), and that the
class of maps which are both eo(2) and cofibrations are closed under cobase
change and finite sums.
To prove the first claim, note that, even without it, we have a closed
model structure. Further, by the proof of proposition 6.5, the class SF sac
consists precisely of the retracts of the filtered colimits of maps which
are both eo(2) and cofibrations, and these are closed under smashing with
pointed simplicial sets. Thus the map X ^K ! Y ^K and the map
X ^L ! Y ^L are in SF sw. Since SF sacis closed under cobase change,
__
the map X ^L ! X ^L [X ^K Y ^K is also in SF sw, thus so is i__^j. |_|
9.8. Proposition: The class SF sf consists of these maps f : X ! Y
in SF hf for which the square
XK ! XK
# #
Y K ! Y K
is homotopy cartesian for each K in Sfin.
__
Proof. The proof is similar to the proof of proposition 8.8. |_|
Proof of lemma 9.4. Fix a map f : X ! Y in SF haf = SF paf. The
diagram in the statement of proposition 9.8 is homotopy cartesian, since
the vertical maps are acyclic fibrations. Thus f is in SF sf(and therefore
in SF saf), by proposition 9.8.
29
Conversely, fix f : X ! Y in SF saf. The diagram
XK ! Y K
# #
XK ! Y K
# #
XRK ! Y RK
# #
RXRK ! RY RK
# #
XStK ! Y StK
consists of four squares, which we label, starting from the top, as a,
b, c, and d. Square c is homotopy cartesian, since f is in SF pf, since
the functors X 7! X and X 7! RX preserve pointwise fibrations,
since the functor X 7! X preserves pullbacks, and, finally, since the
map X ! RX is in SF pw. Square b is homotopy cartesian, since f is
in SF hf, and since the functor X 7! X preserves pointwise fibrations
and pullbacks. Square a is homotopy cartesian, since f is in SF sf, by
proposition 9.8. Consider the composed square abc. This square, i.e.,
the one determined, in the obvious way, from the maps X ! T RXR,
Y ! T RY R, and f, is homotopy cartesian. The previous sentence
remains true, if we substitute a finite iterate T ninstead of T , by similar
arguments. Since fibrations of simplicial sets are closed under sequential
colimits, and since pullbacks of simplicial sets commute with sequential
colimits, the composed square abcd is homotopy cartesian. Since f is
in SF sw, the bottom horizontal map is in Sw , thus so is the top horizontal
__
map. Thus f is in SF pw, i.e., f is in SF paf. |_|
10. The stable model structure of spectra
In this section we give an exposition of the part of stable homotopy theory
that we need.
10.1. Convention: All representable simplicial functors in this section
are spectra (i.e., the value of (Sn)* on Sm equals Sph (Sn; Sm )).
10.2. Definition: Let n be a non-negative integer. The canonical map
S0 ! S0 is a vertex in Sph (Sn; Sn+1 ) = (Sn)*Sn+1 . Let s0ndenote
the map from (Sn+1 )*^ S1 to (Sn)* corresponding to the above vertex
in (Sn)*Sn+1 under the isomorphisms of corollary 3.15 and proposi-
tion 3.16. Let sn denote the canonical map from (Sn+1 )*^ S1 to the map-
30
ping cylinder of s0n. We define the set Sp extrasacto consist of the maps sn
with n a non-negative integer.
__ __
A spectrum_E can also be viewed as a diagram E 0! E 1! . .i.n S, such
that E n is given by applying n to_a pointed_simplicial set En (which
is part of the data), and the map E n! E n+1 is given by applying n to
a map En ! En+1. In this description,_a_map_E_! E0 is a map of as-
sociated diagrams, such that the map E n! E0nis given by applying n
to a map En ! E0n. This description shows that, given a spectrum E,
there exists_a spectrum T E, whose associated diagram is given by for-
getting E 0. There is a natural map E ! T E, given for each n by the
canonical map En ! En+1. Define the spectrum T 1E as the colimit
of the diagram E ! T E ! T (T E) ! . ... A map f : E ! E0 in Sp
is called a a stable weak equivalence provided that the map T 1Rf is a
pointwise weak equivalence. We denote the class of these maps by Sp sw.
We set Sp sc= Sp pc, Sp gsac= Sp gpac[ Sp extrasac, and Sp sf= (Sp gsac)r.
10.3. Remark: Given a spectrum E, let E[1] denote the canonical
spectrum with (E[1])n = En+1. Then it is not true that T E equals E[1].
In fact, if E0 = E, then the canonical map E00! E01is not given
by applying to the canonical map E^(0) : E0 ! E1 (it is given by
postcomposing E^(0) with the obvious involution; to see this, write down
explicitly the definition of EK , and specialize to K = S1).
10.4. Theorem: (i) The classes Sp sw, Sp sf, and Sp scform a model
structure Sp s for Sp .
(ii) The class of cofibrations generated by Sp gsacequals Sp sac.
10.5. Definition: We refer to Sp s as the stable model structure of
spectra.
10.6. Lemma: The classes of acyclic fibrations for Sp s and Sp p coin-
cide.
10.7. Definition: We say that a map f in Sp is eventually a pointwise
weak equivalence if there exists a non-negative integer N such that, for
all integers n N, the map fn is a weak equivalence.
10.8. Proposition: Every map in Sp gsacis eventually a pointwise weak
equivalence.
31
Proof. This is non-trivial only for maps of the form sn, and it is suffi-
cient to consider the maps s0n. But these are eventually isomorphisms,
since (Sn+1 )*^ S1 and (Sn)* differ, up to isomorphism, only on the value
__
at Sn. |_|
10.9. Proposition: If a map of spectra is eventually a pointwise weak
__
equivalence, then it is a stable weak equivalence. |_|
Proof of theorem 10.4. The proof of theorem 9.2 applies word for word,
after making the obvious changes (this is literally true; here are the
changes: replace "eo(2)" by "eventually pointwise weak equivalence", and
__
replace SF by Sp ). |_|
10.10. Proposition: The class Sp sfconsists of these maps f : E ! E0
in Sp pf for which the square
En ! En+1
# #
E0n ! E0n+1
is homotopy cartesian for each non-negative integer n.
__
Proof. The proof is similar to the proof of proposition 8.8. |_|
Proof of lemma 10.6. The proof is similar to the proof of lemma 9.4 (in
fact, it is easier than the proof of lemma 9.4, because for spectra we do
__
not have to worry about evaluating on fibrant objects). |_|
We now prove some standard facts about spectra, which we need in the
rest of the paper.
10.11. Definition: An -spectrum is a spectrum which is fibrant
in Sp s.
10.12. Proposition: For any spectrum E, the spectrum T 1RE is an
__
-spectrum. |_|
10.13. Definition: Let n be a non-negative integer. The n-skeleton E(n)
of the spectrum E is defined to be the spectrum obtained from E by let-
ting the map E(n)(m) be equal to either E(m) if m - n < 0, or the
identity map of m-n+1 En if m - n 0.
32
10.14. Proposition: For any spectrum E and any non-negative inte-
ger n, the following hold.
(i) There is a canonical diagram E(0) ! E(1) ! . . .! E, which ex-
presses E as the sequential colimit of its skeleta.
(ii) There is a canonical map (Sn)*^ En ! E(n), which is eventually an
__
isomorphism. |_|
10.15. Definition: Let n be an integer. The n-th homotopy group ssnE
of the spectrum E is the abelian group given by the colimit over m
of ssn+m |Em |.
10.16. Proposition: The following hold.
(i) There are natural isomorphisms between the n-th homotopy group
of the spectrum E and the (n + m)-th homotopy group of the pointed
topological space |(T 1RE)m |, provided that n + m is positive.
(ii) A map of spectra is a stable weak equivalence if and only if it induces
an isomorphism on all homotopy groups.
(iii) There are natural isomorphisms between the n-th homotopy group
of the -spectrum E and the (n + m)-th homotopy group of the pointed
topological space |Em |, provided that n + m is positive.
(iv) A map of -spectra is a stable weak equivalence if and only if it is a
__
pointwise weak equivalence. |_|
10.17. Proposition: Given E0 ! E1 an injective map of spectra,
with E2 its cofiber, there is a natural exact sequence
. .!.ssn+1E2 ! ssnE0 ! ssnE1 ! ssnE2 ! ssn-1E0 ! . .;.
where the maps ssnEi ! ssnEi+1 are the canonical maps.
Proof. Since sequential colimits of abelian groups preserve exact se-
quences, and since the functor E 7! E(n)preserves injections and cofibers,
it suffices to show that the required exact sequence exists on all n-skeleta.
By proposition 10.14, we may assume that all Ei are given by (Sn)*^ Ki
for some Ki in S, and the maps between the Ei are induced by maps
between the Ki. The conclusion follows from the Blakers-Massey the-
orem (the version given in Spanier's book [S] as theorem 9.3.5 is suffi-
cient), which provides a constant c such that, for m k + c, the fol-
lowing two statements are true. First, there are natural isomorphisms
between ssk-nEi and ssk+m Sm ^ Ki. Second, there is a natural exact se-
33
quence
. .!.ssk+m+1 (Sm ^ K2) ! ssk+m (Sm ^ K0) ! ssk+m (Sm ^ K1)
! ssk+m (Sm ^ K2) ! ssk+m-1 (Sm ^ K0) ! . .;.
where the maps ssk+m (Sm ^ Ki) ! ssk+m (Sm ^ Ki+1) are the canonical
__
maps. |_|
10.18. Corollary: Let fi : Ei0! Ei1and gi : E0i! E1ibe maps of
spectra, for i = 0; 1, such that g1f0 = f1g0, and the gi are injective.
Then, if any two of f0, f1=f0, and f1 are stable weak equivalences, so is
__
the third. |_|
10.19. Definition: Let o be the natural isomorphism that interchanges
the second and third factors in the smash product K ^L ^M ^N in S.
10.20. Definition: The smash product of the spectra E and E0 is
the spectrum E ^E0 defined as follows. Let (E ^E0)2n = En ^E0n, and
(E ^E0)2n+1 = S1 ^(E ^E0)2n. Finally, let the map (E ^E0)(2n) be the
identity, and the map (E ^E0)(2n + 1) equal (E(n) ^E0(n)) O t.
10.21. Proposition: Smashing with a spectrum preserves stable weak
equivalences.
Proof. Smashing with a pointed simplicial set preserves stable weak
equivalences: It preserves acyclic cofibrations since Sp s is a simplicial
model category, and it preserves acyclic fibrations, since these are point-
wise weak equivalences. But, as the isomorphisms
ssk(E ^E0) ~=colimm ssk+2m |Em ^E0m| ~=colimm;n ssk+m+n |Em ^E0n|
show, for fixed k and E, the group ssk(E ^E0) depends functorially on a
__
diagram involving only the groups ssk(Em ^E0). |_|
11. Comparison of the stable model structures
In this section we compare the stable model structures for spectra and
simplicial functors.
11.1. Definition: Let C and D be simplicial categories, each with a
model structure, and let (L; R) be an adjoint pair of simplicial functors,
with L : C ! D. The pair (L; R) is a Quillen equivalence, provided
34
that L preserves cofibrations,that R preserves fibrations, and that the
following condition is satisfied. For every cofibrant X in C and fibrant Y
in D, a map X ! RY is a weak equivalence if and only if so is its adjoint.
Quillen equivalences induce adjoint equivalences of homotopy categories,
but we will not need this here. One should think of the equivalence
relation generated by "being Quillen equivalent to" as "the right way" to
partition model structures on simplicial categories, so that each partition
consists of model categories that are "really the same".
11.2. Definition: Let i be the inclusion of Sph in S, and let (i*; i*) be
the adjoint pair of proposition 3.23.
11.3. Theorem: The pair (i*; i*) is a Quillen equivalence between the
stable model structure of simplicial functors and the stable model struc-
ture of spectra.
11.4. Convention: By definition, the model structures SF p, SF h,
and SF s share a common class of cofibrations. We refer to a map in this
common class simply as a "cofibration of simplicial functors". We agree
to a similar convention for the term "cofibrant simplicial functor", and
to similar conventions for cofibrations of spectra.
11.5. Proposition: The canonical map E ! i*((i*E)R) is a stable
weak equivalence, for every cofibrant spectrum E.
11.6. Proposition: A map X ! Y of simplicial functors is a stable
weak equivalence if and only if the induced map i*(XR) ! i*(Y R) of
spectra is a stable weak equivalence.
Proof of theorem 11.3. By propositions 10.10 and 9.8, i* maps SF sf
to Sp sf, and, by lemmas 10.6 and 9.4, it maps SF saf, to Sp saf. Proposi-
tion 6.8 implies that its left adjoint i* maps SF sac, to Sp sac, and that it
maps SF sc, to Sp sc. It remains to prove that, given a fibrant X in SF s
and a cofibrant E in Sp s, a map E ! i*X is a stable weak equivalence
if and only if so is the map i*E ! X.
In case E ! i*X is a stable weak equivalence, by proposition 11.6, we
have to show that i*((i*E)R) ! i*(XR) is a stable weak equivalence.
By proposition 11.5, it suffices to show that the map E ! i*(XR) is
a stable weak equivalence. But the map X ! XR is a pointwise weak
equivalence, (X is fibrant in SF , and therefore also in H) thus the map
35
i*X ! i*(XR) is a pointwise weak equivalence.
In case i*E ! X is a stable weak equivalence, so is i*((i*E)R) ! i*(XR)
(by proposition 11.6), and so is the map E ! i*(XR) (by proposi-
tion 11.5). But, as we have seen above, the map i*X ! i*(XR) is a
__
pointwise weak equivalence. |_|
Proof of proposition 11.5. We treat first the special case E = (Sn)*^ K.
We claim that the map E ! i*((i*E)R) is o(2), where we say that a
map f of spectra is o(2) if, for some c, the map fn is (2n - c)-connected.
Since smashing with K preserves connectivity, we may assume that K =
S0. By proposition 3.24, the associated map Ek+n ! (i*((i*E)R))k+n
is isomorphic to the canonical map Sk ! S(Sn; RSk+n ), and the claim
follows. Since the class of o(2) maps is included in Sp sw, the special case
follows.
In the general case, apply proposition 6.5 to the map f : * ! E with
S = Sp c, and let E0 be the target of the resulting i. Thus E0 is a
cofibrant spectrum, and the map E0 ! E is a pointwise weak equiv-
alence. Recall from the proof of proposition 6.5 that E0 is built by
a combination of cobase changes of finite sums and filtered colimits,
and observe that E 7! i*((i*E)R) commutes with colimits. Recall also
that all the cobase changes of finite sums involved are obtained from
maps between spectra of the form (Sn)*^ K. By a Mayer-Vietoris argu-
ment and induction, the map E0 ! i*((i*E0)R) is a filtered colimit of
o(2) maps, in particular a weak equivalence of spectra. But i* preserves
pointwise acyclic cofibrations, thus also pointwise weak equivalences be-
tween cofibrant objects, by a mapping cylinder argument. Thus the
map i*E0 ! i*E is a pointwise weak equivalence, and therefore so is the
map i*((i*E)R) ! i*((i*E0)R). The conclusion follows immediately from
__
this. |_|
11.7. Proposition: Given K in Sfin and X in SF , the canonical map
i*K*XR ! i*XRK* is a stable weak equivalence.
11.8. Proposition: Given K in Sfin and X in SF , the following hold.
(i) The map i*XRK* ! i*T (RXRK*) is a stable weak equivalence.
(ii) The spectrum i*XsK* is an -spectrum.
Proof of proposition 11.6. We first remark that the "only if" part would
be trivial, if the functors T for spectra and simplicial functors would cor-
36
respond under i*. This is, however, not true (in fact, the spectrum i*T X
is canonically isomorphic to (i*X)[1], cf. remark 10.3).
Suppose first that X ! Y is a stable weak equivalence. By proposi-
tion 11.8(i), the maps i*XR ! i*Xs and i*Y R ! i*Y sare stable weak
equivalences of spectra. It follows that i*XR ! i*Y R is a stable weak
equivalence, since i*Xs ! i*Y sis even a pointwise weak equivalence.
Suppose now that i*XR ! i*Y R is a stable weak equivalence of spec-
tra. Since smashing by K preserves stable weak equivalences of spectra
(for a simple proof of this last fact, see the proof of proposition 10.21),
proposition 11.7 implies that i*XRK* ! i*Y RK* is a stable weak equiv-
alence, for each K in Sfin. Thus, by proposition 11.8(i), so is the map
i*T 1(RXRK*) ! i*T 1(RY RK*). In fact, by proposition 10.16(iv) and
proposition 11.8(ii) (and since XsK* = (T 1RXR)K* is canonically iso-
morphic to T 1(RXRK*)), this map is a pointwise weak equivalence.
__
Thus the map XsK ! Y sK is a weak equivalence. |_|
11.9. Lemma: The map of proposition 11.7 is o(2), if X = L* for
some L in Sfin.
Proof of proposition 11.7. Call the map f. The proof will be divided in
five cases, each more special than the next.
Case 1. X is a finite sum of simplicial functors of the form K* ^L for
some K and L in Sfin. Then lemma 11.9 implies that f is o(2).
Case 2. X is any sum of simplicial functors of the form K* ^L for some K
and L in Sfin. Then f is a filtered colimit of maps which are weak
equivalences, by case 1.
Case 3. X is a colimit of a sequence X0 ! X1 ! . .,.such that X0 = *
and each map is given by a cobase change of a map having the follow-
ing form: It is a sum of maps of the form K* ^g for some K in Sfin
and some cofibration g in Sfin. Since stable equivalences are closed un-
der filtered colimits, we may assume that the sequence is in fact finite.
The conclusion follows by induction on the length of the sequence, using
corollary 10.18 and case 2.
Case 4. X is any cofibrant simplicial functor. By definition, this means
that there exists a sequence X0 ! X1 ! . .s.atisfying all the conditions
of the similar sequence in case 3, except possibly the condition that X0 is
trivial, and with the map * ! X a retract of the composition X0 ! X1 .
But then X is a retract of X1 =X0, and the map Xn=X0 ! Xn+1=X0 is
still a cobase change of the same map as Xn ! Xn+1. The conclusion
37
follows from case 3.
Case 5. General case (X is any simplicial functor). Factor the map
* ! X as pj, with j a cofibration and p an acyclic fibration (the model
structure is irrelevant, cf. convention 11.4). In particular, p is a pointwise
weak equivalence, thus i*pR is a pointwise weak equivalence of spectra,
__
and the conclusion follows from case 4. |_|
Before we prove proposition 11.8, we need the following technical lemma.
The lemma allows us to view any simplicial functor as a filtered colimit
of simplicial functors which are finite in a strong sense.
11.10. Lemma: Let A be the partially ordered set given by the finite
sets of objects of Sfin. Every X in H determines Xc : A ! H satisfying
the following two conditions. First, for every A in A, the object XcA is
the diagonal of a simplicial object in H, whose value at [q] is
_
K*q^S(Kq-1; Kq) ^. . .^S(K0; K1) ^X(K0)
where the sum is over (K0; : :;:Kq) in Aq+1. Second, if Xc1 denotes the
colimit of Xc, then there exists a pointwise weak equivalence Xc1 ! X.
Finally, the construction X 7! Xc is functorial, and the map Xc1 ! X
is natural.
Proof of proposition 11.8. By lemma 11.10, we may assume that X is of
the form XcA, with the notation as in the lemma. Thus X satisfies the fol-
lowing two conditions. First, the simplicial functor XR is o(1). Second,
the canonical map XR ! XR is o(2). (The proof of these two facts
uses that sums, smashing by pointed simplicial sets, and diagonalization,
preserve connectivity of pointed simplicial sets. The proof of the second
fact uses lemma 11.9. It is essential that the set of K* involved is finite,
so that we may assume that the constants c, involved in the definition
of the relevant o(k), are all equal.) Since XRK* is still o(1), the map
XRK* ! RXRK* is o(2). Thus, since the map XRK* ! XRK*
__
is still o(2), so is the map XRK* ! T RXRK*. |_|
Proof of lemma 11.9. Call the map f. We first prove the following
reduction: Given injective maps L0 ! L1 in Sfin, if the lemma is true
for L0 and L1=L0, then it is true for L1. We need the following definition:
In case X, Y , and Z are o(j) simplicial functors in SF (n)and X ! Y ! Z
is a sequence with trivial composition, then we say it is an o(l)-fibration
sequence if the map from X to the pointwise homotopy fiber of Y ! Z
38
is o(l), and we say it is an o(l)-cofibration sequence if the map from the
pointwise homotopy cofiber of X ! Y to Z is o(l). In case l = 2j, the
Blakers-Massey theorem (again, [S, theorem 9.3.5] is sufficient) implies
that the sequence is an o(l)-fibration sequence if and only if it is an o(l)-
cofibration sequence. (In verifying the above, note that, by definition
of o(l), in order to check that a map F in SF (n) is o(l) it suffices to
consider only m-connected K1; : :;:Kn, for any constant m. If further
both the source and the target of F are o(j), then, by taking m large
enough, we can arrange that the values of the source and the target of F
are highly connected.) Since the sequence (L1=L0)*R ! L*1R ! L*0R
is a fibration sequence, smashing it with K yields an o(2) cofibration
sequence in SF . There is a map, given by f, from this sequence to the
sequence (L1=L0)*R(? ^K) ! L*1R(? ^K) ! L*0R(? ^K) in SF . Since
each L* is o(1), the target sequence is an o(2) cofibration sequence (since
it is a fibration sequence), and the reduction follows.
The lemma holds for L = Sn, since f factors as the composition
K ^nR? ! nR(K ^nnR?) ! nR(K ^R?);
and both maps are o(2). Thus the lemma is true for n=@n. The
conclusion follows by induction on the simplices of L, using the above
__
reduction. |_|
Proof of lemma 11.10. Let I be a small category and F a functor from I
to sets. Given i in I, let C(i) be the following category. Its objects
are the pairs (f; x) such that f : i0 ! i and x 2 F (i0). There is one
morphism from (f; x) to (g; y) for each h in I such that gh = f and
h*(x) = y. Define F 0(i) as the nerve of C(i). There is a map F 0(i) ! F (i)
taking (f; x) to f*(x), and is the disjoint union of projections of nerves
of categories to their terminal objects, in particular a weak equivalence.
Note that
a
Fq0~= i*qx I(iq-1; iq) x . .x.I(i0; i1) x F (i0):
(i0;:::;iq)
Suppose now that I is a pointed category and that F is a pointed functor
from I to pointed sets. Let
_
Fq00= i*q^I(iq-1; iq) ^. . .^I(i0; i1) ^F (i0)
(i0;:::;iq)
so that we have a canonical map F 0! F 00. Then it is still true that there
is a map F 00! F , and that, for fixed i, there is a section F (i) ! F 00(i),
39
as well as a map F 00(i) 1 ! F 00(i) which is a homotopy between the
identity and the composition F 00(i) ! F (i) ! F 00(i). All these maps
are compatible with the canonical map F 0! F 00, and exist essentially
because I is a pointed category and F is a pointed functor.
Fix X in H and apply the above construction with I = Sfinpand F = Xp
(the functor Xp from Sfinpto E is defined in 3.10). The resulting X00p,
evaluated at L, has q-simplices
_
K*qL ^Sp(Kq-1; Kq) ^. . .^Sp(K0; K1) ^Xp(K0)
where the sum is over all (K0; : :;:Kq), and K*qdenotes the functor
from Sfinpto E represented by Kq. The conclusion follows immediately
__
from this. |_|
12. Comparison of the smash products
In this section we compare the smash products of spectra and simplicial
functors, cf. theorem 12.5. We also prove that the smash product of
simplicial functors is compatible with all three model structures on sim-
plicial functors considered in previous sections, cf. theorems 12.3, 12.4,
and 12.6.
12.1. Definition: Given maps f : X ! Y and g : Z ! W in SF ,
let f__^g be the canonical map from X ^W [X ^Z Y ^W to Z ^W .
12.2. Proposition: Let f, g, and h be maps in SF . If g is a retract
of f, then g__^h is a retract of f__^h. If f is the composition . .f.1f0,
then f__^h is the composition . .(.f1__^h)(f0__^h). If the map of arrows
f ! g corresponds to a cocartesian square (expressing g as a cobase
change of f), then the map of arrows f__^h ! g__^h also corresponds to a
cocartesian square. Finally, if f is the sum of fs, where {fs} is a family
of maps indexed by s in some indexing set S, then f__^h is the the sum
__
of fs__^h. |_|
12.3. Theorem: Let f and h be maps in SF with f a cofibration. Then
the map f__^h is injective, if so is h, and it is a cofibration, if so is h.
Proof. Suppose first that h is injective. Consider first the case where f
is in the set SF gpcof generating cofibrations, i.e., for some K in Sfin and
some cofibration g in Sfin, we have f = K* ^g. Using proposition 5.13
to write (X ^K*)L as X(K*L), and the fact that injective maps in S are
closed under cobase change, we conclude that f__^h is injective.
40
If f is any cofibration, the conclusion follows from proposition 12.2, using
the fact that injective maps in S are closed under retracts, infinite com-
position, cobase change, and sums, as well as the definition of the class
of cofibrations in SF (this last class is defined as the class of cofibrations
generated by SF gpc; see also definition 6.4).
Suppose now that h is a cofibration. Consider first the case where both f
and h are in the set SF gpcof generating cofibrations, i.e., for some K and L
in Sfin and some cofibrations g and g0 in Sfin, we have f = K* ^g and
h = L*^ g0. By lemma 5.7, (K* ^L*) is isomorphic to (K ^L)*. Thus f__^h
is isomorphic to (K ^L)*^ (g__^g0), which is again a generating cofibration.
If f and h are any cofibrations, the conclusion follows from proposi-
tion 12.2, using the fact that cofibrations are closed under retracts, in-
finite composition, cobase change, and sums, as well as the definition of
__
the class of cofibrations in SF . |_|
12.4. Theorem: Let f be a map in SF and X a cofibrant simplicial
functor. Then X ^f is a pointwise weak equivalence, if so is f, and it is
a weak equivalence of homotopy functors, if so is f.
Proof. If X = K* for some K in Sfin, the conclusion follows from
proposition 5.13, togetherWwith the fact that K*L is fibrant if so is L.
The special case X = sK*s^Ls follows immediately.
If X is any cofibrant simplicial functor, by proceeding as in the proof
of proposition 11.7, we can write X as a retract of the colimit X1 of
a sequence * = X0 ! X1W! . .,.with the map Xn ! Xn+1 a cobase
change of a map gn = s K*s^hs with hs a cofibration in Sfin. We
may assume that X = X1 , since weak equivalences in S are closed
under retracts. We may assume that X equals some Xn, since weak
equivalences in S are closed under sequential colimits. The conlusion
now follows by induction on n, the gluing lemma, and the special case
treated in the previous paragraph. (In order to apply the gluing lemma,
we need to know that g ^Y is injective for any simplicial functor Y , and
any cofibration g of simplicial functors. This follows from theorem 12.3,
__
by considering the injective map * ! Y .) |_|
12.5. Theorem: There is a map of spectra i*XR ^i*Y R ! i*(X ^Y )R,
natural in the simplicial functors X and Y , which is a stable weak equiv-
alence if one of the factors is cofibrant.
12.6. Theorem: Smashing with a cofibrant simplicial functor preserves
41
stable weak equivalences.
Proof. This follows from theorem 12.5, proposition 11.6, and proposi-
__
tion 10.21. |_|
12.7. Definition: An S2-spectrum E consists of a sequence of pointed
simplicial sets E2n and a sequence of maps E(2n) : S2 ^E2n ! E2n+2,
for n = 0; 1; : :.: We define maps and stable weak equivalences of such
objects, so that forgetting the odd terms of a spectrum gives a func-
tor E 7! E* that preserves stable weak equivalences.
Let X be a simplicial functor. The S2-spectrum XSo is defined as follows.
Let (XSo)2n = XS2n. Let (XSo)(2n) be the composition
S1 ^S1 ^X(Sn ^Sn) ! X(S1 ^S1 ^Sn ^Sn) ! X(S1 ^Sn ^S1 ^Sn);
where the first map is the special assembly map, and the second map is
induced by o (for the definition of o , see 10.19).
12.8. Proposition: The S2-spectra (i*X)* and XSo are naturally iso-
morphic.
Proof. Let the sphere spectrum S equal i*S, and the S2-spectrum So
equal S(So). The conclusion of the proposition is true in the special
case X = S, i.e., S* and So are isomorphic. The general case follows
because there is a functor which associates the S2-spectrum XE to a
simplicial functor X and an S2-spectrum E, such that (i*X)*, resp. XSo,
__
equals XE with E = S*, resp. E = So. |_|
The only reason we consider XSo (in fact, the only reason we consider S2-
spectra) is to be able to write the map in the statement of theorem 12.5
as a composition of simpler maps. One of these simpler maps is given
by the proposition above, and another one is given by the proposition
below.
12.9. Proposition: Given simplicial functors X and Y , there is a
non-trivial natural map (i*XR ^i*Y R)* ! (X ^Y )RSo.
Proof. By definition of X ^Y , there is a canonical map from XM ^Y N
to (X ^Y )(M ^N). The map we want is obtained from this map by
letting M and N be RSn (this uses the map RR ! R, coming from the
__
adjunction between singular complex and realization). |_|
42
12.10. Proposition: If X and Y are simplicial functors and X is cofi-
brant, then the map (i*XR ^i*Y R)* ! (X ^Y )RSo is a weak equivalence.
Proof of theorem 12.5. Note that the forgetful functor E 7! E* from
spectra to S2-spectra has a left adjoint L, such that (LE)2n = E2n,
(LE)2n+1 = S1 ^E2n, (LE)(2n) = 1, and (LE)(2n + 1) = E(2n). The
conclusion now follows from proposition 12.8, proposition 12.10, and the
fact that, for any spectra E and E0, the equality E ^E0 = L(E ^E0)*
__
holds. |_|
12.11. Lemma: The canonical map XRM ^Y RN ! (X ^Y )R(M ^N)
is o(3), if X = K* and Y = L* for some K and L in Sfin.
Proof of proposition 12.10. We say that a map of S2-spectra is o(3), if
there exists an integer c such that, for all non-negative integers n the
map of (2n)-terms is (3n - c)-connected. It follows immediately that an
o(3)-map is a stable weak equivalence. Call the map f. The proof will
be divided in five cases, each more special than the next.
Case 1. X and Y are a finite sum of simplicial functors of the form K* ^L
for some K and L in Sfin. Then lemma 12.11 implies that f is o(3).
Case 2. X and Y are any sum of simplicial functors of the form K* ^L
for some K and L in Sfin. Then f is a filtered colimit of maps which are
weak equivalences, by case 1.
Case 3. X and Y are given by colimits of sequences Z0 ! Z1 ! . .,.
such that Z0 = * and each map is given by a cobase change of a map
having the following form: It is a sum of maps of the form K* ^g for
some K in Sfinand some cofibration g in Sfin. Since stable equivalences
are closed under filtered colimits, we may assume that both compositions
are in fact finite. The conclusion follows by a double induction on the
length of the compositions, using corollary 10.18 and case 2 (theorem 12.3
guarantees that the injectivity hypothesis in corollary 10.18 is satisfied).
Case 4. X and Y are any cofibrant simplicial functors. The conclusion
follows by proceeding as in the proof of proposition 11.7, to write X
and Y as retracts of simplicial functors corresponding to the case 3.
Case 5. General case (X and Y are any simplicial functors, and X is
cofibrant). Factor the map * ! Y as pj, with j a cofibration and p an
acyclic fibration (the model structure is irrelevant, cf. convention 11.4).
In particular, p is a pointwise weak equivalence, thus i*p is a stable weak
equivalence of spectra (even a pointwise one). The conclusion follows
43
__
from theorem 12.4, proposition 10.21, and case 4. |_|
Proof of lemma 12.11. Call the map f. We first prove the following
reduction: Given injective maps K0 ! K1 in Sfin, if the lemma is true
for (K0; L) and (K1=K0; L), then it is true for (K1; L). Since the sequence
(K1=K0)* ! K*1 ! K*0is a fibration sequence, external-smashing it
with L* yields an o(3) cofibration sequence in SF (2) (cf. the proof of
lemma 11.9). There is a map, given by f, from this sequence to the
sequence
(K1=K0^ L)*R(M ^N) ! (K1^ L)*R(M ^N) ! (K0^ L)*R(M ^N)
in SF (2). Since each (K1^ L)*R(M ^N) is o(2), the target sequence is an
o(4) cofibration sequence (since it is a fibration sequence; cf. the proof of
lemma 11.9 again), and the reduction follows.
The lemma holds for (K; L) = (Sm ; Sn), since f factors as the composi-
tion
m RM ^nRN ! m+n R(Sm m RM ^SnnRN) ! m+n R(M ^N);
and both maps are o(4). Thus the lemma is true for (m =@m ; n=@n).
The conclusion follows by induction on the simplices of K and L, using
__
the above reduction. |_|
13. Linear functors
We end this paper by relating the stable model structure of simplicial
functors to the calculus of homotopy functors, cf. [G1], [G2], and [G3].
For our purposes, the second section of [G1] is sufficient. In fact, here
we are only interested in the part of this calculus having to do with the
linear approximation at *. We show in proposition 13.4 that, by using
simplicial functors instead of homotopy functors (this is "no restriction,
up to weak equivalence", cf. remark 4.6), a rigid universal property of
this linear approximation is possible. The corresponding results in [G1]
are stated for "stably excisive" functors only, a restriction that turns out
to be unnecessary.
13.1. Definition: A simplicial functor is linear, provided that it is a
homotopy functor and that it takes any sequence K0 ! K1 ! K2, such
that K0 ! K1 is injective with cofiber K1 ! K2, to a fibration sequence.
13.2. Proposition: A simplicial functor is fibrant in the stable model
structure if and only if it is linear and pointwise fibrant.
44
Proof. Suppose X is linear and pointwise fibrant. Let D be the pointed
simplicial set obtained by choosing a basepoint for 1. It suffices to
exhibit a weak equivalence from the sequence XK ! X(D ^K) ! XK
to the sequence XK ! (XK)D ! XK. For, since the source is a
fibration sequence, so is the target. But, since (XK)D ! XK is a
fibration with fiber XK, and the resulting map from XK to the fiber
is the canonical one, the conclusion follows from proposition 9.8. We
exhibit a map which is the identity on the fiber and the base. Choose
the contraction of 1 yielding a map D ^D ! D (the choice depends on
the choice of the basepoint of D), and postcompose with the canonical
map D ! S1 to obtain a map D ^D ! S1. The required map is the
adjoint of the composition D ^X(D ^K) ! X(D ^D ^K) ! XK.
Suppose X is fibrant in the stable model structure, and fix a sequence
K0 ! K1 ! K2 in Sfin, such that K0 ! K1 is injective with cofiber
K1 ! K2. Note that X is fibrant in SFh thus, by proposition 8.7,
the map X ! XR is a pointwise weak equivalence. Proposition 11.7
now implies that the canonical map from i*X ^K = i*K*X to i*XK*
is a stable weak equivalence. The proof of proposition 10.17 actually
shows that (T 1(i*X ^K0))n ! (T 1(i*X ^K1))n ! (T 1(i*X ^K2))n is
a fibration sequence for any n. By propositions 10.16 (iv), and 10.12,
and since by proposition 9.8 the spectrum i*(XK*) is an -spectrum,
this sequence for n = 0 is related by a chain of weak equivalences to the
sequence XK0 ! XK1 ! XK2, therefore this last sequence is also a
__
fibration sequence. |_|
13.3. Definition: Given a closed model category D, for which there
exists a set S of maps in D with small domains such that the class of
acyclic cofibrations equals the class of cofibrations generated by S, and
given an object X of D, the canonical fibrant replacement X ! Xf is the
acyclic cofibration provided by the small object argument applied to the
map X ! *. Thus Xf is a fibrant object depending functorially on X,
and the map X ! Xf is both natural and an acyclic cofibration.
13.4. Proposition: Given a simplicial functor X, let X ! P X be the
canonical fibrant replacement in the stable model structure. Then P X
is linear, and X ! P X is a natural stable weak equivalence. Further,
if Y is any linear simplicial functor which is pointwise fibrant, any map
X ! Y factors through X ! P X, and any two factorizations P X ! Y
can be connected by a homotopy P X 1 ! Y relative to X.
45
Proof. This follows immediately from proposition 13.2 and standard
__
properties of simplicial model categories. |_|
Acknowledgements_ The author is grateful to the SFB 343 at the University of
Bielefeld, especially Friedhelm Waldhausen, for their hospitality while this pa*
*per was
written.
46
References
[B] B"okstedt, M.: Topological Hochschild homology, preprint, 1985.
[BF] Bousfield, A. K., Friedlander, E. M.: Homotopy theory of -
spaces, spectra, and bisimplicial sets, Springer Lecture Notes in Ma*
*th.,
Vol. 658, Springer, Berlin, 1978, pp. 80-130.
[D] Day, B.: On closed categories of functors, Reports of the Midwest
Category Seminar IV, Springer Lecture Notes in Math., Vol. 137,
Springer, Berlin, 1970, pp. 1-38.
[EKMM] Elmendorf, A. D., Kriz, I., Mandell, M. A., May, J. P.: Rings,
modules, and algebras in stable homotopy theory. With an appendix by
M. Cole, Mathematical Surveys and Monographs, Vol. 47, American
Mathematical Society, Providence, 1997.
[G1] Goodwillie, T: Calculus I: The first derivative of pseudoisotopy the-
ory, K-Theory 4 (1990), 1-27.
[G2] Goodwillie, T: Calculus II: Analytic functors, K-Theory 5 (1992),
295-332.
[G3] Goodwillie, T: Calculus III: The Taylor series of a homotopy func-
tor, preprint, 1993.
[HSS] Hovey, M., Shipley, B., Smith, J.: Symmetric spectra, preprint,
1998.
[L1] Lydakis, M.: Homotopy limits of categories, J. Pure Appl. Algebra
97 (1994), 73-80.
[L2] Lydakis, M.: Smash products and -spaces, Math. Proc. Cam. Phil.
Soc., to appear.
[LS] Lydakis, M., Schwede, S.: Topological Hochschild homology and
-spaces, in preparation.
[Q] Quillen, D. G.: Homotopical Algebra, Springer Lecture Notes in
Math., Vol. 43, Springer, Berlin, 1967.
[S] Spanier, E.: Algebraic Topology, Springer, New York, 1966.
[Sch] Schwede, S.: Stable homotopical algebra and -spaces, Math. Proc.
Cam. Phil. Soc., to appear.
[W] Waldhausen, F.: Algebraic K-theory of spaces, Springer Lecture
Notes in Math., Vol. 1126, Springer, Berlin, 1984, pp. 318-419.
47