Locality and descent
In its categorical meaning descent is the study of generalizations of the sheaf condition on presheaves to presheaves with values in higher categories. Those higher presheaves that satisfy descent are called infinity-stacks.
More generally, descent theory studies existence and (non)uniqueness of an object in a (possibly higher) category provided some “inverse image” functor which applied to produces an object in some (possibly higher) category (or a collection of inverse image functors in is given). In favourable cases, the nonuniqueness is parametrized by equipping the object with additional “gluing” data . The pair is called a descent datum, the existence of a reconstruction procedure of from is also called a descent, and it describes the property that the (higher) category of descent data in is equivalent to the category , or at least that it embeds via a canonical fully faithful functor. Descent theory in 1-categorical context has been first formulated by Grothendieck in FGA using pseudofunctors and in SGA1 using fibered categories.
The most important case is when there is a descent (in the sense of equivalence of higher categories) along an inverse image functor along every cover of a Grothendieck topology or its higher analogue; though many cases (for example descent in noncommutative algebraic geometry) do not fit into this framework. These cases of descent along all covers is also called (higher) stack theory and may be phrased in modern viewpoint as a characterization of -sheaves (i.e. -stacks) among all -presheaves as those -presheaves which are local objects with respect to certain morphisms which are to be regarded as covers or hypercover of the -presheaf : the idea is that an -sheaf “descends from the cover down to ”.
every (∞,1)-category of (∞,1)-sheaves is characterized as being a sub-(∞,1)-topos of the -topos of (∞,1)-presheaves on some (small) (∞,1)-category ;
every such -topos is a reflective (∞,1)-subcategory of , hence a localization of an (∞,1)-category at a collection of morphisms which are sent to equivalences by the left adjoint of the inclusion;
and the sheaves in are precisely the local objects with respect to this collection of morphisms, i.e. precisely those objects such that is an isomorphism in the homotopy category, which we shall write in the following paragraphs.
This condition is essentially the descent conditon.
In concrete models for the (∞,1)-category of (∞,1)-sheaves – notably in terms of the model structure on simplicial presheaves – the morphisms in usually come from hypercovers ;
in this case the above condition becomes which is equivalent to . This in turn is usually equivalently written
And this is the form of the local object-condition which is usually called descent condition.
Descent for ordinary sheaves
Descent is best understood as a direct generalization of the situation for 0-stacks, i.e. ordinary sheaves, which we briefly recall in a language suitable for the following generalization.
For any small category and Set the category of small sets, write for the category of presheaves on . Categories of this form enjoy various nice properties which are familiar from itself, and which are summarized by saying that is a topos. The relevance of this for the present purpose is that there is a natural notion of morphisms of topoi, which are functors respecting this structure in some sense: these are called geometric morphisms.
A category of sheaves on is a sub-topos of in that it is a full and faithful functor which is a geometric morphism.
One finds that the reflective subcategory of sheaves inside presheaves is the localization of at morphisms called local isomorphisms, which are determined by and determing the choice of topos-inclusion. A presheaf is a sheaf precisely if it is a local object with respect to these local isomorphisms, that is precisely if
is an isomorphism for all local isomorphisms .
This locality condition is in fact the descent condition: the sheaf has to descend from down to . More concretely, this condition is called a descent condition when evaluated on morphisms which are hypercovers:
namely if is a local epimorphism with respect to the coverage that corresponds to the localization and if is a local epimorphism, then with
being the two canonical morphisms out of , it follows that the canonical morphism
is a local isomorphism.
(This is excercise 16.6 in Categories and Sheaves).
Therefore for a presheaf to be a sheaf, it is necessary that
is an isomorphism. The colimit may be taken out of the hom-functor to make this equivalently
It is convenient, suggestive and common to write , , following the spirit of the Yoneda lemma whether or not and/or are representable. In that notation the above finally becomes
This is the form of the condition that is most commonly called the descent condition.
Descent for simplicial presheaves
For more references and background on the following see descent for simplicial presheaves.
A well-studied class of models/presentations for an (∞,1)-category of (∞,1)-sheaves is obtained using the model structure on simplicial presheaves on an ordinary (1-categorical) site , as follows.
Let be the SSet-enriched category of simplicial presheaves on .
Recall from model structure on simplicial presheaves that there is the global and the local injective simplicial model structure on which makes it a simplicial model category and that the local model structure is a (Bousfield-)localization of the global model structure.
So in terms of simplicial presheaves the localization of an (∞,1)-category that we want to describe, namely ∞-stackification, is modeled as the localization of a simplicial model category.
Recall that the (∞,1)-category modeled/presented by a simplicial model category is the full SSet-subcategory on fibrant-cofibrant objects. According to section 6.5.2 of HTT we have:
the full simplicial subcategory on fibrant-cofibrant objects of with respect to the global injective model structure is (the SSet-enriched category realization of) the -category of (∞,1)-presheaves on .
the full simplicial subcategory on fibrant-cofibrant objects of with respect to the local injective model structure is (the SSet-enriched category realization of) the -category which is the hypercompletion of the -category of (∞,1)-sheaves on .
Since with respect to the local or global injective model structure all objects are automatically cofibrant, this means that is the full sub--category of on simplicial presheaves which are fibrant with respect to the local injective model structure: these are the ∞-stacks in this model.
By the general properties of localization of an (∞,1)-category there should be a class of morphisms in – hence between injective-fibrant objects in – such that the simplicial presheaves representing -stacks are precisely the local objects with respect to these morphisms.
The general idea of descent in this simplicial context is the precise analog of the situation for ordinary sheaves, but with ordinary (co)limits replaced everywhere with the (∞,1)-categorical (co)limits, which in terms of the presentation by the model structure on simplicial presheaves amounts to the homotopy (co)limit.
So for a morphism of simplicial presheaves, the condition that a simplicial presheaf is local with respect to it, hence satisfies descent with respect to it, is generally that
is a weak equivalence, where denotes the corresponding -categorical hom, i.e. the derived hom with respect to the model structure on simplicial presheaves – for instance the ordinary simplicial hom if both and are fibrant with respect to the given model structure.
The details on which morphisms one needs to check against here have been worked out in
- D. Dugger, S. Hollander, D. Isaksen, Hypercovers and simplicial presheaves (pdf)
We now describe central results of that article.
For an object in the site regarded as a simplicial presheaf and a simplicial presheaf on , a morphism is a hypercover if it is a local acyclic fibration, i.e. of for all and all diagrams
there exists a covering sieve of with respect to the given Grothendieck topology on such that for every in that sieve the pullback of the abve diagram to has a lift
If is a Verdier site then every such hypercover has a refinement by a hypercover which is cofibrant with respect to the projective global model structure on simplicial presheaves. We shall from now on make the assumption that the hypercovers we discuss are cofibrant in this sense. These are called split hypercovers. (This works in many cases that arise in practice, see the discussion after DHI, def. 9.1.)
The objects of – i.e. the fibrant objects with respect to the projective model structure on – are precisely those objects of – i.e. Kan complex-valued simplicial presheaves – which satisfy descent for all split hypercovers, i.e. those for which for all split hypercover in we have that
is a weak equivalence of simplicial sets.
Notice that by the co-Yoneda lemma every simplicial presheaf , which we may regard as a presheaf , is isomorphic to the weighted colimit
which is equivalently the coend
where is the Set-valued presheaf of -cells of regarded as an -valued presheaf under the inclusion , and where the SSet-weight is the canonical cosimplicial simplicial set , i.e. for all
In particular therefore for a Kan complex-valued presheaf the descent condition reads
With the shorthand notation introduced above the descent condition finally reads, for all global-injective fibrant simplicial presheaves and hypercovers :
The right hand here is often denoted , in which case this reads
Descent for strict -groupoid valued presheaves
While simplicial sets are a very convenient model for general reasoning about higher weak categories and ∞-groupoids, often concrete computations in particular with -groupoids are more convenient in the context of more strictified models.
Notably, by the generalized Dold-Kan correspondence the ω nerve injects crossed complexes – nonabelian generalizations of chain complexes of abelian groups which are equivalent to strict ω-groupoids – to simplicial sets
Since for instance something as simple as an abelian group regarded as a complex of groups in degree (hence as an -group) already bcomes a somewhat involved object to understand under the nervet operation,
it is desireable to have a means to control descent for simplicial presheaves which happen to factor through the -nerve directly in the context of .
In his work on descent
Ross Street considered presheaves with values in strict ω-categories
and declared the descent -category with respect to a simplicial object to be the weighted limit in -enriched category theory
where are the orientals, i.e. the free -categories on the simplicial simplices
where is the right adjoint to the ω-nerve .
The two precscriptions
have a very similar appearance. The following theorem asserts if and when they are actually equivalent.
Theorem (Dominic Verity)
There exists a canonical comparison map
This is a weak equivalence of Kan complexes if the cosimplicial simplicial set is Reedy fibrant.
Descent in terms of gluing conditions
We unwrap the expression
for the descent data for a presheaf with respect to a (hyper)cover
This weighted limit (whether taken in - or in -enriched category theory) is given by the coend
Unwrapping what this means one finds that an object/vertex of this is a choice of -simplex in each , subject to conditions which say that the boundary of this -simplex must be obtained from pullback of along the maps of the -simplex in
Namely an object in
is a commuting diagram
where the vertical arrows indicate all the simplicial maps of the cosimplicial objects and .
So this is
on an object ;
on “double intersections” (might be a cover of double intersections) a gluing isomorphism which identifies the two copies of obtained by pullback along the two projection maps .
on “triple intersections” a gluing 2-isomorphism which identifies the different gluing 1-isomorphisms.
And so on.
Gluing for ordinary stacks
- Sharon Hollander, A Homotopy Theory for Stacks (arXiv)
spells out how the familiar formulation of the descent condition for ordinary stacks is equivalent to the corresponding descent condition for simplicial presheaves, duscussed above.
Sometimes one wishes to compute the descent objects for presheaves of the form
where is a given presheaf-valued co-presheaf. For instance in the context of differential nonabelian cohomology one is interested in the co-presheaf that assigns fundamental ∞-groupoids
in which case the presheaf
would assign to the pre--stack of “trivial -principial bundles with flat connection”.
For a given (hyper)cover, the descent object for can be expressed as
This way the descent for on the object is reexpressed as descent for of the -modified object . Following Street, this we may call the codescent object, as it co-represents descent.
In some context the descent condion may algebraically be encoded in an adjunction. This leads to the notion of monadic descent. See there for more details.
- A. Grothendieck, M. Raynaud et al. Revêtements étales et groupe fondamental (SGA1), Lecture Notes in Mathematics 224, Springer 1971 (retyped as math.AG/0206203; published version Documents Mathématiques 3, Société Mathématique de France, Paris 2003)
- Angelo Vistoli, Grothendieck topologies, fibered categories and descent theory MR2223406; math.AG/0412512 pp. 1–104 in Barbara Fantechi, Lothar Göttsche, Luc Illusie, Steven L. Kleiman, Nitin Nitsure, Angelo Vistoli, Fundamental algebraic geometry. Grothendieck’s FGA explained, Mathematical Surveys and Monographs 123, Amer. Math. Soc. 2005. x+339 pp. MR2007f:14001
- Ross Street, Categorical and combinatorial aspects of descent theory, arXiv:math.CT/0303175
- Jacob Lurie, Descent Theorems
- Daniel Schäppi, Descent via Tannaka duality, arxiv/1505.05681
A connection between the monadic descent and the descent in the language of fibered categories is proved in
- Jean Bénabou, Jacques Roubaud, Monades et descente, C. R. Acad. Sc. Paris, t. 270 (12 Janvier 1970), Serie A, 96–98, (link, Bibliothèque nationale de France)