Professional Documents
Culture Documents
Abstract
Dwyer and Kan developed a homotopical version of the calculus of
fractions in order to get a handle on simplicial localisations of categories. We use this to show that, for a category of fibrant objects (in the
sense of Brown), Jardines cocycle categories functorially compute the
homotopy type of the hom-spaces in the simplicial localisation. As an
application, we deduce a non-abelian version of Verdiers hypercovering
theorem suggested by Rezk.
Introduction
Given a category C and a subcategory W C, the category C[W 1 ] obtained
from C by freely inverting the morphisms in W is straightforward to construct:
its objects are the objects in C and its morphisms are equivalence classes of
zigzags of arrows in C where the backward-pointing arrows are in W. The
simplicial and hammock localisations introduced by Dwyer and Kan [1980a,b]
are homotopy-theoretic versions of this construction and can be shown to have
the appropriate universal property in the context of (, 1)-categories.
Although the hom-spaces of the hammock localisation already have a fairly
simple explicit description, just as in the case of ordinary localisation, one can
sometimes obtain an even simpler description when the pair (C, W) has good
properties. For instance, Dwyer and Kan [1980c] have shown that it suces
to consider only zigzags of the form
and the main goal of this paper is to show that the hom-spaces of the hammock
localisation of (C, W) have the homotopy type of the nerve of the cocycle
categories.[1]
As an application, we consider the category of (locally brant) simplicial
presheaves on a site. From the point of view of homotopy theory, the central problem of sheaf theory is essentially the determination of the homotopy
type the space of sections (over a given object in the site) of the hypersheaf
associated with a given simplicial presheaf: for example, as Brown [1973, 3]
observed, sheaf cohomology can be paraphrased in these terms via the formula
below,
Rn (, A)
= mn R(, K(A, m))
where A is an abelian (pre)sheaf, K(A, m) is the simplicial (pre)sheaf corresponding (under DoldKan) to the chain complex consisting of just A in degree
m, and m n. Verdiers hypercovering theorem in its classical form is a colimit
formula for sheaf cohomology in terms of generalised ech cochain complexes,
and following a suggestion of Rezk [2014], we derive a non-abelian version
that computes (up to weak homotopy equivalence) R(, X) for any locally
brant simplicial presheaf X in terms of a homotopy colimit of simplicial sets
of generalised sections of X.
Here is a brief outline of what follows:
In 1, we collect some miscellaneous facts about homotopy colimits.
In 2, we review the denitions and fundamental results regarding zigzags in relative categories.
In 3, we introduce a sucient condition for a category with weak equivalences to admit a homotopical calculus of right fractions.
[1] A related claim previously appeared as Proposition 3.23 in [Cisinski, 2010b], but the proof
at loc. cit. has a gap: while it is true that the hammock localisation of C with respect to
weak equivalences is DwyerKan equivalent to the hammock localisation of C with respect
to trivial fibrations, the hypotheses of Proposition 8.1 in [Dwyer and Kan, 1980b] are not
satisfied, so it does not follow that we have a homotopy calculus of right fractions. Instead,
we must use a somewhat more complicated argument.
Conventions
It will be convenient to implicitly assume that categories are small, especially
in 25 and A. Since the categories of interest are usually not small, it
is not possible to apply these results as stated literally; one way to work
around this to adopt a suitable universe axiom. Alternatively, because most
of the categories under consideration in 6 are essentially small, one could just
replace them with small skeletons where necessary, thereby avoiding the use
of universes.
Acknowledgements
Discussions with Aaron Mazel-Gee led to the discovery and correction of some
errors in earlier drafts of this paper. His comments also helped improve the
exposition.
The author gratefully acknowledges nancial support from the Cambridge
Commonwealth, European and International Trust and the Department of
Pure Mathematics and Mathematical Statistics.
Homotopy colimits
The following denition is due to Bouseld and Kan [1972].
Definition 1.1. Let X : C op sSet be a small simplicially enriched diagram.
The homotopy colimit holimC op X is the diagonal of the bisimplicial set
E (e0 , R(e1 ))
E L e0 , e1
E(L(e0 ), e1 )
B (b0 , G(b1 ))
B(F (b0 ), b1 )
B F b0 , b1
where both squares and the outer rectangle are pullback diagrams; but the
composite of the bottom row is a bijection, so the composite of the top row is
also a bijection. Thus, L : E E indeed has a right adjoint.
Lemma 1.8. Let X : C op sSet be a small simplicially enriched diagram. If
C has cotensor products m c for every standard simplex m and every object
c in C, then (regarding the underlying category C as a simplicially enriched
category with discrete hom-spaces) the canonical comparison morphism
holimC op X holimC op X
=
Ym,n
(c0 ,...,cn )
Definition 2.1.
A relative category is a pair C = (und C, weq C) where und C is a category and weq C is a (usually non-full) subcategory of und C containing
all the objects.
Given a relative category C, a weak equivalence in C is a morphism in
weq C.
The homotopy category of a relative category C is the category Ho C
obtained by freely inverting the weak equivalences in C.
Given relative categories C and D, a relative functor C D is a
functor und C und D that restricts to a functor weq C weq D, and
the relative functor category [C, D]h is the relative category whose
underlying category is the full subcategory of the ordinary functor category [und C, und D] spanned by the relative functors, with the weak
equivalences being the natural transformations whose components are
weak equivalences in D.
Remark 2.2. The 2-category of (small) categories admits several 2-fully faithful embeddings into the 2-category of (small) relative categories; unless otherwise stated, we will regard an ordinary category as minimal relative category where the only weak equivalences are the identity morphisms. In particular, given an ordinary category C and a relative category D, we will often
tacitly identify the ordinary functor category [C, D] with the relative functor
category [C, D]h .
Definition 2.3.
A zigzag type is a nite sequence of non-zero integers (k0 , . . . , kn ),
where n 0, such that for 0 i < n, the sign of ki is the opposite
of the sign of ki+1 .
Given a nite sequence of integers (k0 , . . . , kn ), [k0 ; . . . ; kn ] is the relative
category whose underlying category is freely generated by the graph
0
|k0 | + + |kn |
where (counting from the left) the rst |k0 | arrows point rightward (resp.
leftward) if k0 > 0 (resp. k0 < 0), the next |k2 | arrows point rightward
(resp. leftward) if k1 > 0 (resp. k1 < 0), etc., with the weak equivalences
being generated by the leftward-pointing arrows.
A zigzag in a relative category C of type [k0 ; . . . ; kn ] is a relative functor
[k0 ; . . . ; kn ] C; given a zigzag, its length is m = |k0 | + + |kn |, its
7
where the top row is the domain, the bottom row is the codomain, and
the vertical arrows are weak equivalences in C.
Composition and identities are inherited from C.
Remark. In other words, the morphisms in C [k0 ;...;kn ] (X, Y ) are certain hammocks of width 1, in the sense of Dwyer and Kan [1980b].
For brevity, let us say that a weak homotopy equivalence of categories
is a functor F : C D such that N(F ) : N(C) N(D) (i.e. the induced
morphism of nerves) is a weak homotopy equivalence of simplicial sets. The
following is a variation on the homotopy calculus of right fractions introduced
in [Dwyer and Kan, 1980b].
LH C(X, Y )
N C [1;1] (X , Y )
LH C(X , Y )
where the left vertical arrow is defined by composition and the right vertical arrow is defined by concatenation.
9
X is a morphism in V.
where v : X
[1;1]
We write CV
(X, Y ) for the full subcategory of C [1;1] (X, Y ) spanned
by the V-cocycles.
If V = weq C, then we may simply say cocycle instead of V-cocycle.
Remark 3.2. In other words, a cocycle in C is a zigzag of type [1; 1].
Proposition 3.3. Let C be a relative category and let : C Ho C be the
localising functor. If C admits a homotopical calculus of right fractions, then:
(i) Every morphism X Y in Ho C can be factored as (f ) (w)1 for
some cocycle (f, w) : X
[ Y in C.
(ii) Two cocycles X
[ Y represent the same morphism X Y in Ho C if
and only if they are in the same connected component of C [1;1] (X, Y ).
10
X
v
hdom, codomi
weq C weq C
12
D fun
C fun
hdom, codomi
hdom, codomi
weq D weq D
weq C weq C
so by the pullback pasting lemma, the outer rectangle in the diagram below is
also a pullback diagram in Cat:
D fun
C fun
hdom, codomi
weq D weq D
hdom, codomi
weq C weq C
id
p1
Path(Y )
p0
13
C fun (X, Y )
C fun (X, Y )
CV
UX,Y
(X, Y )
C [1;1] (X, Y )
Y
u
id
14
Y
w
id
Y
u
Y
id
r
g
id
id
Y
w
w
x
15
H2 (X, Y )
R
P
d
Y/
H1 (X, Y )
Y/
W is the
16
N s2
N(H2 (X, Y ))
N r2
id
N(d)
N(H0 (X, Y ))
N(s)
N(H1 (X, Y ))
and (by the 2-out-of-6 property) it will follow that s2 : H0 (X, Y ) H2 (X, Y )
is indeed a weak homotopy equivalence of categories.
First, observe that every object in H2 (X, Y ) has an underlying commutative diagram in C of the form below:
Y
id
X
k
vk
0
X
k1
X
fk1
k
X
j X
j+1 in the above diagram.
For 0 j < k, write fj for the morphism X
X
k is in V, we may functorially construct the following
Since vk : X
k
commutative diagram in C,
X
0
X
k1
X
k
vk1
v0
0
X
fk1
k1
X
vk
fk1
k
X
X
0
X
k1
id
X
0
X
k1
vk1
v0
0
X
fk1
k1
X
X
k
vk
fk1
k
X
fj
j+1
X
uj
id
X
j
uj+1
fj
X
j+1
vj
j
X
vj+1
fj
j+1
X
0
X
X
0
fk1
uk1
u0
k1
X
X
k1
q fk1
LH C(X , Y )
where the left vertical arrow is defined as in remark 3.5 and the right
vertical arrow is defined by concatenation.
(iii) There is an isomorphism
[1;1]
N CV
(, )
= LH C(, )
of functors Ho C op Ho C Ho sSet.
18
Path(X)
X X
where : X X X is the diagonal morphism, i : X Path(X) is
in W, and Path(X) X X is in F .
(E) For any object X in C, the unique morphism X 1 is in F .
In a category of brant objects as above,
a weak equivalence is a morphism in W,
a fibration is a morphism in F , and
a trivial fibration (or acyclic fibration) is a morphism in W F .
Example 4.2. Of course, the full subcategory of brant objects in a model
category is a category of brant objects, with weak equivalences and brations
inherited from the model structure.
19
X
p
Y X is a bration.
such that hp, vi : X
fun
We write C (resp. C fun (X, Y )) for the full subcategory of weq [[1; 1], C]h
(resp. C [1;1] (X, Y )) spanned by the functional correspondences.
Remark 4.5. Since product projections in a category of brant objects are
X is a trivial bration and p : X
Y is a
brations, it follows that v : X
bration. However, the converse is not true: for instance, (idY , idY ) : Y [ Y
is rarely a functional correspondence.
Lemma 4.6. Let C be a category of fibrant objects and let W = weq C.
(i) The functor C fun W W sending functional correspondences X [ Y
to the pair (X, Y ) is a Grothendieck fibration.
(ii) The inclusion C fun weq [[1; 1], C]h preserves cartesian morphisms.
Proof. Consider a functional correspondence in C, say:
v
hp , v i
Y X
X
hp, vi
gf
20
Y X
Y
g
denes a cartesian morphism in both C fun and weq [[1; 1], C]h .
id
Ef
Lemma 4.9. Let C be a category of fibrant objects and let (X, Y ) be a pair
of objects in C. If C has functorial path objects, then the inclusion UX,Y :
C fun (X, Y ) C [1;1] (X, Y ) is homotopy cofinal.
21
To extend the above result to the case where C is not assumed to have functorial path objects, we would have to prove lemma 4.9 without using functorial
factorisations. We will do this in the appendix.
22
idK p : K X K Y
LX
LY
i idX
i idY
K X
idK p
KY
K U
K X p
idK p
K X
X
23
idL p
LX
i idU
i idX
K U
idK p
KX
LU
K X U
i idU
ip
i X U
(K U) KX (L X)
Z X p
K U
idK p
LX
i idX
K X
Q(v, p)
Q(p , p)
24
Q(p , p)
L X p
i X p
f
K X p
Q(p , p)
Q(v, p)
Q(p , p)
25
Moreover, recalling theorem 4.11 and proposition 5.3, we may apply theorem 2.10
and lemma 3.9 to reduce the problem to showing that
[1;1]
holimQop C(U, ) N CV
(X, )
[4] See paragraph 4.3 in [Bousfield and Kan, 1972, Ch. XII] or Theorem 18.7.4 in [Hirschhorn,
2003].
26
27
28
R(C, )
L
E
h
,
N
(E
)
h
,
N
(E
)
h
,
=
=
=
< C
<
C
< V
C
of functors Ho E< Ho sSet.
Proof. Combine theorem 2.10 and lemmas 3.9, 6.3, and 6.4.
One may then derive a homotopy colimit formula for R(C, ) analogous
to Verdiers original colimit formula (cf. Thorme 7.4.1 in [SGA 4b, Expos V]
or Theorem 8.16 in [Artin and Mazur, 1969]):
Proposition 6.6. Let C be an object in C, let Q be the simplicially enriched
full subcategory of the simplicially enriched slice category (E< )/hC spanned by
the J-local trivial fibrations, and let U : Q E< be the evident projection
functor. Then there is an isomorphism
E (U , )
R(C, )
= holim
Qop <
of functors Ho E< Ho sSet.
Proof. Apply theorem 5.7 and proposition 6.5.
Corollary 6.7. Let K be a Kan complex of cardinality < and let K be the
constant simplicial presheaf on C with value K. With other notation as above,
we have
R(C, K)
lim
U,
K
sSet
= holim
C op
Qop
as objects in Ho sSet, and this is natural in K.
Proof. As usual, we have the following isomorphism of simplicially enriched
functors Qop sSet:
sSet limC op U, K
= E< (U, K)
References
Artin, Michael, Alexander Grothendieck and Jean-Louis Verdier
[SGA 4b] Thorie des topos et cohomologie tale des schmas. Tome 2. Lecture
Notes in Mathematics 270. Berlin: Springer-Verlag, 1972. iv+418. isbn:
3-540-06012-X.
Artin, Michael and Barry Mazur
[1969] Etale homotopy. Lecture Notes in Mathematics 100. Berlin: Springer-Verlag,
1969. iii+169.
Barnea, Ilan, Yonatan Harpaz and Geoffroy Horel
[BHH] On the pro-category of a weak fibration category. In preparation.
Barwick, Clark and Daniel M. Kan
[2012] Relative categories: another model for the homotopy theory of homotopy
theories. In: Indag. Math. (N.S.) 23.1-2 (2012), pp. 4268. issn: 0019-3577.
doi: 10.1016/j.indag.2011.10.002.
Bousfield, A. K. and Daniel M. Kan
[1972] Homotopy limits, completions and localizations. Lecture Notes in
Mathematics 304. Springer-Verlag, Berlin-New York, 1972. v+348.
Brown, Kenneth S.
[1973] Abstract homotopy theory and generalized sheaf cohomology. In: Trans.
Amer. Math. Soc. 186 (1973), pp. 419458. issn: 0002-9947.
Cisinski, Denis-Charles
[2010a] Catgories drivables. In: Bull. Soc. Math. France 138.3 (2010),
pp. 317393. issn: 0037-9484.
[2010b] Invariance de la K-thorie par quivalences drives. In: J. K-Theory 6.3
(2010), pp. 505546. issn: 1865-2433. doi: 10.1017/is009010008jkt094.
Dwyer, William G. and Daniel M. Kan
[1980a] Simplicial localizations of categories. In: J. Pure Appl. Algebra 17.3
(1980), pp. 267284. issn: 0022-4049. doi: 10.1016/0022-4049(80)90049-3.
[1980b] Calculating simplicial localizations. In: J. Pure Appl. Algebra 18.1 (1980),
pp. 1735. issn: 0022-4049. doi: 10.1016/0022-4049(80)90113-9.
[1980c] Function complexes in homotopical algebra. In: Topology 19.4 (1980),
pp. 427440. issn: 0040-9383. doi: 10.1016/0040-9383(80)90025-7.
Hirschhorn, Philip S.
[2003] Model categories and their localizations. Mathematical Surveys and
32
33
Cambridge Philos. Soc. 85.1 (1979), pp. 91109. issn: 0305-0041. doi:
10.1017/S0305004100055535.
34